sparc64/80410: [netgraph] netgraph is causing crash with mpd on sparc64

[Marius Strobl]

The following reply was made to PR sparc64/80410; it has been noted by GNATS.

From: Marius Strobl <marius at alchemy.franken.de>
To: Sergey Mokryshev <mokr at mokr.net>, bug-followup at FreeBSD.org
Cc:  
Subject: Re: sparc64/80410: [netgraph] netgraph is causing crash with mpd on sparc64
Date: Thu, 25 Mar 2010 00:15:40 +0100

 On Wed, Mar 24, 2010 at 05:30:05PM +0000, Sergey Mokryshev wrote:
 > The following reply was made to PR sparc64/80410; it has been noted by GNATS.
 > 
 > From: "Sergey Mokryshev" <mokr at mokr.net>
 > To: <bug-followup at FreeBSD.org>, <bel at orel.ru>
 > Cc:  
 > Subject: Re: sparc64/80410: [netgraph] netgraph is causing crash with mpd on sparc64
 > Date: Wed, 24 Mar 2010 20:07:47 +0300
 > 
 >  FreeBSD-8 with mpd-5.1 still panics on sun4u (Netra X1).
 >  Coredump and server access available on request.
 >  
 >  FreeBSD 8.0-STABLE #5: Fri Mar 19 09:23:54 MSK 2010
 >      root at netra-x1.*****.ru:/usr/obj/usr/src/sys/MOKR-S64 sparc64
 >  
 >  # kldstat
 >  Id Refs Address            Size     Name
 >   1   37 0xc0000000 7455d8   kernel
 >   2    3 0xc0746000 1554a0   usb.ko
 >   3    1 0xc089c000 117448   ohci.ko
 >   4    1 0xc09b4000 120708   ehci.ko
 >   5    1 0xc0ad6000 17d6d0   ipl.ko
 >   6    1 0xc1314000 2a8000   zfs.ko
 >   7    1 0xc15bc000 122000   krpc.ko
 >   8    1 0xc16e2000 104000   opensolaris.ko
 >   9    1 0xc1e1c000 108000   if_vlan.ko
 >  10    1 0xc1f24000 106000   ng_socket.ko
 >  11    3 0xc202a000 116000   netgraph.ko
 >  12    1 0xc2144000 104000   ng_mppc.ko
 >  13    1 0xc2248000 102000   rc4.ko
 >  14    1 0xc238c000 108000   ng_ksocket.ko
 >  
 >  
 >  Mar 24 20:01:12 netra-x1 mpd: [L-1] LCP: SendConfigReq #3
 >  Mar 24 20:01:12 netra-x1 mpd: [L-1]   ACFCOMP
 >  Mar 24 20:01:12 netra-x1 mpd: [L-1]   PROTOCOMP
 >  Mar 24 20:01:12 netra-x1 mpd: [L-1]   MRU 1500
 >  Mar 24 20:01:12 netra-x1 mpd: [L-1]   MAGICNUM e361cfb8
 >  Mar 24 20:01:12 netra-x1 mpd: [L-1]   AUTHPROTO CHAP MSOFTv2
 >  Mar 24 20:01:12 net
 >  panic: trap: memory address not aligned
 >  cpuid = 0
 >  KDB: stack backtrace:
 >  panic() at 0xc01ce588 = panic+0x1c8
 >  trap() at 0xc046da10 = trap+0x4d0
 >  -- memory address not aligned sfar=0xfffff800011fad8e sfsr=0x4002d 
 >  %o7=0xc02cdd58 --
 >  rip_output() at 0xc02cdb44 = rip_output+0x124
 >  rip_send() at 0xc02cddec = rip_send+0x4c
 >  sosend_generic() at 0xc0249164 = sosend_generic+0x5e4
 >  sosend() at 0xc0245a4c = sosend+0x2c
 >  ng_ksocket_rcvdata() at 0xc2390218 = ng_ksocket_rcvdata+0xb8
 >  ng_apply_item() at 0xc2037ff0 = ng_apply_item+0x250
 >  ngthread() at 0xc203944c = ngthread+0x16c
 >  fork_exit() at 0xc0197510 = fork_exit+0xd0
 >  fork_trampoline() at 0xc00891e8 = fork_trampoline+0x8
 
 This is a bit of a blind shot due to insufficient debugging
 information but could you please give the following patch
 a try:
 http://people.freebsd.org/~marius/raw_ip_alignment.diff
 
 Note that this backtrace is different from the one originally
 reported in this PR, i.e. is caused by a different problem.
 
 Marius