linux-kernel - RE:[PATCH 1/1] blk-mq: added case for cpu offline during send_ipi in rq

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Thu, 29 Sep 2022 12:34:28 +0900
From:   Manjong Lee <mj0123.lee@...sung.com>
To:     mj0123.lee@...sung.com, bvanassche@....org, ming.lei@...hat.com,
        hch@....de
Cc:     axboe@...nel.dk, jisoo2146.oh@...sung.com, junho89.kim@...sung.com,
        linux-block@...r.kernel.org, linux-kernel@...r.kernel.org,
        nanich.lee@...sung.com, sbeom16.kim@...sung.com,
        seunghwan.hyun@...sung.com, sookwan7.kim@...sung.com,
        yt0928.kim@...sung.com
Subject: RE:[PATCH 1/1] blk-mq: added case for cpu offline during send_ipi
 in rq_complete

Hello. I think there seems to be a problem that the request
complete processing will be omitted due to the follow routine.
Please give me a advice or opinion if I get a mistake or missing something.
Thank you.
>
>When a request complete then send ipi to original cpu which issued request.
>If cpu offline during this process, send_ipi might fail.
>However, there is currently no code to handle this error case.
>This may cause in missing request complete.
>Therefore, if send_ipi fails due to cpu offline, the request complete
>has to be processed directly from the cpu where it is running.
>
>
>
>Signed-off-by: Manjong Lee <mj0123.lee@...sung.com>
>Signed-off-by: Changheun Lee <nanich.lee@...sung.com>
>Signed-off-by: Junho Kim <junho89.kim@...sung.com>
>---
> block/blk-mq.c | 16 ++++++++++------
> 1 file changed, 10 insertions(+), 6 deletions(-)
>
>diff --git a/block/blk-mq.c b/block/blk-mq.c
>index 3c1e6b6d991d..f2ce79708c5e 100644
>--- a/block/blk-mq.c
>+++ b/block/blk-mq.c
>@@ -1064,17 +1064,22 @@ static inline bool blk_mq_complete_need_ipi(struct request *rq)
> 	return cpu_online(rq->mq_ctx->cpu);
> }
> 
>-static void blk_mq_complete_send_ipi(struct request *rq)
>+static int blk_mq_complete_send_ipi(struct request *rq)
> {
> 	struct llist_head *list;
> 	unsigned int cpu;
>+	int ret = 0;
> 
> 	cpu = rq->mq_ctx->cpu;
> 	list = &per_cpu(blk_cpu_done, cpu);
> 	if (llist_add(&rq->ipi_list, list)) {
> 		INIT_CSD(&rq->csd, __blk_mq_complete_request_remote, rq);
>-		smp_call_function_single_async(cpu, &rq->csd);
>+		ret = smp_call_function_single_async(cpu, &rq->csd);
>+		if (ret)
>+			llist_del_all(list);
> 	}
>+
>+	return ret;
> }
> 
> static void blk_mq_raise_softirq(struct request *rq)
>@@ -1099,10 +1104,9 @@ bool blk_mq_complete_request_remote(struct request *rq)
> 	if (rq->cmd_flags & REQ_POLLED)
> 		return false;
> 
>-	if (blk_mq_complete_need_ipi(rq)) {
>-		blk_mq_complete_send_ipi(rq);
>-		return true;
>-	}
>+	if (blk_mq_complete_need_ipi(rq))
>+		if (!blk_mq_complete_send_ipi(rq))
>+			return true;
> 
> 	if (rq->q->nr_hw_queues == 1) {
> 		blk_mq_raise_softirq(rq);
>-- 
>2.32.0
>