[PATCH] D33251: [lld][ELF]Add option to make .dynamic read only

[Rui Ueyama via llvm-commits]

Roland,

What I was trying to do is to understand it, and I'm not pushing it back
blindly. I was posing questions to understand the proposal.

The original proposal didn't say much about the reason why we want a new
flag. The first reason I was told was memory saving, which is dubious as we
discussed here. And then vDSO's story came in. And then it is now clear
that you guys just want to make all .dynamic read-only in the ABI of the
new OS.

So the true reason why you want to add `-z rodynamic` is because in the
Fuchsia ABI, .dynamic is read-only, and we want to have some way to make it
actually so in ELF files. Probably virtually no memory savings, no security
improvement, but this should be done because that's the right thing to do.
If that's the reason, I think I'm OK with the new flag.

On Thu, May 25, 2017 at 2:12 PM, Roland McGrath <mcgrathr at google.com> wrote:

> Look, we have a new system, and we're defining what the ABI will be
> for our system.
> We've decided we want .dynamic to be read-only and we are confident in
> that choice.
> It's OK if you aren't convinced that's the decision you would make in
> your ABI, but it's the decision we've made for our ABI.
>
> I don't see why you feel the need to push back on a new platform's ABI
> decisions when we've done the implementation
> work for LLD to support the ABI and the actual impact on the LLD code
> base is tiny.
>
> Since LLD doesn't support selecting its behavior based on target
> triple, we've added a switch.
> If you'd prefer this switch were spelled --target=fuchsia rather than
> -z rodynamic, that's fine with us.
>
>
> Thanks,
> Roland
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170525/bf518305/attachment.html>