nut/docs/sock-protocol.txt

224 lines
5.7 KiB
Text
Raw Permalink Normal View History

2011-01-26 09:35:08 +00:00
Driver/server socket protocol
=============================
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
Here's a brief explanation of the text-based protocol which is used
2010-03-25 23:20:59 +00:00
between the drivers and server.
The drivers may send things on the socket at any time. They will send
out changes to their local storage immediately, without any sort of
prompting from the server. As a result, the server must always check on
any driver sockets for activity.
Formatting
2011-01-26 09:35:08 +00:00
----------
2010-03-25 23:20:59 +00:00
All parsing on either side of the socket is done by parseconf, so the
same rules about escaping characters and "quoting multi-word elements"
apply here. Values which may contain odd characters are typically
sent through pconf_encode to apply \ characters where necessary.
The "" construct is used throughout to force a multi-word value to stay
together on its way to the other end.
Commands used by the drivers
2011-01-26 09:35:08 +00:00
----------------------------
2010-03-25 23:20:59 +00:00
SETINFO
2011-01-26 09:35:08 +00:00
~~~~~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
SETINFO <varname> "<value>"
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
SETINFO ups.status "OB LB"
2010-03-25 23:20:59 +00:00
There is no "ADDINFO" - if a given variable does not exist, it is
created upon receiving the first SETINFO command.
DELINFO
2011-01-26 09:35:08 +00:00
~~~~~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
DELINFO <varname>
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
DELINFO ups.temperature
2010-03-25 23:20:59 +00:00
ADDENUM
2011-01-26 09:35:08 +00:00
~~~~~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
ADDENUM <varname> "<value>"
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
ADDENUM input.transfer.low "95"
2010-03-25 23:20:59 +00:00
DELENUM
2011-01-26 09:35:08 +00:00
~~~~~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
DELENUM <varname> "<value>"
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
DELENUM input.transfer.low "98"
2010-03-25 23:20:59 +00:00
SETAUX
2011-01-26 09:35:08 +00:00
~~~~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
SETAUX <varname> <numeric value>
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
SETAUX ups.id 8
2010-03-25 23:20:59 +00:00
This overrides any previous value. The auxiliary value is presently
used as a length byte for read-write variables that are strings.
SETFLAGS
2011-01-26 09:35:08 +00:00
~~~~~~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
SETFLAGS <varname> <flag>...
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
SETFLAGS ups.id RW STRING
2010-03-25 23:20:59 +00:00
Note that this command takes a variable number of arguments, as multiple
flags are supported. Also note that they are not crammed together in
"", since "RW STRING" would mean something completely different.
This also replaces any previous flags for a given variable.
ADDCMD
2011-01-26 09:35:08 +00:00
~~~~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
ADDCMD <cmdname>
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
ADDCMD load.off
2010-03-25 23:20:59 +00:00
DELCMD
2011-01-26 09:35:08 +00:00
~~~~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
DELCMD <cmdname>
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
DELCMD load.on
2010-03-25 23:20:59 +00:00
DUMPDONE
2011-01-26 09:35:08 +00:00
~~~~~~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
DUMPDONE
2010-03-25 23:20:59 +00:00
This is only used to tell the server that every possible item has been
transmitted in response to its DUMPALL request. Once this has been
received by the server, it can be sure that it knows everything that the
driver does.
PONG
2011-01-26 09:35:08 +00:00
~~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
PONG
2010-03-25 23:20:59 +00:00
This is sent in response to a PING from the server. It is only used as
a sanity check to make sure that the driver has not gotten stuck
somewhere.
DATAOK
2011-01-26 09:35:08 +00:00
~~~~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
DATAOK
2010-03-25 23:20:59 +00:00
This means that the driver is able to communicate with the UPS, and the
data should be treated as usable. It is always sent at the end of the
dump if the data is not stale. It may also be sent at other times.
DATASTALE
2011-01-26 09:35:08 +00:00
~~~~~~~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
DATASTALE
2010-03-25 23:20:59 +00:00
This is sent by the driver to inform any listeners that the data is no
longer usable. This usually means that the driver is unable to get any
sort of meaningful response from the UPS. You must not rely on any
status information once this has been sent.
This will be sent in the beginning of a dump if the data is stale, and
may be repeated. It is cleared by DATAOK.
Commands sent by the server
2011-01-26 09:35:08 +00:00
---------------------------
2010-03-25 23:20:59 +00:00
PING
2011-01-26 09:35:08 +00:00
~~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
PING
2010-03-25 23:20:59 +00:00
This is sent to check on the health of a driver. The server should only
send this when it hasn't heard anything valid from a driver recently.
Some drivers have very little to say in terms of updates, and this may
be the only communications they have with the server on a normal basis.
If a driver does not respond with the PONG within a few seconds at the
most, it should be treated as dead/unavailable. Data stored in the
server must not be passed on to the clients when this happens.
INSTCMD
2011-01-26 09:35:08 +00:00
~~~~~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
INSTCMD <cmdname>
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
INSTCMD panel.test.start
2010-03-25 23:20:59 +00:00
SET
2011-01-26 09:35:08 +00:00
~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
SET <varname> "<value>"
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
SET ups.id "Data room"
2010-03-25 23:20:59 +00:00
DUMPALL
2011-01-26 09:35:08 +00:00
~~~~~~~
2010-03-25 23:20:59 +00:00
2011-01-26 09:35:08 +00:00
DUMPALL
2010-03-25 23:20:59 +00:00
The server uses this to request a complete copy of everything the driver
knows. This is returned in the form of the same commands (SETINFO,
etc.) that would be used if they were being updated normally. As a
result, the same parsing happens either way.
The server can tell when it has a full copy of the data by waiting for
DUMPDONE. That special response from the driver is sent once the entire
set has been transmitted.
Design notes
2011-01-26 09:35:08 +00:00
------------
2010-03-25 23:20:59 +00:00
Requests
2011-01-26 09:35:08 +00:00
~~~~~~~~
2010-03-25 23:20:59 +00:00
There is no way to request just one variable. This was done on purpose
to limit the complexity of the drivers. Their job is to send out
updates and handle a few simple requests. DUMPALL is provided to give
the server a known foundation.
To track a limited set of variables, a server just needs to do DUMPALL,
then only have handlers that remember values for the variables that
matter. Anything else should be ignored.
Access/Security
2011-01-26 09:35:08 +00:00
~~~~~~~~~~~~~~~
2010-03-25 23:20:59 +00:00
There are no access controls in the drivers. Anything that can connect
to their sockets can make requests, including SET and INSTCMD if
supported by the driver and hardware. These sockets must be kept
secure. If your operating system does not honor permissions or modes on
sockets, then you must store them in a directory with suitable
permissions to limit access.
Command limitations
2011-01-26 09:35:08 +00:00
~~~~~~~~~~~~~~~~~~~
2010-03-25 23:20:59 +00:00
As parseconf is used to handle decoding and chunking of the data, there
are some limits on what may be used. These default to 32 arguments of
512 characters each, which should be more than enough for everything
which is currently needed by the software.
These limits are strictly for sanity purposes, and may be raised if
necessary. parseconf itself can handle vast numbers of arguments and
characters, with some speed penalty as things get really big.
Re-establishing communications
2011-01-26 09:35:08 +00:00
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2010-03-25 23:20:59 +00:00
If the server loses its connection to the driver and later reconnects,
it must flush any local storage and start again with DUMPALL. The
driver may have changed the internal state considerably during that
time, and anything other approach could leave old elements behind.