Re: kernel panic when accessing digi serial multi io.

On Monday 02 August 2004 10:27 am, Klaus Robert Suetterlin wrote:
> Dear all,
>
> I use a Digi Xem (16 serial ports) with FreeBSD Current.
>
> geclab2# kldload digi
> digi0 mem 0xb0800000-0xb0bfffff irq 30 at device 4.0 on pci0
> digi0: Digiboard PCI PC/Xem ASIC, 16 ports found
>
> This worked ok until I updated last week.
>
> geclab2# uname -a
> FreeBSD geclab2.cips.mpg.de 5.2-CURRENT FreeBSD 5.2-CURRENT #0: Thu Jul 29
> 17:14:49 CEST 2004     krs_at_geclab2.cips.mpg.de:/usr/obj/usr/src/sys/GECLAB2
>  i386
>
> Now every time I access any serial port on the digi I get a kernel
> panic.  My normal (sio0, sio1) serial ports work ok.  It's not
> always the same panic but always about mutexes or Giant etc.  Here
> is one as an example:
>
> geclab2# cu -l /dev/ttyD0.2
> panicpanic: process 530(cu):2 holds Giant but isn't blocked on a lock
>
> KDB: enter: panic
> [thread 100006]
> Stopped at      kdb_enter+0x2b: nop
> db> where
> kdb_enter(c0632982) at kdb_enter+0x2b
> panic(c063567d,212,c20be83c,2,c0642312) at panic+0xbb
> propagate_priority(c1de3840,c068cc70,c0688aa0,c1de3840,c1dd8340) at
> propagate_pr iority+0x142
> turnstile_wait(c1dd8340,c0688aa0,c20bd2c0,c0688aa0,2,c0631d0e,212) at
> turnstile_ wait+0x2de
> _mtx_lock_sleep(c0688aa0,0,c06302e4,21f) at _mtx_lock_sleep+0x11b
> _mtx_lock_flags(c0688aa0,0,c06302e4,21f) at _mtx_lock_flags+0x83
> ithread_loop(c1de0b80,d8f6bd48,c1de0b80,c04ae198,0) at ithread_loop+0x11c
> fork_exit(c04ae198,c1de0b80,d8f6bd48) at fork_exit+0xa4
> fork_trampoline() at fork_trampoline+0x8
> --- trap 0x1, eip = 0, esp = 0xd8f6bd7c, ebp = 0 ---

Do you have an SMP machine?  If so, does adding 'options NO_ADAPTIVE_MUTEXES' 
to your kernel fix the panic?

-- 
John Baldwin <jhb_at_FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve"  =  http://www.FreeBSD.org