[arch-general] Better NILFS2 Support

Alex Matviychuk

23 Jul 2010 23 Jul '10

12:47 a.m.

I just spent some time setting up my new laptop with an SSD to use NILFS2 as its root filesystem and I'm very pleased with the performance. In an effort to get better support for NILFS2 in arch I've added a request to move nilfs-utils to the base group and submitted a patch to initscripts to bring up the nlfs_cleanerd daemon if root is formatted as NILFS2. If NILFS2 is used on other partitions the mount command brings up the daemon on it's own. These two bugs should make it easier to add NILFS2 support to the Arch Installation Framework. Related bugs here, patch comments welcome: http://bugs.archlinux.org/task/20260 http://bugs.archlinux.org/task/20261 Cheers, alexmat

Show replies by thread

Dieter Plaetinck

1 Aug 1 Aug

11:51 a.m.

On Thu, 22 Jul 2010 15:47:24 -0700 Alex Matviychuk wrote:

...

I just spent some time setting up my new laptop with an SSD to use NILFS2 as its root filesystem and I'm very pleased with the performance.

In an effort to get better support for NILFS2 in arch I've added a request to move nilfs-utils to the base group and submitted a patch to initscripts to bring up the nlfs_cleanerd daemon if root is formatted as NILFS2. If NILFS2 is used on other partitions the mount command brings up the daemon on it's own.

These two bugs should make it easier to add NILFS2 support to the Arch Installation Framework.

Related bugs here, patch comments welcome:

http://bugs.archlinux.org/task/20260 http://bugs.archlinux.org/task/20261

Cheers, alexmat

good stuff Alex. Now who wants to patch aif? (should be trivial if the FS behaves like the others). Dieter

Heiko Baums

4:46 p.m.

...

On Thu, 22 Jul 2010 15:47:24 -0700 Alex Matviychuk wrote:

...
In an effort to get better support for NILFS2 in arch I've added a request to move nilfs-utils to the base group

I don't think that nilfs-utils should be moved to the base group. I agree with moving it to [core] but not to base, because base is assumed to be installed on every computer and packages in the base group are usually not listed in the depends array of a PKGBUILD. On the contrary I think there could be some other file system tools like jfsutils, lvm2 and xfsprogs be removed from the base group (not from [core]). They probably should be automatically installed if a partition is formatted with one of these filesystems by AIF. But usually people who are using these filesystems know what they are doing and know that they have to select these packages for installation. Heiko

Dieter Plaetinck

4:58 p.m.

On Sun, 1 Aug 2010 16:46:33 +0200 Heiko Baums wrote:

...

I don't think that nilfs-utils should be moved to the base group. I agree with moving it to [core] but not to base, because base is assumed to be installed on every computer and packages in the base group are usually not listed in the depends array of a PKGBUILD.

On the contrary I think there could be some other file system tools like jfsutils, lvm2 and xfsprogs be removed from the base group (not from [core]).

I think this explanation makes sense. I just found this back: http://wiki.archlinux.org/index.php/User:Allan/Base_Cleanup So indeed, it seems like the goal is to remove all non-essential things (reiserfs, xfs, ..) from base.

...

They probably should be automatically installed if a partition is formatted with one of these filesystems by AIF. But usually people who are using these filesystems know what they are doing and know that they have to select these packages for installation.

Actually it's fairly trivial to "preselect" those packagas based on the filesystems used during the filesystem step. I believe some of it is already implemented in aif. Dieter

Tobias Powalowski

5:30 p.m.

...

On Sun, 1 Aug 2010 16:46:33 +0200

Heiko Baums wrote:

...
I don't

...

...
agree with moving it to [core] but not to base, because base is assumed to be installed on every computer and packages in the base group are usually not listed in the depends array of a PKGBUILD.

On the contrary I

...

...
like jfsutils, lvm2 and xfsprogs be removed from the base group (not from [core]).

I

Am Sonntag 01 August 2010 schrieb Dieter Plaetinck: think that nilfs-utils should be moved to the base group. I think there could be some other file system tools think this explanation makes sense. I just found this back:

...

So indeed, it seems like the goal is to remove all non-essential things (reiserfs, xfs, ..) from base.

...
They probably should be automatically installed if a

...

...
formatted with one of these filesystems by AIF. But usually

...

...
are using these filesystems know what they are doing and know

...

...
they have to select these packages for installation.

Actually it's fairly trivial to "preselect" those packagas based on the filesystems used during the filesystem step. I believe some of it is already implemented in aif.

Dieter My 2 cents about nilfs, I have added it to archboot environment some time ago. To integrate it into an installer it is

http://wiki.archlinux.org/index.php/User:Allan/Base_Cleanup partition is people who that plain not usable, because the kernel doesn't know it. All usual filesystem utilities like blkid etc. don't recognize it. Sure it will be an optional filesystem in the future, but it should be supported by the kernel and major filesystem and core utiltities in a sane way. greetings tpowa -- Tobias Powalowski Archlinux Developer & Package Maintainer (tpowa) http://www.archlinux.org tpowa@archlinux.org

Dan McGee

5:38 p.m.

On Sun, Aug 1, 2010 at 10:30 AM, Tobias Powalowski wrote:

...

...
On Sun, 1 Aug 2010 16:46:33 +0200

Heiko Baums wrote:

...
I don't

...
...
agree with moving it to [core] but not to base, because base is assumed to be installed on every computer and packages in the base group are usually not listed in the depends array of a PKGBUILD.

On the contrary I

...
...
like jfsutils, lvm2 and xfsprogs be removed from the base group (not from [core]).

I

Am Sonntag 01 August 2010 schrieb Dieter Plaetinck: think that nilfs-utils should be moved to the base group. I think there could be some other file system tools think this explanation makes sense. I just found this back:

...
...
So indeed, it seems like the goal is to remove all non-essential things (reiserfs, xfs, ..) from base.

...
They probably should be automatically installed if a

...
...
formatted with one of these filesystems by AIF. But usually

...
...
are using these filesystems know what they are doing and know

...
...
they have to select these packages for installation.

Actually it's fairly trivial to "preselect" those packagas based on the filesystems used during the filesystem step. I believe some of it is already implemented in aif.

Dieter My 2 cents about nilfs, I have added it to archboot environment some time ago. To integrate it into an installer it is

http://wiki.archlinux.org/index.php/User:Allan/Base_Cleanup partition is people who that plain not usable, because the kernel doesn't know it. All usual filesystem utilities like blkid etc. don't recognize it. Sure it will be an optional filesystem in the future, but it should be supported by the kernel and major filesystem and core utiltities in a sane way.

Except that I've seen it running on his root partition on a booting laptop without any of these problems you speak of... -Dan

Tobias Powalowski

8:30 p.m.

...

On Sun, Aug 1, 2010 at 10:30 AM, Tobias Powalowski wrote:

...
Am Sonntag 01 August 2010 schrieb Dieter Plaetinck:

...
On Sun, 1 Aug 2010

16:46:33 +0200

...
...
Heiko Baums wrote:

...
I don't

Am Sonntag 01 August 2010 schrieb Dan McGee: think that nilfs-utils should be moved to the base group. I

...

...
...
...
agree with

...
moving it to [core] but not to base, because base is

...

...
...
...
assumed to be

installed on every computer and packages in the base

...
...
group are usually

not listed in the depends array of a PKGBUILD.

...
...
On the contrary I

think there could be some other file system tools

...
...
like jfsutils, lvm2

and xfsprogs be removed from the base group (not

...
...
...
from

[core]).

...
...
I

think this explanation makes sense. I just found this back:

http://wiki.archlinux.org/index.php/User:Allan/Base_Cleanup

...
...
So

indeed, it

...
seems like the goal is to remove all non-essential

things

...
...
(reiserfs, xfs,

..) from base.

...
...
They

probably should be automatically installed if a

...
partition is

...
...
formatted with one of these filesystems by AIF. But usually

people who

...
...
...
are using these filesystems know what they are doing

and know

...
that

...
...
they have to select these packages for

installation.

...
...
Actually

it's fairly trivial to "preselect" those packagas based on the

...
filesystems

used during the filesystem step. I believe some of it is

...
already

...

...
implemented in aif.

...
Dieter

My 2 cents about nilfs,

...
archboot environment some time ago. To integrate it into an installer it is plain not usable, because the kernel doesn't know it. All usual filesystem utilities like blkid etc. don't recognize it. Sure it will be an optional filesystem in

I have added it to the future, but it should be supported by the kernel and

...
major filesystem and core utiltities in a sane way.

Except that I've seen it running on his root partition on a booting laptop without any of these problems you speak of...

-Dan Yes sure i believe it works, my concerns are more that the main tools just don't support it, which makes it imho more complicated to implement the installation support.

greetings tpowa -- Tobias Powalowski Archlinux Developer & Package Maintainer (tpowa) http://www.archlinux.org tpowa@archlinux.org

Alex Matviychuk

2 Aug 2 Aug

12:56 a.m.

...

Yes sure i believe it works, my concerns are more that the main tools just don't support it, which makes it imho more complicated to implement the installation support.

So here is what I needed to mess with to get it working: Kernel: "Version 2 of the filesystem, known as NILFS2, is included in Linux kernel 2.6.30." Works out of the box with the most recent arch kernel. mkinitcpio.conf: I just added: MODULES="nilfs2" I wonder if this is something that can be easily added to the filesystems hook. /boot/grub/menu.lst: kernel /vmlinuz26 root=/dev/sda3 ro rootfstype=nilfs2 AIF would need to take this into account, but seems easy enough. Initscripts: I provided a patch here: http://bugs.archlinux.org/task/20260 No other troubles as far as I know. Is there anything I'm missing? Let me know where I can help, this is an awesome filesystem. Blazing fast and when I delete or overwrite things on accident, it's so cool to just mount a snapshot from like an hour ago without having to set anything up in advance, and without even unmounting my root fs. Cheers, Alex

Tobias Powalowski

8:27 a.m.

Am Montag 02 August 2010 schrieb Alex Matviychuk:

...

...
Yes sure i believe it works, my concerns are more that the main tools just don't support it, which makes it imho more complicated to implement the installation support.

So here is what I needed to mess with to get it working:

...

"Version 2 of the filesystem, known as NILFS2, is included in Linux kernel 2.6.30." Works out of the box with the most recent arch kernel.

mkinitcpio.conf: I just added: MODULES="nilfs2" I wonder if

Kernel: this is something that can be easily added to the filesystems

...

hook.

...

kernel /vmlinuz26 root=/dev/sda3 ro rootfstype=nilfs2 AIF would need to take this into account, but seems easy enough.

Initscripts: I provided a patch here: http://bugs.archlinux.org/task/20260

No other troubles as far as I know. Is there anything I'm missing?

Let me know where I can help, this is an awesome filesystem. Blazing fast and when I delete or overwrite

/boot/grub/menu.lst: things on accident, it's so cool

...

to just mount a snapshot from like an hour ago without having to set anything up in advance, and without even unmounting my root fs.

Cheers, Alex According to the util-linux git it seems new version will have nilfs2 support, if this happens archboot support will happen very soon.

greetings tpowa -- Tobias Powalowski Archlinux Developer & Package Maintainer (tpowa) http://www.archlinux.org tpowa@archlinux.org

Pierre Chapuis

3:31 p.m.

On Sun, 1 Aug 2010 16:58:00 +0200, Dieter Plaetinck wrote:

...

On Sun, 1 Aug 2010 16:46:33 +0200 Heiko Baums wrote:

...
I don't think that nilfs-utils should be moved to the base group. I agree with moving it to [core] but not to base, because base is assumed to be installed on every computer and packages in the base group are usually not listed in the depends array of a PKGBUILD.

On the contrary I think there could be some other file system tools like jfsutils, lvm2 and xfsprogs be removed from the base group (not from [core]).

I think this explanation makes sense. I just found this back: http://wiki.archlinux.org/index.php/User:Allan/Base_Cleanup So indeed, it seems like the goal is to remove all non-essential things (reiserfs, xfs, ..) from base.

Just for information, the opposite point of view recently came up on the suckless mailing-list: http://lists.suckless.org/dev/1007/5256.html I prefer Arch's approach but it is probably true that a large base system reduces the workload of package maintainers. -- Pierre 'catwell' Chapuis

Dieter Plaetinck

4:44 p.m.

On Mon, 02 Aug 2010 15:31:47 +0200 Pierre Chapuis wrote:

...

On Sun, 1 Aug 2010 16:58:00 +0200, Dieter Plaetinck wrote:

...
On Sun, 1 Aug 2010 16:46:33 +0200 Heiko Baums wrote:

...
I don't think that nilfs-utils should be moved to the base group. I agree with moving it to [core] but not to base, because base is assumed to be installed on every computer and packages in the base group are usually not listed in the depends array of a PKGBUILD.

On the contrary I think there could be some other file system tools like jfsutils, lvm2 and xfsprogs be removed from the base group (not from [core]).

I think this explanation makes sense. I just found this back: http://wiki.archlinux.org/index.php/User:Allan/Base_Cleanup So indeed, it seems like the goal is to remove all non-essential things (reiserfs, xfs, ..) from base.

Just for information, the opposite point of view recently came up on the suckless mailing-list: http://lists.suckless.org/dev/1007/5256.html

I prefer Arch's approach but it is probably true that a large base system reduces the workload of package maintainers.

I don't think packaging becomes much harder when you remove optional filesystems and configuration tools from base. Dieter

Heiko Baums

6:08 p.m.

Am Mon, 02 Aug 2010 15:31:47 +0200 schrieb Pierre Chapuis :

...

Just for information, the opposite point of view recently came up on the suckless mailing-list: http://lists.suckless.org/dev/1007/5256.html

I prefer Arch's approach but it is probably true that a large base system reduces the workload of package maintainers.

I don't know Source Mage, but I think that a large base system (as I understand this posting, all in one big package) has only disadvantages. It's far less flexible, wastes much more disk space for unneeded tools, makes a lot more work for the users and maintainers, because the base package - as I understand the posting - needs to be updated more often (a big package needs to be updated every time one small tool is updated), which forces the users to regularly download and the maintainers to regularly rebuild and upload a bigger package instead of sometimes down- and uploading a few much smaller packages, etc. Not really a good idea. Heiko

Dieter Plaetinck

6:10 p.m.

On Mon, 2 Aug 2010 18:08:50 +0200 Heiko Baums wrote:

...

Am Mon, 02 Aug 2010 15:31:47 +0200 schrieb Pierre Chapuis :

...
Just for information, the opposite point of view recently came up on the suckless mailing-list: http://lists.suckless.org/dev/1007/5256.html

I prefer Arch's approach but it is probably true that a large base system reduces the workload of package maintainers.

I don't know Source Mage, but I think that a large base system (as I understand this posting, all in one big package) has only disadvantages. It's far less flexible, wastes much more disk space for unneeded tools, makes a lot more work for the users and maintainers, because the base package - as I understand the posting - needs to be updated more often (a big package needs to be updated every time one small tool is updated), which forces the users to regularly download and the maintainers to regularly rebuild and upload a bigger package instead of sometimes down- and uploading a few much smaller packages, etc.

Not really a good idea.

Heiko

I think they mean keeping regular packages like usual, but just making more packages part of the base group. And like I said earlier, I don't see the point either. Dieter

Heiko Baums

6:19 p.m.

Am Mon, 2 Aug 2010 18:10:22 +0200 schrieb Dieter Plaetinck :

...

I think they mean keeping regular packages like usual, but just making more packages part of the base group. And like I said earlier, I don't see the point either.

This indeed doesn't make much sense, too. I totally agree with you. Heiko

5013

Age (days ago)

5024

Last active (days ago)

List overview

Download

13 comments

6 participants

participants (6)

Alex Matviychuk
Dan McGee
Dieter Plaetinck
Heiko Baums
Pierre Chapuis
Tobias Powalowski