linux-kernel - Re: [PATCH] scsi: do not requeue requests unaligned with device sector size

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:   Wed, 21 Dec 2016 09:01:11 -0500
From:   "Martin K. Petersen" <martin.petersen@...cle.com>
To:     Mauricio Faria de Oliveira <mauricfo@...ux.vnet.ibm.com>
Cc:     jejb@...ux.vnet.ibm.com, martin.petersen@...cle.com,
        linux-scsi@...r.kernel.org, linux-block@...r.kernel.org,
        linux-kernel@...r.kernel.org, gpiccoli@...ux.vnet.ibm.com
Subject: Re: [PATCH] scsi: do not requeue requests unaligned with device sector size

>>>>> "Mauricio" == Mauricio Faria de Oliveira <mauricfo@...ux.vnet.ibm.com> writes:

Mauricio,

Mauricio> When a SCSI command (e.g., read operation) is partially
Mauricio> completed with good status and residual bytes (i.e., not all
Mauricio> the bytes from the specified transfer length were transferred)
Mauricio> the SCSI midlayer will update the request/bios with the
Mauricio> completed bytes and requeue the request in order to complete
Mauricio> the remainder/pending bytes.

I agree with Christoph and Hannes. Some of this falls into the gray area
that's outside of the T10 spec (HBA programming interface guarantees)
but it seems like a deficiency in the HBA to report a byte count that's
not a multiple of the logical block size. A block can't be partially
written. Either it made it or it didn't. Regardless of how the I/O is
being broken up into frames at the transport level and at which offset
the transfer was interrupted.

I am also not a fan of the delayed retry stuff which seems somewhat
orthogonal to the problem you're describing.

-- 
Martin K. Petersen	Oracle Linux Engineering