Re: [PATCH 2.6] serialization of kernelcapi work

[Armin Schindler]

On Thu, 18 Mar 2004, Andrew Morton wrote:
> Armin Schindler <armin@xxxxxxxxxx> wrote:
> >
> > Hi all,
> >
> > the ISDN kernelcapi function recv_handler() is triggered by
> > schedule_work() and dispatches the CAPI messages to the applications.
> >
> > Since a workqueue function may run on another CPU at the same time,
> > reordering of CAPI messages may occur.
>
> TCP has the same problem.
>
> > For serialization I suggest a mutex semaphore in recv_handler(),
> > patch is appended (yet untested).
>
> It will work OK.  It isn't very scalable of course, but I assume you're
> dealing with relatively low bandwidths.

Right, compared with network, I guess it is low bandwith.

> I would suggest that you look at avoiding the global semaphore.  Suppose
> someone has 64 interfaces or something.  Is that possible?  It might be
> better to put the semaphore into struct capi_ctr so you can at least
> process frames from separate cards in parallel.

This was just mentioned on the i4l-developer list too.
But not on capi_ctr basis, a semaphore per application (struct capi20_appl)
should be better.

The reason for the serialization is a possible re-ordering of messages
per application, so only the application needs the semaphore.

Thanks Andrew, I will prepare a new patch.

Armin


-
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/