Re: [PATCH -V14 05/11] vfs: Support null pathname in readlink

["Aneesh Kumar K. V" <aneesh.kumar@linux.vnet.ibm.com>]

On Thu, 8 Jul 2010 01:27:06 +1000, Nick Piggin <npiggin@suse.de> wrote:
> On Tue, Jun 15, 2010 at 10:42:55PM +0530, Aneesh Kumar K.V wrote:
> > From: NeilBrown <neilb@suse.de>
> > 
> > This enables to use readlink to get the link target name
> > from a file descriptor point to the link. This can be used
> > with open_by_handle syscall that returns a file descriptor for a link.
> > We can then use this file descriptor to get the target name.
> > 
> > Signed-off-by: NeilBrown <neilb@suse.de>
> > Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
> > ---
> >  fs/stat.c |   30 ++++++++++++++++++++++--------
> >  1 files changed, 22 insertions(+), 8 deletions(-)
> > 
> > diff --git a/fs/stat.c b/fs/stat.c
> > index c4ecd52..49b95a7 100644
> > --- a/fs/stat.c
> > +++ b/fs/stat.c
> > @@ -284,26 +284,40 @@ SYSCALL_DEFINE2(newfstat, unsigned int, fd, struct stat __user *, statbuf)
> >  SYSCALL_DEFINE4(readlinkat, int, dfd, const char __user *, pathname,
> >  		char __user *, buf, int, bufsiz)
> >  {
> > -	struct path path;
> > -	int error;
> > +	int error = 0;
> > +	struct path path, *pp;
> > +	struct file *file = NULL;
> >  
> >  	if (bufsiz <= 0)
> >  		return -EINVAL;
> >  
> > -	error = user_path_at(dfd, pathname, 0, &path);
> > +	if (pathname == NULL && dfd != AT_FDCWD) {
> > +		file = fget(dfd);
> > +
> > +		if (file)
> > +			pp = &file->f_path;
> > +		else
> > +			error = -EBADF;
> > +	} else {
> > +		error = user_path_at(dfd, pathname, 0, &path);
> > +		pp = &path;
> > +	}
> 
> This (and all the others) is really ugly overloading of syscall
> arguments IMO, and the changelog is seriously lacking for such
> changes.

Initially we had  freadlink

http://lkml.org/lkml/2010/5/12/222

We updated the patches to use the existing readlinkat interface because
utimensat(2) already exposed a similar interface. So it should be ok to
expect that other *at call behaved in a similar way ?

> 
> This also changes the the syscall API of existing calls; from reading
> the path at NULL, to switching to a completely different syscall.
> Perhaps you're assuming nobody relies on SIGSEGV / mmapped NULL address
> there, but even then you surely need to document the changed semantics
> somewhere (and document the new syscall semantics properly).


Yes this would need a documentation update. But i guess since we already
have utimensat(2) behaving similarly we are ok to extent readlinkat,
linkat and faccessat on similar lines ?

-aneesh