1
0
mirror of https://github.com/torvalds/linux.git synced 2024-12-30 14:52:05 +00:00
linux/fs/ioprio.c
Greg Thelen d69b78ba1d 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:
   :  (tasklist_lock){.+.+..}, at: [<ffffffff8115bc09>] sys_ioprio_set+0x7f/0x29e

  stack backtrace:
  Pid: 1355, comm: ureadahead Not tainted 2.6.37-dbg-DEV 
  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: Greg Thelen <gthelen@google.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Oleg 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: Jens Axboe <jaxboe@fusionio.com>
2010-11-15 10:23:31 +01:00

251 lines
5.0 KiB
C

/*
* fs/ioprio.c
*
* Copyright (C) 2004 Jens Axboe <axboe@kernel.dk>
*
* Helper functions for setting/querying io priorities of processes. The
* system calls closely mimmick getpriority/setpriority, see the man page for
* those. The prio argument is a composite of prio class and prio data, where
* the data argument has meaning within that class. The standard scheduling
* classes have 8 distinct prio levels, with 0 being the highest prio and 7
* being the lowest.
*
* IOW, setting BE scheduling class with prio 2 is done ala:
*
* unsigned int prio = (IOPRIO_CLASS_BE << IOPRIO_CLASS_SHIFT) | 2;
*
* ioprio_set(PRIO_PROCESS, pid, prio);
*
* See also Documentation/block/ioprio.txt
*
*/
#include <linux/gfp.h>
#include <linux/kernel.h>
#include <linux/ioprio.h>
#include <linux/blkdev.h>
#include <linux/capability.h>
#include <linux/syscalls.h>
#include <linux/security.h>
#include <linux/pid_namespace.h>
int set_task_ioprio(struct task_struct *task, int ioprio)
{
int err;
struct io_context *ioc;
const struct cred *cred = current_cred(), *tcred;
rcu_read_lock();
tcred = __task_cred(task);
if (tcred->uid != cred->euid &&
tcred->uid != cred->uid && !capable(CAP_SYS_NICE)) {
rcu_read_unlock();
return -EPERM;
}
rcu_read_unlock();
err = security_task_setioprio(task, ioprio);
if (err)
return err;
task_lock(task);
do {
ioc = task->io_context;
/* see wmb() in current_io_context() */
smp_read_barrier_depends();
if (ioc)
break;
ioc = alloc_io_context(GFP_ATOMIC, -1);
if (!ioc) {
err = -ENOMEM;
break;
}
task->io_context = ioc;
} while (1);
if (!err) {
ioc->ioprio = ioprio;
ioc->ioprio_changed = 1;
}
task_unlock(task);
return err;
}
EXPORT_SYMBOL_GPL(set_task_ioprio);
SYSCALL_DEFINE3(ioprio_set, int, which, int, who, int, ioprio)
{
int class = IOPRIO_PRIO_CLASS(ioprio);
int data = IOPRIO_PRIO_DATA(ioprio);
struct task_struct *p, *g;
struct user_struct *user;
struct pid *pgrp;
int ret;
switch (class) {
case IOPRIO_CLASS_RT:
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
/* fall through, rt has prio field too */
case IOPRIO_CLASS_BE:
if (data >= IOPRIO_BE_NR || data < 0)
return -EINVAL;
break;
case IOPRIO_CLASS_IDLE:
break;
case IOPRIO_CLASS_NONE:
if (data)
return -EINVAL;
break;
default:
return -EINVAL;
}
ret = -ESRCH;
rcu_read_lock();
switch (which) {
case IOPRIO_WHO_PROCESS:
if (!who)
p = current;
else
p = find_task_by_vpid(who);
if (p)
ret = set_task_ioprio(p, ioprio);
break;
case IOPRIO_WHO_PGRP:
if (!who)
pgrp = task_pgrp(current);
else
pgrp = find_vpid(who);
do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
ret = set_task_ioprio(p, ioprio);
if (ret)
break;
} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
break;
case IOPRIO_WHO_USER:
if (!who)
user = current_user();
else
user = find_user(who);
if (!user)
break;
do_each_thread(g, p) {
if (__task_cred(p)->uid != who)
continue;
ret = set_task_ioprio(p, ioprio);
if (ret)
goto free_uid;
} while_each_thread(g, p);
free_uid:
if (who)
free_uid(user);
break;
default:
ret = -EINVAL;
}
rcu_read_unlock();
return ret;
}
static int get_task_ioprio(struct task_struct *p)
{
int ret;
ret = security_task_getioprio(p);
if (ret)
goto out;
ret = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_NONE, IOPRIO_NORM);
if (p->io_context)
ret = p->io_context->ioprio;
out:
return ret;
}
int ioprio_best(unsigned short aprio, unsigned short bprio)
{
unsigned short aclass = IOPRIO_PRIO_CLASS(aprio);
unsigned short bclass = IOPRIO_PRIO_CLASS(bprio);
if (aclass == IOPRIO_CLASS_NONE)
aclass = IOPRIO_CLASS_BE;
if (bclass == IOPRIO_CLASS_NONE)
bclass = IOPRIO_CLASS_BE;
if (aclass == bclass)
return min(aprio, bprio);
if (aclass > bclass)
return bprio;
else
return aprio;
}
SYSCALL_DEFINE2(ioprio_get, int, which, int, who)
{
struct task_struct *g, *p;
struct user_struct *user;
struct pid *pgrp;
int ret = -ESRCH;
int tmpio;
rcu_read_lock();
switch (which) {
case IOPRIO_WHO_PROCESS:
if (!who)
p = current;
else
p = find_task_by_vpid(who);
if (p)
ret = get_task_ioprio(p);
break;
case IOPRIO_WHO_PGRP:
if (!who)
pgrp = task_pgrp(current);
else
pgrp = find_vpid(who);
do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
tmpio = get_task_ioprio(p);
if (tmpio < 0)
continue;
if (ret == -ESRCH)
ret = tmpio;
else
ret = ioprio_best(ret, tmpio);
} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
break;
case IOPRIO_WHO_USER:
if (!who)
user = current_user();
else
user = find_user(who);
if (!user)
break;
do_each_thread(g, p) {
if (__task_cred(p)->uid != user->uid)
continue;
tmpio = get_task_ioprio(p);
if (tmpio < 0)
continue;
if (ret == -ESRCH)
ret = tmpio;
else
ret = ioprio_best(ret, tmpio);
} while_each_thread(g, p);
if (who)
free_uid(user);
break;
default:
ret = -EINVAL;
}
rcu_read_unlock();
return ret;
}