Marvel AOC-SASLP-MV8 - 8 Sata PCI-E 4x non-raid controller card

@pirivan

Im afraid you have some unlucky hardware mobo-card-hdd combination, since here I have just tried hooking up 8 drives on saslp-mv8 card and no problems, have restarted a few times, copied a few gigs of data onto server without a hickup..

im ruining this on asus p5bv/sas mobo with 2× sat2-mv8 and 1× saslp-mv8 card

You running WHS or something else?
 
Could someone who has one of these try booting OpenSolaris 2008.11 and see if this card is recognized at all? I'm not terribly optimistic about its chances of success, but if it works I'll buy one, and I know a few other people who'd be interested.
 
Well, I set the BIOS to IDE emulation mode for the SATA ports and reinstalled with only one controller card in the system. Same symptoms as before, after I installed the controller card drivers I was able to see the 7 drives connected to the controller in the WHS console for a brief moment before it hard locked. A reboot had the same symptoms as usual, WHS loading screen and then to black.

I am going to try another WHS reinstall with NO drives connected other than the OS drive and internal DVD/RW (I will just pull everything out of the hot swap drive bays) AND neither of the AOC-SASLP-MV8 controllers installed in the PC. After boot I will install the motherboard drivers, reboot, install the controller card physically (and drivers), reboot, then slide all the controller card drives in to the hot swap drive bays.

***UPDATE***

I tried another WHS install with NO drives connected other than the OS drive and DVD-RW/RW, in SATA IDE mode, not AHCI. After WHS was up and running I installed the motherboard drivers and then rebooted. Afterwards I shut down and physically installed the AOC-SASLP-MV8 in the primary PCI-E slot, then booted up and installed the drivers. With no drives connected to the AOC-SASLP-MV8 I was able to boot into WHS without any issues.

After rebooting with 7 drives connected to the AOC-SASLP-MV8 it hung at a black screen after the WHS loading screen again. Another reboot with 4 drives on the primary controller controller allowed it to boot. I then shutdown and inserted another drive on the primary controller, total of 5 on the first controller (6 total in the system counting the OS drive). The system then booted without any issues. I shut down again and inserted another drive on the primary controller, total of 6 on the first controller (7 total in the system counting the OS drive). This is where it hung on loading at a black screen. So as soon as I hit 6 drives on the primary controller, it hangs.

So, I moved the drive in bay 6 (the Norco 4020 has 5 rows of 4 drives, I am counting starting on the left going right, top to bottom so bay 6 is row 2 drive 2) to bay 7 and it booted! So I slowly added drives to each bay until all of the first two rows EXCEPT bay 6 was full and it continued to boot. I added drives to the onboard ports (the two that I could use), still boots (I was shutting down in between adding each drive). Then, I shut down and installed the other AOC-SASLP-MV8. Afterwards I continued to slowly add drives until I had the machine up and booting with 16 drives (I ran out of drives to use)!

Currently all slots are filled except bay 6, 1 bay on controller 2 (ran out of drives) and 2 bays on the bottom row (these are connected to AHCI only ports and I couldn't plug anything into them for the moment as I had installed the OS in SATA IDE mode). If I had 3 more drives and was operating in AHCI mode I suspect I could fill all the ports. After it was up and running I moved a drive back to bay 6 and it worked! My only guess is that the drives I had been placing in it before (I have tried a variety of drives in that bay) during my testing were not fitting right in the connectors to the backplane.

Tomorrow I intend to mess around quite a bit with drives in bay 6 and see what was causing it to malfunction with certain drives that would prevent WHS from booting. I also plan to reinstall WHS in AHCI mode, which I fully expect will work this time. It looks like that damn bay 6 was messing everything up. I will also try to borrow a few more drives, so that I can make sure it works if all 20 bays are filled. Also, for some reason on this install the onboard NIC isn't working right but I hardly care about that right now, I suspect it will be fixed when I do another fresh reinstall with AHCI.

It looks like I am finally getting somewhere; I am going to try to not be excited.
 
Could someone who has one of these try booting OpenSolaris 2008.11 and see if this card is recognized at all? I'm not terribly optimistic about its chances of success, but if it works I'll buy one, and I know a few other people who'd be interested.

I can verify that OpenSolaris 2008.11 does NOT pick these cards up, at least not "out of the box". I do not have much experience with OpenSolaris so I did not spend any time on it. All I know is that it did pick up my AOC-SAT2-MV8 automatically and I built a test ZFS array on that card before deciding WHS was a better fit for me.
 
I can also add to the success stories of this card fully loaded running WHS, but only with Int13h disabled in the card's bios.

Before I tried disabling it, WHS would not even install much less boot with the AOC-SASLP-MV8 in my system (an Asus K8N-LR). The installer hangs at the "Loading Windows Files" screen and never gets to the GUI portion of the install. At the time I hadn't figured out the Int3h fix, so I just removed the card to install onto a drive on the motherboard's controller hoping it would work after WHS was installed. It did not. Once it was installed and I reinserted the AOC-SASLP-MV8, I would just get a black screen after the POST process. No WHS splash loading screen at all.

This was when I tried disabling Int13h and it immediately booted. I have been running this config for several days now with numerous reboots and it has always come right back up without issue. I have one system drive on the MB and 8 1TB drives on the AOC-SASLP-MV8 for 9 total. I have loaded up about 4TB of data so far with 0 stability issues.
 
I can also add to the success stories of this card fully loaded running WHS, but only with Int13h disabled in the card's bios.

Not the case here or at least I dont know if this is disabled by default at my card.. whats the command for entering card bios (ctrl+c, ctrl+s?) as I dont see any messages during boot time as I did with my adaptec card or LSI chip on the MB
 
Last edited:
I'm not at home and can't double check, but I seem to remember it being CTRL-M to enter the bios. The message showing the key combination shows up at the very end, after all the drives have been spun up and detected. I think it was only on the screen for a second or so before it moved on and starts showing the WHS splash screen.

Int13h was enabled by default for me.
 
Yes, it is ctrl + M to enter the controller BIOS and int13h was enabled by default for me as well. I had to disable it so that I could select my hard drive connected to the motherboard controller as a boot device. With int13h enabled on the 'bootable devices' list in my motherboard BIOS all I could select was 1 of the 8 drives attached to the AOC-SASLP-MV8.

Anyone have any success slipstreaming the AOC-SASLP-MV8 drivers into their WHS install? I am going to reinstall with my slipstreamed AHCI drivers but I thought it might be worth a shot adding in the AOC-SASLP-MV8 drivers as well.
 
just checked, int13h enabled, but still no problems with 8 drives hooked to controller
 
just checked, int13h enabled, but still no problems with 8 drives hooked to controller

How many hard drives are listed as bootable choices in your motherboard BIOS? Also, what is your OS drive connected to? The AOC-SASLP-MV8 or a motherboard SATA port?
 
will check tomorrow, as I'm copying large batch of data onto server atm & will leave it run overnight.. im booting whs from mb port
 
hmm, I can't come into card bios no matter what, have restarted at least 5 times in row now while pressing ctrl+m..
strange, as I cant even see no option or listed drives on boot up (only the sat2-mv8 and LSI sas controller are listing drives on boot)
 
Excellent, thanks for the test.

I was hoping this would work under Solaris as well, but I checked the ZFS mailing lists as well, and the news was glum indeed.

I picked up the AOC-USAS-L8i for $80 off eBay instead; stock BIOS lacks RAID functionality allowing to act as a simple HBA. Overkill, but nearly $50 cheaper than either the AOC-SAT2-MV8, AOC-SAS(LP)-MV8, SAS 5/i, 5/e, 6/iR, HP SC44GE.

I'd go as far to say that if you don't want a bottleneck running a SAT2-MV8 in a PCI slot, and aren't planning on Solaris, the SAS(LP)-MV8 is probably a safe bet. Else, the USAS-L8i and its LSI 1086E based chip can also function as a HBA, albeit with all the components flipped (its a backwards PCIe card).

Note that another option (that kinda sucks) is using a cheap PERC 5/i and running 8 disks all as single-disk RAID0s. I was about to do that before I scored the L8i.
 
In case this might be helpful to anyone I was able to slipstream both my motherboard AHCI drivers (NVIDIA 780a drivers) and the AOC-SASLP-MV8 drivers into a WHS installation. I followed the slipstream install guide here and copied the drivers into the DOS folder. However, as both drivers have a txtsetup.oem file I "joined" those two files together so that both drivers would install. I was able to see drives connected to all the controllers during WHS setup with this method. The install worked perfectly. After boot I simply installed the motherboard drivers which worked fine (though I believe at some point my onboard NIC went bad as it doesn't appear to work, oh well).

Here are the contents of the txtsetup.oem file I joined together/created on my own (obviously this will be different depending on what motherboard controller drivers you have):

Code:
[Disks]
d1 = "NVIDIA AHCI DRIVER (SCSI) disk 1",\disk1,\
d2 = "NVIDIA AHCI DRIVER (SCSI) disk 2",\disk2,\
d3 = "Marvell SAS Driver Diskette", \mv64xx, \
d4 = "Marvell SAS Driver Diskette", \mv64xx, \i386
d5 = "Marvell SAS Driver Diskette", \mv64xx, \amd64

[Defaults]
SCSI = BUSDRV

[SCSI]
BUSDRV = "NVIDIA nForce Storage Controller (required)"
i386_64xx   = "Marvell 64xx/63xx SAS Controller 32bit Driver", mv64xx 
amd64_64xx  = "Marvell 64xx/63xx SAS Controller 64bit Driver", mv64xx

[Files.scsi.BUSDRV]
driver = d1,nvgts.sys,BUSDRV
inf    = d1, nvgts.inf
catalog = d1, nvata.cat
dll    = d1,nvraidco.dll
dll     = d1,NvRCoENU.dll
dll     = d2,NvRCoAr.dll
dll     = d2,NvRCoCs.dll
dll     = d2,NvRCoDa.dll
dll     = d2,NvRCoDe.dll
dll     = d2,NvRCoEl.dll
dll     = d2,NvRCoEng.dll
dll     = d2,NvRCoEs.dll
dll     = d2,NvRCoEsm.dll
dll     = d2,NvRCoFi.dll
dll     = d2,NvRCoFr.dll
dll     = d2,NvRCoHe.dll
dll     = d2,NvRCoHu.dll
dll     = d2,NvRCoIt.dll
dll     = d2,NvRCoJa.dll
dll     = d2,NvRCoKo.dll
dll     = d2,NvRCoNl.dll
dll     = d2,NvRCoNo.dll
dll     = d2,NvRCoPl.dll
dll     = d2,NvRCoPt.dll
dll     = d2,NvRCoPtb.dll
dll     = d2,NvRCoRu.dll
dll     = d2,NvRCoSk.dll
dll     = d2,NvRCoSl.dll
dll     = d2,NvRCoSv.dll
dll     = d2,NvRCoTh.dll
dll     = d2,NvRCoTr.dll
dll     = d2,NvRCoZhc.dll
dll     = d2,NvRCoZht.dll

[Files.SCSI.i386_64xx] 
inf	= d4, mv64xx.inf
inf 	= d4, mvnodrv.inf
driver	= d4, mv64xx.sys, mv64xx
catalog = d4, mv64xx.cat

[Files.SCSI.amd64_64xx] 
inf	= d5, mv64xx.inf
inf 	= d5, mvnodrv.inf
driver	= d5, mv64xx.sys, mv64xx
catalog = d5, mv64xx.cat

[Config.BUSDRV]
value = parameters\PnpInterface,5,REG_DWORD,1

[Config.mv64xx]
value = parameters\PnpInterface,5,REG_DWORD,1

[HardwareIds.scsi.BUSDRV]
id = "PCI\VEN_10DE&DEV_0036", "nvgts"
id = "PCI\VEN_10DE&DEV_003E", "nvgts"
id = "PCI\VEN_10DE&DEV_0054", "nvgts"
id = "PCI\VEN_10DE&DEV_0055", "nvgts"
id = "PCI\VEN_10DE&DEV_0266", "nvgts"
id = "PCI\VEN_10DE&DEV_0267", "nvgts"
id = "PCI\VEN_10DE&DEV_037E", "nvgts"
id = "PCI\VEN_10DE&DEV_037F", "nvgts"
id = "PCI\VEN_10DE&DEV_036F", "nvgts"
id = "PCI\VEN_10DE&DEV_03F6", "nvgts"
id = "PCI\VEN_10DE&DEV_03F7", "nvgts"
id = "PCI\VEN_10DE&DEV_03E7", "nvgts"
id = "PCI\VEN_10DE&DEV_044D", "nvgts"
id = "PCI\VEN_10DE&DEV_044E", "nvgts"
id = "PCI\VEN_10DE&DEV_044F", "nvgts"
id = "PCI\VEN_10DE&DEV_0554", "nvgts"
id = "PCI\VEN_10DE&DEV_0555", "nvgts"
id = "PCI\VEN_10DE&DEV_0556", "nvgts"
id = "PCI\VEN_10DE&DEV_07F4", "nvgts"
id = "PCI\VEN_10DE&DEV_07F5", "nvgts"
id = "PCI\VEN_10DE&DEV_07F6", "nvgts"
id = "PCI\VEN_10DE&DEV_07F7", "nvgts"
id = "PCI\VEN_10DE&DEV_0768", "nvgts"
id = "PCI\VEN_10DE&DEV_0AD5", "nvgts"
id = "PCI\VEN_10DE&DEV_0AD4", "nvgts"
id = "PCI\VEN_10DE&DEV_0AB9", "nvgts"
id = "PCI\VEN_10DE&DEV_0AB8", "nvgts"

[HardwareIds.SCSI.i386_64xx] 
id = "PCI\VEN_11AB&DEV_6440", "mv64xx" 
id = "PCI\VEN_11AB&DEV_6485", "mv64xx"

[HardwareIds.SCSI.amd64_64xx] 
id = "PCI\VEN_11AB&DEV_6440", "mv64xx" 
id = "PCI\VEN_11AB&DEV_6485", "mv64xx"
 
In case this might be helpful to anyone I was able to slipstream both my motherboard AHCI drivers (NVIDIA 780a drivers) and the AOC-SASLP-MV8 drivers into a WHS installation. I followed the slipstream install guide here and copied the drivers into the DOS folder. However, as both drivers have a txtsetup.oem file I "joined" those two files together so that both drivers would install. I was able to see drives connected to all the controllers during WHS setup with this method. The install worked perfectly. After boot I simply installed the motherboard drivers which worked fine (though I believe at some point my onboard NIC went bad as it doesn't appear to work, oh well).

Here are the contents of the txtsetup.oem file I joined together/created on my own (obviously this will be different depending on what motherboard controller drivers you have):

Slip streaming works well. But you should take the advice and to never install an operating system with all your storage drives connected. Mistakes can happen :eek: If you want to take it a step further, automate the slip stream process, it works great, then you never have to worry, unless if you have drives connected (been there done that one, oops).
 
I just ordered one from wiredzone.com. I'm planning to use it with Ubuntu 9.04 and 8 1TB drives. I'll send an update when I get it all built.

Dan
 
I just ordered one from wiredzone.com. I'm planning to use it with Ubuntu 9.04 and 8 1TB drives. I'll send an update when I get it all built.

Dan

Good luck on getting it to work on Ubuntu.

Also got a aoc-saslp-mv8 last week. Installed it, fresh install Debian, .... hmmm... Why are my 8 Seagates SATA disks not showing up in Debian's /dev/xxx folder... They showed up during boot time...

After some digging around, it looks like the mvsas, or sata_mv drivers are not working ( tried both to be sure ).

From what i can tell:

HD's shows up in Bios
Controllers Bios screen loads & show the drives.
Debian can find the controller:
A quick lspci shows this ->
02:00.0 SCSI storage controller: Marvell Technology Group Ltd. MV64460/64461/64462 System Controller, Revision B (rev 01)

But, the chipset on it is the Marvel 88SE6480. And it looks like the drivers do not support it. In other words, they can't recognize it, and thus, do not work...

I found that Marvell released a patch adding support for the 64bit controller used in the aoc-saslp-mv8, about 8 weeks ago, but the first patch was broken ( several in fact have been problematic/broken based on the scsi kernel newsgroup ). And the current used kernels ( 2.6.26 ) only have the old limited drivers.

From what i can tell from the changelogs, even up to 2.6.29 don't have the updated drivers...

I'm relative new to Linux, so, if i'm wrong, please accept my apologies, but as far as i can tell, this card can not function on a Debian based system, unless Marvell has the complete patch source code finished and it will need a manual build of the code.

Already contacted Marvell, Supermicro, etc about it. Got only one reply ( from supermicro ), and up to now, did not get answer back on it. I mean, come on. They release it with Windows, Redhat, etc drivers, but Debian or Debian based Distro's make up more then 50% off the server market, and whoeps, no separate drivers included ( and the readme sucks, written by somebody from Asia, you can tell ;) ) and no source with a clear build instruction ...

Even the opensolaris guys who have one reported a no go.

So dan.b.mann, do not expect a: Plop into a PCI-e slot, attach hd's, boot pc, and here are your drives under Ubunutu ( what is a Debian based distro ).
 
Already contacted Marvell, Supermicro, etc about it. Got only one reply ( from supermicro ), and up to now, did not get answer back on it. I mean, come on. They release it with Windows, Redhat, etc drivers, but Debian or Debian based Distro's make up more then 50% off the server market, and whoeps, no separate drivers included ( and the readme sucks, written by somebody from Asia, you can tell ;) ) and no source with a clear build instruction ...

From the Supermicro manual:

"Use the source code from the folder: \Linux_3_1_0_7 to compile."

They look to include source, so you'll have to build it yourself.
 
From the Supermicro manual:

"Use the source code from the folder: \Linux_3_1_0_7 to compile."

They look to include source, so you'll have to build it yourself.

No such luck it seems... Already checked out the drivers cd...

The cd its linux files are under the following

system:/media/scd0/Driver/Linux/3.1.0.7/
system:/media/scd0/Driver/Linux/3.1.0.7/3.1.0.7-rpms

This is identical to the Supermicro's ftp site files, and structure:
ftp://ftp.supermicro.com/driver/SAS/Marvell/MV8/Linux/

It holds rpm's for redhat, suse ( /3.1.0.7-rpms ) and .ko files for redhat, suze ( wrapped in img files )... I don't see a source, unless i'm blind ;)...

The manual itself is almost pointless... Its missing a major chunk of information, like the controllers bios screen, that has several options ( like enabling int13 ) etc. Fyi, enabling int13 allows the hd's to be used as boot device. Or in reverse, in those MB bios that it does not show up, disabling allows it... You know, details that belong in the manual ;)

http://www.supermicro.com/manuals/other/AOC-SASLP-MV8.pdf
Did i mention that i love that manual, 20 pages, 2 about the cards layout, and 4 about windows/redhat/suse driver installation. All the rest is blank pages, or just plain useless... No information about the bios etc...

Just reading the readme on the cd/ftp, making my heat hurt, definably not written by somebody fluid in English ;)
ftp://ftp.supermicro.com/driver/SAS/Marvell/MV8/Linux/readme.txt
 
You appear correct, that is kinda weird. Most manufactures will provide an open source build driver, especially if they have RPMs etc. I bet it is coming.
 
You appear correct, that is kinda weird. Most manufactures will provide an open source build driver, especially if they have RPMs etc. I bet it is coming.

Send another mail to Supermicro, as they are the manufactures. Contacting Marvell has turned out a total dead end.

Time to put some fire under there feet ;)

When they sell a piece of hardware, its expected that the customer can use it. Not that its totally useless because they "forgot" the drivers or the source. Its not like Debian or even OpenSolaris are some small unknown distro's.

Anyway, we are fighting the good fight ;) So they better come over the bridge with working source code. Its rather bad that this hardware has gone unused for over a week & half, and that we can't deploy it to production.
 
Send another mail to Supermicro, as they are the manufactures. Contacting Marvell has turned out a total dead end.

Time to put some fire under there feet ;)

+1 Email to Supermicro, we will get our open source driver!
 
Instead of emailing them, try calling them up and actually talk to one of the people there? Their support is great and if they don't know the answer, they'll go find one of the engineers and ask them. Might be a bit quicker that way, and you can't exactly be ignored.
 
Wow, that sucks to hear about Debian, which means I'd probably be out of luck with Ubuntu unless I could compile a module from somewhere. But over the weekend I found that my Technet subscription gets me access to Windows Home Server so I thought I'd give that a try.

The card showed up today by the way, pretty damn fast from the time I ordered it. I give two thumbs up to Wiredzone.com. I thought I'd be waiting like 2 weeks for it because it is low/no stock.

I ordered the cables from Newegg, long story short paypal had an old cc on file and so my order went on hold without me noticing. So the cables and the 6 additional drives never shipped. Then I took that money and put together an Intel G45 micro ATX board to connect to my new 58 inch Samsung plasma. Right now I'm fighting with getting VMC (Vista Media Center) to recognize Video_TS folders.

I should have the card in by next week at the latest with one drive attached. I have my fingers crossed that I can run it in my PCIe slot that I normally use for graphics and just stick an old PCI video card in. It's an AMD 770 board.

Let me know what you find RE: those linux drivers, because I may want to go that route some day.

PS - the supermicro board itself looks to be of excellent quality. PCB labeling and what not are great. LED's are the icing on the cake.

Dan
 
For what its worth, here's a reply from Marvell ( with supermicro as the go between ) that i got 20min ago:

We do not provide the Debian driver binary, but the customer is free to use our open source driver for this kernel.

There are other customers who adopt this solution, so it is a known working configuration.

Its interesting. Every time you are faced with a dead end, getting angry speeds up the communication all off a sudden.

But, it did not exactly answer my questions. We know that the old mvsas does not support. We wanted to know about the new drivers for the 64bit controllers ( what this card is ).

This is the important link, and the answer to the questions asked ( yes, the customer needed to do the work in finding out ).Yes, the source driver is finally available on the svn: http://thread.gmane.org/gmane.linux.scsi/49940/focus=49957

A small tutorial i made for dan.b.mann, and others that like to not wast a lot of time on communicating with 3th party's ( and end up doing it yourself ;) ).

# Be sure that all the required files are available
apt-get install build-essential subversion module-assistant
module-assistant prepare

# Make a directory, and get the svn files
mkdir sas
cd sas
svn co https://scst.svn.sourceforge.net/svnroot/scst

# Remove the old driver ( or move out out of the way. Note the directory will differ based on your kernel version ).
rm /lib/modules/2.6.26-1-openvz-amd64/kernel/drivers/scsi/mvsas.ko

# Enter the directory to the correct subdir, and make the driver
cd /sas/scst/trunk/mvsas_tgt/
/sas/scst/trunk/mvsas_tgt# make

# copy over the new driver to the directory
cp mvsas.ko /lib/modules/2.6.26-1-openvz-amd64/kernel/drivers/scsi/

# Load up the driver, and verify that it works.
modprobe -v mvsas

# If you have drives attached, you will need to see new drives under the /dev/sd*

# Add the module to your kernel loading
vi /etc/modules

# Add the line
mvsas

# Save the file
:qw

#Thats about it...

Last time i wast my time waiting for answer from 3th party, its faster finding out things yourself. I hope my instructions are clear. I'm not exactly a linux genius, but the above will do the trick...

So yes, the HD's now show up. Finally...

Update ...

Noticed that the array sync was way below what it needed to be ( 30 to 140KB/s <- yes, you did read this correct ) for a 8 SATA drive set with raid 6 ...

hdparm -tT /dev/sdc:
Timing cached reads: 2 MB in 21.35 seconds = 95.90 kB/sec
Timing buffered disk reads: 2 MB in 22.37 seconds = 91.57 kB/sec
hdparm -tT /dev/sdd
... Still busy after 10min when i opened the case.

*uch* ... Great, more problem solving. Possible chipset overheating?
 
Last edited:
don't feel like reading this whole thread

what OS does this work with?
 
Yea I was being a smartass :p

It works with some peoples WHS....some people have had bad luck with hardware combos it seems and there is something going on with some bios configuration but I havent been paying too much attention to that.
 
I'm not reading 163 posts
does it work with WHS? yes or no

Well ... You won't get the answer from me... Using Linux ;)

This controller is a piece of ***** to get working properly on Linux ( Debian anyway ).

Connected to the AOC-SASLP-MV8 Controller:

/dev/sdc:
Timing cached reads: 2 MB in 21.45 seconds = 95.48 kB/sec
Timing buffered disk reads: 2 MB in 22.44 seconds = 91.25 kB/sec

Connected to the MB Controller ( system reboot ):

/dev/sdc:
Timing cached reads: 4546 MB in 2.00 seconds = 2274.25 MB/sec
Timing buffered disk reads: 368 MB in 3.02 seconds = 122.05 MB/sec

I know, hdparm is not a benchmark tool, but the results speak for them selfs...

Just a slight performance difference off ... heuuu ... 1000 times? :)

Note: Can't even get any smart information from it ( even when using the sas, ata overrides ) to check if its some dma problem or not.

Lets say, that after about 3 weeks, getting this **** to work, i feel like trowing it out of the window. And driving to the nearest computer shop, and getting a cheap intel or something controller...
 
For what it's worth, here is a other update.

Support from Supermicro/Marvell has stopped. Well, they do not bother answering anymore.

It seems that the "new" kernel driver does not play nice with this controller, resulting in a constant resetting of the connection. You know, up to now, will all the communication we have been sending to Supermicro / Marvell, there have been only 2 what can be considered "answers":

1. In response to why there are no Debian drivers ( yet there are redhat/suse drivers included ). And the questioning to where to find the complete source, and other information.

We do not provide the Debian driver binary, but the customer is free to use our open source driver for this kernel.
There are other customers who adopt this solution, so it is a known working configuration.

Yes, those two lines constitute the responds to a 3 page mail with details.

In the end, we needed to find out ourself crawling the web, if the new "working" code was online in the svn, etc. The old driver in the kernels up to 2.6.29 all will not work with this card! So you can only get the last version from svn, and compile it yourself.

Support: 1/10. Reason for the one, is they at leased answered.

2.

In questioning why the responding speeds are only in the 40 to 140KB range, when directly connected to a ICH10 controller, the drive's do 100+MB/s

Below is our test result for CentOS and SAS HDDs:

Linux Performance

Linux Edition CentOS 4.7 x86_64 Kernel Version 2.6.9-78.Elsmp
Linux Test SW Bonnie++ 1.03a

Ref.6A3B6

I. SINGLE HDD

|CUT|

You can see that our seq read and write above.

Below is our test result in RH and SATA HDDs:

Linux Performance

Linux Edition RHEL5U2 x86_64 Kernel Version 2.6.18-92.el5

Linux Test SW Bonnie++ 1.03a

Ref.3D2B6

I. SINGLE HDD

|CUT|

Tx,

-Farid

And some stats from both tests...

Support: 0/10. This is not support. As you can tell, a response that can only be considered "cover our asses". Any technical support worth there money, will have asked for log dumps, etc... All this showed that on there specific versions, that it worked ( fun thing is, CentOs 2.6.9 kernel? ).

Trust me, except for the performance numbers, i have removed nothing from the communication... Rather meager again on a 2 page mail...

Anyway... Not to beat a dead horse to much. The driver is not worth production on a Debian based kernel.

The reason why the stats are in the KB range is the constant resetting of the driver when performing some actions, and generating a error. One's that happens, it keeps generating errors non stop until you unload the driver ( or reboot ).

In the end, we have spend maybe 30hours or more on getting this to work. We even ended up trying to debug the driver source... At that moment, we concluded "fuck this". We are not going to do Marvell's job...

So, we are dumping this worthless card. Personally, I have no fait into this card. Maybe in the future when the Linux kernel driver has gone true some more testing ( and bug fixing! ) it can be usable, but we are not going to wait a few months or more for that.

Supermicro? Well, at leased they tried to play courier between us & Marvell. But it was no level off support you expect for server hardware. They manufacture the pcb,

And Marvell? Well, to sound very impolite: Fuck them! I hate using terms like that, but talk about a company that does not like to communicate at all. And there is never going to come a piece off hardware manufactured ( and especially driver design ) from them coming in our servers again...

So, we have ordered a raid controller and said bye bye to this card. So this will probably be also my last update regarding this topic/card.
 
I'm not reading 163 posts
does it work with WHS? yes or no

It's a crap shoot. It may work great or may not work at all for you depending on your hardware configuration.
 
such a shame i think ill be giving this card a miss also then

are there any alternitives available within the same price per port ratio
 
such a shame i think ill be giving this card a miss also then

are there any alternitives available within the same price per port ratio

besides the PCI-X version, nope. This is really as cheap as it gets for port density.
 
Back
Top