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) 2000-2001,2005 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 07:51:50 +08:00
|
|
|
#include "xfs_format.h"
|
2013-10-23 07:50:10 +08:00
|
|
|
#include "xfs_log_format.h"
|
|
|
|
#include "xfs_trans_resv.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"
|
2013-10-15 06:17:51 +08:00
|
|
|
#include "xfs_da_format.h"
|
2005-11-02 11:38:42 +08:00
|
|
|
#include "xfs_da_btree.h"
|
2005-04-17 06:20:36 +08:00
|
|
|
#include "xfs_inode.h"
|
2013-10-23 07:50:10 +08:00
|
|
|
#include "xfs_trans.h"
|
2005-11-02 11:38:42 +08:00
|
|
|
#include "xfs_inode_item.h"
|
2005-04-17 06:20:36 +08:00
|
|
|
#include "xfs_bmap.h"
|
2013-08-12 18:49:37 +08:00
|
|
|
#include "xfs_dir2.h"
|
2011-07-13 19:43:48 +08:00
|
|
|
#include "xfs_dir2_priv.h"
|
2017-10-18 12:37:34 +08:00
|
|
|
#include "xfs_ialloc.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"
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2013-08-12 18:50:09 +08:00
|
|
|
struct xfs_name xfs_name_dotdot = { (unsigned char *)"..", 2, XFS_DIR3_FT_DIR };
|
|
|
|
|
2014-12-04 06:43:17 +08:00
|
|
|
/*
|
2017-01-18 03:41:43 +08:00
|
|
|
* Convert inode mode to directory entry filetype
|
2014-12-04 06:43:17 +08:00
|
|
|
*/
|
2017-10-18 12:37:44 +08:00
|
|
|
unsigned char
|
|
|
|
xfs_mode_to_ftype(
|
|
|
|
int mode)
|
2017-01-18 03:41:43 +08:00
|
|
|
{
|
|
|
|
switch (mode & S_IFMT) {
|
|
|
|
case S_IFREG:
|
|
|
|
return XFS_DIR3_FT_REG_FILE;
|
|
|
|
case S_IFDIR:
|
|
|
|
return XFS_DIR3_FT_DIR;
|
|
|
|
case S_IFCHR:
|
|
|
|
return XFS_DIR3_FT_CHRDEV;
|
|
|
|
case S_IFBLK:
|
|
|
|
return XFS_DIR3_FT_BLKDEV;
|
|
|
|
case S_IFIFO:
|
|
|
|
return XFS_DIR3_FT_FIFO;
|
|
|
|
case S_IFSOCK:
|
|
|
|
return XFS_DIR3_FT_SOCK;
|
|
|
|
case S_IFLNK:
|
|
|
|
return XFS_DIR3_FT_SYMLINK;
|
|
|
|
default:
|
|
|
|
return XFS_DIR3_FT_UNKNOWN;
|
|
|
|
}
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2008-05-21 14:58:55 +08:00
|
|
|
/*
|
|
|
|
* ASCII case-insensitive (ie. A-Z) support for directories that was
|
|
|
|
* used in IRIX.
|
|
|
|
*/
|
|
|
|
STATIC xfs_dahash_t
|
|
|
|
xfs_ascii_ci_hashname(
|
|
|
|
struct xfs_name *name)
|
|
|
|
{
|
|
|
|
xfs_dahash_t hash;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0, hash = 0; i < name->len; i++)
|
|
|
|
hash = tolower(name->name[i]) ^ rol32(hash, 7);
|
|
|
|
|
|
|
|
return hash;
|
|
|
|
}
|
|
|
|
|
|
|
|
STATIC enum xfs_dacmp
|
|
|
|
xfs_ascii_ci_compname(
|
|
|
|
struct xfs_da_args *args,
|
2010-01-20 07:47:17 +08:00
|
|
|
const unsigned char *name,
|
|
|
|
int len)
|
2008-05-21 14:58:55 +08:00
|
|
|
{
|
|
|
|
enum xfs_dacmp result;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if (args->namelen != len)
|
|
|
|
return XFS_CMP_DIFFERENT;
|
|
|
|
|
|
|
|
result = XFS_CMP_EXACT;
|
|
|
|
for (i = 0; i < len; i++) {
|
|
|
|
if (args->name[i] == name[i])
|
|
|
|
continue;
|
|
|
|
if (tolower(args->name[i]) != tolower(name[i]))
|
|
|
|
return XFS_CMP_DIFFERENT;
|
|
|
|
result = XFS_CMP_CASE;
|
|
|
|
}
|
|
|
|
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2016-11-28 11:57:42 +08:00
|
|
|
static const struct xfs_nameops xfs_ascii_ci_nameops = {
|
2008-05-21 14:58:55 +08:00
|
|
|
.hashname = xfs_ascii_ci_hashname,
|
|
|
|
.compname = xfs_ascii_ci_compname,
|
|
|
|
};
|
|
|
|
|
2014-06-06 13:01:58 +08:00
|
|
|
int
|
|
|
|
xfs_da_mount(
|
|
|
|
struct xfs_mount *mp)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2014-06-06 13:01:58 +08:00
|
|
|
struct xfs_da_geometry *dageo;
|
|
|
|
int nodehdr_size;
|
2013-08-26 12:13:30 +08:00
|
|
|
|
|
|
|
|
2014-05-20 05:46:55 +08:00
|
|
|
ASSERT(mp->m_sb.sb_versionnum & XFS_SB_VERSION_DIRV2BIT);
|
2018-01-09 02:51:27 +08:00
|
|
|
ASSERT(xfs_dir2_dirblock_bytes(&mp->m_sb) <= XFS_MAX_BLOCKSIZE);
|
2013-10-29 19:11:51 +08:00
|
|
|
|
|
|
|
mp->m_dir_inode_ops = xfs_dir_get_ops(mp, NULL);
|
|
|
|
mp->m_nondir_inode_ops = xfs_nondir_get_ops(mp, NULL);
|
|
|
|
|
2013-10-30 06:15:02 +08:00
|
|
|
nodehdr_size = mp->m_dir_inode_ops->node_hdr_size;
|
2014-06-06 13:01:58 +08:00
|
|
|
mp->m_dir_geo = kmem_zalloc(sizeof(struct xfs_da_geometry),
|
|
|
|
KM_SLEEP | KM_MAYFAIL);
|
|
|
|
mp->m_attr_geo = kmem_zalloc(sizeof(struct xfs_da_geometry),
|
|
|
|
KM_SLEEP | KM_MAYFAIL);
|
|
|
|
if (!mp->m_dir_geo || !mp->m_attr_geo) {
|
|
|
|
kmem_free(mp->m_dir_geo);
|
|
|
|
kmem_free(mp->m_attr_geo);
|
2014-06-25 12:58:08 +08:00
|
|
|
return -ENOMEM;
|
2014-06-06 13:01:58 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/* set up directory geometry */
|
|
|
|
dageo = mp->m_dir_geo;
|
|
|
|
dageo->blklog = mp->m_sb.sb_blocklog + mp->m_sb.sb_dirblklog;
|
|
|
|
dageo->fsblog = mp->m_sb.sb_blocklog;
|
2018-01-09 02:51:27 +08:00
|
|
|
dageo->blksize = xfs_dir2_dirblock_bytes(&mp->m_sb);
|
2014-06-06 13:01:58 +08:00
|
|
|
dageo->fsbcount = 1 << mp->m_sb.sb_dirblklog;
|
2014-06-06 13:08:18 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Now we've set up the block conversion variables, we can calculate the
|
|
|
|
* segment block constants using the geometry structure.
|
|
|
|
*/
|
|
|
|
dageo->datablk = xfs_dir2_byte_to_da(dageo, XFS_DIR2_DATA_OFFSET);
|
|
|
|
dageo->leafblk = xfs_dir2_byte_to_da(dageo, XFS_DIR2_LEAF_OFFSET);
|
|
|
|
dageo->freeblk = xfs_dir2_byte_to_da(dageo, XFS_DIR2_FREE_OFFSET);
|
2014-06-06 13:01:58 +08:00
|
|
|
dageo->node_ents = (dageo->blksize - nodehdr_size) /
|
2013-08-26 12:13:30 +08:00
|
|
|
(uint)sizeof(xfs_da_node_entry_t);
|
2014-06-06 13:01:58 +08:00
|
|
|
dageo->magicpct = (dageo->blksize * 37) / 100;
|
|
|
|
|
|
|
|
/* set up attribute geometry - single fsb only */
|
|
|
|
dageo = mp->m_attr_geo;
|
|
|
|
dageo->blklog = mp->m_sb.sb_blocklog;
|
|
|
|
dageo->fsblog = mp->m_sb.sb_blocklog;
|
|
|
|
dageo->blksize = 1 << dageo->blklog;
|
|
|
|
dageo->fsbcount = 1;
|
|
|
|
dageo->node_ents = (dageo->blksize - nodehdr_size) /
|
2013-08-26 12:13:30 +08:00
|
|
|
(uint)sizeof(xfs_da_node_entry_t);
|
2014-06-06 13:01:58 +08:00
|
|
|
dageo->magicpct = (dageo->blksize * 37) / 100;
|
2013-08-26 12:13:30 +08:00
|
|
|
|
2008-05-21 14:58:55 +08:00
|
|
|
if (xfs_sb_version_hasasciici(&mp->m_sb))
|
|
|
|
mp->m_dirnameops = &xfs_ascii_ci_nameops;
|
|
|
|
else
|
|
|
|
mp->m_dirnameops = &xfs_default_nameops;
|
2013-10-29 19:11:46 +08:00
|
|
|
|
2014-06-06 13:01:58 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
xfs_da_unmount(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
|
|
|
kmem_free(mp->m_dir_geo);
|
|
|
|
kmem_free(mp->m_attr_geo);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Return 1 if directory contains only "." and "..".
|
|
|
|
*/
|
2006-06-20 11:04:51 +08:00
|
|
|
int
|
|
|
|
xfs_dir_isempty(
|
|
|
|
xfs_inode_t *dp)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2011-07-08 20:35:13 +08:00
|
|
|
xfs_dir2_sf_hdr_t *sfp;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2016-02-09 13:54:58 +08:00
|
|
|
ASSERT(S_ISDIR(VFS_I(dp)->i_mode));
|
2006-06-20 11:04:51 +08:00
|
|
|
if (dp->i_d.di_size == 0) /* might happen during shutdown. */
|
2005-04-17 06:20:36 +08:00
|
|
|
return 1;
|
|
|
|
if (dp->i_d.di_size > XFS_IFORK_DSIZE(dp))
|
|
|
|
return 0;
|
2011-07-08 20:35:13 +08:00
|
|
|
sfp = (xfs_dir2_sf_hdr_t *)dp->i_df.if_u1.if_data;
|
|
|
|
return !sfp->count;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2006-06-20 11:04:51 +08:00
|
|
|
/*
|
|
|
|
* Validate a given inode number.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
xfs_dir_ino_validate(
|
|
|
|
xfs_mount_t *mp,
|
|
|
|
xfs_ino_t ino)
|
|
|
|
{
|
2017-10-18 12:37:34 +08:00
|
|
|
bool ino_ok = xfs_verify_dir_ino(mp, ino);
|
|
|
|
|
2017-06-21 08:54:47 +08:00
|
|
|
if (unlikely(XFS_TEST_ERROR(!ino_ok, mp, XFS_ERRTAG_DIR_INO_VALIDATE))) {
|
2011-03-07 07:05:35 +08:00
|
|
|
xfs_warn(mp, "Invalid inode number 0x%Lx",
|
2006-06-20 11:04:51 +08:00
|
|
|
(unsigned long long) ino);
|
|
|
|
XFS_ERROR_REPORT("xfs_dir_ino_validate", XFS_ERRLEVEL_LOW, mp);
|
2014-06-25 12:58:08 +08:00
|
|
|
return -EFSCORRUPTED;
|
2006-06-20 11:04:51 +08:00
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Initialize a directory with its "." and ".." entries.
|
|
|
|
*/
|
2006-06-20 11:04:51 +08:00
|
|
|
int
|
|
|
|
xfs_dir_init(
|
|
|
|
xfs_trans_t *tp,
|
|
|
|
xfs_inode_t *dp,
|
|
|
|
xfs_inode_t *pdp)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2014-02-27 13:51:26 +08:00
|
|
|
struct xfs_da_args *args;
|
2006-06-20 11:04:51 +08:00
|
|
|
int error;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2016-02-09 13:54:58 +08:00
|
|
|
ASSERT(S_ISDIR(VFS_I(dp)->i_mode));
|
2014-02-27 13:51:26 +08:00
|
|
|
error = xfs_dir_ino_validate(tp->t_mountp, pdp->i_ino);
|
|
|
|
if (error)
|
2005-04-17 06:20:36 +08:00
|
|
|
return error;
|
2014-02-27 13:51:26 +08:00
|
|
|
|
|
|
|
args = kmem_zalloc(sizeof(*args), KM_SLEEP | KM_NOFS);
|
|
|
|
if (!args)
|
2014-06-25 12:58:08 +08:00
|
|
|
return -ENOMEM;
|
2014-02-27 13:51:26 +08:00
|
|
|
|
2014-06-06 13:01:58 +08:00
|
|
|
args->geo = dp->i_mount->m_dir_geo;
|
2014-02-27 13:51:26 +08:00
|
|
|
args->dp = dp;
|
|
|
|
args->trans = tp;
|
|
|
|
error = xfs_dir2_sf_create(args, pdp->i_ino);
|
|
|
|
kmem_free(args);
|
|
|
|
return error;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2014-09-09 09:58:07 +08:00
|
|
|
* Enter a name in a directory, or check for available space.
|
|
|
|
* If inum is 0, only the available space test is performed.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2006-06-20 11:04:51 +08:00
|
|
|
int
|
|
|
|
xfs_dir_createname(
|
|
|
|
xfs_trans_t *tp,
|
|
|
|
xfs_inode_t *dp,
|
2008-04-10 10:22:07 +08:00
|
|
|
struct xfs_name *name,
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_ino_t inum, /* new entry inode number */
|
|
|
|
xfs_fsblock_t *first, /* bmap's firstblock */
|
2016-08-03 09:19:29 +08:00
|
|
|
struct xfs_defer_ops *dfops, /* bmap's freeblock list */
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_extlen_t total) /* bmap's total block count */
|
|
|
|
{
|
2014-02-27 13:51:26 +08:00
|
|
|
struct xfs_da_args *args;
|
2006-06-20 11:04:51 +08:00
|
|
|
int rval;
|
2005-04-17 06:20:36 +08:00
|
|
|
int v; /* type-checking value */
|
|
|
|
|
2016-02-09 13:54:58 +08:00
|
|
|
ASSERT(S_ISDIR(VFS_I(dp)->i_mode));
|
2014-09-09 09:58:07 +08:00
|
|
|
if (inum) {
|
|
|
|
rval = xfs_dir_ino_validate(tp->t_mountp, inum);
|
|
|
|
if (rval)
|
|
|
|
return rval;
|
2015-10-12 15:21:22 +08:00
|
|
|
XFS_STATS_INC(dp->i_mount, xs_dir_create);
|
2014-09-09 09:58:07 +08:00
|
|
|
}
|
2006-06-20 11:04:51 +08:00
|
|
|
|
2014-02-27 13:51:26 +08:00
|
|
|
args = kmem_zalloc(sizeof(*args), KM_SLEEP | KM_NOFS);
|
|
|
|
if (!args)
|
2014-06-25 12:58:08 +08:00
|
|
|
return -ENOMEM;
|
2014-02-27 13:51:26 +08:00
|
|
|
|
2014-06-06 13:01:58 +08:00
|
|
|
args->geo = dp->i_mount->m_dir_geo;
|
2014-02-27 13:51:26 +08:00
|
|
|
args->name = name->name;
|
|
|
|
args->namelen = name->len;
|
|
|
|
args->filetype = name->type;
|
|
|
|
args->hashval = dp->i_mount->m_dirnameops->hashname(name);
|
|
|
|
args->inumber = inum;
|
|
|
|
args->dp = dp;
|
|
|
|
args->firstblock = first;
|
2016-08-03 09:19:29 +08:00
|
|
|
args->dfops = dfops;
|
2014-02-27 13:51:26 +08:00
|
|
|
args->total = total;
|
|
|
|
args->whichfork = XFS_DATA_FORK;
|
|
|
|
args->trans = tp;
|
|
|
|
args->op_flags = XFS_DA_OP_ADDNAME | XFS_DA_OP_OKNOENT;
|
2014-09-09 09:58:07 +08:00
|
|
|
if (!inum)
|
|
|
|
args->op_flags |= XFS_DA_OP_JUSTCHECK;
|
2014-02-27 13:51:26 +08:00
|
|
|
|
|
|
|
if (dp->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
|
|
|
|
rval = xfs_dir2_sf_addname(args);
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
|
2014-06-06 13:20:32 +08:00
|
|
|
rval = xfs_dir2_isblock(args, &v);
|
2014-02-27 13:51:26 +08:00
|
|
|
if (rval)
|
|
|
|
goto out_free;
|
|
|
|
if (v) {
|
|
|
|
rval = xfs_dir2_block_addname(args);
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
|
2014-06-06 13:20:32 +08:00
|
|
|
rval = xfs_dir2_isleaf(args, &v);
|
2014-02-27 13:51:26 +08:00
|
|
|
if (rval)
|
|
|
|
goto out_free;
|
|
|
|
if (v)
|
|
|
|
rval = xfs_dir2_leaf_addname(args);
|
2005-04-17 06:20:36 +08:00
|
|
|
else
|
2014-02-27 13:51:26 +08:00
|
|
|
rval = xfs_dir2_node_addname(args);
|
|
|
|
|
|
|
|
out_free:
|
|
|
|
kmem_free(args);
|
2005-04-17 06:20:36 +08:00
|
|
|
return rval;
|
|
|
|
}
|
|
|
|
|
2008-05-21 14:58:22 +08:00
|
|
|
/*
|
|
|
|
* If doing a CI lookup and case-insensitive match, dup actual name into
|
|
|
|
* args.value. Return EEXIST for success (ie. name found) or an error.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
xfs_dir_cilookup_result(
|
|
|
|
struct xfs_da_args *args,
|
2010-01-20 07:47:25 +08:00
|
|
|
const unsigned char *name,
|
2008-05-21 14:58:22 +08:00
|
|
|
int len)
|
|
|
|
{
|
|
|
|
if (args->cmpresult == XFS_CMP_DIFFERENT)
|
2014-06-25 12:58:08 +08:00
|
|
|
return -ENOENT;
|
2008-05-21 14:58:22 +08:00
|
|
|
if (args->cmpresult != XFS_CMP_CASE ||
|
|
|
|
!(args->op_flags & XFS_DA_OP_CILOOKUP))
|
2014-06-25 12:58:08 +08:00
|
|
|
return -EEXIST;
|
2008-05-21 14:58:22 +08:00
|
|
|
|
2009-07-19 06:14:57 +08:00
|
|
|
args->value = kmem_alloc(len, KM_NOFS | KM_MAYFAIL);
|
2008-05-21 14:58:22 +08:00
|
|
|
if (!args->value)
|
2014-06-25 12:58:08 +08:00
|
|
|
return -ENOMEM;
|
2008-05-21 14:58:22 +08:00
|
|
|
|
|
|
|
memcpy(args->value, name, len);
|
|
|
|
args->valuelen = len;
|
2014-06-25 12:58:08 +08:00
|
|
|
return -EEXIST;
|
2008-05-21 14:58:22 +08:00
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Lookup a name in a directory, give back the inode number.
|
2008-05-21 14:58:22 +08:00
|
|
|
* If ci_name is not NULL, returns the actual name in ci_name if it differs
|
|
|
|
* to name, or ci_name->name is set to NULL for an exact match.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2008-05-21 14:58:22 +08:00
|
|
|
|
2006-06-20 11:04:51 +08:00
|
|
|
int
|
|
|
|
xfs_dir_lookup(
|
|
|
|
xfs_trans_t *tp,
|
|
|
|
xfs_inode_t *dp,
|
2008-04-10 10:22:07 +08:00
|
|
|
struct xfs_name *name,
|
2008-05-21 14:58:22 +08:00
|
|
|
xfs_ino_t *inum, /* out: inode number */
|
|
|
|
struct xfs_name *ci_name) /* out: actual name if CI match */
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2014-02-27 13:51:26 +08:00
|
|
|
struct xfs_da_args *args;
|
2006-06-20 11:04:51 +08:00
|
|
|
int rval;
|
2005-04-17 06:20:36 +08:00
|
|
|
int v; /* type-checking value */
|
xfs: stop holding ILOCK over filldir callbacks
The recent change to the readdir locking made in 40194ec ("xfs:
reinstate the ilock in xfs_readdir") for CXFS directory sanity was
probably the wrong thing to do. Deep in the readdir code we
can take page faults in the filldir callback, and so taking a page
fault while holding an inode ilock creates a new set of locking
issues that lockdep warns all over the place about.
The locking order for regular inodes w.r.t. page faults is io_lock
-> pagefault -> mmap_sem -> ilock. The directory readdir code now
triggers ilock -> page fault -> mmap_sem. While we cannot deadlock
at this point, it inverts all the locking patterns that lockdep
normally sees on XFS inodes, and so triggers lockdep. We worked
around this with commit 93a8614 ("xfs: fix directory inode iolock
lockdep false positive"), but that then just moved the lockdep
warning to deeper in the page fault path and triggered on security
inode locks. Fixing the shmem issue there just moved the lockdep
reports somewhere else, and now we are getting false positives from
filesystem freezing annotations getting confused.
Further, if we enter memory reclaim in a readdir path, we now get
lockdep warning about potential deadlocks because the ilock is held
when we enter reclaim. This, again, is different to a regular file
in that we never allow memory reclaim to run while holding the ilock
for regular files. Hence lockdep now throws
ilock->kmalloc->reclaim->ilock warnings.
Basically, the problem is that the ilock is being used to protect
the directory data and the inode metadata, whereas for a regular
file the iolock protects the data and the ilock protects the
metadata. From the VFS perspective, the i_mutex serialises all
accesses to the directory data, and so not holding the ilock for
readdir doesn't matter. The issue is that CXFS doesn't access
directory data via the VFS, so it has no "data serialisaton"
mechanism. Hence we need to hold the IOLOCK in the correct places to
provide this low level directory data access serialisation.
The ilock can then be used just when the extent list needs to be
read, just like we do for regular files. The directory modification
code can take the iolock exclusive when the ilock is also taken,
and this then ensures that readdir is correct excluded while
modifications are in progress.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-08-19 08:33:00 +08:00
|
|
|
int lock_mode;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2016-02-09 13:54:58 +08:00
|
|
|
ASSERT(S_ISDIR(VFS_I(dp)->i_mode));
|
2015-10-12 15:21:22 +08:00
|
|
|
XFS_STATS_INC(dp->i_mount, xs_dir_lookup);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2014-02-27 13:51:26 +08:00
|
|
|
/*
|
|
|
|
* We need to use KM_NOFS here so that lockdep will not throw false
|
|
|
|
* positive deadlock warnings on a non-transactional lookup path. It is
|
|
|
|
* safe to recurse into inode recalim in that case, but lockdep can't
|
|
|
|
* easily be taught about it. Hence KM_NOFS avoids having to add more
|
|
|
|
* lockdep Doing this avoids having to add a bunch of lockdep class
|
|
|
|
* annotations into the reclaim path for the ilock.
|
|
|
|
*/
|
|
|
|
args = kmem_zalloc(sizeof(*args), KM_SLEEP | KM_NOFS);
|
2014-06-06 13:01:58 +08:00
|
|
|
args->geo = dp->i_mount->m_dir_geo;
|
2014-02-27 13:51:26 +08:00
|
|
|
args->name = name->name;
|
|
|
|
args->namelen = name->len;
|
|
|
|
args->filetype = name->type;
|
|
|
|
args->hashval = dp->i_mount->m_dirnameops->hashname(name);
|
|
|
|
args->dp = dp;
|
|
|
|
args->whichfork = XFS_DATA_FORK;
|
|
|
|
args->trans = tp;
|
|
|
|
args->op_flags = XFS_DA_OP_OKNOENT;
|
2008-05-21 14:58:22 +08:00
|
|
|
if (ci_name)
|
2014-02-27 13:51:26 +08:00
|
|
|
args->op_flags |= XFS_DA_OP_CILOOKUP;
|
2006-06-20 11:04:51 +08:00
|
|
|
|
xfs: stop holding ILOCK over filldir callbacks
The recent change to the readdir locking made in 40194ec ("xfs:
reinstate the ilock in xfs_readdir") for CXFS directory sanity was
probably the wrong thing to do. Deep in the readdir code we
can take page faults in the filldir callback, and so taking a page
fault while holding an inode ilock creates a new set of locking
issues that lockdep warns all over the place about.
The locking order for regular inodes w.r.t. page faults is io_lock
-> pagefault -> mmap_sem -> ilock. The directory readdir code now
triggers ilock -> page fault -> mmap_sem. While we cannot deadlock
at this point, it inverts all the locking patterns that lockdep
normally sees on XFS inodes, and so triggers lockdep. We worked
around this with commit 93a8614 ("xfs: fix directory inode iolock
lockdep false positive"), but that then just moved the lockdep
warning to deeper in the page fault path and triggered on security
inode locks. Fixing the shmem issue there just moved the lockdep
reports somewhere else, and now we are getting false positives from
filesystem freezing annotations getting confused.
Further, if we enter memory reclaim in a readdir path, we now get
lockdep warning about potential deadlocks because the ilock is held
when we enter reclaim. This, again, is different to a regular file
in that we never allow memory reclaim to run while holding the ilock
for regular files. Hence lockdep now throws
ilock->kmalloc->reclaim->ilock warnings.
Basically, the problem is that the ilock is being used to protect
the directory data and the inode metadata, whereas for a regular
file the iolock protects the data and the ilock protects the
metadata. From the VFS perspective, the i_mutex serialises all
accesses to the directory data, and so not holding the ilock for
readdir doesn't matter. The issue is that CXFS doesn't access
directory data via the VFS, so it has no "data serialisaton"
mechanism. Hence we need to hold the IOLOCK in the correct places to
provide this low level directory data access serialisation.
The ilock can then be used just when the extent list needs to be
read, just like we do for regular files. The directory modification
code can take the iolock exclusive when the ilock is also taken,
and this then ensures that readdir is correct excluded while
modifications are in progress.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-08-19 08:33:00 +08:00
|
|
|
lock_mode = xfs_ilock_data_map_shared(dp);
|
2014-02-27 13:51:26 +08:00
|
|
|
if (dp->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
|
|
|
|
rval = xfs_dir2_sf_lookup(args);
|
|
|
|
goto out_check_rval;
|
|
|
|
}
|
|
|
|
|
2014-06-06 13:20:32 +08:00
|
|
|
rval = xfs_dir2_isblock(args, &v);
|
2014-02-27 13:51:26 +08:00
|
|
|
if (rval)
|
|
|
|
goto out_free;
|
|
|
|
if (v) {
|
|
|
|
rval = xfs_dir2_block_lookup(args);
|
|
|
|
goto out_check_rval;
|
|
|
|
}
|
|
|
|
|
2014-06-06 13:20:32 +08:00
|
|
|
rval = xfs_dir2_isleaf(args, &v);
|
2014-02-27 13:51:26 +08:00
|
|
|
if (rval)
|
|
|
|
goto out_free;
|
|
|
|
if (v)
|
|
|
|
rval = xfs_dir2_leaf_lookup(args);
|
2005-04-17 06:20:36 +08:00
|
|
|
else
|
2014-02-27 13:51:26 +08:00
|
|
|
rval = xfs_dir2_node_lookup(args);
|
|
|
|
|
|
|
|
out_check_rval:
|
2014-06-25 12:58:08 +08:00
|
|
|
if (rval == -EEXIST)
|
2005-04-17 06:20:36 +08:00
|
|
|
rval = 0;
|
2008-05-21 14:58:22 +08:00
|
|
|
if (!rval) {
|
2014-02-27 13:51:26 +08:00
|
|
|
*inum = args->inumber;
|
2008-05-21 14:58:22 +08:00
|
|
|
if (ci_name) {
|
2014-02-27 13:51:26 +08:00
|
|
|
ci_name->name = args->value;
|
|
|
|
ci_name->len = args->valuelen;
|
2008-05-21 14:58:22 +08:00
|
|
|
}
|
|
|
|
}
|
2014-02-27 13:51:26 +08:00
|
|
|
out_free:
|
xfs: stop holding ILOCK over filldir callbacks
The recent change to the readdir locking made in 40194ec ("xfs:
reinstate the ilock in xfs_readdir") for CXFS directory sanity was
probably the wrong thing to do. Deep in the readdir code we
can take page faults in the filldir callback, and so taking a page
fault while holding an inode ilock creates a new set of locking
issues that lockdep warns all over the place about.
The locking order for regular inodes w.r.t. page faults is io_lock
-> pagefault -> mmap_sem -> ilock. The directory readdir code now
triggers ilock -> page fault -> mmap_sem. While we cannot deadlock
at this point, it inverts all the locking patterns that lockdep
normally sees on XFS inodes, and so triggers lockdep. We worked
around this with commit 93a8614 ("xfs: fix directory inode iolock
lockdep false positive"), but that then just moved the lockdep
warning to deeper in the page fault path and triggered on security
inode locks. Fixing the shmem issue there just moved the lockdep
reports somewhere else, and now we are getting false positives from
filesystem freezing annotations getting confused.
Further, if we enter memory reclaim in a readdir path, we now get
lockdep warning about potential deadlocks because the ilock is held
when we enter reclaim. This, again, is different to a regular file
in that we never allow memory reclaim to run while holding the ilock
for regular files. Hence lockdep now throws
ilock->kmalloc->reclaim->ilock warnings.
Basically, the problem is that the ilock is being used to protect
the directory data and the inode metadata, whereas for a regular
file the iolock protects the data and the ilock protects the
metadata. From the VFS perspective, the i_mutex serialises all
accesses to the directory data, and so not holding the ilock for
readdir doesn't matter. The issue is that CXFS doesn't access
directory data via the VFS, so it has no "data serialisaton"
mechanism. Hence we need to hold the IOLOCK in the correct places to
provide this low level directory data access serialisation.
The ilock can then be used just when the extent list needs to be
read, just like we do for regular files. The directory modification
code can take the iolock exclusive when the ilock is also taken,
and this then ensures that readdir is correct excluded while
modifications are in progress.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-08-19 08:33:00 +08:00
|
|
|
xfs_iunlock(dp, lock_mode);
|
2014-02-27 13:51:26 +08:00
|
|
|
kmem_free(args);
|
2005-04-17 06:20:36 +08:00
|
|
|
return rval;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Remove an entry from a directory.
|
|
|
|
*/
|
2006-06-20 11:04:51 +08:00
|
|
|
int
|
|
|
|
xfs_dir_removename(
|
|
|
|
xfs_trans_t *tp,
|
|
|
|
xfs_inode_t *dp,
|
2008-04-10 10:22:07 +08:00
|
|
|
struct xfs_name *name,
|
2006-06-20 11:04:51 +08:00
|
|
|
xfs_ino_t ino,
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_fsblock_t *first, /* bmap's firstblock */
|
2016-08-03 09:19:29 +08:00
|
|
|
struct xfs_defer_ops *dfops, /* bmap's freeblock list */
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_extlen_t total) /* bmap's total block count */
|
|
|
|
{
|
2014-02-27 13:51:26 +08:00
|
|
|
struct xfs_da_args *args;
|
2006-06-20 11:04:51 +08:00
|
|
|
int rval;
|
2005-04-17 06:20:36 +08:00
|
|
|
int v; /* type-checking value */
|
|
|
|
|
2016-02-09 13:54:58 +08:00
|
|
|
ASSERT(S_ISDIR(VFS_I(dp)->i_mode));
|
2015-10-12 15:21:22 +08:00
|
|
|
XFS_STATS_INC(dp->i_mount, xs_dir_remove);
|
2006-06-20 11:04:51 +08:00
|
|
|
|
2014-02-27 13:51:26 +08:00
|
|
|
args = kmem_zalloc(sizeof(*args), KM_SLEEP | KM_NOFS);
|
|
|
|
if (!args)
|
2014-06-25 12:58:08 +08:00
|
|
|
return -ENOMEM;
|
2014-02-27 13:51:26 +08:00
|
|
|
|
2014-06-06 13:01:58 +08:00
|
|
|
args->geo = dp->i_mount->m_dir_geo;
|
2014-02-27 13:51:26 +08:00
|
|
|
args->name = name->name;
|
|
|
|
args->namelen = name->len;
|
|
|
|
args->filetype = name->type;
|
|
|
|
args->hashval = dp->i_mount->m_dirnameops->hashname(name);
|
|
|
|
args->inumber = ino;
|
|
|
|
args->dp = dp;
|
|
|
|
args->firstblock = first;
|
2016-08-03 09:19:29 +08:00
|
|
|
args->dfops = dfops;
|
2014-02-27 13:51:26 +08:00
|
|
|
args->total = total;
|
|
|
|
args->whichfork = XFS_DATA_FORK;
|
|
|
|
args->trans = tp;
|
|
|
|
|
|
|
|
if (dp->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
|
|
|
|
rval = xfs_dir2_sf_removename(args);
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
|
2014-06-06 13:20:32 +08:00
|
|
|
rval = xfs_dir2_isblock(args, &v);
|
2014-02-27 13:51:26 +08:00
|
|
|
if (rval)
|
|
|
|
goto out_free;
|
|
|
|
if (v) {
|
|
|
|
rval = xfs_dir2_block_removename(args);
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
|
2014-06-06 13:20:32 +08:00
|
|
|
rval = xfs_dir2_isleaf(args, &v);
|
2014-02-27 13:51:26 +08:00
|
|
|
if (rval)
|
|
|
|
goto out_free;
|
|
|
|
if (v)
|
|
|
|
rval = xfs_dir2_leaf_removename(args);
|
2005-04-17 06:20:36 +08:00
|
|
|
else
|
2014-02-27 13:51:26 +08:00
|
|
|
rval = xfs_dir2_node_removename(args);
|
|
|
|
out_free:
|
|
|
|
kmem_free(args);
|
2005-04-17 06:20:36 +08:00
|
|
|
return rval;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Replace the inode number of a directory entry.
|
|
|
|
*/
|
2006-06-20 11:04:51 +08:00
|
|
|
int
|
|
|
|
xfs_dir_replace(
|
|
|
|
xfs_trans_t *tp,
|
|
|
|
xfs_inode_t *dp,
|
2008-04-10 10:22:07 +08:00
|
|
|
struct xfs_name *name, /* name of entry to replace */
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_ino_t inum, /* new inode number */
|
|
|
|
xfs_fsblock_t *first, /* bmap's firstblock */
|
2016-08-03 09:19:29 +08:00
|
|
|
struct xfs_defer_ops *dfops, /* bmap's freeblock list */
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_extlen_t total) /* bmap's total block count */
|
|
|
|
{
|
2014-02-27 13:51:26 +08:00
|
|
|
struct xfs_da_args *args;
|
2006-06-20 11:04:51 +08:00
|
|
|
int rval;
|
2005-04-17 06:20:36 +08:00
|
|
|
int v; /* type-checking value */
|
|
|
|
|
2016-02-09 13:54:58 +08:00
|
|
|
ASSERT(S_ISDIR(VFS_I(dp)->i_mode));
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2014-02-27 13:51:26 +08:00
|
|
|
rval = xfs_dir_ino_validate(tp->t_mountp, inum);
|
|
|
|
if (rval)
|
2005-04-17 06:20:36 +08:00
|
|
|
return rval;
|
2006-06-20 11:04:51 +08:00
|
|
|
|
2014-02-27 13:51:26 +08:00
|
|
|
args = kmem_zalloc(sizeof(*args), KM_SLEEP | KM_NOFS);
|
|
|
|
if (!args)
|
2014-06-25 12:58:08 +08:00
|
|
|
return -ENOMEM;
|
2014-02-27 13:51:26 +08:00
|
|
|
|
2014-06-06 13:01:58 +08:00
|
|
|
args->geo = dp->i_mount->m_dir_geo;
|
2014-02-27 13:51:26 +08:00
|
|
|
args->name = name->name;
|
|
|
|
args->namelen = name->len;
|
|
|
|
args->filetype = name->type;
|
|
|
|
args->hashval = dp->i_mount->m_dirnameops->hashname(name);
|
|
|
|
args->inumber = inum;
|
|
|
|
args->dp = dp;
|
|
|
|
args->firstblock = first;
|
2016-08-03 09:19:29 +08:00
|
|
|
args->dfops = dfops;
|
2014-02-27 13:51:26 +08:00
|
|
|
args->total = total;
|
|
|
|
args->whichfork = XFS_DATA_FORK;
|
|
|
|
args->trans = tp;
|
|
|
|
|
|
|
|
if (dp->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
|
|
|
|
rval = xfs_dir2_sf_replace(args);
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
|
2014-06-06 13:20:32 +08:00
|
|
|
rval = xfs_dir2_isblock(args, &v);
|
2014-02-27 13:51:26 +08:00
|
|
|
if (rval)
|
|
|
|
goto out_free;
|
|
|
|
if (v) {
|
|
|
|
rval = xfs_dir2_block_replace(args);
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
|
2014-06-06 13:20:32 +08:00
|
|
|
rval = xfs_dir2_isleaf(args, &v);
|
2014-02-27 13:51:26 +08:00
|
|
|
if (rval)
|
|
|
|
goto out_free;
|
|
|
|
if (v)
|
|
|
|
rval = xfs_dir2_leaf_replace(args);
|
2005-04-17 06:20:36 +08:00
|
|
|
else
|
2014-02-27 13:51:26 +08:00
|
|
|
rval = xfs_dir2_node_replace(args);
|
|
|
|
out_free:
|
|
|
|
kmem_free(args);
|
2005-04-17 06:20:36 +08:00
|
|
|
return rval;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* See if this entry can be added to the directory without allocating space.
|
|
|
|
*/
|
2006-06-20 11:04:51 +08:00
|
|
|
int
|
|
|
|
xfs_dir_canenter(
|
|
|
|
xfs_trans_t *tp,
|
|
|
|
xfs_inode_t *dp,
|
2014-09-09 09:57:52 +08:00
|
|
|
struct xfs_name *name) /* name of entry to add */
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2014-09-09 09:58:07 +08:00
|
|
|
return xfs_dir_createname(tp, dp, name, 0, NULL, NULL, 0);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Utility routines.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Add a block to the directory.
|
2011-07-13 19:43:49 +08:00
|
|
|
*
|
|
|
|
* This routine is for data and free blocks, not leaf/node blocks which are
|
|
|
|
* handled by xfs_da_grow_inode.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2006-06-20 11:04:51 +08:00
|
|
|
int
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_dir2_grow_inode(
|
2011-07-13 19:43:49 +08:00
|
|
|
struct xfs_da_args *args,
|
|
|
|
int space, /* v2 dir's space XFS_DIR2_xxx_SPACE */
|
|
|
|
xfs_dir2_db_t *dbp) /* out: block number added */
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2011-07-13 19:43:49 +08:00
|
|
|
struct xfs_inode *dp = args->dp;
|
|
|
|
struct xfs_mount *mp = dp->i_mount;
|
|
|
|
xfs_fileoff_t bno; /* directory offset of new block */
|
|
|
|
int count; /* count of filesystem blocks */
|
|
|
|
int error;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2009-12-15 07:14:59 +08:00
|
|
|
trace_xfs_dir2_grow_inode(args, space);
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Set lowest possible block in the space requested.
|
|
|
|
*/
|
|
|
|
bno = XFS_B_TO_FSBT(mp, space * XFS_DIR2_SPACE_SIZE);
|
2014-06-06 13:14:11 +08:00
|
|
|
count = args->geo->fsbcount;
|
2011-07-13 19:43:49 +08:00
|
|
|
|
|
|
|
error = xfs_da_grow_inode_int(args, &bno, count);
|
|
|
|
if (error)
|
2005-04-17 06:20:36 +08:00
|
|
|
return error;
|
2008-10-30 14:38:12 +08:00
|
|
|
|
2014-06-06 13:07:53 +08:00
|
|
|
*dbp = xfs_dir2_da_to_db(args->geo, (xfs_dablk_t)bno);
|
2008-10-30 14:38:12 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Update file's size if this is the data space and it grew.
|
|
|
|
*/
|
|
|
|
if (space == XFS_DIR2_DATA_SPACE) {
|
|
|
|
xfs_fsize_t size; /* directory file (data) size */
|
|
|
|
|
|
|
|
size = XFS_FSB_TO_B(mp, bno + count);
|
|
|
|
if (size > dp->i_d.di_size) {
|
|
|
|
dp->i_d.di_size = size;
|
2011-07-13 19:43:49 +08:00
|
|
|
xfs_trans_log_inode(args->trans, dp, XFS_ILOG_CORE);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* See if the directory is a single-block form directory.
|
|
|
|
*/
|
2006-06-20 11:04:51 +08:00
|
|
|
int
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_dir2_isblock(
|
2014-06-06 13:20:32 +08:00
|
|
|
struct xfs_da_args *args,
|
|
|
|
int *vp) /* out: 1 is block, 0 is not block */
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2014-06-06 13:20:32 +08:00
|
|
|
xfs_fileoff_t last; /* last file offset */
|
|
|
|
int rval;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2014-06-06 13:20:32 +08:00
|
|
|
if ((rval = xfs_bmap_last_offset(args->dp, &last, XFS_DATA_FORK)))
|
2005-04-17 06:20:36 +08:00
|
|
|
return rval;
|
2014-06-06 13:20:32 +08:00
|
|
|
rval = XFS_FSB_TO_B(args->dp->i_mount, last) == args->geo->blksize;
|
2017-01-18 03:41:41 +08:00
|
|
|
if (rval != 0 && args->dp->i_d.di_size != args->geo->blksize)
|
|
|
|
return -EFSCORRUPTED;
|
2005-04-17 06:20:36 +08:00
|
|
|
*vp = rval;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* See if the directory is a single-leaf form directory.
|
|
|
|
*/
|
2006-06-20 11:04:51 +08:00
|
|
|
int
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_dir2_isleaf(
|
2014-06-06 13:20:32 +08:00
|
|
|
struct xfs_da_args *args,
|
|
|
|
int *vp) /* out: 1 is block, 0 is not block */
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2014-06-06 13:20:32 +08:00
|
|
|
xfs_fileoff_t last; /* last file offset */
|
|
|
|
int rval;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2014-06-06 13:20:32 +08:00
|
|
|
if ((rval = xfs_bmap_last_offset(args->dp, &last, XFS_DATA_FORK)))
|
2005-04-17 06:20:36 +08:00
|
|
|
return rval;
|
2014-06-06 13:20:32 +08:00
|
|
|
*vp = last == args->geo->leafblk + args->geo->fsbcount;
|
2005-04-17 06:20:36 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Remove the given block from the directory.
|
|
|
|
* This routine is used for data and free blocks, leaf/node are done
|
|
|
|
* by xfs_da_shrink_inode.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
xfs_dir2_shrink_inode(
|
2006-06-20 11:04:51 +08:00
|
|
|
xfs_da_args_t *args,
|
|
|
|
xfs_dir2_db_t db,
|
2012-06-22 16:50:14 +08:00
|
|
|
struct xfs_buf *bp)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
xfs_fileoff_t bno; /* directory file offset */
|
|
|
|
xfs_dablk_t da; /* directory file offset */
|
|
|
|
int done; /* bunmap is finished */
|
2006-06-20 11:04:51 +08:00
|
|
|
xfs_inode_t *dp;
|
|
|
|
int error;
|
|
|
|
xfs_mount_t *mp;
|
|
|
|
xfs_trans_t *tp;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2009-12-15 07:14:59 +08:00
|
|
|
trace_xfs_dir2_shrink_inode(args, db);
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
dp = args->dp;
|
|
|
|
mp = dp->i_mount;
|
|
|
|
tp = args->trans;
|
2014-06-06 13:07:53 +08:00
|
|
|
da = xfs_dir2_db_to_da(args->geo, db);
|
2015-07-29 09:51:01 +08:00
|
|
|
|
|
|
|
/* Unmap the fsblock(s). */
|
|
|
|
error = xfs_bunmapi(tp, dp, da, args->geo->fsbcount, 0, 0,
|
2016-08-03 09:19:29 +08:00
|
|
|
args->firstblock, args->dfops, &done);
|
2015-07-29 09:51:01 +08:00
|
|
|
if (error) {
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
2015-07-29 09:51:01 +08:00
|
|
|
* ENOSPC actually can happen if we're in a removename with no
|
|
|
|
* space reservation, and the resulting block removal would
|
|
|
|
* cause a bmap btree split or conversion from extents to btree.
|
|
|
|
* This can only happen for un-fragmented directory blocks,
|
|
|
|
* since you need to be punching out the middle of an extent.
|
|
|
|
* In this case we need to leave the block in the file, and not
|
|
|
|
* binval it. So the block has to be in a consistent empty
|
|
|
|
* state and appropriately logged. We don't free up the buffer,
|
|
|
|
* the caller can tell it hasn't happened since it got an error
|
|
|
|
* back.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
ASSERT(done);
|
|
|
|
/*
|
|
|
|
* Invalidate the buffer from the transaction.
|
|
|
|
*/
|
2012-06-22 16:50:14 +08:00
|
|
|
xfs_trans_binval(tp, bp);
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* If it's not a data block, we're done.
|
|
|
|
*/
|
2014-06-06 13:08:18 +08:00
|
|
|
if (db >= xfs_dir2_byte_to_db(args->geo, XFS_DIR2_LEAF_OFFSET))
|
2005-04-17 06:20:36 +08:00
|
|
|
return 0;
|
|
|
|
/*
|
|
|
|
* If the block isn't the last one in the directory, we're done.
|
|
|
|
*/
|
2014-06-06 13:06:53 +08:00
|
|
|
if (dp->i_d.di_size > xfs_dir2_db_off_to_byte(args->geo, db + 1, 0))
|
2005-04-17 06:20:36 +08:00
|
|
|
return 0;
|
|
|
|
bno = da;
|
|
|
|
if ((error = xfs_bmap_last_before(tp, dp, &bno, XFS_DATA_FORK))) {
|
|
|
|
/*
|
|
|
|
* This can't really happen unless there's kernel corruption.
|
|
|
|
*/
|
|
|
|
return error;
|
|
|
|
}
|
2014-06-06 13:11:18 +08:00
|
|
|
if (db == args->geo->datablk)
|
2005-04-17 06:20:36 +08:00
|
|
|
ASSERT(bno == 0);
|
|
|
|
else
|
|
|
|
ASSERT(bno > 0);
|
|
|
|
/*
|
|
|
|
* Set the size to the new last block.
|
|
|
|
*/
|
|
|
|
dp->i_d.di_size = XFS_FSB_TO_B(mp, bno);
|
|
|
|
xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);
|
|
|
|
return 0;
|
|
|
|
}
|