Re: block: sleeping in atomic warnings

[Jens Axboe]

On 2/7/23 7:06?AM, Dan Carpenter wrote:
> block/blk-mq.c:206 blk_freeze_queue() warn: sleeping in atomic context
> rexmit_timer() <- disables preempt
> -> aoedev_downdev()
>    -> blk_mq_freeze_queue()
>       -> blk_freeze_queue()

That is definitely a legit bug, aoe should punt to a workqueue or
similar.

> block/blk-mq.c:4083 blk_mq_destroy_queue() warn: sleeping in atomic context
> nvme_fc_match_disconn_ls() <- disables preempt
> -> nvme_fc_ctrl_put()
>    -> nvme_fc_ctrl_free()
>       -> nvme_remove_admin_tag_set()
> nvme_fc_ctrl_free() <duplicate>
> -> nvme_remove_io_tag_set()
>          -> blk_mq_destroy_queue()

Also looks like a legitimate bug.

> block/blk-mq.c:2174 __blk_mq_run_hw_queue() warn: sleeping in atomic context
> __blk_mq_run_hw_queue() <duplicate>
> -> blk_mq_sched_dispatch_requests()
>    -> __blk_mq_sched_dispatch_requests()
>       -> blk_mq_do_dispatch_sched()
> blk_mq_do_dispatch_sched() <duplicate>
> -> __blk_mq_do_dispatch_sched()
>    -> blk_mq_dispatch_hctx_list()
> __blk_mq_do_dispatch_sched() <duplicate>
> __blk_mq_sched_dispatch_requests() <duplicate>
> -> blk_mq_do_dispatch_ctx()
> __blk_mq_sched_dispatch_requests() <duplicate>
>       -> blk_mq_dispatch_rq_list()
> __blk_mq_do_dispatch_sched() <duplicate>
> blk_mq_do_dispatch_ctx() <duplicate>
> -> blk_mq_delay_run_hw_queues()
>          -> blk_mq_delay_run_hw_queue()

This one I'm not really following... Would it be possible to expand the
reporting to be a bit more verbose?

> sg_remove_sfp_usercontext() <- disables preempt
> block/blk-wbt.c:843 wbt_init() warn: sleeping in atomic context
> ioc_qos_write() <- disables preempt
> -> wbt_enable_default()
>    -> wbt_init()

Also definitely a bug.

-- 
Jens Axboe