Closed Bug 70520 Opened 24 years ago Closed 23 years ago

SSL through a proxy is still problematic.

Categories

(Core :: Networking, defect)

Product:
Core
Component:
Networking
Platform:
x86
Linux
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
VERIFIED FIXED

People

(Reporter: mrsam, Assigned: neeti)

Details

I am still having problems using SSL through a junkbuster proxy (see bug 31174).
 Using the last nightly build, with the following hack:

user_pref("network.http.proxy.keep-alive", false);

Mozilla is configured to talk through a junkbuster proxy.

About 50-60% of the time I can usually load something like
https://sourceforge.net without any problems.  If I continue to browse
Sourceforge in SSL, after about 2-3 pages I always begin to experience problems.
Sometimes it happens right away, on the very first SSL page, about 4-5 SSL pages
before I hit a problem.

Initially, Mozilla seems to stall for about 1-2 minutes for each request, with
the throbber running, but doing absolutely nothing otherwise.  The junkbuster
proxy is on a different machine, and while Mozilla's spinning, I do not see any
connections from Mozilla to the proxy, and no outbound connections to sourceforge.

After about a minute, Mozilla decides to finally load and display the page
(during this delay, browsing Sourceforge from Netscape 4.76 in SSL shows nothing
unusual, which eliminates a server problem on the remote end).

A few clicks later, Mozilla decides to simply stop working, and one of the
mozilla-bin processes decides to start eating 100% of the CPU.  stracing the
process shows that this is what it's doing:

recv(23, "", 8, 0)                      = 0
recv(23, "", 8, 0)                      = 0
recv(23, "", 8, 0)                      = 0
recv(23, "", 8, 0)                      = 0
recv(23, "", 8, 0)                      = 0
recv(23, "", 8, 0)                      = 0
recv(23, "", 8, 0)                      = 0
recv(23, "", 8, 0)                      = 0
recv(23, "", 8, 0)                      = 0
recv(23, "", 8, 0)                      = 0
recv(23, "", 8, 0)                      = 0

Ho-hum.  Poking around shows that file descriptor 23 is connected to a UNIX
domain socket, and that socket is also opened by another 5-6 mozilla-bin
processes who are idling, and not doing anything much.

This is 100% repeatable, with the only random factor being how many SSL pages I
try to load before Mozilla starts failing.
Related to bug 65703? (Runaway PSM hogs CPU time and generates threads.)
Related to bug 64621 (SourceForge SSL site freezes browser).
Maybe we best wait for PSM 2.0 
(http://www.mozilla.org/projects/security/pki/psm/plan_20.html).
CC'ing dougt.. sounds familiar?
Reporter this still a problem in the latest nightlies?
Build 2001-04-10 is working for me.  Haven't tried anything more recent.


Marking WORKSFORME based on reporters comments.
Status: UNCONFIRMED → RESOLVED
Closed: 23 years ago
Resolution: --- → WORKSFORME
SSL through a proxy broke for me again. <sigh>

To reproduce:

1. Configure mozilla to use a proxy.

2. Go to sourceforge.net's https login page:

https://sourceforge.net/account/login.php

3. After submitting the form, mozilla will appear to try to contact the server,
but will abort, clear the screen, and display the following in the main window:

<html><body></body></html>

I reproduce this both in build 0.9.2, and 2001070612.

NOTE: I have both keep alive and http pipelining TURNED OFF in the debug menu.

NOTE: I can use https to login and browse other sites.  This is something that's
particular to sourceforge.
Status: RESOLVED → UNCONFIRMED
Resolution: WORKSFORME → ---
Bug appears to be fixed again in build 2001070916.
Status: UNCONFIRMED → RESOLVED
Closed: 23 years ago → 23 years ago
Resolution: --- → FIXED
Verified.
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.