I have finally got by RedHat 9.0 system to reliably boot from a drive on the HPT374 controller, due mainly in part to information I gathered at this site. Here's some information that may be helpful to others:

1) The only way to upgrade the HPT374 BIOS is to do a BIOS upgrade on the IT7 mobo. A new IT7 BIOS doesn't necessarily have a newer version of HPT374 BIOS, but the only way to upgrade the HPT374 BIOS is to install an IT7 BIOS upgrade that includes an HPT374 BIOS upgrade. I don't remember the version numbers, but there were a few versions available at www.abit.com that were newer than what my mobo had. I downloaded the lastest one.

2) On my particular system, the ONLY way I could boot from a drive on the HPT374 using the v1.2 (standard) driver that comes in binary form, was to shut the system down, wait 15-20 minutes, then boot. Presuambly something needs to cool down before it will boot..? By watching the kernel load, I could tell when it was going to fail, because the hpt.o driver would try to talk to the hpt374, even though I had set all 8 of its drives to "noprobe" in grub.conf.

3) If you send an email to Highpoint Tech Support, they'll tell you where to get a beta version of the v2.10 open source driver. It comes with instructions on how to compile it.

Because of the problem with the wrong driver talking to the HPT374, I ended up creating a custom kernel where I disabled the HPT 3xx IDE drivers (there's two of them) and disabled all IDE RAID controller support. I then copied the hpt374.o driver that I compiled from the v2.10 source into the /lib/modules tree and created a new initrd that included this driver.

My system reliably boots without the need to power down and cool down, and the problem with the hpt_wt process being in the wrong state, causing a +1 offset in the load average, is not present.

YAHOO!!

With a little encouragement, I may be coerced into writing a mini-HOWTO...

Greg

From dmesg:

RocketRAID 154x/1640 S-ATA Controller driver
Version 2.1, Compiled Sep 26 2003 13:55:27
scsi0 : hpt374
Vendor: WDC WD80 Model: 0JB-00CRA1 Rev: 17.0
Type: Direct-Access ANSI SCSI revision: 00
Vendor: WDC WD80 Model: 0JB-00CRA1 Rev: 17.0
Type: Direct-Access ANSI SCSI revision: 00
Vendor: Maxtor 4 Model: D040H2 Rev: DAH0
Type: Direct-Access ANSI SCSI revision: 00
Attached scsi disk sda at scsi0, channel 0, id 0, lun 0
Attached scsi disk sdb at scsi0, channel 0, id 1, lun 0
Attached scsi disk sdc at scsi0, channel 0, id 2, lun 0
SCSI device sda: 156301477 512-byte hdwr sectors (80026 MB)
Partition check:
sda: sda1
SCSI device sdb: 156301477 512-byte hdwr sectors (80026 MB)
sdb: sdb1 sdb2 sdb3 sdb4 < sdb5 >
SCSI device sdc: 80043253 512-byte hdwr sectors (40982 MB)
sdc: sdc1
sdc1: <bsd: sdc5 sdc6 sdc7 sdc8 sdc9 >

After determining that software raid is significantly faster than the hpt's "hardware" raid, I went ahead and switched to linux's own hpt* drivers, which present the drives as IDE disks instead of SCSI disks, thus allowing me to remove the SCSI layer entirely.

This works fine.

However, a problem that I've seen, using any drivers, is that the 3rd drive fails for no apparent reason. Using highpoint's driver, the failed disk is sdc; using linux's driver, it's hdi. The failure isn't caused by a bad cable or bad disk. It occurs when I mkraid a new mirror and they are syncing up. It also seems to always happen a little bit after the sync is 50% complete although I'm not sure if that means anything.

Repeatedly running bonnie++ on the array doesn't trigger the problem .. I think it has something to do with heavy usage of dma transfers.

From the kernel:

Sep 23 18:32:42 franklin kernel: hdi: dma_timer_expiry: dma status == 0x25
Sep 23 18:32:52 franklin kernel: hdi: DMA interrupt recovery
Sep 23 18:32:52 franklin kernel: hdi: lost interrupt
Sep 23 18:32:52 franklin kernel: hdi: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Sep 23 18:32:52 franklin kernel: hdi: dma_intr: error=0x01 { AddrMarkNotFound }, LBAsect=224004573, high=13, low=5900765, sector=224004480


=========================
/proc/mdstat :

Personalities : [raid0] [raid1]
read_ahead 1024 sectors
md2 : active raid0 md1[1] md0[0]
390715392 blocks 512k chunks

md1 : active raid1 hdk1[1] hdi1[0]
195358208 blocks [2/2] [UU]

md0 : active raid1 hdg1[1] hde1[0]
195358208 blocks [2/2] [UU]

unused devices: <none>
==========================
And bonnie++ results:

Version 1.03 ------Sequential Output------ --Sequential Input- --Random-
-Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
franklin.imgen.b 3G 15922 92 38527 12 25900 10 16835 90 103144 20 529.6 1
------Sequential Create------ --------Random Create--------
-Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 4075 97 +++++ +++ +++++ +++ 4290 97 +++++ +++ 10690 98
==============

Also any ideas why the write speeds aren't faster? Shouldn't it be closer to 50MB/sec ?

I'm not using RAID arrays at the moment. When I built this system a year ago, the only OS I could get to work using RAID was WinXP. Since then, I gave up on RAID arrays and was just trying to get all OSes to boot from drives on the HPT374. WinXP and FreeBSD were easy, just give the driver when installing the OS. RedHat 9.0 required a custom kernel to get it to work reliably.

I can't get RedHat to boot from a drive on the HPT374 using the RedHat driver unless I shut my machine off and let it cool down for ~15 minutes. Otherwise the hpt driver RedHat provides talks to the drive and I get a kernel panic. Even though I have "hd?=noprobe" for hde through hdl as kernel boot parameters.

Some previous posts address the pointlessness (is that a word) of using these drivers for RAID. The CPU is doing all the work anyway, so why bother with extra layers when Linux suports RAID in the kernel. Of course, when you're booting three different OSes, that's not an option.

Greg

mdmbkr,

Which driver are you using to make the drives appear as IDE drives? The hpt36x/37x driver (or whatever it's called)? The only reason I'm going the SCSI route is because I can't get the thing to boot any other way. Perhaps I'll build a new kernel with just the above driver and see if that works.

Greg

Yes, I am using the driver you're thinking of; it's the one that is included in the kernel along with the rest of the IDE chipset drivers. I didn't include the "IDE RAID" driver, just the 'normal' ide driver.

Running 2.4.22.

I'm just confirming which driver I'm using, not whether I think it can boot. I don't boot from my hpt374.

Dear Folks,

I'm new in this place, but here are so many replies to that hellmade hpt37x2.o driver that i thought, registration could be useful.

My machine is a PIV (2GHz? matters?) intel 845 1HE Box with a RocketRAID133 on his Risercard.

I _had_ a functioning installed RH9 with the hpt37x2.o driver from Highpoint, booting from a RAID1 device and installed like Highpoint suggest. For some reasons i had to build a new kernel so took the 2.4.20-8 Src package, made my conf and compiled it. i placed the bzimage, System.map in /boot (with a extra <kernel-version> info, cause i wanted to be able to boot both kernels), placed the hpt37x2.o in /lib/modules/2.4.20-8<my-kernel>/ ..., made a depmod and an own initrd. (I think, man can use the installed initrd as well, since there are no other requirements for the ramdisk, but i just wanted to get sure...). Ok, lets go to grub.conf, make an extra entry for my new kernel and run "grub-install /dev/sda". no errors reported.

I rebooted, and never could boot again

grub says Error 21 (which means selected disc does not exists).

I booted the rescue system, inserted the driver manually and mounted my filesystem - reconfigured the grub.conf to only one entry (the old functioning) - with no success!


As the rescue tried to find a linux installation, i had suprisingly 2 options.../dev/sda2 and /dev/sdb2???? i rebooted and entered the Highpiont BIOS and my RAID1 was gone, i had now 2 single discs!!! Could grub have broken my RAID1 Array???

And how can make the box booting again. Where shall i install grub? i'm a bit confused, when grub is loading, the hpt37x2.o isn't loaded (because it's in initrd), so therefore the intall Destination can't be /dev/sda, can it?

is there something like "grub-install (hd0)" ??

My old configuration was :

/dev/sda1 /boot
/dev/sda2 /


I'm now merging the two discs together and mirrorin Disc one to disc 2 (take about 3 hours... so i went home), so i hope, i'm having my data on an array called sda tomorrow

Can anyone help me?

Someone please help me out!!

I'm a Linux newbie and i just installd Suse 9.1 on a normal IDE harddrive.
Nou the problem is i have windows XP and all of my backups on my two serial ata raid harddisk's.(Whit's are NTFS of course)
And i have a Abit At7-MAX2 mainbord whit a hpt374 raid controller.

My first quastion is can i normalie see these harddisk's and see my data becouse i can see them when i go to my computer in Suse 9.2 but when i open them there is no data on these disk's.(There is NTFS support in Suse 9.2 is there not????)

So it's my hpt374 controller whit's suse 9.2 can't see and i downloaden the drivers from the HighPoint site but these where only until versio 8.2 for suse.
And when i testet them whit the command 'insmod hpt374.o' it did indeed not work.

So where can i get the correct hpt374 driver for suse 9.2 OR what aim i doing wrong????