From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner@vger.kernel.org>
MIME-Version: 1.0
In-Reply-To: <CAOQ4uxjKJHnM_ad9TnyQ8i0fUjEqCEG8ocPAs_L0oSraKhSEPA@mail.gmail.com>
References: <20180412150826.20988-1-mszeredi@redhat.com> <20180412150826.20988-20-mszeredi@redhat.com>
 <CAOQ4uxjKJHnM_ad9TnyQ8i0fUjEqCEG8ocPAs_L0oSraKhSEPA@mail.gmail.com>
From: Amir Goldstein <amir73il@gmail.com>
Date: Wed, 18 Apr 2018 11:39:57 +0300
Message-ID: <CAOQ4uxgMfrA=-cgdoJ3QQe0oTh4-W+gyPBO+nx+rukaaJNd=kA@mail.gmail.com>
Subject: Re: [RFC PATCH 19/35] ovl: readd reflink/copyfile/dedup support
To: Miklos Szeredi <mszeredi@redhat.com>
Cc: overlayfs <linux-unionfs@vger.kernel.org>,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>,
        linux-kernel <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-fsdevel.vger.kernel.org>

On Tue, Apr 17, 2018 at 11:31 PM, Amir Goldstein <amir73il@gmail.com> wrote:
> On Thu, Apr 12, 2018 at 6:08 PM, Miklos Szeredi <mszeredi@redhat.com> wrote:
>> Since set of arguments are so similar, handle in a common helper.
>>
>> Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
>> ---
>>  fs/overlayfs/file.c | 79 +++++++++++++++++++++++++++++++++++++++++++++++++++++
>>  1 file changed, 79 insertions(+)
>>
>> diff --git a/fs/overlayfs/file.c b/fs/overlayfs/file.c
>> index 9670e160967e..39b1b73334ad 100644
>> --- a/fs/overlayfs/file.c
>> +++ b/fs/overlayfs/file.c
>> @@ -352,6 +352,81 @@ long ovl_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
>>         return ret;
>>  }
>>
>> +enum ovl_copyop {
>> +       OVL_COPY,
>> +       OVL_CLONE,
>> +       OVL_DEDUPE,
>> +};
>> +
>> +static ssize_t ovl_copyfile(struct file *file_in, loff_t pos_in,
>> +                           struct file *file_out, loff_t pos_out,
>> +                           u64 len, unsigned int flags, enum ovl_copyop op)
>> +{
>> +       struct inode *inode_out = file_inode(file_out);
>> +       struct fd real_in, real_out;
>> +       const struct cred *old_cred;
>> +       int ret;
>> +
>> +       ret = ovl_real_file(file_out, &real_out);
>> +       if (ret)
>> +               return ret;
>> +
>> +       ret = ovl_real_file(file_in, &real_in);
>> +       if (ret) {
>> +               fdput(real_out);
>> +               return ret;
>> +       }
>> +
>> +       old_cred = ovl_override_creds(file_inode(file_out)->i_sb);
>> +       switch (op) {
>> +       case OVL_COPY:
>> +               ret = vfs_copy_file_range(real_in.file, pos_in,
>> +                                         real_out.file, pos_out, len, flags);
>
> Problem:
> vfs_copy_file_range(ovl_lower_file, ovl_upper_file) on non samefs
> will get -EXDEV from  ovl_copy_file_range(), so will not fall back
> to do_splice_direct().
> We may be better off checking in_sb != out_sb and returning
> -EOPNOTSUPP? not sure.
>
>
>> +               break;
>> +
>> +       case OVL_CLONE:
>> +               ret = vfs_clone_file_range(real_in.file, pos_in,
>> +                                          real_out.file, pos_out, len);
>> +               break;
>> +
>> +       case OVL_DEDUPE:
>> +               ret = vfs_dedupe_file_range_one(real_in.file, pos_in, len,
>> +                                               real_out.file, pos_out);
>
> Problem:
> real_out can be a readonly fd (for is_admin), so we will be deduping
> the lower file.
> I guess this problem is mitigated in current code by may_write_real().
>
> How can we deal with that sort of "write leak" without patching
>  mnt_want_write_file()?
>

Possible solution:
Add interface file_oprations->permission().

At least in rw_verify_area() and clone_verify_area() it is clear
how this would be used. Instead if calling security_file_permission()
call it via a helper file_permission() like with inode_permission.

My understanding in the VFS permission checks model is too
limited to say if this makes sense.

Thanks,
Amir.