From: Shachar Shemesh (linux-il_at_nonexisting.hamakor.org.il)
Date: Sun 25 Jul 2004 - 18:54:23 IDT
Muli Ben-Yehuda wrote:
>On Sun, Jul 25, 2004 at 01:03:47AM +0300, Shachar Shemesh wrote:
>
>
>>Hi all,
>>
>>I'm trying to patch 2.6.6 to add support for the 4GB user space/ 4GB
>>kernel space split.
>>
>>
>
>I'm curious why?
>
>
Don't ask. Really. It is a very long story that has to do with HUGE
text-based databases, Wine's inability to allocate memory below the 1GB
line due to its reliance on mmap, and Windows Advanced Server. Do you
still want to know?
>>The patch I have is against 2.6.6-B7, whatever that
>>is. It does not apply cleanly on vanilla 2.6.6.
>>
>>
>
>The -B7 is just Ingo's numbering scheme. 2.6.6-B7 means it's likely to
>be against 2.6.6-vanilla. Where does it not apply?
>
>
Lot's of places. Mostly changes in the patch that were already in the
vanilla source, but also some symbols removed.
>>I have shaked, twisted and hit the patch into place, until it now
>>applies cleanly on stock vanilla 2.6.6 (attached). There are still a few
>>rough edges on it. However, that does not appear to be the problem. The
>>problem is that when I try to compile this kernel using the Debian tools
>>(make-kpkg), I get compilation error ON THE DOCS. It claims that
>>"direct_strncpy_from_user" is defined twice in kernel_api.sgml. I tried
>>tracing it back to the source, and found nothing meaningful -
>>direct_strncpy_from_user is only defined once there.
>>
>>
>
>Since you obviously don't care about the docs, how about using any of
>several brute force methods?
>- rm kernel_api.sgml; touch kenel_api.sgml
>
>
It's a generated file. I'd have to rm it higher up, and I'm not sure
where "higher up" actually is.
>- find out where make-kpkg compiles the docs, and remove that (the
>docs are not compiled by default- if make-kgkg compiles them, it is
>doing it explicitly).
>
>
Tried that as well. failed.
>etc
>
>
Do elaborate.
The thing that makes me really curious about this is the fact that it
obviously fails due to the patch. A really vanilla kernel passes. Also,
it fails on a function not defined in the vanilla kernel (vanilla has
"strncpy_from_user", without "direct", instead). I did manage to find
that the SGMLs were generated directly from the C sources, but that's as
far as I got before switching to compiling by hand.
>Cheers,
>Muli
>
>
Oh well. Thanks anyways.
-- Shachar Shemesh Lingnu Open Source Consulting ltd. http://www.lingnu.com/ ================================================================= To unsubscribe, send mail to linux-il-request_at_linux.org.il with the word "unsubscribe" in the message body, e.g., run the command echo unsubscribe | mail linux-il-request_at_linux.org.il
This archive was generated by hypermail 2.1.7 : Sun 25 Jul 2004 - 19:03:01 IDT