Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[Laurent GUERBY]

Hi,

When linking a very large (> 100MB executable) application on
powerpc-linux with trunk we get linker errors:

.../lib/gcc/powerpc-unknown-linux-gnu/4.4.0/crtbegin.o:(.fini+0x0):
relocation truncated to fit: R_PPC_REL24 against `.text'
.../lib/gcc/powerpc-unknown-linux-gnu/4.4.0/crtend.o:(.init+0x0):
relocation truncated to fit: R_PPC_REL24 against `.text' 

The application itself is compiled with -mlongcall,
would adding -mlongcall to crtstuff.c Makefile rule help here?
If no, what is the proper solution GCC-wise?

Thanks in advance,

Laurent

Re: Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[Alan Modra]

On Mon, Jun 16, 2008 at 01:27:58PM +0200, Laurent GUERBY wrote:
> Hi,
> 
> When linking a very large (> 100MB executable) application on
> powerpc-linux with trunk we get linker errors:
> 
> .../lib/gcc/powerpc-unknown-linux-gnu/4.4.0/crtbegin.o:(.fini+0x0):
> relocation truncated to fit: R_PPC_REL24 against `.text'
> .../lib/gcc/powerpc-unknown-linux-gnu/4.4.0/crtend.o:(.init+0x0):
> relocation truncated to fit: R_PPC_REL24 against `.text' 
> 
> The application itself is compiled with -mlongcall,
> would adding -mlongcall to crtstuff.c Makefile rule help here?

It ought to.  You could also try GNU ld's --relax option, which might
also allow you to dispense with -mlongcall for your app.

> If no, what is the proper solution GCC-wise?

I'll note that this problem is exacerbated by the fact that crtbegin.o
defines the destructor function run from .fini and crtend.o defines
the constructor run from .init.  It really should be the other way
around, since this arrangement results in maximmum offset branches,
from the .init section located before .text to a function at the end
of .text, and from the .fini section located after .text to a function
at the beginning of .text.

-- 
Alan Modra
Australia Development Lab, IBM

Re: Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[Laurent GUERBY]

On Mon, 2008-06-16 at 23:36 +0930, Alan Modra wrote:
> On Mon, Jun 16, 2008 at 01:27:58PM +0200, Laurent GUERBY wrote:
> > Hi,
> > 
> > When linking a very large (> 100MB executable) application on
> > powerpc-linux with trunk we get linker errors:
> > 
> > .../lib/gcc/powerpc-unknown-linux-gnu/4.4.0/crtbegin.o:(.fini+0x0):
> > relocation truncated to fit: R_PPC_REL24 against `.text'
> > .../lib/gcc/powerpc-unknown-linux-gnu/4.4.0/crtend.o:(.init+0x0):
> > relocation truncated to fit: R_PPC_REL24 against `.text' 
> > 
> > The application itself is compiled with -mlongcall,
> > would adding -mlongcall to crtstuff.c Makefile rule help here?
> 
> It ought to.  You could also try GNU ld's --relax option, which might
> also allow you to dispense with -mlongcall for your app.

You mean adding to config/rs6000/t-linux64 and create
config/rs6000/t-linux with:

CRTSTUFF_T_CFLAGS += -mlongcalls
CRTSTUFF_T_CFLAGS_S += -mlongcalls

Would a patch with this be approved on trunk?

I noticed another target config/xtensa/t-elf which has something very
close:

# Build CRT files and libgcc with the "longcalls" option
CRTSTUFF_T_CFLAGS += -mlongcalls
CRTSTUFF_T_CFLAGS_S += -mlongcalls
TARGET_LIBGCC2_CFLAGS += -mlongcalls

Is the "TARGET_LIBGCC2_CFLAGS" necessary as well?

> > If no, what is the proper solution GCC-wise?
> 
> I'll note that this problem is exacerbated by the fact that crtbegin.o
> defines the destructor function run from .fini and crtend.o defines
> the constructor run from .init.  It really should be the other way
> around, since this arrangement results in maximmum offset branches,
> from the .init section located before .text to a function at the end
> of .text, and from the .fini section located after .text to a function
> at the beginning of .text.

Too arcane for me :). Let's hope someone else can comment on this.

Thanks for your help, I'll get some results tomorrow.

Laurent

Re: Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[Andreas Jaeger]

[-- Attachment #1: Type: text/plain, Size: 813 bytes --]


I had the same problem on my PowerPC Linux machine and the workaround
I'm using now is to run make with the additional flag STAGE1_CFLAGS=-O.
The problem (thanks to Andreas Schwab and Michael Matz for help) is that
.glink and .text is too far away from each other (> 32 MB in the
non-optimized case) and therefore building of e.g. cc1-dummy fails.
Adding -Wl,--relax to the cc1-dummy build helped for that binary but I
found no way to get this properly into the Makefiles (and make
LDFLAGS="-Wl,--relax" gave an error in configure later on),

Andreas
-- 
 Andreas Jaeger, Director Platform / openSUSE, aj@suse.de
  SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
   Maxfeldstr. 5, 90409 Nürnberg, Germany
    GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

[-- Attachment #2: Type: application/pgp-signature, Size: 188 bytes --]

Re: Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[David Edelsohn]

>>>>> Andreas Jaeger writes:

Andreas> I had the same problem on my PowerPC Linux machine and the workaround
Andreas> I'm using now is to run make with the additional flag STAGE1_CFLAGS=3D-O.
Andreas> The problem (thanks to Andreas Schwab and Michael Matz for help) is that
Andreas> .glink and .text is too far away from each other (> 32 MB in the
Andreas> non-optimized case) and therefore building of e.g. cc1-dummy fails.
Andreas> Adding -Wl,--relax to the cc1-dummy build helped for that binary but I
Andreas> found no way to get this properly into the Makefiles (and make
Andreas> LDFLAGS=3D"-Wl,--relax" gave an error in configure later on),

	Can you add

cc1-dummy$(exeext) : override LDFLAGS += -Wl,--relax

to an appropriate tmake_file in gcc/config/rs6000 ?

David

Re: Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[Andreas Schwab]

David Edelsohn <dje@watson.ibm.com> writes:

>>>>>> Andreas Jaeger writes:
>
> Andreas> I had the same problem on my PowerPC Linux machine and the workaround
> Andreas> I'm using now is to run make with the additional flag STAGE1_CFLAGS=3D-O.
> Andreas> The problem (thanks to Andreas Schwab and Michael Matz for help) is that
> Andreas> .glink and .text is too far away from each other (> 32 MB in the
> Andreas> non-optimized case) and therefore building of e.g. cc1-dummy fails.
> Andreas> Adding -Wl,--relax to the cc1-dummy build helped for that binary but I
> Andreas> found no way to get this properly into the Makefiles (and make
> Andreas> LDFLAGS=3D"-Wl,--relax" gave an error in configure later on),
>
> 	Can you add
>
> cc1-dummy$(exeext) : override LDFLAGS += -Wl,--relax
>
> to an appropriate tmake_file in gcc/config/rs6000 ?

It should be put in a host_xmake_file fragment.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[Andreas Jaeger]

On Sun, Jun 29, 2008 at 03:08:43PM +0200, Andreas Schwab wrote:
> David Edelsohn <dje@watson.ibm.com> writes:
> 
> >>>>>> Andreas Jaeger writes:
> >
> > Andreas> I had the same problem on my PowerPC Linux machine and the workaround
> > Andreas> I'm using now is to run make with the additional flag STAGE1_CFLAGS=3D-O.
> > Andreas> The problem (thanks to Andreas Schwab and Michael Matz for help) is that
> > Andreas> .glink and .text is too far away from each other (> 32 MB in the
> > Andreas> non-optimized case) and therefore building of e.g. cc1-dummy fails.
> > Andreas> Adding -Wl,--relax to the cc1-dummy build helped for that binary but I
> > Andreas> found no way to get this properly into the Makefiles (and make
> > Andreas> LDFLAGS=3D"-Wl,--relax" gave an error in configure later on),
> >
> > 	Can you add
> >
> > cc1-dummy$(exeext) : override LDFLAGS += -Wl,--relax
> >
> > to an appropriate tmake_file in gcc/config/rs6000 ?
> 
> It should be put in a host_xmake_file fragment.

I'm not a build machinery expert - if anybody has a patch, I'll happily test it,
I'm building on Linux/Powerpc64.

Adding the above line to rs6000/x-linux64 did not help me.  Btw.
we would need it for other files - like gnat1 - as well,

Andreas
-- 
 Andreas Jaeger, Director Platform / openSUSE, aj@suse.de
  SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
   Maxfeldstr. 5, 90409 Nürnberg, Germany
    GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

Re: Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[David Edelsohn]

>>>>> Andreas Jaeger writes:

Andreas> I'm not a build machinery expert - if anybody has a patch, I'll happily test it,
Andreas> I'm building on Linux/Powerpc64.

Andreas> Adding the above line to rs6000/x-linux64 did not help me.  Btw.
Andreas> we would need it for other files - like gnat1 - as well,

	Andreas Schwab removed x-linux64 yesterday because it was not
referenced by any configuration.

	I know that the tmake_files are the wrong place, but just to test
that it works, try adding the line to t-linux64, which is included in the
powerpc64-linux configuration.

	Once we confirm the command is appended to LDFLAGS for executables
that need it, we can figure out the correct Makefile fragment.

David

Re: Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[Andreas Jaeger]

[-- Attachment #1: Type: text/plain, Size: 2052 bytes --]

David Edelsohn <dje@watson.ibm.com> writes:

>>>>>> Andreas Jaeger writes:
>
> Andreas> I'm not a build machinery expert - if anybody has a patch, I'll happily test it,
> Andreas> I'm building on Linux/Powerpc64.
>
> Andreas> Adding the above line to rs6000/x-linux64 did not help me.  Btw.
> Andreas> we would need it for other files - like gnat1 - as well,
>
> 	Andreas Schwab removed x-linux64 yesterday because it was not
> referenced by any configuration.

Ah, that explains it.

>
> 	I know that the tmake_files are the wrong place, but just to test
> that it works, try adding the line to t-linux64, which is included in the
> powerpc64-linux configuration.
>
> 	Once we confirm the command is appended to LDFLAGS for executables
> that need it, we can figure out the correct Makefile fragment.

I added now to t-linux64:

cc1-dummy$(exeext) : override LDFLAGS += -Wl,--relax
cc1$(exeext) : override LDFLAGS += -Wl,--relax
gnat1$(exeext) : override LDFLAGS += -Wl,--relax

This gets me a bit further but I do get later a build failure in
configure:

checking for powerpc64-suse-linux-gnu-strip... strip
checking whether ln -s works... yes
checking for powerpc64-suse-linux-gnu-gcc... /abuild/aj/gcc-tst/./gcc/xgcc -B/abuild/aj/gcc-tst/./gcc/ -B/opt/gcc/4.4-devel/powerpc64-suse-linux-gnu/bin/ -B/opt/gcc/4.4-devel/powerpc64-suse-linux-gnu/lib/ -isystem /opt/gcc/4.4-devel/powerpc64-suse-linux-gnu/include -isystem /opt/gcc/4.4-devel/powerpc64-suse-linux-gnu/sys-include
checking for suffix of object files... configure: error: in `/abuild/aj/gcc-tst/powerpc64-suse-linux-gnu/libgcc':
configure: error: cannot compute suffix of object files: cannot compile
See `config.log' for more details.
make[2]: *** [configure-stage1-target-libgcc] Error 1

Andreas
-- 
 Andreas Jaeger, Director Platform / openSUSE, aj@suse.de
  SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
   Maxfeldstr. 5, 90409 Nürnberg, Germany
    GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

[-- Attachment #2: Type: application/pgp-signature, Size: 188 bytes --]

Re: Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[laurent]

Selon Andreas Jaeger <aj@suse.de>:

> This gets me a bit further but I do get later a build failure in
> configure:
>
> checking for powerpc64-suse-linux-gnu-strip... strip
> checking whether ln -s works... yes
> checking for powerpc64-suse-linux-gnu-gcc... /abuild/aj/gcc-tst/./gcc/xgcc
> -B/abuild/aj/gcc-tst/./gcc/
> -B/opt/gcc/4.4-devel/powerpc64-suse-linux-gnu/bin/
> -B/opt/gcc/4.4-devel/powerpc64-suse-linux-gnu/lib/ -isystem
> /opt/gcc/4.4-devel/powerpc64-suse-linux-gnu/include -isystem
> /opt/gcc/4.4-devel/powerpc64-suse-linux-gnu/sys-include
> checking for suffix of object files... configure: error: in
> `/abuild/aj/gcc-tst/powerpc64-suse-linux-gnu/libgcc':
> configure: error: cannot compute suffix of object files: cannot compile
> See `config.log' for more details.
> make[2]: *** [configure-stage1-target-libgcc] Error 1

If you've run configure even once by mistake in the source directory this is
usually what you get as error when you configure and build in a build dir later
on. The solution is to restart from a clean source dir.

Other possibility is that you linked a base compiler with mpfr as dynamic
library and you don't have them anymore in your LD_LIBRARY_PATH so it doesn't
work. There's a configure flag --disable-shared or some such for MPFR,
otherwise you can move the lib/*mpfr.so* somewhere else.

Hope this helps,

Laurent

Re: Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[Andreas Jaeger]

[-- Attachment #1: Type: text/plain, Size: 1995 bytes --]

laurent@guerby.net writes:

> Selon Andreas Jaeger <aj@suse.de>:
>
>> This gets me a bit further but I do get later a build failure in
>> configure:
>>
>> checking for powerpc64-suse-linux-gnu-strip... strip
>> checking whether ln -s works... yes
>> checking for powerpc64-suse-linux-gnu-gcc... /abuild/aj/gcc-tst/./gcc/xgcc
>> -B/abuild/aj/gcc-tst/./gcc/
>> -B/opt/gcc/4.4-devel/powerpc64-suse-linux-gnu/bin/
>> -B/opt/gcc/4.4-devel/powerpc64-suse-linux-gnu/lib/ -isystem
>> /opt/gcc/4.4-devel/powerpc64-suse-linux-gnu/include -isystem
>> /opt/gcc/4.4-devel/powerpc64-suse-linux-gnu/sys-include
>> checking for suffix of object files... configure: error: in
>> `/abuild/aj/gcc-tst/powerpc64-suse-linux-gnu/libgcc':
>> configure: error: cannot compute suffix of object files: cannot compile
>> See `config.log' for more details.
>> make[2]: *** [configure-stage1-target-libgcc] Error 1
>
> If you've run configure even once by mistake in the source directory this is
> usually what you get as error when you configure and build in a build dir later
> on. The solution is to restart from a clean source dir.

I'm not building inside the source directory.

> Other possibility is that you linked a base compiler with mpfr as dynamic
> library and you don't have them anymore in your LD_LIBRARY_PATH so it doesn't
> work. There's a configure flag --disable-shared or some such for MPFR,
> otherwise you can move the lib/*mpfr.so* somewhere else.
>
> Hope this helps,

The bootstrap works with make STAGE1_CFLAGS but fails when I add the
LDFLAGS lines to t-linux64, so this looks different.

config.log has:
xgcc: Internal error: Segmentation fault (program cc1)
Please submit a full bug report.


Andreas
-- 
 Andreas Jaeger, Director Platform / openSUSE, aj@suse.de
  SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
   Maxfeldstr. 5, 90409 Nürnberg, Germany
    GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

[-- Attachment #2: Type: application/pgp-signature, Size: 188 bytes --]

Re: Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[David Edelsohn]

>> checking for suffix of object files... configure: error: in
>> `/abuild/aj/gcc-tst/powerpc64-suse-linux-gnu/libgcc':
>> configure: error: cannot compute suffix of object files: cannot compile
>> See `config.log' for more details.

	This is the friendly way the GCC build process reports that the
compiler built during the previous stage does not work.  In other words,
-Wl,--relax allowed cc1 to link, but the resulting executable does not
run.

David

Re: Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[Andreas Jaeger]

[-- Attachment #1: Type: text/plain, Size: 959 bytes --]

David Edelsohn <dje@watson.ibm.com> writes:

>>> checking for suffix of object files... configure: error: in
>>> `/abuild/aj/gcc-tst/powerpc64-suse-linux-gnu/libgcc':
>>> configure: error: cannot compute suffix of object files: cannot compile
>>> See `config.log' for more details.
>
> 	This is the friendly way the GCC build process reports that the
> compiler built during the previous stage does not work.  In other words,
> -Wl,--relax allowed cc1 to link, but the resulting executable does not
> run.

So, it means that --relax is not the right solution for the problem.

I'll continue with the STAGE1_CFLAG flag but if anybody else wants me to
test something, please tell me,

Andreas
-- 
 Andreas Jaeger, Director Platform / openSUSE, aj@suse.de
  SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
   Maxfeldstr. 5, 90409 Nürnberg, Germany
    GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

[-- Attachment #2: Type: application/pgp-signature, Size: 188 bytes --]

Re: Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[David Edelsohn]

>>>>> Andreas Jaeger writes:

Andreas> So, it means that --relax is not the right solution for the problem.

Andreas> I'll continue with the STAGE1_CFLAG flag but if anybody else wants me to
Andreas> test something, please tell me,

	Maybe Alan will have some insight about --relax not working.

	Otherwise, in the past Alan has had some suggestions for swapping
around the crt file order or using linker scripts to place those sections
more effectively.

David

Re: Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[Alan Modra]

On Mon, Jun 30, 2008 at 11:04:43AM -0400, David Edelsohn wrote:
> >>>>> Andreas Jaeger writes:
> 
> Andreas> So, it means that --relax is not the right solution for the problem.
> 
> Andreas> I'll continue with the STAGE1_CFLAG flag but if anybody else wants me to
> Andreas> test something, please tell me,
> 
> 	Maybe Alan will have some insight about --relax not working.

I'm definitely interested in trying to reproduce the problem as it
sounds like there might be a linker bug.  Andreas, can you send me
your configure options?

> 	Otherwise, in the past Alan has had some suggestions for swapping
> around the crt file order or using linker scripts to place those sections
> more effectively.

These tricks would only help the particular case of reloc overflow in
branches to __do_global_{c,d}tors_aux.  Something like the following
(untested!) in place of the .text output section description in
ld/scripttempl/elf.sc ought to work.

  .text         ${RELOCATING-0} :
  {
    ${RELOCATING+${TEXT_START_SYMBOLS}}
    *crtend.o(.text .stub${RELOCATING+ .text.* .gnu.linkonce.t.*})
    *crtend?.o(.text .stub${RELOCATING+ .text.* .gnu.linkonce.t.*})
    *(EXCLUDE_FILE (*crtbegin.o *crtbegin?.o) .text .stub${RELOCATING+ .text.* .gnu.linkonce.t.*})
    *crtbegin.o(.text .stub${RELOCATING+ .text.* .gnu.linkonce.t.*})
    *crtbegin?.o(.text .stub${RELOCATING+ .text.* .gnu.linkonce.t.*})
    /* .gnu.warning sections are handled specially by elf32.em.  */
    *(.gnu.warning)
    ${RELOCATING+${OTHER_TEXT_SECTIONS}}
  } =${NOP-0}

-- 
Alan Modra
Australia Development Lab, IBM

Re: Linking very large application with GCC trunk on powerpc-linux leads to relocation error of crtbegin/end.o

[Alan Modra]

Found it, finally.  powerpc ld --relax had a bug affecting -fPIC
code.  See http://sourceware.org/ml/binutils/2008-08/msg00040.html

-- 
Alan Modra
Australia Development Lab, IBM