| Message ID | 20210312124830.1344255-1-ribalda@chromium.org |
|---|---|
| Headers | show |
| Series | Fix v4l2-compliance errors | expand |
Hi Ricardo, Thank you for the patch. On Fri, Mar 12, 2021 at 01:48:29PM +0100, Ricardo Ribalda wrote: > All the entities must have a unique name. And now that we are at it, we > append the entity->id to the name to avoid collisions on multi-chain > devices. > > Fixes v4l2-compliance: > Media Controller ioctls: > fail: v4l2-test-media.cpp(205): v2_entity_names_set.find(key) != v2_entity_names_set.end() > test MEDIA_IOC_G_TOPOLOGY: FAIL > fail: v4l2-test-media.cpp(394): num_data_links != num_links > test MEDIA_IOC_ENUM_ENTITIES/LINKS: FAIL > > Signed-off-by: Ricardo Ribalda <ribalda@chromium.org> > --- > drivers/media/usb/uvc/uvc_driver.c | 21 ++++++++++++++++++++- > 1 file changed, 20 insertions(+), 1 deletion(-) > > diff --git a/drivers/media/usb/uvc/uvc_driver.c b/drivers/media/usb/uvc/uvc_driver.c > index 35873cf2773d..6c928e708615 100644 > --- a/drivers/media/usb/uvc/uvc_driver.c > +++ b/drivers/media/usb/uvc/uvc_driver.c > @@ -2154,6 +2154,18 @@ static void uvc_unregister_video(struct uvc_device *dev) > #endif > } > > +static int uvc_oterm_id(struct uvc_video_chain *chain) > +{ > + struct uvc_entity *entity; > + > + list_for_each_entry(entity, &chain->entities, chain) { > + if (UVC_ENTITY_IS_OTERM(entity)) > + return entity->id; It can also be an ITERM for output devices. You can drop this function and use stream>header.bTerminalLink below (see uvc_stream_by_id() and its usage in uvc_register_terms()). > + } > + > + return -1; > +} > + > int uvc_register_video_device(struct uvc_device *dev, > struct uvc_streaming *stream, > struct video_device *vdev, > @@ -2162,6 +2174,8 @@ int uvc_register_video_device(struct uvc_device *dev, > const struct v4l2_file_operations *fops, > const struct v4l2_ioctl_ops *ioctl_ops) > { > + char prefix[sizeof(vdev->name) - 9]; > + const char *suffix; > int ret; > > /* Initialize the video buffers queue. */ > @@ -2190,16 +2204,21 @@ int uvc_register_video_device(struct uvc_device *dev, > case V4L2_BUF_TYPE_VIDEO_CAPTURE: > default: > vdev->device_caps = V4L2_CAP_VIDEO_CAPTURE | V4L2_CAP_STREAMING; > + suffix = "video"; > break; > case V4L2_BUF_TYPE_VIDEO_OUTPUT: > vdev->device_caps = V4L2_CAP_VIDEO_OUTPUT | V4L2_CAP_STREAMING; > + suffix = "out"; I wonder if these two should be video-cap and video-out (or vid-cap and vid-out if you want to shorten them) ? > break; > case V4L2_BUF_TYPE_META_CAPTURE: > vdev->device_caps = V4L2_CAP_META_CAPTURE | V4L2_CAP_STREAMING; > + suffix = "meta"; > break; > } > > - strscpy(vdev->name, dev->name, sizeof(vdev->name)); > + strscpy(prefix, dev->name, sizeof(prefix)); > + snprintf(vdev->name, sizeof(vdev->name), "%s-%d %s", prefix, The unit ID is never negative, so %u ? > + uvc_oterm_id(stream->chain), suffix); Truncating the device name at the beginning of the video node name isn't very nice :-S How about the following ? snprintf(vdev->name, sizeof(vdev->name), "%s-%u (%s)", type_name, uvc_oterm_id(stream->chain), dev->name); with the suffix variable renamed to type_name ? Thinking some more about it, vdev->name serves two purposes in the driver: creating the entity name, and reporting the card name in querycap. The former is done in the V4L2 core, which uses vdev->name as-is. In this context, we con't need to add dev->name, it would be redundant as the media controller device already reports it. The latter is done in uvc_ioctl_querycap(). How about dropping dev->name from vdev->name, and modifying uvc_ioctl_querycap() to use dev->name instead of cap->card ? > > /* > * Set the driver data before calling video_register_device, otherwise
HI Laurent Thanks for the review On Fri, Mar 12, 2021 at 11:30 PM Laurent Pinchart <laurent.pinchart@ideasonboard.com> wrote: > > Hi Ricardo, > > Thank you for the patch. > > On Fri, Mar 12, 2021 at 01:48:29PM +0100, Ricardo Ribalda wrote: > > All the entities must have a unique name. And now that we are at it, we > > append the entity->id to the name to avoid collisions on multi-chain > > devices. > > > > Fixes v4l2-compliance: > > Media Controller ioctls: > > fail: v4l2-test-media.cpp(205): v2_entity_names_set.find(key) != v2_entity_names_set.end() > > test MEDIA_IOC_G_TOPOLOGY: FAIL > > fail: v4l2-test-media.cpp(394): num_data_links != num_links > > test MEDIA_IOC_ENUM_ENTITIES/LINKS: FAIL > > > > Signed-off-by: Ricardo Ribalda <ribalda@chromium.org> > > --- > > drivers/media/usb/uvc/uvc_driver.c | 21 ++++++++++++++++++++- > > 1 file changed, 20 insertions(+), 1 deletion(-) > > > > diff --git a/drivers/media/usb/uvc/uvc_driver.c b/drivers/media/usb/uvc/uvc_driver.c > > index 35873cf2773d..6c928e708615 100644 > > --- a/drivers/media/usb/uvc/uvc_driver.c > > +++ b/drivers/media/usb/uvc/uvc_driver.c > > @@ -2154,6 +2154,18 @@ static void uvc_unregister_video(struct uvc_device *dev) > > #endif > > } > > > > +static int uvc_oterm_id(struct uvc_video_chain *chain) > > +{ > > + struct uvc_entity *entity; > > + > > + list_for_each_entry(entity, &chain->entities, chain) { > > + if (UVC_ENTITY_IS_OTERM(entity)) > > + return entity->id; > > It can also be an ITERM for output devices. You can drop this function > and use stream>header.bTerminalLink below (see uvc_stream_by_id() and > its usage in uvc_register_terms()). > > > + } > > + > > + return -1; > > +} > > + > > int uvc_register_video_device(struct uvc_device *dev, > > struct uvc_streaming *stream, > > struct video_device *vdev, > > @@ -2162,6 +2174,8 @@ int uvc_register_video_device(struct uvc_device *dev, > > const struct v4l2_file_operations *fops, > > const struct v4l2_ioctl_ops *ioctl_ops) > > { > > + char prefix[sizeof(vdev->name) - 9]; > > + const char *suffix; > > int ret; > > > > /* Initialize the video buffers queue. */ > > @@ -2190,16 +2204,21 @@ int uvc_register_video_device(struct uvc_device *dev, > > case V4L2_BUF_TYPE_VIDEO_CAPTURE: > > default: > > vdev->device_caps = V4L2_CAP_VIDEO_CAPTURE | V4L2_CAP_STREAMING; > > + suffix = "video"; > > break; > > case V4L2_BUF_TYPE_VIDEO_OUTPUT: > > vdev->device_caps = V4L2_CAP_VIDEO_OUTPUT | V4L2_CAP_STREAMING; > > + suffix = "out"; > > I wonder if these two should be video-cap and video-out (or vid-cap and > vid-out if you want to shorten them) ? > > > break; > > case V4L2_BUF_TYPE_META_CAPTURE: > > vdev->device_caps = V4L2_CAP_META_CAPTURE | V4L2_CAP_STREAMING; > > + suffix = "meta"; > > break; > > } > > > > - strscpy(vdev->name, dev->name, sizeof(vdev->name)); > > + strscpy(prefix, dev->name, sizeof(prefix)); > > + snprintf(vdev->name, sizeof(vdev->name), "%s-%d %s", prefix, > > The unit ID is never negative, so %u ? > > > + uvc_oterm_id(stream->chain), suffix); > > Truncating the device name at the beginning of the video node name isn't > very nice :-S How about the following ? > > snprintf(vdev->name, sizeof(vdev->name), "%s-%u (%s)", type_name, > uvc_oterm_id(stream->chain), dev->name); > > with the suffix variable renamed to type_name ? > > Thinking some more about it, vdev->name serves two purposes in the > driver: creating the entity name, and reporting the card name in > querycap. The former is done in the V4L2 core, which uses vdev->name > as-is. In this context, we con't need to add dev->name, it would be > redundant as the media controller device already reports it. The latter > is done in uvc_ioctl_querycap(). How about dropping dev->name from > vdev->name, and modifying uvc_ioctl_querycap() to use dev->name instead > of cap->card ? > Something like ? https://git.kernel.org/pub/scm/linux/kernel/git/ribalda/linux.git/commit/?h=uvc-compliance-v4&id=d4f7363455837116268152c96bf4b78d9761ad1e https://git.kernel.org/pub/scm/linux/kernel/git/ribalda/linux.git/commit/?h=uvc-compliance-v4&id=ee3916f12b30f56c03d5622ba8a599b9c610a055 I need to work on the V4L2_CTRL_FLAG_GRABBED issue and then I will send the whole v4 series that can pass all the v4l2-compliance test :) Thanks! > > > > /* > > * Set the driver data before calling video_register_device, otherwise > > -- > Regards, > > Laurent Pinchart
Hi Laurent Thanks! On Sat, Mar 13, 2021 at 12:29 AM Laurent Pinchart <laurent.pinchart@ideasonboard.com> wrote: > > Hi Ricardo, > > On Sat, Mar 13, 2021 at 12:17:50AM +0100, Ricardo Ribalda Delgado wrote: > > On Fri, Mar 12, 2021 at 11:30 PM Laurent Pinchart wrote: > > > On Fri, Mar 12, 2021 at 01:48:29PM +0100, Ricardo Ribalda wrote: > > > > All the entities must have a unique name. And now that we are at it, we > > > > append the entity->id to the name to avoid collisions on multi-chain > > > > devices. > > > > > > > > Fixes v4l2-compliance: > > > > Media Controller ioctls: > > > > fail: v4l2-test-media.cpp(205): v2_entity_names_set.find(key) != v2_entity_names_set.end() > > > > test MEDIA_IOC_G_TOPOLOGY: FAIL > > > > fail: v4l2-test-media.cpp(394): num_data_links != num_links > > > > test MEDIA_IOC_ENUM_ENTITIES/LINKS: FAIL > > > > > > > > Signed-off-by: Ricardo Ribalda <ribalda@chromium.org> > > > > --- > > > > drivers/media/usb/uvc/uvc_driver.c | 21 ++++++++++++++++++++- > > > > 1 file changed, 20 insertions(+), 1 deletion(-) > > > > > > > > diff --git a/drivers/media/usb/uvc/uvc_driver.c b/drivers/media/usb/uvc/uvc_driver.c > > > > index 35873cf2773d..6c928e708615 100644 > > > > --- a/drivers/media/usb/uvc/uvc_driver.c > > > > +++ b/drivers/media/usb/uvc/uvc_driver.c > > > > @@ -2154,6 +2154,18 @@ static void uvc_unregister_video(struct uvc_device *dev) > > > > #endif > > > > } > > > > > > > > +static int uvc_oterm_id(struct uvc_video_chain *chain) > > > > +{ > > > > + struct uvc_entity *entity; > > > > + > > > > + list_for_each_entry(entity, &chain->entities, chain) { > > > > + if (UVC_ENTITY_IS_OTERM(entity)) > > > > + return entity->id; > > > > > > It can also be an ITERM for output devices. You can drop this function > > > and use stream>header.bTerminalLink below (see uvc_stream_by_id() and > > > its usage in uvc_register_terms()). > > > > > > > + } > > > > + > > > > + return -1; > > > > +} > > > > + > > > > int uvc_register_video_device(struct uvc_device *dev, > > > > struct uvc_streaming *stream, > > > > struct video_device *vdev, > > > > @@ -2162,6 +2174,8 @@ int uvc_register_video_device(struct uvc_device *dev, > > > > const struct v4l2_file_operations *fops, > > > > const struct v4l2_ioctl_ops *ioctl_ops) > > > > { > > > > + char prefix[sizeof(vdev->name) - 9]; > > > > + const char *suffix; > > > > int ret; > > > > > > > > /* Initialize the video buffers queue. */ > > > > @@ -2190,16 +2204,21 @@ int uvc_register_video_device(struct uvc_device *dev, > > > > case V4L2_BUF_TYPE_VIDEO_CAPTURE: > > > > default: > > > > vdev->device_caps = V4L2_CAP_VIDEO_CAPTURE | V4L2_CAP_STREAMING; > > > > + suffix = "video"; > > > > break; > > > > case V4L2_BUF_TYPE_VIDEO_OUTPUT: > > > > vdev->device_caps = V4L2_CAP_VIDEO_OUTPUT | V4L2_CAP_STREAMING; > > > > + suffix = "out"; > > > > > > I wonder if these two should be video-cap and video-out (or vid-cap and > > > vid-out if you want to shorten them) ? > > > > > > > break; > > > > case V4L2_BUF_TYPE_META_CAPTURE: > > > > vdev->device_caps = V4L2_CAP_META_CAPTURE | V4L2_CAP_STREAMING; > > > > + suffix = "meta"; > > > > break; > > > > } > > > > > > > > - strscpy(vdev->name, dev->name, sizeof(vdev->name)); > > > > + strscpy(prefix, dev->name, sizeof(prefix)); > > > > + snprintf(vdev->name, sizeof(vdev->name), "%s-%d %s", prefix, > > > > > > The unit ID is never negative, so %u ? > > > > > > > + uvc_oterm_id(stream->chain), suffix); > > > > > > Truncating the device name at the beginning of the video node name isn't > > > very nice :-S How about the following ? > > > > > > snprintf(vdev->name, sizeof(vdev->name), "%s-%u (%s)", type_name, > > > uvc_oterm_id(stream->chain), dev->name); > > > > > > with the suffix variable renamed to type_name ? > > > > > > Thinking some more about it, vdev->name serves two purposes in the > > > driver: creating the entity name, and reporting the card name in > > > querycap. The former is done in the V4L2 core, which uses vdev->name > > > as-is. In this context, we con't need to add dev->name, it would be > > > redundant as the media controller device already reports it. The latter > > > is done in uvc_ioctl_querycap(). How about dropping dev->name from > > > vdev->name, and modifying uvc_ioctl_querycap() to use dev->name instead > > > of cap->card ? > > > > > > > Something like ? > > https://git.kernel.org/pub/scm/linux/kernel/git/ribalda/linux.git/commit/?h=uvc-compliance-v4&id=d4f7363455837116268152c96bf4b78d9761ad1e > > I would have moved the sprintf() after the switch/case, but otherwise > it's the idea. https://git.kernel.org/pub/scm/linux/kernel/git/ribalda/linux.git/commit/?h=uvc-compliance-v4&id=fe287f9d764457f5bcd48780b99e54df161f33f6 Yep, it looks better > > > https://git.kernel.org/pub/scm/linux/kernel/git/ribalda/linux.git/commit/?h=uvc-compliance-v4&id=ee3916f12b30f56c03d5622ba8a599b9c610a055 > > Looks good. > > > I need to work on the V4L2_CTRL_FLAG_GRABBED issue and then I will > > send the whole v4 series that can pass all the v4l2-compliance test :) > > > > > > > > > > /* > > > > * Set the driver data before calling video_register_device, otherwise > > -- > Regards, > > Laurent Pinchart
In my computer I am getting this output for v4l2-compliance -m /dev/media0 -a -f Total for uvcvideo device /dev/media0: 8, Succeeded: 6, Failed: 2, Warnings: 0 Total for uvcvideo device /dev/video0: 54, Succeeded: 50, Failed: 4, Warnings: 2 Total for uvcvideo device /dev/video1: 46, Succeeded: 46, Failed: 0, Warnings: 0 Grand Total for uvcvideo device /dev/media0: 108, Succeeded: 102, Failed: 6, Warnings: 2 After fixing all of them we go down to: Total for uvcvideo device /dev/media0: 8, Succeeded: 8, Failed: 0, Warnings: 0 Total for uvcvideo device /dev/video0: 54, Succeeded: 54, Failed: 0, Warnings: 9 Total for uvcvideo device /dev/video1: 46, Succeeded: 46, Failed: 0, Warnings: 0 Grand Total for uvcvideo device /dev/media0: 108, Succeeded: 108, Failed: 0, Warnings: 9 We are still not compliant with v4l2-compliance -s: Streaming ioctls: test read/write: OK (Not Supported) test blocking wait: OK fail: v4l2-test-buffers.cpp(1265): node->streamon(q.g_type()) != EINVAL test MMAP (no poll): FAIL fail: v4l2-test-buffers.cpp(1265): node->streamon(q.g_type()) != EINVAL test MMAP (select): FAIL fail: v4l2-test-buffers.cpp(1265): node->streamon(q.g_type()) != EINVAL test MMAP (epoll): FAIL But fixing that will probably require a lot of changes in the driver that are already implemented in the vb2 helpers. It is better to continue Hans work on that: https://git.linuxtv.org/hverkuil/media_tree.git/commit/?h=uvc-4.19&id=a6a0a05f643521d29a4c1e551b0be73ce66b7108 Changelog v2 (Thanks to Hans and Laurent) - Reimplement the CTRL_CLASS as a patch on queryctl - Do not return -EIO for case 8 - Handle request bug and which_def multiclass in core Hans Verkuil (1): uvc: use vb2 ioctl and fop helpers Ricardo Ribalda (7): media: v4l2-ioctl: Fix check_ext_ctrls media: uvcvideo: Set capability in s_param media: uvcvideo: Return -EIO for control errors media: uvcvideo: set error_idx to count on EACCESS media: uvcvideo: refactor __uvc_ctrl_add_mapping media: uvcvideo: Add support for V4L2_CTRL_TYPE_CTRL_CLASS media: uvcvideo: Set a different name for the metadata entity drivers/media/usb/uvc/uvc_ctrl.c | 137 +++++++++++-- drivers/media/usb/uvc/uvc_driver.c | 36 +++- drivers/media/usb/uvc/uvc_metadata.c | 8 +- drivers/media/usb/uvc/uvc_queue.c | 131 ------------- drivers/media/usb/uvc/uvc_v4l2.c | 280 +++------------------------ drivers/media/usb/uvc/uvc_video.c | 5 + drivers/media/usb/uvc/uvcvideo.h | 34 +--- drivers/media/v4l2-core/v4l2-ioctl.c | 25 ++- 8 files changed, 206 insertions(+), 450 deletions(-)