1177 replies to this topic

[Ziegenmelker]

I will create a SD-Card with OpenATV 6 this Weekend. Then I will check it...

[Ziegenmelker]

So,

 

ich habe mal OpenATV 6.0 installiert. Die benutzen bei der Wetek den GStreamer 0.10.36 (GIT). So steht es zumindest in der Info.

 

Ich muss sagen, dass OpenATV 6.0 sehr gut funktioniert. Alles für mich Wichtige (Mediaportal, OSCam) konnte ich problemlos installieren und bislang konnte ich keine Fehler feststellen, selbst Timeshift funktioniert. Mal schauen, ob ich wie bei älteren Versionen alle paar Tage die Box neu starten muss. Werde aber erst einmal dabei bleiben...

 

MfG Ziegenmelker

[littlesat]

ich habe mal OpenATV 6.0 installiert. Die benutzen bei der Wetek den GStreamer 0.10.36 (GIT). So steht es zumindest in der Info.

Das bedeutet komischen work-a-rounds... 

[peti]

with the current develop branch I'm getting GSOD when static IPv6 is configured:

No handlers could be found for logger "AT"
Traceback (most recent call last):
  File "/usr/lib/enigma2/python/Components/Console.py", line 37, in finishedCB
  File "/usr/lib/enigma2/python/Components/Network.py", line 129, in routeFinished
  File "/usr/lib/enigma2/python/Components/Network.py", line 191, in loadNetworkConfig
ValueError: invalid literal for int() with base 10: '2a01:xxxx:xxxx:84d0::20' 

in the config file I have:

iface eth0 inet6 static
        address 2a01:xxxx:xxxx:84d0::20
        netmask 80 

[WanWizard]

There is no IPv6 support in Enigma, and there never was. It's high on my todo list.

[peti]

hm. it was working correctly prior to some recent update. 

[WanWizard]

Very weird. The code it fails on:

ifaces[currif]["address"] = map(int, split[1].split('.'))

hardcoded splits on a dot and converts the individual values to integers, which is IPv4 only. This code hasn't changed in 10 years. 

 

It's full of stuff like this as well

ipRegexp = "[0-9]{1,3}\.[0-9]{1,3}\.[0-9]{1,3}\.[0-9]{1,3}"

which doesn't work at all with IPv6 either.

 

I've worked around the issue by using DHCPv6 in my LAN, until I find time to work on it (which is going to be a major change).

[peti]

ok, I tried to remember, I think when I set this up first, I haven't rebooted the box, just the networking service, so most probably it was crashing on e2 start all the time. 

Indeed, with autoconfig is running fine, and static config is rather marginal (I have multiple local subnets, and that's why I'm splitting my /64 prefix to /80s and my router doesn't support stateful DHCP), however would be nice to fix at some point. 

[SpaceRat]

OpenPLi might want to cherry-pick some of my OpenATV commits to Network.py which avoid exactly these parsing errors:

1. The output of busybox ip is junk, but E2 relies on it
When using the full "ip" command from iproute, E2 will fail.
https://github.com/o...64f33f0ebcfdd28

2. Network.py can not properly handle IPv6 addresses
https://github.com/o...6774edafade3dee

3. ip addr flush -> flush scope global only
https://github.com/o...bbc8c5c6523b3d7
https://github.com/o...84ef5919659f544 (Fixes previous commit)


Those commits are only workarounds to keep the crappy Network.py happy when using IPv6 without really supporting it.
Network.py needs a lot more work to properly support IPv4 and IPv6, for the beginning an option to simply disable one or the other.

I didn't implement any IPv6 support yet, because I wanted to avoid unneeded differences in handling and API between OpenATV and OpenPLi.

Edited by SpaceRat, 4 July 2017 - 09:06.

[littlesat]

Nice clean patches. nie we need to decide if we put it i. 6.0 or 6.1... at least I understand where the issue is coming from...

Edited by littlesat, 4 July 2017 - 09:13.

[SpaceRat]

ok, I tried to remember, I think when I set this up first, I haven't rebooted the box, just the networking service, so most probably it was crashing on e2 start all the time.

You won't get a crash from that code unless you go to the Network settings in E2.
And as one doesn't need to go there at all, it took years for me to discover the problems there

Indeed, with autoconfig is running fine, and static config is rather marginal (I have multiple local subnets, and that's why I'm splitting my /64 prefix to /80s

Yikes.
Your ISP is supposed to give you at least a /58 prefix so that you can create multiple full featured /64 subnets.

and my router doesn't support stateful DHCP)

... and neither does OpenPLi I guess.

OpenEmbedded does have neither dibbler nor any other DHCPv6 client.
I have integrated odhcp6c (The DHCPv6 client from OpenWrt/LEDE) in oe-a to overcome this limitation.

As the next limitation, also crappybox (busybox) has very limited support for IPv6 in ifup/ifdown, I had to patch that in too.

Result:

root@solo2 / # ifconfig
eth0      Link encap:Ethernet  HWaddr 00:1D:EC:00:12:34
          inet6 addr: 2001:db8:dead:0:21d:ecff:fe00:1234%4933648/64 Scope:Global
          inet6 addr: fdc1:beef:affe:0:21d:ecff:fe00:1234%4933648/64 Scope:Global
          inet6 addr: fdc1:beef:affe::75%4933648/128 Scope:Global
          inet6 addr: fe80::21d:ecff:fe00:1234%4933648/64 Scope:Link
          inet6 addr: 2001:db8:dead::75%4933648/128 Scope:Global

That box gets IPv6 addresses via DHCPv6 (The ones ending with :75) and SLAAC (The ones ending with :1234) at the same time (It can be configured using /etc/network/interfaces to use DHCPv6 or SLAAC or both).

IPv4 is disabled entirely on two of my machines to encounter and fight show-stoppers for the sunset of IPv4.

(Some) known show-stoppers are:
- E2 allows zero configuration of IPv6
- NetworkBrowser creates silly IP(v4)-based entries for autofs/fstab, rather than using hostnames
- NetworkBrowser relies on sunset detection methods not supported via IPv6 and deprecated since Windows 7/Windows Server 2012.
- NetworkBrowser relies on IPv4 port-scan (oe-a only)
- MediaPortal allows configuration of HLS-Server using a dotted IPv4 only (The HLS server socket is created as a Dual-Stack socket however)
- XMLTVImport uses IPv4-only code for downloading (OpenPLi only, fixed in oe-a)
- 99.9% of all pre-compiled oscams, e.g. those on feeds, are still IPv4-only

[littlesat]

Ot should be arranged oe adapt it... using work a rounds is note done... and I know sometimes it's a pain

Edited by littlesat, 4 July 2017 - 10:53.

[peti]

@SpaceRat: it's crashing on startup. I assume because the webif (to deny requests from different subnets without auth)

[peti]

ok, so the second part of this patch should fix the startup issue (the first is for the network settings):

--- a/lib/python/Components/Network.py	2017-07-02 09:47:54.000000000 +0200
+++ b/lib/python/Components/Network.py	2017-07-04 13:24:29.000000000 +0200
@@ -61,7 +61,7 @@
 		return [ int(n) for n in ip.split('.') ]
 
 	def getAddrInet(self, iface, callback):
-		cmd = ("/sbin/ip", "/sbin/ip", "-o", "addr", "show", "dev", iface)
+		cmd = "busybox ip -o addr show dev " + iface + " | grep -v inet6"
 		self.console.ePopen(cmd, self.IPaddrFinished, [iface, callback])
 
 	def IPaddrFinished(self, result, retval, extra_args):
@@ -179,7 +179,7 @@
 		currif = ""
 		for i in interfaces:
 			split = i.strip().split(' ')
-			if split[0] == "iface":
+			if split[0] == "iface" and split[2] != "inet6":
 				currif = split[1]
 				ifaces[currif] = {}
 				if len(split) == 4 and split[3] == "dhcp":

[SpaceRat]

The second part is redundant:
As grep filters out lines with "inet6", split[2] can not be inet6.

But it's ok for later enhancements.

[peti]

The second part is redundant:
As grep filters out lines with "inet6", split[2] can not be inet6.

But it's ok for later enhancements.

The second part is for interfaces file parsing

[SpaceRat]

Ah, ok.

I thought it was the parser of the "busybox ip" output.

[SpaceRat]

ok, so the second part of this patch should fix the startup issue (the first is for the network settings):

--- a/lib/python/Components/Network.py	2017-07-02 09:47:54.000000000 +0200
+++ b/lib/python/Components/Network.py	2017-07-04 13:24:29.000000000 +0200
@@ -179,7 +179,7 @@
 		currif = ""
 		for i in interfaces:
 			split = i.strip().split(' ')
-			if split[0] == "iface":
+			if split[0] == "iface" and split[2] != "inet6":
 				currif = split[1]
 				ifaces[currif] = {}
 				if len(split) == 4 and split[3] == "dhcp":

Merged ... in OpenATV
https://github.com/o...a1c38e542352c40

[Erik Slagter]

3. ip addr flush -> flush scope global only
https://github.com/o...bbc8c5c6523b3d7
https://github.com/o...84ef5919659f544 (Fixes previous commit)

I fixed this obvious error long ago. Not only does it prevent SLAAC from working, it also messes up the IPv6 stack in the kernel, it only starts working again once the interface is set down and up again.

[Erik Slagter]

Proper ipv6 handling (and with that, for the first time, I hope, proper ipv4 handling) in enigma is something that needs to be thought through carefully. For a start enigma should not call external binaries and trust their output, it should address the kernel interface itself, either using the C++ code or using some Python library (which apparently exists).

 

So it won't be in 6.1, more like 7 or 8.

Edited by Erik Slagter, 4 July 2017 - 17:57.