Skip to contentSkip to conversation

rcu_read_lock/unlock protect find_task_by_vpid call in posix_cpu_timer_create

PATCH Dave Jones 4 participants 7 messages Nov 25, 2010

lore
Share & tools
lore

Dave Jones

author
#0 lore raw

===================================================
[ 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 = 1
1 lock held by scrashme/20820:
#0: (tasklist_lock){.?.?..}, at: [<ffffffff8106e30f>] posix_cpu_timer_create+0x50/0xee

stack backtrace:
Pid: 20820, comm: scrashme Not tainted 2.6.37-rc3+ #7
Call Trace:
 [<ffffffff8107cfd5>] lockdep_rcu_dereference+0x9d/0xa5
 [<ffffffff81069d08>] find_task_by_pid_ns+0x44/0x5d
 [<ffffffff81069d43>] find_task_by_vpid+0x22/0x24
 [<ffffffff8106e32d>] posix_cpu_timer_create+0x6e/0xee
 [<ffffffff8106eb88>] do_cpu_nanosleep+0x83/0x1ad
 [<ffffffff8106f50a>] posix_cpu_nsleep+0x6d/0xf6
 [<ffffffff810f9a64>] ? might_fault+0xa5/0xac
 [<ffffffff810f9a1b>] ? might_fault+0x5c/0xac
 [<ffffffff8106bf57>] sys_clock_nanosleep+0x7c/0xcb
 [<ffffffff81009cb2>] system_call_fastpath+0x16/0x1b
kernel/posix-cpu-timers.c +2 -0
index 6842eeb..2658955 100644
@@ -391,6 +391,7 @@ int posix_cpu_timer_create(struct k_itimer *new_timer)
 	INIT_LIST_HEAD(&new_timer->it.cpu.entry);
 
 	read_lock(&tasklist_lock);
+	rcu_read_lock();
 	if (CPUCLOCK_PERTHREAD(new_timer->it_clock)) {
 		if (pid == 0) {
 			p = current;
@@ -414,6 +415,7 @@ int posix_cpu_timer_create(struct k_itimer *new_timer)
 	} else {
 		ret = -EINVAL;
 	}
+	rcu_read_unlock();
 	read_unlock(&tasklist_lock);
 
 	return ret;
author
#1 lore raw

rcu_read_lock/unlock protect find_task_by_vpid call in check_clock

===================================================
[ 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 = 1
1 lock held by scrashme/13382:
#0: (tasklist_lock){.?.?..}, at: [<ffffffff8106ddea>] check_clock+0x46/0x9a

stack backtrace:
Pid: 13382, comm: scrashme Not tainted 2.6.37-rc3+ #8
Call Trace:
 [<ffffffff8107cfe1>] lockdep_rcu_dereference+0x9d/0xa5
 [<ffffffff81069d08>] find_task_by_pid_ns+0x44/0x5d
 [<ffffffff81069d43>] find_task_by_vpid+0x22/0x24
 [<ffffffff8106ddf2>] check_clock+0x4e/0x9a
 [<ffffffff8106deac>] posix_cpu_clock_getres+0x16/0x41
 [<ffffffff8106be74>] sys_clock_getres+0x39/0xa0
 [<ffffffff81009cb2>] system_call_fastpath+0x16/0x1b
kernel/posix-cpu-timers.c +2 -0
index 6842eeb..4bef9aa 100644
@@ -38,11 +38,13 @@ static int check_clock(const clockid_t which_clock)
 		return 0;
 
 	read_lock(&tasklist_lock);
+	rcu_read_lock();
 	p = find_task_by_vpid(pid);
 	if (!p || !(CPUCLOCK_PERTHREAD(which_clock) ?
 		   same_thread_group(p, current) : thread_group_leader(p))) {
 		error = -EINVAL;
 	}
+	rcu_read_unlock();
 	read_unlock(&tasklist_lock);
 
 	return error;
#3 lore raw
In reply to Dave Jones
Quoted 41 lines · Dave Jones

===================================================
[ 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 = 1
1 lock held by scrashme/20820:
#0: (tasklist_lock){.?.?..}, at: [<ffffffff8106e30f>] posix_cpu_timer_create+0x50/0xee

stack backtrace:
Pid: 20820, comm: scrashme Not tainted 2.6.37-rc3+ #7
Call Trace:
 [<ffffffff8107cfd5>] lockdep_rcu_dereference+0x9d/0xa5
 [<ffffffff81069d08>] find_task_by_pid_ns+0x44/0x5d
 [<ffffffff81069d43>] find_task_by_vpid+0x22/0x24
 [<ffffffff8106e32d>] posix_cpu_timer_create+0x6e/0xee
 [<ffffffff8106eb88>] do_cpu_nanosleep+0x83/0x1ad
 [<ffffffff8106f50a>] posix_cpu_nsleep+0x6d/0xf6
 [<ffffffff810f9a64>] ? might_fault+0xa5/0xac
 [<ffffffff810f9a1b>] ? might_fault+0x5c/0xac
 [<ffffffff8106bf57>] sys_clock_nanosleep+0x7c/0xcb
 [<ffffffff81009cb2>] system_call_fastpath+0x16/0x1b
kernel/posix-cpu-timers.c +1 -0
index 6842eeb..2658955 100644
@@ -391,6 +391,7 @@ int posix_cpu_timer_create(struct k_itimer *new_timer)
 	INIT_LIST_HEAD(&new_timer->it.cpu.entry);
 
 	read_lock(&tasklist_lock);
+	rcu_read_lock();
 	if (CPUCLOCK_PERTHREAD(new_timer->it_clock)) {
 		if (pid == 0) {
 			p = current;

kernel/posix-cpu-timers.c int posix_cpu_timer_create(struct k_itimer *new_timer) · line 421

Quoted 7 lines
kernel/posix-cpu-timers.c +1 -0
@@ -414,6 +415,7 @@ int posix_cpu_timer_create(struct k_itimer *new_timer)
 	} else {
 		ret = -EINVAL;
 	}
+	rcu_read_unlock();
 	read_unlock(&tasklist_lock);
 
 	return ret;

Do we still need the tasklist_lock in this case?

Also, why is that think complaining, surely the tasklist_lock pins any and all PID objects?

#2 lore raw

Re: rcu_read_lock/unlock protect find_task_by_vpid call in check_clock

In reply to Dave Jones
Quoted 33 lines · Dave Jones

===================================================
[ 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 = 1
1 lock held by scrashme/13382:
#0: (tasklist_lock){.?.?..}, at: [<ffffffff8106ddea>] check_clock+0x46/0x9a

stack backtrace:
Pid: 13382, comm: scrashme Not tainted 2.6.37-rc3+ #8
Call Trace:
 [<ffffffff8107cfe1>] lockdep_rcu_dereference+0x9d/0xa5
 [<ffffffff81069d08>] find_task_by_pid_ns+0x44/0x5d
 [<ffffffff81069d43>] find_task_by_vpid+0x22/0x24
 [<ffffffff8106ddf2>] check_clock+0x4e/0x9a
 [<ffffffff8106deac>] posix_cpu_clock_getres+0x16/0x41
 [<ffffffff8106be74>] sys_clock_getres+0x39/0xa0
 [<ffffffff81009cb2>] system_call_fastpath+0x16/0x1b
kernel/posix-cpu-timers.c +0 -0
index 6842eeb..4bef9aa 100644
@@ -38,11 +38,13 @@ static int check_clock(const clockid_t which_clock)
 		return 0;
 
 	read_lock(&tasklist_lock);

kernel/posix-cpu-timers.c static int check_clock(const clockid_t which_clock) · line 50

Quoted 10 lines
kernel/posix-cpu-timers.c +2 -0
+	rcu_read_lock();
 	p = find_task_by_vpid(pid);
 	if (!p || !(CPUCLOCK_PERTHREAD(which_clock) ?
 		   same_thread_group(p, current) : thread_group_leader(p))) {
 		error = -EINVAL;
 	}
+	rcu_read_unlock();
 	read_unlock(&tasklist_lock);
 
 	return error;

Pretty much the same comment as the other patch..

<copy/paste>

Do we still need the tasklist_lock in this case?

Also, why is that think complaining, surely the tasklist_lock pins any and all PID objects?

#4 lore raw
In reply to Peter Zijlstra

(add Sergey)

Quoted 19 lines · Peter Zijlstra
Quoted 17 lines · Dave Jones
kernel/posix-cpu-timers.c +2 -0
@@ -391,6 +391,7 @@ int posix_cpu_timer_create(struct k_itimer *new_timer)
 	INIT_LIST_HEAD(&new_timer->it.cpu.entry);
 
 	read_lock(&tasklist_lock);
+	rcu_read_lock();
 	if (CPUCLOCK_PERTHREAD(new_timer->it_clock)) {
 		if (pid == 0) {
 			p = current;
@@ -414,6 +415,7 @@ int posix_cpu_timer_create(struct k_itimer *new_timer)
 	} else {
 		ret = -EINVAL;
 	}
+	rcu_read_unlock();
 	read_unlock(&tasklist_lock);
 
Quoted 3 lines

kernel/posix-cpu-timers.c int posix_cpu_timer_create(struct k_itimer *new_timer) · line 421

Quoted 1 line
kernel/posix-cpu-timers.c +0 -0
 	return ret;

Do we still need the tasklist_lock in this case?

No. posix-cpu-timer.c shouldn't use tasklist at all. But it is not completely trivial to remove it.

In particular, this patch is not exactly right, we can't trust thread_group_leader() without tasklist.

Sergey already sent the patch which removes tasklist from posix_cpu_timer_create() and posix_cpu_timer_create(), and iirc Thomas queued it.

Quoted 2 lines

Also, why is that think complaining, surely the tasklist_lock pins any and all PID objects?

The only problem is: if copy_process() fails, it does free_pid() lockless. This means, without rcu lock it is not safe to scan the rcu-protected lists.

We can change copy_process() (in fact I sent the patch several years ago), but everybody think that find_pid/etc should always take rcu_read_lock() instead. I tend to agree.

Oleg.

#5 lore raw
In reply to Peter Zijlstra

(another try, actually add Sergey)

Quoted 19 lines · Peter Zijlstra
Quoted 17 lines · Dave Jones
kernel/posix-cpu-timers.c +2 -0
@@ -391,6 +391,7 @@ int posix_cpu_timer_create(struct k_itimer *new_timer)
 	INIT_LIST_HEAD(&new_timer->it.cpu.entry);
 
 	read_lock(&tasklist_lock);
+	rcu_read_lock();
 	if (CPUCLOCK_PERTHREAD(new_timer->it_clock)) {
 		if (pid == 0) {
 			p = current;
@@ -414,6 +415,7 @@ int posix_cpu_timer_create(struct k_itimer *new_timer)
 	} else {
 		ret = -EINVAL;
 	}
+	rcu_read_unlock();
 	read_unlock(&tasklist_lock);
 
Quoted 3 lines

kernel/posix-cpu-timers.c int posix_cpu_timer_create(struct k_itimer *new_timer) · line 421

Quoted 1 line
kernel/posix-cpu-timers.c +0 -0
 	return ret;

Do we still need the tasklist_lock in this case?

No. posix-cpu-timer.c shouldn't use tasklist at all. But it is not completely trivial to remove it.

In particular, this patch is not exactly right, we can't trust thread_group_leader() without tasklist.

Sergey already sent the patch which removes tasklist from posix_cpu_timer_create() and posix_cpu_timer_create(), and iirc Thomas queued it.

Quoted 2 lines

Also, why is that think complaining, surely the tasklist_lock pins any and all PID objects?

The only problem is: if copy_process() fails, it does free_pid() lockless. This means, without rcu lock it is not safe to scan the rcu-protected lists.

We can change copy_process() (in fact I sent the patch several years ago), but everybody think that find_pid/etc should always take rcu_read_lock() instead. I tend to agree.

Oleg.

#6 lore raw

Hello,

Quoted 2 lines · Oleg Nesterov

(another try, actually add Sergey)

Thank you.

Quoted 27 lines · Peter Zijlstra
Quoted 19 lines · Dave Jones
Quoted 17 lines
kernel/posix-cpu-timers.c +2 -0
@@ -391,6 +391,7 @@ int posix_cpu_timer_create(struct k_itimer *new_timer)
 	INIT_LIST_HEAD(&new_timer->it.cpu.entry);
 
 	read_lock(&tasklist_lock);
+	rcu_read_lock();
 	if (CPUCLOCK_PERTHREAD(new_timer->it_clock)) {
 		if (pid == 0) {
 			p = current;
@@ -414,6 +415,7 @@ int posix_cpu_timer_create(struct k_itimer *new_timer)
 	} else {
 		ret = -EINVAL;
 	}
+	rcu_read_unlock();
 	read_unlock(&tasklist_lock);
 
Quoted 3 lines

kernel/posix-cpu-timers.c int posix_cpu_timer_create(struct k_itimer *new_timer) · line 421

Quoted 1 line
kernel/posix-cpu-timers.c +0 -0
 	return ret;

Do we still need the tasklist_lock in this case?

No. posix-cpu-timer.c shouldn't use tasklist at all. But it is not completely trivial to remove it.

Quoted 7 lines

In particular, this patch is not exactly right, we can't trust thread_group_leader() without tasklist.

Sergey already sent the patch which removes tasklist from posix_cpu_timer_create() and posix_cpu_timer_create(), and iirc Thomas queued it.

You're right, Oleg.

Commit-ID:  c0deae8c9587419ab13874b74425ce2eb2e18508
Gitweb:     http://git.kernel.org/tip/c0deae8c9587419ab13874b74425ce2eb2e18508
Author:     Sergey Senozhatsky <[email protected]>
AuthorDate: Wed, 3 Nov 2010 18:52:56 +0200
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Wed, 10 Nov 2010 13:07:06 +0100

posix-cpu-timers: Rcu_read_lock/unlock protect find_task_by_vpid call

queued (15 days so far).

Quoted 7 lines
Quoted 2 lines

Also, why is that think complaining, surely the tasklist_lock pins any and all PID objects?

The only problem is: if copy_process() fails, it does free_pid() lockless. This means, without rcu lock it is not safe to scan the rcu-protected lists.

Quoted 6 lines

We can change copy_process() (in fact I sent the patch several years ago), but everybody think that find_pid/etc should always take rcu_read_lock() instead. I tend to agree.

Oleg.

Sergey