the inconsistency in the zSeries ELF ABI get fixed?
I'll see to get this fixed.
Thanks for reporting the problem!
Bye,
Ulrich
--
Dr. Ulrich Weigand
GNU Toolchain for Linux on System z and Cell BE
[EMAIL PROTECTED]
--
For LINUX-390
.)
Likewise for ^d and ^z.
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
GNU compiler/toolchain for Linux on System z and Cell BE
IBM Deutschland Entwicklung GmbH, Schoenaicher Str. 220, 71032
Boeblingen
Phone: +49-7031/16-3727 --- Email: [EMAIL PROTECTED
recommend to use GDB to track this down further,
either on the live process or else on a core dump.
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
GNU compiler/toolchain for Linux on System z and Cell BE
IBM Deutschland Entwicklung GmbH, Schoenaicher Str. 220, 71032
--
Dr. Ulrich Weigand
Linux on zSeries Development
[EMAIL PROTECTED]
--
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http
is
invoked with the -march=z9-109 option. The first official release to
contain this support will be GCC 4.1.
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux for S/390 Design Development
IBM Deutschland Entwicklung GmbH, Schoenaicher Str. 220, 71032
/ Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux for S/390 Design Development
IBM Deutschland Entwicklung GmbH, Schoenaicher Str. 220, 71032 Boeblingen
Phone: +49-7031/16-3727 --- Email: [EMAIL PROTECTED
at all, but you may in some cases see performance
improvements of 5% or more.
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux for S/390 Design Development
IBM Deutschland Entwicklung GmbH, Schoenaicher Str. 220, 71032 Boeblingen
Phone: +49-7031/16
to handle your very problem (analysing SIGFPEs
resulting from divide), you may have a look at
libjava/include/s390-signal.h (HANDLE_DIVIDE_OVERFLOW)
residing in current GCC sources.
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux for S/390 Design
. AFAIK the 2.6 kernel also has a read-once detection
heuristic of some kind, though I'm not sure how reliable that
one is ...
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux for S/390 Design Development
IBM Deutschland Entwicklung GmbH, Schoenaicher
Mark Post wrote:
I wanted to let everyone know that Andreas Herrmann of the Boeblingen lab
was able to help me with this problem. He did some diagnostic work, and
then contacted Ulrich Weigand to confirm that there was a gcc compiler bug
at work. The fix for the gcc-3.3 series is documented
official kernel releases, because that would
be a lot of additional work. Either use the service stream, wait
for inclusion in the next official release, or backport yourself ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED
,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
--
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex
with in-tree
builds right now (AFAIK); I was just suggesting to prefer
out-of-tree builds if you have the choice, given the
circumstances ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
--
For LINUX-390 subscribe / signoff
guess we can provide that patch ...)
It is certainly not our intention to make building your own kernel
difficult; in fact with the 'release-often' model we're trying to
follow with 2.6 Bitkeeper head, we thought it would be easier to
do so than it was with 2.4 ...
Bye,
Ulrich
--
Dr. Ulrich
potential problems (and it's a good idea in general, anyway).
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
--
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message
trying to rebuild them ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
--
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http
are
placed from the bottom of the stack on downwards.
There is currently discussion ongoing whether to add this logic
to the standard kernel. In any case, it is not present in any
distribution currently available for zSeries ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED
home page where their
bug tracking system now resides ...
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux for S/390 Design Development
IBM Deutschland Entwicklung GmbH, Schoenaicher Str. 220, 71032 Boeblingen
Phone: +49-7031/16-3727 --- Email: [EMAIL
.
(-m64 doesn't hurt, it will simply be ignored.)
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux for S/390 Design Development
IBM Deutschland Entwicklung GmbH, Schoenaicher Str. 220, 71032 Boeblingen
Phone: +49-7031/16-3727 --- Email: [EMAIL
with e.g. the SLES-8 gdb.
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux for S/390 Design Development
IBM Deutschland Entwicklung GmbH, Schoenaicher Str. 220, 71032 Boeblingen
Phone: +49-7031/16-3727 --- Email: [EMAIL PROTECTED
Linux Symposium ;-)
See that paper for details why it is not OK.
We've gotten rid of chandev.conf for the 2.6 kernel. Network
devices are now configured via sysfs attributes, hopefully
abstracted away via a hotplug framework.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED
Gordon Wolfe wrote:
Is there a way for a Linux server running under z/VM to determine its =
own VM userid? Something using hcp perhaps? I can't use the IDENTIFY =
command since that's a CMS command.
/proc/sysinfo will show this (amongst other info) ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
by the address space limit of 2 GB).
I'd suggest to reduce the thread stack size (pthread_attr_setstacksize or so);
if you really need many large thread stacks, you can try to play with
/proc/.../map_base on distros that have it, or else go to 64 bit ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL
Weigand
--
Dr. Ulrich Weigand
Linux for S/390 Design Development
IBM Deutschland Entwicklung GmbH, Schoenaicher Str. 220, 71032 Boeblingen
Phone: +49-7031/16-3727 --- Email: [EMAIL PROTECTED]
is defined by the author
of the corresponding piece of code, I'd suggest to do the same as
you'd do for proposed code changes: contact either the author
of the device driver directly (if you know who that is), or else
write to [EMAIL PROTECTED]
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL
.
Only if your makefile is broken. I'm performing nearly all my
builds at -j6 or higher, normally without any problems.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
/messages file will contain a line
Out of Memory: Killed process ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
/cvsweb.cgi/gcc/gcc/config/s390/s390.c.diff?r1=1.70.2.8r2=1.70.2.9
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
managed to inline
everything into main due to better inlining heuristics,
which in turn allowed much more aggressive dead code detection.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
, you could simply
delete the one_cmpqi2 pattern without replacement.)
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
eye of some serious performance tester.
Well, if you have a specific example that shows this
behaviour, I'd be happy to have a look and see what
I can find out ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
. This
would argue for placing the literal pool apart from the code
that's using it.
[ Note that on older machines (pre-z900) with just one single
L1 cache for both code and data, the cache aspects would also
argue for placing the literal pool near to the code. ]
Bye,
Ulrich
--
Dr. Ulrich
on machines with separate i-cache and d-cache (i.e.
all zSeries machines). I don't have specific measurements to
prove that point, though.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
. Ulrich Weigand
[EMAIL PROTECTED]
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
look similar to
what you should expect from a 3.3 build? E.g. compare:
http://gcc.gnu.org/ml/gcc-testresults/2003-07/msg01260.html
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
can set up an account for
you, and you can SSH in.
If you have a specific problem you think I should look at, I'd be
happy to do so ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
, but I guess we can find out if
you're interested. What tool-chain (binutils, glibc) and kernel levels
are you using?
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
the number space, who should enforce consistency etc.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
. The exact value to choose depends a bit on
the circumstances; some experimentation may be required.
This should allow use of contiguous regions of nearly up to
2 GB ... (Of course, you might prefer to just use a 64-bit
kernel if your address space is this cramped ;-))
Bye,
Ulrich
--
Dr. Ulrich
the System.map), you
can at least find out which panic it was if for some reason the
console output didn't make it through anymore ...
What 'common cases' are you thinking about?
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
between 31-bit and 64-bit code that would need thunks. (The
only exception is that system calls need to convert the 31-bit
syscall arguments to the format the 64-bit kernel expects.)
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
'. However,
this application certainly overflowed all caches.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
applications
under a 64-bit kernel, which can in some cases combine the
benefit of faster and smaller 31-bit code with a larger amount
of physical memory.)
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
.
To my knowledge, no further documentation (apart from the actual source
code) will be released. In any case, the SIGA handling is done by the
QDIO infrastructure (which is already released as open source), not the
QETH network driver.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
, the largest logical vol with striping is 103GB (105600 MB).
PE size of 32M, same as 16M PE, 103GB.
This is weird. What does vgdisplay of that volume group show?
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
-shared read/write devices.
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux for S/390 Design Development
IBM Deutschland Entwicklung GmbH, Schoenaicher Str. 220, 71032 Boeblingen
Phone: +49-7031/16-3727 --- Email: [EMAIL PROTECTED]
make sure that the devices
are spread out over the internal paths on the storage subsystem;
see e.g. Klaus Bergmann's SHARE presentation for details on how
to do so.
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux for S/390 Design Development
IBM
-).
Is there anything specific you have in mind? (Things that are applicable
to modern storage subsystems -- we are not particularly interested in
supporting all the quirks of real 3390 devices at this point ...)
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
/Intel).
Do you have performance data comparing the scenarios?
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
or else a deficiency in the Linux kernel that
we'd need to fix.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
. Ulrich Weigand
[EMAIL PROTECTED]
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
2.4.17 doesn't have that any more as
far as I can see.
Once this is gone, the 5-second interval for kswapd is the next one to
address.
I'm not sure what you are talking about here ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
factor that is
broken for some reason. B.t.w. does this symptom change if you
switch the timer patch on/off?
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
be necessary. Could you send me
the .i file so I can check out what's going wrong here?
Thanks,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
if they can reproduce this, I can send them a copy
of the program that we wrote to do Step 4 from above. It's less than
100 lines of C code.
I'd appreciate it if you could send me this program.
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux for S/390
Hi Mark,
That did seem to correct the problem. Is this documented in the latest
Device Drivers manual?
I'm not sure. I'll check once I'm back from vacation ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
you try naming the interfaces ctc0 and ctc1?
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
are the
combinations that we can be reasonable sure are working ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
that the function
won't be found by the linker.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
code to use interruptible sleep instead.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
());
}
This calls ostream::write with a second argument of type
size_t, which is long on s390, not streamsize, which is
int on s390 ...
Does it help if you change that line to:
return o.write (s.data (), (streamsize) s.length ());
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
' facility; Linux definitely cannot
run on machines lacking that facility, not even the 'vintage'
patches will help you there ;-))
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
with kernel 2.4.17, so it depends on
your kernel level whether you have it or not.
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux for S/390 Design Development
IBM Deutschland Entwicklung GmbH, Schoenaicher Str. 220, 71032 Boeblingen
Phone: +49-7031/16
in, this is most likely a bug.)
Also, sometimes the configure/make scripts are not 100% cross-compile
ready, and make some decisions depending on the host architecture which
should really depend on the target ...
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux
variables internally to the compiler (e.g. for branch probabilities)
where different rounding could cause different optimization
decisions and thus different emitted code in a cross vs. a
native compilation.
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux
to tell the RPM that you only
want the X client libraries I don't know offhand ...
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux for S/390 Design Development
IBM Deutschland Entwicklung GmbH, Schoenaicher Str. 220, 71032 Boeblingen
Phone: +49-7031/16
would need at least as much effort as changing it to a multi-process
regime.
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux for S/390 Design Development
IBM Deutschland Entwicklung GmbH, Schoenaicher Str. 220, 71032 Boeblingen
Phone: +49-7031/16-3727
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
to implement some sort of authorization domains
on a finer granularity than processes, and even if this would
all work out somehow, what *benefit* this would bring over and
above what is already now possible in Unix using multiple
process as just described ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL
, what's in them, what pieces of code run under which
PSW key and/or mask, and at what places the keys are switched?)
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
to run on the same
machine, just send the raw binary image of your data structures
over the pipe and you're done.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
calls
in Linux for S/390 user space ...)
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
Mark Post wrote:
I'm trying to compile gdb 5.2 [...]
Please don't; use gdb 5.2.1 instead (ideally with the developerWorks patch).
gdb 5.2 is known not to build out-of-the-box on s390.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
*even in the presence of bugs* is ultimately
futile.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
only 2 GB
instead of 3 GB (and even *that* difference actually caused applications
to break).
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
just
fine. You might need new s390-tools, however.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
for
IS_ERR (msg) follows anyway, but this time under the proper lock.
You should also report this to the linux-kernel mailing list,
this is an architecture-indepdent bug.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
with the apparent solar noon ;-)
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
for example?)
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
-gcc.
One workaround is to first build a 'minimal' cross-gcc (without
shared libary or thread support, no other languages but C), use
this first-stage gcc to build a cross-glibc, and once that is
installed, rebuild a full cross-gcc.
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr
the original code.
The problem is that the semantics of the VM are identical to x86
semantics (i.e. the ADD ucode op sets just the same flag bits as
the x86 ADD instruction). That's the shortcuts you can take as
long as you stay on the same platform ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL
. Ulrich Weigand
[EMAIL PROTECTED]
,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
... ;-/)
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
that version (it should work for 2.4.7 as well).
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
Gerard Monteleone wrote:
../../gcc-3.2/gcc/unwind-dw2.c:1167: warning: implicit declaration of
function `__gthread_once'
Is this really the first error you see? I think there
should be some other messages leading up to this ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
as well ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
.
This is not a complete solution, but it will work on /usr/lib64-style
distros like SuSE. (B.t.w. if you look into the SuSE gcc source RPM,
you'll find a patch doing just that.)
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
Eddie Chen wrote:
Are there patches for GCC 3,2 under S390???
No, gcc 3.2 works as is on s390 and s390x.
Why, are you seeing any problems?
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
\
--host=s390-slackware-linux
You really should add --enable-__cxa_atexit to the configure line,
or else the resulting compiler will generate object code that does
not fully conform to the C++ ABI. See
http://gcc.gnu.org/gcc-3.2/c++-abi.html
for details.
Bye,
Ulrich
--
Dr. Ulrich
someone wants to port Mosix to s390,
maybe?)
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
?)
That's what the dynamic loader would be saying if
you had one. (Then again, it might be that ld.so also
just says 'No such file or directory' if a library is
missing unless you run it in ldd mode - I don't recall
right now.)
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
) I'd need a more specific bug report to help ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
only link 64-bit objects.
So, you'd need to perform the install step on a 32-bit system;
once the linking has been done, the resulting executable should
run on a 64-bit system.
Mit freundlichen Gruessen / Best Regards
Ulrich Weigand
--
Dr. Ulrich Weigand
Linux for S/390 Design Development
to
discuss this with the glibc maintainers; I'd like
to wait until Martin is back from vacation before
doing so, however.
For now, you can just make this one-line change;
that should fix the problem on s390.
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
1 - 100 of 122 matches
Mail list logo