Re: [f2fs-dev] [PATCH 1/4] f2fs: convert inline_dir early before starting rename
From: Jaegeuk Kim
Date: Mon Dec 30 2019 - 19:06:53 EST
On 12/28, Chao Yu wrote:
> On 2019/12/19 4:09, Jaegeuk Kim wrote:
> > If we hit an error during rename, we'll get two dentries in different
> > directories.
> >
> > Signed-off-by: Jaegeuk Kim <jaegeuk@xxxxxxxxxx>
> > ---
> > fs/f2fs/f2fs.h | 1 +
> > fs/f2fs/inline.c | 30 ++++++++++++++++++++++++++++--
> > fs/f2fs/namei.c | 36 +++++++++++++-----------------------
> > 3 files changed, 42 insertions(+), 25 deletions(-)
> >
> > diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> > index dbc20d33d0e1..8d64525743cb 100644
> > --- a/fs/f2fs/f2fs.h
> > +++ b/fs/f2fs/f2fs.h
> > @@ -3490,6 +3490,7 @@ void f2fs_truncate_inline_inode(struct inode *inode,
> > int f2fs_read_inline_data(struct inode *inode, struct page *page);
> > int f2fs_convert_inline_page(struct dnode_of_data *dn, struct page *page);
> > int f2fs_convert_inline_inode(struct inode *inode);
> > +int f2fs_convert_inline_dir(struct inode *dir);
> > int f2fs_write_inline_data(struct inode *inode, struct page *page);
> > bool f2fs_recover_inline_data(struct inode *inode, struct page *npage);
> > struct f2fs_dir_entry *f2fs_find_in_inline_dir(struct inode *dir,
> > diff --git a/fs/f2fs/inline.c b/fs/f2fs/inline.c
> > index 52f85ed07a15..f82c3d9cf333 100644
> > --- a/fs/f2fs/inline.c
> > +++ b/fs/f2fs/inline.c
> > @@ -530,7 +530,7 @@ static int f2fs_move_rehashed_dirents(struct inode *dir, struct page *ipage,
> > return err;
> > }
> >
> > -static int f2fs_convert_inline_dir(struct inode *dir, struct page *ipage,
> > +static int do_convert_inline_dir(struct inode *dir, struct page *ipage,
> > void *inline_dentry)
> > {
> > if (!F2FS_I(dir)->i_dir_level)
> > @@ -539,6 +539,32 @@ static int f2fs_convert_inline_dir(struct inode *dir, struct page *ipage,
> > return f2fs_move_rehashed_dirents(dir, ipage, inline_dentry);
> > }
> >
> > +int f2fs_convert_inline_dir(struct inode *dir)
> > +{
> > + struct f2fs_sb_info *sbi = F2FS_I_SB(dir);
> > + struct page *ipage;
> > + void *inline_dentry = NULL;
> > + int err;
> > +
> > + if (!f2fs_has_inline_dentry(dir))
> > + return 0;
> > +
> > + f2fs_lock_op(sbi);
> > +
> > + ipage = f2fs_get_node_page(sbi, dir->i_ino);
> > + if (IS_ERR(ipage))
> > + return PTR_ERR(ipage);
> > +
> > + inline_dentry = inline_data_addr(dir, ipage);
> > +
> > + err = do_convert_inline_dir(dir, ipage, inline_dentry);
> > + if (!err)
> > + f2fs_put_page(ipage, 1);
> > +
> > + f2fs_unlock_op(sbi);
> > + return err;
> > +}
> > +
> > int f2fs_add_inline_entry(struct inode *dir, const struct qstr *new_name,
> > const struct qstr *orig_name,
> > struct inode *inode, nid_t ino, umode_t mode)
> > @@ -562,7 +588,7 @@ int f2fs_add_inline_entry(struct inode *dir, const struct qstr *new_name,
> >
> > bit_pos = f2fs_room_for_filename(d.bitmap, slots, d.max);
> > if (bit_pos >= d.max) {
> > - err = f2fs_convert_inline_dir(dir, ipage, inline_dentry);
> > + err = do_convert_inline_dir(dir, ipage, inline_dentry);
> > if (err)
> > return err;
> > err = -EAGAIN;
> > diff --git a/fs/f2fs/namei.c b/fs/f2fs/namei.c
> > index 5d9584281935..61615ab466c2 100644
> > --- a/fs/f2fs/namei.c
> > +++ b/fs/f2fs/namei.c
> > @@ -855,7 +855,6 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
> > struct f2fs_dir_entry *old_dir_entry = NULL;
> > struct f2fs_dir_entry *old_entry;
> > struct f2fs_dir_entry *new_entry;
> > - bool is_old_inline = f2fs_has_inline_dentry(old_dir);
> > int err;
> >
> > if (unlikely(f2fs_cp_error(sbi)))
> > @@ -868,6 +867,19 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
> > F2FS_I(old_dentry->d_inode)->i_projid)))
> > return -EXDEV;
> >
> > + /*
> > + * old entry and new entry can locate in the same inline
> > + * dentry in inode, when attaching new entry in inline dentry,
> > + * it could force inline dentry conversion, after that,
> > + * old_entry and old_page will point to wrong address, in
> > + * order to avoid this, let's do the check and update here.
> > + */
>
> The comment is out-of-update here...
>
> If there is enough room for the new dentry in old directory's inline space,
> how about just keeping inline dir state for old directory?
It'd work, but how can we deal with race conditions?
>
>
> ---
> fs/f2fs/dir.c | 14 ++++++++++++++
> fs/f2fs/f2fs.h | 4 +++-
> fs/f2fs/inline.c | 22 +++++++++++++++++-----
> fs/f2fs/namei.c | 2 +-
> 4 files changed, 35 insertions(+), 7 deletions(-)
>
> diff --git a/fs/f2fs/dir.c b/fs/f2fs/dir.c
> index c967cacf979e..b56f6060c1a6 100644
> --- a/fs/f2fs/dir.c
> +++ b/fs/f2fs/dir.c
> @@ -578,6 +578,20 @@ int f2fs_room_for_filename(const void *bitmap, int slots, int max_slots)
> goto next;
> }
>
> +bool f2fs_has_enough_room(struct inode *dir, struct page *ipage,
> + struct fscrypt_name *fname)
> +{
> + struct f2fs_dentry_ptr d;
> + unsigned int bit_pos;
> + int slots = GET_DENTRY_SLOTS(fname_len(fname));
> +
> + make_dentry_ptr_inline(dir, &d, inline_data_addr(dir, ipage));
> +
> + bit_pos = f2fs_room_for_filename(d.bitmap, slots, d.max);
> +
> + return bit_pos < d.max;
> +}
> +
> void f2fs_update_dentry(nid_t ino, umode_t mode, struct f2fs_dentry_ptr *d,
> const struct qstr *name, f2fs_hash_t name_hash,
> unsigned int bit_pos)
> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> index 5d55cef66410..ce4ce33a40ce 100644
> --- a/fs/f2fs/f2fs.h
> +++ b/fs/f2fs/f2fs.h
> @@ -3120,6 +3120,8 @@ ino_t f2fs_inode_by_name(struct inode *dir, const struct qstr *qstr,
> struct page **page);
> void f2fs_set_link(struct inode *dir, struct f2fs_dir_entry *de,
> struct page *page, struct inode *inode);
> +bool f2fs_has_enough_room(struct inode *dir, struct page *ipage,
> + struct fscrypt_name *fname);
> void f2fs_update_dentry(nid_t ino, umode_t mode, struct f2fs_dentry_ptr *d,
> const struct qstr *name, f2fs_hash_t name_hash,
> unsigned int bit_pos);
> @@ -3662,7 +3664,7 @@ void f2fs_truncate_inline_inode(struct inode *inode,
> int f2fs_read_inline_data(struct inode *inode, struct page *page);
> int f2fs_convert_inline_page(struct dnode_of_data *dn, struct page *page);
> int f2fs_convert_inline_inode(struct inode *inode);
> -int f2fs_convert_inline_dir(struct inode *dir);
> +int f2fs_try_convert_inline_dir(struct inode *dir, struct dentry *dentry);
> int f2fs_write_inline_data(struct inode *inode, struct page *page);
> bool f2fs_recover_inline_data(struct inode *inode, struct page *npage);
> struct f2fs_dir_entry *f2fs_find_in_inline_dir(struct inode *dir,
> diff --git a/fs/f2fs/inline.c b/fs/f2fs/inline.c
> index f82c3d9cf333..4167e5408151 100644
> --- a/fs/f2fs/inline.c
> +++ b/fs/f2fs/inline.c
> @@ -539,28 +539,40 @@ static int do_convert_inline_dir(struct inode *dir, struct page *ipage,
> return f2fs_move_rehashed_dirents(dir, ipage, inline_dentry);
> }
>
> -int f2fs_convert_inline_dir(struct inode *dir)
> +int f2fs_try_convert_inline_dir(struct inode *dir, struct dentry *dentry)
> {
> struct f2fs_sb_info *sbi = F2FS_I_SB(dir);
> struct page *ipage;
> + struct fscrypt_name fname;
> void *inline_dentry = NULL;
> - int err;
> + int err = 0;
>
> if (!f2fs_has_inline_dentry(dir))
> return 0;
>
> f2fs_lock_op(sbi);
>
> + err = fscrypt_setup_filename(dir, &dentry->d_name, 0, &fname);
> + if (err)
> + goto out;
> +
> ipage = f2fs_get_node_page(sbi, dir->i_ino);
> - if (IS_ERR(ipage))
> - return PTR_ERR(ipage);
> + if (IS_ERR(ipage)) {
> + err = PTR_ERR(ipage);
> + goto out;
> + }
> +
> + if (f2fs_has_enough_room(dir, ipage, &fname)) {
> + f2fs_put_page(ipage, 1);
> + goto out;
> + }
>
> inline_dentry = inline_data_addr(dir, ipage);
>
> err = do_convert_inline_dir(dir, ipage, inline_dentry);
> if (!err)
> f2fs_put_page(ipage, 1);
> -
> +out:
> f2fs_unlock_op(sbi);
> return err;
> }
> diff --git a/fs/f2fs/namei.c b/fs/f2fs/namei.c
> index 1daa54c8d8d8..856f0f984549 100644
> --- a/fs/f2fs/namei.c
> +++ b/fs/f2fs/namei.c
> @@ -927,7 +927,7 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
> * order to avoid this, let's do the check and update here.
> */
> if (old_dir == new_dir && !new_inode) {
> - err = f2fs_convert_inline_dir(old_dir);
> + err = f2fs_try_convert_inline_dir(old_dir, new_dentry);
> if (err)
> return err;
> }
> --
> 2.18.0.rc1
>
> Thanks,
>
> > + if (old_dir == new_dir && !new_inode) {
> > + err = f2fs_convert_inline_dir(old_dir);
> > + if (err)
> > + return err;
> > + }
> > +
> > if (flags & RENAME_WHITEOUT) {
> > err = f2fs_create_whiteout(old_dir, &whiteout);
> > if (err)
> > @@ -954,28 +966,6 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
> >
> > if (old_dir_entry)
> > f2fs_i_links_write(new_dir, true);
> > -
> > - /*
> > - * old entry and new entry can locate in the same inline
> > - * dentry in inode, when attaching new entry in inline dentry,
> > - * it could force inline dentry conversion, after that,
> > - * old_entry and old_page will point to wrong address, in
> > - * order to avoid this, let's do the check and update here.
> > - */
> > - if (is_old_inline && !f2fs_has_inline_dentry(old_dir)) {
> > - f2fs_put_page(old_page, 0);
> > - old_page = NULL;
> > -
> > - old_entry = f2fs_find_entry(old_dir,
> > - &old_dentry->d_name, &old_page);
> > - if (!old_entry) {
> > - err = -ENOENT;
> > - if (IS_ERR(old_page))
> > - err = PTR_ERR(old_page);
> > - f2fs_unlock_op(sbi);
> > - goto out_dir;
> > - }
> > - }
> > }
> >
> > down_write(&F2FS_I(old_inode)->i_sem);
> >