On Thu, Oct 20, 2016 at 5:04 AM, Bart Van Assche
[off-list ref] wrote:

On 10/18/2016 02:50 PM, Bart Van Assche wrote:

quoted

blk_mq_quiesce_queue() waits until ongoing .queue_rq() invocations
have finished. This function does *not* wait until all outstanding
requests have finished (this means invocation of request.end_io()).

(replying to my own e-mail)

The zero-day kernel test infrastructure reported to me that this patch
causes a build failure with CONFIG_SRCU=n. Should I add "select SRCU" to
block/Kconfig (excludes TINY_RCU) or should I rather modify this patch such

Select SRCU is fine, and you can see it is done in lots of
places(btrfs, net, quota, kvm, power,...)

that a mutex or rwsem is used instead of SRCU?

Both should be much worse than SRCU, even not as good as atomic_t.

Thanks,
Ming

Thanks,

Bart.

-- 
Ming Lei