Make PartitionAllocDeathTest.GuardPages pass on Win64.

Make PartitionAllocDeathTest.GuardPages pass on Win64.

On Windows, the allocation granularity is 64kB instead of 4kB elsewhere,
so if a multiple of 4kB data is requested from partitionAlloc() there
is some free space at the end.  On 32-bit 4kB following the allocation
is mapped as inaccessible, but on 64-bit this isn't done since ASLR
should cover writes that are far enough out-of-bounds.  So make sure
GuardPages writes outside of the allocated memory, and not into the
free space at the end of the allocation.

BUG=524308
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=201453

[Nico, 2015-08-28 23:07:57]

thakis@chromium.org changed reviewers:
+ haraken@chromium.org, scottmg@chromium.org, tsepez@chromium.org

[Nico, 2015-08-28 23:07:58]

scottmg, if  you're comfortable stamping this (since I mail-bombed you through
the bug) I'd take that. If not, tsepez or haraken can look after the weekend.

[scottmg, 2015-08-29 01:34:17]

The updated logic of the test/comment smells-lgtm, but Tom should really do a
thorough review when he returns.

[haraken, 2015-08-29 07:58:13]

LGTM

I guess NaCl will hit the same problem. NaCl is using a system page of 64 KB as
well.

[Nico, 2015-08-29 15:45:14]

On 2015/08/29 07:58:13, haraken wrote:
> LGTM
> 
> I guess NaCl will hit the same problem. NaCl is using a system page of 64 KB
as
> well.

Thanks!

Do we build blink with the NaCl toolchain somewhere? From what I can
tell,kPageAllocationGranularity is only set to 64k on win; if we build blink
with NaCl we should probably set it to 64k in that case too?

[Nico, 2015-08-29 15:45:27]

The CQ bit was checked by thakis@chromium.org

[commit-bot: I haz the power, 2015-08-29 15:45:30]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1314843008/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1314843008/20001

[commit-bot: I haz the power, 2015-08-29 15:49:42]

Committed patchset #2 (id:20001) as
https://src.chromium.org/viewvc/blink?view=rev&revision=201453

[haraken, 2015-08-30 01:21:53]

On 2015/08/29 15:45:14, Nico wrote:
> On 2015/08/29 07:58:13, haraken wrote:
> > LGTM
> > 
> > I guess NaCl will hit the same problem. NaCl is using a system page of 64 KB
> as
> > well.
> 
> Thanks!
> 
> Do we build blink with the NaCl toolchain somewhere? From what I can
> tell,kPageAllocationGranularity is only set to 64k on win; if we build blink
> with NaCl we should probably set it to 64k in that case too?

Not yet but planned. I don't know the latest status but the NaCl team is working
on that now.