Skip to contentSkip to conversation

RFC: rename arch/arm/mach-s3c2410 to arch/arm/mach-s3c24xx

DISCUSSION Dimitry Andric 5 participants 5 messages Apr 19, 2006 – Apr 20, 2006 1 message not in the archive

lore
Share & tools
lore

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

#1 lore raw
Quoted 5 lines · Ben Dooks

With the advent of the s3c2410 port adding support for more of the samsung SoC product line (s3c2440, s3c2442, s3c2400) there have been several requests by other people to rename the (in their opinion) increasingly inaccurate arch/arm/mach-s3c2410 to arch/arm/mach-s3c24xx.

Well, if you start this way, you might also consider renaming it to mach-s3cxxxx, since Samsung also seems to have S3C3410, S3C44B0 and who knows what else. Otherwise you'd maybe have to do such an operation again in the future...

Also, I've always found the dichotomy of having "include/asm-arm/arch-s3c2410" and "arch/arm/mach-s3c2410" rather weird. Isn't s3cxxxx an "architecture", instead of a specific machine? If so, arch/arm/arch-s3cxxxx would be more logical.

Anyway, by starting to rename directories, you start a never-ending quest, and you'll stress the abilities of most version control systems too. Your huge diff for just one rename operation already shows this.

There are certainly a lot more directories (also not specifically arm-related ones) in the Linux kernel source that could be renamed to be more logical, but I'd say the cost is rather large. E.g. difficulty merging patches on older kernels and other version control difficulties.

Cheers,
Dimitry

#2 lore raw
Quoted 13 lines · --- Dimitry Andric
Quoted 5 lines · Ben Dooks

With the advent of the s3c2410 port adding support for more of the samsung SoC product line (s3c2440, s3c2442, s3c2400) there have been several requests by other people to rename the (in their opinion) increasingly inaccurate arch/arm/mach-s3c2410 to arch/arm/mach-s3c24xx.

Well, if you start this way, you might also consider renaming it to mach-s3cxxxx, since Samsung also seems to have S3C3410, S3C44B0 and who
knows what else. Otherwise you'd maybe have to do such an operation again in the future...

Quoted 6 lines

Also, I've always found the dichotomy of having "include/asm-arm/arch-s3c2410" and "arch/arm/mach-s3c2410" rather weird.
Isn't s3cxxxx an "architecture", instead of a specific machine? If so,
arch/arm/arch-s3cxxxx would be more logical.

I am not sure, this will apply to Samsung chips or not. But in case of TI OMAP we had separated the directories based on the "Generation" of processors produced by TI, which is OMAP1 (OMAP1510, OMAP1610, OMAP1710, OMAP730, OMAP5910, OMAP5912) and OMAP2 (OMAP2420, OMAP2430 - (2430 support is not added to git tree yet)). So directory structure now becomes like

arch/arm/mach-omap1
arch/arm/mach-omap2
include/asm-arm/arch-omap
arch/arm/plat-omap - This directory contains the things which are common between OMAP1 and OMAP2 generation of processors and can be #ifdefed with specific CONFIG_ARCH_OMAP1/2. e.g gpio, dma apis etc.

---Komal Shah
http://komalshah.blogspot.com/

__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

#3 lore raw
In reply to Dimitry Andric

Dimitry:

Quoted 16 lines · Dimitry Andric
Quoted 7 lines · Ben Dooks

With the advent of the s3c2410 port adding support for more of the samsung SoC product line (s3c2440, s3c2442, s3c2400) there have been several requests by other people to rename the (in their opinion) increasingly inaccurate arch/arm/mach-s3c2410 to arch/arm/mach-s3c24xx.

Well, if you start this way, you might also consider renaming it to mach-s3cxxxx, since Samsung also seems to have S3C3410, S3C44B0 and who knows what else. Otherwise you'd maybe have to do such an operation again in the future...

Quoted 6 lines

Also, I've always found the dichotomy of having "include/asm-arm/arch-s3c2410" and "arch/arm/mach-s3c2410" rather weird. Isn't s3cxxxx an "architecture", instead of a specific machine? If so, arch/arm/arch-s3cxxxx would be more logical.

I always interpreted the arch/arm directories to be "machines based on the s3cxxxx", etc. Thus, in my world there's no dichotomy. But hey, that's just one person's world. :)

Quoted 5 lines

Anyway, by starting to rename directories, you start a never-ending quest, and you'll stress the abilities of most version control systems too. Your huge diff for just one rename operation already shows this.

It doesn't stress GNU Arch, and I bet it doesn't stress SVN or Cogito. What it does do is make the kernel code appear more obvious and better organized, which I see as being a good thing for future maintainability's sake alone. So I'm all for these changes.

Quoted 6 lines

There are certainly a lot more directories (also not specifically arm-related ones) in the Linux kernel source that could be renamed to be more logical, but I'd say the cost is rather large. E.g. difficulty merging patches on older kernels and other version control difficulties.

Well, now's our chance to find out whose VC systems we break. :) And I don't see the "moving patches forward from older kernels" as being an argument for locking down the current/future state of the kernel sources.

Respectfully,

b.g.

Signature · 2 lines
Bill Gatliff
[email protected]
#4 lore raw
In reply to Dimitry Andric
Quoted 9 lines · Dimitry Andric
Quoted 5 lines · Ben Dooks

With the advent of the s3c2410 port adding support for more of the samsung SoC product line (s3c2440, s3c2442, s3c2400) there have been several requests by other people to rename the (in their opinion) increasingly inaccurate arch/arm/mach-s3c2410 to arch/arm/mach-s3c24xx.

Also, I've always found the dichotomy of having "include/asm-arm/arch-s3c2410" and "arch/arm/mach-s3c2410" rather weird.

There's a reason for this (this has actually been covered and discussed at length in the past on the linux-arm mailing lists.)

Folk convinced me that the only thing which we should call "architecture" is the CPU - so things like "PPC", "ARM", "i386" are architectures, and not implementations of these (AT91RM9200, S3C2410).

The things in arch/arm/mach* are machine classes which support a variety of machines which are all essentially similar. Inside these directories you have the core support for the individual machines.

However, the problem is that we can't rename include/asm-arm/arch-* to include/asm-arm/mach-*, because we need a symlink to select the right one. If we renamed include/asm-arm/arch-* to include/asm-arm/mach-*, we'd want the symlink to be called include/asm-arm/mach.

Unfortunately, we have an include/asm-arm/mach directory, so we'd end up having to symlink include/asm-arm/mach-* to include/asm-arm/arch. This would be even more confusing than leaving the include/asm-arm as currently is.

Signature · 3 lines
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 Serial core
#5 lore raw

Hi Russell!

Quoted 3 lines · Russell King

Folk convinced me that the only thing which we should call "architecture" is the CPU - so things like "PPC", "ARM", "i386" are architectures, and not implementations of these (AT91RM9200, S3C2410).

And if we use ARM nomenclature there is also a difference between the architecture (e.g. v4, v4t, v5, ...) and the implementation (ARM7, ARM7T, ARM9T, Amulet, StrongARM, Xscale, ...) of the CPU core. So we have to distinguish between the core and the SoC/ASSP architecture. The register model is defined by the core architecture but the co-processors (MMU, CP15) by the implementation. And should we handle OMAP as a standard ARM9/10/11 implementation or another core? And should StrongARM be inherited from v4 or ARM8?

Does it make sense to reflect this also in the directory naming conventions? Hmmm, I am not sure. We could end with the complete ARM company's history.

With best regards
Andreas Schweigstill

Signature · 5 lines
Dipl.-Phys. Andreas Schweigstill
Schweigstill IT | Embedded Systems
Schauenburgerstraße 116, D-24118 Kiel, Germany
Phone: (+49) 431 5606-435, Fax: (+49) 431 5606-436
Mobile: (+49) 171 6921973, Web: http://www.schweigstill.de/