From: Pete Zaitcev <zaitcev@redhat.com>
Subject: [RHEL5 patch] Sun/AMI virtual floppy issue
Date: Mon, 8 Jan 2007 18:42:14 -0800
Bugzilla: 219628
Message-Id: <20070108184214.e52a33ef.zaitcev@redhat.com>
Changelog: usb: Sun/AMI virtual floppy issue


This is a fix for bug 219628. Sun servers use a virtual floppy which gets
detected 8 times instead of one, and produces a bunch of 30-second timeouts
while every LUN is being detected. Since USB storage only accepts one SCSI
command at a time, floppy device is not accessible for 24 minutes after
a reboot.

The root cause has lots of boring details, but the main issue is that
the firmware behaves poorly when it receives MODE SENSE command, which
is where timeouts come from. Also, it reports its SCSI level as zero,
or "legacy", which completely caught us (and writers of USB storage
specs) by surprise. Such devices went out of fashion long before USB
was invented, so nobody knows for sure what to do with them.

Upstream, we used this case to streamline USB our stack a bit, getting
rid of SCSI level remapping, but the blacklist was still necessary in
the end. So for RHEL 5 I only want to add a blacklist which solves
both problems. I don't like grabbing the really bleeding edge upstream
for RHEL in case something crops up.

I tested the patch on the hardware and it works.

Please ack.

-- Pete

diff -urp -X dontdiff linux-2.6.18-1.2960.el5.bz219628.5/drivers/usb/storage/unusual_devs.h linux-2.6.18-1.2960.el5.bz219628.5a/drivers/usb/storage/unusual_devs.h
--- linux-2.6.18-1.2960.el5.bz219628.5/drivers/usb/storage/unusual_devs.h	2006-09-19 23:42:06.000000000 -0400
+++ linux-2.6.18-1.2960.el5.bz219628.5a/drivers/usb/storage/unusual_devs.h	2007-01-08 15:06:02.000000000 -0500
@@ -223,6 +223,21 @@ UNUSUAL_DEV(  0x045a, 0x5210, 0x0101, 0x
 		"Rio Karma",
 		US_SC_SCSI, US_PR_BULK, rio_karma_init, 0),
 
+/*
+ * This thing is found in a lot of Sun equipment (x4600, x4200m2, etc.)
+ * Reported by Pete Zaitcev <zaitcev@redhat.com>
+ * This device follows "legacy SCSI" (the one before SCSI 2), for which
+ * our USB stack is not set up well. We remap this to SCSI 2 and then
+ * device ignores LUN bits inside the CDB, where CBI transfer puts them.
+ * It also NAKs instead of STALL when transferring data for a failed command,
+ * so we must avoid sending MODE SENSE of any kind.
+ */
+UNUSUAL_DEV(  0x046b, 0xff40, 0x0100, 0x0100,
+		"AMI",
+		"Virtual Floppy",
+		US_SC_DEVICE, US_PR_DEVICE, NULL,
+		US_FL_SINGLE_LUN | US_FL_NOT_LOCKABLE | US_FL_NO_WP_DETECT),
+
 /* Patch submitted by Philipp Friedrich <philipp@void.at> */
 UNUSUAL_DEV(  0x0482, 0x0100, 0x0100, 0x0100,
 		"Kyocera",