Re: [PATCH] synchronize_irq needs a barrier

On Fri, 2007-10-19 at 12:20 +0800, Herbert Xu wrote:

That's why I think this patch is in fact the only one that
solves all the races in this thread.  The case that it solves
which the lock/unlock patch does not is the one where action
flows downwards past the clearing of IRQ_INPROGRESS.  I missed
this case earlier.

In fact this bug exists elsewhere too.  For example, the network
stack does this in net/sched/sch_generic.c:

        /* Wait for outstanding qdisc_run calls. */
        while (test_bit(__LINK_STATE_QDISC_RUNNING, &dev->state))
                yield();

This has the same problem as the current synchronize_irq code.

The network stack always made be nervous with it's bitops use as locks.
Any loop of test_bit() alone as a synchronisation method, without locks
or barriers is, imho, inherently buggy.

Ben.