On Wed, May 25, 2011 at 10:02:37AM +0800, Yongqiang Yang wrote:

path[i]p_idx is mutated to EXT_MAX_INDEX(path[i].p_hdr) + 1, it is
meaningless.  ext4_ext_split() is used only by
ext4_ext_create_new_leaf() which drops the path after ext4_ext_split()
returns, so the path[i].p_idx is no longer used.

I had tested the patch with fsx.  I am not sure if this test is
enough.  If you want more tests, I can do it.

OK, thanks for the explanation.  Sometimes testing doesn't find
problems, so I try to understand the patch and make sure it is sane.
I noticed that your path[i].p_ext and path[i].p_idx were no longer
being updated with your patch; I was able to convince myself that
p_ext didn't matter, but I wasn't entirely sure about p_idx, so that's
why I asked you.

I've added it to the ext4 dev tree.

					- Ted