[arch-general] Kernel ext2 driver
Hi list, just wanted to ask, why ext2 isn't built directly into the kernel anymore? Is there any special reason? I would prefer to have ext2 built in again, because I'm dealing a lot with special configurations of archlinux, like thinclient installations where the whole image is running in an ext2 formated ramdisk and so on. Currently I'm building the complete kernel everytime I need to do an upgrade. But I think the ext2 driver isn't such a big ballast for standard systems, so you might want to do me a favour and set it as a built in driver of the standard kernel again. Would save me a lot of compile time. kind regards, --peter
2008/3/1, Peter Feuerer <peter.feuerer@linux-gamers.net>:
I don't understand where the problem is. You can just add it to initcpio image. -- Roman Kyrylych (Роман Кирилич)
On Sat, 2008-03-01 at 20:28 +0200, Roman Kyrylych wrote:
I don't use the arch initcpio generated initramfs. I create a standalone "old" initrd containing the complete root filesystem of the arch installation myself. And due to the fact that the initrd is using ext2 as filesystem the kernel can't read it unless i compile ext2 in. --peter
Peter Feuerer schrieb:
Just don't use a filesystem for your initrd. Create a cpio image and use it as an initramfs image. Apart from the fact that it is easier to create, there is also no unnecessary overhead like in case of a ramdisk. Usage: gen_init_cpio <cpio_list> <cpio_list> is a file containing newline separated entries that describe the files to be included in the initramfs archive: # a comment file <name> <location> <mode> <uid> <gid> dir <name> <mode> <uid> <gid> nod <name> <mode> <uid> <gid> <dev_type> <maj> <min> slink <name> <target> <mode> <uid> <gid> pipe <name> <mode> <uid> <gid> sock <name> <mode> <uid> <gid> <name> name of the file/dir/nod/etc in the archive <location> location of the file in the current filesystem <target> link target <mode> mode/permissions of the file <uid> user id (0=root) <gid> group id (0=root) <dev_type> device type (b=block, c=character) <maj> major number of nod <min> minor number of nod example: # A simple initramfs dir /dev 0755 0 0 nod /dev/console 0600 0 0 c 5 1 dir /root 0700 0 0 dir /sbin 0755 0 0 file /sbin/kinit /usr/src/klibc/kinit/kinit 0755 0 0
On Sat, 2008-03-01 at 20:27 +0100, Thomas Bächler wrote:
Thanks for this answer, creating initramfs instead of initrd's was an option I already thought about. But the thing is, that I'm using the initrd-way already for many years. And it turned out to be a stable and deterministic solution. Switching to initramfs is not just creating the cpio archive. It is also about testing the target systems. It begins with testing the realtime behaviour of solutions like rtai or the rt-preempt patch and ends up with the testing of the daily usage of my thin clients. I mean, I could only switch for my recent projects. But then I have to take care about two solutions, what does not make much sense at all. That's why I want to stick to the old initrd way until I found time to do all the testing stuff with initramfs and I can completely switch. Anyway, my question was not about finding a way to replace the initrd. I asked if it would be possible to add ext2 into the standard kernel's config of archlinux again. Through many ppl use ext2 anyway, e.g. for their /boot partition. And the driver is not that big (the module is 74k). I do understand, you don't want to have every little driver in the kernel, but come on, ext2 is a very common filesystem... --peter
On Sat, 2008-03-01 at 22:42 +0100, Peter Feuerer wrote:
Ext2 is a dead filesystem and has been replaced by ext3. We don't have use for any filesystem in our kernel, as our kernel doesn't need any filesystem driver to boot. Adding ext2 as static configuration option will bring up the next question: "I don't want to use initcpio, I want to have <insert random filesystem> built in the kernel".
On Sat, Mar 1, 2008 at 7:51 PM, Jan de Groot <jan@jgc.homeip.net> wrote:
Your entire rest of statement was valid, but why on earth did you make a claim like this? I have to speak up and say something here. ext2 is nowhere near dead. I use it on my /boot partition on every Linux install, and every filesystem on my Eee is currently ext2. Hardly dead- unless "stable as hell" = dead.
This is a valid justification- thanks for giving the best answer on this thread yet. -Dan
On Sat, 2008-03-01 at 20:50 -0600, Dan McGee wrote:
Ext2 is stable as hell, so is ext3. The point is that nowadays almost nobody wants to use a filesystem without journal, that's why I consider it dead. If journalling wasn't important (and if I didn't need filesizes bigger than 2 or 4GB), then FAT32 wouldn't be a dead filesystem either. Ext3 is basically just Ext2 with an added journal, so that's why Ext2 is still developed and supported.
On Sunday 02 March 2008, Jan de Groot wrote:
then FAT32 wouldn't be a dead filesystem either.
Well, in my experience almost every USB stick or newly buyed hard disk comes formated on FAT32. I don't wanna get into why is that, but it's just the way it is. By the way, I know many people that having a perfectly stable systen with a UPS doesn't want to use journaling.
On Sun, 2008-03-02 at 14:11 +0100, Marc Deop i Argemí wrote:
External harddisks and USB sticks come preformatted with FAT32 because NTFS isn't compatible with Mac OS X. FAT32 is the only choice for these vendors if they want cross-compatibility. When you install Windows on a >32GB partition, the setup doesn't allow you to format FAT32, only NTFS. Reasoning behind this is that you get inefficient cluster sizes on FAT32 then, and scandisk operations at bootup take ages to complete with many errors due to a missing journal.
On Sunday 02 March 2008, Jan de Groot wrote:
As I said, i don't want to get into why the things are how they are. The fact is that neither FAT32 nor ext2 are dead and are widely used.
This thread is stupid. If you want ext2 compiled into the kernel by default.. for goodness sake, do it yourself. It would be what.. a two line change in the ABS pkgbuild for the kernel? The original poster said he had a thin client farm using the old initrd image, so he has likely diverged from the current kernel install enough that he should be using (and testing) his own kernel PKGBUILD anyway. Storm in a teacup. /me facepalms
On Sun, 2008-03-02 at 13:59 -0800, eliott wrote:
don't be so aggressive... I just asked friendly to add ext2 to the standard kernel, because that is the only change I need to do, to run my systems. And the answer of the arch kernel devs is clear and I do understand that. So I will build my own kernel with abs, like i did since ext2 was set to module. please don't do further discussion on this thread! ty, --peter
My apologies. I reread my post and it does appear a bit cranky. Not sure about 'agressive', but it was definitely 'grumpy'.
Um,..... I *ALSO* use ext2 for my /boot, and for ALL the reasons pointed out in this thread. PLEASE do NOT remove it, and *IF* it was removed, please restore it. Thank you. and Very best regards; Bob Finch Liviu Librescu - În veci pomenirea lui. (May his memory be eternal.)
OOPSS.... Sometimes I read my email backwards (most recently received first), and it seems I replied to some email just a few minutes ago without a good review of the situation. Sorry if my comments aobut the kernel offended anyone. and Very best regards; Bob Finch Liviu Librescu - În veci pomenirea lui. (May his memory be eternal.)
On Sat, Mar 1, 2008 at 12:34 PM, Peter Feuerer <peter.feuerer@linux-gamers.net> wrote:
On Sat, 2008-03-01 at 17:13 -0500, Loui wrote:
Nonono, don't be afraid, when you've got a normal archlinux installation, everything is fine :) (the driver is in your initramfs which is loaded by your kernel at boot up)
participants (9)
-
Dan McGee
-
eliott
-
Jan de Groot
-
Loui
-
Marc Deop i Argemí
-
Peter Feuerer
-
Roman Kyrylych
-
Thomas Bächler
-
w9ya＠qrparci.net