Skip to contentSkip to conversation

futex: Detect mismatched requeue targets

PATCH In review John Kacur 2 participants 4 messages Oct 20, 2009

lore
Share & tools
lore

John Kacur

author
#0 lore raw

PATCH 1/1 futex: Detect mismatched requeue targets

Hello Darren

I took your patch from commit 84bc4af59081ee974dd80210e694ab59ebe51ce8 and I tried to git-cherry-pick it for v2.6.31.4-rt14

I had a little merge-commit to resolve. That wasn't too hard, but as the code is a bit different between the two versions, I would appreciate it if you could review the patch, and make sure that it still makes sense for v2.6.31.r-rt14

Thanks!
John

Quoted 1 line

From 7135729cbbc5bf16c95abdca5366ec5de0a01909 Mon Sep 17 00:00:00 2001

From: Darren Hart <[email protected]>
Date: Thu, 13 Aug 2009 17:36:53 -0700
Subject: [PATCH] futex: Detect mismatched requeue targets

There is currently no check to ensure that userspace uses the same futex requeue target (uaddr2) in futex_requeue() that the waiter used in futex_wait_requeue_pi(). A mismatch here could very unexpected results as the waiter assumes it either wakes on uaddr1 or uaddr2. We could detect this on wakeup in the waiter, but the cleanup is more intense after the improper requeue has occured.

This patch stores the waiter's expected requeue target in a new requeue_pi_key pointer in the futex_q which futex_requeue() checks prior to attempting to do a proxy lock acquistion or a requeue when requeue_pi=1. If they don't match, return -EINVAL from futex_requeue, aborting the requeue of any remaining waiters.

Cc Peter Zijlstra <[email protected]>
Cc Eric Dumazet <[email protected]>
Cc John Kacur <[email protected]>
Cc Dinakar Guniguntala <[email protected]>
Cc John Stultz <[email protected]>

LKML-Reference: <20090814003650.14634.63916.stgit@Aeon>

---
 kernel/futex.c |   24 ++++++++++++++++++++----
 1 files changed, 20 insertions(+), 4 deletions(-)
kernel/futex.c +15 -3
index 7c4a6ac..5978a84 100644
@@ -115,6 +115,9 @@ struct futex_q {
 	/* rt_waiter storage for requeue_pi: */
 	struct rt_mutex_waiter *rt_waiter;
 
+	/* The expected requeue pi target futex key: */
+	union futex_key *requeue_pi_key;
+
 	/* Bitset for the optional bitmasked wakeup */
 	u32 bitset;
 };
@@ -1089,6 +1092,10 @@ static int futex_proxy_trylock_atomic(u32 __user *pifutex,
 	if (!top_waiter)
 		return 0;
 
+	/* Ensure we requeue to the expected futex. */
+	if (!match_futex(top_waiter->requeue_pi_key, key2))
+		return -EINVAL;
+
 	/*
 	 * Try to take the lock for top_waiter.  Set the FUTEX_WAITERS bit in
 	 * the contended case or if set_waiters is 1.  The pi_state is returned
@@ -1276,6 +1283,12 @@ retry_private:
 			continue;
 		}
 
+		/* Ensure we requeue to the expected futex for requeue_pi. */
+		if (requeue_pi && !match_futex(this->requeue_pi_key, &key2)) {
+			ret = -EINVAL;
+			break;
+		}
+
 		/*
 		 * Requeue nr_requeue waiters and possibly one more in the case
 		 * of requeue_pi if we couldn't acquire the lock atomically.
@@ -1742,6 +1755,7 @@ static int futex_wait(u32 __user *uaddr, int fshared,
 	q.pi_state = NULL;
 	q.bitset = bitset;
 	q.rt_waiter = NULL;
+	q.requeue_pi_key = NULL;
 
 	if (abs_time) {
 		to = &timeout;
@@ -1855,6 +1869,7 @@ static int futex_lock_pi(u32 __user *uaddr, int fshared,
 
 	q.pi_state = NULL;
 	q.rt_waiter = NULL;
+	q.requeue_pi_key = NULL;
 retry:
 	q.key = FUTEX_KEY_INIT;
 	ret = get_futex_key(uaddr, fshared, &q.key, VERIFY_WRITE);
@@ -2167,16 +2182,17 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, int fshared,
 	debug_rt_mutex_init_waiter(&rt_waiter);
 	rt_waiter.task = NULL;
 
-	q.pi_state = NULL;
-	q.bitset = bitset;
-	q.rt_waiter = &rt_waiter;
#1 lore raw
Quoted 6 lines · John Kacur

Hello Darren

I took your patch from commit 84bc4af59081ee974dd80210e694ab59ebe51ce8 and I tried to git-cherry-pick it for v2.6.31.4-rt14

I had a little merge-commit to resolve. That wasn't too hard, but as the code is a bit different between the two versions, I would appreciate it if you could review the patch, and make sure that it still makes sense for v2.6.31.r-rt14

Hi John,

It looks good to me. Where did you have the conflicts?

--
Darren

Quoted 111 lines

Thanks!
John

From 7135729cbbc5bf16c95abdca5366ec5de0a01909 Mon Sep 17 00:00:00 2001
From: Darren Hart <[email protected]>
Date: Thu, 13 Aug 2009 17:36:53 -0700
Subject: [PATCH] futex: Detect mismatched requeue targets

There is currently no check to ensure that userspace uses the same futex requeue target (uaddr2) in futex_requeue() that the waiter used in futex_wait_requeue_pi(). A mismatch here could very unexpected results as the waiter assumes it either wakes on uaddr1 or uaddr2. We could detect this on wakeup in the waiter, but the cleanup is more intense after the improper requeue has occured.

This patch stores the waiter's expected requeue target in a new requeue_pi_key pointer in the futex_q which futex_requeue() checks prior to attempting to do a proxy lock acquistion or a requeue when requeue_pi=1. If they don't match, return -EINVAL from futex_requeue, aborting the requeue of any remaining waiters.

Cc Peter Zijlstra <[email protected]>
Cc Eric Dumazet <[email protected]>
Cc John Kacur <[email protected]>
Cc Dinakar Guniguntala <[email protected]>
Cc John Stultz <[email protected]>

LKML-Reference: <20090814003650.14634.63916.stgit@Aeon>

---
 kernel/futex.c |   24 ++++++++++++++++++++----
 1 files changed, 20 insertions(+), 4 deletions(-)
kernel/futex.c +20 -4
index 7c4a6ac..5978a84 100644
@@ -115,6 +115,9 @@ struct futex_q {
 	/* rt_waiter storage for requeue_pi: */
 	struct rt_mutex_waiter *rt_waiter;
 
+	/* The expected requeue pi target futex key: */
+	union futex_key *requeue_pi_key;
+
 	/* Bitset for the optional bitmasked wakeup */
 	u32 bitset;
 };
@@ -1089,6 +1092,10 @@ static int futex_proxy_trylock_atomic(u32 __user *pifutex,
 	if (!top_waiter)
 		return 0;
 
+	/* Ensure we requeue to the expected futex. */
+	if (!match_futex(top_waiter->requeue_pi_key, key2))
+		return -EINVAL;
+
 	/*
 	 * Try to take the lock for top_waiter.  Set the FUTEX_WAITERS bit in
 	 * the contended case or if set_waiters is 1.  The pi_state is returned
@@ -1276,6 +1283,12 @@ retry_private:
 			continue;
 		}
 
+		/* Ensure we requeue to the expected futex for requeue_pi. */
+		if (requeue_pi && !match_futex(this->requeue_pi_key, &key2)) {
+			ret = -EINVAL;
+			break;
+		}
+
 		/*
 		 * Requeue nr_requeue waiters and possibly one more in the case
 		 * of requeue_pi if we couldn't acquire the lock atomically.
@@ -1742,6 +1755,7 @@ static int futex_wait(u32 __user *uaddr, int fshared,
 	q.pi_state = NULL;
 	q.bitset = bitset;
 	q.rt_waiter = NULL;
+	q.requeue_pi_key = NULL;
 
 	if (abs_time) {
 		to = &timeout;
@@ -1855,6 +1869,7 @@ static int futex_lock_pi(u32 __user *uaddr, int fshared,
 
 	q.pi_state = NULL;
 	q.rt_waiter = NULL;
+	q.requeue_pi_key = NULL;
 retry:
 	q.key = FUTEX_KEY_INIT;
 	ret = get_futex_key(uaddr, fshared, &q.key, VERIFY_WRITE);
@@ -2167,16 +2182,17 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, int fshared,
 	debug_rt_mutex_init_waiter(&rt_waiter);
 	rt_waiter.task = NULL;
 
-	q.pi_state = NULL;
-	q.bitset = bitset;
-	q.rt_waiter = &rt_waiter;
-
 retry:
 	key2 = FUTEX_KEY_INIT;
 	ret = get_futex_key(uaddr2, fshared, &key2, VERIFY_WRITE);
 	if (unlikely(ret != 0))
 		goto out;
 
+	q.pi_state = NULL;
+	q.bitset = bitset;
+	q.rt_waiter = &rt_waiter;
+	q.requeue_pi_key = &key2;
+
 	/* Prepare to wait on uaddr. */
 	ret = futex_wait_setup(uaddr, val, fshared, &q, &hb);
 	if (ret)
Signature · 3 lines
Darren Hart
IBM Linux Technology Center
Real-Time Linux Team
author
#2 lore raw
Quoted 12 lines · ----- "Darren Hart
Quoted 3 lines · John Kacur

Hello Darren

I took your patch from commit

84bc4af59081ee974dd80210e694ab59ebe51ce8

Quoted 3 lines

and I tried to git-cherry-pick it for v2.6.31.4-rt14

I had a little merge-commit to resolve. That wasn't too hard, but as

the

Quoted 1 line

code is a bit different between the two versions, I would appreciate

it if

Quoted 1 line

you could review the patch, and make sure that it still makes sense

Quoted 8 lines

for

Quoted 1 line

v2.6.31.r-rt14

Hi John,

It looks good to me. Where did you have the conflicts?

--

The conflict was in the futex_wait_requeue_pi function.
That's where you really need to double check that the logic still makes sense.

Thanks.

#3 lore raw
Quoted 16 lines · John Kacur
Quoted 14 lines · ----- "Darren Hart
Quoted 3 lines · John Kacur

Hello Darren

I took your patch from commit

84bc4af59081ee974dd80210e694ab59ebe51ce8

Quoted 3 lines

and I tried to git-cherry-pick it for v2.6.31.4-rt14

I had a little merge-commit to resolve. That wasn't too hard, but as

the

Quoted 1 line

code is a bit different between the two versions, I would appreciate

it if

Quoted 1 line

you could review the patch, and make sure that it still makes sense

for

Quoted 1 line

v2.6.31.r-rt14

Quoted 8 lines
Quoted 5 lines

Hi John,

It looks good to me. Where did you have the conflicts?

--

The conflict was in the futex_wait_requeue_pi function.
That's where you really need to double check that the logic still makes sense.

Hi John,

Yes, I suspect the conflict was do to the retry: label. You should have already pulled in the spurious wakeup patch, so you should be fine. The patch looks good to me.

Thanks,

Signature · 3 lines
Darren Hart
IBM Linux Technology Center
Real-Time Linux Team