[Top] [All Lists]

Re: [PATCH] Introduce SEEK_DATA/SEEK_HOLE support V8

To: jeff.liu@xxxxxxxxxx
Subject: Re: [PATCH] Introduce SEEK_DATA/SEEK_HOLE support V8
From: Mark Tinguely <tinguely@xxxxxxx>
Date: Wed, 09 May 2012 12:42:57 -0500
Cc: xfs@xxxxxxxxxxx, Christoph Hellwig <hch@xxxxxxxxxxxxx>, Ben Myers <bpm@xxxxxxx>, Chris Mason <chris.mason@xxxxxxxxxx>
In-reply-to: <4F3E532E.6000708@xxxxxxxxxx>
References: <4F3E532E.6000708@xxxxxxxxxx>
User-agent: Mozilla/5.0 (X11; FreeBSD amd64; rv:9.0) Gecko/20120122 Thunderbird/9.0
On 02/17/12 07:16, Jeff Liu wrote:

This is the revised patch according to Dave's comments for V7.

Changes to V8:
1. If there is an internal error raised at extent reading routine, just
return it rather than ENXIO.
2. Add the commit message.
3. Remove the for(;;) loop since there is no continuous holes shown even
if create a Petabyte sparse file with hole extent length longer than
32-bit.  Thanks Dave for helping verify that!
4. In xfs_seek_data(), s/len/end/, looks 'end' is more meaningful here
to indicate the range of extents mapped.
5. Remove BUG() from xfs_seek_data() since xfs_bmapi_read() have found
any corruption during the lookup, it should not occurred at all.

Any comments are appreciated!


Signed-off-by: Jie Liu<jeff.liu@xxxxxxxxxx>
  This patch adds lseek(2) SEEK_DATA/SEEK_HOLE functionality.

+STATIC loff_t
+       struct file             *file,
+       loff_t                  start,
+       u32                     type)
+       struct inode            *inode = file->f_mapping->host;
+       struct xfs_inode        *ip = XFS_I(inode);
+       struct xfs_mount        *mp = ip->i_mount;
+       loff_t                  uninitialized_var(offset);
+       loff_t                  holeoff;
+       xfs_fsize_t             isize;
+       xfs_fileoff_t           fsbno;
+       uint                    lock;
+       int                     error;
+       lock = xfs_ilock_map_shared(ip);
+       isize = i_size_read(inode);
+       if (start>= isize) {
+               error = ENXIO;
+               goto out_unlock;
+       }
+       fsbno = XFS_B_TO_FSBT(mp, start);
+       error = xfs_bmap_first_unused(NULL, ip, 1,&fsbno, XFS_DATA_FORK);
+       if (error)
+               goto out_unlock;
+       holeoff = XFS_FSB_TO_B(mp, fsbno);
+       if (holeoff<= start)
+               offset = start;
+       else
+               offset = min_t(loff_t, holeoff, isize);
+       if (offset != file->f_pos)
+               file->f_pos = offset;
+       xfs_iunlock_map_shared(ip, lock);
+       if (error)
+               return -error;
+       return offset;

Sorry this got set aside. This is being dusted off for inclusion.

Refresh my memory, do we need a XFS_FORCED_SHUTDOWN test in
xfs_seek_hole()? In the earlier versions it was covered in the

A comment on the min_t() that the xfs_bmap_first_unused() could
return a value bigger than the isize if there are no more holes
may help the forgetful such as myself.

Otherwise it looks good.


Mark Tinguely <tinguely@xxxxxxx>

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