yaffs update#2

yaffs update#2

[Nick Bane]

[-- Attachment #1: Type: text/plain, Size: 762 bytes --]

Hi folks,

Some mods that get yaffs running properly on 2.6.8.1. The superblock name issue is now resolved.

For kernel integration examples in 2.4.25 and 2.6.8.1-tcl1, see trees at http://husaberg.toby-churchill.com/balloon/. I am also starting a cvs repository for 2.6 with viewcvs access not unlike the hh site.

There are moves afoot to submit yaffs to mainstream linux. All opinions about suitability/maturity welcome. The kernel mtd layer and yaffs_mtdif have now harmonised so now seems a good moment.

Non-yaffs: Current balloon board is supported in 2.6.8.1-tcl1 as is Mainstone (pxa270) though 2.6.9-rcx may superceed it.

Nick
cc mtd list, balloonboard list
-----------------------
Nick Bane
nick@cecomputing.co.uk
+44 (0)1954 719270 

[-- Attachment #2: diff --]
[-- Type: application/octet-stream, Size: 2085 bytes --]

diff -urN yaffs-clean/yaffs_fs.c yaffs-2.6.8/yaffs_fs.c
--- yaffs-clean/yaffs_fs.c	Thu Nov 20 08:54:05 2003
+++ yaffs-2.6.8/yaffs_fs.c	Thu Sep  2 18:37:17 2004
@@ -52,7 +52,8 @@
 #include <asm/statfs.h>
 #define UnlockPage(p) unlock_page(p)
 #define Page_Uptodate(page)	test_bit(PG_uptodate, &(page)->flags)
-#define kdevname(x) cdevname(to_kdev_t(x))
+//#define kdevname(x) cdevname(to_kdev_t(x))
+//#define sbname(sb) sb->s_id	// temporary fix
 #else
 #include <linux/locks.h>
 #endif
@@ -582,7 +583,7 @@
 		inode->i_gid = obj->st_gid;
 		inode->i_blksize = inode->i_sb->s_blocksize;
 #if defined(CONFIG_KERNEL_2_5)
-		inode->i_rdev = to_kdev_t(obj->st_rdev);
+		inode->i_rdev = old_decode_dev(obj->st_rdev);
 		inode->i_atime.tv_sec = (time_t)(obj->st_atime);
 		inode->i_atime.tv_nsec = 0;
 		inode->i_mtime.tv_sec = (time_t)obj->st_mtime;
@@ -1241,10 +1242,14 @@
 		printk(KERN_INFO"yaffs: sb is NULL\n");
 	else if(!sb->s_dev)
 		printk(KERN_INFO"yaffs: sb->s_dev is NULL\n");
+#if defined(CONFIG_KERNEL_2_5)
+		printk(KERN_INFO"yaffs: dev is %d name is \"%s\"\n", sb->s_dev, sb->s_id);
+#else
 	else if(! kdevname(sb->s_dev))
 		printk(KERN_INFO"yaffs: kdevname is NULL\n");
 	else
 		printk(KERN_INFO"yaffs: dev is %d name is \"%s\"\n", sb->s_dev, kdevname(sb->s_dev));
+#endif
 
 	
 
@@ -1303,7 +1308,11 @@
 		struct mtd_info *mtd;
 		
 		printk(KERN_DEBUG "yaffs: Attempting MTD mount on %u.%u, \"%s\"\n",
+#if defined(CONFIG_KERNEL_2_5)
+		 MAJOR(sb->s_dev),MINOR(sb->s_dev),sb->s_id);
+#else
 		 MAJOR(sb->s_dev),MINOR(sb->s_dev),kdevname(sb->s_dev));
+#endif
 			
 		// Check it's an mtd device.....
 		if(MAJOR(sb->s_dev) != MTD_BLOCK_MAJOR)
diff -urN yaffs-clean/yaffs_mtdif.c yaffs-2.6.8/yaffs_mtdif.c
--- yaffs-clean/yaffs_mtdif.c	Sun Jul 20 15:03:30 2003
+++ yaffs-2.6.8/yaffs_mtdif.c	Thu Sep  2 13:15:34 2004
@@ -31,6 +31,8 @@
 
 struct nand_oobinfo yaffs_oobinfo = {
 	useecc: 1,
+// this is for versions of mtd nand driver in kernel 2.6.8 and later
+	eccbytes: 6,
 	eccpos: {8, 9, 10, 13, 14, 15}
 };
 

Re: yaffs update#2

[Thomas Gleixner]

On Thu, 2004-09-02 at 19:56, Nick Bane wrote:
> Hi folks,
> There are moves afoot to submit yaffs to mainstream linux. 
> All opinions about suitability/maturity welcome. The kernel 
> mtd layer and yaffs_mtdif have now harmonised so now seems 
> a good moment.

Can we look into use the automatic placement of OOB data. Then it would
be using the same mechanisms as JFFS2 and it would be easy to use YAFFS
on DOC and other devices which cannot use the fixed OOB layout scheme ?

tglx