The problem is due to the max number of avaliable tags is
max(total tags / active_queues, 4).
We fix it by ensuring the following:
a. don't restrict if the utilization rate of total tags is less than
100%
b. if the utilization rate of total tags is 100%, and someone failed
to get driver tag, start to restrict as before.
b. if the utilization rate of total tags goes below 100%, it can
stop to restrict quickly.
Yu Kuai (3):
blk-mq: allow hardware queue to get more tag while sharing a tag set
blk-mq: clear 'active_queues' immediately when 'nr_active' is
decreased to 0
blk-mq: decrease pending_queues when it expires
block/blk-mq-debugfs.c | 2 ++
block/blk-mq-tag.c | 61 +++++++++++++++++++++++++++++++++++++++++-
block/blk-mq-tag.h | 23 ++++++++++++++++
block/blk-mq.c | 9 +++++--
block/blk-mq.h | 11 +++++---
include/linux/blk-mq.h | 10 ++++++-
include/linux/blkdev.h | 1 +
7 files changed, 110 insertions(+), 7 deletions(-)
--
2.25.4
