[LU-14805] Do not do parallel DIO except when lockless

• Type: Bug
• Resolution: Fixed
• Priority: Major
• Fix Version/s: Lustre 2.15.0
• Severity: 3
• Assignee: Patrick Farrell
• Reporter: Patrick Farrell

Description

If we do parallel dio when we are not lockless, we release the OSC and LDLM locks early, leading to the following LBUG:

00000008:00020000:3.0:1625183932.837186:0:2823:0:(osc_object.c:419:osc_req_attr_set()) page@00000000e4e028d1[2 000000001b7edf2c 3 2 000000001ecefbe3]
00000008:00020000:3.0:1625183932.839117:0:2823:0:(osc_object.c:419:osc_req_attr_set()) vvp-page@000000000f84e320(0:0) vm@00000000ea4c0139 17ffffc004006c 2:1 0 23127069812 lru
00000008:00020000:3.0:1625183932.840866:0:2823:0:(osc_object.c:419:osc_req_attr_set()) lov-page@00000000b307a48c
00000008:00020000:3.0:1625183932.842047:0:2823:0:(osc_object.c:419:osc_req_attr_set()) osc-page@000000009712b563 152: 1< 0x845fed 258 + + > 2< 622592 0 4096 0x7 0x1aa | 000000001ecefbe3 000000006f42d6bb 000000008b4c789d > 3< 1 5 0 > 4< 2 1 8 283635712 - | - - + - > 5< - - + - | 0 - | 9 - ->
00000008:00020000:3.0:1625183932.844433:0:2823:0:(osc_object.c:419:osc_req_attr_set()) end page@00000000e4e028d1
00000008:00020000:3.0:1625183932.845629:0:2823:0:(osc_object.c:419:osc_req_attr_set()) uncovered page!

Issue Links

• is related to: LU-14843 Allow parallel dio under a group lock (and possibly locked in general?)

Comments

[Peter Jones]

Landed for 2.15

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/44131/
Subject: LU-14805 llite: No locked parallel DIO
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 0f8db7e06abbc341e1ecc6ae164fca7b4a040c4a

[Gerrit Updater]

Patrick Farrell ([e-mail]) uploaded a new patch: https://review.whamcloud.com/44131
Subject: LU-14805 llite: No locked parallel DIO
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 658552c3dd120a9c24474240e8a3237cb8221089