Messages in this thread Patch in this message | | | From | "Rafael J. Wysocki" <> | Subject | Re: [Bug #11989] Suspend failure on NForce4-based boards due to chanes in stop_machine | Date | Sat, 15 Nov 2008 14:37:45 +0100 |
| |
On Wednesday, 12 of November 2008, Rusty Russell wrote: > On Tuesday 11 November 2008 21:22:14 Ingo Molnar wrote: > > * Rafael J. Wysocki <rjw@sisk.pl> wrote: > > > So, it evidently fails while re-enabling the non-boot CPU and not > > > during disabling it as I thought before. > > (Resend, due to HTML version previously) > > But what is calling stop_machine in that path? > > There *is* a race, but I don't think it could cause this (we should make a > copy of active.fnret inside the lock before returning it).
Still, that seems to be the case.
> Two patches: one fixes that race, the next adds debugging spew. > > stop_machine: fix race with return value
With this patch applied (reproduced below for clarity) the problem is not reproducible any more.
Care to push it upstream ASAP?
Thanks, Rafael
--- stop_machine: fix race with return value
We should not access active.fnret outside the lock; in theory the next stop_machine could overwrite it.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> --- kernel/stop_machine.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
diff -r d7c9a15da615 kernel/stop_machine.c --- a/kernel/stop_machine.c Mon Nov 10 09:47:45 2008 +1100 +++ b/kernel/stop_machine.c Tue Nov 11 23:19:47 2008 +1030 @@ -112,7 +112,7 @@ int __stop_machine(int (*fn)(void *), void *data, const cpumask_t *cpus) { struct work_struct *sm_work; - int i; + int i, ret; /* Set up initial state. */ mutex_lock(&lock); @@ -137,8 +137,9 @@ /* This will release the thread on our CPU. */ put_cpu(); flush_workqueue(stop_machine_wq); + ret = active.fnret; mutex_unlock(&lock); - return active.fnret; + return ret; } int stop_machine(int (*fn)(void *), void *data, const cpumask_t *cpus)
| |