linux-kernel - Re: [PATCH] scsi: core: Cap shost cmd_per_lun at 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]

Date:   Thu, 20 May 2021 18:24:02 +0100
From:   John Garry <john.garry@...wei.com>
To:     Bart Van Assche <bvanassche@....org>, <jejb@...ux.ibm.com>,
        <martin.petersen@...cle.com>
CC:     <linux-scsi@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <ming.lei@...hat.com>
Subject: Re: [PATCH] scsi: core: Cap shost cmd_per_lun at can_queue

On 20/05/2021 17:57, Bart Van Assche wrote:
>> not be limited to 16b?
> Maybe I'm missing something but it is not clear to me why different
> structures in the SCSI headers use different data types for can_queue
> and cmd_per_lun?

For cmd_per_lun, is it related to SCSI task tag limit? SAM-3 says upto 
64b for task tag, but then SAS uses 16b for TMF tag, so not sure.

Someone with more SCSI spec knowledge than we can clarify this.

> 
> $ git grep -nHEw '(cmd_per_lun|can_queue);' include/scsi
> include/scsi/scsi_device.h:318:	unsigned int		can_queue;
> include/scsi/scsi_host.h:372:	int can_queue;
> include/scsi/scsi_host.h:425:	short cmd_per_lun;
> include/scsi/scsi_host.h:612:	int can_queue;
> include/scsi/scsi_host.h:613:	short cmd_per_lun;
> 
>> It seems intentional that can_queue is int and cmd_per_lun is short.
> Intentional? It is not clear to me why? Even high-performance drivers
> like iSER and SRP set can_queue by default to a value that fits well in
> a 16-bit variable (512 and 64 respectively). The highest value that I
> found after a quick search is the following:
> 
>   #define ISCSI_TOTAL_CMDS_MAX		4096

I guess int was used for can_queue as an arbitrarily big number.

And if we try to use 16b for can_queue, reducing size of 
variables/structure members sometimes breaks things, from my experience.

Thanks,
John