On Sun, 26 Nov 2000 00:53:29 +0100,
Andi Kleen <ak@xxxxxxx> wrote:
>On Sun, Nov 26, 2000 at 10:24:23AM +1100, Keith Owens wrote:
>> On Sat, 25 Nov 2000 20:40:31 +0100,
>> Andi Kleen <ak@xxxxxxx> wrote:
>> >I needed this patch from Russel King to load xfs again (without it the
>> >module
>> >loading put zeroes where support_init should be, causing a crash)
>> >This was with modutils 2.3.14.
>>
>> The fact that Documentation/Changes requires modutils 2.3.18 for
>> kernels >= 2.4.0-test10 has obviously escaped everybody's notice.
>> You should be using modutils 2.3.21 anyway, to fix the local root
>> exploit.
>
>The code in module.c was obviously wrong anyways though, otherwise
>it would be rather useless to pass the structure size in.
module.c was wrong but it only trips when you use old modutils on a
current kernel. BTW, the patch was mine, not Russel King's. Russel
reported the problem first with a patch but his patch was in the wrong
place, I did the correct patch.
>I also do not care about local root exploits on my test machines.
But you should care about upgrading according to Documentation/Changes.
modutils 2.3.18 was required for kernel 2.4.0-test10, 2.3.14 is too
old. Before introducing significant kernel changes I issue new
modutils and bump the required version in Changes then wait one or two
kernel releases before adding the corresponding kernel patch. That way
people get a chance to upgrade modutils gracefully instead of
immediately hitting problems when a new kernel is released. But that
assumes people read Documentation/Changes.
|