From mboxrd@z Thu Jan 1 00:00:00 1970 From: Hans de Goede Subject: Re: [PATCH 1/2] uas: Set no_report_opcodes Date: Wed, 10 Sep 2014 10:17:04 +0200 Message-ID: <54100900.2050109@redhat.com> References: <1410274800-11578-1-git-send-email-hdegoede@redhat.com> <1410274800-11578-2-git-send-email-hdegoede@redhat.com> <20140909160150.GA15724@infradead.org> <540F5333.80201@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 8bit Return-path: In-Reply-To: <540F5333.80201@redhat.com> Sender: stable-owner@vger.kernel.org To: Christoph Hellwig Cc: Greg Kroah-Hartman , linux-usb@vger.kernel.org, linux-scsi@vger.kernel.org, stable@vger.kernel.org List-Id: linux-scsi@vger.kernel.org Hi, On 09/09/2014 09:21 PM, Hans de Goede wrote: > Hi, > > On 09/09/2014 06:01 PM, Christoph Hellwig wrote: >> On Tue, Sep 09, 2014 at 04:59:59PM +0200, Hans de Goede wrote: >>> asm1051e usb <-> sata bridges hang when receiving a report opcodes scsi cmnd. >>> Take a page out of the usb-storage book, and simple disable no_report_opcodes >>> outright. >> >> Given that this device also seems broken in other ways can we wait a bit >> before using the big hammer? > > Actually the big hammer I would like to avoid is disabling uas all together > on these devices, as they work fine with 2 out of 3 of the 3 disks I've > tested with, as long as no report opcodes is not used. > > Which is why my other patch also includes a log message to explain how > to enable uas despite the blacklist, but that will only work if we don't > send report opcodes. > >> I'm still hoping UAS might give us some >> better SCSI implementation so that we don't have to disable any kind of >> advanced feature. > > I understand, so an alternative would be to make this a quirk and only > set it for ASM1051/ASM1053 bridges. So further (stress) testing of ASM1051 bridges with disks with which they do work under normal conditions have shown that even with those disks they can get stuck / hang in a state which can only be recovered by an unplug + replug (power cycle). So basically ASM1051 bridges should just never be used together with uas, which means that this patch can be dropped. So: self-nak. I'll respin the second patch in this set to change the log messages a bit to reflect this new insight. Regards, Hans