Fixes for padding in OpenFlow header match fields

Currently, we assume that OpenFlow 1.3 set_field actions can contain multiple header match fields (OXMs). According to specification, a set_field action contains exactly one OXM, followed by padding to align the action field in a OpenFlow message to 64 bits. This diff changes how we handle OpenFlo

switch(4): port protection man page updates

Hi, These are the manpage updates to go with the port protection diffs for switch(4). Thanks, Ayaka Index: sbin/ifconfig/ifconfig.8 === RCS file: /cvs/src/sbin/ifconfig/ifconfig.8,v retrieving revision 1.334 diff -u -p -u -r1.334

switch(4): add support for protected domains

Hi, The following adds port protection support to switch(4). The behavior copies that of bridge(4), where ports can be added to up to 31 protected domains with a list of IDs using ifconfig(8): # ifconfig switch0 protected pair1 1,2,.. Domain membership is checked for unicast, flooded (broadcast)

switchd.conf(5): mention default values used for global configurations

Hi all, After a second thought, I've split off the diff for switchd.conf(5) from the diffs for changing switchd(8)'s default listen port. As suggested by kn@, it now notes the default listen port and address. Comments/OKs? Thanks, Ayaka Index: switchd.conf.5 ==

Re: switchd(8): change default listen port to the standardized OpenFlow port

On Wed, Nov 07, 2018 at 10:52:59AM +0100, Klemens Nanni wrote: > On Tue, Nov 06, 2018 at 11:01:37PM -0800, Ayaka Koshibe wrote: > > Currently, switchd(8) defaults to listening on port 6633, which was > > the defacto port value used by OpenFlow. A decent chunk of OpenFlow >

switchd(8): change default listen port to the standardized OpenFlow port

Hi all, Currently, switchd(8) defaults to listening on port 6633, which was the defacto port value used by OpenFlow. A decent chunk of OpenFlow controllers have switched over to the IANA standardized OpenFlow port, 6653. While testing the change, I noticed that switchd(8) will listen on random por

switchctl(8): additional pledge for 'show *' commands

Hi all, I've taken claudio@'s advice and restricted the pledge further. All of the promises except stdio were for parsing args and opening sockets. OK? Thanks, Ayaka Index: switchctl.c === RCS file: /cvs/src/usr.sbin/switchctl/sw

Re: switchctl(8): Add ability to reference control device for 'dump' option

On Mon, Oct 15, 2018 at 03:17:28PM -0700, Ayaka Koshibe wrote: > Hi all, > > Currently, switchctl(8) is able to query a switch for information by > specifying > the switch's network address. The following adds support for querying a > switch(4) instance via its control

switchctl(8): Add ability to reference control device for 'dump' option

Hi all, Currently, switchctl(8) is able to query a switch for information by specifying the switch's network address. The following adds support for querying a switch(4) instance via its control device: $ doas switchctl switch /dev/switch0 dump features any > /dev/switch0: version 1_3 type FEATU

Re: vmd stdio /dev/null

On Mon, Sep 10, 2018 at 01:48:23PM +0200, Alexander Bluhm wrote: > On Mon, Sep 10, 2018 at 09:59:31AM +0200, Reyk Floeter wrote: > > I already gave my OK for relayd but asked to adjust all proc.c instances. > > I have fixed relayd, httpd, vmd. > akoshibe@ cares about switchd. switchd is also now

switchd(8) stdio /dev/null

Hi, Mirroring bluhm's fixes for proc.c daemons to dup /dev/null for child processes, for switchd(8). OK? Thanks, Ayaka Index: proc.c === RCS file: /cvs/src/usr.sbin/switchd/proc.c,v retrieving revision 1.13 diff -u -p -u -r1.13 pr

switchd(8): don't set output port to OFP*_PORT_ANY for input == output

Hi, This is a new version of a previous diff that I had for making switchd(8) ignore PACKET_IN messages generated from looped traffic. Currently, it will respond to the PACKET_IN with an invalid PACKET_OUT onto OFP*_PORT_ANY, resulting in the switch responding with an error that makes switchd disc

[patch] switchd(8) on sparc64: panic: trap type 0x34

Hi, At BSDCan, someone reported that a sparc64 machine would panic if it was receiving any traffic on a member interface of a switch(4) during reboot. We got as far as getting this trace: panic: trap type 0x34 (mem address not aligned): pc=15864ec npc=15864f0 pstate=99820006 Stopped at db_en

Check lo(4) rdomain before attempting to move it to another

Hi, I noticed that if you try to move an immovable lo(4) to a nonexistent rdomain, a new routing table will be created despite the operation failing with an EPERM. The following moves the rdomain/index check for lo(4) to before the creation of a nonexisting target routing table. OK? Thanks, Ay

remove pledge(2) cpath from switchd(8)

On Sat, Aug 04, 2018 at 12:18:45PM +0100, Ricardo Mestre wrote: > Hi, > > Here's another one for switchd(8) which removes cpath promise from pledge(2) > > OK? This makes sense to me, and my usual tests worked without any issues. ok akoshibe@ > > Index: control.c > ==

switchd(8): ignore packet_out with dstport = OFP_PORT_ANY

On Fri, Apr 20, 2018 at 10:18:34AM -0700, Ayaka Koshibe wrote: > Hi, > > switchd(8) sends packet_outs with OFP_PORT_ANY as the dstport, causing > switches receiving them to disconnect due to validation failure of the said > message. The comments indicate that looping packets sh

switchd(8): ignore packet_out with dstport = OFP_PORT_ANY

On Fri, Apr 20, 2018 at 10:18:34AM -0700, Ayaka Koshibe wrote: > Hi, > > switchd(8) sends packet_outs with OFP_PORT_ANY as the dstport, causing > switches receiving them to disconnect due to validation failure of the said > message. The comments indicate that looping packets sh

ifconfig: add -tunneldomain option

On Wed, Apr 25, 2018 at 11:58:30AM +0100, Jason McIntyre wrote: > On Wed, Apr 25, 2018 at 12:20:35PM +0200, Sebastian Benoit wrote: > > ok > > > > while doing it, can you move the tunneldomain options (both) to after > tunneldf instead of before it (both in the mini synopsis and the list)? > if n

ifconfig: add -tunneldomain option

Hi, A while back I got a suggestion to add a '-tunneldomain' option as shorthand for 'tunneldomain 0' to be consistent with '-rdomain'. OK? Thanks, Ayaka Index: ifconfig.8 === RCS file: /cvs/src/sbin/ifconfig/ifconfig.8,v retrievin

switchd(8): ignore packet_out with dstport = OFP_PORT_ANY

Hi, switchd(8) sends packet_outs with OFP_PORT_ANY as the dstport, causing switches receiving them to disconnect due to validation failure of the said message. The comments indicate that looping packets should be ignored, which I agree is the expected behavior for a controller implementing a learn

Re: ifconfig: add -rdomain option

On Thu, Feb 22, 2018 at 4:54 AM, David Gwynne wrote: > > >> On 22 Feb 2018, at 5:00 pm, Ayaka Koshibe wrote: >> >> On Wed, Feb 21, 2018 at 03:42:58PM +0100, Sebastian Benoit wrote: >>> Ayaka Koshibe(akosh...@openbsd.org) on 2018.02.20 21:20:20 -0800: >>&

Re: ifconfig: add -rdomain option

On Wed, Feb 21, 2018 at 03:42:58PM +0100, Sebastian Benoit wrote: > Ayaka Koshibe(akosh...@openbsd.org) on 2018.02.20 21:20:20 -0800: > > On Tue, Feb 20, 2018 at 4:48 AM, Reyk Floeter wrote: > > > > > >> Am 20.02.2018 um 11:15 schrieb Klemens Nanni : > > >

Re: ifconfig: add -rdomain option

On Tue, Feb 20, 2018 at 5:08 AM, Stefan Sperling wrote: > On Mon, Feb 19, 2018 at 05:09:58PM -0800, Ayaka Koshibe wrote: >> Index: sbin/ifconfig/ifconfig.c >> === >> RCS file: /cvs/src/sbin/ifconfig/ifcon

Re: ifconfig: add -rdomain option

On Tue, Feb 20, 2018 at 4:48 AM, Reyk Floeter wrote: > >> Am 20.02.2018 um 11:15 schrieb Klemens Nanni : >> >>> On Mon, Feb 19, 2018 at 05:09:58PM -0800, Ayaka Koshibe wrote: >>> This diff would allow saying 'ifconfig foo -rdomain' instead of '

ifconfig: add -rdomain option

Hi, This diff would allow saying 'ifconfig foo -rdomain' instead of 'ifconfig foo rdomain 0'. OK? Thanks, Ayaka Index: sbin/ifconfig/ifconfig.8 === RCS file: /cvs/src/sbin/ifconfig/ifconfig.8,v retrieving revision 1.292 diff -u -