SATA through Syba 3112-150R in Fedora 5

JohnLocke · 07-02-2006, 09:05 PM

Just got the newest distro of Fedora C5, and thought I'd try out the sata support they're supposed to have. I ran through several problems (asked on the newbie forum here: http://www.linuxquestions.org/questi...d.php?t=460078 )

In short, I ruled out raid as the problem. The last install I did I tried just to install the two 120Gb sata drives as part of the linux system (I even tried a few different mount points: /data1 and /data2, /mnt/sata1 /mnt/sata2 ... mount point didn't seem to matter). The install hung while trying to format the drives (in the beginning).

I also had that error when I left out the main portion of the drives from the install. I installed with 1Gb of each drive as raid 0 swap space, and that worked fine ... until I used fdisk and partitioned the rest of the space (once as ext3 and once as linux raid space). As soon as I tried to format the large volume, either the OS hung completely, or at the very least, the terminal window locked up and the process went rogue. I looked in the dmesg and got the following:

Code:

Jul  2 12:26:44 localhost kernel: ata2: command 0x35 timeout, stat 0xd9 host_stat 0x61
Jul  2 12:26:44 localhost kernel: ata2: status=0xd9 { Busy }
Jul  2 12:26:44 localhost kernel: sd 1:0:0:0: SCSI error: return code = 0x8000002
Jul  2 12:26:44 localhost kernel: sdb: Current: sense key: Aborted Command
Jul  2 12:26:44 localhost kernel:     Additional sense: Scsi parity error
Jul  2 12:26:44 localhost kernel: end_request: I/O error, dev sdb, sector 2749187
Jul  2 12:26:44 localhost kernel: ATA: abnormal status 0xD9 on port 0xE0832CC7

I did some research on that error code and came across a post in linuxforums:
http://www.linuxforums.org/forum/sla...-too-well.html

All in all, I'm led to the conclusion that there's either some kind large space allocation issue (doubtful, I even tried it as LVM spaces), or, more likely, some kind of driver issue.

Since I'm certainly not an expert, I don't know where to go from here.

Best solution, if I can find one, is one that allows me to install the main files on my IDE drive (20Gb ... having / and /boot on it), my 1-2 Gb of swap as raid 0 on /dev/sda and /dev/sdb, and the remainder as a raid 1 on /dev/sda and /dev/sdb.

Even better solution is if I can assume I can do all this without having to reinstall again ... I'm going to install what I currently have only on the IDE drive (since the install will hang on format of sda and sdb anyway), and hope that I can use fdsik and mdadm to get my raid on.

Anyone seen these issues before or know of a solution?

ErrorBound · 07-04-2006, 10:08 PM

Quote:

Originally Posted by JohnLocke

... until I used fdisk and partitioned the rest of the space (once as ext3 and once as linux raid space). As soon as I tried to format the large volume, either the OS hung completely, or at the very least, the terminal window locked up and the process went rogue.

Have you tried running this without X in a text login? I added an extra 300GB SATA drive not too long ago. Formatting large volumes may cause the OS to appear to hang when it's really just taking forever to format. I'd try doing a text-login, and going away and making a sandwich or whatever while fdisk is running.

JohnLocke · 07-04-2006, 10:17 PM

Quote:

Originally Posted by ErrorBound

Have you tried running this without X in a text login? I added an extra 300GB SATA drive not too long ago. Formatting large volumes may cause the OS to appear to hang when it's really just taking forever to format. I'd try doing a text-login, and going away and making a sandwich or whatever while fdisk is running.

I thought about that and left the format running for about 3 hours (just in case ... even though my 40Gb format only took about 10 minutes on the IDE drive).

That's when I was seeing the above errors in dmesg (and always at the same address, which made me think it was most certainly hung in just one spot).

I'm hearing other people having similar issues and needing to recompile the kernel. For instance:

Quote:

ok, i recompiled the kernel with both nv_sata and SiI sata. powered
down and swapped SATA drives from the nv ports to the sil3512 ports.
rebooted.

success!

from:

http://www.uwsg.iu.edu/hypermail/lin...06.2/1010.html

Of course, being somewhat new at this, I have no idea how to even /check/ which driver it's using, and /really/ have no idea how to recompile the kernel (especially as I have only one linux box capable of doing that ATM, and that's the one what would need the new kernel).

Getting rather frustrated with all this

. It seems it's on the verge of working each time and then ...

(nothing)

ErrorBound · 07-04-2006, 11:30 PM

If you have a Knoppix CD then I'd suggest trying that. I don't have much experience with fedora and I don't know how well it supports things, but knoppix seems to have excellent hardware support. I've heard of some people using live CDs to repair their systems, but I've only ever got them to work read-only. However you might just stick it in and have everything magically work, which would narrow down your problem to be surely in the Fedora kernel or drivers.

You could also check out your drivers, and whether your current kernel was compiled with SATA support, both in your kernel configuration file, which is usually located under /boot.

JohnLocke · 07-05-2006, 12:06 AM

I just tried with the latest version of knoppix, same thing. I get to (roughly) inode 125 (it's always between 100 and 150 so far) of 896, and then the mke2fs hangs ... locking the entire system.

I /think/ I upgraded the controller drivers (ran their "update drivers" script for linux 2.6+ and it appeared to be doing something). I also checked the config file (I have two for some reason ... maybe I'm supposed to ... it has 2.6.15.xxxx and 2.6.17.xxxx) and I don't really know what I'm looking for. There's certainly a bunch of stuff in there about SATA and SCSI that's enabled. There's also a lot set to "m" instead of "y", and that's not a font error where it's supposed to be a lowercase "n".

I've heard a lot about people switching their drivers from or to the nvidia and sil drivers, but I have zero idea what they're talking about or how to even try it

.

ErrorBound · 07-05-2006, 09:10 AM

Out of curiosity, what is the brand of the SATA drives? Are they new?

The "m"s in your kernel config file mean that those options were included as modules when the kernel was compiled. It's normal to have more than one file, you'll probably find that you have more than one kernel in there as well, of varying versions (vmlinuz-versionnumber).

JohnLocke · 07-05-2006, 06:30 PM

They're Western Digital ... which I understood to be very compatable (my IDE drive is also Western). They're WD1200JS drives.

Is there another stage of debug I can go through at least to figure whether it's the card, driver, or drive?

ErrorBound · 07-06-2006, 09:06 AM

You could surely narrow down your problem to either drive or card if you had access to another (different) drive that you could try. Other than that, I've never had to switch the drivers and so I'm not sure how it is done.

JohnLocke · 07-06-2006, 11:50 PM

Ok ... I just ran across a patch:

http://www.kernel.org/pub/linux/kern...id-2.6.17-rc4/

But I have no idea how to install this (nor do I have a great idea whether it's right, but the verbage in there looks promising. I'm on 2.6.17-1.2145_FC5, btw.

Anyone know how to install this patch? I somewhat assume it means recompiling the kernel with this information added, but I have no idea how to do that (though I'm not terribly scared to try), especially how to recompile and install the kernel on the very box that has the problem.

As to the status of the drives (whether they work or not), I just formatted and installed one in my windows XP box, so the drives themselves are fine. Also, since the drivers (from what I can guess) in the linux box look like they specifically call out WD drives, but I don't see the Silicon Image card anywhere, I've got to think it's problems with the drivers for the card, or some such thing like that.

ErrorBound · 07-08-2006, 01:32 AM

Well you can head over to http://www.kernel.org and download the source for the kernel version you want. It's probably a good idea to build a kernel of the same version that you already have, so you can copy over the config file. Unzip the source and save it to /usr/src. Copy the patch to the source directory and apply it using the patch command (might have to look up the man pages for use). Then you should look up some distribution-specific instructions for compiling and installing your kernel. I've only done it before under Debian, where it is very easy to make and install a .deb kernel image, but the procedure will be different under Fedora. An easy way to configure is to configure just the options for your patch using "make menuconfig" and then copying over the rest of your settings from your (working) kernel config file (in /boot) into your new configuration file (/usr/src/linux-version/.config).