On Wed, May 16, 2012 at 3:43 PM, Matthew Wilcox <willy@xxxxxxxxxxxxxxx> wrote: > On Wed, May 16, 2012 at 03:37:56PM -0400, chetan loke wrote: >> On Wed, May 16, 2012 at 3:35 PM, Matthew Wilcox <willy@xxxxxxxxxxxxxxx> wrote: >> > On Wed, May 16, 2012 at 03:31:55PM -0400, chetan loke wrote: >> >> > + if (list_empty(&ram_device->commands)) >> >> > + wake_up_process(ram_device->thread); >> >> >> >> Didn't look in detail but if the queue is empty then why would you >> >> want to wake up the kthread? What if you just wake_up after >> >> list_add_tail below? >> > >> > If the list is non-empty, then the kthread has already been woken up >> > and doesn't need to be woken again. >> >> Sorry, not able to follow. wait_even_interruptible will put kthread to >> sleep. So how will it be already awake? > > Consider the following: > > CPU 0 CPU 1 > ->queuecommand > lock > wakes kthread > queues command 1 > unlock > > ->queuecommand > lock > kthread wakes > lock > queues command 2 > unlock > dequeues command 1 > dequeues command 2 > unlock > > > See? No need to wake the kthread *if there's already something on the > queue*, because you know it was already woken by whoever put the first > command on the queue. I thought that 'scsi_ram_device_thread' has a 'wait_event_interruptible' call if it sees that the list is empty. I don't see that call that's why the confusion. Either queuecmd or kthread will get the lock and so why keep the kthread spinning in that while loop if there's nothing to do? -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
