AU: Don't mark booted kernel good if we've applied an update.

AU: Don't mark booted kernel good if we've applied an update.

Otherwise, we may end up reverting the update.

BUG=7553
TEST=unit tests; tested on device with modified chromeos-setgoodkernel script.

Change-Id: I52bd025947ddfc9146644fe9449a4446ea679b3e

Committed: http://chrome-svn/viewvc/chromeos?view=rev&revision=eee26ee

[adlr, 2010-10-21 23:51:12]

http://codereview.chromium.org/3980005/diff/1/2
File main.cc (right):

http://codereview.chromium.org/3980005/diff/1/2#newcode150
main.cc:150: g_timeout_add_seconds(45, &chromeos_update_engine::UpdateBootFlags,
NULL);
race condition:

- status is DOWNLOADING
- set timout for update boot flags
- status changes to NEED_REBOOT
- boot flags get updated.

[adlr, 2010-10-21 23:52:42]

err, i didn't quite get that race condition exactly right, but the point is
we should check the status closer to running set good kernel.

On Thu, Oct 21, 2010 at 4:51 PM, <adlr@chromium.org> wrote:

>
> http://codereview.chromium.org/3980005/diff/1/2
> File main.cc (right):
>
> http://codereview.chromium.org/3980005/diff/1/2#newcode150
> main.cc:150: g_timeout_add_seconds(45,
>
> &chromeos_update_engine::UpdateBootFlags, NULL);
> race condition:
>
> - status is DOWNLOADING
> - set timout for update boot flags
> - status changes to NEED_REBOOT
> - boot flags get updated.
>
>
> http://codereview.chromium.org/3980005/show
>

[petkov, 2010-10-21 23:57:55]

Yeah, I thought this was unlikely but true -- it could happen if the
user initiates an update check through the UI.

On 2010/10/21 23:52:42, adlr wrote:
> err, i didn't quite get that race condition exactly right, but the point is
> we should check the status closer to running set good kernel.
> 
> On Thu, Oct 21, 2010 at 4:51 PM, <mailto:adlr@chromium.org> wrote:
> 
> >
> > http://codereview.chromium.org/3980005/diff/1/2
> > File main.cc (right):
> >
> > http://codereview.chromium.org/3980005/diff/1/2#newcode150
> > main.cc:150: g_timeout_add_seconds(45,
> >
> > &chromeos_update_engine::UpdateBootFlags, NULL);
> > race condition:
> >
> > - status is DOWNLOADING
> > - set timout for update boot flags
> > - status changes to NEED_REBOOT
> > - boot flags get updated.
> >
> >
> > http://codereview.chromium.org/3980005/show
> >
>

[petkov, 2010-10-22 00:10:35]

PTAL. I did some basic testing and I'll do some more testing tomorrow. I'll push
tomorrow if it LGTY.

[adlr, 2010-10-22 00:45:28]

LGTM

http://codereview.chromium.org/3980005/diff/7001/8001
File main.cc (right):

http://codereview.chromium.org/3980005/diff/7001/8001#newcode49
main.cc:49: if (attempter->status() != UPDATE_STATUS_IDLE) {
since it's single threaded, and this and postinst are synchronous, I think it's
okay to go ahead and call setgoodkernel when status != NEED_REBOOT or
FINALIZING, but your version makes fewer requirements about things being
synchronous, which is good.