free_initmem() bug

[Benjamin Herrenschmidt]

>
>free_initmem() must not be __init because that means sawing the branch
>underneat itselves, which may cause a crash.
>
>I also optimized the freeing of the various sections, both from a soure and
>object view.
>
>The patch is half tested, i.e. it compiles. Since I won't be able to do
bk for
>a while, can someone please take care of this? Thanks!

Ah, this might explain some problems I had with 2.4 hangs on freeing init
mem...


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/