Si-pex40062

Newegg: N82E16816124062

Got this when I upgraded one of my arrays to 4 TB drives. Wanted to re-use that old server without an expensive RAID card. RAID-Z2 was the solution. Seemed simple enough, I do tons of Linux builds, etc. No problem. Snagged up the card.

FreeNAS, no joy. FreeBSD, same. CentOS LiveDVD however does show the disks without any issues. The Free's show a drive timeout like: (I'm typing this by hand so it might have a typo... This is BSD's reporting:))

Code:
ahcich6:Timeout on slot0 port0
ahcich6: is 00000002 cs 00000000 ss 00000000 rs 00000001 tfd 50 serr 00000000 cmd 00046017
(aprobe1:ahchich6:0:0:0): ATA_IDENTITY ACB: EC 00 00 00 00 40 00 00 00 00 00
(aprobe1:ahchich6:0:0:0): CAM status: Command Timeout
(aprobe1:ahchich6:0:0:0): Error 5 Retry was blocked

So it seems for whatever reason FreeBSD 9's driver isn't working with this particular card. Is there a tweaked AHCI driver somewhere? I'd rather use FreeBSD for this RAID-Z2 setup (8-drives) but... CentOS can do it... I'd rather not. Although it was rather odd having to setup the main boot drive also (it was an IDE drive and I had to jump through hurdles with the CD-ROM drive since it was SATA, it didn't want to load, I had to drop it to IDE mode. I need to fix that.)

Either way; I can easily setup a RAID 6 (though likely not Z2, have to check) in CentOS directly through the tool without a controller card on all 8 drives. But TBH, I'd rather have the lower overhead and more pure direct NAS-ness of FreeBSD for this system. Anyone have any ideas? I've done tons of Linux distributions before, at home and professionally but never FreeBSD.
 
Hmm, a Marvell 88SE9235. I have not used those, but have used multiple SIL3124-based SY-PEX40008 cards with success. And the two-port SIL3128 version, I think the same as SD-SA2PEX-2IR works, but the SATA connectors do not have latches.

Have you checked for firmware updates for the card's BIOS?
 
Checked, there are no firmware updates. Mainboard has only one PCIe port so I need four on one card. Or some PCI cards... Still it works on other Linux builds without an issue. I tried it on Ubuntu Live and RedHat also, all work no problem. It seems to be the stripped down flavors that aren't playing nicely. Actually the darn thing is fast too. The same drives on the board are slower than the ones on the card, 4 and 4... Not by much.

I am building two servers for work, I'll need two of these cards also (twelve drives each, four on the board). So this is also a test project for how those servers are going to happen.
 
I also had a ticket in with the manufacturer for what it's worth, but at this time I don't think it's in their court. It also has no latches I do believe. I don't really care about those since once it's on it won't be messed with and cables push on nicely and do not slide off without a nice tug.
 
Well, so far I've done additional tests with CentOS and Ubuntu, both see the card just peachy. It has to be a missing driver piece from FreeBSD unfortunately. Which also means that at least for this project, if I can't figure out how to add said driver (and find it), I'm out of luck and have to use another flavor...
 
The driver developers sometimes read and post here, but it's intermittent. They tend to read and respond much more frequently on the mailing lists.

I did a little bit of searching, and the type of errors you're seeing can sometimes be caused by a SATA level mismatch, the drive being SATA 2 and the controller trying to use SATA 3. But whether there is a way to override that in that driver, or if it is an easily-fixed bug, I don't know.
 
I have emailed that list you mentioned. I wanted to make sure I dotted all my T's and crossed all my I's.. ;-)

Anyhow, I triple checked.. All of the full flavored 'Nixes worked without any problems. Just the stripped down BSD's didn't seem to have the drivers. Makes perfect sense actually. I ended up using Ubuntu Server version with ZFS this time around to get the system up and running for the time being (CentOS would have been preferred but ZFS on CentOS is a bit Wonky...).

As for the card itself, working swell for what it is. Loaded up the now usable 10.8 TB RAID-Z2 array last night and have it populated over 40% already today as a test (from my active 24 TB array via an actual card in another server). It's running without a hitch as far as that's concerned, I can't complain although writing speed isn't horribly fast but reading speed is acceptable. I do need to find some performance measurement tools for non-RAID/GUI Linux systems to get better measurements though...
 
Ohh, and to test that theory also, I tried some new SSDs and a few 1 TB V-Raptor Drives on the cards also. Same results... So the mismatch wasn't it.
 
Marvell chips have issues with AHCI compatibility. Make sure you have fresh enough system that includes respective quirks for that chip in ahci(4) driver. Make sure that system detects controller as Marvell 88SE9235 and not generic AHCI. Otherwise -- update to 9-STABLE.
 
Huh? As I said before, it works just peachy on every full Linux build. Just not the FreeBSD and FreeNAS flavors. But in CentOS, Ubunto, Redhat, etc. it worked flawlessly. and in the BIOS POST it is seen and recognized flawlessly also. It is during Linux boot reading initialization that it times out and fails in FreeBSD.

And this was with a fresh install of 9-STABLE. Nothing on anything else. No updates, etc. The system is running Ubuntu with ZFS right now and has been flawlessly since a few days after this post was made and the array is 92% loaded. It's the 12 TB server (two parity) for my 32 TB total array and quite necessary. I could do some reboot testing with say a spare HDD for you simple enough to see if it could recognize something but at this point I can't reinstall it anymore at this point. Somewhere I have a whole video of the boot process on my phone if you really want to see it. (S4, it's in 1080, but it's a recording of the screen so we know how crappy that can be in return). At least up to the failure point. That is if I still kept it after I captured the error for making this post (probably).
 
When you say "9 stable", do you mean you installed a snapshot? Or installed a release and updated by building from source? I ask because people often misunderstand what "-STABLE" means.
 
I got the ISO for FreeBSD 9.1-RELEASE, burned it to a DVD, then installed it to a blank HDD. At the time, all of the other drives were also blank and clean also. No partition information at all. Clean slates to make a nice new RAIDZ2 volume on (which is nice cause Ubuntu won't destroy any partitions when it's creating them... I had to re-destroy it before installing Ubuntu. CentOS and Redhat both let me stomp on the old partition.)

After that I did, I forget, it's been so many.. csup something? to update the system to the latest updates?.. tweak it all out... rebooted... same result. And it works with the rest so I got tired of messing with it... Simple as that. If you want me to test something, I'd be glad to. Or try something even. I have plenty of blank drives and I can take the server offline for a spell to test some theories in the offtime. But other then that.

I did not try the FreeBSD 9.2-BETA1 release, perhaps I should have. If I can make a bootable USB stick off FreeBSD, that would be the easiest way to do testing, but other then that it'll be the weekend or longer before I can do anything.
 
Sorry for the delay guys, I have Primary Progressive Multiple Sclerosis, and it slammed me into the Hospital Friday night. Been here ever since. Now I got my laptop with me but that's little good to do testing on the servers... I'll be here it seems for at least another week. But I'll get some testing for ya as soon as I can get home and get stuff settled down.
 
Back
Top