[PATCH]indistinguishable devices with broken and unbroken firmware

[PATCH]indistinguishable devices with broken and unbroken firmware

[Oliver Neukum]

Hi,

there's a USB mass storage device which exists in two version. One
reports the correct size and the other does not. Apart from that they
are identical and cannot be told apart. Here's a heuristic based on the
empirical finding that drives have even sizes. I know it's ugly, but it fixes
an ugly bug.

	Regards
		Oliver

Signed-off-by: Oliver Neukum <oneukum@suse.de>
----

--- a/drivers/usb/storage/unusual_devs.h	2007-02-06 14:14:49.000000000 +0100
+++ b/drivers/usb/storage/unusual_devs.h	2007-02-07 15:14:01.000000000 +0100
@@ -1430,7 +1430,7 @@
 		"DataStor",
 		"USB4500 FW1.04",
 		US_SC_DEVICE, US_PR_DEVICE, NULL,
-		US_FL_FIX_CAPACITY),
+		US_FL_CAPACITY_HEURISTICS),
 
 /* Control/Bulk transport for all SubClass values */
 USUAL_DEV(US_SC_RBC, US_PR_CB, USB_US_TYPE_STOR),
--- a/drivers/usb/storage/scsiglue.c	2007-02-06 14:13:49.000000000 +0100
+++ b/drivers/usb/storage/scsiglue.c	2007-02-07 15:14:01.000000000 +0100
@@ -170,6 +170,12 @@
 		if (us->flags & US_FL_FIX_CAPACITY)
 			sdev->fix_capacity = 1;
 
+		/* A few disks have two indistinguishable version, one of
+		 * which reports the correct capacity and the other does not.
+		 * The sd driver has to guess which is the case. */
+		if (us->flags & US_FL_CAPACITY_HEURISTICS)
+			sdev->guess_capacity = 1;
+
 		/* Some devices report a SCSI revision level above 2 but are
 		 * unable to handle the REPORT LUNS command (for which
 		 * support is mandatory at level 3).  Since we already have
--- a/drivers/scsi/sd.c	2007-02-06 14:14:48.000000000 +0100
+++ b/drivers/scsi/sd.c	2007-02-07 15:14:01.000000000 +0100
@@ -1273,6 +1273,13 @@
 	if (sdp->fix_capacity)
 		--sdkp->capacity;
 
+	/* Some devices have version which report the correct sizes
+	 * and others which do not. We guess size according to a heuristic
+	 * and err on the side of lowering the capacity. */
+	if (sdp->guess_capacity)
+		if (sdkp->capacity & 0x01) /* odd sizes are odd */
+			--sdkp->capacity;
+
 got_data:
 	if (sector_size == 0) {
 		sector_size = 512;
--- a/include/linux/usb_usual.h	2007-02-06 14:14:07.000000000 +0100
+++ b/include/linux/usb_usual.h	2007-02-07 15:14:01.000000000 +0100
@@ -46,7 +46,9 @@
 	US_FLAG(MAX_SECTORS_64,	0x00000400)			\
 		/* Sets max_sectors to 64    */			\
 	US_FLAG(IGNORE_DEVICE,	0x00000800)			\
-		/* Don't claim device */
+		/* Don't claim device */			\
+	US_FLAG(CAPACITY_HEURISTICS,	0x00001000)		\
+		/* sometimes sizes is too big */
 
 #define US_FLAG(name, value)	US_FL_##name = value ,
 enum { US_DO_ALL_FLAGS };
--- a/include/scsi/scsi_device.h	2007-02-06 14:14:57.000000000 +0100
+++ b/include/scsi/scsi_device.h	2007-02-07 15:14:01.000000000 +0100
@@ -122,6 +122,7 @@
 	unsigned no_uld_attach:1; /* disable connecting to upper level drivers */
 	unsigned select_no_atn:1;
 	unsigned fix_capacity:1;	/* READ_CAPACITY is too high by 1 */
+	unsigned guess_capacity:1;	/* READ_CAPACITY might be too high by 1 */
 	unsigned retry_hwerror:1;	/* Retry HARDWARE_ERROR */
 
 	unsigned int device_blocked;	/* Device returned QUEUE_FULL. */

Re: [PATCH]indistinguishable devices with broken and unbroken firmware

[Christoph Hellwig]

On Wed, Feb 07, 2007 at 03:22:39PM +0100, Oliver Neukum wrote:
> Hi,
> 
> there's a USB mass storage device which exists in two version. One
> reports the correct size and the other does not. Apart from that they
> are identical and cannot be told apart. Here's a heuristic based on the
> empirical finding that drives have even sizes. I know it's ugly, but it fixes
> an ugly bug.

Do we really need a second flag?  Alternatively fix_capacity could be
changed to mean fix size if odd.

Re: [PATCH]indistinguishable devices with broken and unbroken firmware

[Oliver Neukum]

Am Mittwoch, 7. Februar 2007 16:05 schrieb Christoph Hellwig:
> On Wed, Feb 07, 2007 at 03:22:39PM +0100, Oliver Neukum wrote:
> > Hi,
> > 
> > there's a USB mass storage device which exists in two version. One
> > reports the correct size and the other does not. Apart from that they
> > are identical and cannot be told apart. Here's a heuristic based on the
> > empirical finding that drives have even sizes. I know it's ugly, but it fixes
> > an ugly bug.
> 
> Do we really need a second flag?  Alternatively fix_capacity could be
> changed to mean fix size if odd.

Until we learn about devices which really do have an odd number of sectors.
And then we'll have a problem. I'd like to avoid heuristics when we know
for sure.

	Regards
		Oliver

Re: [usb-storage] [PATCH]indistinguishable devices with broken and unbroken firmware

[Alan Stern]

On Wed, 7 Feb 2007, Oliver Neukum wrote:

> --- a/drivers/scsi/sd.c	2007-02-06 14:14:48.000000000 +0100
> +++ b/drivers/scsi/sd.c	2007-02-07 15:14:01.000000000 +0100
> @@ -1273,6 +1273,13 @@
>  	if (sdp->fix_capacity)
>  		--sdkp->capacity;
>  
> +	/* Some devices have version which report the correct sizes
> +	 * and others which do not. We guess size according to a heuristic
> +	 * and err on the side of lowering the capacity. */
> +	if (sdp->guess_capacity)
> +		if (sdkp->capacity & 0x01) /* odd sizes are odd */
> +			--sdkp->capacity;

Please make this "else if ...".  Just in case somebody ever manages to set 
both sdp->fix_capacity and sdp->guess_capacity.

Otherwise it looks reasonable to me.

Alan Stern

Re: [usb-storage] [PATCH]indistinguishable devices with broken and unbroken firmware

[Alan Stern]

On Wed, 7 Feb 2007, Oliver Neukum wrote:

> Am Mittwoch, 7. Februar 2007 16:05 schrieb Christoph Hellwig:
> > On Wed, Feb 07, 2007 at 03:22:39PM +0100, Oliver Neukum wrote:
> > > Hi,
> > > 
> > > there's a USB mass storage device which exists in two version. One
> > > reports the correct size and the other does not. Apart from that they
> > > are identical and cannot be told apart. Here's a heuristic based on the
> > > empirical finding that drives have even sizes. I know it's ugly, but it fixes
> > > an ugly bug.
> > 
> > Do we really need a second flag?  Alternatively fix_capacity could be
> > changed to mean fix size if odd.
> 
> Until we learn about devices which really do have an odd number of sectors.
> And then we'll have a problem. I'd like to avoid heuristics when we know
> for sure.

In fact there are devices which have an odd number of sectors.  I once 
thought that would be a good heuristic to apply to all devices, but it 
won't work.

Alan Stern