Instead of terminating after five retries, commands terminated by
ABORTED_COMMAND sense are retrying forever. The problem was
introduced by:
commit b60af5b0ad
Author: Alan Stern <stern@rowland.harvard.edu>
Date: Mon Nov 3 15:56:47 2008 -0500
[SCSI] simplify scsi_io_completion()
Which introduced an error whereby ABORTED_COMMAND now gets erroneously
retried in scsi_io_completion. Fix this by returning the behaviour
back to the default no retry.
Reported-by: Sitsofe Wheeler <sitsofe@yahoo.com>
Tested-by: Sitsofe Wheeler <sitsofe@yahoo.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Commit f27bac2761 which converted sd to
use ida instead of idr incorrectly removed sd_index_lock around id
allocation and free. idr/ida do have internal locks but they protect
their free object lists not the allocation itself. The caller is
responsible for that. This missing synchronization led to the same id
being assigned to multiple devices leading to oops.
Reported and tracked down by Stuart Hayes of Dell.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Stable Tree <stable@kernel.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Need to make sure the outgoing pdu can fit into a single skb. When
calulating the max. outgoing pdu payload size, take into consideration
of
- data can be held in the skb's fragment list, assume 512 bytes per
fragment, and
- data can be held in the headroom.
Signed-off-by: Karen Xie <kxie@chelsio.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
added per-task struct cxgb3i_task_data to track the data transmiting
progress and the state of the pdus to be transmitted.
Signed-off-by: Karen Xie <kxie@chelsio.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
- resize the work-request credit array to be based on skb's MAX_SKB_FRAGS.
- split the skb cb into tx and rx portion
- increase the default transmit window to 128K.
- stop queueing up the outgoing pdus if transmit window is full.
Signed-off-by: Karen Xie <kxie@chelsio.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Bits 31-8 are marked as reserved and should be ignored while
interpreting a region's code.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
The clearing of a vha's req_ques were overrunning during vport
creation. During deletion, vport queues should be torn-down
after all cleanup has occurred.
Signed-off-by: Anirban Chakraborty <anirban.chakraborty@qlogic.com>
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
To ensure smooth operations amongst the FCoE and NIC side
components of the ISP81xx chip, the FCoE driver (qla2xxx) must
ensure the 10gb NIC driver (qlge) does not timeout waiting for
IDC (Inter-Driver Communication) acknowledgments. The
acknowledgment requirements are trivial -- a simple mirroring of
incoming mailbox registers during the AEN to a process-context
capable mailbox command.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Correct response-queue-0 processing by instructing the firmware
to run with interrupt-handshaking disabled, similarly to what is
now done for all non-0 response queues. Since all
response-queues now run in the same mode, the driver no longer
needs the hot-path 'is-disabled-HCCR' test.
Signed-off-by: Anirban Chakraborty <anirban.chakraborty@qlogic.com>
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Error handling code following a kmalloc should free the allocated data.
The semantic match that finds the problem is as follows:
(http://www.emn.fr/x-info/coccinelle/)
// <smpl>
@r exists@
local idexpression x;
statement S;
expression E;
identifier f,l;
position p1,p2;
expression *ptr != NULL;
@@
(
if ((x@p1 = \(kmalloc\|kzalloc\|kcalloc\)(...)) == NULL) S
|
x@p1 = \(kmalloc\|kzalloc\|kcalloc\)(...);
...
if (x == NULL) S
)
<... when != x
when != if (...) { <+...x...+> }
x->f = E
...>
(
return \(0\|<+...x...+>\|ptr\);
|
return@p2 ...;
)
@script:python@
p1 << r.p1;
p2 << r.p2;
@@
print "* file: %s kmalloc %s return %s" % (p1[0].file,p1[0].line,p2[0].line)
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Acked-by: James Smart <james.smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Yanling Qi from LSI found the root cause of the panic, below is his
analysis:
Problem description: the open iscsi driver installs eh_timed_out handler
to the
blank_transport_template of the scsi middle level that causes panic of
timed
out command of other host
Here are the details
Iscsi Session creation
During iscsi session creation time, the iscsi_tcp_session_create() of
iscsi_tpc.c will create a scsi-host for the session. See the statement
marked
with the label A. The statement B replaces the shost->transportt point
with a
local struct variable.
static struct iscsi_cls_session *
iscsi_tcp_session_create(struct iscsi_endpoint *ep, uint16_t cmds_max,
uint16_t qdepth, uint32_t initial_cmdsn,
uint32_t *hostno)
{
struct iscsi_cls_session *cls_session;
struct iscsi_session *session;
struct Scsi_Host *shost;
int cmd_i;
if (ep) {
printk(KERN_ERR "iscsi_tcp: invalid ep %p.\n", ep);
return NULL;
}
A shost = iscsi_host_alloc(&iscsi_sht, 0, qdepth);
if (!shost)
return NULL;
B shost->transportt = iscsi_tcp_scsi_transport;
shost->max_lun = iscsi_max_lun;
Please note the scsi host is allocated by invoking isccsi_host_alloc()
in
libiscsi.c
Polluting the middle level blank_transport_template in
iscsi_host_alloc() of
libiscsi.c
The iscsi_host_alloc() invokes the middle level function
scsi_host_alloc() in
hosts.c for allocating a scsi_host. Then the statement marked with C
assigns
the iscsi_eh_cmd_timed_out handler to the eh_timed_out callback
function.
struct Scsi_Host *iscsi_host_alloc(struct scsi_host_template *sht,
int dd_data_size, uint16_t qdepth)
{
struct Scsi_Host *shost;
struct iscsi_host *ihost;
shost = scsi_host_alloc(sht, sizeof(struct iscsi_host) +
dd_data_size);
if (!shost)
return NULL;
C shost->transportt->eh_timed_out = iscsi_eh_cmd_timed_out;
Please note the shost->transport is the middle level
blank_transport_template
as shown in the code segment below. We see two problems here. 1.
iscsi_eh_cmd_timed_out is installed to the blank_transport_template that
will
cause some body else problem. 2. iscsi_eh_cmd_timed_out will never be
invoked
when iscsi command gets timeout because the statement B resets the
pointer.
Middle level blank_transport_template
In the middle level function scsi_host_alloc() of hosts.c, the middle
level
assigns a blank_transport_template for those hosts not implementing its
transport layer. All HBAs without supporting a specific scsi_transport
will
share the middle level blank_transport_template. Please see the
statement D
struct Scsi_Host *scsi_host_alloc(struct scsi_host_template *sht, int
privsize)
{
struct Scsi_Host *shost;
gfp_t gfp_mask = GFP_KERNEL;
int rval;
if (sht->unchecked_isa_dma && privsize)
gfp_mask |= __GFP_DMA;
shost = kzalloc(sizeof(struct Scsi_Host) + privsize, gfp_mask);
if (!shost)
return NULL;
shost->host_lock = &shost->default_lock;
spin_lock_init(shost->host_lock);
shost->shost_state = SHOST_CREATED;
INIT_LIST_HEAD(&shost->__devices);
INIT_LIST_HEAD(&shost->__targets);
INIT_LIST_HEAD(&shost->eh_cmd_q);
INIT_LIST_HEAD(&shost->starved_list);
init_waitqueue_head(&shost->host_wait);
mutex_init(&shost->scan_mutex);
shost->host_no = scsi_host_next_hn++; /* XXX(hch): still racy */
shost->dma_channel = 0xff;
/* These three are default values which can be overridden */
shost->max_channel = 0;
shost->max_id = 8;
shost->max_lun = 8;
/* Give each shost a default transportt */
D shost->transportt = &blank_transport_template;
Why we see panic at iscsi_eh_cmd_timed_out()
The mpp virtual HBA doesn’t have a specific scsi_transport. Therefore,
the
blank_transport_template will be assigned to the virtual host of the MPP
virtual HBA by SCSI middle level. Please note that the statement C has
assigned
iscsi-transport eh_timedout handler to the blank_transport_template.
When a mpp
virtual command gets timedout, the iscsi_eh_cmd_timed_out() will be
invoked to
handle mpp virtual command timeout from the middle level
scsi_times_out()
function of the scsi_error.c.
enum blk_eh_timer_return scsi_times_out(struct request *req)
{
struct scsi_cmnd *scmd = req->special;
enum blk_eh_timer_return (*eh_timed_out)(struct scsi_cmnd *);
enum blk_eh_timer_return rtn = BLK_EH_NOT_HANDLED;
scsi_log_completion(scmd, TIMEOUT_ERROR);
if (scmd->device->host->transportt->eh_timed_out)
E eh_timed_out =
scmd->device->host->transportt->eh_timed_out;
else if (scmd->device->host->hostt->eh_timed_out)
eh_timed_out = scmd->device->host->hostt->eh_timed_out;
else
eh_timed_out = NULL;
if (eh_timed_out) {
rtn = eh_timed_out(scmd);
It is very easy to understand why we get panic in the
iscsi_eh_cmd_timed_out().
A scsi_cmnd from a no-iscsi device definitely can not resolve out a
session and
session->lock. The panic can be happed anywhere during the differencing.
static enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*scmd)
{
struct iscsi_cls_session *cls_session;
struct iscsi_session *session;
struct iscsi_conn *conn;
enum blk_eh_timer_return rc = BLK_EH_NOT_HANDLED;
cls_session = starget_to_session(scsi_target(scmd->device));
session = cls_session->dd_data;
debug_scsi("scsi cmd %p timedout\n", scmd);
spin_lock(&session->lock);
This patch fixes the problem by moving the setting of the
iscsi_eh_cmd_timed_out to iscsi_add_host, which is after the LLDs
have set their transport template to shost->transportt.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
During cancel testing it has been shown that 15 seconds is not
nearly long enough for the VIOS to respond to a cancel under
loaded situations. Increasing this timeout to 60 seconds allows
time for the VIOS to cancel the outstanding commands and prevents
us from escalating to a full host reset, which can take much longer.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
The ibmvfc driver has a bug in its SCN handling. If it receives
an ELS event such asn an N-Port SCN event or an unsolicited PLOGI,
or any other SCN event which causes ibmvfc_reinit_host to be called,
it is possible that we will call fc_remote_port_add for a target
that already has an rport added, which can result in duplicate
rports getting created for the same targets. Fix this by calling
fc_remote_port_rolechg in this scenario instead to report any possible
role change that may have occurred.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Currently the ibmvfc driver sets the IBMVFC_CLASS_3_ERR flag
in the VFC Frame if both the adapter and the device claim support
for Class 3. However, this bit actually refers to Class 3 Error
Recovery, which is currently not supported by the VIOS. Setting this
bit can cause lots of command timeout responses from the VIOS resulting
in general instability. Fix this by never setting this bit.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Hi,
we have run into an issue with blktrace being started for sg devices.
Please apply.
Thanks,
Martin
From: Martin Peschke <mpeschke@linux.vnet.ibm.com>
The device number denoting a generic SCSI devices (sg) in a blktrace
trace is broken; major and minor are always 0. It looks like
sdp->device->sdev_gendev.devt is not initialized properly.
The fix below uses other data to make up a valid device number,
similar to the way an sg device number is generated for sysfs output.
Reported-by: Stefan Raspl <raspl@linux.vnet.ibm.com>
Signed-off-by: Martin Peschke <mpeschke@linux.vnet.ibm.com>
Acked-by: Douglas Gilbert <dgilbert@interlog.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
We were running i/o and performing a bunch of hba resets in a loop.
This forces a lot of target removes and then rescans. Since the
resets are occuring during scan it's causing the scan i/o to timeout,
invoking error recovery, etc. We end up getting some nasty crashing
in scsi_scan.c due to references to old sdevs that are failing
but had some lingering references that kept them around.
Fix by setting device state to SDEV_DEL if the LLD's slave_alloc
fails.
Signed-off-by: James Smart <james.smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
The ibmvscsi client driver is not unmapping the SCSI command after
encountering a DMA mapping error while trying to map an indirect
scattergather list for the event pool. This leads to a leak of DMA
entitlement that could result in the device failing future DMA operations
in a CMO environment.
Signed-off-by: Robert Jennings <rcj@linux.vnet.ibm.com>
Acked-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Currently, netlink_broadcast() reports errors to the caller if no
messages at all were delivered:
1) If, at least, one message has been delivered correctly, returns 0.
2) Otherwise, if no messages at all were delivered due to skb_clone()
failure, return -ENOBUFS.
3) Otherwise, if there are no listeners, return -ESRCH.
With this patch, the caller knows if the delivery of any of the
messages to the listeners have failed:
1) If it fails to deliver any message (for whatever reason), return
-ENOBUFS.
2) Otherwise, if all messages were delivered OK, returns 0.
3) Otherwise, if no listeners, return -ESRCH.
In the current ctnetlink code and in Netfilter in general, we can add
reliable logging and connection tracking event delivery by dropping the
packets whose events were not successfully delivered over Netlink. Of
course, this option would be settable via /proc as this approach reduces
performance (in terms of filtered connections per seconds by a stateful
firewall) but providing reliable logging and event delivery (for
conntrackd) in return.
This patch also changes some clients of netlink_broadcast() that
may report ENOBUFS errors via printk. This error handling is not
of any help. Instead, the userspace daemons that are listening to
those netlink messages should resync themselves with the kernel-side
if they hit ENOBUFS.
BTW, netlink_broadcast() clients include those that call
cn_netlink_send(), nlmsg_multicast() and genlmsg_multicast() since they
internally call netlink_broadcast() and return its error value.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is currently a DMA mapping leak that can occur in the ibmvfc
driver if we fail to allocate a scatterlist. Fix this by unmapping
the scatterlist in the failure path. Additionally, only log an error
for a scatterlist allocation failure if the log level is greater
than the default, since this can occur when running Active Memory
Sharing and this is not considered an error.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Commit f78badb1ae ([SCSI] fc
transport: pre-emptively terminate i/o upon dev_loss_tmo timeout)
changed the callback semantics of dev_loss_tmo and
terminate_rport_io such that repeated calls could be made. This
could result in the the driver using stale (NULLed-out, in
dev_loss_tmo) data from the rport. Correct this by addint a
simple check to ensure a valid fcport is attached.
Signed-off-by: Seokmann Ju <seokmann.ju@qlogic.com>
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Jeremy Higdon noted
(http://marc.info/?l=linux-scsi&m=123262143131788&w=2) that the
rework done in commit e315cd28b9
was not setting the proper consistent and streaming DMA masks
prior to memory allocations. Correct this and remove the
unnecessary prototype.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
When clearing the flash device's SR, the comment is incorrect...
clearing the SR is 2 steps:
1. the SR protect bit is 1, so the first write zero clears only
that bit,
2. the SR protect bit is now 0, so the next write zero clears the
remaining bits.
The sector erase debug print more correctly identifies that the erase failed.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Commit 73208dfd7a (qla2xxx: add
support for multi-queue adapter) inadvertently backed-out the fix
in 5bff55db3d (qla2xxx: Return a
FAILED status when abort mailbox-command fails.).
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
The flash data was incorrectly being converted (cpu_to_le32())
when using the bulk-flash-write mailbox command (ISP25xx and
above).
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Pre-ISP81XX parts (including ISP24xx and ISP25xx) could contain a
firmware image within a segment of flash, driver would fallback
to loading this firmware if the request-firmware interface failed
(userspace .bin file). Moving forward, all ISP81XX parts will
ship with a suggested-to-be-used firmware image within flash
which all driver should first attempt to load. If the flash
firmware load fails, the driver will then fallback to loading
firmware via the request-firmware interface (ql8100_fw.bin).
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Original code would incorrectly bypass serialization if the DPC
thread were performing a big-hammer operation (ISP abort). This
short circuit, though rare, would subsequently stomp on a
secondary thread's mailbox command execution. Found during
ISP81XX testing.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Original code should work as well given qla24xx_reset_adapter()
is only called in extreme cases where the HBA is taken offline.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Also removes unneeded 'findex' local variable within routine.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
qla4xxx does not check the I_T nexus values correctly
so it ends up creating one session to the target. If
a portal should disappear or they should be reported
in different order the driver will think it is already
logged in when it could now be speaking to a different
target portal or accessing it through a different
initiator port (iscsi initiator port is not tied to
hardware and is just the initiator name plus isid
so you could end up with multiple ports through one
host).
This patch has the driver check the iscsi scsi port
values when matching sessions (we do not check
the initiator name because that is static). It results
in a portal from each target portal group getting
logged into instead of just one per target. In the future
the firmware should hopefully send us notification of other
sessions that are created to other portals within the
same tpgt and the sessions should have different isids.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
I am not sure what happened. It looks like we have always leaked
the q->queue that is allocated from the kfifo_init call. nab finally
noticed that we were leaking and this patch fixes it by adding a
kfree call to iscsi_pool_free. kfifo_free is not used per kfifo_init's
instructions to use kfree.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
This patch makes cxgb3 invoke the GRO hooks instead of LRO. As
GRO has a compatible external interface to LRO this is a very
straightforward replacement.
I've kept the ioctl controls for per-queue LRO switches. However,
we should not encourage anyone to use these.
Because of that, I've also kept the skb construction code in
cxgb3. Hopefully we can phase out those per-queue switches
and then kill this too.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
serial: Add 16850 uart type support to OF uart driver
hvc_console: Remove tty->low_latency
powerpc: Get the number of SLBs from "slb-size" property
powerpc: is_hugepage_only_range() must account for both 4kB and 64kB slices
powerpc/ps3: printing fixups for l64 to ll64 conversion drivers/video
powerpc/ps3: Printing fixups for l64 to ll64 conversion drivers/scsi
powerpc/ps3: Printing fixups for l64 to ll64 conversion drivers/ps3
powerpc/ps3: Printing fixups for l64 to ll64 conversion sound/ppc
powerpc/ps3: Printing fixups for l64 to ll64 conversion drivers/char
powerpc/ps3: Printing fixups for l64 to ll64 conversion drivers/block
powerpc/ps3: Printing fixups for l64 to ll64 conversion arch/powerpc
powerpc/ps3: ps3_repository_read_mm_info() takes u64 * arguments
powerpc/ps3: clear_bit()/set_bit() operate on unsigned longs
powerpc/ps3: The lv1_ routines have u64 parameters
powerpc/ps3: Use dma_addr_t down through the stack
powerpc/ps3: set_dabr() takes an unsigned long
powerpc: Cleanup from l64 to ll64 change drivers/scsi
* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
sata_fsl: Return non-zero on error in probe()
drivers/ata/pata_ali.c: s/isa_bridge/ali_isa_bridge/ to fix alpha build
libata: New driver for OCTEON SOC Compact Flash interface (v7).
libata: Add another column to the ata_timing table.
sata_via: Add VT8261 support
pata_atiixp: update port enabledness test handling
[libata] get-identity ioctl: Fix use of invalid memory pointer
for SAS drivers.
Caught by Ke Wei (and team?) at Marvell.
Also, move the ata_scsi_ioctl export to libata-scsi.c, as that seems to be the
general trend.
Acked-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This is a powerpc specific driver.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
__scsi_device_lookup_by_target() will always return
the first sdev with a matching LUN, regardless of
the state. However, when this sdev is in SDEV_DEL
scsi_device_lookup_by_target() will ignore this
device and so any valid device on the list after
the deleted device will never be found.
So we have to modify __scsi_device_lookup_by_target()
to skip any device in SDEV_DEL.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
David S. Miller