GitLab

commit 919148f94db54fc04d287eb6a42c0c36b166bbfa
Merge: e2ca3c3 84a5bd9
Author: Jonathan 'Wolf' Rentzsch <jwr.git@redshed.net>
Date:   Sun May 11 21:51:20 2014 -0500

    Merge pull request #16 from mark-chromium/patch-1

    [FIX] Stop using mach_host_self and host_page_size, fixing a port
    right leak. (Mark Mentovai)

commit 84a5bd929213b9e0f059d3bc8c5b738e9fe4e620
Author: Mark Mentovai <mark@chromium.org>
Date:   Fri May 9 16:40:10 2014 -0400

    Stop using mach_host_self and host_page_size, fixing a port right
    leak

    It is incorrect to use mach_host_self without disposing of the send
    send right to the host port with mach_port_deallocate when done with
    it. http://crbug.com/105513 shows the sorts of problems that can
    arise when send rights aren’t properly deallocated.

    mach_host_self was only used by mach_override to be able to call
    host_page_size. host_page_size is unnecessary, because it always
    returns a constant value, PAGE_SIZE, which is also known at
    user-land compile time. See libsyscall/mach/mach_init.c. User code
    is better off just using this macro directly, and not fumbling with
    the system calls to obtain and properly dispose of a send right to
    the host port.

    (You need to mach_port_deallocate the ports you get from
    mach_host_self and mach_thread_self, but you must not normally
    deallocate the one from mach_task_self, because mach_task_self is
    actually just a macro that references a global variable. It doesn’t
    add any port rights at all. See <mach/mach_init.h>. If you bypass
    the macro and call the real mach_task_self system call, you do need
    to call mach_port_deallocate, but this situation is incredibly
    rare.)

R=rsesek@chromium.org

Review URL: https://codereview.chromium.org/282523004

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@269785 0039d316-1c4b-4281-b951-d872f2087c98