On Wed, 2008-04-23 at 13:47 +0200, Heiko Carstens wrote: > I don't think that works: the old code also relied on the fact that there > could be only one cpu sending an smp_call_* IPI (serialized by call_lock). > The current etr code exploits that to serialize all cpus: > > Do smp_call_function() to force all other cpus into clock_sync_cpu_start() > where they sort of busy wait(!). Then the master cpu is syncing the clock > and when it finished the other cpus may continue again and finally leave > the smp_call_function interrupt handler. > > Your patch series however doesn't make sure anymore that there is only > one cpu doing an smp_call_function*, so we may deadlock as soon as two > cpus are trying to synchronize all cpus this way. > After all I think the etr code should be converted to use stop_machine_run > instead. However that doesn't allow for master/slave cpus. Changing the etr code to stop_machine_run() will require a common code change. We can use the fn() parameter of stop_machine_run() for the cpu that does the clock synchronization but all the other cpus are not just disabled. They need to do something before going back to normal operation, namely waiting for the clock to get back into sync and then reprogram the clock comparator before we can allow interrupts again. Looking at the stop_machine() code it doesn't seem to be to hard to get this implemented. stop_machine_run() would get a second function argument that replaces the call to cpu_relax() in stopmachine(). The default if the additional function argument is NULL would be cpu_relax(). For the etr sync code it would be etr_sync_cpu_start(). -- blue skies, Martin. "Reality continues to ruin my life." - Calvin. -- To unsubscribe from this list: send the line "unsubscribe linux-arch" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
