On Thu, Oct 27, 2016 at 10:40 AM, Bart Van Assche [off-list ref] wrote:

On 10/26/16 19:31, Ming Lei wrote:

quoted

On Thu, Oct 27, 2016 at 10:04 AM, Bart Van Assche [off-list ref]
wrote:

quoted

On 10/26/16 18:30, Ming Lei wrote:

quoted

On Thu, Oct 27, 2016 at 6:53 AM, Bart Van Assche
[off-list ref] 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()).
The algorithm used by blk_mq_quiesce_queue() is as follows:
* Hold either an RCU read lock or an SRCU read lock around
  .queue_rq() calls. The former is used if .queue_rq() does not
  block and the latter if .queue_rq() may block.
* blk_mq_quiesce_queue() calls synchronize_srcu() or
  synchronize_rcu() to wait for .queue_rq() invocations that
  started before blk_mq_quiesce_queue() was called.
* The blk_mq_hctx_stopped() calls that control whether or not
  .queue_rq() will be called are called with the (S)RCU read lock
  held. This is necessary to avoid race conditions against
  the "blk_mq_stop_hw_queues(q); blk_mq_quiesce_queue(q);"
  sequence from another thread.

Signed-off-by: Bart Van Assche <redacted>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Ming Lei <tom.leiming@gmail.com>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Johannes Thumshirn <redacted>
---
 block/Kconfig          |  1 +
 block/blk-mq.c         | 69
+++++++++++++++++++++++++++++++++++++++++++++-----
 include/linux/blk-mq.h |  3 +++
 include/linux/blkdev.h |  1 +
 4 files changed, 67 insertions(+), 7 deletions(-)

diff --git a/block/Kconfig b/block/Kconfig
index 1d4d624..0562ef9 100644
--- a/block/Kconfig
+++ b/block/Kconfig

@@ -5,6 +5,7 @@ menuconfig BLOCK
        bool "Enable the block layer" if EXPERT
        default y
        select SBITMAP
+       select SRCU
        help
         Provide block layer support for the kernel.

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 0cf21c2..4945437 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c

@@ -115,6 +115,31 @@ void blk_mq_unfreeze_queue(struct request_queue

*q)
 }
 EXPORT_SYMBOL_GPL(blk_mq_unfreeze_queue);

+/**
+ * blk_mq_quiesce_queue() - wait until all ongoing queue_rq calls have
finished
+ * @q: request queue.
+ *
+ * Note: this function does not prevent that the struct request
end_io()
+ * callback function is invoked. Additionally, it is not prevented
that
+ * new queue_rq() calls occur unless the queue has been stopped first.
+ */
+void blk_mq_quiesce_queue(struct request_queue *q)
+{
+       struct blk_mq_hw_ctx *hctx;
+       unsigned int i;
+       bool rcu = false;

Before synchronizing SRCU/RCU, we have to set a per-hctx flag
or per-queue flag to block comming .queue_rq(), seems I mentioned
that before:

   https://www.spinics.net/lists/linux-rdma/msg41389.html

Hello Ming,

Thanks for having included an URL to an archived version of that
discussion.
What I remember about that discussion is that I proposed to use the
existing
flag BLK_MQ_S_STOPPED instead of introducing a
new QUEUE_FLAG_QUIESCING flag and that you agreed with that proposal. See
also https://www.spinics.net/lists/linux-rdma/msg41430.html.

Yes, I am fine with either one, but the flag need to set in
blk_mq_quiesce_queue(), doesnt't it?

Hello Ming,

If you have a look at the later patches in this series then you will see
that the dm core and the NVMe driver have been modified such that
blk_mq_stop_hw_queues(q) is called immediately before
blk_mq_quiesce_queue(q) is called.

Cause any current and future users of blk_mq_quiesce_queue(q)
have to set the flag via blk_mq_stop_hw_queues(q), why not set
the flag explicitly in blk_mq_quiesce_queue(q)?


thanks,
Ming Lei