On Mon, Sep 3, 2018 at 1:20 PM, Anthony J. Bentley <anth...@anjbe.name> wrote: > Hi, > > I've been using an APU2 with OpenBSD as my home wired/wireless router > for a couple of years now. > > Initially I used bridge(4), and never experienced any problems. About a > release ago I replaced bridge(4) with switch(4). It works just the same, > except that every so often my net goes down for seemingly no reason, > because switchd has exited. > > "rcctl start switchd && switchctl connect /dev/switch0" starts up my > network again, but switchd always exits again, within a few minutes or > hours or days or weeks. I've never noticed any direct cause or been > able to reliably trigger it, although I'll say it seems to happen > more often with lots of traffic (e.g., torrents or large HTTP > downloads). > > I ran switchd -dvvv with script(1) and got a huge typescript, ending > in the following (more context available if necessary): > > switch_learn: updated mac 10:60:4b:7b:0a:44 on switch 1 port 3 > packet_input: 10:60:4b:7b:0a:44 -> 01:80:c2:00:00:0e, port 3 -> 4294967295 > any > /dev/switch0: version 1_3 type PACKET_OUT length 100 xid 8433 > buffer NO_BUFFER in_port <3> actions_len 16 > action OUTPUT len 16 port FLOOD max_len NO_BUFFER > 040a006e 0000606f > ffffffff 003c0000 00000000 00000000 00010018 80000004 00000003 80000408 > 00000000 00000000 00000180 c2000001 00000000 00008808 0001ffff 00000000 > 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 > 00000000 0000 > ofrelay_input_done: connection 1.1: 110 bytes from switch 1 > 040a006e 0000606f ffffffff 003c0000 00000000 00000000 00010018 80000004 > 00000003 80000408 00000000 00000000 00000180 c2000001 00000000 00008808 > 0001ffff 00000000 00000000 00000000 00000000 00000000 00000000 00000000 > 00000000 00000000 00000000 0000 > /dev/switch0 > any: version 1_3 type PACKET_IN length 110 xid 24687 > buffer NO_BUFFER length 60 reason REASON_NO_MATCH table <0> cookie > 0x000 > 0000000000000 > match type OXM length 24 (padded to 26) > ox match class OPENFLOW_BASIC type IN_PORT hasmask no length 4 > 3 > ox match class OPENFLOW_BASIC type META hasmask no length 8 > 0 > 22881 parent exiting > control exiting, pid 20472 > ofcconn exiting, pid 93222 > #
It looks like switchd died midway through processing a message from the switch, though I can't say if the cause was the message itself, or something else. > There's nothing in /var/log/messages. I don't know why the parent > process exits this way. Do you see anything in /var/log/daemon? Thanks, Ayaka