IBM PS/2 SCSI-Subsystem Linux Support

This page contains the latest information about the development of the IBM Microchannel SCSI-subsystem low level driver for Linux. See the subpages for detailed description about this part of Linux.

Contents of These Pages

 * * * Latest News * * *
  • 18 Sep 2003: A new driver release is in test. Several unexpected locks occured.
  • 11 Jan 2002: One year has passed and Kernel 2.5 is already deep in its development. Some changes have been applied to the driver in Kernel 2.5. There are some modifications for the system locking. Please do not use Kernel 2.5.45 or previous releases of Kernel 2.5, as the SCSI driver won't work or it will just run very slow. I am currently reshaping the driver and yes, indeed, RAID support is still not out (sigh)...
  • 18 Jan 2002: PSU replaced and model 8595 back on track. Thanks to everybody for your support and recommendations.
  • 26 Dec 2001: Call for emergency support concerning broken power supply on model 8595-AK9 (see message above).
  • 12 Sep 2001: Terrorism in USA - statement.
  • 14 Jan 2001: Kernel 2.4.0 is now out for one week, including driver version 4.0a, but it is not working correctly with the new kernel. Yes, I am very sorry for that, but new technology never works on the first stage... anyway, no excuses. The new age defines things different, so it can happen, that the driver is not called and/or none of your SCSI adapters gets detected!. The answer to this is version 4.0b. This is located on the driver page here. Model 80 problems with the SCSI driver are described in the FAQ of README.ibmmca.
  • 4 Jan 2001: In order to better know, how the IBM MCA SCSI driver is used and how much people really use it, it would be great, if every user could just drop his usage data into the formular under Driver User Statistics. It is absolutely anonym, so no danger for you at all.
  • 3 Jan 2001: The 4.0 drivercode was not working as a module. This is solved by version 4.0a. Now it is working fine with kernel 2.4.x and module support. In addition, some code got optimized. The knowledge, applied in version 4.0a will soon appear also in 3.2d driver for older kernels. But before this, the driverpage will be renovated or else you will not find the right driver for your kernel anymore.
  • 28 Dec 2000: Strange, I would bet, I placed some note here still after 29 July... However, driver version 4.0 is released today. There are from now on two drivers available for your support. People which still use kernel older than 2.4.0 must use 3.xx drivers, while from kernel 2.4.0, version 4.xx is required. You cannot do wrong. If you have the wrong driver for your kernel release, the driver code stops the compilation of the kernel. Major reasons for the new driver 4.0 were troubles with the interrupt handler (my favourite bugsource) and with the DEVICE_INQUIRY command. These bugs exist in 3.2, but they will be removed in 3.2c completely. See the history section of README.ibmmca for latest changes.
  • 29 Jul 2000: I took the last 10 days to finish minor work on the 3.2 and the F/W support. So, today I hereby present release 3.2 of the driver that is submitted to the current kernel developments of 2.2, 2.3 and 2.4.
  • 18 Jul 2000: The ANSI-device-order problem is now solved with version 3.2pre9, just released on the driver-page. In addition, the possibility that the IBM SCSI-2 F/W adapter could have an adapter PUN of 4-bit is now taken into account, sticking strictly to the F/W manual.
  • 17 Jul 2000: Warning! There could be a bug with the device access in the PUN-mapping, when you have selected ANSI-device-order for mapping the SCSI-devices! I hope to fix that during today. If you use standard device order, you should not get any problems.
  • 17 Jul 2000: 3.2pre8 is release today! All reported problems seem to be solved, now. Ghost-SCSI-interface on 9595 is now avoided by a more restrictive check of the POS registers. In principal, if the driver is not sure, if the device, shown in the integrated SCSI-POS is some integrated SCSI-controller, it will be ignored. IBM SCSI F/W adapter support is now working fine (autoprobing maximum synchronous transfer rate and allowing wide-addressing (id=0..16)). COMMAND ERRORs, caused by overestimated synchronous speed setting got corrected. Only dual-bus-mode to allow to slow down speed on external SCSI-devices is not yet provided.
  • 13 Jul 2000: Added a link to Louis Ohlands great MCA-page and extended the latest README-file. The request for your POS registers in the SCSI: line of the pos-file is no joke. I really need that info from as much as possible people.
  • 1 Jul 2000: Thanks to everybody who contributed hardware for my test-setup. I have now the capabilities to test the driver on a broad variety of PS/2 systems and types. This allows a much better bughunting than ever before.
  • 4 Jun 2000: Back again from Italy and today I release a new 3.2pre-release, that is 3.2pre7. This release behaves slightly faster during I/O, because unused and complicated parts of the code got deleted or optimized. A contribution from J. Shorney to activate the harddrive activity LED on models 95A is added and common complaints by P. Hahn conerning the various KERNEL_VERSION derivates is taken into account in this release. To activate the harddrive LED on the model 95A in addition to the alphanumeric LED panel, just use the commandline parameter activity now. Concerning the bootproblems on various machines, we got THE big step ahead. I have some (yet unofficial) patch here which modifies scsi_scan.c.
  • 22 May 2000: Thomas Persson contributed a Bootdisk for S.u.S.E. 6.3 to install this Linux-distribution on MCA-machines. In addition, some links were wrongly set on the driver-page. This got corrected now.
  • 27 Apr 2000: Version 3.2pre1 got replaced by version 3.2pre2. The reason was still a slight bug in the interrupt-handling of the driver, when the PS/2 is cold-booted and the first device on the bus to be detected is of sequential type, like tape-drives and kernel 2.2.x is used. In addition, internal synchronous transfer at adapter maximum speed is guaranteed at boottime. It is possible to slow down this synchronous rate for troubleshooting (see README). The bug-report form WWW-page call for e-mail was buggy, this has been repaired and now your bug reports won't get swallowed by the mailing software.
  • 25 Apr 2000: Added some comments about supported hardware.
  • 24 Apr 2000: Version 3.2pre1 of the driver is out! Please download it for testing and send bugreports to the usual address. In addition, I provide missing ADF-files for your adapters here. On the link-page, you may find (finally) a link to David Weinehall's homepage, too.
  • 23 Apr 2000: Version 3.2 is now in test. Concerning E-Mail, you send to me, I am sometimes burried in requests and questions, so please, if you do not get any reply within one week, just send your E-Mail request again.
  • 4 Apr 2000: Version 3.2 of the IBM SCSI-subsystem driver is coming soon. It will contain the following optimizations:
    • Better interrupt-handling for more watertight SCSI-command processing.
    • Solution to the problem, why the driver reports errors, when everything was obviously ok (hopefully).
    • Recognition of commandline parameters in Kernel 2.3 and 2.4.
  • 31 Oct 1999: Users having problems with the driver on boottime and who have non-SCSI-II devices should try the kernel bootparameter ibmmcascsi=bypass first. If the problem remains, please contact me.
  • 28 Oct 1999: The tool scsiinfo as also kernel 2.3.23 cause a severe fatal kernel panic from the IBM SCSI-subsystem. It appears some kind of sequencing error. If somebody can offer bugreports, they are highly welcome!
  • 1 Oct 1999: I placed a new bootdisk for SuSE-distribution on the bootdisk page. It has been provided by Peter Jakobsson. In any case, if you think, you made a nice bootdisk for some Linux-distribution, please send it to me. Requested bootdisks are for Debian and Redhat.
  • 31 Aug 1999: Ok, today I was so happy, that I got positive feedback from people using more than one IBM-SCSI-Adapter in their box. Regrettably, there are still a lot of people out there who do not agree and receive heavy problems with multiple IBM-SCSI-adapters! Please, if so, send me your problem- and bug-report with details on your problem and when the system crashes. I think that sometimes weired things do happen in multiple host mode.
  • 1 Jul 1999: The release 3.1e of the SCSI-driver is included from Kernel release 2.2.9.
  • 18 Mar 1999: Test on the driver continues. By the way, IBM and SAP now officially support the distribution and support for Linux customers, users and developers. I already checked the developers support out on IBM, and it worked! See this page for further info!
  • 16 Mar 1999: Added a new page containing bootimages which come up on IBM Micrcochannel machines with SCSI-adapters. It is all still under construction.
  • 3 Mar 1999: With version 3.1e, some problems with interrupt handling can still appear, if you have more than one IBM SCSI-subsystem. To get this under control, I am right now working on an improvement in order to get it proof for multiple adapters. This will become version 3.1f and should be ready within the next days.
  • 23 Feb 1999: Even though, the latest driver-source has been submitted to Linus, Kernel 2.2.2 contains still the 3.0f release. I have again to warn everybody NOT(!!) to use 3.0f and to upgrade all relevant files with the latest version from this server here. 3.0f will not work on most PS/2 machines!
  • 20 Feb 1999: Ok, I am careful with enthusiasm, but it seems, the latest release of the driver is running on models 95XX, really. The latest release is version 3.1e and it is offered on the driver-page. The reason for the success is based on some unassigned variable memory for the SCBs, passed to the SCSI-adapter. Since this bug in the implementation is solved, even the problem with the Ooops-message, when looking inside /proc/mca has vanished. Therefore, the ignorecmd and forgiveall is now obsolete! Read the new README.ibmmca to be informed about the latest steps.
  • 17 Feb 1999: Even though I found severe problems with the variable handling in the driver, the version 3.1b does not yet appear to solve the problems with 95XX series PS/2! Now, it shows a OOPS-message, really strange, but do not give up hope, as I do not give up, either :-) !
  • 15 Feb 1999: After some sequence of coincident events, I think, I got the bug with models 95XX series solved!!! As I do not have one, try the version 3.1b (is published on the page here ), and tell me about your results :-), but this time, I found really severe reasons in the variable handling of the scb-structure, which was quite buggy.
  • 13 Feb 1999: The problem, that still exists with some integrated SCSI controllers seems to be limited to models of type 95XX. The 85XX series should work with integrated SCSI-controllers. If somebody knows about the technical differences in detail between the integrated SCSI-subsystem on models 85XX and 95XX, feel free to send me some e-mail! If you are owner of a model of type 95XX, where the integrated SCSI-driver won't work with the linux driver, send e-mail to me with subject: NOTIFY 95XX PROBLEM and an empty mail-body to me, so you will be added to a list, and when the solution is found, you get immediate e-mail in order to inform you about the problems' solution.
  • 11 Feb 1999: Linux Kernel releases from 2.1.87 to 2.2.1 will not work with IBM SCSI-Adapter and the IBM SCSI-Adapter w/cache. If you have a kernel version, lower than 2.2.2(!), get the latest ibmmca.c SCSI-driver source file from this page! It will work with both types!

Last content changes: 18 September 2003

Content created and/or collected by:
Louis Ohland, Peter Wendt, William Walsh, Kevin Bowling, Jim Shorney, Tim Clarke, David Beem, Tatsuo Sunagawa, Tomáš Slavotínek, and many others.

Ardent Tool of Capitalism - MAD Edition! is maintained by Tomáš Slavotínek.
Last update: 11 Apr 2021 - Changes & Credits | Legal Info & Contact