When XFS creates a new symlink, it writes its size to disk but not to the
VFS inode. This causes i_size_read() to return 0 for that symlink until
it is re-read from disk, for example when the system is rebooted.
I found this inconsistency while protecting directories with eCryptFS.
The command "stat path/to/symlink/in/ecryptfs" will report "Size: 0" if
the symlink was created after the last reboot on an XFS root.
Call i_size_write() in xfs_symlink()
Signed-off-by: Jeffrey Mitchell <[email protected]>
---
fs/xfs/xfs_symlink.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/fs/xfs/xfs_symlink.c b/fs/xfs/xfs_symlink.c
index 1f43fd7f3209..c835827ae389 100644
--- a/fs/xfs/xfs_symlink.c
+++ b/fs/xfs/xfs_symlink.c
@@ -300,6 +300,7 @@ xfs_symlink(
}
ASSERT(pathlen == 0);
}
+ i_size_write(VFS_I(ip), ip->i_d.di_size);
/*
* Create the directory entry for the symlink.
--
2.25.1
On Thu, Jan 21, 2021 at 09:19:12AM -0600, Jeffrey Mitchell wrote:
> When XFS creates a new symlink, it writes its size to disk but not to the
> VFS inode. This causes i_size_read() to return 0 for that symlink until
> it is re-read from disk, for example when the system is rebooted.
>
> I found this inconsistency while protecting directories with eCryptFS.
Heh. No doubt caused by the fact that we only call i_size_write for
regular files, ecryptfs copies i_size_read() of the lower dentry into
the ecryptfs inode, and ecryptfs_getattr_link and doesn't call ->getattr
on the lower inode.
Do directories have the same problem? I'm guessing no because every
other file type calls vfs_getattr on the lower dentry, which should link
us to xfs_vn_getattr.
Reviewed-by: Darrick J. Wong <[email protected]>
--D
> The command "stat path/to/symlink/in/ecryptfs" will report "Size: 0" if
> the symlink was created after the last reboot on an XFS root.
>
> Call i_size_write() in xfs_symlink()
>
> Signed-off-by: Jeffrey Mitchell <[email protected]>
> ---
> fs/xfs/xfs_symlink.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/fs/xfs/xfs_symlink.c b/fs/xfs/xfs_symlink.c
> index 1f43fd7f3209..c835827ae389 100644
> --- a/fs/xfs/xfs_symlink.c
> +++ b/fs/xfs/xfs_symlink.c
> @@ -300,6 +300,7 @@ xfs_symlink(
> }
> ASSERT(pathlen == 0);
> }
> + i_size_write(VFS_I(ip), ip->i_d.di_size);
>
> /*
> * Create the directory entry for the symlink.
> --
> 2.25.1
>
On Thu, Jan 21, 2021 at 10:41:37AM -0800, Darrick J. Wong wrote:
> Do directories have the same problem?
Yes, I just checked in a VM. While ecryptfs does call vfs_getattr(), it
only uses the "blocks" value to supplement the data from an identical
generic_fillattr() call to what ecryptfs_getattr_link() uses. The reported
size still comes from i_size_read().
V/R,
Jeffrey Mitchell
On Thu, Jan 21, 2021 at 09:19:12AM -0600, Jeffrey Mitchell wrote:
> When XFS creates a new symlink, it writes its size to disk but not to the
> VFS inode. This causes i_size_read() to return 0 for that symlink until
> it is re-read from disk, for example when the system is rebooted.
>
> I found this inconsistency while protecting directories with eCryptFS.
> The command "stat path/to/symlink/in/ecryptfs" will report "Size: 0" if
> the symlink was created after the last reboot on an XFS root.
>
> Call i_size_write() in xfs_symlink()
>
> Signed-off-by: Jeffrey Mitchell <[email protected]>
> ---
Reviewed-by: Brian Foster <[email protected]>
> fs/xfs/xfs_symlink.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/fs/xfs/xfs_symlink.c b/fs/xfs/xfs_symlink.c
> index 1f43fd7f3209..c835827ae389 100644
> --- a/fs/xfs/xfs_symlink.c
> +++ b/fs/xfs/xfs_symlink.c
> @@ -300,6 +300,7 @@ xfs_symlink(
> }
> ASSERT(pathlen == 0);
> }
> + i_size_write(VFS_I(ip), ip->i_d.di_size);
>
> /*
> * Create the directory entry for the symlink.
> --
> 2.25.1
>
Looks good,
Reviewed-by: Christoph Hellwig <[email protected]>