blk-mq: use get_cpu/put_cpu instead of preempt_disable/preempt_enable
authorPaolo Bonzini <pbonzini@redhat.com>
Fri, 7 Nov 2014 22:04:00 +0000 (23:04 +0100)
committerJens Axboe <axboe@fb.com>
Tue, 11 Nov 2014 18:04:49 +0000 (11:04 -0700)
commit2a90d4aae5509e9cf1ba848c5d0b3458201160a0
treef327e5db3649df99d7f957ff6932eb7e7b5fc4ed
parent398205b8391b208f0034a392242867b28ad8af3d
blk-mq: use get_cpu/put_cpu instead of preempt_disable/preempt_enable

blk-mq is using preempt_disable/enable in order to ensure that the
queue runners are placed on the right CPU.  This does not work with
the RT patches, because __blk_mq_run_hw_queue takes a non-raw
spinlock with the preemption-disabled region.  If there is contention
on the lock, this violates the rules for preemption-disabled regions.

While this should be easily fixable within the RT patches just by doing
migrate_disable/enable, we can do better and document _why_ this
particular region runs with disabled preemption.  After the previous
patch, it is trivial to switch it to get/put_cpu; the RT patches then
can change it to get_cpu_light, which lets virtio-blk run under RT
kernels.

Cc: Jens Axboe <axboe@kernel.dk>
Cc: Thomas Gleixner <tglx@linutronix.de>
Reported-by: Clark Williams <williams@redhat.com>
Tested-by: Clark Williams <williams@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
block/blk-mq.c