Messages in this thread | | | Date | Fri, 4 Jan 2013 07:35:34 -0500 | From | Josh Boyer <> | Subject | Re: Oops on aoe module removal |
| |
On Thu, Jan 03, 2013 at 04:00:46PM -0500, Josh Boyer wrote: > On Thu, Jan 03, 2013 at 02:50:46PM -0600, Ed Cashin wrote: > > >>> The blk_alloc_queue has already done a bdi_init, so do not bdi_init again in > > >>> aoeblk_gdalloc. > > >>> > > >>> The patch below applies to v3.5.6, with its v47 aoe driver. On my system it > > >>> eliminates the list_del corruption messages. > > >> > > >> Since the patch doesn't apply to current -git, does the problem not > > >> exist there? > > > > > > The original post is about an older kernel with the v47 aoe driver. The current mainline has a v81 aoe driver, so the patch for v3.5.6 isn't expected to apply to the mainline. > > > > > > I'm currently investigating the state of the mainline with relation to this issue. > > > > I don't see the extra call to bdi_init in aoe driver v81 in the mainline git tree, and > > I don't see the symptoms under discussion, either, when performing the same > > test steps. > > > > Kernels with aoe v47 can use the patch I just posted. > > > > I'm going to go through the stable kernels and check, but I believe kernels after > > v47 but before commit 0a41409c5180 should apply the fix in 0a41409c5180: > > > > commit 0a41409c518083133e79015092585d68915865be > > Author: Ed Cashin <ecashin@coraid.com> > > Date: Mon Dec 17 16:03:58 2012 -0800 > > > > aoe: remove vestigial request queue allocation > > > > Josh, can you confirm that the patch I posted in this thread today works for > > your customer? > > Sure. I'll get a test kernel built with that patch and ask them to > test.
Dimitar confirmed in the bug that the kernel I built with the patch no longer oopses on rmmod. Thanks for the quick turn around!
josh
| |