Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

(one is just a comment update: most of that stanza was not true any more)
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

A WR socket registers using a specific mac address and the receiving code
uses the mac address to match the frame.  This is going to be a problem
with the new PTP code, as pdelay frames use a different mcast mac address.
Moreover, if we go to UDP, stuff gets worse.

Packet filter rules are already in place for the new mac addresses, and
we now need to relax the checking in the software stack.

As a side effect, we avoid a number of broadcast frames, because wrpc-sw
now replies to unicast ARP requests.

What happened before:

   16:27:07 00:0b:ab:7c:8e:4a > 00:26:7b:00:04:18 who-has ...107 tell ...200
   16:27:08 00:0b:ab:7c:8e:4a > 00:26:7b:00:04:18 who-has ...107 tell ...200
   16:27:09 00:0b:ab:7c:8e:4a > 00:26:7b:00:04:18 who-has ...107 tell ...200
   16:27:12 00:0b:ab:7c:8e:4a > ff:ff:ff:ff:ff:ff who-has ...107 tell ...200
   16:27:12 00:26:7b:00:04:18 > 00:0b:ab:7c:8e:4a ...107 is-at 00:26:7b:00:04:18

what happens now:

   16:42:27 00:0b:ab:7c:8e:4a > 00:26:7b:00:04:18 who-has ...107 tell ...200
   16:42:27 00:26:7b:00:04:18 > 00:0b:ab:7c:8e:4a ...107 is-at 00:26:7b:00:04:18
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This commit removes the requirement for ARP frames to be broadcast
in the no-vlan case.  The simplified rule-set of a vlan setup already
accepts ARP unicast, and no harm is done.

Although we do not make ARP  requests so we won't receive unicast
replies, hosts do send unicast requests, to refresh an ARP entry
that is going to expire.

The next commit enacts this feature in software, as a side effect
of supporting pdelay PTP.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This is incompatible with the previous implementation

 - sequence number is 32 bits
 - frames are 1,2,3 not 0,1,2 (so to be *more* incompatible)
 - no fake latency values are printed if frames are lost
 - the message is one line instead of two, and cleaner too
 - the sender reminds it's sending every 10s.

The receiver's messages now include the sequence number, so we can
match different receivers on a network.  Also, The sender's message
report both TAI and sequence, so we can related frames to the time of
day.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This adds a tool to check the latency in the network. It is selected
at build time (config) time.  If you select CONFIG_LATENCY_PROBE you
are then asked for a CONFIG_LATENCY_ETHTYPE.

The sender node sends three frames, as broadcast in that ethertype
within the current vlan (or no vlan).  The first frame is at priority
7, the second at priority 6, and the third includes the egress
timestamps for the former ones. Each group has a sequence number
and group identifier.

If the tool is built in wrc.bin, every node acts as a receiver. To
enable a sender issuing the set of frames every 5 seconds, use "ltest 5".
To disable use "ltest 0".

The receiver prints the measured latency, and warns if it receives
frames that are not in the proper order, using the sequence and frame
type fields.

This is a successful run crossing two wr switches:

    latency prio7: 5142.775
    latency prio6: 4998.775
    latency prio7: 4822.773
    latency prio6: 5142.773
    latency prio7: 5142.771
    latency prio6: 5286.771

Later on, we'll make the thing less verbose, and we'll have threshold
to signal a problem if the network went slow. Currently we just print
all collected data.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This commit adds the Kconfig mechanism and the pfilter rules.

The latency test socket transmists frames of a specific ether type;
the receiver must receive all such frames.  To minimize the load on the
packet filter, this adds a single rule: we already accept all PTPv2
frames, so we now "OR" to this the new ethernet type.   If the feature
is configured off, the rule uses the same PTPv2 ethertye, thus
turning in practice into a NOP.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This is used by the latency measurement, to send high-priority frames
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

One rule only!
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This shortens the rule-set by one rule, making space for pdelay.

IP is acceptable whether unicast or broadcast. Thus, remove the IP_UNI
intermediate calculation and use only IP_OK.

Most of the lines in this commit are a renaming of FRAME_OUR_MAC to
FRAME_MAC_OK, with no real effect, because we'll coalesce all valid
macs in a single bit (I planned to use it in this commit, but I did
not, still I don't want to split the patch).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This shortens the rule-set by one rule, making space for pdelay.

There is no need to use a temporary register once FOR_CPU is done.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This is a safe measure before adding UDP support, which adds a socket.
And we have other ongoing net-related activities too, so I'd better
spend 16 bytes now and feel safer.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This is needed to build ppsi on the host with arch-wrpc, that
I'm using to developer udp on ppsi::wrpc-sw
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Trigger start/stop of PTP from wrc_check_link job instead of PTP's job.
This picks ppsi's new master that matches the change.

We need to remove prototype for an unused function, now that the one
in ppsi changed arguments and we'd otherwise get

  error: conflicting types for 'wr_servo_reset'
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>
Signed-off-by: Alessandro Rubini <a.rubini@gsi.de>

Although we love to build-test stuff and discard at link time, as soon
as we move tasks to be sections, we need to avoid building the file of
the task if not configured in.  This means lib/arp.c and lib/ipv4.c.

But, for consistency, avoid building icmp.c and bootp.c as well.
The alternative, which is ifconfig'ing the task itself results in
warnings like "'ipv4_init' defined but not used".
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This allows several functions to be static in the files where they
are defined. See temperature.c for an example.

But mainly, the idea is allowing ease addition of features (likely
Kconfig'd ones) for specific WR users, without any modification to the
main function and other generic code.  This will happen soon for snmp,
for example.

We may push this a little further, by moving all shell stuff into
shell.c, the ptp task within ppsi itself, and so on. But now my time
is over.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Socket matching depended on the order or sockets. With the new ELF
section (next commit) we get the ICMP socket before some UDP ones, so
the code matched IPV4 and the UDP service did not get the data.

Now we give priority to UDP sockets. So this commit
fixes the problem beforehand.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

The threshold and the repeat interval is set by Kconfig.
The temperatures are read using the temperature framework.

Tested in host environment:

   wrc# faketemp 90 80 10
   wrc# faketemp 90 40 20
   wrc# faketemp 20 30 40

   Jan  1 10:32:09 192.168.16.7 Temperature high: \
                                  roof:90.0000 core:80.0000 case:10.0000
   Jan  1 10:33:09 192.168.16.7 Temperature high: \
                                  roof:90.0000 core:40.0000 case:20.0000
   Jan  1 10:33:18 192.168.16.7 Temperature ok: \
                                  roof:20.0000 core:30.0000 case:40.0000
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This shortens the rule-set by two rules, making space for pdelay.

By accepting all ports from 0 to 511, we get all of our services,
including snmp that will be tested soon.  This relaxed check will give
us possibly some false positives (then discarded by the CPU) while
still protecting from most of an hostile or just probing  port sweep.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

The configuration is not called *_defconfig, because we can't just
build it in ./MAKEALL -- we need ppsi to be built with a different
configuration, and this is not autodetected yet (but building from a fresh
clone works).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>