On 7/30/19 6:46 PM, Ming Lei wrote:
On Tue, Jul 30, 2019 at 11:17:56AM -0700, Bart Van Assche wrote:
It is required to protect blkg_lookup() calls with a blk_queue_enter() /
blk_queue_exit() pair. Since it is nontrivial to verify whether this is
Could you explain the reason why the blk_queue_enter()/blk_queue_exit()
pair is required for blkg_lookup()? And comment on blkg_lookup() only
mentioned RCU read lock is needed.
The request queue reference counter is supposed to be held for any
caller of submit_bio(), why isn't that ref count enough?
This patch was added to my own tree before commit
47cdee29ef9d ("block: move blk_exit_queue into __blk_release_queue")
went upstream. Since blk_exit_queue() has been moved from
blk_cleanup_queue() into __blk_release_queue() I don' think that we
need this patch series. I will drop these two patches.
Bart.
