[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20170707184843.GA3113@redhat.com>
Date: Fri, 7 Jul 2017 14:48:44 -0400
From: Jerome Glisse <jglisse@...hat.com>
To: Igor Stoppa <igor.stoppa@...wei.com>
Cc: keescook@...omium.org, mhocko@...nel.org, jmorris@...ei.org,
labbott@...hat.com, hch@...radead.org,
penguin-kernel@...ove.SAKURA.ne.jp, paul@...l-moore.com,
sds@...ho.nsa.gov, casey@...aufler-ca.com,
linux-security-module@...r.kernel.org, linux-mm@...ck.org,
linux-kernel@...r.kernel.org, kernel-hardening@...ts.openwall.com
Subject: Re: [PATCH 1/3] Protectable memory support
On Fri, Jul 07, 2017 at 11:42:09AM +0300, Igor Stoppa wrote:
> On 06/07/17 19:27, Jerome Glisse wrote:
> > On Wed, Jul 05, 2017 at 04:46:26PM +0300, Igor Stoppa wrote:
[...]
> > Yet another way is to use some of the free struct page fields ie
> > when a page is allocated for vmalloc i think most of struct page
> > fields are unuse (mapping, index, lru, ...). It would be better
> > to use those rather than adding a page flag.
>
> Like introducing an unnamed union? Some sort of vmalloc_page_subtype?
> If that is what you are proposing, I agree that it would work in a
> similar fashion as what I have now, but without introducing the overhead
> of the extra page flag.
No need to introduce unamed union or anything. Just use one of the
existing field for install you can make page->mapping point to the
pmalloc pool structure. Or you can store a unique key value.
I believe there is enough unuse field that for vmalloc pages that
you should find one you can use. Just add some documentation in
mm_types.h so people are aware of alternate use for the field you
are using.
Jérôme
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
