All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=775378

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=cba098db29906db691770e03b3b1ad70123575bdb4779c1a90380f7b940831ff


Bot(s) for this bug's original alert(s):

linux-release

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/148fc64b780000

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/148fc64b780000

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8965476822773063872

=== Auto-CCing suspected CL author kerrnel@chromium.org ===

Hi kerrnel@chromium.org, the bisect results pointed to your CL, please take a look at the
results.


=== BISECT JOB RESULTS ===
Perf regression found with culprit

Suspected Commit
  Author : Greg Kerr
  Commit : fcd7b94faa40dfc3e4e4be26f4e854dd14ef650c
  Date   : Thu Oct 12 22:48:50 2017
  Subject: Enable Finch Trial Test for MacV2Sandbox.

Bisect Details
  Configuration: mac_10_11_perf_bisect
  Benchmark    : loading.desktop
  Metric       : cpuTimeToFirstMeaningfulPaint_avg/cold/Economist
  Change       : 18.57% | 284.352928571 -> 337.152333333

Revision             Result                  N
chromium@508397      284.353 +- 35.8217      14      good
chromium@508468      283.132 +- 22.4576      14      good
chromium@508504      283.524 +- 24.1958      14      good
chromium@508509      283.755 +- 34.916       14      good
chromium@508511      282.663 +- 18.1448      14      good
chromium@508512      282.766 +- 21.0022      14      good
chromium@508513      335.213 +- 29.5751      3       bad       <--
chromium@508521      332.871 +- 11.4897      5       bad
chromium@508540      334.072 +- 9.56954      3       bad
chromium@508681      337.152 +- 13.7394      3       bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=Economist loading.desktop

More information on addressing performance regressions:
  http://g.co/ChromePerformanceRegressions

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8965476822773063872


For feedback, file a bug with component Speed>Bisection

So this regressed 50ms on macOS 10.11? Let's see what 10.12 and 10.13 look like.

 Issue 772881  has been merged into this issue.

 Issue 776491  has been merged into this issue.

 Issue 776744  has been merged into this issue.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/c770680d7df3c5949627a51cb0a25381400f2ef0

commit c770680d7df3c5949627a51cb0a25381400f2ef0
Author: Greg Kerr <kerrnel@chromium.org>
Date: Fri Oct 20 23:47:26 2017

macOS V2 Sandbox: Improve performance by avoiding unncessary denials.

This improves performance by avoiding calls to diagnosticd,
and allow access to /etc/passwd, which doesn't contain any password
hashes. It also allows access to /S/L/LinguisticData.

Bug:  775378 
Change-Id: I79e934086507ea959d2aab45fb9290c3c8bef721
Reviewed-on: https://chromium-review.googlesource.com/729152
Reviewed-by: Nasko Oskov <nasko@chromium.org>
Reviewed-by: Robert Sesek <rsesek@chromium.org>
Commit-Queue: Greg Kerr <kerrnel@chromium.org>
Cr-Commit-Position: refs/heads/master@{#510615}
[modify] https://crrev.com/c770680d7df3c5949627a51cb0a25381400f2ef0/content/browser/child_process_launcher_helper_mac.cc
[modify] https://crrev.com/c770680d7df3c5949627a51cb0a25381400f2ef0/services/service_manager/sandbox/mac/renderer_v2.sb

 Issue 776740  has been merged into this issue.

It appears that the experimental CL has caused 10.11 to return to baseline, but 10.12 has not budged. I will investigate further why no change is visible in 10.12 and then we'll make final decisions about what system services to allow.

Looking at  ChromiumPerf/chromium-rel-mac12/loading.desktop / timeToFirstMeaningfulPaint_avg / cold / Economist it looks like it that benchmark got about 100ms slower.

Is that the only story?

Tim/Ned - wdyt?

For the next step, I think we should look at UMA metrics to understand the impact of MacV2Sandbox at real world scale.

Sounds good. I deployed a finch trial and then on my TODO list I need to understand, through local testing, why 10.12 didn't respond to the experimental CL. Thanks so much for the help in all this.

This is not surprising, but you can see the 10.12 slow downs in the bots on the low end hardware, but not as much on the normal hardware:
https://chromeperf.appspot.com/report?sid=b7740daab5ddb9ec6b9848b6aef9837f45075ef8bba1660eb8bda59c85061de8

I think that just supports the current plan which is to let the finch trial run and then let's analyze the actual histograms from the field. 

The canary and dev runs on macOS are tough because of how small the user base is, but I see nothing in the results that should prevent this from going to beta for a better sample set.

SG.

Looks like there are results from the field now; any update on this bug?

There is < 50ms slowdown in the cold startup time for some users with the V2 sandbox, and after talking to jschuh@, we've decided to accept the security/performance tradeoff and ship the feature.

Marking WontFix per comment #22.