[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20210629082542.vm3yh6k36d2zh3k5@beryllium.lan>
Date: Tue, 29 Jun 2021 10:25:42 +0200
From: Daniel Wagner <dwagner@...e.de>
To: Ming Lei <ming.lei@...hat.com>
Cc: Wen Xiong <wenxiong@...ibm.com>, james.smart@...adcom.com,
linux-kernel@...r.kernel.org, sagi@...mberg.me,
wenxiong@...ux.vnet.ibm.com
Subject: Re: [PATCH 1/1] block: System crashes when cpu hotplug + bouncing
port
On Tue, Jun 29, 2021 at 11:47:55AM +0800, Ming Lei wrote:
> > >>data.ctx = __blk_mq_get_ctx(q, cpu);
> > cpu=2048 if hctx_idx = 4
>
> Yeah, that is the issue I mentioned, any CPU in hctx->cpumask becomes
> offline, please try the following patch and see if it makes a
> difference:
Given that cpumask_first_and() will return nr_cpu_ids in this case,
can't we just bail out here and have to caller handle the error? I am
able to reproduce the crash you reported in [1] and the fix [2] here
works for me:
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -469,6 +469,8 @@ struct request *blk_mq_alloc_request_hctx(struct request_queue *q,
if (!blk_mq_hw_queue_mapped(data.hctx))
goto out_queue_exit;
cpu = cpumask_first_and(data.hctx->cpumask, cpu_online_mask);
+ if (cpu >= nr_cpu_ids)
+ goto out_queue_exit;
data.ctx = __blk_mq_get_ctx(q, cpu);
if (!q->elevator)
[1] https://lore.kernel.org/linux-block/20191117041233.GA30615@ming.t460p/
[2] https://lore.kernel.org/linux-block/20210608183339.70609-1-dwagner@suse.de/
Powered by blists - more mailing lists