Messages in this thread | | | Date | Mon, 7 Nov 2011 17:43:51 +0100 | From | Heiko Carstens <> | Subject | Re: [GIT PULL] Queue free fix (was Re: [PATCH] block: Free queue resources at blk_release_queue()) |
| |
On Mon, Nov 07, 2011 at 10:36:49AM -0500, Mike Snitzer wrote: > > struct pgpath is freed before dm_target when tearing down dm table. > > So if the problematic completion was being done after freeing pgpath > > but before freeing dm_target, crash would look like that > > and what's happening seems the same for these dm crashes: > > dm table was somehow destroyed while I/O was in-flight. > > Could be the block layer's onstack plugging changes are at the heart of > this. > > I voiced onstack plugging concerns relative to DM some time ago > (https://lkml.org/lkml/2011/3/9/450) but somehow convinced myself DM was > fine to no longer need dm_table_unplug_all() etc. Unfortunately I > cannot recall _why_ I felt that was the case. > > So DM needs further review relative to block's onstack plugging changes > and DM IO completion. > > > It's interesting that your test started to crash in dm with v3.0.7. > > Have you gotten these dm crashes with v3.0.6 or before? > > Have you hit the initially-reported scsi oops with v3.0.7? > > Are your v3.0.6 and v3.0.7 compiled with same config and the tests > > ran on same system? > > If all 3.0.x fail: it would be interesting to know if 2.6.39 (which > introduced the onstack plugging) also has these problems.
At least 2.6.39.1 failed as well, as reported already in June. At least the crash looked similar back then:
http://marc.info/?l=linux-scsi&m=130824013229933&w=2
> Testing with > 2.6.38 would be very insightful because it obviously doesn't have any of > the onstack plugging churn.
Ok, we might try 2.6.38.8 (latest stable) and see what happens.
| |