2018-06-06 10:42:14 +08:00
|
|
|
// SPDX-License-Identifier: GPL-2.0
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
2005-11-02 11:58:39 +08:00
|
|
|
* Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
|
|
|
|
* All Rights Reserved.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
|
|
|
#include "xfs.h"
|
2005-11-02 11:38:42 +08:00
|
|
|
#include "xfs_fs.h"
|
2013-10-23 07:36:05 +08:00
|
|
|
#include "xfs_format.h"
|
2013-10-23 07:50:10 +08:00
|
|
|
#include "xfs_log_format.h"
|
2013-10-23 07:36:05 +08:00
|
|
|
#include "xfs_shared.h"
|
2013-10-23 07:50:10 +08:00
|
|
|
#include "xfs_trans_resv.h"
|
2005-11-02 11:38:42 +08:00
|
|
|
#include "xfs_bit.h"
|
2005-04-17 06:20:36 +08:00
|
|
|
#include "xfs_mount.h"
|
2016-08-03 09:15:38 +08:00
|
|
|
#include "xfs_defer.h"
|
2005-04-17 06:20:36 +08:00
|
|
|
#include "xfs_btree.h"
|
2016-08-03 09:33:43 +08:00
|
|
|
#include "xfs_rmap.h"
|
2013-10-23 07:51:50 +08:00
|
|
|
#include "xfs_alloc_btree.h"
|
2005-04-17 06:20:36 +08:00
|
|
|
#include "xfs_alloc.h"
|
2012-04-29 18:39:43 +08:00
|
|
|
#include "xfs_extent_busy.h"
|
2017-11-01 03:04:49 +08:00
|
|
|
#include "xfs_errortag.h"
|
2005-04-17 06:20:36 +08:00
|
|
|
#include "xfs_error.h"
|
2009-12-15 07:14:59 +08:00
|
|
|
#include "xfs_trace.h"
|
2013-10-23 07:50:10 +08:00
|
|
|
#include "xfs_trans.h"
|
2013-04-03 13:11:13 +08:00
|
|
|
#include "xfs_buf_item.h"
|
2013-10-23 07:50:10 +08:00
|
|
|
#include "xfs_log.h"
|
2021-06-02 08:48:24 +08:00
|
|
|
#include "xfs_ag.h"
|
xfs: set up per-AG free space reservations
One unfortunate quirk of the reference count and reverse mapping
btrees -- they can expand in size when blocks are written to *other*
allocation groups if, say, one large extent becomes a lot of tiny
extents. Since we don't want to start throwing errors in the middle
of CoWing, we need to reserve some blocks to handle future expansion.
The transaction block reservation counters aren't sufficient here
because we have to have a reserve of blocks in every AG, not just
somewhere in the filesystem.
Therefore, create two per-AG block reservation pools. One feeds the
AGFL so that rmapbt expansion always succeeds, and the other feeds all
other metadata so that refcountbt expansion never fails.
Use the count of how many reserved blocks we need to have on hand to
create a virtual reservation in the AG. Through selective clamping of
the maximum length of allocation requests and of the length of the
longest free extent, we can make it look like there's less free space
in the AG unless the reservation owner is asking for blocks.
In other words, play some accounting tricks in-core to make sure that
we always have blocks available. On the plus side, there's nothing to
clean up if we crash, which is contrast to the strategy that the rough
draft used (actually removing extents from the freespace btrees).
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-09-19 08:30:52 +08:00
|
|
|
#include "xfs_ag_resv.h"
|
2018-05-08 08:38:47 +08:00
|
|
|
#include "xfs_bmap.h"
|
|
|
|
|
2021-10-13 05:17:01 +08:00
|
|
|
struct kmem_cache *xfs_extfree_item_cache;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2012-03-22 13:15:07 +08:00
|
|
|
struct workqueue_struct *xfs_alloc_wq;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
#define XFS_ABSDIFF(a,b) (((a) <= (b)) ? ((b) - (a)) : ((a) - (b)))
|
|
|
|
|
|
|
|
#define XFSA_FIXUP_BNO_OK 1
|
|
|
|
#define XFSA_FIXUP_CNT_OK 2
|
|
|
|
|
2018-03-07 09:08:32 +08:00
|
|
|
/*
|
|
|
|
* Size of the AGFL. For CRC-enabled filesystes we steal a couple of slots in
|
|
|
|
* the beginning of the block for a proper header with the location information
|
|
|
|
* and CRC.
|
|
|
|
*/
|
|
|
|
unsigned int
|
|
|
|
xfs_agfl_size(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
|
|
|
unsigned int size = mp->m_sb.sb_sectsize;
|
|
|
|
|
2021-08-19 09:46:37 +08:00
|
|
|
if (xfs_has_crc(mp))
|
2018-03-07 09:08:32 +08:00
|
|
|
size -= sizeof(struct xfs_agfl);
|
|
|
|
|
|
|
|
return size / sizeof(xfs_agblock_t);
|
|
|
|
}
|
|
|
|
|
2016-10-04 00:11:17 +08:00
|
|
|
unsigned int
|
|
|
|
xfs_refc_block(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
2021-08-19 09:46:37 +08:00
|
|
|
if (xfs_has_rmapbt(mp))
|
2016-10-04 00:11:17 +08:00
|
|
|
return XFS_RMAP_BLOCK(mp) + 1;
|
2021-08-19 09:46:37 +08:00
|
|
|
if (xfs_has_finobt(mp))
|
2016-10-04 00:11:17 +08:00
|
|
|
return XFS_FIBT_BLOCK(mp) + 1;
|
|
|
|
return XFS_IBT_BLOCK(mp) + 1;
|
|
|
|
}
|
|
|
|
|
2016-08-03 09:31:47 +08:00
|
|
|
xfs_extlen_t
|
|
|
|
xfs_prealloc_blocks(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
2021-08-19 09:46:37 +08:00
|
|
|
if (xfs_has_reflink(mp))
|
2016-10-04 00:11:17 +08:00
|
|
|
return xfs_refc_block(mp) + 1;
|
2021-08-19 09:46:37 +08:00
|
|
|
if (xfs_has_rmapbt(mp))
|
2016-08-03 09:31:47 +08:00
|
|
|
return XFS_RMAP_BLOCK(mp) + 1;
|
2021-08-19 09:46:37 +08:00
|
|
|
if (xfs_has_finobt(mp))
|
2016-08-03 09:31:47 +08:00
|
|
|
return XFS_FIBT_BLOCK(mp) + 1;
|
|
|
|
return XFS_IBT_BLOCK(mp) + 1;
|
|
|
|
}
|
|
|
|
|
2016-08-03 09:38:24 +08:00
|
|
|
/*
|
2022-03-17 02:38:27 +08:00
|
|
|
* The number of blocks per AG that we withhold from xfs_mod_fdblocks to
|
|
|
|
* guarantee that we can refill the AGFL prior to allocating space in a nearly
|
2022-07-23 01:58:17 +08:00
|
|
|
* full AG. Although the space described by the free space btrees, the
|
2022-03-17 02:38:27 +08:00
|
|
|
* blocks used by the freesp btrees themselves, and the blocks owned by the
|
|
|
|
* AGFL are counted in the ondisk fdblocks, it's a mistake to let the ondisk
|
|
|
|
* free space in the AG drop so low that the free space btrees cannot refill an
|
|
|
|
* empty AGFL up to the minimum level. Rather than grind through empty AGs
|
|
|
|
* until the fs goes down, we subtract this many AG blocks from the incore
|
|
|
|
* fdblocks to ensure user allocation does not overcommit the space the
|
|
|
|
* filesystem needs for the AGFLs. The rmap btree uses a per-AG reservation to
|
|
|
|
* withhold space from xfs_mod_fdblocks, so we do not account for that here.
|
|
|
|
*/
|
|
|
|
#define XFS_ALLOCBT_AGFL_RESERVE 4
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Compute the number of blocks that we set aside to guarantee the ability to
|
|
|
|
* refill the AGFL and handle a full bmap btree split.
|
|
|
|
*
|
2016-08-03 09:38:24 +08:00
|
|
|
* In order to avoid ENOSPC-related deadlock caused by out-of-order locking of
|
|
|
|
* AGF buffer (PV 947395), we place constraints on the relationship among
|
|
|
|
* actual allocations for data blocks, freelist blocks, and potential file data
|
|
|
|
* bmap btree blocks. However, these restrictions may result in no actual space
|
|
|
|
* allocated for a delayed extent, for example, a data block in a certain AG is
|
|
|
|
* allocated but there is no additional block for the additional bmap btree
|
|
|
|
* block due to a split of the bmap btree of the file. The result of this may
|
|
|
|
* lead to an infinite loop when the file gets flushed to disk and all delayed
|
|
|
|
* extents need to be actually allocated. To get around this, we explicitly set
|
|
|
|
* aside a few blocks which will not be reserved in delayed allocation.
|
|
|
|
*
|
2022-03-17 02:38:27 +08:00
|
|
|
* For each AG, we need to reserve enough blocks to replenish a totally empty
|
|
|
|
* AGFL and 4 more to handle a potential split of the file's bmap btree.
|
2016-08-03 09:38:24 +08:00
|
|
|
*/
|
|
|
|
unsigned int
|
|
|
|
xfs_alloc_set_aside(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
2022-03-17 02:38:27 +08:00
|
|
|
return mp->m_sb.sb_agcount * (XFS_ALLOCBT_AGFL_RESERVE + 4);
|
2016-08-03 09:38:24 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* When deciding how much space to allocate out of an AG, we limit the
|
|
|
|
* allocation maximum size to the size the AG. However, we cannot use all the
|
|
|
|
* blocks in the AG - some are permanently used by metadata. These
|
|
|
|
* blocks are generally:
|
|
|
|
* - the AG superblock, AGF, AGI and AGFL
|
|
|
|
* - the AGF (bno and cnt) and AGI btree root blocks, and optionally
|
|
|
|
* the AGI free inode and rmap btree root blocks.
|
|
|
|
* - blocks on the AGFL according to xfs_alloc_set_aside() limits
|
|
|
|
* - the rmapbt root block
|
|
|
|
*
|
|
|
|
* The AG headers are sector sized, so the amount of space they take up is
|
|
|
|
* dependent on filesystem geometry. The others are all single blocks.
|
|
|
|
*/
|
|
|
|
unsigned int
|
|
|
|
xfs_alloc_ag_max_usable(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
|
|
|
unsigned int blocks;
|
|
|
|
|
|
|
|
blocks = XFS_BB_TO_FSB(mp, XFS_FSS_TO_BB(mp, 4)); /* ag headers */
|
2022-03-17 02:38:27 +08:00
|
|
|
blocks += XFS_ALLOCBT_AGFL_RESERVE;
|
2016-08-03 09:38:24 +08:00
|
|
|
blocks += 3; /* AGF, AGI btree root blocks */
|
2021-08-19 09:46:37 +08:00
|
|
|
if (xfs_has_finobt(mp))
|
2016-08-03 09:38:24 +08:00
|
|
|
blocks++; /* finobt root block */
|
2021-08-19 09:46:37 +08:00
|
|
|
if (xfs_has_rmapbt(mp))
|
2022-03-17 02:38:27 +08:00
|
|
|
blocks++; /* rmap root block */
|
2021-08-19 09:46:37 +08:00
|
|
|
if (xfs_has_reflink(mp))
|
2016-10-04 00:11:24 +08:00
|
|
|
blocks++; /* refcount root block */
|
2016-08-03 09:38:24 +08:00
|
|
|
|
|
|
|
return mp->m_sb.sb_agblocks - blocks;
|
|
|
|
}
|
|
|
|
|
2008-10-30 13:56:09 +08:00
|
|
|
/*
|
|
|
|
* Lookup the record equal to [bno, len] in the btree given by cur.
|
|
|
|
*/
|
|
|
|
STATIC int /* error */
|
|
|
|
xfs_alloc_lookup_eq(
|
|
|
|
struct xfs_btree_cur *cur, /* btree cursor */
|
|
|
|
xfs_agblock_t bno, /* starting block of extent */
|
|
|
|
xfs_extlen_t len, /* length of extent */
|
|
|
|
int *stat) /* success/failure */
|
|
|
|
{
|
2019-10-14 08:10:31 +08:00
|
|
|
int error;
|
|
|
|
|
2008-10-30 13:56:09 +08:00
|
|
|
cur->bc_rec.a.ar_startblock = bno;
|
|
|
|
cur->bc_rec.a.ar_blockcount = len;
|
2019-10-14 08:10:31 +08:00
|
|
|
error = xfs_btree_lookup(cur, XFS_LOOKUP_EQ, stat);
|
2020-03-11 08:57:51 +08:00
|
|
|
cur->bc_ag.abt.active = (*stat == 1);
|
2019-10-14 08:10:31 +08:00
|
|
|
return error;
|
2008-10-30 13:56:09 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Lookup the first record greater than or equal to [bno, len]
|
|
|
|
* in the btree given by cur.
|
|
|
|
*/
|
2012-03-22 13:15:12 +08:00
|
|
|
int /* error */
|
2008-10-30 13:56:09 +08:00
|
|
|
xfs_alloc_lookup_ge(
|
|
|
|
struct xfs_btree_cur *cur, /* btree cursor */
|
|
|
|
xfs_agblock_t bno, /* starting block of extent */
|
|
|
|
xfs_extlen_t len, /* length of extent */
|
|
|
|
int *stat) /* success/failure */
|
|
|
|
{
|
2019-10-14 08:10:31 +08:00
|
|
|
int error;
|
|
|
|
|
2008-10-30 13:56:09 +08:00
|
|
|
cur->bc_rec.a.ar_startblock = bno;
|
|
|
|
cur->bc_rec.a.ar_blockcount = len;
|
2019-10-14 08:10:31 +08:00
|
|
|
error = xfs_btree_lookup(cur, XFS_LOOKUP_GE, stat);
|
2020-03-11 08:57:51 +08:00
|
|
|
cur->bc_ag.abt.active = (*stat == 1);
|
2019-10-14 08:10:31 +08:00
|
|
|
return error;
|
2008-10-30 13:56:09 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Lookup the first record less than or equal to [bno, len]
|
|
|
|
* in the btree given by cur.
|
|
|
|
*/
|
2018-01-17 10:52:12 +08:00
|
|
|
int /* error */
|
2008-10-30 13:56:09 +08:00
|
|
|
xfs_alloc_lookup_le(
|
|
|
|
struct xfs_btree_cur *cur, /* btree cursor */
|
|
|
|
xfs_agblock_t bno, /* starting block of extent */
|
|
|
|
xfs_extlen_t len, /* length of extent */
|
|
|
|
int *stat) /* success/failure */
|
|
|
|
{
|
2019-10-14 08:10:31 +08:00
|
|
|
int error;
|
2008-10-30 13:56:09 +08:00
|
|
|
cur->bc_rec.a.ar_startblock = bno;
|
|
|
|
cur->bc_rec.a.ar_blockcount = len;
|
2019-10-14 08:10:31 +08:00
|
|
|
error = xfs_btree_lookup(cur, XFS_LOOKUP_LE, stat);
|
2020-03-11 08:57:51 +08:00
|
|
|
cur->bc_ag.abt.active = (*stat == 1);
|
2019-10-14 08:10:31 +08:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline bool
|
|
|
|
xfs_alloc_cur_active(
|
|
|
|
struct xfs_btree_cur *cur)
|
|
|
|
{
|
2020-03-11 08:57:51 +08:00
|
|
|
return cur && cur->bc_ag.abt.active;
|
2008-10-30 13:56:09 +08:00
|
|
|
}
|
|
|
|
|
2008-10-30 13:56:32 +08:00
|
|
|
/*
|
|
|
|
* Update the record referred to by cur to the value given
|
|
|
|
* by [bno, len].
|
|
|
|
* This either works (return 0) or gets an EFSCORRUPTED error.
|
|
|
|
*/
|
|
|
|
STATIC int /* error */
|
|
|
|
xfs_alloc_update(
|
|
|
|
struct xfs_btree_cur *cur, /* btree cursor */
|
|
|
|
xfs_agblock_t bno, /* starting block of extent */
|
|
|
|
xfs_extlen_t len) /* length of extent */
|
|
|
|
{
|
|
|
|
union xfs_btree_rec rec;
|
|
|
|
|
|
|
|
rec.alloc.ar_startblock = cpu_to_be32(bno);
|
|
|
|
rec.alloc.ar_blockcount = cpu_to_be32(len);
|
|
|
|
return xfs_btree_update(cur, &rec);
|
|
|
|
}
|
2008-10-30 13:56:09 +08:00
|
|
|
|
2023-04-12 10:00:01 +08:00
|
|
|
/* Convert the ondisk btree record to its incore representation. */
|
|
|
|
void
|
|
|
|
xfs_alloc_btrec_to_irec(
|
|
|
|
const union xfs_btree_rec *rec,
|
|
|
|
struct xfs_alloc_rec_incore *irec)
|
|
|
|
{
|
|
|
|
irec->ar_startblock = be32_to_cpu(rec->alloc.ar_startblock);
|
|
|
|
irec->ar_blockcount = be32_to_cpu(rec->alloc.ar_blockcount);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Simple checks for free space records. */
|
|
|
|
xfs_failaddr_t
|
|
|
|
xfs_alloc_check_irec(
|
|
|
|
struct xfs_btree_cur *cur,
|
|
|
|
const struct xfs_alloc_rec_incore *irec)
|
|
|
|
{
|
|
|
|
struct xfs_perag *pag = cur->bc_ag.pag;
|
|
|
|
|
|
|
|
if (irec->ar_blockcount == 0)
|
|
|
|
return __this_address;
|
|
|
|
|
|
|
|
/* check for valid extent range, including overflow */
|
|
|
|
if (!xfs_verify_agbext(pag, irec->ar_startblock, irec->ar_blockcount))
|
|
|
|
return __this_address;
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2023-04-12 10:00:04 +08:00
|
|
|
static inline int
|
|
|
|
xfs_alloc_complain_bad_rec(
|
|
|
|
struct xfs_btree_cur *cur,
|
|
|
|
xfs_failaddr_t fa,
|
|
|
|
const struct xfs_alloc_rec_incore *irec)
|
|
|
|
{
|
|
|
|
struct xfs_mount *mp = cur->bc_mp;
|
|
|
|
|
|
|
|
xfs_warn(mp,
|
|
|
|
"%s Freespace BTree record corruption in AG %d detected at %pS!",
|
|
|
|
cur->bc_btnum == XFS_BTNUM_BNO ? "Block" : "Size",
|
|
|
|
cur->bc_ag.pag->pag_agno, fa);
|
|
|
|
xfs_warn(mp,
|
|
|
|
"start block 0x%x block count 0x%x", irec->ar_startblock,
|
|
|
|
irec->ar_blockcount);
|
|
|
|
return -EFSCORRUPTED;
|
|
|
|
}
|
|
|
|
|
2008-10-30 13:58:11 +08:00
|
|
|
/*
|
|
|
|
* Get the data from the pointed-to record.
|
|
|
|
*/
|
2011-01-07 21:02:04 +08:00
|
|
|
int /* error */
|
2008-10-30 13:58:11 +08:00
|
|
|
xfs_alloc_get_rec(
|
|
|
|
struct xfs_btree_cur *cur, /* btree cursor */
|
|
|
|
xfs_agblock_t *bno, /* output: starting block of extent */
|
|
|
|
xfs_extlen_t *len, /* output: length of extent */
|
|
|
|
int *stat) /* output: success/failure */
|
|
|
|
{
|
2023-04-12 10:00:01 +08:00
|
|
|
struct xfs_alloc_rec_incore irec;
|
2008-10-30 13:58:11 +08:00
|
|
|
union xfs_btree_rec *rec;
|
2023-04-12 10:00:01 +08:00
|
|
|
xfs_failaddr_t fa;
|
2008-10-30 13:58:11 +08:00
|
|
|
int error;
|
|
|
|
|
|
|
|
error = xfs_btree_get_rec(cur, &rec, stat);
|
2018-06-04 07:10:14 +08:00
|
|
|
if (error || !(*stat))
|
|
|
|
return error;
|
|
|
|
|
2023-04-12 10:00:01 +08:00
|
|
|
xfs_alloc_btrec_to_irec(rec, &irec);
|
|
|
|
fa = xfs_alloc_check_irec(cur, &irec);
|
|
|
|
if (fa)
|
2023-04-12 10:00:04 +08:00
|
|
|
return xfs_alloc_complain_bad_rec(cur, fa, &irec);
|
2018-06-06 10:42:13 +08:00
|
|
|
|
2023-04-12 10:00:01 +08:00
|
|
|
*bno = irec.ar_startblock;
|
|
|
|
*len = irec.ar_blockcount;
|
2018-06-06 10:42:13 +08:00
|
|
|
return 0;
|
2008-10-30 13:58:11 +08:00
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Compute aligned version of the found extent.
|
|
|
|
* Takes alignment and min length into account.
|
|
|
|
*/
|
2017-02-08 06:06:57 +08:00
|
|
|
STATIC bool
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_alloc_compute_aligned(
|
2011-03-04 20:59:54 +08:00
|
|
|
xfs_alloc_arg_t *args, /* allocation argument structure */
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_agblock_t foundbno, /* starting block in found extent */
|
|
|
|
xfs_extlen_t foundlen, /* length in found extent */
|
|
|
|
xfs_agblock_t *resbno, /* result block number */
|
2017-02-08 06:06:57 +08:00
|
|
|
xfs_extlen_t *reslen, /* result length */
|
|
|
|
unsigned *busy_gen)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2017-02-08 06:06:57 +08:00
|
|
|
xfs_agblock_t bno = foundbno;
|
|
|
|
xfs_extlen_t len = foundlen;
|
xfs: support min/max agbno args in block allocator
The block allocator supports various arguments to tweak block allocation
behavior and set allocation requirements. The sparse inode chunk feature
introduces a new requirement not supported by the current arguments.
Sparse inode allocations must convert or merge into an inode record that
describes a fixed length chunk (64 inodes x inodesize). Full inode chunk
allocations by definition always result in valid inode records. Sparse
chunk allocations are smaller and the associated records can refer to
blocks not owned by the inode chunk. This model can result in invalid
inode records in certain cases.
For example, if a sparse allocation occurs near the start of an AG, the
aligned inode record for that chunk might refer to agbno 0. If an
allocation occurs towards the end of the AG and the AG size is not
aligned, the inode record could refer to blocks beyond the end of the
AG. While neither of these scenarios directly result in corruption, they
both insert invalid inode records and at minimum cause repair to
complain, are unlikely to merge into full chunks over time and set land
mines for other areas of code.
To guarantee sparse inode chunk allocation creates valid inode records,
support the ability to specify an agbno range limit for
XFS_ALLOCTYPE_NEAR_BNO block allocations. The min/max agbno's are
specified in the allocation arguments and limit the block allocation
algorithms to that range. The starting 'agbno' hint is clamped to the
range if the specified agbno is out of range. If no sufficient extent is
available within the range, the allocation fails. For backwards
compatibility, the min/max fields can be initialized to 0 to disable
range limiting (e.g., equivalent to min=0,max=agsize).
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-05-29 06:53:00 +08:00
|
|
|
xfs_extlen_t diff;
|
2017-02-08 06:06:57 +08:00
|
|
|
bool busy;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2011-04-25 03:06:15 +08:00
|
|
|
/* Trim busy sections out of found extent */
|
2017-02-08 06:06:57 +08:00
|
|
|
busy = xfs_extent_busy_trim(args, &bno, &len, busy_gen);
|
2011-04-25 03:06:15 +08:00
|
|
|
|
xfs: support min/max agbno args in block allocator
The block allocator supports various arguments to tweak block allocation
behavior and set allocation requirements. The sparse inode chunk feature
introduces a new requirement not supported by the current arguments.
Sparse inode allocations must convert or merge into an inode record that
describes a fixed length chunk (64 inodes x inodesize). Full inode chunk
allocations by definition always result in valid inode records. Sparse
chunk allocations are smaller and the associated records can refer to
blocks not owned by the inode chunk. This model can result in invalid
inode records in certain cases.
For example, if a sparse allocation occurs near the start of an AG, the
aligned inode record for that chunk might refer to agbno 0. If an
allocation occurs towards the end of the AG and the AG size is not
aligned, the inode record could refer to blocks beyond the end of the
AG. While neither of these scenarios directly result in corruption, they
both insert invalid inode records and at minimum cause repair to
complain, are unlikely to merge into full chunks over time and set land
mines for other areas of code.
To guarantee sparse inode chunk allocation creates valid inode records,
support the ability to specify an agbno range limit for
XFS_ALLOCTYPE_NEAR_BNO block allocations. The min/max agbno's are
specified in the allocation arguments and limit the block allocation
algorithms to that range. The starting 'agbno' hint is clamped to the
range if the specified agbno is out of range. If no sufficient extent is
available within the range, the allocation fails. For backwards
compatibility, the min/max fields can be initialized to 0 to disable
range limiting (e.g., equivalent to min=0,max=agsize).
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-05-29 06:53:00 +08:00
|
|
|
/*
|
|
|
|
* If we have a largish extent that happens to start before min_agbno,
|
|
|
|
* see if we can shift it into range...
|
|
|
|
*/
|
|
|
|
if (bno < args->min_agbno && bno + len > args->min_agbno) {
|
|
|
|
diff = args->min_agbno - bno;
|
|
|
|
if (len > diff) {
|
|
|
|
bno += diff;
|
|
|
|
len -= diff;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-04-25 03:06:15 +08:00
|
|
|
if (args->alignment > 1 && len >= args->minlen) {
|
|
|
|
xfs_agblock_t aligned_bno = roundup(bno, args->alignment);
|
xfs: support min/max agbno args in block allocator
The block allocator supports various arguments to tweak block allocation
behavior and set allocation requirements. The sparse inode chunk feature
introduces a new requirement not supported by the current arguments.
Sparse inode allocations must convert or merge into an inode record that
describes a fixed length chunk (64 inodes x inodesize). Full inode chunk
allocations by definition always result in valid inode records. Sparse
chunk allocations are smaller and the associated records can refer to
blocks not owned by the inode chunk. This model can result in invalid
inode records in certain cases.
For example, if a sparse allocation occurs near the start of an AG, the
aligned inode record for that chunk might refer to agbno 0. If an
allocation occurs towards the end of the AG and the AG size is not
aligned, the inode record could refer to blocks beyond the end of the
AG. While neither of these scenarios directly result in corruption, they
both insert invalid inode records and at minimum cause repair to
complain, are unlikely to merge into full chunks over time and set land
mines for other areas of code.
To guarantee sparse inode chunk allocation creates valid inode records,
support the ability to specify an agbno range limit for
XFS_ALLOCTYPE_NEAR_BNO block allocations. The min/max agbno's are
specified in the allocation arguments and limit the block allocation
algorithms to that range. The starting 'agbno' hint is clamped to the
range if the specified agbno is out of range. If no sufficient extent is
available within the range, the allocation fails. For backwards
compatibility, the min/max fields can be initialized to 0 to disable
range limiting (e.g., equivalent to min=0,max=agsize).
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-05-29 06:53:00 +08:00
|
|
|
|
|
|
|
diff = aligned_bno - bno;
|
2011-04-25 03:06:15 +08:00
|
|
|
|
|
|
|
*resbno = aligned_bno;
|
|
|
|
*reslen = diff >= len ? 0 : len - diff;
|
2005-04-17 06:20:36 +08:00
|
|
|
} else {
|
2011-04-25 03:06:15 +08:00
|
|
|
*resbno = bno;
|
|
|
|
*reslen = len;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2017-02-08 06:06:57 +08:00
|
|
|
|
|
|
|
return busy;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Compute best start block and diff for "near" allocations.
|
|
|
|
* freelen >= wantlen already checked by caller.
|
|
|
|
*/
|
|
|
|
STATIC xfs_extlen_t /* difference value (absolute) */
|
|
|
|
xfs_alloc_compute_diff(
|
|
|
|
xfs_agblock_t wantbno, /* target starting block */
|
|
|
|
xfs_extlen_t wantlen, /* target length */
|
|
|
|
xfs_extlen_t alignment, /* target alignment */
|
xfs: remote attribute blocks aren't really userdata
When adding a new remote attribute, we write the attribute to the
new extent before the allocation transaction is committed. This
means we cannot reuse busy extents as that violates crash
consistency semantics. Hence we currently treat remote attribute
extent allocation like userdata because it has the same overwrite
ordering constraints as userdata.
Unfortunately, this also allows the allocator to incorrectly apply
extent size hints to the remote attribute extent allocation. This
results in interesting failures, such as transaction block
reservation overruns and in-memory inode attribute fork corruption.
To fix this, we need to separate the busy extent reuse configuration
from the userdata configuration. This changes the definition of
XFS_BMAPI_METADATA slightly - it now means that allocation is
metadata and reuse of busy extents is acceptible due to the metadata
ordering semantics of the journal. If this flag is not set, it
means the allocation is that has unordered data writeback, and hence
busy extent reuse is not allowed. It no longer implies the
allocation is for user data, just that the data write will not be
strictly ordered. This matches the semantics for both user data
and remote attribute block allocation.
As such, This patch changes the "userdata" field to a "datatype"
field, and adds a "no busy reuse" flag to the field.
When we detect an unordered data extent allocation, we immediately set
the no reuse flag. We then set the "user data" flags based on the
inode fork we are allocating the extent to. Hence we only set
userdata flags on data fork allocations now and consider attribute
fork remote extents to be an unordered metadata extent.
The result is that remote attribute extents now have the expected
allocation semantics, and the data fork allocation behaviour is
completely unchanged.
It should be noted that there may be other ways to fix this (e.g.
use ordered metadata buffers for the remote attribute extent data
write) but they are more invasive and difficult to validate both
from a design and implementation POV. Hence this patch takes the
simple, obvious route to fixing the problem...
Reported-and-tested-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-09-26 06:21:28 +08:00
|
|
|
int datatype, /* are we allocating data? */
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_agblock_t freebno, /* freespace's starting block */
|
|
|
|
xfs_extlen_t freelen, /* freespace's length */
|
|
|
|
xfs_agblock_t *newbnop) /* result: best start block from free */
|
|
|
|
{
|
|
|
|
xfs_agblock_t freeend; /* end of freespace extent */
|
|
|
|
xfs_agblock_t newbno1; /* return block number */
|
|
|
|
xfs_agblock_t newbno2; /* other new block number */
|
|
|
|
xfs_extlen_t newlen1=0; /* length with newbno1 */
|
|
|
|
xfs_extlen_t newlen2=0; /* length with newbno2 */
|
|
|
|
xfs_agblock_t wantend; /* end of target extent */
|
2019-10-31 03:25:00 +08:00
|
|
|
bool userdata = datatype & XFS_ALLOC_USERDATA;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
ASSERT(freelen >= wantlen);
|
|
|
|
freeend = freebno + freelen;
|
|
|
|
wantend = wantbno + wantlen;
|
2013-04-12 04:09:56 +08:00
|
|
|
/*
|
|
|
|
* We want to allocate from the start of a free extent if it is past
|
|
|
|
* the desired block or if we are allocating user data and the free
|
|
|
|
* extent is before desired block. The second case is there to allow
|
|
|
|
* for contiguous allocation from the remaining free space if the file
|
|
|
|
* grows in the short term.
|
|
|
|
*/
|
|
|
|
if (freebno >= wantbno || (userdata && freeend < wantend)) {
|
2005-04-17 06:20:36 +08:00
|
|
|
if ((newbno1 = roundup(freebno, alignment)) >= freeend)
|
|
|
|
newbno1 = NULLAGBLOCK;
|
|
|
|
} else if (freeend >= wantend && alignment > 1) {
|
|
|
|
newbno1 = roundup(wantbno, alignment);
|
|
|
|
newbno2 = newbno1 - alignment;
|
|
|
|
if (newbno1 >= freeend)
|
|
|
|
newbno1 = NULLAGBLOCK;
|
|
|
|
else
|
|
|
|
newlen1 = XFS_EXTLEN_MIN(wantlen, freeend - newbno1);
|
|
|
|
if (newbno2 < freebno)
|
|
|
|
newbno2 = NULLAGBLOCK;
|
|
|
|
else
|
|
|
|
newlen2 = XFS_EXTLEN_MIN(wantlen, freeend - newbno2);
|
|
|
|
if (newbno1 != NULLAGBLOCK && newbno2 != NULLAGBLOCK) {
|
|
|
|
if (newlen1 < newlen2 ||
|
|
|
|
(newlen1 == newlen2 &&
|
|
|
|
XFS_ABSDIFF(newbno1, wantbno) >
|
|
|
|
XFS_ABSDIFF(newbno2, wantbno)))
|
|
|
|
newbno1 = newbno2;
|
|
|
|
} else if (newbno2 != NULLAGBLOCK)
|
|
|
|
newbno1 = newbno2;
|
|
|
|
} else if (freeend >= wantend) {
|
|
|
|
newbno1 = wantbno;
|
|
|
|
} else if (alignment > 1) {
|
|
|
|
newbno1 = roundup(freeend - wantlen, alignment);
|
|
|
|
if (newbno1 > freeend - wantlen &&
|
|
|
|
newbno1 - alignment >= freebno)
|
|
|
|
newbno1 -= alignment;
|
|
|
|
else if (newbno1 >= freeend)
|
|
|
|
newbno1 = NULLAGBLOCK;
|
|
|
|
} else
|
|
|
|
newbno1 = freeend - wantlen;
|
|
|
|
*newbnop = newbno1;
|
|
|
|
return newbno1 == NULLAGBLOCK ? 0 : XFS_ABSDIFF(newbno1, wantbno);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Fix up the length, based on mod and prod.
|
|
|
|
* len should be k * prod + mod for some k.
|
|
|
|
* If len is too small it is returned unchanged.
|
|
|
|
* If len hits maxlen it is left alone.
|
|
|
|
*/
|
|
|
|
STATIC void
|
|
|
|
xfs_alloc_fix_len(
|
|
|
|
xfs_alloc_arg_t *args) /* allocation argument structure */
|
|
|
|
{
|
|
|
|
xfs_extlen_t k;
|
|
|
|
xfs_extlen_t rlen;
|
|
|
|
|
|
|
|
ASSERT(args->mod < args->prod);
|
|
|
|
rlen = args->len;
|
|
|
|
ASSERT(rlen >= args->minlen);
|
|
|
|
ASSERT(rlen <= args->maxlen);
|
|
|
|
if (args->prod <= 1 || rlen < args->mod || rlen == args->maxlen ||
|
|
|
|
(args->mod == 0 && rlen < args->prod))
|
|
|
|
return;
|
|
|
|
k = rlen % args->prod;
|
|
|
|
if (k == args->mod)
|
|
|
|
return;
|
2014-06-06 14:06:37 +08:00
|
|
|
if (k > args->mod)
|
|
|
|
rlen = rlen - (k - args->mod);
|
|
|
|
else
|
|
|
|
rlen = rlen - args->prod + (args->mod - k);
|
2015-02-24 07:16:04 +08:00
|
|
|
/* casts to (int) catch length underflows */
|
2014-06-06 14:06:37 +08:00
|
|
|
if ((int)rlen < (int)args->minlen)
|
|
|
|
return;
|
|
|
|
ASSERT(rlen >= args->minlen && rlen <= args->maxlen);
|
|
|
|
ASSERT(rlen % args->prod == args->mod);
|
2017-01-10 05:44:30 +08:00
|
|
|
ASSERT(args->pag->pagf_freeblks + args->pag->pagf_flcount >=
|
|
|
|
rlen + args->minleft);
|
2005-04-17 06:20:36 +08:00
|
|
|
args->len = rlen;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Update the two btrees, logically removing from freespace the extent
|
|
|
|
* starting at rbno, rlen blocks. The extent is contained within the
|
|
|
|
* actual (current) free extent fbno for flen blocks.
|
|
|
|
* Flags are passed in indicating whether the cursors are set to the
|
|
|
|
* relevant records.
|
|
|
|
*/
|
|
|
|
STATIC int /* error code */
|
|
|
|
xfs_alloc_fixup_trees(
|
2021-09-17 03:18:47 +08:00
|
|
|
struct xfs_btree_cur *cnt_cur, /* cursor for by-size btree */
|
|
|
|
struct xfs_btree_cur *bno_cur, /* cursor for by-block btree */
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_agblock_t fbno, /* starting block of free extent */
|
|
|
|
xfs_extlen_t flen, /* length of free extent */
|
|
|
|
xfs_agblock_t rbno, /* starting block of returned extent */
|
|
|
|
xfs_extlen_t rlen, /* length of returned extent */
|
|
|
|
int flags) /* flags, XFSA_FIXUP_... */
|
|
|
|
{
|
|
|
|
int error; /* error code */
|
|
|
|
int i; /* operation results */
|
|
|
|
xfs_agblock_t nfbno1; /* first new free startblock */
|
|
|
|
xfs_agblock_t nfbno2; /* second new free startblock */
|
|
|
|
xfs_extlen_t nflen1=0; /* first new free length */
|
|
|
|
xfs_extlen_t nflen2=0; /* second new free length */
|
2015-02-23 19:39:13 +08:00
|
|
|
struct xfs_mount *mp;
|
|
|
|
|
|
|
|
mp = cnt_cur->bc_mp;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Look up the record in the by-size tree if necessary.
|
|
|
|
*/
|
|
|
|
if (flags & XFSA_FIXUP_CNT_OK) {
|
|
|
|
#ifdef DEBUG
|
|
|
|
if ((error = xfs_alloc_get_rec(cnt_cur, &nfbno1, &nflen1, &i)))
|
|
|
|
return error;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp,
|
|
|
|
i != 1 ||
|
|
|
|
nfbno1 != fbno ||
|
|
|
|
nflen1 != flen))
|
|
|
|
return -EFSCORRUPTED;
|
2005-04-17 06:20:36 +08:00
|
|
|
#endif
|
|
|
|
} else {
|
|
|
|
if ((error = xfs_alloc_lookup_eq(cnt_cur, fbno, flen, &i)))
|
|
|
|
return error;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1))
|
|
|
|
return -EFSCORRUPTED;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Look up the record in the by-block tree if necessary.
|
|
|
|
*/
|
|
|
|
if (flags & XFSA_FIXUP_BNO_OK) {
|
|
|
|
#ifdef DEBUG
|
|
|
|
if ((error = xfs_alloc_get_rec(bno_cur, &nfbno1, &nflen1, &i)))
|
|
|
|
return error;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp,
|
|
|
|
i != 1 ||
|
|
|
|
nfbno1 != fbno ||
|
|
|
|
nflen1 != flen))
|
|
|
|
return -EFSCORRUPTED;
|
2005-04-17 06:20:36 +08:00
|
|
|
#endif
|
|
|
|
} else {
|
|
|
|
if ((error = xfs_alloc_lookup_eq(bno_cur, fbno, flen, &i)))
|
|
|
|
return error;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1))
|
|
|
|
return -EFSCORRUPTED;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2008-10-30 14:14:34 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
#ifdef DEBUG
|
2008-10-30 14:14:34 +08:00
|
|
|
if (bno_cur->bc_nlevels == 1 && cnt_cur->bc_nlevels == 1) {
|
|
|
|
struct xfs_btree_block *bnoblock;
|
|
|
|
struct xfs_btree_block *cntblock;
|
|
|
|
|
2021-09-17 03:24:04 +08:00
|
|
|
bnoblock = XFS_BUF_TO_BLOCK(bno_cur->bc_levels[0].bp);
|
|
|
|
cntblock = XFS_BUF_TO_BLOCK(cnt_cur->bc_levels[0].bp);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp,
|
|
|
|
bnoblock->bb_numrecs !=
|
|
|
|
cntblock->bb_numrecs))
|
|
|
|
return -EFSCORRUPTED;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
#endif
|
2008-10-30 14:14:34 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Deal with all four cases: the allocated record is contained
|
|
|
|
* within the freespace record, so we can have new freespace
|
|
|
|
* at either (or both) end, or no freespace remaining.
|
|
|
|
*/
|
|
|
|
if (rbno == fbno && rlen == flen)
|
|
|
|
nfbno1 = nfbno2 = NULLAGBLOCK;
|
|
|
|
else if (rbno == fbno) {
|
|
|
|
nfbno1 = rbno + rlen;
|
|
|
|
nflen1 = flen - rlen;
|
|
|
|
nfbno2 = NULLAGBLOCK;
|
|
|
|
} else if (rbno + rlen == fbno + flen) {
|
|
|
|
nfbno1 = fbno;
|
|
|
|
nflen1 = flen - rlen;
|
|
|
|
nfbno2 = NULLAGBLOCK;
|
|
|
|
} else {
|
|
|
|
nfbno1 = fbno;
|
|
|
|
nflen1 = rbno - fbno;
|
|
|
|
nfbno2 = rbno + rlen;
|
|
|
|
nflen2 = (fbno + flen) - nfbno2;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Delete the entry from the by-size btree.
|
|
|
|
*/
|
2008-10-30 13:58:01 +08:00
|
|
|
if ((error = xfs_btree_delete(cnt_cur, &i)))
|
2005-04-17 06:20:36 +08:00
|
|
|
return error;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1))
|
|
|
|
return -EFSCORRUPTED;
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Add new by-size btree entry(s).
|
|
|
|
*/
|
|
|
|
if (nfbno1 != NULLAGBLOCK) {
|
|
|
|
if ((error = xfs_alloc_lookup_eq(cnt_cur, nfbno1, nflen1, &i)))
|
|
|
|
return error;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 0))
|
|
|
|
return -EFSCORRUPTED;
|
2008-10-30 13:57:40 +08:00
|
|
|
if ((error = xfs_btree_insert(cnt_cur, &i)))
|
2005-04-17 06:20:36 +08:00
|
|
|
return error;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1))
|
|
|
|
return -EFSCORRUPTED;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
if (nfbno2 != NULLAGBLOCK) {
|
|
|
|
if ((error = xfs_alloc_lookup_eq(cnt_cur, nfbno2, nflen2, &i)))
|
|
|
|
return error;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 0))
|
|
|
|
return -EFSCORRUPTED;
|
2008-10-30 13:57:40 +08:00
|
|
|
if ((error = xfs_btree_insert(cnt_cur, &i)))
|
2005-04-17 06:20:36 +08:00
|
|
|
return error;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1))
|
|
|
|
return -EFSCORRUPTED;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Fix up the by-block btree entry(s).
|
|
|
|
*/
|
|
|
|
if (nfbno1 == NULLAGBLOCK) {
|
|
|
|
/*
|
|
|
|
* No remaining freespace, just delete the by-block tree entry.
|
|
|
|
*/
|
2008-10-30 13:58:01 +08:00
|
|
|
if ((error = xfs_btree_delete(bno_cur, &i)))
|
2005-04-17 06:20:36 +08:00
|
|
|
return error;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1))
|
|
|
|
return -EFSCORRUPTED;
|
2005-04-17 06:20:36 +08:00
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* Update the by-block entry to start later|be shorter.
|
|
|
|
*/
|
|
|
|
if ((error = xfs_alloc_update(bno_cur, nfbno1, nflen1)))
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
if (nfbno2 != NULLAGBLOCK) {
|
|
|
|
/*
|
|
|
|
* 2 resulting free entries, need to add one.
|
|
|
|
*/
|
|
|
|
if ((error = xfs_alloc_lookup_eq(bno_cur, nfbno2, nflen2, &i)))
|
|
|
|
return error;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 0))
|
|
|
|
return -EFSCORRUPTED;
|
2008-10-30 13:57:40 +08:00
|
|
|
if ((error = xfs_btree_insert(bno_cur, &i)))
|
2005-04-17 06:20:36 +08:00
|
|
|
return error;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1))
|
|
|
|
return -EFSCORRUPTED;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2023-06-05 12:48:15 +08:00
|
|
|
/*
|
|
|
|
* We do not verify the AGFL contents against AGF-based index counters here,
|
|
|
|
* even though we may have access to the perag that contains shadow copies. We
|
|
|
|
* don't know if the AGF based counters have been checked, and if they have they
|
|
|
|
* still may be inconsistent because they haven't yet been reset on the first
|
|
|
|
* allocation after the AGF has been read in.
|
|
|
|
*
|
|
|
|
* This means we can only check that all agfl entries contain valid or null
|
|
|
|
* values because we can't reliably determine the active range to exclude
|
|
|
|
* NULLAGBNO as a valid value.
|
|
|
|
*
|
|
|
|
* However, we can't even do that for v4 format filesystems because there are
|
|
|
|
* old versions of mkfs out there that does not initialise the AGFL to known,
|
|
|
|
* verifiable values. HEnce we can't tell the difference between a AGFL block
|
|
|
|
* allocated by mkfs and a corrupted AGFL block here on v4 filesystems.
|
|
|
|
*
|
|
|
|
* As a result, we can only fully validate AGFL block numbers when we pull them
|
|
|
|
* from the freelist in xfs_alloc_get_freelist().
|
|
|
|
*/
|
2018-01-09 02:51:03 +08:00
|
|
|
static xfs_failaddr_t
|
2012-11-14 14:52:32 +08:00
|
|
|
xfs_agfl_verify(
|
2012-11-12 19:54:06 +08:00
|
|
|
struct xfs_buf *bp)
|
|
|
|
{
|
2019-06-29 10:27:29 +08:00
|
|
|
struct xfs_mount *mp = bp->b_mount;
|
2012-11-12 19:54:06 +08:00
|
|
|
struct xfs_agfl *agfl = XFS_BUF_TO_AGFL(bp);
|
2020-03-10 23:57:28 +08:00
|
|
|
__be32 *agfl_bno = xfs_buf_to_agfl_bno(bp);
|
2012-11-12 19:54:06 +08:00
|
|
|
int i;
|
|
|
|
|
2021-08-19 09:46:37 +08:00
|
|
|
if (!xfs_has_crc(mp))
|
2018-01-09 02:51:08 +08:00
|
|
|
return NULL;
|
|
|
|
|
2019-02-08 02:45:48 +08:00
|
|
|
if (!xfs_verify_magic(bp, agfl->agfl_magicnum))
|
2018-01-09 02:51:03 +08:00
|
|
|
return __this_address;
|
2019-02-08 02:45:48 +08:00
|
|
|
if (!uuid_equal(&agfl->agfl_uuid, &mp->m_sb.sb_meta_uuid))
|
2018-01-09 02:51:03 +08:00
|
|
|
return __this_address;
|
2013-04-03 13:11:14 +08:00
|
|
|
/*
|
|
|
|
* during growfs operations, the perag is not fully initialised,
|
|
|
|
* so we can't use it for any useful checking. growfs ensures we can't
|
|
|
|
* use it by using uncached buffers that don't have the perag attached
|
|
|
|
* so we can detect and avoid this problem.
|
|
|
|
*/
|
|
|
|
if (bp->b_pag && be32_to_cpu(agfl->agfl_seqno) != bp->b_pag->pag_agno)
|
2018-01-09 02:51:03 +08:00
|
|
|
return __this_address;
|
2013-04-03 13:11:14 +08:00
|
|
|
|
2018-03-07 09:08:32 +08:00
|
|
|
for (i = 0; i < xfs_agfl_size(mp); i++) {
|
2020-03-10 23:57:28 +08:00
|
|
|
if (be32_to_cpu(agfl_bno[i]) != NULLAGBLOCK &&
|
|
|
|
be32_to_cpu(agfl_bno[i]) >= mp->m_sb.sb_agblocks)
|
2018-01-09 02:51:03 +08:00
|
|
|
return __this_address;
|
2012-11-12 19:54:06 +08:00
|
|
|
}
|
2015-10-12 12:59:25 +08:00
|
|
|
|
2018-01-09 02:51:03 +08:00
|
|
|
if (!xfs_log_check_lsn(mp, be64_to_cpu(XFS_BUF_TO_AGFL(bp)->agfl_lsn)))
|
|
|
|
return __this_address;
|
|
|
|
return NULL;
|
2013-04-03 13:11:14 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
xfs_agfl_read_verify(
|
|
|
|
struct xfs_buf *bp)
|
|
|
|
{
|
2019-06-29 10:27:29 +08:00
|
|
|
struct xfs_mount *mp = bp->b_mount;
|
2018-01-09 02:51:03 +08:00
|
|
|
xfs_failaddr_t fa;
|
2013-04-03 13:11:14 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* There is no verification of non-crc AGFLs because mkfs does not
|
|
|
|
* initialise the AGFL to zero or NULL. Hence the only valid part of the
|
|
|
|
* AGFL is what the AGF says is active. We can't get to the AGF, so we
|
|
|
|
* can't verify just those entries are valid.
|
|
|
|
*/
|
2021-08-19 09:46:37 +08:00
|
|
|
if (!xfs_has_crc(mp))
|
2013-04-03 13:11:14 +08:00
|
|
|
return;
|
|
|
|
|
2014-02-27 12:23:10 +08:00
|
|
|
if (!xfs_buf_verify_cksum(bp, XFS_AGFL_CRC_OFF))
|
2018-01-09 02:51:03 +08:00
|
|
|
xfs_verifier_error(bp, -EFSBADCRC, __this_address);
|
|
|
|
else {
|
|
|
|
fa = xfs_agfl_verify(bp);
|
|
|
|
if (fa)
|
|
|
|
xfs_verifier_error(bp, -EFSCORRUPTED, fa);
|
|
|
|
}
|
2012-11-14 14:52:32 +08:00
|
|
|
}
|
|
|
|
|
2012-11-14 14:54:40 +08:00
|
|
|
static void
|
2012-11-14 14:52:32 +08:00
|
|
|
xfs_agfl_write_verify(
|
|
|
|
struct xfs_buf *bp)
|
|
|
|
{
|
2019-06-29 10:27:29 +08:00
|
|
|
struct xfs_mount *mp = bp->b_mount;
|
2018-01-25 05:38:48 +08:00
|
|
|
struct xfs_buf_log_item *bip = bp->b_log_item;
|
2018-01-09 02:51:03 +08:00
|
|
|
xfs_failaddr_t fa;
|
2012-11-14 14:52:32 +08:00
|
|
|
|
2013-04-03 13:11:14 +08:00
|
|
|
/* no verification of non-crc AGFLs */
|
2021-08-19 09:46:37 +08:00
|
|
|
if (!xfs_has_crc(mp))
|
2013-04-03 13:11:14 +08:00
|
|
|
return;
|
|
|
|
|
2018-01-09 02:51:03 +08:00
|
|
|
fa = xfs_agfl_verify(bp);
|
|
|
|
if (fa) {
|
|
|
|
xfs_verifier_error(bp, -EFSCORRUPTED, fa);
|
2013-04-03 13:11:14 +08:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (bip)
|
|
|
|
XFS_BUF_TO_AGFL(bp)->agfl_lsn = cpu_to_be64(bip->bli_item.li_lsn);
|
|
|
|
|
2014-02-27 12:18:23 +08:00
|
|
|
xfs_buf_update_cksum(bp, XFS_AGFL_CRC_OFF);
|
2012-11-12 19:54:06 +08:00
|
|
|
}
|
|
|
|
|
2012-11-14 14:54:40 +08:00
|
|
|
const struct xfs_buf_ops xfs_agfl_buf_ops = {
|
2016-01-04 13:10:19 +08:00
|
|
|
.name = "xfs_agfl",
|
2019-02-08 02:45:48 +08:00
|
|
|
.magic = { cpu_to_be32(XFS_AGFL_MAGIC), cpu_to_be32(XFS_AGFL_MAGIC) },
|
2012-11-14 14:54:40 +08:00
|
|
|
.verify_read = xfs_agfl_read_verify,
|
|
|
|
.verify_write = xfs_agfl_write_verify,
|
2018-01-09 02:51:08 +08:00
|
|
|
.verify_struct = xfs_agfl_verify,
|
2012-11-14 14:54:40 +08:00
|
|
|
};
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Read in the allocation group free block array.
|
|
|
|
*/
|
2022-07-07 17:08:15 +08:00
|
|
|
int
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_alloc_read_agfl(
|
2022-07-07 17:08:15 +08:00
|
|
|
struct xfs_perag *pag,
|
|
|
|
struct xfs_trans *tp,
|
|
|
|
struct xfs_buf **bpp)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2022-07-07 17:08:15 +08:00
|
|
|
struct xfs_mount *mp = pag->pag_mount;
|
|
|
|
struct xfs_buf *bp;
|
|
|
|
int error;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
error = xfs_trans_read_buf(
|
|
|
|
mp, tp, mp->m_ddev_targp,
|
2022-07-07 17:08:15 +08:00
|
|
|
XFS_AG_DADDR(mp, pag->pag_agno, XFS_AGFL_DADDR(mp)),
|
2012-11-14 14:54:40 +08:00
|
|
|
XFS_FSS_TO_BB(mp, 1), 0, &bp, &xfs_agfl_buf_ops);
|
2005-04-17 06:20:36 +08:00
|
|
|
if (error)
|
|
|
|
return error;
|
2011-10-11 00:52:45 +08:00
|
|
|
xfs_buf_set_ref(bp, XFS_AGFL_REF);
|
2005-04-17 06:20:36 +08:00
|
|
|
*bpp = bp;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2011-03-04 20:59:55 +08:00
|
|
|
STATIC int
|
|
|
|
xfs_alloc_update_counters(
|
|
|
|
struct xfs_trans *tp,
|
|
|
|
struct xfs_buf *agbp,
|
|
|
|
long len)
|
|
|
|
{
|
2020-03-10 23:57:29 +08:00
|
|
|
struct xfs_agf *agf = agbp->b_addr;
|
2011-03-04 20:59:55 +08:00
|
|
|
|
2020-07-14 00:13:00 +08:00
|
|
|
agbp->b_pag->pagf_freeblks += len;
|
2011-03-04 20:59:55 +08:00
|
|
|
be32_add_cpu(&agf->agf_freeblks, len);
|
|
|
|
|
|
|
|
if (unlikely(be32_to_cpu(agf->agf_freeblks) >
|
2019-11-03 00:40:53 +08:00
|
|
|
be32_to_cpu(agf->agf_length))) {
|
2020-03-12 01:37:54 +08:00
|
|
|
xfs_buf_mark_corrupt(agbp);
|
2014-06-25 12:58:08 +08:00
|
|
|
return -EFSCORRUPTED;
|
2019-11-03 00:40:53 +08:00
|
|
|
}
|
2011-03-04 20:59:55 +08:00
|
|
|
|
|
|
|
xfs_alloc_log_agf(tp, agbp, XFS_AGF_FREEBLKS);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
2019-10-14 08:10:31 +08:00
|
|
|
* Block allocation algorithm and data structures.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2019-10-14 08:10:31 +08:00
|
|
|
struct xfs_alloc_cur {
|
|
|
|
struct xfs_btree_cur *cnt; /* btree cursors */
|
|
|
|
struct xfs_btree_cur *bnolt;
|
|
|
|
struct xfs_btree_cur *bnogt;
|
2019-10-14 08:10:36 +08:00
|
|
|
xfs_extlen_t cur_len;/* current search length */
|
2019-10-14 08:10:32 +08:00
|
|
|
xfs_agblock_t rec_bno;/* extent startblock */
|
|
|
|
xfs_extlen_t rec_len;/* extent length */
|
|
|
|
xfs_agblock_t bno; /* alloc bno */
|
|
|
|
xfs_extlen_t len; /* alloc len */
|
|
|
|
xfs_extlen_t diff; /* diff from search bno */
|
2019-10-14 08:10:32 +08:00
|
|
|
unsigned int busy_gen;/* busy state */
|
|
|
|
bool busy;
|
2019-10-14 08:10:31 +08:00
|
|
|
};
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Set up cursors, etc. in the extent allocation cursor. This function can be
|
|
|
|
* called multiple times to reset an initialized structure without having to
|
|
|
|
* reallocate cursors.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
xfs_alloc_cur_setup(
|
|
|
|
struct xfs_alloc_arg *args,
|
|
|
|
struct xfs_alloc_cur *acur)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
int i;
|
|
|
|
|
2019-10-14 08:10:36 +08:00
|
|
|
acur->cur_len = args->maxlen;
|
2019-10-14 08:10:32 +08:00
|
|
|
acur->rec_bno = 0;
|
|
|
|
acur->rec_len = 0;
|
|
|
|
acur->bno = 0;
|
|
|
|
acur->len = 0;
|
2019-10-14 08:10:33 +08:00
|
|
|
acur->diff = -1;
|
2019-10-14 08:10:32 +08:00
|
|
|
acur->busy = false;
|
|
|
|
acur->busy_gen = 0;
|
|
|
|
|
2019-10-14 08:10:31 +08:00
|
|
|
/*
|
|
|
|
* Perform an initial cntbt lookup to check for availability of maxlen
|
|
|
|
* extents. If this fails, we'll return -ENOSPC to signal the caller to
|
|
|
|
* attempt a small allocation.
|
|
|
|
*/
|
|
|
|
if (!acur->cnt)
|
|
|
|
acur->cnt = xfs_allocbt_init_cursor(args->mp, args->tp,
|
2021-06-02 08:48:24 +08:00
|
|
|
args->agbp, args->pag, XFS_BTNUM_CNT);
|
2019-10-14 08:10:31 +08:00
|
|
|
error = xfs_alloc_lookup_ge(acur->cnt, 0, args->maxlen, &i);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Allocate the bnobt left and right search cursors.
|
|
|
|
*/
|
|
|
|
if (!acur->bnolt)
|
|
|
|
acur->bnolt = xfs_allocbt_init_cursor(args->mp, args->tp,
|
2021-06-02 08:48:24 +08:00
|
|
|
args->agbp, args->pag, XFS_BTNUM_BNO);
|
2019-10-14 08:10:31 +08:00
|
|
|
if (!acur->bnogt)
|
|
|
|
acur->bnogt = xfs_allocbt_init_cursor(args->mp, args->tp,
|
2021-06-02 08:48:24 +08:00
|
|
|
args->agbp, args->pag, XFS_BTNUM_BNO);
|
2019-10-14 08:10:31 +08:00
|
|
|
return i == 1 ? 0 : -ENOSPC;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
xfs_alloc_cur_close(
|
|
|
|
struct xfs_alloc_cur *acur,
|
|
|
|
bool error)
|
|
|
|
{
|
|
|
|
int cur_error = XFS_BTREE_NOERROR;
|
|
|
|
|
|
|
|
if (error)
|
|
|
|
cur_error = XFS_BTREE_ERROR;
|
|
|
|
|
|
|
|
if (acur->cnt)
|
|
|
|
xfs_btree_del_cursor(acur->cnt, cur_error);
|
|
|
|
if (acur->bnolt)
|
|
|
|
xfs_btree_del_cursor(acur->bnolt, cur_error);
|
|
|
|
if (acur->bnogt)
|
|
|
|
xfs_btree_del_cursor(acur->bnogt, cur_error);
|
|
|
|
acur->cnt = acur->bnolt = acur->bnogt = NULL;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2019-10-14 08:10:33 +08:00
|
|
|
/*
|
|
|
|
* Check an extent for allocation and track the best available candidate in the
|
|
|
|
* allocation structure. The cursor is deactivated if it has entered an out of
|
|
|
|
* range state based on allocation arguments. Optionally return the extent
|
|
|
|
* extent geometry and allocation status if requested by the caller.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
xfs_alloc_cur_check(
|
|
|
|
struct xfs_alloc_arg *args,
|
|
|
|
struct xfs_alloc_cur *acur,
|
|
|
|
struct xfs_btree_cur *cur,
|
|
|
|
int *new)
|
|
|
|
{
|
|
|
|
int error, i;
|
|
|
|
xfs_agblock_t bno, bnoa, bnew;
|
|
|
|
xfs_extlen_t len, lena, diff = -1;
|
|
|
|
bool busy;
|
|
|
|
unsigned busy_gen = 0;
|
|
|
|
bool deactivate = false;
|
2019-10-14 08:10:33 +08:00
|
|
|
bool isbnobt = cur->bc_btnum == XFS_BTNUM_BNO;
|
2019-10-14 08:10:33 +08:00
|
|
|
|
|
|
|
*new = 0;
|
|
|
|
|
|
|
|
error = xfs_alloc_get_rec(cur, &bno, &len, &i);
|
|
|
|
if (error)
|
|
|
|
return error;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(args->mp, i != 1))
|
|
|
|
return -EFSCORRUPTED;
|
2019-10-14 08:10:33 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Check minlen and deactivate a cntbt cursor if out of acceptable size
|
|
|
|
* range (i.e., walking backwards looking for a minlen extent).
|
|
|
|
*/
|
|
|
|
if (len < args->minlen) {
|
2019-10-14 08:10:33 +08:00
|
|
|
deactivate = !isbnobt;
|
2019-10-14 08:10:33 +08:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
busy = xfs_alloc_compute_aligned(args, bno, len, &bnoa, &lena,
|
|
|
|
&busy_gen);
|
|
|
|
acur->busy |= busy;
|
|
|
|
if (busy)
|
|
|
|
acur->busy_gen = busy_gen;
|
|
|
|
/* deactivate a bnobt cursor outside of locality range */
|
2019-10-14 08:10:33 +08:00
|
|
|
if (bnoa < args->min_agbno || bnoa > args->max_agbno) {
|
|
|
|
deactivate = isbnobt;
|
2019-10-14 08:10:33 +08:00
|
|
|
goto out;
|
2019-10-14 08:10:33 +08:00
|
|
|
}
|
2019-10-14 08:10:33 +08:00
|
|
|
if (lena < args->minlen)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
args->len = XFS_EXTLEN_MIN(lena, args->maxlen);
|
|
|
|
xfs_alloc_fix_len(args);
|
|
|
|
ASSERT(args->len >= args->minlen);
|
|
|
|
if (args->len < acur->len)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We have an aligned record that satisfies minlen and beats or matches
|
|
|
|
* the candidate extent size. Compare locality for near allocation mode.
|
|
|
|
*/
|
|
|
|
diff = xfs_alloc_compute_diff(args->agbno, args->len,
|
|
|
|
args->alignment, args->datatype,
|
|
|
|
bnoa, lena, &bnew);
|
|
|
|
if (bnew == NULLAGBLOCK)
|
|
|
|
goto out;
|
2019-10-14 08:10:33 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Deactivate a bnobt cursor with worse locality than the current best.
|
|
|
|
*/
|
|
|
|
if (diff > acur->diff) {
|
|
|
|
deactivate = isbnobt;
|
2019-10-14 08:10:33 +08:00
|
|
|
goto out;
|
2019-10-14 08:10:33 +08:00
|
|
|
}
|
2019-10-14 08:10:33 +08:00
|
|
|
|
|
|
|
ASSERT(args->len > acur->len ||
|
|
|
|
(args->len == acur->len && diff <= acur->diff));
|
|
|
|
acur->rec_bno = bno;
|
|
|
|
acur->rec_len = len;
|
|
|
|
acur->bno = bnew;
|
|
|
|
acur->len = args->len;
|
|
|
|
acur->diff = diff;
|
|
|
|
*new = 1;
|
|
|
|
|
2019-10-14 08:10:34 +08:00
|
|
|
/*
|
|
|
|
* We're done if we found a perfect allocation. This only deactivates
|
|
|
|
* the current cursor, but this is just an optimization to terminate a
|
|
|
|
* cntbt search that otherwise runs to the edge of the tree.
|
|
|
|
*/
|
|
|
|
if (acur->diff == 0 && acur->len == args->maxlen)
|
|
|
|
deactivate = true;
|
2019-10-14 08:10:33 +08:00
|
|
|
out:
|
|
|
|
if (deactivate)
|
2020-03-11 08:57:51 +08:00
|
|
|
cur->bc_ag.abt.active = false;
|
2019-10-14 08:10:33 +08:00
|
|
|
trace_xfs_alloc_cur_check(args->mp, cur->bc_btnum, bno, len, diff,
|
|
|
|
*new);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2019-10-14 08:10:35 +08:00
|
|
|
/*
|
|
|
|
* Complete an allocation of a candidate extent. Remove the extent from both
|
|
|
|
* trees and update the args structure.
|
|
|
|
*/
|
|
|
|
STATIC int
|
|
|
|
xfs_alloc_cur_finish(
|
|
|
|
struct xfs_alloc_arg *args,
|
|
|
|
struct xfs_alloc_cur *acur)
|
|
|
|
{
|
2020-03-10 23:57:29 +08:00
|
|
|
struct xfs_agf __maybe_unused *agf = args->agbp->b_addr;
|
2019-10-14 08:10:35 +08:00
|
|
|
int error;
|
|
|
|
|
|
|
|
ASSERT(acur->cnt && acur->bnolt);
|
|
|
|
ASSERT(acur->bno >= acur->rec_bno);
|
|
|
|
ASSERT(acur->bno + acur->len <= acur->rec_bno + acur->rec_len);
|
2020-03-10 23:57:29 +08:00
|
|
|
ASSERT(acur->rec_bno + acur->rec_len <= be32_to_cpu(agf->agf_length));
|
2019-10-14 08:10:35 +08:00
|
|
|
|
|
|
|
error = xfs_alloc_fixup_trees(acur->cnt, acur->bnolt, acur->rec_bno,
|
|
|
|
acur->rec_len, acur->bno, acur->len, 0);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
|
|
|
args->agbno = acur->bno;
|
|
|
|
args->len = acur->len;
|
|
|
|
args->wasfromfl = 0;
|
|
|
|
|
|
|
|
trace_xfs_alloc_cur(args);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2019-10-14 08:10:36 +08:00
|
|
|
/*
|
|
|
|
* Locality allocation lookup algorithm. This expects a cntbt cursor and uses
|
|
|
|
* bno optimized lookup to search for extents with ideal size and locality.
|
|
|
|
*/
|
|
|
|
STATIC int
|
|
|
|
xfs_alloc_cntbt_iter(
|
|
|
|
struct xfs_alloc_arg *args,
|
|
|
|
struct xfs_alloc_cur *acur)
|
|
|
|
{
|
|
|
|
struct xfs_btree_cur *cur = acur->cnt;
|
|
|
|
xfs_agblock_t bno;
|
|
|
|
xfs_extlen_t len, cur_len;
|
|
|
|
int error;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if (!xfs_alloc_cur_active(cur))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* locality optimized lookup */
|
|
|
|
cur_len = acur->cur_len;
|
|
|
|
error = xfs_alloc_lookup_ge(cur, args->agbno, cur_len, &i);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
if (i == 0)
|
|
|
|
return 0;
|
|
|
|
error = xfs_alloc_get_rec(cur, &bno, &len, &i);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
|
|
|
/* check the current record and update search length from it */
|
|
|
|
error = xfs_alloc_cur_check(args, acur, cur, &i);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
ASSERT(len >= acur->cur_len);
|
|
|
|
acur->cur_len = len;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We looked up the first record >= [agbno, len] above. The agbno is a
|
|
|
|
* secondary key and so the current record may lie just before or after
|
|
|
|
* agbno. If it is past agbno, check the previous record too so long as
|
|
|
|
* the length matches as it may be closer. Don't check a smaller record
|
|
|
|
* because that could deactivate our cursor.
|
|
|
|
*/
|
|
|
|
if (bno > args->agbno) {
|
|
|
|
error = xfs_btree_decrement(cur, 0, &i);
|
|
|
|
if (!error && i) {
|
|
|
|
error = xfs_alloc_get_rec(cur, &bno, &len, &i);
|
|
|
|
if (!error && i && len == acur->cur_len)
|
|
|
|
error = xfs_alloc_cur_check(args, acur, cur,
|
|
|
|
&i);
|
|
|
|
}
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Increment the search key until we find at least one allocation
|
|
|
|
* candidate or if the extent we found was larger. Otherwise, double the
|
|
|
|
* search key to optimize the search. Efficiency is more important here
|
|
|
|
* than absolute best locality.
|
|
|
|
*/
|
|
|
|
cur_len <<= 1;
|
|
|
|
if (!acur->len || acur->cur_len >= cur_len)
|
|
|
|
acur->cur_len++;
|
|
|
|
else
|
|
|
|
acur->cur_len = cur_len;
|
|
|
|
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2019-06-29 10:30:19 +08:00
|
|
|
/*
|
|
|
|
* Deal with the case where only small freespaces remain. Either return the
|
|
|
|
* contents of the last freespace record, or allocate space from the freelist if
|
|
|
|
* there is nothing in the tree.
|
|
|
|
*/
|
|
|
|
STATIC int /* error */
|
|
|
|
xfs_alloc_ag_vextent_small(
|
|
|
|
struct xfs_alloc_arg *args, /* allocation argument structure */
|
|
|
|
struct xfs_btree_cur *ccur, /* optional by-size cursor */
|
|
|
|
xfs_agblock_t *fbnop, /* result block number */
|
|
|
|
xfs_extlen_t *flenp, /* result length */
|
|
|
|
int *stat) /* status: 0-freelist, 1-normal/none */
|
|
|
|
{
|
2020-03-10 23:57:29 +08:00
|
|
|
struct xfs_agf *agf = args->agbp->b_addr;
|
2019-06-29 10:30:19 +08:00
|
|
|
int error = 0;
|
|
|
|
xfs_agblock_t fbno = NULLAGBLOCK;
|
|
|
|
xfs_extlen_t flen = 0;
|
2019-06-29 10:30:20 +08:00
|
|
|
int i = 0;
|
2019-06-29 10:30:19 +08:00
|
|
|
|
2019-06-29 10:30:20 +08:00
|
|
|
/*
|
|
|
|
* If a cntbt cursor is provided, try to allocate the largest record in
|
|
|
|
* the tree. Try the AGFL if the cntbt is empty, otherwise fail the
|
|
|
|
* allocation. Make sure to respect minleft even when pulling from the
|
|
|
|
* freelist.
|
|
|
|
*/
|
|
|
|
if (ccur)
|
|
|
|
error = xfs_btree_decrement(ccur, 0, &i);
|
2019-06-29 10:30:19 +08:00
|
|
|
if (error)
|
|
|
|
goto error;
|
|
|
|
if (i) {
|
|
|
|
error = xfs_alloc_get_rec(ccur, &fbno, &flen, &i);
|
|
|
|
if (error)
|
|
|
|
goto error;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(args->mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error;
|
|
|
|
}
|
2019-06-29 10:30:19 +08:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (args->minlen != 1 || args->alignment != 1 ||
|
|
|
|
args->resv == XFS_AG_RESV_AGFL ||
|
2020-03-10 23:57:29 +08:00
|
|
|
be32_to_cpu(agf->agf_flcount) <= args->minleft)
|
2019-06-29 10:30:19 +08:00
|
|
|
goto out;
|
|
|
|
|
2022-07-07 17:08:01 +08:00
|
|
|
error = xfs_alloc_get_freelist(args->pag, args->tp, args->agbp,
|
|
|
|
&fbno, 0);
|
2019-06-29 10:30:19 +08:00
|
|
|
if (error)
|
|
|
|
goto error;
|
|
|
|
if (fbno == NULLAGBLOCK)
|
|
|
|
goto out;
|
|
|
|
|
2021-06-02 08:48:24 +08:00
|
|
|
xfs_extent_busy_reuse(args->mp, args->pag, fbno, 1,
|
2019-10-31 03:25:00 +08:00
|
|
|
(args->datatype & XFS_ALLOC_NOBUSY));
|
2019-06-29 10:30:19 +08:00
|
|
|
|
2019-10-31 03:25:00 +08:00
|
|
|
if (args->datatype & XFS_ALLOC_USERDATA) {
|
2019-06-29 10:30:19 +08:00
|
|
|
struct xfs_buf *bp;
|
|
|
|
|
2020-01-24 09:01:19 +08:00
|
|
|
error = xfs_trans_get_buf(args->tp, args->mp->m_ddev_targp,
|
|
|
|
XFS_AGB_TO_DADDR(args->mp, args->agno, fbno),
|
|
|
|
args->mp->m_bsize, 0, &bp);
|
|
|
|
if (error)
|
2019-06-29 10:30:19 +08:00
|
|
|
goto error;
|
|
|
|
xfs_trans_binval(args->tp, bp);
|
|
|
|
}
|
2019-06-29 10:30:20 +08:00
|
|
|
*fbnop = args->agbno = fbno;
|
|
|
|
*flenp = args->len = 1;
|
2020-03-10 23:57:29 +08:00
|
|
|
if (XFS_IS_CORRUPT(args->mp, fbno >= be32_to_cpu(agf->agf_length))) {
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error;
|
|
|
|
}
|
2019-06-29 10:30:19 +08:00
|
|
|
args->wasfromfl = 1;
|
|
|
|
trace_xfs_alloc_small_freelist(args);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we're feeding an AGFL block to something that doesn't live in the
|
|
|
|
* free space, we need to clear out the OWN_AG rmap.
|
|
|
|
*/
|
2021-06-02 08:48:24 +08:00
|
|
|
error = xfs_rmap_free(args->tp, args->agbp, args->pag, fbno, 1,
|
2019-06-29 10:30:19 +08:00
|
|
|
&XFS_RMAP_OINFO_AG);
|
|
|
|
if (error)
|
|
|
|
goto error;
|
|
|
|
|
|
|
|
*stat = 0;
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
out:
|
|
|
|
/*
|
|
|
|
* Can't do the allocation, give up.
|
|
|
|
*/
|
|
|
|
if (flen < args->minlen) {
|
|
|
|
args->agbno = NULLAGBLOCK;
|
|
|
|
trace_xfs_alloc_small_notenough(args);
|
|
|
|
flen = 0;
|
|
|
|
}
|
|
|
|
*fbnop = fbno;
|
|
|
|
*flenp = flen;
|
|
|
|
*stat = 1;
|
|
|
|
trace_xfs_alloc_small_done(args);
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
error:
|
|
|
|
trace_xfs_alloc_small_error(args);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Allocate a variable extent at exactly agno/bno.
|
|
|
|
* Extent's length (returned in *len) will be between minlen and maxlen,
|
|
|
|
* and of the form k * prod + mod unless there's nothing that large.
|
|
|
|
* Return the starting a.g. block (bno), or NULLAGBLOCK if we can't do it.
|
|
|
|
*/
|
|
|
|
STATIC int /* error */
|
|
|
|
xfs_alloc_ag_vextent_exact(
|
|
|
|
xfs_alloc_arg_t *args) /* allocation argument structure */
|
|
|
|
{
|
2020-03-10 23:57:29 +08:00
|
|
|
struct xfs_agf __maybe_unused *agf = args->agbp->b_addr;
|
2021-09-17 03:18:47 +08:00
|
|
|
struct xfs_btree_cur *bno_cur;/* by block-number btree cursor */
|
|
|
|
struct xfs_btree_cur *cnt_cur;/* by count btree cursor */
|
2005-04-17 06:20:36 +08:00
|
|
|
int error;
|
|
|
|
xfs_agblock_t fbno; /* start block of found extent */
|
|
|
|
xfs_extlen_t flen; /* length of found extent */
|
2017-02-08 06:06:57 +08:00
|
|
|
xfs_agblock_t tbno; /* start block of busy extent */
|
|
|
|
xfs_extlen_t tlen; /* length of busy extent */
|
|
|
|
xfs_agblock_t tend; /* end block of busy extent */
|
2005-04-17 06:20:36 +08:00
|
|
|
int i; /* success/failure of operation */
|
2017-02-08 06:06:57 +08:00
|
|
|
unsigned busy_gen;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
ASSERT(args->alignment == 1);
|
2010-12-10 23:03:57 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Allocate/initialize a cursor for the by-number freespace btree.
|
|
|
|
*/
|
2008-10-30 13:53:59 +08:00
|
|
|
bno_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
|
2021-06-02 08:48:24 +08:00
|
|
|
args->pag, XFS_BTNUM_BNO);
|
2010-12-10 23:03:57 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Lookup bno and minlen in the btree (minlen is irrelevant, really).
|
|
|
|
* Look for the closest free block <= bno, it must contain bno
|
|
|
|
* if any free block does.
|
|
|
|
*/
|
2010-12-10 23:03:57 +08:00
|
|
|
error = xfs_alloc_lookup_le(bno_cur, args->agbno, args->minlen, &i);
|
|
|
|
if (error)
|
2005-04-17 06:20:36 +08:00
|
|
|
goto error0;
|
2010-12-10 23:03:57 +08:00
|
|
|
if (!i)
|
|
|
|
goto not_found;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Grab the freespace record.
|
|
|
|
*/
|
2010-12-10 23:03:57 +08:00
|
|
|
error = xfs_alloc_get_rec(bno_cur, &fbno, &flen, &i);
|
|
|
|
if (error)
|
2005-04-17 06:20:36 +08:00
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(args->mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
ASSERT(fbno <= args->agbno);
|
2010-12-10 23:03:57 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
2011-04-25 03:06:15 +08:00
|
|
|
* Check for overlapping busy extents.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2017-02-08 06:06:57 +08:00
|
|
|
tbno = fbno;
|
|
|
|
tlen = flen;
|
|
|
|
xfs_extent_busy_trim(args, &tbno, &tlen, &busy_gen);
|
2011-04-25 03:06:15 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Give up if the start of the extent is busy, or the freespace isn't
|
|
|
|
* long enough for the minimum request.
|
|
|
|
*/
|
|
|
|
if (tbno > args->agbno)
|
|
|
|
goto not_found;
|
|
|
|
if (tlen < args->minlen)
|
|
|
|
goto not_found;
|
|
|
|
tend = tbno + tlen;
|
|
|
|
if (tend < args->agbno + args->minlen)
|
2010-12-10 23:03:57 +08:00
|
|
|
goto not_found;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* End of extent will be smaller of the freespace end and the
|
|
|
|
* maximal requested end.
|
2010-12-10 23:03:57 +08:00
|
|
|
*
|
2005-04-17 06:20:36 +08:00
|
|
|
* Fix the length according to mod and prod if given.
|
|
|
|
*/
|
2011-06-10 00:47:49 +08:00
|
|
|
args->len = XFS_AGBLOCK_MIN(tend, args->agbno + args->maxlen)
|
|
|
|
- args->agbno;
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_alloc_fix_len(args);
|
2011-06-10 00:47:49 +08:00
|
|
|
ASSERT(args->agbno + args->len <= tend);
|
2010-12-10 23:03:57 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
2011-06-10 00:47:49 +08:00
|
|
|
* We are allocating agbno for args->len
|
2005-04-17 06:20:36 +08:00
|
|
|
* Allocate/initialize a cursor for the by-size btree.
|
|
|
|
*/
|
2008-10-30 13:53:59 +08:00
|
|
|
cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
|
2021-06-02 08:48:24 +08:00
|
|
|
args->pag, XFS_BTNUM_CNT);
|
2020-03-10 23:57:29 +08:00
|
|
|
ASSERT(args->agbno + args->len <= be32_to_cpu(agf->agf_length));
|
2010-12-10 23:03:57 +08:00
|
|
|
error = xfs_alloc_fixup_trees(cnt_cur, bno_cur, fbno, flen, args->agbno,
|
|
|
|
args->len, XFSA_FIXUP_BNO_OK);
|
|
|
|
if (error) {
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_btree_del_cursor(cnt_cur, XFS_BTREE_ERROR);
|
|
|
|
goto error0;
|
|
|
|
}
|
2010-12-10 23:03:57 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_btree_del_cursor(bno_cur, XFS_BTREE_NOERROR);
|
|
|
|
xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
|
2009-12-15 07:14:59 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
args->wasfromfl = 0;
|
2010-12-10 23:03:57 +08:00
|
|
|
trace_xfs_alloc_exact_done(args);
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
not_found:
|
|
|
|
/* Didn't find it, return null. */
|
|
|
|
xfs_btree_del_cursor(bno_cur, XFS_BTREE_NOERROR);
|
|
|
|
args->agbno = NULLAGBLOCK;
|
|
|
|
trace_xfs_alloc_exact_notfound(args);
|
2005-04-17 06:20:36 +08:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
error0:
|
|
|
|
xfs_btree_del_cursor(bno_cur, XFS_BTREE_ERROR);
|
2009-12-15 07:14:59 +08:00
|
|
|
trace_xfs_alloc_exact_error(args);
|
2005-04-17 06:20:36 +08:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2010-12-10 23:04:11 +08:00
|
|
|
/*
|
2019-10-14 08:10:34 +08:00
|
|
|
* Search a given number of btree records in a given direction. Check each
|
|
|
|
* record against the good extent we've already found.
|
2010-12-10 23:04:11 +08:00
|
|
|
*/
|
|
|
|
STATIC int
|
2019-10-14 08:10:34 +08:00
|
|
|
xfs_alloc_walk_iter(
|
2019-10-14 08:10:33 +08:00
|
|
|
struct xfs_alloc_arg *args,
|
|
|
|
struct xfs_alloc_cur *acur,
|
|
|
|
struct xfs_btree_cur *cur,
|
2019-10-14 08:10:34 +08:00
|
|
|
bool increment,
|
|
|
|
bool find_one, /* quit on first candidate */
|
|
|
|
int count, /* rec count (-1 for infinite) */
|
|
|
|
int *stat)
|
2010-12-10 23:04:11 +08:00
|
|
|
{
|
|
|
|
int error;
|
|
|
|
int i;
|
|
|
|
|
2019-10-14 08:10:34 +08:00
|
|
|
*stat = 0;
|
|
|
|
|
2010-12-10 23:04:11 +08:00
|
|
|
/*
|
2019-10-14 08:10:33 +08:00
|
|
|
* Search so long as the cursor is active or we find a better extent.
|
|
|
|
* The cursor is deactivated if it extends beyond the range of the
|
|
|
|
* current allocation candidate.
|
2010-12-10 23:04:11 +08:00
|
|
|
*/
|
2019-10-14 08:10:34 +08:00
|
|
|
while (xfs_alloc_cur_active(cur) && count) {
|
2019-10-14 08:10:33 +08:00
|
|
|
error = xfs_alloc_cur_check(args, acur, cur, &i);
|
2010-12-10 23:04:11 +08:00
|
|
|
if (error)
|
2019-10-14 08:10:33 +08:00
|
|
|
return error;
|
2019-10-14 08:10:34 +08:00
|
|
|
if (i == 1) {
|
|
|
|
*stat = 1;
|
|
|
|
if (find_one)
|
|
|
|
break;
|
|
|
|
}
|
2019-10-14 08:10:33 +08:00
|
|
|
if (!xfs_alloc_cur_active(cur))
|
|
|
|
break;
|
2010-12-10 23:04:11 +08:00
|
|
|
|
2019-10-14 08:10:33 +08:00
|
|
|
if (increment)
|
|
|
|
error = xfs_btree_increment(cur, 0, &i);
|
2010-12-10 23:04:11 +08:00
|
|
|
else
|
2019-10-14 08:10:33 +08:00
|
|
|
error = xfs_btree_decrement(cur, 0, &i);
|
2010-12-10 23:04:11 +08:00
|
|
|
if (error)
|
2019-10-14 08:10:33 +08:00
|
|
|
return error;
|
|
|
|
if (i == 0)
|
2020-03-11 08:57:51 +08:00
|
|
|
cur->bc_ag.abt.active = false;
|
2019-10-14 08:10:34 +08:00
|
|
|
|
|
|
|
if (count > 0)
|
|
|
|
count--;
|
2019-10-14 08:10:33 +08:00
|
|
|
}
|
2010-12-10 23:04:11 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2019-10-14 08:10:35 +08:00
|
|
|
/*
|
2019-10-14 08:10:36 +08:00
|
|
|
* Search the by-bno and by-size btrees in parallel in search of an extent with
|
|
|
|
* ideal locality based on the NEAR mode ->agbno locality hint.
|
2019-10-14 08:10:35 +08:00
|
|
|
*/
|
|
|
|
STATIC int
|
2019-10-14 08:10:36 +08:00
|
|
|
xfs_alloc_ag_vextent_locality(
|
2019-10-14 08:10:35 +08:00
|
|
|
struct xfs_alloc_arg *args,
|
|
|
|
struct xfs_alloc_cur *acur,
|
|
|
|
int *stat)
|
|
|
|
{
|
|
|
|
struct xfs_btree_cur *fbcur = NULL;
|
|
|
|
int error;
|
|
|
|
int i;
|
|
|
|
bool fbinc;
|
|
|
|
|
|
|
|
ASSERT(acur->len == 0);
|
|
|
|
|
|
|
|
*stat = 0;
|
|
|
|
|
2019-10-14 08:10:36 +08:00
|
|
|
error = xfs_alloc_lookup_ge(acur->cnt, args->agbno, acur->cur_len, &i);
|
|
|
|
if (error)
|
|
|
|
return error;
|
2019-10-14 08:10:35 +08:00
|
|
|
error = xfs_alloc_lookup_le(acur->bnolt, args->agbno, 0, &i);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
error = xfs_alloc_lookup_ge(acur->bnogt, args->agbno, 0, &i);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
|
|
|
/*
|
2019-10-14 08:10:36 +08:00
|
|
|
* Search the bnobt and cntbt in parallel. Search the bnobt left and
|
|
|
|
* right and lookup the closest extent to the locality hint for each
|
|
|
|
* extent size key in the cntbt. The entire search terminates
|
|
|
|
* immediately on a bnobt hit because that means we've found best case
|
|
|
|
* locality. Otherwise the search continues until the cntbt cursor runs
|
|
|
|
* off the end of the tree. If no allocation candidate is found at this
|
|
|
|
* point, give up on locality, walk backwards from the end of the cntbt
|
|
|
|
* and take the first available extent.
|
|
|
|
*
|
|
|
|
* The parallel tree searches balance each other out to provide fairly
|
|
|
|
* consistent performance for various situations. The bnobt search can
|
|
|
|
* have pathological behavior in the worst case scenario of larger
|
|
|
|
* allocation requests and fragmented free space. On the other hand, the
|
|
|
|
* bnobt is able to satisfy most smaller allocation requests much more
|
|
|
|
* quickly than the cntbt. The cntbt search can sift through fragmented
|
|
|
|
* free space and sets of free extents for larger allocation requests
|
|
|
|
* more quickly than the bnobt. Since the locality hint is just a hint
|
|
|
|
* and we don't want to scan the entire bnobt for perfect locality, the
|
|
|
|
* cntbt search essentially bounds the bnobt search such that we can
|
|
|
|
* find good enough locality at reasonable performance in most cases.
|
2019-10-14 08:10:35 +08:00
|
|
|
*/
|
|
|
|
while (xfs_alloc_cur_active(acur->bnolt) ||
|
2019-10-14 08:10:36 +08:00
|
|
|
xfs_alloc_cur_active(acur->bnogt) ||
|
|
|
|
xfs_alloc_cur_active(acur->cnt)) {
|
|
|
|
|
|
|
|
trace_xfs_alloc_cur_lookup(args);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Search the bnobt left and right. In the case of a hit, finish
|
|
|
|
* the search in the opposite direction and we're done.
|
|
|
|
*/
|
2019-10-14 08:10:35 +08:00
|
|
|
error = xfs_alloc_walk_iter(args, acur, acur->bnolt, false,
|
|
|
|
true, 1, &i);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
if (i == 1) {
|
|
|
|
trace_xfs_alloc_cur_left(args);
|
|
|
|
fbcur = acur->bnogt;
|
|
|
|
fbinc = true;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
error = xfs_alloc_walk_iter(args, acur, acur->bnogt, true, true,
|
|
|
|
1, &i);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
if (i == 1) {
|
|
|
|
trace_xfs_alloc_cur_right(args);
|
|
|
|
fbcur = acur->bnolt;
|
|
|
|
fbinc = false;
|
|
|
|
break;
|
|
|
|
}
|
2019-10-14 08:10:36 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Check the extent with best locality based on the current
|
|
|
|
* extent size search key and keep track of the best candidate.
|
|
|
|
*/
|
|
|
|
error = xfs_alloc_cntbt_iter(args, acur);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
if (!xfs_alloc_cur_active(acur->cnt)) {
|
|
|
|
trace_xfs_alloc_cur_lookup_done(args);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we failed to find anything due to busy extents, return empty
|
|
|
|
* handed so the caller can flush and retry. If no busy extents were
|
|
|
|
* found, walk backwards from the end of the cntbt as a last resort.
|
|
|
|
*/
|
|
|
|
if (!xfs_alloc_cur_active(acur->cnt) && !acur->len && !acur->busy) {
|
|
|
|
error = xfs_btree_decrement(acur->cnt, 0, &i);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
if (i) {
|
2020-03-11 08:57:51 +08:00
|
|
|
acur->cnt->bc_ag.abt.active = true;
|
2019-10-14 08:10:36 +08:00
|
|
|
fbcur = acur->cnt;
|
|
|
|
fbinc = false;
|
|
|
|
}
|
2019-10-14 08:10:35 +08:00
|
|
|
}
|
|
|
|
|
2019-10-14 08:10:36 +08:00
|
|
|
/*
|
|
|
|
* Search in the opposite direction for a better entry in the case of
|
|
|
|
* a bnobt hit or walk backwards from the end of the cntbt.
|
|
|
|
*/
|
2019-10-14 08:10:35 +08:00
|
|
|
if (fbcur) {
|
|
|
|
error = xfs_alloc_walk_iter(args, acur, fbcur, fbinc, true, -1,
|
|
|
|
&i);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (acur->len)
|
|
|
|
*stat = 1;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2019-11-08 04:29:00 +08:00
|
|
|
/* Check the last block of the cnt btree for allocations. */
|
|
|
|
static int
|
|
|
|
xfs_alloc_ag_vextent_lastblock(
|
|
|
|
struct xfs_alloc_arg *args,
|
|
|
|
struct xfs_alloc_cur *acur,
|
|
|
|
xfs_agblock_t *bno,
|
|
|
|
xfs_extlen_t *len,
|
|
|
|
bool *allocated)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
#ifdef DEBUG
|
|
|
|
/* Randomly don't execute the first algorithm. */
|
2022-10-10 10:44:02 +08:00
|
|
|
if (get_random_u32_below(2))
|
2019-11-08 04:29:00 +08:00
|
|
|
return 0;
|
|
|
|
#endif
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Start from the entry that lookup found, sequence through all larger
|
|
|
|
* free blocks. If we're actually pointing at a record smaller than
|
|
|
|
* maxlen, go to the start of this block, and skip all those smaller
|
|
|
|
* than minlen.
|
|
|
|
*/
|
2020-03-14 04:57:58 +08:00
|
|
|
if (*len || args->alignment > 1) {
|
2021-09-17 03:24:04 +08:00
|
|
|
acur->cnt->bc_levels[0].ptr = 1;
|
2019-11-08 04:29:00 +08:00
|
|
|
do {
|
|
|
|
error = xfs_alloc_get_rec(acur->cnt, bno, len, &i);
|
|
|
|
if (error)
|
|
|
|
return error;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(args->mp, i != 1))
|
|
|
|
return -EFSCORRUPTED;
|
2019-11-08 04:29:00 +08:00
|
|
|
if (*len >= args->minlen)
|
|
|
|
break;
|
|
|
|
error = xfs_btree_increment(acur->cnt, 0, &i);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
} while (i);
|
|
|
|
ASSERT(*len >= args->minlen);
|
|
|
|
if (!i)
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
error = xfs_alloc_walk_iter(args, acur, acur->cnt, true, false, -1, &i);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* It didn't work. We COULD be in a case where there's a good record
|
|
|
|
* somewhere, so try again.
|
|
|
|
*/
|
|
|
|
if (acur->len == 0)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
trace_xfs_alloc_near_first(args);
|
|
|
|
*allocated = true;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Allocate a variable extent near bno in the allocation group agno.
|
|
|
|
* Extent's length (returned in len) will be between minlen and maxlen,
|
|
|
|
* and of the form k * prod + mod unless there's nothing that large.
|
|
|
|
* Return the starting a.g. block, or NULLAGBLOCK if we can't do it.
|
|
|
|
*/
|
2019-10-14 08:10:31 +08:00
|
|
|
STATIC int
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_alloc_ag_vextent_near(
|
2023-06-29 02:04:32 +08:00
|
|
|
struct xfs_alloc_arg *args,
|
|
|
|
uint32_t alloc_flags)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2019-10-14 08:10:31 +08:00
|
|
|
struct xfs_alloc_cur acur = {};
|
2019-10-14 08:10:33 +08:00
|
|
|
int error; /* error code */
|
|
|
|
int i; /* result code, temporary */
|
|
|
|
xfs_agblock_t bno;
|
|
|
|
xfs_extlen_t len;
|
2011-04-25 03:06:15 +08:00
|
|
|
|
2019-11-08 05:24:52 +08:00
|
|
|
/* handle uninitialized agbno range so caller doesn't have to */
|
xfs: support min/max agbno args in block allocator
The block allocator supports various arguments to tweak block allocation
behavior and set allocation requirements. The sparse inode chunk feature
introduces a new requirement not supported by the current arguments.
Sparse inode allocations must convert or merge into an inode record that
describes a fixed length chunk (64 inodes x inodesize). Full inode chunk
allocations by definition always result in valid inode records. Sparse
chunk allocations are smaller and the associated records can refer to
blocks not owned by the inode chunk. This model can result in invalid
inode records in certain cases.
For example, if a sparse allocation occurs near the start of an AG, the
aligned inode record for that chunk might refer to agbno 0. If an
allocation occurs towards the end of the AG and the AG size is not
aligned, the inode record could refer to blocks beyond the end of the
AG. While neither of these scenarios directly result in corruption, they
both insert invalid inode records and at minimum cause repair to
complain, are unlikely to merge into full chunks over time and set land
mines for other areas of code.
To guarantee sparse inode chunk allocation creates valid inode records,
support the ability to specify an agbno range limit for
XFS_ALLOCTYPE_NEAR_BNO block allocations. The min/max agbno's are
specified in the allocation arguments and limit the block allocation
algorithms to that range. The starting 'agbno' hint is clamped to the
range if the specified agbno is out of range. If no sufficient extent is
available within the range, the allocation fails. For backwards
compatibility, the min/max fields can be initialized to 0 to disable
range limiting (e.g., equivalent to min=0,max=agsize).
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-05-29 06:53:00 +08:00
|
|
|
if (!args->min_agbno && !args->max_agbno)
|
|
|
|
args->max_agbno = args->mp->m_sb.sb_agblocks - 1;
|
|
|
|
ASSERT(args->min_agbno <= args->max_agbno);
|
|
|
|
|
|
|
|
/* clamp agbno to the range if it's outside */
|
|
|
|
if (args->agbno < args->min_agbno)
|
|
|
|
args->agbno = args->min_agbno;
|
|
|
|
if (args->agbno > args->max_agbno)
|
|
|
|
args->agbno = args->max_agbno;
|
|
|
|
|
xfs: don't block in busy flushing when freeing extents
If the current transaction holds a busy extent and we are trying to
allocate a new extent to fix up the free list, we can deadlock if
the AG is entirely empty except for the busy extent held by the
transaction.
This can occur at runtime processing an XEFI with multiple extents
in this path:
__schedule+0x22f at ffffffff81f75e8f
schedule+0x46 at ffffffff81f76366
xfs_extent_busy_flush+0x69 at ffffffff81477d99
xfs_alloc_ag_vextent_size+0x16a at ffffffff8141711a
xfs_alloc_ag_vextent+0x19b at ffffffff81417edb
xfs_alloc_fix_freelist+0x22f at ffffffff8141896f
xfs_free_extent_fix_freelist+0x6a at ffffffff8141939a
__xfs_free_extent+0x99 at ffffffff81419499
xfs_trans_free_extent+0x3e at ffffffff814a6fee
xfs_extent_free_finish_item+0x24 at ffffffff814a70d4
xfs_defer_finish_noroll+0x1f7 at ffffffff81441407
xfs_defer_finish+0x11 at ffffffff814417e1
xfs_itruncate_extents_flags+0x13d at ffffffff8148b7dd
xfs_inactive_truncate+0xb9 at ffffffff8148bb89
xfs_inactive+0x227 at ffffffff8148c4f7
xfs_fs_destroy_inode+0xb8 at ffffffff81496898
destroy_inode+0x3b at ffffffff8127d2ab
do_unlinkat+0x1d1 at ffffffff81270df1
do_syscall_64+0x40 at ffffffff81f6b5f0
entry_SYSCALL_64_after_hwframe+0x44 at ffffffff8200007c
This can also happen in log recovery when processing an EFI
with multiple extents through this path:
context_switch() kernel/sched/core.c:3881
__schedule() kernel/sched/core.c:5111
schedule() kernel/sched/core.c:5186
xfs_extent_busy_flush() fs/xfs/xfs_extent_busy.c:598
xfs_alloc_ag_vextent_size() fs/xfs/libxfs/xfs_alloc.c:1641
xfs_alloc_ag_vextent() fs/xfs/libxfs/xfs_alloc.c:828
xfs_alloc_fix_freelist() fs/xfs/libxfs/xfs_alloc.c:2362
xfs_free_extent_fix_freelist() fs/xfs/libxfs/xfs_alloc.c:3029
__xfs_free_extent() fs/xfs/libxfs/xfs_alloc.c:3067
xfs_trans_free_extent() fs/xfs/xfs_extfree_item.c:370
xfs_efi_recover() fs/xfs/xfs_extfree_item.c:626
xlog_recover_process_efi() fs/xfs/xfs_log_recover.c:4605
xlog_recover_process_intents() fs/xfs/xfs_log_recover.c:4893
xlog_recover_finish() fs/xfs/xfs_log_recover.c:5824
xfs_log_mount_finish() fs/xfs/xfs_log.c:764
xfs_mountfs() fs/xfs/xfs_mount.c:978
xfs_fs_fill_super() fs/xfs/xfs_super.c:1908
mount_bdev() fs/super.c:1417
xfs_fs_mount() fs/xfs/xfs_super.c:1985
legacy_get_tree() fs/fs_context.c:647
vfs_get_tree() fs/super.c:1547
do_new_mount() fs/namespace.c:2843
do_mount() fs/namespace.c:3163
ksys_mount() fs/namespace.c:3372
__do_sys_mount() fs/namespace.c:3386
__se_sys_mount() fs/namespace.c:3383
__x64_sys_mount() fs/namespace.c:3383
do_syscall_64() arch/x86/entry/common.c:296
entry_SYSCALL_64() arch/x86/entry/entry_64.S:180
To avoid this deadlock, we should not block in
xfs_extent_busy_flush() if we hold a busy extent in the current
transaction.
Now that the EFI processing code can handle requeuing a partially
completed EFI, we can detect this situation in
xfs_extent_busy_flush() and return -EAGAIN rather than going to
sleep forever. The -EAGAIN get propagated back out to the
xfs_trans_free_extent() context, where the EFD is populated and the
transaction is rolled, thereby moving the busy extents into the CIL.
At this point, we can retry the extent free operation again with a
clean transaction. If we hit the same "all free extents are busy"
situation when trying to fix up the free list, we can safely call
xfs_extent_busy_flush() and wait for the busy extents to resolve
and wake us. At this point, the allocation search can make progress
again and we can fix up the free list.
This deadlock was first reported by Chandan in mid-2021, but I
couldn't make myself understood during review, and didn't have time
to fix it myself.
It was reported again in March 2023, and again I have found myself
unable to explain the complexities of the solution needed during
review.
As such, I don't have hours more time to waste trying to get the
fix written the way it needs to be written, so I'm just doing it
myself. This patchset is largely based on Wengang Wang's last patch,
but with all the unnecessary stuff removed, split up into multiple
patches and cleaned up somewhat.
Reported-by: Chandan Babu R <chandanrlinux@gmail.com>
Reported-by: Wengang Wang <wen.gang.wang@oracle.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
2023-06-29 02:04:33 +08:00
|
|
|
/* Retry once quickly if we find busy extents before blocking. */
|
|
|
|
alloc_flags |= XFS_ALLOC_FLAG_TRYFLUSH;
|
2011-04-25 03:06:15 +08:00
|
|
|
restart:
|
2019-10-14 08:10:33 +08:00
|
|
|
len = 0;
|
2011-04-25 03:06:15 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
2019-10-14 08:10:31 +08:00
|
|
|
* Set up cursors and see if there are any free extents as big as
|
|
|
|
* maxlen. If not, pick the last entry in the tree unless the tree is
|
|
|
|
* empty.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2019-10-14 08:10:31 +08:00
|
|
|
error = xfs_alloc_cur_setup(args, &acur);
|
|
|
|
if (error == -ENOSPC) {
|
2019-10-14 08:10:33 +08:00
|
|
|
error = xfs_alloc_ag_vextent_small(args, acur.cnt, &bno,
|
|
|
|
&len, &i);
|
2019-10-14 08:10:31 +08:00
|
|
|
if (error)
|
|
|
|
goto out;
|
2019-10-14 08:10:33 +08:00
|
|
|
if (i == 0 || len == 0) {
|
2011-04-25 03:06:15 +08:00
|
|
|
trace_xfs_alloc_near_noentry(args);
|
2019-10-14 08:10:31 +08:00
|
|
|
goto out;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
ASSERT(i == 1);
|
2019-10-14 08:10:31 +08:00
|
|
|
} else if (error) {
|
|
|
|
goto out;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2011-04-25 03:06:15 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* First algorithm.
|
|
|
|
* If the requested extent is large wrt the freespaces available
|
|
|
|
* in this a.g., then the cursor will be pointing to a btree entry
|
|
|
|
* near the right edge of the tree. If it's in the last btree leaf
|
|
|
|
* block, then we just examine all the entries in that block
|
|
|
|
* that are big enough, and pick the best one.
|
|
|
|
*/
|
2019-11-08 04:29:00 +08:00
|
|
|
if (xfs_btree_islastblock(acur.cnt, 0)) {
|
|
|
|
bool allocated = false;
|
2019-10-14 08:10:34 +08:00
|
|
|
|
2019-11-08 04:29:00 +08:00
|
|
|
error = xfs_alloc_ag_vextent_lastblock(args, &acur, &bno, &len,
|
|
|
|
&allocated);
|
2019-10-14 08:10:34 +08:00
|
|
|
if (error)
|
|
|
|
goto out;
|
2019-11-08 04:29:00 +08:00
|
|
|
if (allocated)
|
|
|
|
goto alloc_finish;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2019-10-14 08:10:31 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
2019-10-14 08:10:36 +08:00
|
|
|
* Second algorithm. Combined cntbt and bnobt search to find ideal
|
|
|
|
* locality.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2019-10-14 08:10:36 +08:00
|
|
|
error = xfs_alloc_ag_vextent_locality(args, &acur, &i);
|
2019-10-14 08:10:31 +08:00
|
|
|
if (error)
|
|
|
|
goto out;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* If we couldn't get anything, give up.
|
|
|
|
*/
|
2019-10-14 08:10:33 +08:00
|
|
|
if (!acur.len) {
|
2019-10-14 08:10:32 +08:00
|
|
|
if (acur.busy) {
|
xfs: don't block in busy flushing when freeing extents
If the current transaction holds a busy extent and we are trying to
allocate a new extent to fix up the free list, we can deadlock if
the AG is entirely empty except for the busy extent held by the
transaction.
This can occur at runtime processing an XEFI with multiple extents
in this path:
__schedule+0x22f at ffffffff81f75e8f
schedule+0x46 at ffffffff81f76366
xfs_extent_busy_flush+0x69 at ffffffff81477d99
xfs_alloc_ag_vextent_size+0x16a at ffffffff8141711a
xfs_alloc_ag_vextent+0x19b at ffffffff81417edb
xfs_alloc_fix_freelist+0x22f at ffffffff8141896f
xfs_free_extent_fix_freelist+0x6a at ffffffff8141939a
__xfs_free_extent+0x99 at ffffffff81419499
xfs_trans_free_extent+0x3e at ffffffff814a6fee
xfs_extent_free_finish_item+0x24 at ffffffff814a70d4
xfs_defer_finish_noroll+0x1f7 at ffffffff81441407
xfs_defer_finish+0x11 at ffffffff814417e1
xfs_itruncate_extents_flags+0x13d at ffffffff8148b7dd
xfs_inactive_truncate+0xb9 at ffffffff8148bb89
xfs_inactive+0x227 at ffffffff8148c4f7
xfs_fs_destroy_inode+0xb8 at ffffffff81496898
destroy_inode+0x3b at ffffffff8127d2ab
do_unlinkat+0x1d1 at ffffffff81270df1
do_syscall_64+0x40 at ffffffff81f6b5f0
entry_SYSCALL_64_after_hwframe+0x44 at ffffffff8200007c
This can also happen in log recovery when processing an EFI
with multiple extents through this path:
context_switch() kernel/sched/core.c:3881
__schedule() kernel/sched/core.c:5111
schedule() kernel/sched/core.c:5186
xfs_extent_busy_flush() fs/xfs/xfs_extent_busy.c:598
xfs_alloc_ag_vextent_size() fs/xfs/libxfs/xfs_alloc.c:1641
xfs_alloc_ag_vextent() fs/xfs/libxfs/xfs_alloc.c:828
xfs_alloc_fix_freelist() fs/xfs/libxfs/xfs_alloc.c:2362
xfs_free_extent_fix_freelist() fs/xfs/libxfs/xfs_alloc.c:3029
__xfs_free_extent() fs/xfs/libxfs/xfs_alloc.c:3067
xfs_trans_free_extent() fs/xfs/xfs_extfree_item.c:370
xfs_efi_recover() fs/xfs/xfs_extfree_item.c:626
xlog_recover_process_efi() fs/xfs/xfs_log_recover.c:4605
xlog_recover_process_intents() fs/xfs/xfs_log_recover.c:4893
xlog_recover_finish() fs/xfs/xfs_log_recover.c:5824
xfs_log_mount_finish() fs/xfs/xfs_log.c:764
xfs_mountfs() fs/xfs/xfs_mount.c:978
xfs_fs_fill_super() fs/xfs/xfs_super.c:1908
mount_bdev() fs/super.c:1417
xfs_fs_mount() fs/xfs/xfs_super.c:1985
legacy_get_tree() fs/fs_context.c:647
vfs_get_tree() fs/super.c:1547
do_new_mount() fs/namespace.c:2843
do_mount() fs/namespace.c:3163
ksys_mount() fs/namespace.c:3372
__do_sys_mount() fs/namespace.c:3386
__se_sys_mount() fs/namespace.c:3383
__x64_sys_mount() fs/namespace.c:3383
do_syscall_64() arch/x86/entry/common.c:296
entry_SYSCALL_64() arch/x86/entry/entry_64.S:180
To avoid this deadlock, we should not block in
xfs_extent_busy_flush() if we hold a busy extent in the current
transaction.
Now that the EFI processing code can handle requeuing a partially
completed EFI, we can detect this situation in
xfs_extent_busy_flush() and return -EAGAIN rather than going to
sleep forever. The -EAGAIN get propagated back out to the
xfs_trans_free_extent() context, where the EFD is populated and the
transaction is rolled, thereby moving the busy extents into the CIL.
At this point, we can retry the extent free operation again with a
clean transaction. If we hit the same "all free extents are busy"
situation when trying to fix up the free list, we can safely call
xfs_extent_busy_flush() and wait for the busy extents to resolve
and wake us. At this point, the allocation search can make progress
again and we can fix up the free list.
This deadlock was first reported by Chandan in mid-2021, but I
couldn't make myself understood during review, and didn't have time
to fix it myself.
It was reported again in March 2023, and again I have found myself
unable to explain the complexities of the solution needed during
review.
As such, I don't have hours more time to waste trying to get the
fix written the way it needs to be written, so I'm just doing it
myself. This patchset is largely based on Wengang Wang's last patch,
but with all the unnecessary stuff removed, split up into multiple
patches and cleaned up somewhat.
Reported-by: Chandan Babu R <chandanrlinux@gmail.com>
Reported-by: Wengang Wang <wen.gang.wang@oracle.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
2023-06-29 02:04:33 +08:00
|
|
|
/*
|
|
|
|
* Our only valid extents must have been busy. Flush and
|
|
|
|
* retry the allocation again. If we get an -EAGAIN
|
|
|
|
* error, we're being told that a deadlock was avoided
|
|
|
|
* and the current transaction needs committing before
|
|
|
|
* the allocation can be retried.
|
|
|
|
*/
|
2011-04-25 03:06:15 +08:00
|
|
|
trace_xfs_alloc_near_busy(args);
|
xfs: don't block in busy flushing when freeing extents
If the current transaction holds a busy extent and we are trying to
allocate a new extent to fix up the free list, we can deadlock if
the AG is entirely empty except for the busy extent held by the
transaction.
This can occur at runtime processing an XEFI with multiple extents
in this path:
__schedule+0x22f at ffffffff81f75e8f
schedule+0x46 at ffffffff81f76366
xfs_extent_busy_flush+0x69 at ffffffff81477d99
xfs_alloc_ag_vextent_size+0x16a at ffffffff8141711a
xfs_alloc_ag_vextent+0x19b at ffffffff81417edb
xfs_alloc_fix_freelist+0x22f at ffffffff8141896f
xfs_free_extent_fix_freelist+0x6a at ffffffff8141939a
__xfs_free_extent+0x99 at ffffffff81419499
xfs_trans_free_extent+0x3e at ffffffff814a6fee
xfs_extent_free_finish_item+0x24 at ffffffff814a70d4
xfs_defer_finish_noroll+0x1f7 at ffffffff81441407
xfs_defer_finish+0x11 at ffffffff814417e1
xfs_itruncate_extents_flags+0x13d at ffffffff8148b7dd
xfs_inactive_truncate+0xb9 at ffffffff8148bb89
xfs_inactive+0x227 at ffffffff8148c4f7
xfs_fs_destroy_inode+0xb8 at ffffffff81496898
destroy_inode+0x3b at ffffffff8127d2ab
do_unlinkat+0x1d1 at ffffffff81270df1
do_syscall_64+0x40 at ffffffff81f6b5f0
entry_SYSCALL_64_after_hwframe+0x44 at ffffffff8200007c
This can also happen in log recovery when processing an EFI
with multiple extents through this path:
context_switch() kernel/sched/core.c:3881
__schedule() kernel/sched/core.c:5111
schedule() kernel/sched/core.c:5186
xfs_extent_busy_flush() fs/xfs/xfs_extent_busy.c:598
xfs_alloc_ag_vextent_size() fs/xfs/libxfs/xfs_alloc.c:1641
xfs_alloc_ag_vextent() fs/xfs/libxfs/xfs_alloc.c:828
xfs_alloc_fix_freelist() fs/xfs/libxfs/xfs_alloc.c:2362
xfs_free_extent_fix_freelist() fs/xfs/libxfs/xfs_alloc.c:3029
__xfs_free_extent() fs/xfs/libxfs/xfs_alloc.c:3067
xfs_trans_free_extent() fs/xfs/xfs_extfree_item.c:370
xfs_efi_recover() fs/xfs/xfs_extfree_item.c:626
xlog_recover_process_efi() fs/xfs/xfs_log_recover.c:4605
xlog_recover_process_intents() fs/xfs/xfs_log_recover.c:4893
xlog_recover_finish() fs/xfs/xfs_log_recover.c:5824
xfs_log_mount_finish() fs/xfs/xfs_log.c:764
xfs_mountfs() fs/xfs/xfs_mount.c:978
xfs_fs_fill_super() fs/xfs/xfs_super.c:1908
mount_bdev() fs/super.c:1417
xfs_fs_mount() fs/xfs/xfs_super.c:1985
legacy_get_tree() fs/fs_context.c:647
vfs_get_tree() fs/super.c:1547
do_new_mount() fs/namespace.c:2843
do_mount() fs/namespace.c:3163
ksys_mount() fs/namespace.c:3372
__do_sys_mount() fs/namespace.c:3386
__se_sys_mount() fs/namespace.c:3383
__x64_sys_mount() fs/namespace.c:3383
do_syscall_64() arch/x86/entry/common.c:296
entry_SYSCALL_64() arch/x86/entry/entry_64.S:180
To avoid this deadlock, we should not block in
xfs_extent_busy_flush() if we hold a busy extent in the current
transaction.
Now that the EFI processing code can handle requeuing a partially
completed EFI, we can detect this situation in
xfs_extent_busy_flush() and return -EAGAIN rather than going to
sleep forever. The -EAGAIN get propagated back out to the
xfs_trans_free_extent() context, where the EFD is populated and the
transaction is rolled, thereby moving the busy extents into the CIL.
At this point, we can retry the extent free operation again with a
clean transaction. If we hit the same "all free extents are busy"
situation when trying to fix up the free list, we can safely call
xfs_extent_busy_flush() and wait for the busy extents to resolve
and wake us. At this point, the allocation search can make progress
again and we can fix up the free list.
This deadlock was first reported by Chandan in mid-2021, but I
couldn't make myself understood during review, and didn't have time
to fix it myself.
It was reported again in March 2023, and again I have found myself
unable to explain the complexities of the solution needed during
review.
As such, I don't have hours more time to waste trying to get the
fix written the way it needs to be written, so I'm just doing it
myself. This patchset is largely based on Wengang Wang's last patch,
but with all the unnecessary stuff removed, split up into multiple
patches and cleaned up somewhat.
Reported-by: Chandan Babu R <chandanrlinux@gmail.com>
Reported-by: Wengang Wang <wen.gang.wang@oracle.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
2023-06-29 02:04:33 +08:00
|
|
|
error = xfs_extent_busy_flush(args->tp, args->pag,
|
|
|
|
acur.busy_gen, alloc_flags);
|
|
|
|
if (error)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
alloc_flags &= ~XFS_ALLOC_FLAG_TRYFLUSH;
|
2011-04-25 03:06:15 +08:00
|
|
|
goto restart;
|
|
|
|
}
|
2009-12-15 07:14:59 +08:00
|
|
|
trace_xfs_alloc_size_neither(args);
|
2005-04-17 06:20:36 +08:00
|
|
|
args->agbno = NULLAGBLOCK;
|
2019-10-14 08:10:31 +08:00
|
|
|
goto out;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2010-12-10 23:04:11 +08:00
|
|
|
|
2019-11-08 04:29:00 +08:00
|
|
|
alloc_finish:
|
2019-10-14 08:10:35 +08:00
|
|
|
/* fix up btrees on a successful allocation */
|
|
|
|
error = xfs_alloc_cur_finish(args, &acur);
|
2009-12-15 07:14:59 +08:00
|
|
|
|
2019-10-14 08:10:31 +08:00
|
|
|
out:
|
|
|
|
xfs_alloc_cur_close(&acur, error);
|
2005-04-17 06:20:36 +08:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Allocate a variable extent anywhere in the allocation group agno.
|
|
|
|
* Extent's length (returned in len) will be between minlen and maxlen,
|
|
|
|
* and of the form k * prod + mod unless there's nothing that large.
|
|
|
|
* Return the starting a.g. block, or NULLAGBLOCK if we can't do it.
|
|
|
|
*/
|
2023-06-29 02:04:32 +08:00
|
|
|
static int
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_alloc_ag_vextent_size(
|
2023-06-29 02:04:32 +08:00
|
|
|
struct xfs_alloc_arg *args,
|
|
|
|
uint32_t alloc_flags)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2023-06-29 02:04:32 +08:00
|
|
|
struct xfs_agf *agf = args->agbp->b_addr;
|
|
|
|
struct xfs_btree_cur *bno_cur;
|
|
|
|
struct xfs_btree_cur *cnt_cur;
|
|
|
|
xfs_agblock_t fbno; /* start of found freespace */
|
|
|
|
xfs_extlen_t flen; /* length of found freespace */
|
|
|
|
xfs_agblock_t rbno; /* returned block number */
|
|
|
|
xfs_extlen_t rlen; /* length of returned extent */
|
|
|
|
bool busy;
|
|
|
|
unsigned busy_gen;
|
|
|
|
int error;
|
|
|
|
int i;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
xfs: don't block in busy flushing when freeing extents
If the current transaction holds a busy extent and we are trying to
allocate a new extent to fix up the free list, we can deadlock if
the AG is entirely empty except for the busy extent held by the
transaction.
This can occur at runtime processing an XEFI with multiple extents
in this path:
__schedule+0x22f at ffffffff81f75e8f
schedule+0x46 at ffffffff81f76366
xfs_extent_busy_flush+0x69 at ffffffff81477d99
xfs_alloc_ag_vextent_size+0x16a at ffffffff8141711a
xfs_alloc_ag_vextent+0x19b at ffffffff81417edb
xfs_alloc_fix_freelist+0x22f at ffffffff8141896f
xfs_free_extent_fix_freelist+0x6a at ffffffff8141939a
__xfs_free_extent+0x99 at ffffffff81419499
xfs_trans_free_extent+0x3e at ffffffff814a6fee
xfs_extent_free_finish_item+0x24 at ffffffff814a70d4
xfs_defer_finish_noroll+0x1f7 at ffffffff81441407
xfs_defer_finish+0x11 at ffffffff814417e1
xfs_itruncate_extents_flags+0x13d at ffffffff8148b7dd
xfs_inactive_truncate+0xb9 at ffffffff8148bb89
xfs_inactive+0x227 at ffffffff8148c4f7
xfs_fs_destroy_inode+0xb8 at ffffffff81496898
destroy_inode+0x3b at ffffffff8127d2ab
do_unlinkat+0x1d1 at ffffffff81270df1
do_syscall_64+0x40 at ffffffff81f6b5f0
entry_SYSCALL_64_after_hwframe+0x44 at ffffffff8200007c
This can also happen in log recovery when processing an EFI
with multiple extents through this path:
context_switch() kernel/sched/core.c:3881
__schedule() kernel/sched/core.c:5111
schedule() kernel/sched/core.c:5186
xfs_extent_busy_flush() fs/xfs/xfs_extent_busy.c:598
xfs_alloc_ag_vextent_size() fs/xfs/libxfs/xfs_alloc.c:1641
xfs_alloc_ag_vextent() fs/xfs/libxfs/xfs_alloc.c:828
xfs_alloc_fix_freelist() fs/xfs/libxfs/xfs_alloc.c:2362
xfs_free_extent_fix_freelist() fs/xfs/libxfs/xfs_alloc.c:3029
__xfs_free_extent() fs/xfs/libxfs/xfs_alloc.c:3067
xfs_trans_free_extent() fs/xfs/xfs_extfree_item.c:370
xfs_efi_recover() fs/xfs/xfs_extfree_item.c:626
xlog_recover_process_efi() fs/xfs/xfs_log_recover.c:4605
xlog_recover_process_intents() fs/xfs/xfs_log_recover.c:4893
xlog_recover_finish() fs/xfs/xfs_log_recover.c:5824
xfs_log_mount_finish() fs/xfs/xfs_log.c:764
xfs_mountfs() fs/xfs/xfs_mount.c:978
xfs_fs_fill_super() fs/xfs/xfs_super.c:1908
mount_bdev() fs/super.c:1417
xfs_fs_mount() fs/xfs/xfs_super.c:1985
legacy_get_tree() fs/fs_context.c:647
vfs_get_tree() fs/super.c:1547
do_new_mount() fs/namespace.c:2843
do_mount() fs/namespace.c:3163
ksys_mount() fs/namespace.c:3372
__do_sys_mount() fs/namespace.c:3386
__se_sys_mount() fs/namespace.c:3383
__x64_sys_mount() fs/namespace.c:3383
do_syscall_64() arch/x86/entry/common.c:296
entry_SYSCALL_64() arch/x86/entry/entry_64.S:180
To avoid this deadlock, we should not block in
xfs_extent_busy_flush() if we hold a busy extent in the current
transaction.
Now that the EFI processing code can handle requeuing a partially
completed EFI, we can detect this situation in
xfs_extent_busy_flush() and return -EAGAIN rather than going to
sleep forever. The -EAGAIN get propagated back out to the
xfs_trans_free_extent() context, where the EFD is populated and the
transaction is rolled, thereby moving the busy extents into the CIL.
At this point, we can retry the extent free operation again with a
clean transaction. If we hit the same "all free extents are busy"
situation when trying to fix up the free list, we can safely call
xfs_extent_busy_flush() and wait for the busy extents to resolve
and wake us. At this point, the allocation search can make progress
again and we can fix up the free list.
This deadlock was first reported by Chandan in mid-2021, but I
couldn't make myself understood during review, and didn't have time
to fix it myself.
It was reported again in March 2023, and again I have found myself
unable to explain the complexities of the solution needed during
review.
As such, I don't have hours more time to waste trying to get the
fix written the way it needs to be written, so I'm just doing it
myself. This patchset is largely based on Wengang Wang's last patch,
but with all the unnecessary stuff removed, split up into multiple
patches and cleaned up somewhat.
Reported-by: Chandan Babu R <chandanrlinux@gmail.com>
Reported-by: Wengang Wang <wen.gang.wang@oracle.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
2023-06-29 02:04:33 +08:00
|
|
|
/* Retry once quickly if we find busy extents before blocking. */
|
|
|
|
alloc_flags |= XFS_ALLOC_FLAG_TRYFLUSH;
|
2011-04-25 03:06:15 +08:00
|
|
|
restart:
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Allocate and initialize a cursor for the by-size btree.
|
|
|
|
*/
|
2008-10-30 13:53:59 +08:00
|
|
|
cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
|
2021-06-02 08:48:24 +08:00
|
|
|
args->pag, XFS_BTNUM_CNT);
|
2005-04-17 06:20:36 +08:00
|
|
|
bno_cur = NULL;
|
2011-04-25 03:06:15 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Look for an entry >= maxlen+alignment-1 blocks.
|
|
|
|
*/
|
|
|
|
if ((error = xfs_alloc_lookup_ge(cnt_cur, 0,
|
|
|
|
args->maxlen + args->alignment - 1, &i)))
|
|
|
|
goto error0;
|
2011-04-25 03:06:15 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
2017-02-08 06:06:57 +08:00
|
|
|
* If none then we have to settle for a smaller extent. In the case that
|
|
|
|
* there are no large extents, this will return the last entry in the
|
|
|
|
* tree unless the tree is empty. In the case that there are only busy
|
|
|
|
* large extents, this will return the largest small extent unless there
|
2011-04-25 03:06:15 +08:00
|
|
|
* are no smaller extents available.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2017-02-08 06:06:57 +08:00
|
|
|
if (!i) {
|
2011-04-25 03:06:15 +08:00
|
|
|
error = xfs_alloc_ag_vextent_small(args, cnt_cur,
|
|
|
|
&fbno, &flen, &i);
|
|
|
|
if (error)
|
2005-04-17 06:20:36 +08:00
|
|
|
goto error0;
|
|
|
|
if (i == 0 || flen == 0) {
|
|
|
|
xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
|
2009-12-15 07:14:59 +08:00
|
|
|
trace_xfs_alloc_size_noentry(args);
|
2005-04-17 06:20:36 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
ASSERT(i == 1);
|
2017-02-08 06:06:57 +08:00
|
|
|
busy = xfs_alloc_compute_aligned(args, fbno, flen, &rbno,
|
|
|
|
&rlen, &busy_gen);
|
2011-04-25 03:06:15 +08:00
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* Search for a non-busy extent that is large enough.
|
|
|
|
*/
|
|
|
|
for (;;) {
|
|
|
|
error = xfs_alloc_get_rec(cnt_cur, &fbno, &flen, &i);
|
|
|
|
if (error)
|
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(args->mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2011-04-25 03:06:15 +08:00
|
|
|
|
2017-02-08 06:06:57 +08:00
|
|
|
busy = xfs_alloc_compute_aligned(args, fbno, flen,
|
|
|
|
&rbno, &rlen, &busy_gen);
|
2011-04-25 03:06:15 +08:00
|
|
|
|
|
|
|
if (rlen >= args->maxlen)
|
|
|
|
break;
|
|
|
|
|
|
|
|
error = xfs_btree_increment(cnt_cur, 0, &i);
|
|
|
|
if (error)
|
|
|
|
goto error0;
|
xfs: don't block in busy flushing when freeing extents
If the current transaction holds a busy extent and we are trying to
allocate a new extent to fix up the free list, we can deadlock if
the AG is entirely empty except for the busy extent held by the
transaction.
This can occur at runtime processing an XEFI with multiple extents
in this path:
__schedule+0x22f at ffffffff81f75e8f
schedule+0x46 at ffffffff81f76366
xfs_extent_busy_flush+0x69 at ffffffff81477d99
xfs_alloc_ag_vextent_size+0x16a at ffffffff8141711a
xfs_alloc_ag_vextent+0x19b at ffffffff81417edb
xfs_alloc_fix_freelist+0x22f at ffffffff8141896f
xfs_free_extent_fix_freelist+0x6a at ffffffff8141939a
__xfs_free_extent+0x99 at ffffffff81419499
xfs_trans_free_extent+0x3e at ffffffff814a6fee
xfs_extent_free_finish_item+0x24 at ffffffff814a70d4
xfs_defer_finish_noroll+0x1f7 at ffffffff81441407
xfs_defer_finish+0x11 at ffffffff814417e1
xfs_itruncate_extents_flags+0x13d at ffffffff8148b7dd
xfs_inactive_truncate+0xb9 at ffffffff8148bb89
xfs_inactive+0x227 at ffffffff8148c4f7
xfs_fs_destroy_inode+0xb8 at ffffffff81496898
destroy_inode+0x3b at ffffffff8127d2ab
do_unlinkat+0x1d1 at ffffffff81270df1
do_syscall_64+0x40 at ffffffff81f6b5f0
entry_SYSCALL_64_after_hwframe+0x44 at ffffffff8200007c
This can also happen in log recovery when processing an EFI
with multiple extents through this path:
context_switch() kernel/sched/core.c:3881
__schedule() kernel/sched/core.c:5111
schedule() kernel/sched/core.c:5186
xfs_extent_busy_flush() fs/xfs/xfs_extent_busy.c:598
xfs_alloc_ag_vextent_size() fs/xfs/libxfs/xfs_alloc.c:1641
xfs_alloc_ag_vextent() fs/xfs/libxfs/xfs_alloc.c:828
xfs_alloc_fix_freelist() fs/xfs/libxfs/xfs_alloc.c:2362
xfs_free_extent_fix_freelist() fs/xfs/libxfs/xfs_alloc.c:3029
__xfs_free_extent() fs/xfs/libxfs/xfs_alloc.c:3067
xfs_trans_free_extent() fs/xfs/xfs_extfree_item.c:370
xfs_efi_recover() fs/xfs/xfs_extfree_item.c:626
xlog_recover_process_efi() fs/xfs/xfs_log_recover.c:4605
xlog_recover_process_intents() fs/xfs/xfs_log_recover.c:4893
xlog_recover_finish() fs/xfs/xfs_log_recover.c:5824
xfs_log_mount_finish() fs/xfs/xfs_log.c:764
xfs_mountfs() fs/xfs/xfs_mount.c:978
xfs_fs_fill_super() fs/xfs/xfs_super.c:1908
mount_bdev() fs/super.c:1417
xfs_fs_mount() fs/xfs/xfs_super.c:1985
legacy_get_tree() fs/fs_context.c:647
vfs_get_tree() fs/super.c:1547
do_new_mount() fs/namespace.c:2843
do_mount() fs/namespace.c:3163
ksys_mount() fs/namespace.c:3372
__do_sys_mount() fs/namespace.c:3386
__se_sys_mount() fs/namespace.c:3383
__x64_sys_mount() fs/namespace.c:3383
do_syscall_64() arch/x86/entry/common.c:296
entry_SYSCALL_64() arch/x86/entry/entry_64.S:180
To avoid this deadlock, we should not block in
xfs_extent_busy_flush() if we hold a busy extent in the current
transaction.
Now that the EFI processing code can handle requeuing a partially
completed EFI, we can detect this situation in
xfs_extent_busy_flush() and return -EAGAIN rather than going to
sleep forever. The -EAGAIN get propagated back out to the
xfs_trans_free_extent() context, where the EFD is populated and the
transaction is rolled, thereby moving the busy extents into the CIL.
At this point, we can retry the extent free operation again with a
clean transaction. If we hit the same "all free extents are busy"
situation when trying to fix up the free list, we can safely call
xfs_extent_busy_flush() and wait for the busy extents to resolve
and wake us. At this point, the allocation search can make progress
again and we can fix up the free list.
This deadlock was first reported by Chandan in mid-2021, but I
couldn't make myself understood during review, and didn't have time
to fix it myself.
It was reported again in March 2023, and again I have found myself
unable to explain the complexities of the solution needed during
review.
As such, I don't have hours more time to waste trying to get the
fix written the way it needs to be written, so I'm just doing it
myself. This patchset is largely based on Wengang Wang's last patch,
but with all the unnecessary stuff removed, split up into multiple
patches and cleaned up somewhat.
Reported-by: Chandan Babu R <chandanrlinux@gmail.com>
Reported-by: Wengang Wang <wen.gang.wang@oracle.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
2023-06-29 02:04:33 +08:00
|
|
|
if (i)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Our only valid extents must have been busy. Flush and
|
|
|
|
* retry the allocation again. If we get an -EAGAIN
|
|
|
|
* error, we're being told that a deadlock was avoided
|
|
|
|
* and the current transaction needs committing before
|
|
|
|
* the allocation can be retried.
|
|
|
|
*/
|
|
|
|
trace_xfs_alloc_size_busy(args);
|
|
|
|
error = xfs_extent_busy_flush(args->tp, args->pag,
|
|
|
|
busy_gen, alloc_flags);
|
|
|
|
if (error)
|
|
|
|
goto error0;
|
|
|
|
|
|
|
|
alloc_flags &= ~XFS_ALLOC_FLAG_TRYFLUSH;
|
|
|
|
xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
|
|
|
|
goto restart;
|
2011-04-25 03:06:15 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2011-04-25 03:06:15 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* In the first case above, we got the last entry in the
|
|
|
|
* by-size btree. Now we check to see if the space hits maxlen
|
|
|
|
* once aligned; if not, we search left for something better.
|
|
|
|
* This can't happen in the second case above.
|
|
|
|
*/
|
|
|
|
rlen = XFS_EXTLEN_MIN(args->maxlen, rlen);
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(args->mp,
|
|
|
|
rlen != 0 &&
|
|
|
|
(rlen > flen ||
|
|
|
|
rbno + rlen > fbno + flen))) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
if (rlen < args->maxlen) {
|
|
|
|
xfs_agblock_t bestfbno;
|
|
|
|
xfs_extlen_t bestflen;
|
|
|
|
xfs_agblock_t bestrbno;
|
|
|
|
xfs_extlen_t bestrlen;
|
|
|
|
|
|
|
|
bestrlen = rlen;
|
|
|
|
bestrbno = rbno;
|
|
|
|
bestflen = flen;
|
|
|
|
bestfbno = fbno;
|
|
|
|
for (;;) {
|
2008-10-30 13:55:58 +08:00
|
|
|
if ((error = xfs_btree_decrement(cnt_cur, 0, &i)))
|
2005-04-17 06:20:36 +08:00
|
|
|
goto error0;
|
|
|
|
if (i == 0)
|
|
|
|
break;
|
|
|
|
if ((error = xfs_alloc_get_rec(cnt_cur, &fbno, &flen,
|
|
|
|
&i)))
|
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(args->mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
if (flen < bestrlen)
|
|
|
|
break;
|
2017-02-08 06:06:57 +08:00
|
|
|
busy = xfs_alloc_compute_aligned(args, fbno, flen,
|
|
|
|
&rbno, &rlen, &busy_gen);
|
2005-04-17 06:20:36 +08:00
|
|
|
rlen = XFS_EXTLEN_MIN(args->maxlen, rlen);
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(args->mp,
|
|
|
|
rlen != 0 &&
|
|
|
|
(rlen > flen ||
|
|
|
|
rbno + rlen > fbno + flen))) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
if (rlen > bestrlen) {
|
|
|
|
bestrlen = rlen;
|
|
|
|
bestrbno = rbno;
|
|
|
|
bestflen = flen;
|
|
|
|
bestfbno = fbno;
|
|
|
|
if (rlen == args->maxlen)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if ((error = xfs_alloc_lookup_eq(cnt_cur, bestfbno, bestflen,
|
|
|
|
&i)))
|
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(args->mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
rlen = bestrlen;
|
|
|
|
rbno = bestrbno;
|
|
|
|
flen = bestflen;
|
|
|
|
fbno = bestfbno;
|
|
|
|
}
|
|
|
|
args->wasfromfl = 0;
|
|
|
|
/*
|
|
|
|
* Fix up the length.
|
|
|
|
*/
|
|
|
|
args->len = rlen;
|
2011-04-25 03:06:15 +08:00
|
|
|
if (rlen < args->minlen) {
|
2017-02-08 06:06:57 +08:00
|
|
|
if (busy) {
|
xfs: don't block in busy flushing when freeing extents
If the current transaction holds a busy extent and we are trying to
allocate a new extent to fix up the free list, we can deadlock if
the AG is entirely empty except for the busy extent held by the
transaction.
This can occur at runtime processing an XEFI with multiple extents
in this path:
__schedule+0x22f at ffffffff81f75e8f
schedule+0x46 at ffffffff81f76366
xfs_extent_busy_flush+0x69 at ffffffff81477d99
xfs_alloc_ag_vextent_size+0x16a at ffffffff8141711a
xfs_alloc_ag_vextent+0x19b at ffffffff81417edb
xfs_alloc_fix_freelist+0x22f at ffffffff8141896f
xfs_free_extent_fix_freelist+0x6a at ffffffff8141939a
__xfs_free_extent+0x99 at ffffffff81419499
xfs_trans_free_extent+0x3e at ffffffff814a6fee
xfs_extent_free_finish_item+0x24 at ffffffff814a70d4
xfs_defer_finish_noroll+0x1f7 at ffffffff81441407
xfs_defer_finish+0x11 at ffffffff814417e1
xfs_itruncate_extents_flags+0x13d at ffffffff8148b7dd
xfs_inactive_truncate+0xb9 at ffffffff8148bb89
xfs_inactive+0x227 at ffffffff8148c4f7
xfs_fs_destroy_inode+0xb8 at ffffffff81496898
destroy_inode+0x3b at ffffffff8127d2ab
do_unlinkat+0x1d1 at ffffffff81270df1
do_syscall_64+0x40 at ffffffff81f6b5f0
entry_SYSCALL_64_after_hwframe+0x44 at ffffffff8200007c
This can also happen in log recovery when processing an EFI
with multiple extents through this path:
context_switch() kernel/sched/core.c:3881
__schedule() kernel/sched/core.c:5111
schedule() kernel/sched/core.c:5186
xfs_extent_busy_flush() fs/xfs/xfs_extent_busy.c:598
xfs_alloc_ag_vextent_size() fs/xfs/libxfs/xfs_alloc.c:1641
xfs_alloc_ag_vextent() fs/xfs/libxfs/xfs_alloc.c:828
xfs_alloc_fix_freelist() fs/xfs/libxfs/xfs_alloc.c:2362
xfs_free_extent_fix_freelist() fs/xfs/libxfs/xfs_alloc.c:3029
__xfs_free_extent() fs/xfs/libxfs/xfs_alloc.c:3067
xfs_trans_free_extent() fs/xfs/xfs_extfree_item.c:370
xfs_efi_recover() fs/xfs/xfs_extfree_item.c:626
xlog_recover_process_efi() fs/xfs/xfs_log_recover.c:4605
xlog_recover_process_intents() fs/xfs/xfs_log_recover.c:4893
xlog_recover_finish() fs/xfs/xfs_log_recover.c:5824
xfs_log_mount_finish() fs/xfs/xfs_log.c:764
xfs_mountfs() fs/xfs/xfs_mount.c:978
xfs_fs_fill_super() fs/xfs/xfs_super.c:1908
mount_bdev() fs/super.c:1417
xfs_fs_mount() fs/xfs/xfs_super.c:1985
legacy_get_tree() fs/fs_context.c:647
vfs_get_tree() fs/super.c:1547
do_new_mount() fs/namespace.c:2843
do_mount() fs/namespace.c:3163
ksys_mount() fs/namespace.c:3372
__do_sys_mount() fs/namespace.c:3386
__se_sys_mount() fs/namespace.c:3383
__x64_sys_mount() fs/namespace.c:3383
do_syscall_64() arch/x86/entry/common.c:296
entry_SYSCALL_64() arch/x86/entry/entry_64.S:180
To avoid this deadlock, we should not block in
xfs_extent_busy_flush() if we hold a busy extent in the current
transaction.
Now that the EFI processing code can handle requeuing a partially
completed EFI, we can detect this situation in
xfs_extent_busy_flush() and return -EAGAIN rather than going to
sleep forever. The -EAGAIN get propagated back out to the
xfs_trans_free_extent() context, where the EFD is populated and the
transaction is rolled, thereby moving the busy extents into the CIL.
At this point, we can retry the extent free operation again with a
clean transaction. If we hit the same "all free extents are busy"
situation when trying to fix up the free list, we can safely call
xfs_extent_busy_flush() and wait for the busy extents to resolve
and wake us. At this point, the allocation search can make progress
again and we can fix up the free list.
This deadlock was first reported by Chandan in mid-2021, but I
couldn't make myself understood during review, and didn't have time
to fix it myself.
It was reported again in March 2023, and again I have found myself
unable to explain the complexities of the solution needed during
review.
As such, I don't have hours more time to waste trying to get the
fix written the way it needs to be written, so I'm just doing it
myself. This patchset is largely based on Wengang Wang's last patch,
but with all the unnecessary stuff removed, split up into multiple
patches and cleaned up somewhat.
Reported-by: Chandan Babu R <chandanrlinux@gmail.com>
Reported-by: Wengang Wang <wen.gang.wang@oracle.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
2023-06-29 02:04:33 +08:00
|
|
|
/*
|
|
|
|
* Our only valid extents must have been busy. Flush and
|
|
|
|
* retry the allocation again. If we get an -EAGAIN
|
|
|
|
* error, we're being told that a deadlock was avoided
|
|
|
|
* and the current transaction needs committing before
|
|
|
|
* the allocation can be retried.
|
|
|
|
*/
|
2011-04-25 03:06:15 +08:00
|
|
|
trace_xfs_alloc_size_busy(args);
|
xfs: don't block in busy flushing when freeing extents
If the current transaction holds a busy extent and we are trying to
allocate a new extent to fix up the free list, we can deadlock if
the AG is entirely empty except for the busy extent held by the
transaction.
This can occur at runtime processing an XEFI with multiple extents
in this path:
__schedule+0x22f at ffffffff81f75e8f
schedule+0x46 at ffffffff81f76366
xfs_extent_busy_flush+0x69 at ffffffff81477d99
xfs_alloc_ag_vextent_size+0x16a at ffffffff8141711a
xfs_alloc_ag_vextent+0x19b at ffffffff81417edb
xfs_alloc_fix_freelist+0x22f at ffffffff8141896f
xfs_free_extent_fix_freelist+0x6a at ffffffff8141939a
__xfs_free_extent+0x99 at ffffffff81419499
xfs_trans_free_extent+0x3e at ffffffff814a6fee
xfs_extent_free_finish_item+0x24 at ffffffff814a70d4
xfs_defer_finish_noroll+0x1f7 at ffffffff81441407
xfs_defer_finish+0x11 at ffffffff814417e1
xfs_itruncate_extents_flags+0x13d at ffffffff8148b7dd
xfs_inactive_truncate+0xb9 at ffffffff8148bb89
xfs_inactive+0x227 at ffffffff8148c4f7
xfs_fs_destroy_inode+0xb8 at ffffffff81496898
destroy_inode+0x3b at ffffffff8127d2ab
do_unlinkat+0x1d1 at ffffffff81270df1
do_syscall_64+0x40 at ffffffff81f6b5f0
entry_SYSCALL_64_after_hwframe+0x44 at ffffffff8200007c
This can also happen in log recovery when processing an EFI
with multiple extents through this path:
context_switch() kernel/sched/core.c:3881
__schedule() kernel/sched/core.c:5111
schedule() kernel/sched/core.c:5186
xfs_extent_busy_flush() fs/xfs/xfs_extent_busy.c:598
xfs_alloc_ag_vextent_size() fs/xfs/libxfs/xfs_alloc.c:1641
xfs_alloc_ag_vextent() fs/xfs/libxfs/xfs_alloc.c:828
xfs_alloc_fix_freelist() fs/xfs/libxfs/xfs_alloc.c:2362
xfs_free_extent_fix_freelist() fs/xfs/libxfs/xfs_alloc.c:3029
__xfs_free_extent() fs/xfs/libxfs/xfs_alloc.c:3067
xfs_trans_free_extent() fs/xfs/xfs_extfree_item.c:370
xfs_efi_recover() fs/xfs/xfs_extfree_item.c:626
xlog_recover_process_efi() fs/xfs/xfs_log_recover.c:4605
xlog_recover_process_intents() fs/xfs/xfs_log_recover.c:4893
xlog_recover_finish() fs/xfs/xfs_log_recover.c:5824
xfs_log_mount_finish() fs/xfs/xfs_log.c:764
xfs_mountfs() fs/xfs/xfs_mount.c:978
xfs_fs_fill_super() fs/xfs/xfs_super.c:1908
mount_bdev() fs/super.c:1417
xfs_fs_mount() fs/xfs/xfs_super.c:1985
legacy_get_tree() fs/fs_context.c:647
vfs_get_tree() fs/super.c:1547
do_new_mount() fs/namespace.c:2843
do_mount() fs/namespace.c:3163
ksys_mount() fs/namespace.c:3372
__do_sys_mount() fs/namespace.c:3386
__se_sys_mount() fs/namespace.c:3383
__x64_sys_mount() fs/namespace.c:3383
do_syscall_64() arch/x86/entry/common.c:296
entry_SYSCALL_64() arch/x86/entry/entry_64.S:180
To avoid this deadlock, we should not block in
xfs_extent_busy_flush() if we hold a busy extent in the current
transaction.
Now that the EFI processing code can handle requeuing a partially
completed EFI, we can detect this situation in
xfs_extent_busy_flush() and return -EAGAIN rather than going to
sleep forever. The -EAGAIN get propagated back out to the
xfs_trans_free_extent() context, where the EFD is populated and the
transaction is rolled, thereby moving the busy extents into the CIL.
At this point, we can retry the extent free operation again with a
clean transaction. If we hit the same "all free extents are busy"
situation when trying to fix up the free list, we can safely call
xfs_extent_busy_flush() and wait for the busy extents to resolve
and wake us. At this point, the allocation search can make progress
again and we can fix up the free list.
This deadlock was first reported by Chandan in mid-2021, but I
couldn't make myself understood during review, and didn't have time
to fix it myself.
It was reported again in March 2023, and again I have found myself
unable to explain the complexities of the solution needed during
review.
As such, I don't have hours more time to waste trying to get the
fix written the way it needs to be written, so I'm just doing it
myself. This patchset is largely based on Wengang Wang's last patch,
but with all the unnecessary stuff removed, split up into multiple
patches and cleaned up somewhat.
Reported-by: Chandan Babu R <chandanrlinux@gmail.com>
Reported-by: Wengang Wang <wen.gang.wang@oracle.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
2023-06-29 02:04:33 +08:00
|
|
|
error = xfs_extent_busy_flush(args->tp, args->pag,
|
|
|
|
busy_gen, alloc_flags);
|
|
|
|
if (error)
|
|
|
|
goto error0;
|
|
|
|
|
|
|
|
alloc_flags &= ~XFS_ALLOC_FLAG_TRYFLUSH;
|
|
|
|
xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
|
2011-04-25 03:06:15 +08:00
|
|
|
goto restart;
|
|
|
|
}
|
|
|
|
goto out_nominleft;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2011-04-25 03:06:15 +08:00
|
|
|
xfs_alloc_fix_len(args);
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
rlen = args->len;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(args->mp, rlen > flen)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Allocate and initialize a cursor for the by-block tree.
|
|
|
|
*/
|
2008-10-30 13:53:59 +08:00
|
|
|
bno_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
|
2021-06-02 08:48:24 +08:00
|
|
|
args->pag, XFS_BTNUM_BNO);
|
2005-04-17 06:20:36 +08:00
|
|
|
if ((error = xfs_alloc_fixup_trees(cnt_cur, bno_cur, fbno, flen,
|
|
|
|
rbno, rlen, XFSA_FIXUP_CNT_OK)))
|
|
|
|
goto error0;
|
|
|
|
xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
|
|
|
|
xfs_btree_del_cursor(bno_cur, XFS_BTREE_NOERROR);
|
|
|
|
cnt_cur = bno_cur = NULL;
|
|
|
|
args->len = rlen;
|
|
|
|
args->agbno = rbno;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(args->mp,
|
|
|
|
args->agbno + args->len >
|
2020-03-10 23:57:29 +08:00
|
|
|
be32_to_cpu(agf->agf_length))) {
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2009-12-15 07:14:59 +08:00
|
|
|
trace_xfs_alloc_size_done(args);
|
2005-04-17 06:20:36 +08:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
error0:
|
2009-12-15 07:14:59 +08:00
|
|
|
trace_xfs_alloc_size_error(args);
|
2005-04-17 06:20:36 +08:00
|
|
|
if (cnt_cur)
|
|
|
|
xfs_btree_del_cursor(cnt_cur, XFS_BTREE_ERROR);
|
|
|
|
if (bno_cur)
|
|
|
|
xfs_btree_del_cursor(bno_cur, XFS_BTREE_ERROR);
|
|
|
|
return error;
|
2011-04-25 03:06:15 +08:00
|
|
|
|
|
|
|
out_nominleft:
|
|
|
|
xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
|
|
|
|
trace_xfs_alloc_size_nominleft(args);
|
|
|
|
args->agbno = NULLAGBLOCK;
|
|
|
|
return 0;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Free the extent starting at agno/bno for length.
|
|
|
|
*/
|
2016-08-03 09:33:42 +08:00
|
|
|
STATIC int
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_free_ag_extent(
|
2018-12-13 00:46:23 +08:00
|
|
|
struct xfs_trans *tp,
|
|
|
|
struct xfs_buf *agbp,
|
|
|
|
xfs_agnumber_t agno,
|
|
|
|
xfs_agblock_t bno,
|
|
|
|
xfs_extlen_t len,
|
|
|
|
const struct xfs_owner_info *oinfo,
|
|
|
|
enum xfs_ag_resv_type type)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2018-12-13 00:46:23 +08:00
|
|
|
struct xfs_mount *mp;
|
|
|
|
struct xfs_btree_cur *bno_cur;
|
|
|
|
struct xfs_btree_cur *cnt_cur;
|
|
|
|
xfs_agblock_t gtbno; /* start of right neighbor */
|
|
|
|
xfs_extlen_t gtlen; /* length of right neighbor */
|
|
|
|
xfs_agblock_t ltbno; /* start of left neighbor */
|
|
|
|
xfs_extlen_t ltlen; /* length of left neighbor */
|
|
|
|
xfs_agblock_t nbno; /* new starting block of freesp */
|
|
|
|
xfs_extlen_t nlen; /* new length of freespace */
|
|
|
|
int haveleft; /* have a left neighbor */
|
|
|
|
int haveright; /* have a right neighbor */
|
|
|
|
int i;
|
|
|
|
int error;
|
2021-06-02 08:48:24 +08:00
|
|
|
struct xfs_perag *pag = agbp->b_pag;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2016-08-03 09:33:43 +08:00
|
|
|
bno_cur = cnt_cur = NULL;
|
2005-04-17 06:20:36 +08:00
|
|
|
mp = tp->t_mountp;
|
2016-08-03 09:33:43 +08:00
|
|
|
|
2017-12-08 11:07:27 +08:00
|
|
|
if (!xfs_rmap_should_skip_owner_update(oinfo)) {
|
2021-06-02 08:48:24 +08:00
|
|
|
error = xfs_rmap_free(tp, agbp, pag, bno, len, oinfo);
|
2016-08-03 09:33:43 +08:00
|
|
|
if (error)
|
|
|
|
goto error0;
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Allocate and initialize a cursor for the by-block btree.
|
|
|
|
*/
|
2021-06-02 08:48:24 +08:00
|
|
|
bno_cur = xfs_allocbt_init_cursor(mp, tp, agbp, pag, XFS_BTNUM_BNO);
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Look for a neighboring block on the left (lower block numbers)
|
|
|
|
* that is contiguous with this space.
|
|
|
|
*/
|
|
|
|
if ((error = xfs_alloc_lookup_le(bno_cur, bno, len, &haveleft)))
|
|
|
|
goto error0;
|
|
|
|
if (haveleft) {
|
|
|
|
/*
|
|
|
|
* There is a block to our left.
|
|
|
|
*/
|
|
|
|
if ((error = xfs_alloc_get_rec(bno_cur, <bno, <len, &i)))
|
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* It's not contiguous, though.
|
|
|
|
*/
|
|
|
|
if (ltbno + ltlen < bno)
|
|
|
|
haveleft = 0;
|
|
|
|
else {
|
|
|
|
/*
|
|
|
|
* If this failure happens the request to free this
|
|
|
|
* space was invalid, it's (partly) already free.
|
|
|
|
* Very bad.
|
|
|
|
*/
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, ltbno + ltlen > bno)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Look for a neighboring block on the right (higher block numbers)
|
|
|
|
* that is contiguous with this space.
|
|
|
|
*/
|
2008-10-30 13:55:45 +08:00
|
|
|
if ((error = xfs_btree_increment(bno_cur, 0, &haveright)))
|
2005-04-17 06:20:36 +08:00
|
|
|
goto error0;
|
|
|
|
if (haveright) {
|
|
|
|
/*
|
|
|
|
* There is a block to our right.
|
|
|
|
*/
|
|
|
|
if ((error = xfs_alloc_get_rec(bno_cur, >bno, >len, &i)))
|
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* It's not contiguous, though.
|
|
|
|
*/
|
|
|
|
if (bno + len < gtbno)
|
|
|
|
haveright = 0;
|
|
|
|
else {
|
|
|
|
/*
|
|
|
|
* If this failure happens the request to free this
|
|
|
|
* space was invalid, it's (partly) already free.
|
|
|
|
* Very bad.
|
|
|
|
*/
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, bno + len > gtbno)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Now allocate and initialize a cursor for the by-size tree.
|
|
|
|
*/
|
2021-06-02 08:48:24 +08:00
|
|
|
cnt_cur = xfs_allocbt_init_cursor(mp, tp, agbp, pag, XFS_BTNUM_CNT);
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Have both left and right contiguous neighbors.
|
|
|
|
* Merge all three into a single free block.
|
|
|
|
*/
|
|
|
|
if (haveleft && haveright) {
|
|
|
|
/*
|
|
|
|
* Delete the old by-size entry on the left.
|
|
|
|
*/
|
|
|
|
if ((error = xfs_alloc_lookup_eq(cnt_cur, ltbno, ltlen, &i)))
|
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2008-10-30 13:58:01 +08:00
|
|
|
if ((error = xfs_btree_delete(cnt_cur, &i)))
|
2005-04-17 06:20:36 +08:00
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Delete the old by-size entry on the right.
|
|
|
|
*/
|
|
|
|
if ((error = xfs_alloc_lookup_eq(cnt_cur, gtbno, gtlen, &i)))
|
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2008-10-30 13:58:01 +08:00
|
|
|
if ((error = xfs_btree_delete(cnt_cur, &i)))
|
2005-04-17 06:20:36 +08:00
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Delete the old by-block entry for the right block.
|
|
|
|
*/
|
2008-10-30 13:58:01 +08:00
|
|
|
if ((error = xfs_btree_delete(bno_cur, &i)))
|
2005-04-17 06:20:36 +08:00
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Move the by-block cursor back to the left neighbor.
|
|
|
|
*/
|
2008-10-30 13:55:58 +08:00
|
|
|
if ((error = xfs_btree_decrement(bno_cur, 0, &i)))
|
2005-04-17 06:20:36 +08:00
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
#ifdef DEBUG
|
|
|
|
/*
|
|
|
|
* Check that this is the right record: delete didn't
|
|
|
|
* mangle the cursor.
|
|
|
|
*/
|
|
|
|
{
|
|
|
|
xfs_agblock_t xxbno;
|
|
|
|
xfs_extlen_t xxlen;
|
|
|
|
|
|
|
|
if ((error = xfs_alloc_get_rec(bno_cur, &xxbno, &xxlen,
|
|
|
|
&i)))
|
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp,
|
|
|
|
i != 1 ||
|
|
|
|
xxbno != ltbno ||
|
|
|
|
xxlen != ltlen)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
/*
|
|
|
|
* Update remaining by-block entry to the new, joined block.
|
|
|
|
*/
|
|
|
|
nbno = ltbno;
|
|
|
|
nlen = len + ltlen + gtlen;
|
|
|
|
if ((error = xfs_alloc_update(bno_cur, nbno, nlen)))
|
|
|
|
goto error0;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Have only a left contiguous neighbor.
|
|
|
|
* Merge it together with the new freespace.
|
|
|
|
*/
|
|
|
|
else if (haveleft) {
|
|
|
|
/*
|
|
|
|
* Delete the old by-size entry on the left.
|
|
|
|
*/
|
|
|
|
if ((error = xfs_alloc_lookup_eq(cnt_cur, ltbno, ltlen, &i)))
|
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2008-10-30 13:58:01 +08:00
|
|
|
if ((error = xfs_btree_delete(cnt_cur, &i)))
|
2005-04-17 06:20:36 +08:00
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Back up the by-block cursor to the left neighbor, and
|
|
|
|
* update its length.
|
|
|
|
*/
|
2008-10-30 13:55:58 +08:00
|
|
|
if ((error = xfs_btree_decrement(bno_cur, 0, &i)))
|
2005-04-17 06:20:36 +08:00
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
nbno = ltbno;
|
|
|
|
nlen = len + ltlen;
|
|
|
|
if ((error = xfs_alloc_update(bno_cur, nbno, nlen)))
|
|
|
|
goto error0;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Have only a right contiguous neighbor.
|
|
|
|
* Merge it together with the new freespace.
|
|
|
|
*/
|
|
|
|
else if (haveright) {
|
|
|
|
/*
|
|
|
|
* Delete the old by-size entry on the right.
|
|
|
|
*/
|
|
|
|
if ((error = xfs_alloc_lookup_eq(cnt_cur, gtbno, gtlen, &i)))
|
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2008-10-30 13:58:01 +08:00
|
|
|
if ((error = xfs_btree_delete(cnt_cur, &i)))
|
2005-04-17 06:20:36 +08:00
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Update the starting block and length of the right
|
|
|
|
* neighbor in the by-block tree.
|
|
|
|
*/
|
|
|
|
nbno = bno;
|
|
|
|
nlen = len + gtlen;
|
|
|
|
if ((error = xfs_alloc_update(bno_cur, nbno, nlen)))
|
|
|
|
goto error0;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* No contiguous neighbors.
|
|
|
|
* Insert the new freespace into the by-block tree.
|
|
|
|
*/
|
|
|
|
else {
|
|
|
|
nbno = bno;
|
|
|
|
nlen = len;
|
2008-10-30 13:57:40 +08:00
|
|
|
if ((error = xfs_btree_insert(bno_cur, &i)))
|
2005-04-17 06:20:36 +08:00
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
xfs_btree_del_cursor(bno_cur, XFS_BTREE_NOERROR);
|
|
|
|
bno_cur = NULL;
|
|
|
|
/*
|
|
|
|
* In all cases we need to insert the new freespace in the by-size tree.
|
|
|
|
*/
|
|
|
|
if ((error = xfs_alloc_lookup_eq(cnt_cur, nbno, nlen, &i)))
|
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 0)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2008-10-30 13:57:40 +08:00
|
|
|
if ((error = xfs_btree_insert(cnt_cur, &i)))
|
2005-04-17 06:20:36 +08:00
|
|
|
goto error0;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, i != 1)) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto error0;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
|
|
|
|
cnt_cur = NULL;
|
2011-03-04 20:59:55 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Update the freespace totals in the ag and superblock.
|
|
|
|
*/
|
2020-07-14 00:13:00 +08:00
|
|
|
error = xfs_alloc_update_counters(tp, agbp, len);
|
|
|
|
xfs_ag_resv_free_extent(agbp->b_pag, type, tp, len);
|
2011-03-04 20:59:55 +08:00
|
|
|
if (error)
|
|
|
|
goto error0;
|
|
|
|
|
2015-10-12 15:21:22 +08:00
|
|
|
XFS_STATS_INC(mp, xs_freex);
|
|
|
|
XFS_STATS_ADD(mp, xs_freeb, len);
|
2009-12-15 07:14:59 +08:00
|
|
|
|
2018-03-10 06:01:59 +08:00
|
|
|
trace_xfs_free_extent(mp, agno, bno, len, type, haveleft, haveright);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
error0:
|
2018-03-10 06:01:59 +08:00
|
|
|
trace_xfs_free_extent(mp, agno, bno, len, type, -1, -1);
|
2005-04-17 06:20:36 +08:00
|
|
|
if (bno_cur)
|
|
|
|
xfs_btree_del_cursor(bno_cur, XFS_BTREE_ERROR);
|
|
|
|
if (cnt_cur)
|
|
|
|
xfs_btree_del_cursor(cnt_cur, XFS_BTREE_ERROR);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Visible (exported) allocation/free functions.
|
|
|
|
* Some of these are used just by xfs_alloc_btree.c and this file.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
2021-10-14 01:02:19 +08:00
|
|
|
* Compute and fill in value of m_alloc_maxlevels.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
|
|
|
void
|
|
|
|
xfs_alloc_compute_maxlevels(
|
|
|
|
xfs_mount_t *mp) /* file system mount structure */
|
|
|
|
{
|
2021-10-14 01:02:19 +08:00
|
|
|
mp->m_alloc_maxlevels = xfs_btree_compute_maxlevels(mp->m_alloc_mnr,
|
2016-06-21 09:53:28 +08:00
|
|
|
(mp->m_sb.sb_agblocks + 1) / 2);
|
2021-09-24 01:32:06 +08:00
|
|
|
ASSERT(mp->m_alloc_maxlevels <= xfs_allocbt_maxlevels_ondisk());
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2009-03-16 15:29:46 +08:00
|
|
|
/*
|
xfs: set up per-AG free space reservations
One unfortunate quirk of the reference count and reverse mapping
btrees -- they can expand in size when blocks are written to *other*
allocation groups if, say, one large extent becomes a lot of tiny
extents. Since we don't want to start throwing errors in the middle
of CoWing, we need to reserve some blocks to handle future expansion.
The transaction block reservation counters aren't sufficient here
because we have to have a reserve of blocks in every AG, not just
somewhere in the filesystem.
Therefore, create two per-AG block reservation pools. One feeds the
AGFL so that rmapbt expansion always succeeds, and the other feeds all
other metadata so that refcountbt expansion never fails.
Use the count of how many reserved blocks we need to have on hand to
create a virtual reservation in the AG. Through selective clamping of
the maximum length of allocation requests and of the length of the
longest free extent, we can make it look like there's less free space
in the AG unless the reservation owner is asking for blocks.
In other words, play some accounting tricks in-core to make sure that
we always have blocks available. On the plus side, there's nothing to
clean up if we crash, which is contrast to the strategy that the rough
draft used (actually removing extents from the freespace btrees).
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-09-19 08:30:52 +08:00
|
|
|
* Find the length of the longest extent in an AG. The 'need' parameter
|
|
|
|
* specifies how much space we're going to need for the AGFL and the
|
|
|
|
* 'reserved' parameter tells us how many blocks in this AG are reserved for
|
|
|
|
* other callers.
|
2009-03-16 15:29:46 +08:00
|
|
|
*/
|
|
|
|
xfs_extlen_t
|
|
|
|
xfs_alloc_longest_free_extent(
|
2015-06-22 08:04:31 +08:00
|
|
|
struct xfs_perag *pag,
|
xfs: set up per-AG free space reservations
One unfortunate quirk of the reference count and reverse mapping
btrees -- they can expand in size when blocks are written to *other*
allocation groups if, say, one large extent becomes a lot of tiny
extents. Since we don't want to start throwing errors in the middle
of CoWing, we need to reserve some blocks to handle future expansion.
The transaction block reservation counters aren't sufficient here
because we have to have a reserve of blocks in every AG, not just
somewhere in the filesystem.
Therefore, create two per-AG block reservation pools. One feeds the
AGFL so that rmapbt expansion always succeeds, and the other feeds all
other metadata so that refcountbt expansion never fails.
Use the count of how many reserved blocks we need to have on hand to
create a virtual reservation in the AG. Through selective clamping of
the maximum length of allocation requests and of the length of the
longest free extent, we can make it look like there's less free space
in the AG unless the reservation owner is asking for blocks.
In other words, play some accounting tricks in-core to make sure that
we always have blocks available. On the plus side, there's nothing to
clean up if we crash, which is contrast to the strategy that the rough
draft used (actually removing extents from the freespace btrees).
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-09-19 08:30:52 +08:00
|
|
|
xfs_extlen_t need,
|
|
|
|
xfs_extlen_t reserved)
|
2009-03-16 15:29:46 +08:00
|
|
|
{
|
2015-06-22 08:04:31 +08:00
|
|
|
xfs_extlen_t delta = 0;
|
2009-03-16 15:29:46 +08:00
|
|
|
|
xfs: set up per-AG free space reservations
One unfortunate quirk of the reference count and reverse mapping
btrees -- they can expand in size when blocks are written to *other*
allocation groups if, say, one large extent becomes a lot of tiny
extents. Since we don't want to start throwing errors in the middle
of CoWing, we need to reserve some blocks to handle future expansion.
The transaction block reservation counters aren't sufficient here
because we have to have a reserve of blocks in every AG, not just
somewhere in the filesystem.
Therefore, create two per-AG block reservation pools. One feeds the
AGFL so that rmapbt expansion always succeeds, and the other feeds all
other metadata so that refcountbt expansion never fails.
Use the count of how many reserved blocks we need to have on hand to
create a virtual reservation in the AG. Through selective clamping of
the maximum length of allocation requests and of the length of the
longest free extent, we can make it look like there's less free space
in the AG unless the reservation owner is asking for blocks.
In other words, play some accounting tricks in-core to make sure that
we always have blocks available. On the plus side, there's nothing to
clean up if we crash, which is contrast to the strategy that the rough
draft used (actually removing extents from the freespace btrees).
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-09-19 08:30:52 +08:00
|
|
|
/*
|
|
|
|
* If the AGFL needs a recharge, we'll have to subtract that from the
|
|
|
|
* longest extent.
|
|
|
|
*/
|
2009-03-16 15:29:46 +08:00
|
|
|
if (need > pag->pagf_flcount)
|
|
|
|
delta = need - pag->pagf_flcount;
|
|
|
|
|
xfs: set up per-AG free space reservations
One unfortunate quirk of the reference count and reverse mapping
btrees -- they can expand in size when blocks are written to *other*
allocation groups if, say, one large extent becomes a lot of tiny
extents. Since we don't want to start throwing errors in the middle
of CoWing, we need to reserve some blocks to handle future expansion.
The transaction block reservation counters aren't sufficient here
because we have to have a reserve of blocks in every AG, not just
somewhere in the filesystem.
Therefore, create two per-AG block reservation pools. One feeds the
AGFL so that rmapbt expansion always succeeds, and the other feeds all
other metadata so that refcountbt expansion never fails.
Use the count of how many reserved blocks we need to have on hand to
create a virtual reservation in the AG. Through selective clamping of
the maximum length of allocation requests and of the length of the
longest free extent, we can make it look like there's less free space
in the AG unless the reservation owner is asking for blocks.
In other words, play some accounting tricks in-core to make sure that
we always have blocks available. On the plus side, there's nothing to
clean up if we crash, which is contrast to the strategy that the rough
draft used (actually removing extents from the freespace btrees).
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-09-19 08:30:52 +08:00
|
|
|
/*
|
|
|
|
* If we cannot maintain others' reservations with space from the
|
|
|
|
* not-longest freesp extents, we'll have to subtract /that/ from
|
|
|
|
* the longest extent too.
|
|
|
|
*/
|
|
|
|
if (pag->pagf_freeblks - pag->pagf_longest < reserved)
|
|
|
|
delta += reserved - (pag->pagf_freeblks - pag->pagf_longest);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the longest extent is long enough to satisfy all the
|
|
|
|
* reservations and AGFL rules in place, we can return this extent.
|
|
|
|
*/
|
2009-03-16 15:29:46 +08:00
|
|
|
if (pag->pagf_longest > delta)
|
2019-10-22 00:26:34 +08:00
|
|
|
return min_t(xfs_extlen_t, pag->pag_mount->m_ag_max_usable,
|
|
|
|
pag->pagf_longest - delta);
|
xfs: set up per-AG free space reservations
One unfortunate quirk of the reference count and reverse mapping
btrees -- they can expand in size when blocks are written to *other*
allocation groups if, say, one large extent becomes a lot of tiny
extents. Since we don't want to start throwing errors in the middle
of CoWing, we need to reserve some blocks to handle future expansion.
The transaction block reservation counters aren't sufficient here
because we have to have a reserve of blocks in every AG, not just
somewhere in the filesystem.
Therefore, create two per-AG block reservation pools. One feeds the
AGFL so that rmapbt expansion always succeeds, and the other feeds all
other metadata so that refcountbt expansion never fails.
Use the count of how many reserved blocks we need to have on hand to
create a virtual reservation in the AG. Through selective clamping of
the maximum length of allocation requests and of the length of the
longest free extent, we can make it look like there's less free space
in the AG unless the reservation owner is asking for blocks.
In other words, play some accounting tricks in-core to make sure that
we always have blocks available. On the plus side, there's nothing to
clean up if we crash, which is contrast to the strategy that the rough
draft used (actually removing extents from the freespace btrees).
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-09-19 08:30:52 +08:00
|
|
|
|
|
|
|
/* Otherwise, let the caller try for 1 block if there's space. */
|
2009-03-16 15:29:46 +08:00
|
|
|
return pag->pagf_flcount > 0 || pag->pagf_longest > 0;
|
|
|
|
}
|
|
|
|
|
2019-12-19 03:09:55 +08:00
|
|
|
/*
|
|
|
|
* Compute the minimum length of the AGFL in the given AG. If @pag is NULL,
|
|
|
|
* return the largest possible minimum length.
|
|
|
|
*/
|
2015-06-22 08:13:30 +08:00
|
|
|
unsigned int
|
|
|
|
xfs_alloc_min_freelist(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
struct xfs_perag *pag)
|
|
|
|
{
|
2019-12-19 03:09:55 +08:00
|
|
|
/* AG btrees have at least 1 level. */
|
|
|
|
static const uint8_t fake_levels[XFS_BTNUM_AGF] = {1, 1, 1};
|
|
|
|
const uint8_t *levels = pag ? pag->pagf_levels : fake_levels;
|
2015-06-22 08:13:30 +08:00
|
|
|
unsigned int min_free;
|
|
|
|
|
2021-10-14 01:02:19 +08:00
|
|
|
ASSERT(mp->m_alloc_maxlevels > 0);
|
2019-12-19 03:09:55 +08:00
|
|
|
|
xfs: fix internal error from AGFL exhaustion
commit f63a5b3769ad7659da4c0420751d78958ab97675 upstream.
We've been seeing XFS errors like the following:
XFS: Internal error i != 1 at line 3526 of file fs/xfs/libxfs/xfs_btree.c. Caller xfs_btree_insert+0x1ec/0x280
...
Call Trace:
xfs_corruption_error+0x94/0xa0
xfs_btree_insert+0x221/0x280
xfs_alloc_fixup_trees+0x104/0x3e0
xfs_alloc_ag_vextent_size+0x667/0x820
xfs_alloc_fix_freelist+0x5d9/0x750
xfs_free_extent_fix_freelist+0x65/0xa0
__xfs_free_extent+0x57/0x180
...
This is the XFS_IS_CORRUPT() check in xfs_btree_insert() when
xfs_btree_insrec() fails.
After converting this into a panic and dissecting the core dump, I found
that xfs_btree_insrec() is failing because it's trying to split a leaf
node in the cntbt when the AG free list is empty. In particular, it's
failing to get a block from the AGFL _while trying to refill the AGFL_.
If a single operation splits every level of the bnobt and the cntbt (and
the rmapbt if it is enabled) at once, the free list will be empty. Then,
when the next operation tries to refill the free list, it allocates
space. If the allocation does not use a full extent, it will need to
insert records for the remaining space in the bnobt and cntbt. And if
those new records go in full leaves, the leaves (and potentially more
nodes up to the old root) need to be split.
Fix it by accounting for the additional splits that may be required to
refill the free list in the calculation for the minimum free list size.
P.S. As far as I can tell, this bug has existed for a long time -- maybe
back to xfs-history commit afdf80ae7405 ("Add XFS_AG_MAXLEVELS macros
...") in April 1994! It requires a very unlucky sequence of events, and
in fact we didn't hit it until a particular sparse mmap workload updated
from 5.12 to 5.19. But this bug existed in 5.12, so it must've been
exposed by some other change in allocation or writeback patterns. It's
also much less likely to be hit with the rmapbt enabled, since that
increases the minimum free list size and is unlikely to split at the
same time as the bnobt and cntbt.
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Chandan Babu R <chandanbabu@kernel.org>
Signed-off-by: Catherine Hoang <catherine.hoang@oracle.com>
Acked-by: Chandan Babu R <chandanbabu@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2024-02-09 07:20:48 +08:00
|
|
|
/*
|
|
|
|
* For a btree shorter than the maximum height, the worst case is that
|
|
|
|
* every level gets split and a new level is added, then while inserting
|
|
|
|
* another entry to refill the AGFL, every level under the old root gets
|
|
|
|
* split again. This is:
|
|
|
|
*
|
|
|
|
* (full height split reservation) + (AGFL refill split height)
|
|
|
|
* = (current height + 1) + (current height - 1)
|
|
|
|
* = (new height) + (new height - 2)
|
|
|
|
* = 2 * new height - 2
|
|
|
|
*
|
|
|
|
* For a btree of maximum height, the worst case is that every level
|
|
|
|
* under the root gets split, then while inserting another entry to
|
|
|
|
* refill the AGFL, every level under the root gets split again. This is
|
|
|
|
* also:
|
|
|
|
*
|
|
|
|
* 2 * (current height - 1)
|
|
|
|
* = 2 * (new height - 1)
|
|
|
|
* = 2 * new height - 2
|
|
|
|
*/
|
|
|
|
|
2015-06-22 08:13:30 +08:00
|
|
|
/* space needed by-bno freespace btree */
|
2019-12-19 03:09:55 +08:00
|
|
|
min_free = min_t(unsigned int, levels[XFS_BTNUM_BNOi] + 1,
|
xfs: fix internal error from AGFL exhaustion
commit f63a5b3769ad7659da4c0420751d78958ab97675 upstream.
We've been seeing XFS errors like the following:
XFS: Internal error i != 1 at line 3526 of file fs/xfs/libxfs/xfs_btree.c. Caller xfs_btree_insert+0x1ec/0x280
...
Call Trace:
xfs_corruption_error+0x94/0xa0
xfs_btree_insert+0x221/0x280
xfs_alloc_fixup_trees+0x104/0x3e0
xfs_alloc_ag_vextent_size+0x667/0x820
xfs_alloc_fix_freelist+0x5d9/0x750
xfs_free_extent_fix_freelist+0x65/0xa0
__xfs_free_extent+0x57/0x180
...
This is the XFS_IS_CORRUPT() check in xfs_btree_insert() when
xfs_btree_insrec() fails.
After converting this into a panic and dissecting the core dump, I found
that xfs_btree_insrec() is failing because it's trying to split a leaf
node in the cntbt when the AG free list is empty. In particular, it's
failing to get a block from the AGFL _while trying to refill the AGFL_.
If a single operation splits every level of the bnobt and the cntbt (and
the rmapbt if it is enabled) at once, the free list will be empty. Then,
when the next operation tries to refill the free list, it allocates
space. If the allocation does not use a full extent, it will need to
insert records for the remaining space in the bnobt and cntbt. And if
those new records go in full leaves, the leaves (and potentially more
nodes up to the old root) need to be split.
Fix it by accounting for the additional splits that may be required to
refill the free list in the calculation for the minimum free list size.
P.S. As far as I can tell, this bug has existed for a long time -- maybe
back to xfs-history commit afdf80ae7405 ("Add XFS_AG_MAXLEVELS macros
...") in April 1994! It requires a very unlucky sequence of events, and
in fact we didn't hit it until a particular sparse mmap workload updated
from 5.12 to 5.19. But this bug existed in 5.12, so it must've been
exposed by some other change in allocation or writeback patterns. It's
also much less likely to be hit with the rmapbt enabled, since that
increases the minimum free list size and is unlikely to split at the
same time as the bnobt and cntbt.
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Chandan Babu R <chandanbabu@kernel.org>
Signed-off-by: Catherine Hoang <catherine.hoang@oracle.com>
Acked-by: Chandan Babu R <chandanbabu@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2024-02-09 07:20:48 +08:00
|
|
|
mp->m_alloc_maxlevels) * 2 - 2;
|
2015-06-22 08:13:30 +08:00
|
|
|
/* space needed by-size freespace btree */
|
2019-12-19 03:09:55 +08:00
|
|
|
min_free += min_t(unsigned int, levels[XFS_BTNUM_CNTi] + 1,
|
xfs: fix internal error from AGFL exhaustion
commit f63a5b3769ad7659da4c0420751d78958ab97675 upstream.
We've been seeing XFS errors like the following:
XFS: Internal error i != 1 at line 3526 of file fs/xfs/libxfs/xfs_btree.c. Caller xfs_btree_insert+0x1ec/0x280
...
Call Trace:
xfs_corruption_error+0x94/0xa0
xfs_btree_insert+0x221/0x280
xfs_alloc_fixup_trees+0x104/0x3e0
xfs_alloc_ag_vextent_size+0x667/0x820
xfs_alloc_fix_freelist+0x5d9/0x750
xfs_free_extent_fix_freelist+0x65/0xa0
__xfs_free_extent+0x57/0x180
...
This is the XFS_IS_CORRUPT() check in xfs_btree_insert() when
xfs_btree_insrec() fails.
After converting this into a panic and dissecting the core dump, I found
that xfs_btree_insrec() is failing because it's trying to split a leaf
node in the cntbt when the AG free list is empty. In particular, it's
failing to get a block from the AGFL _while trying to refill the AGFL_.
If a single operation splits every level of the bnobt and the cntbt (and
the rmapbt if it is enabled) at once, the free list will be empty. Then,
when the next operation tries to refill the free list, it allocates
space. If the allocation does not use a full extent, it will need to
insert records for the remaining space in the bnobt and cntbt. And if
those new records go in full leaves, the leaves (and potentially more
nodes up to the old root) need to be split.
Fix it by accounting for the additional splits that may be required to
refill the free list in the calculation for the minimum free list size.
P.S. As far as I can tell, this bug has existed for a long time -- maybe
back to xfs-history commit afdf80ae7405 ("Add XFS_AG_MAXLEVELS macros
...") in April 1994! It requires a very unlucky sequence of events, and
in fact we didn't hit it until a particular sparse mmap workload updated
from 5.12 to 5.19. But this bug existed in 5.12, so it must've been
exposed by some other change in allocation or writeback patterns. It's
also much less likely to be hit with the rmapbt enabled, since that
increases the minimum free list size and is unlikely to split at the
same time as the bnobt and cntbt.
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Chandan Babu R <chandanbabu@kernel.org>
Signed-off-by: Catherine Hoang <catherine.hoang@oracle.com>
Acked-by: Chandan Babu R <chandanbabu@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2024-02-09 07:20:48 +08:00
|
|
|
mp->m_alloc_maxlevels) * 2 - 2;
|
2016-08-03 09:38:24 +08:00
|
|
|
/* space needed reverse mapping used space btree */
|
2021-08-19 09:46:55 +08:00
|
|
|
if (xfs_has_rmapbt(mp))
|
2019-12-19 03:09:55 +08:00
|
|
|
min_free += min_t(unsigned int, levels[XFS_BTNUM_RMAPi] + 1,
|
xfs: fix internal error from AGFL exhaustion
commit f63a5b3769ad7659da4c0420751d78958ab97675 upstream.
We've been seeing XFS errors like the following:
XFS: Internal error i != 1 at line 3526 of file fs/xfs/libxfs/xfs_btree.c. Caller xfs_btree_insert+0x1ec/0x280
...
Call Trace:
xfs_corruption_error+0x94/0xa0
xfs_btree_insert+0x221/0x280
xfs_alloc_fixup_trees+0x104/0x3e0
xfs_alloc_ag_vextent_size+0x667/0x820
xfs_alloc_fix_freelist+0x5d9/0x750
xfs_free_extent_fix_freelist+0x65/0xa0
__xfs_free_extent+0x57/0x180
...
This is the XFS_IS_CORRUPT() check in xfs_btree_insert() when
xfs_btree_insrec() fails.
After converting this into a panic and dissecting the core dump, I found
that xfs_btree_insrec() is failing because it's trying to split a leaf
node in the cntbt when the AG free list is empty. In particular, it's
failing to get a block from the AGFL _while trying to refill the AGFL_.
If a single operation splits every level of the bnobt and the cntbt (and
the rmapbt if it is enabled) at once, the free list will be empty. Then,
when the next operation tries to refill the free list, it allocates
space. If the allocation does not use a full extent, it will need to
insert records for the remaining space in the bnobt and cntbt. And if
those new records go in full leaves, the leaves (and potentially more
nodes up to the old root) need to be split.
Fix it by accounting for the additional splits that may be required to
refill the free list in the calculation for the minimum free list size.
P.S. As far as I can tell, this bug has existed for a long time -- maybe
back to xfs-history commit afdf80ae7405 ("Add XFS_AG_MAXLEVELS macros
...") in April 1994! It requires a very unlucky sequence of events, and
in fact we didn't hit it until a particular sparse mmap workload updated
from 5.12 to 5.19. But this bug existed in 5.12, so it must've been
exposed by some other change in allocation or writeback patterns. It's
also much less likely to be hit with the rmapbt enabled, since that
increases the minimum free list size and is unlikely to split at the
same time as the bnobt and cntbt.
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Chandan Babu R <chandanbabu@kernel.org>
Signed-off-by: Catherine Hoang <catherine.hoang@oracle.com>
Acked-by: Chandan Babu R <chandanbabu@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2024-02-09 07:20:48 +08:00
|
|
|
mp->m_rmap_maxlevels) * 2 - 2;
|
2015-06-22 08:13:30 +08:00
|
|
|
|
|
|
|
return min_free;
|
|
|
|
}
|
|
|
|
|
2015-06-22 08:04:42 +08:00
|
|
|
/*
|
|
|
|
* Check if the operation we are fixing up the freelist for should go ahead or
|
|
|
|
* not. If we are freeing blocks, we always allow it, otherwise the allocation
|
|
|
|
* is dependent on whether the size and shape of free space available will
|
|
|
|
* permit the requested allocation to take place.
|
|
|
|
*/
|
|
|
|
static bool
|
|
|
|
xfs_alloc_space_available(
|
|
|
|
struct xfs_alloc_arg *args,
|
|
|
|
xfs_extlen_t min_free,
|
|
|
|
int flags)
|
|
|
|
{
|
|
|
|
struct xfs_perag *pag = args->pag;
|
2017-01-10 05:39:35 +08:00
|
|
|
xfs_extlen_t alloc_len, longest;
|
xfs: set up per-AG free space reservations
One unfortunate quirk of the reference count and reverse mapping
btrees -- they can expand in size when blocks are written to *other*
allocation groups if, say, one large extent becomes a lot of tiny
extents. Since we don't want to start throwing errors in the middle
of CoWing, we need to reserve some blocks to handle future expansion.
The transaction block reservation counters aren't sufficient here
because we have to have a reserve of blocks in every AG, not just
somewhere in the filesystem.
Therefore, create two per-AG block reservation pools. One feeds the
AGFL so that rmapbt expansion always succeeds, and the other feeds all
other metadata so that refcountbt expansion never fails.
Use the count of how many reserved blocks we need to have on hand to
create a virtual reservation in the AG. Through selective clamping of
the maximum length of allocation requests and of the length of the
longest free extent, we can make it look like there's less free space
in the AG unless the reservation owner is asking for blocks.
In other words, play some accounting tricks in-core to make sure that
we always have blocks available. On the plus side, there's nothing to
clean up if we crash, which is contrast to the strategy that the rough
draft used (actually removing extents from the freespace btrees).
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-09-19 08:30:52 +08:00
|
|
|
xfs_extlen_t reservation; /* blocks that are still reserved */
|
2015-06-22 08:04:42 +08:00
|
|
|
int available;
|
2019-04-12 22:39:21 +08:00
|
|
|
xfs_extlen_t agflcount;
|
2015-06-22 08:04:42 +08:00
|
|
|
|
|
|
|
if (flags & XFS_ALLOC_FLAG_FREEING)
|
|
|
|
return true;
|
|
|
|
|
xfs: set up per-AG free space reservations
One unfortunate quirk of the reference count and reverse mapping
btrees -- they can expand in size when blocks are written to *other*
allocation groups if, say, one large extent becomes a lot of tiny
extents. Since we don't want to start throwing errors in the middle
of CoWing, we need to reserve some blocks to handle future expansion.
The transaction block reservation counters aren't sufficient here
because we have to have a reserve of blocks in every AG, not just
somewhere in the filesystem.
Therefore, create two per-AG block reservation pools. One feeds the
AGFL so that rmapbt expansion always succeeds, and the other feeds all
other metadata so that refcountbt expansion never fails.
Use the count of how many reserved blocks we need to have on hand to
create a virtual reservation in the AG. Through selective clamping of
the maximum length of allocation requests and of the length of the
longest free extent, we can make it look like there's less free space
in the AG unless the reservation owner is asking for blocks.
In other words, play some accounting tricks in-core to make sure that
we always have blocks available. On the plus side, there's nothing to
clean up if we crash, which is contrast to the strategy that the rough
draft used (actually removing extents from the freespace btrees).
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-09-19 08:30:52 +08:00
|
|
|
reservation = xfs_ag_resv_needed(pag, args->resv);
|
|
|
|
|
2015-06-22 08:04:42 +08:00
|
|
|
/* do we have enough contiguous free space for the allocation? */
|
2017-01-10 05:39:35 +08:00
|
|
|
alloc_len = args->minlen + (args->alignment - 1) + args->minalignslop;
|
2018-04-07 01:09:42 +08:00
|
|
|
longest = xfs_alloc_longest_free_extent(pag, min_free, reservation);
|
2017-01-10 05:39:35 +08:00
|
|
|
if (longest < alloc_len)
|
2015-06-22 08:04:42 +08:00
|
|
|
return false;
|
|
|
|
|
2019-04-12 22:39:21 +08:00
|
|
|
/*
|
|
|
|
* Do we have enough free space remaining for the allocation? Don't
|
|
|
|
* account extra agfl blocks because we are about to defer free them,
|
|
|
|
* making them unavailable until the current transaction commits.
|
|
|
|
*/
|
|
|
|
agflcount = min_t(xfs_extlen_t, pag->pagf_flcount, min_free);
|
|
|
|
available = (int)(pag->pagf_freeblks + agflcount -
|
2017-01-10 05:44:30 +08:00
|
|
|
reservation - min_free - args->minleft);
|
2017-01-10 05:39:35 +08:00
|
|
|
if (available < (int)max(args->total, alloc_len))
|
2015-06-22 08:04:42 +08:00
|
|
|
return false;
|
|
|
|
|
2017-01-10 05:44:30 +08:00
|
|
|
/*
|
|
|
|
* Clamp maxlen to the amount of free space available for the actual
|
|
|
|
* extent allocation.
|
|
|
|
*/
|
|
|
|
if (available < (int)args->maxlen && !(flags & XFS_ALLOC_FLAG_CHECK)) {
|
|
|
|
args->maxlen = available;
|
|
|
|
ASSERT(args->maxlen > 0);
|
|
|
|
ASSERT(args->maxlen >= args->minlen);
|
|
|
|
}
|
|
|
|
|
2015-06-22 08:04:42 +08:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2018-05-08 08:38:46 +08:00
|
|
|
int
|
|
|
|
xfs_free_agfl_block(
|
|
|
|
struct xfs_trans *tp,
|
|
|
|
xfs_agnumber_t agno,
|
|
|
|
xfs_agblock_t agbno,
|
|
|
|
struct xfs_buf *agbp,
|
|
|
|
struct xfs_owner_info *oinfo)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
struct xfs_buf *bp;
|
|
|
|
|
|
|
|
error = xfs_free_ag_extent(tp, agbp, agno, agbno, 1, oinfo,
|
|
|
|
XFS_AG_RESV_AGFL);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
2020-01-24 09:01:19 +08:00
|
|
|
error = xfs_trans_get_buf(tp, tp->t_mountp->m_ddev_targp,
|
|
|
|
XFS_AGB_TO_DADDR(tp->t_mountp, agno, agbno),
|
|
|
|
tp->t_mountp->m_bsize, 0, &bp);
|
|
|
|
if (error)
|
|
|
|
return error;
|
2018-05-08 08:38:46 +08:00
|
|
|
xfs_trans_binval(tp, bp);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2018-03-16 01:51:58 +08:00
|
|
|
/*
|
2023-06-05 12:48:15 +08:00
|
|
|
* Check the agfl fields of the agf for inconsistency or corruption.
|
|
|
|
*
|
|
|
|
* The original purpose was to detect an agfl header padding mismatch between
|
|
|
|
* current and early v5 kernels. This problem manifests as a 1-slot size
|
|
|
|
* difference between the on-disk flcount and the active [first, last] range of
|
|
|
|
* a wrapped agfl.
|
|
|
|
*
|
|
|
|
* However, we need to use these same checks to catch agfl count corruptions
|
|
|
|
* unrelated to padding. This could occur on any v4 or v5 filesystem, so either
|
|
|
|
* way, we need to reset the agfl and warn the user.
|
2018-03-16 01:51:58 +08:00
|
|
|
*
|
|
|
|
* Return true if a reset is required before the agfl can be used, false
|
|
|
|
* otherwise.
|
|
|
|
*/
|
|
|
|
static bool
|
|
|
|
xfs_agfl_needs_reset(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
struct xfs_agf *agf)
|
|
|
|
{
|
|
|
|
uint32_t f = be32_to_cpu(agf->agf_flfirst);
|
|
|
|
uint32_t l = be32_to_cpu(agf->agf_fllast);
|
|
|
|
uint32_t c = be32_to_cpu(agf->agf_flcount);
|
|
|
|
int agfl_size = xfs_agfl_size(mp);
|
|
|
|
int active;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The agf read verifier catches severe corruption of these fields.
|
|
|
|
* Repeat some sanity checks to cover a packed -> unpacked mismatch if
|
|
|
|
* the verifier allows it.
|
|
|
|
*/
|
|
|
|
if (f >= agfl_size || l >= agfl_size)
|
|
|
|
return true;
|
|
|
|
if (c > agfl_size)
|
|
|
|
return true;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check consistency between the on-disk count and the active range. An
|
|
|
|
* agfl padding mismatch manifests as an inconsistent flcount.
|
|
|
|
*/
|
|
|
|
if (c && l >= f)
|
|
|
|
active = l - f + 1;
|
|
|
|
else if (c)
|
|
|
|
active = agfl_size - f + l + 1;
|
|
|
|
else
|
|
|
|
active = 0;
|
|
|
|
|
|
|
|
return active != c;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Reset the agfl to an empty state. Ignore/drop any existing blocks since the
|
|
|
|
* agfl content cannot be trusted. Warn the user that a repair is required to
|
|
|
|
* recover leaked blocks.
|
|
|
|
*
|
|
|
|
* The purpose of this mechanism is to handle filesystems affected by the agfl
|
|
|
|
* header padding mismatch problem. A reset keeps the filesystem online with a
|
|
|
|
* relatively minor free space accounting inconsistency rather than suffer the
|
|
|
|
* inevitable crash from use of an invalid agfl block.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
xfs_agfl_reset(
|
|
|
|
struct xfs_trans *tp,
|
|
|
|
struct xfs_buf *agbp,
|
|
|
|
struct xfs_perag *pag)
|
|
|
|
{
|
|
|
|
struct xfs_mount *mp = tp->t_mountp;
|
2020-03-10 23:57:29 +08:00
|
|
|
struct xfs_agf *agf = agbp->b_addr;
|
2018-03-16 01:51:58 +08:00
|
|
|
|
2023-02-13 06:14:52 +08:00
|
|
|
ASSERT(xfs_perag_agfl_needs_reset(pag));
|
2018-03-16 01:51:58 +08:00
|
|
|
trace_xfs_agfl_reset(mp, agf, 0, _RET_IP_);
|
|
|
|
|
|
|
|
xfs_warn(mp,
|
|
|
|
"WARNING: Reset corrupted AGFL on AG %u. %d blocks leaked. "
|
|
|
|
"Please unmount and run xfs_repair.",
|
|
|
|
pag->pag_agno, pag->pagf_flcount);
|
|
|
|
|
|
|
|
agf->agf_flfirst = 0;
|
|
|
|
agf->agf_fllast = cpu_to_be32(xfs_agfl_size(mp) - 1);
|
|
|
|
agf->agf_flcount = 0;
|
|
|
|
xfs_alloc_log_agf(tp, agbp, XFS_AGF_FLFIRST | XFS_AGF_FLLAST |
|
|
|
|
XFS_AGF_FLCOUNT);
|
|
|
|
|
|
|
|
pag->pagf_flcount = 0;
|
2023-02-13 06:14:52 +08:00
|
|
|
clear_bit(XFS_AGSTATE_AGFL_NEEDS_RESET, &pag->pag_opstate);
|
2018-03-16 01:51:58 +08:00
|
|
|
}
|
|
|
|
|
2018-05-08 08:38:47 +08:00
|
|
|
/*
|
|
|
|
* Defer an AGFL block free. This is effectively equivalent to
|
2021-10-13 05:17:01 +08:00
|
|
|
* xfs_free_extent_later() with some special handling particular to AGFL blocks.
|
2018-05-08 08:38:47 +08:00
|
|
|
*
|
|
|
|
* Deferring AGFL frees helps prevent log reservation overruns due to too many
|
|
|
|
* allocation operations in a transaction. AGFL frees are prone to this problem
|
|
|
|
* because for one they are always freed one at a time. Further, an immediate
|
|
|
|
* AGFL block free can cause a btree join and require another block free before
|
|
|
|
* the real allocation can proceed. Deferring the free disconnects freeing up
|
|
|
|
* the AGFL slot from freeing the block.
|
|
|
|
*/
|
2023-06-05 12:48:15 +08:00
|
|
|
static int
|
2018-05-08 08:38:47 +08:00
|
|
|
xfs_defer_agfl_block(
|
2018-08-01 22:20:34 +08:00
|
|
|
struct xfs_trans *tp,
|
2018-05-08 08:38:47 +08:00
|
|
|
xfs_agnumber_t agno,
|
2023-06-29 02:04:34 +08:00
|
|
|
xfs_agblock_t agbno,
|
2018-05-08 08:38:47 +08:00
|
|
|
struct xfs_owner_info *oinfo)
|
|
|
|
{
|
2018-08-01 22:20:34 +08:00
|
|
|
struct xfs_mount *mp = tp->t_mountp;
|
2023-02-02 02:16:02 +08:00
|
|
|
struct xfs_extent_free_item *xefi;
|
2023-06-29 02:04:34 +08:00
|
|
|
xfs_fsblock_t fsbno = XFS_AGB_TO_FSB(mp, agno, agbno);
|
2018-05-08 08:38:47 +08:00
|
|
|
|
2021-10-13 05:17:01 +08:00
|
|
|
ASSERT(xfs_extfree_item_cache != NULL);
|
2018-05-08 08:38:47 +08:00
|
|
|
ASSERT(oinfo != NULL);
|
|
|
|
|
2023-06-29 02:04:34 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, !xfs_verify_fsbno(mp, fsbno)))
|
|
|
|
return -EFSCORRUPTED;
|
|
|
|
|
2023-02-02 02:16:02 +08:00
|
|
|
xefi = kmem_cache_zalloc(xfs_extfree_item_cache,
|
2020-07-23 00:23:04 +08:00
|
|
|
GFP_KERNEL | __GFP_NOFAIL);
|
2023-06-29 02:04:34 +08:00
|
|
|
xefi->xefi_startblock = fsbno;
|
2023-02-02 02:16:02 +08:00
|
|
|
xefi->xefi_blockcount = 1;
|
|
|
|
xefi->xefi_owner = oinfo->oi_owner;
|
2023-06-29 02:04:32 +08:00
|
|
|
xefi->xefi_agresv = XFS_AG_RESV_AGFL;
|
2018-05-08 08:38:47 +08:00
|
|
|
|
|
|
|
trace_xfs_agfl_free_defer(mp, agno, 0, agbno, 1);
|
|
|
|
|
2023-04-12 09:59:54 +08:00
|
|
|
xfs_extent_free_get_group(mp, xefi);
|
2023-02-02 02:16:02 +08:00
|
|
|
xfs_defer_add(tp, XFS_DEFER_OPS_TYPE_AGFL_FREE, &xefi->xefi_list);
|
2023-06-05 12:48:15 +08:00
|
|
|
return 0;
|
2018-05-08 08:38:47 +08:00
|
|
|
}
|
|
|
|
|
2021-10-13 05:17:01 +08:00
|
|
|
/*
|
|
|
|
* Add the extent to the list of extents to be free at transaction end.
|
|
|
|
* The list is maintained sorted (by block number).
|
|
|
|
*/
|
2023-06-05 12:48:15 +08:00
|
|
|
int
|
2021-10-13 05:17:01 +08:00
|
|
|
__xfs_free_extent_later(
|
|
|
|
struct xfs_trans *tp,
|
|
|
|
xfs_fsblock_t bno,
|
|
|
|
xfs_filblks_t len,
|
|
|
|
const struct xfs_owner_info *oinfo,
|
2023-06-29 02:04:32 +08:00
|
|
|
enum xfs_ag_resv_type type,
|
2021-10-13 05:17:01 +08:00
|
|
|
bool skip_discard)
|
|
|
|
{
|
2023-02-02 02:16:02 +08:00
|
|
|
struct xfs_extent_free_item *xefi;
|
2021-10-13 05:17:01 +08:00
|
|
|
struct xfs_mount *mp = tp->t_mountp;
|
2023-04-12 09:59:54 +08:00
|
|
|
#ifdef DEBUG
|
2021-10-13 05:17:01 +08:00
|
|
|
xfs_agnumber_t agno;
|
|
|
|
xfs_agblock_t agbno;
|
|
|
|
|
|
|
|
ASSERT(bno != NULLFSBLOCK);
|
|
|
|
ASSERT(len > 0);
|
2021-08-09 14:35:22 +08:00
|
|
|
ASSERT(len <= XFS_MAX_BMBT_EXTLEN);
|
2021-10-13 05:17:01 +08:00
|
|
|
ASSERT(!isnullstartblock(bno));
|
|
|
|
agno = XFS_FSB_TO_AGNO(mp, bno);
|
|
|
|
agbno = XFS_FSB_TO_AGBNO(mp, bno);
|
|
|
|
ASSERT(agno < mp->m_sb.sb_agcount);
|
|
|
|
ASSERT(agbno < mp->m_sb.sb_agblocks);
|
|
|
|
ASSERT(len < mp->m_sb.sb_agblocks);
|
|
|
|
ASSERT(agbno + len <= mp->m_sb.sb_agblocks);
|
|
|
|
#endif
|
|
|
|
ASSERT(xfs_extfree_item_cache != NULL);
|
2023-06-29 02:04:32 +08:00
|
|
|
ASSERT(type != XFS_AG_RESV_AGFL);
|
2021-10-13 05:17:01 +08:00
|
|
|
|
2023-06-05 12:48:15 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, !xfs_verify_fsbext(mp, bno, len)))
|
|
|
|
return -EFSCORRUPTED;
|
|
|
|
|
2023-02-02 02:16:02 +08:00
|
|
|
xefi = kmem_cache_zalloc(xfs_extfree_item_cache,
|
2021-10-13 05:17:01 +08:00
|
|
|
GFP_KERNEL | __GFP_NOFAIL);
|
2023-02-02 02:16:02 +08:00
|
|
|
xefi->xefi_startblock = bno;
|
|
|
|
xefi->xefi_blockcount = (xfs_extlen_t)len;
|
2023-06-29 02:04:32 +08:00
|
|
|
xefi->xefi_agresv = type;
|
2021-10-13 06:55:54 +08:00
|
|
|
if (skip_discard)
|
2023-02-02 02:16:02 +08:00
|
|
|
xefi->xefi_flags |= XFS_EFI_SKIP_DISCARD;
|
2021-10-13 06:55:54 +08:00
|
|
|
if (oinfo) {
|
|
|
|
ASSERT(oinfo->oi_offset == 0);
|
|
|
|
|
|
|
|
if (oinfo->oi_flags & XFS_OWNER_INFO_ATTR_FORK)
|
2023-02-02 02:16:02 +08:00
|
|
|
xefi->xefi_flags |= XFS_EFI_ATTR_FORK;
|
2021-10-13 06:55:54 +08:00
|
|
|
if (oinfo->oi_flags & XFS_OWNER_INFO_BMBT_BLOCK)
|
2023-02-02 02:16:02 +08:00
|
|
|
xefi->xefi_flags |= XFS_EFI_BMBT_BLOCK;
|
|
|
|
xefi->xefi_owner = oinfo->oi_owner;
|
2021-10-13 06:55:54 +08:00
|
|
|
} else {
|
2023-02-02 02:16:02 +08:00
|
|
|
xefi->xefi_owner = XFS_RMAP_OWN_NULL;
|
2021-10-13 06:55:54 +08:00
|
|
|
}
|
2023-04-12 09:59:54 +08:00
|
|
|
trace_xfs_bmap_free_defer(mp,
|
2021-10-13 05:17:01 +08:00
|
|
|
XFS_FSB_TO_AGNO(tp->t_mountp, bno), 0,
|
|
|
|
XFS_FSB_TO_AGBNO(tp->t_mountp, bno), len);
|
2023-04-12 09:59:54 +08:00
|
|
|
|
|
|
|
xfs_extent_free_get_group(mp, xefi);
|
2023-02-02 02:16:02 +08:00
|
|
|
xfs_defer_add(tp, XFS_DEFER_OPS_TYPE_FREE, &xefi->xefi_list);
|
2023-06-05 12:48:15 +08:00
|
|
|
return 0;
|
2021-10-13 05:17:01 +08:00
|
|
|
}
|
|
|
|
|
2021-01-23 08:48:17 +08:00
|
|
|
#ifdef DEBUG
|
|
|
|
/*
|
|
|
|
* Check if an AGF has a free extent record whose length is equal to
|
|
|
|
* args->minlen.
|
|
|
|
*/
|
|
|
|
STATIC int
|
|
|
|
xfs_exact_minlen_extent_available(
|
|
|
|
struct xfs_alloc_arg *args,
|
|
|
|
struct xfs_buf *agbp,
|
|
|
|
int *stat)
|
|
|
|
{
|
|
|
|
struct xfs_btree_cur *cnt_cur;
|
|
|
|
xfs_agblock_t fbno;
|
|
|
|
xfs_extlen_t flen;
|
|
|
|
int error = 0;
|
|
|
|
|
|
|
|
cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, agbp,
|
2021-06-02 08:48:24 +08:00
|
|
|
args->pag, XFS_BTNUM_CNT);
|
2021-01-23 08:48:17 +08:00
|
|
|
error = xfs_alloc_lookup_ge(cnt_cur, 0, args->minlen, stat);
|
|
|
|
if (error)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (*stat == 0) {
|
|
|
|
error = -EFSCORRUPTED;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
error = xfs_alloc_get_rec(cnt_cur, &fbno, &flen, stat);
|
|
|
|
if (error)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (*stat == 1 && flen != args->minlen)
|
|
|
|
*stat = 0;
|
|
|
|
|
|
|
|
out:
|
|
|
|
xfs_btree_del_cursor(cnt_cur, error);
|
|
|
|
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Decide whether to use this allocation group for this allocation.
|
|
|
|
* If so, fix up the btree freelist's size.
|
|
|
|
*/
|
2016-01-04 13:10:42 +08:00
|
|
|
int /* error */
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_alloc_fix_freelist(
|
2015-06-22 08:13:19 +08:00
|
|
|
struct xfs_alloc_arg *args, /* allocation argument structure */
|
2023-06-29 02:04:32 +08:00
|
|
|
uint32_t alloc_flags)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2015-06-22 08:13:19 +08:00
|
|
|
struct xfs_mount *mp = args->mp;
|
|
|
|
struct xfs_perag *pag = args->pag;
|
|
|
|
struct xfs_trans *tp = args->tp;
|
|
|
|
struct xfs_buf *agbp = NULL;
|
|
|
|
struct xfs_buf *agflbp = NULL;
|
|
|
|
struct xfs_alloc_arg targs; /* local allocation arguments */
|
|
|
|
xfs_agblock_t bno; /* freelist block */
|
|
|
|
xfs_extlen_t need; /* total blocks needed in freelist */
|
2015-08-25 08:05:13 +08:00
|
|
|
int error = 0;
|
2015-06-22 08:13:19 +08:00
|
|
|
|
2019-04-23 22:54:06 +08:00
|
|
|
/* deferred ops (AGFL block frees) require permanent transactions */
|
|
|
|
ASSERT(tp->t_flags & XFS_TRANS_PERM_LOG_RES);
|
|
|
|
|
2023-02-13 06:14:52 +08:00
|
|
|
if (!xfs_perag_initialised_agf(pag)) {
|
2023-06-29 02:04:32 +08:00
|
|
|
error = xfs_alloc_read_agf(pag, tp, alloc_flags, &agbp);
|
2020-01-24 09:01:19 +08:00
|
|
|
if (error) {
|
|
|
|
/* Couldn't lock the AGF so skip this AG. */
|
|
|
|
if (error == -EAGAIN)
|
|
|
|
error = 0;
|
2015-06-22 08:13:19 +08:00
|
|
|
goto out_no_agbp;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2015-06-22 08:13:19 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2006-08-10 12:40:41 +08:00
|
|
|
/*
|
2015-06-22 08:13:19 +08:00
|
|
|
* If this is a metadata preferred pag and we are user data then try
|
|
|
|
* somewhere else if we are not being asked to try harder at this
|
|
|
|
* point
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2023-02-13 06:14:52 +08:00
|
|
|
if (xfs_perag_prefers_metadata(pag) &&
|
|
|
|
(args->datatype & XFS_ALLOC_USERDATA) &&
|
2023-06-29 02:04:32 +08:00
|
|
|
(alloc_flags & XFS_ALLOC_FLAG_TRYLOCK)) {
|
|
|
|
ASSERT(!(alloc_flags & XFS_ALLOC_FLAG_FREEING));
|
2015-06-22 08:13:19 +08:00
|
|
|
goto out_agbp_relse;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2015-06-22 08:13:30 +08:00
|
|
|
need = xfs_alloc_min_freelist(mp, pag);
|
2023-06-29 02:04:32 +08:00
|
|
|
if (!xfs_alloc_space_available(args, need, alloc_flags |
|
2017-01-10 05:44:30 +08:00
|
|
|
XFS_ALLOC_FLAG_CHECK))
|
2015-06-22 08:13:19 +08:00
|
|
|
goto out_agbp_relse;
|
2006-08-10 12:40:41 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Get the a.g. freespace buffer.
|
|
|
|
* Can fail if we're not blocking on locks, and it's held.
|
|
|
|
*/
|
2015-06-22 08:13:19 +08:00
|
|
|
if (!agbp) {
|
2023-06-29 02:04:32 +08:00
|
|
|
error = xfs_alloc_read_agf(pag, tp, alloc_flags, &agbp);
|
2020-01-24 09:01:19 +08:00
|
|
|
if (error) {
|
|
|
|
/* Couldn't lock the AGF so skip this AG. */
|
|
|
|
if (error == -EAGAIN)
|
|
|
|
error = 0;
|
2015-06-22 08:13:19 +08:00
|
|
|
goto out_no_agbp;
|
2006-08-10 12:40:41 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2015-06-22 08:04:31 +08:00
|
|
|
|
2018-03-16 01:51:58 +08:00
|
|
|
/* reset a padding mismatched agfl before final free space check */
|
2023-02-13 06:14:52 +08:00
|
|
|
if (xfs_perag_agfl_needs_reset(pag))
|
2018-03-16 01:51:58 +08:00
|
|
|
xfs_agfl_reset(tp, agbp, pag);
|
|
|
|
|
2015-06-22 08:04:31 +08:00
|
|
|
/* If there isn't enough total space or single-extent, reject it. */
|
2015-06-22 08:13:30 +08:00
|
|
|
need = xfs_alloc_min_freelist(mp, pag);
|
2023-06-29 02:04:32 +08:00
|
|
|
if (!xfs_alloc_space_available(args, need, alloc_flags))
|
2015-06-22 08:13:19 +08:00
|
|
|
goto out_agbp_relse;
|
2015-06-22 08:04:42 +08:00
|
|
|
|
2021-01-23 08:48:17 +08:00
|
|
|
#ifdef DEBUG
|
|
|
|
if (args->alloc_minlen_only) {
|
|
|
|
int stat;
|
|
|
|
|
|
|
|
error = xfs_exact_minlen_extent_available(args, agbp, &stat);
|
|
|
|
if (error || !stat)
|
|
|
|
goto out_agbp_relse;
|
|
|
|
}
|
|
|
|
#endif
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Make the freelist shorter if it's too long.
|
2015-06-22 08:04:31 +08:00
|
|
|
*
|
2015-06-22 08:13:19 +08:00
|
|
|
* Note that from this point onwards, we will always release the agf and
|
|
|
|
* agfl buffers on error. This handles the case where we error out and
|
|
|
|
* the buffers are clean or may not have been joined to the transaction
|
|
|
|
* and hence need to be released manually. If they have been joined to
|
|
|
|
* the transaction, then xfs_trans_brelse() will handle them
|
|
|
|
* appropriately based on the recursion count and dirty state of the
|
|
|
|
* buffer.
|
|
|
|
*
|
2015-06-22 08:04:31 +08:00
|
|
|
* XXX (dgc): When we have lots of free space, does this buy us
|
|
|
|
* anything other than extra overhead when we need to put more blocks
|
|
|
|
* back on the free list? Maybe we should only do this when space is
|
|
|
|
* getting low or the AGFL is more than half full?
|
2016-08-03 10:19:53 +08:00
|
|
|
*
|
|
|
|
* The NOSHRINK flag prevents the AGFL from being shrunk if it's too
|
|
|
|
* big; the NORMAP flag prevents AGFL expand/shrink operations from
|
|
|
|
* updating the rmapbt. Both flags are used in xfs_repair while we're
|
|
|
|
* rebuilding the rmapbt, and neither are used by the kernel. They're
|
|
|
|
* both required to ensure that rmaps are correctly recorded for the
|
|
|
|
* regenerated AGFL, bnobt, and cntbt. See repair/phase5.c and
|
|
|
|
* repair/rmap.c in xfsprogs for details.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2016-08-03 10:19:53 +08:00
|
|
|
memset(&targs, 0, sizeof(targs));
|
2018-12-13 00:46:23 +08:00
|
|
|
/* struct copy below */
|
2023-06-29 02:04:32 +08:00
|
|
|
if (alloc_flags & XFS_ALLOC_FLAG_NORMAP)
|
2018-12-13 00:46:23 +08:00
|
|
|
targs.oinfo = XFS_RMAP_OINFO_SKIP_UPDATE;
|
2016-08-03 10:19:53 +08:00
|
|
|
else
|
2018-12-13 00:46:23 +08:00
|
|
|
targs.oinfo = XFS_RMAP_OINFO_AG;
|
2023-06-29 02:04:32 +08:00
|
|
|
while (!(alloc_flags & XFS_ALLOC_FLAG_NOSHRINK) &&
|
|
|
|
pag->pagf_flcount > need) {
|
2022-07-07 17:08:01 +08:00
|
|
|
error = xfs_alloc_get_freelist(pag, tp, agbp, &bno, 0);
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 13:26:31 +08:00
|
|
|
if (error)
|
2015-06-22 08:13:19 +08:00
|
|
|
goto out_agbp_relse;
|
2018-05-08 08:38:46 +08:00
|
|
|
|
2018-08-01 22:20:35 +08:00
|
|
|
/* defer agfl frees */
|
2023-06-05 12:48:15 +08:00
|
|
|
error = xfs_defer_agfl_block(tp, args->agno, bno, &targs.oinfo);
|
|
|
|
if (error)
|
|
|
|
goto out_agbp_relse;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2015-06-22 08:04:31 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
targs.tp = tp;
|
|
|
|
targs.mp = mp;
|
|
|
|
targs.agbp = agbp;
|
|
|
|
targs.agno = args->agno;
|
xfs: set up per-AG free space reservations
One unfortunate quirk of the reference count and reverse mapping
btrees -- they can expand in size when blocks are written to *other*
allocation groups if, say, one large extent becomes a lot of tiny
extents. Since we don't want to start throwing errors in the middle
of CoWing, we need to reserve some blocks to handle future expansion.
The transaction block reservation counters aren't sufficient here
because we have to have a reserve of blocks in every AG, not just
somewhere in the filesystem.
Therefore, create two per-AG block reservation pools. One feeds the
AGFL so that rmapbt expansion always succeeds, and the other feeds all
other metadata so that refcountbt expansion never fails.
Use the count of how many reserved blocks we need to have on hand to
create a virtual reservation in the AG. Through selective clamping of
the maximum length of allocation requests and of the length of the
longest free extent, we can make it look like there's less free space
in the AG unless the reservation owner is asking for blocks.
In other words, play some accounting tricks in-core to make sure that
we always have blocks available. On the plus side, there's nothing to
clean up if we crash, which is contrast to the strategy that the rough
draft used (actually removing extents from the freespace btrees).
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-09-19 08:30:52 +08:00
|
|
|
targs.alignment = targs.minlen = targs.prod = 1;
|
2005-04-17 06:20:36 +08:00
|
|
|
targs.pag = pag;
|
2022-07-07 17:08:15 +08:00
|
|
|
error = xfs_alloc_read_agfl(pag, tp, &agflbp);
|
2015-06-22 08:04:31 +08:00
|
|
|
if (error)
|
2015-06-22 08:13:19 +08:00
|
|
|
goto out_agbp_relse;
|
2015-06-22 08:04:31 +08:00
|
|
|
|
|
|
|
/* Make the freelist longer if it's too short. */
|
|
|
|
while (pag->pagf_flcount < need) {
|
2005-04-17 06:20:36 +08:00
|
|
|
targs.agbno = 0;
|
2015-06-22 08:04:31 +08:00
|
|
|
targs.maxlen = need - pag->pagf_flcount;
|
2018-03-10 06:02:32 +08:00
|
|
|
targs.resv = XFS_AG_RESV_AGFL;
|
2015-06-22 08:04:31 +08:00
|
|
|
|
|
|
|
/* Allocate as many blocks as possible at once. */
|
2023-06-29 02:04:32 +08:00
|
|
|
error = xfs_alloc_ag_vextent_size(&targs, alloc_flags);
|
2015-06-22 08:13:19 +08:00
|
|
|
if (error)
|
|
|
|
goto out_agflbp_relse;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Stop if we run out. Won't happen if callers are obeying
|
|
|
|
* the restrictions correctly. Can happen for free calls
|
|
|
|
* on a completely full ag.
|
|
|
|
*/
|
2006-06-09 12:55:18 +08:00
|
|
|
if (targs.agbno == NULLAGBLOCK) {
|
2023-06-29 02:04:32 +08:00
|
|
|
if (alloc_flags & XFS_ALLOC_FLAG_FREEING)
|
2006-08-10 12:40:41 +08:00
|
|
|
break;
|
2015-06-22 08:13:19 +08:00
|
|
|
goto out_agflbp_relse;
|
2006-06-09 12:55:18 +08:00
|
|
|
}
|
2023-02-13 06:14:53 +08:00
|
|
|
|
|
|
|
if (!xfs_rmap_should_skip_owner_update(&targs.oinfo)) {
|
|
|
|
error = xfs_rmap_alloc(tp, agbp, pag,
|
|
|
|
targs.agbno, targs.len, &targs.oinfo);
|
|
|
|
if (error)
|
|
|
|
goto out_agflbp_relse;
|
|
|
|
}
|
|
|
|
error = xfs_alloc_update_counters(tp, agbp,
|
|
|
|
-((long)(targs.len)));
|
|
|
|
if (error)
|
|
|
|
goto out_agflbp_relse;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Put each allocated block on the list.
|
|
|
|
*/
|
|
|
|
for (bno = targs.agbno; bno < targs.agbno + targs.len; bno++) {
|
2022-07-07 17:08:08 +08:00
|
|
|
error = xfs_alloc_put_freelist(pag, tp, agbp,
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 13:26:31 +08:00
|
|
|
agflbp, bno, 0);
|
|
|
|
if (error)
|
2015-06-22 08:13:19 +08:00
|
|
|
goto out_agflbp_relse;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
}
|
2006-05-08 17:51:58 +08:00
|
|
|
xfs_trans_brelse(tp, agflbp);
|
2005-04-17 06:20:36 +08:00
|
|
|
args->agbp = agbp;
|
|
|
|
return 0;
|
2015-06-22 08:13:19 +08:00
|
|
|
|
|
|
|
out_agflbp_relse:
|
|
|
|
xfs_trans_brelse(tp, agflbp);
|
|
|
|
out_agbp_relse:
|
|
|
|
if (agbp)
|
|
|
|
xfs_trans_brelse(tp, agbp);
|
|
|
|
out_no_agbp:
|
|
|
|
args->agbp = NULL;
|
|
|
|
return error;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Get a block from the freelist.
|
|
|
|
* Returns with the buffer for the block gotten.
|
|
|
|
*/
|
2021-06-02 08:48:51 +08:00
|
|
|
int
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_alloc_get_freelist(
|
2022-07-07 17:08:01 +08:00
|
|
|
struct xfs_perag *pag,
|
2021-06-02 08:48:51 +08:00
|
|
|
struct xfs_trans *tp,
|
|
|
|
struct xfs_buf *agbp,
|
|
|
|
xfs_agblock_t *bnop,
|
|
|
|
int btreeblk)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2021-06-02 08:48:51 +08:00
|
|
|
struct xfs_agf *agf = agbp->b_addr;
|
|
|
|
struct xfs_buf *agflbp;
|
|
|
|
xfs_agblock_t bno;
|
|
|
|
__be32 *agfl_bno;
|
|
|
|
int error;
|
2022-04-21 08:46:16 +08:00
|
|
|
uint32_t logflags;
|
2021-06-02 08:48:51 +08:00
|
|
|
struct xfs_mount *mp = tp->t_mountp;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Freelist is empty, give up.
|
|
|
|
*/
|
|
|
|
if (!agf->agf_flcount) {
|
|
|
|
*bnop = NULLAGBLOCK;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Read the array of free blocks.
|
|
|
|
*/
|
2022-07-07 17:08:15 +08:00
|
|
|
error = xfs_alloc_read_agfl(pag, tp, &agflbp);
|
2013-04-03 13:11:14 +08:00
|
|
|
if (error)
|
2005-04-17 06:20:36 +08:00
|
|
|
return error;
|
2013-04-03 13:11:14 +08:00
|
|
|
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Get the block number and update the data structures.
|
|
|
|
*/
|
2020-03-10 23:57:28 +08:00
|
|
|
agfl_bno = xfs_buf_to_agfl_bno(agflbp);
|
2013-04-03 13:11:14 +08:00
|
|
|
bno = be32_to_cpu(agfl_bno[be32_to_cpu(agf->agf_flfirst)]);
|
2023-06-05 12:48:15 +08:00
|
|
|
if (XFS_IS_CORRUPT(tp->t_mountp, !xfs_verify_agbno(pag, bno)))
|
|
|
|
return -EFSCORRUPTED;
|
|
|
|
|
2008-02-14 07:03:29 +08:00
|
|
|
be32_add_cpu(&agf->agf_flfirst, 1);
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_trans_brelse(tp, agflbp);
|
2018-03-07 09:08:32 +08:00
|
|
|
if (be32_to_cpu(agf->agf_flfirst) == xfs_agfl_size(mp))
|
2005-04-17 06:20:36 +08:00
|
|
|
agf->agf_flfirst = 0;
|
2010-01-11 19:47:41 +08:00
|
|
|
|
2023-02-13 06:14:52 +08:00
|
|
|
ASSERT(!xfs_perag_agfl_needs_reset(pag));
|
2008-02-14 07:03:29 +08:00
|
|
|
be32_add_cpu(&agf->agf_flcount, -1);
|
2005-04-17 06:20:36 +08:00
|
|
|
pag->pagf_flcount--;
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 13:26:31 +08:00
|
|
|
|
|
|
|
logflags = XFS_AGF_FLFIRST | XFS_AGF_FLCOUNT;
|
|
|
|
if (btreeblk) {
|
2008-02-14 07:03:29 +08:00
|
|
|
be32_add_cpu(&agf->agf_btreeblks, 1);
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 13:26:31 +08:00
|
|
|
pag->pagf_btreeblks++;
|
|
|
|
logflags |= XFS_AGF_BTREEBLKS;
|
|
|
|
}
|
|
|
|
|
|
|
|
xfs_alloc_log_agf(tp, agbp, logflags);
|
2005-04-17 06:20:36 +08:00
|
|
|
*bnop = bno;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Log the given fields from the agf structure.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
xfs_alloc_log_agf(
|
2022-04-21 08:46:16 +08:00
|
|
|
struct xfs_trans *tp,
|
|
|
|
struct xfs_buf *bp,
|
|
|
|
uint32_t fields)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
int first; /* first byte offset */
|
|
|
|
int last; /* last byte offset */
|
|
|
|
static const short offsets[] = {
|
|
|
|
offsetof(xfs_agf_t, agf_magicnum),
|
|
|
|
offsetof(xfs_agf_t, agf_versionnum),
|
|
|
|
offsetof(xfs_agf_t, agf_seqno),
|
|
|
|
offsetof(xfs_agf_t, agf_length),
|
|
|
|
offsetof(xfs_agf_t, agf_roots[0]),
|
|
|
|
offsetof(xfs_agf_t, agf_levels[0]),
|
|
|
|
offsetof(xfs_agf_t, agf_flfirst),
|
|
|
|
offsetof(xfs_agf_t, agf_fllast),
|
|
|
|
offsetof(xfs_agf_t, agf_flcount),
|
|
|
|
offsetof(xfs_agf_t, agf_freeblks),
|
|
|
|
offsetof(xfs_agf_t, agf_longest),
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 13:26:31 +08:00
|
|
|
offsetof(xfs_agf_t, agf_btreeblks),
|
2013-04-03 13:11:13 +08:00
|
|
|
offsetof(xfs_agf_t, agf_uuid),
|
2016-08-17 06:31:49 +08:00
|
|
|
offsetof(xfs_agf_t, agf_rmap_blocks),
|
2016-10-04 00:11:19 +08:00
|
|
|
offsetof(xfs_agf_t, agf_refcount_blocks),
|
|
|
|
offsetof(xfs_agf_t, agf_refcount_root),
|
|
|
|
offsetof(xfs_agf_t, agf_refcount_level),
|
2016-08-26 13:59:31 +08:00
|
|
|
/* needed so that we don't log the whole rest of the structure: */
|
|
|
|
offsetof(xfs_agf_t, agf_spare64),
|
2005-04-17 06:20:36 +08:00
|
|
|
sizeof(xfs_agf_t)
|
|
|
|
};
|
|
|
|
|
2020-03-10 23:57:29 +08:00
|
|
|
trace_xfs_agf(tp->t_mountp, bp->b_addr, fields, _RET_IP_);
|
2009-12-15 07:14:59 +08:00
|
|
|
|
2013-04-03 13:11:30 +08:00
|
|
|
xfs_trans_buf_set_type(tp, bp, XFS_BLFT_AGF_BUF);
|
2013-04-03 13:11:13 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_btree_offsets(fields, offsets, XFS_AGF_NUM_BITS, &first, &last);
|
|
|
|
xfs_trans_log_buf(tp, bp, (uint)first, (uint)last);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Put the block on the freelist for the allocation group.
|
|
|
|
*/
|
2021-06-02 08:48:51 +08:00
|
|
|
int
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_alloc_put_freelist(
|
2022-07-07 17:08:08 +08:00
|
|
|
struct xfs_perag *pag,
|
2021-06-02 08:48:51 +08:00
|
|
|
struct xfs_trans *tp,
|
|
|
|
struct xfs_buf *agbp,
|
|
|
|
struct xfs_buf *agflbp,
|
|
|
|
xfs_agblock_t bno,
|
|
|
|
int btreeblk)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2020-03-10 23:57:29 +08:00
|
|
|
struct xfs_mount *mp = tp->t_mountp;
|
|
|
|
struct xfs_agf *agf = agbp->b_addr;
|
2021-06-02 08:48:51 +08:00
|
|
|
__be32 *blockp;
|
2005-04-17 06:20:36 +08:00
|
|
|
int error;
|
2022-04-21 08:46:16 +08:00
|
|
|
uint32_t logflags;
|
2013-04-03 13:11:14 +08:00
|
|
|
__be32 *agfl_bno;
|
|
|
|
int startoff;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2022-07-07 17:08:15 +08:00
|
|
|
if (!agflbp) {
|
|
|
|
error = xfs_alloc_read_agfl(pag, tp, &agflbp);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2008-02-14 07:03:29 +08:00
|
|
|
be32_add_cpu(&agf->agf_fllast, 1);
|
2018-03-07 09:08:32 +08:00
|
|
|
if (be32_to_cpu(agf->agf_fllast) == xfs_agfl_size(mp))
|
2005-04-17 06:20:36 +08:00
|
|
|
agf->agf_fllast = 0;
|
2010-01-11 19:47:41 +08:00
|
|
|
|
2023-02-13 06:14:52 +08:00
|
|
|
ASSERT(!xfs_perag_agfl_needs_reset(pag));
|
2008-02-14 07:03:29 +08:00
|
|
|
be32_add_cpu(&agf->agf_flcount, 1);
|
2005-04-17 06:20:36 +08:00
|
|
|
pag->pagf_flcount++;
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 13:26:31 +08:00
|
|
|
|
|
|
|
logflags = XFS_AGF_FLLAST | XFS_AGF_FLCOUNT;
|
|
|
|
if (btreeblk) {
|
2008-02-14 07:03:29 +08:00
|
|
|
be32_add_cpu(&agf->agf_btreeblks, -1);
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 13:26:31 +08:00
|
|
|
pag->pagf_btreeblks--;
|
|
|
|
logflags |= XFS_AGF_BTREEBLKS;
|
|
|
|
}
|
|
|
|
|
|
|
|
xfs_alloc_log_agf(tp, agbp, logflags);
|
|
|
|
|
2018-03-07 09:08:32 +08:00
|
|
|
ASSERT(be32_to_cpu(agf->agf_flcount) <= xfs_agfl_size(mp));
|
2013-04-03 13:11:14 +08:00
|
|
|
|
2020-03-10 23:57:28 +08:00
|
|
|
agfl_bno = xfs_buf_to_agfl_bno(agflbp);
|
2013-04-03 13:11:14 +08:00
|
|
|
blockp = &agfl_bno[be32_to_cpu(agf->agf_fllast)];
|
2006-09-28 08:56:51 +08:00
|
|
|
*blockp = cpu_to_be32(bno);
|
2013-04-03 13:11:14 +08:00
|
|
|
startoff = (char *)blockp - (char *)agflbp->b_addr;
|
|
|
|
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 13:26:31 +08:00
|
|
|
xfs_alloc_log_agf(tp, agbp, logflags);
|
2013-04-03 13:11:14 +08:00
|
|
|
|
2013-04-03 13:11:30 +08:00
|
|
|
xfs_trans_buf_set_type(tp, agflbp, XFS_BLFT_AGFL_BUF);
|
2013-04-03 13:11:14 +08:00
|
|
|
xfs_trans_log_buf(tp, agflbp, startoff,
|
|
|
|
startoff + sizeof(xfs_agblock_t) - 1);
|
2005-04-17 06:20:36 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2023-06-30 01:15:45 +08:00
|
|
|
/*
|
|
|
|
* Check that this AGF/AGI header's sequence number and length matches the AG
|
|
|
|
* number and size in fsblocks.
|
|
|
|
*/
|
|
|
|
xfs_failaddr_t
|
|
|
|
xfs_validate_ag_length(
|
|
|
|
struct xfs_buf *bp,
|
|
|
|
uint32_t seqno,
|
|
|
|
uint32_t length)
|
|
|
|
{
|
|
|
|
struct xfs_mount *mp = bp->b_mount;
|
|
|
|
/*
|
|
|
|
* During growfs operations, the perag is not fully initialised,
|
|
|
|
* so we can't use it for any useful checking. growfs ensures we can't
|
|
|
|
* use it by using uncached buffers that don't have the perag attached
|
|
|
|
* so we can detect and avoid this problem.
|
|
|
|
*/
|
|
|
|
if (bp->b_pag && seqno != bp->b_pag->pag_agno)
|
|
|
|
return __this_address;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Only the last AG in the filesystem is allowed to be shorter
|
|
|
|
* than the AG size recorded in the superblock.
|
|
|
|
*/
|
|
|
|
if (length != mp->m_sb.sb_agblocks) {
|
|
|
|
/*
|
|
|
|
* During growfs, the new last AG can get here before we
|
|
|
|
* have updated the superblock. Give it a pass on the seqno
|
|
|
|
* check.
|
|
|
|
*/
|
|
|
|
if (bp->b_pag && seqno != mp->m_sb.sb_agcount - 1)
|
|
|
|
return __this_address;
|
|
|
|
if (length < XFS_MIN_AG_BLOCKS)
|
|
|
|
return __this_address;
|
|
|
|
if (length > mp->m_sb.sb_agblocks)
|
|
|
|
return __this_address;
|
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2023-06-05 12:48:15 +08:00
|
|
|
/*
|
|
|
|
* Verify the AGF is consistent.
|
|
|
|
*
|
|
|
|
* We do not verify the AGFL indexes in the AGF are fully consistent here
|
|
|
|
* because of issues with variable on-disk structure sizes. Instead, we check
|
|
|
|
* the agfl indexes for consistency when we initialise the perag from the AGF
|
|
|
|
* information after a read completes.
|
|
|
|
*
|
|
|
|
* If the index is inconsistent, then we mark the perag as needing an AGFL
|
|
|
|
* reset. The first AGFL update performed then resets the AGFL indexes and
|
|
|
|
* refills the AGFL with known good free blocks, allowing the filesystem to
|
|
|
|
* continue operating normally at the cost of a few leaked free space blocks.
|
|
|
|
*/
|
2018-01-09 02:51:03 +08:00
|
|
|
static xfs_failaddr_t
|
2012-11-14 14:52:32 +08:00
|
|
|
xfs_agf_verify(
|
2018-01-09 02:51:08 +08:00
|
|
|
struct xfs_buf *bp)
|
|
|
|
{
|
2019-06-29 10:27:29 +08:00
|
|
|
struct xfs_mount *mp = bp->b_mount;
|
2020-03-10 23:57:29 +08:00
|
|
|
struct xfs_agf *agf = bp->b_addr;
|
2023-06-30 01:15:45 +08:00
|
|
|
xfs_failaddr_t fa;
|
|
|
|
uint32_t agf_seqno = be32_to_cpu(agf->agf_seqno);
|
2023-06-29 10:43:55 +08:00
|
|
|
uint32_t agf_length = be32_to_cpu(agf->agf_length);
|
2012-11-14 14:44:56 +08:00
|
|
|
|
2021-08-19 09:46:37 +08:00
|
|
|
if (xfs_has_crc(mp)) {
|
2015-10-12 12:59:25 +08:00
|
|
|
if (!uuid_equal(&agf->agf_uuid, &mp->m_sb.sb_meta_uuid))
|
2018-01-09 02:51:03 +08:00
|
|
|
return __this_address;
|
2020-03-10 23:57:29 +08:00
|
|
|
if (!xfs_log_check_lsn(mp, be64_to_cpu(agf->agf_lsn)))
|
2018-01-09 02:51:03 +08:00
|
|
|
return __this_address;
|
2015-10-12 12:59:25 +08:00
|
|
|
}
|
2012-11-14 14:44:56 +08:00
|
|
|
|
2019-02-08 02:45:48 +08:00
|
|
|
if (!xfs_verify_magic(bp, agf->agf_magicnum))
|
|
|
|
return __this_address;
|
|
|
|
|
2023-06-29 10:43:55 +08:00
|
|
|
if (!XFS_AGF_GOOD_VERSION(be32_to_cpu(agf->agf_versionnum)))
|
2018-01-09 02:51:03 +08:00
|
|
|
return __this_address;
|
2012-11-14 14:44:56 +08:00
|
|
|
|
2023-06-29 10:43:55 +08:00
|
|
|
/*
|
|
|
|
* Both agf_seqno and agf_length need to validated before anything else
|
|
|
|
* block number related in the AGF or AGFL can be checked.
|
|
|
|
*/
|
2023-06-30 01:15:45 +08:00
|
|
|
fa = xfs_validate_ag_length(bp, agf_seqno, agf_length);
|
|
|
|
if (fa)
|
|
|
|
return fa;
|
2023-06-29 10:43:55 +08:00
|
|
|
|
|
|
|
if (be32_to_cpu(agf->agf_flfirst) >= xfs_agfl_size(mp))
|
|
|
|
return __this_address;
|
|
|
|
if (be32_to_cpu(agf->agf_fllast) >= xfs_agfl_size(mp))
|
|
|
|
return __this_address;
|
|
|
|
if (be32_to_cpu(agf->agf_flcount) > xfs_agfl_size(mp))
|
xfs: add agf freeblocks verify in xfs_agf_verify
We recently used fuzz(hydra) to test XFS and automatically generate
tmp.img(XFS v5 format, but some metadata is wrong)
xfs_repair information(just one AG):
agf_freeblks 0, counted 3224 in ag 0
agf_longest 536874136, counted 3224 in ag 0
sb_fdblocks 613, counted 3228
Test as follows:
mount tmp.img tmpdir
cp file1M tmpdir
sync
In 4.19-stable, sync will stuck, the reason is:
xfs_mountfs
xfs_check_summary_counts
if ((!xfs_sb_version_haslazysbcount(&mp->m_sb) ||
XFS_LAST_UNMOUNT_WAS_CLEAN(mp)) &&
!xfs_fs_has_sickness(mp, XFS_SICK_FS_COUNTERS))
return 0; -->just return, incore sb_fdblocks still be 613
xfs_initialize_perag_data
cp file1M tmpdir -->ok(write file to pagecache)
sync -->stuck(write pagecache to disk)
xfs_map_blocks
xfs_iomap_write_allocate
while (count_fsb != 0) {
nimaps = 0;
while (nimaps == 0) { --> endless loop
nimaps = 1;
xfs_bmapi_write(..., &nimaps) --> nimaps becomes 0 again
xfs_bmapi_write
xfs_bmap_alloc
xfs_bmap_btalloc
xfs_alloc_vextent
xfs_alloc_fix_freelist
xfs_alloc_space_available -->fail(agf_freeblks is 0)
In linux-next, sync not stuck, cause commit c2b3164320b5 ("xfs:
use the latest extent at writeback delalloc conversion time") remove
the above while, dmesg is as follows:
[ 55.250114] XFS (loop0): page discard on page ffffea0008bc7380, inode 0x1b0c, offset 0.
Users do not know why this page is discard, the better soultion is:
1. Like xfs_repair, make sure sb_fdblocks is equal to counted
(xfs_initialize_perag_data did this, who is not called at this mount)
2. Add agf verify, if fail, will tell users to repair
This patch use the second soultion.
Signed-off-by: Zheng Bin <zhengbin13@huawei.com>
Signed-off-by: Ren Xudong <renxudong1@huawei.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2020-02-21 23:38:20 +08:00
|
|
|
return __this_address;
|
|
|
|
|
|
|
|
if (be32_to_cpu(agf->agf_freeblks) < be32_to_cpu(agf->agf_longest) ||
|
2023-06-29 10:43:55 +08:00
|
|
|
be32_to_cpu(agf->agf_freeblks) > agf_length)
|
xfs: add agf freeblocks verify in xfs_agf_verify
We recently used fuzz(hydra) to test XFS and automatically generate
tmp.img(XFS v5 format, but some metadata is wrong)
xfs_repair information(just one AG):
agf_freeblks 0, counted 3224 in ag 0
agf_longest 536874136, counted 3224 in ag 0
sb_fdblocks 613, counted 3228
Test as follows:
mount tmp.img tmpdir
cp file1M tmpdir
sync
In 4.19-stable, sync will stuck, the reason is:
xfs_mountfs
xfs_check_summary_counts
if ((!xfs_sb_version_haslazysbcount(&mp->m_sb) ||
XFS_LAST_UNMOUNT_WAS_CLEAN(mp)) &&
!xfs_fs_has_sickness(mp, XFS_SICK_FS_COUNTERS))
return 0; -->just return, incore sb_fdblocks still be 613
xfs_initialize_perag_data
cp file1M tmpdir -->ok(write file to pagecache)
sync -->stuck(write pagecache to disk)
xfs_map_blocks
xfs_iomap_write_allocate
while (count_fsb != 0) {
nimaps = 0;
while (nimaps == 0) { --> endless loop
nimaps = 1;
xfs_bmapi_write(..., &nimaps) --> nimaps becomes 0 again
xfs_bmapi_write
xfs_bmap_alloc
xfs_bmap_btalloc
xfs_alloc_vextent
xfs_alloc_fix_freelist
xfs_alloc_space_available -->fail(agf_freeblks is 0)
In linux-next, sync not stuck, cause commit c2b3164320b5 ("xfs:
use the latest extent at writeback delalloc conversion time") remove
the above while, dmesg is as follows:
[ 55.250114] XFS (loop0): page discard on page ffffea0008bc7380, inode 0x1b0c, offset 0.
Users do not know why this page is discard, the better soultion is:
1. Like xfs_repair, make sure sb_fdblocks is equal to counted
(xfs_initialize_perag_data did this, who is not called at this mount)
2. Add agf verify, if fail, will tell users to repair
This patch use the second soultion.
Signed-off-by: Zheng Bin <zhengbin13@huawei.com>
Signed-off-by: Ren Xudong <renxudong1@huawei.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2020-02-21 23:38:20 +08:00
|
|
|
return __this_address;
|
|
|
|
|
2016-12-05 09:32:50 +08:00
|
|
|
if (be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNO]) < 1 ||
|
|
|
|
be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]) < 1 ||
|
2021-10-14 01:02:19 +08:00
|
|
|
be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNO]) >
|
|
|
|
mp->m_alloc_maxlevels ||
|
|
|
|
be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]) >
|
|
|
|
mp->m_alloc_maxlevels)
|
2018-01-09 02:51:03 +08:00
|
|
|
return __this_address;
|
2014-09-09 09:47:24 +08:00
|
|
|
|
2023-06-29 10:43:55 +08:00
|
|
|
if (xfs_has_lazysbcount(mp) &&
|
|
|
|
be32_to_cpu(agf->agf_btreeblks) > agf_length)
|
xfs: add agf freeblocks verify in xfs_agf_verify
We recently used fuzz(hydra) to test XFS and automatically generate
tmp.img(XFS v5 format, but some metadata is wrong)
xfs_repair information(just one AG):
agf_freeblks 0, counted 3224 in ag 0
agf_longest 536874136, counted 3224 in ag 0
sb_fdblocks 613, counted 3228
Test as follows:
mount tmp.img tmpdir
cp file1M tmpdir
sync
In 4.19-stable, sync will stuck, the reason is:
xfs_mountfs
xfs_check_summary_counts
if ((!xfs_sb_version_haslazysbcount(&mp->m_sb) ||
XFS_LAST_UNMOUNT_WAS_CLEAN(mp)) &&
!xfs_fs_has_sickness(mp, XFS_SICK_FS_COUNTERS))
return 0; -->just return, incore sb_fdblocks still be 613
xfs_initialize_perag_data
cp file1M tmpdir -->ok(write file to pagecache)
sync -->stuck(write pagecache to disk)
xfs_map_blocks
xfs_iomap_write_allocate
while (count_fsb != 0) {
nimaps = 0;
while (nimaps == 0) { --> endless loop
nimaps = 1;
xfs_bmapi_write(..., &nimaps) --> nimaps becomes 0 again
xfs_bmapi_write
xfs_bmap_alloc
xfs_bmap_btalloc
xfs_alloc_vextent
xfs_alloc_fix_freelist
xfs_alloc_space_available -->fail(agf_freeblks is 0)
In linux-next, sync not stuck, cause commit c2b3164320b5 ("xfs:
use the latest extent at writeback delalloc conversion time") remove
the above while, dmesg is as follows:
[ 55.250114] XFS (loop0): page discard on page ffffea0008bc7380, inode 0x1b0c, offset 0.
Users do not know why this page is discard, the better soultion is:
1. Like xfs_repair, make sure sb_fdblocks is equal to counted
(xfs_initialize_perag_data did this, who is not called at this mount)
2. Add agf verify, if fail, will tell users to repair
This patch use the second soultion.
Signed-off-by: Zheng Bin <zhengbin13@huawei.com>
Signed-off-by: Ren Xudong <renxudong1@huawei.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2020-02-21 23:38:20 +08:00
|
|
|
return __this_address;
|
|
|
|
|
2023-06-29 10:43:55 +08:00
|
|
|
if (xfs_has_rmapbt(mp)) {
|
|
|
|
if (be32_to_cpu(agf->agf_rmap_blocks) > agf_length)
|
|
|
|
return __this_address;
|
2012-11-14 14:44:56 +08:00
|
|
|
|
2023-06-29 10:43:55 +08:00
|
|
|
if (be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAP]) < 1 ||
|
|
|
|
be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAP]) >
|
|
|
|
mp->m_rmap_maxlevels)
|
|
|
|
return __this_address;
|
|
|
|
}
|
2013-04-03 13:11:13 +08:00
|
|
|
|
2023-06-29 10:43:55 +08:00
|
|
|
if (xfs_has_reflink(mp)) {
|
|
|
|
if (be32_to_cpu(agf->agf_refcount_blocks) > agf_length)
|
|
|
|
return __this_address;
|
xfs: add agf freeblocks verify in xfs_agf_verify
We recently used fuzz(hydra) to test XFS and automatically generate
tmp.img(XFS v5 format, but some metadata is wrong)
xfs_repair information(just one AG):
agf_freeblks 0, counted 3224 in ag 0
agf_longest 536874136, counted 3224 in ag 0
sb_fdblocks 613, counted 3228
Test as follows:
mount tmp.img tmpdir
cp file1M tmpdir
sync
In 4.19-stable, sync will stuck, the reason is:
xfs_mountfs
xfs_check_summary_counts
if ((!xfs_sb_version_haslazysbcount(&mp->m_sb) ||
XFS_LAST_UNMOUNT_WAS_CLEAN(mp)) &&
!xfs_fs_has_sickness(mp, XFS_SICK_FS_COUNTERS))
return 0; -->just return, incore sb_fdblocks still be 613
xfs_initialize_perag_data
cp file1M tmpdir -->ok(write file to pagecache)
sync -->stuck(write pagecache to disk)
xfs_map_blocks
xfs_iomap_write_allocate
while (count_fsb != 0) {
nimaps = 0;
while (nimaps == 0) { --> endless loop
nimaps = 1;
xfs_bmapi_write(..., &nimaps) --> nimaps becomes 0 again
xfs_bmapi_write
xfs_bmap_alloc
xfs_bmap_btalloc
xfs_alloc_vextent
xfs_alloc_fix_freelist
xfs_alloc_space_available -->fail(agf_freeblks is 0)
In linux-next, sync not stuck, cause commit c2b3164320b5 ("xfs:
use the latest extent at writeback delalloc conversion time") remove
the above while, dmesg is as follows:
[ 55.250114] XFS (loop0): page discard on page ffffea0008bc7380, inode 0x1b0c, offset 0.
Users do not know why this page is discard, the better soultion is:
1. Like xfs_repair, make sure sb_fdblocks is equal to counted
(xfs_initialize_perag_data did this, who is not called at this mount)
2. Add agf verify, if fail, will tell users to repair
This patch use the second soultion.
Signed-off-by: Zheng Bin <zhengbin13@huawei.com>
Signed-off-by: Ren Xudong <renxudong1@huawei.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2020-02-21 23:38:20 +08:00
|
|
|
|
2023-06-29 10:43:55 +08:00
|
|
|
if (be32_to_cpu(agf->agf_refcount_level) < 1 ||
|
|
|
|
be32_to_cpu(agf->agf_refcount_level) > mp->m_refc_maxlevels)
|
|
|
|
return __this_address;
|
|
|
|
}
|
2016-10-04 00:11:16 +08:00
|
|
|
|
2018-01-09 02:51:03 +08:00
|
|
|
return NULL;
|
2012-11-14 14:52:32 +08:00
|
|
|
}
|
|
|
|
|
2012-11-14 14:54:40 +08:00
|
|
|
static void
|
|
|
|
xfs_agf_read_verify(
|
2012-11-14 14:52:32 +08:00
|
|
|
struct xfs_buf *bp)
|
|
|
|
{
|
2019-06-29 10:27:29 +08:00
|
|
|
struct xfs_mount *mp = bp->b_mount;
|
2018-01-09 02:51:03 +08:00
|
|
|
xfs_failaddr_t fa;
|
2013-04-03 13:11:13 +08:00
|
|
|
|
2021-08-19 09:46:37 +08:00
|
|
|
if (xfs_has_crc(mp) &&
|
2014-02-27 12:23:10 +08:00
|
|
|
!xfs_buf_verify_cksum(bp, XFS_AGF_CRC_OFF))
|
2018-01-09 02:51:03 +08:00
|
|
|
xfs_verifier_error(bp, -EFSBADCRC, __this_address);
|
|
|
|
else {
|
2018-01-09 02:51:08 +08:00
|
|
|
fa = xfs_agf_verify(bp);
|
2018-01-09 02:51:03 +08:00
|
|
|
if (XFS_TEST_ERROR(fa, mp, XFS_ERRTAG_ALLOC_READ_AGF))
|
|
|
|
xfs_verifier_error(bp, -EFSCORRUPTED, fa);
|
|
|
|
}
|
2012-11-14 14:52:32 +08:00
|
|
|
}
|
2012-11-14 14:44:56 +08:00
|
|
|
|
2012-11-14 14:53:49 +08:00
|
|
|
static void
|
2012-11-14 14:54:40 +08:00
|
|
|
xfs_agf_write_verify(
|
2012-11-14 14:52:32 +08:00
|
|
|
struct xfs_buf *bp)
|
|
|
|
{
|
2019-06-29 10:27:29 +08:00
|
|
|
struct xfs_mount *mp = bp->b_mount;
|
2018-01-25 05:38:48 +08:00
|
|
|
struct xfs_buf_log_item *bip = bp->b_log_item;
|
2020-03-10 23:57:29 +08:00
|
|
|
struct xfs_agf *agf = bp->b_addr;
|
2018-01-09 02:51:03 +08:00
|
|
|
xfs_failaddr_t fa;
|
2013-04-03 13:11:13 +08:00
|
|
|
|
2018-01-09 02:51:08 +08:00
|
|
|
fa = xfs_agf_verify(bp);
|
2018-01-09 02:51:03 +08:00
|
|
|
if (fa) {
|
|
|
|
xfs_verifier_error(bp, -EFSCORRUPTED, fa);
|
2013-04-03 13:11:13 +08:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2021-08-19 09:46:37 +08:00
|
|
|
if (!xfs_has_crc(mp))
|
2013-04-03 13:11:13 +08:00
|
|
|
return;
|
|
|
|
|
|
|
|
if (bip)
|
2020-03-10 23:57:29 +08:00
|
|
|
agf->agf_lsn = cpu_to_be64(bip->bli_item.li_lsn);
|
2013-04-03 13:11:13 +08:00
|
|
|
|
2014-02-27 12:18:23 +08:00
|
|
|
xfs_buf_update_cksum(bp, XFS_AGF_CRC_OFF);
|
2012-11-14 14:44:56 +08:00
|
|
|
}
|
|
|
|
|
2012-11-14 14:54:40 +08:00
|
|
|
const struct xfs_buf_ops xfs_agf_buf_ops = {
|
2016-01-04 13:10:19 +08:00
|
|
|
.name = "xfs_agf",
|
2019-02-08 02:45:48 +08:00
|
|
|
.magic = { cpu_to_be32(XFS_AGF_MAGIC), cpu_to_be32(XFS_AGF_MAGIC) },
|
2012-11-14 14:54:40 +08:00
|
|
|
.verify_read = xfs_agf_read_verify,
|
|
|
|
.verify_write = xfs_agf_write_verify,
|
2018-01-09 02:51:08 +08:00
|
|
|
.verify_struct = xfs_agf_verify,
|
2012-11-14 14:54:40 +08:00
|
|
|
};
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Read in the allocation group header (free/alloc section).
|
|
|
|
*/
|
2022-07-07 17:07:54 +08:00
|
|
|
int
|
2008-11-28 11:23:38 +08:00
|
|
|
xfs_read_agf(
|
2022-07-07 17:07:54 +08:00
|
|
|
struct xfs_perag *pag,
|
|
|
|
struct xfs_trans *tp,
|
|
|
|
int flags,
|
|
|
|
struct xfs_buf **agfbpp)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2022-07-07 17:07:54 +08:00
|
|
|
struct xfs_mount *mp = pag->pag_mount;
|
|
|
|
int error;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2022-07-07 17:07:54 +08:00
|
|
|
trace_xfs_read_agf(pag->pag_mount, pag->pag_agno);
|
2013-11-01 12:27:19 +08:00
|
|
|
|
2020-01-24 09:01:16 +08:00
|
|
|
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
|
2022-07-07 17:07:54 +08:00
|
|
|
XFS_AG_DADDR(mp, pag->pag_agno, XFS_AGF_DADDR(mp)),
|
|
|
|
XFS_FSS_TO_BB(mp, 1), flags, agfbpp, &xfs_agf_buf_ops);
|
2005-04-17 06:20:36 +08:00
|
|
|
if (error)
|
|
|
|
return error;
|
2008-11-28 11:23:38 +08:00
|
|
|
|
2022-07-07 17:07:54 +08:00
|
|
|
xfs_buf_set_ref(*agfbpp, XFS_AGF_REF);
|
2008-11-28 11:23:38 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2022-07-07 17:07:32 +08:00
|
|
|
* Read in the allocation group header (free/alloc section) and initialise the
|
|
|
|
* perag structure if necessary. If the caller provides @agfbpp, then return the
|
|
|
|
* locked buffer to the caller, otherwise free it.
|
2008-11-28 11:23:38 +08:00
|
|
|
*/
|
2022-07-07 17:07:40 +08:00
|
|
|
int
|
2008-11-28 11:23:38 +08:00
|
|
|
xfs_alloc_read_agf(
|
2022-07-07 17:07:40 +08:00
|
|
|
struct xfs_perag *pag,
|
|
|
|
struct xfs_trans *tp,
|
|
|
|
int flags,
|
2022-07-07 17:07:32 +08:00
|
|
|
struct xfs_buf **agfbpp)
|
2008-11-28 11:23:38 +08:00
|
|
|
{
|
2022-07-07 17:07:32 +08:00
|
|
|
struct xfs_buf *agfbp;
|
2022-07-07 17:07:40 +08:00
|
|
|
struct xfs_agf *agf;
|
2008-11-28 11:23:38 +08:00
|
|
|
int error;
|
2021-04-29 06:05:50 +08:00
|
|
|
int allocbt_blks;
|
2008-11-28 11:23:38 +08:00
|
|
|
|
2022-07-07 17:07:40 +08:00
|
|
|
trace_xfs_alloc_read_agf(pag->pag_mount, pag->pag_agno);
|
2008-11-28 11:23:38 +08:00
|
|
|
|
2020-01-24 09:01:19 +08:00
|
|
|
/* We don't support trylock when freeing. */
|
|
|
|
ASSERT((flags & (XFS_ALLOC_FLAG_FREEING | XFS_ALLOC_FLAG_TRYLOCK)) !=
|
|
|
|
(XFS_ALLOC_FLAG_FREEING | XFS_ALLOC_FLAG_TRYLOCK));
|
2022-07-07 17:07:54 +08:00
|
|
|
error = xfs_read_agf(pag, tp,
|
2010-01-19 17:56:44 +08:00
|
|
|
(flags & XFS_ALLOC_FLAG_TRYLOCK) ? XBF_TRYLOCK : 0,
|
2022-07-07 17:07:32 +08:00
|
|
|
&agfbp);
|
2008-11-28 11:23:38 +08:00
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
2022-07-07 17:07:32 +08:00
|
|
|
agf = agfbp->b_addr;
|
2023-02-13 06:14:52 +08:00
|
|
|
if (!xfs_perag_initialised_agf(pag)) {
|
2005-11-02 12:11:25 +08:00
|
|
|
pag->pagf_freeblks = be32_to_cpu(agf->agf_freeblks);
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 13:26:31 +08:00
|
|
|
pag->pagf_btreeblks = be32_to_cpu(agf->agf_btreeblks);
|
2005-11-02 12:11:25 +08:00
|
|
|
pag->pagf_flcount = be32_to_cpu(agf->agf_flcount);
|
|
|
|
pag->pagf_longest = be32_to_cpu(agf->agf_longest);
|
2005-04-17 06:20:36 +08:00
|
|
|
pag->pagf_levels[XFS_BTNUM_BNOi] =
|
2005-11-02 12:11:25 +08:00
|
|
|
be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNOi]);
|
2005-04-17 06:20:36 +08:00
|
|
|
pag->pagf_levels[XFS_BTNUM_CNTi] =
|
2005-11-02 12:11:25 +08:00
|
|
|
be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNTi]);
|
2016-08-03 09:30:32 +08:00
|
|
|
pag->pagf_levels[XFS_BTNUM_RMAPi] =
|
|
|
|
be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAPi]);
|
2016-10-04 00:11:16 +08:00
|
|
|
pag->pagf_refcount_level = be32_to_cpu(agf->agf_refcount_level);
|
2023-02-13 06:14:52 +08:00
|
|
|
if (xfs_agfl_needs_reset(pag->pag_mount, agf))
|
|
|
|
set_bit(XFS_AGSTATE_AGFL_NEEDS_RESET, &pag->pag_opstate);
|
xfs: clear incore AGFL_RESET state if it's not needed
Prior to commit 7ac2ff8bb371, when we loaded the incore perag structure
with information from the AGF header, we would set or clear the
pagf_agfl_reset field based on whether or not the AGFL list was
misaligned within the block. IOWs, it's an incore state bit that's
supposed to cache something in the ondisk metadata. Therefore, the code
still needs to support clearing the incore bit if (somehow) the AGFL
were to correct itself.
It turns out that xfs_repair does exactly this -- phase 4 loads the AGF
to scan the rmapbt for corrupt records, which can set NEEDS_AGFL_RESET.
The scan unsets AGF_INIT but doesn't unset NEEDS_AGFL_RESET. Phase 5
totally rewrites the AGFL and fixes the alignment problem, didn't clear
NEEDS_AGFL_RESET historically, and reloads the perag state to fix the
freelist. This results in the AGFL being reset based on stale data,
which then causes the new AGFL blocks to be leaked. A subsequent
xfs_repair -n then complains about the leaks.
One could argue that phase 5 ought to clear this bit directly when it
reloads the perag AGF data after rewriting the AGFL, but libxfs used to
handle this for us, so it should go back to doing that.
Found by fuzzing flfirst = ones in xfs/352.
Fixes: 7ac2ff8bb371 ("xfs: perags need atomic operational state")
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
2023-03-22 07:33:20 +08:00
|
|
|
else
|
|
|
|
clear_bit(XFS_AGSTATE_AGFL_NEEDS_RESET, &pag->pag_opstate);
|
2021-04-29 06:05:50 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Update the in-core allocbt counter. Filter out the rmapbt
|
|
|
|
* subset of the btreeblks counter because the rmapbt is managed
|
|
|
|
* by perag reservation. Subtract one for the rmapbt root block
|
|
|
|
* because the rmap counter includes it while the btreeblks
|
|
|
|
* counter only tracks non-root blocks.
|
|
|
|
*/
|
|
|
|
allocbt_blks = pag->pagf_btreeblks;
|
2022-07-07 17:07:40 +08:00
|
|
|
if (xfs_has_rmapbt(pag->pag_mount))
|
2021-04-29 06:05:50 +08:00
|
|
|
allocbt_blks -= be32_to_cpu(agf->agf_rmap_blocks) - 1;
|
|
|
|
if (allocbt_blks > 0)
|
2022-07-07 17:07:40 +08:00
|
|
|
atomic64_add(allocbt_blks,
|
|
|
|
&pag->pag_mount->m_allocbt_blks);
|
2023-02-13 06:14:52 +08:00
|
|
|
|
|
|
|
set_bit(XFS_AGSTATE_AGF_INIT, &pag->pag_opstate);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
#ifdef DEBUG
|
2022-07-07 17:07:40 +08:00
|
|
|
else if (!xfs_is_shutdown(pag->pag_mount)) {
|
2005-11-02 12:11:25 +08:00
|
|
|
ASSERT(pag->pagf_freeblks == be32_to_cpu(agf->agf_freeblks));
|
2008-10-30 14:05:49 +08:00
|
|
|
ASSERT(pag->pagf_btreeblks == be32_to_cpu(agf->agf_btreeblks));
|
2005-11-02 12:11:25 +08:00
|
|
|
ASSERT(pag->pagf_flcount == be32_to_cpu(agf->agf_flcount));
|
|
|
|
ASSERT(pag->pagf_longest == be32_to_cpu(agf->agf_longest));
|
2005-04-17 06:20:36 +08:00
|
|
|
ASSERT(pag->pagf_levels[XFS_BTNUM_BNOi] ==
|
2005-11-02 12:11:25 +08:00
|
|
|
be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNOi]));
|
2005-04-17 06:20:36 +08:00
|
|
|
ASSERT(pag->pagf_levels[XFS_BTNUM_CNTi] ==
|
2005-11-02 12:11:25 +08:00
|
|
|
be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNTi]));
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
#endif
|
2022-07-07 17:07:32 +08:00
|
|
|
if (agfbpp)
|
|
|
|
*agfbpp = agfbp;
|
|
|
|
else
|
|
|
|
xfs_trans_brelse(tp, agfbp);
|
2005-04-17 06:20:36 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2023-02-13 06:14:53 +08:00
|
|
|
* Pre-proces allocation arguments to set initial state that we don't require
|
|
|
|
* callers to set up correctly, as well as bounds check the allocation args
|
|
|
|
* that are set up.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2023-02-13 06:14:53 +08:00
|
|
|
static int
|
|
|
|
xfs_alloc_vextent_check_args(
|
2023-02-13 06:14:53 +08:00
|
|
|
struct xfs_alloc_arg *args,
|
2023-02-13 06:14:54 +08:00
|
|
|
xfs_fsblock_t target,
|
|
|
|
xfs_agnumber_t *minimum_agno)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2023-02-13 06:14:53 +08:00
|
|
|
struct xfs_mount *mp = args->mp;
|
|
|
|
xfs_agblock_t agsize;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
args->fsbno = NULLFSBLOCK;
|
2023-02-13 06:14:53 +08:00
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
*minimum_agno = 0;
|
|
|
|
if (args->tp->t_highest_agno != NULLAGNUMBER)
|
|
|
|
*minimum_agno = args->tp->t_highest_agno;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Just fix this up, for the case where the last a.g. is shorter
|
|
|
|
* (or there's only one a.g.) and the caller couldn't easily figure
|
|
|
|
* that out (xfs_bmap_alloc).
|
|
|
|
*/
|
|
|
|
agsize = mp->m_sb.sb_agblocks;
|
|
|
|
if (args->maxlen > agsize)
|
|
|
|
args->maxlen = agsize;
|
|
|
|
if (args->alignment == 0)
|
|
|
|
args->alignment = 1;
|
2023-02-13 06:14:54 +08:00
|
|
|
|
|
|
|
ASSERT(args->minlen > 0);
|
|
|
|
ASSERT(args->maxlen > 0);
|
|
|
|
ASSERT(args->alignment > 0);
|
|
|
|
ASSERT(args->resv != XFS_AG_RESV_AGFL);
|
|
|
|
|
2023-02-13 06:14:53 +08:00
|
|
|
ASSERT(XFS_FSB_TO_AGNO(mp, target) < mp->m_sb.sb_agcount);
|
|
|
|
ASSERT(XFS_FSB_TO_AGBNO(mp, target) < agsize);
|
2005-04-17 06:20:36 +08:00
|
|
|
ASSERT(args->minlen <= args->maxlen);
|
|
|
|
ASSERT(args->minlen <= agsize);
|
|
|
|
ASSERT(args->mod < args->prod);
|
2023-02-13 06:14:54 +08:00
|
|
|
|
2023-02-13 06:14:53 +08:00
|
|
|
if (XFS_FSB_TO_AGNO(mp, target) >= mp->m_sb.sb_agcount ||
|
|
|
|
XFS_FSB_TO_AGBNO(mp, target) >= agsize ||
|
2005-04-17 06:20:36 +08:00
|
|
|
args->minlen > args->maxlen || args->minlen > agsize ||
|
|
|
|
args->mod >= args->prod) {
|
2009-12-15 07:14:59 +08:00
|
|
|
trace_xfs_alloc_vextent_badargs(args);
|
2023-02-13 06:14:53 +08:00
|
|
|
return -ENOSPC;
|
|
|
|
}
|
2023-02-13 06:14:54 +08:00
|
|
|
|
|
|
|
if (args->agno != NULLAGNUMBER && *minimum_agno > args->agno) {
|
|
|
|
trace_xfs_alloc_vextent_skip_deadlock(args);
|
|
|
|
return -ENOSPC;
|
|
|
|
}
|
2023-02-13 06:14:53 +08:00
|
|
|
return 0;
|
2023-02-13 06:14:54 +08:00
|
|
|
|
2023-02-13 06:14:53 +08:00
|
|
|
}
|
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
/*
|
|
|
|
* Prepare an AG for allocation. If the AG is not prepared to accept the
|
|
|
|
* allocation, return failure.
|
|
|
|
*
|
|
|
|
* XXX(dgc): The complexity of "need_pag" will go away as all caller paths are
|
|
|
|
* modified to hold their own perag references.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
xfs_alloc_vextent_prepare_ag(
|
2023-06-05 02:06:27 +08:00
|
|
|
struct xfs_alloc_arg *args,
|
2023-06-29 02:04:32 +08:00
|
|
|
uint32_t alloc_flags)
|
2023-02-13 06:14:54 +08:00
|
|
|
{
|
|
|
|
bool need_pag = !args->pag;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
if (need_pag)
|
|
|
|
args->pag = xfs_perag_get(args->mp, args->agno);
|
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
args->agbp = NULL;
|
2023-06-29 02:04:32 +08:00
|
|
|
error = xfs_alloc_fix_freelist(args, alloc_flags);
|
2023-02-13 06:14:54 +08:00
|
|
|
if (error) {
|
|
|
|
trace_xfs_alloc_vextent_nofix(args);
|
|
|
|
if (need_pag)
|
|
|
|
xfs_perag_put(args->pag);
|
|
|
|
args->agbno = NULLAGBLOCK;
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
if (!args->agbp) {
|
|
|
|
/* cannot allocate in this AG at all */
|
|
|
|
trace_xfs_alloc_vextent_noagbp(args);
|
|
|
|
args->agbno = NULLAGBLOCK;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
args->wasfromfl = 0;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2023-02-13 06:14:53 +08:00
|
|
|
/*
|
2023-02-13 06:14:54 +08:00
|
|
|
* Post-process allocation results to account for the allocation if it succeed
|
|
|
|
* and set the allocated block number correctly for the caller.
|
2023-02-13 06:14:53 +08:00
|
|
|
*
|
2023-02-13 06:14:54 +08:00
|
|
|
* XXX: we should really be returning ENOSPC for ENOSPC, not
|
2023-02-13 06:14:53 +08:00
|
|
|
* hiding it behind a "successful" NULLFSBLOCK allocation.
|
|
|
|
*/
|
2023-02-13 06:14:54 +08:00
|
|
|
static int
|
|
|
|
xfs_alloc_vextent_finish(
|
2023-02-13 06:14:53 +08:00
|
|
|
struct xfs_alloc_arg *args,
|
2023-02-13 06:14:54 +08:00
|
|
|
xfs_agnumber_t minimum_agno,
|
|
|
|
int alloc_error,
|
|
|
|
bool drop_perag)
|
2023-02-13 06:14:53 +08:00
|
|
|
{
|
|
|
|
struct xfs_mount *mp = args->mp;
|
2023-02-13 06:14:54 +08:00
|
|
|
int error = 0;
|
2023-02-13 06:14:53 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* We can end up here with a locked AGF. If we failed, the caller is
|
|
|
|
* likely going to try to allocate again with different parameters, and
|
|
|
|
* that can widen the AGs that are searched for free space. If we have
|
|
|
|
* to do BMBT block allocation, we have to do a new allocation.
|
|
|
|
*
|
|
|
|
* Hence leaving this function with the AGF locked opens up potential
|
|
|
|
* ABBA AGF deadlocks because a future allocation attempt in this
|
|
|
|
* transaction may attempt to lock a lower number AGF.
|
|
|
|
*
|
|
|
|
* We can't release the AGF until the transaction is commited, so at
|
|
|
|
* this point we must update the "first allocation" tracker to point at
|
|
|
|
* this AG if the tracker is empty or points to a lower AG. This allows
|
|
|
|
* the next allocation attempt to be modified appropriately to avoid
|
|
|
|
* deadlocks.
|
|
|
|
*/
|
|
|
|
if (args->agbp &&
|
|
|
|
(args->tp->t_highest_agno == NULLAGNUMBER ||
|
|
|
|
args->agno > minimum_agno))
|
|
|
|
args->tp->t_highest_agno = args->agno;
|
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
/*
|
|
|
|
* If the allocation failed with an error or we had an ENOSPC result,
|
|
|
|
* preserve the returned error whilst also marking the allocation result
|
|
|
|
* as "no extent allocated". This ensures that callers that fail to
|
|
|
|
* capture the error will still treat it as a failed allocation.
|
|
|
|
*/
|
|
|
|
if (alloc_error || args->agbno == NULLAGBLOCK) {
|
2023-02-13 06:14:53 +08:00
|
|
|
args->fsbno = NULLFSBLOCK;
|
2023-02-13 06:14:54 +08:00
|
|
|
error = alloc_error;
|
|
|
|
goto out_drop_perag;
|
2023-02-13 06:14:53 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
args->fsbno = XFS_AGB_TO_FSB(mp, args->agno, args->agbno);
|
2023-02-13 06:14:54 +08:00
|
|
|
|
2023-02-13 06:14:53 +08:00
|
|
|
ASSERT(args->len >= args->minlen);
|
|
|
|
ASSERT(args->len <= args->maxlen);
|
|
|
|
ASSERT(args->agbno % args->alignment == 0);
|
|
|
|
XFS_AG_CHECK_DADDR(mp, XFS_FSB_TO_DADDR(mp, args->fsbno), args->len);
|
2023-02-13 06:14:54 +08:00
|
|
|
|
|
|
|
/* if not file data, insert new block into the reverse map btree */
|
|
|
|
if (!xfs_rmap_should_skip_owner_update(&args->oinfo)) {
|
|
|
|
error = xfs_rmap_alloc(args->tp, args->agbp, args->pag,
|
|
|
|
args->agbno, args->len, &args->oinfo);
|
|
|
|
if (error)
|
|
|
|
goto out_drop_perag;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!args->wasfromfl) {
|
|
|
|
error = xfs_alloc_update_counters(args->tp, args->agbp,
|
|
|
|
-((long)(args->len)));
|
|
|
|
if (error)
|
|
|
|
goto out_drop_perag;
|
|
|
|
|
|
|
|
ASSERT(!xfs_extent_busy_search(mp, args->pag, args->agbno,
|
|
|
|
args->len));
|
|
|
|
}
|
|
|
|
|
|
|
|
xfs_ag_resv_alloc_extent(args->pag, args->resv, args);
|
|
|
|
|
|
|
|
XFS_STATS_INC(mp, xs_allocx);
|
|
|
|
XFS_STATS_ADD(mp, xs_allocb, args->len);
|
|
|
|
|
2023-03-16 08:30:33 +08:00
|
|
|
trace_xfs_alloc_vextent_finish(args);
|
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
out_drop_perag:
|
2023-02-13 06:14:54 +08:00
|
|
|
if (drop_perag && args->pag) {
|
|
|
|
xfs_perag_rele(args->pag);
|
2023-02-13 06:14:54 +08:00
|
|
|
args->pag = NULL;
|
|
|
|
}
|
|
|
|
return error;
|
2023-02-13 06:14:53 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2023-02-13 06:14:54 +08:00
|
|
|
* Allocate within a single AG only. This uses a best-fit length algorithm so if
|
|
|
|
* you need an exact sized allocation without locality constraints, this is the
|
|
|
|
* fastest way to do it.
|
|
|
|
*
|
|
|
|
* Caller is expected to hold a perag reference in args->pag.
|
2023-02-13 06:14:53 +08:00
|
|
|
*/
|
2023-02-13 06:14:53 +08:00
|
|
|
int
|
2023-02-13 06:14:53 +08:00
|
|
|
xfs_alloc_vextent_this_ag(
|
2023-02-13 06:14:54 +08:00
|
|
|
struct xfs_alloc_arg *args,
|
|
|
|
xfs_agnumber_t agno)
|
2023-02-13 06:14:53 +08:00
|
|
|
{
|
|
|
|
struct xfs_mount *mp = args->mp;
|
2023-02-13 06:14:54 +08:00
|
|
|
xfs_agnumber_t minimum_agno;
|
2023-06-29 02:04:32 +08:00
|
|
|
uint32_t alloc_flags = 0;
|
2023-02-13 06:14:53 +08:00
|
|
|
int error;
|
|
|
|
|
2023-03-11 05:42:08 +08:00
|
|
|
ASSERT(args->pag != NULL);
|
|
|
|
ASSERT(args->pag->pag_agno == agno);
|
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
args->agno = agno;
|
|
|
|
args->agbno = 0;
|
2023-03-16 08:30:33 +08:00
|
|
|
|
|
|
|
trace_xfs_alloc_vextent_this_ag(args);
|
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
error = xfs_alloc_vextent_check_args(args, XFS_AGB_TO_FSB(mp, agno, 0),
|
|
|
|
&minimum_agno);
|
2023-02-13 06:14:53 +08:00
|
|
|
if (error) {
|
|
|
|
if (error == -ENOSPC)
|
|
|
|
return 0;
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2023-06-29 02:04:32 +08:00
|
|
|
error = xfs_alloc_vextent_prepare_ag(args, alloc_flags);
|
2023-02-13 06:14:54 +08:00
|
|
|
if (!error && args->agbp)
|
2023-06-29 02:04:32 +08:00
|
|
|
error = xfs_alloc_ag_vextent_size(args, alloc_flags);
|
2023-02-13 06:14:54 +08:00
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
return xfs_alloc_vextent_finish(args, minimum_agno, error, false);
|
2023-02-13 06:14:53 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Iterate all AGs trying to allocate an extent starting from @start_ag.
|
|
|
|
*
|
|
|
|
* If the incoming allocation type is XFS_ALLOCTYPE_NEAR_BNO, it means the
|
|
|
|
* allocation attempts in @start_agno have locality information. If we fail to
|
|
|
|
* allocate in that AG, then we revert to anywhere-in-AG for all the other AGs
|
|
|
|
* we attempt to allocation in as there is no locality optimisation possible for
|
|
|
|
* those allocations.
|
|
|
|
*
|
2023-02-13 06:14:54 +08:00
|
|
|
* On return, args->pag may be left referenced if we finish before the "all
|
|
|
|
* failed" return point. The allocation finish still needs the perag, and
|
|
|
|
* so the caller will release it once they've finished the allocation.
|
|
|
|
*
|
2023-02-13 06:14:53 +08:00
|
|
|
* When we wrap the AG iteration at the end of the filesystem, we have to be
|
|
|
|
* careful not to wrap into AGs below ones we already have locked in the
|
|
|
|
* transaction if we are doing a blocking iteration. This will result in an
|
|
|
|
* out-of-order locking of AGFs and hence can cause deadlocks.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
xfs_alloc_vextent_iterate_ags(
|
|
|
|
struct xfs_alloc_arg *args,
|
|
|
|
xfs_agnumber_t minimum_agno,
|
|
|
|
xfs_agnumber_t start_agno,
|
2023-02-13 06:14:54 +08:00
|
|
|
xfs_agblock_t target_agbno,
|
2023-06-29 02:04:32 +08:00
|
|
|
uint32_t alloc_flags)
|
2023-02-13 06:14:53 +08:00
|
|
|
{
|
|
|
|
struct xfs_mount *mp = args->mp;
|
2023-03-16 08:30:33 +08:00
|
|
|
xfs_agnumber_t restart_agno = minimum_agno;
|
2023-02-13 06:14:54 +08:00
|
|
|
xfs_agnumber_t agno;
|
2023-02-13 06:14:53 +08:00
|
|
|
int error = 0;
|
2023-02-11 01:07:06 +08:00
|
|
|
|
2023-06-29 02:04:32 +08:00
|
|
|
if (alloc_flags & XFS_ALLOC_FLAG_TRYLOCK)
|
2023-03-16 08:30:33 +08:00
|
|
|
restart_agno = 0;
|
2023-02-13 06:14:54 +08:00
|
|
|
restart:
|
2023-03-16 08:30:33 +08:00
|
|
|
for_each_perag_wrap_range(mp, start_agno, restart_agno,
|
2023-02-13 06:14:54 +08:00
|
|
|
mp->m_sb.sb_agcount, agno, args->pag) {
|
|
|
|
args->agno = agno;
|
2023-06-29 02:04:32 +08:00
|
|
|
error = xfs_alloc_vextent_prepare_ag(args, alloc_flags);
|
2023-02-13 06:14:54 +08:00
|
|
|
if (error)
|
2005-04-17 06:20:36 +08:00
|
|
|
break;
|
2023-02-13 06:14:54 +08:00
|
|
|
if (!args->agbp) {
|
|
|
|
trace_xfs_alloc_vextent_loopfailed(args);
|
|
|
|
continue;
|
2023-02-13 06:14:54 +08:00
|
|
|
}
|
2023-02-13 06:14:53 +08:00
|
|
|
|
|
|
|
/*
|
2023-02-13 06:14:54 +08:00
|
|
|
* Allocation is supposed to succeed now, so break out of the
|
|
|
|
* loop regardless of whether we succeed or not.
|
2023-02-13 06:14:53 +08:00
|
|
|
*/
|
2023-02-13 06:14:54 +08:00
|
|
|
if (args->agno == start_agno && target_agbno) {
|
2023-02-13 06:14:54 +08:00
|
|
|
args->agbno = target_agbno;
|
2023-06-29 02:04:32 +08:00
|
|
|
error = xfs_alloc_ag_vextent_near(args, alloc_flags);
|
2023-02-13 06:14:54 +08:00
|
|
|
} else {
|
|
|
|
args->agbno = 0;
|
2023-06-29 02:04:32 +08:00
|
|
|
error = xfs_alloc_ag_vextent_size(args, alloc_flags);
|
2023-02-13 06:14:53 +08:00
|
|
|
}
|
2023-02-13 06:14:54 +08:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (error) {
|
|
|
|
xfs_perag_rele(args->pag);
|
2023-02-13 06:14:53 +08:00
|
|
|
args->pag = NULL;
|
2023-02-13 06:14:54 +08:00
|
|
|
return error;
|
2023-02-13 06:14:53 +08:00
|
|
|
}
|
2023-02-13 06:14:54 +08:00
|
|
|
if (args->agbp)
|
|
|
|
return 0;
|
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
/*
|
2023-02-13 06:14:54 +08:00
|
|
|
* We didn't find an AG we can alloation from. If we were given
|
|
|
|
* constraining flags by the caller, drop them and retry the allocation
|
|
|
|
* without any constraints being set.
|
2023-02-13 06:14:54 +08:00
|
|
|
*/
|
2023-06-29 02:04:32 +08:00
|
|
|
if (alloc_flags & XFS_ALLOC_FLAG_TRYLOCK) {
|
|
|
|
alloc_flags &= ~XFS_ALLOC_FLAG_TRYLOCK;
|
2023-03-16 08:30:33 +08:00
|
|
|
restart_agno = minimum_agno;
|
2023-02-13 06:14:54 +08:00
|
|
|
goto restart;
|
|
|
|
}
|
|
|
|
|
|
|
|
ASSERT(args->pag == NULL);
|
|
|
|
trace_xfs_alloc_vextent_allfailed(args);
|
|
|
|
return 0;
|
2023-02-13 06:14:53 +08:00
|
|
|
}
|
2009-12-15 07:14:59 +08:00
|
|
|
|
2023-02-13 06:14:53 +08:00
|
|
|
/*
|
|
|
|
* Iterate from the AGs from the start AG to the end of the filesystem, trying
|
|
|
|
* to allocate blocks. It starts with a near allocation attempt in the initial
|
|
|
|
* AG, then falls back to anywhere-in-ag after the first AG fails. It will wrap
|
|
|
|
* back to zero if allowed by previous allocations in this transaction,
|
|
|
|
* otherwise will wrap back to the start AG and run a second blocking pass to
|
|
|
|
* the end of the filesystem.
|
|
|
|
*/
|
2023-02-13 06:14:53 +08:00
|
|
|
int
|
2023-02-13 06:14:53 +08:00
|
|
|
xfs_alloc_vextent_start_ag(
|
|
|
|
struct xfs_alloc_arg *args,
|
2023-02-13 06:14:53 +08:00
|
|
|
xfs_fsblock_t target)
|
2023-02-13 06:14:53 +08:00
|
|
|
{
|
|
|
|
struct xfs_mount *mp = args->mp;
|
2023-02-13 06:14:54 +08:00
|
|
|
xfs_agnumber_t minimum_agno;
|
2023-02-13 06:14:53 +08:00
|
|
|
xfs_agnumber_t start_agno;
|
|
|
|
xfs_agnumber_t rotorstep = xfs_rotorstep;
|
|
|
|
bool bump_rotor = false;
|
2023-06-29 02:04:32 +08:00
|
|
|
uint32_t alloc_flags = XFS_ALLOC_FLAG_TRYLOCK;
|
2023-02-13 06:14:53 +08:00
|
|
|
int error;
|
2023-02-11 01:07:06 +08:00
|
|
|
|
2023-03-11 05:42:08 +08:00
|
|
|
ASSERT(args->pag == NULL);
|
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
args->agno = NULLAGNUMBER;
|
|
|
|
args->agbno = NULLAGBLOCK;
|
2023-03-16 08:30:33 +08:00
|
|
|
|
2023-03-25 04:14:48 +08:00
|
|
|
trace_xfs_alloc_vextent_start_ag(args);
|
2023-03-16 08:30:33 +08:00
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
error = xfs_alloc_vextent_check_args(args, target, &minimum_agno);
|
2023-02-13 06:14:53 +08:00
|
|
|
if (error) {
|
|
|
|
if (error == -ENOSPC)
|
|
|
|
return 0;
|
|
|
|
return error;
|
|
|
|
}
|
2023-02-11 01:07:06 +08:00
|
|
|
|
2023-02-13 06:14:53 +08:00
|
|
|
if ((args->datatype & XFS_ALLOC_INITIAL_USER_DATA) &&
|
|
|
|
xfs_is_inode32(mp)) {
|
2023-02-13 06:14:53 +08:00
|
|
|
target = XFS_AGB_TO_FSB(mp,
|
2023-02-13 06:14:53 +08:00
|
|
|
((mp->m_agfrotor / rotorstep) %
|
|
|
|
mp->m_sb.sb_agcount), 0);
|
|
|
|
bump_rotor = 1;
|
|
|
|
}
|
2017-01-10 05:36:19 +08:00
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
start_agno = max(minimum_agno, XFS_FSB_TO_AGNO(mp, target));
|
2023-02-13 06:14:53 +08:00
|
|
|
error = xfs_alloc_vextent_iterate_ags(args, minimum_agno, start_agno,
|
2023-06-29 02:04:32 +08:00
|
|
|
XFS_FSB_TO_AGBNO(mp, target), alloc_flags);
|
2023-02-13 06:14:54 +08:00
|
|
|
|
2023-02-13 06:14:53 +08:00
|
|
|
if (bump_rotor) {
|
|
|
|
if (args->agno == start_agno)
|
|
|
|
mp->m_agfrotor = (mp->m_agfrotor + 1) %
|
|
|
|
(mp->m_sb.sb_agcount * rotorstep);
|
|
|
|
else
|
|
|
|
mp->m_agfrotor = (args->agno * rotorstep + 1) %
|
|
|
|
(mp->m_sb.sb_agcount * rotorstep);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2015-11-03 09:27:22 +08:00
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
return xfs_alloc_vextent_finish(args, minimum_agno, error, true);
|
2023-02-13 06:14:53 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2023-02-13 06:14:54 +08:00
|
|
|
* Iterate from the agno indicated via @target through to the end of the
|
2023-02-13 06:14:53 +08:00
|
|
|
* filesystem attempting blocking allocation. This does not wrap or try a second
|
2023-02-13 06:14:54 +08:00
|
|
|
* pass, so will not recurse into AGs lower than indicated by the target.
|
2023-02-13 06:14:53 +08:00
|
|
|
*/
|
2023-02-13 06:14:53 +08:00
|
|
|
int
|
2023-02-13 06:14:53 +08:00
|
|
|
xfs_alloc_vextent_first_ag(
|
|
|
|
struct xfs_alloc_arg *args,
|
2023-02-13 06:14:53 +08:00
|
|
|
xfs_fsblock_t target)
|
|
|
|
{
|
2023-02-13 06:14:53 +08:00
|
|
|
struct xfs_mount *mp = args->mp;
|
2023-02-13 06:14:54 +08:00
|
|
|
xfs_agnumber_t minimum_agno;
|
2023-02-13 06:14:53 +08:00
|
|
|
xfs_agnumber_t start_agno;
|
2023-06-29 02:04:32 +08:00
|
|
|
uint32_t alloc_flags = XFS_ALLOC_FLAG_TRYLOCK;
|
2023-02-13 06:14:53 +08:00
|
|
|
int error;
|
|
|
|
|
2023-03-11 05:42:08 +08:00
|
|
|
ASSERT(args->pag == NULL);
|
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
args->agno = NULLAGNUMBER;
|
|
|
|
args->agbno = NULLAGBLOCK;
|
2023-03-16 08:30:33 +08:00
|
|
|
|
2023-03-25 04:14:48 +08:00
|
|
|
trace_xfs_alloc_vextent_first_ag(args);
|
2023-03-16 08:30:33 +08:00
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
error = xfs_alloc_vextent_check_args(args, target, &minimum_agno);
|
2023-02-13 06:14:53 +08:00
|
|
|
if (error) {
|
|
|
|
if (error == -ENOSPC)
|
|
|
|
return 0;
|
|
|
|
return error;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2023-02-11 01:07:06 +08:00
|
|
|
|
2023-02-13 06:14:53 +08:00
|
|
|
start_agno = max(minimum_agno, XFS_FSB_TO_AGNO(mp, target));
|
2023-02-13 06:14:54 +08:00
|
|
|
error = xfs_alloc_vextent_iterate_ags(args, minimum_agno, start_agno,
|
2023-06-29 02:04:32 +08:00
|
|
|
XFS_FSB_TO_AGBNO(mp, target), alloc_flags);
|
2023-02-13 06:14:54 +08:00
|
|
|
return xfs_alloc_vextent_finish(args, minimum_agno, error, true);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
/*
|
2023-02-13 06:14:54 +08:00
|
|
|
* Allocate at the exact block target or fail. Caller is expected to hold a
|
|
|
|
* perag reference in args->pag.
|
2023-02-13 06:14:54 +08:00
|
|
|
*/
|
|
|
|
int
|
|
|
|
xfs_alloc_vextent_exact_bno(
|
|
|
|
struct xfs_alloc_arg *args,
|
|
|
|
xfs_fsblock_t target)
|
|
|
|
{
|
|
|
|
struct xfs_mount *mp = args->mp;
|
2023-02-13 06:14:54 +08:00
|
|
|
xfs_agnumber_t minimum_agno;
|
2023-02-13 06:14:54 +08:00
|
|
|
int error;
|
|
|
|
|
2023-03-11 05:42:08 +08:00
|
|
|
ASSERT(args->pag != NULL);
|
|
|
|
ASSERT(args->pag->pag_agno == XFS_FSB_TO_AGNO(mp, target));
|
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
args->agno = XFS_FSB_TO_AGNO(mp, target);
|
|
|
|
args->agbno = XFS_FSB_TO_AGBNO(mp, target);
|
2023-03-16 08:30:33 +08:00
|
|
|
|
2023-03-25 04:14:48 +08:00
|
|
|
trace_xfs_alloc_vextent_exact_bno(args);
|
2023-03-16 08:30:33 +08:00
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
error = xfs_alloc_vextent_check_args(args, target, &minimum_agno);
|
2023-02-13 06:14:54 +08:00
|
|
|
if (error) {
|
|
|
|
if (error == -ENOSPC)
|
|
|
|
return 0;
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2023-06-05 02:06:27 +08:00
|
|
|
error = xfs_alloc_vextent_prepare_ag(args, 0);
|
2023-02-13 06:14:54 +08:00
|
|
|
if (!error && args->agbp)
|
2023-02-13 06:14:54 +08:00
|
|
|
error = xfs_alloc_ag_vextent_exact(args);
|
2023-02-13 06:14:54 +08:00
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
return xfs_alloc_vextent_finish(args, minimum_agno, error, false);
|
2023-02-13 06:14:54 +08:00
|
|
|
}
|
|
|
|
|
2023-02-13 06:14:53 +08:00
|
|
|
/*
|
2023-02-13 06:14:54 +08:00
|
|
|
* Allocate an extent as close to the target as possible. If there are not
|
|
|
|
* viable candidates in the AG, then fail the allocation.
|
2023-02-13 06:14:54 +08:00
|
|
|
*
|
|
|
|
* Caller may or may not have a per-ag reference in args->pag.
|
2023-02-13 06:14:53 +08:00
|
|
|
*/
|
|
|
|
int
|
2023-02-13 06:14:54 +08:00
|
|
|
xfs_alloc_vextent_near_bno(
|
|
|
|
struct xfs_alloc_arg *args,
|
|
|
|
xfs_fsblock_t target)
|
2023-02-13 06:14:53 +08:00
|
|
|
{
|
2023-02-13 06:14:54 +08:00
|
|
|
struct xfs_mount *mp = args->mp;
|
2023-02-13 06:14:54 +08:00
|
|
|
xfs_agnumber_t minimum_agno;
|
2023-02-13 06:14:54 +08:00
|
|
|
bool needs_perag = args->pag == NULL;
|
2023-06-29 02:04:32 +08:00
|
|
|
uint32_t alloc_flags = 0;
|
2023-02-13 06:14:53 +08:00
|
|
|
int error;
|
2023-02-13 06:14:53 +08:00
|
|
|
|
2023-03-11 05:42:08 +08:00
|
|
|
if (!needs_perag)
|
|
|
|
ASSERT(args->pag->pag_agno == XFS_FSB_TO_AGNO(mp, target));
|
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
args->agno = XFS_FSB_TO_AGNO(mp, target);
|
|
|
|
args->agbno = XFS_FSB_TO_AGBNO(mp, target);
|
2023-03-16 08:30:33 +08:00
|
|
|
|
2023-03-25 04:14:48 +08:00
|
|
|
trace_xfs_alloc_vextent_near_bno(args);
|
2023-03-16 08:30:33 +08:00
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
error = xfs_alloc_vextent_check_args(args, target, &minimum_agno);
|
2023-02-13 06:14:54 +08:00
|
|
|
if (error) {
|
|
|
|
if (error == -ENOSPC)
|
|
|
|
return 0;
|
|
|
|
return error;
|
2023-02-13 06:14:53 +08:00
|
|
|
}
|
2023-02-13 06:14:54 +08:00
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
if (needs_perag)
|
2023-02-13 06:14:54 +08:00
|
|
|
args->pag = xfs_perag_grab(mp, args->agno);
|
2023-02-13 06:14:54 +08:00
|
|
|
|
2023-06-29 02:04:32 +08:00
|
|
|
error = xfs_alloc_vextent_prepare_ag(args, alloc_flags);
|
2023-02-13 06:14:54 +08:00
|
|
|
if (!error && args->agbp)
|
2023-06-29 02:04:32 +08:00
|
|
|
error = xfs_alloc_ag_vextent_near(args, alloc_flags);
|
2023-02-13 06:14:54 +08:00
|
|
|
|
2023-02-13 06:14:54 +08:00
|
|
|
return xfs_alloc_vextent_finish(args, minimum_agno, error, needs_perag);
|
2023-02-13 06:14:53 +08:00
|
|
|
}
|
|
|
|
|
2016-06-21 09:53:28 +08:00
|
|
|
/* Ensure that the freelist is at full capacity. */
|
|
|
|
int
|
|
|
|
xfs_free_extent_fix_freelist(
|
|
|
|
struct xfs_trans *tp,
|
2021-06-02 08:48:24 +08:00
|
|
|
struct xfs_perag *pag,
|
2016-06-21 09:53:28 +08:00
|
|
|
struct xfs_buf **agbp)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2016-06-21 09:53:28 +08:00
|
|
|
struct xfs_alloc_arg args;
|
|
|
|
int error;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2016-06-21 09:53:28 +08:00
|
|
|
memset(&args, 0, sizeof(struct xfs_alloc_arg));
|
2005-04-17 06:20:36 +08:00
|
|
|
args.tp = tp;
|
|
|
|
args.mp = tp->t_mountp;
|
2021-06-02 08:48:24 +08:00
|
|
|
args.agno = pag->pag_agno;
|
|
|
|
args.pag = pag;
|
2011-04-08 10:45:07 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* validate that the block number is legal - the enables us to detect
|
|
|
|
* and handle a silent filesystem corruption rather than crashing.
|
|
|
|
*/
|
|
|
|
if (args.agno >= args.mp->m_sb.sb_agcount)
|
2014-06-25 12:58:08 +08:00
|
|
|
return -EFSCORRUPTED;
|
2011-04-08 10:45:07 +08:00
|
|
|
|
|
|
|
error = xfs_alloc_fix_freelist(&args, XFS_ALLOC_FLAG_FREEING);
|
|
|
|
if (error)
|
2021-06-02 08:48:24 +08:00
|
|
|
return error;
|
2016-06-21 09:53:28 +08:00
|
|
|
|
|
|
|
*agbp = args.agbp;
|
2021-06-02 08:48:24 +08:00
|
|
|
return 0;
|
2016-06-21 09:53:28 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Free an extent.
|
|
|
|
* Just break up the extent address and hand off to xfs_free_ag_extent
|
|
|
|
* after fixing up the freelist.
|
|
|
|
*/
|
2018-12-13 00:46:23 +08:00
|
|
|
int
|
2018-05-09 23:45:04 +08:00
|
|
|
__xfs_free_extent(
|
2018-12-13 00:46:23 +08:00
|
|
|
struct xfs_trans *tp,
|
2023-04-12 09:59:53 +08:00
|
|
|
struct xfs_perag *pag,
|
|
|
|
xfs_agblock_t agbno,
|
2018-12-13 00:46:23 +08:00
|
|
|
xfs_extlen_t len,
|
|
|
|
const struct xfs_owner_info *oinfo,
|
|
|
|
enum xfs_ag_resv_type type,
|
|
|
|
bool skip_discard)
|
2016-06-21 09:53:28 +08:00
|
|
|
{
|
2018-12-13 00:46:23 +08:00
|
|
|
struct xfs_mount *mp = tp->t_mountp;
|
|
|
|
struct xfs_buf *agbp;
|
2020-03-10 23:57:29 +08:00
|
|
|
struct xfs_agf *agf;
|
2018-12-13 00:46:23 +08:00
|
|
|
int error;
|
|
|
|
unsigned int busy_flags = 0;
|
2016-06-21 09:53:28 +08:00
|
|
|
|
|
|
|
ASSERT(len != 0);
|
2018-03-10 06:02:32 +08:00
|
|
|
ASSERT(type != XFS_AG_RESV_AGFL);
|
2016-06-21 09:53:28 +08:00
|
|
|
|
2016-08-03 09:26:33 +08:00
|
|
|
if (XFS_TEST_ERROR(false, mp,
|
2017-06-21 08:54:47 +08:00
|
|
|
XFS_ERRTAG_FREE_EXTENT))
|
2016-08-03 09:26:33 +08:00
|
|
|
return -EIO;
|
|
|
|
|
2021-06-02 08:48:24 +08:00
|
|
|
error = xfs_free_extent_fix_freelist(tp, pag, &agbp);
|
2016-06-21 09:53:28 +08:00
|
|
|
if (error)
|
2023-04-12 09:59:53 +08:00
|
|
|
return error;
|
2020-03-10 23:57:29 +08:00
|
|
|
agf = agbp->b_addr;
|
2016-06-21 09:53:28 +08:00
|
|
|
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, agbno >= mp->m_sb.sb_agblocks)) {
|
|
|
|
error = -EFSCORRUPTED;
|
2021-06-02 08:48:24 +08:00
|
|
|
goto err_release;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
}
|
2011-04-08 10:45:07 +08:00
|
|
|
|
|
|
|
/* validate the extent size is legal now we have the agf locked */
|
2020-03-10 23:57:29 +08:00
|
|
|
if (XFS_IS_CORRUPT(mp, agbno + len > be32_to_cpu(agf->agf_length))) {
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
error = -EFSCORRUPTED;
|
2021-06-02 08:48:24 +08:00
|
|
|
goto err_release;
|
xfs: kill the XFS_WANT_CORRUPT_* macros
The XFS_WANT_CORRUPT_* macros conceal subtle side effects such as the
creation of local variables and redirections of the code flow. This is
pretty ugly, so replace them with explicit XFS_IS_CORRUPT tests that
remove both of those ugly points. The change was performed with the
following coccinelle script:
@@
expression mp, test;
identifier label;
@@
- XFS_WANT_CORRUPTED_GOTO(mp, test, label);
+ if (XFS_IS_CORRUPT(mp, !test)) { error = -EFSCORRUPTED; goto label; }
@@
expression mp, test;
@@
- XFS_WANT_CORRUPTED_RETURN(mp, test);
+ if (XFS_IS_CORRUPT(mp, !test)) return -EFSCORRUPTED;
@@
expression mp, lval, rval;
@@
- XFS_IS_CORRUPT(mp, !(lval == rval))
+ XFS_IS_CORRUPT(mp, lval != rval)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 && e2))
+ XFS_IS_CORRUPT(mp, !e1 || !e2)
@@
expression e1, e2;
@@
- !(e1 == e2)
+ e1 != e2
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 && e3 == e4) || e5 != e6
+ e1 != e2 || e3 != e4 || e5 != e6
@@
expression e1, e2, e3, e4, e5, e6;
@@
- !(e1 == e2 || (e3 <= e4 && e5 <= e6))
+ e1 != e2 && (e3 > e4 || e5 > e6)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2))
+ XFS_IS_CORRUPT(mp, e1 > e2)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 < e2))
+ XFS_IS_CORRUPT(mp, e1 >= e2)
@@
expression mp, e1;
@@
- XFS_IS_CORRUPT(mp, !!e1)
+ XFS_IS_CORRUPT(mp, e1)
@@
expression mp, e1, e2;
@@
- XFS_IS_CORRUPT(mp, !(e1 || e2))
+ XFS_IS_CORRUPT(mp, !e1 && !e2)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 == e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 != e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 <= e2) || !(e3 >= e4))
+ XFS_IS_CORRUPT(mp, e1 > e2 || e3 < e4)
@@
expression mp, e1, e2, e3, e4;
@@
- XFS_IS_CORRUPT(mp, !(e1 == e2) && !(e3 <= e4))
+ XFS_IS_CORRUPT(mp, e1 != e2 && e3 > e4)
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2019-11-12 04:52:18 +08:00
|
|
|
}
|
2011-04-08 10:45:07 +08:00
|
|
|
|
2023-04-12 09:59:53 +08:00
|
|
|
error = xfs_free_ag_extent(tp, agbp, pag->pag_agno, agbno, len, oinfo,
|
|
|
|
type);
|
2016-06-21 09:53:28 +08:00
|
|
|
if (error)
|
2021-06-02 08:48:24 +08:00
|
|
|
goto err_release;
|
2016-06-21 09:53:28 +08:00
|
|
|
|
2018-05-09 23:45:04 +08:00
|
|
|
if (skip_discard)
|
|
|
|
busy_flags |= XFS_EXTENT_BUSY_SKIP_DISCARD;
|
2021-06-02 08:48:24 +08:00
|
|
|
xfs_extent_busy_insert(tp, pag, agbno, len, busy_flags);
|
2016-06-21 09:53:28 +08:00
|
|
|
return 0;
|
|
|
|
|
2021-06-02 08:48:24 +08:00
|
|
|
err_release:
|
2016-06-21 09:53:28 +08:00
|
|
|
xfs_trans_brelse(tp, agbp);
|
2005-04-17 06:20:36 +08:00
|
|
|
return error;
|
|
|
|
}
|
2017-03-29 05:56:35 +08:00
|
|
|
|
|
|
|
struct xfs_alloc_query_range_info {
|
|
|
|
xfs_alloc_query_range_fn fn;
|
|
|
|
void *priv;
|
|
|
|
};
|
|
|
|
|
|
|
|
/* Format btree record and pass to our callback. */
|
|
|
|
STATIC int
|
|
|
|
xfs_alloc_query_range_helper(
|
|
|
|
struct xfs_btree_cur *cur,
|
2021-08-11 08:02:16 +08:00
|
|
|
const union xfs_btree_rec *rec,
|
2017-03-29 05:56:35 +08:00
|
|
|
void *priv)
|
|
|
|
{
|
|
|
|
struct xfs_alloc_query_range_info *query = priv;
|
|
|
|
struct xfs_alloc_rec_incore irec;
|
2023-04-12 10:00:04 +08:00
|
|
|
xfs_failaddr_t fa;
|
2017-03-29 05:56:35 +08:00
|
|
|
|
2023-04-12 10:00:01 +08:00
|
|
|
xfs_alloc_btrec_to_irec(rec, &irec);
|
2023-04-12 10:00:04 +08:00
|
|
|
fa = xfs_alloc_check_irec(cur, &irec);
|
|
|
|
if (fa)
|
|
|
|
return xfs_alloc_complain_bad_rec(cur, fa, &irec);
|
2023-04-12 10:00:01 +08:00
|
|
|
|
2017-03-29 05:56:35 +08:00
|
|
|
return query->fn(cur, &irec, query->priv);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Find all free space within a given range of blocks. */
|
|
|
|
int
|
|
|
|
xfs_alloc_query_range(
|
|
|
|
struct xfs_btree_cur *cur,
|
2021-08-11 08:02:15 +08:00
|
|
|
const struct xfs_alloc_rec_incore *low_rec,
|
|
|
|
const struct xfs_alloc_rec_incore *high_rec,
|
2017-03-29 05:56:35 +08:00
|
|
|
xfs_alloc_query_range_fn fn,
|
|
|
|
void *priv)
|
|
|
|
{
|
2023-06-30 08:39:46 +08:00
|
|
|
union xfs_btree_irec low_brec = { .a = *low_rec };
|
|
|
|
union xfs_btree_irec high_brec = { .a = *high_rec };
|
|
|
|
struct xfs_alloc_query_range_info query = { .priv = priv, .fn = fn };
|
2017-03-29 05:56:35 +08:00
|
|
|
|
|
|
|
ASSERT(cur->bc_btnum == XFS_BTNUM_BNO);
|
|
|
|
return xfs_btree_query_range(cur, &low_brec, &high_brec,
|
|
|
|
xfs_alloc_query_range_helper, &query);
|
|
|
|
}
|
2017-03-29 05:56:35 +08:00
|
|
|
|
|
|
|
/* Find all free space records. */
|
|
|
|
int
|
|
|
|
xfs_alloc_query_all(
|
|
|
|
struct xfs_btree_cur *cur,
|
|
|
|
xfs_alloc_query_range_fn fn,
|
|
|
|
void *priv)
|
|
|
|
{
|
|
|
|
struct xfs_alloc_query_range_info query;
|
|
|
|
|
|
|
|
ASSERT(cur->bc_btnum == XFS_BTNUM_BNO);
|
|
|
|
query.priv = priv;
|
|
|
|
query.fn = fn;
|
|
|
|
return xfs_btree_query_all(cur, xfs_alloc_query_range_helper, &query);
|
|
|
|
}
|
2017-10-18 12:37:32 +08:00
|
|
|
|
2023-04-12 10:00:10 +08:00
|
|
|
/*
|
|
|
|
* Scan part of the keyspace of the free space and tell us if the area has no
|
|
|
|
* records, is fully mapped by records, or is partially filled.
|
|
|
|
*/
|
2018-01-17 10:52:12 +08:00
|
|
|
int
|
2023-04-12 10:00:10 +08:00
|
|
|
xfs_alloc_has_records(
|
2018-01-17 10:52:12 +08:00
|
|
|
struct xfs_btree_cur *cur,
|
|
|
|
xfs_agblock_t bno,
|
|
|
|
xfs_extlen_t len,
|
2023-04-12 10:00:10 +08:00
|
|
|
enum xbtree_recpacking *outcome)
|
2018-01-17 10:52:12 +08:00
|
|
|
{
|
|
|
|
union xfs_btree_irec low;
|
|
|
|
union xfs_btree_irec high;
|
|
|
|
|
|
|
|
memset(&low, 0, sizeof(low));
|
|
|
|
low.a.ar_startblock = bno;
|
|
|
|
memset(&high, 0xFF, sizeof(high));
|
|
|
|
high.a.ar_startblock = bno + len - 1;
|
|
|
|
|
2023-04-12 10:00:11 +08:00
|
|
|
return xfs_btree_has_records(cur, &low, &high, NULL, outcome);
|
2018-01-17 10:52:12 +08:00
|
|
|
}
|
2018-05-14 21:34:34 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Walk all the blocks in the AGFL. The @walk_fn can return any negative
|
2019-07-03 00:39:38 +08:00
|
|
|
* error code or XFS_ITER_*.
|
2018-05-14 21:34:34 +08:00
|
|
|
*/
|
|
|
|
int
|
|
|
|
xfs_agfl_walk(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
struct xfs_agf *agf,
|
|
|
|
struct xfs_buf *agflbp,
|
|
|
|
xfs_agfl_walk_fn walk_fn,
|
|
|
|
void *priv)
|
|
|
|
{
|
|
|
|
__be32 *agfl_bno;
|
|
|
|
unsigned int i;
|
|
|
|
int error;
|
|
|
|
|
2020-03-10 23:57:28 +08:00
|
|
|
agfl_bno = xfs_buf_to_agfl_bno(agflbp);
|
2018-05-14 21:34:34 +08:00
|
|
|
i = be32_to_cpu(agf->agf_flfirst);
|
|
|
|
|
|
|
|
/* Nothing to walk in an empty AGFL. */
|
|
|
|
if (agf->agf_flcount == cpu_to_be32(0))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* Otherwise, walk from first to last, wrapping as needed. */
|
|
|
|
for (;;) {
|
|
|
|
error = walk_fn(mp, be32_to_cpu(agfl_bno[i]), priv);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
if (i == be32_to_cpu(agf->agf_fllast))
|
|
|
|
break;
|
|
|
|
if (++i == xfs_agfl_size(mp))
|
|
|
|
i = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
2021-10-13 05:17:01 +08:00
|
|
|
|
|
|
|
int __init
|
|
|
|
xfs_extfree_intent_init_cache(void)
|
|
|
|
{
|
|
|
|
xfs_extfree_item_cache = kmem_cache_create("xfs_extfree_intent",
|
|
|
|
sizeof(struct xfs_extent_free_item),
|
|
|
|
0, 0, NULL);
|
|
|
|
|
|
|
|
return xfs_extfree_item_cache != NULL ? 0 : -ENOMEM;
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
xfs_extfree_intent_destroy_cache(void)
|
|
|
|
{
|
|
|
|
kmem_cache_destroy(xfs_extfree_item_cache);
|
|
|
|
xfs_extfree_item_cache = NULL;
|
|
|
|
}
|