[v3,0/3] blk-mq: Don't complete in IRQ, use llist_head
mbox series

Message ID 20210123201027.3262800-1-bigeasy@linutronix.de
Headers show
Series
  • blk-mq: Don't complete in IRQ, use llist_head
Related show

Message

Sebastian Andrzej Siewior Jan. 23, 2021, 8:10 p.m. UTC
Patch 2+3 were applied and then dropped by Jens due to a NOHZ+softirq
related warning [0]. Turns out a successful wakeup via
set_nr_if_polling() will not process any softirqs and the CPU may go
back to idle. This is addressed by patch #1.

smpcfd_dying_cpu() will also invoke SMP-functions calls via
flush_smp_call_function_queue() but the block layer shouldn't queue
anything because the CPU isn't online anymore.
The two caller of flush_smp_call_function_from_idle() look fine with
opening interrupts from within do_softirq().

[0] https://lkml.kernel.org/r/1ee4b31b-350e-a9f5-4349-cfb34b89829a@kernel.dk

Sebastian

Comments

Jens Axboe Jan. 25, 2021, 4:27 a.m. UTC | #1
On 1/23/21 1:10 PM, Sebastian Andrzej Siewior wrote:
> Patch 2+3 were applied and then dropped by Jens due to a NOHZ+softirq
> related warning [0]. Turns out a successful wakeup via
> set_nr_if_polling() will not process any softirqs and the CPU may go
> back to idle. This is addressed by patch #1.
> 
> smpcfd_dying_cpu() will also invoke SMP-functions calls via
> flush_smp_call_function_queue() but the block layer shouldn't queue
> anything because the CPU isn't online anymore.
> The two caller of flush_smp_call_function_from_idle() look fine with
> opening interrupts from within do_softirq().
> 
> [0] https://lkml.kernel.org/r/1ee4b31b-350e-a9f5-4349-cfb34b89829a@kernel.dk

I can queue up the block side once the IPI fix is in some stable branch
that I can pull in.
Jens Axboe Feb. 10, 2021, 2:43 p.m. UTC | #2
On 1/23/21 1:10 PM, Sebastian Andrzej Siewior wrote:
> Patch 2+3 were applied and then dropped by Jens due to a NOHZ+softirq
> related warning [0]. Turns out a successful wakeup via
> set_nr_if_polling() will not process any softirqs and the CPU may go
> back to idle. This is addressed by patch #1.
> 
> smpcfd_dying_cpu() will also invoke SMP-functions calls via
> flush_smp_call_function_queue() but the block layer shouldn't queue
> anything because the CPU isn't online anymore.
> The two caller of flush_smp_call_function_from_idle() look fine with
> opening interrupts from within do_softirq().

Applied, thanks.