Perplexing PPP Problem

Another Not So Anonymous Coward is pulling hair out over this issue: "I think this may not be the most appropriate forum for this, but I am at my wits end. I have been asking around and presenting this to the top minds I know, and it baffles all of them. So far, not a single person has been able to explain this, even on the Linux newsgroups. I am setting up PPP. And I have a problem. I am not a complete fool, so I know how to set up PPP correctly (Linux RH6.0, if you were wondering). Everything works perfectly, it connects, it pings it traces routes as far as the eye can see. But, if you ask for a http request, pppd gets a brain fart and starts acting like a mad man. I added the argument "record /tmp/ppp.log" to pppd to see what was happening, and I found the answer. Whenever I make a connection, pppd recieves the first 3-4000 characters of the page, I can see it in the log. I ususally get about 1/2 a page, BUT then it coughs. It waits a couple of seconds, and then STARTS OVER. It recieves again the first 3000 characters and hiccups again. So, I can use the web if I want to look at pages with less than 4000 characters, but anything more than that and it just hangs forever calling up the heads of web-pages. Does anybody have a clue as to why this is happening? (I know there is no porblems on the remote side becuase when I use windows 98, everything works peachy. It has to be my shiny new Linux box... grrrr!) " This sounds really odd. I admit that I'm stumped. Do any of you kind folks have any clues?

Comments

MTU Settings?

[Anonymous Coward]

Have you tried setting the MTU/MRU to something sane? Try putting 'mtu 1500 mru 1500' into your /etc/ppp/options file.

Post the options you're using

[Anonymous Coward]

It would be helpful if you could post all the options you are using from your /etc/ppp/options file and from the command line. Wait, youre not using that icky redhat config tool are you? Eek! :)

Some questions - and one suggestion

[Anonymous Coward]

Err .. I don't know if this will help but here goes with some questions

- What browsers have you tried? Do you get this with Netscape*, lynx and others?

- Have you any specific url's that fail or pass?

I have noticed very occasional pages which work with Windows* browsers but rarely if ever with my Netscape and lynx on Linux*

The cause seems to be a large number of lines overwhelming the storage capabilties of the processes on my small Linux machine.

Specifically there are a lot of useless carriage returns (0x013) embedded in Javascript widgets in this page - of which the first 2K will load before the http link fails "stalled"

http://ds.dial.pipex.com/town/plaza/ec91/

Try this link at your own risk! Oddly it seems to be working for me today.

AC David _

* Linux is a registered trademark of Linux Torvalds

Windows is a registered trademark of Microsoft Netscape is a registered tradqemark of America Online

AC is a trademark of slashdot

What is the slashdot policy on trademark attributions?

Re:Some questions - and one suggestion

[Anonymous Coward]

AC David revising his deliberate mistake

The suspect is not carrage returns, but line feeds as far as I remember

I am suspicious because it is possible that Windows is handling line feeds differently.

Anyway I didn't get near the bottom of this problem - so take my comments for what they are with.

Re:Some questions - and one suggestion

[Royster]

What is the slashdot policy on trademark attributions?

It can be found at the bottom of every page. Ergo:

All trademarks and copyrights on this page are owned by their respective owners. Comments are owned by the Poster. The Rest © 1997-99 Andover.Net.

HTTP Stall

[Julz]

Does the command line http transfer work okay?
I was wondering if it might be something to do with your browser.

Re:HTTP Stall

[rips]

Does FTP data seem to work?? When you say it stops and resends, have you had a look to see if its resending at the PPP level or at the TCP/IP level? If your socket is being disconnected at the TCP/IP level then I'd be trying a different browser or something at the application level. If not and FTP works, then I'm stumped too. :) Its hard to do much with the amount of information you've provided.

MTU and Redhat6/2.2x kernel

[Anonymous Coward]

I read on the IP Masq How-to, that MTU's of anything other than 1500 on older 2.0.x kernels caused problems if the web server sent a "don't fragment" flag on any of the packets. It would cause them to be dropped. Supposedly, this problem doesn't exist in the 2.2.x kernel, so redhat 6 shoudln't be directly affected by that bug, at least. I know hotmail.com was giving me issues about that until I changed my MTU on my firewall back to 1500. (I liked 576)

PPP or TCP

[snookums]

Is this perhaps some strange behaviour of the TCP layer? (or a bad interaction between PPP and TCP layers?) Do you have a packet trace of the connection? If you get one then you might be able to narrow down just who it is that is asking for the re-sends.

Hi, this is me, the original poster.

[clambake]

I found the solution, well, kind of. My MTU and MRU settings were fine, and I tried many different browsers. I was using pppd 2.3.7 (or whatever is standard issue with RH 6.0). I updated it to pppd 2.3.>9particular problem, so it is possible that whatever bug caused it may crop up again.

Thanks for all your help!

Re:Hi, this is me, the original poster.

[Anonymous Coward]

Do you have an ethernet card as well in that machine? If so, and you are using the redhat control-panel thing, try deactivating the ethernet card before you activate your ppp device. My computer exhibits much the same problem and that takes care of it. It has something to do with the dns settings.

Re:Hi, this is me, the original poster.

[Anonymous Coward]

No, DNS doesn't have anything to do with it. Why should something stop at the same point through an active connection and make a DNS lookup? If your machine does that midway through a connection, is it trying to balance the load? Doubtful, but possible.

flow control?

[Anonymous Coward]

Check the flow control settings on your modem. My roommate had a similar problem, where it would only download the first 4k of binary files (though text, such as html, was not a problem). Fixing the flow control solved the problem. Don't know what the correct settings were, however.

Re:flow control?

[Anonymous Coward]

i have had the same experience. this can often be remedied by using the proper init string for the modem. before you get too deep into the various hayes commands, just try "AT&F" and "AT&F1"

Whats your asyncmap setting? compression?

[TeddyR]

Try adding

asyncmap 200a0000

to the "extra options"


if the other side requires compression, you may want to try modprobing bsd_comp and/or ppp_deflate


https://www.mav.net/teddyr/syousif/

I'll bet it doesn't happen on handcoded pages...

[Any_doom?_a_cow_runs]

...all these automatic pages are the only ones that have ever given me any trouble. Full of so much GARBAGE :) might confuse something.

I hand code the very few pages I do, and nary a problem with anyone.

Anonymous Coward, get it? :)