Whamcloud - gitweb
LU-11986 lnet: don't read debugfs lnet stats when shutting down 04/39404/7
authorJames Simmons <jsimmons@infradead.org>
Sun, 30 Aug 2020 16:45:42 +0000 (12:45 -0400)
committerOleg Drokin <green@whamcloud.com>
Tue, 8 Sep 2020 18:08:05 +0000 (18:08 +0000)
commitf53eea15d470c9bb29a7b867b733db9249aec95b
treea48a3bbb3a8cb839a0c519f64d7c71973656410a
parentad444ed9836320c6ae8b770ff96edd6b0fe4f0d4
LU-11986 lnet: don't read debugfs lnet stats when shutting down

A race exist on shutdown with an external application reading
the debugfs file containing lnet stats which causes an kernel
crash.

[  257.192117] BUG: unable to handle kernel paging request at fffffffffffffff0
[  257.194859] IP: [<ffffffffc0bb95c6>] cfs_percpt_number+0x6/0x10 [libcfs]
[  257.196863] PGD 7c14067 PUD 7c16067 PMD 0
[  257.198665] Oops: 0000 [#1] SMP
[  257.200431] Modules linked in: ksocklnd(OE) lnet(OE) libcfs(OE) dm_service_time iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi sunrpc zfs(POE) zunicode(POE) zavl(POE) icp(POE) zcommon(POE) znvpair(POE) spl(OE) ppdev iosf_mbi crc32_pclmul crc32c_intel ghash_clmulni_intel aesni_intel lrw gf128mul glue_helper ablk_helper cryptd pcspkr sg e1000 video parport_pc parport i2c_piix4 dm_multipath dm_mod ip_tables xfs libcrc32c sd_mod crc_t10dif crct10dif_generic ata_generic pata_acpi crct10dif_pclmul crct10dif_common ata_piix serio_raw libata [last unloaded: obdclass]
[  257.222895] CPU: 0 PID: 7331 Comm: lctl Tainted: P           OE  ------------   3.10.0-957.el7_lustre.x86_64 #1
[  257.229312] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006
[  257.233659] task: ffff9c9fbaf15140 ti: ffff9c9fbabcc000 task.ti: ffff9c9fbabcc000
[  257.238388] RIP: 0010:[<ffffffffc0bb95c6>]  [<ffffffffc0bb95c6>] cfs_percpt_number+0x6/0x10 [libcfs]
[  257.243851] RSP: 0018:ffff9c9fbabcfdb0  EFLAGS: 00010296
[  257.246400] RAX: 0000000000000000 RBX: ffff9c9fba2a5200 RCX: 0000000000000000
[  257.250304] RDX: 0000000000000001 RSI: 00000000ffffffff RDI: 0000000000000000
[  257.253677] RBP: ffff9c9fbabcfdd0 R08: 000000000001f120 R09: ffff9c9fbe001700
[  257.257073] R10: ffffffffc0c376db R11: 0000000000000246 R12: 0000000000000000
[  257.260339] R13: 0000000000000000 R14: 0000000000001000 R15: ffff9c9fba2a5200
[  257.263204] FS:  00007fbdc89c6740(0000) GS:ffff9c9fbfc00000(0000) knlGS:0000000000000000
[  257.266409] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  257.269105] CR2: fffffffffffffff0 CR3: 0000000022e36000 CR4: 00000000000606f0
[  257.272529] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  257.275209] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[  257.277936] Call Trace:
[  257.279245]  [<ffffffffc0c0a88b>] ? lnet_counters_get_common+0xeb/0x150 [lnet]
[  257.283071]  [<ffffffffc0c0a95c>] lnet_counters_get+0x6c/0x150 [lnet]
[  257.286224]  [<ffffffffc0c3771b>] __proc_lnet_stats+0xfb/0x810 [lnet]
[  257.288975]  [<ffffffffc0ba6602>] lprocfs_call_handler+0x22/0x50 [libcfs]
[  257.292387]  [<ffffffffc0c36bf5>] proc_lnet_stats+0x25/0x30 [lnet]
[  257.295184]  [<ffffffffc0ba665d>] lnet_debugfs_read+0x2d/0x40 [libcfs]

The solution is to only allow reading of the lnet stats when the
lnet state is LNET_STATE_RUNNING.

Test-Parameters: trivial testlist=sanity-lnet
Change-Id: I8720a51ec358e4f6ae121acb34cc23020054ab84
Signed-off-by: James Simmons <jsimmons@infradead.org>
Reviewed-on: https://review.whamcloud.com/39404
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Chris Horn <chris.horn@hpe.com>
Reviewed-by: Serguei Smirnov <ssmirnov@whamcloud.com>
Reviewed-by: Nathaniel Clark <nclark@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
lnet/include/lnet/lib-lnet.h
lnet/lnet/api-ni.c
lnet/lnet/router_proc.c