Hi Anatolij Sorry for a delay On Tue, 6 Nov 2012, Anatolij Gustschin wrote: > On Tue, 6 Nov 2012 12:45:51 +0100 (CET) > Guennadi Liakhovetski <g.liakhovetski@xxxxxx> wrote: > > > On Tue, 6 Nov 2012, Anatolij Gustschin wrote: > > > > > VIDIOC_S_GROP ioctl doesn't work, soc-camera driver reports: > > > > > > soc-camera-pdrv soc-camera-pdrv.0: S_CROP denied: getting current crop failed > > > > > > The issue is caused by checking for V4L2_BUF_TYPE_VIDEO_CAPTURE type > > > in driver's g_crop callback. This check should be in s_crop instead, > > > g_crop should just set the type field to V4L2_BUF_TYPE_VIDEO_CAPTURE > > > as other drivers do. Move the V4L2_BUF_TYPE_VIDEO_CAPTURE type check > > > to s_crop callback. > > > > I'm not sure this is correct: > > > > http://linuxtv.org/downloads/v4l-dvb-apis/vidioc-g-crop.html > > > > Or is the .g_crop() subdev operation using a different semantics? Where is > > that documented? > > I do not know if it is documented somewhere. But it seems natural to me > that a sensor driver sets the type field to V4L2_BUF_TYPE_VIDEO_CAPTURE > in its .g_crop(). A sensor is a capture device, not an output or overlay > device. And this ioctl is a query operation. > > OTOH I'm fine with this type checking in .g_crop() and it can help > to discover bugs in user space apps. The VIDIOC_G_CROP documentation > states that the type field needs to be set to the respective buffer type > when querying, so the check in .g_crop() is perfectly valid. But then > I need following patch to fix the observed issue: > > --- a/drivers/media/platform/soc_camera/soc_camera.c > +++ b/drivers/media/platform/soc_camera/soc_camera.c > @@ -902,6 +902,8 @@ static int soc_camera_s_crop(struct file *file, void *fh, > dev_dbg(icd->pdev, "S_CROP(%ux%u@%u:%u)\n", > rect->width, rect->height, rect->left, rect->top); > > + current_crop.type = a->type; > + > /* If get_crop fails, we'll let host and / or client drivers decide */ > ret = ici->ops->get_crop(icd, ¤t_crop); > > What do you think? Yes, this makes sense. Please, submit a patch. > And the type field should be checked in .s_crop() anyway, I think. It is checked in soc_camera_s_crop() just a couple of lines above the snippet above. Or what do you mean? Thanks Guennadi --- Guennadi Liakhovetski, Ph.D. Freelance Open-Source Software Developer http://www.open-technology.de/ -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
