blk-mq: don't keep offline CPUs mapped to hctx 0
authorMing Lei <ming.lei@redhat.com>
Sun, 8 Apr 2018 09:48:08 +0000 (17:48 +0800)
committerJens Axboe <axboe@kernel.dk>
Tue, 10 Apr 2018 14:38:46 +0000 (08:38 -0600)
From commit 4b855ad37194 ("blk-mq: Create hctx for each present CPU),
blk-mq doesn't remap queue after CPU topo is changed, that said when
some of these offline CPUs become online, they are still mapped to
hctx 0, then hctx 0 may become the bottleneck of IO dispatch and
completion.

This patch sets up the mapping from the beginning, and aligns to
queue mapping for PCI device (blk_mq_pci_map_queues()).

Cc: Stefan Haberland <sth@linux.vnet.ibm.com>
Cc: Keith Busch <keith.busch@intel.com>
Cc: stable@vger.kernel.org
Fixes: 4b855ad37194 ("blk-mq: Create hctx for each present CPU)
Tested-by: Christian Borntraeger <borntraeger@de.ibm.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
block/blk-mq-cpumap.c

index 9f8cffc..3eb169f 100644 (file)
 
 static int cpu_to_queue_index(unsigned int nr_queues, const int cpu)
 {
-       /*
-        * Non present CPU will be mapped to queue index 0.
-        */
-       if (!cpu_present(cpu))
-               return 0;
        return cpu % nr_queues;
 }