Re: [PATCH 2.4.23-pre1] /proc/ikconfig support

Re: [PATCH 2.4.23-pre1] /proc/ikconfig support

[Marcelo Tosatti]

On Tue, 26 Aug 2003 18:36:59 +0200 Marc-Christian Petersen
<m.c.p@wolk-project.de> wrote:

| On Monday 25 August 2003 19:50, Marc-Christian Petersen wrote:
|
| Hi Marcelo,
|
| > Idea/Patch from Randy Dunlap including fixes/updates:
| > From: Daniele Bellucci <bellucda@tiscali.it>
| >     put_user() to return -EFAULT on error.
| > From: Randy.Dunlap" <rddunlap@osdl.org>
| >     Updated 'extract-ikconfig' script
| > From me:
| >     /proc/ikconfig should only be allowed if CONFIG_PROC_FS=y
| > Attached is /proc/ikconfig support.
...
|
| something not ok with this or do you just need some time to review it?
:)

> I have the same question about the seq_file "single" additions
> patch that I sent yesterday.... ???

The seq_file patch needs EXPORT_SYMBOL right?

And about ikconfig, hum, I'm not sure if I want that. Its nice, yes, but I
still wonder. You are free to convince me though: I think people usually
know what they compile in their kernels, dont they?

Re: [PATCH 2.4.23-pre1] /proc/ikconfig support

[Willy Tarreau]

On Tue, Aug 26, 2003 at 04:33:09PM -0300, Marcelo Tosatti wrote:
 
> And about ikconfig, hum, I'm not sure if I want that. Its nice, yes, but I
> still wonder. You are free to convince me though: I think people usually
> know what they compile in their kernels, dont they?

Marcelo,

I can say for sure that I would really like to have ikconfig (I've been using
other sorts of proc_config patches for a long time now). It's not something
strictly necessary (otherwise it would already have been in), but it's really
userful in some situations because we're humans and not fail-safe :

  - when a default kernel and a default config doesn't work on a customer's
    system, and I insist on the fact it's someone else, you end up trying
    several tricks till the time it finally runs. When you do this at a
    customer's and you're compiling your 30rd kernel at 23h40 with the man
    behind you with red eyes, you know for sure that at the very moment he
    will see it boot, he will say "ok, thanks a lot, now let's go to bed. bye!"
    When you come again several weeks after, he tells you that /usr/src/linux
    took lots of useless space on his tapes and he removed it. Given the late
    hour the last time, you didn't save the .config and definitely lost it.
    Believe me, I was happy to see that I still had the .config for the 500+
    days uptime kernel I reported a few days ago because that's not always the
    case (even if I didn't need it this time).

  - when trying patches, config options, optimizations, etc... it's common for
    some (most ?) of us to quickly copy a bzimage.test under /boot, which entry
    is already filled in lilo.conf, reboot it with init=/bin/sh and trying lots
    of dirty tricks. After a while, you don't remember which .config was used
    for the only one which causes trouble or which works, or you simply lost it
    in a quick "make distclean" or "cd ..&&rm -rf linux&&tar x&&cd -", and being
    able to recover it from bzimage or at run time would be great.

As you see, nothing much important in my opinion, but definitely compensates
for my unreliable brain and saves me much time. I don't really matter whether
you merge it or not, since I will use it anyway, but I see no reason it couldn't
serve others as well.

Cheers,
Willy

Re: [PATCH 2.4.23-pre1] /proc/ikconfig support

[Marc-Christian Petersen]

On Tuesday 26 August 2003 21:33, Marcelo Tosatti wrote:

Hi Marcelo,

> > I have the same question about the seq_file "single" additions
> > patch that I sent yesterday.... ???
> The seq_file patch needs EXPORT_SYMBOL right?

I've send the export_symbol patch yesterday to you cc'ed lkml and randy etc.

> And about ikconfig, hum, I'm not sure if I want that. Its nice, yes, but I
> still wonder. You are free to convince me though: I think people usually
> know what they compile in their kernels, dont they?

grmpf. Absolutely wrong argumentation. People like me compile kernels a lot, 
for tons of different machines, for customers and so on. My brain is not that 
good that I'll remember each and every customers kernel/machine config ;)

Many people forget to copy the current .config to somewhere else in /boot or 
such. And I agree 100% with Willy :)

P.S.: Alan planed it for .23-pre1 ;-))

ciao, Marc

Re: [PATCH 2.4.23-pre1] /proc/ikconfig support

[Matthias Andree]

On Tue, 26 Aug 2003, Marcelo Tosatti wrote:

> And about ikconfig, hum, I'm not sure if I want that. Its nice, yes, but I
> still wonder. You are free to convince me though: I think people usually
> know what they compile in their kernels, dont they?

Precedent: admin at a site changes, original .config was snatched from
some other machine, and kernel sources have been removed from $NOTEBOOK
because space was tight... In such cases, a config store is pretty
handy. Extract, make oldconfig, there you go.

If you're collecting user votes, count mine as AYE.

-- 
Matthias Andree

Re: [PATCH 2.4.23-pre1] /proc/ikconfig support

[Matt Gibson]

On Tuesday 26 Aug 2003 23:33, Matthias Andree wrote:
>
> Precedent: admin at a site changes, original .config was snatched from
> some other machine, and kernel sources have been removed from $NOTEBOOK
> because space was tight... In such cases, a config store is pretty
> handy. Extract, make oldconfig, there you go.

Or, as happened with me on SuSE kernels (which had/have a similar arrangement 
patched in) -- your distro ships with a precompiled kernel, you download a 
new one from kernel.org, and you can configure it very quickly to match what 
you've already got.  Far quicker than starting from scratch.

M

-- 
"It's the small gaps between the rain that count,
 and learning how to live amongst them."
	      -- Jeff Noon

Re: [PATCH 2.4.23-pre1] /proc/ikconfig support

[Herbert Pötzl]

On Tue, Aug 26, 2003 at 04:33:09PM -0300, Marcelo Tosatti wrote:
> 
> On Tue, 26 Aug 2003 18:36:59 +0200 Marc-Christian Petersen
> <m.c.p@wolk-project.de> wrote:
> 
> | On Monday 25 August 2003 19:50, Marc-Christian Petersen wrote:
> |
> | Hi Marcelo,
> |
> | > Idea/Patch from Randy Dunlap including fixes/updates:
> | > From: Daniele Bellucci <bellucda@tiscali.it>
> | >     put_user() to return -EFAULT on error.
> | > From: Randy.Dunlap" <rddunlap@osdl.org>
> | >     Updated 'extract-ikconfig' script
> | > From me:
> | >     /proc/ikconfig should only be allowed if CONFIG_PROC_FS=y
> | > Attached is /proc/ikconfig support.
> ...
> |
> | something not ok with this or do you just need some time to review it?
> :)
> 
> > I have the same question about the seq_file "single" additions
> > patch that I sent yesterday.... ???
> 
> The seq_file patch needs EXPORT_SYMBOL right?
> 
> And about ikconfig, hum, I'm not sure if I want that. Its nice, yes, but I
> still wonder. You are free to convince me though: I think people usually
> know what they compile in their kernels, dont they?

not if you do 20 kernels a week, and then want to know
the config for machine XY, to compile a new kernel ...

since early 2.4.18 I'm happy with kconfig[1] but I guess
ikconfig will do almost the same ...

best,
Herbert

[1] http://www.13thfloor.at/VServer/patches-2.4.22-p10c17/03_kconfig-2.4.22-pre3.patch.bz2