coredump: set_dumpable: fix the theoretical race with itself
authorOleg Nesterov <oleg@redhat.com>
Thu, 23 Jan 2014 23:55:31 +0000 (15:55 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 24 Jan 2014 00:37:01 +0000 (16:37 -0800)
commitabacd2fe3ca10b3ade57f3634053241a660002c2
treee3d8d77e9b12d2e57a90a026a225313ccca9ddb1
parentf3c73a99a1fac2db992b6879b8a78a3ae2fcc06e
coredump: set_dumpable: fix the theoretical race with itself

set_dumpable() updates MMF_DUMPABLE_MASK in a non-trivial way to ensure
that get_dumpable() can't observe the intermediate state, but this all
can't help if multiple threads call set_dumpable() at the same time.

And in theory commit_creds()->set_dumpable(SUID_DUMP_ROOT) racing with
sys_prctl()->set_dumpable(SUID_DUMP_DISABLE) can result in SUID_DUMP_USER.

Change this code to update both bits atomically via cmpxchg().

Note: this assumes that it is safe to mix bitops and cmpxchg.  IOW, if,
say, an architecture implements cmpxchg() using the locking (like
arch/parisc/lib/bitops.c does), then it should use the same locks for
set_bit/etc.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Kees Cook <keescook@chromium.org>
Cc: Alex Kelly <alex.page.kelly@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Petr Matousek <pmatouse@redhat.com>
Cc: Vasily Kulikov <segoon@openwall.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/exec.c