linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Alexandre Courbot <acourbot@chromium.org>
To: Hans Verkuil <hverkuil@xs4all.nl>
Cc: Tiffany Lin <tiffany.lin@mediatek.com>,
	Andrew-CT Chen <andrew-ct.chen@mediatek.com>,
	Rui Wang <gtk_ruiwang@mediatek.com>,
	Yunfei Dong <yunfei.dong@mediatek.com>,
	Pi-Hsun Shih <pihsun@chromium.org>,
	Maoguang Meng <maoguang.meng@mediatek.com>,
	Linux Media Mailing List <linux-media@vger.kernel.org>,
	"moderated list:ARM/Mediatek SoC support" 
	<linux-mediatek@lists.infradead.org>,
	LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v2 16/18] media: mtk-vcodec: venc: make S_PARM return -ENOTTY for CAPTURE queue
Date: Tue, 7 Jul 2020 13:44:19 +0900	[thread overview]
Message-ID: <CAPBb6MWT=aALmuQLkaKLYeXiyPViQCDZLdUUGdi_NFFXCo6Vcw@mail.gmail.com> (raw)
In-Reply-To: <74652f57-8dd8-6b24-9889-59bec389464e@xs4all.nl>

On Mon, Jul 6, 2020 at 9:43 PM Hans Verkuil <hverkuil@xs4all.nl> wrote:
>
> On 04/07/2020 14:36, Alexandre Courbot wrote:
> > On Fri, Jul 3, 2020 at 5:30 PM Hans Verkuil <hverkuil@xs4all.nl> wrote:
> >>
> >> On 26/06/2020 10:04, Alexandre Courbot wrote:
> >>> v4l2-compliance expects ENOTTY to be returned when a given queue does
> >>> not support S_PARM.
> >>>
> >>> Signed-off-by: Alexandre Courbot <acourbot@chromium.org>
> >>> ---
> >>>  drivers/media/platform/mtk-vcodec/mtk_vcodec_enc.c | 2 +-
> >>>  1 file changed, 1 insertion(+), 1 deletion(-)
> >>>
> >>> diff --git a/drivers/media/platform/mtk-vcodec/mtk_vcodec_enc.c b/drivers/media/platform/mtk-vcodec/mtk_vcodec_enc.c
> >>> index aae610e6d4e8..346a33c6869d 100644
> >>> --- a/drivers/media/platform/mtk-vcodec/mtk_vcodec_enc.c
> >>> +++ b/drivers/media/platform/mtk-vcodec/mtk_vcodec_enc.c
> >>> @@ -200,7 +200,7 @@ static int vidioc_venc_s_parm(struct file *file, void *priv,
> >>>       struct mtk_vcodec_ctx *ctx = fh_to_ctx(priv);
> >>>
> >>>       if (a->type != V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE)
> >>> -             return -EINVAL;
> >>> +             return -ENOTTY;
> >>
> >> This doesn't look right: S_PARM *is* supported, just not for this buffer type.
> >> So -EINVAL is the correct error code.
> >>
> >> What is the exact v4l2-compliance failure? It might be a bug in the test.
> >
> > The error is as follows:
> >
> > Format ioctls:
> >         test VIDIOC_ENUM_FMT/FRAMESIZES/FRAMEINTERVALS: OK
> >                 fail: v4l2-test-formats.cpp(1336): got error 22 when
> > setting parms for buftype 9
> >         test VIDIOC_G/S_PARM: FAIL
>
> This is due to bugs in the compliance test that do not take into account the
> stateful encoder spec.
>
> The compliance test should check for the following:
>
> If ENUM_FRAMEINTERVALS is implemented, then G/S_PARM shall be supported for OUTPUT.
>
> If ENUM_FRAMEINTERVALS is not implemented, then G/S_PARM is optional for OUTPUT. But if
> G_PARM is implemented, then S_PARM shall also be implemented. In this case the
> frame interval range is determined by the codec standard.
>
> If V4L2_FMT_FLAG_ENC_CAP_FRAME_INTERVAL is set for one or more pixel formats,
> then G/S_PARM shall be implemented for both CAPTURE and OUTPUT.
>
> For backwards compatibility: if G/S_PARM is supported for OUTPUT, then it may be
> supported for CAPTURE as well.
>
> Can you try with the following patch?
>
> With this v4l2-compliance patch you should be able to drop patches 15/18 and 16/18
> of your series.

Indeed, with this patch v4l2-compliance passes on the encoder device
without 15/18 and 16/18. Thanks, I'll remove them from the next
iteration.

>
> Regards,
>
>         Hans
>
> -------------------------- cut here ------------------------------
> diff --git a/utils/v4l2-compliance/v4l2-compliance.cpp b/utils/v4l2-compliance/v4l2-compliance.cpp
> index b5bde2f4..d0441651 100644
> --- a/utils/v4l2-compliance/v4l2-compliance.cpp
> +++ b/utils/v4l2-compliance/v4l2-compliance.cpp
> @@ -1250,6 +1250,7 @@ void testNode(struct node &node, struct node &node_m2m_cap, struct node &expbuf_
>                 node.frmsizes.clear();
>                 node.frmsizes_count.clear();
>                 node.has_frmintervals = false;
> +               node.has_enc_cap_frame_interval = false;
>                 node.valid_buftypes = 0;
>                 node.valid_memorytype = 0;
>                 node.buf_caps = 0;
> diff --git a/utils/v4l2-compliance/v4l2-compliance.h b/utils/v4l2-compliance/v4l2-compliance.h
> index 21e31872..ae10bdf9 100644
> --- a/utils/v4l2-compliance/v4l2-compliance.h
> +++ b/utils/v4l2-compliance/v4l2-compliance.h
> @@ -141,6 +141,7 @@ struct base_node {
>         frmsizes_set frmsizes;
>         frmsizes_count_map frmsizes_count;
>         bool has_frmintervals;
> +       bool has_enc_cap_frame_interval;
>         __u32 valid_buftypes;
>         __u32 valid_buftype;
>         __u32 valid_memorytype;
> diff --git a/utils/v4l2-compliance/v4l2-test-formats.cpp b/utils/v4l2-compliance/v4l2-test-formats.cpp
> index 3dfc593e..edf1536e 100644
> --- a/utils/v4l2-compliance/v4l2-test-formats.cpp
> +++ b/utils/v4l2-compliance/v4l2-test-formats.cpp
> @@ -71,6 +71,8 @@ static int testEnumFrameIntervals(struct node *node, __u32 pixfmt,
>                 ret = doioctl(node, VIDIOC_ENUM_FRAMEINTERVALS, &frmival);
>                 if (ret == ENOTTY)
>                         return ret;
> +               // M2M devices don't support this, except for stateful encoders
> +               fail_on_test(node->is_m2m && !(node->codec_mask & STATEFUL_ENCODER));
>                 if (f == 0 && ret == EINVAL) {
>                         if (type == V4L2_FRMSIZE_TYPE_DISCRETE)
>                                 warn("found framesize %dx%d, but no frame intervals\n", w, h);
> @@ -264,16 +266,22 @@ static int testEnumFormatsType(struct node *node, unsigned type)
>                         return fail("drivers must never set the emulated flag\n");
>                 if (fmtdesc.flags & ~(V4L2_FMT_FLAG_COMPRESSED | V4L2_FMT_FLAG_EMULATED |
>                                       V4L2_FMT_FLAG_CONTINUOUS_BYTESTREAM |
> -                                     V4L2_FMT_FLAG_DYN_RESOLUTION))
> +                                     V4L2_FMT_FLAG_DYN_RESOLUTION |
> +                                     V4L2_FMT_FLAG_ENC_CAP_FRAME_INTERVAL))
>                         return fail("unknown flag %08x returned\n", fmtdesc.flags);
>                 if (!(fmtdesc.flags & V4L2_FMT_FLAG_COMPRESSED))
>                         fail_on_test(fmtdesc.flags & (V4L2_FMT_FLAG_CONTINUOUS_BYTESTREAM |
> -                                                     V4L2_FMT_FLAG_DYN_RESOLUTION));
> +                                                     V4L2_FMT_FLAG_DYN_RESOLUTION |
> +                                                     V4L2_FMT_FLAG_ENC_CAP_FRAME_INTERVAL));
>                 ret = testEnumFrameSizes(node, fmtdesc.pixelformat);
>                 if (ret)
>                         fail_on_test(node->codec_mask & STATEFUL_ENCODER);
>                 if (ret && ret != ENOTTY)
>                         return ret;
> +               if (fmtdesc.flags & V4L2_FMT_FLAG_ENC_CAP_FRAME_INTERVAL) {
> +                       fail_on_test(!(node->codec_mask & STATEFUL_ENCODER));
> +                       node->has_enc_cap_frame_interval = true;
> +               }
>                 f++;
>                 if (type == V4L2_BUF_TYPE_PRIVATE)
>                         continue;
> @@ -1222,6 +1230,7 @@ int testSlicedVBICap(struct node *node)
>
>  static int testParmStruct(struct node *node, struct v4l2_streamparm &parm)
>  {
> +       bool is_stateful_enc = node->codec_mask & STATEFUL_ENCODER;
>         struct v4l2_captureparm *cap = &parm.parm.capture;
>         struct v4l2_outputparm *out = &parm.parm.output;
>         int ret;
> @@ -1239,6 +1248,7 @@ static int testParmStruct(struct node *node, struct v4l2_streamparm &parm)
>                         fail_on_test(!cap->readbuffers);
>                 fail_on_test(cap->capability & ~V4L2_CAP_TIMEPERFRAME);
>                 fail_on_test(node->has_frmintervals && !cap->capability);
> +               fail_on_test(is_stateful_enc && !cap->capability);
>                 fail_on_test(cap->capturemode & ~V4L2_MODE_HIGHQUALITY);
>                 if (cap->capturemode & V4L2_MODE_HIGHQUALITY)
>                         warn("V4L2_MODE_HIGHQUALITY is poorly defined\n");
> @@ -1257,6 +1267,7 @@ static int testParmStruct(struct node *node, struct v4l2_streamparm &parm)
>                 else if (node->g_caps() & V4L2_CAP_STREAMING)
>                         fail_on_test(!out->writebuffers);
>                 fail_on_test(out->capability & ~V4L2_CAP_TIMEPERFRAME);
> +               fail_on_test(is_stateful_enc && !out->capability);
>                 fail_on_test(out->outputmode);
>                 fail_on_test(out->extendedmode);
>                 if (out->capability & V4L2_CAP_TIMEPERFRAME)
> @@ -1271,6 +1282,7 @@ static int testParmStruct(struct node *node, struct v4l2_streamparm &parm)
>  static int testParmType(struct node *node, unsigned type)
>  {
>         struct v4l2_streamparm parm;
> +       bool is_stateful_enc = node->codec_mask & STATEFUL_ENCODER;
>         int ret;
>
>         memset(&parm, 0, sizeof(parm));
> @@ -1288,10 +1300,10 @@ static int testParmType(struct node *node, unsigned type)
>         case V4L2_BUF_TYPE_VIDEO_CAPTURE_MPLANE:
>         case V4L2_BUF_TYPE_VIDEO_OUTPUT:
>         case V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE:
> -               if (type && (node->g_caps() & buftype2cap[type]))
> +               if (node->g_caps() & buftype2cap[type]) {
>                         fail_on_test(ret && node->has_frmintervals);
> -               if (ret)
> -                       break;
> +                       fail_on_test(ret && node->has_enc_cap_frame_interval);
> +               }
>                 break;
>         default:
>                 fail_on_test(ret == 0);
> @@ -1300,8 +1312,15 @@ static int testParmType(struct node *node, unsigned type)
>                 fail_on_test(!doioctl(node, VIDIOC_S_PARM, &parm));
>                 break;
>         }
> +       if (ret == ENOTTY) {
> +               memset(&parm, 0, sizeof(parm));
> +               parm.type = type;
> +               fail_on_test(doioctl(node, VIDIOC_S_PARM, &parm) != ENOTTY);
> +       }
>         if (ret == ENOTTY)
>                 return ret;
> +       // M2M devices don't support this, except for stateful encoders
> +       fail_on_test(node->is_m2m && !is_stateful_enc);
>         if (ret == EINVAL)
>                 return ENOTTY;
>         if (ret)
> @@ -1327,11 +1346,29 @@ static int testParmType(struct node *node, unsigned type)
>                 cap = parm.parm.output.capability;
>         else
>                 cap = parm.parm.capture.capability;
> -       fail_on_test(ret && node->has_frmintervals);
> -       if (!ret && (cap & V4L2_CAP_TIMEPERFRAME) && !node->has_frmintervals)
> +
> +       if (is_stateful_enc) {
> +               fail_on_test(ret && node->has_enc_cap_frame_interval);
> +               if (V4L2_TYPE_IS_OUTPUT(type))
> +                       fail_on_test(ret);
> +       } else {
> +               fail_on_test(ret && node->has_frmintervals);
> +       }
> +
> +       /*
> +        * Stateful encoders can support S_PARM without ENUM_FRAMEINTERVALS.
> +        * being present. In that case the limits of the chosen codec apply.
> +        */
> +       if (!ret && (cap & V4L2_CAP_TIMEPERFRAME) && !node->has_frmintervals && !is_stateful_enc)
>                 warn("S_PARM is supported for buftype %d, but not for ENUM_FRAMEINTERVALS\n", type);
>         if (ret == ENOTTY)
>                 return 0;
> +       /*
> +        * S_PARM(CAPTURE) is optional for stateful encoders, so EINVAL is a
> +        * valid error code in that case.
> +        */
> +       if (is_stateful_enc && !V4L2_TYPE_IS_OUTPUT(type) && ret == -EINVAL)
> +               return 0;
>         if (ret)
>                 return fail("got error %d when setting parms for buftype %d\n", ret, type);
>         fail_on_test(parm.type != type);
> -------------------------- cut here ------------------------------

  reply	other threads:[~2020-07-07  4:44 UTC|newest]

Thread overview: 28+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-06-26  8:04 [PATCH v2 00/18] media: mtk-vcodec: venc: support for MT8183 and v4l2-compliance fixes Alexandre Courbot
2020-06-26  8:04 ` [PATCH v2 01/18] media: mtk-vcodec: abstract firmware interface Alexandre Courbot
2020-07-01  3:19   ` Tiffany Lin
2020-06-26  8:04 ` [PATCH v2 02/18] media: mtk-vcodec: add SCP firmware ops Alexandre Courbot
2020-06-26  8:04 ` [PATCH v2 03/18] media: mtk-vcodec: venc: support SCP firmware Alexandre Courbot
2020-06-26  8:04 ` [PATCH v2 04/18] media: mtk-vcodec: venc: handle firmware version field Alexandre Courbot
2020-06-26  8:04 ` [PATCH v2 05/18] media: mtk-vcodec: venc: specify bitrate range per-chip Alexandre Courbot
2020-06-26  8:04 ` [PATCH v2 06/18] media: mtk-vcodec: venc: specify supported formats per-chip Alexandre Courbot
2020-07-01  3:15   ` Tiffany Lin
2020-06-26  8:04 ` [PATCH v2 07/18] media: mtk-vcodec: add support for MT8183 encoder Alexandre Courbot
2020-06-26  8:04 ` [PATCH v2 08/18] media: dt-bindings: mtk-vcodec: specify SCP node Alexandre Courbot
2020-06-26  8:04 ` [PATCH v2 09/18] media: dt-bindings: mtk-vcodec: document mediatek,mt8183-vcodec-enc Alexandre Courbot
2020-06-26  8:04 ` [PATCH v2 10/18] Revert "media: mtk-vcodec: Remove extra area allocation in an input buffer on encoding" Alexandre Courbot
2020-07-01  3:30   ` Tiffany Lin
2020-06-26  8:04 ` [PATCH v2 11/18] media: mtk-vcodec: venc support MIN_OUTPUT_BUFFERS control Alexandre Courbot
2020-07-01  3:30   ` Tiffany Lin
2020-06-26  8:04 ` [PATCH v2 12/18] media: mtk-vcodec: venc: set OUTPUT buffers field to V4L2_FIELD_NONE Alexandre Courbot
2020-07-01  3:46   ` Tiffany Lin
2020-06-26  8:04 ` [PATCH v2 13/18] media: mtk-vcodec: venc: use platform data for ENUM_FRAMESIZES Alexandre Courbot
2020-06-26  8:04 ` [PATCH v2 14/18] media: mtk-vcodec: venc: support ENUM_FRAMESIZES on OUTPUT formats Alexandre Courbot
2020-06-26  8:04 ` [PATCH v2 15/18] media: mtk-vcodec: venc: support G_PARM on CAPTURE queue Alexandre Courbot
2020-06-26  8:04 ` [PATCH v2 16/18] media: mtk-vcodec: venc: make S_PARM return -ENOTTY for " Alexandre Courbot
2020-07-03  8:30   ` Hans Verkuil
2020-07-04 12:36     ` Alexandre Courbot
2020-07-06 12:41       ` Hans Verkuil
2020-07-07  4:44         ` Alexandre Courbot [this message]
2020-06-26  8:04 ` [PATCH v2 17/18] media: mtk-vcodec: venc: set default time per frame Alexandre Courbot
2020-06-26  8:04 ` [PATCH v2 18/18] media: mtk-vcodec: venc: fix invalid time per frame in S_PARM Alexandre Courbot

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to='CAPBb6MWT=aALmuQLkaKLYeXiyPViQCDZLdUUGdi_NFFXCo6Vcw@mail.gmail.com' \
    --to=acourbot@chromium.org \
    --cc=andrew-ct.chen@mediatek.com \
    --cc=gtk_ruiwang@mediatek.com \
    --cc=hverkuil@xs4all.nl \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-media@vger.kernel.org \
    --cc=linux-mediatek@lists.infradead.org \
    --cc=maoguang.meng@mediatek.com \
    --cc=pihsun@chromium.org \
    --cc=tiffany.lin@mediatek.com \
    --cc=yunfei.dong@mediatek.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).