On Sun, 10 Apr 2011 14:05:54 +0400
Pavel Shilovsky <piastry@xxxxxxxxxxx> wrote:
> Add cifs_match_super to use in sget to share superblock between mounts
> that have the same //server/sharename, credentials and mount options.
> It helps us to improve performance on work with future SMB2.1 leases.
>
> Signed-off-by: Pavel Shilovsky <piastry@xxxxxxxxxxx>
> ---
> fs/cifs/cifsfs.c | 6 ++-
> fs/cifs/connect.c | 93 +++++++++++++++++++++++++++++++++++++++++++++++++++++
> 2 files changed, 97 insertions(+), 2 deletions(-)
>
> diff --git a/fs/cifs/cifsfs.c b/fs/cifs/cifsfs.c
> index c40c1f4..abc8436 100644
> --- a/fs/cifs/cifsfs.c
> +++ b/fs/cifs/cifsfs.c
> @@ -577,14 +577,16 @@ cifs_do_mount(struct file_system_type *fs_type,
> if (rc)
> return ERR_PTR(rc);
>
> - sb = sget(fs_type, NULL, set_anon_super, NULL);
> + sb = sget(fs_type, cifs_match_super, set_anon_super, volume_info);
> if (IS_ERR(sb)) {
> cifs_cleanup_volume_info(&volume_info);
> return ERR_CAST(sb);
> }
>
> - if (sb->s_fs_info)
> + if (sb->s_fs_info) {
> + cFYI(1, "Use existing superblock");
> goto out;
> + }
>
> sb->s_flags = flags;
>
> diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c
> index 53b79e7..7c98c36 100644
> --- a/fs/cifs/connect.c
> +++ b/fs/cifs/connect.c
> @@ -2193,6 +2193,99 @@ cifs_put_tlink(struct tcon_link *tlink)
> return;
> }
>
> +static void setup_cifs_sb(struct smb_vol *vol, struct cifs_sb_info *sb);
> +static inline struct tcon_link *
> +cifs_sb_master_tlink(struct cifs_sb_info *cifs_sb);
> +
> +static const u32 ACTUAL_MOUNT_FLAGS = CIFS_MOUNT_NO_PERM | CIFS_MOUNT_SET_UID |
> + CIFS_MOUNT_SERVER_INUM | CIFS_MOUNT_DIRECT_IO | CIFS_MOUNT_NO_XATTR |
> + CIFS_MOUNT_MAP_SPECIAL_CHR | CIFS_MOUNT_UNX_EMUL | CIFS_MOUNT_NO_BRL |
> + CIFS_MOUNT_CIFS_ACL | CIFS_MOUNT_OVERR_UID | CIFS_MOUNT_OVERR_GID |
> + CIFS_MOUNT_DYNPERM | CIFS_MOUNT_NOPOSIXBRL | CIFS_MOUNT_NOSSYNC |
> + CIFS_MOUNT_FSCACHE | CIFS_MOUNT_MF_SYMLINKS | CIFS_MOUNT_MULTIUSER |
> + CIFS_MOUNT_STRICT_IO;
> +
> +static int
> +compare_mount_options(struct super_block *sb, struct smb_vol *vol)
> +{
> + struct cifs_sb_info *old = CIFS_SB(sb), new;
> +
> + setup_cifs_sb(vol, &new);
> +
> + if ((old->mnt_cifs_flags & ACTUAL_MOUNT_FLAGS) !=
> + (new.mnt_cifs_flags & ACTUAL_MOUNT_FLAGS))
> + return 0;
> +
> + if (old->rsize != new.rsize || old->wsize != new.wsize)
> + return 0;
> +
> + if (old->mnt_uid != new.mnt_uid || old->mnt_gid != new.mnt_gid)
> + return 0;
> +
> + if (old->mnt_file_mode != new.mnt_file_mode ||
> + old->mnt_dir_mode != new.mnt_dir_mode)
> + return 0;
> +
> + if (old->prepathlen != new.prepathlen ||
> + strncmp(old->prepath, new.prepath, old->prepathlen))
> + return 0;
> +
I don't think this is really what you want...
Suppose I mount the same share twice, but with 2 different prepaths.
For instance:
//server/share/foo
...and...
//server/share/foo/bar
...shouldn't the inodes for files/directories accessable via both
mounts be shared in that case?
--
Jeff Layton <jlayton@xxxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe linux-cifs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
