Skip to contentSkip to conversation

m32r: Cleanup linker script

GIT PULL Hirokazu Takata 2 participants 5 messages Sep 22, 2009 – Sep 25, 2009

lore
Share & tools
lore
#0 lore raw

Hi Linus,

This patchset cleans up the linker script for m32r.

From: Tim Abbott <[email protected]>

This cross-architecture linker script cleanup project is in preparation for adding support for building the kernel with -ffunction-sections -fdata-sections, which is a prerequisite for Ksplice.

Please pull the following updates.

---
The following changes since commit 43c1266ce4dc06bfd236cec31e11e9ecd69c0bef:
Linus Torvalds (1):
Merge branch 'perfcounters-rename-for-linus' of git://git.kernel.org/.../tip/linux-2.6-tip

are available in the git repository at:

git://www.linux-m32r.org/git/takata/linux-2.6_dev.git for-linus

Tim Abbott (6):
m32r: make PAGE_SIZE available to assembly.
m32r: Define THREAD_SIZE only once.
m32r: Move GET_THREAD_INFO definition out of asm/thread_info.h.
m32r: Remove unused .altinstructions and .exit.* code from linker script.
m32r: Move the spi_stack_top and spu_stack_top into .init.data section.
m32r: Cleanup linker script using new linker script macros.

 arch/m32r/include/asm/page.h        |    4 +-
 arch/m32r/include/asm/processor.h   |    2 -
 arch/m32r/include/asm/thread_info.h |   15 +------
 arch/m32r/kernel/entry.S            |    7 +++
 arch/m32r/kernel/head.S             |    4 +-
 arch/m32r/kernel/vmlinux.lds.S      |   78 ++++-------------------------------
 6 files changed, 23 insertions(+), 87 deletions(-)

---
commit 85233c43f7fece10a3ea8ed34f0d546b8dd3a435

Author: Tim Abbott <[email protected]>
Date:   Fri Sep 18 16:32:49 2009 -0400

m32r: Cleanup linker script using new linker script macros.

This patch is largely a straightforward conversion. One thing to note is that the new macros use fewer separate output sections than the old code; this should have no functional impact but is relevant for people objdumping vmlinux files.

Also note that it moves the .data.init_task output sections inside _edata.

commit 743486dff1d03eb4a67bd75402699629f70cd5b4

Author: Tim Abbott <[email protected]>
Date:   Fri Sep 18 16:32:48 2009 -0400

m32r: Move the spi_stack_top and spu_stack_top into .init.data section.

Since these get squashed into the .data output section by the m32r linker script, it seems likely that they don't need their own input sections.

At Hirokazu Takata's suggestion, we place these structures in .init.data rather than just placing them in .data (since they are only used at init time).

This patch is preparation for cleaning up the m32r architecture to use the new macros in vmlinux.lds.h; if these sections are indeed not needed, then we can use the RW_DATA_SECTION macro for m32r and save a bunch of redundant code.

commit 2a67d26284c31e676bcf3891ad8cee75e87fbada

Author: Tim Abbott <[email protected]>
Date:   Fri Sep 18 16:32:47 2009 -0400

m32r: Remove unused .altinstructions and .exit.* code from linker script.

It appears that m32r copied the .altinstructions definition from x86 when the architecture was first merged into Linux. m32r doesn't put anything in .altinstructions, so this is just dead code.

The following block affecting .exit.text/.exit.data, which has a comment also copied from x86, should also be deleted; the linker script later discards the .exit.text and .exit.data sections.

commit 00b01b246bc2e28763cbd85f0dc949d6c0d38c13

Author: Tim Abbott <[email protected]>
Date:   Fri Sep 18 16:32:46 2009 -0400

m32r: Move GET_THREAD_INFO definition out of asm/thread_info.h.

Previously, asm/thread_info.h was not usable from linker scripts because it contains a piece of .macro code. Since that code was only used in the m32r entry.S, the right fix is probably to move the macro there.

commit a7efb879ab5b07726873ace8305373bdf68be8a8

Author: Tim Abbott <[email protected]>
Date:   Fri Sep 18 16:32:45 2009 -0400

m32r: Define THREAD_SIZE only once.

Previously, m32r's asm/thread_info.h defined THREAD_SIZE differently for assembly and C code; now that PAGE_SIZE is usable from assembly, these can be combined. Also, m32r's asm/processor.h redefines THREAD_SIZE to the same value; remove this redundant definition.

commit 8b1c9befbfbd443bd95e77db01c70f1741aa2511

Author: Tim Abbott <[email protected]>
Date:   Fri Sep 18 16:32:44 2009 -0400

m32r: make PAGE_SIZE available to assembly.

page.h includes ifndef __ASSEMBLY__ guards, but PAGE_SIZE is defined using "1UL", which the assembler does not support. Use the _AC macro
from const.h to make it available to assembly (and linker scripts).

--
Hirokazu Takata <[email protected]>
Linux/M32R Project: http://www.linux-m32r.org/

#1 lore raw

Hi, Linus,

Sorry, there was a cross in the post.

I was not conscious of Sam's acknowledgement mail. And a newer patchset has been sent by Tim Abbott on Sep. 22, 2009 - 10:23:21 EST.

Quoted 1 line

[PATCH 00/13] Linker script cleanup patches for various architectures

The newer patchset includes updates for not only m32r but also parisc, xtenza, h8300, mn10300 and um.

What do I deal with such a cross architectural patchset?

-- Takata

At Wed, 23 Sep 2009 02:55:19 +0900,

Quoted 141 lines · Hirokazu Takata

Hi Linus,

This patchset cleans up the linker script for m32r.

From: Tim Abbott <[email protected]>

This cross-architecture linker script cleanup project is in preparation for adding support for building the kernel with -ffunction-sections -fdata-sections, which is a prerequisite for Ksplice.

Please pull the following updates.

---
The following changes since commit 43c1266ce4dc06bfd236cec31e11e9ecd69c0bef:
Linus Torvalds (1):
Merge branch 'perfcounters-rename-for-linus' of git://git.kernel.org/.../tip/linux-2.6-tip

are available in the git repository at:

git://www.linux-m32r.org/git/takata/linux-2.6_dev.git for-linus

Tim Abbott (6):
m32r: make PAGE_SIZE available to assembly.
m32r: Define THREAD_SIZE only once.
m32r: Move GET_THREAD_INFO definition out of asm/thread_info.h.
m32r: Remove unused .altinstructions and .exit.* code from linker script.
m32r: Move the spi_stack_top and spu_stack_top into .init.data section.
m32r: Cleanup linker script using new linker script macros.

 arch/m32r/include/asm/page.h        |    4 +-
 arch/m32r/include/asm/processor.h   |    2 -
 arch/m32r/include/asm/thread_info.h |   15 +------
 arch/m32r/kernel/entry.S            |    7 +++
 arch/m32r/kernel/head.S             |    4 +-
 arch/m32r/kernel/vmlinux.lds.S      |   78 ++++-------------------------------
 6 files changed, 23 insertions(+), 87 deletions(-)

---
commit 85233c43f7fece10a3ea8ed34f0d546b8dd3a435

Author: Tim Abbott <[email protected]>
Date:   Fri Sep 18 16:32:49 2009 -0400

m32r: Cleanup linker script using new linker script macros.

This patch is largely a straightforward conversion. One thing to note is that the new macros use fewer separate output sections than the old code; this should have no functional impact but is relevant for people objdumping vmlinux files.

Also note that it moves the .data.init_task output sections inside _edata.

commit 743486dff1d03eb4a67bd75402699629f70cd5b4

Author: Tim Abbott <[email protected]>
Date:   Fri Sep 18 16:32:48 2009 -0400

m32r: Move the spi_stack_top and spu_stack_top into .init.data section.

Since these get squashed into the .data output section by the m32r linker script, it seems likely that they don't need their own input sections.

At Hirokazu Takata's suggestion, we place these structures in .init.data rather than just placing them in .data (since they are only used at init time).

This patch is preparation for cleaning up the m32r architecture to use the new macros in vmlinux.lds.h; if these sections are indeed not needed, then we can use the RW_DATA_SECTION macro for m32r and save a bunch of redundant code.

commit 2a67d26284c31e676bcf3891ad8cee75e87fbada

Author: Tim Abbott <[email protected]>
Date:   Fri Sep 18 16:32:47 2009 -0400

m32r: Remove unused .altinstructions and .exit.* code from linker script.

It appears that m32r copied the .altinstructions definition from x86 when the architecture was first merged into Linux. m32r doesn't put anything in .altinstructions, so this is just dead code.

The following block affecting .exit.text/.exit.data, which has a comment also copied from x86, should also be deleted; the linker script later discards the .exit.text and .exit.data sections.

commit 00b01b246bc2e28763cbd85f0dc949d6c0d38c13

Author: Tim Abbott <[email protected]>
Date:   Fri Sep 18 16:32:46 2009 -0400

m32r: Move GET_THREAD_INFO definition out of asm/thread_info.h.

Previously, asm/thread_info.h was not usable from linker scripts because it contains a piece of .macro code. Since that code was only used in the m32r entry.S, the right fix is probably to move the macro there.

commit a7efb879ab5b07726873ace8305373bdf68be8a8

Author: Tim Abbott <[email protected]>
Date:   Fri Sep 18 16:32:45 2009 -0400

m32r: Define THREAD_SIZE only once.

Previously, m32r's asm/thread_info.h defined THREAD_SIZE differently for assembly and C code; now that PAGE_SIZE is usable from assembly, these can be combined. Also, m32r's asm/processor.h redefines THREAD_SIZE to the same value; remove this redundant definition.

commit 8b1c9befbfbd443bd95e77db01c70f1741aa2511

Author: Tim Abbott <[email protected]>
Date:   Fri Sep 18 16:32:44 2009 -0400

m32r: make PAGE_SIZE available to assembly.

page.h includes ifndef __ASSEMBLY__ guards, but PAGE_SIZE is defined using "1UL", which the assembler does not support. Use the _AC macro
from const.h to make it available to assembly (and linker scripts).

--
Hirokazu Takata <[email protected]>
Linux/M32R Project: http://www.linux-m32r.org/

#2 lore raw
Quoted 7 lines · Hirokazu Takata

Hi, Linus,

Sorry, there was a cross in the post.

I was not conscious of Sam's acknowledgement mail. And a newer patchset has been sent by Tim Abbott on Sep. 22, 2009 - 10:23:21 EST.

Quoted 6 lines
Quoted 1 line

[PATCH 00/13] Linker script cleanup patches for various architectures

The newer patchset includes updates for not only m32r but also parisc, xtenza, h8300, mn10300 and um.

What do I deal with such a cross architectural patchset?

Sorry for the miscommunication.

My patchset is just 5 orthogonal patchsets included together; there's no dependence between the various architectures. These were just the subset of the architecture linker script cleanups that were ready to be merged at that time.

The 6 m32r patches I sent to Linus are exactly the same as what's in your pull request, so it doesn't matter which one gets merged.

-Tim Abbott

#3 lore raw
Quoted 13 lines · Tim Abbott
Quoted 11 lines · Hirokazu Takata

Hi, Linus,

Sorry, there was a cross in the post.

I was not conscious of Sam's acknowledgement mail. And a newer patchset has been sent by Tim Abbott on Sep. 22, 2009 - 10:23:21 EST.

Quoted 1 line

[PATCH 00/13] Linker script cleanup patches for various architectures

The newer patchset includes updates for not only m32r but also parisc, xtenza, h8300, mn10300 and um.

...

Quoted 2 lines

The 6 m32r patches I sent to Linus are exactly the same as what's in your pull request, so it doesn't matter which one gets merged.

Since I'm resending that patch series to Linus for other reasons today, I'm going to drop the m32r patches from it (so Linus should merge these m32r linker script cleanups from the pull request).

Sorry again for the confusion.

Best regards,
-Tim Abbott

#4 lore raw

At Thu, 24 Sep 2009 10:27:06 -0400 (EDT),

Quoted 20 lines · Tim Abbott
Quoted 13 lines · Tim Abbott
Quoted 11 lines · Hirokazu Takata

Hi, Linus,

Sorry, there was a cross in the post.

I was not conscious of Sam's acknowledgement mail. And a newer patchset has been sent by Tim Abbott on Sep. 22, 2009 - 10:23:21 EST.

Quoted 1 line

[PATCH 00/13] Linker script cleanup patches for various architectures

The newer patchset includes updates for not only m32r but also parisc, xtenza, h8300, mn10300 and um.

...

Quoted 2 lines

The 6 m32r patches I sent to Linus are exactly the same as what's in your pull request, so it doesn't matter which one gets merged.

Quoted 9 lines

Since I'm resending that patch series to Linus for other reasons today, I'm going to drop the m32r patches from it (so Linus should merge these m32r linker script cleanups from the pull request).

Sorry again for the confusion.

Best regards,
-Tim Abbott

Thank you very much for your additional effort.

I will send a updated pull request for m32r to Linus again.

Best regards,

-- Takata