[arch-releng] next snapshot: 2012.10

[Gerardo Exequiel Pozzi]

On 10/03/2012 11:26 AM, Dave Reisner wrote:
> On Wed, Oct 03, 2012 at 11:12:24AM -0300, Gerardo Exequiel Pozzi wrote:
>> On 10/03/2012 10:54 AM, Dave Reisner wrote:
>>> On Wed, Oct 03, 2012 at 02:37:52AM -0300, Gerardo Exequiel Pozzi wrote:
>>>> Hello
>>>>
>>>> This is a list of user-view-changes from archiso side:
>>>> * New packages: xl2tpd, refind-efi-x86_64, gummiboot-efi,
>>>> fsarchiver, mc, partclone, partimage, testdisk
>>>> * Re-added pkgs: sudo, wget
>>>> * Simplified steps to make an USB PC-EFI from ISO
>>>> * Usage of gummiboot for fancy menu when booting in EFI mode.
>>>> * The /script/ boot parameter works again (really really now!)
>>>>
>>>> * systemd (PENDING) [ready for merge to master] I want to do this,
>>>> otherwise there is no user feedback.
>>> Is there anything we can do to unroll the rats' nest of squashfs loops
>>> to avoid the spammy mess at shutdown?
>> There is no way to disable this cleanup-loop in systemd. This loop
> It's obviously not a deal breaker since it's R/O loops on an equally R/O
> media, but my point wasn't to fix this in systemd. We have squashes
> inside squashes (inside squashes?). It really doesn't seem necessary,
> and other liveCDs don't have such complex "requirements". Can you
> explain why this setup is needed/wanted?

jeje, where is done this? :P Because we need a writable / filesystem, 
and Linux does not support any type of overflayfs at this moment. Lots 
and lots of intents was made during years without success acceptation.
Also we need a compression at the same time :)

So the way to do this is using a block based device. What is done:

[rw-fs] (ext4 /) -> [dm-snapshot] -> {ro-fs-loop} (origin) + 
{rw-cow-loop} (COW)
     {ro-fs-loop} -> [fs-image] -> [ro-sfs-loop] -> [sfs-image] -> 
[ro-fs] (iso9660 / extfs / vfat)
     {rw-cow-loop} -> [cow-image] -> [rw-fs] (tmpfs/non-persistent | 
extfs/persistent)

Where "origin" is one of "root-image" (i686/x86_64) "usr-share" (any) 
and "usr-lib-modules"(*) (i686/x86_64)

As you can see the only thing inside an squashfs images is only one file 
that is an ext4 image.


(*) This split allows (unofficially) boot x86_64 kernel with i686 userspace.

>
>> is only disabled when running inside a container [#1].
>> Maybe we can submit to upstream a patch that in some way, we can
>> instruct to systemd to skip this logic. For example if some files
>> are found on /run/initramfs like "skip_loop_detach" "skip_dm_detach"
> I've already gone down this road, and it's a dead end.
>
>> If you look at Fedora Live-CD, you can see similar shutdown messages.
> Lucky for us, we aren't Fedora =)
cuac!
>> [#1] http://cgit.freedesktop.org/systemd/systemd/tree/src/core/shutdown.c#n184
>>>> Good Luck
>>>>
>>>> -- 
>>>> Gerardo Exequiel Pozzi
>>>> \cos^2\alpha + \sin^2\alpha = 1
>>>>
>>
>> -- 
>> Gerardo Exequiel Pozzi
>> \cos^2\alpha + \sin^2\alpha = 1
>>


-- 
Gerardo Exequiel Pozzi
\cos^2\alpha + \sin^2\alpha = 1