port-amd64/48651: Lock up or panic under what appears to be heavy kevent load

[mbowie%rocket-space.com@localhost]

>Number:         48651
>Category:       port-amd64
>Synopsis:       Lock up or panic under what appears to be heavy kevent load
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    port-amd64-maintainer
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Mon Mar 10 20:50:00 +0000 2014
>Originator:     Mike Bowie
>Release:        6.1.3
>Organization:
RocketSpace, Inc.
>Environment:
NetBSD sfo3-nms01.rocketstre.am 6.1.3 NetBSD 6.1.3 (GENERIC) amd64
>Description:
Generic kernel with polling Java (openjdk7-1.7.51) process (OpenNMS) running 
either panic's or locks up after a varying amount of time. (May be up to five 
days... may be less than five hours.)
>How-To-Repeat:
Not exactly sure how to make this sure reproducible, but we just cold-boot the 
box and start the monitoring process; then wait for it to drop off the network.

In our case the device is headless in a remote DC, so connecting back to the 
DRAC console either reveals the db{0}> prompt or an unresponsive console. (With 
no scroll-back.)

db{0}> bt
filt_sowrite() at netbsd:filt_sowrite+0x22
kevent1() at netbsd:kevent1+0x61f
sys___kevent50() at netbsd:sys___kevent50+0x33
syscall() at netbsd:syscall+0xc4
db{0}> cont
uvm_fault(0xfffffe842a143328, 0x0, 1) -> e
fatal page fault in supervisor mode
trap type 6 code 0 rip ffffffff807901d8 cs 8 rflags 10287 cr2  0 cpl 0 rsp 
fffffe811dfe3990
kernel: page fault trap, code=0
Stopped in pid 2618.141268 (java) at    netbsd:filt_sowrite+0x22:       movq    0
(%rbx),%r14
db{0}> 

>Fix: