Enable check for DEP / NX page protection

Call into NaClCheckDEP() wherever we call NaClOsIsSupported.
- Currently requires passing the -Q flag through sel_universal
  to sel_ldr (maybe in a future CL that can be done as part the
  sel_ldr flags portion of sel_universal's flags instead)
- Renamed the arch-specific files to make gyp/windows build happy.
  Perhaps they don't like that the files under the OS-dirs have
  the same name.
- Turn off ARM dep check for now. Otherwise, the arm HW bots timeout
  on {app_lib_test, egyptian_cotton, selenium, ...} (bug filed).
- If disable_hardy64_vmware_failures is set, skip the fake_browser* tests.


BUG= http://code.google.com/p/nativeclient/issues/detail?id=239
TEST= gcl try
and locally (Hardy64) run smoke_tests for platform={arm, x86-32, x86-64}.


Committed: http://src.chromium.org/viewvc/native_client?view=rev&revision=3479

[jvoung - send to chromium..., 2010-10-07 02:28:22]

Hi Cliff -- Trybots now pass, but I had to cheat a bit; I turned off the DEP
check for ARM. 

With the DEP check *on* for ARM, the hardware bots would quite consistently
timeout while running things like "app_lib_test". Note that this test normally
finishes in under a second, with qemu. As for properly debugging this, the
chrome-bot login doesn't seem to work for nacl-arm1/nacl-arm2. I've emailed the
chrome-infrastructure folks and will have to see what they say.

Any objections to turning that on later? Other comments?

[Cliff L. Biffle, 2010-10-08 23:04:31]

LGTM.  I saw the bug you filed to track the disabled ARM tests; thank you for
doing that.

[jvoung - send to chromium..., 2010-10-11 23:01:37]

+Mark

Is there a clean way to send the "-Q" sel_ldr flag from fake_browser_test (and
the ppapi fake_browser_test) to sel_main?

If not, would you be against skipping this test on the hardy64 vmware buildbots
(otherwise NaClCheckDEP might fail on the bot
http://buildbot.jail.google.com/buildbot/nacl-tryserver/builders/nacl-hardy64...

[Mark Seaborn, 2010-10-12 10:10:45]

On 2010/10/11 23:01:37, jvoung wrote:
> +Mark
> 
> Is there a clean way to send the "-Q" sel_ldr flag from fake_browser_test (and
> the ppapi fake_browser_test) to sel_main?

No, I'm afraid there isn't.  It might be worth adding an env var (for either the
plugin or sel_ldr to interpret) or something like that.

> If not, would you be against skipping this test on the hardy64 vmware
buildbots
> (otherwise NaClCheckDEP might fail on the bot
>
http://buildbot.jail.google.com/buildbot/nacl-tryserver/builders/nacl-hardy64...

That's OK.  I'm surprised the problem is still there.

BTW, a better title for the changeset would be "Enable check for DEP / NX page
protection" (assuming that's what it does).  That would be easier to understand
than references to particular functions.

Cheers,
Mark

[Cliff L. Biffle, 2010-10-12 22:22:47]

Jan, are we good to go here?

[jvoung - send to chromium..., 2010-10-12 22:26:58]

On 2010/10/12 22:22:47, Cliff L. Biffle wrote:
> Jan, are we good to go here?

Getting some trybot errors on the windows 64 bots after some gclient syncs.
Trying again with gcl try -rHEAD. Also trying the tests on my own windows
machine in parallel.

I'll commit if the tests pass.

[jvoung - send to chromium..., 2010-10-12 23:15:10]

Ok, the trybot errors are reproduce-able on my machine as well (and only with
the DEP check turned on). 

** Fault in NaCl trusted code.

Will try to debug this...

Btw, total stab in the dark, but do you happen to know if factoring in Noel's
nacl_signal change would mean that this check needs to be updated (sorry for
naming you Noel)?

[jvoung - send to chromium..., 2010-10-13 22:04:09]

Hi Cliff,

New version of the win/nacl_dep_qualify using Noel's signal interface. Will send
off a try-job once Noel's CL lands and NaClSignalHandlerRemove is actually
available (smoke_tests pass on my windows machine with that part commented out).

[jvoung - send to chromium..., 2010-10-14 18:12:56]

On 2010/10/13 22:04:09, jvoung wrote:
> Hi Cliff,
> 
> New version of the win/nacl_dep_qualify using Noel's signal interface. Will
send
> off a try-job once Noel's CL lands and NaClSignalHandlerRemove is actually
> available (smoke_tests pass on my windows machine with that part commented
out).

Try bots pass now.

[Cliff L. Biffle, 2010-10-14 18:27:07]

LGTM, and thanks.