Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Weaver]

If you can get IP MTU 1500 + 8 = 1508 then you might as well have it, no danger of any incompatibility with remote apps (just theoretical as I said) and there is less than zero cost provided the overheads you have do not push you above 32 cells ie 1536 bytes in total, which is the limit of 32 ATM cell payloads * 48 bytes capacity per cell = 1536. Do if you have say 36 bytes overhead and 1500 bytes of IP then you are fine. More than 36 bytes of overhead and you need to reduce IP MTU to say 1492 to get the total eg 1492 + overhead below 1536.

In my case the overhead for PPPoEoA including PPP and the AAL5 CPCS is 32, so 32 + 1500 is not above 1536 so that is fine. The figure PPP + PPPoE + ethernet mac with no FCS + RFC 2684 (para. ‘bridged’, VC-MUX ) + AAL5 CPCS = 2 + 6 + 14 + 2 + 8 = 32. See wikipedia article on PPPoE, section on PPPoEoA overheads, for (what I hope is) a proper explanation.

[johnson]

So I have been running an 8324 with the jumbo frames + x1 dsl code for about a week now and am happy to report that it has solved the issue (I'm 99% sure is) related to vectoring. Thinking about making a post detailing that issue, but otherwise, have those that tried it had at least stable lines?

Would be great to find other users with vectoring enabled to confirm, but it seems like very few have it.

[j0hn]

Would be handy if anyone could work out how to allow multiple telnet instances. Could be another handy feature to add to customized firmware builds.

Earlier versions of Zyxel firmwares allowed this, then Zyxel randomly limited it to 1 session. It means if something like DslStats hangs then the modem effectively locks you out of telnet until the session eventually times out.

I'm looking at you Johnson 

[underzone]

£30 per feature I reckon 

[spring]

So I have been running an 8324 with the jumbo frames + x1 dsl code for about a week now and am happy to report that it has solved the issue (I'm 99% sure is) related to vectoring. Thinking about making a post detailing that issue, but otherwise, have those that tried it had at least stable lines?

Would be great to find other users with vectoring enabled to confirm, but it seems like very few have it.

Well I've flashed it today and it's better by 0.1 SNR etc and don't ask why but I ended up resynching 8 times and as always there's no sign of DLM.
Also definitely not more errors. We'll see.

[Weaver]

@spring do you find that you get changes of sync speed between resynchs when you haven’t changed anything? This used to happen for me, but nowadays such changes are really small.

[johnson]

Would be handy if anyone could work out how to allow multiple telnet instances. Could be another handy feature to add to customized firmware builds.

Earlier versions of Zyxel firmwares allowed this, then Zyxel randomly limited it to 1 session. It means if something like DslStats hangs then the modem effectively locks you out of telnet until the session eventually times out.

I'm looking at you Johnson 

So I had a play with what could be changed to allow more than one telnet/ssh session. Unfortunately messing with these firmwares is not much fun as all the important features and settings are hidden inside binaries with no source supplied in the GPL package.

Seems the change to telnet sessions was hardcoded in the telnetd binary as the "A telnet session already exist, cannot creat new telnet session" message is contained within it.

As far as I can tell this seems to be the only change to it since the version which allowed multiple sessions (all the way back with AAJZ.3 from 2014) so I just swapped in that version and it seems to work fine. This is probably both bad and wrong, but given we are using telnet anyway and it seems they have not upgraded it since release, I guess we are not that concerned about its possible security holes.

No joy trying the same approach with ssh sessions, as far as I can tell ssh doesnt even work for me with the AAJZ.03 release. 

So here are 4 more firmwares with the old telnetd binary to allow multiple sessions:

https://www.dropbox.com/sh/q4y2y2kbhb7mb1i/AAA6C520hnZVD0RtStqvRz_da?dl=0

As per the last link the -oldtelnet.bin just have the jumbo frames patches and old binary, with -3925.bin the updated adsl_phy.bin.

Hope someone finds them useful, but a warning that I have only tried the 1312-B10A version out, so maybe wait until I have tried the 8324, might do it in a minute actually.

[Weaver]

Many thanks johnson, that is very good of you.

I think £30 per feature is a bit cheap, ought to be more. 

[johnson]

Nah, its just fun messing with things to make them better.

Flashed the 8324 image to my running box btw, all good and now no more holes in my DSLstats graphs when I forget to close a telnet session!


Its probably a dumb idea as there are several other great stats monitoring methods, but I was looking into how to add slightly better monitoring to the web GUI. Its the same state of affairs as telnet & ssh though, everything hard coded into the httpd binary. No server side scripts to mess with. So would probably need add another small web server, something like libmicrohttpd, but could then serve more detailed things like snr or bitloading. Instantaneous things I guess as logging anything over time is tricky with the tiny amount of memory available.

Does that sound stupid? I feel like it would be quite nice to be able to have a quick look at the bit loading or SNR graph straight from the modem. Its probably harder than it seems and not worth it, but I'l look anyway.

[johnson]

Also Weaver, did you get another browser running that works properly with the GUI? Saw theres a nasty minified javascript file in the source that does something to do with "smoothscrolling" that could be the culprit of your safari woes.

[Weaver]

I tried Chrome but that was just the same, as I expected. Another Kitizen claimed that my opinion was out of date - I had read that all iOS web browsers are based on Webkit code or bits of core code used by Safari or whatever - and claimed that this was no longer the case. But I just don’t know the truth of it. It actually is perhaps not so simple. Regardless of the browser’s code, it could be that there is either capability detection or browser sniffing code in the javascript and maybe it detects one or the other browser but classes them all as one and the same ‘mobile’ nonsense thing and then this triggers broken behaviour.

The issue is one where eg the DSL statistics page is unreadable because part of it goes off the bottom of the screen. Unfortunately I can’t fix by turning Javascript off, not even turning it off selectively + temporarily, as the page in question is then not drawn at all.

[j0hn]

Sweet!
Zyxel limiting all their devices to 1 telnet session was a real pain.

I'm sure there's a few users still on the 6b1 firmware of the 8923-b10a as that's the last 1 that offered multiple telnet sessions.
This gives them the option to upgrade without losing that functionality.

Cheers Johnson!

[johnson]

I'm sure there's a few users still on the 6b1 firmware of the 8923-b10a as that's the last 1 that offered multiple telnet sessions.

I went back through them, but the latest available from the zyxel ftp folder without the silly telnet restriction was AAJZ.5C0, AAJZ.6C1 had it.

But like I say, if you compare the binaries (clumsily with 'strings') the only real change seems to be adding the "feature" of 1 session allowed. So dont think we're missing out on anything with the older version.

[johnson]

I tried Chrome but that was just the same, as I expected. Another Kitizen claimed that my opinion was out of date - I had read that all iOS web browsers are based on Webkit code or bits of core code used by Safari or whatever - and claimed that this was no longer the case. But I just don’t know the truth of it. It actually is perhaps not so simple. Regardless of the browser’s code, it could be that there is either capability detection or browser sniffing code in the javascript and maybe it detects one or the other browser but classes them all as one and the same ‘mobile’ nonsense thing and then this triggers broken behaviour.

The issue is one where eg the DSL statistics page is unreadable because part of it goes off the bottom of the screen. Unfortunately I can’t fix by turning Javascript off, not even turning it off selectively + temporarily, as the page in question is then not drawn at all.

Yeah without js it will not even be able to fetch the data for the stats page.

I had a go using the chrome debugger deleting all references to "overflow" which is I believe the style option that adds scroll bars, in the vain hope it would just move the buttons to the bottom of a long page, but all I ended up with was a broken mess.

Wish I could have proper go at fixing it but as I cant replicated it on any of my android devices it would just be stabbing in the dark.

[Weaver]

The page cannot be dragged to make it scroll. It moves upwards so far and then twangs back down again, as I may have said before. Sometimes you can see a tiny bit more content that is visible further down the window (ie in the -y direction) below the bar containing icons when you manage to pull the whole lot upwards for a second or so.