Issue with pflogd consuming 100% of CPU

["David H. Gutteridge" <dhgutteridge%sympatico.ca@localhost>]

Hi all,

On one of my machines I'm finding that pflogd consistently ends up
getting stuck in a loop where it consumes close to 100% of the CPU.
This is with the netbsd-7 branch, on i386. This never happens on an
evbarm machine, also running netbsd-7, with a similar use case, which
is curious.

I tried using ktrace, which doesn't really tell me all that much
beyond that it's looping trying to access a resource that never
becomes available. I don't see any command line options that would
increase pflogd's verbosity or such. And it doesn't log anything
useful. All I get from ktrace is:

1055      1 pflogd   RET   read -1 errno 35 Resource temporarily unavailable
1055      1 pflogd   CALL  read(3,0xbb912000,0x80000)

Has anyone else seen this? Or has any insight? (Other than suggesting
alternate packet filters...)

I've been using PF on NetBSD for over ten years and have never seen
this issue before. I have to manually restart pflogd regularly to work
around this, which is kind of annoying.

Regards,

Dave