Re: [PATCH 00/75] media: imx: Miscellaneous fixes and cleanups for i.MX7

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

Hi Fabio,

On Tue, Jan 05, 2021 at 02:45:40PM -0300, Fabio Estevam wrote:
> On Tue, Jan 5, 2021 at 12:31 PM Laurent Pinchart wrote:
> >
> > Hello,
> >
> > This large patch series has been sitting in my tree for way too long. I
> > haven't posted it yet as I'm running into an issue on my test hardware
> > that I can't prove is not a regression from this series, but the
> > pressure has grown and the patches are better on the list for review.
> >
> > There's really not much to detail in the cover letter as there are
> > "just" fixes and cleanups I developed while bringing up camera support
> > for an i.MX7D platform, and later on an i.MX8MM that shares the same
> > MIPI-CSI2 and CSI IP cores (with some differences).
> >
> > The issue I've noticed is that the CSI writes two images consecutively
> > to the same buffer, overwritting memory after the end of the buffer. I
> > believe this bug to already be present in mainline, but I can't prove it
> > as my sensor won't work without some of the patches in this series. The
> > problem could also be sensor-specific.
> >
> > Rui, would you be able to test this on your i.MX7 hardware to make sure
> > there's no regression ?
> 
> Thanks for your series.
> 
> I tested it on a imx6ul-evk board.

Thank you.

> There is a build error introduced by patch 74/75. I fixed it like this:
> 
> --- a/drivers/staging/media/imx/imx7-mipi-csis.c
> +++ b/drivers/staging/media/imx/imx7-mipi-csis.c
> @@ -193,6 +193,8 @@
>  #define MIPI_CSIS_SDW_RESOL_CH(n)              (0x84 + (n) * 0x10)
>  #define MIPI_CSIS_SDW_SYNC_CH(n)               (0x88 + (n) * 0x10)
> 
> +/* Debug Control register */
> +#define MIPI_CSIS_DBG_CTRL                     0x20
>  /* Non-image packet data buffers */
>  #define MIPI_CSIS_PKTDATA_ODD                  0x2000
>  #define MIPI_CSIS_PKTDATA_EVEN                 0x3000

Oops. I have a debug patch in my branch on top of the series that adds
the macro, that's why I haven't noticed compilation broke. Sorry about
that.

> Then I applied my patch and Rui's to fix the imx6ul regression as per
> the other thread we have been discussing, but I was not able to
> capture:

Would you be able to bisect this ?

> # gst-launch-1.0 -v  v4l2src device=/dev/video1 ! v4l2convert ! fbdevsink
> Setting pipeline to PAUSED ...
> Pipeline is live and does not need PREROLL ...
> Pipeline is PREROLLED ...
> Setting pipeline to PLAYING ...
> New clock: GstSystemClock
> /GstPipeline:pipeline0/GstV4l2Src:v4l2src0.GstPad:src: caps =
> video/x-raw, format=(string)BGRx, width=(int)3840, height=(int)2160,
> framerate=(fraction)120/1, interlace-mode=(string)progressive,
> colorimetr
> y=(string)1:1:5:1
> /GstPipeline:pipeline0/v4l2convert:v4l2convert0.GstPad:src: caps =
> video/x-raw, format=(string)BGRx, width=(int)3840, height=(int)2160,
> framerate=(fraction)120/1, interlace-mode=(string)progressive, color
> imetry=(string)1:1:5:1
> /GstPipeline:pipeline0/GstFBDEVSink:fbdevsink0.GstPad:sink: caps =
> video/x-raw, format=(string)BGRx, width=(int)3840, height=(int)2160,
> framerate=(fraction)120/1, interlace-mode=(string)progressive, color
> imetry=(string)1:1:5:1
> /GstPipeline:pipeline0/v4l2convert:v4l2convert0.GstPad:sink: caps =
> video/x-raw, format=(string)BGRx, width=(int)3840, height=(int)2160,
> framerate=(fraction)120/1, interlace-mode=(string)progressive, colo
> rimetry=(string)1:1:5:1
> [   32.783736] cma: cma_alloc: alloc failed, req-size: 8100 pages, ret: -12
> [   32.791332] imx7-csi 21c4000.csi: dma_alloc_coherent of size 33177600 failed
> ERROR: from element /GstPipeline:pipeline0/GstV4l2Src:v4l2src0: Failed
> to allocate required memory.
> Additional debug info:
> ../sys/v4l2/gstv4l2src.c(659): gst_v4l2src_decide_allocation ():
> /GstPipeline:pipeline0/GstV4l2Src:v4l2src0:
> Buffer pool activation failed
> Execution ended after 0:00:00.214658125
> Setting pipeline to NULL ...
> Freeing pipeline ...
> 
> As shown above the dimensions and framerate are incorrectly reported
> as: width=(int)3840, height=(int)2160, framerate=(fraction)120/1
> 
> Previously it was:
> 
> # gst-launch-1.0 -v  v4l2src device=/dev/video1 ! v4l2convert ! fbdevsink
> Setting pipeline to PAUSED ...
> Pipeline is live and does not need PREROLL ...
> Pipeline is PREROLLED ...
> Setting pipeline to PLAYING ...
> New clock: GstSystemClock
> /GstPipeline:pipeline0/GstV4l2Src:v4l2src0.GstPad:src: caps =
> video/x-raw, format=(string)UYVY, width=(int)320, height=(int)240,
> framerate=(fraction)30000/1001, interlace-mode=(string)progressive,
> colorim
> etry=(string)1:4:7:1
> /GstPipeline:pipeline0/v4l2convert:v4l2convert0.GstPad:src: caps =
> video/x-raw, format=(string)BGRx, width=(int)320, height=(int)240,
> framerate=(fraction)30000/1001, interlace-mode=(string)progressive
> /GstPipeline:pipeline0/GstFBDEVSink:fbdevsink0.GstPad:sink: caps =
> video/x-raw, format=(string)BGRx, width=(int)320, height=(int)240,
> framerate=(fraction)30000/1001, interlace-mode=(string)progressive
> /GstPipeline:pipeline0/v4l2convert:v4l2convert0.GstPad:sink: caps =
> video/x-raw, format=(string)UYVY, width=(int)320, height=(int)240,
> framerate=(fraction)30000/1001, interlace-mode=(string)progressive, c
> olorimetry=(string)1:4:7:1

-- 
Regards,

Laurent Pinchart