Re: post 2.6.21 regression in F_GETLK

[Doug Chapman <doug.chapman@hp.com>]

On Thu, 2007-05-10 at 15:38 -0400, J. Bruce Fields wrote:
> On Thu, May 10, 2007 at 03:30:50PM -0400, bfields wrote:
> > On Thu, May 10, 2007 at 02:56:15PM -0400, Doug Chapman wrote:
> > > A recent regression (introduced after 2.6.21) was caught by the LTP test
> > > fcntl11.  It appears that F_GETLK is not properly checking for existing
> > > F_RDLCK and allows taking out a write lock.
> > 
> > Ouch.
> > 
> > > This can be demonstrated by either running fcntl11 from the LTP suite or
> > > I have hacked up a much shorter version which demonstrates the issue and
> > > am attaching it.
> > > 
> > > Using git bisect I came up with this commit as the one that introduced
> > > the issue.
> > 
> > Thanks for the report--investigating....
> 
> Argh.  Looks like a cut-n-paste error.  Does this fix it?
> 
> --b.
> 
> diff --git a/fs/locks.c b/fs/locks.c
> index 671a034..909f454 100644
> --- a/fs/locks.c
> +++ b/fs/locks.c
> @@ -1632,6 +1632,7 @@ static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
>  	flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
>  		fl->fl_end - fl->fl_start + 1;
>  	flock->l_whence = 0;
> +	flock->l_type = fl->fl_type;
>  	return 0;
>  }
>  

Bruce,

This doesn't fix the problem but it does look like it should be there.
I imagine this would have been the next bug we tripped over once the
original one is fixed.

- Doug