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

From: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
To: Joseph Myers <joseph@codesourcery.com>
Cc: carlos <carlos@redhat.com>, Florian Weimer <fweimer@redhat.com>,
	Thomas Gleixner <tglx@linutronix.de>, Ben Maurer <bmaurer@fb.com>,
	Peter Zijlstra <peterz@infradead.org>,
	"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>,
	Boqun Feng <boqun.feng@gmail.com>,
	Will Deacon <will.deacon@arm.com>,
	Dave Watson <davejwatson@fb.com>, Paul Turner <pjt@google.com>,
	libc-alpha <libc-alpha@sourceware.org>,
	linux-kernel <linux-kernel@vger.kernel.org>,
	linux-api <linux-api@vger.kernel.org>
Subject: Re: [RFC PATCH] glibc: Perform rseq(2) registration at nptl init and thread creation
Date: Thu, 20 Sep 2018 16:04:26 -0400 (EDT)	[thread overview]
Message-ID: <381061785.8994.1537473866841.JavaMail.zimbra@efficios.com> (raw)
In-Reply-To: <alpine.DEB.2.21.1809191630260.26757@digraph.polyomino.org.uk>

----- On Sep 19, 2018, at 12:37 PM, Joseph Myers joseph@codesourcery.com wrote:

> On Wed, 19 Sep 2018, Mathieu Desnoyers wrote:
> 
>> Here is a rough prototype registering rseq(2) TLS for each thread
>> (including main), and unregistering for each thread (excluding
>> main). "rseq" stands for Restartable Sequences.
> 
> A final patch would need to add documentation and tests and a NEWS entry
> and fix various coding style issues.

Sure,

> 
>> diff --git a/nptl/Versions b/nptl/Versions
>> index e7f691da7a..7316c2815d 100644
>> --- a/nptl/Versions
>> +++ b/nptl/Versions
>> @@ -275,6 +275,7 @@ libpthread {
>>      mtx_init; mtx_lock; mtx_timedlock; mtx_trylock; mtx_unlock; mtx_destroy;
>>      call_once; cnd_broadcast; cnd_destroy; cnd_init; cnd_signal;
>>      cnd_timedwait; cnd_wait; tss_create; tss_delete; tss_get; tss_set;
>> +    __rseq_abi; __rseq_refcount;
> 
> That's the GLIBC_2.28 section, but 2.28 is already out.  New symbols would
> need to go in GLIBC_2.29 or later (and the ABI test baselines would all
> need updating).

OK

> 
>> diff --git a/sysdeps/unix/sysv/linux/rseq.h b/sysdeps/unix/sysv/linux/rseq.h
> 
> This looks like it's coming from the Linux kernel.  Can't the relevant
> uapi header just be used directly without copying into glibc (with due
> care to ensure that glibc still builds if the kernel headers used for the
> build are too old - you need such conditionals anyway if they don't define
> the relevant syscall number)?

Something like this in pthreadP.h ?

+#ifdef __NR_rseq
+#include <sysdeps/unix/sysv/linux/rseq-internal.h>
+#else
+#include <sysdeps/nptl/rseq-internal.h>
+#endif  /* __NR_rseq.  */

where sysdeps/unix/sysv/linux/rseq-internal.h contains the linux
implementation of rseq_register_current_thread () and
rseq_unregister_current_thread (), and sysdeps/nptl/rseq-internal.h
contains stubs.

Am I on the right track ?

Thanks,

Mathieu

> 
> --
> Joseph S. Myers
> joseph@codesourcery.com

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