1373365 - Page load stall, related to Avast antivirus

Status: RESOLVED FIXED

(External Software Affecting Firefox :: Other, defect, P2)

Description

[Alon Zakai (:azakai)]

Details:

1. This is on https://www.kitely.com/
2. When the bug happens, loading the page on Firefox stalls and takes tens of seconds before the page is rendered and loaded, while on other browsers it's very snappy.
3. This happens on my friend's machine but not mine. It appears to show up in his google analytics reports too, where Firefox load time is 3.6x slower than other browsers, so it's not just a problem on his machine, and he's worried about losing users that are on Firefox.
4. Disabling his antivirus (Avast) gets rid of the bug. He doesn't have any Firefox plugin or addon from the antivirus, or any other suspicious addon or plugin.
5. Nothing odd shows up in the network tab in devtools. Using the profiler, the stall time is reported as spent in XRE_InitChildProcess.

This looks like a regression on firefox, since older firefox is still ok today (so it's not a regression in the antivirus or something else). I got my friend to run mozregression, and it bisected to

https://hg.mozilla.org/integration/mozilla-inbound/rev/48ad3b64dea6

which is bug 1349699. To be sure, he bisected twice to confirm and got the exact same result.

Comment 1

[Benjamin Smedberg]

[Tracking Requested - why for this release]: FF55 regression with one of the most popular antivirus.

Comment 2

[Bill McCloskey [inactive unless it's an emergency] (:billm)]

There's not much I can do here. This is a problem with Avast.

Comment 3

[Jim Mathies [:jimm]]

Alon, can you provide the details on what version of avast your friend is running?

Comment 4

[Jim Mathies [:jimm]]

Also, your friend's about:support text would be helpful.

Comment 5

[Alon Zakai (:azakai)]

Attachment: about.support.txt

Attached is his about:support.

Avast version is Avast 17.4.2294 (should be the latest).

Comment 6

[Jim Mathies [:jimm]]

How does this Kitely service interact with the browser? I keep getting a message: "Start your virtual world viewer, and login using the following information". According to this support page I need a 3rd party app - 

https://kitely.atlassian.net/wiki/display/doc/How+to+Enter+Kitely+Virtual+Worlds

Can you provide steps to reproduce?

Comment 7

[Alon Zakai (:azakai)]

The steps are just to load the page at that url, https://www.kitely.com/ . When the bug happens, a very long delay occurs before the page is loaded (during which you'll see either nothing in the page or just a spinner).

(To actually use the virtual world service you do need a virtual world viewer app, but all that's unrelated to the bug here.)

Comment 8

[Julien Cristau [:jcristau]]

Tracking as a regression when using popular third party software.

Comment 9

[Jim Mathies [:jimm]]

Alon, do you run the same rev of Avast? I tried testing in a vm with their latest installed, no issues.

Comment 10

[Alon Zakai (:azakai)]

I actually don't have a Windows machine or VM to test on, so I couldn't test Avast myself. I only saw the bug on my friend's machine, which was using that version.

Comment 11

[Mike Taylor [:miketaylr]]

Andrei, can you find someone who can reproduce so we can move this forward? THanks.

Comment 12

[Emil Ghitta, QA [:emilghitta]]

Hi everyone!

I managed to reproduce the issue described in comment 0 with Firefox 55.0b7 (BuildId:20170706085221) using Avast Program Version: 17.5.2302 (build 17.5.3559.0) and having the Web Shield component on.

Steps to reproduce:
1.Activate Avast's Web Shield component.
2.Launch Firefox 
3.Go to https://www.kitely.com/ webpage

I saw that this issue occurs only when having Web Shield component activated and this may not be reproducible at all times.

Please observe the attached video for further details (Note that I am pressing the F5 key to constantly refresh the webpage).

Alon is this somehow the same as what you have encountered?

Comment 13

[Emil Ghitta, QA [:emilghitta]]

Attachment: Avast Web Shield

Comment 14

[Alon Zakai (:azakai)]

Yes, that looks just like the issue I saw on my friend's machine.

Comment 15

[Mike Taylor [:miketaylr]]

Jim, can you reproduce w/ the steps from Comment #12.

Comment 16

[Jim Mathies [:jimm]]

Dee's, do we have any contacts at avast?

Comment 17

[Desigan Chinniah [:cyberdees] [:dees] [London - GMT]]

(In reply to Jim Mathies [:jimm] from comment #16)
> Dee's, do we have any contacts at avast?

Unfortunately all of my Avast contacts have since moved on. Adam could likely assist with the outreach to a relevant contact — and forging a more solid relationship for this issue as well as the future.

Comment 18

[Adam Stevenson [:adamopenweb]]

Reaching out via LinkedIn.

Comment 19

[Lukas Rypacek]

Hi, Lukas from Avast here. We've been able to reproduce the problem with https://www.kitely.com/ and tracked it down to a potential problem with the handling of padding inside HTTP/2 data frames. When the last data frame on a connection had a non-zero padding we delayed the padded bytes. We are testing the fix now and should release a fixed version of the HTTP/2 parser (part of Avast AV WebShield) in the next release 17.6.

Comment 20

[Alon Zakai (:azakai)]

Great progress! :)

Out of curiosity, do we know why this only affected Firefox, and not Chrome or other browsers?

Comment 21

[Jim Mathies [:jimm]]

(In reply to Lukas Rypacek from comment #19)
> Hi, Lukas from Avast here. We've been able to reproduce the problem with
> https://www.kitely.com/ and tracked it down to a potential problem with the
> handling of padding inside HTTP/2 data frames. When the last data frame on a
> connection had a non-zero padding we delayed the padded bytes. We are
> testing the fix now and should release a fixed version of the HTTP/2 parser
> (part of Avast AV WebShield) in the next release 17.6.

Hey Lukas, can you confirm this fix shipped?

Comment 22

[Lukas Rypacek]

Hello Jim,
I can confirm the fix has been released.
Lukas.