Re: [PATCH 1/5] glibc: Perform rseq(2) registration at C startup and thread creation (v10)

[Mathieu Desnoyers <mathieu dot desnoyers at efficios dot com>]

----- On Jun 14, 2019, at 3:53 PM, Florian Weimer fweimer@redhat.com wrote:

> * Mathieu Desnoyers:
> 
>> ----- On Jun 14, 2019, at 3:42 PM, Florian Weimer fweimer@redhat.com wrote:
>>
>>> * Mathieu Desnoyers:
>>> 
>>>> +  /* Publicize rseq registration ownership.  This must be performed
>>>> +     after rtld re-relocation, before invoking constructors of
>>>> +     preloaded libraries.  */
>>>> +  rseq_init ();
>>> 
>>> Please add a comment that IFUNC resolvers do not see the initialized
>>> value.  I think this is okay because we currently do not support access
>>> to extern variables in IFUNC resolvers.
>>
>> Do IFUNC resolvers happen to observe the __rseq_handled address that
>> was internal to ld.so ?
> 
> They should observe the correct address, but they can access the
> variable before initialization.  An initializer in ld.so will not have
> an effect if an interposed definition initalized the variable to
> something else.
> 
>> If so, we could simply initialize __rseq_handled twice: early before calling
>> IFUNC resolvers, and after ld.so re-relocation.
> 
> No, I don't think this will make a difference.

So comment it is:

  /* Publicize rseq registration ownership.  This must be performed
     after rtld re-relocation, before invoking constructors of
     preloaded libraries. IFUNC resolvers are called before this
     initialization, so they may not observe the initialized state.  */
  rseq_init ();

Thanks,

Mathieu


-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com