Hi,
On Tue, May 07, 2019 at 12:41:48PM +0800, hongjiefang wrote:
> if the directory had been removed, should not set policy for it.
>
> Signed-off-by: hongjiefang <hongjiefang@xxxxxxxxxxxx>
Can you explain the motivation for this change? It makes some sense, but I
don't see why it's really needed. If you look at all the other IS_DEADDIR()
checks in the kernel, they're not for operations on the directory inode itself,
but rather for creating/finding/listing entries in the directory. I think
FS_IOC_SET_ENCRYPTION_POLICY is more like the former (though it does have to
check whether the directory is empty).
> ---
> fs/crypto/policy.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/fs/crypto/policy.c b/fs/crypto/policy.c
> index bd7eaf9..82900a4 100644
> --- a/fs/crypto/policy.c
> +++ b/fs/crypto/policy.c
> @@ -77,6 +77,12 @@ int fscrypt_ioctl_set_policy(struct file *filp, const void __user *arg)
>
> inode_lock(inode);
>
> + /* don't set policy for a dead directory */
> + if (IS_DEADDIR(inode)) {
> + ret = -ENOENT;
> + goto deaddir_out;
> + }
> +
> ret = inode->i_sb->s_cop->get_context(inode, &ctx, sizeof(ctx));
> if (ret == -ENODATA) {
> if (!S_ISDIR(inode->i_mode))
> @@ -96,6 +102,7 @@ int fscrypt_ioctl_set_policy(struct file *filp, const void __user *arg)
> ret = -EEXIST;
> }
>
> +deaddir_out:
> inode_unlock(inode);
Call this label 'out_unlock' instead?
>
> mnt_drop_write_file(filp);
> --
> 1.9.1
>
Thanks,
- Eric
|