Skip to contentSkip to conversation

MAINTAINERS file out of date?

DISCUSSION Michael McConnell 5 participants 14 messages Apr 26, 2007 – Apr 27, 2007

lore
Share & tools
lore
#0 lore raw

Hi folks,

I noticed the MAINTAINERS file still lists Riley Williams as the maintainer of the i386 boot code, presumably as no-one else has taken it up in his absence (though, I'm sure it's probably been touched since late 2004).

If I knew the first thing about the i386 boot process I would volunteer, but sadly I don't....

This article below details the reason for his absence, but please make sure you're sitting down before clicking. (It is work-safe.)
http://www.chorleytoday.co.uk/ViewArticle2.aspx?SectionID=81&ArticleID=905248

Sorry to be the bearer of bad news.

Signature · 2 lines
-- Michael "Soruk" McConnell
   Eridani Star System - http://www.eridani.co.uk/
#1 lore raw
Quoted 7 lines · Michael McConnell

Hi folks,

I noticed the MAINTAINERS file still lists Riley Williams as the maintainer of the i386 boot code, presumably as no-one else has taken it up in his absence (though, I'm sure it's probably been touched since late 2004).

Quoted 7 lines

If I knew the first thing about the i386 boot process I would volunteer, but sadly I don't....

This article below details the reason for his absence, but please make sure you're sitting down before clicking. (It is work-safe.)
http://www.chorleytoday.co.uk/ViewArticle2.aspx?SectionID=81&ArticleID=905248

Dear.

I have considered myself as a rather unofficial maintainer of this code, and wouldn't mind make it official now when I actually have a job which both cares about and actually can support my upstream Linux activities, which was a major pain for a while.

Overall, there is a lot of cleanup which really is needed in the i386 boot process; I have done some work on it already, but more is needed.

-hpa

#2 lore raw
Quoted 18 lines · H. Peter Anvin
Quoted 7 lines · Michael McConnell

Hi folks,

I noticed the MAINTAINERS file still lists Riley Williams as the maintainer of the i386 boot code, presumably as no-one else has taken it up in his absence (though, I'm sure it's probably been touched since late 2004).

Quoted 7 lines

If I knew the first thing about the i386 boot process I would volunteer, but sadly I don't....

This article below details the reason for his absence, but please make sure you're sitting down before clicking. (It is work-safe.)
http://www.chorleytoday.co.uk/ViewArticle2.aspx?SectionID=81&ArticleID=905248

Dear.

Quoted 8 lines

I have considered myself as a rather unofficial maintainer of this code, and wouldn't mind make it official now when I actually have a job which both cares about and actually can support my upstream Linux activities, which was a major pain for a while.

Overall, there is a lot of cleanup which really is needed in the i386 boot process; I have done some work on it already, but more is needed.

Sounds interesting. Can you point me what needs to be done exactly? Maybe I can help you. ;)

Have fun!

#3 lore raw
Quoted 10 lines · WANG Cong
Quoted 8 lines

I have considered myself as a rather unofficial maintainer of this code, and wouldn't mind make it official now when I actually have a job which both cares about and actually can support my upstream Linux activities, which was a major pain for a while.

Overall, there is a lot of cleanup which really is needed in the i386 boot process; I have done some work on it already, but more is needed.

Sounds interesting. Can you point me what needs to be done exactly? Maybe I can help you. ;)

There was a long thread on the linux-virtualization list (@lists.linux-foundation.org) just a few days ago. The biggest single issue right now is probably how we transition from the bootup page tables to the "real" init_mm page tables, but the real-mode code also needs a massive overhaul (especially so since this code can and should be shared with x86-64); in particular I really want to get rid of the insane segment handling, where segments are constantly recalculated for no good reason.

For the real-mode stuff, I have some patches already in the works for this. Eric W. Biederman has also done a lot of work in this area.

-hpa

#4 lore raw
Quoted 21 lines · H. Peter Anvin
Quoted 10 lines · WANG Cong
Quoted 8 lines

I have considered myself as a rather unofficial maintainer of this code, and wouldn't mind make it official now when I actually have a job which both cares about and actually can support my upstream Linux activities, which was a major pain for a while.

Overall, there is a lot of cleanup which really is needed in the i386 boot process; I have done some work on it already, but more is needed.

Sounds interesting. Can you point me what needs to be done exactly? Maybe I can help you. ;)

There was a long thread on the linux-virtualization list (@lists.linux-foundation.org) just a few days ago. The biggest single issue right now is probably how we transition from the bootup page tables to the "real" init_mm page tables, but the real-mode code also needs a massive overhaul (especially so since this code can and should be shared with x86-64); in particular I really want to get rid of the insane segment handling, where segments are constantly recalculated for no good reason.

Quoted 4 lines

For the real-mode stuff, I have some patches already in the works for this. Eric W. Biederman has also done a lot of work in this area.

-hpa

Thanks for your point.
I know little about virtualization, maybe can't help much. But I am interested in other things you mentioned. AFAIK, segments can't be avoided on i386, and Linux uses them very little, how are they recalculated constantly?

Regards!

#5 lore raw
Quoted 5 lines · WANG Cong

Thanks for your point.
I know little about virtualization, maybe can't help much. But I am interested in other things you mentioned. AFAIK, segments can't be avoided on i386, and Linux uses them very little, how are they recalculated constantly?

Look in arch/i386/boot/setup.S it runs in 16bit mode. We are talking about real mode segments not 16bit segments.

16bit real mode is a completely different ball game, and why we keep BIOS calls isolated to that one dinky file.

Eric

#6 lore raw
Quoted 7 lines · Eric W. Biederman
Quoted 4 lines · WANG Cong

Thanks for your point.
I know little about virtualization, maybe can't help much. But I am interested in other things you mentioned. AFAIK, segments can't be avoided on i386, and Linux uses them very little, how are they recalculated constantly?

Quoted 6 lines

Look in arch/i386/boot/setup.S it runs in 16bit mode. We are talking about real mode segments not 16bit segments.

16bit real mode is a completely different ball game, and why we keep BIOS calls isolated to that one dinky file.

A lot of that code (although, of course, not all) could be written in C, though. I'm thinking of taking a stab at rewriting it that way.

-hpa

#7 lore raw
Quoted 2 lines · H. Peter Anvin

A lot of that code (although, of course, not all) could be written in C, though. I'm thinking of taking a stab at rewriting it that way.

Is this using the .code16gcc? Or are you thinking of some other technique. Requiring another C compiler to build the kernel would be a pain to use.

Eric

#8 lore raw
Quoted 8 lines · Eric W. Biederman
Quoted 2 lines · H. Peter Anvin

A lot of that code (although, of course, not all) could be written in C, though. I'm thinking of taking a stab at rewriting it that way.

Is this using the .code16gcc? Or are you thinking of some other technique. Requiring another C compiler to build the kernel would be a pain to use.

.code16gcc was what I was using. There is a GSoC project that I'm mentoring to get 16-bit support for gcc, that will be possible to eventually migrate to (for code size) if/when it gets implemented and gets pushed out far enough, but that's for the future.

-hpa

#9 lore raw
Quoted 10 lines · H. Peter Anvin
Quoted 8 lines · Eric W. Biederman
Quoted 2 lines · H. Peter Anvin

A lot of that code (although, of course, not all) could be written in C, though. I'm thinking of taking a stab at rewriting it that way.

Is this using the .code16gcc? Or are you thinking of some other technique. Requiring another C compiler to build the kernel would be a pain to use.

Quoted 6 lines

.code16gcc was what I was using. There is a GSoC project that I'm mentoring to get 16-bit support for gcc, that will be possible to eventually migrate to (for code size) if/when it gets implemented and gets pushed out far enough, but that's for the future.

-hpa

Thanks! I will take a look at that file.

Maybe we can rewrite them in C, use a 16-bit C compiler to generate AT&T asm code and finally push the asm code in the kernel source tree. But perhaps there is no such ideal compiler. ;)

#11 lore raw
In reply to H. Peter Anvin
Quoted 4 lines

A lot of that code (although, of course, not all) could be written in C, though. I'm thinking of taking a stab at rewriting it that way.

That would require a new compiler, right? I don't think that would make users very happy.

Besides the code is not exactly that maintenance intensive and only changes rarely so I don't need a pressing need to rewrite it

-Andi

#10 lore raw
In reply to WANG Cong
Quoted 6 lines · WANG Cong

On Thu, Apr 26, 2007 at 10:27:47PM -0700, H. Peter Anvin wrote:

Thanks! I will take a look at that file.

Maybe we can rewrite them in C, use a 16-bit C compiler to generate AT&T asm code and finally push the asm code in the kernel source tree. But perhaps there is no such ideal compiler. ;)

No, that would be bad.

-hpa

#12 lore raw
In reply to Andi Kleen
Quoted 8 lines · Andi Kleen
Quoted 2 lines

A lot of that code (although, of course, not all) could be written in C, though. I'm thinking of taking a stab at rewriting it that way.

That would require a new compiler, right? I don't think that would make users very happy.

Besides the code is not exactly that maintenance intensive and only changes rarely so I don't need a pressing need to rewrite it

No, it would not need a new compiler. All it requires is gcc plus a reasonably recent binutils which you need anyway.

-hpa

#13 lore raw
Quoted 4 lines · H. Peter Anvin
Quoted 3 lines · Andi Kleen
Quoted 2 lines

A lot of that code (although, of course, not all) could be written in C, though. I'm thinking of taking a stab at rewriting it that way.

Quoted 8 lines
Quoted 5 lines

That would require a new compiler, right? I don't think that would make users very happy.

Besides the code is not exactly that maintenance intensive and only changes rarely so I don't need a pressing need to rewrite it

No, it would not need a new compiler. All it requires is gcc plus a reasonably recent binutils which you need anyway.

There opportunities to enhance this code without writing it in C. Such as building the code out comprehensible single of subroutines, with a well defined calling sequence.

The big benefit when you can go to C is that you can include headers from elsewhere in the kernel and since setup.S is increasingly becoming optional it has a fixed interface to the rest of the kernel, so there is much less opportunity for enhancement there.

Eric