5.10 code - race in the HDA codec probe when SOF probe fails (no video driver bound)

[perexg]

We have got the oops bellow on the Tigerlake machine with the current 5.10 code when the video driver is not bound for a reason. It seems that the hdac_ext_link structure is corrupted (late access?) on probe or so. Any idea?

[  111.582047] sof-audio-pci 0000:00:1f.3: couldn't bind with audio component 
[  111.589044] sof-audio-pci 0000:00:1f.3: init of i915 and HDMI codec failed 
[  111.598257] sof-audio-pci 0000:00:1f.3: use msi interrupt mode 
[  111.609641] sof-audio-pci 0000:00:1f.3: codec #2 probe error, ret: -2 
[  111.616492] sof-audio-pci 0000:00:1f.3: No SoundWire machine driver found 
[  111.623374] sof-audio-pci 0000:00:1f.3: no hda codecs found! 
[  111.629127] sof-audio-pci 0000:00:1f.3: warning: No matching ASoC machine driver found 
[  111.637137] sof-audio-pci 0000:00:1f.3: error: no matching ASoC machine driver found - aborting probe 
[  111.646454] sof-audio-pci 0000:00:1f.3: error: failed to get machine info -19 
[  111.653938] sof-audio-pci 0000:00:1f.3: error: sof_probe_work failed err: -19 
[  111.653991] BUG: unable to handle kernel paging request at ffffaf3542098c84 
[  111.668180] PGD 107c7b067 P4D 107c7b067 PUD 107c7c067 PMD 487f49067 PTE 0 
[  111.675043] Oops: 0000 [#1] SMP NOPTI 
[  111.678768] CPU: 2 PID: 20629 Comm: systemd-udevd Kdump: loaded Not tainted 5.10.x86_64 #1 
[  111.687988] Hardware name: Intel Corporation Tiger Lake Client Platform/TigerLake Y LPDDR4x T4 Crb, BIOS TGLSFWI1.R00.3313.A02.2007300246 07/30/2020 
[  111.701387] RIP: 0010:check_hdac_link_power_active+0x3a/0x60 [snd_hda_ext_core] 
[  111.708776] Code: 89 f5 53 bb 96 00 00 00 e8 a3 7c de e4 eb 16 a9 00 00 80 00 75 23 bf 55 32 00 00 e8 90 7c de e4 83 eb 01 74 1b 49 8b 44 24 10 <8b> 40 04 40 84 ed 75 dd a9 00 00 80 00 75 dd 5b 31 c0 5d 41 5c c3 
[  111.727659] RSP: 0018:ffffaf3547e7fb48 EFLAGS: 00010202 
[  111.732952] RAX: ffffaf3542098c80 RBX: 0000000000000085 RCX: 0000004456f80453 
[  111.740159] RDX: 0000000000001af3 RSI: 0000000000000002 RDI: 00000000000019a7 
[  111.747375] RBP: 0000000000000001 R08: 0000000000000001 R09: 0000000000000002 
[  111.754585] R10: 000000000000000a R11: f000000000000000 R12: ffff9661fa087100 
[  111.761799] R13: 0000000000000000 R14: 0000000000000034 R15: 0000000000000000 
[  111.769017] FS:  00007f0f1184c980(0000) GS:ffff96621f880000(0000) knlGS:0000000000000000 
[  111.777204] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033 
[  111.783016] CR2: ffffaf3542098c84 CR3: 0000000477622006 CR4: 0000000000760ee0 
[  111.790232] PKRU: 55555554 
[  111.793002] Call Trace: 
[  111.795519]  snd_hdac_ext_bus_link_get+0x70/0xf0 [snd_hda_ext_core] 
[  111.801865]  hdac_hda_dev_probe+0x3c/0xc0 [snd_soc_hdac_hda] 
[  111.807592]  really_probe+0x212/0x440 
[  111.811325]  driver_probe_device+0x49/0xc0 
[  111.815489]  device_driver_attach+0x50/0x60 
[  111.819737]  __driver_attach+0x61/0x130 
[  111.823639]  ? device_driver_attach+0x60/0x60 
[  111.828058]  bus_for_each_dev+0x77/0xc0 
[  111.831971]  ? klist_add_tail+0x3b/0x70 
[  111.835895]  bus_add_driver+0x14d/0x1e0 
[  111.839799]  ? 0xffffffffc04df000 
[  111.843176]  driver_register+0x6b/0xb0 
[  111.846987]  ? 0xffffffffc04df000 
[  111.850368]  do_one_initcall+0x46/0x1c3 
[  111.854282]  ? do_init_module+0x22/0x220 
[  111.858267]  ? preempt_schedule_common+0xa/0x20 
[  111.862871]  ? _cond_resched+0x1d/0x30 
[  111.866683]  ? kmem_cache_alloc_trace+0x140/0x1c0 
[  111.871464]  do_init_module+0x5a/0x220 
[  111.875278]  load_module+0x14b4/0x17e0 
[  111.879092]  ? __do_sys_init_module+0x13b/0x180 
[  111.883686]  __do_sys_init_module+0x13b/0x180 
[  111.888109]  do_syscall_64+0x5b/0x1a0 
[  111.891836]  entry_SYSCALL_64_after_hwframe+0x65/0xca 
[  111.896954] RIP: 0033:0x7f0f1043e1ae 
[  111.900601] Code: 48 8b 0d dd ec 2b 00 f7 d8 64 89 01 48 83 c8 ff c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa 49 89 ca b8 af 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d aa ec 2b 00 f7 d8 64 89 01 48 
[  111.919482] RSP: 002b:00007ffc3776fce8 EFLAGS: 00000246 ORIG_RAX: 00000000000000af 
[  111.927128] RAX: ffffffffffffffda RBX: 000056284670e350 RCX: 00007f0f1043e1ae 
[  111.934344] RDX: 00007f0f10fa982d RSI: 0000000000023370 RDI: 0000562846823b90 
[  111.941561] RBP: 00007f0f10fa982d R08: 000056284669401a R09: 0000000000000006 
[  111.948771] R10: 0000562846694010 R11: 0000000000000246 R12: 0000562846823b90 
[  111.955991] R13: 0000562846707390 R14: 0000000000020000 R15: 0000000000000000 

[keyonjie]

Thanks for reporting @perexg This looks a valid issue which I actually saw randomly on TGL platform.
@kv2019i any comments on this?

[kv2019i]

I'll take a look at this. The following -ENOENT case is a good hint:
[ 111.609641] sof-audio-pci 0000:00:1f.3: codec #2 probe error, ret: -2

I think this is related to "ASoC: SOF: fix a runtime pm issue in SOF when HDMI codec doesn't work". The patch addresses a problem where we do not clean up the hdac device after error and this blocks runtime-pm. But in this case, the cleanup is done too early.

[kv2019i]

I have a tentative fix, but it needs some more testing before ready for submission. I was tempted to clean up some of the related code, but for stable we're probably better off with a more clean and smaller fix.

[kv2019i]

Patch in sof-dev and sent to upstream, closing.