RE: Patch for 3.5/0.5 address space split

RE: Patch for 3.5/0.5 address space split

[gigag]

>I would try to use the whole aa patchset, I'm using the 3.5/0.5 
>split on one of the servers at work and it handles it fine with the 
>whole aa patchset
Could you be more specific on the version of the kernel and the 
patch that you applied? 
I tried 2.4.21 and 2.4.21rc8aa1 from Andrea and I couldn't boot the 
system with 3.5 user space.

Thanks
Giga

Re: Patch for 3.5/0.5 address space split

[gigag]

>As far as I remember I saw some places in the code which 
>should be fixed before split 1/3 can be used. Split 0.5/3.5 can be 
>not-working if you are using PAE-enabled kernel (in some places 
>pgds are copied directly - i.e. 1GB granularity is used).
Right. I've seen a mention on this all over the Net.

Anyway, I'm not using PAE since I have only 4GB of real memory. 
Also, I'm using a patch obtained at 
http://www.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.4/2
.4.20pre5aa1/00_3.5G-address-space-5

I need this split to be able to fully utilize the memory for my 
application, which I hope will work. Are you working on the patch 
for 2.4 or 2.5 kernel?

Could anybody suggest a configuration what is working?

Thanks
Giga

Re: Patch for 3.5/0.5 address space split

[Brian Jackson]

I would try to use the whole aa patchset, I'm using the 3.5/0.5 split on one 
of the servers at work and it handles it fine with the whole aa patchset.

--Brian Jackson

On Sunday 06 July 2003 11:00 am, gigag@bezeqint.net wrote:
> >As far as I remember I saw some places in the code which
> >should be fixed before split 1/3 can be used. Split 0.5/3.5 can be
> >not-working if you are using PAE-enabled kernel (in some places
> >pgds are copied directly - i.e. 1GB granularity is used).
>
> Right. I've seen a mention on this all over the Net.
>
> Anyway, I'm not using PAE since I have only 4GB of real memory.
> Also, I'm using a patch obtained at
> http://www.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.4/2
> .4.20pre5aa1/00_3.5G-address-space-5
>
> I need this split to be able to fully utilize the memory for my
> application, which I hope will work. Are you working on the patch
> for 2.4 or 2.5 kernel?
>
> Could anybody suggest a configuration what is working?
>
> Thanks
> Giga
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

-- 
OpenGFS -- http://opengfs.sourceforge.net
Home -- http://www.brianandsara.net

Patch for 3.5/0.5 address space split

[gigag]

Could anybody point out to patches available for 3.5/0.5 address 
space split for 2.4 and 2.5 kernels?

Any other working options? I managed to compile 2.4.21 kernel 
with 1/3 split, but not with 0.5/3.5. The last one simply doesn't 
boot. What could I be doing wrong?

Thanks
Giga

Re: Patch for 3.5/0.5 address space split

[Kirill Korotaev]

> Could anybody point out to patches available for 3.5/0.5 address
> space split for 2.4 and 2.5 kernels?
> Any other working options? I managed to compile 2.4.21 kernel
> with 1/3 split, but not with 0.5/3.5. The last one simply doesn't
> boot. What could I be doing wrong?

As far as I remember I saw some places in the code which should be fixed 
before split 1/3 can be used. Split 0.5/3.5 can be not-working if you are 
using PAE-enabled kernel (in some places pgds are copied directly - i.e. 1GB 
granularity is used).

Anyway due to bug in glibc pthread lib some apps are not working properly with 
non-standart 3/1 split (i.e. java).

I'm working on 4GB/4GB split now (kernel space is separated from user space 
completely). If someone wants to have a look at it I will probably post it 
later.

Kirill

Re: Patch for 3.5/0.5 address space split

[Martin Waitz]

[-- Attachment #1: Type: text/plain, Size: 712 bytes --]

hi :)

On Sun, Jul 06, 2003 at 07:47:49PM +0400, Kirill Korotaev wrote:
> I'm working on 4GB/4GB split now (kernel space is separated from user space 
> completely). If someone wants to have a look at it I will probably post it 
> later.
that would be very interesting.

could you please provide some details on how you plan to split the
address space?
thank you very much in advance!

-- 
CU,		  / Friedrich-Alexander University Erlangen, Germany
Martin Waitz	//  Department of Computer Science 3       _________
______________/// - - - - - - - - - - - - - - - - - - - - ///
dies ist eine manuell generierte mail, sie beinhaltet    //
tippfehler und ist auch ohne grossbuchstaben gueltig.   /

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: Patch for 3.5/0.5 address space split

[Martin J. Bligh]

> Could anybody point out to patches available for 3.5/0.5 address 
> space split for 2.4 and 2.5 kernels?

It's in 2.4-aa and 2.5-mjb trees. 2.5 has the added feature that it
can now do that for PAE (> 4GB) machines (from Dave Hansen). 

> Any other working options? I managed to compile 2.4.21 kernel 
> with 1/3 split, but not with 0.5/3.5. The last one simply doesn't 
> boot. What could I be doing wrong?

You can chage PAGE_OFFSET yourself, but there's a few places to change
it ... do a grep -r for "C0000000", and hack all those - one of them
is in some .lds file or something, I forget.

M.

Re: Patch for 3.5/0.5 address space split

[Herbert Poetzl]

On Sun, Jul 06, 2003 at 09:47:20AM -0700, Martin J. Bligh wrote:
> > Could anybody point out to patches available for 3.5/0.5 address 
> > space split for 2.4 and 2.5 kernels?
> 
> It's in 2.4-aa and 2.5-mjb trees. 2.5 has the added feature that it
> can now do that for PAE (> 4GB) machines (from Dave Hansen). 
> 
> > Any other working options? I managed to compile 2.4.21 kernel 
> > with 1/3 split, but not with 0.5/3.5. The last one simply doesn't 
> > boot. What could I be doing wrong?
> 
> You can chage PAGE_OFFSET yourself, but there's a few places to change
> it ... do a grep -r for "C0000000", and hack all those - one of them
> is in some .lds file or something, I forget.

I change it to use QEMU an x86 emulator ...

arch/i386/vmlinux.lds
include/asm-i386/page.h

the following patch changes it from 0xC0000000 to 0x90000000

HTH,
Herbert


diff -NurbP --minimal linux-2.4.21-P1/arch/i386/vmlinux.lds linux-2.4.21-P1-qemu/arch/i386/vmlinux.lds
--- linux-2.4.21-P1/arch/i386/vmlinux.lds       Mon Feb 25 20:37:53 2002
+++ linux-2.4.21-P1-qemu/arch/i386/vmlinux.lds  Fri Jun 27 15:48:35 2003
@@ -6,7 +6,7 @@
 ENTRY(_start)
 SECTIONS
 {
-  . = 0xC0000000 + 0x100000;
+  . = 0x90000000 + 0x100000;
   _text = .;                   /* Text and read-only data */
   .text : {
        *(.text)
diff -NurbP --minimal linux-2.4.21-P1/include/asm-i386/page.h linux-2.4.21-P1-qemu/include/asm-i386/page.h
--- linux-2.4.21-P1/include/asm-i386/page.h     Thu Jun 26 23:40:59 2003
+++ linux-2.4.21-P1-qemu/include/asm-i386/page.h        Fri Jun 27 15:47:52 2003
@@ -78,7 +78,7 @@
  * and CONFIG_HIGHMEM64G options in the kernel configuration.
  */
 
-#define __PAGE_OFFSET          (0xC0000000)
+#define __PAGE_OFFSET          (0x90000000)
 
 /*
  * This much address space is reserved for vmalloc() and iomap()



> 
> M.
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/