[PATCH v3 3/5] fat (exportfs): rebuild inode if ilookup() fails

[PATCH v3 3/5] fat (exportfs): rebuild inode if ilookup() fails

[Namjae Jeon]

From: Namjae Jeon <namjae.jeon@samsung.com>

Since the previous patch in this patch-set uses i_pos as the inode
number, we can use it to find the directory entry of the inode and
subsequently rebuild the inode if the cache lookups fail.

Since this involves  accessing the FAT media,it is better to do this
only if the 'nfs' mount option is enabled with limited_stable.

Signed-off-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Ravishankar N <ravi.n1@samsung.com>
Signed-off-by: Amit Sahrawat <a.sahrawat@samsung.com>
---
 fs/fat/nfs.c |   25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)

diff --git a/fs/fat/nfs.c b/fs/fat/nfs.c
index ef4b5fa..3cf5412 100644
--- a/fs/fat/nfs.c
+++ b/fs/fat/nfs.c
@@ -52,6 +52,31 @@ static struct inode *fat_nfs_get_inode(struct super_block *sb,
 		iput(inode);
 		inode = NULL;
 	}
+	if (inode == NULL && MSDOS_SB(sb)->options.nfs == FAT_NFS_LIMITED) {
+		struct buffer_head *bh = NULL;
+		struct msdos_dir_entry *de ;
+		loff_t i_pos = (loff_t)ino;
+		int bits = MSDOS_SB(sb)->dir_per_block_bits;
+		loff_t blocknr = i_pos >> bits;
+		bh = sb_bread(sb, blocknr);
+		if (!bh) {
+			fat_msg(sb, KERN_ERR,
+				 "unable to read block(%llu) for building NFS inode",
+				(llu)blocknr);
+			return inode;
+		}
+		de = (struct msdos_dir_entry *)bh->b_data;
+		/* If a file is deleted on server and client is not updated
+		 * yet, we must not build the inode upon a lookup call.
+		 */
+		if (de[i_pos&((1<<bits)-1)].name[0] == DELETED_FLAG)
+
+			inode = NULL;
+		else
+			inode = fat_build_inode(sb, &de[i_pos&((1<<bits)-1)],
+					i_pos);
+		brelse(bh);
+	}
 
 	return inode;
 }
-- 
1.7.9.5

Re: [PATCH v3 3/5] fat (exportfs): rebuild inode if ilookup() fails

[OGAWA Hirofumi]

Namjae Jeon <linkinjeon@gmail.com> writes:

> +	if (inode == NULL && MSDOS_SB(sb)->options.nfs == FAT_NFS_LIMITED) {
> +		struct buffer_head *bh = NULL;
> +		struct msdos_dir_entry *de ;
> +		loff_t i_pos = (loff_t)ino;
> +		int bits = MSDOS_SB(sb)->dir_per_block_bits;
> +		loff_t blocknr = i_pos >> bits;
> +		bh = sb_bread(sb, blocknr);
> +		if (!bh) {
> +			fat_msg(sb, KERN_ERR,
> +				 "unable to read block(%llu) for building NFS inode",
> +				(llu)blocknr);
> +			return inode;
> +		}
> +		de = (struct msdos_dir_entry *)bh->b_data;
> +		/* If a file is deleted on server and client is not updated
> +		 * yet, we must not build the inode upon a lookup call.
> +		 */
> +		if (de[i_pos&((1<<bits)-1)].name[0] == DELETED_FLAG)

IS_FREE() would be better.

> +
> +			inode = NULL;
> +		else
> +			inode = fat_build_inode(sb, &de[i_pos&((1<<bits)-1)],
> +					i_pos);
> +		brelse(bh);
> +	}
>  
>  	return inode;
>  }

1 << bits == sbi->dir_per_block

It would be better to add helper to decode i_pos to blocknr and offset.
And it should be shared on fat_write_inode and this.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: [PATCH v3 3/5] fat (exportfs): rebuild inode if ilookup() fails

[Namjae Jeon]

2012/9/22, OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>:
> Namjae Jeon <linkinjeon@gmail.com> writes:
>
>> +	if (inode == NULL && MSDOS_SB(sb)->options.nfs == FAT_NFS_LIMITED) {
>> +		struct buffer_head *bh = NULL;
>> +		struct msdos_dir_entry *de ;
>> +		loff_t i_pos = (loff_t)ino;
>> +		int bits = MSDOS_SB(sb)->dir_per_block_bits;
>> +		loff_t blocknr = i_pos >> bits;
>> +		bh = sb_bread(sb, blocknr);
>> +		if (!bh) {
>> +			fat_msg(sb, KERN_ERR,
>> +				 "unable to read block(%llu) for building NFS inode",
>> +				(llu)blocknr);
>> +			return inode;
>> +		}
>> +		de = (struct msdos_dir_entry *)bh->b_data;
>> +		/* If a file is deleted on server and client is not updated
>> +		 * yet, we must not build the inode upon a lookup call.
>> +		 */
>> +		if (de[i_pos&((1<<bits)-1)].name[0] == DELETED_FLAG)
>
> IS_FREE() would be better.
Okay~ I will change.
>
>> +
>> +			inode = NULL;
>> +		else
>> +			inode = fat_build_inode(sb, &de[i_pos&((1<<bits)-1)],
>> +					i_pos);
>> +		brelse(bh);
>> +	}
>>
>>  	return inode;
>>  }
>
> 1 << bits == sbi->dir_per_block
>
> It would be better to add helper to decode i_pos to blocknr and offset.
> And it should be shared on fat_write_inode and this.
Okay, I will change.

Thanks.
> --
> OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
>