Re: PR 9725 and the m68k pmap

[Hauke Fath <hauke%Espresso.Rhein-Neckar.DE@localhost>]

At 9:16 Uhr -0600 12.5.2009, Michael L. Hitch wrote:
>   A quick and dirty workaround is to reduce the user address space by
>lowering the USERSTACK in vmparam.h (I can't remember for sure what value
>it would need to be, but it might be 0x0e000000 or 0x0c000000).

At 20:54 Uhr +0900 14.5.2009, Izumi Tsutsui wrote:
>I guess Michael's hack for current pmap requires less efforts
>to fix (or workaround) the PR.

I am not sure what it is supposed to achieve... I built and ran a few test
kernels  on a Quadra 650 with 136 MB RAM with

Index: vmparam.h
===================================================================
RCS file: /cvsroot/src/sys/arch/mac68k/include/vmparam.h,v
retrieving revision 1.36
diff -u -r1.36 vmparam.h
--- vmparam.h   11 Dec 2005 12:18:03 -0000      1.36
+++ vmparam.h   20 May 2009 07:35:45 -0000
@@ -127,7 +127,9 @@
  * NOTE: HP300 uses HIGHPAGES == (0x100000/PAGE_SIZE) for HP/UX compatibility.
  * Do we care?  Obviously not at the moment.
  */
+#ifndef USRSTACK
 #define        USRSTACK        (-HIGHPAGES*PAGE_SIZE)  /* Start of user
stack */
+#endif
 #define        BTOPUSRSTACK    (0x100000-HIGHPAGES)    /* btop(USRSTACK) */
 #define P1PAGES                0x100000
 #define HIGHPAGES      3                       /* UPAGES */


and config options

<snip>
# see mac68k/include/vmparam.h
options         MAXDSIZ="(192*1024*1024)" # Bump max data size
options         USRSTACK=0x0c000000     # PR 9725
</snip>

varying USRSTACK between 0x0c000000 and 0x0f00000 (can anybody point me to
a memory map?), and MAXDSIZ between 96 MB and 192 MB. Running the PR 9715
test app, what I got was either an 'out of memory' core dump for small
values of MAXDSIZ (< phys memory), or the usual kernel panic, or - for low
values of USRSTACK - /sbin/init being not executable.

        hauke

--
"It's never straight up and down"     (DEVO)