[PATCH] ufs: make solaris fsck happy

[PATCH] ufs: make solaris fsck happy

[Evgeniy Dushistov]

Alex Viskovatoff let me know that after copying data to solaris's ufs from
linux, solaris's fsck sees some errors in cylinder summary information.
This is because of solaris expects find some data on another places,
then curernt implementation save it.
This patch fixes this issue. It is tested by me, and also Alex
reported that it works for him.


Signed-off-by: Evgeniy Dushistov <dushistov@mail.ru>
Cc: Alex Viskovatoff <viskovatoff@imap.cc>
---
 fs/ufs/ufs_fs.h |    9 +++++++--
 1 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/fs/ufs/ufs_fs.h b/fs/ufs/ufs_fs.h
index 54bde18..f221c5e 100644
--- a/fs/ufs/ufs_fs.h
+++ b/fs/ufs/ufs_fs.h
@@ -227,11 +227,16 @@ typedef __u16 __bitwise __fs16;
  */
 #define ufs_cbtocylno(bno) \
 	((bno) * uspi->s_nspf / uspi->s_spc)
-#define ufs_cbtorpos(bno) \
+#define ufs_cbtorpos(bno)				      \
+	((UFS_SB(sb)->s_flags & UFS_CG_SUN) ?		      \
+	 (((((bno) * uspi->s_nspf % uspi->s_spc) %	      \
+	    uspi->s_nsect) *				      \
+	   uspi->s_nrpos) / uspi->s_nsect)		      \
+	 :						      \
 	((((bno) * uspi->s_nspf % uspi->s_spc / uspi->s_nsect \
 	* uspi->s_trackskew + (bno) * uspi->s_nspf % uspi->s_spc \
 	% uspi->s_nsect * uspi->s_interleave) % uspi->s_nsect \
-	* uspi->s_nrpos) / uspi->s_npsect)
+	  * uspi->s_nrpos) / uspi->s_npsect))
 
 /*
  * The following macros optimize certain frequently calculated
-- 
1.6.5.6

-- 
/Evgeniy

Re: [PATCH] ufs: make solaris fsck happy

[Andrew Morton]

On Thu, 25 Feb 2010 21:47:43 +0300
Evgeniy Dushistov <dushistov@mail.ru> wrote:

> --- a/fs/ufs/ufs_fs.h
> +++ b/fs/ufs/ufs_fs.h
> @@ -227,11 +227,16 @@ typedef __u16 __bitwise __fs16;
>   */
>  #define ufs_cbtocylno(bno) \
>  	((bno) * uspi->s_nspf / uspi->s_spc)
> -#define ufs_cbtorpos(bno) \
> +#define ufs_cbtorpos(bno)				      \
> +	((UFS_SB(sb)->s_flags & UFS_CG_SUN) ?		      \
> +	 (((((bno) * uspi->s_nspf % uspi->s_spc) %	      \
> +	    uspi->s_nsect) *				      \
> +	   uspi->s_nrpos) / uspi->s_nsect)		      \
> +	 :						      \
>  	((((bno) * uspi->s_nspf % uspi->s_spc / uspi->s_nsect \
>  	* uspi->s_trackskew + (bno) * uspi->s_nspf % uspi->s_spc \
>  	% uspi->s_nsect * uspi->s_interleave) % uspi->s_nsect \
> -	* uspi->s_nrpos) / uspi->s_npsect)
> +	  * uspi->s_nrpos) / uspi->s_npsect))

yikes, that macro should be killed with a stick before it becomes
self-aware and starts breeding.

Or converted into a C function which humans have a chance of
understanding and maintaining.