[PATCH] media: cedrus: Propagate OUTPUT resolution to CAPTURE

[PATCH] media: cedrus: Propagate OUTPUT resolution to CAPTURE

[Nicolas Dufresne]

As per spec, the CAPTURE resolution should be automatically set based on
the OTUPUT resolution. This patch properly propagate width/height to the
capture when the OUTPUT format is set and override the user provided
width/height with configured OUTPUT resolution when the CAPTURE fmt is
updated.

This also prevents userspace from selecting a CAPTURE resolution that is
too small, avoiding unwanted page faults.

Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
---
 drivers/staging/media/sunxi/cedrus/cedrus_video.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/staging/media/sunxi/cedrus/cedrus_video.c b/drivers/staging/media/sunxi/cedrus/cedrus_video.c
index 16d82309e7b6..a6d6b15adc2e 100644
--- a/drivers/staging/media/sunxi/cedrus/cedrus_video.c
+++ b/drivers/staging/media/sunxi/cedrus/cedrus_video.c
@@ -247,6 +247,8 @@ static int cedrus_try_fmt_vid_cap(struct file *file, void *priv,
 		return -EINVAL;
 
 	pix_fmt->pixelformat = fmt->pixelformat;
+	pix_fmt->width = ctx->src_fmt.width;
+	pix_fmt->height = ctx->src_fmt.height;
 	cedrus_prepare_format(pix_fmt);
 
 	return 0;
@@ -319,11 +321,14 @@ static int cedrus_s_fmt_vid_out(struct file *file, void *priv,
 		break;
 	}
 
-	/* Propagate colorspace information to capture. */
+	/* Propagate format information to capture. */
 	ctx->dst_fmt.colorspace = f->fmt.pix.colorspace;
 	ctx->dst_fmt.xfer_func = f->fmt.pix.xfer_func;
 	ctx->dst_fmt.ycbcr_enc = f->fmt.pix.ycbcr_enc;
 	ctx->dst_fmt.quantization = f->fmt.pix.quantization;
+	ctx->dst_fmt.width = ctx->src_fmt.width;
+	ctx->dst_fmt.height = ctx->src_fmt.height;
+	cedrus_prepare_format(&ctx->dst_fmt);
 
 	return 0;
 }
-- 
2.26.2


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH] media: cedrus: Propagate OUTPUT resolution to CAPTURE

[Nicolas Dufresne]

Le jeudi 14 mai 2020 à 11:39 -0400, Nicolas Dufresne a écrit :
> As per spec, the CAPTURE resolution should be automatically set based on
> the OTUPUT resolution. This patch properly propagate width/height to the
> capture when the OUTPUT format is set and override the user provided
> width/height with configured OUTPUT resolution when the CAPTURE fmt is
> updated.
> 
> This also prevents userspace from selecting a CAPTURE resolution that is
> too small, avoiding unwanted page faults.

Side note, this patch is based on top of "Samuel Holland <samuel@sholland.org>"
patches:

	[PATCH v3 1/2] media: cedrus: Program output format during each run
	[PATCH v3 2/2] media: cedrus: Implement runtime PM

As the patchset also fixes concurrent decoding issues. This was tested using
GStreamer implementation, which strictly follow the Stateless CODEC spec and
expect OUTPUT to CAPTURE width/height propagation.

> 
> Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
> ---
>  drivers/staging/media/sunxi/cedrus/cedrus_video.c | 7 ++++++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> b/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> index 16d82309e7b6..a6d6b15adc2e 100644
> --- a/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> +++ b/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> @@ -247,6 +247,8 @@ static int cedrus_try_fmt_vid_cap(struct file *file, void
> *priv,
>  		return -EINVAL;
>  
>  	pix_fmt->pixelformat = fmt->pixelformat;
> +	pix_fmt->width = ctx->src_fmt.width;
> +	pix_fmt->height = ctx->src_fmt.height;
>  	cedrus_prepare_format(pix_fmt);
>  
>  	return 0;
> @@ -319,11 +321,14 @@ static int cedrus_s_fmt_vid_out(struct file *file, void
> *priv,
>  		break;
>  	}
>  
> -	/* Propagate colorspace information to capture. */
> +	/* Propagate format information to capture. */
>  	ctx->dst_fmt.colorspace = f->fmt.pix.colorspace;
>  	ctx->dst_fmt.xfer_func = f->fmt.pix.xfer_func;
>  	ctx->dst_fmt.ycbcr_enc = f->fmt.pix.ycbcr_enc;
>  	ctx->dst_fmt.quantization = f->fmt.pix.quantization;
> +	ctx->dst_fmt.width = ctx->src_fmt.width;
> +	ctx->dst_fmt.height = ctx->src_fmt.height;
> +	cedrus_prepare_format(&ctx->dst_fmt);
>  
>  	return 0;
>  }


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH] media: cedrus: Propagate OUTPUT resolution to CAPTURE

[Ezequiel Garcia]

It seems this one felt thru the cracks. Sorry for the delay.

On Thu, 2020-05-14 at 11:39 -0400, Nicolas Dufresne wrote:
> As per spec, the CAPTURE resolution should be automatically set based on
> the OTUPUT resolution. This patch properly propagate width/height to the
> capture when the OUTPUT format is set and override the user provided
> width/height with configured OUTPUT resolution when the CAPTURE fmt is
> updated.
> 
> This also prevents userspace from selecting a CAPTURE resolution that is
> too small, avoiding unwanted page faults.
> 
> Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>

This looks correct.

Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>

Thanks,
Ezequiel

> ---
>  drivers/staging/media/sunxi/cedrus/cedrus_video.c | 7 ++++++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/staging/media/sunxi/cedrus/cedrus_video.c b/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> index 16d82309e7b6..a6d6b15adc2e 100644
> --- a/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> +++ b/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> @@ -247,6 +247,8 @@ static int cedrus_try_fmt_vid_cap(struct file *file, void *priv,
>  		return -EINVAL;
>  
>  	pix_fmt->pixelformat = fmt->pixelformat;
> +	pix_fmt->width = ctx->src_fmt.width;
> +	pix_fmt->height = ctx->src_fmt.height;
>  	cedrus_prepare_format(pix_fmt);
>  
>  	return 0;
> @@ -319,11 +321,14 @@ static int cedrus_s_fmt_vid_out(struct file *file, void *priv,
>  		break;
>  	}
>  
> -	/* Propagate colorspace information to capture. */
> +	/* Propagate format information to capture. */
>  	ctx->dst_fmt.colorspace = f->fmt.pix.colorspace;
>  	ctx->dst_fmt.xfer_func = f->fmt.pix.xfer_func;
>  	ctx->dst_fmt.ycbcr_enc = f->fmt.pix.ycbcr_enc;
>  	ctx->dst_fmt.quantization = f->fmt.pix.quantization;
> +	ctx->dst_fmt.width = ctx->src_fmt.width;
> +	ctx->dst_fmt.height = ctx->src_fmt.height;
> +	cedrus_prepare_format(&ctx->dst_fmt);
>  
>  	return 0;
>  }
> -- 
> 2.26.2
> 
> 



_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH] media: cedrus: Propagate OUTPUT resolution to CAPTURE

[Paul Kocialkowski]

[-- Attachment #1.1: Type: text/plain, Size: 2560 bytes --]

Hi,

On Wed 15 Jul 20, 08:57, Ezequiel Garcia wrote:
> It seems this one felt thru the cracks. Sorry for the delay.
> 
> On Thu, 2020-05-14 at 11:39 -0400, Nicolas Dufresne wrote:
> > As per spec, the CAPTURE resolution should be automatically set based on
> > the OTUPUT resolution. This patch properly propagate width/height to the
> > capture when the OUTPUT format is set and override the user provided
> > width/height with configured OUTPUT resolution when the CAPTURE fmt is
> > updated.
> > 
> > This also prevents userspace from selecting a CAPTURE resolution that is
> > too small, avoiding unwanted page faults.
> > 
> > Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
> 
> This looks correct.

Looks like there's no issue with alignment (that will get applied as well),
which was a possible concern.

Thanks for reviewing!

Acked-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>

Cheers,

Paul

> Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
> 
> Thanks,
> Ezequiel
> 
> > ---
> >  drivers/staging/media/sunxi/cedrus/cedrus_video.c | 7 ++++++-
> >  1 file changed, 6 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/staging/media/sunxi/cedrus/cedrus_video.c b/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> > index 16d82309e7b6..a6d6b15adc2e 100644
> > --- a/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> > +++ b/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> > @@ -247,6 +247,8 @@ static int cedrus_try_fmt_vid_cap(struct file *file, void *priv,
> >  		return -EINVAL;
> >  
> >  	pix_fmt->pixelformat = fmt->pixelformat;
> > +	pix_fmt->width = ctx->src_fmt.width;
> > +	pix_fmt->height = ctx->src_fmt.height;
> >  	cedrus_prepare_format(pix_fmt);
> >  
> >  	return 0;
> > @@ -319,11 +321,14 @@ static int cedrus_s_fmt_vid_out(struct file *file, void *priv,
> >  		break;
> >  	}
> >  
> > -	/* Propagate colorspace information to capture. */
> > +	/* Propagate format information to capture. */
> >  	ctx->dst_fmt.colorspace = f->fmt.pix.colorspace;
> >  	ctx->dst_fmt.xfer_func = f->fmt.pix.xfer_func;
> >  	ctx->dst_fmt.ycbcr_enc = f->fmt.pix.ycbcr_enc;
> >  	ctx->dst_fmt.quantization = f->fmt.pix.quantization;
> > +	ctx->dst_fmt.width = ctx->src_fmt.width;
> > +	ctx->dst_fmt.height = ctx->src_fmt.height;
> > +	cedrus_prepare_format(&ctx->dst_fmt);
> >  
> >  	return 0;
> >  }
> > -- 
> > 2.26.2
> > 
> > 
> 
> 

-- 
Paul Kocialkowski, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

[-- Attachment #2: Type: text/plain, Size: 176 bytes --]

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH] media: cedrus: Propagate OUTPUT resolution to CAPTURE

[Hans Verkuil]

On 14/05/2020 17:39, Nicolas Dufresne wrote:
> As per spec, the CAPTURE resolution should be automatically set based on
> the OTUPUT resolution. This patch properly propagate width/height to the
> capture when the OUTPUT format is set and override the user provided
> width/height with configured OUTPUT resolution when the CAPTURE fmt is
> updated.
> 
> This also prevents userspace from selecting a CAPTURE resolution that is
> too small, avoiding unwanted page faults.
> 
> Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
> ---
>  drivers/staging/media/sunxi/cedrus/cedrus_video.c | 7 ++++++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/staging/media/sunxi/cedrus/cedrus_video.c b/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> index 16d82309e7b6..a6d6b15adc2e 100644
> --- a/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> +++ b/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> @@ -247,6 +247,8 @@ static int cedrus_try_fmt_vid_cap(struct file *file, void *priv,
>  		return -EINVAL;
>  
>  	pix_fmt->pixelformat = fmt->pixelformat;
> +	pix_fmt->width = ctx->src_fmt.width;
> +	pix_fmt->height = ctx->src_fmt.height;
>  	cedrus_prepare_format(pix_fmt);
>  
>  	return 0;
> @@ -319,11 +321,14 @@ static int cedrus_s_fmt_vid_out(struct file *file, void *priv,
>  		break;
>  	}
>  
> -	/* Propagate colorspace information to capture. */
> +	/* Propagate format information to capture. */
>  	ctx->dst_fmt.colorspace = f->fmt.pix.colorspace;
>  	ctx->dst_fmt.xfer_func = f->fmt.pix.xfer_func;
>  	ctx->dst_fmt.ycbcr_enc = f->fmt.pix.ycbcr_enc;
>  	ctx->dst_fmt.quantization = f->fmt.pix.quantization;
> +	ctx->dst_fmt.width = ctx->src_fmt.width;
> +	ctx->dst_fmt.height = ctx->src_fmt.height;

You can only do this if the capture queue isn't busy.

See also hantro_reset_raw_fmt() where it does that check.

So this needs a v2.

Regards,

	Hans

> +	cedrus_prepare_format(&ctx->dst_fmt);
>  
>  	return 0;
>  }
> 


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH] media: cedrus: Propagate OUTPUT resolution to CAPTURE

[Nicolas Dufresne]

[-- Attachment #1.1: Type: text/plain, Size: 5731 bytes --]

Le jeudi 17 septembre 2020 à 12:39 +0200, Hans Verkuil a écrit :
> On 14/05/2020 17:39, Nicolas Dufresne wrote:
> > As per spec, the CAPTURE resolution should be automatically set based on
> > the OTUPUT resolution. This patch properly propagate width/height to the
> > capture when the OUTPUT format is set and override the user provided
> > width/height with configured OUTPUT resolution when the CAPTURE fmt is
> > updated.
> > 
> > This also prevents userspace from selecting a CAPTURE resolution that is
> > too small, avoiding unwanted page faults.
> > 
> > Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
> > ---
> >  drivers/staging/media/sunxi/cedrus/cedrus_video.c | 7 ++++++-
> >  1 file changed, 6 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/staging/media/sunxi/cedrus/cedrus_video.c b/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> > index 16d82309e7b6..a6d6b15adc2e 100644
> > --- a/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> > +++ b/drivers/staging/media/sunxi/cedrus/cedrus_video.c
> > @@ -247,6 +247,8 @@ static int cedrus_try_fmt_vid_cap(struct file *file, void *priv,
> >  		return -EINVAL;
> >  
> >  	pix_fmt->pixelformat = fmt->pixelformat;
> > +	pix_fmt->width = ctx->src_fmt.width;
> > +	pix_fmt->height = ctx->src_fmt.height;
> >  	cedrus_prepare_format(pix_fmt);
> >  
> >  	return 0;
> > @@ -319,11 +321,14 @@ static int cedrus_s_fmt_vid_out(struct file *file, void *priv,
> >  		break;
> >  	}
> >  
> > -	/* Propagate colorspace information to capture. */
> > +	/* Propagate format information to capture. */
> >  	ctx->dst_fmt.colorspace = f->fmt.pix.colorspace;
> >  	ctx->dst_fmt.xfer_func = f->fmt.pix.xfer_func;
> >  	ctx->dst_fmt.ycbcr_enc = f->fmt.pix.ycbcr_enc;
> >  	ctx->dst_fmt.quantization = f->fmt.pix.quantization;
> > +	ctx->dst_fmt.width = ctx->src_fmt.width;
> > +	ctx->dst_fmt.height = ctx->src_fmt.height;
> 
> You can only do this if the capture queue isn't busy.
> 
> See also hantro_reset_raw_fmt() where it does that check.
> 
> So this needs a v2.

I missed this reply, wasn't expecting a negative third review. I'll
copy and paste from there. So basically:

 - EBUSY if peer vq is busy
 - EBUSY if pixel format is changed while vq is busy
 - EBUSY if streaming

Obviously, I don't have userspace to exercise any of these cases, but
it seems to make sense. I remembered about this patch today as someone
reported the kernel crash we get without this patch:

---

I tried testing cedrus with gstreamer 0.18 and it managed to crash the
kernel on
A64. I used:

  gst-launch-1.0 filesrc location=test.mkv ! matroskademux ! queue !
h264parse ! v4l2slh264dec ! filesink location=aaa.dat

Unable to handle kernel paging request at virtual address
8080808080808088
Mem abort info:
  ESR = 0x96000044
  EC = 0x25: DABT (current EL), IL = 32 bits
  SET = 0, FnV = 0
  EA = 0, S1PTW = 0
Data abort info:
  ISV = 0, ISS = 0x00000044
  CM = 0, WnR = 1
[8080808080808088] address between user and kernel address ranges
Internal error: Oops: 96000044 [#1] SMP
Modules linked in: modem_power hci_uart btrtl bluetooth ecdh_generic
ecc sunxi_cedrus(C) sun8i_ce crypto_engine snd_soc_bt_sco
snd_soc_simple_card snd_soc_simple_card_utils snd_soc_simple_amplifier
sun50i_codec_analog sun8i_codec sun8i_adda_pr_regmap snd_soc_ec25
sun4i_i2s snd_soc_core snd_pcm_dmaengine snd_pcm snd_timer snd
soundcore stk3310 inv_mpu6050_i2c inv_mpu6050 st_magn_i2c
st_sensors_i2c st_magn st_sensors industrialio_triggered_buffer
kfifo_buf regmap_i2c option usb_wwan usbserial anx7688 ohci_platform
ohci_hcd ehci_platform ehci_hcd g_cdc usb_f_acm u_serial usb_f_ecm
u_ether libcomposite sunxi phy_generic musb_hdrc udc_core usbcore
sun8i_rotate v4l2_mem2mem gc2145 ov5640 sun6i_csi videobuf2_dma_contig
v4l2_fwnode videobuf2_memops videobuf2_v4l2 videobuf2_common videodev
mc 8723cs(C) cfg80211 rfkill lima gpu_sched goodix
CPU: 0 PID: 0 Comm: swapper/0 Tainted: G         C        5.9.0-rc5-
00386-g4fe2ef82bd0b #62
Hardware name: Pine64 PinePhone (1.2) (DT)
pstate: 80000085 (Nzcv daIf -PAN -UAO BTYPE=--)
pc : v4l2_m2m_buf_remove+0x44/0x90 [v4l2_mem2mem]
lr : v4l2_m2m_buf_remove+0x18/0x90 [v4l2_mem2mem]
sp : ffffffc010c8be20
x29: ffffffc010c8be20 x28: ffffffc010bb2fc0 
x27: 0000000000000060 x26: ffffffc010935e58 
x25: ffffffc010c06a5a x24: 0000000000000080 
x23: 0000000000000005 x22: ffffffc010c8bf4c 
x21: ffffff806ba0d088 x20: ffffff80687d1800 
x19: ffffff8066c40298 x18: 0000000000000000 
x17: 0000000000000000 x16: 0000000000000000 
x15: 000001b66678fd80 x14: 0000000000000204 
x13: 0000000000000001 x12: 0000000000000040 
x11: ffffff806f0c0248 x10: ffffff806f0c024a 
x9 : ffffffc010bbdac8 x8 : ffffff806f000270 
x7 : 0000000000000000 x6 : dead000000000100 
x5 : dead000000000122 x4 : 0000000000000000 
x3 : 8080808080808080 x2 : 8080808080808080 
x1 : ffffff80641327a8 x0 : 0000000000000080 
Call trace:
 v4l2_m2m_buf_remove+0x44/0x90 [v4l2_mem2mem]
 v4l2_m2m_buf_done_and_job_finish+0x34/0x140 [v4l2_mem2mem]
 cedrus_irq+0x8c/0xc0 [sunxi_cedrus]
 __handle_irq_event_percpu+0x54/0x150
 handle_irq_event+0x4c/0xec
 handle_fasteoi_irq+0xbc/0x1c0
 __handle_domain_irq+0x78/0xdc
 gic_handle_irq+0x50/0xa0
 el1_irq+0xb8/0x140
 arch_cpu_idle+0x10/0x14
 cpu_startup_entry+0x24/0x60
 rest_init+0xb0/0xbc
 arch_call_rest_init+0xc/0x14
 start_kernel+0x690/0x6b0
Code: f2fbd5a6 f2fbd5a5 52800004 a9400823 (f9000462) 
---[ end trace 88233b9a76cdb261 ]---
Kernel panic - not syncing: Fatal exception in interrupt


> 
> Regards,
> 
> 	Hans
> 
> > +	cedrus_prepare_format(&ctx->dst_fmt);
> >  
> >  	return 0;
> >  }
> > 

[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

[-- Attachment #2: Type: text/plain, Size: 176 bytes --]

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel