lkml.org 
[lkml]   [2006]   [Nov]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH -mm 4/5][AIO] - AIO completion signal notification
On Wed, 29 Nov 2006 10:51:50 +0000, Christoph Hellwig <hch@infradead.org> wrote:

> I'm a little bit unhappy about the usage of the notify flag. The usage
> seems correct but very confusing:

Well, I followed the logic from posix-timers.c, but it may be a poor
choice ;-)

For a start, the SIGEV_* flags are quite confusing (for me at least).
SIGEV_SIGNAL is defined as 0, SIGEV_NONE as 1 and SIGEV_THREAD_ID as 4. I
would rather have seen SIGEV_NONE defined as 0 to avoid all this.

I also wish I knew why those SIGEV_* constants were defined that way.

>
> In io_submit_one we set ki_notify.notify to SIGEV_NONE and possibly
> call aio_setup_sigevent:
>
> > + /* handle setting up the sigevent for POSIX AIO signals */
> > + req->ki_notify.notify = SIGEV_NONE;
> > +
> > + if (iocb->aio_sigeventp) {
> > + ret = aio_setup_sigevent(&req->ki_notify,
> > + (struct sigevent __user *)(unsigned
> > long)
> > + iocb->aio_sigeventp);
> > + if (ret)
> > + goto out_put_req;
> > + }
> > +
>
> aio_setup_sigevent then checks the user passed even for which notify type
> we have, and returns if it's none or otherwise sets notify->notify to it.
>
> > + if (event.sigev_notify == SIGEV_NONE)
> > + return 0;
> > +
> > + notify->notify = event.sigev_notify;
>
> Later aio_setup_sigevent gets a reference to the target task_structure
> if notify->notify is (SIGEV_SIGNAL|SIGEV_THREAD_ID) but _always_ stores
> the target pointer.

Yep, as SIGEV_SIGNAL is 0, this in fact checks that notify is SIGEV_THREAD_ID.
It could have been written as:

if (notify->notify == SIGEV_THREAD_ID)

And the target pointer is always store because at this point notify is either
SIGEV_SIGNAL or SIGEV_THREAD_ID.

>
> > + if (notify->notify == (SIGEV_SIGNAL|SIGEV_THREAD_ID)) {
> > + /*
> > + * This reference will be dropped in really_put_req() when
> > + * we're done with the request.
> > + */
> > + get_task_struct(target);
> > + }
> > +
> > + notify->target = target;
>
>
> Once we're done with the iocb aio_complete aclls aio_send_signal if
> notify.notify is not SIGEV_NONE.

Again, if it's not SIGEV_NONE, then it's either SIGEV_SIGNAL or
SIGEV_THREAD_ID.

>
> > + if (iocb->ki_notify.notify != SIGEV_NONE) {
> > + ret = aio_send_signal(&iocb->ki_notify);
> > +
> > + /* If signal generation failed, release the sigqueue */
> > + if (ret)
> > + sigqueue_free(iocb->ki_notify.sigq);
> > + }
> > +
>
> Which then uses notify->target to send the signal:
> > + if (notify->notify & SIGEV_THREAD_ID)
> > + ret = send_sigqueue(notify->signo, sigq, notify->target);
> > + else
> > + ret = send_group_sigqueue(notify->signo, sigq, notify->target);
>
> And finally really_put_req puts the target if notify.notify contains
> either SIGEV_SIGNAL or SIGEV_THREAD_ID.
>
> > + /* Release task ref */
> > + if (req->ki_notify.notify == (SIGEV_SIGNAL|SIGEV_THREAD_ID))
> > + put_task_struct(req->ki_notify.target);

Could have been if (req->ki_notify.notify == SIGEV_THREAD_ID)

>
> Do you see the confusing? I think all the notify.notify != SIGEV_NONE
> in the above code should be replaces by the much more descriptive
> notify.notify == (SIGEV_SIGNAL|SIGEV_THREAD_ID). In addition we should
> only store the target pointer inside the (SIGEV_SIGNAL|SIGEV_THREAD_ID)
> if block that gets a reference to it.

No, I don't think so, notify == (SIGEV_SIGNAL|SIGEV_THREAD_ID) really means
notify == SIGEV_THREAD_ID which leaves out notify == SIGEV_SIGNAL. Or
am I grossly mistaken?

Thanks,

Sébastien.



> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2006-11-29 14:13    [W:0.053 / U:3.172 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site