cashew: add local byte counters

cashew: add local byte counters

The data provided by the carrier usage API isn't always updated in real
time after a cellular device has connected to the network, and carriers
would prefer if we didn't make frequent API requests.

So, instead of making periodic API requests and relying on the returned
data to be fresh, we hit the API once on connect to set a baseline and
then track usage relative to this baseline by looking at local network
interface counters maintained by the system.

This change should be transparent to our clients, who trust us to provide
the best info possible regardless of its source.

BUG=chromium-os:7770
TEST=unit tests, manual testing on device

Change-Id: Ib88dee5ee52d0e781f853356a3d68da714627bc8

Committed: http://chrome-svn/viewvc/chromeos?view=rev&revision=9de77a5

[Vince Laviano, 2010-11-18 22:14:35]

First cut at local counters. 

Remaining work:
- Do something more sensible with the update timer. This CL brings it down from
10 mins to 1 min since it's no longer being used to make periodic requests
during normal operation but rather as a retry timer in the event of a failure.
However, we should only retry a limited number of times in quick succession
before backing off to a longer interval.
- Automated unit tests where possible: data_plan_unittest.cc and
procfs_byte_counter_unittest.cc (an empty shell right now)
- Needs more testing on a device in a realistic environment

[Jason Glasgow, 2010-11-19 02:22:32]

Looking good.  See comment below about flimflam CL which changes DBUS
properties.

I still need to review the procfs files.

http://codereview.chromium.org/5180003/diff/1/src/device.cc
File src/device.cc (right):

http://codereview.chromium.org/5180003/diff/1/src/device.cc#newcode15
src/device.cc:15: static const char *kFlimflamDeviceName =
"org.chromium.flimflam";
Why this change?  I think google style is const char*.
flimflam style is const char *

http://codereview.chromium.org/5180003/diff/1/src/service.h
File src/service.h (right):

http://codereview.chromium.org/5180003/diff/1/src/service.h#newcode207
src/service.h:207: // we've received updated Cellular.RestrictedPool info from
Flimflam
This is out of date.  See http://codereview.chromium.org/5167004/

[Vince Laviano, 2010-11-19 02:44:58]

http://codereview.chromium.org/5180003/diff/1/src/device.cc
File src/device.cc (right):

http://codereview.chromium.org/5180003/diff/1/src/device.cc#newcode15
src/device.cc:15: static const char *kFlimflamDeviceName =
"org.chromium.flimflam";
On 2010/11/19 02:22:32, Jason Glasgow wrote:
> Why this change?  I think google style is const char*.
> flimflam style is const char *

According to the style guide:

"When declaring a pointer variable or argument, you may place the asterisk
adjacent to either the type or to the variable name: [snip example]. You should
do this consistently within a single file, so, when modifying an existing file,
use the style in that file."

I made this change, because I'm trying to converge cashew towards consistent use
of the "Type *variable_name" style. (I've made similar changes in previous CLs.)
The reason that I prefer this style over "Type* variable_name" is that it's less
misleading when multiple variables are declared on a single line.  Compare:

char* a, b;
char *a, b;

In both cases, a is a pointer and b is not, but I think that the second style
makes this clearer.

http://codereview.chromium.org/5180003/diff/1/src/service.h
File src/service.h (right):

http://codereview.chromium.org/5180003/diff/1/src/service.h#newcode207
src/service.h:207: // we've received updated Cellular.RestrictedPool info from
Flimflam
On 2010/11/19 02:22:32, Jason Glasgow wrote:
> This is out of date.  See http://codereview.chromium.org/5167004/

Thanks for the heads-up. I'll revise my CL to look for the new enum.

[Jason Glasgow, 2010-11-19 21:51:07]

LGTM - with just one nit.

http://codereview.chromium.org/5180003/diff/1/src/procfs_byte_counter.h
File src/procfs_byte_counter.h (right):

http://codereview.chromium.org/5180003/diff/1/src/procfs_byte_counter.h#newco...
src/procfs_byte_counter.h:22: // an procfs-based implementation of the
ByteCounter interface
a procfs

[Vince Laviano, 2010-11-20 00:32:00]

Includes modification to monitor flimflam ConnectivityState property instead of
Cellular.RestrictedPool property.

http://codereview.chromium.org/5180003/diff/1/src/procfs_byte_counter.h
File src/procfs_byte_counter.h (right):

http://codereview.chromium.org/5180003/diff/1/src/procfs_byte_counter.h#newco...
src/procfs_byte_counter.h:22: // an procfs-based implementation of the
ByteCounter interface
On 2010/11/19 21:51:08, Jason Glasgow wrote:
> a procfs

Done.