Issue metadata
Sign in to add a comment
|
ERR_QUIC_PROTOCOL_ERROR accessing google sites as non-admin user
Reported by
hkpa...@gmail.com,
Aug 25 2016
|
||||||||||||||||||||||||
Issue descriptionUserAgent: Mozilla/5.0 (Windows NT 6.1; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/54.0.2839.2 Safari/537.36 Example URL: https://www.google.com/search?q=windows+update+command+line&oq=windows+update+command+line&aqs=chrome..69i57j69i64l2.4767j0j7&sourceid=chrome&ie=UTF-8 Steps to reproduce the problem: 1. Open Chrome Canary as regular user 2. Try to navigate to any google sites - search, gmail, youtube etc. What is the expected behavior? - Site should load What went wrong? This site can’t be reached The webpage at https://www.google.com/search?q=windows+update+command+line&oq=windows+update+command+line&aqs=chrome..69i57j69i64l2.4767j0j7&sourceid=chrome&ie=UTF-8 might be temporarily down or it may have moved permanently to a new web address. ERR_QUIC_PROTOCOL_ERROR Does it occur on multiple sites: Yes Is it a problem with a plugin? No Did this work before? Yes The last version also had similar issue. Last to last probably worked. Does this work in other browsers? Yes Chrome version: 54.0.2839.2 Channel: canary OS Version: 6.1 (Windows 7, Windows Server 2008 R2) Flash Version: Shockwave Flash 22.0 r0 Browser works fine when run as admin/elevated user. I've tried reinstalling a fresh copy and also tried deleting my data and reinstalling. Tried launching incognito mode as well but same issue persists. So far, only elevated mode seems to work.
,
Aug 25 2016
Its a corporate machine and has antivirus and stuff. Hard to tell if they did something to the effect that you mention. But this issue only happens for google sites. I don't think they are doing anything special for google sites. Non google sites with https load just fine.
,
Aug 25 2016
Can you collect a net-internals trace? https://dev.chromium.org/for-testers/providing-network-details The QUIC protocol is only supported by Google (and a few other servers) so if there are any network devices which interfere with QUIC (UDP traffic) that would end up mostly affecting Google.
,
Aug 25 2016
I had to restart my box a bunch of times due to Windows Update installing all those updates. After that, I haven't been able to reproduce this issue. I'll keep an eye out. Maybe it happens after a while? Will provide the net-internals trace when it happens. If not, maybe we can close this issue in a day or two.
,
Aug 25 2016
It did happen after a while. So here's the trace.
,
Aug 26 2016
Found something this morning - if I have Chrome stable open and I try google sites, I run into this issue - if I close Chrome stable, wait for all chrome processes to exit (via task manager) and then open a new Canary instance and try, google sites load fine. - I've pinned gmail tabs (work gmail) on Chrome stable. So if I open Chrome stable and shut it down quickly, even before the gmail tab gets a chance to load fully - and then open Canary and try google sites, it fails. So tl;dr - Ensure gmail tab is loaded fully (in Chrome stable). Then clean exit Chrome stable. And then open Canary and google sites work.
,
Aug 26 2016
I've looked at your net-internals trace, and it looks like something in your network is blocking QUIC traffic after the handshake completes. In each of the QUIC_SESSION events, within 100ms of the connection being established no more packets are received by the server. Do you know what firewall or NAT box you are going through? Who is your ISP? Also, can you send a net-internals from Canary? I *suspect* it's merely a coincidence that it's working, but I'd love to confirm.
,
Aug 26 2016
The trace was from Canary. Also, if something was blocking QUIC, then it'd block it always, wouldn't it? Like I said in my earlier comment, if Chrome stable is not running in parallel, this issue doesn't happen.
,
Aug 26 2016
> Also, if something was blocking QUIC, then it'd block it always, wouldn't it? You'd be surprised at the variety of bizarre behavior we see from network devices. In particular, the mode of dropping UDP after less than a second has been observed elsewhere. Fascinating that the blocking does not happen to Canary if Chrome stable is not running at the same time.
,
Aug 29 2016
Upgraded to Canary 55 today and could not connect to google sites with the above workaround (shutting down Chrome stable workaround). Finally, running as Admin worked. I'm going to check Virus scanner logs to see if there is any mention of QUIC (I doubt it though). But is it possible that running as admin bypasses the blocks? Also, isn't Chrome stable using QUIC? Any blockages would apply to the Stable version as well, no?
,
Aug 29 2016
OK I give up - there is no method to this madness :). Sometimes one way works, sometimes the other. Luckily Chrome stable works so far (maybe I shouldn't jinx it ;) ).
,
Aug 29 2016
The idea it may be a virus scanner sounds reasonable. Is it possible for you to disable any virus scanning software from blocking network traffic for a period of time? Also, would you mind sharing what virus scanners you're currently running, in case we spot a trend?
,
Aug 29 2016
> Also, isn't Chrome stable using QUIC? Any blockages would apply to the Stable version as well, no? Yes, chrome stable typically uses QUIC too and so I would expect it to be subject to the same blocking. I'd love to see a net-internals from stable where things work just after a net-internals from canary when they didn't. Perhaps the virus scanner has some mechanism to not block "Google Chrome" which does not apply to "Google Chrome Canary".
,
Aug 29 2016
Here's the net internals trace from stable. As for virus scanner, I'm using McAffee and its details are: ----------------------------------------------------------- System Information Computer Name: <redacted> McAfee Agent Version number: 4.8.0.1500 Managed Last security update check: 8/29/2016 1:00:10 PM Last agent-to-server communication: 8/29/2016 1:10:53 PM Agent to Server Communication Interval (every): 1 hours 40 minutes 0 seconds Policy Enforcement Interval (every): 5 minutes Agent ID: <redacted> ePO Server/Agent Handler DNS Name: <redacted> IP Address: <redacted> Port Number: 443 McAfee VirusScan Enterprise + AntiSpyware Enterprise Version number: 8.8.0 (8.8.0.1247) Build date: 1/16/2014 Anti-virus License Type: licensed Scan engine version (32-bit): 5800.7501 Scan engine version (64-bit): 5800.7501 DAT version: 8271.0000 DAT Created on: 8/28/2016 Number of Signatures in extra.dat: 0 Name of threats that extra.dat can detect: None Buffer Overflow and Access Protection DAT version: 659 Installed Patches: 4 Installed Modules: Copyright © 1995-2013 McAfee, Inc. All Rights Reserved. www.mcafee.com --------------------------------------
,
Aug 29 2016
Wow. That is perplexing. The stable trace shows perfectly reasonable connectivity whereas the canary trace shows basically 100% loss after less than 1 second. I suspect Ian might be on the right track with the virus scanner.
,
Aug 29 2016
A virus scanner definitely seems like a good thesis. Can you check if there is a policy whitelist for the scanner that includes "Google Chrome"?
,
Aug 29 2016
https://community.mcafee.com/thread/50328?tstart=0 is a bit dated but suggests that you may be able to add Apps to the firewall whitelist policy. Can you add Chrome Canary to this list, and see if that fixes it?
,
Aug 29 2016
I checked the logs and I don't see any mentions of QUIC or Chrome (of any sorts) in there. I also don't see a firewall neither do I see any rules that block any browsers/processes (let alone canary). I don't see us using Security Center (as shown in the community thread) so that seems moot. Anything else I should look for?
,
Aug 29 2016
Just tried running with Virus scan disabled and no dice. Right now, the run as admin hack is working.
,
Sep 6 2016
Thank you for providing more feedback. Adding requester "rch@chromium.org" for another review and adding "Needs-Review" label for tracking. For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
,
Oct 19 2016
|
|||||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||||
Comment 1 by juliatut...@chromium.org
, Aug 25 2016