Re: XFS bug with ftruncate, mmap and holes.

[Andrew Morton <akpm@xxxxxxxxxx>]

Ethan Benson wrote:
> 
> someone i was talking to recently has been using the 2.4.17 xfs
> patches with 2.4.18pre kernels and found apt-get failed due to errors
> returned by msync() (which appears to be returning 1 (not -1)).
> msync is involved in writing these apt binary cache files.
> 

Interesting.

In 2.4.18-pre3 I changed msync.  Previously, it was ignoring things
like EIO and ENOSPC on writepage() operations.  The change was to
correctly propagate these error codes all the way back to the
caller, all over the place.

The raw diff is at
http://www.zip.com.au/~akpm/linux/2.4/2.4.18-pre3/msync-ret.patch

If someone is returning `1' from msync() then probably an
underlying filesystem is returning `1' from its get_block()
or writepage() method.  It shouldn't.  It should return
either zero or a negative errno.