Linux drivers for Datafab-based and Lexar Jumpshot CompactFlash readers

Change Log:

11 Oct 2000: Initial release
23 Nov 2000: Updated for test11 kernel.

Now in patch format. Thanks to Thomas Eschenbacher.
Now correctly interprets the IDENTIFY DEVICE results. That means we don't have to muck with the MBR to compute media size now.

19 Dec 2000: I've been busy lately and haven't kept up with the changes being made to the USB repository. Unfortunately those changes will make the patch break. So I've modified the patch so that it applies directly to a vanilla 2.4.0-test11 kernel.
14 Jan 2001: Fixed a bug that can cause the Datafab driver to fail to initialize. I've updated the patch so that it can be applied directly to Linux 2.4.0. Linus has changed his policy of accepting a flood of new-function patches following a .0 kernel release so it might be a while before these are officially integrated into the main kernel tree. Until then, I'll try to keep these patches up to date.
17 Jan 2001: Feh. The 2.4.0 patch was apparently kind of broken. That's what I get for trying to add a couple lines to the patch directly rather than simply rebuilding the patch. Hopefully it's fixed now.
1 Feb 2001: Hopefully added support for devices with IDs ranging from 0xA003 through 0xA006. Consider support for these devices "beta" at best.
10 Feb 2001: Created a patch against 2.4.1-ac2 with the drivers/usb tree replaced with what's currently in the Linux-USB repository. This patch itself will probably apply directly to a 2.4.1 kernel but I've not tried it. I submitted this patch for inclusion in the kernel usb tree. Woohoo! So this might be the last patch I'll place on this website...
27 May 2001: Created a patch against 2.4.5 since it appears that Config.in fails to patch...
19 July 2001: Continue to get reports of both successes and failures with various incarnations of Datafab-based readers. The Linux-USB webpage's supported hardware list is useful for most hardware but doesn't seem well-suited for hardware that has as many revisions as the Datafab chipset so I've added a hardware list here.
Late update. These drivers became part of the main kernel tree somewhere around 2.4.10 or 2.4.11 but the drivers/usb/Config.in file wasn't updated to reflect this. So if you're trying to use one of these kernels, you can take a look at the 2.4.5 patch below and make the Config.in mods manually or you can try applying the patch. It'll make the appropriate mods to Config.in (by luck, it's the first file patched). Patch will then complain about previously applied patches for the other files at which point you can break out using control-C.

This patch actually contains two drivers that enable Lexar "Jumpshot" CompactFlash readers and a number of readers based on Datafab chipsets to be used with Linux. These devices implement variations of the ATA command set. We, in fact, treat them as SCSI devices so there is some degree of IDE-SCSI translation occuring within the driver. This means that you'll need to build SCSI disk support into your kernel or as a module. Currently, only a single LUN is supported so if your build your SCSI layer to probe all SCSI LUNs, you might see timeouts/delays when you load the usb-storage module due to probing non-existant LUNs.

Before I go further, I want to empasize that these drivers are EXPERIMENTAL. While they appear to work fine in my tests for the two readers that I own, bugs may crop up. When creating these drivers, I was quite unable to get any sort of information from their respective companies so they were 'reverse engineered' by observing USB traffic on a Windows machine. If you use these drivers for mission-critical file storage, you do so at your own peril. Also, be aware that error-checking is still sparse. This means that if something fails, you'll get a generic error that's probably less than useful.

Datafab-based Readers

Datafab is an interesting story. Not only do they manufacture readers but they're apparently an OEM provider of USB-ATA chipsets (though recently it was pointed out to me that these chips might actually be rebranded OnSpec or Shuttle chips). The driver was originally created for my reader (a Datafab MDCFE-B 0x07c4/0xa000). I've since received a mixed bag of reports of other brands of readers working (or not working) with this driver. Enough that it makes sense to maintain a list of hardware on this page based on user feedback.

A number of Datafab-based readers are so-called "combination" readers that are capable of handling CompactFlash and SmartMedia. In most cases, SmartMedia will not be functional with this driver. There are some exceptions noted below. Please be aware that I personally do not have any SmartMedia hardware so I have to rely on user feedback.

Also be aware that product name itself isn't a guarantee that the reader will be supported. Though the marketed name might be the same, it's quite possible that the underlying hardware is different. (witness the PNY Flashback readers). Unfortunately manufacturers tend not to print the device IDs on the packaging so you can't be sure what you bought until you plug it into your computer.

Manufacturer	Name	Device ID	Works?	Comments
Datafab	MDCFE-B	07c4:a000	Yes	This is the device the driver was originally created for
PNY	Flashback	07c4:a000	Yes	(1)
Integral	Integral CompactFlash USB	07c4:a000	Yes	(1)
Datafab	MDCFSM-B	07c4:a001	Yes	Reportedly device is sometimes unmountable (1)
PNY	Unknown CF+SM combo reader	07c4:a005	No	Last I heard, this device still was not working
Datafab	CF+SM combo reader	07c4:a005	Yes for CF; no for SM	Reportedly works with CF but not SM (1)
PNY	Unknown	07c4:a006	Yes	Reportedly works with SmartMedia (1)
Jenoptik	Unkown SmartMedia reader	07c4:a006	Yes	Reportedly works with SmartMedia (1)
Simple Technologies	CF Type I/II reader	(unknown)	Yes	Reported to work with IBM microdrive as well (1)
Datafab	MDSM-B-USB	07c4:a103	No	(1)
Apacer	AP-CSCRU10 CF+SM	07c4:a109	No	Last I heard, this device was not working (1)
Greenhouse	GH-CRSC CF+SM USB reader	07c4:a109	Yes	Only tested with CompactFlash (1)
Datafab	DSCFSM-USB	07c4:a109	Yes	Not tested with SmartMedia. Reportedly works for CompactFlash (1)
Datafab	USB to CF + SM Combo (LC1)	07c4:a109	Yes	Not tested with SmartMedia. Reportedly works with IBM microdrive. This might be the same as the DSCFSM-USB reader. (1)
PNY	Flashback	07c4:b000	No	There are apparently different versions of the Flashback. This one (b000) does not work (1)

Notes

(1) = feedback reported by a user
I'm sure I've missed a couple feedback reports
Lately, I've been getting reports of 'b000' devices in the wild. These appear to be quite different from previous readers in that they act like SCSI devices instead of ATA devices. Thus, this driver has no chance of working with these readers. In fact, the usb-storage module detects the reader as a SCSI device and attempts to use the built-in transparent SCSI routines to access the device instead of this driver.. Obviously there are some problems here since b000 readers are essentially non-functional under Linux. Since I don't have access to one of these readers, I'm afraid you're on your own if you have one. :-/

Lexar Jumpshot Reader

The Jumpshot is essentially a USB cable with one end modified to accept a "USB-enabled" CF card. All the actual USB circuits are supposedly onboard the CF card itself. This means that you cannot use a vanilla CompactFlash card with this reader. Still, it's a pretty neat design. So far as I can tell there is only one hardware revision so things are alot simpler than with the Datafab-based stuff.

Testing. I currently have 3 CF cards at my disposal: 8mb, 48mb and 64mb. I've verified that the Datafab reader works correctly with all three. I can only test the Lexar reader with the 48mb card since it's the only one that's compatible. I've tested both reading and writing and for grins, even created an EXT2 filesystem on one card with no noted problems.

TODO

Add support for non-x86 architecture. Make sure no alignment problems for SPARC.
Flesh out the SCSI error reporting.
Any more readers out there based on a Datafab chipset?
Anybody care to backport these to 2.2.x? :-)

Installation: 3 steps.

Extract the appropriate kernel source
Apply my patch using "patch -p1 < foo.diff"
When you configure the kernel, there will now be 2 additional options in the USB config menu for these drivers.

Usage:

Make sure there is a memory card in your reader
Load the SCSI disk modules if necessary
Load the usbcore and UHCI/OHCI modules if necessary
Load the usb-storage module. The SCSI layer will dump a message to the console (check dmesg) indicating which SCSI device it's using. We'll assume /dev/sda here.
Now, just mount the filesystem. For instance, mount -t msdos /dev/sda1 /mnt/reader

Contact: In addition to the email address at the bottom of this page, you can find me on IRC (server: haybaler.sackheads.org:6667 channel #usb).

Note: I've received some reports from users who say that their reader is being detected multiple times. Unfortunately I don't see this behavior with my readers. I haven't heard any reports of this being dangerous but if it annoys you greatly, you might be able to avoid it by adding US_FL_SINGLE_LUN to the flags section of your device entry in unusual_devs.h.

Grab the patch for 2.4.0 here.

Grab the patch for 2.4.1 here.

Grab the patch for 2.4.5 here.

Back to the software page

Questions or comments? Drop me a line.