On Sep 14, 4:57 pm, David Given wrote:
> If I was going to have overcommit enabled, I'd much rather have the
> allocation failures exposed to the application in the form of the
> appropriate signal when the application fails to access the page.
[...]
> Incidentally, does Android use rlimits to im
On Mon, Sep 14, 2009 at 4:57 PM, David Given wrote:
> In fact, I find the OOM killer behaviour deeply suspect. I do not
> believe that randomly killing processes is *ever* the right thing to do
> on a reliable system, as it completely denies the application the
> ability to properly clean up afte
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
David Turner wrote:
[...]
> That can only be said for files that are mapped read-only. It doesn't apply
> for
> anything that is shared copy-on-write (e.g. the Zygote process pages, as
> well
> as initial heap) and non-file mappings.
Ah, I'd forgotte
On Mon, Sep 14, 2009 at 3:53 PM, David Given wrote:
>
> -BEGIN PGP SIGNED MESSAGE-
> Hash: SHA1
>
> Dianne Hackborn wrote:
> [...]
> > If you turn off over-commit, I believe Android won't even boot on a G1,
> > because Linux would need to assume that RAM is needed for every mmapped()
> >
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Dianne Hackborn wrote:
[...]
> If you turn off over-commit, I believe Android won't even boot on a G1,
> because Linux would need to assume that RAM is needed for every mmapped()
> thing, and we run out of it well before the system is fully up.
Not n
On Sun, Sep 13, 2009 at 10:12 AM, David Given wrote:
> Yes, but that doesn't really answer my question --- *why* is memory
> overcommit used on Android? It's normally used on systems with huge
> amounts of swap to make more efficient use of physical memory, but
> Android devices don't have any sw
David Turner wrote:
[...]
> memory overcommit is used by design, there is no plan to remove it. The fact
> that the OOM killer will nuke processes to make room for others is part
> of the
> platform's design. As far as I know, the OOM changes were to make the killer
> a bit smarter about what kin
On Sun, Sep 13, 2009 at 4:34 AM, David Given wrote:
>
> Hello,
>
> I was rather startled recently to notice that the standard Android
> kernel appears to have the memory overcommit setting set to 1. This is
> --- as far as I can tell, the numbering got changed not long ago and not
> all the docum
On Sun, Sep 13, 2009 at 4:34 AM, David Given wrote:
>
> Hello,
>
> I was rather startled recently to notice that the standard Android
> kernel appears to have the memory overcommit setting set to 1. This is
> --- as far as I can tell, the numbering got changed not long ago and not
> all the docum
