Re: lfs heads up

[David Holland]

On Sun, Jun 30, 2013 at 01:47:28AM +0900, NAKAJIMA Yoshihiro wrote:
  After LFS split, ufs_readwrite.c is no longer shared by FFS and LFS.
  So I think:
  
  (1) Simplify them (see attached file)
  
  And (2a) merge into {ffs,lfs}_vnops.c
  or (2b) make them standalone like ext2fs_readwrite.c (and rename
  ufs_readwrite.c to ffs_readwrite.c).

Please commit the ufs one; I have been meaning to do that but I
haven't had time in the past week. (And yes, it should be moved into
ffs/ and named ffs_readwrite.c, but please do that as a separate
commit.)

Don't throw away the lfs part of the patch, but give me a few days to
think about it. I think we may want to hang onto some bits (like the
places the snapshot hooks go) and I'd also like to do a general pass
through lfs looking for unreachable code while it's still very clear
what came from ufs and what didn't.

I think we want to keep at least some of the ufs code for 64-bit
(LFS2) filesystems even though it's never been runnable with LFS; I
also want to hang onto all the quota code until we can reach a
decision about the on-disk quota format, even though AFAIK quotas have
never worked on LFS either.

There's a lot of other ufs-derived cruft though.

-- 
David A. Holland
dholl...@netbsd.org

Re: lfs heads up

[Christos Zoulas]

In article 20130630.014728.361007319.nakay...@leto.eonet.ne.jp,
NAKAJIMA Yoshihiro  nakay...@leto.eonet.ne.jp wrote:
-=-=-=-=-=-

Hello,


After LFS split, ufs_readwrite.c is no longer shared by FFS and LFS.
So I think:

(1) Simplify them (see attached file)

And (2a) merge into {ffs,lfs}_vnops.c
or (2b) make them standalone like ext2fs_readwrite.c (and rename
ufs_readwrite.c to ffs_readwrite.c).


LGTM, but use %s: __func__ instead of hard-coding.

christos

Re: lfs heads up

[NAKAJIMA Yoshihiro]

Hello,

I guess LFS no longer depends on FFS, so:

diff -u ./src/sys/ufs/lfs/lfs_vfsops.c.ORIG ./src/sys/ufs/lfs/lfs_vfsops.c
--- ./src/sys/ufs/lfs/lfs_vfsops.c.ORIG 2013-06-10 18:25:05.0 +0900
+++ ./src/sys/ufs/lfs/lfs_vfsops.c  2013-06-17 23:04:02.0 +0900
@@ -114,7 +114,7 @@
 #include miscfs/genfs/genfs.h
 #include miscfs/genfs/genfs_node.h
 
-MODULE(MODULE_CLASS_VFS, lfs, ffs);
+MODULE(MODULE_CLASS_VFS, lfs, NULL);
 
 static int lfs_gop_write(struct vnode *, struct vm_page **, int, int);
 static bool lfs_issequential_hole(const struct ulfsmount *,


Regards,
-- 
nakayosh

Re: lfs heads up

[NAKAJIMA Yoshihiro]

Please modify also src/sys/modules/lfs/Makefile.

Regards,
-- 
nakayosh

Re: lfs heads up

[Tetsuya Isaki]

At Thu, 6 Jun 2013 05:57:17 +,
David Holland wrote:
 I have committed the lfs/ufs split that's been hanging fire for the
 past 3.5+ years. It is not fully done in the sense that the header
 files still need substantial rearranging; plus there's a lot of
 consolidation and pruning to do (that's the point) but the bulk of
 it's in.
 
 If I broke anything, let me know.

build of sys/arch/x68k/stand/boot_ufs and sys/arch/hp700/stand/xxboot
fails.  Could you please see them?


http://releng.netbsd.org/builds/HEAD/201306121730Z/x68k.build.failed
dependall === sys/arch/x68k/stand/boot_ufs
/home/builds/ab/HEAD/src/sys/arch/x68k/stand/boot_ufs/readufs_lfs.c: In 
function 'get_lfs_inode':
/home/builds/ab/HEAD/src/sys/arch/x68k/stand/boot_ufs/readufs_lfs.c:203:9: 
error: 'struct ufs1_dinode' has no member named 'di_inumber'
 

http://releng.netbsd.org/builds/HEAD/201306121730Z/hp700.build.failed
dependall === sys/arch/hp700/stand/xxboot
/home/builds/ab/HEAD/src/sys/arch/hp700/stand/xxboot/readufs_lfs.c: In function 
'get_lfs_inode':
/home/builds/ab/HEAD/src/sys/arch/hp700/stand/xxboot/readufs_lfs.c:193:9: 
error: 'struct ufs1_dinode' has no member named 'di_inumber'

---
Tetsuya Isaki is...@pastel-flower.jp / is...@netbsd.org