Re: Locking Between User Context and Soft IRQs in 2.4.0

[Jeff Garzik <jgarzik@mandrakesoft.com>]

Andi Kleen wrote:
> 
> On Sat, Nov 04, 2000 at 04:45:33AM -0500, Jeff Garzik wrote:
> >
> > > *       What about dev->open and dev->stop ?
> >
> > Sleep all you want, we'll leave the light on for ya.
> 
> ... but make sure you have no module unload races (or at least not too
> huge holes, some are probably unavoidable with the current network
> driver interface, e.g. without moving module count management a bit up).
> This means you should do MOD_INC_USE_COUNT very early at least to
> minimize the windows (and DEC_USE_COUNT very late)

Can you provide a trace of a race or deadlock?  I do not see where there
are races in the current 2.4.x code.


> > dev->do_ioctl:
> >       Locking: Inside rtnl_lock() semaphore.
> >       Sleeping: OK
> 
> Just make sure you lock against your interrupt and xmit threads.

But of course :)  My doc only covered locks external to the driver.


> >       Locking: Inside dev->xmit_lock spinlock.
> >       Sleeping: NO[1]
> >
> >
> > NOTE [1]: On principle, you should not sleep when a spinlock is held.
> > However, since this spinlock is per-net-device, we only block ourselves
> > if we sleep, so the effect is mitigated.
> 
> Sounds like dangerous advice.

Probably... I changed the doc "just say no" :)

	Jeff


-- 
Jeff Garzik             | Dinner is ready when
Building 1024           | the smoke alarm goes off.
MandrakeSoft            |	-/usr/games/fortune
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/