"Sharpening Support" reported but using it, will crash GPU

[Quantomax]

Hi,

I've a problem with sharpening support on Intel J3160 (CherryView 2):
If I enable Sharpening, e.g. switch from 0 -> 1, I'll get something like a frozen not-deinterlaced flickering picture (seems like three frames joined), crashing GPU some moments later. Need to detach/attach to resume.

vainfo:
error: XDG_RUNTIME_DIR not set in the environment.
libva info: VA-API version 1.0.0
libva info: va_getDriverName() returns 0
libva info: Trying to open /usr/lib/x86_64-linux-gnu/dri/i965_drv_video.so
libva info: Found init function __vaDriverInit_1_0
libva info: va_openDriver() returns 0
vainfo: VA-API version: 1.0 (libva 2.0.0)
vainfo: Driver version: Intel i965 driver for Intel(R) CherryView - 2.0.0
vainfo: Supported profile and entrypoints
VAProfileMPEG2Simple : VAEntrypointVLD
VAProfileMPEG2Simple : VAEntrypointEncSlice
VAProfileMPEG2Main : VAEntrypointVLD
VAProfileMPEG2Main : VAEntrypointEncSlice
VAProfileH264ConstrainedBaseline: VAEntrypointVLD
VAProfileH264ConstrainedBaseline: VAEntrypointEncSlice
VAProfileH264Main : VAEntrypointVLD
VAProfileH264Main : VAEntrypointEncSlice
VAProfileH264High : VAEntrypointVLD
VAProfileH264High : VAEntrypointEncSlice
VAProfileH264MultiviewHigh : VAEntrypointVLD
VAProfileH264MultiviewHigh : VAEntrypointEncSlice
VAProfileH264StereoHigh : VAEntrypointVLD
VAProfileH264StereoHigh : VAEntrypointEncSlice
VAProfileVC1Simple : VAEntrypointVLD
VAProfileVC1Main : VAEntrypointVLD
VAProfileVC1Advanced : VAEntrypointVLD
VAProfileNone : VAEntrypointVideoProc
VAProfileJPEGBaseline : VAEntrypointVLD
VAProfileJPEGBaseline : VAEntrypointEncPicture
VAProfileVP8Version0_3 : VAEntrypointVLD
VAProfileHEVCMain : VAEntrypointVLD

Logs:
[21959] VDR version 2.3.8 started
[21959] loading plugin: /usr/lib/vdr/plugins/libvdr-vaapidevice.so.2.3.8
video/vaapi: libva 1.0 (Intel i965 driver for Intel(R) CherryView - 2.0.0) initialized
video/vaapi: background-color is unsupported
video/vaapi: supports video processing
video/vaapi: supports unscaled osd
[vaapidevice] ready
..
codec: YUV 420 supported
Using entrypoint for vpp: 10
video/vaapi: can't destroy postproc context!
can't destroy config!
video/vaapi: noise reduction supported
video/vaapi: 0,00 - 1,00 ++ 0,03 = 0,50
Enabling denoise filter (pos = 0)
video/vaapi: deinterlacing supported
video/vaapi: bob deinterlace supported
motion adaptive deinterlace supported
motion compensated deinterlace supported
Enabling Deint (pos = 1)
video/vaapi: sharpening supported
video/vaapi: 0,00 - 1,00 ++ 0,03 = 0,50
Enabling sharpening filter (pos = 0)
video/vaapi: enabling color balance filters
Supported color balance filter count: 4
Hue (-180,00 - 180,00 ++ 1,00 = 0,00) (pos = 2)
Saturation (0,00 - 10,00 ++ 0,10 = 1,00) (pos = 2)
Brightness (-100,00 - 100,00 ++ 1,00 = 0,00) (pos = 2)
Contrast (0,00 - 10,00 ++ 0,10 = 1,00) (pos = 2)
video/vaapi: skin tone enhancement supported
video/vaapi: 0,00 - 4,00 ++ 1,00 = 0,00
Enabling skin tone filter (pos = 3)
Allocating 1 forward reference surfaces for postprocessing
Allocating 0 backward reference surfaces for postprocessing
video: display buffer empty, duping frame (46/0) 0
Surface 0 query status = 0x6: invalid VASurfaceID
Surface 0 query status = 0x6: invalid VASurfaceID
...

While switching on sharpening via menu
video/vaapi: gpu hung -4294949ms 78208

kernel: [53772.319113] [drm] GPU HANG: ecode 8:0:0x85ddfffb, in vdr [21959], reason: Hang on rcs0, action: reset
kernel: [53772.319118] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
kernel: [53772.319120] [drm] Please file a new bug report on bugs.freedesktop.org against DRI -> DRM/Intel
kernel: [53772.319121] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
kernel: [53772.319122] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
kernel: [53772.319123] [drm] GPU crash dump saved to /sys/class/drm/card0/error
kernel: [53772.319136] i915 0000:00:02.0: Resetting rcs0 after gpu hang
kernel: [53776.454505] asynchronous wait on fence i915:vdr[21959]/0:38eff timed out
kernel: [53780.294596] i915 0000:00:02.0: Resetting rcs0 after gpu hang
kernel: [53788.294494] i915 0000:00:02.0: Resetting rcs0 after gpu hang
kernel: [53796.294490] i915 0000:00:02.0: Resetting rcs0 after gpu hang
`
Kernel is a 4.14.10.

[rofafor]

IMO, this looks like a bug in Intel drivers - you should make an issue for them instead.

[pesintta]

You can also run this simple sharpening test tool I've written some years ago when I reported similar bug to Intel devs: https://bugs.freedesktop.org/attachment.cgi?id=110316

It may make the crash easier to reproduce or not, YMMV.