Re: [PATCH 1/2] kthread: Move prio/affinite change into the newly created thread

From: Peter Zijlstra
Date: Tue Nov 17 2020 - 07:45:13 EST


On Tue, Nov 10, 2020 at 12:38:47PM +0100, Sebastian Andrzej Siewior wrote:
> With enabled threaded interrupts the nouveau driver reported the
> following:
> | Chain exists of:
> | &mm->mmap_lock#2 --> &device->mutex --> &cpuset_rwsem
> |
> | Possible unsafe locking scenario:
> |
> | CPU0 CPU1
> | ---- ----
> | lock(&cpuset_rwsem);
> | lock(&device->mutex);
> | lock(&cpuset_rwsem);
> | lock(&mm->mmap_lock#2);
>
> The device->mutex is nvkm_device::mutex.
>
> Unblocking the lockchain at `cpuset_rwsem' is probably the easiest thing
> to do.
> Move the priority reset to the start of the newly created thread.
>
> Fixes: 710da3c8ea7df ("sched/core: Prevent race condition between cpuset and __sched_setscheduler()")
> Reported-by: Mike Galbraith <efault@xxxxxx>
> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>
> Link: https://lkml.kernel.org/r/a23a826af7c108ea5651e73b8fbae5e653f16e86.camel@xxxxxx

Moo... yes this is certainly the easiest solution, because nouveau is a
horrible rats nest. But when I spoke to Greg KH about this, he suggested
nouveau ought to be fixed.

Ben, I got terminally lost when trying to untangle nouvea init, is there
any chance this can be fixed to not hold that nvkm_device::mutex thing
while doing request_irq() ?

> ---
> kernel/kthread.c | 16 ++++++++--------
> 1 file changed, 8 insertions(+), 8 deletions(-)
>
> diff --git a/kernel/kthread.c b/kernel/kthread.c
> index 933a625621b8d..4a31127c6efbf 100644
> --- a/kernel/kthread.c
> +++ b/kernel/kthread.c
> @@ -243,6 +243,7 @@ EXPORT_SYMBOL_GPL(kthread_parkme);
>
> static int kthread(void *_create)
> {
> + static const struct sched_param param = { .sched_priority = 0 };
> /* Copy data: it's on kthread's stack */
> struct kthread_create_info *create = _create;
> int (*threadfn)(void *data) = create->threadfn;
> @@ -273,6 +274,13 @@ static int kthread(void *_create)
> init_completion(&self->parked);
> current->vfork_done = &self->exited;
>
> + /*
> + * The new thread inherited kthreadd's priority and CPU mask. Reset
> + * back to default in case they have been changed.
> + */
> + sched_setscheduler_nocheck(current, SCHED_NORMAL, &param);
> + set_cpus_allowed_ptr(current, housekeeping_cpumask(HK_FLAG_KTHREAD));
> +
> /* OK, tell user we're spawned, wait for stop or wakeup */
> __set_current_state(TASK_UNINTERRUPTIBLE);
> create->result = current;
> @@ -370,7 +378,6 @@ struct task_struct *__kthread_create_on_node(int (*threadfn)(void *data),
> }
> task = create->result;
> if (!IS_ERR(task)) {
> - static const struct sched_param param = { .sched_priority = 0 };
> char name[TASK_COMM_LEN];
>
> /*
> @@ -379,13 +386,6 @@ struct task_struct *__kthread_create_on_node(int (*threadfn)(void *data),
> */
> vsnprintf(name, sizeof(name), namefmt, args);
> set_task_comm(task, name);
> - /*
> - * root may have changed our (kthreadd's) priority or CPU mask.
> - * The kernel thread should not inherit these properties.
> - */
> - sched_setscheduler_nocheck(task, SCHED_NORMAL, &param);
> - set_cpus_allowed_ptr(task,
> - housekeeping_cpumask(HK_FLAG_KTHREAD));
> }
> kfree(create);
> return task;
> --
> 2.29.2
>