proc: make sure mem_open() doesn't pin the target's memory
authorOleg Nesterov <oleg@redhat.com>
Tue, 31 Jan 2012 16:15:11 +0000 (17:15 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Mon, 13 Feb 2012 19:16:53 +0000 (11:16 -0800)
commit43904e95ba660b59db5899a4d58a00e4ac4d3663
tree57c33753841224ffd19eeb962d19a41fb648eda0
parent034089b6f4e2ae0d0df38f3409cd73c386ad069a
proc: make sure mem_open() doesn't pin the target's memory

commit 6d08f2c7139790c268820a2e590795cb8333181a upstream.

Once /proc/pid/mem is opened, the memory can't be released until
mem_release() even if its owner exits.

Change mem_open() to do atomic_inc(mm_count) + mmput(), this only
pins mm_struct. Change mem_rw() to do atomic_inc_not_zero(mm_count)
before access_remote_vm(), this verifies that this mm is still alive.

I am not sure what should mem_rw() return if atomic_inc_not_zero()
fails. With this patch it returns zero to match the "mm == NULL" case,
may be it should return -EINVAL like it did before e268337d.

Perhaps it makes sense to add the additional fatal_signal_pending()
check into the main loop, to ensure we do not hold this memory if
the target task was oom-killed.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/proc/base.c