Re: [PATCHSET v5] blk-mq: reimplement timeout handling

[jianchao.wang]

On 01/13/2018 05:19 AM, Bart Van Assche wrote:
> Sorry but I only retrieved the blk-mq debugfs several minutes after the hang
> started so I'm not sure the state information is relevant. Anyway, I have attached
> it to this e-mail. The most remarkable part is the following:
> 
> ./000000009ddfa913/requeue_list:000000009646711c {.op=READ, .state=idle, gen=0x1
> 18, abort_gen=0x0, .cmd_flags=, .rq_flags=SORTED|1|SOFTBARRIER|IO_STAT, complete
> =0, .tag=-1, .internal_tag=217}
> 
> The hexadecimal number at the start is the request_queue pointer (I modified the
> blk-mq-debugfs code such that queues are registered with there address just after
> creation and until a name is assigned). This is a dm-mpath queue.

There seems to be something wrong in hctx->nr_active.
./sde/hctx2/cpu2/completed:2 3
./sde/hctx2/cpu2/merged:0
./sde/hctx2/cpu2/dispatched:2 3
./sde/hctx2/active:5

./sde/hctx1/cpu1/completed:2 38
./sde/hctx1/cpu1/merged:0
./sde/hctx1/cpu1/dispatched:2 38
./sde/hctx1/active:40

./sde/hctx0/cpu0/completed:20 11
./sde/hctx0/cpu0/merged:0
./sde/hctx0/cpu0/dispatched:20 11
./sde/hctx0/active:31
...
./sdc/hctx1/cpu1/completed:14 13
./sdc/hctx1/cpu1/merged:0
./sdc/hctx1/cpu1/dispatched:14 13
./sdc/hctx1/active:21

./sdc/hctx0/cpu0/completed:1 41
./sdc/hctx0/cpu0/merged:0
./sdc/hctx0/cpu0/dispatched:1 41
./sdc/hctx0/active:36
....
Then hctx_may_queue return false.

Thanks
Jianchao