RE: [Dri-devel] chown /dev/dri/card0, system crashes

[Jim C. <ji...@ma...>]

On Mon, 28 Jan 2002, Alexander Stohr <Ale...@at...> wrote:
> so what have we learnt? better do not manipulate rights of files
> that are in currently in use by some other application.
> someone with something better to pop up?

On Mon, 28 Jan 2002 14:19:45 -0800, Ian Romanick <id...@us...> wrote:
> Unless I am mistaken, the short answer is don't do that. /dev/dri/card0
> is only directly accessed by the X server, which runs with root
> permissions. So, you shouldn't ever need to do such a thing.

Thank you for your suggestions, and indeed, I've removed /dev/dri/card0
from /etc/logindevperm, specifying mode 0666 in the DRI section of
XF86config.  Since this is a one-user system, inter-user paranoia is not
really an issue, but I like to run a tight ship, to have good habits where
it is a problem.

However, when the current user did not have permission to write on card0
(due to mode 0660 group video, the default provided by SuSE SaX2) he could
not use DRI.  At one point, with mode 0666, I ran strace, and the user
application really did map all of video RAM.  I'm not 100% up on all the
internals, but my understanding was that openGL would proceed to write all
over the video RAM, and do magic stuff with the graphics engine, using that
filehandle and bypassing the X-server completely (even though the X-server
also has card0 open).

It's puzzling that chown-chmod would have any baleful effect.  With a plain
file, once you've opened it you can monkey with the inode all you want and
the filehandle remains valid, and similarly with devfs the various device
inodes (/dev/misc/psaux, modem TTY, etc.) can be chowned even if, in the
case of the mouse, the X-server has it open.

James F. Carter          Voice 310 825 2897    FAX 310 206 6673
UCLA-Mathnet;  6115 MSA; 405 Hilgard Ave.; Los Angeles, CA, USA  90095-1555
Email: ji...@ma...    http://www.math.ucla.edu/~jimc (q.v. for PGP key)