From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754484AbdDFPn0 (ORCPT ); Thu, 6 Apr 2017 11:43:26 -0400 Received: from lb1-smtp-cloud3.xs4all.net ([194.109.24.22]:44960 "EHLO lb1-smtp-cloud3.xs4all.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753737AbdDFPnS (ORCPT ); Thu, 6 Apr 2017 11:43:18 -0400 Subject: Re: [PATCH] [media] imx: csi: retain current field order and colorimetry setting as default To: Philipp Zabel References: <1490661656-10318-1-git-send-email-steve_longerbeam@mentor.com> <1490661656-10318-22-git-send-email-steve_longerbeam@mentor.com> <1491486929.2392.29.camel@pengutronix.de> <0f9690f8-c7f6-59ff-9e3e-123af9972d4b@xs4all.nl> <1491490451.2392.70.camel@pengutronix.de> Cc: Steve Longerbeam , robh+dt@kernel.org, mark.rutland@arm.com, shawnguo@kernel.org, kernel@pengutronix.de, fabio.estevam@nxp.com, linux@armlinux.org.uk, mchehab@kernel.org, nick@shmanahar.org, markus.heiser@darmarIT.de, laurent.pinchart+renesas@ideasonboard.com, bparrot@ti.com, geert@linux-m68k.org, arnd@arndb.de, sudipm.mukherjee@gmail.com, minghsiu.tsai@mediatek.com, tiffany.lin@mediatek.com, jean-christophe.trotin@st.com, horms+renesas@verge.net.au, niklas.soderlund+renesas@ragnatech.se, robert.jarzmik@free.fr, songjun.wu@microchip.com, andrew-ct.chen@mediatek.com, gregkh@linuxfoundation.org, shuah@kernel.org, sakari.ailus@linux.intel.com, pavel@ucw.cz, devicetree@vger.kernel.org, linux-kernel@vger.kernel.org, linux-arm-kernel@lists.infradead.org, linux-media@vger.kernel.org, devel@driverdev.osuosl.org, Steve Longerbeam From: Hans Verkuil Message-ID: <59e72974-bfb0-6061-8b13-5f13f8723ba6@xs4all.nl> Date: Thu, 6 Apr 2017 17:43:00 +0200 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101 Thunderbird/45.8.0 MIME-Version: 1.0 In-Reply-To: <1491490451.2392.70.camel@pengutronix.de> Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 04/06/2017 04:54 PM, Philipp Zabel wrote: > On Thu, 2017-04-06 at 16:20 +0200, Hans Verkuil wrote: >> On 04/06/2017 03:55 PM, Philipp Zabel wrote: >>> If the the field order is set to ANY in set_fmt, choose the currently >>> set field order. If the colorspace is set to DEFAULT, choose the current >>> colorspace. If any of xfer_func, ycbcr_enc or quantization are set to >>> DEFAULT, either choose the current setting, or the default setting for the >>> new colorspace, if non-DEFAULT colorspace was given. >>> >>> This allows to let field order and colorimetry settings be propagated >>> from upstream by calling media-ctl on the upstream entity source pad, >>> and then call media-ctl on the sink pad to manually set the input frame >>> interval, without changing the already set field order and colorimetry >>> information. >>> >>> Signed-off-by: Philipp Zabel >>> --- >>> This is based on imx-media-staging-md-v14, and it is supposed to allow >>> configuring the pipeline with media-ctl like this: >>> >>> 1) media-ctl --set-v4l2 "'tc358743 1-000f':0[fmt:UYVY8_1X16/1920x1080]" >>> 2) media-ctl --set-v4l2 "'imx6-mipi-csi2':1[fmt:UYVY8_1X16/1920x108]" >>> 3) media-ctl --set-v4l2 "'ipu1_csi0_mux':2[fmt:UYVY8_1X16/1920x1080]" >>> 4) media-ctl --set-v4l2 "'ipu1_csi0':0[fmt:UYVY8_1X16/1920x1080@1/60]" >>> 5) media-ctl --set-v4l2 "'ipu1_csi0':2[fmt:AYUV32/1920x1080@1/30]" >>> >>> Without having step 4) overwrite the colorspace and field order set on >>> 'ipu1_csi0':0 by the propagation in step 3). >>> --- >>> drivers/staging/media/imx/imx-media-csi.c | 34 +++++++++++++++++++++++++++++++ >>> 1 file changed, 34 insertions(+) >>> >>> diff --git a/drivers/staging/media/imx/imx-media-csi.c b/drivers/staging/media/imx/imx-media-csi.c >>> index 64dc454f6b371..d94ce1de2bf05 100644 >>> --- a/drivers/staging/media/imx/imx-media-csi.c >>> +++ b/drivers/staging/media/imx/imx-media-csi.c >>> @@ -1325,6 +1325,40 @@ static int csi_set_fmt(struct v4l2_subdev *sd, >>> csi_try_fmt(priv, sensor, cfg, sdformat, crop, compose, &cc); >>> >>> fmt = __csi_get_fmt(priv, cfg, sdformat->pad, sdformat->which); >>> + >>> + /* Retain current field setting as default */ >>> + if (sdformat->format.field == V4L2_FIELD_ANY) >>> + sdformat->format.field = fmt->field; >> >> sdformat->format.field should never be FIELD_ANY. If it is, then that's a >> subdev bug and I'm pretty sure FIELD_NONE was intended. > > This is the subdev. sdformat is passed in from userspace, so we have to > deal with it being set to ANY. I'm trying hard right now not to return > ANY though. The values in sdformat->format are applied to fmt down > below. Do you have a git tree with this patch? It is really hard to review without having the full imx-media-csi.c source. I think one problem is that it is not clearly defined how subdevs and colorspace information should work. Regards, Hans > >>> + >>> + /* Retain current colorspace setting as default */ >>> + if (sdformat->format.colorspace == V4L2_COLORSPACE_DEFAULT) { >>> + sdformat->format.colorspace = fmt->colorspace; >> >> No! Subdevs should never return COLORSPACE_DEFAULT. If they do, then fix >> them. If this happens a lot (I'm not sure how reliably subdevs fill this >> in) you could set it to COLORSPACE_RAW. Perhaps with a WARN_ON_ONCE. > > Same here, if userspace calls VIDIOC_SUBDEV_S_FMT with DEFAULT > colorspace, I don't want to return that unchanged, but overwrite it with > the currently set value. > >>> + if (sdformat->format.xfer_func == V4L2_XFER_FUNC_DEFAULT) >>> + sdformat->format.xfer_func = fmt->xfer_func; >>> + if (sdformat->format.ycbcr_enc == V4L2_YCBCR_ENC_DEFAULT) >>> + sdformat->format.ycbcr_enc = fmt->ycbcr_enc; >>> + if (sdformat->format.quantization == V4L2_QUANTIZATION_DEFAULT) >>> + sdformat->format.quantization = fmt->quantization; >> >> Nack. This is meaningless. > > It isn't. It may be wrong, but it is not without effect. If the > colorimetry info (in fmt) is currently set to something that is > non-standard, calling VIDIOC_SUBDEV_S_FMT with DEFAULT in xfer_func, > ycbcr_enc or quantization will cause those old values to be retained, > instead of overwriting them to DEFAULT. > >>> + } else { >>> + if (sdformat->format.xfer_func == V4L2_XFER_FUNC_DEFAULT) { >>> + sdformat->format.xfer_func = >>> + V4L2_MAP_XFER_FUNC_DEFAULT( >>> + sdformat->format.colorspace); >>> + } >>> + if (sdformat->format.ycbcr_enc == V4L2_YCBCR_ENC_DEFAULT) { >>> + sdformat->format.ycbcr_enc = >>> + V4L2_MAP_YCBCR_ENC_DEFAULT( >>> + sdformat->format.colorspace); >>> + } >>> + if (sdformat->format.quantization == V4L2_QUANTIZATION_DEFAULT) { >>> + sdformat->format.quantization = >>> + V4L2_MAP_QUANTIZATION_DEFAULT( >>> + cc->cs != IPUV3_COLORSPACE_YUV, >>> + sdformat->format.colorspace, >>> + sdformat->format.ycbcr_enc); >>> + } >> >> This isn't wrong, but it is perfectly fine to keep the DEFAULT here and let >> the application call V4L2_MAP_. >> >> I get the feeling this patch is a workaround for subdev errors. Either that, >> or the commit log doesn't give me enough information to really understand the >> problem that's being addressed here. > > It's the latter. I should have written VIDIOC_SUBDEV_S_FMT instead of > just set_fmt in the comment. > >> Regards, >> >> Hans >> >>> + } >>> + >>> *fmt = sdformat->format; > > Here sdformat is applied to the subdev pad. > > regards > Philipp > From mboxrd@z Thu Jan 1 00:00:00 1970 From: Hans Verkuil Subject: Re: [PATCH] [media] imx: csi: retain current field order and colorimetry setting as default Date: Thu, 6 Apr 2017 17:43:00 +0200 Message-ID: <59e72974-bfb0-6061-8b13-5f13f8723ba6@xs4all.nl> References: <1490661656-10318-1-git-send-email-steve_longerbeam@mentor.com> <1490661656-10318-22-git-send-email-steve_longerbeam@mentor.com> <1491486929.2392.29.camel@pengutronix.de> <0f9690f8-c7f6-59ff-9e3e-123af9972d4b@xs4all.nl> <1491490451.2392.70.camel@pengutronix.de> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1491490451.2392.70.camel@pengutronix.de> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: driverdev-devel-bounces@linuxdriverproject.org Sender: "devel" To: Philipp Zabel Cc: mark.rutland@arm.com, andrew-ct.chen@mediatek.com, minghsiu.tsai@mediatek.com, nick@shmanahar.org, songjun.wu@microchip.com, Steve Longerbeam , pavel@ucw.cz, robert.jarzmik@free.fr, devel@driverdev.osuosl.org, markus.heiser@darmarIT.de, laurent.pinchart+renesas@ideasonboard.com, shuah@kernel.org, linux@armlinux.org.uk, geert@linux-m68k.org, Steve Longerbeam , linux-media@vger.kernel.org, devicetree@vger.kernel.org, sakari.ailus@linux.intel.com, arnd@arndb.de, mchehab@kernel.org, bparrot@ti.com, robh+dt@kernel.org, horms+renesas@verge.net.au, tiffany.lin@mediatek.com, linux-arm-kernel@lists.infradead.org, niklas.soderlund+renesas@ragnatech.se, gregkh@linuxfoundation.org, linux-kernel@vger.kernel.org, jean-christophe.trotin@st.com, kernel@pengutronix.de, fabio.estevam@nxp.com, shawnguo@kernel.org, sudipm.mukherjee@gmail.c List-Id: devicetree@vger.kernel.org On 04/06/2017 04:54 PM, Philipp Zabel wrote: > On Thu, 2017-04-06 at 16:20 +0200, Hans Verkuil wrote: >> On 04/06/2017 03:55 PM, Philipp Zabel wrote: >>> If the the field order is set to ANY in set_fmt, choose the currently >>> set field order. If the colorspace is set to DEFAULT, choose the current >>> colorspace. If any of xfer_func, ycbcr_enc or quantization are set to >>> DEFAULT, either choose the current setting, or the default setting for the >>> new colorspace, if non-DEFAULT colorspace was given. >>> >>> This allows to let field order and colorimetry settings be propagated >>> from upstream by calling media-ctl on the upstream entity source pad, >>> and then call media-ctl on the sink pad to manually set the input frame >>> interval, without changing the already set field order and colorimetry >>> information. >>> >>> Signed-off-by: Philipp Zabel >>> --- >>> This is based on imx-media-staging-md-v14, and it is supposed to allow >>> configuring the pipeline with media-ctl like this: >>> >>> 1) media-ctl --set-v4l2 "'tc358743 1-000f':0[fmt:UYVY8_1X16/1920x1080]" >>> 2) media-ctl --set-v4l2 "'imx6-mipi-csi2':1[fmt:UYVY8_1X16/1920x108]" >>> 3) media-ctl --set-v4l2 "'ipu1_csi0_mux':2[fmt:UYVY8_1X16/1920x1080]" >>> 4) media-ctl --set-v4l2 "'ipu1_csi0':0[fmt:UYVY8_1X16/1920x1080@1/60]" >>> 5) media-ctl --set-v4l2 "'ipu1_csi0':2[fmt:AYUV32/1920x1080@1/30]" >>> >>> Without having step 4) overwrite the colorspace and field order set on >>> 'ipu1_csi0':0 by the propagation in step 3). >>> --- >>> drivers/staging/media/imx/imx-media-csi.c | 34 +++++++++++++++++++++++++++++++ >>> 1 file changed, 34 insertions(+) >>> >>> diff --git a/drivers/staging/media/imx/imx-media-csi.c b/drivers/staging/media/imx/imx-media-csi.c >>> index 64dc454f6b371..d94ce1de2bf05 100644 >>> --- a/drivers/staging/media/imx/imx-media-csi.c >>> +++ b/drivers/staging/media/imx/imx-media-csi.c >>> @@ -1325,6 +1325,40 @@ static int csi_set_fmt(struct v4l2_subdev *sd, >>> csi_try_fmt(priv, sensor, cfg, sdformat, crop, compose, &cc); >>> >>> fmt = __csi_get_fmt(priv, cfg, sdformat->pad, sdformat->which); >>> + >>> + /* Retain current field setting as default */ >>> + if (sdformat->format.field == V4L2_FIELD_ANY) >>> + sdformat->format.field = fmt->field; >> >> sdformat->format.field should never be FIELD_ANY. If it is, then that's a >> subdev bug and I'm pretty sure FIELD_NONE was intended. > > This is the subdev. sdformat is passed in from userspace, so we have to > deal with it being set to ANY. I'm trying hard right now not to return > ANY though. The values in sdformat->format are applied to fmt down > below. Do you have a git tree with this patch? It is really hard to review without having the full imx-media-csi.c source. I think one problem is that it is not clearly defined how subdevs and colorspace information should work. Regards, Hans > >>> + >>> + /* Retain current colorspace setting as default */ >>> + if (sdformat->format.colorspace == V4L2_COLORSPACE_DEFAULT) { >>> + sdformat->format.colorspace = fmt->colorspace; >> >> No! Subdevs should never return COLORSPACE_DEFAULT. If they do, then fix >> them. If this happens a lot (I'm not sure how reliably subdevs fill this >> in) you could set it to COLORSPACE_RAW. Perhaps with a WARN_ON_ONCE. > > Same here, if userspace calls VIDIOC_SUBDEV_S_FMT with DEFAULT > colorspace, I don't want to return that unchanged, but overwrite it with > the currently set value. > >>> + if (sdformat->format.xfer_func == V4L2_XFER_FUNC_DEFAULT) >>> + sdformat->format.xfer_func = fmt->xfer_func; >>> + if (sdformat->format.ycbcr_enc == V4L2_YCBCR_ENC_DEFAULT) >>> + sdformat->format.ycbcr_enc = fmt->ycbcr_enc; >>> + if (sdformat->format.quantization == V4L2_QUANTIZATION_DEFAULT) >>> + sdformat->format.quantization = fmt->quantization; >> >> Nack. This is meaningless. > > It isn't. It may be wrong, but it is not without effect. If the > colorimetry info (in fmt) is currently set to something that is > non-standard, calling VIDIOC_SUBDEV_S_FMT with DEFAULT in xfer_func, > ycbcr_enc or quantization will cause those old values to be retained, > instead of overwriting them to DEFAULT. > >>> + } else { >>> + if (sdformat->format.xfer_func == V4L2_XFER_FUNC_DEFAULT) { >>> + sdformat->format.xfer_func = >>> + V4L2_MAP_XFER_FUNC_DEFAULT( >>> + sdformat->format.colorspace); >>> + } >>> + if (sdformat->format.ycbcr_enc == V4L2_YCBCR_ENC_DEFAULT) { >>> + sdformat->format.ycbcr_enc = >>> + V4L2_MAP_YCBCR_ENC_DEFAULT( >>> + sdformat->format.colorspace); >>> + } >>> + if (sdformat->format.quantization == V4L2_QUANTIZATION_DEFAULT) { >>> + sdformat->format.quantization = >>> + V4L2_MAP_QUANTIZATION_DEFAULT( >>> + cc->cs != IPUV3_COLORSPACE_YUV, >>> + sdformat->format.colorspace, >>> + sdformat->format.ycbcr_enc); >>> + } >> >> This isn't wrong, but it is perfectly fine to keep the DEFAULT here and let >> the application call V4L2_MAP_. >> >> I get the feeling this patch is a workaround for subdev errors. Either that, >> or the commit log doesn't give me enough information to really understand the >> problem that's being addressed here. > > It's the latter. I should have written VIDIOC_SUBDEV_S_FMT instead of > just set_fmt in the comment. > >> Regards, >> >> Hans >> >>> + } >>> + >>> *fmt = sdformat->format; > > Here sdformat is applied to the subdev pad. > > regards > Philipp > From mboxrd@z Thu Jan 1 00:00:00 1970 From: hverkuil@xs4all.nl (Hans Verkuil) Date: Thu, 6 Apr 2017 17:43:00 +0200 Subject: [PATCH] [media] imx: csi: retain current field order and colorimetry setting as default In-Reply-To: <1491490451.2392.70.camel@pengutronix.de> References: <1490661656-10318-1-git-send-email-steve_longerbeam@mentor.com> <1490661656-10318-22-git-send-email-steve_longerbeam@mentor.com> <1491486929.2392.29.camel@pengutronix.de> <0f9690f8-c7f6-59ff-9e3e-123af9972d4b@xs4all.nl> <1491490451.2392.70.camel@pengutronix.de> Message-ID: <59e72974-bfb0-6061-8b13-5f13f8723ba6@xs4all.nl> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On 04/06/2017 04:54 PM, Philipp Zabel wrote: > On Thu, 2017-04-06 at 16:20 +0200, Hans Verkuil wrote: >> On 04/06/2017 03:55 PM, Philipp Zabel wrote: >>> If the the field order is set to ANY in set_fmt, choose the currently >>> set field order. If the colorspace is set to DEFAULT, choose the current >>> colorspace. If any of xfer_func, ycbcr_enc or quantization are set to >>> DEFAULT, either choose the current setting, or the default setting for the >>> new colorspace, if non-DEFAULT colorspace was given. >>> >>> This allows to let field order and colorimetry settings be propagated >>> from upstream by calling media-ctl on the upstream entity source pad, >>> and then call media-ctl on the sink pad to manually set the input frame >>> interval, without changing the already set field order and colorimetry >>> information. >>> >>> Signed-off-by: Philipp Zabel >>> --- >>> This is based on imx-media-staging-md-v14, and it is supposed to allow >>> configuring the pipeline with media-ctl like this: >>> >>> 1) media-ctl --set-v4l2 "'tc358743 1-000f':0[fmt:UYVY8_1X16/1920x1080]" >>> 2) media-ctl --set-v4l2 "'imx6-mipi-csi2':1[fmt:UYVY8_1X16/1920x108]" >>> 3) media-ctl --set-v4l2 "'ipu1_csi0_mux':2[fmt:UYVY8_1X16/1920x1080]" >>> 4) media-ctl --set-v4l2 "'ipu1_csi0':0[fmt:UYVY8_1X16/1920x1080 at 1/60]" >>> 5) media-ctl --set-v4l2 "'ipu1_csi0':2[fmt:AYUV32/1920x1080 at 1/30]" >>> >>> Without having step 4) overwrite the colorspace and field order set on >>> 'ipu1_csi0':0 by the propagation in step 3). >>> --- >>> drivers/staging/media/imx/imx-media-csi.c | 34 +++++++++++++++++++++++++++++++ >>> 1 file changed, 34 insertions(+) >>> >>> diff --git a/drivers/staging/media/imx/imx-media-csi.c b/drivers/staging/media/imx/imx-media-csi.c >>> index 64dc454f6b371..d94ce1de2bf05 100644 >>> --- a/drivers/staging/media/imx/imx-media-csi.c >>> +++ b/drivers/staging/media/imx/imx-media-csi.c >>> @@ -1325,6 +1325,40 @@ static int csi_set_fmt(struct v4l2_subdev *sd, >>> csi_try_fmt(priv, sensor, cfg, sdformat, crop, compose, &cc); >>> >>> fmt = __csi_get_fmt(priv, cfg, sdformat->pad, sdformat->which); >>> + >>> + /* Retain current field setting as default */ >>> + if (sdformat->format.field == V4L2_FIELD_ANY) >>> + sdformat->format.field = fmt->field; >> >> sdformat->format.field should never be FIELD_ANY. If it is, then that's a >> subdev bug and I'm pretty sure FIELD_NONE was intended. > > This is the subdev. sdformat is passed in from userspace, so we have to > deal with it being set to ANY. I'm trying hard right now not to return > ANY though. The values in sdformat->format are applied to fmt down > below. Do you have a git tree with this patch? It is really hard to review without having the full imx-media-csi.c source. I think one problem is that it is not clearly defined how subdevs and colorspace information should work. Regards, Hans > >>> + >>> + /* Retain current colorspace setting as default */ >>> + if (sdformat->format.colorspace == V4L2_COLORSPACE_DEFAULT) { >>> + sdformat->format.colorspace = fmt->colorspace; >> >> No! Subdevs should never return COLORSPACE_DEFAULT. If they do, then fix >> them. If this happens a lot (I'm not sure how reliably subdevs fill this >> in) you could set it to COLORSPACE_RAW. Perhaps with a WARN_ON_ONCE. > > Same here, if userspace calls VIDIOC_SUBDEV_S_FMT with DEFAULT > colorspace, I don't want to return that unchanged, but overwrite it with > the currently set value. > >>> + if (sdformat->format.xfer_func == V4L2_XFER_FUNC_DEFAULT) >>> + sdformat->format.xfer_func = fmt->xfer_func; >>> + if (sdformat->format.ycbcr_enc == V4L2_YCBCR_ENC_DEFAULT) >>> + sdformat->format.ycbcr_enc = fmt->ycbcr_enc; >>> + if (sdformat->format.quantization == V4L2_QUANTIZATION_DEFAULT) >>> + sdformat->format.quantization = fmt->quantization; >> >> Nack. This is meaningless. > > It isn't. It may be wrong, but it is not without effect. If the > colorimetry info (in fmt) is currently set to something that is > non-standard, calling VIDIOC_SUBDEV_S_FMT with DEFAULT in xfer_func, > ycbcr_enc or quantization will cause those old values to be retained, > instead of overwriting them to DEFAULT. > >>> + } else { >>> + if (sdformat->format.xfer_func == V4L2_XFER_FUNC_DEFAULT) { >>> + sdformat->format.xfer_func = >>> + V4L2_MAP_XFER_FUNC_DEFAULT( >>> + sdformat->format.colorspace); >>> + } >>> + if (sdformat->format.ycbcr_enc == V4L2_YCBCR_ENC_DEFAULT) { >>> + sdformat->format.ycbcr_enc = >>> + V4L2_MAP_YCBCR_ENC_DEFAULT( >>> + sdformat->format.colorspace); >>> + } >>> + if (sdformat->format.quantization == V4L2_QUANTIZATION_DEFAULT) { >>> + sdformat->format.quantization = >>> + V4L2_MAP_QUANTIZATION_DEFAULT( >>> + cc->cs != IPUV3_COLORSPACE_YUV, >>> + sdformat->format.colorspace, >>> + sdformat->format.ycbcr_enc); >>> + } >> >> This isn't wrong, but it is perfectly fine to keep the DEFAULT here and let >> the application call V4L2_MAP_. >> >> I get the feeling this patch is a workaround for subdev errors. Either that, >> or the commit log doesn't give me enough information to really understand the >> problem that's being addressed here. > > It's the latter. I should have written VIDIOC_SUBDEV_S_FMT instead of > just set_fmt in the comment. > >> Regards, >> >> Hans >> >>> + } >>> + >>> *fmt = sdformat->format; > > Here sdformat is applied to the subdev pad. > > regards > Philipp >