Re: kernel/2211: PF / Kernel problem, it filtered all ports

[John Kerbawy <john@maKintosh.com>]

On Wed, Nov 28, 2001 at 08:56:04PM +0100, Daniel Hartmeier wrote:
> I can't reproduce this, neither on a 512MB macppc nor on a 64MB i386. On
> the latter, after about 60000 states, pf runs out of memory. The running
> nmap prints
> 
> sendto in send_tcp_raw: sendto(3, packet, 40, 0, 62.65.145.30, 16)
>   => No route to host
> 
> at this point, and I get 'pf: ERROR! insert failed' messages in
> /var/log/messages. The system remains responsive. Existing connections
> continue to work. When I abort the scan the state entries time out and
> are removed. New connections work again. 
> 
> With 512MB of memory, I can't exhaust memory even running several
> concurrent scans, the states time out faster than the scans produce
> them (10mbps network).

I'm seeing similar things.

(gdb) target kcore /var/crash/bsd.0.core
panic: malloc: out of space in kmem_map
#0  0x1000 in ?? ()
(gdb) bt
#0  0x1000 in ?? ()
#1  0xe02d10b9 in boot ()
#2  0xe017e40b in db_boot_crash_cmd ()
#3  0xe017dfc8 in db_command ()
#4  0xe017e24b in db_command_loop ()
#5  0xe01814ce in db_trap ()
#6  0xe02cc262 in kdb_trap ()
#7  0xe02d9e3c in trap ()
#8  0xe0100dcd in alltraps ()
#9  0xe01b0d89 in panic ()
#10 0xe01a2ffe in malloc ()
#11 0xe0293efc in amap_extend ()
#12 0xe029ae8f in uvm_map ()
#13 0xe02a670d in sys_obreak ()
#14 0xe02da616 in syscall ()
#15 0xe0100e79 in Xosyscall_end ()
can not access 0xdfbb7b90, invalid address (dfbb7b90)
can not access 0xdfbb7b90, invalid address (dfbb7b90)
Cannot access memory at address 0xdfbb7b90.
(gdb) 

This box is running 3.0-release and has 256MB of RAM.