Compile on Severn, runs on RH9, segfaults on Severn

[M A Young]

On Wed, 10 Sep 2003, M A Young wrote:

> I may be seeing something similar while trying to run uml kernels. In my
> case the segfault is in a memcpy very early on in the uml boot, no
> libraries are involved, 2040 does exactly the same thing, and the same
> thing happens with the 2039 kernel on an RH9 system. The kernels boot fine
> with a 2.6 kernel under severn, or on a standard RH9 system.

Actually further investigation reveals the uml kernel is doing something
not very exec-shield friendly. It tries to copy some of its executable
bits to new bits of memory. Unfortunately the final chunk is everything
from a certain point to sbrk(0), ie. the start of the data stack, and it
seems in severn kernels that the code and data segments are no longer
adjacent, so it tries to copy 26Mb rather than under 1Mb, including bits
it presumably has no right to copy.

	Michael Young