Re: TLS descriptors for ARM

[Alexandre Oliva <aoliva at redhat dot com>]

On Jan  5, 2011, Nathan Sidwell <nathan@codesourcery.com> wrote:

> This patch adds support for TLS descriptors for ARM.  Refer to
> http://www.codesourcery.com/publications/RFC-TLSDESC-ARM.txt for the
> current specification.

Neat!


It refers to R_ARM_TLS_DESCSEQ16 and R_ARM_TLS_DESCSEQ32, but no such
relocations are defined in the document.  You may want to add a blurb
about that like you did to the e-mail.  

I suppose you meant to refer to R_ARM_THM_TLS_DESCSEQ too, although I
don't quite understand the rationale for the THM-specific relocations;
wouldn't non-THM relocations with odd addresses have worked?


The description of IE relaxation for TLS_CALLs gives me the impression
that a single instruction could be replaced with two.  What am I
missing?


As for the special handling for local symbols in the TLS_DESC
relocation...  I *think* refraining from decaying local TLS symbols to
section + offset would have worked, although this arrangement is
certainly more efficient.  One inconsistency I kind of dislike is that
it permits addends for local symbols but not for global symbols, which
can be confusing (and only reported at link time?), but I guess that's
not too bad.  Limiting oneself to REL relocations, rather than RELA,
it's hard to do better.

-- 
Alexandre Oliva, freedom fighter    http://FSFLA.org/~lxoliva/
You must be the change you wish to see in the world. -- Gandhi
Be Free! -- http://FSFLA.org/   FSF Latin America board member
Free Software Evangelist      Red Hat Brazil Compiler Engineer