[LU-18170] conf-sanity test_123ad: 'llog print only showed 92/93 records'

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.17.0
• Severity: 3
• Assignee: Etienne Aujames
• Reporter: Maloo

Description

This issue was created by maloo for emoly <[e-mail]>

This issue relates to the following test suite run: https://testing.whamcloud.com/test_sets/46914488-02f0-47cd-89b0-ab0907e5fe34

Test session details:
clients: https://build.whamcloud.com/job/lustre-reviews/106986 - 5.14.0-362.24.1.el9_3.x86_64
servers: https://build.whamcloud.com/job/lustre-reviews/106986 - 5.14.0-362.24.1_lustre.el9.x86_64

This issue is very easy to reproduce by setting ONLY_REPEAT to n(n>=10).

VVVVVVV DO NOT REMOVE LINES BELOW, Added by Maloo for auto-association VVVVVVV
conf-sanity test_123ad - llog print only showed 92/93 records

Issue Links

• is related to: LU-18764 Crash in conf-sanity 123H in interop with 2.16.1
• is related to: LU-18114 lctl and lfs command "groups" should be split into subcommands

Comments

[Andreas Dilger]

Patch https://review.whamcloud.com/58366 ("LU-18764 tests: fix conf-sanity 123H for interop testing") is fixing the test that was added in this ticket.

[Oleg Drokin]

I think we need to add an interop check to test 123H

(first hit Feb 18) we get this hit regularly now with 2.16.1 servers

[27373.539546] Lustre: DEBUG MARKER: /usr/sbin/lctl mark == conf-sanity test 123H: check concurent accesses with \'lctl llog_print ========================================================== 22:07:19 \(1740780439\)
[27373.780325] Lustre: DEBUG MARKER: == conf-sanity test 123H: check concurent accesses with 'lctl llog_print ========================================================== 22:07:19 (1740780439)
[27438.099364] Lustre: DEBUG MARKER: /usr/sbin/lctl llog_print -r params
[27438.542994] Lustre: DEBUG MARKER: seq 1 20 |  xargs -P20 -I{} bash -c '/usr/sbin/lctl llog_print -r params | wc -l' |  sort | uniq -c
[27438.824080] ------------[ cut here ]------------
[27438.824088] WARNING: CPU: 0 PID: 969973 at lib/vsprintf.c:2726 vsnprintf+0x4bf/0x570
[27438.825864] Modules linked in: xfs libcrc32c ofd(OE) ost(OE) loop osp(OE) mdd(OE) lod(OE) mdt(OE) lfsck(OE) mgs(OE) osd_ldiskfs(OE) ldiskfs(OE) lquota(OE) lustre(OE) obdecho(OE) mgc(OE) mdc(OE) lov(OE) osc(OE) lmv(OE) fid(OE) fld(OE) ptlrpc_gss(OE) ptlrpc(OE) obdclass(OE) ksocklnd(OE) lnet(OE) libcfs(OE) dm_flakey tls dm_mod rpcsec_gss_krb5 auth_rpcgss nfsv4 dns_resolver nfs lockd grace fscache netfs rfkill sunrpc intel_rapl_msr intel_rapl_common i2c_piix4 virtio_balloon pcspkr joydev drm fuse ext4 mbcache jbd2 ata_generic crct10dif_pclmul crc32_pclmul crc32c_intel ata_piix libata virtio_net ghash_clmulni_intel virtio_blk net_failover failover serio_raw [last unloaded: libcfs]
[27438.832027] CPU: 0 PID: 969973 Comm: llog_process_th Kdump: loaded Tainted: G           OE     -------  ---  5.14.0-427.31.1_lustre.el9.x86_64 #1
[27438.833428] Hardware name: Red Hat KVM, BIOS 0.5.1 01/01/2011
[27438.834084] RIP: 0010:vsnprintf+0x4bf/0x570
[27438.834603] Code: 44 24 10 8b 03 4d 8d 7c 0f 01 31 c9 83 f8 2f 0f 87 50 ff ff ff e9 08 fc ff ff 89 d0 83 c2 08 48 03 43 10 89 13 e9 98 fe ff ff <0f> 0b e9 cc fc ff ff 41 0f b6 04 24 4d 89 e6 e9 c2 fd ff ff 83 fa
[27438.836542] RSP: 0000:ffffa1ccc2d33c70 EFLAGS: 00010292
[27438.837140] RAX: 0000000000000000 RBX: ffff92c10181b824 RCX: ffffa1ccc2d33cc8
[27438.837954] RDX: ffffffffc0b75150 RSI: fffffffffffffffd RDI: ffff92c10181b869
[27438.838762] RBP: ffffa1ccc2d33d18 R08: 000000000000000a R09: ffff92c20181b861
[27438.839566] R10: ffffffffffffffff R11: 000000000000000f R12: 0000000000000007
[27438.840386] R13: ffff92c10181b869 R14: ffff92c1014736bf R15: 0000000000000510
[27438.841176] FS:  0000000000000000(0000) GS:ffff92c17fc00000(0000) knlGS:0000000000000000
[27438.842055] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[27438.842718] CR2: 0000560664894008 CR3: 0000000033b90001 CR4: 00000000000606f0
[27438.843521] Call Trace:
[27438.843847]  <TASK>
[27438.844136]  ? show_trace_log_lvl+0x1c4/0x2df
[27438.844708]  ? show_trace_log_lvl+0x1c4/0x2df
[27438.845241]  ? snprintf+0x49/0x70
[27438.845662]  ? vsnprintf+0x4bf/0x570
[27438.846095]  ? __warn+0x81/0x110
[27438.846558]  ? vsnprintf+0x4bf/0x570
[27438.846992]  ? report_bug+0x10a/0x140
[27438.847474]  ? handle_bug+0x3c/0x70
[27438.847939]  ? exc_invalid_op+0x14/0x70
[27438.848422]  ? asm_exc_invalid_op+0x16/0x20
[27438.848945]  ? vsnprintf+0x4bf/0x570
[27438.849413]  snprintf+0x49/0x70
[27438.849814]  class_config_yaml_output+0x3b6/0x600 [obdclass]
[27438.850874]  llog_print_cb+0x5e2/0x670 [obdclass]
[27438.851516]  llog_process_thread+0xd97/0x1920 [obdclass]
[27438.852185]  ? __pfx_llog_process_thread_daemonize+0x10/0x10 [obdclass]
[27438.852977]  llog_process_thread_daemonize+0x99/0xe0 [obdclass]
[27438.853715]  kthread+0xe0/0x100
[27438.854127]  ? __pfx_kthread+0x10/0x10
[27438.854601]  ret_from_fork+0x2c/0x50
[27438.855082]  </TASK>
[27438.855421] ---[ end trace 708d17dd91c9a88b ]---
[27438.988112] LustreError: 970099:0:(llog_ioctl.c:296:llog_print_cb()) not enough space for print log records
[27439.098846] LustreError: 970208:0:(llog.c:525:llog_process_thread()) ASSERTION( is_power_of_2(chunk_size) ) failed:
[27439.098852] LustreError: 970208:0:(llog.c:525:llog_process_thread()) LBUG
[27439.098855] CPU: 0 PID: 970208 Comm: llog_process_th Kdump: loaded Tainted: G        W  OE     -------  ---  5.14.0-427.31.1_lustre.el9.x86_64 #1
[27439.098858] Hardware name: Red Hat KVM, BIOS 0.5.1 01/01/2011
[27439.098859] Call Trace:
[27439.098883]  <TASK>
[27439.098886]  dump_stack_lvl+0x34/0x48
[27439.098914]  lbug_with_loc.cold+0x5/0x58 [libcfs]
[27439.098990]  llog_process_thread+0xa9e/0x1920 [obdclass]
[27439.099065]  ? __pfx_llog_process_thread_daemonize+0x10/0x10 [obdclass]
[27439.099120]  ? mgs_key_init+0xa8/0x130 [mgs]
[27439.099200]  ? __pfx_llog_process_thread_daemonize+0x10/0x10 [obdclass]
[27439.099248]  llog_process_thread_daemonize+0x99/0xe0 [obdclass]
[27439.099297]  kthread+0xe0/0x100
[27439.099301]  ? __pfx_kthread+0x10/0x10
[27439.099304]  ret_from_fork+0x2c/0x50
[27439.099309]  </TASK>
[27445.568983] Kernel panic - not syncing: LBUG
[27445.568988] CPU: 0 PID: 970208 Comm: llog_process_th Kdump: loaded Tainted: G        W  OE     -------  ---  5.14.0-427.31.1_lustre.el9.x86_64 #1
[27445.571030] Hardware name: Red Hat KVM, BIOS 0.5.1 01/01/2011
[27445.571676] Call Trace:
[27445.571994]  <TASK>
[27445.572295]  dump_stack_lvl+0x34/0x48
[27445.572739]  panic+0x107/0x2f7
[27445.573157]  lbug_with_loc.cold+0x2f/0x58 [libcfs]
[27445.573725]  llog_process_thread+0xa9e/0x1920 [obdclass]
[27445.574409]  ? __pfx_llog_process_thread_daemonize+0x10/0x10 [obdclass]
[27445.575186]  ? mgs_key_init+0xa8/0x130 [mgs]
[27445.575707]  ? __pfx_llog_process_thread_daemonize+0x10/0x10 [obdclass]
[27445.576486]  llog_process_thread_daemonize+0x99/0xe0 [obdclass]
[27445.577216]  kthread+0xe0/0x100
[27445.577615]  ? __pfx_kthread+0x10/0x10
[27445.578080]  ret_from_fork+0x2c/0x50
[27445.578532]  </TASK>

First hit: https://testing.whamcloud.com/test_sets/3f66b52d-b4d7-4d30-884e-ba82f8340698

more hits:

https://testing.whamcloud.com/test_sets/3a04df19-c183-4e94-85d8-c3f00d71aaf0

https://testing.whamcloud.com/test_sets/db2be02b-8282-4de1-8d1f-a5e5c367695a

https://testing.whamcloud.com/test_sets/2de30589-9fa9-4047-9763-a19616f19601

https://testing.whamcloud.com/test_sets/c32f7d41-35cb-4df1-b64f-dd98d009a41a

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/57354/
Subject: LU-18170 obdclass: replace round_up(_inllenX) with ALIGN
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 4a7db8f29b7f3e9962b5260b3bd51a4f3685dcbb

[Gerrit Updater]

"Etienne AUJAMES <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/57354
Subject: LU-18170 obdclass: replace round_up(_inllenX) with ALIGN
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 3331fd8408b11d04e1bb828bcdb613c6bc451a46

[Peter Jones]

Merged for 2.17

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/56256/
Subject: LU-18170 obdclass: fix llog_print_cb()
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 5ad117fc6f28389cb800953094b5acb0532f702b

[Andreas Dilger]

Etienne, thanks for looking into this. I've also seen this failure a few times.

Config log records are not size fixed, so one record can take several llog indexes (see llog_osd_pad()).

It is possible to have llogs with either fixed or variable-sized records. The llog_osd_pad() function is adding an empty record at the end of the block if it is too small to fit another record. This is done so that the next record will start at the beginning of a new disk block (no partial records crossing block boundaries). The llog file must have a whole record at the start of each block to allow backward and forward seeking.

In this failure case, it looks like there may be a padding record written to fill the block, but this does not get printed by llog_print?

One possible solution on this case is to change the test to add another config record and confirm that the new record is printed properly. This test was to catch a bug when the config llog was truncated incorrectly, but if the new record is printed then that is not the case.