Re: [PATCH] Async signal safe TLS accesses

["Carlos O'Donell" <carlos at redhat dot com>]

On 12/06/2013 05:50 PM, Andrew Hunter wrote:
> I'm looking into fixing TLDDESC.  Is this the best available documentation:
> 
> http://www.fsfla.org/~lxoliva/writeups/TLS/RFC-TLSDESC-x86.txt
> 
> ?

As far as I know it is the canonical documentation for them.

> I note with some interest that using mask_all_signals whenever we take
> dl_load_lock would (very nearly, except for STB_GNU_UNIQUE symbols)
> solve the problem outright.  I welcome comment from others as to
> whether this carries acceptable overhead.
> 
> If not, I see two obvious possibilities
> 
> - deprecate lazy relocations; I'm unclear if the gains from not doing
> a single dl_lookup_symbol_x per reloc at dlopen time are actually
> significant (though maybe I'm missing the point of lazy relocations?)
> AFAICT, the current scheme is more than capable of supporting the
> TLSDESC indirections that actually use __tls_get_addr and friends.  I
> favor this scheme unless someone can tell me a good reason not to do
> so.
> 
> - mask signals in _dl_tlsdesc_resolve_rela_fixup;  use a dedicated
> lock instead of dl_load_lock in _dl_tlsdesc_resolve_early_return_p
> (easy). make dl_lookup_symbol_x async-signal-safe (ugh).
> 
> Open to other suggestions as well.

Yes, I suggest:

- Declare GNU2 dialect TLS symbols AS-unsafe for the time being
  and tackle it later.

Cheers,
Carlos.