On Mon June 18 2012 11:47:07 Laurent Pinchart wrote:
> Hi Hans,
>
> Thanks for the patch.
>
> On Sunday 10 June 2012 12:25:26 Hans Verkuil wrote:
> > From: Hans Verkuil <[email protected]>
> >
> > Add the necessary plumbing to make it possible to replace the switch by a
> > table driven implementation.
> >
> > Signed-off-by: Hans Verkuil <[email protected]>
> > ---
> > drivers/media/video/v4l2-ioctl.c | 91
> > ++++++++++++++++++++++++++++++++------ 1 file changed, 78 insertions(+), 13
> > deletions(-)
> >
> > diff --git a/drivers/media/video/v4l2-ioctl.c
> > b/drivers/media/video/v4l2-ioctl.c index a9602db..a4115ce 100644
> > --- a/drivers/media/video/v4l2-ioctl.c
> > +++ b/drivers/media/video/v4l2-ioctl.c
> > @@ -563,23 +595,40 @@ static long __video_do_ioctl(struct file *file,
> > }
> >
> > if (v4l2_is_known_ioctl(cmd)) {
> > - struct v4l2_ioctl_info *info = &v4l2_ioctls[_IOC_NR(cmd)];
> > + info = &v4l2_ioctls[_IOC_NR(cmd)];
> >
> > if (!test_bit(_IOC_NR(cmd), vfd->valid_ioctls) &&
> > !((info->flags & INFO_FL_CTRL) && vfh && vfh->ctrl_handler))
> > - return -ENOTTY;
> > + goto error;
> >
> > if (use_fh_prio && (info->flags & INFO_FL_PRIO)) {
> > ret = v4l2_prio_check(vfd->prio, vfh->prio);
> > if (ret)
> > - return ret;
> > + goto error;
> > }
> > + } else {
> > + default_info.ioctl = cmd;
> > + default_info.flags = 0;
> > + default_info.debug = NULL;
> > + info = &default_info;
> > }
> >
> > - if ((vfd->debug & V4L2_DEBUG_IOCTL) &&
> > - !(vfd->debug & V4L2_DEBUG_IOCTL_ARG)) {
> > + write_only = _IOC_DIR(cmd) == _IOC_WRITE;
> > + if (info->debug && write_only && vfd->debug > V4L2_DEBUG_IOCTL) {
> > v4l_print_ioctl(vfd->name, cmd);
> > - printk(KERN_CONT "\n");
> > + pr_cont(": ");
> > + info->debug(arg);
> > + }
>
> Shouldn't you print the ioctl name and information even if info->debug is
> NULL
> ?
The 'info->debug' test is temporary. Once the conversion is finished this test
will be removed since then info->debug will never be NULL.
>
> > + if (info->flags & INFO_FL_STD) {
> > + typedef int (*vidioc_op)(struct file *file, void *fh, void *p);
> > + const void *p = vfd->ioctl_ops;
> > + const vidioc_op *vidioc = p + info->offset;
> > +
> > + ret = (*vidioc)(file, fh, arg);
> > + goto error;
> > + } else if (info->flags & INFO_FL_FUNC) {
> > + ret = info->func(ops, file, fh, arg);
> > + goto error;
> > }
> >
> > switch (cmd) {
> > @@ -2100,10 +2149,26 @@ static long __video_do_ioctl(struct file *file,
> > break;
> > } /* switch */
> >
> > - if (vfd->debug & V4L2_DEBUG_IOCTL_ARG) {
> > - if (ret < 0) {
> > - v4l_print_ioctl(vfd->name, cmd);
> > - printk(KERN_CONT " error %ld\n", ret);
> > +error:
>
> This isn't an error, is it ? I'd call it done instead.
True, I'll change this.
>
> > + if (vfd->debug) {
> > + if (write_only && vfd->debug > V4L2_DEBUG_IOCTL) {
>
> vfd->debug is a bitmask (or at least is documented as being a bitmask in
> include/media/v4l2-ioctl.h).
Yeah, and that makes no sense. Having V4L2_DEBUG_IOCTL_ARG without
V4L2_DEBUG_IOCTL is undefined. It is much more logical to have 0 for no
debugging, 1 for just ioctl name debugging and >= 2 for extensive debugging.
It's just weird to have to set the debug sysfs entry to 0, 1 or 3.
In a later patch I want to change the few drivers that use this accordingly.
>
> > + if (ret)
>
> Shouldn't you test for < 0 instead ? Driver-specific ioctls might return a >
> 0
> value in case of success.
Good catch. Yes, I need to do that.
>
> > + pr_info("%s: error %ld\n",
> > + video_device_node_name(vfd), ret);
> > + return ret;
> > + }
> > + v4l_print_ioctl(vfd->name, cmd);
> > + if (ret)
> > + pr_cont(": error %ld\n", ret);
> > + else if (vfd->debug == V4L2_DEBUG_IOCTL)
> > + pr_cont("\n");
> > + else if (!info->debug)
> > + return ret;
> > + else if (_IOC_DIR(cmd) == _IOC_NONE)
> > + info->debug(arg);
> > + else {
> > + pr_cont(": ");
> > + info->debug(arg);
> > }
>
> Ouch. What are you trying to do here ? Can't we simplify debug messages ?
It's simplified a bit in a later patch when the temporary info->debug test
can be dropped.
Regards,
Hans
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html