Re: [PATCH 1/4] xfs: fix bogus minleft manipulations

[Christoph Hellwig <hch@lst.de>]

On Fri, Dec 16, 2016 at 09:09:38AM +1100, Dave Chinner wrote:
> > @@ -3812,7 +3812,6 @@ xfs_bmap_btalloc(
> >  		args.fsbno = 0;
> >  		args.type = XFS_ALLOCTYPE_FIRST_AG;
> >  		args.total = ap->minlen;
> > -		args.minleft = 0;
> >  		if ((error = xfs_alloc_vextent(&args)))
> >  			return error;
> >  		ap->dfops->dop_low = true;
> 
> But this looks wrong. when combined with the next patch that sets:
> 
> 	args.minleft = xfs_bmap_worst_indlen(ap->ip, args.maxlen);
> 
> we've got a minleft value that might be for multigigabyte
> allocation, but now we are only asking for a total allocation
> of minlen, and that might be 1 block. IOWs, shouldn't this "last,
> final attempt" code actually do something like this:
> 
> 		args.maxlen = ap->minlen;
> 		args.total = ap->minlen;
> 		args.minleft = xfs_bmap_worst_indlen(ap->ip, args.maxlen);

Yes.  And I actually had that in a previous iteration, and it got
dropped at some point.