VFS: fixes a bug in sys_linkat()
PATCH 1/1 VFS: fixes a bug in sys_linkat()
Hardlink using sys_linkat() returns EXDEV when the source and the destination point to the same filesystem, residing under different mounts.
An example scenario is:
mount /dev/sda /mnt
mount /dev/sdb /mnt1
mkdir /mnt1/src
mkdir /mnt/dest
mount --bind /mnt1/src /mnt/dest (note: at this point the filesystems
under /mnt1/src and /mnt/dest is same)
touch /mnt1/src/testfile
ln /mnt1/src/testfile /mnt/dest/sametestfile (note: this should succeed).
The ln command fails. It does not correctly recognize that both the source and destination point to the same filesystem.
The following patch fixes the problem.
Thanks to Nishanth Aravamudan([email protected]) for identifying the bug.
Signed by Ram Pai ([email protected])
--- fs/namei.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
fs/namei.c +1 -1
@@ -2263,7 +2263,7 @@ asmlinkage long sys_linkat(int olddfd, cif (error) goto out; error = -EXDEV ; -if (old_nd.mnt != nd.mnt) +if (old_nd.mnt->mnt_sb != nd.mnt->mnt_sb) goto out_release; new_dentry =lookup_create (&nd,0 ); error =PTR_ERR (new_dentry);
Quoted 2 lines · Ram Pai
Hardlink using
sys_linkat()returns EXDEV when the source and the destination point to the same filesystem, residing under different mounts.
... and that is absolutely deliberate. Ability to restrict links to a subtree is a deliberate feature.
Quoted 6 lines · Al Viro
Quoted 2 lines · Ram Pai
Hardlink using
sys_linkat()returns EXDEV when the source and the destination point to the same filesystem, residing under different mounts.... and that is absolutely deliberate. Ability to restrict links to a subtree is a deliberate feature.
But wondering what side effects will it have if we allowed it to create hardlinks?
RP