Hi,

I have set up 3 systems with RH7.1, but I decided to build a new, faster computer, and every time I try to install RH7.1, it hangs while trying to format partitions. I originally thought it was a problem with the HighPoint HPT370 on the motherboard, but I have ruled that out. I can't even install when the drive is on the regular IDE bus.

Here is the hardware configuration:

Epox 8KTA3+ (w/ HTP370)
BIOS: 09/06 (Latest) w/ 2.1 HPT370 (not in use)
Processor: 1.4GHz Athlon, 266FSB, AYHJA core
RAM: 1GB
Drive: Quantum FireBall Plus AS40 40Gb ATA100 (primary master)
Video Card: Riva TNT2 AGP 4x (SIIG)
NIC: NetGear FA310Tx
CDR: Creative Blaster 8-4-32x (secondary master)
Floppy Drive
No Audio

I don't appear to be having any problem with the initial stages of the installer or setting up the partition map, etc. The partition map is approximately as follows:

/boot - 100M
/swap - 2047M
/var - 5000M
/home - 4000M
/usr - 4000M
/ - 3000M

When I reach the end of the RedHat 7.1 Installer and it starts the actual install, it hangs while trying to format the root filesystem.

I have tried many different combinations, disabled all non-essential I/O. Set everything in the BIOS to slow mode, like IDE Mode 0, for example, just to see if I could get it to work, and it hangs up every time - the old lockarooney...mouse frozen, etc.

------------
AHA!
------------
OK...I just typed all that and then I thought...hmmmm I wonder if it could be the RAM or the kernel command...so I ripped out all but DIMM 1 and added the hda=ide-scsi to the command line in the LILO setup screen, and the @#*@&#^$#(*@ thing is installing now! So, maybe that's why it was locking up with the HPT370 also - I have to figure out if it was bad RAM or the ide-scsi command. The bad RAM would explain the lock-up, but so would a failed IDE transaction or unhandled interrupt, etc.

I guess I'll leave the info above in case anyone has a similar problem. I'll also follow-up here if I am able to get the 8KTA3+ and the HPT370 to work with a mirror raid (I have 2 of the quantum drives, but went to 1 for troubleshooting purposes).

I hope it works!!!!!!!! I guess I'll know by tomorrow. If it works, I'll post the BIOS hack I have for the 2.1 HPT370 for the 8KTA3+.

LinuxElf

So far, so good...

I have tried installing using one of the AS40's on the HPT370 interface and it works! I followed the patch instructions that I got from highpoint-tech.com - you have to make sure you do the RH install in text mode so that you can enter the shell at the end to run the 370PostInstall script.

When I reboot, the 370 startup config is there, and linux shows that it has correctly detected the RAID controller. Pretty ! I'm about to debug the DIMMs to see if it's a bad DIMM or bad socket, then re-establish the RAID and try 2 drives. I hope it's not a socket problem because I'm already on MB #2. The first 8KTA3+ motherboard I had had a C1 error at POST, so I had to swap it. Then, the replacement board had a "locked" BIOS due to a manufacturing error on a run of 8KTA3+ boards, so I had to wait for a replacement BIOS. Grrr... *does a little ju-ju dance to ward off evil spirits*

One thing to remember: Whenever you do an expert install, don't forget to add the root=/dev/xxx command to your kernel line - otherwise you won't be able to boot...hee hee

I STILL have not yet begun to install!

LinuxElf

Well...I've run a bunch of tests:

1. Tested all DIMMs in socket 1 - NO PROBLEMS
2. Tested 256, 512, 768 (up to 3 sockets filled) - NO PROBLEMS
3. Tested 1Gb in PC100 mode w/ interleave off - NO PROBLEMS (but processor locked at 1GHz because of RAM timing)

What does fail is all 4 sockets populated in PC133 mode, even with the "fail safe" BIOS parameters set (or any for that matter - I tried different manual settings). During this testing, I realized that I am seeing signal 11 errors, which are failed RAM accesses. After swapping the RAM around, the kernel reported back instead of just freezing the machine.

Based on the tests, it does not appear that any individual DIMM is at fault, or that I have a bad socket (PC100 test). It's just that when all 4 DIMM sockets are populated, whatever the board is doing when it runs the timing/interleave based on the RAM spd info appears to be out of spec for my crucial PC133 RAM (too fast). As above, I played around with the BIOS settings in PC133 mode, and nothing seemed to make the problem go away, except to step down to PC100. For some of the RAM tests, I used memtest-86 v2.7 to verify that the RAM was working - it takes over 2 hours to complete a full sweep but seems like a great utility for testing RAM.

This morning, I called Crucial and talked it over with them. I currently have the only single-sided 256Mb PC133 DIMM they supply, which is Part No. CT32M64S4D75.8T. They don't have anything faster, unfortunately. I will also add that they have great customer service - I really like them. They initially sent me double-sided RAM and swapped it with no problem.

I also called Epox and they are checking with Taiwan for me about using 4 256Mb DIMMs. Their manual is somewhat ambiguous, although they have said that I would have to use 2 512's if I wanted 1Gb. My PC100 mode test implies that 1Gb should work if the timing is correct (and is also fastest - best interleave). They are checking to see if it is fixable with a BIOS update or if they can find out any other info. Epox support is also friendly and helpful. (BTW, I have definitely made them aware of the need for an upgraded BIOS to include the HPT370 2.1 BIOS for the linux RAID driver).

I also called Mushkin (mushkin.com). The only single-sided PC133 RAM they make is also CAS=3 (similar to Crucial) - the "Sammy Special." They are going to test it out with 1Gb in their 8KTA3+ board. All of their higher performance RAM is double-sided, which won't work in my shared DIMM sockets (3 & 4). For now, I'm not going to pursue that path until I know what Epox says.

I've also considered using registered memory, but I'll check on that after I hear back from Epox about basic support for the configuration.

The saga continues...

BTW, I did try installing RH7.1 using my updated BIOS with the HPT2.1 stuff and it worked with 1 drive installed - haven't tried the mirror set yet.

oops...attachments aren't supported here...can't post the BIOS.

More to follow...

Hmmm...sorry about the paragraphs in that last post...looks like they got munched.

Well...I set up RAID yesterday with each Quantum FireBall Plus ATA100 AS40 40Gb drive on its own HPT370 UDMA channel as a master. (I'm sure it would work as master/slave on the same channel, but I thought master/master on dedicated channels would yield better performance.)

I used the HPT370 boot and driver disks from the driver package at the highpoint site and did the post install at the end. NOTE: you MUST do a text install to do this...but it's no big deal... Everything went smoothly.

The RAID 1 set is working great, I have a 370 boot configuration at startup, and RH7.1 detects the RAID controller and runs the drives in the native mode - I assume ATA/100.

I installed on the first drive and then created the RAID, and it worked fine. The new HPT370 2.1 BIOS has some new features which make it a tiny bit easier to use.

I'm running the machine with 768Mb pending the results of the answer from Epox in Taiwan, and it's running like a champ.

When I find out the result of my 1Gb quest, I'll post it. Other than that, I hope this thread is helpful to someone who is trying to get RH7.1 working with RAID on the Epox 8KTA3+ board. It works!

I just noticed last night that RH7.2 is out. After all my struggles for 7.1, I think I'll wait until the next driver update. 7.1 works fine for now and is stable...and their site is choked with ftp action. 1K/second for the update? NOT.

Hopefully, Epox will provide an official BIOS update with the HPT2.1 code so that it's not necessary to make the "home brew" version...

Happy tweaking!

LinuxElf

ps. I still have to install all the patches, so technically, "I have only just begun to install!"

BTW - if anyone wants my 8KTA3+ 09/06/01 + HPT370 2.1 BIOS hybrid to upload to your motherboard, just email me. You can use it at your own risk...all I guarantee is that it programmed fine and works on my system. It was created from the shipping 09/06 BIOS on the Epox site and the latest HPT370 BIOS from the highpoint-tech.com site included with the RH driver package.

LinuxElf

Linux_Elf: If you could post a link to the download site for that BIOS update that would be great.

I see the following non-fatal error below when I boot. Everything seems to start up OK, but does anyone here know why I am seeing the -16 and -2 errors? Is it some weird kernel bug, or something that I can fix by editing modules.conf, etc.? How can I get rid of the "trying to unmount root" error?

Thanks for the help.

LinuxElf

ps. As far as I know, I'm the only one who has the hybrid bios for the 8KTA3+ - I'd be happy to send it over for you to post here. It's not on the web.

-----------------------------------------------

Vendor: QUANTUM Model: FIREBALLP AS Rev: A1Y.
Type: Direct-Access ANSI SCSI revision: 02
Vendor: QUANTUM Model: FIREBALLP AS Rev: A1Y.
Type: Direct-Access ANSI SCSI revision: 02
Attached scsi disk sda at scsi0, channel 2, id 0, lun 0
Attached scsi disk sdb at scsi0, channel 2, id 1, lun 0
SCSI device sda: 78177792 512-byte hdwr sectors (40027 MB)
Partition check:
sda: sda1 sda2 < sda5 sda6 sda7 sda8 sda9 sda10 sda11 >
SCSI device sdb: 78177792 512-byte hdwr sectors (40027 MB)
sdb:
VFS: Mounted root (ext2 filesystem) readonly.
change_root: old root has d_count=6
Trying to unmount old root ... <3>error -16
Change root to /initrd: error -2
Freeing unused kernel memory: 236k freed

Anyone have any ideas on my question in the last post?

Also - the machine is completely installed and updated now and running great ...except for the startup error...

I'd appreciate any suggestions about the startup errors (-2 & -16) and how to get rid of them...

Hi,

I thought I would post the final results of my RAM config research for the 8KTA3+, in case anyone is interested. After learning that I was not going to be able to run 1Gb as 4 256Mb DIMMs, I have returned them and am now using 3 512Mb DIMMs for a total of 1.5Gb.

The 8KTA3+ supports 6 banks of RAM, and the 512's I got are all dual-banked, which means that I am maximizing the interleave as well as the physical RAM size. So, with 3 dual-bank 512's, I am using all 6 banks.

The specific 512Mb DIMMs I got from Crucial are CT64M64S4D75.16LT

I have just starting running a boot-level MemTest on them and will probably let it run until sometime tomorrow just to make sure everything is OK, but they seem to be fine.

I just want to say again that I am extremely impressed with Crucial's customer service and they have been absolutely great in helping me to resolve these problems across every department - tech support, RMA, shipping, etc. They send out great order and RMA status emails so that I always know what's up with the status, and any email I send back automatically gets appended to my record so anyone who checks can see it. Anyway - they're great and I highy recommend them. I've used their parts for some years and have always been a satisfied customer.

LinuxElf

Hi,

I wrote to EPoX in Taiwan asking about the 4-DIMM thing, and they sent me the KT133 spec back in return which says:

− Supports PC133 and PC100 SDRAM and Virtual Channel Memory (VCM) SDRAM up to 3 DIMMs

So, it's a little perplexing that they would design a motherboard with 4 sockets that has a controller that only supports 3 DIMMs.

The 512Mb DIMMs are working great! I used MemTest to run a standard test, which passed (I think it was about 5 hrs.), and for the past 35 hrs., it has run a complete, extended test on the entire 1.5Gb (1 successful pass!).

I have noticed that EPoX has released a new BIOS (October) for the 8KTA3+, which still does not include the latest HPT370 BIOS, so I asked Taiwan about that also. They said that the 2.1 BIOS has a problem, so I followed up with both HighPoint and EPoX to ask what's wrong and when they're going to fix it.

So, far, 2.1 seems to be working fine on my system, but there is a note in the release notes about SMP not working...maybe that's it...

LinuxElf

Hi
After several days of grief, I seem to have resolved my own install problems by reducing the fsb speed in BIOS to 100 (200), instead of 266 (133). I am eagerly looking forward to the arrival of the customer service team at Watford electronics where I got the CPU.
Jim

Linuxelf
I'm seeing the same errors now that my machine actually works. Did you figure it out?? I wonder if its something to do with ramdisks?????
Jim