Re: [PATCH 0/5] media: hws: add HDMI audio capture support

[hverkuil+cisco@kernel.org]

Hi Ben,

Something went wrong here when posting: instead of having a proper patch series,
all the patches are appended to the cover letter.

Please repost!

Regards,

	Hans

On 27/06/2026 03:40, hoff.benjamin.k@gmail.com wrote:
> From: Ben Hoff <hoff.benjamin.k@gmail.com>
> 
> This series prepares the AVMatrix HWS PCIe capture driver for HDMI audio
> capture by making the video DMA programming use the hardware BAR remap
> windows consistently, adding the shared scratch DMA arena required by the
> remap model, adding a video bounce path for the shared remap window,
> hardening queue ownership, and wiring up ALSA PCM capture support.
> 
> This driver is derived from a GPL out-of-tree driver. The baseline used
> for comparison is available at:
> 
>   https://github.com/benhoff/hws/tree/baseline
> 
> This series supersedes the earlier standalone submission at:
> 
>   https://patchwork.linuxtv.org/project/linux-media/patch/20260620173448.36279-1-hoff.benjamin.k@gmail.com/
> 
> Supersedes: <20260620173448.36279-1-hoff.benjamin.k@gmail.com>
> Assisted-by: codex:5.5
> 
> Validation:
> 
>   - Build checks passed with W=1 for the HWS driver as a module
>     (VIDEO_HWS=m) and through the built-in directory target
>     (VIDEO_HWS=y, producing drivers/media/pci/hws/built-in.a). C=1
>     builds completed for both configurations
> 
>   - v4l2-compliance 1.32.0, 64 bits, 64-bit time_t, passed
>       Total: 51, Succeeded: 51, Failed: 0, Warnings: 1
> 
>   - HDMI audio progressive validation passed. Coverage included ALSA
>     capture-device enumeration, advertised hardware-parameter checks,
>     48 kHz S16_LE stereo capture from an HDMI playback, WAV
>     header/size/nonzero-sample checks, captured rate/channel/format
>     verification, no reported XRUNs, no kernel warnings/errors during
>     capture, repeated start/stop capture cycles, 100 consecutive short
>     captures, and 1000 consecutive device open/close cycles.
> 
>   - ALSA period/buffer matrix validation passed at 48 kHz S16_LE stereo
>     with HDMI playback tone. Tested period/buffer frame pairs were
>     256/1024, 256/4096, 1024/4096, and 4096/16384; each case checked
>     that arecord accepted the requested hw params, capture completed,
>     the WAV artifact had valid header/size/non-silent content and the
>     expected rate/channel/format, no XRUN was reported, and no kernel
>     warning/error appeared during the case.
> 
>   - Mixed audio/video validation passed. V4L2 MMAP streaming was kept
>     active on an HWS video node while ALSA captured 48 kHz S16_LE stereo
>     audio from the matching HWS PCM device. The test checked that audio
>     capture completed, video streaming stayed active for the full run,
>     the WAV artifact was present, non-truncated, and non-silent, and the
>     kernel log delta did not contain warnings, errors, DMA faults, XRUNs,
>     overrun/underrun reports, or HWS timeout/stall diagnostics.
> 
>   - video tested with enabled_audio=0
> 
>   - pm_test=devices passes with and without D3hot.
> 
> Full platform S3/deep suspend is not considered validated, my system is
> unstable that I can't resume for deep suspend even without the driver
> 
> Ben Hoff (5):
>   media: hws: program video DMA through remap windows
>   media: hws: add shared scratch DMA arena
>   media: hws: add video bounce path for shared remap windows
>   media: hws: harden video DMA queue ownership
>   media: hws: add HDMI audio capture support
>