[LU-9859] libcfs simplification

• Type: Improvement
• Resolution: Unresolved
• Priority: Minor
• Fix Version/s: Upstream
• Affects Version/s: Lustre 2.11.0
• Labels: patch
• Assignee: James A Simmons
• Reporter: James A Simmons

Description

Now that libcfs has been sorted out with user land we can look to simplify the code and use kernel function that already exist. Good example are prng.c or the cfs_alloc_array_* functions. Also only place in the linux directory only code that is needed to support various distros.

Issue Links

• is related to: LU-17053 t-f's log() needs lnet
• is related to: LU-12838 ptlrpc watchdog ratelimiting is broken
• is related to: LU-14463 stack traces not being printed to console on RHEL8
• is related to: LU-17788 restore ptlrpc watchdog timer message
• is related to: LU-8304 libcfs_debug_dumplog calls potentially sleeping function in wrong state
• is related to: LU-12952 oops in cfs_trace_copyin_string
• is related to: LU-12923 replace CLASSERT() with BUILD_BUG_ON()
• is related to: LU-12511 Prepare lustre for adoption into the linux kernel
• is related to: LU-14290 Convert fault-injection framework to match the model using in linux
• is related to: LU-14428 Convert tracefile to use ring_buffer from linux
• is related to: LU-16796 Replace LASSERT_ATOMIC_* (atomic_t) with reference counters (refcount)
• is related to: LU-17242 Clean up and Improve Lustre Debugging
• is related to: LU-9019 Migrate lustre to standard 64 bit time kernel API
• is related to: LU-4922 Fix cfs_str2mask to better deal with garbage input
• is related to: LU-13258 Bind linux workqueues to specific core
• is related to: LU-14289 Restrict libcfs to compatibility code only
• is related to: LU-8703 rework CPU partition code
• is related to: LU-14680 remove strict dependency between user tools and kernel RPMs

Comments

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/55187/
Subject: LU-9859 libcfs: move crypto wrappers to lnet
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: a4d358a94496ffb4cd628f3dc10efa712c416f64

[Gerrit Updater]

"James Simmons <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/55187
Subject: LU-9859 libcfs: move crypto wrappers to lnet
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 7d9f12e442b01aa097d5a3abe048212ea05b4539

[James A Simmons]

Yes but it is mostly done.

[Peter Jones]

James

Is there still more to come here?

Peter

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/50845/
Subject: LU-9859 lnet: move expr parsing from libcfs
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 84d3585890c4758178cedbef0e4f032bf0eeaa03

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/52923/
Subject: LU-9859 lnet: move CPT handling to LNet
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 7f8cde3b77ada95e8b96dee1996f8d40bd17a538

[Gerrit Updater]

"James Simmons <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/54178
Subject: LU-9859 libcfs: move libcfs hash to obdclass
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 44901b543321d443a5c83a363fa28c5cc9eee2a1

[Gerrit Updater]

"James Simmons <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/54177
Subject: LU-9859 libcfs: move libcfs hash to obdclass
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 2da0309e3481134f25bd72a57c5e15c1e808e8e9

[James A Simmons]

Yes I can reproduce it. Also patch https://review.whamcloud.com/c/fs/lustre-release/+/41664 resolves this

[James A Simmons]

Thanks. Let me try to reproduce this problem.

[Vladimir Saveliev]

Did you change anything in your cfg/local.sh file to expose this bug?

It seems that something got broken with libcfs module parameters.
I have libcfs_debug defined in /etc/modprobe.d/lustre.conf:

[root@vmcentos7-1 ~]# cat /etc/modprobe.d/lustre.conf
options lnet networks=tcp0(ens33)
options libcfs libcfs_debug=33965841

Then with

[root@vmcentos7-1 ~]# modprobe libcfs
[root@vmcentos7-1 ~]# rmmod libcfs

I get:

[   50.390835] BUG: unable to handle kernel NULL pointer dereference at 00000000000000cc
[   50.393177] IP: [<ffffffffc07933fa>] cfs_trace_lock_tcd+0xa/0x80 [libcfs]
...
[   50.422078] CPU: 1 PID: 2517 Comm: rmmod Kdump: loaded Tainted: G           OE  ------------   3.10.0-1160.88.1.el7.x86_64 #4
...
[   50.446143] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[   50.447885] Call Trace:
[   50.448517]  [<ffffffffc0793837>] libcfs_debug_msg+0xd7/0xa30 [libcfs]
[   50.449904]  [<ffffffff9d5621ed>] ? call_rcu_sched+0x1d/0x30
[   50.450999]  [<ffffffff9d680af1>] ? mnt_get_count+0x51/0x70
[   50.452449]  [<ffffffff9d688935>] ? simple_release_fs+0x45/0x50
[   50.453652]  [<ffffffffc07a1692>] ? libcfs_exit+0x27/0x995 [libcfs]
[   50.455271]  [<ffffffffc07a1706>] libcfs_exit+0x9b/0x995 [libcfs]
[   50.456848]  [<ffffffff9d52329e>] SyS_delete_module+0x19e/0x320