Re: [PATCH v13 2/9] f2fs: Simplify the handling of cached insensitive names

[Eugen Hristev]

On 3/14/24 16:41, Gabriel Krisman Bertazi wrote:
> Eugen Hristev <eugen.hristev@xxxxxxxxxxxxx> writes:
> 
>>> Please, make sure you actually stress test this patchset with fstests
>>> against both f2fs and ext4 before sending each new version.
>>
>> I did run the xfstests, however, maybe I did not run the full suite, or maybe I am
>> running it in a wrong way ?
> 
> No worries.  Did you manage to reproduce it?

Yes, thank you, using qemu on the x86_64 with your commands below.

While the oops was caused by that wrong kfree call, fixing it and moving further
got me into further problems.
I am unsure though how these patches cause it.

Here is a snippet of the problem that occurs now :

generic/417 12s ... [  616.265444] run fstests generic/417 at 2024-03-18 09:22:48
[  616.768435] ------------[ cut here ]------------
[  616.769493] WARNING: CPU: 4 PID: 133 at block/blk-merge.c:580
__blk_rq_map_sg+0x46a/0x480
[  616.771253] Modules linked in:
[  616.771873] CPU: 4 PID: 133 Comm: kworker/4:1H Not tainted
6.7.0-09941-g554c4640dff5 #18
[  616.773660] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.15.0-1
04/01/2014
[  616.775573] Workqueue: kblockd blk_mq_run_work_fn
[  616.776570] RIP: 0010:__blk_rq_map_sg+0x46a/0x480
[  616.777547] Code: 17 fe ff ff 44 89 58 0c 48 8b 01 e9 ec fc ff ff 43 8d 3c 06 48
8b 14 24 81 ff 00 10 00 00 0f 86 af fc ff ff e9 02 fe ff ff 90 <0f> 0b 90 e9 76 fd
ff ff 90 0f 0b 90 0f 0b 0f 1f 84 00 00 00 00 00
[  616.781245] RSP: 0018:ffff97e4804f3b98 EFLAGS: 00010216
[  616.782322] RAX: 000000000000005e RBX: 0000000000000f10 RCX: ffff8f5701eed000
[  616.783929] RDX: ffffdc0c4052df82 RSI: 0000000000001000 RDI: 00000000fffffffc
[  616.785426] RBP: 000000000000005e R08: 0000000000000000 R09: ffff8f5702120000
[  616.787065] R10: ffffdc0c4052df80 R11: 0000000000000000 R12: ffff8f5702118000
[  616.788650] R13: 0000000000000000 R14: 0000000000001000 R15: ffffdc0c4052df80
[  616.790129] FS:  0000000000000000(0000) GS:ffff8f577db00000(0000)
knlGS:0000000000000000
[  616.791826] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  616.793069] CR2: 00007fbe596adc98 CR3: 000000001162e000 CR4: 0000000000350ef0
[  616.794528] Call Trace:
[  616.795093]  <TASK>
[  616.795541]  ? __warn+0x7f/0x130
[  616.796242]  ? __blk_rq_map_sg+0x46a/0x480
[  616.797101]  ? report_bug+0x199/0x1b0
[  616.797843]  ? handle_bug+0x3d/0x70
[  616.798656]  ? exc_invalid_op+0x18/0x70
[  616.799461]  ? asm_exc_invalid_op+0x1a/0x20
[  616.800313]  ? __blk_rq_map_sg+0x46a/0x480
[  616.801207]  ? __blk_rq_map_sg+0xfc/0x480
[  616.802213]  scsi_alloc_sgtables+0xae/0x2b0
[  616.803258]  sd_init_command+0x181/0x860
[  616.804111]  scsi_queue_rq+0x7c3/0xae0
[  616.804910]  blk_mq_dispatch_rq_list+0x2bf/0x7c0
[  616.805962]  __blk_mq_sched_dispatch_requests+0x40a/0x5c0
[  616.807226]  blk_mq_sched_dispatch_requests+0x34/0x60
[  616.808389]  blk_mq_run_work_fn+0x5f/0x70
[  616.809332]  process_one_work+0x136/0x2f0
[  616.810268]  ? __pfx_worker_thread+0x10/0x10
[  616.811320]  worker_thread+0x2ef/0x400
[  616.812215]  ? __pfx_worker_thread+0x10/0x10
[  616.813205]  kthread+0xd5/0x100
[  616.813907]  ? __pfx_kthread+0x10/0x10
[  616.814787]  ret_from_fork+0x2f/0x50
[  616.815598]  ? __pfx_kthread+0x10/0x10
[  616.816394]  ret_from_fork_asm+0x1b/0x30
[  616.817210]  </TASK>
[  616.817658] ---[ end trace 0000000000000000 ]---
[  616.818687] ------------[ cut here ]------------
[  616.819697] kernel BUG at drivers/scsi/scsi_lib.c:1068!


Do you have any ideas ?

Thanks !
Eugen

> 
>> How are you running the kvm-xfstests with qemu ? Can you share your command
>> arguments please ?
> 
> I don't use kvm-xfstests.  I run ./check directly:
> 
> export SCRATCH_DEV=/dev/loop1
> export SCRATCH_MNT=$BASEMNT/scratch
> export TEST_DEV=/dev/loop0
> export TEST_DIR=$BASEMNT/test
> export RESULT_BASE=${BASEMNT}/results
> export REPORT_DIR=${BASEMNT}/report
> export FSTYP=f2fs
> 
> mkfs.f2fs -f -C utf8 -O casefold ${TEST_DEV}
> 
> ./check -g encrypt,quick
>