[lkml]   [2002]   [Feb]   [9]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [V4L] [PATCH/RFC] videodev.[ch] redesign
Gerd Knorr wrote:

>It also provides a ioctl wrapper function which handles copying the
>ioctl args from/to userspace, so we have this at one place can drop all
>the copy_from/to_user calls within the v4l device driver ioctl handlers.

Excellent work. I have no complaints, just a few questions:

1. Would it be better to memset the temp buffer in video_generic_ioctl()
rather than in the driver? I've seen so many drivers forget to do this,
and it's a potential (albeit very small) security hole.

2. In skeleton_open(), couldn't the device_data lookup code be replaced

struct video_device *vdev = video_devdata(file);
struct device_data *dev = vdev->priv;

3. In skeleton_initdev(), shouldn't...

dev->vdev = skeleton_template;

memcpy(&dev->vdev, &skeleton_template, sizeof(skeleton_template);

4. Is it safe to keep even 128 bytes on the stack in
video_generic_ioctl()? Consider that devices might spend a relatively
long time blocking on VIDIOCSYNC. With 32 devices in use at once, you'd
be coming dangerously close to a stack overflow. IMHO it would be better
to only allocate as much as MCAPTURE and SYNC need, and fall back on
kmalloc for the less time-critical ones (if necessary).

Other than that, I extremely happy with what you've done!

Mark McClelland

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2005-03-22 13:24    [W:0.111 / U:12.388 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site