If you think a bug might affect users in the 57 release, please set the correct tracking and status flags for Release Management.

Windows freezes when updating multiple tabs in rapid succession

RESOLVED INVALID

Status

()

Firefox
General
--
critical
RESOLVED INVALID
8 years ago
7 years ago

People

(Reporter: preciseinfo, Unassigned)

Tracking

({hang})

3.6 Branch
x86_64
Windows XP
Points:
---

Firefox Tracking Flags

(Not tracked)

Details

(Whiteboard: [sg:needinfo])

(Reporter)

Description

8 years ago
User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.1.7) Gecko/20091221 Firefox/3.5.7
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.1.7) Gecko/20091221 Firefox/3.5.7

When reloading pages on several tabs in a rapid succession,
windows freezes. No mouse movement, no keyboard input possible.
Task manager shows 100% CPU use on 2 out of 4 cores.
You have to reboot. The system is totally dead.

But system is still running as task manager keeps updating
the permormance tab and keeps showing the cpu usage being update.

That may mean this is a serious security issue.

Reproducible: Sometimes

Steps to Reproduce:
To reproduce the problem, I'd have to provide you the login/passwd
info. I'd be willing to do that provided it is guaranteed to remain
confidential.

Actual Results:  
When you refresh pages in rapid succession before some page
has been fully loaded, what you get is total freeze of the box.

Expected Results:  
All these pages should complete loading without problems.
(Reporter)

Comment 1

8 years ago
The system is Athlon X4 6200 quad core processor
with Asus M4A78 motherboard running Win XP SP3.

This looks like a multi-core processor issue.
Everything works fine with a single core Athlon 2600 processor.

There are no addons for firefox except of iMacros.
You mentioned providing a login/password to be able to access a site that can reliably reproduce the behavior.  Would it be possible, instead, to save a copy of the page locally and verify that you can reproduce the hang by reloading the local copy?  If this is the case, you can attach the local copy to this bug report.
Whiteboard: [sg:needinfo]
(Reporter)

Comment 3

8 years ago
Ok, you can try to verify this issue by opening the following tabs
then doing a rapid page refresh for each tab that completed loading
the page.

Not guaranteed to happen on the 1st try.
But, as long as you keep refreshing those that completed
the page load, it should happen. Happens for me consistently.

http://webstats.motigo.com/s?tab=1&link=1&id=4585653&last=more
http://webstats.motigo.com/s?tab=1&link=1&id=4587547&last=more
http://webstats.motigo.com/s?tab=1&link=1&id=4584457&last=more
http://webstats.motigo.com/s?tab=1&link=1&id=4637611&last=more
http://webstats.motigo.com/s?tab=1&link=1&id=4644519&last=more
http://webstats.motigo.com/s?tab=1&link=1&id=4704184&last=more
http://www.sitemeter.com/?a=stats&s=s42mfcart&r=5
http://www.sitemeter.com/?a=stats&s=s42cppart&r=5
http://www.sitemeter.com/?a=stats&s=s42javaart&r=5
http://www.sitemeter.com/?a=stats&s=s32phpart&r=5
(Reporter)

Comment 4

8 years ago
Btw, memtest86 was ran for 20 hrs. with no problems.
Also, Prime95 load test was ran for several hours. No problems either.
BIOS was updated to the latest version.
(Reporter)

Updated

8 years ago
Version: unspecified → 3.5 Branch
(Reporter)

Comment 5

8 years ago
The same problem exists with ff 3.6 branch.
Version: 3.5 Branch → 3.6 Branch
do you see this when started in safe mode?
http://support.mozilla.com/en-US/kb/Safe+Mode
Keywords: hang
(Reporter)

Comment 7

8 years ago
The problems still exists in safe mode.
Windows task manager shows one of cores going up to 100% of CPU time.
But task manager still functions and updates the network and
performance tab charts.

Also, my monitoring firewall, that has traffic lights for incoming/
outgoing traffic, keeps flashing the lights. Meaning the program is
running normally.

But sometimes when it hangs, the monitoring firewall keeps the traffic
light lit up without flashing it, meaning it probably can not process
events.

Btw, all the links above are no longer password protected.
So you can try and see if it happens on your box.
Just open all those links in different tabs and keep refreshing
them before page completes loading.

Looks like a multi-core issue somewhere.
Does ff safe mode run a single thread?
(Reporter)

Comment 8

8 years ago
Btw, this hang only happens in ff so far.
Do not recall seeing anything else hang.
This is a relatively new box (< 2 months old).
But I do have some multithreaded programs that load up
the cpu, file system and networking pretty bad.
When they are run, the system is loaded up the hilt,
as heavy as it gets.

Never seen it hang with those apps.
I cannot reproduce using URLs in comment 3 and constant "reload all"s

Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9.3a1pre) Gecko/20100123 Minefield/3.7a1pre (.NET CLR 3.5.30729)
and
    File: NPSWF32.dll
    Version: 10.0.42.34
    Shockwave Flash 10.0 r42

perhaps your firewall is involved. do you see this problem with firewall off?
WFM as well.
Mozilla/5.0 (Windows; U; Windows NT 6.1; en-US; rv:1.9.3a2pre) Gecko/20100208 Minefield/3.7a2pre ID:20100208050429, which is running on two cores. New profile maybe?
(Reporter)

Comment 11

8 years ago
I doubt it can possibly have anything to do with the firewall.
It does have an intermediate NDIS driver (kernel mode driver).
But it runs for years without any problem.
I suspect if it was a firewall issue, I'd have a blue screen orgasm.

Sometimes you do have to go through refreshing pages about 5 times
to get it. The thing is hitting the refresh button on all the pages
as soon as page gets fully loaded. That way, most of the pages keep
showing the refresh rotating wheel.

Don't know what to think of if you are not getting the same issue
confirmed.

Someone recommended to run it under kernel debugger and see where
it concs out. The problem is I only have a single machine. I used
Softice kernel debugger but it no longer works under xp sp3, and
if I recall correctly windbg requires 2 machines to run.

This one looks like a tough one. Really nasty bug.
Looks like a dead loop somewhere. But that somewhere MUST have something
to do with the kernel level. Else, sooner or later, the events would get
processed even if one of cores got totally tied up.
But it is hard to say.
https://developer.mozilla.org/En/How_to_get_a_stacktrace_for_a_bug_report
https://developer.mozilla.org/en/How_to_get_a_stacktrace_with_WinDbg
 !analyze -v -hang

As far as I know you only need one machine.
Tyler: You need 2 system if you want to debug a freeze on the kernel mode level. How Do you want to run another application (the debugger) on the same system if the whole system is frozen ?

Reporter:
Do you tried to reproduce this on another system ?
In most cases this isn't caused by the kernel itself. I would try to disable/replace your kernel mode drivers like graphic card/sound/Firewall/av drivers.
Note: Disabling third party Firewalls often doesn't disable the driver.
This is a whole system kernel freeze then? Reporter was able to enter task manager and get it to respond, or so it seems from comment 1. Or I might be misunderstanding the issue behind this.
(Reporter)

Comment 15

8 years ago
Matthias,

>Do you tried to reproduce this on another system ?

Nope. It is not reproducible on my other system.
But that other system is a SINGLE core processor.

I am aware of device drivers needed to be updated.
I did what I could to get the latest version.

Tyler Downer,

>This is a whole system kernel freeze then?

Well, does not look like a whole kernel freeze.
Since task manager is updating its charts and a monitoring
firewall keeps flashing the traffic lights, the system is
still up. But the keyboard and mouse are totally unresponsive.
You need to reboot.

What I did as far as task manager and monitoring firewall goes
is to click on them in the windows task bar so they are on the top
BEFORE doing a load of all tabs that cause the freeze.
This way I could see at which exact point the freeze happens.

And every time it happens, one of the CPU cores shows up a 100%
load. Not sure which thread this is. So, it is impossible to say
which thread hangs.

The very fact that one of cores goes up to 100% of CPU time tells
me this is a dead loop. It is not a passive wait on some kernel
object. Otherwise, CPU core would not get totally loaded from what
I see.

But the system is still functional as long as you don't need the
keyboard/mouse input.
(Reporter)

Comment 16

8 years ago
Tyler Downer,

I can not get a windbg info because keyboard and mouse is not responsive.
(Reporter)

Comment 17

8 years ago
As far as crash reports go, the latest entry I see in
Application Data\Mozilla\Firefox\Crash Reports is
InstallTime20100115144158, which contains 1264722279.

There are no other crash reports I see.
All I see is 4 install entries (as they look to me).

FF does not crash. I hangs. Or rather SOMETHING hangs,
at least as far as keyboard/mouse go.
(Reporter)

Comment 18

8 years ago
Well, what I see is this could be a wait on a spin lock
in the kernel for one thing.

Is there anything that ff does on a kernel level?
(Such as some optimization for rendering performance
that uses a kernel mode driver)
Firefox itself is a full usermode application and it calls only the windows API function .
Do you see the looping process in the taskmanager or is it impossible to get into this state if the taskmanager is open and in front ?

There is nothing that we as triagers could do in this case and your symptoms make this bug invalid because Firefox triggers only a bug somewhere in your system. I think it would be worth to add a workaround if this would be a bug in the OS itself but I can't reproduce this on my win2003 system (quadcore) and on my dualcore vista system.
I'm inclined to agree that FF wouldn't lock the OS and that this is invalid against Firefox.

I would suggest, in no particular order, disabling firewall, install flashblock extension and use it to test your "reloads" with no flash images running, if you are reporting this as a regression (i.e. you didn't see this in a prior release) to retest the prior release.

removing sec-sensitive flag
Group: core-security
(Reporter)

Comment 21

8 years ago
So, from what I see, none of you guys, running the multicore XP SP3
can reproduce this.

And you did try to keep updating the pages several rounds.

Did I get it right?
(Reporter)

Comment 22

8 years ago
Yep, looks like it is not a FF bug.
This lockup has been seen without running FF for the first time.
FF simply makes it occur much more often for whatever reason.
Status: UNCONFIRMED → RESOLVED
Last Resolved: 8 years ago
Resolution: --- → INVALID
preciseinfo ... were you able to determine the cause?
You need to log in before you can comment on or make changes to this bug.