GitLab

SetSystemPagesAccess is more or less a wrapper around mprotect on posix
systems, and VirtualFree and VirtualAlloc on Windows.  We see crashes,
e.g. in the referenced issue, where the return value of mprotect is not
equal to 0. However, because we do not CHECK the return value of
mprotect directly, we cannot see the reason why mprotect failed. With
this CL I move the CHECK of the return value of mprotect to the
mprotect, so if the CHECK fails, we also see the error reason.

Some caller of SetSystemPagesAccess do not CHECK the result of
SetSystemPagesAccess but deal with it or forward it. For these callers I
introduced a new function, TrySetSystemPagesAccess, which does exactly
the same as SetSystemPagesAccess did until now.

Bug: chromium:839036
Change-Id: I774e648cc6968202805a495fc0b3c3b7d9974b02
Reviewed-on: https://chromium-review.googlesource.com/c/1336130Reviewed-by: Jeremy Roman <jbroman@chromium.org>
Reviewed-by: Kentaro Hara <haraken@chromium.org>
Commit-Queue: Andreas Haas <ahaas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#608310}