From mboxrd@z Thu Jan 1 00:00:00 1970 From: James Bottomley Subject: Re: [Bug 80711] [PATCH]SG_FLAG_LUN_INHIBIT is no longer implemented and there's not way to prevent the kernel from using the 2nd cdb byte for the LUN Date: Mon, 25 Aug 2014 12:39:07 -0700 Message-ID: <1408995547.3629.7.camel@HansenPartnership.com> References: Mime-Version: 1.0 Content-Type: text/plain; charset="ISO-8859-15" Content-Transfer-Encoding: 7bit Return-path: Received: from bedivere.hansenpartnership.com ([66.63.167.143]:45045 "EHLO bedivere.hansenpartnership.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S933108AbaHYTjJ (ORCPT ); Mon, 25 Aug 2014 15:39:09 -0400 In-Reply-To: Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: Alan Stern Cc: Christoph Hellwig , "Martin K. Petersen" , Douglas Gilbert , Tiziano Bacocco , bugzilla-daemon@bugzilla.kernel.org, SCSI development list , USB list On Mon, 2014-08-25 at 10:44 -0400, Alan Stern wrote: > James, can you explain how the INQUIRY command in scsi_probe_lun() > managed to work back in the days when multi-lun SCSI-2 devices were > common? sdev->scsi_level doesn't get set when sdev is allocated, so it > initially contains 0, so the LUN bits won't get filled in when the > first INQUIRY command is sent. Then how could the target know which > logical unit the INQUIRY was meant for? Best guess, some patches over the course of time altered the way we do this and no-one noticed. I think it was probably the introduction of the unknown SCSI data level that caused the breakage. Historically, the LUN in CMD bits is left over from SCSI-1; it was incorporated into SCSI-2 for backward compatibility (even though SCSI-2 moved the LUN specification to the identify message). In SCSI-3 and beyond, those bits were obsoleted and transports took sole responsibility for LUN handling. I'm fairly certain all the SCSI-1 devices relying on this behaviour have long ago migrated to the great data centre in the sky. Alan's fix looks reasonable because we probe LUN 0 first (for SCSI-1 and 2 which has parallel scanning), which is why it doesn't matter (the bits are set to zero) and once we have LUN 0 we propagate the data to the target and make it the basis for future checks. I'd like to see a comment explaining this in the code, though ... James