Re: SUMMARY: GGI/X : the other way??

Albert D. Cahalan (acahalan@cs.uml.edu)
Mon, 6 Apr 1998 20:34:25 -0400 (EDT)

>> An fb device would solve this, but as linux didn't have a revoke
>> syscall I never sent my DGA compatible /dev/fb to linus. The
>> security issues of a publically readable /dev/fb are very compicated
>> (I would like to know how ggi has tried to get round them).
>
> Not cleanly. Its worse than this btw. The BSD revoke() doesnt revoke
> existing mmap()'s. This leaves a gaping hole here and in the sound drivers
>
> Please don't advertise that too far. CERT are aware of it and it seems
> to open holes all over the place (snooping with camera interfaces etc)

If revoke() can kill a mmap on a file, users can make servers crash.
It might be better to convert mappings to /dev/zero or /dev/null
so that servers won't get SIGSEGV or SIGBUS.

There is a patch:
http://linuxwww.db.erau.edu/mail_archives/linux-kernel/Dec_97/0946.html

I notice that not even Theodore Ts'o is able to get patches in now.
Perhaps Linus 2.0 could help Linus 1.0 some more.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu