Re: Problem found, but why???

[Boaz Harrosh <bharrosh@panasas.com>]

115e19c53501edc11f730191f7f047736815ae3d is the first bad commit
commit 115e19c53501edc11f730191f7f047736815ae3d
Author: Boaz Harrosh <Boaz Harrosh bharrosh@panasas.com>

    exofs: Set i_mapping->backing_dev_info anyway

OK It's stupid ME. But why please someone explain this issue?
For now I'll send a revert, but I would like this explained,
perhaps a more solid fix is needed.

Please look at this patch:

diff --git a/fs/exofs/inode.c b/fs/exofs/inode.c
index 24ab327..0ba9886 100644
--- a/fs/exofs/inode.c
+++ b/fs/exofs/inode.c
@@ -1030,6 +1030,7 @@ struct inode *exofs_iget(struct super_block *sb, unsigned long ino)
 		memcpy(oi->i_data, fcb.i_data, sizeof(fcb.i_data));
 	}
 
+	inode->i_mapping->backing_dev_info = sb->s_bdi;
 	if (S_ISREG(inode->i_mode)) {
 		inode->i_op = &exofs_file_inode_operations;
 		inode->i_fop = &exofs_file_operations;
@@ -1129,6 +1130,7 @@ struct inode *exofs_new_inode(struct inode *dir, int mode)
 
 	sbi = sb->s_fs_info;
 
+	inode->i_mapping->backing_dev_info = sb->s_bdi;
 	sb->s_dirt = 1;
 	inode_init_owner(inode, dir, mode);
 	inode->i_ino = sbi->s_nextid++;


All it does is set the inode's bdi to point to the super-block bdi. In stead
of this default "null" bdi from VFS. When we do so the read-ahead stops
coming, completely. Is there something that needs to be implemented farther
for a super-block's bdi to support read-ahead? Please help?

Thanks
Boaz