[LU-12460] lli_trunc_sem can lead to a readlock

• Type: Bug
• Resolution: Fixed
• Priority: Major
• Fix Version/s: Lustre 2.14.0
• Affects Version/s: Lustre 2.13.0, Lustre 2.12.2
• Environment: Any lustre client
• Severity: 3
• Assignee: Neil Brown
• Reporter: James A Simmons

Description

Neil Brown discovered during the linux lustre client work a potential read lock condition involving lli_trunc_sem. If you mmap a lustre file, then read into that mapped memory from the file they can shared the same mmap_sem which does is okay since they are down_read() calls. The read lock can happen if truncate is called in between memory mapping the file and reading into the that mapped memory from the file.

Issue Links

• is related to: LU-12511 Prepare lustre for adoption into the linux kernel

Comments

[Peter Jones]

Landed for 2.14

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/35271/
Subject: LU-12460 llite: replace lli_trunc_sem
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: e5914a61ac7705f5768c9c75094a456dafefb04f

[Patrick Farrell (Inactive)]

Right, thank you - Obviously, I did not read through the full series......

[Neil Brown]

Patrick,

 that patch has been in my lustre-testing branch for a while, but I hadn't posted it because I really want there to be a better way - so I was procrastinating.

James did post it to the list on 20th May with a bunch of patches which were patches from my lustre-testing which he had modified to varying extents (I've merged the changes I liked into my lustre-testing).

You might  remember discussing your "ll_fault fixes" patch that was in the series and I had questions about it.  It was patch 1 of the series.  This patch was patch  3 (of 29).

[Gerrit Updater]

Patrick Farrell ([e-mail]) uploaded a new patch: https://review.whamcloud.com/35283
Subject: LU-12460 llite: replace lli_trunc_sem
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: e0561ca36a6062d423a447070eda688d0194b2cf

[Patrick Farrell (Inactive)]

James,

I asked this in Gerrit but didn't see a response - Has this patch been on a mailing list and I missed it?  Or is this the first appearance of it?  If it's been on a mailing list, which one?

[James A Simmons]

Yeah the details were lacking at the time I opened the ticket. It was based on what was in the patch. As the discussion continues this ticket description can be updated.

[Andreas Dilger]

James, your bug description doesn't quite make sense. Is there supposed to be a write lock somewhere in there?

I haven't looked at the CLIO code in a while, but we used to have a check for the source pages coming from mmapped memory of the target file, or vice versa.

[Gerrit Updater]

James Simmons ([e-mail]) uploaded a new patch: https://review.whamcloud.com/35271
Subject: LU-12460 llite: replace lli_trunc_sem
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: a234469c691056b1eb81ae0a22ab9f105cffa8a9