From time to time, I get a burst of four error messages within 1 to 2 seconds, always the same problem and always on /dev/hda. Here's one such burst from my kernel.warnings log:

Mar 12 22:01:45 JKP200 kernel: hda: timeout waiting for DMA
Mar 12 22:01:45 JKP200 kernel: ide_dmaproc: chipset supported ide_dma_timeout func only: 14
Mar 12 22:01:45 JKP200 kernel: hda: status error: status=0x58 { DriveReady SeekComplete DataRequest }
Mar 12 22:01:45 JKP200 kernel: hda: drive not ready for command

This drive is a fairly new Maxtor 20-GB 7200 RPM unit, that's partitioned to all be in vfat format (it was originally the Win95 drive in a dual-boot setup but I subsequently dumped Windows entirely from that box; hdc is my Linux drive, another 20-GB but only 5400 RPM). I'm guessing that the rotation speed has something to do with it, and that the driver is simply missing a signal somewhere. So far as I can tell no data corruption is taking place, but I sometimes have to try twice to open a file instead of having success on the first try.

My question is simply how can I cure this situation? I've used hdparm to enable DMA and unmask the IRQs, to try to take full advantage of the drive's speed, but I was getting these same errors before I did that tweak. I'm also running SMART on both drives, and it's reporting "no errors" on both. Any ideas, anyone? Incidentally I'm familiar with device interfacing at the ATA register level, having worked for years with the plant that's now Seagate but which began as General Electric's Peripherals Department...

Well, I've never worked in a fab plant, so you probably know more about this than I do, but I'll tell you what I'd do if it were me...

1) Use the Maxtor PowerMax (http://www.maxtor.com/softwaredownload/default.htm) to check the drive. SMART isn't the best technology, as it sometimes does not report correctable drive errors (bad sectors that can be disabled, etc.). The software has a large array of tests and seems to be very complete. If the full scale test comes back clean you probably have a good drive.

2) If you're still running a FAT filesystem, you could try Steve Gibson's SpinRite utility (www.grc.com). I haven't used it in years, but apparently it's still a very useful program, and there's a lot of detailed information for you to read.

3) Now that that the diagnostics are done, try playing with hdparm. This program allows you to test and set all sorts of drive parameters. Like anything else, the disclaimer says it has the potential to corrupt data, but I've used it for a long time to enable UDMA (UDMA settings could be important in your case) on my drives and I've never had a problem. There's an article on some of it's uses here: http://linux.oreillynet.com/pub/a/li...9/hdparm.html.

So, if worse comes to worse and the drive is defective, you can count on Maxtor to replace it for free without any haggling. Otherwise, it's probably something that you can tweak with hdparm. I run a 20GB 7200 Maxtor on one of my systems without any special configurations (other than enabling UDMA with hdparm), and several other 7200 drives from various vendors in other systems.

I hope something here helps. Again, I'm not an expert in this particular field, so please excuse anything that may come off as trivial. Good luck...

-Joe

Thanks for the ideas! I've downloaded the PowerMax package per your link and will run it next time I take down the Linux box (which is my firewall-router-FTP server for my business; the hda drive is the file server). I already have SpinRite 5 and find it quite useful; hadn't thought of using it!

The situation could well be a bad sector; it could equally well be something in the vfat driver. I really should offload all the data I have on that drive and convert it over to ext2 or ext3 filesystem since I no longer run Windows at all on that machine. A third possibility could be that my CPU there (a Pentium 200 MMX dating from 1997 or so) isn't fast enough to keep up with things.

I've considered switching from the current DMA2 interface on this drive, to UDMA; "hdparm -i" reports that it has UDMA-66 capability. Since the error seems to be DMA-related that might take care of it all by itself!