[LU-16807] Resolve newer debug kernel warnings

• Type: Bug
• Resolution: Unresolved
• Priority: Minor
• Fix Version/s: Lustre 2.17.0
• Affects Version/s: Lustre 2.16.0
• Labels: janitor9x
• Environment: Newer distro running a debug kernel.
• Severity: 3
• Assignee: James A Simmons
• Reporter: James A Simmons

Description

Currently we use RHEL7 debug kernels for collecting potential kernel issues. This has two limitations. One is that for Lustre 2.17 we will be dropping RHEL7 support and second RHEL7 kernels has unique issues that are not present in newer kernels. The reason for the delay to using a newer debug kernel like RHEL8 is that it reports far more problems that needs to be resolved before we move to this target for general testing.

Issue Links

• is blocked by: LU-17419 KASAN: slab-out-of-bounds in string_nocheck in server_start_targets
• is blocked by: LU-15899 BUG: KASAN: slab-out-of-bounds in mdt_hsm_release
• is blocked by: LU-17651 KASAN: use-after-free in lnet_net_remove_cpts
• is related to: LU-18657 Bad page state in sanity pcc test 21j
• is related to: LU-18658 circular locking dependency for lod QOS code
• is related to: LU-16783 DEBUG_LOCKS_WARN_ON(!lockdep_enabled()) for quota
• is related to: LU-18612 Quota over-softlimit flag remained after large file removal
• is related to: LU-18664 distribute_txn_commit_batchid_update() is called in an atomic context
• is related to: LU-15899 BUG: KASAN: slab-out-of-bounds in mdt_hsm_release
• is related to: LU-18743 sanity-scrub test fail with debug enabled kernels.
• is related to: LU-15808 ptlrpc lockdep warning "do not call blocking ops when !TASK_RUNNING"
• is related to: LU-17650 KASAN: slab-out-of-bounds in unix_find_other on RHEL9.3
• is related to: LU-17794 UBSAN: array-index-out-of-bounds errors
• is related to: LU-18485 use-after-free in batch_send_update_req()
• is related to: LU-18516 do not call blocking ops when !TASK_RUNNING occurs in osd-ldisk / quota path
• is related to: LU-17386 BUG: KASAN: stack-out-of-bounds in iov_iter_advance
• is related to: LU-18767 qsd_id_glimpse_ast might_sleep due to multiple layers of locking
• is related to: LU-18009 Remove o_fid_init/o_fid_fini

Comments

[Andreas Dilger]

We are using the janitor9x label for tickets in Jira and as a hashtag in Gerrit to track issues related to this ticket.

[Peter Jones]

As per discussion on the LWG call today, moving tickets that do not appear to be essential to fix version 2.17. If the fix lands before code freeze we will update the fix version to reflect that but we want to focus on activities on the critical path. Please speak up if you think that this issue definitely needs to be fixed before we could issue a 2.16 release.

[James A Simmons]

Are you seeing other lockdep issues beside LU-15808 and LU-16783?

[Alex Zhuravlev]

well, even with ldiskfs lockdep turns off after the very first mount.

[James A Simmons]

The ZFS work hasn't started. Technically lockdep doesn't work with ZFS due to license issues so no one is really looking at it. I was wanting to see if RHEL9 had the fix for ldiskfs quota but that is not the case. 

[Alex Zhuravlev]

simmonsja please tell the current status? are you able to run Lustre+RHEL8 with lockdep enabled using the patches above?

[James A Simmons]

ZFS lockdep issue is here: https://github.com/openzfs/zfs/issues/15111

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/51192/
Subject: LU-16807 llite: make lsm_sem nested for ll_update_dir_depth
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: fe44198b7fb710fce0f3daf669952c11ea0f4189

[Gerrit Updater]

"James Simmons <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/51192
Subject: LU-16807 llite: make lsm_sem nested for ll_update_dir_depth
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 9de949e168d61c549b78ca0077af481feedea24b

[James A Simmons]

Several more patches are needed.