Print

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

[roseway]

For Nodens:

I've made some more small changes and produced a debug version (Windows only) for you to try:

http://s446074245.websitehome.co.uk/files/rs-w32-2.54-debug.zip

[Nodens]

 Log files are attached. It still seems like there's something strange with the flow control or something.. I don't know. Logs still look similar although now
I've seen the password getting "cut" in 2.. in some cases log is showing 2 chars of the password sent and then the other 6. And in a couple of cases
it seems like the last 7 chars from the password sent and then the first char of the password appended to the "enable" string?!? I've marked it in the log
but for a clearer example let's say the password is "FGi7j872", password was sent as "Gi7j872" and the enable line after that was "Fenable".. I find this
very strange..

On a sidenote, you will see that it managed to connect once (but I believe it a purely random event, as I'm unable to reproduce it) but even in that case
the log does not show "Username:", "Password:" or rs-w32 answers. I wonder if the initial greeting screen is causing the issue? Encoding issue maybe?

The first log is with the default options. The other 3 (including the one that managed to connect) were made with :/: as Username:/Password: expressions.

I wonder if a timed option of let's say 2000 milliseconds between each response would manage to make it login. That would verify a certain issue in parsing
modem telnet output..
I know the modem's firmware has been modified by the Greek ISP as they purchase these in bulk and modify the firmware for default settings and web interface
theme. Perhaps they messed up with the char encoding?

Also what happens if you use only user/pass of Special Login to telnet in another modem (one of those supported in normal login sequence). Does it work this way?
This test would show Special Login flow control is working properly..

Thank you for your time!

[roseway]

Also what happens if you use only user/pass of Special Login to telnet in another modem (one of those supported in normal login sequence). Does it work this way?

It works fine with the Huawei HG612 which I use. This model does use a second stage of logging in to get to the BusyBox shell, and the special login deals with it with no problems.

Your suggestion of a time delay between commands is a good one. As some of the commands seem to lose their first character, this may simply be the result of a timing issue.

I won't be able to do any more until tomorrow, but I'll have a look at your logs and your other points then.

[roseway]

Here's another version with debug symbols. I've added some options to the special login setup, so you can enable/disable a time delay before the login commands, and set the time to different values. You can also enable/disable the diagnostic logging, although it will be best to keep it enabled at this stage.

http://s446074245.websitehome.co.uk/files/rs-w32-2.55-debug.zip

[Nodens]

Ok this is even stranger:/ As long as I enable the millisecond delay, it fails to logon immediately (<=1sec). No matter what I set the value to. 10000 produces the same result as 100ms. Also with the delay enabled, it doesn't log anything..

In order to rule any Win 8 operating system quirks out of the equation and possibly anything in my LSP chain interfering (NOD32 etc) I also tried to run under the VMWare WinXP VM that I use to test my software under (I should probably have done that sooner). No difference at all. The Win XP VM, is stock XP, latest service pack no locale settings whatsoever or anything other than VMWare Tools.

Then I did something else I should have tried sooner as well. I run the Linux 2.51 version on the Fedora box I also use for development. Same results... On exit it also threw an unhandled exception:

[ntt@wintermute rs-ux64-2.51]$ ./rsux
TLResourceList.Sort 53 DUPLICATE RESOURCE FOUND: TBitchart:PNG
TLResourceList.Sort 54 DUPLICATE RESOURCE FOUND: TBitchart:PNG
TLResourceList.Sort 57 DUPLICATE RESOURCE FOUND: TLinechart:PNG
TLResourceList.Sort 58 DUPLICATE RESOURCE FOUND: TLinechart:PNG
TLResourceList.Sort 60 DUPLICATE RESOURCE FOUND: TLinediagchart:PNG
TLResourceList.Sort 61 DUPLICATE RESOURCE FOUND: TLinediagchart:PNG
An unhandled exception occurred at $0000008312D00000 :
EAccessViolation : Access violation
  $0000008312D00000
  $0000000000518227
  $0000000000518245
  $0000000000518227
  $0000000000518227
  $0000000000518245
  $00000000005181D4
  $000000000051928D
  $00000000005707EE
  $000000000057082E
  $0000000000831342
  $0000000000838462

[ntt@wintermute rs-ux64-2.51]$

This proves the issue is not operating system dependent and since Special Login works properly with your modem, it seems it's specific to the ZTE device. I'm leaning again on failure to parse modem output due to encoding? I managed to get it to connect one more time randomly, telnet data was visible (connection stats and bitloading) but SNR Margin graph was still blank. I'm taking a wild guess that rs-w could not understand the output. I can not think of a way to check what character encoding is used:/


EDIT: The only thing I can find in the shell is the terminal emulation is set to VT102 (According to /proc/self/environ). $LANG is not set.

[GigabitEthernet]

Eric, I am a bit confused as to why you have included the 'Linksys X3000' as a pre-configured modem, because it doesn't support telnet!

In addition, on the SNRM tweaking page, at least on Linux, the instructions end with 'new'.

[roseway]

Eric, I am a bit confused as to why you have included the 'Linksys X3000' as a pre-configured modem, because it doesn't support telnet!

In addition, on the SNRM tweaking page, at least on Linux, the instructions end with 'new'.

The answer to the first question is very simple - I made a mistake. I'll remove it from the next version.

Your second point is a fonts issue - the fonts on your system are using more vertical space than I expected. I'll increase the space in the next version to deal with this.

[GigabitEthernet]

Thanks Eric!

[roseway]

Ok this is even stranger:/ As long as I enable the millisecond delay, it fails to logon immediately (<=1sec). No matter what I set the value to. 10000 produces the same result as 100ms. Also with the delay enabled, it doesn't log anything..

That's a result of a handled exception, so it appears that any delay added at that point causes an exception.

Then I did something else I should have tried sooner as well. I run the Linux 2.51 version on the Fedora box I also use for development. Same results... On exit it also threw an unhandled exception:

[ntt@wintermute rs-ux64-2.51]$ ./rsux
TLResourceList.Sort 53 DUPLICATE RESOURCE FOUND: TBitchart:PNG
TLResourceList.Sort 54 DUPLICATE RESOURCE FOUND: TBitchart:PNG
TLResourceList.Sort 57 DUPLICATE RESOURCE FOUND: TLinechart:PNG
TLResourceList.Sort 58 DUPLICATE RESOURCE FOUND: TLinechart:PNG
TLResourceList.Sort 60 DUPLICATE RESOURCE FOUND: TLinediagchart:PNG
TLResourceList.Sort 61 DUPLICATE RESOURCE FOUND: TLinediagchart:PNG
An unhandled exception occurred at $0000008312D00000 :
EAccessViolation : Access violation
  $0000008312D00000
  $0000000000518227
  $0000000000518245
  $0000000000518227
  $0000000000518227
  $0000000000518245
  $00000000005181D4
  $000000000051928D
  $00000000005707EE
  $000000000057082E
  $0000000000831342
  $0000000000838462

[ntt@wintermute rs-ux64-2.51]$

This proves the issue is not operating system dependent and since Special Login works properly with your modem, it seems it's specific to the ZTE device. I'm leaning again on failure to parse modem output due to encoding? I managed to get it to connect one more time randomly, telnet data was visible (connection stats and bitloading) but SNR Margin graph was still blank. I'm taking a wild guess that rs-w could not understand the output. I can not think of a way to check what character encoding is used:/

When I run it in Linux 64 from the command line I see the same "DUPLICATE RESOURCE" lines but no other information, so it does appear that the problem is related to the ZTE router. I don't think it's likely to be due to character encoding, because in the last two versions I've removed the checking of the prompts, and I simply wait for a response (any response) from the router before sending the next command. The fact that it sometimes half-works in a random way strongly suggested that it was a timing issue. I think if it was character encoding it wouldn't have this random effect.

EDIT: The only thing I can find in the shell is the terminal emulation is set to VT102 (According to /proc/self/environ). $LANG is not set.

The only small point that occurs to me is the possibility of the terminal emulation using a different newline sequence, but this seems very unlikely. I'm assuming Unix line endings. As before, I don't think this would have a random effect.

[krypton]

@Nodens

Do you use a wireless connection?

Does it work if you manually start another telnet server, for example at port 2323, and connect to it:

Code: [Select]

telnetd -p 2323 -l /bin/sh

[roseway]

Ok this is even stranger:/ As long as I enable the millisecond delay, it fails to logon immediately (<=1sec). No matter what I set the value to. 10000 produces the same result as 100ms. Also with the delay enabled, it doesn't log anything..

I've had another look at this, and I've realised that the way in which I implemented the delay may be responsible for this result. The program is multithreaded, and the time delay code that I added broke multithreading rules, so it's not surprising that it threw an exception. I'm still not clear why it works with my router and not yours though, so there's more work to be done.

[Nodens]

Ok this is even stranger:/ As long as I enable the millisecond delay, it fails to logon immediately (<=1sec). No matter what I set the value to. 10000 produces the same result as 100ms. Also with the delay enabled, it doesn't log anything..

I've had another look at this, and I've realised that the way in which I implemented the delay may be responsible for this result. The program is multithreaded, and the time delay code that I added broke multithreading rules, so it's not surprising that it threw an exception. I'm still not clear why it works with my router and not yours though, so there's more work to be done.

I was just looking for the reason the delay fails in the debugger (by the way rsw.exe 2.55 is not built with symbols either). I noticed a thread calling _SockWaitForSingleObject() and _WSPSelect() that is dying very fast. I could not discern why though as I could not see any exception handler test jumps taken (although it's possible I missed something). I'm glad you figured it out though!

@morphium

I use ethernet connections for everything other than an android device.

For some reason I can't understand, I'm unable to start a new telnetd process:/ I get no error message but no new server starts. I tried several port numbers in case I was hitting ports that were in use by some other process..

[roseway]

I've corrected the time delay code so it remains within the telnet login thread, so it shouldn't throw an exception now. I'm going to make this version a general release tomorrow, because I don't want to delay the fixing of several other little issues. The "special login" section is still definitely experimental.

[ryant704]

Is the error reporting for FEC broken or am I actually getting 15, 000+ per min? This leading to 353m in 3 days?

[ZeeGy]

Is the error reporting for FEC broken or am I actually getting 15, 000+ per min? This leading to 353m in 3 days?

Quite possibly, im getting on average <60CRC's per minute, and no FEC's at all