Arch Linux

Please read this before reporting a bug:

Do NOT report bugs when a package is just outdated, or it is in Unsupported. Use the 'flag out of date' link on the package page, or the Mailing List.

REPEAT: Do NOT report bugs for outdated packages!

FS#24259 - [kernel26] AUFS2 is against "The Arch Way"

Attached to Project: Arch Linux
Opened by Jamie Nguyen (jnguyen) - Friday, 13 May 2011, 15:03 GMT
Last edited by Thomas Bächler (brain0) - Sunday, 03 July 2011, 21:12 GMT
Task Type Bug Report
Category Packages: Core
Status Closed
Assigned To Tobias Powalowski (tpowa)
Thomas Bächler (brain0)
Architecture All
Severity Low
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 4
Private No


The Arch Linux kernel is currently patched with aufs2. This is against "The Arch Way". Quotes:

"Arch provides non-patched, vanilla software; packages are offered from pure upstream sources, how the author originally intended it be distributed. Patching only occurs in extremely rare cases, to prevent severe breakage in the instance of version mismatches that may occur within a rolling release model."

"Arch Linux defines simplicity as without unnecessary additions, modifications, or complications..."

"clean, correct, simple code, rather than unnecessary patching"

"Software patches are therefore kept to an absolute minimum; ideally, never."

Judging from these quotes, either the aufs2 patches need to be removed, or the above quotes need to be modified. Not only is aufs2 an unneccessary modification, it is modification to the kernel itself which is the very heart of what makes Arch Linux a Linux distribution.

I know there are people that use aufs2, but aufs2 should never have been in the kernel in the first place. It should be an AUR kernel and people that need it should build their own kernel instead, just like users that build kernel26-ck or kernel26-pf or kernel26-ccs. Those are patches that are _not_ mainline, and consequently they are in the AUR.

I accept that this might be marked as "won't fix" and aufs2 users will vote against this, but one can't deny that it's contradictory to say "no unnecessary patching" but then to patch the Linux kernel. Don't worry, I don't plan to start a massive flamewar campaign (it's not that big a deal)... consider this as just a reminder that maybe we shouldn't forget the ideals that Arch Linux is built on, and if those ideals have changed (which I hope they haven't), then the Wiki should reflect this.
This task depends upon

Closed by  Thomas Bächler (brain0)
Sunday, 03 July 2011, 21:12 GMT
Reason for closing:  None
Comment by Ionut Biru (wonder) - Friday, 13 May 2011, 15:53 GMT
so what are you suggesting is to make the iso and the installation media obsolete?
Comment by Jamie Nguyen (jnguyen) - Friday, 13 May 2011, 16:01 GMT
I sense some sarcasm in your reply :-)

Allow me to retort:
Comment by Gerardo Exequiel Pozzi (djgera) - Friday, 13 May 2011, 16:47 GMT
I will skip the "philosophical" part.

The patchs called "aufs2*" just only makes some functions visible for buildtime and runtime of external modules & There is no aufs2 code in these patchs. => There is no "aufs2" patch in kernel26 package. QED.
Comment by Jamie Nguyen (jnguyen) - Friday, 13 May 2011, 17:04 GMT
Gerardo Exequiel Pozzi wrote:
> I will skip the "philosophical" part.

I suppose the debate over whether the aufs2* patches count as patches is itself another philosophical debate :p . For the record, I still do count them as unnecessary modifications. From the ML link above, it seems that tpowa wanted aufs2 to be dropped and brain0 seemed indifferent (the two main kernel devs AFAIK). I don't know if they still have the same opinion.
Comment by Gerardo Exequiel Pozzi (djgera) - Friday, 13 May 2011, 17:35 GMT
Sure it can be dropped if using unionfs-fuse on archiso. Some time ago I sent a RFC patch for that to arch-releng ML.
Comment by Thomas Bächler (brain0) - Friday, 13 May 2011, 22:02 GMT
Gerardo, does any of that fuse-based unionfs stuff actually work? From what I saw, none of these systems (and there are a few) are actually actively maintained and feature-complete.

Jamie, what you are doing is what I call "destructive criticism". You want something to be removed for ideological reasons, yet you provide no alternative solutions. And you do not seem to care for the reason why we actually have aufs2. Please do your research, then complain.
Comment by Jamie Nguyen (jnguyen) - Friday, 13 May 2011, 22:29 GMT
Thomas wrote:
> Please do your research, then complain.

Hi Thomas, I tried my best not to be inflammatory, honest ;-) I do realise that the ISO requires aufs2. The mailing list thread I posted above I declare as my prior research. In hindsight, I should have mentioned this in the original bug report (if only my skills of foresight were as great as my skills of hindsight!). I apologise that I appeared to "not seem to care for the reason why we actually have aufs2", but that is simply not true.

Thomas wrote:
> yet you provide no alternative solutions

I believe you yourself provided a solution:

Thomas wrote:
> We could maintain an extra kernel only for the live CD if that is easier.

I would of course not always be so willing to suggest a solution that requires the developers to perform more work than they already are, especially considering this is a voluntary project done in free time for no money, but it was you yourself that suggested this solution :-) Perhaps you no longer consider that solution as viable, but I was unable to find any further public communication related to this.
Comment by Gerardo Exequiel Pozzi (djgera) - Friday, 13 May 2011, 22:58 GMT
@Thomas: When I tested unionfs-fuse [#1] just works, not extensivelly tested, but for the purpose of archiso it does the right job. Last update is one year ago.

Comment by orbisvicis (orbisvicis) - Tuesday, 14 June 2011, 01:47 GMT
I miss aufs2.

Let me retort:

as per aufs webapge:

'Note: it becomes clear that "Aufs was rejected. Let's give it up."
According to Christoph Hellwig, linux rejects all union-type filesystems
but UnionMount.'

Unfortunately the "kosher" approach is in no better condition:

'Union mounts require patches to the kernel, e2fsprogs, and util-linux.'

The last modification was nearly a year ago, and only support ext2, jffs2 and tmpfs.

So either approach seems to offer philosophical quandaries.

As for the practical aspect, I need to create a read-only overlay from multiple directories. The more generic, the simpler the approach, the better.
Comment by Pascal Lemazurier (pascal.lemazurier) - Wednesday, 15 June 2011, 22:33 GMT
Good evening, I got an hard problem after this "bug" resolution.

You wanted to remove the kernel patch of AUFS2, to be more "KISS rules" conform, well.

I thought firstly It was a temporary remove of aufs2 and aufs2-tool packages because the kernel was updated to 2.6.39.

I manage a french ArchLinux Live-CD community project... And now I can't simply build any working .iso. with the standard kernel26 package, and there is no more aufs2 aufs2-toom official packages...

I haven't worked on an alternative solution yet. Is there another simple solution to make AUFS2 works ?
Comment by Gerardo Exequiel Pozzi (djgera) - Wednesday, 15 June 2011, 22:38 GMT
@Pascal: Please see arch-releng ML, and test archiso with dm-snapshot.
Comment by GI Jack (GI_Jack) - Sunday, 03 July 2011, 18:52 GMT
you troll, you just unexpectantly broke my live CD. At very least, release a stable live-cd friendly kernel package, like kernel26-live or something