Skip to contentSkip to conversation

PowerPC: Change -fpic flag to -fPIC flag in the config.mk

PATCH New Chunhe Lan 5 participants 11 messages Apr 18, 2012 – Apr 27, 2012 1 message not in the archive PowerPC

lore
Share & tools
lore

Chunhe Lan

author
#0 lore raw

PATCH 1/1 PowerPC: Change -fpic flag to -fPIC flag in the config.mk

The -fPIC/-fpic flag belongs with -mrelocatable. The -fpic flag can limit the size of the GOT and produce smaller binaries, so it causes some GOT entries to be lost in the gcc 4.6 version. But -fPIC flag allows the maximum possible size of the GOT entries.

However, currently -mrelocatable promotes -fpic flag to -fPIC flag.

This reverts that portion of the
commit 33ee4c92339ee386662c0ee2d221098c5cc8b07e.

Signed-off-by Chunhe Lan <[email protected]>
---
 arch/powerpc/config.mk |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
arch/powerpc/config.mk +1 -1
index a307154..7c14ff6 100644
@@ -25,7 +25,7 @@ CROSS_COMPILE ?= ppc_8xx-
 
 CONFIG_STANDALONE_LOAD_ADDR ?= 0x40000
 LDFLAGS_FINAL += --gc-sections
-PLATFORM_RELFLAGS += -fpic -mrelocatable -ffunction-sections -fdata-sections
+PLATFORM_RELFLAGS += -fPIC -mrelocatable -ffunction-sections -fdata-sections
 PLATFORM_CPPFLAGS += -DCONFIG_PPC -D__powerpc__
 PLATFORM_LDFLAGS  += -n
Signature · 1 line
1.5.6.5
#1 lore raw

PATCH 1/1 PowerPC: Change -fpic flag to -fPIC flag in the config.mk

Dear Chunhe Lan,

Quoted 9 lines · In message <[email protected]> you…

The -fPIC/-fpic flag belongs with -mrelocatable. The -fpic flag can limit the size of the GOT and produce smaller binaries, so it causes some GOT entries to be lost in the gcc 4.6 version. But -fPIC flag allows the maximum possible size of the GOT entries.

However, currently -mrelocatable promotes -fpic flag to -fPIC flag.

This reverts that portion of the
commit 33ee4c92339ee386662c0ee2d221098c5cc8b07e.

As you state yourself, your modification has the negative impact of increasing the image size. What would be the benefits of it?

Are you trying to fix any specific problem? Which one? I am not aware of any related isses for any of the mainline PowerPC systems...

Best regards,

Wolfgang Denk

Signature · 4 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
If ignorance is bliss, why aren't there more happy people?

A message that is not in the archive (sent off-list or not yet received): [email protected]

#3 lore raw

PATCH 1/1 PowerPC: Change -fpic flag to -fPIC flag in the config.mk

Dear Chunhe Lan,

Quoted 4 lines · In message <[email protected]> you wrote:

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"> <html>
<head>
<meta content=3D"text/html;charset=3DUTF-8" http-equiv=3D"Content-Type"=

Message unreadable, ignored. Please post plain text only.

Wolfgang Denk

Signature · 5 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: [email protected]
I have the simplest tastes.  I am always satisfied with the best.
                                                       -- Oscar Wilde
#4 lore raw

PATCH 1/1 PowerPC: Change -fpic flag to -fPIC flag in the config.mk

In reply to Wolfgang Denk
Quoted 11 lines · Wolfgang Denk

Dear Chunhe Lan,

In message <[email protected]>

Quoted 7 lines · you

The -fPIC/-fpic flag belongs with -mrelocatable. The -fpic flag can limit the size of the GOT and produce smaller binaries, so it causes some GOT entries to be lost in the gcc 4.6 version. But -fPIC flag allows the maximum possible size of the GOT entries.

However, currently -mrelocatable promotes -fpic flag to -fPIC flag.

Quoted 8 lines
Quoted 2 lines

This reverts that portion of the
commit 33ee4c92339ee386662c0ee2d221098c5cc8b07e.

As you state yourself, your modification has the negative impact of increasing the image size. What would be the benefits of it?

Are you trying to fix any specific problem? Which one? I am not aware of any related isses for any of the mainline PowerPC systems...

Yes, I have fixed the booting issue of nand u-boot of
P1010/P1022/P1023/P2020 with this patch.

When use gcc 4.5 version, produce the size of u-boot-nand.bin to be
same with -fpic or -fPIC.
The u-boot-nand.bin is OK.

But when use gcc 4.6 version, produce the size of u-boot-nand.bin to
be different with -fpic or -fPIC.
The some GOT entries of u-boot-nand.bin are lost with -fpic(because
limit the size), so
u-boot-nand.bin hangs when booting. But use -fPIC, the
u-boot-nand.bin is OK.

And this patch reverts that portion of the commit
33ee4c92339ee386662c0ee2d221098c5cc8b07e.

Thanks,
Chunhe

Quoted 5 lines

Best regards,

Wolfgang Denk

#5 lore raw

PATCH 1/1 PowerPC: Change -fpic flag to -fPIC flag in the config.mk

Dear Chunhe Lan,

Quoted 8 lines · In message <[email protected]> you wrote:
Quoted 6 lines

As you state yourself, your modification has the negative impact of increasing the image size. What would be the benefits of it?

Are you trying to fix any specific problem? Which one? I am not aware of any related isses for any of the mainline PowerPC systems...

Yes, I have fixed the booting issue of nand u-boot of
P1010/P1022/P1023/P2020 with this patch.

What exactly is "the booting issue" you mention here? Are you referring to any messages previously posted here? Which?

Quoted 6 lines

But when use gcc 4.6 version, produce the size of u-boot-nand.bin to
be different with -fpic or -fPIC.
The some GOT entries of u-boot-nand.bin are lost with -fpic(because
limit the size), so
u-boot-nand.bin hangs when booting. But use -fPIC, the
u-boot-nand.bin is OK.

What do you mean "GOT entries ... are lost"? Are there any error messages?

If yes, what exactly are these?

If not, then this has to be considered a serious bug in either the compiler and the linker, and it should be fixed there rather than being papered over.

Best regards,

Wolfgang Denk

Signature · 5 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
I think it's a new feature. Don't tell anyone it was an accident. :-)
  -- Larry Wall on s/foo/bar/eieio in <[email protected]>
#6 lore raw

PATCH 1/1 PowerPC: Change -fpic flag to -fPIC flag in the config.mk

In reply to Chunhe Lan
Quoted 6 lines
Quoted 2 lines · Wolfgang Denk

Dear Chunhe Lan,

Quoted 8 lines
Quoted 8 lines

In message <[email protected]>

Quoted 6 lines · you

The -fPIC/-fpic flag belongs with -mrelocatable. The -fpic flag can limit the size of the GOT and produce smaller binaries, so it causes some GOT entries to be lost in the gcc 4.6 version. But -fPIC flag allows the maximum possible size of the GOT entries.

However, currently -mrelocatable promotes -fpic flag to -fPIC flag.

Not in newer gcc's. I added this to gcc some time ago and I think it got into gcc 4.6.

Quoted 16 lines
Quoted 9 lines
Quoted 3 lines

This reverts that portion of the
commit 33ee4c92339ee386662c0ee2d221098c5cc8b07e.

As you state yourself, your modification has the negative impact of increasing the image size. What would be the benefits of it?

Are you trying to fix any specific problem? Which one? I am not aware of any related isses for any of the mainline PowerPC systems...

Yes, I have fixed the booting issue of nand u-boot of
P1010/P1022/P1023/P2020 with this patch.

When use gcc 4.5 version, produce the size of u-boot-nand.bin to be
same with -fpic or -fPIC.
The u-boot-nand.bin is OK.

Quoted 6 lines

But when use gcc 4.6 version, produce the size of u-boot-nand.bin to
be different with -fpic or -fPIC.
The some GOT entries of u-boot-nand.bin are lost with -fpic(because
limit the size), so
u-boot-nand.bin hangs when booting. But use -fPIC, the
u-boot-nand.bin is OK.

Your linker file is buggy I think. I found u-boot-nand_spl.lds, is that the one? Check out that files reloc part:

.reloc : {
		_GOT2_TABLE_ = .;
		KEEP(*(.got2))
		_FIXUP_TABLE_ = .;
		KEEP(*(.fixup))
	}
	__got2_entries = (_FIXUP_TABLE_ - _GOT2_TABLE_) >> 2;
	__fixup_entries = (. - _FIXUP_TABLE_) >> 2;

Compare that with(from u-boot.lds):
.reloc :

  {
    _GOT2_TABLE_ = .;
    KEEP(*(.got2))
    KEEP(*(.got))
    PROVIDE(_GLOBAL_OFFSET_TABLE_ = . + 4);
    _FIXUP_TABLE_ = .;
    KEEP(*(.fixup))

}
__got2_entries = ((_GLOBAL_OFFSET_TABLE_ - _GOT2_TABLE_) >> 2) - 1; __fixup_entries = (. - _FIXUP_TABLE_) >> 2;

You may have to look at start.S for nand too

#7 lore raw

PATCH 1/1 PowerPC: Change -fpic flag to -fPIC flag in the config.mk

Quoted 1 line · Joakim Tjernlund

Your linker file is buggy I think. I found u-boot-nand_spl.lds, is that the one?

That's the one for the SPL part.

Quoted 19 lines

Check out that files reloc part:

.reloc : {
		_GOT2_TABLE_ = .;
		KEEP(*(.got2))
		_FIXUP_TABLE_ = .;
		KEEP(*(.fixup))
	}
	__got2_entries = (_FIXUP_TABLE_ - _GOT2_TABLE_) >> 2;
	__fixup_entries = (. - _FIXUP_TABLE_) >> 2;

Compare that with(from u-boot.lds):
.reloc :

  {
    _GOT2_TABLE_ = .;
    KEEP(*(.got2))
    KEEP(*(.got))
    PROVIDE(_GLOBAL_OFFSET_TABLE_ = . + 4);
    _FIXUP_TABLE_ = .;
    KEEP(*(.fixup))
Quoted 3 lines

}
__got2_entries = ((_GLOBAL_OFFSET_TABLE_ - _GOT2_TABLE_) >> 2) - 1; __fixup_entries = (. - _FIXUP_TABLE_) >> 2;

I notice a difference between u-boot.lds and u-boot-nand.lds -- the latter (used for the main part of U-Boot when loaded from SPL) has:

Quoted 10 lines
  .reloc   :
  {
    KEEP(*(.got))
    _GOT2_TABLE_ = .;
    KEEP(*(.got2))
    _FIXUP_TABLE_ = .;
    KEEP(*(.fixup))
  }
  __got2_entries = (_FIXUP_TABLE_ - _GOT2_TABLE_) >> 2;
  __fixup_entries = (. - _FIXUP_TABLE_) >> 2;

Is this wrong as well?

-Scott

#8 lore raw

PATCH 1/1 PowerPC: Change -fpic flag to -fPIC flag in the config.mk

Scott Wood <[email protected]> wrote on 2012/04/25 21:01:00:

Quoted 5 lines · Joakim Tjernlund
Quoted 1 line

Your linker file is buggy I think. I found u-boot-nand_spl.lds, is that the one?

That's the one for the SPL part.

OK, good.

Quoted 38 lines
Quoted 20 lines

Check out that files reloc part:

.reloc : {
      _GOT2_TABLE_ = .;
      KEEP(*(.got2))
      _FIXUP_TABLE_ = .;
      KEEP(*(.fixup))
   }
   __got2_entries = (_FIXUP_TABLE_ - _GOT2_TABLE_) >> 2;
   __fixup_entries = (. - _FIXUP_TABLE_) >> 2;

Compare that with(from u-boot.lds):
.reloc :

  {
    _GOT2_TABLE_ = .;
    KEEP(*(.got2))
    KEEP(*(.got))
    PROVIDE(_GLOBAL_OFFSET_TABLE_ = . + 4);
    _FIXUP_TABLE_ = .;
    KEEP(*(.fixup))
Quoted 3 lines

}
__got2_entries = ((_GLOBAL_OFFSET_TABLE_ - _GOT2_TABLE_) >> 2) - 1; __fixup_entries = (. - _FIXUP_TABLE_) >> 2;

I notice a difference between u-boot.lds and u-boot-nand.lds -- the latter (used for the main part of U-Boot when loaded from SPL) has:

Quoted 10 lines
  .reloc   :
  {
    KEEP(*(.got))
    _GOT2_TABLE_ = .;
    KEEP(*(.got2))
    _FIXUP_TABLE_ = .;
    KEEP(*(.fixup))
  }
  __got2_entries = (_FIXUP_TABLE_ - _GOT2_TABLE_) >> 2;
  __fixup_entries = (. - _FIXUP_TABLE_) >> 2;
Quoted 1 line

Is this wrong as well?

Yes, I think so. This one does not care about the .got entries at all so I suspect never gcc's wont work.

Jocke

#9 lore raw

PATCH 1/1 PowerPC: Change -fpic flag to -fPIC flag in the config.mk

Quoted 47 lines · Joakim Tjernlund

Scott Wood <[email protected]> wrote on 2012/04/25 21:01:00:

Quoted 5 lines · Joakim Tjernlund
Quoted 1 line

Your linker file is buggy I think. I found u-boot-nand_spl.lds, is that the one?

That's the one for the SPL part.

OK, good.

Quoted 38 lines
Quoted 20 lines

Check out that files reloc part:

.reloc : {
      _GOT2_TABLE_ = .;
      KEEP(*(.got2))
      _FIXUP_TABLE_ = .;
      KEEP(*(.fixup))
   }
   __got2_entries = (_FIXUP_TABLE_ - _GOT2_TABLE_) >> 2;
   __fixup_entries = (. - _FIXUP_TABLE_) >> 2;

Compare that with(from u-boot.lds):
.reloc :

  {
    _GOT2_TABLE_ = .;
    KEEP(*(.got2))
    KEEP(*(.got))
    PROVIDE(_GLOBAL_OFFSET_TABLE_ = . + 4);
    _FIXUP_TABLE_ = .;
    KEEP(*(.fixup))
Quoted 3 lines

}
__got2_entries = ((_GLOBAL_OFFSET_TABLE_ - _GOT2_TABLE_) >> 2) - 1; __fixup_entries = (. - _FIXUP_TABLE_) >> 2;

I notice a difference between u-boot.lds and u-boot-nand.lds -- the latter (used for the main part of U-Boot when loaded from SPL) has:

Quoted 10 lines
  .reloc   :
  {
    KEEP(*(.got))
    _GOT2_TABLE_ = .;
    KEEP(*(.got2))
    _FIXUP_TABLE_ = .;
    KEEP(*(.fixup))
  }
  __got2_entries = (_FIXUP_TABLE_ - _GOT2_TABLE_) >> 2;
  __fixup_entries = (. - _FIXUP_TABLE_) >> 2;
Quoted 4 lines
Quoted 1 line

Is this wrong as well?

Yes, I think so. This one does not care about the .got entries at all so I suspect never gcc's wont work.

Thanks.

I'm not entirely sure about the _GLOBAL_OFFSET_TABLE_ stuff, though -- what's wrong with the simpler version that uses _FIXUP_TABLE_?

-Scott

#10 lore raw

PATCH 1/1 PowerPC: Change -fpic flag to -fPIC flag in the config.mk

Scott Wood <[email protected]> wrote on 2012/04/26 23:26:52:

Quoted 51 lines · Joakim Tjernlund
Quoted 49 lines

Scott Wood <[email protected]> wrote on 2012/04/25 21:01:00:

Quoted 5 lines · Joakim Tjernlund
Quoted 1 line

Your linker file is buggy I think. I found u-boot-nand_spl.lds, is that the one?

That's the one for the SPL part.

OK, good.

Quoted 39 lines
Quoted 20 lines

Check out that files reloc part:

.reloc : {
      _GOT2_TABLE_ = .;
      KEEP(*(.got2))
      _FIXUP_TABLE_ = .;
      KEEP(*(.fixup))
   }
   __got2_entries = (_FIXUP_TABLE_ - _GOT2_TABLE_) >> 2;
   __fixup_entries = (. - _FIXUP_TABLE_) >> 2;

Compare that with(from u-boot.lds):
.reloc :

  {
    _GOT2_TABLE_ = .;
    KEEP(*(.got2))
    KEEP(*(.got))
    PROVIDE(_GLOBAL_OFFSET_TABLE_ = . + 4);
    _FIXUP_TABLE_ = .;
    KEEP(*(.fixup))
Quoted 3 lines

}
__got2_entries = ((_GLOBAL_OFFSET_TABLE_ - _GOT2_TABLE_) >> 2) - 1; __fixup_entries = (. - _FIXUP_TABLE_) >> 2;

I notice a difference between u-boot.lds and u-boot-nand.lds -- the latter (used for the main part of U-Boot when loaded from SPL) has:

Quoted 10 lines
  .reloc   :
  {
    KEEP(*(.got))
    _GOT2_TABLE_ = .;
    KEEP(*(.got2))
    _FIXUP_TABLE_ = .;
    KEEP(*(.fixup))
  }
  __got2_entries = (_FIXUP_TABLE_ - _GOT2_TABLE_) >> 2;
  __fixup_entries = (. - _FIXUP_TABLE_) >> 2;

Is this wrong as well?

Quoted 7 lines
Quoted 2 lines

Yes, I think so. This one does not care about the .got entries at all so I suspect never gcc's wont work.

Thanks.

I'm not entirely sure about the _GLOBAL_OFFSET_TABLE_ stuff, though -- what's wrong with the simpler version that uses _FIXUP_TABLE_?

_GLOBAL_OFFSET_TABLE_ is a -fpic thing, it marks the beginning of reloc entrires and gcc/binutils defines it if there are -fpic relocs present. If -fpic is present, there is are some extra entrires. If no -fpic I need to manually recreate space, otherwise the reloc routine will be off.

Jocke

#11 lore raw

PATCH 1/1 PowerPC: Change -fpic flag to -fPIC flag in the config.mk

Quoted 56 lines

Scott Wood <[email protected]> wrote on 2012/04/26 23:26:52:

Quoted 54 lines · Joakim Tjernlund
Quoted 51 lines

Scott Wood <[email protected]> wrote on 2012/04/25 21:01:00:

Quoted 5 lines · Joakim Tjernlund
Quoted 1 line

Your linker file is buggy I think. I found u-boot-nand_spl.lds, is that the one?

That's the one for the SPL part.

OK, good.

Quoted 38 lines
Quoted 20 lines

Check out that files reloc part:

.reloc : {
      _GOT2_TABLE_ = .;
      KEEP(*(.got2))
      _FIXUP_TABLE_ = .;
      KEEP(*(.fixup))
   }
   __got2_entries = (_FIXUP_TABLE_ - _GOT2_TABLE_) >> 2;
   __fixup_entries = (. - _FIXUP_TABLE_) >> 2;

Compare that with(from u-boot.lds):
.reloc :

  {
    _GOT2_TABLE_ = .;
    KEEP(*(.got2))
    KEEP(*(.got))
    PROVIDE(_GLOBAL_OFFSET_TABLE_ = . + 4);
    _FIXUP_TABLE_ = .;
    KEEP(*(.fixup))
Quoted 3 lines

}
__got2_entries = ((_GLOBAL_OFFSET_TABLE_ - _GOT2_TABLE_) >> 2) - 1; __fixup_entries = (. - _FIXUP_TABLE_) >> 2;

I notice a difference between u-boot.lds and u-boot-nand.lds -- the latter (used for the main part of U-Boot when loaded from SPL) has:

Quoted 10 lines
  .reloc   :
  {
    KEEP(*(.got))
    _GOT2_TABLE_ = .;
    KEEP(*(.got2))
    _FIXUP_TABLE_ = .;
    KEEP(*(.fixup))
  }
  __got2_entries = (_FIXUP_TABLE_ - _GOT2_TABLE_) >> 2;
  __fixup_entries = (. - _FIXUP_TABLE_) >> 2;
Quoted 1 line

Is this wrong as well?

Yes, I think so. This one does not care about the .got entries at all so I suspect never gcc's wont work.

Quoted 8 lines
Quoted 4 lines

Thanks.

I'm not entirely sure about the _GLOBAL_OFFSET_TABLE_ stuff, though -- what's wrong with the simpler version that uses _FIXUP_TABLE_?

_GLOBAL_OFFSET_TABLE_ is a -fpic thing, it marks the beginning of reloc entrires and gcc/binutils defines it if there are -fpic relocs present. If -fpic is present, there is are some extra entrires. If no -fpic I need to manually recreate space, otherwise the reloc routine will be off.

PS. note the -1 diff too:
__got2_entries = ((_GLOBAL_OFFSET_TABLE_ - _GOT2_TABLE_) >> 2) - 1;

Jocke