[PATCH] futex: add WARN_ON to catch certain assymetric (get|put)_futex_key calls

[PATCH] futex: add WARN_ON to catch certain assymetric (get|put)_futex_key calls

[Darren Hart]

Following up on my previous key reference accounting patches, this patch
will catch puts on keys that haven't been "got".  This won't catch nested
get/put mismatches though.

Build and boot tested, with minimal desktop activity and a run of the
open_posix_testsuite in LTP for testing.  No warnings logged.

Signed-off-by: Darren Hart <dvhltc@us.ibm.com>
Cc: "Peter Zijlstra" <peterz@infradead.org>
Cc: "Thomas Gleixner" <tglx@linutronix.de>
Cc: "Rusty Russell" <rusty@au1.ibm.com>
---

 kernel/futex.c |    5 ++++-
 1 files changed, 4 insertions(+), 1 deletions(-)

diff --git a/kernel/futex.c b/kernel/futex.c
index 748778a..5288acf 100644
--- a/kernel/futex.c
+++ b/kernel/futex.c
@@ -170,8 +170,11 @@ static void get_futex_key_refs(union futex_key *key)
  */
 static void drop_futex_key_refs(union futex_key *key)
 {
-	if (!key->both.ptr)
+	if (!key->both.ptr) {
+		/* If we're here then we tried to put a key we failed to get */
+		WARN_ON(1);
 		return;
+	}
 
 	switch (key->both.offset & (FUT_OFF_INODE|FUT_OFF_MMSHARED)) {
 	case FUT_OFF_INODE:

Re: [PATCH] futex: add WARN_ON to catch certain assymetric (get|put)_futex_key calls

[Ingo Molnar]

* Darren Hart <dvhltc@us.ibm.com> wrote:

> Following up on my previous key reference accounting patches, this patch
> will catch puts on keys that haven't been "got".  This won't catch nested
> get/put mismatches though.
> 
> Build and boot tested, with minimal desktop activity and a run of the
> open_posix_testsuite in LTP for testing.  No warnings logged.
> 
> Signed-off-by: Darren Hart <dvhltc@us.ibm.com>
> Cc: "Peter Zijlstra" <peterz@infradead.org>
> Cc: "Thomas Gleixner" <tglx@linutronix.de>
> Cc: "Rusty Russell" <rusty@au1.ibm.com>
> ---
> 
>  kernel/futex.c |    5 ++++-
>  1 files changed, 4 insertions(+), 1 deletions(-)

applied to tip/core/futexes, thanks!

I changed this:

> +		WARN_ON(1);
>  		return;

to WARN_ON_ONCE() - warning once per bootup is good enough for 
kerneloops.org to pick it up, while a ton of warnings add no extra (other 
than to inconvenience the user, or even make the system unworkable).

	Ingo