linux-kernel - Re: [PATCH 0/3] Remove scsi

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: <6c83bd7f-9fd2-1b43-627f-615467fa55d4@huawei.com>
Date:   Thu, 19 Aug 2021 10:09:27 +0100
From:   John Garry <john.garry@...wei.com>
To:     Hannes Reinecke <hare@...e.de>,
        Bart Van Assche <bvanassche@....org>,
        "Martin K. Petersen" <martin.petersen@...cle.com>
CC:     <satishkh@...co.com>, <sebaddel@...co.com>, <kartilak@...co.com>,
        <jejb@...ux.ibm.com>, <linux-scsi@...r.kernel.org>,
        <linux-kernel@...r.kernel.org>, <hch@....de>
Subject: Re: [PATCH 0/3] Remove scsi_cmnd.tag

On 19/08/2021 08:50, Hannes Reinecke wrote:
>>>    select CPU_32v4 if ARCH_RPC
>>
>> Does that build fully for xconfig or any others which you tried?
>>
>  > Yep, xconfig and full build works.
> 
> Well.
> 
> Would've worked if you hadn't messed up tag handling for acornscsi :-)
>  > Besides: tag handling in acornscsi (and fas216, for that matter) seems
> to be completely broken.
> 
> Consider this beauty:
> 
> #ifdef CONFIG_SCSI_ACORNSCSI_TAGGED_QUEUE
>         /*
>          * tagged queueing - allocate a new tag to this command
>          */
>         if (SCpnt->device->simple_tags) {
>             SCpnt->device->current_tag += 1;
>             if (SCpnt->device->current_tag == 0)
>                 SCpnt->device->current_tag = 1;
>             SCpnt->tag = SCpnt->device->current_tag;
>         } else
> #endif

So isn't this just using the scsi_cmnd.tag as it own scribble?

> 
> which is broken on _soo many_ counts.
> Not only does it try to allocate its own tags, the code also assumes 
> that a tag value of '0' indicates that tagged queueing is not active:
> 

In case you missed it, Arnd B tried to clear out some old platforms 
earlier this year. With respect to rpc, Russell King apparently still 
uses it and has some SCSI patches:

https://lore.kernel.org/lkml/20210109174357.GB1551@shell.armlinux.org.uk/

I wonder what they are and maybe we can check. Anyway... I'd run any 
changes by him...

> static
> void acornscsi_abortcmd(AS_Host *host, unsigned char tag)
> {
>      host->scsi.phase = PHASE_ABORTED;
>      sbic_arm_write(host, SBIC_CMND, CMND_ASSERTATN);
> 
>      msgqueue_flush(&host->scsi.msgs);
> #ifdef CONFIG_SCSI_ACORNSCSI_TAGGED_QUEUE
>      if (tag)
>          msgqueue_addmsg(&host->scsi.msgs, 2, ABORT_TAG, tag);
>      else
> #endif
>          msgqueue_addmsg(&host->scsi.msgs, 1, ABORT);
> }
> 
> And, of course, there's the usual confusion about when to check for
> sdev->tagged_supported and sdev->simple_tags.
> 
> Drop me a note if I can give a hand.

Thanks! Let's see what happens to the series which you just sent.