Cheetah under linux
Alex J Burke
Hope you are all well.
I have spoken to many people that feel that if we had a linux driver for the
Cheetah under linux (Passplay may come under this heading also), then it
would be much more useable for them. I myself have had to turn away from
using linux because it will not go on my RAID systems.
Firstly, I wonder if I could clarify some things:
1) Firstly, everybody says that if we had the newer firmware on the Cheetah,
then the current linux driver would be able to use it. What is different
about the newer firmware?
2) Apart from the IBM adapters, I believe there were some other Mylex own
brand MCA adapters. Do these work with the current linux driver?
3) Is this purely a firmware issue, or is the current linux driver also
inadvertently designed for PCI only Mylex adapters?
It has occured to me that maybe the easiest way to a good working driver
would be to work backwards from the current linux driver. In which case, I
ask people to share any info they may have about what changed so we can
document it and maybe kindly ask one of our members to help with the coding
itself. If anyone has any technical references that may be of use about
Cheetah (and maybe one of the later adapters for comparison). That type of
stuff :-)
Lastly, I thought of writing a new site devoted to MCA Linux (maybe someone
can tell me where to contact the creator of the original website so I can
see if I could maybe update his pages and put them up on the inet). Would
there be any support for this?
Thanks, Alex Burke.
Laust Brock-Nannestad
> 1) Firstly, everybody says that if we had the newer firmware on the Cheetah,
> then the current linux driver would be able to use it. What is different
> about the newer firmware?
Paging Peter, I think :)
But just having a newer firmware is not quite enough. Read below.
> 2) Apart from the IBM adapters, I believe there were some other Mylex own
> brand MCA adapters. Do these work with the current linux driver?
No.
> 3) Is this purely a firmware issue, or is the current linux driver also
> inadvertently designed for PCI only Mylex adapters?
It runs PCI and EISA adapters. Given that no MCA adapters are supported, the
driver doesn't even _look_ for them. You will need to add MCA support to the
driver before it will find the adapter, even if the firmware was supported.
(This probably isn't a major problem, however)
> It has occured to me that maybe the easiest way to a good working driver
> would be to work backwards from the current linux driver. In which case, I
> ask people to share any info they may have about what changed so we can
> document it and maybe kindly ask one of our members to help with the coding
I agree. I think the best would be to get technical docs for one of the MCA
adapters and then for the DAC960-P, the smallest PCI based adapter (with
the lowest supported firmware) and then see how they compare. Unfortunately,
IBM doesn't seem to be too willing to give out such information (I recall
Alfred Arnold trying to get hold of documentation for the sadly unsupported
Etherstreamer), but maybe Mylex (now owned by IBM, but still) would be more
helpful?
Regards,
Laust
shinguz
of the MCA
> adapters and then for the DAC960-P, the smallest PCI based adapter
(with
> the lowest supported firmware) and then see how they compare.
This is somewhat of a pain, in my mind.The DAC960 driver suppports
firmware 2.73 that is only available from Mylex while the latest IBM
firmware 2.43 is not supported. I have hacked the firmware revision
control on the drivers to support 2.43, but the driver refuses to
work. Could be that my C is at such a rudimentary level that I
missed something, or the firmware itself has something weird on it.
The other thing is that I have not been able to get my
MCA520/DAC960P-2 combination to boot off the array with any
operating system while running firmware 2.73. It does boot NT and
SCO nicely when firmware 2.43 is installed, so I suspect that IBM
has done something to the firmware in order for it to work on MCA.
--
Hit reply to experience the cybervoid
shinguz <at> phreaker <dot> net
Alex J Burke
I am glad you agree with my working down proposition. I must admit that one
of the reasons I posted this article is that maybe someone has the technical
docs of the Cheetah/Passplay (very unlikely) or even the Mylex own brand MCA
RAID adapter. This seems a more likely bet, Mylex probably arent so fussed
about people getting tech docs as IBM seem to be (hard to find, I hear).
Then, just as we said, we would compare the MCA RAID to a newer Mylex DAC960
PCI based card with correct firmware and make the appropriate corrections to
the driver.
Thanks for confirming my thought that it would not really look for an MCA
adapter, but in a way that could be seen useful because then any driver we
make will be the ole thing working with MCA RAID - that means no driver
conflicts. Maybe our eventual goal should be to have a seperate MCA RAID
linux driver, but obviously developed from the PCI/EISA one?
Thanks for replying, Alex J Burke.
Alex J Burke
You troubles seem to suggest that IBM has done something to
firmware/hardware to make it work better/work with MCA altogether.
Could you just clarify -- are you trying to use a PCI Mylex adapter in your
PC Server 520MCA or a Mylex own brand MCA adapter?
Thanks, Alex Burke.
shinguz
adapter in your
> PC Server 520MCA or a Mylex own brand MCA adapter?
Sure. The adapter is the original PCI version that came with the 520
to begin with. I believe IBM calls it Savannah. Two internal scsi2
ports, one external, 4M cache and an Intel 960 processor. Raid, of
course :)
It's a run off the mill pci Mylex DAC960P-2 with a small thingy
plugged into one of the small ports on the board. Says something
like LBA or something on the plug-like thing IIRC.
The adapter accepts both the IBM supplied firmwareflash 2.43 as well
as the Mylex-supplied 2.73, configures with both but boots only with
the IBM version.
The box in question is a pci/mca 520 8641-mze, upgraded with a
second cpu and some additional memory, can't remember the exact
types and amounts at right now.
William R. Walsh
> but maybe Mylex (now owned by IBM, but still) would be more
> helpful?
Mylex is now part of LSI Logic, so maybe our odds just improved?
William
Jim Shorney
>
> It has occured to me that maybe the easiest way to a good working driver
> would be to work backwards from the current linux driver. In which case, I
> ask people to share any info they may have about what changed so we can
> document it and maybe kindly ask one of our members to help with the coding
> itself. If anyone has any technical references that may be of use about
> Cheetah (and maybe one of the later adapters for comparison). That type of
> stuff :-)
Contact Michael Lang, he has done some work on this.
http://www.uni-mainz.de/~langm000/linux.html
-----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
-----== Over 80,000 Newsgroups - 16 Different Servers! =-----
Laust Brock-Nannestad
> Mylex is now part of LSI Logic, so maybe our odds just improved?
Oh, LSI not IBM. I guess I remembered incorrectly.
Regards,
Laust
William R. Walsh
No, you were right. The change in the ownership of Mylex has been recent. I
only noticed it out of the corner of my eye after visiting their website.
William
Unal Z
> > inadvertently designed for PCI only Mylex adapters?
>
> It runs PCI and EISA adapters. Given that no MCA adapters are supported,
the
> driver doesn't even _look_ for them. You will need to add MCA support to
the
> driver before it will find the adapter, even if the firmware was
supported.
> (This probably isn't a major problem, however)
One has to simply replace the EISA detect/config logic with that of the MCA.
This isn't a big deal, nor so difficult, changes will be required in one or
two procedures (so I assume) and after that all can proceed as usual
provided it is only the MCA adapter interface that is different than that of
the PCI/EISA variants.
The procedure: make sure you are on MCA, query the slots to detect the
adapter (or use eventually available system config info), read out the
needed number of POS bytes (as defined in the ADF), interpret the byte
values and feed them to the driver routines. See below the AHA-154x NT
driver example, incorporating the AHA-1640 detection and interpetation
logic.
QUMC works similarly to show the adapter configuration but it stops after
interpreting the bytes, that is, 'converting' them to meaningfull text
messages.
Here are some codelets from the AHA NT drivers, "to whom it may be useful":
// MCA specific definitions.
#define NUMBER_POS_SLOTS 8
#define POS_IDENTIFIER 0x0F1F
// ** Ed: 0F1F Adaptec AHA-1640 SCSI Host Adapter
#define POS_PORT_MASK 0xC7
#define POS_PORT_130 0x01
#define POS_PORT_134 0x41
#define POS_PORT_230 0x02
#define POS_PORT_234 0x42
#define POS_PORT_330 0x03
#define POS_PORT_334 0x43
typedef struct _POS_DATA {
USHORT AdapterId;
UCHAR BiosEnabled;
UCHAR IoPortInformation;
UCHAR ScsiInformation;
UCHAR DmaInformation;
} POS_DATA, *PPOS_DATA;
typedef struct _INIT_DATA {
ULONG AdapterId;
ULONG CardSlot;
POS_DATA PosData[NUMBER_POS_SLOTS];
} INIT_DATA, *PINIT_DATA;
...................................................
if (Mca == TRUE) {
INIT_DATA initData;
LONG slot;
LONG i;
for (slot = 0; slot < NUMBER_POS_SLOTS; slot++) {
i = ScsiPortGetBusData(HwDeviceExtension,
Pos,
0,
slot,
&initData.PosData[slot],
sizeof(POS_DATA));
if (i < (sizeof(POS_DATA))) {
initData.PosData[slot].AdapterId = 0xffff;
}
}
for (slot = 0; slot < NUMBER_POS_SLOTS; slot++) {
if (initData.PosData[slot].AdapterId == POS_IDENTIFIER) {
switch (initData.PosData[slot].IoPortInformation &
POS_PORT_MASK) {
case POS_PORT_130:
if (IoPort == 0x0130) {
return TRUE;
}
break;
case POS_PORT_134:
// Ed: .... same as above, below too ...
case POS_PORT_230:
//.....................
case POS_PORT_234:
//........................
case POS_PORT_330:
//....................
case POS_PORT_334:
if (IoPort == 0x334) {
return TRUE;
}
break;
}
}
}
return FALSE;
}
This one utilizes only the IoPort byte, but the routine shows how it is
done. There are no any other major MCA dependencies to check for. Hence,
154x and 1640 are equivalent for the driver and it makes more efforts to
check for buggy firmware and bad boards than to detect the 1640 .. ;)
--
UZ
Laust Brock-Nannestad
> Here are some codelets from the AHA NT drivers, "to whom it may be useful":
I don't suppose you have the source to the NT DAC960 driver? Now that would
be useful ;)
Regards,
Laust
Charles Lasitter
163445.news.dfncis.de:
> The other thing is that I have not been able to get my
> MCA520/DAC960P-2 combination to boot off the array with any
> operating system while running firmware 2.73.
I've got one of these DAC960P-2 PCI RAID adapters not in use in the
moment if anyone else is in need (see email in sig)
-- CL.
+-----------------------------------------+
| Charles Lasitter | Mailing / Shipping |
| 401/728-1987 | 14 Cooke St |
| cl+at+ncdm+dot+com | Pawtucket RI 02860 |
+-----------------------------------------+
Unal Z
would
> be useful ;)
I haven't seen it in the DDK. AHA and NCR are about the only ones who have
MCA drivers in the NT DDK.
--
UZ
shinguz
who have
> MCA drivers in the NT DDK.
Could IBM have it? I mean, when installing NT onto the 520/DAC960
combination, the instructions call for the IBM driver disk *not* the
original NT DAC driver.
Unal Z
> combination, the instructions call for the IBM driver disk *not* the
> original NT DAC driver.
Laust was referring to the driver source code ....
--
UZ
shinguz
> > Could IBM have it?
> Laust was referring to the driver source code ....
So was I. I was just thinking that IBM might have source available
for the modified drivers.
--
shinguz the sleepless
Guenter Harp
> Contact Michael Lang, he has done some work on this.
Hi Alex,
I have contacted Michael Lang, the maintainer of the MCA-Linux SCSI
driver.
(lan...@kph.uni-mainz.de)and asked about the status of the driver for
the Cheetah. Here is his translated answer:
"At the moment there is no Linux driver for the adapter. However a
very considerable demand for a driver exists. I am still seeking for
technical documentaion for that type of adapter imploringly. The
hardware for testing the driver is no problem, because he already have
it. I am very sorry not to have the technical documentation. IBM
remain silent due to that information. As soon as I have all the
missing information I will start the work to generate the driver.
Best regards
Michael Lang"
I can imagine, that IBM would not like to make a MCA Linux driver for
the Cheetah by themselves. But do you think it may help, to make a
joint request of all interested MCA Linux friends to IBM to get the
necessary documentation for making Michael Lang able generating the
missing driver? I would like to know, what others will think about
that.
Best regards
Guenter
shinguz
Alex J Burke
I got in touch with Michael Lang a few days ago, he sent me an email
answering some points I raised and asked me for some information such as
adapter FRUs and who owns who today. I have sent that information along with
my background thoughts and gathering of other peoples comments, so I wait on
that. We also discussed strategies for making the drivers, and the tech docs
required for the driver building.
I would also like to ask anyone! who has any tech docs with regards to the
MCA RAID adapters, particularly the Cheetah (this is because it power the
majority ofthe biggest MCA systems) to please get in contact with us!
Also, anyone who can volunteer their support, it would be greatly
appreciated. People can help in all ways, testing drivers, website,
documentation - really anything. Any support is welcome! It would be nice to
get more members of the group behind this project, as the more voices we
have asking for it the harder we can push the companies (if need be) to give
us the tech references.
I think it would be fantastic for the group to have the option of running
their high end MCA systems under linux, I certainly have one 500 that is
just waiting to be used as such :-)
Also, I wonder what people think of possibly making a dedicated site for the
newer linux driver projects, like a central store of info and news - or, as
I mentioned before, maybe contacting the original MCA Linux website writer
and asking whether we can update it and place it back on the internet for
all to see. Any thoughts welcome!
Thanks very much, Alex J Burke.
shinguz
DAC960 that usually seems to come in the PCI flavor?
I've never seen a Cheetah...<sheepishly>
Kevin Bowling
"Alex J Burke" <ajb...@blueyonder.co.uk> wrote in message
news:koVra.198$Dd4...@news-binary.blueyonder.co.uk...
> Also, anyone who can volunteer their support, it would be greatly
> appreciated. People can help in all ways, testing drivers, website,
> documentation - really anything. Any support is welcome! It would be nice
to
> get more members of the group behind this project, as the more voices we
> have asking for it the harder we can push the companies (if need be) to
give
> us the tech references.
for IBM to release documentation on this and other projects that we then
send to them? I can make if nessesary.
> I think it would be fantastic for the group to have the option of running
> their high end MCA systems under linux, I certainly have one 500 that is
> just waiting to be used as such :-)
I'll help in any ways needed by those who write the driver, and am very
familier with the Linux system workings though I currently do not own a
card.
> Also, I wonder what people think of possibly making a dedicated site for
the
> newer linux driver projects, like a central store of info and news - or,
as
> I mentioned before, maybe contacting the original MCA Linux website writer
> and asking whether we can update it and place it back on the internet for
> all to see. Any thoughts welcome!
former are rarely updated and we could easily change that. If no one else
is interested, I have PHP/MySQL, Perl/DBI, HTML and Linux/basic web/system
administration to my name from several projects.
Alex J Burke
Thanks very much for your supoprt.
I must say, making a place where people can log their suport that could be
used to get docs at some stage would begood, but in the mean time just to
see the numbers of people who want to help with the project would also be
great.
About the coding itself, that too will be very useful in future, but I
reckon our first goal should be to get suport and get the tech refs and docs
we require. Then we can go into that side further.
Lastly, I am glad you approve of the website idea If people post what they
want to see up, maybe we could coordinate it?
Thanks, Alex J Burke.
Louis Ohland
possibly an OEM as well, I've seen at least three microchannel RAID
adapters listed by Mylex.
The real answer (been a conspiracy to keep the truth away from you) is
that the Linux support for the Mylex RAID adapters is for a higher level
of firmware, 3.xx (PCI versions). The Passplay/DAC960M/Cheetah use 2.73
(IIRC).
Mylex DAC960M Firmware
The Mylex Manufacturing Part ("D040") number can be located on the back
of the DAC960
controller, and uniquely identifies the model and number of channels on
the controller. It does not identify the amount of memory installed, or
the FW/BIOS versions, since these can be updated.
When referring to this D040 number, please use the entire number, since
this will help Technical Support identify specific features.
Mfg.No. Mylex Model
D040322 DAC960M
D040325 DBX960M
D040331 DAC960M-2
I have an older controller with version 2.xx FW, can I update the FW to
the 3.xx?
Not all boards will support the upgrade to 3.xx firmware. If the
controller has a revision number of D040347 or greater, the board will
support the upgrade. This revision label is usually found on the back
(non-component side of the board).
Kevin Bowling
"Louis Ohland" <ohl...@charter.net> wrote in message
news:3EB059C1...@charter.net...
> The Passplay is a spittin' image of the DAC960M. The Cheetah is
> possibly an OEM as well, I've seen at least three microchannel RAID
> adapters listed by Mylex.
>
> The real answer (been a conspiracy to keep the truth away from you) is
> that the Linux support for the Mylex RAID adapters is for a higher level
> of firmware, 3.xx (PCI versions). The Passplay/DAC960M/Cheetah use 2.73
> (IIRC).
card does work (brilliant). It MUST be possible for Linux drivers to be
written for this device, just that the coder of the current drivers only
knew how to go back to FW 3.xx, most likely due to lack of hardware/docs.
shinguz
> possibly an OEM as well,
In that case :)
<snip>
Linux support for the Mylex RAID adapters is for a higher level
> of firmware, 3.xx (PCI versions). The Passplay/DAC960M/Cheetah use
2.73
> (IIRC).
<snip>
The Linux drivers support from 2.73 which can be obtained from the
mylex website while Ibm supplies only 2.43 afaik. My pci-based
DAC960 (IBM Savannah) takes the 2.73 firmware without a glitch.
After flashing, the adapter works just fine.
The problems come when trying to get the adapter to boot. It simply
will not boot with firmware 2.73 from mylex, it does boot with 2.43
from IBM. All this on a 520 server pci/mca. If the DAC960 is
installed into any other pci box without mca, it boots with either
firmware. I suspect the mca screws things up somehow.
Laust Brock-Nannestad
> The real answer (been a conspiracy to keep the truth away from you) is
> that the Linux support for the Mylex RAID adapters is for a higher level
> of firmware, 3.xx (PCI versions). The Passplay/DAC960M/Cheetah use 2.73
> (IIRC).
Update your knowledge base, Louis, that's outdated information ;-)
The Linux driver has supported firmware 2.73 for some time now. The change
meant that the oldest DAC960-P adapters (single flash chip) are now
supported. Still, that's not MCA.
The Passplay has a max firmware of 1.99 (yep, low). I don't remember what
the Cheetah has, since I don't have one, but it's lower than 2.73.
I don't suppose anyone has actually tried flashing the Cheetah with the 2.73
firmware?
Regards,
Laust