[jwhat]

Hi SGI'er,

I know this has been covered before in Nekochan, but can someone (Jan-Jaap, Hamei, Ian ??) advise on whether anyone has found SGI / IRIX Firmware for LSI 1068 PCI-X SAS/SATA Card.

I have two LSI SAS3442X-R boards, one in O350 and one I have just installed in Fuel,

The fuel complains on boot about the firmware version installed on the LSI board:



>> Sep 29 21:11:18 1A:pink unix: ALERT: /hw/module/001c01/Ibrick/xtalk/14/pci/2/scsi_ctlr/0:  Old SAS firmware.  Download new fw with lsflash

As others have found there is no 1068 firmware in /usr/firmware/ls .

So is there a patch (other than patchSG0007209 which has firmware for LSI 1030 & 949) which has SGI / IRIX preferred firmware ?

Needless to say I got this card for $30 on eBay and it had very old firmware and even complained when put into PC that it had "Beta Software" installed.

The Broadcom Support site does not appear to have firmware for a card this old (mine is reporting as LSI 1068(A0) ) and I cannot find any more recent A0 version firmware (just BX & CX versions).

Advise from those who have been down this route before would be helpful.

Cheers from Oz,



John.

[jan-jaap]

This was discussed on Nekochan, and it was captured

https://gainos.org/~elf/sgi/nekonomicon/...486/3.html

2nd post in the thread:

1. The 1068 A0 rev is indeed a prototype chip, can't run new versions of firmware. I would not suggest buying any cards without knowing it had an A3 rev or greater 1064/1068 chipset, the previous revs were protos and were never released to the public.
2. Most of the SAS3041X-R boards I've seen have the A3 rev chip on them. (These are all low profile cards and come with a low profile bracket or full height bracket.
3. There was an issue with SATA dvd drives that I believe was fixed in the newest versions of the firmware for the 1064/1068 A3 and 1064e/1068e (pci express versions of the same chip) B2 & B3. No idea if this would fix the problem in IRIX though.
4. There should be no problems using the firmware posted on the LSI website for the Sun & HP OEM versions of the cards as long as you pick the correct chipset type, you can flash in a PC booted into DOS with the files on the LSI website. These flash updates also update the fcode for using the cards in a Sun SPARC system.

It looks like you're out of luck with A0 silicon.

To the best of my knowledge, SGI never published firmware for the 1064/1068 SAS cards, only for the FC version (949X). Then again, I don't think the firmware is SGI specific. What sometimes works is looking for OEM-ed versions of a card, e.g. from HPE or IBM. But with remark #1 above I think I'd just bin it.

[jan-jaap]

So I was at the Broadcom site for something else and dug around for the SAS3442X-R as well.

I can't find a product support page anymore (it's retired), but it's possible to find individual download links:

[Unknown filetype] SAS3442XR_P20_Supports_SAS_1-5G_3G_6G_and_SATA_1-5G_3G_HDD

and

[Unknown filetype] SAS3442XR_P20_Supports_SAS_1-5G_3G_6G_and_SATA_3G_6G_HDD

So, not only does newer firmware exist, it seems you need to choose a firmware depending on your (SATA) disks. I downloaded the firmware and can confirm that (1) the firmware inside both zip files is different and (2) there's firmware inside for B0 and B1 silicon, but not A0.

Finally: I don't know what firmware the SAS3442X-R in my O350 has, but I have no desire to flash it. Maybe one day if I feel like striping a couple of SATA3 SSDs. 

Edit: this is only for the 3442X-R, there are other downloads for the 3800X etc. Search and you'll find.

[jwhat]

Hi Jan-Jaap,

thanks very much for weighing in on this one.

Your Nekochan posting on putting SAS3442X-R into O350 motivated me to put one into my O350.
It was a really nice job.

I also did some further digging around and as you say there is so much stuff on the "revised" BroadCom support page is is hard to know where to start...

I downloaded the following:
- Installer_P19_for_Linux
- SAS3442XR_ Package_P21_IR_IT_Firmware_BIOS_for_MSDOS_Windows
- SAS3442XR_P20_Supports_SAS_1-5G_3G_6G_and_SATA_1-5G_3G_HDD

I used the Installer for LINUX and that got me to point of getting diagnostics etc but none of the firmware packages have AO version firmware.

So I next tired the SuperMicro site which had various LSI image and managed to get the BIOS updated (removing the "Beta" software message that come up on boot) but could not find any A0 firmware again.

So backtracking I decided to see what HP had, as I come across a posting that HP frequently changed LSI boards within its product life-cycle without any change in part no.
The result was I found reference to an HP Support Pack (SP45154) that had a A0 firmware version for what HP sold as "HP 8 Internal Port SAS HBA with RAID".

This pack came with DOS sasflash tool, so I had to dig out old MS DevNet CDs with MS-DOS 6.22 floppy images and created a virtual floppy via VM which I then dd'ed over to USB floppy drive.

Result is I have board with firmware 1.23.46.0 which seems to be acceptable for the Fuel as it does not complain about firmware version any more.

Getting to this point has taken a lot of web surfing and reading up on LSI SAS3442X-R boards and software, but now at least I know a bit about it ;-)

I have yet to connect a disk to board, as I have spent some considerable time just playing with setting up a flashing machine (I got an old HP DL380 server off Ebay for AUD $60), getting firmware, installing linux, msdos and testing.

The funny thing is that it looks like HP never released 442 variation and so my board reports now as SAS3080 (8 internal).

I did see that there is a file that allows you to provide and override the SW configuration details that get loaded into NVP via "sasflash -o -concat IN.fw SPEC.dat OUT.fw" but I could not get this to work so it reported the correct 442 variation.

Also with this firmware variation you change whether it should act as RAID or HBA via the BIOS boot configuration rather than like in later BX / CX variations where you load in either RAID (-R) or IT (Initiator Target) (-IT) Firmware variations to manage mode of operation.

So it will be interesting to see if it works on Fuel or not. Here is boot into single user mode with 1068 Firmware version notice:




Because the SNAPHAT is dead I have to boot into single user mode and reset the date/time every time after I unplug Fuel from the power.
I have ordered enough SNAPHATs to go into all my O350 Chassis and the Fuel.

I will report back once I put disk in and test...

BTW I raised return request from Ebay seller as this board is obviously not right as you say and got refund.

I will use the $$ to find a BX / CX variation, which should avoid the firmware issues.

Cheers from Oz,


John.

[jwhat]

Hi Jan-Jaap and other Fuel/O350 Owners,

I have some good news 

I pulled the SAS3442X-R that was in my O350 and plugged it into Fuel, where it reported as having Firmware 1.26.

I then pulled it back out and did Firmware / BIOS flash to update it to IT (HBA) based versions from the Broadcom support site P21 package with: 1.33 FW / 6.36 BIOS.

Flashing from the RAID to IT (HBA) version, requires that you first erase the ROM which is only possible with MS-DOS version of sasflash utility.

NOTE: Erasing ROM also erases the SAS Address (so make sure you save this before flashing).
So once you update fw/bios you need to restore the SAS Address with: "sasflash -o -sasadd XXXXX" .

I got this version of SAS3442X (reports as "1068(B1)") as an "new old stock" retail package, so it came with all the required cabling (SAS/SATA & MOLEX) which means it was trivial to put into the Fuel.

The results are excellent. Here are at set of diskperf result, showing the SAS performance relative to an SSD based Octane and Spinning Disk Fuel

>> ---
>> --- 1. Here is diskperf of in built Ultra 160 SCSI boot disk
>> ---      with a Seagate Cheetah U160 spinning disk
>> ---
>> 
>> # diskperf -W -D -c4g -n "fuel/pink UW160 ST336706LW" testfile
>> #---------------------------------------------------------
>> # Disk Performance Test Results Generated By Diskperf V1.2
>> #
>> # Test name     : fuel/pink UW160 ST336706LW
>> # Test date     : Sat Oct  3 22:52:00 2020
>> # Test machine  : IRIX64 pink 6.5 07202013 IP35
>> # Test type     : XFS data subvolume
>> # Test path     : testfile
>> # Request sizes : min=16384 max=4194304
>> # Parameters    : direct=1 time=10 scale=1.000 delay=0.000
>> # XFS file size : 4294967296 bytes
>> #---------------------------------------------------------
>> # req_size  fwd_wt  fwd_rd  bwd_wt  bwd_rd  rnd_wt  rnd_rd
>> #  (bytes)  (MB/s)  (MB/s)  (MB/s)  (MB/s)  (MB/s)  (MB/s)
>> #---------------------------------------------------------
>>       16384    2.60   52.58    2.88    2.89    2.32    2.45
>>       32768    4.98   52.84    6.08    6.10    4.38    4.62
>>       65536    9.14   52.85   13.70   13.72    8.03    8.32
>>      131072   15.68   52.51   31.43   28.23   14.27   14.76
>>      262144   24.39   51.92   36.51   36.63   22.76   22.36
>>      524288   33.61   51.29   36.42   36.74   31.47   31.79
>>     1048576   41.44   51.62   46.25   41.76   39.28   37.59
>>     2097152   46.84   51.23   46.36   46.41   45.73   43.57
>>     4194304   50.30   50.96   49.32   49.69   49.13   47.92
>>     
>> ---
>> --- 2. And here is an Octane2 with an ACARD SAS/SATA adaptor
>> ---      with a Samsung 850 EVO SATA SSD
>> ---
>> 
>> # diskperf -W -D -c4g -n "octane2/porcipine scsi/sata/acard 850 EVO" testfile
>> #---------------------------------------------------------
>> # Disk Performance Test Results Generated By Diskperf V1.2
>> #
>> # Test name     : octane2/porcipine scsi/sata/acard 850 EVO
>> # Test date     : Sat Oct  3 22:32:06 2020
>> # Test machine  : IRIX64 porcipine 6.5 07202013 IP30
>> # Test type     : XFS data subvolume
>> # Test path     : testfile
>> # Request sizes : min=16384 max=4194304
>> # Parameters    : direct=1 time=10 scale=1.000 delay=0.000
>> # XFS file size : 4294967296 bytes
>> #---------------------------------------------------------
>> # req_size  fwd_wt  fwd_rd  bwd_wt  bwd_rd  rnd_wt  rnd_rd
>> #  (bytes)  (MB/s)  (MB/s)  (MB/s)  (MB/s)  (MB/s)  (MB/s)
>> #---------------------------------------------------------
>>       16384   23.72   20.57   23.65   19.16   23.67   19.08
>>       32768   29.38   24.93   29.36   23.86   29.34   23.73
>>       65536   33.03   29.02   33.02   28.26   33.01   28.15
>>      131072   35.43   30.43   35.40   29.61   35.43   29.49
>>      262144   36.94   31.26   36.93   30.41   36.92   30.32
>>      524288   37.67   31.53   37.66   30.66   37.67   30.56
>>     1048576   37.99   31.43   37.98   30.81   37.98   30.79
>>     2097152   38.18   31.00   38.14   30.55   38.18   30.67
>>     4194304   38.27   30.04   38.28   29.83   38.28   29.89
>>     
>> ---
>> --- 3. Now here is the LSI Logic SAS3442X-R with
>> ---       Samsung 840 EVO SATA SSD
>> ---
>> 
>> # diskperf -W -D -c4g -n "fuel/pink sas3442X-I 840 EVO" test/testfile
>> #---------------------------------------------------------
>> # Disk Performance Test Results Generated By Diskperf V1.2
>> #
>> # Test name     : fuel/pink sas3442X-I 840 EVO
>> # Test date     : Sat Oct  3 22:23:11 2020
>> # Test machine  : IRIX64 pink 6.5 07202013 IP35
>> # Test type     : XFS data subvolume
>> # Test path     : test/testfile
>> # Request sizes : min=16384 max=4194304
>> # Parameters    : direct=1 time=10 scale=1.000 delay=0.000
>> # XFS file size : 4294967296 bytes
>> #---------------------------------------------------------
>> # req_size  fwd_wt  fwd_rd  bwd_wt  bwd_rd  rnd_wt  rnd_rd
>> #  (bytes)  (MB/s)  (MB/s)  (MB/s)  (MB/s)  (MB/s)  (MB/s)
>> #---------------------------------------------------------
>>       16384   96.89  105.67   94.87   63.93   94.41   63.93
>>       32768  138.93  150.77  129.34  103.63  135.62  102.92
>>       65536  180.30  194.71  136.40  149.68   59.84  149.72
>>      131072  205.64  225.30  148.95  191.94   61.98  191.70
>>      262144  224.77  245.83  132.71  224.31   57.87  224.69
>>      524288  228.81  258.08  131.87  246.02   59.23  246.12
>>     1048576  224.70  264.18  111.43  257.66   59.37  257.81
>>     2097152  217.02  267.26  109.98  264.21   57.54  264.55
>>     4194304  179.81  268.91  135.53  267.42   57.62  267.30

So SAS3442X (-IT Flashed) works very well in Fuel and provides much better performance than the spinning U160 SCSI disk.
I have put more details on MS-DOS flashing steps on my blog.

This also confirms what Ian Mapleson advised me of a while ago, that using newer 1608 board with updated firmware allows you to get faster than 1.5 Gb/sec SATA throughput.
This makes for a very cheap disk performance boost (shame this is not a bootable disk through).

Cheers from Oz,


John.

[Irinikus]

Wow!!! That's impressive!!!

[jwhat]

Wow!!! That's impressive!!!

Hi Irinikus,

You should be able to do this easily with Tezro as well I believe.


Cheers,

John.

[Raion]

I tried running SAS but the restriction of it not being a boot disk means that I'm just better off using NFS for the main storage medium of my machines.

[jwhat]

Hi Raion,

yes likewise, I have non SGI machines providing: NFS, CIFS, Rsync, AppleTalk & Time Machine backup for Macs.

I have mostly used Octane lately to provide rsh server for network IRIX Install and IRIX software collection nfs, but now I have Fuel I think maybe I will use it for this purpose and just rsync back to Octane to make sure my IRIX software collection is is being protected with some degree of automation and regularity.

Next for Fuel is:
 - 1. get new Snapcat timer chip, see if this helps with the issues I am having with L1 comms (as per this thread: https://forums.irixnet.org/thread-2477.html )
 - 2. possibly reconfigure storage on Fuel so it has / on disk with U160 interface and /usr on SAS3442X with SATA and then use it as IRIX boot server

Glad this thread has come to happy end ;-)

Cheers from Oz,

John.

[jwhat]

Hi SGI'ers,

final update on this one...

I pulled second Acard ARS-2320S from O350 as this had IRIX boot disk for Onyx 4 ATI Graphics configuration, which for sure I am never going to use again, best just to stick with V12.

As this has SCA 80 pin connector, I again had to wait 2 weeks for shipping from China to get SCA (80 pin) to U160 68 pin converter.

So reconfigured Fuel to boot off Acard based SSD and also consolidated everything onto single SSD disk (usual fx, dvhtool & xfsdump/xfsrestore system clone)  as the fuel came with 2 spinning SCSI disks.

Run diskperf as I was curious if U160 bus with SSD's would get near SAS3442 performance.

Here are results:

>> # diskperf -W -D -c4g -n "fuel/pink UW160 ACARD/SATA/SAMSUNG EVO 850" testfile
>> #---------------------------------------------------------
>> # Disk Performance Test Results Generated By Diskperf V1.2
>> #
>> # Test name     : fuel/pink UW160 ACARD/SATA/SAMSUNG EVO 850
>> # Test date     : Fri Oct 16 16:31:13 2020
>> # Test machine  : IRIX64 pink 6.5 07202013 IP35
>> # Test type     : XFS data subvolume
>> # Test path     : testfile
>> # Request sizes : min=16384 max=4194304
>> # Parameters    : direct=1 time=10 scale=1.000 delay=0.000
>> # XFS file size : 4294967296 bytes
>> #---------------------------------------------------------
>> # req_size  fwd_wt  fwd_rd  bwd_wt  bwd_rd  rnd_wt  rnd_rd
>> #  (bytes)  (MB/s)  (MB/s)  (MB/s)  (MB/s)  (MB/s)  (MB/s)
>> #---------------------------------------------------------
>>       16384   42.49   43.17   42.33   37.55   42.48   37.52
>>       32768   61.65   59.98   61.64   54.22   61.60   53.74
>>      65536   79.52   85.20   79.46   78.93   79.36   78.15
>>      131072   85.45   94.65   85.44   86.95   85.38   86.16
>>      262144   87.64  100.91   87.67   92.18   87.68   91.45
>>      524288   88.07  103.34   88.09   93.01   88.09   92.71
>>     1048576   87.38  103.38   87.38   93.70   87.37   94.86
>>     2097152   85.32  100.92   85.33   97.13   85.32   96.65
>>     4194304   81.13   95.31   81.14   93.11   81.13   93.28

These result are pretty much consistent with those on Ian Mapleson's  fuel page, for 15K RPM SCSI disk (with exception of smaller random reads, where SSD performance significantly better).

So Acard ARS-2320S (connected to U160 bus) with SATA SSD does not provide the equivalent performance boost, that I saw with Octane.

Based on this I reconfigured machine again so it now has: / (root) on ACARD/SATA SSD and /usr with SATA SSD on SAS3442, as it would seem remise not to take advantage of the significant performance improvement provided by SAS3442.

The interactive performance of my fuel now feels almost as snappy as my Mac ;-)

And the pile of retired SCSI HDs continues to grow.


Cheers from Oz,


John.