Skip to contentSkip to conversation

Eliminate most of sparc_ksyms_*.c

PATCH Applied Julian Calaby 4 participants 8 messages Jan 8, 2009 – Jan 9, 2009

lore
Share & tools
lore
#0 lore raw

v2 cover Eliminate most of sparc_ksyms_*.c

Hi all,

This patch set eliminates most of the definitions within arch/sparc/kernel/sparc_ksyms_*.c by moving them to their rightful places.

This is done within patches 1-3.

Note that this work was originally performed by Sam Ravnborg and these patches have been rebased and added to by myself.

I've tried to reflect this in the commit messages but may have gotten it a little wrong.

Patches 4 and 5 are cleanups that I had in my tree for an eventual oplib_*.h unification - but as this screws up my original patches and they're so small, I figured this'd be as good a time as any to send them.

Changes in version 2:
1. Patches 2 and 3 have a few more includes added to the modified files.
2. It shouldn't be word wrapped this time!

Thanks,

Julian Calaby

#1 lore raw
Quoted 13 lines · Julian Calaby

Hi all,

This patch set eliminates most of the definitions within arch/sparc/kernel/sparc_ksyms_*.c by moving them to their rightful places.

This is done within patches 1-3.

Note that this work was originally performed by Sam Ravnborg and these patches have been rebased and added to by myself.

I've tried to reflect this in the commit messages but may have gotten it a little wrong.

Quoted 7 lines

Patches 4 and 5 are cleanups that I had in my tree for an eventual oplib_*.h unification - but as this screws up my original patches and they're so small, I figured this'd be as good a time as any to send them.

Changes in version 2:
1. Patches 2 and 3 have a few more includes added to the modified files.
2. It shouldn't be word wrapped this time!

Hi Julian.

I have tested your patches by applying them one-by-one and done both sparc and sparc64 builds.

Everything worked as expected - good work!

Sam

#2 lore raw
In reply to Julian Calaby
Quoted 22 lines · Sam Ravnborg
Quoted 13 lines · Julian Calaby

Hi all,

This patch set eliminates most of the definitions within arch/sparc/kernel/sparc_ksyms_*.c by moving them to their rightful places.

This is done within patches 1-3.

Note that this work was originally performed by Sam Ravnborg and these patches have been rebased and added to by myself.

I've tried to reflect this in the commit messages but may have gotten it a little wrong.

Quoted 7 lines

Patches 4 and 5 are cleanups that I had in my tree for an eventual oplib_*.h unification - but as this screws up my original patches and they're so small, I figured this'd be as good a time as any to send them.

Changes in version 2:
1. Patches 2 and 3 have a few more includes added to the modified files.
2. It shouldn't be word wrapped this time!

Quoted 6 lines

Hi Julian.

I have tested your patches by applying them one-by-one and done both sparc and sparc64 builds.

Everything worked as expected - good work!

I forgot to mention that I also looked through all the patches and everything looked good(*).

(*) It does not look "good" that we have all the local declarations in ksyms.c but to eliminate them is another set of patches that has lower priority.

Sam

#3 lore raw
In reply to Julian Calaby
Quoted 22 lines · Sam Ravnborg
Quoted 13 lines · Julian Calaby

Hi all,

This patch set eliminates most of the definitions within arch/sparc/kernel/sparc_ksyms_*.c by moving them to their rightful places.

This is done within patches 1-3.

Note that this work was originally performed by Sam Ravnborg and these patches have been rebased and added to by myself.

I've tried to reflect this in the commit messages but may have gotten it a little wrong.

Quoted 7 lines

Patches 4 and 5 are cleanups that I had in my tree for an eventual oplib_*.h unification - but as this screws up my original patches and they're so small, I figured this'd be as good a time as any to send them.

Changes in version 2:
1. Patches 2 and 3 have a few more includes added to the modified files.
2. It shouldn't be word wrapped this time!

Quoted 6 lines

Hi Julian.

I have tested your patches by applying them one-by-one and done both sparc and sparc64 builds.

Everything worked as expected - good work!

Thanks for testing and reviewing them.

Hopefully we can fix the declarations at some point.

My stance on this is: as this functionality that is defined in assembler, having nasty C code is acceptable - providing that it's documented. (which we could improve somewhat)

I plan to merge sparc_ksyms_*.c into (possibly) another ksyms.c, however I have a compile error in a module to fix first.

Thanks for all your help on this and for looking over my patches.

Signature · 3 lines
Julian Calaby

Email: [email protected]
#4 lore raw
In reply to Julian Calaby

From: "Julian Calaby" <[email protected]>
Date: Fri, 9 Jan 2009 08:23:41 +1100

Quoted 3 lines

My stance on this is: as this functionality that is defined in assembler, having nasty C code is acceptable - providing that it's documented. (which we could improve somewhat)

Note, we can probably do this automatically without having to define a symbol exporting macro for assembler code.

Just add some annotation to the assembler files, then some script or similar processes the *.S files looking for the symbol export annotations and uses that to generate the ksyms.c file.

Nice project for someone who is bored :)

#5 lore raw
In reply to Julian Calaby
Quoted 7 lines · David Miller

From: "Julian Calaby" <[email protected]>
Date: Fri, 9 Jan 2009 08:23:41 +1100

Quoted 3 lines

My stance on this is: as this functionality that is defined in assembler, having nasty C code is acceptable - providing that it's documented. (which we could improve somewhat)

Quoted 7 lines

Note, we can probably do this automatically without having to define a symbol exporting macro for assembler code.

Just add some annotation to the assembler files, then some script or similar processes the *.S files looking for the symbol export annotations and uses that to generate the ksyms.c file.

Arnd Bergman once posted a patchset that allow us to do the export in .S files.

Google for "add support for exporting symbols from .S files"

Could we resurface this patchset?

Arnd?

Sam

#6 lore raw
In reply to Julian Calaby
Quoted 5 lines · Sam Ravnborg

Arnd Bergman once posted a patchset that allow us to do the export in .S files.

Google for "add support for exporting symbols from .S files"

Could we resurface this patchset?

The patch should still work. The only concern with this was that it still isn't sufficient to kill ksyms.c on architectures that export symbols from libgcc.

	Arnd <><
#7 lore raw
In reply to Julian Calaby
Quoted 10 lines · Arnd Bergmann
Quoted 5 lines · Sam Ravnborg

Arnd Bergman once posted a patchset that allow us to do the export in .S files.

Google for "add support for exporting symbols from .S files"

Could we resurface this patchset?

The patch should still work. The only concern with this was that it still isn't sufficient to kill ksyms.c on architectures that export symbols from libgcc.

We have the opportunity to move all EXPORT_SYMBOLS() to the place where the functions are defined.
The documentation effect that tell the reader:

"Hey, I may be used by a module so take extra care"

That is enough to warrant the inclusion of this patch. The aim here is IMO not to kill all the *ksyms.c files but rather to document which functions are exported at their definition.

I could see there were some comments about different .align and the like.

Can I ask you to repolish these bits and submit it to linux-arch.

Thanks,
Sam