Subject: Re: Whatever happened to legendary NetBSD reliability?
To: NetBSD/Alpha <>
From: Bill Dorsey <>
List: port-alpha
Date: 07/24/2001 01:50:34

I just upgraded my PWS to NetBSD 1.5.1 and the kernel paniced almost
right away.  Here's what happened.  I transferred a large (>500M byte)
file from a W2K system to the PWS using Samba.  The transfer was
proceeding at a nice pace of about 5 megabytes/second until about
90 meg had finished transferring.  Then, the transfer rate dropped
to a crawl.  This continued for a minute or so and then NetBSD 1.5.1
paniced.  Once again, I have reproduced the information from the panic

/netbsd: panic: lockmgr: no context
/netbsd: syncing disks...
/netbsd: fatal kernel trap:
/netbsd:     trap entry = 0x2 (memory management fault)
/netbsd:     a0         = 0x7f7f7c8085fb0c6c
/netbsd:     a1         = 0x1
/netbsd:     a2         = 0x0
/netbsd:     pc         = 0xfffffc00003bf100
/netbsd:     ra         = 0xfffffc00003bf0c4
/netbsd:     curproc    = 0x0
/netbsd: panic: trap

Here's the kadb trace:

cpu_Debugger() at cpu_Debugger + 0x4
panic() at panic + 0xfc
lockmgr() at lockmgr + 0xac
uvm_fault() at uvm_fault + 0x184
trap() at trap + 0x37c
XentMM() at XentMM + 0x20
--- memory management fault (from ipl 1) ---
schedcpu() at schedcpu + 0x60
softclock() at softclock + 0x1a8
hardclock() at hardclock + 0x660
interrupt() at interrupt + 0x94
XentInt() at XentInt + 0x1c
--- interrupt (from ipl 0) ---
idle() at idle + 0x28
mi_switch() at mi_switch + 0x1a0
ltsleep() at ltsleep + 0x2c8
audio_read() at audio_read + 0x270
audioread() at audioread + 0xb8
spec_read() at spec_read + 0x120
ufsspec_read() at ufsspec_read + 0x50
vn_read() at vn_read + 0x118
dofileread() at dofileread() + 0xd0
sys_read() at sys_read + 0xa0
syscall() at syscall + 0x1dc
XentSys() at XentSys + 0x50
--- syscall (3) ---
--- user mode ---

Oh, and one person has asked me about the 533MHz clock speed.  The
machine was originally a PWS 433a.  I bought a 533MHz CPU for it
and changed the CPU clock via the dip switches to the frequency
the CPU is rated for (533MHz).  Note that it ran for many months
under NetBSD 1.4.2 without problems with the CPU running at this

If there's any other information I can provide, please let me

By the way, I am able to reproduce the network performance problem
at will.  All I need to do is try to copy a large file from my
W2K box to NetBSD.  I just tried it with FTP and after about 100
megabytes, it slowed to a crawl.  At least it didn't panic the
machine this time.  Any thoughts on what could be causing this?

Bill Dorsey