Richard W.M. Jones wrote:
On Mon, Apr 27, 2009 at 03:42:32PM -0500, Eric Sandeen wrote:
> If you didn't want to fall back to the slow behavior on ext3, you might
> consider only using fallocate() if it's there, rather than
> posix_fallocate().
Does this imply there's a fast way to allocate a non-sparse
zero-filled file on ext3?!?
No, there is not... ext3 has no mechanism to track the "uninitialized"
metadata about the block range.
What I meant was (sorry, I seem to be having trouble being clear) is
that if you don't want the slow behavior on ext3, you can use
fallocate(), which will simply -fail- on ext3, and you get no
preallocation at all in that case. For some applications, "fast
prealloc if I can do it, no prealloc if I can't do it fast" may be ok.
-Eric
Rich.