hardware_DiskSize,SsdDetection: fix hardware root dev discovery

hardware_DiskSize,SsdDetection: fix hardware root dev discovery

When root filesystem verification is enabled, the root device discoverable
by walking the kernel command line does not match the physical drive the root
partition is on.  Just recently, we added support for rootdev to resolve
through the device-mapper device (/dev/dm-0) to find the true root device.

These two tests now use rootdev with this functionality.

TEST=ran these on a vboot'd dogfood device (well waiting for a full fresh build, then I will :)
BUG=chromium-os:890, chromium-os:892
Change-Id: Ieda92c297d4b8ec2b93f5bf8a3f0a61e1cedd33a

[bfreed, 2010-09-01 18:13:09]

http://codereview.chromium.org/3349001/diff/1/2
File client/site_tests/hardware_DiskSize/hardware_DiskSize.py (right):

http://codereview.chromium.org/3349001/diff/1/2#newcode16
client/site_tests/hardware_DiskSize/hardware_DiskSize.py:16: match =
re.search(r'/dev/([^ ]+)', devnode)
Hmm...

All our other rootdev-using tests pretty much assume something useful will be
returned here.  This this re search really necessary?  Is -i useful?

I would think that if rootdev reports failure, we might want the test to fail
immediately.

[Will Drewry, 2010-09-01 18:20:29]

A few responses - do they make sense?

http://codereview.chromium.org/3349001/diff/1/2
File client/site_tests/hardware_DiskSize/hardware_DiskSize.py (right):

http://codereview.chromium.org/3349001/diff/1/2#newcode16
client/site_tests/hardware_DiskSize/hardware_DiskSize.py:16: match =
re.search(r'/dev/([^ ]+)', devnode)
On 2010/09/01 18:13:09, bfreed wrote:
> Hmm...
> 
> All our other rootdev-using tests pretty much assume something useful will be
> returned here.  This this re search really necessary?

The re search extracts just the block device sda and not /dev/sda.

> Is -i useful?

It will ignore error if the device node doesn't exist.  However, the entry will
exist in /proc/partitions even if there's no dev node.  rootdev will not return
0 with -i if it doesn't find a matching device in sys.
 
> I would think that if rootdev reports failure, we might want the test to fail
> immediately.

Agreed.

[bfreed, 2010-09-01 18:41:08]

http://codereview.chromium.org/3349001/diff/1/2
File client/site_tests/hardware_DiskSize/hardware_DiskSize.py (right):

http://codereview.chromium.org/3349001/diff/1/2#newcode16
client/site_tests/hardware_DiskSize/hardware_DiskSize.py:16: match =
re.search(r'/dev/([^ ]+)', devnode)
On 2010/09/01 18:20:29, Will Drewry wrote:
> On 2010/09/01 18:13:09, bfreed wrote:
> > Hmm...
> > 
> > All our other rootdev-using tests pretty much assume something useful will
be
> > returned here.  This this re search really necessary?
> 
> The re search extracts just the block device sda and not /dev/sda.

I see.  In that case, os.path.basename(devnode) should be used.

> > Is -i useful?
> 
> It will ignore error if the device node doesn't exist.  However, the entry
will
> exist in /proc/partitions even if there's no dev node.  rootdev will not
return
> 0 with -i if it doesn't find a matching device in sys.

Ok.

> > I would think that if rootdev reports failure, we might want the test to
fail
> > immediately.
> 
> Agreed.

Other than changing the search to basename, LGTM.

[Will Drewry, 2010-09-01 20:35:16]

On Wed, Sep 1, 2010 at 1:41 PM,  <bfreed@chromium.org> wrote:
>
> http://codereview.chromium.org/3349001/diff/1/2
> File client/site_tests/hardware_DiskSize/hardware_DiskSize.py (right):
>
> http://codereview.chromium.org/3349001/diff/1/2#newcode16
> client/site_tests/hardware_DiskSize/hardware_DiskSize.py:16: match =
> re.search(r'/dev/([^ ]+)', devnode)
> On 2010/09/01 18:20:29, Will Drewry wrote:
>>
>> On 2010/09/01 18:13:09, bfreed wrote:
>> > Hmm...
>> >
>> > All our other rootdev-using tests pretty much assume something
>
> useful will be
>>
>> > returned here.  This this re search really necessary?
>
>> The re search extracts just the block device sda and not /dev/sda.
>
> I see.  In that case, os.path.basename(devnode) should be used.

Certainly! Done and pushing.

>
>> > Is -i useful?
>
>> It will ignore error if the device node doesn't exist.  However, the
>
> entry will
>>
>> exist in /proc/partitions even if there's no dev node.  rootdev will
>
> not return
>>
>> 0 with -i if it doesn't find a matching device in sys.
>
> Ok.
>
>> > I would think that if rootdev reports failure, we might want the
>
> test to fail
>>
>> > immediately.
>
>> Agreed.
>
> Other than changing the search to basename, LGTM.
>
> http://codereview.chromium.org/3349001/show
>