Problem installing Promise FastTrak S150 SX4 (2.4 and 2.6.7)

[yota]

Hi,

We have hard time trying to use Promise FastTrak S150 SX4 raid cards on two dell PowerEdge.

We tryied both 2.4.6 and 2.6.7 debian kernel images.

2.4.6 uses sata_promise driver
2.6.7 uses sata_sx4 driver

The problem occurs on both machines.

The driver detect ATA disks but instead of showing one Raid 5 (one Raid 1 for the second machine), it dumps our four disks on the boot log.

Then it trys to access the first ATA disk and timeout (after a long time).

The 2.6.7 kernel refuse to boot after this failure.

The 2.4.6 kernel continue after the failure so we are able to give its syslog which is quite the same as the 2.6 one concerning sata_* drivers.

The full syslog is accessible here: http://yota.dyndns.org/syslog

This part concern sata except the scb2_flash but i'm not sure if it's armfull or not for the raid driver.

Any help or information would be really appreciated :)

best regards
laurent

Aug 26 15:44:54 localhost kernel: scb2_flash: warning - can't reserve rom window, continuing
Aug 26 15:44:54 localhost kernel: CFI: Found no SCB2 BIOS Flash device at location zero
Aug 26 15:44:54 localhost kernel: scb2_flash: flash probe failed!
Aug 26 15:44:54 localhost kernel: libata version 1.02 loaded.
Aug 26 15:44:54 localhost kernel: sata_promise version 0.92
Aug 26 15:44:54 localhost kernel: Local DIMM ECC Enabled
Aug 26 15:44:54 localhost kernel: ata1: SATA max UDMA/133 cmd 0xF8C22200 ctl 0xF8C22238 bmdma 0x0 irq 25
Aug 26 15:44:54 localhost kernel: ata2: SATA max UDMA/133 cmd 0xF8C22280 ctl 0xF8C222B8 bmdma 0x0 irq 25
Aug 26 15:44:54 localhost kernel: ata3: SATA max UDMA/133 cmd 0xF8C22300 ctl 0xF8C22338 bmdma 0x0 irq 25
Aug 26 15:44:54 localhost kernel: ata4: SATA max UDMA/133 cmd 0xF8C22380 ctl 0xF8C223B8 bmdma 0x0 irq 25
Aug 26 15:44:54 localhost kernel: ata1: dev 0 cfg 49:2f00 82:7c6b 83:7f09 84:4003 85:7c69 86:3e01 87:4003 88:207f
Aug 26 15:44:54 localhost kernel: ata1: dev 0 ATA, max UDMA/133, 490234752 sectors (lba48)
Aug 26 15:44:54 localhost kernel: ata1: dev 0 configured for UDMA/133
Aug 26 15:44:54 localhost kernel: ata2: dev 0 cfg 49:2f00 82:7c6b 83:7f09 84:4003 85:7c69 86:3e01 87:4003 88:207f
Aug 26 15:44:54 localhost kernel: ata2: dev 0 ATA, max UDMA/133, 490234752 sectors (lba48)
Aug 26 15:44:54 localhost kernel: ata2: dev 0 configured for UDMA/133
Aug 26 15:44:54 localhost kernel: ata3: dev 0 cfg 49:2f00 82:7c6b 83:7f09 84:4003 85:7c69 86:3e01 87:4003 88:207f
Aug 26 15:44:54 localhost kernel: ata3: dev 0 ATA, max UDMA/133, 490234752 sectors (lba48)
Aug 26 15:44:54 localhost kernel: ata3: dev 0 configured for UDMA/133
Aug 26 15:44:54 localhost kernel: ata4: dev 0 cfg 49:2f00 82:7c6b 83:7f09 84:4003 85:7c69 86:3e01 87:4003 88:207f
Aug 26 15:44:54 localhost kernel: ata4: dev 0 ATA, max UDMA/133, 490234752 sectors (lba48)
Aug 26 15:44:54 localhost kernel: ata4: dev 0 configured for UDMA/133
Aug 26 15:44:54 localhost kernel: scsi2 : sata_promise
Aug 26 15:44:54 localhost kernel: scsi3 : sata_promise
Aug 26 15:44:54 localhost kernel: scsi4 : sata_promise
Aug 26 15:44:54 localhost kernel: scsi5 : sata_promise
Aug 26 15:44:54 localhost kernel: Vendor: ATA Model: Maxtor 7Y250M0 Rev: 1.02
Aug 26 15:44:54 localhost kernel: Type: Direct-Access ANSI SCSI revision: 05
Aug 26 15:44:54 localhost kernel: Vendor: ATA Model: Maxtor 7Y250M0 Rev: 1.02
Aug 26 15:44:54 localhost kernel: Type: Direct-Access ANSI SCSI revision: 05
Aug 26 15:44:54 localhost kernel: Vendor: ATA Model: Maxtor 7Y250M0 Rev: 1.02
Aug 26 15:44:54 localhost kernel: Type: Direct-Access ANSI SCSI revision: 05
Aug 26 15:44:54 localhost kernel: Vendor: ATA Model: Maxtor 7Y250M0 Rev: 1.02
Aug 26 15:44:54 localhost kernel: Type: Direct-Access ANSI SCSI revision: 05

[kkubasik]

Hey, no solution, but a maybe a pointer or 2 in ther right direction

OK at the beginning we have this

Aug 26 15:44:54 localhost kernel: scb2_flash: warning - can't reserve rom window, continuing
Aug 26 15:44:54 localhost kernel: CFI: Found no SCB2 BIOS Flash device at location zero
Aug 26 15:44:54 localhost kernel: scb2_flash: flash probe failed!

Perhaps the BIOS config is not correct and cannot point the kernel to the devices correctly.
---
the part thats getting me is

Aug 26 15:44:54 localhost kernel: scsi2 : sata_promise
Aug 26 15:44:54 localhost kernel: scsi3 : sata_promise
Aug 26 15:44:54 localhost kernel: scsi4 : sata_promise
Aug 26 15:44:54 localhost kernel: scsi5 : sata_promise

inherently when i see scsi on a ata device i wonder.there are several other mentions of scsi, but im no hardware guru, mabey scsi controlers can be used with sata, but it seems unlikely.

---

Another point to keep in mind is that Debian tends to lag behind some more cutting edge Distros as far as device support as it is a much more stable distro, perhaps you might try your hand at Slackware or Gentoo and see what comes of it.

[dade49]

I am having problems also. I am running the same Promise FastTrak S150 SX4-M card in a Dell SC420 server. Running current (testing) Debian image. I actually have a lot of interesting problems. Part of the time during install, it hangs on loading sd_mod during hardware detect. When I actually get it to load, sometimes I get a DMA timeout after first reboot, sometimes I don't. I know my hardware works. I've loaded .net and xp on it just to verify the hardware.
I can't seem to make any linux distro work with this machine yet. Somtimes I can get it loaded, and the sata_sx4 driver (kernel 2.6) starts working. Then I have massive write problems whenever I write to the disk.
As a side note; the S150 SX4-M card does not have a XOR engine. It's a software raid, and linux uses it as such. I am doing a software raid 5 with this card.
Can anyone help me? I need some type of linux disrto on this box, but I have to use kernel 2.6 due to the Dell hardware not being supported by 2.4. The official drivers provided from Promise only support the 2.4 kernel. I'm willing to test anything.

[guest]

Fedora Core 3 recognizes this card but not the RAID function. It will only work as a 4 port SATA controller (RAID-less). You can run Linux Software RAID with this card, but I noticed a performance hit.

Good Luck

[tirsvad]

For Promise Fasttrak SX4-M and Promise Fasttrak SX4100

To get Fasttrak driver to work under linux u have to recompile driver.

I have make a driver disk for opensuse 10.0 only for I386 so far
hint. remember to unload sata-sx module when install linux

link http://software.tirsvad.dk/

I will try to make a driver disk for opensuse 10.1, but i have big problems here.

Jens Tirsvad Nielsen
http://tirsvad.dk

[Nagio]

Hello tirsvad,

Is your driver works with x86_64 installation ?
I think no because in your .img file , I just see linux/suse/i386-10.0/install ...

Is it possible to modify it to work with x86_64 installation ?
I 'm waiting for the new version of this driver to test it on suse 10.1 x86_64 installation
Best regards
Nagio

[Zelmo]

I was using a FastTrak S150 TX2+ or 4+ a couple of years ago, when the 2.6 kernel was first completed. A fellow had written a driver for the 2.4 kernel which supported the card's RAID features, but the 2.6 kernel has a built-in driver that refuses to do RAID. The rationale is that the FastTrak does not use true hardware RAID like the more expensive cards do, and the kernel developers didn't want to put software-controlled RAID in the kernel. They recommended that you use Linux's software RAID instead, which performs equal to, if not slightly better than, FastTrak and similar controllers' RAID.

Of course, like I said, that was a couple of years ago. I haven't used that controller card in a while, and it's possible some ambitious developer has taken up the flag and made a special kernel module. But Google didn't turn up anything for me.

[tirsvad]

Hello tirsvad,

Is your driver works with x86_64 installation ?
I think no because in your .img file , I just see linux/suse/i386-10.0/install ...

Is it possible to modify it to work with x86_64 installation ?
I 'm waiting for the new version of this driver to test it on suse 10.1 x86_64 installation
Best regards
Nagio

Sorry. It's only for I386. Perhaps i can try making it for x86_64, but i cant test it. Please reply if it works or not

Driver for OpenSuse 10.1 I have big problems here !!!!
(Driver is not opensource, so I don't know what TODO.)

[tirsvad]

Promise making cheap raid controllers.

But Fasttrak is a bad chose because it's software raid, with possible xor processor. They only develop driver until release of produkt. So it not working on newer OS

I recommend that to upgrade to Promise Supertrak 8350 (SATA Controller). That card is hardware raid. Driver is opensource.

I will change my "fasttrak SX4-M" to "supertrak 8350" in near future

[daniel.santos]

I own a Promise FastTrak S150 SX4-M and I want to make a few clarifications here. I'll start with saying that I will never buy another Promise product again (I'm not going to a Supertrak to it unworkable in 2 years when there's a new kernel).

The Promise SX4-M is not true "software raid" or "fake raid" -- it's also not "hardware raid". This card, like a few others, lies in the space in between and is more of a "raid accelerator." The two features it offers is a cache (with fairly complex caching controls) and an XOR engine. The XOR engine will take a HUGE load off of your main CPU when using RAID5. The cache helps in either condition. To utilize the full extend of the cache, you need to have a reliable Uninterruptible Power Supply (UPS) because card does not have an on-board battery backup. With a reliable battery backup, you can enable the write behind cache which enables you to make writes and return from the write call before all of the data has been written to disk. The data will not actually be written until idle time has been found to do so or the cache has been filled up, requiring it to do so. The cache is also expandable to 256MB, although there are very few DIMMs compatible with it (expect to pay $60-$70 US). This much cache has an incredible effect on performance!

All in all, the functionality of this rather inexpensive card is VERY nice. What is NOT at all nice is the support! They spit out a few drivers and then shelved it!! (This product is not dis-continued). There is a beta driver for x86_64 (that also supports the 2.6 kernel) released as partial source code (under the GPL), but I cannot get it to work. It is this support that I find completely unacceptable.

A few other notes, the fully open-source sata_sx4 driver (also beta) is not working correctly on my system. Before I upgraded my Kernel, I could mount my RAID0 partitions and write to them, but I experienced corruption. I was only writing from a single process, so this shouldn't fall under a previously reported bug that caused corruption with async writes. After my Kernel upgrade, I can no longer mount a partition of of the array. Also, the sata_sx4 driver does not have support for the raid accelerating functions (i.e., cache and XOR engine), so it's like doing complete software raid. Oh yea, I should mention (although it's already implied) that with the FastTrak SX4/M, the logical operations of the RAID lies in the driver (and thus, on the main CPU). If you don't have a working driver from Promise, there is no difference between the functionality of using sata_sx4 raid functions and using md (the Liniux fake raid software, now part of the Kernel).

I'm currently using Gentoo on an amd64 system (with the ~amd64 keyword enabled, so I'm using more "bleeding edge" packages and kernels). My kernel is 2.6.17-gentoo-r2 (that's the 2.6.17 kernel with Gentoo's 2nd revision of patches).

When I get around to upgrading my RAID card, I'll be looking at another manufactuer with a better reputation for support (i.e., releasing reliable drivers). www.tweakers.net seems like a good place for this info.

Daniel