On Tue, 19 May 2015, Josef Bacik wrote:
If we fallocate past i_size with KEEP_SIZE, extend the file to use some but not
all of this space, and then truncate(i_size) we won't trim the excess
preallocated space. We decided at LSF that we want to truncate the fallocated
bit past i_size when we truncate to i_size, which is what this patch does.
Thanks,
Signed-off-by: Josef Bacik <jbacik@xxxxxx>
Sorry for the delay, it's been on my mind but only now I get to it.
Yes, that was agreed at LSF, and I've checked that indeed tmpfs is
out of line here: thank you for fixing it. But I do prefer your
original more explicit description, so I'll send the patch to akpm
now for v4.2, with that description instead (plus a reference to LSF).