2003-01-12 05:51:46

by Robert P. J. Day

[permalink] [raw]
Subject: some curiosities on the filesystems layout in kernel config


hoping i don't embarrass myself again ...

1) where is the USMDOS selection that's listed in the Kconfig file?
it doesn't appear in the menu

2) shouldn't ext3 depend on ext2?

3) currently, since quotas are only supported for ext2, ext3 and
reiserfs, shouldn't quotas depend on at least one of those
being selected?

rday


2003-01-12 06:40:52

by Valdis Klētnieks

[permalink] [raw]
Subject: Re: some curiosities on the filesystems layout in kernel config

On Sun, 12 Jan 2003 01:00:40 EST, "Robert P. J. Day" <[email protected]> said:

> 2) shouldn't ext3 depend on ext2?

No, because somebody might want ext3 only, and have no intention or
desire to mount a filesystem in ext2 mode. Everything on this laptop
is ext3...

> 3) currently, since quotas are only supported for ext2, ext3 and
> reiserfs, shouldn't quotas depend on at least one of those
> being selected?

Because if we did that, we'd be setting ourselves up for a mess when
fs/xfs/xfs_qm.c eventually shows up - like it already has ;)

Also, from my (possibly incorrect) reading of kernel/sys.c and
fs/quota.c, there won't be a sys_quotactl() in the kernel. As a
result, if you have users who have 'quota -v' in their .login, things
might get interesting. So you might want a config where the quota
system call is there, even if it doesn't do anything incredibly
useful...

--
Valdis Kletnieks
Computer Systems Senior Engineer
Virginia Tech


Attachments:
(No filename) (226.00 B)

2003-01-12 06:50:47

by Joshua Kwan

[permalink] [raw]
Subject: Re: some curiosities on the filesystems layout in kernel config

On Sun, Jan 12, 2003 at 01:00:40AM -0500, Robert P. J. Day wrote:

> 1) where is the USMDOS selection that's listed in the Kconfig file?
> it doesn't appear in the menu

I think you have to select DOS filesystem first and MSDOS fs support,
then it shows up.

>
> 2) shouldn't ext3 depend on ext2?
>

It doesn't currently, but it should.
Patch to 2.4.20 vanilla is attached, it should apply with some fuzz to
patched trees such as -ac and -ck. (I also renamed Second extended fs to
Ext2, a pet peeve of mine..)

> 3) currently, since quotas are only supported for ext2, ext3 and
> reiserfs, shouldn't quotas depend on at least one of those
> being selected?

Not sure whether this is true, I'm not quite sure.
And plus I don't know enough about Config.ins to write that into the
patch (Is it even possible in 2.4's Kconfig?)

Regards
Josh


Attachments:
(No filename) (0.00 B)
(No filename) (189.00 B)
Download all attachments

2003-01-12 06:52:51

by Joshua Kwan

[permalink] [raw]
Subject: Re: some curiosities on the filesystems layout in kernel config

Never mind my patch in this case. I had just hit 'y' to send the email
when I read this! Obviously, I don't know quite enough about how the
kernel works.. I really did think ext3 depended on ext2, since ext3
was simply ext2 + a journal inode.

Oh well. Sorry for the wasted b/w

Regards
Josh

On Sun, Jan 12, 2003 at 01:49:29AM -0500, [email protected]
wrote:
> On Sun, 12 Jan 2003 01:00:40 EST, "Robert P. J. Day" <[email protected]> said:
>
> > 2) shouldn't ext3 depend on ext2?
>
> No, because somebody might want ext3 only, and have no intention or
> desire to mount a filesystem in ext2 mode. Everything on this laptop
> is ext3...
>
> > 3) currently, since quotas are only supported for ext2, ext3 and
> > reiserfs, shouldn't quotas depend on at least one of those
> > being selected?
>
> Because if we did that, we'd be setting ourselves up for a mess when
> fs/xfs/xfs_qm.c eventually shows up - like it already has ;)
>
> Also, from my (possibly incorrect) reading of kernel/sys.c and
> fs/quota.c, there won't be a sys_quotactl() in the kernel. As a
> result, if you have users who have 'quota -v' in their .login, things
> might get interesting. So you might want a config where the quota
> system call is there, even if it doesn't do anything incredibly
> useful...
>
> --
> Valdis Kletnieks
> Computer Systems Senior Engineer
> Virginia Tech
>



Attachments:
(No filename) (1.37 kB)
(No filename) (189.00 B)
Download all attachments

2003-01-12 06:55:52

by Aaron Lehmann

[permalink] [raw]
Subject: Re: some curiosities on the filesystems layout in kernel config

On Sat, Jan 11, 2003 at 10:59:21PM -0800, Joshua M. Kwan wrote:
> > 3) currently, since quotas are only supported for ext2, ext3 and
> > reiserfs, shouldn't quotas depend on at least one of those
> > being selected?
>
> Not sure whether this is true, I'm not quite sure.

You could compile any of these filesystems as modules.

2003-01-13 09:22:11

by Werner Almesberger

[permalink] [raw]
Subject: Re: some curiosities on the filesystems layout in kernel config

Robert P. J. Day wrote:
> 3) currently, since quotas are only supported for ext2, ext3 and
> reiserfs, shouldn't quotas depend on at least one of those
> being selected?

The problem with expressing every last dependency is that you'll
end up hiding too much.

A less intrusive approach that doesn't require changes to the
current configuration framework may be to add a section
"Warnings" at the end, under which some pseudo-options would be
enabled if some unusual combinations are found (e.g. CD-ROM
drivers but no ISO9660 file system, PC architecture but no
keyboard, etc.)

- Werner

--
_________________________________________________________________________
/ Werner Almesberger, Buenos Aires, Argentina [email protected] /
/_http://www.almesberger.net/____________________________________________/

2003-01-13 12:29:55

by Dave Jones

[permalink] [raw]
Subject: Re: some curiosities on the filesystems layout in kernel config

On Sun, Jan 12, 2003 at 01:00:40AM -0500, Robert P. J. Day wrote:

> 1) where is the USMDOS selection that's listed in the Kconfig file?
> it doesn't appear in the menu

currently broken. this needs fixing at some point for 2.6

> 2) shouldn't ext3 depend on ext2?

nope.

> 3) currently, since quotas are only supported for ext2, ext3 and
> reiserfs, shouldn't quotas depend on at least one of those
> being selected?

Could do, but it's the Kconfig gets messy very quickly when you
have realise that any of those fs's can also be modular.

Dave

--
| Dave Jones. http://www.codemonkey.org.uk
| SuSE Labs

2003-02-07 15:13:57

by Jan Kara

[permalink] [raw]
Subject: Re: some curiosities on the filesystems layout in kernel config

> On Sun, 12 Jan 2003 01:00:40 EST, "Robert P. J. Day" <[email protected]> said:
>
> > 3) currently, since quotas are only supported for ext2, ext3 and
> > reiserfs, shouldn't quotas depend on at least one of those
> > being selected?
Quotas work also for other filesystems...

> Because if we did that, we'd be setting ourselves up for a mess when
> fs/xfs/xfs_qm.c eventually shows up - like it already has ;)
>
> Also, from my (possibly incorrect) reading of kernel/sys.c and
> fs/quota.c, there won't be a sys_quotactl() in the kernel. As a
> result, if you have users who have 'quota -v' in their .login, things
> might get interesting. So you might want a config where the quota
> system call is there, even if it doesn't do anything incredibly
> useful...
You're right that it won't be in the kernel but in that case 'quota
-v' will just say 'Disk quotas for user test (uid 1000): none' (in case
you haven't any filesystem mounted with usrquota option which is
reasonable if you haven't quotas in kernel).

Honza
--
Jan Kara <[email protected]>
SuSE CR Labs