2010-04-15 23:04:20

by Valerie Aurora

[permalink] [raw]
Subject: [PATCH 13/35] fallthru: ext2 fallthru support

Add support for fallthru directory entries to ext2.

XXX - Makes up inode number for fallthru entry
XXX - Might be better implemented as special symlinks

Cc: Theodore Tso <[email protected]>
Cc: [email protected]
Signed-off-by: Valerie Aurora <[email protected]>
Signed-off-by: Jan Blunck <[email protected]>
---
fs/ext2/dir.c | 92 ++++++++++++++++++++++++++++++++++++++++++++--
fs/ext2/ext2.h | 1 +
fs/ext2/namei.c | 22 +++++++++++
include/linux/ext2_fs.h | 1 +
4 files changed, 112 insertions(+), 4 deletions(-)

diff --git a/fs/ext2/dir.c b/fs/ext2/dir.c
index 030bd46..f3b4aff 100644
--- a/fs/ext2/dir.c
+++ b/fs/ext2/dir.c
@@ -219,7 +219,8 @@ static inline int ext2_match (int len, const char * const name,
{
if (len != de->name_len)
return 0;
- if (!de->inode && (de->file_type != EXT2_FT_WHT))
+ if (!de->inode && ((de->file_type != EXT2_FT_WHT) &&
+ (de->file_type != EXT2_FT_FALLTHRU)))
return 0;
return !memcmp(name, de->name, len);
}
@@ -256,6 +257,7 @@ static unsigned char ext2_filetype_table[EXT2_FT_MAX] = {
[EXT2_FT_SOCK] = DT_SOCK,
[EXT2_FT_SYMLINK] = DT_LNK,
[EXT2_FT_WHT] = DT_WHT,
+ [EXT2_FT_FALLTHRU] = DT_UNKNOWN,
};

#define S_SHIFT 12
@@ -342,6 +344,24 @@ ext2_readdir (struct file * filp, void * dirent, filldir_t filldir)
ext2_put_page(page);
return 0;
}
+ } else if (de->file_type == EXT2_FT_FALLTHRU) {
+ int over;
+ unsigned char d_type = DT_UNKNOWN;
+
+ offset = (char *)de - kaddr;
+ /* XXX We don't know the inode number
+ * of the directory entry in the
+ * underlying file system. Should
+ * look it up, either on fallthru
+ * creation at first readdir or now at
+ * filldir time. */
+ over = filldir(dirent, de->name, de->name_len,
+ (n<<PAGE_CACHE_SHIFT) | offset,
+ 123 /* Made up ino */, d_type);
+ if (over) {
+ ext2_put_page(page);
+ return 0;
+ }
}
filp->f_pos += ext2_rec_len_from_disk(de->rec_len);
}
@@ -463,6 +483,10 @@ ino_t ext2_inode_by_dentry(struct inode *dir, struct dentry *dentry)
spin_lock(&dentry->d_lock);
dentry->d_flags |= DCACHE_WHITEOUT;
spin_unlock(&dentry->d_lock);
+ } else if(!res && de->file_type == EXT2_FT_FALLTHRU) {
+ spin_lock(&dentry->d_lock);
+ dentry->d_flags |= DCACHE_FALLTHRU;
+ spin_unlock(&dentry->d_lock);
}
ext2_put_page(page);
}
@@ -532,6 +556,7 @@ static ext2_dirent * ext2_append_entry(struct dentry * dentry,
de->name_len = 0;
de->rec_len = ext2_rec_len_to_disk(chunk_size);
de->inode = 0;
+ de->file_type = 0;
goto got_it;
}
if (de->rec_len == 0) {
@@ -545,6 +570,7 @@ static ext2_dirent * ext2_append_entry(struct dentry * dentry,
name_len = EXT2_DIR_REC_LEN(de->name_len);
rec_len = ext2_rec_len_from_disk(de->rec_len);
if (!de->inode && (de->file_type != EXT2_FT_WHT) &&
+ (de->file_type != EXT2_FT_FALLTHRU) &&
(rec_len >= reclen))
goto got_it;
if (rec_len >= name_len + reclen)
@@ -587,7 +613,8 @@ int ext2_add_link (struct dentry *dentry, struct inode *inode)

err = -EEXIST;
if (ext2_match (namelen, name, de)) {
- if (de->file_type == EXT2_FT_WHT)
+ if ((de->file_type == EXT2_FT_WHT) ||
+ (de->file_type == EXT2_FT_FALLTHRU))
goto got_it;
goto out_unlock;
}
@@ -602,7 +629,8 @@ got_it:
&page, NULL);
if (err)
goto out_unlock;
- if (de->inode || ((de->file_type == EXT2_FT_WHT) &&
+ if (de->inode || (((de->file_type == EXT2_FT_WHT) ||
+ (de->file_type == EXT2_FT_FALLTHRU)) &&
!ext2_match (namelen, name, de))) {
ext2_dirent *de1 = (ext2_dirent *) ((char *) de + name_len);
de1->rec_len = ext2_rec_len_to_disk(rec_len - name_len);
@@ -627,6 +655,60 @@ out_unlock:
}

/*
+ * Create a fallthru entry.
+ */
+int ext2_fallthru_entry (struct inode *dir, struct dentry *dentry)
+{
+ const char *name = dentry->d_name.name;
+ int namelen = dentry->d_name.len;
+ unsigned short rec_len, name_len;
+ ext2_dirent * de;
+ struct page *page;
+ loff_t pos;
+ int err;
+
+ de = ext2_append_entry(dentry, &page);
+ if (IS_ERR(de))
+ return PTR_ERR(de);
+
+ err = -EEXIST;
+ if (ext2_match (namelen, name, de))
+ goto out_unlock;
+
+ name_len = EXT2_DIR_REC_LEN(de->name_len);
+ rec_len = ext2_rec_len_from_disk(de->rec_len);
+
+ pos = page_offset(page) +
+ (char*)de - (char*)page_address(page);
+ err = __ext2_write_begin(NULL, page->mapping, pos, rec_len, 0,
+ &page, NULL);
+ if (err)
+ goto out_unlock;
+ if (de->inode || (de->file_type == EXT2_FT_WHT) ||
+ (de->file_type == EXT2_FT_FALLTHRU)) {
+ ext2_dirent *de1 = (ext2_dirent *) ((char *) de + name_len);
+ de1->rec_len = ext2_rec_len_to_disk(rec_len - name_len);
+ de->rec_len = ext2_rec_len_to_disk(name_len);
+ de = de1;
+ }
+ de->name_len = namelen;
+ memcpy(de->name, name, namelen);
+ de->inode = 0;
+ de->file_type = EXT2_FT_FALLTHRU;
+ err = ext2_commit_chunk(page, pos, rec_len);
+ dir->i_mtime = dir->i_ctime = CURRENT_TIME_SEC;
+ EXT2_I(dir)->i_flags &= ~EXT2_BTREE_FL;
+ mark_inode_dirty(dir);
+ /* OFFSET_CACHE */
+out_put:
+ ext2_put_page(page);
+ return err;
+out_unlock:
+ unlock_page(page);
+ goto out_put;
+}
+
+/*
* ext2_delete_entry deletes a directory entry by merging it with the
* previous entry. Page is up-to-date. Releases the page.
*/
@@ -711,7 +793,9 @@ int ext2_whiteout_entry (struct inode * dir, struct dentry * dentry,
*/
if (ext2_match (namelen, name, de))
de->inode = 0;
- if (de->inode || (de->file_type == EXT2_FT_WHT)) {
+ if (de->inode || (((de->file_type == EXT2_FT_WHT) ||
+ (de->file_type == EXT2_FT_FALLTHRU)) &&
+ !ext2_match (namelen, name, de))) {
ext2_dirent *de1 = (ext2_dirent *) ((char *) de + name_len);
de1->rec_len = ext2_rec_len_to_disk(rec_len - name_len);
de->rec_len = ext2_rec_len_to_disk(name_len);
diff --git a/fs/ext2/ext2.h b/fs/ext2/ext2.h
index 44d190c..2fa32b3 100644
--- a/fs/ext2/ext2.h
+++ b/fs/ext2/ext2.h
@@ -108,6 +108,7 @@ extern struct ext2_dir_entry_2 * ext2_find_entry (struct inode *,struct qstr *,
extern int ext2_delete_entry (struct ext2_dir_entry_2 *, struct page *);
extern int ext2_whiteout_entry (struct inode *, struct dentry *,
struct ext2_dir_entry_2 *, struct page *);
+extern int ext2_fallthru_entry (struct inode *, struct dentry *);
extern int ext2_empty_dir (struct inode *);
extern struct ext2_dir_entry_2 * ext2_dotdot (struct inode *, struct page **);
extern void ext2_set_link(struct inode *, struct ext2_dir_entry_2 *, struct page *, struct inode *, int);
diff --git a/fs/ext2/namei.c b/fs/ext2/namei.c
index 12195a5..f28154c 100644
--- a/fs/ext2/namei.c
+++ b/fs/ext2/namei.c
@@ -349,6 +349,7 @@ static int ext2_whiteout(struct inode *dir, struct dentry *dentry,
goto out;

spin_lock(&new_dentry->d_lock);
+ new_dentry->d_flags &= ~DCACHE_FALLTHRU;
new_dentry->d_flags |= DCACHE_WHITEOUT;
spin_unlock(&new_dentry->d_lock);
d_add(new_dentry, NULL);
@@ -367,6 +368,26 @@ out:
return err;
}

+/*
+ * Create a fallthru entry.
+ */
+static int ext2_fallthru (struct inode *dir, struct dentry *dentry)
+{
+ int err;
+
+ dquot_initialize(dir);
+
+ err = ext2_fallthru_entry(dir, dentry);
+ if (err)
+ return err;
+
+ d_instantiate(dentry, NULL);
+ spin_lock(&dentry->d_lock);
+ dentry->d_flags |= DCACHE_FALLTHRU;
+ spin_unlock(&dentry->d_lock);
+ return 0;
+}
+
static int ext2_rename (struct inode * old_dir, struct dentry * old_dentry,
struct inode * new_dir, struct dentry * new_dentry )
{
@@ -470,6 +491,7 @@ const struct inode_operations ext2_dir_inode_operations = {
.rmdir = ext2_rmdir,
.mknod = ext2_mknod,
.whiteout = ext2_whiteout,
+ .fallthru = ext2_fallthru,
.rename = ext2_rename,
#ifdef CONFIG_EXT2_FS_XATTR
.setxattr = generic_setxattr,
diff --git a/include/linux/ext2_fs.h b/include/linux/ext2_fs.h
index 20468bd..cb3d400 100644
--- a/include/linux/ext2_fs.h
+++ b/include/linux/ext2_fs.h
@@ -577,6 +577,7 @@ enum {
EXT2_FT_SOCK = 6,
EXT2_FT_SYMLINK = 7,
EXT2_FT_WHT = 8,
+ EXT2_FT_FALLTHRU = 9,
EXT2_FT_MAX
};

--
1.6.3.3



2010-04-19 12:40:27

by Jan Blunck

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

On Thu, Apr 15, Valerie Aurora wrote:

> Add support for fallthru directory entries to ext2.
>
> XXX - Makes up inode number for fallthru entry
> XXX - Might be better implemented as special symlinks

Better not. David Woodhouse actually convinced me of moving away from the
special symlink approach. The whiteouts have been implemented as special
symlinks before.

What makes you think that it would be beneficial to do so?

Thanks,
Jan

> Cc: Theodore Tso <[email protected]>
> Cc: [email protected]
> Signed-off-by: Valerie Aurora <[email protected]>
> Signed-off-by: Jan Blunck <[email protected]>
> ---
> fs/ext2/dir.c | 92 ++++++++++++++++++++++++++++++++++++++++++++--
> fs/ext2/ext2.h | 1 +
> fs/ext2/namei.c | 22 +++++++++++
> include/linux/ext2_fs.h | 1 +
> 4 files changed, 112 insertions(+), 4 deletions(-)
>
> diff --git a/fs/ext2/dir.c b/fs/ext2/dir.c
> index 030bd46..f3b4aff 100644
> --- a/fs/ext2/dir.c
> +++ b/fs/ext2/dir.c
> @@ -219,7 +219,8 @@ static inline int ext2_match (int len, const char * const name,
> {
> if (len != de->name_len)
> return 0;
> - if (!de->inode && (de->file_type != EXT2_FT_WHT))
> + if (!de->inode && ((de->file_type != EXT2_FT_WHT) &&
> + (de->file_type != EXT2_FT_FALLTHRU)))
> return 0;
> return !memcmp(name, de->name, len);
> }
> @@ -256,6 +257,7 @@ static unsigned char ext2_filetype_table[EXT2_FT_MAX] = {
> [EXT2_FT_SOCK] = DT_SOCK,
> [EXT2_FT_SYMLINK] = DT_LNK,
> [EXT2_FT_WHT] = DT_WHT,
> + [EXT2_FT_FALLTHRU] = DT_UNKNOWN,
> };
>
> #define S_SHIFT 12
> @@ -342,6 +344,24 @@ ext2_readdir (struct file * filp, void * dirent, filldir_t filldir)
> ext2_put_page(page);
> return 0;
> }
> + } else if (de->file_type == EXT2_FT_FALLTHRU) {
> + int over;
> + unsigned char d_type = DT_UNKNOWN;
> +
> + offset = (char *)de - kaddr;
> + /* XXX We don't know the inode number
> + * of the directory entry in the
> + * underlying file system. Should
> + * look it up, either on fallthru
> + * creation at first readdir or now at
> + * filldir time. */
> + over = filldir(dirent, de->name, de->name_len,
> + (n<<PAGE_CACHE_SHIFT) | offset,
> + 123 /* Made up ino */, d_type);
> + if (over) {
> + ext2_put_page(page);
> + return 0;
> + }
> }
> filp->f_pos += ext2_rec_len_from_disk(de->rec_len);
> }
> @@ -463,6 +483,10 @@ ino_t ext2_inode_by_dentry(struct inode *dir, struct dentry *dentry)
> spin_lock(&dentry->d_lock);
> dentry->d_flags |= DCACHE_WHITEOUT;
> spin_unlock(&dentry->d_lock);
> + } else if(!res && de->file_type == EXT2_FT_FALLTHRU) {
> + spin_lock(&dentry->d_lock);
> + dentry->d_flags |= DCACHE_FALLTHRU;
> + spin_unlock(&dentry->d_lock);
> }
> ext2_put_page(page);
> }
> @@ -532,6 +556,7 @@ static ext2_dirent * ext2_append_entry(struct dentry * dentry,
> de->name_len = 0;
> de->rec_len = ext2_rec_len_to_disk(chunk_size);
> de->inode = 0;
> + de->file_type = 0;
> goto got_it;
> }
> if (de->rec_len == 0) {
> @@ -545,6 +570,7 @@ static ext2_dirent * ext2_append_entry(struct dentry * dentry,
> name_len = EXT2_DIR_REC_LEN(de->name_len);
> rec_len = ext2_rec_len_from_disk(de->rec_len);
> if (!de->inode && (de->file_type != EXT2_FT_WHT) &&
> + (de->file_type != EXT2_FT_FALLTHRU) &&
> (rec_len >= reclen))
> goto got_it;
> if (rec_len >= name_len + reclen)
> @@ -587,7 +613,8 @@ int ext2_add_link (struct dentry *dentry, struct inode *inode)
>
> err = -EEXIST;
> if (ext2_match (namelen, name, de)) {
> - if (de->file_type == EXT2_FT_WHT)
> + if ((de->file_type == EXT2_FT_WHT) ||
> + (de->file_type == EXT2_FT_FALLTHRU))
> goto got_it;
> goto out_unlock;
> }
> @@ -602,7 +629,8 @@ got_it:
> &page, NULL);
> if (err)
> goto out_unlock;
> - if (de->inode || ((de->file_type == EXT2_FT_WHT) &&
> + if (de->inode || (((de->file_type == EXT2_FT_WHT) ||
> + (de->file_type == EXT2_FT_FALLTHRU)) &&
> !ext2_match (namelen, name, de))) {
> ext2_dirent *de1 = (ext2_dirent *) ((char *) de + name_len);
> de1->rec_len = ext2_rec_len_to_disk(rec_len - name_len);
> @@ -627,6 +655,60 @@ out_unlock:
> }
>
> /*
> + * Create a fallthru entry.
> + */
> +int ext2_fallthru_entry (struct inode *dir, struct dentry *dentry)
> +{
> + const char *name = dentry->d_name.name;
> + int namelen = dentry->d_name.len;
> + unsigned short rec_len, name_len;
> + ext2_dirent * de;
> + struct page *page;
> + loff_t pos;
> + int err;
> +
> + de = ext2_append_entry(dentry, &page);
> + if (IS_ERR(de))
> + return PTR_ERR(de);
> +
> + err = -EEXIST;
> + if (ext2_match (namelen, name, de))
> + goto out_unlock;
> +
> + name_len = EXT2_DIR_REC_LEN(de->name_len);
> + rec_len = ext2_rec_len_from_disk(de->rec_len);
> +
> + pos = page_offset(page) +
> + (char*)de - (char*)page_address(page);
> + err = __ext2_write_begin(NULL, page->mapping, pos, rec_len, 0,
> + &page, NULL);
> + if (err)
> + goto out_unlock;
> + if (de->inode || (de->file_type == EXT2_FT_WHT) ||
> + (de->file_type == EXT2_FT_FALLTHRU)) {
> + ext2_dirent *de1 = (ext2_dirent *) ((char *) de + name_len);
> + de1->rec_len = ext2_rec_len_to_disk(rec_len - name_len);
> + de->rec_len = ext2_rec_len_to_disk(name_len);
> + de = de1;
> + }
> + de->name_len = namelen;
> + memcpy(de->name, name, namelen);
> + de->inode = 0;
> + de->file_type = EXT2_FT_FALLTHRU;
> + err = ext2_commit_chunk(page, pos, rec_len);
> + dir->i_mtime = dir->i_ctime = CURRENT_TIME_SEC;
> + EXT2_I(dir)->i_flags &= ~EXT2_BTREE_FL;
> + mark_inode_dirty(dir);
> + /* OFFSET_CACHE */
> +out_put:
> + ext2_put_page(page);
> + return err;
> +out_unlock:
> + unlock_page(page);
> + goto out_put;
> +}
> +
> +/*
> * ext2_delete_entry deletes a directory entry by merging it with the
> * previous entry. Page is up-to-date. Releases the page.
> */
> @@ -711,7 +793,9 @@ int ext2_whiteout_entry (struct inode * dir, struct dentry * dentry,
> */
> if (ext2_match (namelen, name, de))
> de->inode = 0;
> - if (de->inode || (de->file_type == EXT2_FT_WHT)) {
> + if (de->inode || (((de->file_type == EXT2_FT_WHT) ||
> + (de->file_type == EXT2_FT_FALLTHRU)) &&
> + !ext2_match (namelen, name, de))) {
> ext2_dirent *de1 = (ext2_dirent *) ((char *) de + name_len);
> de1->rec_len = ext2_rec_len_to_disk(rec_len - name_len);
> de->rec_len = ext2_rec_len_to_disk(name_len);
> diff --git a/fs/ext2/ext2.h b/fs/ext2/ext2.h
> index 44d190c..2fa32b3 100644
> --- a/fs/ext2/ext2.h
> +++ b/fs/ext2/ext2.h
> @@ -108,6 +108,7 @@ extern struct ext2_dir_entry_2 * ext2_find_entry (struct inode *,struct qstr *,
> extern int ext2_delete_entry (struct ext2_dir_entry_2 *, struct page *);
> extern int ext2_whiteout_entry (struct inode *, struct dentry *,
> struct ext2_dir_entry_2 *, struct page *);
> +extern int ext2_fallthru_entry (struct inode *, struct dentry *);
> extern int ext2_empty_dir (struct inode *);
> extern struct ext2_dir_entry_2 * ext2_dotdot (struct inode *, struct page **);
> extern void ext2_set_link(struct inode *, struct ext2_dir_entry_2 *, struct page *, struct inode *, int);
> diff --git a/fs/ext2/namei.c b/fs/ext2/namei.c
> index 12195a5..f28154c 100644
> --- a/fs/ext2/namei.c
> +++ b/fs/ext2/namei.c
> @@ -349,6 +349,7 @@ static int ext2_whiteout(struct inode *dir, struct dentry *dentry,
> goto out;
>
> spin_lock(&new_dentry->d_lock);
> + new_dentry->d_flags &= ~DCACHE_FALLTHRU;
> new_dentry->d_flags |= DCACHE_WHITEOUT;
> spin_unlock(&new_dentry->d_lock);
> d_add(new_dentry, NULL);
> @@ -367,6 +368,26 @@ out:
> return err;
> }
>
> +/*
> + * Create a fallthru entry.
> + */
> +static int ext2_fallthru (struct inode *dir, struct dentry *dentry)
> +{
> + int err;
> +
> + dquot_initialize(dir);
> +
> + err = ext2_fallthru_entry(dir, dentry);
> + if (err)
> + return err;
> +
> + d_instantiate(dentry, NULL);
> + spin_lock(&dentry->d_lock);
> + dentry->d_flags |= DCACHE_FALLTHRU;
> + spin_unlock(&dentry->d_lock);
> + return 0;
> +}
> +
> static int ext2_rename (struct inode * old_dir, struct dentry * old_dentry,
> struct inode * new_dir, struct dentry * new_dentry )
> {
> @@ -470,6 +491,7 @@ const struct inode_operations ext2_dir_inode_operations = {
> .rmdir = ext2_rmdir,
> .mknod = ext2_mknod,
> .whiteout = ext2_whiteout,
> + .fallthru = ext2_fallthru,
> .rename = ext2_rename,
> #ifdef CONFIG_EXT2_FS_XATTR
> .setxattr = generic_setxattr,
> diff --git a/include/linux/ext2_fs.h b/include/linux/ext2_fs.h
> index 20468bd..cb3d400 100644
> --- a/include/linux/ext2_fs.h
> +++ b/include/linux/ext2_fs.h
> @@ -577,6 +577,7 @@ enum {
> EXT2_FT_SOCK = 6,
> EXT2_FT_SYMLINK = 7,
> EXT2_FT_WHT = 8,
> + EXT2_FT_FALLTHRU = 9,
> EXT2_FT_MAX
> };
>
> --
> 1.6.3.3
>
Regards,
Jan

--
Jan Blunck <[email protected]>

2010-04-19 13:02:56

by David Woodhouse

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

On Mon, 2010-04-19 at 14:40 +0200, Jan Blunck wrote:
> On Thu, Apr 15, Valerie Aurora wrote:
>
> > Add support for fallthru directory entries to ext2.
> >
> > XXX - Makes up inode number for fallthru entry
> > XXX - Might be better implemented as special symlinks
>
> Better not. David Woodhouse actually convinced me of moving away from the
> special symlink approach. The whiteouts have been implemented as special
> symlinks before.

I certainly asked whether you really need a real 'struct inode' for
whiteouts, and suggested that they should be represented _purely_ as a
dentry with type DT_WHT.

I don't much like the manifestation of that in this patch though,
especially with the made-up inode number. (ISTR I had other
jffs2-specific objections too, which I'll dig out and forward).

--
David Woodhouse Open Source Technology Centre
[email protected] Intel Corporation


2010-04-19 13:23:47

by Jan Blunck

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

On Mon, Apr 19, David Woodhouse wrote:

> On Mon, 2010-04-19 at 14:40 +0200, Jan Blunck wrote:
> > On Thu, Apr 15, Valerie Aurora wrote:
> >
> > > Add support for fallthru directory entries to ext2.
> > >
> > > XXX - Makes up inode number for fallthru entry
> > > XXX - Might be better implemented as special symlinks
> >
> > Better not. David Woodhouse actually convinced me of moving away from the
> > special symlink approach. The whiteouts have been implemented as special
> > symlinks before.
>
> I certainly asked whether you really need a real 'struct inode' for
> whiteouts, and suggested that they should be represented _purely_ as a
> dentry with type DT_WHT.
>
> I don't much like the manifestation of that in this patch though,
> especially with the made-up inode number. (ISTR I had other
> jffs2-specific objections too, which I'll dig out and forward).

Yes, this patches still have issues that Val and me are aware off. I can't
remember anything jffs2-specific though.

We return that inode number because we don't want to lookup the name on the
other filesystem during readdir. Therefore returning DT_UNKNOWN to let the
userspace decide if it needs to stat the file was the easiest workaround. I
know that POSIX requires d_ino and d_name but on the other hand it does not
require anything more on how long d_ino is valid.

If somebody has an idea how to make this cleaner please speak up.

Regards,
Jan

--
Jan Blunck <[email protected]>

2010-04-19 13:30:41

by Jamie Lokier

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

Jan Blunck wrote:
> On Mon, Apr 19, David Woodhouse wrote:
>
> > On Mon, 2010-04-19 at 14:40 +0200, Jan Blunck wrote:
> > > On Thu, Apr 15, Valerie Aurora wrote:
> > >
> > > > Add support for fallthru directory entries to ext2.
> > > >
> > > > XXX - Makes up inode number for fallthru entry
> > > > XXX - Might be better implemented as special symlinks
> > >
> > > Better not. David Woodhouse actually convinced me of moving away from the
> > > special symlink approach. The whiteouts have been implemented as special
> > > symlinks before.
> >
> > I certainly asked whether you really need a real 'struct inode' for
> > whiteouts, and suggested that they should be represented _purely_ as a
> > dentry with type DT_WHT.
> >
> > I don't much like the manifestation of that in this patch though,
> > especially with the made-up inode number. (ISTR I had other
> > jffs2-specific objections too, which I'll dig out and forward).
>
> Yes, this patches still have issues that Val and me are aware off. I can't
> remember anything jffs2-specific though.
>
> We return that inode number because we don't want to lookup the name on the
> other filesystem during readdir. Therefore returning DT_UNKNOWN to let the
> userspace decide if it needs to stat the file was the easiest workaround. I
> know that POSIX requires d_ino and d_name but on the other hand it does not
> require anything more on how long d_ino is valid.

Although the lifetime of d_ino might very, I know some programs (not
public) that will break if they see a d_ino which is wrongly matching
the st_ino of another file somewhere on the same st_dev. They will
assume the name is a hard link to the other file, without calling
stat(), which I think is a reasonable assumption and a useful optimisation.

So the made-up d_ino should at least be careful to not match an inode
number of another file which has a stable st_ino.

Why not zero for d_ino?

-- Jamie


2010-04-19 14:12:48

by Jan Blunck

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

On Mon, Apr 19, Jamie Lokier wrote:

> Jan Blunck wrote:
> > On Mon, Apr 19, David Woodhouse wrote:
> >
> > > On Mon, 2010-04-19 at 14:40 +0200, Jan Blunck wrote:
> > > > On Thu, Apr 15, Valerie Aurora wrote:
> > > >
> > > > > Add support for fallthru directory entries to ext2.
> > > > >
> > > > > XXX - Makes up inode number for fallthru entry
> > > > > XXX - Might be better implemented as special symlinks
> > > >
> > > > Better not. David Woodhouse actually convinced me of moving away from the
> > > > special symlink approach. The whiteouts have been implemented as special
> > > > symlinks before.
> > >
> > > I certainly asked whether you really need a real 'struct inode' for
> > > whiteouts, and suggested that they should be represented _purely_ as a
> > > dentry with type DT_WHT.
> > >
> > > I don't much like the manifestation of that in this patch though,
> > > especially with the made-up inode number. (ISTR I had other
> > > jffs2-specific objections too, which I'll dig out and forward).
> >
> > Yes, this patches still have issues that Val and me are aware off. I can't
> > remember anything jffs2-specific though.
> >
> > We return that inode number because we don't want to lookup the name on the
> > other filesystem during readdir. Therefore returning DT_UNKNOWN to let the
> > userspace decide if it needs to stat the file was the easiest workaround. I
> > know that POSIX requires d_ino and d_name but on the other hand it does not
> > require anything more on how long d_ino is valid.
>
> Although the lifetime of d_ino might very, I know some programs (not
> public) that will break if they see a d_ino which is wrongly matching
> the st_ino of another file somewhere on the same st_dev. They will
> assume the name is a hard link to the other file, without calling
> stat(), which I think is a reasonable assumption and a useful optimisation.
>
> So the made-up d_ino should at least be careful to not match an inode
> number of another file which has a stable st_ino.
>
> Why not zero for d_ino?
>

Hmm, why not. Or even the ino of the directory we are reading from ...

Regards,
Jan

--
Jan Blunck <[email protected]>

2010-04-19 14:23:15

by Valerie Aurora

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

On Mon, Apr 19, 2010 at 04:12:48PM +0200, Jan Blunck wrote:
> On Mon, Apr 19, Jamie Lokier wrote:
>
> > Jan Blunck wrote:
> > > On Mon, Apr 19, David Woodhouse wrote:
> > >
> > > > On Mon, 2010-04-19 at 14:40 +0200, Jan Blunck wrote:
> > > > > On Thu, Apr 15, Valerie Aurora wrote:
> > > > >
> > > > > > Add support for fallthru directory entries to ext2.
> > > > > >
> > > > > > XXX - Makes up inode number for fallthru entry
> > > > > > XXX - Might be better implemented as special symlinks
> > > > >
> > > > > Better not. David Woodhouse actually convinced me of moving away from the
> > > > > special symlink approach. The whiteouts have been implemented as special
> > > > > symlinks before.
> > > >
> > > > I certainly asked whether you really need a real 'struct inode' for
> > > > whiteouts, and suggested that they should be represented _purely_ as a
> > > > dentry with type DT_WHT.
> > > >
> > > > I don't much like the manifestation of that in this patch though,
> > > > especially with the made-up inode number. (ISTR I had other
> > > > jffs2-specific objections too, which I'll dig out and forward).
> > >
> > > Yes, this patches still have issues that Val and me are aware off. I can't
> > > remember anything jffs2-specific though.
> > >
> > > We return that inode number because we don't want to lookup the name on the
> > > other filesystem during readdir. Therefore returning DT_UNKNOWN to let the
> > > userspace decide if it needs to stat the file was the easiest workaround. I
> > > know that POSIX requires d_ino and d_name but on the other hand it does not
> > > require anything more on how long d_ino is valid.
> >
> > Although the lifetime of d_ino might very, I know some programs (not
> > public) that will break if they see a d_ino which is wrongly matching
> > the st_ino of another file somewhere on the same st_dev. They will
> > assume the name is a hard link to the other file, without calling
> > stat(), which I think is a reasonable assumption and a useful optimisation.
> >
> > So the made-up d_ino should at least be careful to not match an inode
> > number of another file which has a stable st_ino.
> >
> > Why not zero for d_ino?
> >
>
> Hmm, why not. Or even the ino of the directory we are reading from ...

I don't recall there being any technical reason not to look up the
real inode number. I just wrote it that we because I was lazy. So I
like returning the directory's d_ino better than a single magic
number, but I'd at least like to try returning the real inode number
too.

-VAL

2010-04-19 14:53:22

by Miklos Szeredi

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

On Mon, 19 Apr 2010, Valerie Aurora wrote:
> I don't recall there being any technical reason not to look up the
> real inode number. I just wrote it that we because I was lazy. So I
> like returning the directory's d_ino better than a single magic
> number, but I'd at least like to try returning the real inode number
> too.

Note, "struct dirent" doesn't have d_dev, so you really can't return
the "real" inode number, that's on a different filesystem and just a
random number in the context of the the readdir in question.

Thanks,
Miklos

2010-04-20 21:35:10

by Jamie Lokier

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

Miklos Szeredi wrote:
> On Mon, 19 Apr 2010, Valerie Aurora wrote:
> > I don't recall there being any technical reason not to look up the
> > real inode number. I just wrote it that we because I was lazy. So I
> > like returning the directory's d_ino better than a single magic
> > number, but I'd at least like to try returning the real inode number
> > too.
>
> Note, "struct dirent" doesn't have d_dev, so you really can't return
> the "real" inode number, that's on a different filesystem and just a
> random number in the context of the the readdir in question.

Agree. Does this inappropriate inode number for the union mount's
st_dev happen with stat() on the actual files too? That could be bad.

-- Jamie

2010-04-20 21:40:49

by Jamie Lokier

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

Valerie Aurora wrote:
> On Mon, Apr 19, 2010 at 04:12:48PM +0200, Jan Blunck wrote:
> > On Mon, Apr 19, Jamie Lokier wrote:
> >
> > > Jan Blunck wrote:
> > > > On Mon, Apr 19, David Woodhouse wrote:
> > > >
> > > > > On Mon, 2010-04-19 at 14:40 +0200, Jan Blunck wrote:
> > > > > > On Thu, Apr 15, Valerie Aurora wrote:
> > > > > >
> > > > > > > Add support for fallthru directory entries to ext2.
> > > > > > >
> > > > > > > XXX - Makes up inode number for fallthru entry
> > > > > > > XXX - Might be better implemented as special symlinks
> > > > > >
> > > > > > Better not. David Woodhouse actually convinced me of moving away from the
> > > > > > special symlink approach. The whiteouts have been implemented as special
> > > > > > symlinks before.
> > > > >
> > > > > I certainly asked whether you really need a real 'struct inode' for
> > > > > whiteouts, and suggested that they should be represented _purely_ as a
> > > > > dentry with type DT_WHT.
> > > > >
> > > > > I don't much like the manifestation of that in this patch though,
> > > > > especially with the made-up inode number. (ISTR I had other
> > > > > jffs2-specific objections too, which I'll dig out and forward).
> > > >
> > > > Yes, this patches still have issues that Val and me are aware off. I can't
> > > > remember anything jffs2-specific though.
> > > >
> > > > We return that inode number because we don't want to lookup the name on the
> > > > other filesystem during readdir. Therefore returning DT_UNKNOWN to let the
> > > > userspace decide if it needs to stat the file was the easiest workaround. I
> > > > know that POSIX requires d_ino and d_name but on the other hand it does not
> > > > require anything more on how long d_ino is valid.
> > >
> > > Although the lifetime of d_ino might very, I know some programs (not
> > > public) that will break if they see a d_ino which is wrongly matching
> > > the st_ino of another file somewhere on the same st_dev. They will
> > > assume the name is a hard link to the other file, without calling
> > > stat(), which I think is a reasonable assumption and a useful optimisation.
> > >
> > > So the made-up d_ino should at least be careful to not match an inode
> > > number of another file which has a stable st_ino.
> > >
> > > Why not zero for d_ino?
> > >
> >
> > Hmm, why not. Or even the ino of the directory we are reading from ...
>
> I don't recall there being any technical reason not to look up the
> real inode number. I just wrote it that we because I was lazy. So I
> like returning the directory's d_ino better than a single magic
> number, but I'd at least like to try returning the real inode number
> too.

I thought of zero because Bash and GNU Readline both check d_ino != 0
to decide if an entry is valid.

On reflection, that is why zero must _not_ be used :-)

-- Jamie

2010-04-21 08:42:13

by Jan Blunck

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

On Tue, Apr 20, Jamie Lokier wrote:

> Miklos Szeredi wrote:
> > On Mon, 19 Apr 2010, Valerie Aurora wrote:
> > > I don't recall there being any technical reason not to look up the
> > > real inode number. I just wrote it that we because I was lazy. So I
> > > like returning the directory's d_ino better than a single magic
> > > number, but I'd at least like to try returning the real inode number
> > > too.
> >
> > Note, "struct dirent" doesn't have d_dev, so you really can't return
> > the "real" inode number, that's on a different filesystem and just a
> > random number in the context of the the readdir in question.
>
> Agree. Does this inappropriate inode number for the union mount's
> st_dev happen with stat() on the actual files too? That could be bad.

No, for stat() you do a lookup and that is returning the correct dentry/inode
for the filesystem the name is on.

We just return the the fallthru directory entries to give userspace an offset
that they can seekdir() to.

Regards,
Jan

--
Jan Blunck <[email protected]>

2010-04-21 09:22:52

by Jamie Lokier

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

Jan Blunck wrote:
> On Tue, Apr 20, Jamie Lokier wrote:
>
> > Miklos Szeredi wrote:
> > > On Mon, 19 Apr 2010, Valerie Aurora wrote:
> > > > I don't recall there being any technical reason not to look up the
> > > > real inode number. I just wrote it that we because I was lazy. So I
> > > > like returning the directory's d_ino better than a single magic
> > > > number, but I'd at least like to try returning the real inode number
> > > > too.
> > >
> > > Note, "struct dirent" doesn't have d_dev, so you really can't return
> > > the "real" inode number, that's on a different filesystem and just a
> > > random number in the context of the the readdir in question.
> >
> > Agree. Does this inappropriate inode number for the union mount's
> > st_dev happen with stat() on the actual files too? That could be bad.
>
> No, for stat() you do a lookup and that is returning the correct
> dentry/inode for the filesystem the name is on.

Hmm. I smell potential confusion for some otherwise POSIX-friendly
userspaces.

When I open /path/to/foo, call fstat (st_dev=2, st_ino=5678), and then
keep the file open, then later do a readdir which includes foo
(dir.st_dev=1, d_ino=1234), I'm going to immediately assume a rename
or unlink happened, close the file, abort streaming from it, refresh
the GUI windows, refresh application caches for that name entry, etc.

Because in the POSIX world I think open files have stable inode
numbers (as long as they are open), and I don't think that an open
file can have it's name's d_ino not match the inode number unless it's
a mount point, which my program would know about.

This plays into inotify, where you have to know if you are monitoring
every directory that contains a link to a file, to know if you need to
monitor the file itself directly instead.

Now I think it's fair enough that a union mount doesn't play all the
traditional rules :-) C'est la vie.

This mismatch of (dir.st_dev,d_ino) and st_ino strongly resembles a
file-bind-mount. Like bind mounts, it's quite annoying for programs
that like to assume they've seen all of a file's links when they've
seen i_nlink of them.

Bind mounts can be detected by looking in /proc/mounts. st_dev
changing doesn't work because it can be a binding of the same
filesystem.

How would I go about detecting when a union mount's directory entry
has similar behaviour, without calling stat() on each entry? Is it
just a matter of recognising a particular filesystem name in
/proc/mounts, or something more?

Thanks,
-- Jamie

2010-04-21 09:35:21

by Miklos Szeredi

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

On Wed, 21 Apr 2010, Jamie Lokier wrote:
> Hmm. I smell potential confusion for some otherwise POSIX-friendly
> userspaces.
>
> When I open /path/to/foo, call fstat (st_dev=2, st_ino=5678), and then
> keep the file open, then later do a readdir which includes foo
> (dir.st_dev=1, d_ino=1234), I'm going to immediately assume a rename
> or unlink happened, close the file, abort streaming from it, refresh
> the GUI windows, refresh application caches for that name entry, etc.
>
> Because in the POSIX world I think open files have stable inode
> numbers (as long as they are open), and I don't think that an open
> file can have it's name's d_ino not match the inode number unless it's
> a mount point, which my program would know about.
>
> This plays into inotify, where you have to know if you are monitoring
> every directory that contains a link to a file, to know if you need to
> monitor the file itself directly instead.
>
> Now I think it's fair enough that a union mount doesn't play all the
> traditional rules :-) C'est la vie.
>
> This mismatch of (dir.st_dev,d_ino) and st_ino strongly resembles a
> file-bind-mount. Like bind mounts, it's quite annoying for programs
> that like to assume they've seen all of a file's links when they've
> seen i_nlink of them.
>
> Bind mounts can be detected by looking in /proc/mounts. st_dev
> changing doesn't work because it can be a binding of the same
> filesystem.
>
> How would I go about detecting when a union mount's directory entry
> has similar behaviour, without calling stat() on each entry? Is it
> just a matter of recognising a particular filesystem name in
> /proc/mounts, or something more?

Detecting mount points is best done by comparing st_dev for the parent
directory with st_dev of the child. This is much simpler than parsing
/proc/mounts and will work for bind mounts as well as union mounts.

I think there's no question that union mounts might break apps (POSIX
or not). But I think there's hope that they are few and can easily be
fixed.

Thanks,
Miklos

2010-04-21 09:52:21

by Jamie Lokier

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

Miklos Szeredi wrote:
> On Wed, 21 Apr 2010, Jamie Lokier wrote:
> > Hmm. I smell potential confusion for some otherwise POSIX-friendly
> > userspaces.
> >
> > When I open /path/to/foo, call fstat (st_dev=2, st_ino=5678), and then
> > keep the file open, then later do a readdir which includes foo
> > (dir.st_dev=1, d_ino=1234), I'm going to immediately assume a rename
> > or unlink happened, close the file, abort streaming from it, refresh
> > the GUI windows, refresh application caches for that name entry, etc.
> >
> > Because in the POSIX world I think open files have stable inode
> > numbers (as long as they are open), and I don't think that an open
> > file can have it's name's d_ino not match the inode number unless it's
> > a mount point, which my program would know about.
> >
> > This plays into inotify, where you have to know if you are monitoring
> > every directory that contains a link to a file, to know if you need to
> > monitor the file itself directly instead.
> >
> > Now I think it's fair enough that a union mount doesn't play all the
> > traditional rules :-) C'est la vie.
> >
> > This mismatch of (dir.st_dev,d_ino) and st_ino strongly resembles a
> > file-bind-mount. Like bind mounts, it's quite annoying for programs
> > that like to assume they've seen all of a file's links when they've
> > seen i_nlink of them.
> >
> > Bind mounts can be detected by looking in /proc/mounts. st_dev
> > changing doesn't work because it can be a binding of the same
> > filesystem.
> >
> > How would I go about detecting when a union mount's directory entry
> > has similar behaviour, without calling stat() on each entry? Is it
> > just a matter of recognising a particular filesystem name in
> > /proc/mounts, or something more?
>
> Detecting mount points is best done by comparing st_dev for the parent
> directory with st_dev of the child. This is much simpler than parsing
> /proc/mounts and will work for bind mounts as well as union mounts.

Sorry, no: That does not work for bind mounts. Both layers can have
the same st_dev. Nor does O_NOFOLLOW stop traversal in the middle of
a path, there is no handy O_NOCROSSMOUNTS, and no st_mode flag or
d_type to say it's a bind mount. Bind mounts are really a big pain
for i_nlink+inotify name counting.

Besides, calling stat() on every entry in a large directory to check
st_ino can be orders of magnitude slower than readdir() on a large
directory - especially with a cold cache. It is quicker, but much
more complicated, to parse /proc/mounts and apply arcane rules to find
the exceptions.

Can a union mount overlap two parts of the same filesystem?

> I think there's no question that union mounts might break apps (POSIX
> or not). But I think there's hope that they are few and can easily be
> fixed.

I agree, and union moint is a very useful feature that's worth
breaking a few apps for :-)

I'm curious if there's a clear way to go about it in this case, or
if it'll involve a certain amount of pattern recognition in /proc/mounts.

Basically I'm wondering if it's been thought about already.

-- Jamie

2010-04-21 10:18:07

by Miklos Szeredi

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

On Wed, 21 Apr 2010, Jamie Lokier wrote:
> Sorry, no: That does not work for bind mounts. Both layers can have
> the same st_dev.

Okay.

> Nor does O_NOFOLLOW stop traversal in the middle of
> a path, there is no handy O_NOCROSSMOUNTS, and no st_mode flag or
> d_type to say it's a bind mount. Bind mounts are really a big pain
> for i_nlink+inotify name counting.

I'm confused. You are monitoring a specific file and would like to
know if something is happening to any of it's links, right?

Why do you need to know about bind mounts for that?

Count the number of times you encounter that d_ino and if that matches
i_nlink then every directory is monitored. Simple as that, no?

Thanks,
Miklos

2010-04-21 17:36:24

by Jamie Lokier

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

Miklos Szeredi wrote:
> On Wed, 21 Apr 2010, Jamie Lokier wrote:
> > Sorry, no: That does not work for bind mounts. Both layers can have
> > the same st_dev.
>
> Okay.
>
> > Nor does O_NOFOLLOW stop traversal in the middle of
> > a path, there is no handy O_NOCROSSMOUNTS, and no st_mode flag or
> > d_type to say it's a bind mount. Bind mounts are really a big pain
> > for i_nlink+inotify name counting.
>
> I'm confused. You are monitoring a specific file and would like to
> know if something is happening to any of it's links, right?

Not quite. I'm monitoring a million files (say), so I must use
directory watches for most of them. I need directory watches anyway,
when the semantic is "calling open on /path/to/file and reading would
return the same data", because renames and unlinks are also a way to
invalidate monitored file contents.

At a high level, what we're talking about is the ability to cache and
verify the the validity information derived from reading files in the
filesystem, in a manner which efficiently triggers invalidation only
on changes. Being able to answer, as quickly as possible, "if I read
this, that and other, will I get the same results as the last time I
did those operations, without having to actually do them to check".
There are many applications, provided the method is reliable.

> Why do you need to know about bind mounts for that?
>
> Count the number of times you encounter that d_ino and if that matches
> i_nlink then every directory is monitored. Simple as that, no?

When I see a file has i_nlink > 1, I must watch the file directly
using a file-watch (with inotify; polling with stat() with dnotify),
_unless_ I have seen all the links to that file.

When I've seen all the links to a file, I know that my directory
watches on the directories containing those links are sufficient to
detect changes to the file contents. That's because every
file change will get notified on at least one of those paths.

I learn that I've seen all the links by seeing d_ino during readdir as
you suggested, or by st_ino in the cases where I've not had reason to
readdir and I have needed to open the file or call stat.

Let's look at some bind mounts. One where st_ino doesn't work:

/dirA/file1 [hard link to inode 100, i_nlink = 2]
/dirA/bound [bind mount, has /dirA/file1 mounted on it]
/dirB/file2 [hard link to inode 100, i_nlink = 2]

If the program is asked to open /dirA/file1 and /dirA/bound at various
times, and never asked to readdir /dirA, it will have used fstat not
readdir, seen the same (st_dev,st_ino,i_nlink = 2), and _wrongly_
concluded that it is monitoring all directories containing paths to
the file.

To avoid that problem, it parses /proc/mounts and detects that
/dirA/bound does not contributed to the link count. This is faster
than calling readdir in all possible places that it can happen.

Another one, where readdir + d_ino doesn't work anyway:

/dirA/file1 [hard link to inode 100, i_nlink = 2]
/dirB/dirX [bind mount, has /dirA mounted on it]
/dirC/file2 [hard link to inode 100, i_nlink = 2]

This time the program is asked to open /dirA/file1 and
/dirB/dirX/file1 at various times. Suppose it aggressively calls
readdir on all of the places it goes near, and uses d_ino comparisons.

Bear in mind it can't hunt for /dirC because there may be millions of
directories; this is just an example.

Then it will see the same d_ino for /dirA/file1 and /dirB/dirX/file1,
and wrongly conclude that it is monitoring all directories containing
paths to the file.

So again, it must parse /proc/mounts to detect that everything found
under /dirB/dirX mirrors /dirA.

This is a bit more complicated by the fact that inotify/dnotify send
events to the watching dentry parent of the link used to access a
file, not necessarily the parent in the mounted path space.

Although this doesn't make the bind mount problem go away, this is
where union mounts complicate the picture more:

Ideally, the program may assume that d_ino and st_ino match as long as
the file is open (on any filesystem), or that the filesystem type is
in a whitelist of ones with stable inode numbers (most local
filesystems), and it's not a mountpoint. So when it's asked to open
at one path, and something else asks it to readdir at another path, it
could combine the information to learn when it's found all entries,
without having to use redundant readdirs and stats.

I'm thinking that I might have to detect union mounts specially in
/proc/mounts, now that they are a VFS feature, and disable a bunch of
assumptions about d_ino when seeing them. Hopefully it is possible to
unambiguously check for union mount points in /proc/mounts?

d_ino == directory's st_ino sounds neat. Maybe that will be enough,
as a special magical Linux rule. When reading a directory, it's cheap
to get the directory's st_ino with fstat(). It's possible to bind
mount a directory on it's _own_ child, so that st_ino == directory's
st_ino, but d_ino isn't affected so maybe that's the trick to use.

-- Jamie

2010-04-21 21:34:57

by Valerie Aurora

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

On Wed, Apr 21, 2010 at 10:52:21AM +0100, Jamie Lokier wrote:
> Miklos Szeredi wrote:
> > Detecting mount points is best done by comparing st_dev for the parent
> > directory with st_dev of the child. This is much simpler than parsing
> > /proc/mounts and will work for bind mounts as well as union mounts.
>
> Sorry, no: That does not work for bind mounts. Both layers can have
> the same st_dev. Nor does O_NOFOLLOW stop traversal in the middle of
> a path, there is no handy O_NOCROSSMOUNTS, and no st_mode flag or
> d_type to say it's a bind mount. Bind mounts are really a big pain
> for i_nlink+inotify name counting.
>
> Besides, calling stat() on every entry in a large directory to check
> st_ino can be orders of magnitude slower than readdir() on a large
> directory - especially with a cold cache. It is quicker, but much
> more complicated, to parse /proc/mounts and apply arcane rules to find
> the exceptions.
>
> Can a union mount overlap two parts of the same filesystem?

No. Each layer must be a separate file system, the bottom must be
read-only, the top must be writable, and they must be unioned at their
mount points.

> > I think there's no question that union mounts might break apps (POSIX
> > or not). But I think there's hope that they are few and can easily be
> > fixed.
>
> I agree, and union moint is a very useful feature that's worth
> breaking a few apps for :-)
>
> I'm curious if there's a clear way to go about it in this case, or
> if it'll involve a certain amount of pattern recognition in /proc/mounts.

All it takes is looking for the "union" string in the mount options.

> Basically I'm wondering if it's been thought about already.

Not as much as it deserves. :) Do you have any thoughts about better
solutions?

Something to keep in mind is that most of the app issues are already
present with bind mounts. In many cases, if an app doesn't work with
union mounts, it's also not going to work with bind mounts. I think
you have a good point that we could use a more straightforward way to
say, "Hey, you can't use the normal st_dev/st_ino rules right now..."

-VAL

2010-04-21 21:39:04

by Valerie Aurora

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

On Wed, Apr 21, 2010 at 11:34:52AM +0200, Miklos Szeredi wrote:
> On Wed, 21 Apr 2010, Jamie Lokier wrote:
> > Hmm. I smell potential confusion for some otherwise POSIX-friendly
> > userspaces.
> >
> > When I open /path/to/foo, call fstat (st_dev=2, st_ino=5678), and then
> > keep the file open, then later do a readdir which includes foo
> > (dir.st_dev=1, d_ino=1234), I'm going to immediately assume a rename
> > or unlink happened, close the file, abort streaming from it, refresh
> > the GUI windows, refresh application caches for that name entry, etc.
> >
> > Because in the POSIX world I think open files have stable inode
> > numbers (as long as they are open), and I don't think that an open
> > file can have it's name's d_ino not match the inode number unless it's
> > a mount point, which my program would know about.
> >
> > This plays into inotify, where you have to know if you are monitoring
> > every directory that contains a link to a file, to know if you need to
> > monitor the file itself directly instead.
> >
> > Now I think it's fair enough that a union mount doesn't play all the
> > traditional rules :-) C'est la vie.
> >
> > This mismatch of (dir.st_dev,d_ino) and st_ino strongly resembles a
> > file-bind-mount. Like bind mounts, it's quite annoying for programs
> > that like to assume they've seen all of a file's links when they've
> > seen i_nlink of them.
> >
> > Bind mounts can be detected by looking in /proc/mounts. st_dev
> > changing doesn't work because it can be a binding of the same
> > filesystem.
> >
> > How would I go about detecting when a union mount's directory entry
> > has similar behaviour, without calling stat() on each entry? Is it
> > just a matter of recognising a particular filesystem name in
> > /proc/mounts, or something more?
>
> Detecting mount points is best done by comparing st_dev for the parent
> directory with st_dev of the child. This is much simpler than parsing
> /proc/mounts and will work for bind mounts as well as union mounts.
>
> I think there's no question that union mounts might break apps (POSIX
> or not). But I think there's hope that they are few and can easily be
> fixed.

I couldn't have put it better myself.

To expand slightly, if the broken apps are not few and easily fixed,
then we'll go back and make the kernel more complicated. I'd like to
try the simplest version we think will work, first.

Thanks!

-VAL

2010-04-21 22:10:59

by Jamie Lokier

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support

Valerie Aurora wrote:
> > I think there's no question that union mounts might break apps (POSIX
> > or not). But I think there's hope that they are few and can easily be
> > fixed.
>
> I couldn't have put it better myself.
>
> To expand slightly, if the broken apps are not few and easily fixed,
> then we'll go back and make the kernel more complicated. I'd like to
> try the simplest version we think will work, first.

Don't worry, I'm not trying to deviate you from that good plan.

Just throwing questions out to find what's a good and simple answer to
these little open questions to minimise trouble.

-- Jamie

2010-04-22 10:30:55

by J. R. Okajima

[permalink] [raw]
Subject: Re: [PATCH 13/35] fallthru: ext2 fallthru support


Jamie Lokier:
> Hmm. I smell potential confusion for some otherwise POSIX-friendly
> userspaces.
:::
> This plays into inotify, where you have to know if you are monitoring
> every directory that contains a link to a file, to know if you need to
> monitor the file itself directly instead.

Addition to the inode number of fallthru/readdir, hardlink in union
mount may be a problem. If you open a hardlinked file for writing or
try chmod it, the internal copyup will happen and the hardlink will be
destroyed. For instance, when fileA and fileB are hardlinked on the
lower layer, and the contents of fileA is modifed (copyup happens). You
will not see the latest contents via fileB.
And the IN_CREATE event may be fired to the parent dir if you monitor
it, I am afraid.

(I have pointed out this issue before, but the posted document didn't
seem to contain about it)


J. R. Okajima