Subject: Re: Qube2 "crash" every few days during the daily script
To: None <port-cobalt@NetBSD.org>
From: Izumi Tsutsui <tsutsui@ceres.dti.ne.jp>
List: port-cobalt
Date: 06/01/2007 02:06:11
I wrote:

> It got a TLB miss actually in a ksyms function during daily cron job
> with a background job copying files from NFS to local wd0:
> 
> ---
>  :
> colt-% tar clf - . | ( cd /usr/local/tmp ; tar xf - )
> trap: TLB miss (load or instr. fetch) in kernel mode
> status=0x3, cause=0x8008, epc=0x801d3fdc, vaddr=0x7fb7e63a
> pid=1307 cmd=netstat usp=0x7fffd4b0 ksp=0xcc741c28
> Stopped in pid 1307.1 (netstat) at      netbsd:ksyms_getname+0x318:     \
> lb      v1,0(v0)
> db> tr
> ksyms_getname+318 (c0159a80,8039c835,c0112ffa,c0159a80) ra 801d4084 sz 16
> ksyms_getname+3c0 (c0159a80,8039c835,c0112ffa,c0159a80) ra 0 sz 16
> User-level: pid 1307.1
 :
> I'm not sure if this is cobalt specific or MI ksyms(4) problem,
> but this kernel doen't contain symbols for local (static)
> functions so I'll try to reproduce it on a debug kernel
> (with src/sys/arch/mips/conf/Makefile.mips rev 1.46).

After ~3 hour load, now I get:

---
 :
colt-% tar clf - . | ( cd /usr/local/tmp ; tar xf - )
trap: TLB miss (load or instr. fetch) in kernel mode
status=0x3, cause=0x8, epc=0x801d3fdc, vaddr=0x7ffa9bb1
pid=5047 cmd=netstat usp=0x7fffd2e8 ksp=0xcc72dc28
Stopped in pid 5047.1 (netstat) at      netbsd:findsym+0x7c:    \
lb      v1,0(v0)

db> tr
findsym+7c (c019b500,803b1b7b,c0112ffa,c019b500) ra 801d4290 sz 48
ksymsioctl+dc (c019b500,803b1b7b,c0112ffa,c019b500) ra 8024139c sz 56
spec_ioctl+e8 (c019b500,803b1b7b,c0112ffa,c019b500) ra 8023af38 sz 40
VOP_IOCTL+44 (8db3f1f8,80086c01,cc72de50,1) ra 80237cb4 sz 56
vn_ioctl+78 (8db3f1f8,80086c01,cc72de50,1) ra 80202760 sz 184
sys_ioctl+16c (8db3f1f8,80086c01,cc72de50,1) ra 80272b88 sz 208
syscall_plain+138 (8db3f1f8,80086c01,cc72de50,1) ra 8026c674 sz 80
mips3_SystemCall+b4 (8db3f1f8,80086c01,cc72de50,1) ra 7dee49c0 sz 0
PC 0x7dee49c0: not in kernel space
0+7dee49c0 (8db3f1f8,80086c01,cc72de50,1) ra 0 sz 0
User-level: pid 5047.1
db> 
---

Looks MI ksyms(4) problem, but I'll take a look at it later.

---
Izumi Tsutsui