Re: [Intel-gfx] [PATCH] RFC: console: hack up console_lock more v2

[Chris Wilson]

Quoting Daniel Vetter (2019-05-06 08:45:53)
> +/**
> + * printk_safe_up - release the semaphore in console_unlock
> + * @sem: the semaphore to release
> + *
> + * Release the semaphore.  Unlike mutexes, up() may be called from any
> + * context and even by tasks which have never called down().
> + *
> + * NOTE: This is a special version of up() for console_unlock only. It is only
> + * safe if there are no killable, interruptible or timing out down() calls.
> + */
> +void printk_safe_up(struct semaphore *sem)
> +{
> +       unsigned long flags;
> +       struct semaphore_waiter *waiter = NULL;
> +
> +       raw_spin_lock_irqsave(&sem->lock, flags);
> +       if (likely(list_empty(&sem->wait_list))) {
> +               sem->count++;
> +       } else {
> +               waiter = list_first_entry(&sem->wait_list,
> +                                         struct semaphore_waiter, list);
> +               list_del(&waiter->list);
> +               waiter->up = true;
> +       }
> +       raw_spin_unlock_irqrestore(&sem->lock, flags);
> +
> +       if (waiter)
> +               wake_up_process(waiter->task);

>From comparing against __down_common() there's a risk here that as soon
as waiter->up == true, the waiter may complete and make the onstack
struct semaphore_waiter invalid. If you store waiter->task locally under
the spinlock that problem is resolved.

Then there is the issue of an unprotected dereference of the task in
wake_up_process() -- I think you can wrap this function with
rcu_read_lock() to keep that safe, and wake_up_process() should be a
no-op if it races against process termination.
-Chris