threadgroup: extend threadgroup_lock() to cover exit and exec
authorTejun Heo <tj@kernel.org>
Tue, 13 Dec 2011 02:12:21 +0000 (18:12 -0800)
committerTejun Heo <tj@kernel.org>
Tue, 13 Dec 2011 02:12:21 +0000 (18:12 -0800)
commit77e4ef99d1c596a31747668e5fd837f77b6349b6
treeb9dcc8d73f9febf6a18edc0b2efa3ecddc2a7a35
parent257058ae2b971646b96ab3a15605ac69186e562a
threadgroup: extend threadgroup_lock() to cover exit and exec

threadgroup_lock() protected only protected against new addition to
the threadgroup, which was inherently somewhat incomplete and
problematic for its only user cgroup.  On-going migration could race
against exec and exit leading to interesting problems - the symmetry
between various attach methods, task exiting during method execution,
->exit() racing against attach methods, migrating task switching basic
properties during exec and so on.

This patch extends threadgroup_lock() such that it protects against
all three threadgroup altering operations - fork, exit and exec.  For
exit, threadgroup_change_begin/end() calls are added to exit_signals
around assertion of PF_EXITING.  For exec, threadgroup_[un]lock() are
updated to also grab and release cred_guard_mutex.

With this change, threadgroup_lock() guarantees that the target
threadgroup will remain stable - no new task will be added, no new
PF_EXITING will be set and exec won't happen.

The next patch will update cgroup so that it can take full advantage
of this change.

-v2: beefed up comment as suggested by Frederic.

-v3: narrowed scope of protection in exit path as suggested by
     Frederic.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Li Zefan <lizf@cn.fujitsu.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Paul Menage <paul@paulmenage.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
include/linux/sched.h
kernel/signal.c