[PATCH]: for poor sools with old I2 & 64 bits kernel

[PATCH]: for poor sools with old I2 & 64 bits kernel

[Juan Quintela]

Hi
        this small patch made possible to compile a 64bit kernel for
        people that have old proms that only accept ecoff.  As usual
        stolen from the 32 bits version.

        The easiest way is creating the file in arch/mips/boot,
        otherwise we need to copy elf2ecoff.c to mips64.

Later, Juan.

Index: arch/mips64/Makefile
===================================================================
RCS file: /home/cvs/linux/arch/mips64/Makefile,v
retrieving revision 1.22.2.20
diff -u -r1.22.2.20 Makefile
--- arch/mips64/Makefile	26 Nov 2002 11:19:52 -0000	1.22.2.20
+++ arch/mips64/Makefile	19 Dec 2002 19:48:43 -0000
@@ -255,6 +255,9 @@
 	$(OBJCOPY) -O $(64bit-bfd) --remove-section=.reginfo --change-addresses=0xa800000080000000 $< $@
 endif
 
+vmlinux.ecoff: vmlinux
+	$(MAKE) -C arch/mips/boot $@
+
 zImage: vmlinux
 	@$(MAKEBOOT) zImage
 


-- 
In theory, practice and theory are the same, but in practice they 
are different -- Larry McVoy

Re: [PATCH]: for poor sools with old I2 & 64 bits kernel

[Thiemo Seufer]

Juan Quintela wrote:
> 
> Hi
>         this small patch made possible to compile a 64bit kernel for
>         people that have old proms that only accept ecoff.  As usual
>         stolen from the 32 bits version.
>
>         The easiest way is creating the file in arch/mips/boot,
>         otherwise we need to copy elf2ecoff.c to mips64.

Sorry, but I plainly doubt you have tested this.

I tried such a method today, and elf2ecoff failed to work on ELF64 files.
Maybe a recent enough objcopy is the better choice for this purpose.


Thiemo

Re: [PATCH]: for poor sools with old I2 & 64 bits kernel

[Juan Quintela]

>>>>> "thiemo" == Thiemo Seufer <ica2_ts@csv.ica.uni-stuttgart.de> writes:

thiemo> Juan Quintela wrote:
>> 
>> Hi
>> this small patch made possible to compile a 64bit kernel for
>> people that have old proms that only accept ecoff.  As usual
>> stolen from the 32 bits version.
>> 
>> The easiest way is creating the file in arch/mips/boot,
>> otherwise we need to copy elf2ecoff.c to mips64.

thiemo> Sorry, but I plainly doubt you have tested this.

thiemo> I tried such a method today, and elf2ecoff failed to work on ELF64 files.
thiemo> Maybe a recent enough objcopy is the better choice for this purpose.

I tested it and it works :)

Notice that I am using egcs for MIPS64 as I am not able to compile the
kernel with gcc-3.2 :(

Later, Juan.




-- 
In theory, practice and theory are the same, but in practice they 
are different -- Larry McVoy

Re: [PATCH]: for poor sools with old I2 & 64 bits kernel

[Thiemo Seufer]

Juan Quintela wrote:
> >>>>> "thiemo" == Thiemo Seufer <ica2_ts@csv.ica.uni-stuttgart.de> writes:
> 
> thiemo> Juan Quintela wrote:
> >> 
> >> Hi
> >> this small patch made possible to compile a 64bit kernel for
> >> people that have old proms that only accept ecoff.  As usual
> >> stolen from the 32 bits version.
> >> 
> >> The easiest way is creating the file in arch/mips/boot,
> >> otherwise we need to copy elf2ecoff.c to mips64.
> 
> thiemo> Sorry, but I plainly doubt you have tested this.
> 
> thiemo> I tried such a method today, and elf2ecoff failed to work on ELF64 files.
> thiemo> Maybe a recent enough objcopy is the better choice for this purpose.
> 
> I tested it and it works :)

How is this possible, given that elf2ecoff uses Elf32_* variables to
process the ELF file? elf2ecoff might be bad enough at error checking
to produce some binary garbage.

> Notice that I am using egcs for MIPS64 as I am not able to compile the
> kernel with gcc-3.2 :(

I don't think this is compiler dependent. :-)


Thiemo

Re: [PATCH]: for poor sools with old I2 & 64 bits kernel

[Juan Quintela]

>>>>> "thiemo" == Thiemo Seufer <ica2_ts@csv.ica.uni-stuttgart.de> writes:

thiemo> Juan Quintela wrote:
>> >>>>> "thiemo" == Thiemo Seufer <ica2_ts@csv.ica.uni-stuttgart.de> writes:
>> 
thiemo> Juan Quintela wrote:
>> >> 
>> >> Hi
>> >> this small patch made possible to compile a 64bit kernel for
>> >> people that have old proms that only accept ecoff.  As usual
>> >> stolen from the 32 bits version.
>> >> 
>> >> The easiest way is creating the file in arch/mips/boot,
>> >> otherwise we need to copy elf2ecoff.c to mips64.
>> 
thiemo> Sorry, but I plainly doubt you have tested this.
>> 
thiemo> I tried such a method today, and elf2ecoff failed to work on ELF64 files.
thiemo> Maybe a recent enough objcopy is the better choice for this purpose.
>> 
>> I tested it and it works :)

thiemo> How is this possible, given that elf2ecoff uses Elf32_* variables to
thiemo> process the ELF file? elf2ecoff might be bad enough at error checking
thiemo> to produce some binary garbage.

>> Notice that I am using egcs for MIPS64 as I am not able to compile the
>> kernel with gcc-3.2 :(

thiemo> I don't think this is compiler dependent. :-)

from  arch/mips64/Makefile:

#
# Some machines like the Indy need 32-bit ELF binaries for booting purposes.
# Other need ECOFF, so we build a 32-bit ELF binary for them which we then
# convert to ECOFF using elf2ecoff.
#
ifdef CONFIG_BOOT_ELF32
GCCFLAGS += -Wa,-32 $(shell if $(CC) -Wa,-mgp64 -c -o /dev/null -xc /dev/null > /dev/null 2>&1; then echo "-Wa,-mgp64"; fi)
LINKFLAGS += -T arch/mips64/ld.script.elf32
endif

-- 
In theory, practice and theory are the same, but in practice they 
are different -- Larry McVoy

Re: [PATCH]: for poor sools with old I2 & 64 bits kernel

[Thiemo Seufer]

Juan Quintela wrote:
[snip]
> >> Notice that I am using egcs for MIPS64 as I am not able to compile the
> >> kernel with gcc-3.2 :(
> 
> thiemo> I don't think this is compiler dependent. :-)
> 
> from  arch/mips64/Makefile:
> 
> #
> # Some machines like the Indy need 32-bit ELF binaries for booting purposes.
> # Other need ECOFF, so we build a 32-bit ELF binary for them which we then
> # convert to ECOFF using elf2ecoff.

Argh, this hack again! Sorry for the confusion, I'm just used to think
64bit == ELF64.


Thiemo

Re: [PATCH]: for poor sools with old I2 & 64 bits kernel

[Ralf Baechle]

On Thu, Dec 19, 2002 at 09:04:12PM +0100, Juan Quintela wrote:

>         this small patch made possible to compile a 64bit kernel for
>         people that have old proms that only accept ecoff.  As usual
>         stolen from the 32 bits version.
> 
>         The easiest way is creating the file in arch/mips/boot,
>         otherwise we need to copy elf2ecoff.c to mips64.

Applied slightly modified.  I removed two other unused targets.

  Ralf

Re: [PATCH]: for poor sools with old I2 & 64 bits kernel

[Juan Quintela]

>>>>> "ralf" == Ralf Baechle <ralf@linux-mips.org> writes:

ralf> On Thu, Dec 19, 2002 at 09:04:12PM +0100, Juan Quintela wrote:
>> this small patch made possible to compile a 64bit kernel for
>> people that have old proms that only accept ecoff.  As usual
>> stolen from the 32 bits version.
>> 
>> The easiest way is creating the file in arch/mips/boot,
>> otherwise we need to copy elf2ecoff.c to mips64.

ralf> Applied slightly modified.  I removed two other unused targets.

Please, add that back, and things will indeed compile :)

Later, Juan.

Index: arch/mips64/Makefile
===================================================================
RCS file: /home/cvs/linux/arch/mips64/Makefile,v
retrieving revision 1.22.2.21
diff -u -r1.22.2.21 Makefile
--- arch/mips64/Makefile	20 Dec 2002 02:42:26 -0000	1.22.2.21
+++ arch/mips64/Makefile	20 Dec 2002 14:37:18 -0000
@@ -235,7 +235,7 @@
 CORE_FILES := arch/mips64/kernel/kernel.o arch/mips64/mm/mm.o $(CORE_FILES)
 LIBS := arch/mips64/lib/lib.a $(LIBS)
 
-MAKEBOOT = $(MAKE) -C arch/$(ARCH)/boot
+MAKEBOOT = $(MAKE) -C arch/mips/boot
 
 ifdef CONFIG_CPU_LITTLE_ENDIAN
 64bit-bfd = elf64-tradlittlemips



-- 
In theory, practice and theory are the same, but in practice they 
are different -- Larry McVoy

Re: [PATCH]: for poor sools with old I2 & 64 bits kernel

[Maciej W. Rozycki]

On 20 Dec 2002, Juan Quintela wrote:

> ralf> Applied slightly modified.  I removed two other unused targets.
> 
> Please, add that back, and things will indeed compile :)

 Ralf did that right -- it's annoying to have the COFF image for mips64
built in the mips tree.  Unlike the intermediate object files, this target
is visible to a non-developer.  But a few bits are missing, indeed.  How
about the following patch?  It's a trivial modification of what I use for
about half a year now.

-- 
+  Maciej W. Rozycki, Technical University of Gdansk, Poland   +
+--------------------------------------------------------------+
+        e-mail: macro@ds2.pg.gda.pl, PGP key available        +

patch-mips-2.4.20-pre6-20021220-mips64-ecoff-0
diff -up --recursive --new-file linux-mips-2.4.20-pre6-20021220.macro/arch/mips64/boot/Makefile linux-mips-2.4.20-pre6-20021220/arch/mips64/boot/Makefile
--- linux-mips-2.4.20-pre6-20021220.macro/arch/mips64/boot/Makefile	2002-06-26 03:04:47.000000000 +0000
+++ linux-mips-2.4.20-pre6-20021220/arch/mips64/boot/Makefile	2002-12-21 14:23:32.000000000 +0000
@@ -22,11 +22,11 @@ all: vmlinux.ecoff addinitrd
 vmlinux.ecoff:	$(CONFIGURE) elf2ecoff $(TOPDIR)/vmlinux
 	./elf2ecoff $(TOPDIR)/vmlinux vmlinux.ecoff $(E2EFLAGS)
 
-elf2ecoff: elf2ecoff.c
-	$(HOSTCC) -o $@ $^
+elf2ecoff: $(TOPDIR)/arch/mips/boot/elf2ecoff.c
+	$(HOSTCC) -I$(TOPDIR)/arch/mips/boot -I- -o $@ $^
 
-addinitrd: addinitrd.c
-	$(HOSTCC) -o $@ $^
+addinitrd: $(TOPDIR)/arch/mips/boot/addinitrd.c
+	$(HOSTCC) -I$(TOPDIR)/arch/mips/boot -I- -o $@ $^
 
 # Don't build dependencies, this may die if $(CC) isn't gcc
 dep:

Re: [PATCH]: for poor sools with old I2 & 64 bits kernel

[Juan Quintela]

>>>>> "maciej" == Maciej W Rozycki <macro@ds2.pg.gda.pl> writes:

maciej> On 20 Dec 2002, Juan Quintela wrote:
ralf> Applied slightly modified.  I removed two other unused targets.
>> 
>> Please, add that back, and things will indeed compile :)

maciej> Ralf did that right -- it's annoying to have the COFF image for mips64
maciej> built in the mips tree.  Unlike the intermediate object files, this target
maciej> is visible to a non-developer.  But a few bits are missing, indeed.  How
maciej> about the following patch?  It's a trivial modification of what I use for
maciej> about half a year now.

BTW, are you using mips64 in a r4k? If so, do you need any additional
patches?

I am having some memory corruption :(


maciej> -- 
maciej> +  Maciej W. Rozycki, Technical University of Gdansk, Poland   +
maciej> +--------------------------------------------------------------+
maciej> +        e-mail: macro@ds2.pg.gda.pl, PGP key available        +

maciej> patch-mips-2.4.20-pre6-20021220-mips64-ecoff-0
maciej> diff -up --recursive --new-file linux-mips-2.4.20-pre6-20021220.macro/arch/mips64/boot/Makefile linux-mips-2.4.20-pre6-20021220/arch/mips64/boot/Makefile
maciej> --- linux-mips-2.4.20-pre6-20021220.macro/arch/mips64/boot/Makefile	2002-06-26 03:04:47.000000000 +0000
maciej> +++ linux-mips-2.4.20-pre6-20021220/arch/mips64/boot/Makefile	2002-12-21 14:23:32.000000000 +0000
maciej> @@ -22,11 +22,11 @@ all: vmlinux.ecoff addinitrd
maciej> vmlinux.ecoff:	$(CONFIGURE) elf2ecoff $(TOPDIR)/vmlinux
maciej> ./elf2ecoff $(TOPDIR)/vmlinux vmlinux.ecoff $(E2EFLAGS)
 
maciej> -elf2ecoff: elf2ecoff.c
maciej> -	$(HOSTCC) -o $@ $^
maciej> +elf2ecoff: $(TOPDIR)/arch/mips/boot/elf2ecoff.c
maciej> +	$(HOSTCC) -I$(TOPDIR)/arch/mips/boot -I- -o $@ $^
 
maciej> -addinitrd: addinitrd.c
maciej> -	$(HOSTCC) -o $@ $^
maciej> +addinitrd: $(TOPDIR)/arch/mips/boot/addinitrd.c
maciej> +	$(HOSTCC) -I$(TOPDIR)/arch/mips/boot -I- -o $@ $^
 
maciej> # Don't build dependencies, this may die if $(CC) isn't gcc
maciej> dep:

It is better than mine.

Later, Juan.

-- 
In theory, practice and theory are the same, but in practice they 
are different -- Larry McVoy

Re: [PATCH]: for poor sools with old I2 & 64 bits kernel

[Maciej W. Rozycki]

On 21 Dec 2002, Juan Quintela wrote:

> BTW, are you using mips64 in a r4k? If so, do you need any additional
> patches?

 Yep, for quite some time now, running a DECstation 5000/260 with an
R4400SC.  Yep, a few. 

 I'm merging the patches slowly, but it's not that easy.  Errata for the
R4000 and the R4400 require toolchain changes and bits in the patches
depend on fixed tools.  So chances are I won't merge everything until
changes are applied to tools. 

> I am having some memory corruption :(

 What kind of?  And what processor (PRId) have you got? 

 My system seems reasonably stable, but sometimes it crashes under a load.
I have yet to get at tracking the problem down.

-- 
+  Maciej W. Rozycki, Technical University of Gdansk, Poland   +
+--------------------------------------------------------------+
+        e-mail: macro@ds2.pg.gda.pl, PGP key available        +

Re: [PATCH]: for poor sools with old I2 & 64 bits kernel

[Juan Quintela]

>>>>> "maciej" == Maciej W Rozycki <macro@ds2.pg.gda.pl> writes:

maciej> On 21 Dec 2002, Juan Quintela wrote:
>> BTW, are you using mips64 in a r4k? If so, do you need any additional
>> patches?

maciej> Yep, for quite some time now, running a DECstation 5000/260 with an
maciej> R4400SC.  Yep, a few. 

Here it is bigendian (SGI Indigo 2).

ARCH: SGI-IP22                                                                  
PROMLIB: ARC firmware Version 1 Revision 10                                     
CPU: MIPS-R4400 FPU<MIPS-R4400FPC> ICACHE DCACHE SCACHE                         
CPU revision is: 00000450                                                       
FPU revision is: 00000500                                                       
Primary instruction cache 16kb, linesize 16 bytes.                              
Primary data cache 16kb, linesize 16 bytes.                                     
Secondary cache sized at 1024K linesize 128 bytes.                              

I am using the egcs-1.1.2 from ralf for doing compiles, I think that
you have a more modern compiler, if so, I will be happy to download.

maciej> I'm merging the patches slowly, but it's not that easy.  Errata for the
maciej> R4000 and the R4400 require toolchain changes and bits in the patches
maciej> depend on fixed tools.  So chances are I won't merge everything until
maciej> changes are applied to tools. 

>> I am having some memory corruption :(

maciej> What kind of?  And what processor (PRId) have you got? 

PRid show before.

Corruption is that when I do a ssh to that host, I got parts of
/proc/ksyms into the console.


maciej> My system seems reasonably stable, but sometimes it crashes under a load.
maciej> I have yet to get at tracking the problem down.

mine in 32bits is stable, don't crash under load (not very high load
yet).  But in 64bits (with exactly the same userland) got losts of
problems.

Already found a couple of problems in c-r4k.c send to the list, and
now a couple of problems in sgiseeq.c and sgiserial.c.  Notice that I
am running that machine with nfsroot, i.e. I don't have basically more
devices than the serial console and the network card.  No real console
support and not harddisk support either.

If you have any patches, I will like to take a look.

Later, Juan.

-- 
In theory, practice and theory are the same, but in practice they 
are different -- Larry McVoy

Re: [PATCH]: for poor sools with old I2 & 64 bits kernel

[Maciej W. Rozycki]

On 21 Dec 2002, Juan Quintela wrote:

> maciej> Yep, for quite some time now, running a DECstation 5000/260 with an
> maciej> R4400SC.  Yep, a few. 
> 
> Here it is bigendian (SGI Indigo 2).

 The endianness shouldn't matter, but who knows?

> ARCH: SGI-IP22                                                                  
> PROMLIB: ARC firmware Version 1 Revision 10                                     
> CPU: MIPS-R4400 FPU<MIPS-R4400FPC> ICACHE DCACHE SCACHE                         
> CPU revision is: 00000450                                                       
> FPU revision is: 00000500                                                       
> Primary instruction cache 16kb, linesize 16 bytes.                              
> Primary data cache 16kb, linesize 16 bytes.                                     
> Secondary cache sized at 1024K linesize 128 bytes.                              

 Please get: 
'ftp://ftp.ds2.pg.gda.pl/pub/macro/linux/patch-mips-2.4.20-pre6-20021212-mips-bugs-14mod.gz'
try it and send me the boot log, specifically the bug checks.  My PRId is
00000440 and that is definitely buggy revision 1.0.  Yours is probably
revision 2.0 and should work better, but it won't hurt checking.

 Note the patch is unfinished code; I had to modify it a bit for you to
make it work at all with standard tools.  It does not necessarily make
much sense. ;-) 

> I am using the egcs-1.1.2 from ralf for doing compiles, I think that
> you have a more modern compiler, if so, I will be happy to download.

 Indeed, I use heavily patched gcc 2.95.4.  It's available at my site,
too, as RPM packages.  For mips64 there are only source and i386 binary
packages of a bootstrap cross-compiler -- look for
mips64el-linux-boot-gcc.  It should be fairly easy to build big-endian
packages from the sources.

 But there is a drawback -- the packages have a patch to handle the daddiu
erratum and the workaround is unconditional now.  The result is somewhat
worse code is emitted.

> Corruption is that when I do a ssh to that host, I got parts of
> /proc/ksyms into the console.

 SSH works just fine for me.  I haven't tried connecting over IPv6 to
64-bit Linux, though (for 32-bit version it works).

> maciej> My system seems reasonably stable, but sometimes it crashes under a load.
> maciej> I have yet to get at tracking the problem down.
> 
> mine in 32bits is stable, don't crash under load (not very high load
> yet).  But in 64bits (with exactly the same userland) got losts of
> problems.

 Well, for me the 32-bit configuration is rock-solid.  It's the 64-bit one
that causes some problems, but it's stable enough for most of the recent
package builds to be done with it.  I debugged showstopper problems last
Summer.

> Already found a couple of problems in c-r4k.c send to the list, and

 I haven't seen any problems with caching recently.  I might have been
lucky.  But my cache configuration differs a bit:

CPU revision is: 00000440
FPU revision is: 00000500
Loading R4000 MMU routines.
Primary instruction cache 16kb, linesize 16 bytes.
Primary data cache 16kb, linesize 16 bytes.
Secondary cache sized at 1024K linesize 32 bytes.

> now a couple of problems in sgiseeq.c and sgiserial.c.  Notice that I

 That's driver-specific.  The declance.c and zs.c drivers work.

> am running that machine with nfsroot, i.e. I don't have basically more
> devices than the serial console and the network card.  No real console
> support and not harddisk support either.

 Ditto here.  Except a keyboard and a framebuffer also work; the latter
also under X11. 

> If you have any patches, I will like to take a look.

 Nothing ready for inclusion.  And probably nothing critical (except for
the DECstation), but please send me the report first. 

  Maciej

-- 
+  Maciej W. Rozycki, Technical University of Gdansk, Poland   +
+--------------------------------------------------------------+
+        e-mail: macro@ds2.pg.gda.pl, PGP key available        +