linux-kernel - Re: [PATCH v2 2/3] scsi: introduce 'blocked' sysfs api

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]

Message-ID: <fd703370-29be-2822-eaf7-77da2308ee19@oracle.com>
Date:   Tue, 28 Mar 2023 11:06:39 -0500
From:   Mike Christie <michael.christie@...cle.com>
To:     Ye Bin <yebin@...weicloud.com>, jejb@...ux.ibm.com,
        martin.petersen@...cle.com, linux-scsi@...r.kernel.org,
        linux-kernel@...r.kernel.org
Cc:     Ye Bin <yebin10@...wei.com>
Subject: Re: [PATCH v2 2/3] scsi: introduce 'blocked' sysfs api

On 3/28/23 9:34 AM, Ye Bin wrote:
> From: Ye Bin <yebin10@...wei.com>
> 
> Introduce 'blocked' sysfs api to control scsi host blocking IO.
> Use this founction for test. Perhaps we can use this to do some fault
> recovery or firmware upgrades, as long as the driver support is good,

If it's for testing only then do people like a debugfs type of interface
is better?

If it's for actual production use like firmware upgrades and they can't
handle IO while they are doing the upgrade, then I think you need to do
more than just set a bit to prevent new IO. You also need to handle cmds
that have passed the scsi_queue_rq ready checks and have not been processed
by the driver's queuecommand. Also there are some issues like cmds can
still timeout and so you will get scsi_host_template->eh* calls still.

>  
> diff --git a/include/scsi/scsi_host.h b/include/scsi/scsi_host.h
> index 587cc767bb67..3e916dbac1cb 100644
> --- a/include/scsi/scsi_host.h
> +++ b/include/scsi/scsi_host.h
> @@ -659,6 +659,9 @@ struct Scsi_Host {
>  	/* The transport requires the LUN bits NOT to be stored in CDB[1] */
>  	unsigned no_scsi2_lun_in_cdb:1;
>  
> +	/* True host will blocking IO */
> +	unsigned host_blockio:1;
> +

Maybe rename to host_user_blocked to match the host_self_blocked naming.

I would make the comment similar to host_self_blocked's comment but
instead of the host requesting it userspace did.

>  	/*
>  	 * Optional work queue to be utilized by the transport
>  	 */