* [PATCH] bnx2fc: Set no_async_abort to 1 in SCSI host template.
@ 2014-08-05 19:45 Chad Dupuis
2014-08-05 20:38 ` Venkatesh Srinivas
0 siblings, 1 reply; 6+ messages in thread
From: Chad Dupuis @ 2014-08-05 19:45 UTC (permalink / raw)
To: linux-scsi, hch; +Cc: giridhar.malavali
Set this to 1 for now as we've observed crashes when this is set to the default
value of 0.
Signed-off-by: Chad Dupuis <chad.dupuis@qlogic.com>
---
drivers/scsi/bnx2fc/bnx2fc_fcoe.c | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
diff --git a/drivers/scsi/bnx2fc/bnx2fc_fcoe.c b/drivers/scsi/bnx2fc/bnx2fc_fcoe.c
index 785d0d7..4882ad6 100644
--- a/drivers/scsi/bnx2fc/bnx2fc_fcoe.c
+++ b/drivers/scsi/bnx2fc/bnx2fc_fcoe.c
@@ -2789,6 +2789,7 @@ static struct scsi_host_template bnx2fc_shost_template = {
.use_clustering = ENABLE_CLUSTERING,
.sg_tablesize = BNX2FC_MAX_BDS_PER_CMD,
.max_sectors = 1024,
+ .no_async_abort = 1,
};
static struct libfc_function_template bnx2fc_libfc_fcn_templ = {
--
1.7.1
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH] bnx2fc: Set no_async_abort to 1 in SCSI host template.
2014-08-05 19:45 [PATCH] bnx2fc: Set no_async_abort to 1 in SCSI host template Chad Dupuis
@ 2014-08-05 20:38 ` Venkatesh Srinivas
2014-08-05 21:02 ` Chad Dupuis
0 siblings, 1 reply; 6+ messages in thread
From: Venkatesh Srinivas @ 2014-08-05 20:38 UTC (permalink / raw)
To: Chad Dupuis; +Cc: linux-scsi, hch, giridhar.malavali
On Tue, Aug 5, 2014 at 12:45 PM, Chad Dupuis <chad.dupuis@qlogic.com> wrote:
> Set this to 1 for now as we've observed crashes when this is set to the default
> value of 0.
What sorts of crashes have you seen with no_async_abort=0 (default)?
At which kernel versions?
Thanks,
-- vs;
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] bnx2fc: Set no_async_abort to 1 in SCSI host template.
2014-08-05 20:38 ` Venkatesh Srinivas
@ 2014-08-05 21:02 ` Chad Dupuis
2014-08-06 19:17 ` Christoph Hellwig
0 siblings, 1 reply; 6+ messages in thread
From: Chad Dupuis @ 2014-08-05 21:02 UTC (permalink / raw)
To: Venkatesh Srinivas; +Cc: linux-scsi, hch, giridhar.malavali
On Tue, 5 Aug 2014, Venkatesh Srinivas wrote:
> On Tue, Aug 5, 2014 at 12:45 PM, Chad Dupuis <chad.dupuis@qlogic.com> wrote:
>> Set this to 1 for now as we've observed crashes when this is set to the default
>> value of 0.
>
> What sorts of crashes have you seen with no_async_abort=0 (default)?
> At which kernel versions?
We've seen a crash where we use a request that has already been used else
where when requeuing a SCSI command:
May 1 15:50:48 sl-b109 kernel: ------------[ cut here ]------------
May 1 15:50:48 sl-b109 kernel: kernel BUG at block/blk-core.c:1217!
May 1 15:50:48 sl-b109 kernel: invalid opcode: 0000 [#1] SMP
May 1 15:50:48 sl-b109 kernel: Modules linked in: sg ebtable_nat ebtables
nfsv3 rpcsec_gss_krb5 nfsv4 dns_resolver nfs fscache ipt_REJECT
ip6t_REJECT nf_conntrack_ipv4 nf_conntrack_ipv6 nf_defrag_ipv4
nf_defrag_ipv6 xt_conntrack nf_conntrack ip6table_filter iptable_filter
ip_tables ip6_tables dm_round_robin iTCO_wdt iTCO_vendor_support coretemp
kvm_intel kvm crct10dif_pclmul crc32_pclmul crc32c_intel
ghash_clmulni_intel aesni_intel lrw gf128mul glue_helper ablk_helper
cryptd pcspkr serio_raw sb_edac edac_core lpc_ich mfd_core hpilo hpwdt
ioatdma ntb ipmi_si ipmi_msghandler video acpi_power_meter shpchp
pcc_cpufreq mperf nfsd auth_rpcgss nfs_acl lockd sunrpc uinput xfs
dm_service_time 8021q garp stp llc mrp sd_mod crc_t10dif crct10dif_common
mgag200 syscopyarea sysfillrect sysimgblt drm_kms_helper ttm igb drm bnx2x
May 1 15:50:48 sl-b109 kernel: ptp pps_core dca i2c_algo_bit i2c_core
hpsa mdio libcrc32c dm_mirror dm_region_hash dm_log bnx2fc cnic uio fcoe
libfcoe libfc scsi_transport_fc scsi_tgt dm_multipath dm_mod
May 1 15:50:48 sl-b109 kernel: CPU: 3 PID: 515 Comm: bnx2fc_thread/3 Not
tainted 3.10.0-121.el7.x86_64 #1
May 1 15:50:48 sl-b109 kernel: Hardware name: HP ProLiant BL460c Gen8,
BIOS I31 12/20/2013
May 1 15:50:48 sl-b109 kernel: task: ffff880232345b00 ti:
ffff880036b20000 task.ti: ffff880036b20000
May 1 15:50:48 sl-b109 kernel: RIP: 0010:[<ffffffff8129007d>]
[<ffffffff8129007d>] blk_requeue_request+0x8d/0x90
May 1 15:50:48 sl-b109 kernel: RSP: 0018:ffff880237a63e68 EFLAGS:
00010097
May 1 15:50:48 sl-b109 kernel: RAX: ffff880432d9c660 RBX:
ffff88043142c800 RCX: dead000000200200
May 1 15:50:48 sl-b109 kernel: RDX: ffff880432d9c660 RSI:
ffff880432d9c510 RDI: ffff880432d9c660
May 1 15:50:48 sl-b109 kernel: RBP: ffff880237a63e80 R08:
ffff880432d9c660 R09: 0000000000000001
May 1 15:50:48 sl-b109 kernel: R10: ffffffff819f4560 R11:
0000000000001000 R12: ffff880432d9c510
May 1 15:50:48 sl-b109 kernel: R13: ffff880432b9ad00 R14:
ffff88021222cc40 R15: ffff880410ba0c28
May 1 15:50:48 sl-b109 kernel: FS: 0000000000000000(0000)
GS:ffff880237a60000(0000) knlGS:0000000000000000
May 1 15:50:48 sl-b109 kernel: CS: 0010 DS: 0000 ES: 0000 CR0:
0000000080050033
May 1 15:50:48 sl-b109 kernel: CR2: 00000000020ad4c0 CR3:
000000042b166000 CR4: 00000000000407e0
May 1 15:50:48 sl-b109 kernel: DR0: 0000000000000000 DR1:
0000000000000000 DR2: 0000000000000000
May 1 15:50:48 sl-b109 kernel: DR3: 0000000000000000 DR6:
00000000ffff0ff0 DR7: 0000000000000400
May 1 15:50:48 sl-b109 kernel: Stack:
May 1 15:50:48 sl-b109 kernel: ffff88043142c800 ffff880432b9ad00
0000000000000202 ffff880237a63ec8
May 1 15:50:48 sl-b109 kernel: ffffffff813e73c8 ffff880237a63ea0
ffffffff8128f85c ffff88021222cc40
May 1 15:50:48 sl-b109 kernel: 0000000000002001 000000000002bf20
0000000000000006 0000000000000001
May 1 15:50:48 sl-b109 kernel: Call Trace:
May 1 15:50:48 sl-b109 kernel: <IRQ>
May 1 15:50:48 sl-b109 kernel:
May 1 15:50:48 sl-b109 kernel: [<ffffffff813e73c8>]
__scsi_queue_insert+0x98/0x120
May 1 15:50:48 sl-b109 kernel: [<ffffffff8128f85c>] ?
blk_run_queue_async+0x3c/0x40
May 1 15:50:48 sl-b109 kernel: [<ffffffff813e7542>]
scsi_softirq_done+0xd2/0x160
May 1 15:50:48 sl-b109 kernel: [<ffffffff81299b80>]
blk_done_softirq+0x90/0xc0
May 1 15:50:48 sl-b109 kernel: [<ffffffff81067047>]
__do_softirq+0xf7/0x290
May 1 15:50:48 sl-b109 kernel: [<ffffffff815fe15c>]
call_softirq+0x1c/0x30
May 1 15:50:48 sl-b109 kernel: <EOI>
May 1 15:50:48 sl-b109 kernel:
May 1 15:50:48 sl-b109 kernel: [<ffffffff81014d25>] do_softirq+0x55/0x90
May 1 15:50:48 sl-b109 kernel: [<ffffffff81066564>]
local_bh_enable_ip+0x94/0xa0
May 1 15:50:48 sl-b109 kernel: [<ffffffff815f378b>]
_raw_spin_unlock_bh+0x1b/0x40
May 1 15:50:48 sl-b109 kernel: [<ffffffffa0097f58>]
bnx2fc_process_cq_compl+0xf8/0x280 [bnx2fc]
May 1 15:50:48 sl-b109 kernel: [<ffffffffa0092e16>]
bnx2fc_percpu_io_thread+0x116/0x1a0 [bnx2fc]
May 1 15:50:48 sl-b109 kernel: [<ffffffffa0092d00>] ?
bnx2fc_get_src_mac+0x20/0x20 [bnx2fc]
It's possible that this is related to:
http://marc.info/?l=linux-scsi&m=140266091913551&w=2
however we observed no crashes when setting no_async to '1' so for now we
wanted to set this in our host template.
> > Thanks, >
-- vs; >
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] bnx2fc: Set no_async_abort to 1 in SCSI host template.
2014-08-05 21:02 ` Chad Dupuis
@ 2014-08-06 19:17 ` Christoph Hellwig
2014-08-06 20:01 ` Chad Dupuis
0 siblings, 1 reply; 6+ messages in thread
From: Christoph Hellwig @ 2014-08-06 19:17 UTC (permalink / raw)
To: Chad Dupuis; +Cc: Venkatesh Srinivas, linux-scsi, hch, giridhar.malavali
On Tue, Aug 05, 2014 at 05:02:46PM -0400, Chad Dupuis wrote:
>
>
> On Tue, 5 Aug 2014, Venkatesh Srinivas wrote:
>
>> On Tue, Aug 5, 2014 at 12:45 PM, Chad Dupuis <chad.dupuis@qlogic.com> wrote:
>>> Set this to 1 for now as we've observed crashes when this is set to the default
>>> value of 0.
>>
>> What sorts of crashes have you seen with no_async_abort=0 (default)?
>> At which kernel versions?
>
> We've seen a crash where we use a request that has already been used else
> where when requeuing a SCSI command:
What kernel version is this one? The last issues we saw in this area
were before we pulled fixes for these scenarios in:
d555a2abf3481f81303d835046a5ec2c4fb3ca8e
644373a4219add42123df69c8b7ce6a918475ccd
7daf480483e60898f30e0a2a84fecada7a7cfac0
c69e6f812bab0d5442b40e2f1bfbca48d40bc50b
8922a908908ff947f1f211e07e2e97f1169ad3cb
a33c070bced8b283e22e8dbae35177a033b810bf
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] bnx2fc: Set no_async_abort to 1 in SCSI host template.
2014-08-06 19:17 ` Christoph Hellwig
@ 2014-08-06 20:01 ` Chad Dupuis
2014-08-07 6:43 ` Christoph Hellwig
0 siblings, 1 reply; 6+ messages in thread
From: Chad Dupuis @ 2014-08-06 20:01 UTC (permalink / raw)
To: Christoph Hellwig; +Cc: Venkatesh Srinivas, linux-scsi, giridhar.malavali
On Wed, 6 Aug 2014, Christoph Hellwig wrote:
> On Tue, Aug 05, 2014 at 05:02:46PM -0400, Chad Dupuis wrote:
>>
>>
>> On Tue, 5 Aug 2014, Venkatesh Srinivas wrote:
>>
>>> On Tue, Aug 5, 2014 at 12:45 PM, Chad Dupuis <chad.dupuis@qlogic.com> wrote:
>>>> Set this to 1 for now as we've observed crashes when this is set to the default
>>>> value of 0.
>>>
>>> What sorts of crashes have you seen with no_async_abort=0 (default)?
>>> At which kernel versions?
>>
>> We've seen a crash where we use a request that has already been used else
>> where when requeuing a SCSI command:
>
> What kernel version is this one? The last issues we saw in this area
> were before we pulled fixes for these scenarios in:
>
> d555a2abf3481f81303d835046a5ec2c4fb3ca8e
> 644373a4219add42123df69c8b7ce6a918475ccd
> 7daf480483e60898f30e0a2a84fecada7a7cfac0
> c69e6f812bab0d5442b40e2f1bfbca48d40bc50b
> 8922a908908ff947f1f211e07e2e97f1169ad3cb
> a33c070bced8b283e22e8dbae35177a033b810bf
>
>
This was on RHEL 7. Should we retry the test case with the latest
mainline?
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] bnx2fc: Set no_async_abort to 1 in SCSI host template.
2014-08-06 20:01 ` Chad Dupuis
@ 2014-08-07 6:43 ` Christoph Hellwig
0 siblings, 0 replies; 6+ messages in thread
From: Christoph Hellwig @ 2014-08-07 6:43 UTC (permalink / raw)
To: Chad Dupuis
Cc: Christoph Hellwig, Venkatesh Srinivas, linux-scsi,
giridhar.malavali
On Wed, Aug 06, 2014 at 04:01:12PM -0400, Chad Dupuis wrote:
> This was on RHEL 7. Should we retry the test case with the latest
> mainline?
Yes. Please always test your patches against latest mainline. If that is
for some reason not possible please at least very prominently mention that
the issue was reported against a back-level kernel.
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2014-08-07 6:44 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2014-08-05 19:45 [PATCH] bnx2fc: Set no_async_abort to 1 in SCSI host template Chad Dupuis
2014-08-05 20:38 ` Venkatesh Srinivas
2014-08-05 21:02 ` Chad Dupuis
2014-08-06 19:17 ` Christoph Hellwig
2014-08-06 20:01 ` Chad Dupuis
2014-08-07 6:43 ` Christoph Hellwig
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).