|author||Jens Axboe <firstname.lastname@example.org>||2012-07-03 21:45:39 -0600|
|committer||Jens Axboe <email@example.com>||2012-10-04 11:04:35 +0200|
multiqueue: update MQ-TODO
- We can now limit queue depth successfully through our tag map. - We pass in REQ_END on the last request in a batch, this provides the needed information to the driver. Signed-off-by: Jens Axboe <firstname.lastname@example.org>
1 files changed, 1 insertions, 12 deletions
@@ -4,20 +4,9 @@ TODO list for multiqueue/new-queue. In no particular order.
- Better switching scheme for deciding when to run
queues inline and when to punt to kblockd.
-- Support for limiting queue depth
- - Retrieves batches of queue credits from hw_ctx
-- Allow driver to pass in a mq_ops->submit_rq fn
- - If set, we can split the work between ->queue_rq
- and ->submit_rq, where the latter talks to the hw
- to fire off multiple requests at once.
- - Useful for mtip32xx, where the port issue lock is
- really hot. This relies on the fact that we normally
- fire off > 1 request for each run of the queue.
- Implement proper scheme for not having lots of processes
or kblockds hammering on blk_mq_run_hw_queue() at the same
time. If someone else is already running the queue, we need
- not do anything.
+ not do anything. (partially done)