[Sound-open-firmware] Problems booting byt in qemu

[Liam Girdwood]

On Thu, 2019-06-06 at 13:06 +0300, Dragos Tarcatu wrote:
> On 06.06.2019 11:55, Liam Girdwood wrote:
> 
> > On Thu, 2019-06-06 at 09:39 +0300, Dragos Tarcatu wrote:
> > > Hi Kai,
> > > 
> > > Thanks for the hint!
> > > 
> > > On 05.06.2019 14:29, Kai Vehmanen wrote:
> > > > >       [    3.070623] sof-audio-acpi 80860F28:00: booting DSP
> > > > > firmware
> > > > >       [    3.172697] sof-audio-acpi 80860F28:00: error:
> > > > > firmware
> > > > > boot failure
> > > > >       [    3.176674] sof-audio-acpi 80860F28:00: error:
> > > > > unexpected
> > > > > fault
> > > > >       0x70028800 trace 0x00000000
> > > > First a simple test to do is to increase the boot timeout in
> > > > driver loader.c:
> > > > 
> > > > »       ret = wait_event_timeout(sdev->boot_wait, sdev-
> > > > > boot_complete,
> > > > »       »       »       »        msecs_to_jiffies(sdev-
> > > > > boot_timeout));
> > > > I did hit occasionally that the DSP-side QEMU was too slow to
> > > > respond
> > > > and simply bumping this timeout help.
> > > Indeed, after increasing the values for the default ipc and boot
> > > timeouts (TIMEOUT_DEFAULT_IPC_MS, TIMEOUT_DEFAULT_BOOT_MS,
> > > IPC_TIMEOUT_MS) I got it it to boot. I'm still getting errors
> > > when
> > > loading the topology, though:
> > > 
> > Are the topology loading errors always at the same place or does it
> > vary on each attempt ?
> I've ran it several times now, and I haven't got PCM0P to load:
> 
> [    7.608994] sof-audio-acpi 80860F28:00: loading 
> topology:intel/sof-tplg/sof-byt-nocodec.tplg
> [    7.634153] sof-audio-acpi 80860F28:00: Topology: ABI 3:7:0
> Kernel 
> ABI 3:7:0
> [    7.634156] sof-audio-acpi 80860F28:00: ASoC: adding 11 DAPM
> widgets
> [    7.634157] sof-audio-acpi 80860F28:00: ASoC: creating DAPM
> widget 
> codec_out0 id 1
> [    7.634162] sof-audio-acpi 80860F28:00: tplg: ready widget id 0
> pipe 
> 1 type 1 name : codec_out0 stream none
> [    7.634170] sof-audio-acpi 80860F28:00: warning: widget type 1
> name 
> codec_out0 not handled
> [    7.634172] sof-audio-acpi 80860F28:00: ASoC: creating DAPM
> widget 
> PCM0P id 11
> [    7.634175] sof-audio-acpi 80860F28:00: tplg: ready widget id 1
> pipe 
> 1 type 23 name : PCM0P stream Low Latency Playback 0
> [    7.634177] sof-audio-acpi 80860F28:00: loaded host PCM0P
> [    7.634178] sof-audio-acpi 80860F28:00:  config: periods snk 2 src
> 0 
> fmt 0
> [    7.635249] sof-audio-acpi 80860F28:00: ipc tx: 0x30010000: 
> GLB_TPLG_MSG: COMP_NEW
> [   10.661694] sof-audio-acpi 80860F28:00: error: ipc timed out for 
> 0x30010000 size 76
> [   10.665895] sof-audio-acpi 80860F28:00: error: unexpected fault 
> 0x70028800 trace 0x00000000
> [   10.666200] sof-audio-acpi 80860F28:00: error: DSP failed to add 
> widget id 0 type 23 name : PCM0P stream Low Latency Playback 0 reply
> 0
> [   10.666534] sof-audio-acpi 80860F28:00: ASoC: failed to load
> widget PCM0P
> [   10.666783] sof-audio-acpi 80860F28:00: error: tplg component
> load 
> failed -110
> [   10.667079] sof-audio-acpi 80860F28:00: error: failed to load DSP 
> topology -22
> [   10.667321] sof-audio-acpi 80860F28:00: ASoC: failed to probe 
> component -22
> [   10.667562] sof-nocodec sof-nocodec: ASoC: failed to instantiate
> card -22
> 
> > Btw, what kernel version are you using. Recent version of sof-dev
> > branch will print out additional debug for IPC timeouts.
> I'm on top of sof-dev, but in my previous log I had the verbose IPC 
> logging option off.
> I'm currently running with the timeout values set to:
> 
> core.c:
> #define TIMEOUT_DEFAULT_IPC_MS  3000
> #define TIMEOUT_DEFAULT_BOOT_MS 10000
> 
> ipc.c:
> #define IPC_TIMEOUT_MS         3000
> 
> Those seem like pretty high values for me, compared to the initial
> ones, 
> but I'll try to increase those even more - maybe that fixes it.
> 
Ok, I'm suspicious here that we have hit a regression on the host qemu
side. This previously worked and would load the whole topology and
allow playback capture to/from file.

Fwiw, CI currently only tests the FW DSP qemu side (to FW boot today)
and is not testing the host side yet. Xiuli will be working on adding
host side qemu to the CI but I dont think he's started yet. 

Can you add this to the bug tracker and we can resolve.

Thanks

Liam

> Dragoș
>