On Thu, 31 May 2012, Tom Doherty wrote:
> FWIW I can reproduce this on real i386 hardware
Yeah, it appears to be MI; I finally got the bitcoin proto-port you
pointed at installed and have a kernel crash dump on my amd64 box. I'm
just starting to poke at it and the crash is in uvm_map_pageable_pgon()
from RB_NEXT() returning NULL. first and end are both != NULL, but
apparently iterating across the tree from first is running out of tree
instead of hitting end. This bit is also odd: the backtrace in gdb shows
this:
#9 0xffffffff803ddda0 in uvm_map_pageable_pgon (map=0xfffffe80018f82a8,
first=0xffff800014916d9c, end=0xfffffe8006071240, start_addr=193376,
end_addr=0) at ../../../../uvm/uvm_map.c:1887
end_addr == 0 ? Of course, the compiler may have made it all a lie, so
I'm going to try recompiling uvm/* without -O and provoke another crash...
Philip Guenther
[email protected]
