[Top] [All Lists]

Re: [PATCH v2] xfs: Make fiemap works with sparse file.

To: Tao Ma <tao.ma@xxxxxxxxxx>
Subject: Re: [PATCH v2] xfs: Make fiemap works with sparse file.
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Fri, 18 Jun 2010 16:22:25 +1000
Cc: xfs@xxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx, Alex Elder <aelder@xxxxxxx>, Christoph Hellwig <hch@xxxxxx>, Eric Sandeen <sandeen@xxxxxxxxxx>
In-reply-to: <4C1AD9A5.8010600@xxxxxxxxxx>
References: <20100614122912.GD6590@dastard> <1276764799-4837-1-git-send-email-tao.ma@xxxxxxxxxx> <20100618004708.GZ6590@dastard> <4C1AD9A5.8010600@xxxxxxxxxx>
User-agent: Mutt/1.5.20 (2009-06-14)
On Fri, Jun 18, 2010 at 10:27:49AM +0800, Tao Ma wrote:
> >>This patch add a new parameter BMV_IF_NO_HOLES for bvm_iflags.
> >>So with this flags, we don't use our 'out' in xfs_getbmap for
> >>a hole. The solution is a bit ugly by just don't increasing
> >>index of 'out' vector. I felt that it is not easy to skip it
> >>at the very beginning since we have the complicated check and
> >>some function like xfs_getbmapx_fix_eof_hole to adjust 'out'.
> >
> >... because I think we can safely skip xfs_getbmapx_fix_eof_hole()
> >it only modifies the hole. Hence  just adding a check after the
> >attribute fork end check (which needs to detect a hole to terminate)
> >should be fine: e.g something like:
> >
> >                     if (map[i].br_startblock == HOLESTARTBLOCK&&
> >                     whichfork == XFS_ATTR_FORK) {
> >                             /* came to the end of attribute fork */
> >                             out[cur_ext].bmv_oflags |= BMV_OF_LAST;
> >                             goto out_free_map;
> >                     }
> >+                    if (map[i].br_startblock == HOLESTARTBLOCK&&
> >+                             (iflags&  BMV_IF_NO_HOLES)) {
> >+                            memset(&out[cur_ext], 0, sizeof(out[cur_ext]));
> >+                            continue;
> >+                    }
> >
> >Should work and avoid the worst of the ugliness.
> I am afraid it doesn't work, at least from my test. It enters a dead loop.
> I think the root cause is that your change doesn't update bmv_offset
> and bmv_length for a hole. So in the large loop,

Yes, you are right - I didn't consider the outer loop termination
properly. My bad.

I'll add your patch to my qa list and move it onwards. Thanks.


Dave Chinner

<Prev in Thread] Current Thread [Next in Thread>