[BUG] kernel oops when testing kmod on TGLU_UP_HDA-ipc4

[keqiaozhang]

Describe the bug
Observed this issue in CI daily test. It happens on TGLU_UP_HDA-IPC4 platform, the reproduction rate is almost 100%. It should be a regression, I will do the bisection next.

dmesg

[ 4679.663002] kernel: snd_sof_intel_hda_common:hda_init_caps: sof-audio-pci-intel-tgl 0000:00:1f.3: skipping SoundWire, no links enabled
[ 4679.663139] kernel: snd_sof_intel_hda:hda_codec_probe: sof-audio-pci-intel-tgl 0000:00:1f.3: HDA codec #0 probed OK: response: 10ec0888
[ 4679.668197] kernel: BUG: kernel NULL pointer dereference, address: 0000000000000078
[ 4679.668202] kernel: #PF: supervisor read access in kernel mode
[ 4679.668204] kernel: #PF: error_code(0x0000) - not-present page
[ 4679.668205] kernel: PGD 0 P4D 0 
[ 4679.668207] kernel: Oops: 0000 [#1] PREEMPT SMP NOPTI
[ 4679.668210] kernel: CPU: 3 PID: 61983 Comm: systemd-udevd Not tainted 6.7.0-rc2-g266ee31b71f8 #dev

To Reproduce
~/sof-test/test-case/check-kmod-load-unload.sh -l 25

Reproduction Rate
100%

Environment

1. Branch name and commit hash of the 2 repositories: sof (firmware/topology) and linux (kernel driver).
   • Kernel: {266ee31b71f8}
   • SOF: {e72056769c8b}
2. Name of the topology file
   • Topology: {sof-hda-generic-4ch.tplg}
3. Name of the platform(s) on which the bug is observed.
   • Platform: {TGLU_UP_HDA-ipc4}

dmesg.txt

[ujfalusi]

Just a hunch:
3be9a02 ("fixup! ASoC: hdac_hda: Conditionally register dais for HDMI and Analog")

diff --git a/sound/soc/codecs/hdac_hda.c b/sound/soc/codecs/hdac_hda.c
index 72916a387ad6..b075689db2dc 100644
--- a/sound/soc/codecs/hdac_hda.c
+++ b/sound/soc/codecs/hdac_hda.c
@@ -619,6 +619,7 @@ static const struct snd_soc_component_driver hdac_hda_hdmi_codec = {
 
 static int hdac_hda_dev_probe(struct hdac_device *hdev)
 {
+	struct hdac_hda_priv *hda_pvt = dev_get_drvdata(&hdev->dev);
 	struct hdac_ext_link *hlink;
 	int ret;
 
@@ -631,7 +632,7 @@ static int hdac_hda_dev_probe(struct hdac_device *hdev)
 	snd_hdac_ext_bus_link_get(hdev->bus, hlink);
 
 	/* ASoC specific initialization */
-	if (snd_hda_device_is_hdmi(hdev))
+	if (hda_pvt->need_display_power)
 		ret = devm_snd_soc_register_component(&hdev->dev,
 						&hdac_hda_hdmi_codec, hdac_hda_hdmi_dais,
 						ARRAY_SIZE(hdac_hda_hdmi_dais));

but I cannot reproduce locally.

[ujfalusi]

I can:
rmmod snd_sof_pci_intel_tgl
modprobe snd_sof_pci_intel_tgl

[ujfalusi]

it is not that commit, the callstack:

[ 4679.668236] kernel:  ? __die+0x24/0x70
[ 4679.668241] kernel:  ? page_fault_oops+0x15b/0x440
[ 4679.668244] kernel:  ? find_held_lock+0x32/0x90
[ 4679.668247] kernel:  ? snd_hdac_ext_bus_link_get+0x2b/0xe0 [snd_hda_ext_core]
[ 4679.668253] kernel:  ? exc_page_fault+0x69/0x1a0
[ 4679.668256] kernel:  ? asm_exc_page_fault+0x26/0x30
[ 4679.668261] kernel:  ? hdac_hda_dev_probe+0x48/0xf0 [snd_soc_hdac_hda]
[ 4679.668265] kernel:  really_probe+0x1a2/0x410

snd_hdac_ext_bus_link_get

[ujfalusi]

I have a tentative fix for it, I'll create a PR before sending upstream.