1800392 - Firefox resets network connection at the end of the HTTP exchange

Status: RESOLVED WORKSFORME

(Core :: Networking, defect)

Description

[support]

Attachment: Firefox-exchange.pcapng

User Agent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:106.0) Gecko/20100101 Firefox/106.0

Steps to reproduce:

Using a Python socket-based web server. when the page is loaded, Firefox receives the full HTML page displays it for less than a second and then resets the connection and provides a connection reset message. This same web site works fine with Chrome and Safari web browsers.

Actual results:

Firefox provides a "Connection Reset" message. Additionally, it changes the HTTP request following the initial request and response. It receives the data starts to display it and then sends a reset and breaks the connection. See attached packet captures for the full exchange with Chrome and the same exchange with Firefox.

In this case I control both sides of the connection. Additionally, I have tried all of the other suggestions in the archives including disabling all extensions and clearing the cache. Clearly the network connection is fine as Firefox displays the web page for a brief instant prior to resetting the connection. Insights as to how to debug this on the Firefox end to assess what is causing it to issue the reset would be appreciated. At present this site can only be used with browsers other than Firefox.

Note: Since this submission only allows a single file, I have included the Firefox packet exchange. I have both the second Firefox request (when I select "Try Again" and a sample exchange with Chrome. That are not included here given the single file upload limitation.

Expected results:

The web page should have displayed and stayed as displayed initially.

Comment 1

[support]

Attachment: Firefox-exchange2.pcapng

Here is the Firefox exchange packet capture when the "Try Again" button is selected.

Comment 2

[support]

Attachment: Chrome-exchange.pcapng

This is the exchange that occurs when using the Chrome web browser, which works without issue.

Comment 3

[support]

Note: All packet captures have now been included.

Comment 4

[BugBot [:suhaib / :marco/ :calixte]]

The Bugbug bot thinks this bug should belong to the 'DevTools::Netmonitor' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.

Comment 5

[Kershaw Chang [:kershaw]]

Hi Reporter,

Could you try to record a http log for this?
Note that if about:logging page does not work, please try about:networking#logging. Also, the log may contain some privacy information, so you may want to send it to [e-mail].

Thanks.

Comment 6

[support]

I ran the log and mailed it to the address provided.

Comment 7

[support]

Log sent December 12.

Comment 8

[Kershaw Chang [:kershaw]]

Hi Reporter,

I've looked into the log provide at comment #0.
It looks like the TCP connection was reset by the server after Firefox sent a KeepAlive packet.
Not sure if this is related, but could you try to disable network.tcp.keepalive.enabled and see if it helps?

Thanks.

Comment 9

[Kershaw Chang [:kershaw]]

FWIW, the Chrome-exchange.pcapng also shows the TCP connection was reset by the server, so this looks like an issue st server side?

Comment 10

[support]

Attachment: about:config information for the bug

Comment 11

[edgul]

Closing since we're pretty sure it's server-side and we need more info to dig into this.
Please re-open if the info becomes available and the issue persists.