Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit d69b78ba authored by Greg Thelen's avatar Greg Thelen Committed by Jens Axboe
Browse files

ioprio: grab rcu_read_lock in sys_ioprio_{set,get}() 

Using:
- CONFIG_LOCKUP_DETECTOR=y
- CONFIG_PREEMPT=y
- CONFIG_LOCKDEP=y
- CONFIG_PROVE_LOCKING=y
- CONFIG_PROVE_RCU=y
found a missing rcu lock during boot on a 512 MiB x86_64 ubuntu vm:
  ===================================================
  [ INFO: suspicious rcu_dereference_check() usage. ]
  ---------------------------------------------------
  kernel/pid.c:419 invoked rcu_dereference_check() without protection!

  other info that might help us debug this:

  rcu_scheduler_active = 1, debug_locks = 0
  1 lock held by ureadahead/1355:
   #0:  (tasklist_lock){.+.+..}, at: [<ffffffff8115bc09>] sys_ioprio_set+0x7f/0x29e

  stack backtrace:
  Pid: 1355, comm: ureadahead Not tainted 2.6.37-dbg-DEV #1
  Call Trace:
   [<ffffffff8109c10c>] lockdep_rcu_dereference+0xaa/0xb3
   [<ffffffff81088cbf>] find_task_by_pid_ns+0x44/0x5d
   [<ffffffff81088cfa>] find_task_by_vpid+0x22/0x24
   [<ffffffff8115bc3e>] sys_ioprio_set+0xb4/0x29e
   [<ffffffff8147cf21>] ? trace_hardirqs_off_thunk+0x3a/0x3c
   [<ffffffff8105c409>] sysenter_dispatch+0x7/0x2c
   [<ffffffff8147cee2>] ? trace_hardirqs_on_thunk+0x3a/0x3f

The fix is to:
a) grab rcu lock in sys_ioprio_{set,get}() and
b) avoid grabbing tasklist_lock.
Discussion in: http://marc.info/?l=linux-kernel&m=128951324702889



Signed-off-by: default avatarGreg Thelen <gthelen@google.com>
Acked-by: default avatarPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: default avatarOleg Nesterov <oleg@redhat.com>

Modified by Jens to remove the now redundant inner rcu lock and
unlock since they are now protected by the outer lock.

Signed-off-by: default avatarJens Axboe <jaxboe@fusionio.com>
parent 1ff5125f

fs/ioprio.c

+6 −25
Original line number Diff line number Diff line
@@ -103,22 +103,15 @@ SYSCALL_DEFINE3(ioprio_set, int, which, int, who, int, ioprio) 
	}



	ret = -ESRCH;

	/*

	 * We want IOPRIO_WHO_PGRP/IOPRIO_WHO_USER to be "atomic",

	 * so we can't use rcu_read_lock(). See re-copy of ->ioprio

	 * in copy_process().

	 */

	read_lock(&tasklist_lock);

	rcu_read_lock();

	switch (which) {

		case IOPRIO_WHO_PROCESS:

			rcu_read_lock();

			if (!who)

				p = current;

			else

				p = find_task_by_vpid(who);

			if (p)

				ret = set_task_ioprio(p, ioprio);

			rcu_read_unlock();

			break;

		case IOPRIO_WHO_PGRP:

			if (!who)
@@ -141,12 +134,7 @@ SYSCALL_DEFINE3(ioprio_set, int, which, int, who, int, ioprio) 
				break;



			do_each_thread(g, p) {

				int match;



				rcu_read_lock();

				match = __task_cred(p)->uid == who;

				rcu_read_unlock();

				if (!match)

				if (__task_cred(p)->uid != who)

					continue;

				ret = set_task_ioprio(p, ioprio);

				if (ret)
@@ -160,7 +148,7 @@ free_uid: 
			ret = -EINVAL;

	}



	read_unlock(&tasklist_lock);

	rcu_read_unlock();

	return ret;

}
@@ -204,17 +192,15 @@ SYSCALL_DEFINE2(ioprio_get, int, which, int, who) 
	int ret = -ESRCH;

	int tmpio;



	read_lock(&tasklist_lock);

	rcu_read_lock();

	switch (which) {

		case IOPRIO_WHO_PROCESS:

			rcu_read_lock();

			if (!who)

				p = current;

			else

				p = find_task_by_vpid(who);

			if (p)

				ret = get_task_ioprio(p);

			rcu_read_unlock();

			break;

		case IOPRIO_WHO_PGRP:

			if (!who)
@@ -241,12 +227,7 @@ SYSCALL_DEFINE2(ioprio_get, int, which, int, who) 
				break;



			do_each_thread(g, p) {

				int match;



				rcu_read_lock();

				match = __task_cred(p)->uid == user->uid;

				rcu_read_unlock();

				if (!match)

				if (__task_cred(p)->uid != user->uid)

					continue;

				tmpio = get_task_ioprio(p);

				if (tmpio < 0)
@@ -264,6 +245,6 @@ SYSCALL_DEFINE2(ioprio_get, int, which, int, who) 
			ret = -EINVAL;

	}



	read_unlock(&tasklist_lock);

	rcu_read_unlock();

	return ret;

}