Bug 199753

Summary: stalled web access after upgrade to 2.6.17-1.2157_FC5
Product: [Fedora] Fedora Reporter: Claus Olesen <colesen>
Component: kernelAssignee: John W. Linville <linville>
Status: CLOSED NOTABUG QA Contact: Brian Brock <bbrock>
Severity: high Docs Contact:
Priority: medium    
Version: 5CC: davej, wtogami
Target Milestone: ---   
Target Release: ---   
Hardware: i386   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2006-08-28 21:41:08 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:

Description Claus Olesen 2006-07-21 19:09:40 UTC
Description of problem:
the symptom is that web access using a browser stalls with browser saying 
waiting for reply from the web site to which access is attempted.

Version-Release number of selected component (if applicable):


How reproducible:
always

Steps to Reproduce:
1. upgrade to latest kernel 2.6.17-1.2157_FC5
2. try accessing Internet sites (using browser)
3.
  
Actual results:
browser stalls without ever getting a reply from the Internet.

Expected results:
the Internet should reply allowing the browser to finish.

Additional info:
the problem vanishes if instead kernel 2.6.16-1.2133_FC5 is used.

also, when the problem is present then dmesg has this
b44.c:v1.00 (Apr 7, 2006)
eth0: Broadcom 4400 10/100BaseT Ethernet 00:02:e3:53:5d:b2
b44: eth0: Link is up at 100 Mbps, full duplex.
b44: eth0: Flow control is off for TX and off for RX.

when the problem is not present then dmesg has this
b44.c:v0.97 (Nov 30, 2005)
eth0: Broadcom 4400 10/100BaseT Ethernet 00:02:e3:53:5d:b2
b44: eth0: Link is up at 100 Mbps, full duplex.
b44: eth0: Flow control is on for TX and on for RX.

(strangely? google and sbc are immune to the problem).

Comment 1 John W. Linville 2006-07-26 14:06:20 UTC
Would you mind trying the fedora-netdev kernels in order to ensure that this
problem is still pending upstream?

   http://people.redhat.com/linville/kernels/fedora-netdev/

Please give them a try and post the results here...thanks!

Comment 2 Claus Olesen 2006-07-26 15:30:42 UTC
I tried but got no update...

$ sudo yum update
Loading "installonlyn" plugin
Setting up Update Process
Setting up repositories
core                                                                 [1/4]
fedora-netdev                                                        [2/4]
updates                                                              [3/4]
extras                                                               [4/4]
Reading repository metadata in from local files
No Packages marked for Update/Obsoletion


Comment 3 Danny Ciarniello 2006-07-27 15:47:33 UTC
I have also run into this problem but only with one site: http://www.csfm.com

To reproduce, do wget http://www.csfm.com:

wget http://www.csfm.com
--08:48:52--  http://www.csfm.com/
           => `index.html.1'
Resolving www.csfm.com... 216.187.71.116
Connecting to www.csfm.com|216.187.71.116|:80... connected.
HTTP request sent, awaiting response...

I have tried different versions of the 2.6.17 kernel on fc4 (i386) and fc5
(x86-64) and all exhibit this problem.

I have also tried 2.6.16 and 2.6.15 kernels on fc4 and 2.6.9 kernels on fc3 and
RHEL and all work.

Comment 4 Pinnacle Support 2006-08-01 13:00:03 UTC
I have noticed the same problem with a few of our servers. 2.6.16 kernels work
fine, 2.6.17-1.2157_FC5 just hangs when connecting using various protocols (have
had it fail when connecting with http, ftp, pop and smtp).  Network card being
used in this case is an Intel Pro100 (e100 driver). 

Comment 5 John W. Linville 2006-08-01 18:39:21 UTC
Re: comment 2 -- please try the update again, as I have published later 
FC5.netdev kernels...thanks!

Comment 6 Claus Olesen 2006-08-02 17:55:21 UTC
I tried again - and got a new kernel
uname -a
Linux saturn 2.6.17-1.2157.2.1_FC5.netdev.7 #1 Fri Jul 28 17:30:46 EDT 2006 
i686 i686 i386 GNU/Linux

However, the problem is still there - web pages hang.
dmesg
b44.c:v1.01 (Jun 16, 2006)
eth0: Broadcom 4400 10/100BaseT Ethernet 00:02:e3:53:5d:b2
b44: eth0: Link is up at 100 Mbps, full duplex.
b44: eth0: Flow control is off for TX and off for RX.


Comment 7 John W. Linville 2006-08-28 19:47:46 UTC
Can you attach the output of runnint 'sysreport' on the box in question?  
Thanks!

Comment 8 Claus Olesen 2006-08-28 21:16:09 UTC
The problem that I had has been resolved.

To reiterate, after upgrading from
2.6.16-1.2133_FC5
to
2.6.17-1.2157_FC5
most web sites did not respond to web page requests using a web browser.

One of our sys admins looked into the matter. He tells me that based on this 
article
http://www.mail-archive.com/pf%40benzedrine.cx/msg07837.html
he changed the configuration of the firewall (which runs BSD) that I'm behind 
to ensure that the pass rules match these criterions

1) all 'pass' rules that can match TCP have 'flags S/SA'
2) all 'pass' rules have 'keep state'

After that the problem vanished. Looks to me like the latest kernel changed 
something for which the firewall configuration had to be changed (or maybe 
should have been changed to a long time ago).

If you still want the 'sysreport' I will attach it. Or should I close this 
report?


Comment 9 John W. Linville 2006-08-28 21:41:08 UTC
Claus, sysreport will not be necessary...thanks!