On Wed, Apr 15, 2020 at 08:13:56AM -0700, Guenter Roeck wrote:
On Wed, Apr 15, 2020 at 04:49:19PM +0200, Greg Kroah-Hartman wrote:
On Wed, Apr 15, 2020 at 07:21:44AM -0700, Guenter Roeck wrote:
Upstream commit cc41f11a21a5 ("scsi: mpt3sas: Fix kernel panic observed on soft HBA unplug") Fixes: c666d3be99c0 ("scsi: mpt3sas: wait for and flush running commands on shutdown/unload") in linux-4.14.y: 3748694f1b91 Applies to: v4.14.y
This also belongs to 4.19.y, 5.4.y, 5.5.y, and 5.6.y as it is cc: stable. But it doesn't backport cleanly to all, so I need a working backport in order to be able to take it...
I tracked this one down. The offending patch (c666d3be99c0) was applied to v4.16 and to v4.14.y. The script takes that as clue to request a backport; it assumes that the normal stable processs (whatever you and Sasha run to identify patches to apply) takes care of more recent releases, and doesn't look into those. This is intentional. We can change it, but I don't really want to duplicate your and Sasha's work.
Oops, and I completely forgot about 5.5 and 5.6. The script doesn't tell me (and neither about 4.9) because there is no such Chrome OS release, so I have to check those manually.
Either case, the patch applies cleanly to 4.19.y and later for me. Did you see conflicts, or build problems, when trying to apply it ?
When trying to patch 4.19.y: checking file drivers/scsi/mpt3sas/mpt3sas_scsih.c Hunk #1 succeeded at 9919 (offset 11 lines). Hunk #2 FAILED at 9992. 1 out of 2 hunks FAILED
Interesting. Both "git cherry-pick" and "git am" (I don't even need "git am -3") on top of v4.19.115 work fine for me. But, yes, trying to apply the patch using the patch command does indeed fail with this error. Wonder what git does differently when running "git am".
Odd, let me dig into that, I can do a 'cherry-pick' and then export it as a patch, I've had to do that before...
thanks,
greg k-h