GitLab

The size is used only to reserve address space (with mmap(2)). In the
kernel mmap(2) will always silently align to the next page. Hence
lifting this check should be safe.

Recently program headers changed in Chrome DSOs on Android so the amount
of virtual address space to reserve became unaligned (when DFMs are
enabled we need to reserve more address space than what we immediately
require). It does not prevent the system android_dlopen_ext() from
working (as checked by cjgrant@ on the bug), hence this change would
likely make the crazylinker implementation of android_dlopen_ext()
closer to the reference one. Though I did not verify the alignment of
the reserved size for libmonochrome.so (potentially could still stay
aligned because, after all, it is a different library from
libchrome.so).

The amount of testing for this hotfix is inspired by perf season: NO
TESTING BEYOND CQ. Thanks for accepting this :)

Bug: 998712
Change-Id: I330bda4ac47d2cda4b70a15a5a0c26174028723e
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/1784602Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Christopher Grant <cjgrant@chromium.org>
Commit-Queue: Egor Pasko <pasko@chromium.org>
Cr-Commit-Position: refs/heads/master@{#693145}