[LU-14286] fallocate() does not zero on-disk data

• Type: Bug
• Resolution: Fixed
• Priority: Blocker
• Fix Version/s: Lustre 2.14.0
• Affects Version/s: Lustre 2.14.0
• Severity: 3
• Assignee: Arshad Hussain
• Reporter: Arshad Hussain

Description

The ldiskfs osd_fallocate() does not correctly zero the allocated blocks. This is not noticed during autotest and local developer testing because the underlying storage is zero-filled due to using a sparse backing file and will usually read back as zero. Also, the fsx data correctness test was not properly detecting fallocate() support, and as a result this behavior was not detected until after fsx was fixed.

fsx uses test_fallocate() to determine if fallocate() along with its various FLAGS is enabled and could be called. This was discovered while working on LU-14160 that test_fallocate() always returned false as a bug. Resulting is fallocate() being skipped always for fsx runs.

Issue Links

• is related to: LU-14326 sanity-dom test_fsx: crash in osc_extent_make_ready()
• is related to: LU-14333 ASSERTION( oinfo->loi_lvb.lvb_size >= oinfo->loi_kms ) failed
• is related to: LU-14324 fsx fallocate detection is buggy
• is related to: LUDOC-487 Document llite.*.fallocate_zero_blocks tunable

Comments

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/41216/
Subject: LU-14286 osd-ldiskfs: don't read unwritten blocks
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 42c52069cada6837df34a02638717dff709e5a0e

[Gerrit Updater]

Alex Zhuravlev ([e-mail]) uploaded a new patch: https://review.whamcloud.com/41216
Subject: LU-14286 osd: don't read unwritten blocks
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 065e10770fc082ba8ec80aed21a797a34faaaab4

[Gerrit Updater]

Andreas Dilger ([e-mail]) merged in patch https://review.whamcloud.com/41211/
Subject: LU-14286 tests: reduce fallocate space sanity 150e
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: b9274179adf0939909ff2dd78d077515a64fd25f

[Gerrit Updater]

James Nunez ([e-mail]) uploaded a new patch: https://review.whamcloud.com/41211
Subject: LU-14286 tests: reduce fallocate space sanity 150e
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 4be24e98c0d1f1a08c9ca03444a3cb682017303c

[Andreas Dilger]

Still not fixed.

[Gerrit Updater]

Andreas Dilger ([e-mail]) uploaded a new patch: https://review.whamcloud.com/41204
Subject: LU-14286 osd-ldiskfs: fallocate with unwritten extents
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 44c56a496d80a8607d29681ade498fe4fbf85a59

[Peter Jones]

Landed for 2.14

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/41119/
Subject: LU-14286 osd-ldiskfs: fallocate() should zero new blocks
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 72617588ac8cb2e3e5a7b8e5ebc201cab524d938

[Andreas Dilger]

Arshad, I've recharacterized this ticket to be more about the osd_fallocate() bug rather than the fsx bug, since it is essentially a data corruption issue that needs to be fixed as part of 2.14.0, and since this patch was already fixing the problem. Rather than spin up a new patch with the same fix, I instead changed the commit message to make it clear the importance of the patch and the reason that it is being included in 2.14.0 at this late date.

[Gerrit Updater]

Arshad Hussain ([e-mail]) uploaded a new patch: https://review.whamcloud.com/41119
Subject: LU-14286 fsx: 'fsx' calling test_fallocate() always returns false
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 80f2acf55464ebc52f15872223d49481810ffdb0