> >for a general evaluation of the PAGE_OFFSET ... I will also speak to the
> >S390-Linux
> i believe this is the 'old way' of doing things. looking at some of
Maybe it is, there is indeed no clear definition about where kernel
space begins.
> remember if i sent in my patches to make the 2.4 port use kmap() instead
> of the page offset strangeness. i dont know if this can be back ported
to
> early 2.2 kernels (which you shouldnt run for other security related
reasons)
kmap()/kunmap() are not available for 2.2 kernels. But afaik there is no
stable
2.4 kernel available for Linux S390 and our SuSE distribution comes with
kernel 2.2.
But I was told that the variable "PAGE_OFFSET" in include/asm/page.h is
also
valid for the S390 architecture (it seems that x86 is the only architecture
which does not provide a valid "PAGE_OFFSET" variable). Actually the
PAGE_OFFSET
does not make any sense for Linux S390 since the kernel has its own address
space and thus this variable is set to 0x0 (my first patch setting it to
0x80000000 just worked because bit 31 is ignored).
I attached the small patch that should fix the PAGE_OFFSET issue for
Linux-S390.
Carsten Jacobi
(See attached file: openafs-1.0.3-page_offset.patch)
openafs-1.0.3-page_offset.patch
