> No its because whoever wrote the code to do the 2.1.x core dump
> sanity checks was using decaff by mistake. The logic doesnt make
> sense at all
I'm currently using Alexander Kjeldaas patch to allow coredumps, and
its seems sane (dumps with root, not suid, sgid, etc. though).
The current code will follow link though, so a hard or soft link to
/etc/passwd somewhere where you might expect root to dump core is
still very dangerous.
I think this is probably best solved by stat'ing `core' before:
dentry = open_namei(corefile, O_CREAT | 2 | O_TRUNC, 0600);
is executed:
- if the file is a link, coredumps should fail
- if the file is a regular file, only one referrence to the inode
is permissable
Now, this means tricks like `ln -s /dev/null core' no longer work,
but in this instance, it doesn't matter.
Is there some ligitimate reasons why we might want to allow core to
follow links of write to a file thats has multiple referrences?
-cw
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/