Re: [PATCH] mm/shmem: Fix invalid PTR_ERR(NULL) call in shmem_xattr_handler_set()

[Qasim Ijaz]

On Tue, Jan 28, 2025 at 08:37:51PM -0800, Hugh Dickins wrote:
> On Tue, 28 Jan 2025, Qasim Ijaz wrote:
> 
> > In shmem_xattr_handler_set() if simple_xattr_set() succeeds and the pointer
> > returned is not an error pointer, then old_xattr will be set to NULL in
> > the body of the following if statement:
> > 
> >     if (!IS_ERR(old_xattr))
> > 
> > Later on shmem_xattr_handler_set() calls:
> > 
> >     return PTR_ERR(old_xattr);
> > 
> > The PTR_ERR macro is used to extract an error code from an error pointer
> > and NULL is not an error pointer, PTR_ERR(NULL) simply results in 0.
> 
> NULL pointer returns 0 for success: yes, that's what's wanted there.
> 
> > 
> > To improve correctness and readability,
> 
> Correctness? Please explain - I don't see any incorrectness.
> 
> Readability? Perhaps - I should not be the judge of that.
> 
> > refactor the error handling 
> > to have an explicit default return value of 0 (success) in "ret". 
> > If simple_xattr_set() returns an error pointer, store its 
> > error code in "ret".
> > 
> > Signed-off-by: Qasim Ijaz <qasdev00@xxxxxxxxx>
> > ---
> >  mm/shmem.c | 9 +++++++--
> >  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> I prefer how it was written before.

Hi Hugh,

By correctness I mean that NULL is not an error pointer, here are some additional points:

The code goes against the documentation of PTR_ERR() which mentions the argument to PTR_ERR() should be an error pointer. NULL is not an error pointer so I think it is inconsistent in that regard. 

Secondly the current code is hinging on a side effect of the PTR_ERR() API being passed NULL, sure it works but I think it would be better if the API was used clearly and consistently with how it is meant to be used.

Also to add onto my points I decided to run the mm/shmem.c file through smatch (since one of its key features is detecting 0 being passed to PTR_ERR()) to see if it detects what I spotted and it brings up the following message:

mm/shmem.c:4174 shmem_xattr_handler_set() warn: passing zero to 'PTR_ERR'

Best regards,  
Qasim

> 
> > 
> > diff --git a/mm/shmem.c b/mm/shmem.c
> > index 532afd8e049c..3e97c7890aac 100644
> > --- a/mm/shmem.c
> > +++ b/mm/shmem.c
> > @@ -4143,6 +4143,7 @@ static int shmem_xattr_handler_set(const struct xattr_handler *handler,
> >  	struct shmem_sb_info *sbinfo = SHMEM_SB(inode->i_sb);
> >  	struct simple_xattr *old_xattr;
> >  	size_t ispace = 0;
> > +	int ret = 0;
> >  
> >  	name = xattr_full_name(handler, name);
> >  	if (value && sbinfo->max_inodes) {
> > @@ -4158,7 +4156,9 @@ static int shmem_xattr_handler_set(const struct xattr_handler *handler,
> >  	}
> >  
> >  	old_xattr = simple_xattr_set(&info->xattrs, name, value, size, flags);
> > -	if (!IS_ERR(old_xattr)) {
> > +	if (IS_ERR(old_xattr)) {
> > +		ret = PTR_ERR(old_xattr);
> > +	} else {
> >  		ispace = 0;
> >  		if (old_xattr && sbinfo->max_inodes)
> >  			ispace = simple_xattr_space(old_xattr->name,
> > @@ -4168,12 +4171,13 @@ static int shmem_xattr_handler_set(const struct xattr_handler *handler,
> >  		inode_set_ctime_current(inode);
> >  		inode_inc_iversion(inode);
> >  	}
> > +
> >  	if (ispace) {
> >  		raw_spin_lock(&sbinfo->stat_lock);
> >  		sbinfo->free_ispace += ispace;
> >  		raw_spin_unlock(&sbinfo->stat_lock);
> >  	}
> > -	return PTR_ERR(old_xattr);
> > +	return ret;
> >  }
> >  
> >  static const struct xattr_handler shmem_security_xattr_handler = {
> > -- 
> > 2.39.5