linux-kernel - Re: [RFC PATCH v2 09/19] scsi: core: increase/decrease target_busy without check can

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

Message-ID: <daba5c92-2395-4eee-b212-978fbe83b56f@oracle.com>
Date:   Tue, 5 Sep 2023 18:55:09 -0500
From:   Mike Christie <michael.christie@...cle.com>
To:     Wenchao Hao <haowenchao2@...wei.com>,
        "James E . J . Bottomley" <jejb@...ux.ibm.com>,
        "Martin K . Petersen" <martin.petersen@...cle.com>,
        linux-scsi@...r.kernel.org
Cc:     Hannes Reinecke <hare@...e.de>, linux-kernel@...r.kernel.org,
        louhongxiang@...wei.com, lixiaokeng@...wei.com
Subject: Re: [RFC PATCH v2 09/19] scsi: core: increase/decrease target_busy
 without check can_queue

On 9/1/23 4:41 AM, Wenchao Hao wrote:
> This is preparation for a genernal target based error handle strategy
> to check if to wake up actual error handler.
> 
> Signed-off-by: Wenchao Hao <haowenchao2@...wei.com>
> ---
>  drivers/scsi/scsi_lib.c | 11 ++++-------
>  1 file changed, 4 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
> index db0a42fe49c0..4a7fb48aa60f 100644
> --- a/drivers/scsi/scsi_lib.c
> +++ b/drivers/scsi/scsi_lib.c
> @@ -293,8 +293,7 @@ void scsi_device_unbusy(struct scsi_device *sdev, struct scsi_cmnd *cmd)
>  
>  	scsi_dec_host_busy(shost, cmd);
>  
> -	if (starget->can_queue > 0)
> -		atomic_dec(&starget->target_busy);
> +	atomic_dec(&starget->target_busy);
>  

Ming had found that removing the atomics improves perf.
Since most drivers didn't care about the target level counters
it was moved to the can_queue check so only drivers using the
feature suffer the perf hit.

Your patch should do the same.