On Fri, 26 Mar 1999 12:06:04 +0100 
 Manuel Bouyer <bouyer@antioche.eu.org> wrote:

 > With a kernel from wenesday sources, I got several panics
 > 'pmap_zero_page: lock botch'. I also got a uvm_fault and random core dumps
 > of processes. This was under mid-load (6-8 load average, several compilation
 > running).

UH.....

pmap_zero_page() and pmap_copy_page() should be going to splimp()!!

Please try the following patch:

        -- Jason R. Thorpe <thorpej@nas.nasa.gov>

Index: pmap.new.c
===================================================================
RCS file: /cvsroot/src/sys/arch/i386/i386/pmap.new.c,v
retrieving revision 1.22
diff -c -r1.22 pmap.new.c
*** pmap.new.c	1999/03/19 05:13:17	1.22
--- pmap.new.c	1999/03/26 18:24:20
***************
*** 2002,2007 ****
--- 2002,2010 ----
  paddr_t pa;
  
  {
+   int s;
+ 
+   s = splimp();
    simple_lock(&pmap_zero_page_lock);
  #ifdef DIAGNOSTIC
    if (*zero_pte)
***************
*** 2013,2018 ****
--- 2016,2022 ----
    *zero_pte = 0;				/* zap! */
    pmap_update_pg((vaddr_t)zerop);		/* flush TLB */
    simple_unlock(&pmap_zero_page_lock);
+   splx(s);
  }
  
  /*
***************
*** 2024,2029 ****
--- 2028,2036 ----
  paddr_t srcpa, dstpa;
  
  {
+   int s;
+ 
+   s = splimp();
    simple_lock(&pmap_copy_page_lock);
  #ifdef DIAGNOSTIC
    if (*csrc_pte || *cdst_pte)
***************
*** 2036,2041 ****
--- 2043,2049 ----
    *csrc_pte = *cdst_pte = 0;			/* zap! */
    pmap_update_2pg((vaddr_t)csrcp, (vaddr_t)cdstp);
    simple_unlock(&pmap_copy_page_lock);
+   splx(s);
  }
  
  /*