Re: kern/46885: NetBSD 6.0_RC1 spontaneously reboots as kernel starts to load

[Lars Heidieker <lars%heidieker.de@localhost>]

The following reply was made to PR kern/46885; it has been noted by GNATS.

From: Lars Heidieker <lars%heidieker.de@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: David Holland <dholland-bugs%netbsd.org@localhost>, 
kern-bug-people%netbsd.org@localhost, 
 gnats-admin%netbsd.org@localhost, netbsd-bugs%netbsd.org@localhost, 
 dtyson%wirralcavinggroup.org.uk@localhost
Subject: Re: kern/46885: NetBSD 6.0_RC1 spontaneously reboots as kernel starts
 to load
Date: Wed, 05 Sep 2012 22:25:24 +0200

 On 09/03/2012 06:45 PM, David Holland wrote:
 > The following reply was made to PR kern/46885; it has been noted by GNATS.
 > 
 > From: David Holland <dholland-bugs%netbsd.org@localhost>
 > To: gnats-bugs%NetBSD.org@localhost
 > Cc: 
 > Subject: Re: kern/46885: NetBSD 6.0_RC1 spontaneously reboots as kernel
 >  starts to load
 > Date: Mon, 3 Sep 2012 16:40:39 +0000
 > 
 >  On Mon, Sep 03, 2012 at 04:20:05PM +0000, Valeriy E. Ushakov wrote:
 >   >  This seems to be fixed by src/sys/uvm/uvm_km.c at revision 1.131
 >   >  netbsd-6 pull up candidate?
 >  
 >  yes please
 >  
 >  
 
 Hi,
 
 a revised version of the patch it grows the kernel once after the
 kmem_arena is created but before any other arena imports from it.
 Such an import would call through uvm_km_kmem_alloc for backing va with
 memory, hence growing the kernel in there works as well.
 Once another map_entries is inserted we grow beyond kmem_arena anyway,
 so do it right away.
 
 Index: uvm/uvm_km.c
 ===================================================================
 RCS file: /cvsroot/src/sys/uvm/uvm_km.c,v
 retrieving revision 1.134
 diff -u -p -r1.134 uvm_km.c
 --- uvm/uvm_km.c       4 Sep 2012 13:37:41 -0000       1.134
 +++ uvm/uvm_km.c       5 Sep 2012 15:21:02 -0000
 @@ -329,6 +329,18 @@ uvm_km_bootstrap(vaddr_t start, vaddr_t
        kmem_arena = vmem_create("kmem", kmembase, kmemsize, PAGE_SIZE,
            NULL, NULL, NULL,
            0, VM_NOSLEEP | VM_BOOTSTRAP, IPL_VM);
 +#ifdef PMAP_GROWKERNEL
 +      /*
 +       * kmem_arena VA allocations happen independently of uvm_map.
 +       * grow kernel to accommodate the kmem_arena.
 +       */
 +      if (uvm_maxkaddr < kmembase + kmemsize) {
 +              uvm_maxkaddr = pmap_growkernel(kmembase + kmemsize);
 +              KASSERTMSG(uvm_maxkaddr >= kmembase + kmemsize,
 +                  "%#"PRIxVADDR" %#"PRIxVADDR" %#"PRIxVSIZE,
 +                  uvm_maxkaddr, kmembase, kmemsize);
 +      }
 +#endif
 
        vmem_init(kmem_arena);
 
 @@ -782,18 +794,12 @@ again:
 
  #ifdef PMAP_GROWKERNEL
        /*
 -       * These VA allocations happen independently of uvm_map so if this
 allocation
 -       * extends beyond the current limit, then allocate more resources for 
it.
 -       * This can only happen while the kmem_map is the only map entry in the
 -       * kernel_map because as soon as another map entry is created,
 uvm_map_prepare
 -       * will set uvm_maxkaddr to an address beyond the kmem_map.
 -       */
 -      if (uvm_maxkaddr < va + size) {
 -              uvm_maxkaddr = pmap_growkernel(va + size);
 -              KASSERTMSG(uvm_maxkaddr >= va + size,
 -                  "%#"PRIxVADDR" %#"PRIxPTR" %#zx",
 -                  uvm_maxkaddr, va, size);
 -      }
 +       * These VA allocations happen independently of uvm_map
 +       * so this allocation must not extend beyond the current limit.
 +       */
 +      KASSERTMSG(uvm_maxkaddr >= va + size,
 +          "%#"PRIxVADDR" %#"PRIxPTR" %#zx",
 +          uvm_maxkaddr, va, size);
  #endif
 
        loopva = va;