From: Mathieu Desnoyers <mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>
To: Andy Lutomirski <luto-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
Cc: "Paul E. McKenney"
<paulmck-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>,
linux-kernel
<linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>,
linux-api <linux-api-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>,
Peter Zijlstra <peterz-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org>,
Boqun Feng <boqun.feng-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>,
Andrew Hunter <ahh-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>,
maged michael
<maged.michael-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>,
gromer <gromer-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>,
Avi Kivity <avi-VrcmuVmyx1hWk0Htik3J/w@public.gmane.org>,
Benjamin Herrenschmidt
<benh-XVmvHMARGAS8U2dJNN8I7kB+6BGkLq7r@public.gmane.org>,
Paul Mackerras <paulus-eUNUBHrolfbYtjvyW6yDsg@public.gmane.org>,
Michael Ellerman <mpe-Gsx/Oe8HsFggBc27wqDAHg@public.gmane.org>,
Dave Watson <davejwatson-b10kYP2dOMg@public.gmane.org>,
Thomas Gleixner <tglx-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org>,
Ingo Molnar <mingo-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>,
"H. Peter Anvin" <hpa-YMNOUZJC4hwAvxtiuMwx3w@public.gmane.org>,
Andrea Parri
<parri.andrea-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>,
x86 <x86-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
Subject: Re: [RFC PATCH for 4.15 5/6] membarrier: x86: Provide core serializing command
Date: Thu, 9 Nov 2017 19:35:25 +0000 (UTC) [thread overview]
Message-ID: <452803516.12441.1510256125252.JavaMail.zimbra@efficios.com> (raw)
In-Reply-To: <CALCETrX4vZHxkWhLKBebZ+R_UE8pDsmVbpJzpyPt=ZgZB8w8iQ-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
----- On Nov 9, 2017, at 2:07 PM, Andy Lutomirski luto-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org wrote:
> On Wed, Nov 8, 2017 at 10:35 AM, Mathieu Desnoyers
> <mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org> wrote:
>
>> +/*
>> + * x86-64 implements return to user-space through sysret, which is not a
>> + * core-serializing instruction. Therefore, we need an explicit core
>> + * serializing instruction after going from kernel thread back to
>> + * user-space thread (active_mm moved back to current mm).
>> + */
>> +static inline void membarrier_arch_mm_sync_core(struct mm_struct *mm)
>> +{
>> + if (likely(!(atomic_read(&mm->membarrier_state) &
>> + MEMBARRIER_STATE_SYNC_CORE)))
>> + return;
>> + sync_core();
>> +}
>
> IMO there should be an extremely clear specification somewhere for
> what this function is supposed to do.
>
> If I remember correctly, it's supposed to promise that the icache is
> synced before the next time we return to usermode for the current mm
> on this CPU. If that's correct, then let's document it very
> explicitly and let's also drop the "membarrier" from the name -- it's
> a primitive we'll need anyway given the existing migration bug.
I understand that on x86 (specifically), synchronizing the icache and
doing a core serializing instruction may mean the same thing.
However, on architectures like ARM, icache sync differs from core
serialization. Those architectures typically have either a user-space
accessible instruction or a system call to perform the icache flush.
The missing part for JIT is core serialization (also called context
synchronization). icache flush is already handled by pre-existing
means.
So the promise here given by membarrier_arch_mm_sync_core() is that
a core serializing instruction is issued before the next time we return
to usermode on the current thread. However, we only need that guarantee
if the current thread's mm is a registered MEMBARRIER_{...}_SYNC_CORE user.
Regarding the existing migration bug, what I think you want is a kind
of weaker "sync_core()", which ensures that a core serializing
instruction is issued before the next time the current thread returns
to usermode.
It could be e.g.: set_tsk_need_core_sync() which would set a
TIF_NEED_CORE_SYNC thread flag on the current thread.
Clearly, when this kind of thread flag is introduced as an
optimization over sync_core(), I would like to use that. However,
I don't think it replaces the membarrier_arch_mm_sync_core() entirely,
given that it would not check for the mm membarrier "SYNC_CORE"
registration state. It appears to me to be merely an optimization over
directly invoking sync_core.
What I suggest is that I update the comment above
membarrier_arch_mm_sync_core to spell out more clearly that all we
need is to have a core serializing instruction issued before the next
time the current thread returns to user-space. I can still use
sync_core for now, and we can then improve the implementation
whenever a new thread flag is introduced. The new comment would look
like:
/*
* x86-64 implements return to user-space through sysret, which is not a
* core-serializing instruction. Therefore, we need an explicit core
* serializing instruction after going from kernel thread back to
* user-space thread (active_mm moved back to current mm).
*
* This function ensures that a core serializing instruction is issued
* before the current thread returns to user-space.
*/
Thoughts ?
Thanks,
Mathieu
--
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com
next prev parent reply other threads:[~2017-11-09 19:35 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-11-08 18:35 [RFC PATCH for 4.15 0/6] membarrier updates for 4.15 Mathieu Desnoyers
2017-11-08 18:35 ` [RFC PATCH for 4.15 1/6] membarrier: selftest: Test private expedited cmd Mathieu Desnoyers
2017-11-08 18:35 ` [RFC PATCH for 4.15 2/6] membarrier: powerpc: Skip memory barrier in switch_mm() (v6) Mathieu Desnoyers
2017-11-08 18:35 ` [RFC PATCH for 4.15 3/6] membarrier: Document scheduler barrier requirements (v5) Mathieu Desnoyers
[not found] ` <20171108183514.3306-1-mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>
2017-11-08 18:35 ` [RFC PATCH for 4.15 4/6] membarrier: Provide core serializing command Mathieu Desnoyers
2017-11-08 18:35 ` [RFC PATCH for 4.15 5/6] membarrier: x86: " Mathieu Desnoyers
2017-11-09 19:07 ` Andy Lutomirski
[not found] ` <CALCETrX4vZHxkWhLKBebZ+R_UE8pDsmVbpJzpyPt=ZgZB8w8iQ-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2017-11-09 19:35 ` Mathieu Desnoyers [this message]
2017-11-10 1:19 ` Andy Lutomirski
2017-11-10 15:43 ` Mathieu Desnoyers
2017-11-08 18:35 ` [RFC PATCH for 4.15 6/6] membarrier: selftest: Test private expedited sync core cmd Mathieu Desnoyers
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=452803516.12441.1510256125252.JavaMail.zimbra@efficios.com \
--to=mathieu.desnoyers-vg+e7yoek/dwk0htik3j/w@public.gmane.org \
--cc=ahh-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org \
--cc=avi-VrcmuVmyx1hWk0Htik3J/w@public.gmane.org \
--cc=benh-XVmvHMARGAS8U2dJNN8I7kB+6BGkLq7r@public.gmane.org \
--cc=boqun.feng-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org \
--cc=davejwatson-b10kYP2dOMg@public.gmane.org \
--cc=gromer-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org \
--cc=hpa-YMNOUZJC4hwAvxtiuMwx3w@public.gmane.org \
--cc=linux-api-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=luto-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org \
--cc=maged.michael-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org \
--cc=mingo-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org \
--cc=mpe-Gsx/Oe8HsFggBc27wqDAHg@public.gmane.org \
--cc=parri.andrea-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org \
--cc=paulmck-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org \
--cc=paulus-eUNUBHrolfbYtjvyW6yDsg@public.gmane.org \
--cc=peterz-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org \
--cc=tglx-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org \
--cc=x86-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).