Re: [PATCH 2/2] pty: make sure super_block is still valid in final /dev/tty close

[Peter Hurley]

On 01/14/2016 12:03 PM, Herton R. Krzesinski wrote:
> On Wed, Jan 13, 2016 at 09:54:03AM -0800, Peter Hurley wrote:
>> Hi Herton,
>>
>> On 01/11/2016 06:07 AM, Herton R. Krzesinski wrote:
>>> Considering current pty code and multiple devpts instances, it's possible
>>> to umount a devpts file system while a program still has /dev/tty opened
>>> pointing to a previosuly closed pty pair in that instance. In the case all
>>> ptmx and pts/N files are closed, umount can be done. If the program closes
>>> /dev/tty after umount is done, devpts_kill_index will use now an invalid
>>> super_block, which was already destroyed in the umount operation after
>>> running ->kill_sb. This is another "use after free" type of issue, but now
>>> related to the allocated super_block instance.
>>>
>>> To avoid the problem (warning at ida_remove and potential crashes) for
>>> this specific case, I added two functions in devpts which grabs additional
>>> references to the super_block, which pty code now uses so it makes sure
>>> the super block structure is still valid until pty shutdown is done.
>>> I also moved the additional inode references to the same functions, which
>>> also covered similar case with inode being freed before /dev/tty final
>>> close/shutdown.
>>
>> Thanks for discovering and working this problem.
>> Comments below.
> 
> Thanks for looking/reviewing the patches! :)
> 
>>
>>
>>> Signed-off-by: Herton R. Krzesinski <herton@xxxxxxxxxx>
>>> Cc: stable@xxxxxxxxxxxxxxx # 2.6.29+
>>> ---
>>>  drivers/tty/pty.c         |  9 ++++++---
>>>  fs/devpts/inode.c         | 20 ++++++++++++++++++++
>>>  include/linux/devpts_fs.h |  4 ++++
>>>  3 files changed, 30 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/drivers/tty/pty.c b/drivers/tty/pty.c
>>> index 96016e5..7fc1b3e 100644
>>> --- a/drivers/tty/pty.c
>>> +++ b/drivers/tty/pty.c
>>> @@ -688,7 +688,7 @@ static void pty_unix98_shutdown(struct tty_struct *tty)
>>>  	else
>>>  		ptmx_inode = tty->link->driver_data;
>>>  	devpts_kill_index(ptmx_inode, tty->index);
>>> -	iput(ptmx_inode); /* drop reference we acquired at ptmx_open */
>>> +	devpts_iput_sb_deactive(ptmx_inode);
>>>  }
>>>  
>>>  static const struct tty_operations ptm_unix98_ops = {
>>> @@ -785,9 +785,12 @@ static int ptmx_open(struct inode *inode, struct file *filp)
>>>  	 * still have /dev/tty opened pointing to the master/slave pair (ptmx
>>>  	 * is closed/released before /dev/tty), we must make sure that the inode
>>>  	 * is still valid when we call the final pty_unix98_shutdown, thus we
>>> -	 * hold an additional reference to the ptmx inode
>>> +	 * hold an additional reference to the ptmx inode. For the same /dev/tty
>>> +	 * last close case, we also need to make sure the super_block isn't
>>> +	 * destroyed (devpts instance unmounted), before /dev/tty is closed and
>>> +	 * on its release devpts_kill_index is called.
>>>  	 */
>>> -	ihold(inode);
>>> +	devpts_ihold_sb_active(inode);
>>>  
>>>  	tty_add_file(tty, filp);
>>>  
>>> diff --git a/fs/devpts/inode.c b/fs/devpts/inode.c
>>> index c35ffdc..66a5421 100644
>>> --- a/fs/devpts/inode.c
>>> +++ b/fs/devpts/inode.c
>>> @@ -575,6 +575,26 @@ void devpts_kill_index(struct inode *ptmx_inode, int idx)
>>>  	mutex_unlock(&allocated_ptys_lock);
>>>  }
>>>  
>>> +/*
>>> + * pty code needs to hold extra references in case of last /dev/tty close
>>> + */
>>> +
>>> +void devpts_ihold_sb_active(struct inode *ptmx_inode)
>>> +{
>>> +	struct super_block *sb = pts_sb_from_inode(ptmx_inode);
>>> +
>>> +	atomic_inc(&sb->s_active);
>>> +	ihold(ptmx_inode);
>>> +}
>>> +
>>> +void devpts_iput_sb_deactive(struct inode *ptmx_inode)
>>> +{
>>> +	struct super_block *sb = pts_sb_from_inode(ptmx_inode);
>>> +
>>> +	iput(ptmx_inode);
>>> +	deactivate_super(sb);
>>> +}
>>
>> We might as well roll in this functionality into
>> devpts_new_index() and devpts_kill_index().
>>
>> I realize that's muddying the separation of concern.
>>
>> Alternatively, name the functions for the logical operation
>> rather than specifically for what they do (eg. devpts_add_ref())
> 
> I renamed the functions and submitted a v2. Not sure if you also want the
> move to devpts_new_index()/devpts_kill_index(), I left that as is for now.
> I personally don't have much preference here, I left as separate functions
> which are called at pty.c because that seemed to be "more logical", in that it's
> a tty specific requirement and not tied to devpts fs itself, not sure if it makes
> sense (and anyway devpts is only used by tty anyway). It's mostly a cosmetic/
> different way of doing thigs, but if required or preferred I can do a v3 and
> move/incorporate it into devpts_new_index and devpts_kill_index.

Yeah, that fine. Thanks again for your work on this.

Regards,
Peter Hurley


>>> +
>>>  /**
>>>   * devpts_pty_new -- create a new inode in /dev/pts/
>>>   * @ptmx_inode: inode of the master
>>> diff --git a/include/linux/devpts_fs.h b/include/linux/devpts_fs.h
>>> index 251a209..f73ef49 100644
>>> --- a/include/linux/devpts_fs.h
>>> +++ b/include/linux/devpts_fs.h
>>> @@ -19,6 +19,8 @@
>>>  
>>>  int devpts_new_index(struct inode *ptmx_inode);
>>>  void devpts_kill_index(struct inode *ptmx_inode, int idx);
>>> +void devpts_ihold_sb_active(struct inode *ptmx_inode);
>>> +void devpts_iput_sb_deactive(struct inode *ptmx_inode);
>>>  /* mknod in devpts */
>>>  struct inode *devpts_pty_new(struct inode *ptmx_inode, dev_t device, int index,
>>>  		void *priv);
>>> @@ -32,6 +34,8 @@ void devpts_pty_kill(struct inode *inode);
>>>  /* Dummy stubs in the no-pty case */
>>>  static inline int devpts_new_index(struct inode *ptmx_inode) { return -EINVAL; }
>>>  static inline void devpts_kill_index(struct inode *ptmx_inode, int idx) { }
>>> +static inline void devpts_ihold_sb_active(struct inode *ptmx_inode) { }
>>> +static inline void devpts_iput_sb_deactive(struct inode *ptmx_inode) { }
>>>  static inline struct inode *devpts_pty_new(struct inode *ptmx_inode,
>>>  		dev_t device, int index, void *priv)
>>>  {
>>>
>>
>