The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/ee2314de40741b97283ad4325b284c063cc1fa48

commit ee2314de40741b97283ad4325b284c063cc1fa48
Author: Xiaohan Wang <xhwang@chromium.org>
Date: Fri Nov 03 23:23:37 2017

media: Add trace events for MojoCdmFileIO and CdmAdapter

Currently MojoCdmFileIO and CdmAdapter run on the same thread. So it's
possible that file read/write could block audio/video decoding. Adding
tracing events so that we can understand the performance impact. We'll
use this data to decide whether we need to move file read/write
operations to a different thread.

BUG= 774160 

Change-Id: I5f7a216dfb01cb797d628fecdd095cffb682e4e5
Reviewed-on: https://chromium-review.googlesource.com/748882
Reviewed-by: John Rummell <jrummell@chromium.org>
Commit-Queue: Xiaohan Wang <xhwang@chromium.org>
Cr-Commit-Position: refs/heads/master@{#513951}
[modify] https://crrev.com/ee2314de40741b97283ad4325b284c063cc1fa48/media/cdm/cdm_adapter.cc
[modify] https://crrev.com/ee2314de40741b97283ad4325b284c063cc1fa48/media/mojo/services/mojo_cdm_file_io.cc
[modify] https://crrev.com/ee2314de40741b97283ad4325b284c063cc1fa48/media/mojo/services/mojo_cdm_file_io.h

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/adda6e11072e39d0034c3bc795169a6132f0d972

commit adda6e11072e39d0034c3bc795169a6132f0d972
Author: John Rummell <jrummell@chromium.org>
Date: Fri Nov 10 23:02:59 2017

media: Record time taken for CDM to read/write files

Currently MojoCdmFileIO and CdmAdapter run on the same thread. So it's
possible that file read/write could block audio/video decoding. Adding a
UMA to log the actual time taken, so we can understand if the IO takes too
long on actual user machines and should be moved to a seperate thread.

BUG= 774160 

Change-Id: I25a98e2124b431707c7e9efcdbf4da4cbdfccd2f
Reviewed-on: https://chromium-review.googlesource.com/758458
Reviewed-by: Xiaohan Wang <xhwang@chromium.org>
Reviewed-by: Ilya Sherman <isherman@chromium.org>
Commit-Queue: John Rummell <jrummell@chromium.org>
Cr-Commit-Position: refs/heads/master@{#515726}
[modify] https://crrev.com/adda6e11072e39d0034c3bc795169a6132f0d972/media/mojo/services/mojo_cdm_file_io.cc
[modify] https://crrev.com/adda6e11072e39d0034c3bc795169a6132f0d972/tools/metrics/histograms/histograms.xml

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/f45580a8857e186a038745049e595bd31280f1cf

commit f45580a8857e186a038745049e595bd31280f1cf
Author: John Rummell <jrummell@chromium.org>
Date: Fri Nov 17 05:11:49 2017

media: Add additional trace events for MojoCdmFileIO

Adds trace events for opening the file, as well as when Write() and
WriteDone() are called to understand the impact of opening a temporary
file and renaming it after it's written to. Also adds fileName to the
event data saved.

BUG= 774160 

Change-Id: Id0476d028865802e1ddda22eac6620cd6f16f106
Reviewed-on: https://chromium-review.googlesource.com/775654
Commit-Queue: John Rummell <jrummell@chromium.org>
Reviewed-by: Xiaohan Wang <xhwang@chromium.org>
Cr-Commit-Position: refs/heads/master@{#517313}
[modify] https://crrev.com/f45580a8857e186a038745049e595bd31280f1cf/media/mojo/services/mojo_cdm_file_io.cc

Re #2: What does the UMA number say? Let's watch the number periodically and by M-65 timeframe we should be able to make a decision on whether we need to do anything on this.

For Media.EME.CdmFileIO.ReadTime (Windows), 99th percentile is about 1.5ms. In the last 30 days (all OS), less than 0.01% of reports > 135ms. Largest value is <1062ms.

For Media.EME.CdmFileIO.WriteTime (Windows), 99th percentile is around 2ms. In the last 30 days (all OS), less than 0.01% of reports > 190ms. Largest value is <3553ms.

However, this is canary only. Will have to check again once it reaches more users.

Updating with latest results.

For Media.EME.CdmFileIO.ReadTime (Windows), 95th percentile is ~ 1.0ms, 99th percentile is < 1.6ms.
For Media.EME.CdmFileIO.WriteTime (Windows), 95th percentile is ~ 1.4ms, 99th percentile is < 3.0ms.

Numbers look good, so resolving this as done.