linux-fsdevel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Bharata B Rao <bharata@linux.vnet.ibm.com>
To: Jan Blunck <jblunck@suse.de>
Cc: linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [RFC 20/26] union-mount: Simple union-mount readdir implementation
Date: Mon, 6 Aug 2007 16:38:18 +0530	[thread overview]
Message-ID: <20070806110818.GC7489@in.ibm.com> (raw)
In-Reply-To: <20070730161325.004086646@weierstrass.suse.de>

On Mon, Jul 30, 2007 at 06:13:43PM +0200, Jan Blunck wrote:
> --- a/include/linux/union.h
> +++ b/include/linux/union.h
> @@ -53,6 +53,7 @@ extern void __shrink_d_unions(struct den
>  extern int attach_mnt_union(struct vfsmount *, struct vfsmount *,
>  			    struct dentry *);
>  extern void detach_mnt_union(struct vfsmount *);
> +extern int readdir_union(struct file *, void *, filldir_t);
> 
>  #else /* CONFIG_UNION_MOUNT */
> 
> @@ -69,5 +70,29 @@ extern void detach_mnt_union(struct vfsm
>  #define detach_mnt_union(x)		do { } while (0)
> 
>  #endif	/* CONFIG_UNION_MOUNT */
> +
> +static inline int do_readdir(struct file *file, void *buf, filldir_t filler)
> +{
> +	struct inode *inode = file->f_path.dentry->d_inode;
> +	int res;
> +
> +#ifdef CONFIG_UNION_MOUNT
> +	if (IS_MNT_UNION(file->f_path.mnt))
> +		res = readdir_union(file, buf, filler);
> +	else
> +#endif
> +	{
> +		mutex_lock(&inode->i_mutex);
> +		res = -ENOENT;
> +		if (!IS_DEADDIR(inode)) {
> +			res = file->f_op->readdir(file, buf, filler);
> +			file_accessed(file);
> +		}
> +		mutex_unlock(&inode->i_mutex);
> +	}
> +
> +	return res;
> +}

Here you are doing readdir_union for all the directories under a union
mount point, which is an overhead (building the readdir cache). Here is
the fix:

From: Bharata B Rao <bharata@linux.vnet.ibm.com>

Within a union mount point, there can be directories which don't have a
union stack underneath them. And readdir() doesn't have to maintain a cache
of dirents for such directories. But the current patch maintains the cache
for such directories. Fix this.

Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
---
 fs/union.c            |   17 +++++++++++++++++
 include/linux/union.h |   11 ++++++-----
 2 files changed, 23 insertions(+), 5 deletions(-)

--- a/fs/union.c
+++ b/fs/union.c
@@ -366,6 +366,23 @@ int follow_union_mount(struct vfsmount *
 }
 
 /*
+ * is_dir_unioned - check if the directory represented by @mnt and @dentry
+ * has a union stack underneath.
+ *
+ * Returns true if a union stack exists at this directory level, else returns
+ * false.
+ */
+int is_dir_unioned(struct vfsmount *mnt, struct dentry *dentry)
+{
+	struct union_mount *um;
+
+	spin_lock(&union_lock);
+	um = union_lookup(dentry, mnt);
+	spin_unlock(&union_lock);
+	return um ? 1: 0;
+}
+
+/*
  * This must be called when unhashing a dentry. This is called with dcache_lock
  * and unhashes all unions this dentry is in.
  */
--- a/include/linux/union.h
+++ b/include/linux/union.h
@@ -54,6 +54,7 @@ extern int attach_mnt_union(struct vfsmo
 			    struct dentry *);
 extern void detach_mnt_union(struct vfsmount *);
 extern int readdir_union(struct file *, void *, filldir_t);
+extern int is_dir_unioned(struct vfsmount *, struct dentry *);
 extern int union_relookup_topmost(struct nameidata *, int);
 extern struct dentry *union_create_topmost(struct nameidata *, struct qstr *,
 					   struct path *);
@@ -82,13 +83,14 @@ extern int union_copyup(struct nameidata
 
 static inline int do_readdir(struct file *file, void *buf, filldir_t filler)
 {
-	struct inode *inode = file->f_path.dentry->d_inode;
 	int res;
+	struct inode *inode = file->f_path.dentry->d_inode;
 
 #ifdef CONFIG_UNION_MOUNT
-	if (IS_MNT_UNION(file->f_path.mnt))
-		res = readdir_union(file, buf, filler);
-	else
+	if (IS_MNT_UNION(file->f_path.mnt)) {
+		if (is_dir_unioned(file->f_path.mnt, file->f_path.dentry))
+			res = readdir_union(file, buf, filler);
+	} else
 #endif
 	{
 		mutex_lock(&inode->i_mutex);
@@ -99,7 +101,6 @@ static inline int do_readdir(struct file
 		}
 		mutex_unlock(&inode->i_mutex);
 	}
-
 	return res;
 }
 
Regards,
Bharata.

  reply	other threads:[~2007-08-06 11:08 UTC|newest]

Thread overview: 64+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-07-30 16:13 [RFC 00/26] VFS based Union Mount (V2) Jan Blunck
2007-07-30 16:13 ` [RFC 01/26] [PATCH 14/18] shmem: convert to using splice instead of sendfile() Jan Blunck
2007-07-30 16:13 ` [RFC 02/26] VFS: Export dput_path() and path_to_nameidata() Jan Blunck
2007-07-30 16:13 ` [RFC 03/26] VFS: Make lookup_hash() return a struct path Jan Blunck
2007-07-30 16:13 ` [RFC 04/26] VFS: Make lookup_create() " Jan Blunck
2007-07-30 16:13 ` [RFC 05/26] VFS: cache_lookup() cleanup Jan Blunck
2007-07-30 16:13 ` [RFC 06/26] VFS: Make real_lookup() return a struct path Jan Blunck
2007-07-30 16:13 ` [RFC 07/26] VFS: Introduce dput() variante that maintains a kill-list Jan Blunck
2007-07-30 16:13 ` [RFC 08/26] VFS: Export lives_below_in_same_fs() Jan Blunck
2007-07-30 16:13 ` [RFC 09/26] linux/stat.h: Add the filetype white-out Jan Blunck
2007-07-30 16:13 ` [RFC 10/26] VFS white-out handling Jan Blunck
2007-07-30 16:13 ` [RFC 11/26] tmpfs white-out support Jan Blunck
2007-08-01 15:13   ` Hugh Dickins
2007-08-02  2:48     ` Matt Mackall
2007-07-30 16:13 ` [RFC 12/26] ext2 " Jan Blunck
2007-07-31  3:45   ` Theodore Tso
2007-07-31  7:44     ` Jan Blunck
2007-07-31  8:32       ` Andreas Dilger
2007-07-31  9:08         ` Jan Blunck
2007-07-31 10:53       ` Theodore Tso
2007-08-02 19:31         ` Pavel Machek
2007-07-31 16:36   ` Josef Sipek
2007-07-31 17:00     ` Jan Blunck
2007-07-31 17:11       ` Josef Sipek
2007-08-01 15:23         ` Dave Kleikamp
2007-08-01 18:44           ` Josef Sipek
2007-08-01 19:10             ` Dave Kleikamp
2007-08-01 19:33               ` Josef Sipek
2007-08-01 19:52                 ` Dave Kleikamp
2007-08-01 22:06                   ` Erez Zadok
2007-08-02 12:05                     ` Jan Blunck
2007-08-02 11:55                 ` Jan Blunck
2007-08-02 17:50                 ` Jörn Engel
2007-08-02  5:24             ` Ph. Marek
2007-08-02 12:12               ` Jan Blunck
2007-08-02 10:26         ` Jan Blunck
2007-08-01 10:00       ` Hans-Peter Jansen
2007-08-01 11:43         ` Josef Sipek
2007-08-01 18:01         ` Jan Engelhardt
2007-07-31 17:03     ` Mark Williamson
2007-07-31 17:16       ` Josef Sipek
2007-08-01 17:58     ` Jan Engelhardt
2007-08-01 18:03       ` Josef Sipek
2007-07-30 16:13 ` [RFC 13/26] ext3 whiteout support Jan Blunck
2007-07-30 16:13 ` [RFC 14/26] union-mount: Documentation Jan Blunck
2007-07-30 16:13 ` [RFC 15/26] union-mount: Add union-mount mount flag Jan Blunck
2007-07-30 16:13 ` [RFC 16/26] union-mount: Introduce union_mount structure Jan Blunck
2007-08-06  5:57   ` Bharata B Rao
2007-07-30 16:13 ` [RFC 17/26] union-mount: Drive the union cache via dcache Jan Blunck
2007-07-30 16:13 ` [RFC 18/26] union-mount: Changes to the namespace handling Jan Blunck
2007-08-08 10:10   ` Bharata B Rao
2007-07-30 16:13 ` [RFC 19/26] union-mount: Make lookup work for union-mounted file systems Jan Blunck
2007-08-09  5:42   ` Bharata B Rao
2007-07-30 16:13 ` [RFC 20/26] union-mount: Simple union-mount readdir implementation Jan Blunck
2007-08-06 11:08   ` Bharata B Rao [this message]
2007-07-30 16:13 ` [RFC 21/26] union-mount: in-kernel file copy between union mounted filesystems Jan Blunck
2007-07-30 16:13 ` [RFC 22/26] union-mount: white-out changes for copy-on-open Jan Blunck
2007-07-30 16:13 ` [RFC 23/26] union-mount: copyup on rename Jan Blunck
2007-07-30 16:13 ` [RFC 24/26] union-mount: dont report EROFS for union mounts Jan Blunck
2007-07-30 16:13 ` [RFC 25/26] union-mount: Debug Infrastructure Jan Blunck
2007-07-30 16:13 ` [RFC 26/26] union-mount: Debug code Jan Blunck
2007-07-30 18:23 ` [RFC 00/26] VFS based Union Mount (V2) Al Boldi
2007-08-02  6:49 ` Bharata B Rao
2007-08-02 10:17   ` Jan Blunck

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20070806110818.GC7489@in.ibm.com \
    --to=bharata@linux.vnet.ibm.com \
    --cc=jblunck@suse.de \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).