linux-kernel - Re: [PATCH] block: fix io hung by block throttle

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>] [day] [month] [year] [list]

Message-ID: <0f3ffdbe-b893-7ef1-de69-188146b977c2@oracle.com>
Date:   Sat, 17 Apr 2021 14:37:57 -0700
From:   Junxiao Bi <junxiao.bi@...cle.com>
To:     Hillf Danton <hdanton@...a.com>
Cc:     linux-block@...r.kernel.org, linux-kernel@...r.kernel.org,
        axboe@...nel.dk
Subject: Re: [PATCH] block: fix io hung by block throttle

On 4/17/21 3:10 AM, Hillf Danton wrote:

>> --- a/block/blk-rq-qos.c
>> +++ b/block/blk-rq-qos.c
>> @@ -260,19 +260,17 @@ void rq_qos_wait(struct rq_wait *rqw, void *private_data,
>>   		.cb = acquire_inflight_cb,
>>   		.private_data = private_data,
>>   	};
>> -	bool has_sleeper;
>>   
>> -	has_sleeper = wq_has_sleeper(&rqw->wait);
>> -	if (!has_sleeper && acquire_inflight_cb(rqw, private_data))
>> +	if (!wq_has_sleeper(&rqw->wait)
>> +		&& acquire_inflight_cb(rqw, private_data))
>>   		return;
>>   
>>   	prepare_to_wait_exclusive(&rqw->wait, &data.wq, TASK_UNINTERRUPTIBLE);
>> -	has_sleeper = !wq_has_single_sleeper(&rqw->wait);
>>   	do {
>>   		/* The memory barrier in set_task_state saves us here. */
>>   		if (data.got_token)
>>   			break;
>> -		if (!has_sleeper && acquire_inflight_cb(rqw, private_data)) {
>> +		if (acquire_inflight_cb(rqw, private_data)) {
>>   			finish_wait(&rqw->wait, &data.wq);
> Simply removing !has_sleeper is not enough if it is mandatory before
> acquire_inflight_cb() without adding something like a mutex to sieve the
> concurrent sleepers out, see below.
>
>
> --- a/block/blk-rq-qos.c
> +++ b/block/blk-rq-qos.c
> @@ -260,19 +260,18 @@ void rq_qos_wait(struct rq_wait *rqw, vo
>   		.cb = acquire_inflight_cb,
>   		.private_data = private_data,
>   	};
> -	bool has_sleeper;
>   
> -	has_sleeper = wq_has_sleeper(&rqw->wait);
> -	if (!has_sleeper && acquire_inflight_cb(rqw, private_data))
> -		return;
> +	mutex_lock(&rqw->mutex);
> +
> +	if (acquire_inflight_cb(rqw, private_data))

This function is to increase atomic variable rq_wait->inflight. What's 
the mutex for?

Thanks,

Junxiao.

> +		goto out;
>   
>   	prepare_to_wait_exclusive(&rqw->wait, &data.wq, TASK_UNINTERRUPTIBLE);
> -	has_sleeper = !wq_has_single_sleeper(&rqw->wait);
>   	do {
>   		/* The memory barrier in set_task_state saves us here. */
>   		if (data.got_token)
>   			break;
> -		if (!has_sleeper && acquire_inflight_cb(rqw, private_data)) {
> +		if (acquire_inflight_cb(rqw, private_data)) {
>   			finish_wait(&rqw->wait, &data.wq);
>   
>   			/*
> @@ -286,10 +285,11 @@ void rq_qos_wait(struct rq_wait *rqw, vo
>   			break;
>   		}
>   		io_schedule();
> -		has_sleeper = true;
>   		set_current_state(TASK_UNINTERRUPTIBLE);
>   	} while (1);
>   	finish_wait(&rqw->wait, &data.wq);
> +out:
> +	mutex_unlock(&rqw->mutex);
>   }