eSATA writing causes hard lockups

absoluteflatness

So, I have myself one of the much-maligned Seagate FreeAgent Pro external drives, which can connect by either USB or eSATA. I originally had it connected with USB, which I've got working fine, but I wanted it connected with the faster eSATA interface.

My early uses of the drive connected with eSATA (through an eSATA-SATA bracket connected to nvidia sata headers) were positive. Then soon after, I started getting hard lockups, with the entire system going completely unresponsive. I eventually realized that this always happened while I was doing a fairly large write to the eSATA drive. The first time, I was tar-ing up my home folder onto the drive, and the next several times, I was ripping a CD to the drive.

So, to sum up: the USB interface works fine on both read and write, and eSATA works fine for reading, and locks up the system on sizable writes. Anyone have any success with eSATA? The culprit could easily be the bracket, the drive, or Linux. I don't have another eSATA-ified system to test with, and the log files show no errors.

I'm pretty stumped on this one.

schachti · Posted: Tue Jan 29, 2008 11:45 am Post subject:

You could try if it works using a 2.6.24 kernel. In 2.6.24, a per device dirty threshold is used, maybe it fixes your problem.

I refer to this patch: http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=04fbfdc14e5f48463820d6b9807daa5e9c92c51f. I'm not sure whether it applies only to the two examples mentioned there, so just try...
_________________
Never argue with an idiot. He brings you down to his level, then beats you with experience.

How-To: Daten verschlüsselt auf DVD speichern.

Akkara · Posted: Tue Jan 29, 2008 2:01 pm Post subject:

It *might* be due to slightly different electrical specs between sata and esata.

I had a similar problem when I had hooked up a esata drive to a internal sata connector on a older mobo, using a sata to esata cable. The machine ran for slightly longer when I tried using a shorter cable. The problem went away completely after I had upgraded the mobo to one that had a real esata connector.

I don't know if anyone makes a buffered sata-to-esata bracket but that might be something to try if the above-mentioned patch doesn't work for you.

absoluteflatness · Posted: Wed Jan 30, 2008 2:13 pm Post subject:

I had upgraded to 2.6.24, but was experiencing some other issues. I started getting unkillable running processes taking 100% cpu, (cdparanoia and kio_file to be specific). Rather than try to pin down the problem, I switched back to 2.6.23.

I'll switch back again and see if I've got any eSATA joy. I'm still entertaining the possibility that it's the bracket setup that's messing everything up (pretty annoying that a foot or so of cable could bork everything so effectively), but if that's the case, I don't really see myself getting an eSATA controller card. My gut tells me that if reads work, but just not writes, it shouldn't be an electrical problem, but these things are never so simple as one would like.

The main tangible benefit I had with eSATA is that the drive sleep/wakeup feature works 10 times better on eSATA than USB. The filesystem seen by linux liked to get messed up during the wakeup process (probably ntfs-3g's fault) over USB, so I always end up manually un-and-re-mounting. Worked seamlessly on eSATA.

absoluteflatness · Posted: Wed Feb 06, 2008 8:59 pm Post subject:

So, I've finally had time to test out the eSATA drive in kernel 2.6.24. Results: different, but not good.

Mounting through either hal or manually fails, and throws out some unhappy looking segfaults from the kernel.

absoluteflatness · Posted: Thu Feb 07, 2008 8:11 am Post subject:

So, I fixed the segfault from above, I had an old sys-fs/fuse module laying around. I had deleted it from my 2.6.23 modules dir, but forgot to do it for 2.6.24. Unfortunately, that did nothing to fix the "total lockup" problem from before. This time, I was deliberately not using the drive at all, and the system locked up completely. Once again, nothing of note in the syslog, the last thing to get logged before the freeze was the periodical cron check.

So, am I doomed on eSATA here? I'd really rather use it, but does that mean the bracket has to go? I'll look into prices on eSATA controllers, but unless they're pretty cheap, I probably won't be getting one anytime soon. Any suggestions for other options to try, or specific good controller chips to look for?