XFS warning in 2.6.18-rc4

XFS warning in 2.6.18-rc4

[Meelis Roos]

fs/xfs/xfs_bmap.c: In function 'xfs_bmapi':
fs/xfs/xfs_bmap.c:2662: warning: 'rtx' is used uninitialized in this function

-- 
Meelis Roos (mroos@linux.ee)

Re: XFS warning in 2.6.18-rc4

[Nathan Scott]

On Wed, Aug 09, 2006 at 11:04:53PM +0300, Meelis Roos wrote:
> fs/xfs/xfs_bmap.c: In function 'xfs_bmapi':
> fs/xfs/xfs_bmap.c:2662: warning: 'rtx' is used uninitialized in this function

You have a particularly dense compiler, unfortunately.  This code
has always been this way, its just a false cc warning that can be
safely ignored until you upgrade to a fixed compiler (unless I'm
missing something - please enlighten me if so).  It does seem to
be the case that there is no way 'rtx' will be used uninitialised
when xfs_rtpick_extent() doesn't fail... no?

cheers.

-- 
Nathan

Re: XFS warning in 2.6.18-rc4

[Alexey Dobriyan]

On Wed, Aug 09, 2006 at 11:04:53PM +0300, Meelis Roos wrote:
> fs/xfs/xfs_bmap.c: In function 'xfs_bmapi':
> fs/xfs/xfs_bmap.c:2662: warning: 'rtx' is used uninitialized in this function

gcc bug again? Corresponding preprocessed source when XFS realtime allocator
is off:

static int xfs_bmap_rtalloc(xfs_bmalloca_t *ap)
{
		...
	xfs_rtblock_t rtx;

		...
	if (ap->eof && ap->off == 0) {
===>		error = (251);		<===
		if (error)
			return error;
		ap->rval = rtx * mp->m_sb.sb_rextsize;
	} else {
		ap->rval = 0;
	}

This is the only place where rtx is "used".

When realtime allocator is on, xfs_rtpick_extent() is real function
either a) returning error and leaving garbage in "rtx", or b)
initializing "rtx" and returning 0.

Re: XFS warning in 2.6.18-rc4

[Jesper Juhl]

On 10/08/06, Nathan Scott <nathans@sgi.com> wrote:
> On Wed, Aug 09, 2006 at 11:04:53PM +0300, Meelis Roos wrote:
> > fs/xfs/xfs_bmap.c: In function 'xfs_bmapi':
> > fs/xfs/xfs_bmap.c:2662: warning: 'rtx' is used uninitialized in this function
>
> You have a particularly dense compiler, unfortunately.  This code
> has always been this way, its just a false cc warning that can be
> safely ignored until you upgrade to a fixed compiler (unless I'm
> missing something - please enlighten me if so).  It does seem to
> be the case that there is no way 'rtx' will be used uninitialised
> when xfs_rtpick_extent() doesn't fail... no?
>
Ok, I may be reading something wrong here, but I think the warning is
actually not correct.

In fs/xfs/xfs_rtalloc.c::xfs_rtpick_extent() there is this code :

...
        if ((error = xfs_trans_iget(mp, tp, mp->m_sb.sb_rbmino, 0,
                                        XFS_ILOCK_EXCL, &ip)))
                return error;
...
before anything even touches 'pick' (which is the last argument of
type "xfs_rtblock_t  *" which is what "&rtx" is being passed to that
function as, from fs/xfs/xfs_bmap.c::xfs_bmap_rtalloc() at line 2659
before 'rtx' then used at line 2662.

Look at this block of code from fs/xfs/xfs_bmap.c::xfs_bmap_rtalloc()  :

...
 2658:       if (ap->eof && ap->off == 0) {
 2659:                error = xfs_rtpick_extent(mp, ap->tp, ralen, &rtx);
 2660:                if (error)
 2661:                        return error;
 2662:                ap->rval = rtx * mp->m_sb.sb_rextsize;
 2663:        } else {
 2664:                ap->rval = 0;
 2665:        }
..

'rtx' has not been initialized in this function before this block of code.
If the call to xfs_rtpick_extent() should happen to fail from the
check quoted above, then 'rtx' will in fact be used uninitialized in
line 2662 - except that will never happen, because if the check in In
fs/xfs/xfs_rtalloc.c::xfs_rtpick_extent() fails then error will be !=
0 and then well never get down to line 2662.

Or am I missing something ?


PS. Above line numbers etc are from 2.6.18-rc4.


-- 
Jesper Juhl <jesper.juhl@gmail.com>
Don't top-post  http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please      http://www.expita.com/nomime.html

Re: XFS warning in 2.6.18-rc4

[Nathan Scott]

On Fri, Aug 11, 2006 at 01:47:43AM +0200, Jesper Juhl wrote:
> On 10/08/06, Nathan Scott <nathans@sgi.com> wrote:
> > On Wed, Aug 09, 2006 at 11:04:53PM +0300, Meelis Roos wrote:
> > > fs/xfs/xfs_bmap.c: In function 'xfs_bmapi':
> > > fs/xfs/xfs_bmap.c:2662: warning: 'rtx' is used uninitialized in this function
> >
> > You have a particularly dense compiler, unfortunately.  This code
> > has always been this way, its just a false cc warning that can be
> > safely ignored until you upgrade to a fixed compiler (unless I'm
> > missing something - please enlighten me if so).  It does seem to
> > be the case that there is no way 'rtx' will be used uninitialised
> > when xfs_rtpick_extent() doesn't fail... no?
> >
> Ok, I may be reading something wrong here, but I think the warning is
> actually not correct.

Thats how I read it too.  By "dense" I meant "buggy".

> Or am I missing something ?

Nope, thats my understanding too.  The compiler is wrong in this case.
The only open issue I guess is whether its worh rearranging the code
to stop people reporting it as a problem...  *shrug*.

cheers.

-- 
Nathan

Re: XFS warning in 2.6.18-rc4

[Jesper Juhl]

On 11/08/06, Nathan Scott <nathans@sgi.com> wrote:
> On Fri, Aug 11, 2006 at 01:47:43AM +0200, Jesper Juhl wrote:
> > On 10/08/06, Nathan Scott <nathans@sgi.com> wrote:
> > > On Wed, Aug 09, 2006 at 11:04:53PM +0300, Meelis Roos wrote:
> > > > fs/xfs/xfs_bmap.c: In function 'xfs_bmapi':
> > > > fs/xfs/xfs_bmap.c:2662: warning: 'rtx' is used uninitialized in this function
> > >
> > > You have a particularly dense compiler, unfortunately.  This code
> > > has always been this way, its just a false cc warning that can be
> > > safely ignored until you upgrade to a fixed compiler (unless I'm
> > > missing something - please enlighten me if so).  It does seem to
> > > be the case that there is no way 'rtx' will be used uninitialised
> > > when xfs_rtpick_extent() doesn't fail... no?
> > >
> > Ok, I may be reading something wrong here, but I think the warning is
> > actually not correct.
>
> Thats how I read it too.  By "dense" I meant "buggy".
>
> > Or am I missing something ?
>
> Nope, thats my understanding too.  The compiler is wrong in this case.
> The only open issue I guess is whether its worh rearranging the code
> to stop people reporting it as a problem...  *shrug*.
>
I don't see how it would hurt ;)

-- 
Jesper Juhl <jesper.juhl@gmail.com>
Don't top-post  http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please      http://www.expita.com/nomime.html

Re: XFS warning in 2.6.18-rc4

[David Chinner]

On Fri, Aug 11, 2006 at 09:52:10AM +1000, Nathan Scott wrote:
> On Fri, Aug 11, 2006 at 01:47:43AM +0200, Jesper Juhl wrote:
> > On 10/08/06, Nathan Scott <nathans@sgi.com> wrote:
> > > On Wed, Aug 09, 2006 at 11:04:53PM +0300, Meelis Roos wrote:
> > > > fs/xfs/xfs_bmap.c: In function 'xfs_bmapi':
> > > > fs/xfs/xfs_bmap.c:2662: warning: 'rtx' is used uninitialized in this function
> > >
> > > You have a particularly dense compiler, unfortunately.  This code
> > > has always been this way, its just a false cc warning that can be
> > > safely ignored until you upgrade to a fixed compiler (unless I'm
> > > missing something - please enlighten me if so).  It does seem to
> > > be the case that there is no way 'rtx' will be used uninitialised
> > > when xfs_rtpick_extent() doesn't fail... no?
> > >
> > Ok, I may be reading something wrong here, but I think the warning is
> > actually not correct.
> 
> Thats how I read it too.  By "dense" I meant "buggy".
> 
> > Or am I missing something ?
> 
> Nope, thats my understanding too.  The compiler is wrong in this case.
> The only open issue I guess is whether its worh rearranging the code
> to stop people reporting it as a problem...  *shrug*.

No, the compiler needs to be fixed. There are other warnings caused by code
exactly like this that have not been fixed because it is an incorrect warning
(e.g. "idx" in bio_alloc_bioset()).

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group