This is the mail archive of the
mailing list for the glibc project.
Re: protected data in a DSO with copy relocation
- From: OndÅej BÃlka <neleai at seznam dot cz>
- To: "H.J. Lu" <hjl dot tools at gmail dot com>
- Cc: Roland McGrath <roland at hack dot frob dot com>, GNU C Library <libc-alpha at sourceware dot org>
- Date: Thu, 18 Dec 2014 20:50:04 +0100
- Subject: Re: protected data in a DSO with copy relocation
- Authentication-results: sourceware.org; auth=none
- References: <54919EB5 dot 5030209 at redhat dot com> <CAMe9rOrGcuMQ3J9iNwKvZ-H9irdqjataJzt1+Auck5_q6sOCNA at mail dot gmail dot com> <20141217225330 dot 375CB2C3ACD at topped-with-meat dot com> <CAMe9rOoA2a2rXjnNxZXoQ7yuRPGzgi7mC370TZuDGmfyLycR1A at mail dot gmail dot com>
On Wed, Dec 17, 2014 at 03:22:53PM -0800, H.J. Lu wrote:
> On Wed, Dec 17, 2014 at 2:53 PM, Roland McGrath <firstname.lastname@example.org> wrote:
> >> I'd like to have a decision on
> >> https://sourceware.org/bugzilla/show_bug.cgi?id=17711
> >> If we want to support protected data in a DSO with copy relocation,
> >> which will incur run-time and memory overhead, I can prepare ld.so
> >> as well as linker patches. Otherwise, I will make a linker patch
> >> to disallow protected data in a DSO with copy relocation.
> > I think we certainly shouldn't do any rtld changes for this in the current
> > cycle. If we can just punt thinking about it at all until after this
> If we punt, we need to update the elf/vismain test which fails with
> the current ld.
> > release, that would be best. If we need some kind of decision sooner, then
> > my inclination is to say that STV_PROTECTED is not useful enough to spend
> > effort on supporting its corner cases. I'm not aware of uses cases for it
> > that both appear in the real world and are actually worthwhile. But if
> > there is no urgent need, then let's discuss this in detail after the release.
> I agree.
I also think it should not be supported. I see another potential problem
that it makes maintainance harder, for example uses it and forgots bump
Makefile requirements then when somebody with older system will run to