Re: [RFC PATCH glibc 1/4] glibc: Perform rseq(2) registration at nptl init and thread creation (v4)

[Mathieu Desnoyers]

----- On Jan 10, 2019, at 12:31 PM, Mathieu Desnoyers mathieu.desnoyers@xxxxxxxxxxxx wrote:

> ----- On Dec 11, 2018, at 2:40 AM, Florian Weimer fweimer@xxxxxxxxxx wrote:
[...]
>> 
>>> diff --git a/nptl/Versions b/nptl/Versions
>>> index e7f691da7a..f7890f73fc 100644
>>> --- a/nptl/Versions
>>> +++ b/nptl/Versions
>>> @@ -277,6 +277,10 @@ libpthread {
>>>      cnd_timedwait; cnd_wait; tss_create; tss_delete; tss_get; tss_set;
>>>    }
>>>  
>>> +  GLIBC_2.29 {
>>> +    __rseq_refcount;
>>> +  }
>> 
>> Why put this into libpthread, and __rseq_abi into libc?
> 
> The __rseq_abi symbol should be available to the glibc memory allocator.
> I plan to move the __rseq_abi to sysdeps/unix/sysv/linux/Versions instead
> so it becomes Linux-specific.
> 
> The __rseq_refcount symbol only needs to be made available to applications
> and libraries linking against libpthread, because only libpthread actually
> handles the rseq registration/unregistration at thread start/exit and
> library initialization.
> 
> However, considering that we want this to be Linux-specific as well,
> I could move it to sysdeps/unix/sysv/linux/Versions too.
> 
> Then it would make sense to move the __rseq_refcount symbol defined in
> nptl/rseq.c to sysdeps/unix/sysv/linux/rseq.c as well and group
> everything together.
> 
> Therefore, both symbols will end up in sysdeps/unix/sysv/linux/Versions.

Now that I think about it, it's important to move the rseq registration
done at nptl init (in my current code) to some lower-level csu initialiation,
so applications that happen _not_ to link against libpthread also get
registered rseq for the main thread.

Does it make sense ?

Thanks,

Mathieu

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