 |



- It is strongy recommended not to share tape devices. No
assign/unassign functionality is implemented in the tape driver.
Therefore if a tape device is shared between different instances of
operating systems there is no protection against parallel
access.
|
- When there is IO on the CTC connection and the REMOTE side does
not correctly shut down (i.e. IPL without ifconfig
ctcx down or there is a system crash) the local
CTC device driver may enter into a hung condition. To recover from
this situation, the local side must also shut the connection down
(ifconfig ctcx down).
- The current implementation of the CTC network device driver can
handle up to 8 CTC connections and/or 8 ESCON connections. In
autosense mode the driver always picks 16 Parallel / ESCON channels
with the lowest Subchannel IDs. When a CTC definition is in the
kernel parameter file, the channels have to be in the first 256 CTC
/ ESCON Sub/channel ID's. This value (MAX_CHANNEL_DEVICES) can be
changed in the ctc.c file.
- When the CTC driver is statically linked into the kernel CTC
options cannot be specified in the Kernel parameter file which is
used as input for SILO on native systems. Otherwise LINUX for S/390
cannot be booted. A solution for this problem is to use the CTC
module instead of linking the CTC driver statically into the
kernel.
- Setting up a CTC or ESCON connection (with ifconfig up) may
require several retries (calling ifconfig down on both sides
first).
|
- At most 2GB of a swap device will be used
for swapping.
- The size of any swap file may not exceed
2GB.
|
- The limit for the main storage that can be defined is slightly
less than 2GB. (The actual limit seems to be 1919MB) For G5 and
G6 this can be fixed by an micro code upgrade.
|
- The flag '-fno-caller-saves' must be set.
- Many programs are not ANSI C++ conform. gcc-2.95.2 is stricter
about this than previous compilers widely used on Intel systems
(egcs). To compile such programs use -fpermissive and check on
Intel with gcc-2.95.2, before considering it to be a s390 specific
bug.
|
- The Hardware Console is intended to be used only to bring up
the system. After IPL and configuring your network connection, you
should use a Telnet session to operate the system and pass commands
to it.
Applications such as vi are not supported because of the HWC's line
mode nature.
- Displaying large files might cause some missing sections within
the output because of the latency of the hardware interface
employed by the device.
- In native or LPAR environments, you occasionally have to use
the Delete button of the GUI on the Service Element or Hardware
Management Console to enable further output. This is relevant to:
- SE version 1.6.1 or older on G5, G6, and Multiprise 3000.
- SE version 1.5.2 or older on G3, G4, and Multiprise 2000.
- Messages concerning the Hardware console operation that are
generated by the Hardware console driver, cannot be provided to the
syslog and are therefore unavailable with dmesg.
- Output from the head/top is deleted if the amount exceeds
approximately 30 Kbyte per LPAR (or image) on SE or HMC.
|
- While VM allows you to define up to 64 virtual S/390 processors
per guest machine, LINUX for S/390 currently provides support for a
maximum of 32 processors only.
|
- To avoid filling up you disk space and to improve the system
performance in out of memory situations you should consider to set
the corefile size to 0. The command you need to do that is
dependent of the shell you are using. For the bash is the command
"ulimit -c 0" in the file /etc/profile.
|
- To use this driver, a running TCP/IP service machine is
required on the one side, and on the other side, LINUX for S/390
must be running under VM.
- IUCV is limited to 20 connections.
|
- LINUX for S/390 cannot run with a root file system mounted via
NFS when the network connectivity is established via LCS, because
the LCS driver is delivered as an object-code-only module.
- To use OSA-2 Devices when running LINUX for S/390 on a basic
mode machine (no LPARs) you need to specify an ipldelay=xxm
boot parameter. We recommend a value between 2 and 5 for xx
for the OSA-2 card to settle down after LOAD.
|
- If expanded storage is not available, XPRAM cannot be used. Its
initialization fails with a log message reporting the lack of
expanded storage.
|
- Recommendation: We strongly recommend not to share
devices with production system. See here for more details.
|
- The IEEE emulation (running on machines without IEEE floating
point hardware (i.e. pre-G5 machines)) does not support handling of
denormalized IEEE floating point numbers. Furthermore the IEEE
emulation does not provide floating point exception handling.
|
- When IPLing from the virtual reader of VM/ESA, with a parameter
file that spans more than one line, make sure that a blank
character precedes any kernel parameter. To avoid errors you should
start on column 2 of the parameter line.
- When IPLing from disk or tape using an ASCII encoded parameter
file which you have generated on a UNIX or PC operating system,
make sure, that your line contains no special characters like tabs
or newlines. In particular your parmfile cannot span over more than
one line and must not be larger than 1023 Bytes.
|
- Multicasting has to be switched on in the kernel.
- The Gigabit Ethernet Driver needs QDIO base support.
- Starting with microcode level 0146, OSA-Express QDIO require a
portname to be set in the device driver. This portname is specified
using "portname:FOOBAR" in the qeth_options module parameter (see
Device Drivers and Installation Commands).
- HiperSockets support is not available for kernel 2.2, but for
kernel 2.4 only.
- Broadcast functionality is only available on OSA-2 hardware,
not on OSA-Express QDIO.
- For further information see documentation.
|
|
 |
|
 |