2021-12-29 15:40:16

by Rouven Czerwinski

Subject: IOMMU Page Fault with Creative Soundblaster Zx after suspend

Hi,

With current 5.15.11, my soundblaster Zx is not correctly coming back
from suspend. Dmesg prints:

[Mi Dez 29 16:12:14 2021] [drm] UVD and UVD ENC initialized successfully.
[Mi Dez 29 16:12:14 2021] snd_hda_codec_ca0132 hdaudioC0D1: ca0132 DSP downloaded and running
[Mi Dez 29 16:12:15 2021] [drm] VCE initialized successfully.
[Mi Dez 29 16:12:15 2021] usb 1-5: reset full-speed USB device number 2 using xhci_hcd
[Mi Dez 29 16:12:16 2021] ata1.00: Enabling discard_zeroes_data
[Mi Dez 29 16:12:16 2021] OOM killer enabled.
[Mi Dez 29 16:12:16 2021] Restarting tasks ... done.
[Mi Dez 29 16:12:16 2021] PM: suspend exit
[Mi Dez 29 16:12:16 2021] snd_hda_intel 0000:06:00.0: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x000f address=0xdf880000 flags=0x0000]
[Mi Dez 29 16:12:17 2021] igb 0000:08:00.0 enp8s0: igb: enp8s0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: RX/TX

Where the IOMMU Page Fault looks suspicious. Unbinding and rebinding the
driver using sysfs fixes it, thats the current workaround I have
been using.

Thanks and happy holidays,
Rouven Czerwinski

2021-12-30 07:56:13

by Takashi Iwai

Subject: Re: IOMMU Page Fault with Creative Soundblaster Zx after suspend

On Wed, 29 Dec 2021 16:24:13 +0100,
Rouven Czerwinski wrote:
>
> Hi,
>
> With current 5.15.11, my soundblaster Zx is not correctly coming back
> from suspend. Dmesg prints:
>
> [Mi Dez 29 16:12:14 2021] [drm] UVD and UVD ENC initialized successfully.
> [Mi Dez 29 16:12:14 2021] snd_hda_codec_ca0132 hdaudioC0D1: ca0132 DSP downloaded and running
> [Mi Dez 29 16:12:15 2021] [drm] VCE initialized successfully.
> [Mi Dez 29 16:12:15 2021] usb 1-5: reset full-speed USB device number 2 using xhci_hcd
> [Mi Dez 29 16:12:16 2021] ata1.00: Enabling discard_zeroes_data
> [Mi Dez 29 16:12:16 2021] OOM killer enabled.
> [Mi Dez 29 16:12:16 2021] Restarting tasks ... done.
> [Mi Dez 29 16:12:16 2021] PM: suspend exit
> [Mi Dez 29 16:12:16 2021] snd_hda_intel 0000:06:00.0: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x000f address=0xdf880000 flags=0x0000]
> [Mi Dez 29 16:12:17 2021] igb 0000:08:00.0 enp8s0: igb: enp8s0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: RX/TX
>
> Where the IOMMU Page Fault looks suspicious. Unbinding and rebinding the
> driver using sysfs fixes it, thats the current workaround I have
> been using.

Is it a regression from the earlier releases? If yes, please try to
narrow down the regression range, at least, which version started
showing the problem. It'd be great if you can perform git bisection,
of course.


thanks,

Takashi

2022-01-04 08:52:00

by Rouven Czerwinski

Subject: Re: IOMMU Page Fault with Creative Soundblaster Zx after suspend

Hi,

Am 2021-12-30 08:56, schrieb Takashi Iwai:
> On Wed, 29 Dec 2021 16:24:13 +0100,
> Rouven Czerwinski wrote:
>>
>> Hi,
>>
>> With current 5.15.11, my soundblaster Zx is not correctly coming back
>> from suspend. Dmesg prints:
>>
>> [Mi Dez 29 16:12:14 2021] [drm] UVD and UVD ENC initialized
>> successfully.
>> [Mi Dez 29 16:12:14 2021] snd_hda_codec_ca0132 hdaudioC0D1: ca0132 DSP
>> downloaded and running
>> [Mi Dez 29 16:12:15 2021] [drm] VCE initialized successfully.
>> [Mi Dez 29 16:12:15 2021] usb 1-5: reset full-speed USB device number
>> 2 using xhci_hcd
>> [Mi Dez 29 16:12:16 2021] ata1.00: Enabling discard_zeroes_data
>> [Mi Dez 29 16:12:16 2021] OOM killer enabled.
>> [Mi Dez 29 16:12:16 2021] Restarting tasks ... done.
>> [Mi Dez 29 16:12:16 2021] PM: suspend exit
>> [Mi Dez 29 16:12:16 2021] snd_hda_intel 0000:06:00.0: AMD-Vi: Event
>> logged [IO_PAGE_FAULT domain=0x000f address=0xdf880000 flags=0x0000]
>> [Mi Dez 29 16:12:17 2021] igb 0000:08:00.0 enp8s0: igb: enp8s0 NIC
>> Link is Up 1000 Mbps Full Duplex, Flow Control: RX/TX
>>
>> Where the IOMMU Page Fault looks suspicious. Unbinding and rebinding
>> the
>> driver using sysfs fixes it, thats the current workaround I have
>> been using.
>
> Is it a regression from the earlier releases? If yes, please try to
> narrow down the regression range, at least, which version started
> showing the problem. It'd be great if you can perform git bisection,
> of course.

I thought it was but it isn't, I recently switched the profile from
"Analog Stereo Duplex"
to "Analog Surround 5.1" within pavucontrol (underlying sound daemon is
pipewire).
This seems to lead to the IOMMU error on resume, using 2.0 works fine
over a suspend resume
cycle.
Time to find out what the surround handling does differently from plain
stereo for the
soundblaster Z cards.

Kind regards,
Rouven