BEWARE BUG - Disk corruption in Mac OS X 10.4.2 Tiger!

The comments on that site seem to indicate that the problem is with firmware on the Firewire drive, since one of the four posters wasn't experiencing the problem, and another was able to correct the problem by updating his drive's firmware. But the bug has been submitted to Apple, and if they find it is a problem with the OS, they'll fix it.

Oh, and another poster confirmed that the same behavior occurs under Panther.

How can I update the firmware on an OWC Mercury Elite FW800/400 drive? I haven't found any upgrades out there.

here are all the files and a good tutorial on doing it:

http://fwdepot.com/firewiredepot/firmware/firmware.html

Well, it's not the firmware. When I was experiencing the problem I had already upgraded the firmware b/c I was aware of the firmware issue. In fact, it took me a while to notice this problem because there's no error message or anything. It's just that the files contents are corrupted - but not all the time.

I noticed it first when I was converting raw images from my digital camera to 16bit TIFF. A batch of 40 images resulted in about half of them being corrupted.

I also tried lowering the blocksize, lower UDMA modes via the oxford utility.
While they have a negative impact on the drive throughput for me it seems to impact how often data is corrupted. Doing exactly the same test as before I noticed that less files were corrupted. At UDMA 4 and 128K blocksize on the oxford It's about 40%.

-torsten

I agree its not firmware. My drive has an Initio (? - not Oxford anyway) chipset and there are no upgrades available. It does not only corrupt files but possibly the filesystem too - depends if the bad data is application data or filesystem data.

Happened very often with FW800, less often with FW400.

10.3 & 10.4.

the Initio firmware utility and files are also available at

http://www.fwdepot.com/firewiredepot/initio216/Initio2430217.zip

"It's not fixed with oxford 912 firmware 1.02" would have been more correct. I don't know for sure if it's a apple or oxford problem. However, I got around doing more testing - with "iozone".

The subject is a Maxtor L300R0 on a oxford 912 with 1.02 firmware. The bridge has two 1394b ports, one 1394a port and one USB 2.0 port. Hosts are a PowerMac G5 2.3ghz and a PowerBook G4 15" 1.67ghz (as of this writing both are still the "current" model), both running 10.4.2.

All tests are done using 'iozone -w -i 0 -i 1 -r 16m -s 4g -Va'

Tests using the FW400/1394a port often resulted in hangs, ie, the drive is unresponsive. Processes accessing the filesystem it are stuck and not interruptable (including a SIGKILL). When they don't hang, written data is corrupted.

Tests using the FW800/1394b port for me return verification errors from iozone in about 9 out of 10 times I run iozone. It happens on both the Powerbook as well as the PowerMac. Example of iozone's error message:
Error in file: Position 331776 0 0
Error in file: Position 5664538624
Record # 337 Record size 16384 kb
Found pattern: Char >><< Expecting >>¥<<

Now today I also got around to hook up the drive to a FW400/1394a port on a FreeBSD machine (5.4-stable, Dell Optiplex machine with a older Firewire Card using the TI TSB12LV26 chipset). Filesystem was ufs2 (2k fragments, 16k blocks). Running iozone using the same parameters about 30 times did not show any pattern mismatches nor did the drive hang.

Just added another quick test. I still have a firewire 400 cardbus card that I used on a x86 notebook with FreeBSD before I switched and got the PowerBook.

I gave it a try and 10.4.2 sees the OHCI compat. card. Devices are recognized too. However, when performing the same tests I did with the native FW ports in my PowerBook as well as the PowerMac the result was the same too: the drive froze.

-tb

Torsten

are you certain its a 912 chip? did you get this information from the Oxford utility?

the reason I ask is that you mention there are 2 x 1394b ports and 1 x USB port, so its either a 912 chip with a separate USB chip, or its a 922 chip with built in USB

Yes, got that from the oxford utility. It got in fact two 1394b, one 1394a and one USB2.0 port. It's a Sarotech Hardbox case.

I did check the bridge itself. The printing on the board says "FHD-353UFb(FW912)REV1" (looks like the Sarotech product name). On the chip on the top there's a quality check stamp stating "V3.7 OK" (wonder if that's the initial firmware revision; if so, can't be a 912 as 1.02 seems to be the newest release). Just removed the stamp and it says
Oxford Semiconductor
C0351
KT201

On the bottom there are two chips. The TI phy TSB81BA3 driving the FW ports and one Profilic PL2507 handling the USB2.0 port.

-tb

ahhh, good old Sarotech....

ok, yes, thats a 912, and the latest firmware rev is 1.02 dated Jan 13, 2005

I see you experimented with different UDMA settings

did you also change the Maximum Transfer Rate? Is it set to 4096 or 2048?

Interesting. The oxford utility says:
"Currently Running 12:20:47 Sep 16 2004 (v 01.02) firmware."

I did test various transfer rates (I believe it's more like a transfer/block size on the interface level?) after stumbling across http://www.bustrace.com/delayedwrite/

I get the same problem with 4096 or 2048 and 1024. But the more I decrease the maxium transfer size, the less often it happens. Same applies to the UDMA setting.

sigh
I need fast but reliable external storage. I'm currently waiting with ordering a new case b/c of those issues.

-tb

PS: Anyone knows a external metal case for 1-2 disks, with good cooling for fast disks, FW800, USB2.0 and eSATA (for two disks, I know oxford's 924 supports just one disk if eSATA is used for host connections) as well as a 110/230V internal power supply ?

I think LaCie might have one with all those ports

I used the following myself:

http://fwdepot.com/thestore/productinfo.php/productsid/714

and

http://fwdepot.com/thestore/productinfo.php/productsid/1066

Well, couldn't find one at LaCie's Webserver that has FW800, USB2.0 and eSATA. Even if they had something, I wouldn't buy it as they don't sell cases only. I want to choose the disk in there myself and after having read a few things on the net I'm also a bit skeptical about their post sales support.

I know about the other two that you posted. Almost bought the FireXpress (despite the lack of eSATA and a internal power supply) but didn't pursue it as I started to see corrupted files on my existing OX912 external disk.

To me it seems to be a driver issue in the darwin kernel. On OSX 10.4.2 using the FW400 port the communication between the disk and the computer is frozen every time I do the test. If I do that on a FreeBSD machine with FW400 it works perfectly fine. Unfortunately I don't have a FW800 card in that machine so I can test if the corruption I see on my two macs happens there.

-tb