Telemetry: Implement MSR power monitoring on Linux.

Telemetry: Implement MSR power monitoring on Linux.

BUG=336556
Committed: https://crrev.com/b6e00ac9e417148d6d1e2022d858fb5f3956da67
Cr-Commit-Position: refs/heads/master@{#296325}

[Lei Zhang, 2014-09-17 05:44:24]

thestig@chromium.org changed reviewers:
+ dtu@chromium.org

[Lei Zhang, 2014-09-17 05:44:25]

My workstation doesn't support power MSRs, so I can't actually test this. Can
you take a look? I'll find a newer machine in the meantime.

[tonyg, 2014-09-17 23:11:44]

On 2014/09/17 05:44:25, Lei Zhang wrote:
> My workstation doesn't support power MSRs, so I can't actually test this. Can
> you take a look? I'll find a newer machine in the meantime.

I did this:
$ sudo modprobe msr
$ sudo chmod 644 /dev/cpu/*/msr
$ tools/perf/run_benchmark tab_switching.top_10 --browser=release

And got this:
  ...
  _LinuxReadMsr at
tools/telemetry/telemetry/core/platform/power_monitor/msr_power_monitor.py:220
    with open(LINUX_MSR_PATH, 'rb', buffering=0) as msr:
IOError: [Errno 1] Operation not permitted: '/dev/cpu/0/msr'

Locals:
  msr_number : 1553

Here're the permissions on the file:
$ ls -l /dev/cpu/0/msr
crw-r--r-- 1 root root 202, 0 Sep 17 15:59 /dev/cpu/0/msr

[Lei Zhang, 2014-09-17 23:33:07]

On 2014/09/17 23:11:44, tonyg wrote:
> And got this:
>   ...
>   _LinuxReadMsr at
> tools/telemetry/telemetry/core/platform/power_monitor/msr_power_monitor.py:220
>     with open(LINUX_MSR_PATH, 'rb', buffering=0) as msr:
> IOError: [Errno 1] Operation not permitted: '/dev/cpu/0/msr'

Doh, I took that code straight from dtu's sample. I'll find a computer to test +
debug this soon.

[dtu, 2014-09-18 00:01:43]

On 2014/09/17 23:33:07, Lei Zhang wrote:
> On 2014/09/17 23:11:44, tonyg wrote:
> > And got this:
> >   ...
> >   _LinuxReadMsr at
> >
tools/telemetry/telemetry/core/platform/power_monitor/msr_power_monitor.py:220
> >     with open(LINUX_MSR_PATH, 'rb', buffering=0) as msr:
> > IOError: [Errno 1] Operation not permitted: '/dev/cpu/0/msr'
> 
> Doh, I took that code straight from dtu's sample. I'll find a computer to test
+
> debug this soon.

chmod is not sufficient to give you privileges to access this. You need to run
with sudo.

I'm going to move ReadMsr from msr_power_monitor into *_platform_backend. I
think this is a cleaner way to implement multiple platforms - msr_power_monitor
will become OS-independent.
https://codereview.chromium.org/578123002/

[Lei Zhang (Do not use), 2014-09-18 05:36:11]

On 2014/09/18 00:01:43, dtu wrote:
> On 2014/09/17 23:33:07, Lei Zhang wrote:
> > On 2014/09/17 23:11:44, tonyg wrote:
> > > And got this:
> > >   ...
> > >   _LinuxReadMsr at
> > >
> tools/telemetry/telemetry/core/platform/power_monitor/msr_power_monitor.py:220
> > >     with open(LINUX_MSR_PATH, 'rb', buffering=0) as msr:
> > > IOError: [Errno 1] Operation not permitted: '/dev/cpu/0/msr'
> > 
> > Doh, I took that code straight from dtu's sample. I'll find a computer to
test
> +
> > debug this soon.
> 
> chmod is not sufficient to give you privileges to access this. You need to run
> with sudo.

One can read the MSR as a regular user, but the executable that runs must have
the right capabilities. i.e. setcap cap_sys_rawio+ep /path/to/executable. It
can't be an interpreted script because such a script will be executed by the
python binary. The reason this was added was due to a security exploit, which
the kernel fixed. [1] However, the exploit requires writing to the MSR. If we
grant a program that only reads the MSR CAP_SYS_RAWIO, we should be ok since the
msr files are still read-only due to the file system permissions.

[1]
http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=c90...

[Lei Zhang, 2014-09-19 05:14:17]

See patch set 2. I sent out a separate CL to change our bot configuration to
allow MSR reading.

[dtu, 2014-09-19 07:04:19]

lgtm with the bot change

[Lei Zhang, 2014-09-23 23:36:36]

On 2014/09/19 07:04:19, dtu wrote:
> lgtm with the bot change

Bots update CL went in.

[Lei Zhang, 2014-09-23 23:36:39]

The CQ bit was checked by thestig@chromium.org

[commit-bot: I haz the power, 2014-09-23 23:38:37]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/577893003/20001

[commit-bot: I haz the power, 2014-09-24 00:04:43]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2014-09-24 00:04:43]

Try jobs failed on following builders:
  ios_dbg_simulator on tryserver.chromium.mac
(http://build.chromium.org/p/tryserver.chromium.mac/builders/ios_dbg_simulator...)
  ios_rel_device on tryserver.chromium.mac
(http://build.chromium.org/p/tryserver.chromium.mac/builders/ios_rel_device/bu...)
  ios_rel_device_ninja on tryserver.chromium.mac
(http://build.chromium.org/p/tryserver.chromium.mac/builders/ios_rel_device_ni...)

[Lei Zhang, 2014-09-24 00:13:54]

The CQ bit was checked by thestig@chromium.org

[commit-bot: I haz the power, 2014-09-24 00:14:56]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/577893003/20001

[commit-bot: I haz the power, 2014-09-24 00:19:36]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2014-09-24 00:19:38]

Try jobs failed on following builders:
  ios_dbg_simulator on tryserver.chromium.mac
(http://build.chromium.org/p/tryserver.chromium.mac/builders/ios_dbg_simulator...)
  ios_rel_device_ninja on tryserver.chromium.mac
(http://build.chromium.org/p/tryserver.chromium.mac/builders/ios_rel_device_ni...)

[Lei Zhang, 2014-09-24 01:31:31]

The CQ bit was checked by thestig@chromium.org

[commit-bot: I haz the power, 2014-09-24 01:32:04]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/577893003/20001

[commit-bot: I haz the power, 2014-09-24 01:45:18]

Committed patchset #2 (id:20001) as ba94b800dbde4db6a8c6d456645b8c6816c21485

[commit-bot: I haz the power, 2014-09-24 01:45:53]

Patchset 2 (id:??) landed as
https://crrev.com/b6e00ac9e417148d6d1e2022d858fb5f3956da67
Cr-Commit-Position: refs/heads/master@{#296325}