Re: [PATCH 0 of 4] Generic AIO by scheduling stacks

From: Davide Libenzi
Date: Sun Feb 04 2007 - 15:00:35 EST


On Sat, 3 Feb 2007, Davide Libenzi wrote:

> > - Signals. I have no idea what behaviour we want. Help? My first guess is
> > that we'll want signal state to be shared by fibrils by keeping it in the
> > task_struct. If we want something like individual cancellation, we'll augment
> > signal_pending() with some some per-fibril test which will cause it to return
> > from TASK_INTERRUPTIBLE (the only reasonable way to implement generic
> > cancellation, I'll argue) as it would have if a signal was pending.
>
> Fibril should IMO use current thread signal policies. I think a signal
> should hit (wake) any TASK_INTERRUPTIBLE fibril, if the current thread
> policies mandate that. I'd keep a list_head of currently scheduled-out
> TASK_INTERRUPTIBLE fibrils, and I'd make them runnable when a signal is
> delivered to the thread (wake_target bit #1 set to mean wake-all-interruptable-fibrils?).
> The other thing is signal_pending(). The sigpending flag test is not going
> to work as is (cleared at the first do_signal). Setting a bit in each
> fibril would mean walking the whole TASK_INTERRUPTIBLE fibril list. Maybe
> a sequential signal counter in task_struct, matched by one in the fibril.
> A signal would increment the task_struct counter, and a fibril
> schedule-out would save the task_struct counter to the fibril. The
> signal_pending() for a fibril is a compare of the two. Or something
> similar.

Another thing linked to signals that was not talked about, is cancellation
of an in-flight request. We want to give the ability to cancel an
in-flight request, with something like async_cancel(cookie). In my
userspace library I simply disable SA_RESTART of SIGUSR2, and I do a
pthread_kill() on the thread servicing the request. But this will IMO have
other implications (linked to signal delivery) in a kernel fibril-based
implementation, to think about it.



- Davide


-
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/