Tux on VM

Last updated on:
Sunday, July 06, 2008







Valid HTML 4.01!
Linux for Big Iron

Archives of linuxvm.org

03/05/2002 - In a very important posting, Axel Wirbser of IBM reported the following updates to the DeveloperWorks web site:
On the Recommended level 2.2.16 page
- Recommended kernel 2.2.16 source code for the lcs device driver.

On the Recommended level 2.4.7 page
- Recommended kernel 2.4.7 source code for the lcs device driver.

On the Experimental level 2.2.20 page
- Experimental kernel 2.2.20 source code for the lcs device driver.

On the Experimental level 2.4.17 page
- Experimental kernel 2.4.17 source code for the lcs device driver.

On the kernel 2.4 documentation page
- Updated Linux for S/390 & zSeries Device Driver and Installation Commands manuals.

On the new Useful add-on page you will find two new experimental packages:
- snIPL (simple network IPL)
- src_vipa (source VIPA)

03/06/2002 - Evandro Vargas reported that a new IBM Redbook was available, "e-Business Intelligence: Data Mart Solutions with DB2 for Linux on z/Series."

03/07/2002 - Neale Ferguson posted the URL to an interview with Dan Frye, Director of IBM's Linux Technology Center.

03/09/2002 - In response to a question about the location of the IBM JDKs for Linux, Jim Elliott of IBM posted this link.

03/2/2002 - Florian LaRoche of Red Hat reported a new version of their 64-bit beta had been uploaded. He listed some of the changes:

  • kernel contains:
    • GPL lcs driver is now included
    • fix for samba problems
    • order2-3 patch from IBM
  • many bug-fixes for the installation
  • all rpms are now compiled with the newest binutils patch from IBM
  • htdig/squid plus some other rpm updates
He went on to comment:
"We'd be very interested to hear about remaining problems and get feedback about the current version. We don't expect any larger patches between now and gold-master, but only further testing can prove this right or wrong.

03/13/2002 - Sergey Korzhevsky shared his success in getting tape support to work to write a Linux/390 IPL tape. Most systems have 2 special character files and one special block file defined for each "set" of tape devices, e.g., rtibm0 (rewinding character tape 0), ntibm0 (non-rewinding character tape 0) and btibm0 (block tape 0), rtibm1, ntibm1, btibm1, etc. The entries in /dev (on a Red Hat 7.2 system) look like this:

crw-------    1 root     root      37,   0 Sep 23 11:25 rtibm0
crw-------    1 root     root      37,   1 Sep 23 11:25 ntibm0
brw-------    1 root     root      37,   0 Sep 23 11:25 btibm0

crw-------    1 root     root      37,   2 Sep 23 11:25 rtibm1
crw-------    1 root     root      37,   3 Sep 23 11:25 ntibm1
brw-------    1 root     root      37,   1 Sep 23 11:25 btibm1

crw-------    1 root     root      37,   4 Sep 23 11:25 rtibm2
crw-------    1 root     root      37,   5 Sep 23 11:25 ntibm2
brw-------    1 root     root      37,   2 Sep 23 11:25 btibm2
Note that because there is both a rewinding an non-rewinding device that the minor numbers increment by two between "sets" of devices. In the case of the block tape devices, the minor number increments by one.
Note also that other systems use major number 254, which is supposed to be used for local testing purposes, not production use.

Sergey's method was this:
  1. insmod tape390 tape=<address>
  2. cd /dev; mknod ntibm0 c 254 1
    Driver uses major number 254 (in my case)
  3. mt -f /dev/ntibm0 rewind
  4. dd if=image of=/dev/ntibm0 bs=1024
    write the kernel image to tape
  5. mt -f /dev/ntibm0 eof 1
    write an EndOfFile mark
  6. dd if=parmfile of=/dev/ntibm0 bs=1024
    write the parmfile to tape
  7. mt -f /dev/ntibm0 eof 1
    write an EndOfFile mark
  8. dd if=initrd of=/dev/ntibm0 bs=1024
    write the ramdisk image to tape
  9. mt -f /dev/ntibm0 eof 2
    write EndOfFile mark twice - for sure :)
  10. mt -f /dev/ntibm0 rewind

03/13/2002 - Neale Ferguson reported SuSE's announcement of their 64-bit Linux/390 distribution's anticipated GA date: the beginning of May, 2002.

03/14/2002 - In response to a question about setting the boot address for a Linux/390 image installed via z/VM's System Administration Facility, Kurt Acker posted this:
"This is documented on pages 10 and 11 within Chapter 3, Installation Procedures. z/VM System Administration Facility, SC24-6034-00."

03/14/2002 - Dave Jones discovered this IBM press release pre-announcing the general availability of mySAP.com for Linux for zSeries (64-bit) only, on May 31st of 2002. It includes a testimonial from Burghart Pohler, data center manager of Alcatel SEL AG, on the ease of installation and the results of their performance tests. (No numbers were given, of course. This is a press release after all.)

03/14/2002 - James Melin reported that he was unable to boot Linux/390 in an LPAR off a PAV volume on his Shark (IBM ESS) box. He was able to use PAV volumes from Linux/390, just not boot off them. He closed his note with this comment:
"The reason that PAV volumes for boot volumes seem to work under VM is that VM is managing the I/O for the VM guest and it knows enough to make the PAV volumes interoperable. It appears to be only when Linux is native in an LPAR, that the PAV volume as a boot media fails."

03/15/2002 - Marcel Foortjes discovered that Google has a news search page in beta testing. Using it, he found this article about Wisconsin Physicians Service Corp. conversion from Intel servers to Linux/390.

03/15/2002 - In response to some people having problems with Samba on Red Hat's 7.2 system, David Sainty of Red Hat posted this comment:
"I would suggest trying my patched samba package set, available at: ftp://people.redhat.com/dsainty/s390/
These packages should be identical to those you have already, with the exception that there is an added patch to disable use of the dnotify system call."

03/15/2002 - In response an question about obtaining a trial copy of WebSphere Application Server 3.5 for Linux/390, Dave DiCarlo, Sales Manager for IBM Americas Linux Middleware posted this note:
"We are working on providing a permanent vehicle to provide trial copies of WAS 4.0 to anyone who wants to evaluate the product. In the interim if you need an evaluation copy, please send a note to me and we'll try to turn it around and airship you a copy as quickly as we can. Please be sure to include a shipping address and phone number for the Airbill."

03/15/2002 - In response to someone who was having problems, Carlos Ordonez gave a pointer to an IBM Redbook, "Technical Introduction: IBM eserver zSeries 800, SG24-6515-00", that talks about setting up an OSA Express Fast Ethernet card in QDIO mode, specifically with SuSE Linux Enterprise Server 7.2. The excerpt he quoted from page 123 of the book describes the problem and the workaround:

The installation procedure for the SuSE LINUX 2.4.7 distribution has a minor bug that prevents the automatic setup QDIO mode. However, this setup can be performed manually. To do this, select option 0 (no network). For security reasons, set up the root temporary password first:

   passwd root
Then set up the network, using the following commands:
   echo 'noauto ' >>/proc/chandev
   echo 'qeth0,0x0e30,0x0e31,0x0e32,0,0 '>/proc/chandev
   echo 'add_parms,0x10,0x0e30,0x0e32,portname:RPTRFET1 ' >>/proc/chandev
   insmod qdio
   insmod qeth
   ifconfig eth0 netmask broadcast up
   /etc/rc.d/inetd start
   /etc/rc.d/portmap start
Further on in the book, on pages 124 and 125, Carlos quoted this section:

Post YaST setup for network in QDIO mode
Our SuSE version could not automatically configure a QDIO interface. We assume this will be fixed in later SuSE releases. To bypass the problem in our release, we performed the following steps while still running in the Telnet session connected to the ramdisk Linux on the z800. After YaST has finished its installation processes, we issued the following commands:
   mkdir /mnt/newroot
   mount /dev/dasda1 /mnt/newroot
   cd /mnt/newroot/etc
We used vi to edit the chandev.conf and modules.conf files in the /mnt/newroot/etc directory.
  • In modules.conf, we changed:
       alias eth0 off
       alias eth0 qeth
  • In chandev.conf, we added the following lines:
As an alternative to using vi, you may issue the following commands (remember to adjust the parameters to your values) while in the /mnt/newroot/etc directory:
   echo 'alias eth0 qeth '>>modules.conf
   echo 'noauto ' >>>chandev.conf
   echo 'qeth0,0x0e30,0x0e31,0x0e32,0,0 '>>chandev.conf
   echo 'add_parms,0x10,0x0e30,0x0e32,portname:RPTRFET1 ' >>>chandev.conf

03/15/2002 - Samy Rengasamy reported a problem he was having trying to get the Berkely DB 4.01 package to build in his system. According to the support folks he contacted, "The problem is that Berkeley DB does not include assembly language mutexes for the s390 architecture, and Linux POSIX 1003.1 pthread mutexes aren't sufficient for DB as they do not support locking between processes, only between threads in a single process." In response, Neale Ferguson submitted this post:
"Try this for a spinlock. It's similar to something I coded when I ported Postgresql a couple of years ago. (Have we been doing this for > 2 years already?)"

31 bit:

static inline int TAS(int *lock)
  int _res;

  __asm__ __volatile("   lghi    1,1\n"
                     "   lg      2,%0\n"
                     "0: slgr    0,0\n"
                     "   cs      0,1,0(2)\n"
                     "    jnz     0b\n"
                     "   lgr     %1,0"
                     : "+m" (lock), "=d" (_res)
                     : : "0", "1", "2", "cc");


static inline void CLR(int *lock)
  __asm__ __volatile("   slr     0,0\n"
                     "   l       1,%0\n"
                     "   st      0,0(1)\n"
                     : "=m" (lock) : : "0", "1");



static inline int TAS(int *lock)
  int _res;

  __asm__ __volatile("   lghi    1,1\n"
                     "   lg      2,%0\n"
                     "0: slgr    0,0\n"
                     "   cs      0,1,0(2)\n"
                     "    jnz     0b\n"
                     "   lgr     %1,0"
                     : "+m" (lock), "=d" (_res)
                     : : "0", "1", "2", "cc");


static inline void CLR(int *lock)
  __asm__ __volatile("   slgr    0,0\n"
                     "    lg      1,%0\n"
                     "   st      0,0(1)\n"
                     : "=m" (lock) : : "0", "1");


03/16/2002 - Mark Post forwarded a security alert about vulnerabilities in Oracle 8i and 9i servers:
"Multiple vulnerabilities in Oracle Application Server have recently been discovered. These vulnerabilities include buffer overflows, insecure default settings, failures to enforce access controls, and failure to validate input. The impacts of these vulnerabilities include the execution of arbitrary commands or code, denial of service, and unauthorized access to sensitive information.

For complete details, reference CERT Advisory CA-2002-08. This document is available from:

03/17/2002 - Asher Glynn contributed a pointer to a soxy proxy client that can be compiled and run on Linux/390. Asher said that he's been using it for over a year for SSH connections.

03/18/2002 - During a discussion about the differences between the "cdl" and "ldl" Linux/390 DASD layouts, Volker Sameske of IBM provided the following information:
"The Linux partition table in cdl case is hidden in a 390-like volume table (VTOC) in the second track.

"And you can specify a block on an ECKD DASD by specifying cylinder/head/sector or cylinder/track/block. Head and track are equal here, like sector and block."

(0/0/1): 1st IPL record  - size: blocksize
(0/0/2): 2nd IPL record  - size: blocksize
(0/0/3): volume label    - size: blocksize (LNX1)
followed by the one and only partition

(0/0/1) boot code      - size: 28 byte
(0/0/2) boot code      - size: 148 byte
(0/0/3) volume label   - size: 84 byte (VOL1)
(0/0/4) 1st IPL record - size: blocksize
(0/0/5) 2nd IPL record - size: blocksize (only if needed)
(0/0/n) empty          - size: blocksize

(0/1/1) format 4 DSCB  - size: 140 byte (describes VTOC itself)
(0/1/2) format 5 DSCB  - size: 140 byte (contains free space info)
(0/1/3) format 7 DSCB  - size: 140 byte (omitted for small disks)
(0/1/4) format 1 DSCB  - size: 140 byte (describes 1st partition)
(0/1/5) format 1 DSCB  - size: 140 byte (describes 2nd partition)
(0/1/6) format 1 DSCB  - size: 140 byte (describes 3rd partition)
(0/1/n) empty          - size: 140 byte

(0/2/1) data           - size: blocksize

03/18/2002 - Gregory MacDonald was the first to spot the announcement of Oracle's plans to release a generally available version of its "Oracle9i Database with Real Application Clusters, on IBM's mainframe Linux environments."

03/19/2002 - Axel Wirbser of IBM reported the following updates to the DeveloperWorks web site:

  • Recommended gcc 2.95.3 bug fixes
  • Recommended 2.4.7 "Order 2 allocation relief" patch who no longer breaks other architectures and compatible with the timer patch on the same kernel.
  • Experimental 2.4.17 "Order 2 allocation relief" patch who no longer breaks other architectures and compatible with the timer patch on the same kernel.

03/20/2002 - In response to a comment about native Linux performance measurement tools to look at a sendmail problem Alan Cox posted this comment:
"top will show you process data
sard will show you disk statistics providing your vendor ships that feature (look in proc/partitions and you should see stuff like..
22 64 80043264 hdd 23376 219526 1943138 1722140 14309 266197 2244104 2311340 0 801060 4033700 )
vmstat will provide paging rate information (especially handy is vmstat 1)

Sendmail defaults to stopping listening based on load average, the value it picks may be a bit low for the S/390 i/o v cpu.

Look in /etc/sendmail.cf for

      # load average at which we just queue messages
      #O QueueLA=8

      # load average at which we refuse connections
      #O RefuseLA=12
remove the comment, set values, and restart sendmail. The default sendmail tuning is fairly defensive for a small box."
Dave Jones also recommended the use of "netstat -s" to get network statistics.

03/20/2002 - In response to a query about using PAM to provide Linux authentication and resource control through to the LDAP server that is part of RACF, Franco Masone replied "A friend of mine helped me to write a little PHP3 script to test LDAP access from Linux/390 to OS/390 LDAP/RACF. Write to me directly if you want the code." He further commented "Our objective was to demonstrate that it's possible to validate user and pwd to RACF without duplicate users or to re-invent the wheel." That code is now available at:

03/20/2002 - Neale Ferguson found this URL which points to a Debian GNU/Linux for S/390 news site.

03/20/2002 - In response to a question about alternatives to Computer Associates' CA eTrust Access Control product, Alan Cox recommended rsbac, "Rule Set Based Access Control for Linux."

03/20/2002 - paultz (paultz@ucia.gov) found another news article talking about Oracle's pre-announcement of their 9i database for Linux/390.

03/21/2002 - Michael Coffin's write up on creating a second bootable Red Hat 7.2 system from an existing running Red Hat 7.2 system has been added to the HOWTO page.

03/21/2002 - Neale Ferguson found this article about Red Hat. "Red Hat, the top seller of the Linux operating system, has restructured to focus on big customers while cutting jobs elsewhere, executives said Tuesday."

03/25/2002 - Scott Courtney reported finding a reference to build support for Linux/390 being added to XFree86 4.2.0 in the release notes for the product.

03/25/2002 - In response to a question about getting Apache to talk to Tomcat using mod_webapp, Samy Rengasamy responded with this:
I just included the following lines in httpd.conf:

LoadModule   webapp_module
AddModule    mod_webapp.c

WebAppConnection   conn       warp   xxx.xxx.xxx.xxx:8008
WebAppDeploy       mytest     conn   /mytest
Servr.xml does have info on 'mytest' context. The above lines will deploy a context 'mytest' under tomcat/catalina and it will be available at the relative path /mytest. I did build mod_webapp at our installation. The only problem I had was, I could not get tomcat/catalina to start on reboots.

03/25/2002 - In response to a question about finding a Rexx for Linux/390, Samy Rengasamy responded with a pointer to a mod_rexx package that works with the Apache web server for cgi scripting.

03/25/2002 - Scott Courtney posted a link to a ZDNet Tech Update that looks at Linux/390. The writer was a little shy on S/390 background, particularly VM, so Scott said he sent him an email with some more information and a few URLs to look at as well.

03/26/2002 - In response to a question about ways to "push" maintenance to multiple Linux/390 images, Dave Jones recommended taking a look at a product by Aduva:
"Their product maintains a database of what RPM packages, at what VRM level, are installed on each server in a server farm. Then when an RPM package is updated, the product automatically installs and configures the new patch on each server that needs it, and restarts the server. If other installed packages on a server need to be updated as well, Aduva handles that as well.
"There's a rumor that they are porting this application over to z/VM to support large Linux virtual server farms."

Other people recommended using packages such as rsync, rdist, etc, but Bill Barr recommended using expect/autoexpect to automate certain parts of the task. To use expect, you'll also need Tcl installed.
Jay van der Meer said "We are using ant and the ftp task. It uses a XML control file to handle the processing. It can handle different user id's, and other options per FTP site."
http://expect.nist.gov/ - for expect
http://dev.scriptics.com/ - for Tcl

03/26/2002 - Mark Post reported an Infoworld article on IBM announcing a telecommunications carrier grade Linux system.

03/26/2002 - Gordon Wolfe posted a follow up to a problem he reported previously where he couldn't get Konqueror on SLES7 to work with his company's proxy firewall. It turns out that ECN was enabled by default in the TCP/IP stack, and some internal routers were having problems with it. Gordon's network person recommended issuing a

echo "0" > /proc/sys/net/ipv4/tcp_ecn
command in /etc/init.d/boot.local to turn this off at system boot time. That appeared to fix the problem. Alan Cox recommended getting the root cause fixed, and replace/upgrade the routers that don't support ECN, since ECN is now an official internet standard, and more and more traffic using it will be showing up in the future.

03/27/2002 - In response to a question/complaint about DF/HSM not recognizing the Linux/390 DASD format, Volker Sameske of IBM pointed out two HOWTOs that would be of interest to OS/390-z/OS shops that are using the Linux/390 cdl (compatible disk layout) format. The two HOWTOs are for storage administrators to use DFSMShsm and DFSMSdss to backup and restore Linux/390 partitions and volumes. Pointers to these two IBM HOWTOs have been added to the HOWTO page on this site.

03/27/2002 - There was some discussion on the mailing list about the hostid command, and just exactly what the value was that was being returned, and where it came from. Dougie Lawson finally posted a note that showed it was the little endian representation of the system's IP address. For example, his system with IP address shows a hostid of B4091682, B4=180, 09=9, 16=22, 82=130. This can be complicated by whether a particular hostid displays the results in base 10, or base 16.

03/27/2002 - Neale Ferguson reported a story titled "IBM and SuSE Offers New Enterprise-Ready Linux Services:"
"Nuremberg, Germany and Armonk, NY, March 27, 2002 - SuSE Linux and IBM today announced a broad services alliance that will enable both companies to jointly provide Linux support and services to corporate customers around the world."

03/27/2002 - Mark Post reported an article about Ximian Connector. According to the article "Ximian Connector, a client-based product, serves as an extension to the company's Evolution suite of groupware products that enables Linux and Unix users to better manage personal information as well as to collaborate with Windows®-based colleagues using Exchange 2000."
Mark closed out with the comment "So, now we have the Bynari Insight server for shops that want to keep their Outlook clients but replace the Exchange servers, and we have the Ximian connector for people that want to keep their Exchange servers, but have a need for Linux desktop users to be able to connect. Or, you could replace the Exchange servers, and let your users decide which desktop they want, Windows® or Linux, and still be able to communicate effectively."

03/28/2002 - Phil Tully contributed a pointer to an online Forbes article titled "Wall Street Embraces Linux."
The article talks about some of the work going on with Linux at Merrill Lynch and Credit Suisse First Boston.

03/28/2002 - Mark Post reported that SAP had released its database code and development environment under the GPL some time ago. He downloaded them (about 13MB and 3MB, respectively) to see if they would build, but discovered that the development environment had some Intel binaries in it that were needed to do the build. An inquiry to the SAP team that supports the open source product said that the source to the development environment itself should be available in a few weeks.

03/28/2002 - In response to a question about general rules of thumb for estimating resources that might be needed by multiple Linux/390 guests running under VM, Rich Smrcina replied that there are none, but gave some examples of guests he has defined in the past.
Domain Name Server 24 MB
sendmail 36 MB
Apache and network monitoring 40 MB
DB2 Connect 96 MB
WebSphere 512 MB
He commented that the WebSphere machine had not been driven hard enough to come up with a final configuration for it.

03/28/2002 - In response to a question about VM VM performance, Mike MacIssac contributed this pointer to an IBM page with a number of different recommendations.

03/28/2002 - Don Mulvey announced the "first full release of the Enterprise Volume Management System. Package 1.0.0 is now available for download at the project web site."
Highlights for version 1.0.0:

  • Core Kernel
    • Support for /proc.
      • New directory /proc/evms/.
      • File entries: volumes, info, plugins
      • Sysctl entry: /proc/sys/dev/evms/evms_info_level can be used to query and set the EVMS kernel messaging level.
  • GUI
    • Option panel fixes.
    • New columns in most panels: Read-only and Corrupt.
    • Default engine logging level changed from "everything" to "default".
    • Check for minimum required engine API version.
  • Text-Mode UI
    • Added "F5=Commit" to menu to allow saving changes without exiting.
    • Screen refresh fixes.
    • Default engine logging level changed from "everything" to "default".
    • Check for minimum required engine API version.
  • Command Line
    • On-line help cleanup.
  • New Plugin: s390 Segment Manager
    • Recognizes existing CDL, LDL, and CMS partitions.
    • Can build on top of these partitions in the engine, but cannot yet create new s390 partitions.
  • MD Plugin
    • Added proc entry: /proc/evms/mdstat
    • Added sysctl entries: /proc/sys/dev/evms/md/speed_limit_[min|max] for controlling the speed of RAID-1 and RAID-5 sync'ing.
  • BBR Plugin
    • Bug fixes to the I/O error remap handling.
  • AIX Plugin
    • Bug fixes in the discovery path and mirroring I/O path.
  • LVM Plugin
    • Added proc entry: /proc/evms/lvm/global

03/29/2002 - Mark Post forwarded a pointer to an eWeek article about Winnebago Industries' move to running Bynari's Insight server for email. It also covers a lot of other Linux and Linux/390 trends as viewed by various industry analysts.

Site hosting courtesy of Velocity Software