This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [ping] Change static chain to r11 on aarch64

Am Mittwoch, den 12.12.2018, 18:53 +0000 schrieb Wilco Dijkstra:
> Hi,
> > > On 12 Dec 2018,@18:21, Richard Earnshaw (lists) <>
> wrote:
> > > However, that introduces an issue that that
> > > code is potentially used across multiple versions of gcc, with
> > > potentially different choices of the static chain register.  Hmm, this
> > > might need some more careful thought....
> The static chain is only used inside nested functions, so it's not an ABI but a
> function-local agreement. Although it looks like you can take the address of
> a nested function, I think you cannot ever export it in a way that exposes a
> different static chain given each address-taken nested function would emit
> its own trampoline on the stack.
> In fact the trampoline implementation is broken by design since the stack
> should not be executable by default.

Does a non-executable stack actually improve security?

For the alternative implementation using (custom) function
descriptors (-fno-trampolines) the static chain becomes
part of the ABI or not?


> > > I'm also not keen on the fact that we are now seriously eating into the
> > > space of call clobbered registers; what's the argument behind your
> > > selection of r11 as opposed to any other register?
> The static chain register is only used on entry to a nested function.
> That's why I suggested using x9 given x8 is the last argument register.
> > suggested r9, then I discovered that r9 and r10 were used
> > by the stack probing mechanism, so I just picked the following
> > one that didn't seem to be used for other purposes already.
> We could rename those temporaries if we think x9 is better than x11.
> Wilco

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]