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

Sgt_Strider said:
Can you explain why?

So you're saying that the only reason to get this card is if I want to use WHS?
Click to expand...

Or linux software RAID.
Or if you dont want RAID and just need more SATA ports.

If you want RAID5/6 then a RAID controller is better because it will have onboard IOP processor to handle the parity calculations and will perform better.
 
nitrobass24 said:
Or linux software RAID.
Or if you dont want RAID and just need more SATA ports.

If you want RAID5/6 then a RAID controller is better because it will have onboard IOP processor to handle the parity calculations and will perform better.
Click to expand...

What's the disadvantage of using this card with Linux software RAID 5 or 6?

Wouldn't a fast processor mitigiate any performance issues that a software RAID solution would have?
 
Sgt_Strider said:
What's the disadvantage of using this card with Linux software RAID 5 or 6?

Wouldn't a fast processor mitigiate any performance issues that a software RAID solution would have?
Click to expand...

Nothing, Linux software RAID works great and modern processors have plenty of umph.

Only draw back here is a usable driver didn't make it into the 2.6.32 kernel, hopefully it will be in 2.6.33. Or you can build your own from source.
 
I have two cards installed in my board: the gigabyte m57sli-s4, and both appear in the device manager. I don't have drives on all ports yet, but I don't forsee any issues.
 
For some reason I can't seem to set "Poll S.M.A.R.T Status" to "on" in the Marvell RAID Utility web page. I click on the "on" button and then click on "Modify" but it just goes back to "off" for some reason. Anyone experience this too?
 
[H]ydra;1035314194 said:
For some reason I can't seem to set "Poll S.M.A.R.T Status" to "on" in the Marvell RAID Utility web page. I click on the "on" button and then click on "Modify" but it just goes back to "off" for some reason. Anyone experience this too?
Click to expand...

its not a raid card though.

WHS seems to pull smart info just fine on my system.
 
I was wondering if anyone has gotten this to run reliably in Linux? If so, which kernel are you using?
 
Generally recommend this controller:

AOC-USAS-L8i.gif


http://www.supermicro.com/products/accessories/addon/AOC-USAS-L8i.cfm

Its PCIe x8 (2GB/s full duplex) and offers 8 SATA/300 ports - perfect for a ZFS array if you ask me.
Of course, you can mix both ports from this controller, and the ones on your chipset controller (6x SATA) to make one big array if you want.
 
sub.mesa said:
Generally recommend this controller:

http://www.supermicro.com/a_images/products/Accessories/AOC-USAS-L8i.gif[/im]

[url]http://www.supermicro.com/products/accessories/addon/AOC-USAS-L8i.cfm[/url]

Its PCIe x8 (2GB/s full duplex) and offers 8 SATA/300 ports - perfect for a ZFS array if you ask me.
Of course, you can mix both ports from this controller, and the ones on your chipset controller (6x SATA) to make one big array if you want.[/QUOTE]

Has absolutely nothing to do with this topic.
Also its UIO form factor so unless you have a UIO Supermicro case you are going to be using it wont fit.
Click to expand...
 
nitrobass24 said:
Has absolutely nothing to do with this topic.
Also its UIO form factor so unless you have a UIO Supermicro case you are going to be using it wont fit.
Click to expand...

it will fit, but you'd have make a bracket or not use one at all., and not use a card in the slot above it.
 
OmegaAvenger said:
it will fit, but you'd have make a bracket or not use one at all., and not use a card in the slot above it.
Click to expand...

I guess it depends on your definition of fit ;):p

For me if I have to mod it and/or the product removes other functionality/features of my overall design. It doesnt fit.
Especially when there are strikingly similiar products at same $$ that do "Fit" :)
 
nitrobass24 said:
I guess it depends on your definition of fit ;):p

For me if I have to mod it and/or the product removes other functionality/features of my overall design. It doesnt fit.
Especially when there are strikingly similiar products at same $$ that do "Fit" :)
Click to expand...

Well in this forum people have done crazy mods to fit less important stuff. I figured a bracket would be simple.
 
There is a new mvsas patch available. It fixes all of the old I/O issues from the series of 7 patches, and ACTUALLY fixes hotswap issues. Best of all, it's wrapped up into one nice clean patch, and applies cleanly on 2.6.32 (and probably others). One comment in the thread about some minor hotswap issues, but I tested it extensively and things that would have kernel paniced the old driver didn't cause the slightest issue this time even after dozens of tries. I did encounter the issue where smartctl had to be run with -T permissive, but whatever, that's not a big issue to me at all. I did a partial RAID6 resync with the new driver and it seems to fix all of the stability issues.

Here's the thread: http://comments.gmane.org/gmane.linux.scsi/57088
Here's the patch: http://cache.gmane.org//gmane/linux/scsi/57145-001.bin

put it in your 2.6.32 source folder and do a patch -p1 < whatever-you-saved-it-as , recompile, install, good to go!

Note:
-I'm using the supermicro card but I'm assuming it will work just fine on other cards as well
-This patch is VERY new and by the sounds of the thread will undergo some more improvements in the next few weeks as more people test it
 
so i put this card in my motherboard, just hangs the system at "verifying DMI pool".

Move from the x4 slot to the x16 graphics card slot, no issues.

very strange.
 
nitrobass24 said:
Has absolutely nothing to do with this topic.
Also its UIO form factor so unless you have a UIO Supermicro case you are going to be using it wont fit.
Click to expand...
Has absolutely everything to do with this topic.

And UIO is just inverted PCI-express; it will just fit the only problem is the bracket. So simply run without bracket and you got PCI-express x8 (2GB/s full duplex) card for 8 SATA devices that's capable of running as dumb HBA (non-RAID SATA controller).

So its the most perfect solution for ZFS or other software RAID.
 
sub.mesa said:
Has absolutely everything to do with this topic.

And UIO is just inverted PCI-express; it will just fit the only problem is the bracket. So simply run without bracket and you got PCI-express x8 (2GB/s full duplex) card for 8 SATA devices that's capable of running as dumb HBA (non-RAID SATA controller).

So its the most perfect solution for ZFS or other software RAID.
Click to expand...

I'll have to politely disagree. Look at the thread you posted a off-topic suggestion in. This thread is for discussing AOC-SASLP-MV8 not AOC-USAS-L8i.

UIO card is 23% more expensive then the native PCI-E card going by Froogle
UIO card blocks the slot above it, which can be a killer depending on setup

Why would you want to jury rig a more expensive card in place of a cheaper solution?
 
Because the SASLP is useless, if you want to run ZFS in FreeBSD.

The USAS card also packs double the bandwidth, and is supported by FreeBSD 8.0; so you'll be able to run the latest and greatest ZFS version 13 kernel level. With 8 drives you would have 2GB/s full duplex bandwidth.

The PCIe x4 interface of SASLP is okay too, but it wouldn't work properly with BSD so its not really an option i would consider.
 
Also, it wouldnt block any slot above it; the chips arent high enough for that. The only thing you need to do is remove the bracket; and insert without bracket. You don't need a bracket anyway since they pack internal Mini-SAS ports.
 
sub.mesa said:
Because the SASLP is useless, if you want to run ZFS in FreeBSD.

The USAS card also packs double the bandwidth, and is supported by FreeBSD 8.0; so you'll be able to run the latest and greatest ZFS version 13 kernel level. With 8 drives you would have 2GB/s full duplex bandwidth.

The PCIe x4 interface of SASLP is okay too, but it wouldn't work properly with BSD so its not really an option i would consider.
Click to expand...

But no one was talking about BSD, because the card this thread is discussing does not support it.

Thats great that the card has those features, maybe you should start a new thread about it;)

/Back on topic
 
After buying 2 of these for a file server I could not be happier with the performance in Debian with a patched kernel.

Code: 
fs1:/media/md5# mdadm --detail /dev/md5
/dev/md5:
        Version : 00.90
  Creation Time : Thu Feb  4 16:04:18 2010
     Raid Level : raid5
     Array Size : 6837319552 (6520.58 GiB 7001.42 GB)
  Used Dev Size : 976759936 (931.51 GiB 1000.20 GB)
   Raid Devices : 8
  Total Devices : 8
Preferred Minor : 5
    Persistence : Superblock is persistent

    Update Time : Sun Feb 28 13:16:36 2010
          State : clean
 Active Devices : 8
Working Devices : 8
 Failed Devices : 0
  Spare Devices : 0

         Layout : left-symmetric
     Chunk Size : 64K

           UUID : 4edceacd:ef6f851f:5df300a1:eb89baa4 (local to host fs1)
         Events : 0.4990

    Number   Major   Minor   RaidDevice State
       0       8       49        0      active sync   /dev/sdd1
       1       8       65        1      active sync   /dev/sde1
       2       8       81        2      active sync   /dev/sdf1
       3       8      145        3      active sync   /dev/sdj1
       4       8      129        4      active sync   /dev/sdi1
       5       8      113        5      active sync   /dev/sdh1
       6       8       97        6      active sync   /dev/sdg1
       7       8       33        7      active sync   /dev/sdc1
fs1:/media/md5# dd if=/dev/zero of=ddfile2 bs=8K count=1000000
1000000+0 records in
1000000+0 records out
8192000000 bytes (8.2 GB) copied, 40.643 s, 202 MB/s
fs1:/media/md5# dd if=ddfile2 of=/dev/null bs=8k
1000000+0 records in
1000000+0 records out
8192000000 bytes (8.2 GB) copied, 26.9377 s, 304 MB/s
fs1:/media/md5#

I found this forum great in trying to get it to work.

Thanks everyone
 
Installed Fedora 12 x86_64 (kernel 2.6.31.5)
native mvsas driver did not work
with mdadm --create:
Driver froze lighting all drive activity lights.
mdadm non responsive

with mdadm --assemble
ok with arrays created on a different controller

as jobd:
extremely slow performance

Upgrade kernel to 2.6.33
native 2.6.33 showed nearly the same behavior as before

apply patches to 2.6.33 driver
will now let me create and use array
however kernel threw a series of errors 10 on port 7 and one one port 9 within 2 seconds
kernel: ataN: translated ATA stat/err 0x01/04 to SCSI SK/ASC/ASCQ 0xb/00/00
kernel: ataN: status=0x01 { Error }
kernel: ataN: error=0x04 { DriveStatusError }

Continuing stress test will update with more information as it becomes available
Configuration Intel DQ35JO as proof of concept
will try on
Configuration Tyan s5211 with 2 aoc-saslp-mv8
 
Yon89 said:
After buying 2 of these for a file server I could not be happier with the performance in Debian with a patched kernel.

I found this forum great in trying to get it to work.

Thanks everyone
Click to expand...

Could you write a quick "how-to" on getting it to work.

Cheers :)
 
There is NO working driver for this in the mainline linux kernel as of 2.6.33. There are documented corruption, instability, and speed issues with the drivers in their current state.

You need to download your kernel source code, patch the driver with the latest patch, and then build/install the new mvsas module. You should consult your linux distro's documentation for how to do this. Hopefully this driver will be in the next kernel. There is fairly active discussion about it in the linux-scsi mailing lists, and there is work being done to make it even better. It looks like it will be merged into the kernel pretty soon but no firm date that I can find...

In the mean time, if you want to patch the module yourself, I had linked to it in my post here http://www.hardforums.com/showpost.php?p=1035354777&postcount=420 but it seems to be down. I've uploaded it to pastebin. Use download link here: http://pastebin.org/pastebin.php?dl=102760

EDIT: Here's one of the many discussion threads http://kerneltrap.org/mailarchive/linux-scsi/2010/2/16/6780753
 
As others have said mvsas driver does not seem to be reliable now.

MB Tyan S5211
2 AOC-SASLP-MV8
11 Hitachi 2TB drives
intended array of 12 (ordered 13, 2 DOA no spin up-questionable Hitachi QC)

Fedora 12 x86_64 (non functional mvsas)
upgraded to kernel 2.6.33 (non functional mvsas. but better than 2.6.31.5)

patched mvsas driver: http://cache.gmane.org//gmane/linux/scsi/57145-001.bin
thanks gaspimp: http://www.hardforum.com/showthread.php?t=1397855&highlight=mvsas&page=21

mostly functional. Will do everything but operate reliably under heavy load.
on average every 30 min, under load system would log a series of errors:
Mar 7 02:01:06 localhost kernel: sas: command 0xffff8801cf895900, task 0xffff880165a996c0, timed out: BLK_EH_NOT_HANDLED
Mar 7 02:01:06 localhost kernel: sas: command 0xffff8801cf895d00, task 0xffff880165a98e00, timed out: BLK_EH_NOT_HANDLED
Mar 7 02:01:06 localhost kernel: sas: Enter sas_scsi_recover_host
Mar 7 02:01:06 localhost kernel: sas: trying to find task 0xffff880165a996c0
Mar 7 02:01:06 localhost kernel: sas: sas_scsi_find_task: aborting task 0xffff880165a996c0
Mar 7 02:01:06 localhost kernel: sas: sas_scsi_find_task: querying task 0xffff880165a996c0
Mar 7 02:01:06 localhost kernel: sas: sas_scsi_find_task: task 0xffff880165a996c0 failed to abort
Mar 7 02:01:06 localhost kernel: sas: task 0xffff880165a996c0 is not at LU: I_T recover
Mar 7 02:01:06 localhost kernel: sas: I_T nexus reset for dev 0500000000000000
Mar 7 02:01:08 localhost kernel: sas: sas_form_port: phy5 belongs to port5 already(1)!
Mar 7 02:01:08 localhost kernel: sas: I_T 0500000000000000 recovered
Mar 7 02:01:08 localhost kernel: sas: sas_ata_task_done: SAS error 8d
Mar 7 02:01:08 localhost kernel: ata12: translated ATA stat/err 0x01/04 to SCSI SK/ASC/ASCQ 0xb/00/00
Mar 7 02:01:08 localhost kernel: ata12.00: device reported invalid CHS sector 0
Mar 7 02:01:08 localhost kernel: ata12: status=0x01 { Error }
Mar 7 02:01:08 localhost kernel: ata12: error=0x04 { DriveStatusError }
Mar 7 02:01:08 localhost kernel: sas: trying to find task 0xffff880165a98e00
Mar 7 02:01:08 localhost kernel: sas: sas_scsi_find_task: aborting task 0xffff880165a98e00
Mar 7 02:01:08 localhost kernel: sas: sas_scsi_find_task: task 0xffff880165a98e00 is done
Mar 7 02:01:08 localhost kernel: sas: sas_eh_handle_sas_errors: task 0xffff880165a98e00 is done
Mar 7 02:01:08 localhost kernel: sas: sas_ata_task_done: SAS error 8d
Mar 7 02:01:08 localhost kernel: ata14: translated ATA stat/err 0x01/04 to SCSI SK/ASC/ASCQ 0xb/00/00
Mar 7 02:01:08 localhost kernel: ata14.00: device reported invalid CHS sector 0
Mar 7 02:01:08 localhost kernel: ata14: status=0x01 { Error }
Mar 7 02:01:08 localhost kernel: ata14: error=0x04 { DriveStatusError }
Mar 7 02:01:08 localhost kernel: sas: --- Exit sas_scsi_recover_host

the good news:
with 11 Toshiba 2TB HDS722020ALA330 was able to get 780 MB/s sequential read from bonnie++.
Version 1.03e ------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
localhost.local 16G 89399 99 269773 28 134851 20 87214 96 786285 49 503.9 0
------Sequential Create------ --------Random Create--------
-Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 92 0 +++++ +++ 88 0 78 0 +++++ +++ 87 0
localhost.localdomain,16G,89399,99,269773,28,134851,20,87214,96,786285,49,503.9,0,16,92,0,+++++,+++,88,0,78,0,+++++,+++,87,0


don't know how applicable this is going to be to the final production driver.

the patches to smartmontools to set erc/tler/cctl are also functional on the hitachi drives:
http://www.csc.liv.ac.uk/~greg/projects/erc/
thanks r.gregory

At lower load, the driver seems ok. I don't know if the driver failures are due to processing multiple commands or just buggy.

has anyone tried the drivers included on the supermicro disk??
Marvel-odin-redhat-fc9_x86_64-3.1.0.7.tgz
 
I have installed an AOC-SASLP-MV8 on my system (arch linux amd64 2.6.33 with latest mvsas patch from Srinivas). I do not have any mdraid config, every drive has its own filesystem. Everything works fine, unless i try to boot with 8 disks. Then booting stalls during the mount disk phase and eventually i get an "unable to read superblock" message(curiously fscking all the disks seems to work just before that). When that happens, the blue activity light for the first drive is constantly on. This also happened with the stock 2.6.33 kernel.

With 7 or less disks everything works ok, except for this warning that does not seem to interfere with the operation of the card.

Code: 
------------[ cut here ]------------
WARNING: at drivers/ata/libata-core.c:5186 ata_qc_issue+0x31f/0x330 [libata]()
Hardware name:
Modules linked in: ipv6 hwmon_vid xfs exportfs jfs cpufreq_powersave cpufreq_ondemand fan powernow_k8 freq_table firewire_ohci psmouse firewire_core edac_core thermal i2c_nforce2 crc_itu_t serio_raw k8temp skge edac_mce_amd button evdev processor pcspkr fuse forcedeth sg i2c_core rtc_cmos rtc_core rtc_lib ext4 mbcache jbd2 crc16 dm_crypt dm_mod ohci_hcd sd_mod usb_storage ehci_hcd mvsas libsas sata_sil scsi_transport_sas pata_amd sata_nv usbcore sata_via ata_generic pata_via pata_acpi libata scsi_mod
Pid: 2920, comm: hddtemp Not tainted 2.6.33-ARCH #1
Call Trace:
 [<ffffffff810528e8>] warn_slowpath_common+0x78/0xb0
 [<ffffffff8105292f>] warn_slowpath_null+0xf/0x20
 [<ffffffffa002c14f>] ata_qc_issue+0x31f/0x330 [libata]
 [<ffffffffa0006fae>] ? scsi_init_sgtable+0x4e/0x90 [scsi_mod]
 [<ffffffffa0033cd0>] ? ata_scsi_pass_thru+0x0/0x2f0 [libata]
 [<ffffffffa00310c6>] ata_scsi_translate+0xa6/0x180 [libata]
 [<ffffffffa0000b10>] ? scsi_done+0x0/0x20 [scsi_mod]
 [<ffffffffa0000b10>] ? scsi_done+0x0/0x20 [scsi_mod]
 [<ffffffffa0034369>] ata_sas_queuecmd+0x139/0x2b0 [libata]
 [<ffffffffa00e4098>] sas_queuecommand+0x98/0x300 [libsas]
 [<ffffffffa0000c25>] scsi_dispatch_cmd+0xf5/0x230 [scsi_mod]
 [<ffffffffa0006ba2>] scsi_request_fn+0x322/0x3e0 [scsi_mod]
 [<ffffffff81061937>] ? lock_timer_base+0x37/0x70
 [<ffffffff811b6d0c>] __blk_run_queue+0x5c/0x150
 [<ffffffff811b0760>] elv_insert+0x140/0x230
 [<ffffffff811b08ba>] __elv_add_request+0x6a/0xc0
 [<ffffffff811bc960>] blk_execute_rq_nowait+0x60/0xb0
 [<ffffffff811bca31>] blk_execute_rq+0x81/0xf0
 [<ffffffff811bc7e8>] ? blk_rq_map_user+0x198/0x270
 [<ffffffff811c0b3b>] sg_io+0x28b/0x3d0
 [<ffffffff811c14c0>] scsi_cmd_ioctl+0x4d0/0x530
 [<ffffffff8115146b>] ? blkdev_get+0xb/0x10
 [<ffffffff811514dc>] ? blkdev_open+0x6c/0xc0
 [<ffffffffa00023cd>] ? scsi_ioctl+0x27d/0x390 [scsi_mod]
 [<ffffffffa0128cff>] sd_ioctl+0x8f/0xe0 [sd_mod]
 [<ffffffff811be0cf>] __blkdev_driver_ioctl+0x8f/0xb0
 [<ffffffff811be59e>] blkdev_ioctl+0x22e/0x820
 [<ffffffff8114fca7>] block_ioctl+0x37/0x40
 [<ffffffff81131978>] vfs_ioctl+0x38/0xd0
 [<ffffffff81131b20>] do_vfs_ioctl+0x80/0x560
 [<ffffffff81132081>] sys_ioctl+0x81/0xa0
 [<ffffffff8100a002>] system_call_fastpath+0x16/0x1b
---[ end trace 881aaf5eceedc267 ]---

SMART works ok, except for a WD drive that needs a -T permissive to work
Edit:
The other 6 drives on the SASLP are Seagate ST31500541AS and smartctl works without -T permissive on them

Code: 
root ~  #  smartctl -a -d ata /dev/sdk
smartctl 5.39.1 2010-01-28 r3054 [x86_64-unknown-linux-gnu] (local build)
Copyright (C) 2002-10 by Bruce Allen, http://smartmontools.sourceforge.net

=== START OF INFORMATION SECTION ===
Model Family:     Western Digital Caviar Green family
Device Model:     WDC WD15EADS-00S2B0
Serial Number:    WD-WCAVY1089106
Firmware Version: 01.00A01
User Capacity:    1,500,301,910,016 bytes
Device is:        In smartctl database [for details use: -P show]
ATA Version is:   8
ATA Standard is:  Exact ATA specification draft version not indicated
Local Time is:    Sun Mar 14 13:39:47 2010 EET
SMART support is: Available - device has SMART capability.
SMART support is: Enabled

Error SMART Status command failed
Please get assistance from http://smartmontools.sourceforge.net/
Register values returned from SMART Status command are:
ST =0x40
ERR=0x00
NS =0x14
SC =0xa7
CL =0x1f
CH =0xf8
SEL=0x40
A mandatory SMART command failed: exiting. To continue, add one or more '-T permissive' options.

I am now stressing the card, reading and writing data at the same time from every single disk on my system and everything seems to work ok (fingers crossed). The real test for me will be when I install an HP SAS expander that i am waiting for any day now.
 
Last edited:
A small correction here, the WD15EADS-00P8B0 drive is not on the controller but on a motherboard slot, so that explains why smartctl works without -T permissive. The other 6 drives on the SASLP are Seagate ST31500541AS and smartctl works without -T permissive on them too.
 
I have 2 of these cards, drop me a pm if you are interested.
(Bought for my home NAS, but at the end considered OpenSolaris...)
 
I've been struggling to get this card working reliably in one of my CentOS boxes for some time now. I've been running into the same mvsas issues as everyone else.

By chance, I found a RHEL/CentOS 5.3 and 5.4 driver on a Chinese ftp site (it's the "official" binary driver, although it's not provided on the Supermicro driver disc or ftp site). I've been having good luck with it so far (much more reliable than the stock mvsas driver -- although it does taint the kernel and prevent upgrading). Links and more info here: http://hardforum.com/showthread.php?p=1035728553
 
Last edited:
The mainline drivers are still not working. I do not believe the 2.6.34 drivers have changed. The last I saw Srinivas Naga Venkatasatya Pasagadugula who seems to be leading the charge is still trying to resolve the issues above with the driver in the mainline kernel. Godspeed.
http://kerneltrap.org/mailarchive/linux-scsi/2010/5/1/6884795


There is however a second path. There are a second unrelated set of scsi drivers which are kernel instead of being user space based. stgt is the mainline methodology with non functional mvsas drivers. scst contains an mvsas driver which appears to work.

scst
http://scst.sourceforge.net/index.html

without getting into the debate of the relative benefits of either method, I can say that the mvsas driver appears to work.

main reference
http://iscsi-scst.sourceforge.net/iscsi-scst-howto.txt

howto, cookbook
I have tested for a month without disk errors. My svn image is from 4/29. There are some changes in the new svn image. I have not tested them. Try at your own peril. Good luck.
base system distribution fedora 12
assumes you can compile the kernel

0. download the 2.6.33 kernel - compile it make sure it is working.

1. download and compile scst
mkdir scst
cd scst
svn co https://scst.svn.sourceforge.net/svnroot/scst/trunk

2. install scst kernel patches, they claim that these are optional - have not tested without them
cd /usr/src/linux
patch -p1 <path to scst trunk>/iscsi-scst/kernel/patches/put_page_callback-2.6.33.patch
patch -p1 <path to scst trunk>/scst/kernel/scst_exec_req_fifo-2.6.33.patch

3. recompile kernel

4. compile scst
cd <path to scst trunk>
make scst
make scst_install
make scstadm_install (this gave me errors which were ignored -- an installation subsystem is missing on fedora 12)
make iscsi
make iscsi_install
make srpt
make srpt_install
make usr
make usr_install
make scst_local
make scst_local_install

5. build mvsas driver -- there is change in the sas_change_queue_depth definition from 2 to 3 parameters.
http://lime-technology.com/forum/index.php?action=printpage;topic=3382.0
Modify the file <path to scst trunk>mvsas_tgt/mv_sas.c line 423, which reads:
sas_change_queue_depth(sdev, MVS_QUEUE_SIZE);
should be changed to:
sas_change_queue_depth(sdev, MVS_QUEUE_SIZE,SCSI_QDEPTH_DEFAULT);

make mvsas
make mvsas_install

6. make sure you rebuild your init ram disk so that it does not include the mainline driver

7. reboot


RESULTS:

If the correct drive is loaded, you will see in the output of lsmod
lsmod shows scst loaded and mvsas sized 70614. The mainline kernel driver is around 50k
mvsas 70614 16
libsas 49801 1 mvsas
scst 158216 1 mvsas
scsi_transport_sas 26524 2 mvsas,libsas

PERFORMANCE:

500 random iops
read 1.09 GB/sec!
write 349 MB/sec

raid 6, 16 2tb drives
Version 1.03e ------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
localhost.local 16G 89775 99 349372 37 168696 24 81103 98 1091868 63 500.5 0
------Sequential Create------ --------Random Create--------
-Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 218 2 +++++ +++ 218 2 201 2 +++++ +++ 184 1
localhost.localdomain,16G,89775,99,349372,37,168696,24,81103,98,1091868,63,500.5,0,16,218,2,+++++,+++,218,2,201,2,+++++,+++,184,1
 
Hi,

i i really appreciate your cookbook, but when i try to load the modules i get the following errors:

modprobe mvsas
WARNING: Error inserting libsas (/lib/modules/2.6.34-custom/kernel/drivers/scsi/libsas/libsas.ko): Unknown symbol in module, or unknown parameter (see dmesg)
FATAL: Error inserting mvsas (/lib/modules/2.6.34-custom/extra/mvsas.ko): Unknown symbol in module, or unknown parameter (see dmesg)

dmesg:

[ 7456.949539] scst: Unknown symbol blk_rq_map_kern_sg
[ 7456.950665] scst: Unknown symbol blk_rq_unmap_kern_sg
[ 7456.950849] scst: Unknown symbol sg_copy

Any clue what i can do to make this work ?

Thanks in advance !

Dodg
 
You appear to be using 2.6.34. I don't think they have patches for 2.6.34 yet. I used the 2.6.33 kernel.

I started with fedora 12 as a base.
Not that this would have anything to do with the undefined symbols; but, my drivers are placed in /lib/modules/2.6.33/extra, one level up from yours
 
With the scst mvsas driver, each device will issue the HDIO_GET_IDENTITY failed error.
This is more of a warning than an error.
http://lime-technology.com/forum/index.php?topic=5739.0

This error seems to prevent the automatic assembly of the mdadm array during boot. The work around is to modify your rc.local file to:

assemble the array
mount the array
restart any services which depend upon it

I know this is ugly but it works.
 
There are patches for the recent kernels in the scst source. I Applied them without errors and rebuild my kernel successfully.
One last hope may be the last 2.6.35 rc kernel. Obviously some patches have find a way into that kernel for the mvsas driver.
 
Last edited:
From what i can see 2.6.35-rc2 mainline has the mvsas patch's, I am going to give it a go tonight. will let you know how it goes.
 
You must log in or register to reply here.
Back
Top