Re: [RFC] klibc for 2.5.59 bk

Kai Germaschewski (kai@tp1.ruhr-uni-bochum.de)
Sun, 16 Feb 2003 21:06:09 -0600 (CST)

Messages sorted by: [ date ][ thread ][ subject ][ author ]
Next message: Jeff Garzik: "Re: Annoying /proc/net/dev rollovers."
Previous message: Rusty Russell: "Re: [RFC] Migrating net/sched to new module interface"

On Sun, 9 Feb 2003, Greg KH wrote:

> > Greg KH, Fri, Feb 07, 2003 05:59:19 +0100:
> > > Hi all,
> > >
> > > Thanks to Arnd Bergmann, it looks like the klibc and initramfs code is
> > > now working. I've created a patch against Linus's latest bk tree and
> > > put it at:
> > > http://www.kroah.com/linux/klibc/klibc-2.5.59-2.patch.gz
> >
> > was the following part of the patch intended? (hello_world?)
>
> Yes it was. It shows how to add a binary file to the initramfs image,
> and have it executed by the kernel during the early boot process.
>
> The fact that the program didn't really do anything significant isn't
> important here.

I did some work on integrating klibc into kbuild now. I used your patch as
guide line, though I started from scratch with klibc-0.77. The build
should work fine (reminder: "make KBUILD_VERBOSE=0 ..." will give you much
more readable output), but I probably broke some non-x86 architectures
in the process.

To do something more useful than "hello world", I actually moved some part
of finding / mounting the final root system into userspace, though only
conditional on CONFIG_INITRAMFS.

My tree is at

Pull from http://linux-isdn.bkbits.net/linux-2.5.klibc

If people want a GNU patch, I can mail that privately, it was too large to
attach it here.

--Kai

-----------------------------------------------------------------------------
ChangeSet@1.1027.1.2, 2003-02-15 16:55:46-06:00, kai@tp1.ruhr-uni-bochum.de
klibc: Merge klibc-0.77

That's just a cp -r klibc-0.77/klibc/* linux/usr/lib

-----------------------------------------------------------------------------
ChangeSet@1.1027.1.3, 2003-02-15 23:09:41-06:00, kai@tp1.ruhr-uni-bochum.de
kbuild/klibc: Integrate klibc into the build.

Basically, add a scripts/Makefile.user, which does similar things to
scripts/Makefile.build, but compiles userspace for the target instead.

It's tested for a static klibc on i386, building the shared lib works, too,
but is not further integrated.

This patch also adds gregkh's hello test program, which works as well.

-----------------------------------------------------------------------------
ChangeSet@1.1027.1.4, 2003-02-15 23:11:51-06:00, greg@kroah.com
klibc: fix up the hello_world example

stdout doesn't go anywhere useful when spawned from the kernel :)

----------------------------------------------------------------------------
hello.c | 7 ++++++-
1 files changed, 6 insertions(+), 1 deletion(-)

-----------------------------------------------------------------------------
ChangeSet@1.1027.1.5, 2003-02-15 23:16:53-06:00, kai@tp1.ruhr-uni-bochum.de
klibc: Fix a compiler warning

----------------------------------------------------------------------------
sha1hash.c | 2 +-
1 files changed, 1 insertion(+), 1 deletion(-)

-----------------------------------------------------------------------------
ChangeSet@1.1027.1.6, 2003-02-15 23:20:13-06:00, kai@tp1.ruhr-uni-bochum.de
klibc: Fix the "hello" example (for real)

Greg's fix used fwrite on a file descriptor obtained from open(), which
only works by luck, since for klibc FILE * == fd.

Use standard C lib functions for open/close.

----------------------------------------------------------------------------
hello.c | 6 +++---
1 files changed, 3 insertions(+), 3 deletions(-)

-----------------------------------------------------------------------------
ChangeSet@1.1027.1.7, 2003-02-15 23:21:47-06:00, greg@kroah.com
klibc: add file support to gen_init_cpio.c

----------------------------------------------------------------------------
gen_init_cpio.c | 90 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++
1 files changed, 90 insertions(+)

-----------------------------------------------------------------------------
ChangeSet@1.1027.1.8, 2003-02-15 23:24:10-06:00, arndb@de.ibm.com
klibc: gen_init_cpio file generation fix

I found what kept initramfs from working here: While creating
of initramfs_data.cpio.gz, the padding between a file header
and the file contents was wrong, which can be verified by
unpacking the archive by hand.

----------------------------------------------------------------------------
gen_init_cpio.c | 1 +
1 files changed, 1 insertion(+)

-----------------------------------------------------------------------------
ChangeSet@1.1027.1.9, 2003-02-15 23:33:21-06:00, kai@tp1.ruhr-uni-bochum.de
klibc: Stop on error when building the CPIO

gen_init_cpio still referenced hello in usr/hello_world, but I put it
into usr/root. This is now corrected, however it also pointed out the
common problem that the error code of gen_init_cpio is ignored since
it's output is piped to gzip. To fix that, make the generation of the
.cpio.gz a two step process.

----------------------------------------------------------------------------
Makefile | 24 +++++++++++++++++-------
gen_init_cpio.c | 2 +-
2 files changed, 18 insertions(+), 8 deletions(-)

-----------------------------------------------------------------------------
ChangeSet@1.1027.1.10, 2003-02-16 18:06:13-06:00, kai@tp1.ruhr-uni-bochum.de
klibc: Silence too ambitious warnings

gcc complained about unused function parameters and things, that's just
a little too much.

----------------------------------------------------------------------------
Makefile | 2 +-
1 files changed, 1 insertion(+), 1 deletion(-)

-----------------------------------------------------------------------------
ChangeSet@1.1027.1.11, 2003-02-16 18:11:32-06:00, kai@tp1.ruhr-uni-bochum.de
klibc: Move mounting of the root filesystem into userspace

When selecting CONFIG_INITRAMFS, init/do_mounts.c is not compiled anymore,
and it's task is taken over by a small /sbin/init running in initramfs.

However, this is a proof of concept only, the userspace code only handles
mounting a local filesystem, no support for nfs / initrd / devfs yet.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Next message: Jeff Garzik: "Re: Annoying /proc/net/dev rollovers."
Previous message: Rusty Russell: "Re: [RFC] Migrating net/sched to new module interface"