As we know, there is a risk of accesing stale requests when iterate in-flight requests with tags->rqs[] and this has been talked in following thread, [1] https://marc.info/?l=linux-scsi&m=154511693912752&w=2 [2] https://marc.info/?l=linux-block&m=154526189023236&w=2 A typical sence could be blk_mq_get_request blk_mq_queue_tag_busy_iter -> blk_mq_get_tag -> bt_for_each -> bt_iter -> rq = taags->rqs[] -> rq->q -> blk_mq_rq_ctx_init -> data->hctx->tags->rqs[rq->tag] = rq; The root cause is that there is a window between set bit on tag sbitmap and set tags->rqs[]. This patch would fix this issue by iterating requests with tags->static_rqs[] instead of tags->rqs[] which would be changed dynamically. Moreover, we will try to get a non-zero q_usage_counter before access hctxs and tags and thus could avoid the race with updating nr_hw_queues, switching io scheduler and even queue clean up which are all under a frozen and drained queue. The 1st patch get rid of the useless of synchronize_rcu in __blk_mq_update_nr_hw_queues The 2nd patch modify the blk_mq_queue_tag_busy_iter to use tags->static_rqs[] instead of tags->rqs[] to iterate the busy tags. The 3rd ~ 7th patch change the blk_mq_tagset_busy_iter to blk_mq_queue_tag_busy_iter which is safer The 8th patch get rid of the blk_mq_tagset_busy_iter. Change log V1 -> V2: - Add wrapper to hide "inflight" parameter to user based on Sagi's suggestion. - Other misc changes on comment. Jianchao Wang (8) blk-mq: get rid of the synchronize_rcu in blk-mq: use static_rqs instead of rqs to iterate tags blk-mq: use blk_mq_queue_tag_inflight_iter in debugfs mtip32xx: use blk_mq_queue_tag_inflight_iter nbd: use blk_mq_queue_tag_inflight_iter skd: use blk_mq_queue_tag_inflight_iter nvme: use blk_mq_queue_tag_inflight_iter blk-mq: remove blk_mq_tagset_busy_iter diff stat block/blk-mq-debugfs.c | 2 +- block/blk-mq-tag.c | 193 ++++++++++++++------------------------ block/blk-mq-tag.h | 4 +- block/blk-mq.c | 31 ++---- drivers/block/mtip32xx/mtip32xx.c | 6 +- drivers/block/nbd.c | 2 +- drivers/block/skd_main.c | 4 +- drivers/nvme/host/core.c | 12 +++ drivers/nvme/host/fc.c | 10 +- drivers/nvme/host/nvme.h | 2 + drivers/nvme/host/pci.c | 5 +- drivers/nvme/host/rdma.c | 4 +- drivers/nvme/host/tcp.c | 5 +- drivers/nvme/target/loop.c | 4 +- include/linux/blk-mq.h | 7 +- 15 files changed, 119 insertions(+), 172 deletions(-) Thanks Jianchao
