From: Peter Zijlstra <peterz@infradead.org>
To: torvalds@linux-foundation.org, linux-kernel@vger.kernel.org,
tglx@linutronix.de, mingo@kernel.org, pbonzini@redhat.com,
dvyukov@google.com, hpa@zytor.com
Cc: linux-tip-commits@vger.kernel.org, Steven Rostedt <rostedt@goodmis.org>
Subject: Re: [tip:locking/urgent] locking/static_key: Fix concurrent static_key_slow_inc()
Date: Mon, 31 Jul 2017 11:36:12 +0200 [thread overview]
Message-ID: <20170731093612.7v23kxkk47i56io6@hirez.programming.kicks-ass.net> (raw)
In-Reply-To: <tip-4c5ea0a9cd02d6aa8adc86e100b2a4cff8d614ff@git.kernel.org>
On Fri, Jun 24, 2016 at 01:59:06AM -0700, tip-bot for Paolo Bonzini wrote:
> +++ b/kernel/jump_label.c
> @@ -58,13 +58,36 @@ static void jump_label_update(struct static_key *key);
>
> void static_key_slow_inc(struct static_key *key)
> {
> + int v, v1;
> +
> STATIC_KEY_CHECK_USE();
> - if (atomic_inc_not_zero(&key->enabled))
> - return;
> +
> + /*
> + * Careful if we get concurrent static_key_slow_inc() calls;
> + * later calls must wait for the first one to _finish_ the
> + * jump_label_update() process. At the same time, however,
> + * the jump_label_update() call below wants to see
> + * static_key_enabled(&key) for jumps to be updated properly.
> + *
> + * So give a special meaning to negative key->enabled: it sends
> + * static_key_slow_inc() down the slow path, and it is non-zero
> + * so it counts as "enabled" in jump_label_update(). Note that
> + * atomic_inc_unless_negative() checks >= 0, so roll our own.
> + */
> + for (v = atomic_read(&key->enabled); v > 0; v = v1) {
> + v1 = atomic_cmpxchg(&key->enabled, v, v + 1);
> + if (likely(v1 == v))
> + return;
> + }
>
> jump_label_lock();
> - if (atomic_inc_return(&key->enabled) == 1)
> + if (atomic_read(&key->enabled) == 0) {
> + atomic_set(&key->enabled, -1);
> jump_label_update(key);
> + atomic_set(&key->enabled, 1);
> + } else {
> + atomic_inc(&key->enabled);
> + }
> jump_label_unlock();
> }
So I was recently looking at this again and got all paranoid. Do we want
something like so?
---
kernel/jump_label.c | 24 +++++++++++++++++-------
1 file changed, 17 insertions(+), 7 deletions(-)
diff --git a/kernel/jump_label.c b/kernel/jump_label.c
index d11c506a6ac3..69d07e78e48b 100644
--- a/kernel/jump_label.c
+++ b/kernel/jump_label.c
@@ -103,7 +103,7 @@ EXPORT_SYMBOL_GPL(static_key_disable);
void static_key_slow_inc(struct static_key *key)
{
- int v, v1;
+ int v;
STATIC_KEY_CHECK_USE();
@@ -119,18 +119,28 @@ void static_key_slow_inc(struct static_key *key)
* so it counts as "enabled" in jump_label_update(). Note that
* atomic_inc_unless_negative() checks >= 0, so roll our own.
*/
- for (v = atomic_read(&key->enabled); v > 0; v = v1) {
- v1 = atomic_cmpxchg(&key->enabled, v, v + 1);
- if (likely(v1 == v))
+ for (v = atomic_read(&key->enabled); v > 0;) {
+ if (atomic_try_cmpxchg(&key->enabled, &v, v+1))
return;
}
cpus_read_lock();
jump_label_lock();
- if (atomic_read(&key->enabled) == 0) {
- atomic_set(&key->enabled, -1);
+
+ if (atomic_try_cmpxchg(&key->enabled, 0, -1)) {
+ /*
+ * smp_mb implied, must have -1 before proceeding to change
+ * text.
+ */
jump_label_update(key);
- atomic_set(&key->enabled, 1);
+
+ /*
+ * smp_mb, such that we finish modifying text before enabling
+ * the fast path. Probably not needed because modifying text is
+ * likely to serialize everything. Be paranoid.
+ */
+ smp_mb__before_atomic();
+ atomic_add(2, &key->enabled); /* -1 -> 1 */
} else {
atomic_inc(&key->enabled);
}
next prev parent reply other threads:[~2017-07-31 9:36 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-06-21 16:52 [PATCH] static_key: fix concurrent static_key_slow_inc Paolo Bonzini
2016-06-21 19:20 ` Peter Zijlstra
2016-06-22 8:50 ` Christian Borntraeger
2016-06-22 9:52 ` Paolo Bonzini
2016-06-24 8:59 ` [tip:locking/urgent] locking/static_key: Fix concurrent static_key_slow_inc() tip-bot for Paolo Bonzini
2017-07-31 9:36 ` Peter Zijlstra [this message]
2017-07-31 13:04 ` Paolo Bonzini
2017-07-31 13:33 ` Peter Zijlstra
2017-07-31 15:35 ` Paolo Bonzini
2017-07-31 15:45 ` Peter Zijlstra
2017-07-31 19:21 ` Paolo Bonzini
2017-07-31 17:15 ` Dima Zavin
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=20170731093612.7v23kxkk47i56io6@hirez.programming.kicks-ass.net \
--to=peterz@infradead.org \
--cc=dvyukov@google.com \
--cc=hpa@zytor.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-tip-commits@vger.kernel.org \
--cc=mingo@kernel.org \
--cc=pbonzini@redhat.com \
--cc=rostedt@goodmis.org \
--cc=tglx@linutronix.de \
--cc=torvalds@linux-foundation.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).