On Tue, Nov 22, 2011 at 04:19:58PM -0500, Mike Snitzer wrote: [..] > > Storing q->node info at queue allocation time makes sense to me. In fact > > it might make sense to clean it up from blk_init_allocated_queue_node > > and assume that passed queue has queue->node set at the allocation time. > > > > CCing Mike Snitzer who introduced blk_init_allocated_queue_node(). Mike > > what do you think. I am not sure it makes sense to pass in nodeid, both > > at queue allocation and queue initialization time. To me, it should make > > more sense to allocate the queue at one node and that becomes the default > > node for reset of the initialization. > > Yeah, that makes sense to me too: > > From: Mike Snitzer <snitzer at redhat.com> > Subject: block: initialize request_queue's numa node during allocation > > Set request_queue's node in blk_alloc_queue_node() rather than > blk_init_allocated_queue_node(). This avoids blk_throtl_init() using > q->node before it is initialized. > > Rename blk_init_allocated_queue_node() to blk_init_allocated_queue(). > > Signed-off-by: Mike Snitzer <snitzer at redhat.com> Thanks Mike. Looks good to me. Acked-by: Vivek Goyal <vgoyal at redhat.com> Thanks Vivek
