On Wed, Oct 3, 2018 at 6:32 PM Eugene Syromiatnikov [off-list ref] wrote:

On Wed, Oct 03, 2018 at 09:00:04AM -0700, Yu-cheng Yu wrote:

quoted

On Tue, 2018-10-02 at 22:36 -0700, Andy Lutomirski wrote:

quoted

On Tue, Oct 2, 2018 at 9:55 PM Eugene Syromiatnikov [off-list ref] wrote:

quoted

On Fri, Sep 21, 2018 at 08:03:48AM -0700, Yu-cheng Yu wrote:

quoted

Create a guard area between VMAs, to detect memory corruption.

Do I understand correctly that with this patch a user space program
no longer be able to place two mappings back to back? If it is so,
it will likely break a lot of things; for example, it's a common ring
buffer implementations technique, to map buffer memory twice back
to back in order to avoid special handling of items wrapping its end.

I haven't checked what the patch actually does, but it shouldn't have
any affect on MAP_FIXED or the new no-replace MAP_FIXED variant.

--Andy

I did some mmap tests with/without MAP_FIXED, and it works as intended.
In addition to the ring buffer, are there other test cases?

Right, after some more code reading I figured out that it indeed
shouldn't affect MAP_FIXED, thank you for confirmation.

I'm not sure, however, whether such a change that provides no ability
to configure or affect it will go well with all the supported
architectures.

Is there a concrete reason why you think an architecture might not
like this? As far as I can tell, the virtual address space overhead
should be insignificant even for 32-bit systems.