[BUG] ipc tx error on suspend/resume (error loading HDA0.OUT) #5168
Comments
|
@specious how often does this happen ? Fwiw, the changes https://lore.kernel.org/alsa-devel/20211125101520.291581-4-kai.vehmanen@linux.intel.com/T/ were a fix for a similar issue on CML hardware so it may be something is still not right (and this fix was stress tested on a lot of HW). |
|
@kv2019i fyi. |
|
It happens consistently on wake. I can boot and have audio, but after the first suspend/resume, it will require a reboot to get it back. I'd love to help diagnose what is occurring. I can boot into the 5.16-rc7 kernel and test it out. I just installed the How exactly do I run the logger? I'm not sure what the |
|
@specious here are some ideas to help find the root cause a) downgrade to SOF version 1.9.3 . I see that you are using 2:0:0-b678a so we need to rule out a firmware issue in the last release b) prevent dynamic pipelines for HDaudio. this can be done with the quirk options snd_sof sof_debug=0xF /* Flag definitions used in sof_core_debug (sof_debug module parameter) */
#define SOF_DBG_ENABLE_TRACE BIT(0)
#define SOF_DBG_RETAIN_CTX BIT(1) /* prevent DSP D3 on FW exception */
#define SOF_DBG_VERIFY_TPLG BIT(2) /* verify topology during load */
#define SOF_DBG_DYNAMIC_PIPELINES_OVERRIDE BIT(3) /* 0: use topology token
* 1: override topology
*/
#define SOF_DBG_DYNAMIC_PIPELINES_ENABLE BIT(4) /* 0: use static pipelines
* 1: use dynamic pipelines
*/
#define SOF_DBG_DISABLE_MULTICORE BIT(5) /* schedule all pipelines/widgets
* on primary core
*/
#define SOF_DBG_PRINT_ALL_DUMPS BIT(6) /* Print all ipc and dsp dumps */ |
|
I installed In What could I be doing wrong? And to be clear, audio works perfectly with the 5.15.13 kernel (same configuration), no matter how many times I suspend/resume (either by closing the lid or Arch official kernel is now at 5.16.arch1-1, and I confirmed that audio breaks on suspend/resume, and it also worked with the previous official kernel. |
|
@specious These are then populated with the binaries and links. |
|
got the same issue! workaround to avoid reboot: sudo rmmod snd_sof_pci_intel_cnl
sudo modprobe snd_sof_pci_intel_cnl |
|
If the module is built into the kernel, is there also a way to reinitialize it? |
we don't support this case due to dependencies with the firmware loading from the filesystem. The SOF driver has to be module-based. Can you share your kernel config please? |
|
My kernel configuration: config-linux-5.16-20220118.txt Although all drivers are built in, everything works perfectly with the 5.15.13 kernel, and until a sleep/wake also on 5.16.0. I was simply wondering if there's any way to get the driver to reinitialize without doing an |
mengdonglin
added
CML
|
I have same problem on opensuse - https://bugzilla.suse.com/show_bug.cgi?id=1195027 |
|
Hi! I have exactly the same issue but with |
You can unbind and rebind via sysfs. It's a hard-way, equivalent with the device hot-unplug/plug the PCI device, so it might hit other bugs, though. |
|
@dumprop @carmilso do you guys see the same line in your dmesg ? That shows the HDA0.OUT widget loading message. This may point to some IP programming flow in FW that is failing when we try and use that HDA stream. |
Yes, unless I can't see (on boot, if it's make sense) (after resume, so there are no anything related to mei_hdcp) |
|
Hi, to get my audio to work after wakeup from suspend. This is my dmesg: Environment I am running a Lenovo Yoga 7 slim with kernel |
|
A similar bug report on bugzilla.kernel.org https://bugzilla.kernel.org/show_bug.cgi?id=215535 It suggests that the first bad commit is 5fcdbb2d45df6afb654674379546996b0027aa3e |
lgirdwood
changed the title
|
@plbossart @ranj063 is there a kernel config or module option that can disable dynamic pipelines so we can root cause and workaround atm ? |
|
Hi, While trying to reproduce a different issue I think I did hit this issue: In short: 5.16.1-artix1-1 kernel stopping aplay and starting it again will fail and I have this in the dmesg: Checked that vanilla 5.16 and 5.16.1 also have the same issue, but 5.17-rc1 does not. With some bisecting I ended up backporting the attached patches to 5.16.1 from 5.17-rc1 and the resume issue is resolved: @ranj063, @plbossart, @kv2019i I'll try to find a simpler way to fix it, but this is a good start. dmesg from the start of the first playback to resume: dmesg.txt |
|
this 96da174024b9c63bd5d3358668d0bc12677be877 (ASoC: SOF: handle paused streams during system suspend) commit added in 5.16.3 - https://cdn.kernel.org/pub/linux/kernel/v5.x/ChangeLog-5.16.3 |
|
@galibozek, 5.16.3 appears to be working correctly, thanks for the heads up! |
Yes, now I don't notice such a problem |
|
Closing since resolved with kernel update. |
Occasionally audio stops working after waking from sleep. This has been addressed in kernel 5.16 but since we're still pointed at 5.15 we can use this as a temporary workaround. thesofproject/sof#5168
Sound is not restored on suspend/resume with the recent 5.16-rc7 kernel.
This is the kernel log on wake:
It works perfectly with the 5.15.13 kernel.
Perhaps it has something to do with the recent changes.
Impact
Annoyance (need to reboot to restore sound)
Environment
This is on a Thinkpad X1 Carbon gen 8 with Intel Comet Lake PCH-LP cAVS audio.
Log output
dmesg.txt
The text was updated successfully, but these errors were encountered: