DIPC for PowerPC

DIPC for PowerPC

[Kamran Karimi]

Hi,

 I did a blind port of DIPC (http://wallybox.cei.net/dipc) to PowerPC
processors. Here blind means that due to the lack of equipment, I have not
done any tests, not even a kernel compile. I would be grateful if
someone of this list could test the patch, available by anonymous FTP
from orion.cs.uregina.ca /pub/dipc. More adventurous people can test the
whole system (available from the same place). The following is the readme
file accompanying the new patch.

The file dipc-patch-1.1d.gz was prepared against a 2.2.10 linux
kernel. It is based on the DIPC 1.1c kernel patch and has the following
additions:
 
 *) The port to MIPS by Ralf Baechle (not tested in a cluster)
 *) The port to SPARC (blind, not compiled on target machine)
 *) The port to PowerPC (blind, not compiled on target machine)
 *) Some memory management modifications as proposed by David Miller (This
    might have introduced bugs into DIPC's shared memory subsystem)

 The rest of the package, including the user-space dameon dipcd, some tools
and example programs, and also the documentation of DIPC, can be found in
the file dipc-1.1c.tgz.

 Please test the patch, and if possible the whole system in a cluster, and
report the results to me or to linux-dipc@wallybox.cei.net.

-Kamran Karimi

 

[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]

Re: DIPC for PowerPC

[Paul Mackerras]

Kamran Karimi <karimi@cs.uregina.ca> wrote:

>  I did a blind port of DIPC (http://wallybox.cei.net/dipc) to PowerPC

I tried it on the weekend.  I had to make a couple of small changes to
get the kernel to compile: asm-ppc/ipc.h didn't define DIPC, and PPC
systems (at least the vast majority of the systems that Linux will run
on) are 32-bit, not 64-bit.  I have appended below the diffs for
include/asm-ppc (ipc.h and dipc.h).

I had to hack the userland stuff a bit to get it to compile, due to
the fact that we use glibc-2.1 which doesn't define union semun.  It
seems to work fine on one machine, but when I tried using two, dipcd
on the second machine wasn't creating the ~dipcd/referee_backdoor
socket which seems to be necessary.  I haven't figured out why yet.

Paul.

diff -urN linux/include/asm-ppc/dipc.h pmac/include/asm-ppc/dipc.h
--- linux/include/asm-ppc/dipc.h	Thu Jan  1 10:00:00 1970
+++ pmac/include/asm-ppc/dipc.h	Sun Jul 25 19:45:44 1999
@@ -0,0 +1,10 @@
+/*
+ * (C) Kamran Karimi
+ */
+ 
+#ifndef _ASM_DIPC_H
+#define _ASM_DIPC_H
+
+#define MY_DIPC_ARCH LINUX_BIG_32
+
+#endif _ASM_DIPC_H
diff -urN linux/include/asm-ppc/ipc.h pmac/include/asm-ppc/ipc.h
--- linux/include/asm-ppc/ipc.h	Fri May 14 22:08:03 1999
+++ pmac/include/asm-ppc/ipc.h	Sun Jul 25 20:04:09 1999
@@ -23,6 +23,9 @@
 #define SHMGET		23
 #define SHMCTL		24
 
+/* Used by the DIPC package, try and avoid reusing it */
+#define DIPC            25
+
 #define IPCCALL(version,op)	((version)<<16 | (op))
 
 #endif /* __PPC_IPC_H__ */

[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]

Re: DIPC for PowerPC

[Kamran Karimi]

Hi Paul,

On Mon, 26 Jul 1999, Paul Mackerras wrote:

> I tried it on the weekend.  I had to make a couple of small changes to
> get the kernel to compile: asm-ppc/ipc.h didn't define DIPC, and PPC
> systems (at least the vast majority of the systems that Linux will run
> on) are 32-bit, not 64-bit.  I have appended below the diffs for
> include/asm-ppc (ipc.h and dipc.h).

 Thanks. I'll incorporate your changes ASAP. Is there a way to determine 
at kernel compile time whether the machine is 32 bit or 64 bit?
 
> I had to hack the userland stuff a bit to get it to compile, due to
> the fact that we use glibc-2.1 which doesn't define union semun.  It
> seems to work fine on one machine, but when I tried using two, dipcd
> on the second machine wasn't creating the ~dipcd/referee_backdoor
> socket which seems to be necessary.  I haven't figured out why yet.

 Please inform me of the needed changes. Do you know if the changes will 
affect the compilation of the programs in other architectures? 

 It is normal for that UNIX socket to exist only in one machine, as only one 
computer in a DIPC cluster is the referee, and the socket is created there.

 Have you tried the benchmark programs?

-Kamran

[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]

Re: DIPC for PowerPC

[Tom Rini]

On Sun, 25 Jul 1999, Kamran Karimi wrote:

> 
> Hi Paul,
> 
> On Mon, 26 Jul 1999, Paul Mackerras wrote:
> 
> > I tried it on the weekend.  I had to make a couple of small changes to
> > get the kernel to compile: asm-ppc/ipc.h didn't define DIPC, and PPC
> > systems (at least the vast majority of the systems that Linux will run
> > on) are 32-bit, not 64-bit.  I have appended below the diffs for
> > include/asm-ppc (ipc.h and dipc.h).
> 
>  Thanks. I'll incorporate your changes ASAP. Is there a way to determine 
> at kernel compile time whether the machine is 32 bit or 64 bit?

It's called CONFIG_PPC64 right now I think..

---
Tom Rini (TR1265)
http://gate.crashing.org/~trini/


[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]

Re: DIPC for PowerPC

[Kamran Karimi]

On Tue, 27 Jul 1999, Tom Rini wrote:

> >  Thanks. I'll incorporate your changes ASAP. Is there a way to determine 
> > at kernel compile time whether the machine is 32 bit or 64 bit?
> 
> It's called CONFIG_PPC64 right now I think..

 Thanks. I have put a test package (dipc-1.1d.tgz) with the needed 
changes for downloading and test. Thanks to Paul the user-space 
applications in the package should now be compilable under glibc.

 The package is available by anonymous ftp from orion.cs.uregina.ca 
/pub/dipc.

-Kamran

[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]