[Mac] Use raw disks (/dev/rdisk#) in image_writer_mac.

[Mac] Use raw disks (/dev/rdisk#) in image_writer_mac.

For some USB drives, writes on Mac are much slower than using the old image
creator. This is most likely because the implementation in Chrome uses
/dev/disk# whereas the old implementation used /dev/rdisk# whenever
possible.

The main requirement is that writes are in multiples of the block size, but
since ImageWriter::WriteChunk always writes in multiples of
kMemoryAlignment (4096), it should be fine to always use the raw disk.

My brief tests using a Lexar JumpDrive 16GB and a 1.6GB image:
before: ~6.5 minutes to write, same to verify
after: ~5 minutes to write, ~1.5 minutes to verify

BUG=415891
Committed: https://crrev.com/e06a8c6b35efd534d911d9d6382daf5e8986a1d1
Cr-Commit-Position: refs/heads/master@{#296875}

[jackhou1, 2014-09-24 09:05:31]

jackhou@chromium.org changed reviewers:
+ avi@chromium.org

[jackhou1, 2014-09-24 09:05:32]

avi, I think you wrote the code in the old image creator, could you take a look
at this?

[Avi (use Gerrit), 2014-09-24 15:30:46]

lgtm

[Avi (use Gerrit), 2014-09-24 15:31:36]

https://codereview.chromium.org/599853002/diff/1/chrome/utility/image_writer/...
File chrome/utility/image_writer/image_writer_mac.cc (right):

https://codereview.chromium.org/599853002/diff/1/chrome/utility/image_writer/...
chrome/utility/image_writer/image_writer_mac.cc:104: // multiples of
kMemoryAlignment.
I'm not familiar with WriteChunk. Can we DCHECK in there or something to ensure
this constraint that the write size is divisible by the media block size?

[jackhou1, 2014-09-24 23:58:43]

https://codereview.chromium.org/599853002/diff/1/chrome/utility/image_writer/...
File chrome/utility/image_writer/image_writer_mac.cc (right):

https://codereview.chromium.org/599853002/diff/1/chrome/utility/image_writer/...
chrome/utility/image_writer/image_writer_mac.cc:104: // multiples of
kMemoryAlignment.
On 2014/09/24 15:31:36, Avi wrote:
> I'm not familiar with WriteChunk. Can we DCHECK in there or something to
ensure
> this constraint that the write size is divisible by the media block size?

Added a DCHECK. I'm not sure if there's an easy way to get the DAMediaBlockSize,
but searching online, it only ever seems to be 512, so I think DCHECKing against
kMemoryAlignment should be okay.

Judging by the comments, WriteChunk was written with this in mind:
https://code.google.com/p/chromium/codesearch#chromium/src/chrome/utility/ima...

[jackhou1, 2014-09-24 23:59:11]

jackhou@chromium.org changed reviewers:
+ miket@chromium.org

[jackhou1, 2014-09-24 23:59:12]

miket, could you review for OWNERS in chrome/utility/image_writer/

[miket_OOO, 2014-09-26 02:19:59]

lgtm

[jackhou1, 2014-09-26 02:50:46]

The CQ bit was checked by jackhou@chromium.org

[commit-bot: I haz the power, 2014-09-26 02:51:57]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/599853002/20001

[commit-bot: I haz the power, 2014-09-26 04:00:13]

Committed patchset #2 (id:20001) as 9874d6e8043a45762bade1e280dd62aecb4cc544

[commit-bot: I haz the power, 2014-09-26 04:00:48]

Patchset 2 (id:??) landed as
https://crrev.com/e06a8c6b35efd534d911d9d6382daf5e8986a1d1
Cr-Commit-Position: refs/heads/master@{#296875}