New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 803366 link

Starred by 1 user

Issue metadata

Status: Fixed
Owner:
Closed: Jan 2018
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Linux , Windows , Chrome , Mac
Pri: 2
Type: Bug



Sign in to add a comment

Site Isolation: Unexpected *.google.com process sharing in M63 due to remote NTP

Project Member Reported by creis@chromium.org, Jan 18 2018

Issue description

Chrome Version: 63.0.3239.132
OS: Win 10

What steps will reproduce the problem?
(1) Start Chrome with --disable-features=UseGoogleLocalNtp --isolate-origins=https://google.com
(2) Sign into google.com
(3) Open an NTP and navigate to https://maps.google.com
(4) Open another NTP and notice (in Chrome's task manager) that there's a https://google.com subframe entry.
(5) Navigate the second NTP to https://maps.google.com

What is the expected result?

The second Google Maps tab should be in a different process than the first, since the tabs are in different BrowsingInstances.

What happens instead?

Both Google Maps tabs are in the same process.

This seems to happen because the remote NTP has an effective URL, similar to hosted apps.  When the user is signed in, there are other google.com iframes which do not get the same effective URL, so they're kicked out of the process in M63.  (We no longer kick these subframes out of the process in r521188, which was merged to M64.)

The google.com OOPIF aggressively joins an existing same-site renderer process if one exists (to avoid creating processes for OOPIFs when possible).  However, this means that the BrowsingInstance for the new tab now has a SiteInstance pointing to an existing process, so future visits to https://*.google.com end up in the existing process as well, which is undesirable.

This does not happen with the local NTP, or when the user is not signed in, or in Chrome 64.  However, there's a more general version of this problem, which I'll file next.
 

Comment 1 by creis@chromium.org, Jan 18 2018

Status: Fixed (was: Assigned)
I've filed  issue 803367  for the more general case, which may or may not happen as often.

I think we can probably consider this particular NTP issue fixed in Chrome 64 by r521188, but I wanted to file it to track the impact and for comparison to the general case.

Sign in to add a comment