Re: [PATCH] lock_page() doesn't lock if __wait_on_bit_lock returns -EINTR

[Linus Torvalds]

On Sat, Dec 12, 2015 at 10:33 AM, Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> Peter, did that patch also handle just plain "lock_page()" case?

Looking more at it, I think this all goes back to commit 743162013d40
("sched: Remove proliferation of wait_on_bit() action functions").

Before that, we had wait_on_page_bit() doing:

        __wait_on_bit(page_waitqueue(page), &wait, sleep_on_page,
TASK_UNINTERRUPTIBLE);

and after that, the "sleep_on_page" got changed to "bit_wait_io".

But that is bogus, because sleep_on_page() used to look like this:

    static int sleep_on_page(void *word)
    {
        io_schedule();
        return 0;
    }

while bit_wait_io() looks like this:

    __sched int bit_wait_io(void *word)
    {
        if (signal_pending_state(current->state, current))
                return 1;
        io_schedule();
        return 0;
    }

which is ok, because as long as the task state is
TASK_UNINTERRUPTIBLE, the whole signal_pending_state() thing turns
into a no-op.

So far, so fine.

However, then commit 68985633bccb ("sched/wait: Fix signal handling in
bit wait helpers") _really_ screwed up, and changed the function to

    __sched int bit_wait(struct wait_bit_key *word)
    {
        schedule();
        if (signal_pending(current))
                return -EINTR;
        return 0;
    }

so now it returns an error when no error should happen. Which in turn
makes __wait_on_bit() exit the bit-wait loop early.

It looks like PeterZ's pending patch should fix this, by passing in
the proper TASK_UNINTERRUPTIBLE to the bit_wait_io function, and going
back to signal_pending_state(). PeterZ, did I follow the history of
this correctly?

                   Linus
--
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/