All of lore.kernel.org
 help / color / mirror / Atom feed
From: Nish Aravamudan <nish.aravamudan@gmail.com>
To: "John M. King" <jmking1@uiuc.edu>
Cc: linux-kernel@vger.kernel.org
Subject: Re: Possible race in sound/oss/forte.c ?
Date: Thu, 11 Aug 2005 14:31:07 -0700	[thread overview]
Message-ID: <29495f1d05081114312e3dc2fa@mail.gmail.com> (raw)
In-Reply-To: <42FB8522.7080605@uiuc.edu>

On 8/11/05, John M. King <jmking1@uiuc.edu> wrote:
> I know the OSS drivers are deprecated, but I'm trying to figure this out
> for my own understanding.
> 
> Here's code from sound/oss/forte.c, in the write system call handler.  A
> test has already been performed (under the protection of the lock) and
> the driver has decided to sleep.
> 
>     add_wait_queue (&channel->wait, &wait);
> 
>     for (;;) {
>         spin_unlock_irqrestore (&chip->lock, flags);
> 
>         set_current_state (TASK_INTERRUPTIBLE);
>         schedule();
> 
>         spin_lock_irqsave (&chip->lock, flags);
> 
>         if (channel->frag_num - channel->filled_frags)
>             break;
>     }
> 
>     remove_wait_queue (&channel->wait, &wait);
>     set_current_state (TASK_RUNNING);
> 
> The driver's interrupt handler calls wake_up_all().  What if an
> interrupt occurs just after the spin_unlock_irqrestore() but before
> setting TASK_INTERRUPTIBLE (and the interrupt handler does stuff that
> causes the tested conditional to be true as well)?  The interrupt calls
> wake_up_all(), but then when control returns here, the process will mark
> itself TASK_INTERRUPTIBLE right away and sleep, effectively missing the
> wake_up_all().
> 
> Is this a race condition?  If not, can someone point out the error(s) in
> my reasoning?  Please CC me as I'm not subscribed to the list.

This is broken code. You are supposed to set the state before adding
oneself to the wait-queue, if you are going to unconditionally sleep,
I believe.

So, the loop probably should be:

prepare_to_wait(&channel->wait, &wait, TASK_INTERRUPTIBLE);

for (;;) {
         spin_unlock_irqrestore (&chip->lock, flags);
 
         set_current_state (TASK_INTERRUPTIBLE); // redundant in the
first iteration
         schedule();
 
         spin_lock_irqsave (&chip->lock, flags);
 
         if (channel->frag_num - channel->filled_frags)
             break;
}
 
finish_wait(&channel->wait, &wait);

Thanks,
Nish

      reply	other threads:[~2005-08-11 21:31 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2005-08-11 17:04 Possible race in sound/oss/forte.c ? John M. King
2005-08-11 21:31 ` Nish Aravamudan [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=29495f1d05081114312e3dc2fa@mail.gmail.com \
    --to=nish.aravamudan@gmail.com \
    --cc=jmking1@uiuc.edu \
    --cc=linux-kernel@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.