diff mbox series

[f2fs-dev] f2fs: fix to avoid atomicity corruption of atomic file

Message ID 20250314120651.443184-1-youngjin.gil@samsung.com (mailing list archive)
State Accepted
Commit c60b556bfe0a97b7b2b6a144a97790cad103e084
Headers show
Series [f2fs-dev] f2fs: fix to avoid atomicity corruption of atomic file | expand

Commit Message

Yeongjin Gil March 14, 2025, 12:06 p.m. UTC
In the case of the following call stack for an atomic file,
FI_DIRTY_INODE is set, but FI_ATOMIC_DIRTIED is not subsequently set.

f2fs_file_write_iter
  f2fs_map_blocks
    f2fs_reserve_new_blocks
      inc_valid_block_count
        __mark_inode_dirty(dquot)
          f2fs_dirty_inode

If FI_ATOMIC_DIRTIED is not set, atomic file can encounter corruption
due to a mismatch between old file size and new data.

To resolve this issue, I changed to set FI_ATOMIC_DIRTIED when
FI_DIRTY_INODE is set. This ensures that FI_DIRTY_INODE, which was
previously cleared by the Writeback thread during the commit atomic, is
set and i_size is updated.

Fixes: fccaa81de87e ("f2fs: prevent atomic file from being dirtied before commit")
Reviewed-by: Sungjong Seo <sj1557.seo@samsung.com>
Reviewed-by: Sunmin Jeong <s_min.jeong@samsung.com>
Signed-off-by: Yeongjin Gil <youngjin.gil@samsung.com>
---
 fs/f2fs/inode.c | 4 +---
 fs/f2fs/super.c | 4 ++++
 2 files changed, 5 insertions(+), 3 deletions(-)

Comments

Daeho Jeong March 14, 2025, 7:04 p.m. UTC | #1
On Fri, Mar 14, 2025 at 5:28 AM Yeongjin Gil <youngjin.gil@samsung.com> wrote:
>
> In the case of the following call stack for an atomic file,
> FI_DIRTY_INODE is set, but FI_ATOMIC_DIRTIED is not subsequently set.
>
> f2fs_file_write_iter
>   f2fs_map_blocks
>     f2fs_reserve_new_blocks
>       inc_valid_block_count
>         __mark_inode_dirty(dquot)
>           f2fs_dirty_inode
>
> If FI_ATOMIC_DIRTIED is not set, atomic file can encounter corruption
> due to a mismatch between old file size and new data.
>
> To resolve this issue, I changed to set FI_ATOMIC_DIRTIED when
> FI_DIRTY_INODE is set. This ensures that FI_DIRTY_INODE, which was
> previously cleared by the Writeback thread during the commit atomic, is
> set and i_size is updated.
>
> Fixes: fccaa81de87e ("f2fs: prevent atomic file from being dirtied before commit")
> Reviewed-by: Sungjong Seo <sj1557.seo@samsung.com>
> Reviewed-by: Sunmin Jeong <s_min.jeong@samsung.com>
> Signed-off-by: Yeongjin Gil <youngjin.gil@samsung.com>
> ---
>  fs/f2fs/inode.c | 4 +---
>  fs/f2fs/super.c | 4 ++++
>  2 files changed, 5 insertions(+), 3 deletions(-)
>
> diff --git a/fs/f2fs/inode.c b/fs/f2fs/inode.c
> index aa2f41696a88..83f862578fc8 100644
> --- a/fs/f2fs/inode.c
> +++ b/fs/f2fs/inode.c
> @@ -34,10 +34,8 @@ void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
>         if (f2fs_inode_dirtied(inode, sync))
>                 return;
>
> -       if (f2fs_is_atomic_file(inode)) {
> -               set_inode_flag(inode, FI_ATOMIC_DIRTIED);
> +       if (f2fs_is_atomic_file(inode))
>                 return;
> -       }
>
>         mark_inode_dirty_sync(inode);
>  }
> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
> index 397df271885c..c08d52c6467a 100644
> --- a/fs/f2fs/super.c
> +++ b/fs/f2fs/super.c
> @@ -1534,6 +1534,10 @@ int f2fs_inode_dirtied(struct inode *inode, bool sync)
>                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
>         }
>         spin_unlock(&sbi->inode_lock[DIRTY_META]);
> +
> +       if (!ret && f2fs_is_atomic_file(inode))
> +               set_inode_flag(inode, FI_ATOMIC_DIRTIED);
> +
>         return ret;
>  }
>
> --
> 2.34.1
>
>

Reviewed-by: Daeho Jeong <daehojeong@google.com>

Thanks!

>
> _______________________________________________
> Linux-f2fs-devel mailing list
> Linux-f2fs-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
Jaegeuk Kim March 14, 2025, 9:50 p.m. UTC | #2
On 03/14, Yeongjin Gil wrote:
> In the case of the following call stack for an atomic file,
> FI_DIRTY_INODE is set, but FI_ATOMIC_DIRTIED is not subsequently set.
> 
> f2fs_file_write_iter
>   f2fs_map_blocks
>     f2fs_reserve_new_blocks
>       inc_valid_block_count
>         __mark_inode_dirty(dquot)
>           f2fs_dirty_inode
> 
> If FI_ATOMIC_DIRTIED is not set, atomic file can encounter corruption
> due to a mismatch between old file size and new data.
> 
> To resolve this issue, I changed to set FI_ATOMIC_DIRTIED when
> FI_DIRTY_INODE is set. This ensures that FI_DIRTY_INODE, which was
> previously cleared by the Writeback thread during the commit atomic, is
> set and i_size is updated.
> 
> Fixes: fccaa81de87e ("f2fs: prevent atomic file from being dirtied before commit")
> Reviewed-by: Sungjong Seo <sj1557.seo@samsung.com>
> Reviewed-by: Sunmin Jeong <s_min.jeong@samsung.com>
> Signed-off-by: Yeongjin Gil <youngjin.gil@samsung.com>
> ---
>  fs/f2fs/inode.c | 4 +---
>  fs/f2fs/super.c | 4 ++++
>  2 files changed, 5 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/f2fs/inode.c b/fs/f2fs/inode.c
> index aa2f41696a88..83f862578fc8 100644
> --- a/fs/f2fs/inode.c
> +++ b/fs/f2fs/inode.c
> @@ -34,10 +34,8 @@ void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
>  	if (f2fs_inode_dirtied(inode, sync))
>  		return;
>  
> -	if (f2fs_is_atomic_file(inode)) {
> -		set_inode_flag(inode, FI_ATOMIC_DIRTIED);
> +	if (f2fs_is_atomic_file(inode))
>  		return;
> -	}
>  
>  	mark_inode_dirty_sync(inode);
>  }
> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
> index 397df271885c..c08d52c6467a 100644
> --- a/fs/f2fs/super.c
> +++ b/fs/f2fs/super.c
> @@ -1534,6 +1534,10 @@ int f2fs_inode_dirtied(struct inode *inode, bool sync)
>  		inc_page_count(sbi, F2FS_DIRTY_IMETA);
>  	}
>  	spin_unlock(&sbi->inode_lock[DIRTY_META]);
> +
> +	if (!ret && f2fs_is_atomic_file(inode))
> +		set_inode_flag(inode, FI_ATOMIC_DIRTIED);
> +

I'm not sure what's different here.

Before and after this patch, set_inode_flag(inode, FI_ATOMIC_DIRTIED) is called
only if f2fs_inode_dirtied() returns zero and f2fs_is_atomic_file(inode).

Note, f2fs_mark_inode_dirty_sync() looks the single caller of f2fs_inode_dirtied.


>  	return ret;
>  }
>  
> -- 
> 2.34.1
Jaegeuk Kim March 14, 2025, 10:21 p.m. UTC | #3
On 03/14, Jaegeuk Kim wrote:
> On 03/14, Yeongjin Gil wrote:
> > In the case of the following call stack for an atomic file,
> > FI_DIRTY_INODE is set, but FI_ATOMIC_DIRTIED is not subsequently set.
> > 
> > f2fs_file_write_iter
> >   f2fs_map_blocks
> >     f2fs_reserve_new_blocks
> >       inc_valid_block_count
> >         __mark_inode_dirty(dquot)
> >           f2fs_dirty_inode
> > 
> > If FI_ATOMIC_DIRTIED is not set, atomic file can encounter corruption
> > due to a mismatch between old file size and new data.
> > 
> > To resolve this issue, I changed to set FI_ATOMIC_DIRTIED when
> > FI_DIRTY_INODE is set. This ensures that FI_DIRTY_INODE, which was
> > previously cleared by the Writeback thread during the commit atomic, is
> > set and i_size is updated.
> > 
> > Fixes: fccaa81de87e ("f2fs: prevent atomic file from being dirtied before commit")
> > Reviewed-by: Sungjong Seo <sj1557.seo@samsung.com>
> > Reviewed-by: Sunmin Jeong <s_min.jeong@samsung.com>
> > Signed-off-by: Yeongjin Gil <youngjin.gil@samsung.com>
> > ---
> >  fs/f2fs/inode.c | 4 +---
> >  fs/f2fs/super.c | 4 ++++
> >  2 files changed, 5 insertions(+), 3 deletions(-)
> > 
> > diff --git a/fs/f2fs/inode.c b/fs/f2fs/inode.c
> > index aa2f41696a88..83f862578fc8 100644
> > --- a/fs/f2fs/inode.c
> > +++ b/fs/f2fs/inode.c
> > @@ -34,10 +34,8 @@ void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
> >  	if (f2fs_inode_dirtied(inode, sync))
> >  		return;
> >  
> > -	if (f2fs_is_atomic_file(inode)) {
> > -		set_inode_flag(inode, FI_ATOMIC_DIRTIED);
> > +	if (f2fs_is_atomic_file(inode))
> >  		return;
> > -	}
> >  
> >  	mark_inode_dirty_sync(inode);
> >  }
> > diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
> > index 397df271885c..c08d52c6467a 100644
> > --- a/fs/f2fs/super.c
> > +++ b/fs/f2fs/super.c
> > @@ -1534,6 +1534,10 @@ int f2fs_inode_dirtied(struct inode *inode, bool sync)
> >  		inc_page_count(sbi, F2FS_DIRTY_IMETA);
> >  	}
> >  	spin_unlock(&sbi->inode_lock[DIRTY_META]);
> > +
> > +	if (!ret && f2fs_is_atomic_file(inode))
> > +		set_inode_flag(inode, FI_ATOMIC_DIRTIED);
> > +
> 
> I'm not sure what's different here.
> 
> Before and after this patch, set_inode_flag(inode, FI_ATOMIC_DIRTIED) is called
> only if f2fs_inode_dirtied() returns zero and f2fs_is_atomic_file(inode).
> 
> Note, f2fs_mark_inode_dirty_sync() looks the single caller of f2fs_inode_dirtied.

Ok, I missed another caller, f2fs_dirty_inode, per discussion with Daeho.
Let me apply this.

> 
> 
> >  	return ret;
> >  }
> >  
> > -- 
> > 2.34.1
patchwork-bot+f2fs--- via Linux-f2fs-devel March 14, 2025, 10:30 p.m. UTC | #4
Hello:

This patch was applied to jaegeuk/f2fs.git (dev)
by Jaegeuk Kim <jaegeuk@kernel.org>:

On Fri, 14 Mar 2025 21:06:51 +0900 you wrote:
> In the case of the following call stack for an atomic file,
> FI_DIRTY_INODE is set, but FI_ATOMIC_DIRTIED is not subsequently set.
> 
> f2fs_file_write_iter
>   f2fs_map_blocks
>     f2fs_reserve_new_blocks
>       inc_valid_block_count
>         __mark_inode_dirty(dquot)
>           f2fs_dirty_inode
> 
> [...]

Here is the summary with links:
  - [f2fs-dev] f2fs: fix to avoid atomicity corruption of atomic file
    https://git.kernel.org/jaegeuk/f2fs/c/c60b556bfe0a

You are awesome, thank you!
Chao Yu March 17, 2025, 3:19 a.m. UTC | #5
On 3/14/25 20:06, Yeongjin Gil wrote:
> In the case of the following call stack for an atomic file,
> FI_DIRTY_INODE is set, but FI_ATOMIC_DIRTIED is not subsequently set.
> 
> f2fs_file_write_iter
>   f2fs_map_blocks
>     f2fs_reserve_new_blocks
>       inc_valid_block_count
>         __mark_inode_dirty(dquot)
>           f2fs_dirty_inode
> 
> If FI_ATOMIC_DIRTIED is not set, atomic file can encounter corruption
> due to a mismatch between old file size and new data.
> 
> To resolve this issue, I changed to set FI_ATOMIC_DIRTIED when
> FI_DIRTY_INODE is set. This ensures that FI_DIRTY_INODE, which was
> previously cleared by the Writeback thread during the commit atomic, is
> set and i_size is updated.

I guess we need to add a regression testcase to cover this issue.

> 
> Fixes: fccaa81de87e ("f2fs: prevent atomic file from being dirtied before commit")
> Reviewed-by: Sungjong Seo <sj1557.seo@samsung.com>
> Reviewed-by: Sunmin Jeong <s_min.jeong@samsung.com>
> Signed-off-by: Yeongjin Gil <youngjin.gil@samsung.com>

Reviewed-by: Chao Yu <chao@kernel.org>

Thanks,
diff mbox series

Patch

diff --git a/fs/f2fs/inode.c b/fs/f2fs/inode.c
index aa2f41696a88..83f862578fc8 100644
--- a/fs/f2fs/inode.c
+++ b/fs/f2fs/inode.c
@@ -34,10 +34,8 @@  void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
 	if (f2fs_inode_dirtied(inode, sync))
 		return;
 
-	if (f2fs_is_atomic_file(inode)) {
-		set_inode_flag(inode, FI_ATOMIC_DIRTIED);
+	if (f2fs_is_atomic_file(inode))
 		return;
-	}
 
 	mark_inode_dirty_sync(inode);
 }
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 397df271885c..c08d52c6467a 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -1534,6 +1534,10 @@  int f2fs_inode_dirtied(struct inode *inode, bool sync)
 		inc_page_count(sbi, F2FS_DIRTY_IMETA);
 	}
 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
+
+	if (!ret && f2fs_is_atomic_file(inode))
+		set_inode_flag(inode, FI_ATOMIC_DIRTIED);
+
 	return ret;
 }