Bug 1383 - Tons of "rcvd too much data win" messages since upgrading to 4.7p1
Summary: Tons of "rcvd too much data win" messages since upgrading to 4.7p1
Status: CLOSED FIXED
Alias: None
Product: Portable OpenSSH
Classification: Unclassified
Component: ssh (show other bugs)
Version: 4.7p1
Hardware: ix86 Linux
: P2 normal
Assignee: Assigned to nobody
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2007-10-25 07:13 AEST by Jeff Mitchell
Modified: 2008-07-22 12:20 AEST (History)
1 user (show)

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Jeff Mitchell 2007-10-25 07:13:35 AEST
I often use SSH to connect to a system for which I have "DynamicForward 1080" configured, and then use this as a dynamic SOCKS proxy.

Since upgrading to 4.7p1, I get tons and tons of "rcvd too much data X win Y" messages on the console as I browse around the Internet, and pages randomly stop loading, sometimes immediately, sometimes after part of it has been loaded.  X is some number that is not always the same, same for Y.

Sometimes I get a few of these messages on one page.  Sometimes the ssh window just scrolls and scrolls with them.

Reverting to 4.6p1 fixed this issue.

I'm running Gentoo using their "stable" ebuilds, so it's compiled for my system, and not using any dubious package or anything.
Comment 1 Darren Tucker 2007-12-29 09:43:56 AEDT
(In reply to comment #0)
> I often use SSH to connect to a system for which I have "DynamicForward
> 1080" configured, and then use this as a dynamic SOCKS proxy.
> 
> Since upgrading to 4.7p1,

Was that upgrading the server, client or both?

> I get tons and tons of "rcvd too much data X
> win Y" messages on the console as I browse around the Internet, and
> pages randomly stop loading, sometimes immediately, sometimes after
> part of it has been loaded.  X is some number that is not always the
> same, same for Y.

Could you please attach (using "create attachment") the entire output from "ssh -vvv yourserver" showing the errors (and the server side "/path/to/sshd -ddd" if possible)? 

> Sometimes I get a few of these messages on one page.  Sometimes the
> ssh window just scrolls and scrolls with them.
> 
> Reverting to 4.6p1 fixed this issue.
> 
> I'm running Gentoo using their "stable" ebuilds, so it's compiled for
> my system, and not using any dubious package or anything.

We have just found a and fixed a somewhat similar problem (bug #1360).  You may want to try the patch on that bug (attachment #1426 [details])
Comment 2 Jeff Mitchell 2008-01-11 03:22:49 AEDT
This was upgrading the client.

I'll test out with the patch and get the requested information back to you soon.
Comment 3 Darren Tucker 2008-01-11 11:42:17 AEDT
I was not able to replicate this problem.  Is the server OpenSSH (and if so, which version?) or something else?
Comment 4 Jeff Mitchell 2008-01-11 12:40:11 AEDT
The server is OpenSSH, which is why it's on this Bugzilla.  The version is 4.71p1 as stated in the subject and the the Version field on this bug report.

I switched back to 4.7p1 today and I'm trying to replicate this.  It may take a few days...I didn't see it happening *all* the time.  Or maybe I'll never see it again...we'll have to see.
Comment 5 Jeff Mitchell 2008-01-12 03:31:15 AEDT
I unfortunately can't get the server side logs, but here's a log with -vvv where the connection simply died on me with these messages:

debug1: channel 29: new [dynamic-tcpip]
debug2: channel 29: pre_dynamic: have 0
debug2: channel 29: pre_dynamic: have 3
debug2: channel 29: decode socks5
debug2: channel 29: socks5 auth done
debug2: channel 29: pre_dynamic: need more
debug2: channel 29: pre_dynamic: have 0
debug2: channel 29: pre_dynamic: have 5
debug2: channel 29: decode socks5
debug2: channel 29: socks5 post auth
debug2: channel 29: pre_dynamic: need more
debug2: channel 29: pre_dynamic: have 23
debug2: channel 29: decode socks5
debug2: channel 29: socks5 post auth
debug2: channel 29: dynamic request: socks5 host support.fnfr.com port 80 command 1
debug2: channel 29: open confirm rwindow 131072 rmax 32768
debug2: channel 29: window -98786 sent adjust 98797
channel 29: rcvd too much data 2760, win 11
channel 29: rcvd too much data 5520, win 11
channel 29: rcvd too much data 2760, win 11
channel 29: rcvd too much data 4140, win 11
channel 29: rcvd too much data 2760, win 11
channel 29: rcvd too much data 897, win 11
debug2: channel 27: window -99034 sent adjust 99033
debug2: channel 27: window -99909 sent adjust 99908
debug2: channel 27: window -98392 sent adjust 98391
debug2: channel 27: rcvd adjust 65697
debug2: channel 27: window -99555 sent adjust 99554
debug2: channel 27: window -99893 sent adjust 99892
debug2: channel 27: window -98652 sent adjust 98651
debug2: channel 27: window -100137 sent adjust 100136
debug2: channel 27: window -100635 sent adjust 100634
debug2: channel 27: window -100965 sent adjust 100964
debug2: channel 27: window -99478 sent adjust 99477
debug2: channel 29: rcvd eof
debug2: channel 29: output open -> drain
debug2: channel 29: obuf empty
debug2: channel 29: close_write
debug2: channel 29: output drain -> closed
debug2: channel 29: read<=0 rfd 36 len 0
debug2: channel 29: read failed
debug2: channel 29: close_read
debug2: channel 29: input open -> drain
debug2: channel 29: ibuf empty
debug2: channel 29: send eof
debug2: channel 29: input drain -> closed
debug2: channel 29: send close
debug3: channel 29: will not send data after close
debug3: channel 29: will not send data after close
debug2: channel 29: rcvd close
debug3: channel 29: will not send data after close
debug2: channel 29: is dead
debug2: channel 29: garbage collecting

I'll attempt the patch that you linked to and see if that helps.
Comment 6 Jeff Mitchell 2008-01-12 04:16:44 AEDT
The patch did not fix the issue.  Still getting dropped connections.  Went back to OpenSSH 4.6p1 and everything worked again.
Comment 7 Darren Tucker 2008-01-22 15:34:00 AEDT
(In reply to comment #4)
> The server is OpenSSH, which is why it's on this Bugzilla.  The version
> is 4.71p1 as stated in the subject and the the Version field on this
> bug report.

Just so I have this clear:

* the server is an unmodified OpenSSH 4.7p1 and always has been.
* the client *was* an unmodified 4.6p1 and had no problems.
* the client was upgraded to an unmodified 4.7p1 and the problem occurred.
* the client was downgraded to 4.6p1 and the problem stopped.

Please attach (use "Add an attachment") the *ENTIRE* debug output from the client.
Comment 8 Jeff Mitchell 2008-01-22 23:56:46 AEDT
* the server is an unmodified OpenSSH 4.7p1 and always has been.

I made a typo in comment #4.  I meant the client is OpenSSH 4.7p1.  I've just checked what the server is; it's a SunOS 5.10 box running OpenSSH.  Here's the string I get when telnetting to port 22:

SSH-1.99-OpenSSH_4.2

* the client *was* an unmodified 4.6p1 and had no problems.

Yes, with Gentoo patches.

* the client was upgraded to an unmodified 4.7p1 and the problem
occurred.

Yes, with Gentoo patches.  I can disable these patches if you want me to test that.

* the client was downgraded to 4.6p1 and the problem stopped.

Yes.

I can attach the entire debug output with things going haywire if you want (I assume you want -vvv).  Do you want it with the 4.7 I've been using (i.e. with Gentoo patches) or a totally unmodified tarball?
Comment 9 Darren Tucker 2008-01-30 13:35:01 AEDT
(In reply to comment #8)
> I can attach the entire debug output with things going haywire if you
> want (I assume you want -vvv).

Yes, please attach the whole thing via "Add an attachment"/

> Do you want it with the 4.7 I've been
> using (i.e. with Gentoo patches) or a totally unmodified tarball?

Unmodified tarball please.

Basically I need to either get enough information to figure out what's going on from the debug output, or to be able to replicate your problem here.
Comment 10 Jeff Mitchell 2008-04-04 02:25:01 AEDT
Upon further testing, this seems to be a problem with the HPN patch in 4.7.  I was using HPN in 4.6 as well, which was working fine, but after starting with a stock tarball and then slowly adding patches back, one at a time, HPN seems to be the culprit.

I'm currently running the standard Gentoo patches with various features, minus HPN, on 4.7, and things have been well for a while.

The 4.6 ebuild uses the hpn12v17.diff.gz file, the 4.7 ebuild uses the hpn13v1.diff.gz file, and 4.9 uses hpn13v2.diff.gz.  Any known issues with any of those?
Comment 11 Darren Tucker 2008-06-12 08:11:53 AEST
I vaguely recall a problem like this with one of the HPN versions, but you would need to ask the HPN guys about this.  Since the problem isn't with the software that we provide I am closing this bug.
Comment 12 Damien Miller 2008-07-22 12:20:06 AEST
Mass update RESOLVED->CLOSED after release of openssh-5.1