From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
To: Namjae Jeon <linkinjeon@gmail.com>
Cc: akpm@linux-foundation.org, bfields@fieldses.org, viro@zeniv.linux.org.uk,
        linux-kernel@vger.kernel.org, Namjae Jeon <namjae.jeon@samsung.com>,
        Ravishankar N <ravi.n1@samsung.com>,
        Amit Sahrawat <a.sahrawat@samsung.com>
Subject: Re: [PATCH v4 2/4] fat (exportfs): rebuild inode if ilookup() fails
References: <1349598747-3218-1-git-send-email-linkinjeon@gmail.com>
Date: Sat, 13 Oct 2012 18:01:04 +0900
In-Reply-To: <1349598747-3218-1-git-send-email-linkinjeon@gmail.com> (Namjae
	Jeon's message of "Sun, 7 Oct 2012 04:32:27 -0400")
Message-ID: <87bog6925r.fsf@devron.myhome.or.jp>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/24.2.50 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 4249
Lines: 162

Namjae Jeon <linkinjeon@gmail.com> writes:

> --- a/fs/fat/fat.h
> +++ b/fs/fat/fat.h
> @@ -214,6 +214,13 @@ static inline sector_t fat_clus_to_blknr(struct msdos_sb_info *sbi, int clus)
>  		+ sbi->data_start;
>  }
>  
> +static inline void fat_get_blknr_offset(struct msdos_sb_info *sbi,
> +				loff_t i_pos, sector_t *blknr, int *offset)
> +{
> +	*blknr = i_pos >> sbi->dir_per_block_bits;
> +	*offset = i_pos & (sbi->dir_per_block - 1);
> +}
> +

Let's separate fat_get_blknr_offset() cleanup and others.

> +extern loff_t fat_i_pos_read(struct msdos_sb_info *sbi, struct inode *inode);

[...]

> -static inline loff_t fat_i_pos_read(struct msdos_sb_info *sbi,
> +loff_t fat_i_pos_read(struct msdos_sb_info *sbi,
>  				    struct inode *inode)
>  {
>  	loff_t i_pos;

Let's move fat_i_pos_read() to fat.h to make consists with
fat_get_blknr_offset().

>  static const struct export_operations fat_export_ops = {
> +	.encode_fh      = fat_encode_fh,
>  	.fh_to_dentry	= fat_fh_to_dentry,
>  	.fh_to_parent	= fat_fh_to_parent,
>  	.get_parent	= fat_get_parent,
> diff --git a/fs/fat/nfs.c b/fs/fat/nfs.c
> index ef4b5fa..156903b 100644
> --- a/fs/fat/nfs.c
> +++ b/fs/fat/nfs.c
> @@ -14,6 +14,14 @@
>  #include <linux/exportfs.h>
>  #include "fat.h"
>  
> +struct fat_fid {
> +	u32 ino;
> +	u32 gen;
> +	u64 i_pos;
> +	u32 parent_ino;
> +	u32 parent_gen;
> +} __packed;

This is sizeof(fat_fid)/sizoef(u32) == 6. IIRC, nfsv2 is not supporting
FH > 6, true?

> +int
> +fat_encode_fh(struct inode *inode, __u32 *fh, int *lenp, struct inode *parent)
> +{
> +	int len = *lenp;
> +	struct msdos_sb_info *sbi = MSDOS_SB(inode->i_sb);
> +	struct fat_fid *fid = (struct fat_fid *) fh;
> +	loff_t i_pos;
> +	int type = FILEID_INO32_GEN;
> +
> +	if (parent && (len < 5)) {
> +		*lenp = 5;
> +		return 255;
> +	} else if (len < 3) {
> +		*lenp = 3;
> +		return 255;
> +	}
> +
> +	i_pos = fat_i_pos_read(sbi, inode);
> +	*lenp = 3;
> +	fid->ino = inode->i_ino;
> +	fid->gen = inode->i_generation;
> +	fid->i_pos = i_pos;
> +	if (parent) {
> +		fid->parent_ino = parent->i_ino;
> +		fid->parent_gen = parent->i_generation;
> +		type = FILEID_INO32_GEN_PARENT;
> +		*lenp = 5;
> +	}
> +
> +	return type;
> +}

I was also thinking to use same FH though, because of limitation of FH
size. It looks like to be better to separate with "stale_rw".

So, how about to separate at export_operations level?

I.e. (move export_operations to fat/nfs.c)

For stale_rw,

const struct export_operations fat_export_ops = {
	.fh_to_dentry	= fat_fh_to_dentry,
	.fh_to_parent	= fat_fh_to_parent,
	.get_parent	= fat_get_parent,
}

For nostale_ro,

const struct export_operations fat_export_ops = {
	.fh_to_dentry	= fat_encode_fh,
	.fh_to_dentry	= fat_fh_to_dentry,
	.fh_to_parent	= fat_fh_to_parent,
	.get_parent	= fat_get_parent,
}

And we have to think about FH format. I guess we don't need to
inode->i_ino for nostale_ro.

Maximum i_pos is 40bits, and i_gen is 32bit. So, inode and parent inode
fits to FH of 5 len, we have to pack those though.

I.e.

struct fat_fid {
	u32 i_gen;
	u32 i_pos_low;
	u16 i_pos_hi;
	u16 parent_i_pos_hi;
	u32 parent_i_pos_low;
	u32 parent_i_gen;
} __packed;

And define FILEID_FAT_WITHOUT_PARENT and FILEID_FAT_WITH_PARENT.

User of i_ino is only ilookup, right? So, we can add fat_ilookup() for it.

struct inode *fat_ilookup(struct super_block *sb, u64 ino)
{
	if (stale_rw)
        	return ilookup(sb, ino);

	return fat_iget(sb, i_pos);
}

And I noticed we need lock for fat_build_inode() for nostale_ro. Because
fat_nfs_get_inode() doesn't hold i_mutex of parent dir, right? So, add
lock to fat_build_inode()

static inline void fat_build_inode_lock(struct msdos_sb_info *sbi)
{
	if (sbi->options.nfs == FAT_NFS_NOSTALE_RO)
        	mutex_lock(&sbi->nfs_build_inode_lock);
}

static inline void fat_build_inode_unlock(struct msdos_sb_info *sbi)
{
	if (sbi->options.nfs == FAT_NFS_NOSTALE_RO)
        	mutex_unlock(&sbi->nfs_build_inode_lock);
}
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/