[gentoo-user] Building an initramfs into the kernel

[gentoo-user] Building an initramfs into the kernel

[Mark Knecht]

Hi,
   OK, it's the day after Christmas and this little kid wants to play
with the new toys Uncle Neil gave us yesterday - a copy of his well
worn setup file for building an initramfs into the kernel - a copy of
which I place here:

[QUOTE]

This is the file I use on a system that has / on a LUKS filesystem on top
of LVM. The format is documented in the kernel docs at
Documentation/filesystems/ramfs-rootfs-initramfs.txt


dir /bin 755 0 0
file /bin/busybox /bin/busybox 755 0 0
slink /bin/sh busybox 777 0 0

dir /realroot 755 0 0
dir /etc 755 0 0
dir /proc 755 0 0
dir /sys 755 0 0

dir /sbin 755 0 0
file /sbin/lvm.static /sbin/lvm.static 755 0 0
#file /sbin/mdadm /sbin/mdadm 755 0 0
file /sbin/cryptsetup /sbin/cryptsetup 755 0 0

file /sbin/e2fsck /sbin/e2fsck 755 0 0
dir /lib 755 0 0
file /lib/libext2fs.so /usr/lib64/libext2fs.so 755 0 0

dir /dev 755 0 0
nod /dev/console 600 0 0 c 5 1
nod /dev/null 666 0 0 c 1 3
nod /dev/tty 666 0 0 c 5 0
nod /dev/urandom 666 0 0 c 1 9

file /init /usr/src/init.sh 755 0 0

[/QUOTE]


   OK, so reading through this it seems moderately straight forward. My reading:

a) Create some directories
b) Populate them with some executables
c) Make some nodes
d) Execute a script


I do have a few questions:

1) dir /realroot 755 0 0

Is this something required to make the machine boot? Or is it possibly
a mount point in case of problems and just used inside the initramfs
if trouble arises? Something else? Google didn't point me toward
anything meaningful.

2) Contained executables, as I understand them, either need to be
built with the static flag or you have to include all the libraries.
Static seems simpler so (in my case) should I rebuild mdadm &
e2fsprogs with +static? (I don't currently use lvm or any crypt stuff)
I assume from the line

file /lib/libext2fs.so /usr/lib64/libext2fs.so 755 0 0

that running ldd on the new mdadm is still going to require this one
library be in the initramfs?

3) My system uses RAID today. Is there any significant risk in
rebuilding mdadm with static support, rebooting the existing kernel
without an initramfs and then mdadm having trouble?

4) What's in /usr/src/init.sh ? From the Gentoo initramfs wiki I find
this as an example:

rescue_shell() {
    echo "Something went wrong. Dropping you to a shell."
    busybox --install -s
    exec /bin/sh
}

I wonder what Uncle Neil uses? (Or anyone else...)

   Yipee!!! Better than the Red Ryder BB gun complete with a compass
and a sundial I was hoping for! ;-)

   (And if it's not clear, I'm hoping this thread might possibly help
others in the future move from one place about initramfs to another
place about initramfs. We'll see.)

Cheers,
Mark

Re: [gentoo-user] Building an initramfs into the kernel

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 4726 bytes --]

On Wed, 26 Dec 2012 07:58:34 -0800, Mark Knecht wrote:

>    OK, it's the day after Christmas and this little kid wants to play
> with the new toys Uncle Neil gave us yesterday - a copy of his well
> worn setup file for building an initramfs into the kernel - a copy of
> which I place here:
> 
> [QUOTE]
> 
> This is the file I use on a system that has / on a LUKS filesystem on
> top of LVM. The format is documented in the kernel docs at
> Documentation/filesystems/ramfs-rootfs-initramfs.txt
> 
> 
> dir /bin 755 0 0
> file /bin/busybox /bin/busybox 755 0 0
> slink /bin/sh busybox 777 0 0
> 
> dir /realroot 755 0 0
> dir /etc 755 0 0
> dir /proc 755 0 0
> dir /sys 755 0 0
> 
> dir /sbin 755 0 0
> file /sbin/lvm.static /sbin/lvm.static 755 0 0
> #file /sbin/mdadm /sbin/mdadm 755 0 0
> file /sbin/cryptsetup /sbin/cryptsetup 755 0 0
> 
> file /sbin/e2fsck /sbin/e2fsck 755 0 0
> dir /lib 755 0 0
> file /lib/libext2fs.so /usr/lib64/libext2fs.so 755 0 0
> 
> dir /dev 755 0 0
> nod /dev/console 600 0 0 c 5 1
> nod /dev/null 666 0 0 c 1 3
> nod /dev/tty 666 0 0 c 5 0
> nod /dev/urandom 666 0 0 c 1 9
> 
> file /init /usr/src/init.sh 755 0 0
> 
> [/QUOTE]
> 
> 
>    OK, so reading through this it seems moderately straight forward. My
> reading:
> 
> a) Create some directories
> b) Populate them with some executables
> c) Make some nodes
> d) Execute a script
> 
> 
> I do have a few questions:
> 
> 1) dir /realroot 755 0 0
> 
> Is this something required to make the machine boot? Or is it possibly
> a mount point in case of problems and just used inside the initramfs
> if trouble arises? Something else? Google didn't point me toward
> anything meaningful.

It's where init.sh mounts the real root filesystem before running
switchroot.

> 2) Contained executables, as I understand them, either need to be
> built with the static flag or you have to include all the libraries.
> Static seems simpler so (in my case) should I rebuild mdadm &
> e2fsprogs with +static? (I don't currently use lvm or any crypt stuff)

Yes.

> 3) My system uses RAID today. Is there any significant risk in
> rebuilding mdadm with static support, rebooting the existing kernel
> without an initramfs and then mdadm having trouble?

Static mdadm has given me no problems whatsoever, I guarantee that you
will have no fewer problems than me :)

> 4) What's in /usr/src/init.sh ? From the Gentoo initramfs wiki I find
> this as an example:

I use one based on the wiki example


#!/bin/busybox sh

rescue_shell() {
	busybox --install -s
	echo $1
	exec /bin/sh
	}

### Mount the /proc and /sys filesystems.
mount -t proc none /proc
grep -q initdebug </proc/cmdline && set -x && DEBUG=1
mount -t sysfs none /sys
mount -t devtmpfs none /dev

### Get root and init parameters from cmdline
REAL_ROOT="$(sed 's/.*root=\(\S*\).*/\1/' <proc/cmdline)"
if grep -q init= </proc/cmdline; then
	REAL_INIT="$(sed 's/.*init=\(\S*\).*/\1/' <proc/cmdline)"
else
	REAL_INIT="/sbin/init"
	fi

### Assemble RAID
[[ -f /sbin/mdadm ]] && mdadm --assemble --scan

### Initialise LVM
if [[ -f /sbin/lvm.static ]]; then
	lvm.static vgchange -a y
	lvm.static vgscan --mknodes
	[[ -n "$DEBUG" ]] && sleep 10
	fi

### Mount the root filesystem.
if [[ -f /sbin/cryptsetup ]] && cryptsetup isLuks $REAL_ROOT; then
	busybox echo -e "\e[0;32m\n=====================\nEncrypted root device\n=====================\n\e[m"
    cryptsetup luksOpen $REAL_ROOT cryptroot
    REAL_ROOT="/dev/mapper/cryptroot"
    [[ -n "$DEBUG" ]] && echo $REAL_ROOT && sleep 10
    fi
mount -o ro $REAL_ROOT /realroot || rescue_shell "Could not mount REAL_ROOT"
[[ -n "$DEBUG" ]] && df -h /realroot && sleep 10

### Mount /usr if separate
if grep -qE '^[^#].*\s\/usr\s' /realroot/etc/fstab; then
    #mount $(awk '/\s\/usr\s/ {print $1, "/realroot"$2, "-t", $3, "-o", $4 ",ro"}' /realroot/etc/fstab)
    mount $(awk '/^[^#].*\s\/usr\s/ {print $1, "/realroot"$2, "-t", $3, "-o", $4 ",ro"}' /realroot/etc/fstab)
    [[ -n "$DEBUG" ]] && df -h /realroot//usr && sleep 10
    fi

grep -q waitforkey </proc/cmdline && WAITFORKEY=1
grep -q rescue </proc/cmdline && rescue_shell "...because you asked for it."

### Clean up.
umount /dev
umount /sys
umount /proc

### Boot the real thing
[[ -n "$DEBUG" ]] && echo "Ready to switch" && sleep 10
[[ -n "$WAITFORKEY" ]] && busybox showkey -a
exec busybox switch_root /realroot $REAL_INIT || rescue_shell "Failed to switch_root"

Most of it is debug stuff that I haven't used since I first started using
it.


-- 
Neil Bothwick

For security reasons, all text in this mail is double-rot13 encrypted.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-user] Building an initramfs into the kernel

[Mark Knecht]

On Wed, Dec 26, 2012 at 9:09 AM, Neil Bothwick <neil@digimed.co.uk> wrote:
> On Wed, 26 Dec 2012 07:58:34 -0800, Mark Knecht wrote:

>
>> 2) Contained executables, as I understand them, either need to be
>> built with the static flag or you have to include all the libraries.
>> Static seems simpler so (in my case) should I rebuild mdadm &
>> e2fsprogs with +static? (I don't currently use lvm or any crypt stuff)
>
> Yes.
>
<SNIP>

I've done that, played a bit with ldd to see the changes, etc., and it seems to
make sense.

On the mdadm list one comment came up which was to make sure
my mdadm.conf file was up to date and then include it in the initramfs
image so that if for some reason the machine doesn't boot I have all
the info required to start the RAIDs by hand. I thought that was a good
comment.

>
> Most of it is debug stuff that I haven't used since I first started using
> it.

OK, it's a good example of a more complete init script so that's helpful.
Thanks!

One other question came up. So I build the kernel, and the kernel build
creates the initramfs image and embeds it in the kernel. Is there a tool
that would allow me to query the embedded image prior to booting
so that I can check that everything is in there that I think should be in
there?

Scanning around online I found discussion about lsinitramfs but I
haven't found it in portage yet. dracut has lsinitrd. genkernel doesn't
have anything like it. And even if I found it I think it only works for
stand alone initramfs files.

Thanks,
Mark

Re: [gentoo-user] Building an initramfs into the kernel

[Mark Knecht]

On Wed, Dec 26, 2012 at 12:48 PM, Mark Knecht <markknecht@gmail.com> wrote:
<SNIP>
> One other question came up. So I build the kernel, and the kernel build
> creates the initramfs image and embeds it in the kernel. Is there a tool
> that would allow me to query the embedded image prior to booting
> so that I can check that everything is in there that I think should be in
> there?


Actually, I think the initramfs wiki covers this, albeit it's 'not
easy'. Their words.

I think I'm covered for now.

Cheers,
Mark

Re: [gentoo-user] Building an initramfs into the kernel

[Mark Knecht]

On Wed, Dec 26, 2012 at 1:49 PM, Mark Knecht <markknecht@gmail.com> wrote:
<SNIP>
> Actually, I think the initramfs wiki covers this, albeit it's 'not
> easy'. Their words.
>
> I think I'm covered for now.
>
> Cheers,
> Mark

While extracting it from the kernel image may be involved, and for now
using Neil's example init script which probably isn't exactly right
for my needs, there is an intermediate step I quickly discovered.
After building the new kernel there is an initramfs_data_cpio file
created in /usr/src/linux/usr. I copied this file to a new directory
under /root and then ran

cpio -i -d -H newc --no-absolute-filenames < initramfs_data.cpio

which creates all of the initramfs file structure in this directory. I
can then list it and it appears everything is there:


c2stable initramfs-work # ls -lR
.:
total 2620
drwxr-xr-x 2 root root    4096 Dec 26 14:00 bin
drwxr-xr-x 2 root root    4096 Dec 26 14:00 dev
drwxr-xr-x 2 root root    4096 Dec 26 14:00 etc
-rwxr-xr-x 1 root root    2017 Dec 26 14:00 init
-rw-r--r-- 1 root root 2640384 Dec 26 13:59 initramfs_data.cpio
drwxr-xr-x 2 root root    4096 Dec 26 14:00 lib
drwxr-xr-x 2 root root    4096 Dec 26 14:00 proc
drwxr-xr-x 2 root root    4096 Dec 26 14:00 realroot
drwxr-xr-x 2 root root    4096 Dec 26 14:00 sbin
drwxr-xr-x 2 root root    4096 Dec 26 14:00 sys

./bin:
total 2020
-rwxr-xr-x 1 root root 2061600 Dec 26 14:00 busybox
lrwxrwxrwx 1 root root       7 Dec 26 14:00 sh -> busybox

./dev:
total 0
crw------- 1 root root 5, 1 Dec 26 14:00 console
crw-rw-rw- 1 root root 1, 3 Dec 26 14:00 null
crw-rw-rw- 1 root root 5, 0 Dec 26 14:00 tty
crw-rw-rw- 1 root root 1, 9 Dec 26 14:00 urandom

./etc:
total 4
-rwxr-xr-x 1 root root 3017 Dec 26 14:00 mdadm.conf

./lib:
total 4
-rwxr-xr-x 1 root root 528 Dec 26 14:00 libext2fs.so

./proc:
total 0

./realroot:
total 0

./sbin:
total 568
-rwxr-xr-x 1 root root 216744 Dec 26 14:00 e2fsck
-rwxr-xr-x 1 root root 353928 Dec 26 14:00 mdadm

./sys:
total 0
c2stable initramfs-work #


c2stable initramfs-work # cat init
#!/bin/busybox sh

rescue_shell() {
        busybox --install -s
        echo $1
        exec /bin/sh
        }

### Mount the /proc and /sys filesystems.
mount -t proc none /proc
grep -q initdebug </proc/cmdline && set -x && DEBUG=1
mount -t sysfs none /sys
mount -t devtmpfs none /dev

### Get root and init parameters from cmdline
REAL_ROOT="$(sed 's/.*root=\(\S*\).*/\1/' <proc/cmdline)"
if grep -q init= </proc/cmdline; then
        REAL_INIT="$(sed 's/.*init=\(\S*\).*/\1/' <proc/cmdline)"
else
        REAL_INIT="/sbin/init"
        fi

### Assemble RAID
[[ -f /sbin/mdadm ]] && mdadm --assemble --scan

### Initialise LVM
if [[ -f /sbin/lvm.static ]]; then
        lvm.static vgchange -a y
        lvm.static vgscan --mknodes
        [[ -n "$DEBUG" ]] && sleep 10
        fi

### Mount the root filesystem.
if [[ -f /sbin/cryptsetup ]] && cryptsetup isLuks $REAL_ROOT; then
        busybox echo -e "\e[0;32m\n=====================\nEncrypted
root device\n=====================\n\e[m"
    cryptsetup luksOpen $REAL_ROOT cryptroot
    REAL_ROOT="/dev/mapper/cryptroot"
    [[ -n "$DEBUG" ]] && echo $REAL_ROOT && sleep 10
    fi
mount -o ro $REAL_ROOT /realroot || rescue_shell "Could not mount REAL_ROOT"
[[ -n "$DEBUG" ]] && df -h /realroot && sleep 10

### Mount /usr if separate
if grep -qE '^[^#].*\s\/usr\s' /realroot/etc/fstab; then
    #mount $(awk '/\s\/usr\s/ {print $1, "/realroot"$2, "-t", $3,
"-o", $4 ",ro"}' /realroot/etc/fstab)
    mount $(awk '/^[^#].*\s\/usr\s/ {print $1, "/realroot"$2, "-t",
$3, "-o", $4 ",ro"}' /realroot/etc/fstab)
    [[ -n "$DEBUG" ]] && df -h /realroot//usr && sleep 10
    fi

grep -q waitforkey </proc/cmdline && WAITFORKEY=1
grep -q rescue </proc/cmdline && rescue_shell "...because you asked for it."

### Clean up.
umount /dev
umount /sys
umount /proc

### Boot the real thing
[[ -n "$DEBUG" ]] && echo "Ready to switch" && sleep 10
[[ -n "$WAITFORKEY" ]] && busybox showkey -a
exec busybox switch_root /realroot $REAL_INIT || rescue_shell "Failed
to switch_root"
c2stable initramfs-work #



At this point I don't know that 1) the image is actually in the
kernel, or 2) that my "init thingy" ;-) image would work, but at least
the process of putting it together is verifiable.


Cheers,
Mark

Re: [gentoo-user] Building an initramfs into the kernel

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 499 bytes --]

On Wed, 26 Dec 2012 14:09:34 -0800, Mark Knecht wrote:

> At this point I don't know that 1) the image is actually in the
> kernel, or 2) that my "init thingy" ;-) image would work, but at least
> the process of putting it together is verifiable.

That's why I put in all the debug stuff, so I could watch the progress of
the script as it ran.


-- 
Neil Bothwick

In the 60's people took acid to make the world weird.
Now the world is weird and people take Prozac to make it normal.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-user] Building an initramfs into the kernel

[Mark Knecht]

On Wed, Dec 26, 2012 at 3:03 PM, Neil Bothwick <neil@digimed.co.uk> wrote:
> On Wed, 26 Dec 2012 14:09:34 -0800, Mark Knecht wrote:
>
>> At this point I don't know that 1) the image is actually in the
>> kernel, or 2) that my "init thingy" ;-) image would work, but at least
>> the process of putting it together is verifiable.
>
> That's why I put in all the debug stuff, so I could watch the progress of
> the script as it ran.
>


OK, I'll look at combining that part my my scripts, or just using yours, etc.

Thanks for the help,
Mark

Re: [gentoo-user] Building an initramfs into the kernel

[Mark Knecht]

On Wed, Dec 26, 2012 at 3:50 PM, Mark Knecht <markknecht@gmail.com> wrote:
> On Wed, Dec 26, 2012 at 3:03 PM, Neil Bothwick <neil@digimed.co.uk> wrote:
>> On Wed, 26 Dec 2012 14:09:34 -0800, Mark Knecht wrote:
>>
>>> At this point I don't know that 1) the image is actually in the
>>> kernel, or 2) that my "init thingy" ;-) image would work, but at least
>>> the process of putting it together is verifiable.
>>
>> That's why I put in all the debug stuff, so I could watch the progress of
>> the script as it ran.
>>
>
>
> OK, I'll look at combining that part my my scripts, or just using yours, etc.
>
> Thanks for the help,
> Mark

Neil,
   One more question if I might. What's the simplest way to regenerate
the kernel when there are no kernel changes but you have changes to
the programs that are going into the initramfs? make clean seems like
overkill to me, and it's very slow to boot.

   I suspect that one of the simplest failures on an initramfs is that
you update something that's contained in the initramfs, mdadm or
busybox or whatever, and then don't get that new stuff into the
initramfs. Seems like I need to be right on top of what I'm keeping in
it and make sure a new kernel gets generated any time those packages
are updated. Correct?

   Anyway, I'm wondering how you manage that?

   Should be booting my first kernel with the initramfs in it shortly.
We'll see how that goes.

Thanks,
Mark

Re: [gentoo-user] Building an initramfs into the kernel

[Mark Knecht]

On Thu, Dec 27, 2012 at 9:33 AM, Mark Knecht <markknecht@gmail.com> wrote:
> On Wed, Dec 26, 2012 at 3:50 PM, Mark Knecht <markknecht@gmail.com> wrote:
>> On Wed, Dec 26, 2012 at 3:03 PM, Neil Bothwick <neil@digimed.co.uk> wrote:
>>> On Wed, 26 Dec 2012 14:09:34 -0800, Mark Knecht wrote:
>>>
>>>> At this point I don't know that 1) the image is actually in the
>>>> kernel, or 2) that my "init thingy" ;-) image would work, but at least
>>>> the process of putting it together is verifiable.
>>>
>>> That's why I put in all the debug stuff, so I could watch the progress of
>>> the script as it ran.
>>>
>>
>>
>> OK, I'll look at combining that part my my scripts, or just using yours, etc.
>>
>> Thanks for the help,
>> Mark
>
> Neil,
>    One more question if I might. What's the simplest way to regenerate
> the kernel when there are no kernel changes but you have changes to
> the programs that are going into the initramfs? make clean seems like
> overkill to me, and it's very slow to boot.
>
<SNIP>

Answering self: The Gentoo wiki covered this at the end:

http://en.gentoo-wiki.com/wiki/Initramfs

I tried it and it seemed to work, although I've not yet successfully
booted this kernel so there may well be some issues left to deal with.

Cheers

Re: [gentoo-user] Building an initramfs into the kernel

[Mark Knecht]

On Thu, Dec 27, 2012 at 3:24 PM, Mark Knecht <markknecht@gmail.com> wrote:
> On Thu, Dec 27, 2012 at 9:33 AM, Mark Knecht <markknecht@gmail.com> wrote:
>> On Wed, Dec 26, 2012 at 3:50 PM, Mark Knecht <markknecht@gmail.com> wrote:
>>> On Wed, Dec 26, 2012 at 3:03 PM, Neil Bothwick <neil@digimed.co.uk> wrote:
>>>> On Wed, 26 Dec 2012 14:09:34 -0800, Mark Knecht wrote:
>>>>
>>>>> At this point I don't know that 1) the image is actually in the
>>>>> kernel, or 2) that my "init thingy" ;-) image would work, but at least
>>>>> the process of putting it together is verifiable.
>>>>
>>>> That's why I put in all the debug stuff, so I could watch the progress of
>>>> the script as it ran.
>>>>
>>>
>>>
>>> OK, I'll look at combining that part my my scripts, or just using yours, etc.
>>>
>>> Thanks for the help,
>>> Mark
>>
>> Neil,
>>    One more question if I might. What's the simplest way to regenerate
>> the kernel when there are no kernel changes but you have changes to
>> the programs that are going into the initramfs? make clean seems like
>> overkill to me, and it's very slow to boot.
>>
> <SNIP>
>
> Answering self: The Gentoo wiki covered this at the end:
>
> http://en.gentoo-wiki.com/wiki/Initramfs
>
> I tried it and it seemed to work, although I've not yet successfully
> booted this kernel so there may well be some issues left to deal with.
>
> Cheers

OK, this is all finished up and working now. I'll have to give it some
time to make sure nothing weird pops up but it was a good learning
experience for me.

The two biggest problems I had along way:

1) I used an older init script that I had put together a year back
with I played with this for an afternoon. I had trouble back then with
/dev so I copied over pretty much everything into my initramfs. (As
per the Gentoo initramfs wiki) This time around, using that script, I
still had troubles until I discovered the

mount -t devtmpfs none /dev

command in Neil's script. That allowed me to move forward.

2) Turns out the config format for copying files is target/source, not
source/target like most copies. This resulted in my mdadm.conf file
going into the wrong directory.

Once I figured those two out the machine booted cleanly.

I'm going to continue working on this. I'd like to do a better init
script, maybe add some more stuff to play with. That said I'll attach
in-line the stuff I ended up with for anyone else who comes across
this thread in the future.

Cheers,
Mark


c2stable src # ls -la
total 28
drwxr-xr-x  4 root root 4096 Dec 28 15:04 .
drwxr-xr-x 14 root root 4096 Dec 25  2011 ..
-rw-r--r--  1 root root    0 Dec 29  2010 .keep
-rw-r--r--  1 root root  574 Dec 28 11:45 initramfs.config
-rw-r--r--  1 root root  562 Dec 28 10:44 initramfs_init_new.sh
lrwxrwxrwx  1 root root   19 Dec 27 07:49 linux -> linux-3.6.11-gentoo
drwxr-xr-x 24 root root 4096 Dec 28 09:49 linux-3.2.1-gentoo-r2
drwxr-xr-x 24 root root 4096 Dec 28 11:46 linux-3.6.11-gentoo
-rw-r--r--  1 root root   70 Dec 28 06:58 mdadm_initramfs.conf





c2stable src # cat initramfs.config
dir /bin 755 0 0
file /bin/busybox /bin/busybox 755 0 0
slink /bin/sh busybox 777 0 0

dir /realroot 755 0 0
dir /etc 755 0 0
dir /proc 755 0 0
dir /sys 755 0 0

dir /sbin 755 0 0
file /sbin/mdadm /sbin/mdadm 755 0 0

file /sbin/e2fsck /sbin/e2fsck 755 0 0
dir /lib 755 0 0
file /lib/libext2fs.so /usr/lib64/libext2fs.so 755 0 0
file /etc/mdadm.conf /usr/src/mdadm_initramfs.conf 755 0 0

dir /dev 755 0 0
nod /dev/console 600 0 0 c 5 1
nod /dev/null 666 0 0 c 1 3
nod /dev/tty 666 0 0 c 5 0
nod /dev/urandom 666 0 0 c 1 9

file /init /usr/src/initramfs_init_new.sh 755 0 0





c2stable src # cat initramfs_init_new.sh
#!/bin/busybox sh

rescue_shell() {
    echo "Something went wrong. Dropping you to a shell."
    busybox --install -s
    exec /bin/sh
}

# Mount the /proc and /sys filesystems.
mount -t proc none /proc
mount -t sysfs none /sys
mount -t devtmpfs none /dev

# Do your stuff here.
echo "This script mounts rootfs and boots it up, nothing more!"

mdadm --assemble /dev/md3

# Mount the root filesystem.
mount -o ro /dev/md3 /realroot  || rescue_shell

# Clean up.
umount /dev
umount /proc
umount /sys

# Boot the real thing.
exec switch_root /realroot /sbin/init






c2stable src # cat mdadm_initramfs.conf
ARRAY /dev/md/3 metadata=1.2 UUID=de47f991:86d98467:0637635b:9c6d0591
c2stable src #

Re: [gentoo-user] Building an initramfs into the kernel

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 626 bytes --]

On Thu, 27 Dec 2012 09:33:48 -0800, Mark Knecht wrote:

>    One more question if I might. What's the simplest way to regenerate
> the kernel when there are no kernel changes but you have changes to
> the programs that are going into the initramfs? make clean seems like
> overkill to me, and it's very slow to boot.

Plain make works for me, it appears the initramfs is rebuilt every time,
evewn if there are no changes. To be extra sure, you could delete
usr/initramfs_data.cpio from the source tree before running make.


-- 
Neil Bothwick

You are a completely unique individual, just like everybody else.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-user] Building an initramfs into the kernel

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 1069 bytes --]

On Wed, 26 Dec 2012 12:48:38 -0800, Mark Knecht wrote:

> On the mdadm list one comment came up which was to make sure
> my mdadm.conf file was up to date and then include it in the initramfs
> image so that if for some reason the machine doesn't boot I have all
> the info required to start the RAIDs by hand. I thought that was a good
> comment.

That is a good idea.

> One other question came up. So I build the kernel, and the kernel build
> creates the initramfs image and embeds it in the kernel. Is there a tool
> that would allow me to query the embedded image prior to booting
> so that I can check that everything is in there that I think should be
> in there?

Yes. I had to do this once when I corrupted or deleted my init.sh and
needed to retrieve it from the kernel. I don't remember the details, but
I do recall it involved one of the scripts in the kernel source
directory and the use of Google - although not in that order :)


-- 
Neil Bothwick

"I am a Cub Ranger. We dib dib dib for the One. We dob dob dob for the
One."

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]