[PATCH 2/2] RFC: Fix futex_lock_pi fault handling (NOT FOR INCLUSION)

[Darren Hart]

Regardless of whether we call get_user or futex_handle_fault to deal with a
fault, the uaddr doesn't change, so the key won't change.  There doesn't appear
to be a reason to re-get the key after a get_user call, but not after a
futex_handle_fault.  This patch moves both jump points to right after the the
get_futex_key call.  Also fix a missed put_futex_key() call and update the
comment to accurately depict the current code (we don't hold the mm sem now).

Signed-off-by: Darren Hart <dvhltc@xxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Ingo Molnar <mingo@xxxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Rusty Russell <rusty@xxxxxxxxxxxxxxx>
---

 kernel/futex.c |   17 ++++++++---------
 1 files changed, 8 insertions(+), 9 deletions(-)

diff --git a/kernel/futex.c b/kernel/futex.c
index c15c029..cd03229 100644
--- a/kernel/futex.c
+++ b/kernel/futex.c
@@ -1351,13 +1351,13 @@ static int futex_lock_pi(u32 __user *uaddr, int fshared,
 	}
 
 	q.pi_state = NULL;
-retry:
+
 	q.key = FUTEX_KEY_INIT;
 	ret = get_futex_key(uaddr, fshared, &q.key);
 	if (unlikely(ret != 0))
 		goto out;
 
-retry_unlocked:
+retry:
 	hb = queue_lock(&q);
 
 retry_locked:
@@ -1566,19 +1566,18 @@ out:
 
 uaddr_faulted:
 	/*
-	 * We have to r/w  *(int __user *)uaddr, and we have to modify it
-	 * atomically.  Therefore, if we continue to fault after get_user()
-	 * below, we need to handle the fault ourselves, while still holding
-	 * the mmap_sem.  This can occur if the uaddr is under contention as
-	 * we have to drop the mmap_sem in order to call get_user().
+	 * We need to read and write *(int __user *)uaddr2 atomically.
+	 * Therefore, if get_user below is not enough, we need to
+	 * handle the fault ourselves.
 	 */
 	queue_unlock(&q, hb);
+	put_futex_key(fshared, &q.key);
 
 	if (attempt++) {
 		ret = futex_handle_fault((unsigned long)uaddr, attempt);
 		if (ret)
-			goto out_put_key;
-		goto retry_unlocked;
+			goto out;
+		goto retry;
 	}
 
 	ret = get_user(uval, uaddr);

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/