Re: [PATCH] swap_readpage: avoid blk_wake_io_task() if !synchronous

[Hugh Dickins]

On Thu, 4 Jul 2019, Andrew Morton wrote:
> On Thu, 4 Jul 2019 18:03:01 +0200 Oleg Nesterov <oleg@xxxxxxxxxx> wrote:
> 
> > swap_readpage() sets waiter = bio->bi_private even if synchronous = F,
> > this means that the caller can get the spurious wakeup after return. This
> > can be fatal if blk_wake_io_task() does set_current_state(TASK_RUNNING)
> > after the caller does set_special_state(), in the worst case the kernel
> > can crash in do_task_dead().
> 
> I think we need a Fixes: and a cc:stable here?
> 
> IIRC, we're fixing 0619317ff8baa2 ("block: add polled wakeup task helper").

Yes, you are right.

But catch me by surprise: I had been thinking this was a 5.2 regression.
I guess something in 5.2 (doesn't matter what) has made it significantly
easier to hit: but now I look at old records, see that I hit it once on
5.0-rc1, then never again until 5.2.

Thanks, and to Oleg,
Hugh