From: Thomas Gleixner <tglx@linutronix.de>
To: Darren Hart <dvhltc@us.ibm.com>
Cc: linux-kernel@vger.kernel.org,
Peter Zijlstra <peterz@infradead.org>,
Ingo Molnar <mingo@elte.hu>,
Eric Dumazet <eric.dumazet@gmail.com>,
"Peter W. Morreale" <pmorreale@novell.com>,
Rik van Riel <riel@redhat.com>,
Steven Rostedt <rostedt@goodmis.org>,
Gregory Haskins <ghaskins@novell.com>,
Sven-Thorsten Dietrich <sdietrich@novell.com>,
Chris Mason <chris.mason@oracle.com>,
John Cooper <john.cooper@third-harmonic.com>,
Chris Wright <chrisw@sous-sol.org>,
Ulrich Drepper <drepper@gmail.com>,
Alan Cox <alan@lxorguk.ukuu.org.uk>, Avi Kivity <avi@redhat.com>,
Peter Zijlstra <a.p.zijlstra@chello.nl>
Subject: Re: [PATCH 4/4] futex: Add FUTEX_LOCK with optional adaptive spinning
Date: Fri, 7 May 2010 18:20:41 +0200 (CEST) [thread overview]
Message-ID: <alpine.LFD.2.00.1005071818380.3401@localhost.localdomain> (raw)
In-Reply-To: <1273127060-30375-5-git-send-email-dvhltc@us.ibm.com>
On Wed, 5 May 2010, Darren Hart wrote:
> Add a non-pi TID value based futex locking mechanism. This enables the
> use of adaptive spinning which was problematic with the basic FUTEX_WAIT
> operation.
You still do way too much work in that spin code with way too much
code lines.
Can you try the following (completely uncompiled/untested) patch ?
Thanks,
tglx
Subject: futex-simplify.patch
From: Thomas Gleixner <tglx@linutronix.de>
Date: Fri, 07 May 2010 17:56:38 +0200
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
---
kernel/futex.c | 42 ++++++++++++------------------------
kernel/sched.c | 66 ---------------------------------------------------------
2 files changed, 14 insertions(+), 94 deletions(-)
Index: linux-2.6-tip/kernel/futex.c
===================================================================
--- linux-2.6-tip.orig/kernel/futex.c
+++ linux-2.6-tip/kernel/futex.c
@@ -2406,11 +2406,10 @@ out:
* 1 - Futex lock acquired
* <0 - On error
*/
-static int trylock_futex_adaptive(u32 __user *uaddr, ktime_t *timeout)
+static int trylock_futex_adaptive(u32 __user *uaddr, struct hrtimer_sleeper *to)
{
struct thread_info *owner = NULL;
pid_t curtid, ownertid = 0;
- ktime_t now;
int ret = 0;
u32 curval;
@@ -2425,43 +2424,32 @@ static int trylock_futex_adaptive(u32 __
/*
* Try to acquire the lock.
*/
- if (curval == 0)
+ if (curval == 0) {
curval = cmpxchg_futex_value_locked(uaddr, 0, curtid);
- if (curval == 0) {
- ret = 1;
- break;
- }
+ if (curval == 0) {
+ ret = 1;
+ break;
+ }
- if (unlikely(curval == -EFAULT)) {
- ret = -EFAULT;
- break;
+ if (unlikely(curval == -EFAULT)) {
+ ret = -EFAULT;
+ break;
+ }
}
- /*
- * Futex locks manage the owner atomically. We are not in
- * danger of deadlock by preempting a pending owner. Abort the
- * loop if our time slice has expired. RT Threads can spin
- * until they preempt the owner, at which point they will break
- * out of the loop anyway.
- */
- if (need_resched())
+ if (to && !to->task) {
+ ret = -ETIMEOUT;
break;
-
- if (timeout) {
- now = ktime_get();
- if (timeout->tv64 < now.tv64)
- break;
}
- cpu_relax();
-
if ((curval & FUTEX_TID_MASK) != ownertid) {
if (owner)
put_task_struct(owner->task);
owner = futex_owner(uaddr);
}
- if (owner && !futex_spin_on_owner(uaddr, owner))
+
+ if (!owner->oncpu)
break;
}
@@ -2510,9 +2498,7 @@ static int futex_lock(u32 __user *uaddr,
retry:
#ifdef CONFIG_SMP
if (flags & FLAGS_ADAPTIVE) {
- preempt_disable();
ret = trylock_futex_adaptive(uaddr, time);
- preempt_enable();
/* We got the lock. */
if (ret == 1) {
Index: linux-2.6-tip/kernel/sched.c
===================================================================
--- linux-2.6-tip.orig/kernel/sched.c
+++ linux-2.6-tip/kernel/sched.c
@@ -3821,72 +3821,6 @@ out:
}
#endif
-#ifdef CONFIG_FUTEX
-#include <linux/futex.h>
-
-int futex_spin_on_owner(u32 __user *uaddr, struct thread_info *owner)
-{
- u32 ownertid, uval;
- unsigned int cpu;
- struct rq *rq;
-
- if (!sched_feat(OWNER_SPIN))
- return 0;
-
-#ifdef CONFIG_DEBUG_PAGEALLOC
- /*
- * Need to access the cpu field knowing that
- * DEBUG_PAGEALLOC could have unmapped it if
- * the mutex owner just released it and exited.
- */
- if (probe_kernel_address(&owner->cpu, cpu))
- goto out;
-#else
- cpu = owner->cpu;
-#endif
-
- /*
- * Even if the access succeeded (likely case),
- * the cpu field may no longer be valid.
- */
- if (cpu >= nr_cpumask_bits)
- goto out;
-
- /*
- * We need to validate that we can do a
- * get_cpu() and that we have the percpu area.
- */
- if (!cpu_online(cpu))
- goto out;
-
- rq = cpu_rq(cpu);
-
- if (get_user(ownertid, uaddr))
- return -EFAULT;
- ownertid &= FUTEX_TID_MASK;
-
- for (;;) {
- /*
- * Owner changed, break to re-assess state.
- */
- if (get_user(uval, uaddr))
- return -EFAULT;
- if ((uval & FUTEX_TID_MASK) != ownertid)
- break;
-
- /*
- * Is that owner really running on that cpu?
- */
- if (task_thread_info(rq->curr) != owner || need_resched())
- return 0;
-
- cpu_relax();
- }
-out:
- return 1;
-}
-#endif
-
#ifdef CONFIG_PREEMPT
/*
* this is the entry point to schedule() from in-kernel preemption
next prev parent reply other threads:[~2010-05-07 16:22 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-05-06 6:24 [PATCH V6 0/4][RFC] futex: " Darren Hart
2010-05-06 6:24 ` [PATCH 1/4] futex: replace fshared and clockrt with combined flags Darren Hart
2010-05-06 6:24 ` [PATCH 2/4] futex: add futex_q static initializer Darren Hart
2010-05-06 6:24 ` [PATCH 3/4] futex: refactor futex_lock_pi_atomic Darren Hart
2010-05-06 6:24 ` [PATCH 4/4] futex: Add FUTEX_LOCK with optional adaptive spinning Darren Hart
2010-05-07 16:20 ` Thomas Gleixner [this message]
2010-05-07 16:24 ` Peter Zijlstra
2010-05-07 16:30 ` Thomas Gleixner
2010-05-07 16:35 ` Peter Zijlstra
2010-05-07 16:43 ` Thomas Gleixner
2010-05-07 19:05 ` Darren Hart
2010-05-07 16:52 ` Darren Hart
2010-05-07 19:11 ` Darren Hart
-- strict thread matches above, loose matches on Subject: below --
2010-04-09 5:15 [PATCH V5 0/4][RFC] futex: " dvhltc
2010-04-09 5:15 ` [PATCH 4/4] futex: Add " dvhltc
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=alpine.LFD.2.00.1005071818380.3401@localhost.localdomain \
--to=tglx@linutronix.de \
--cc=a.p.zijlstra@chello.nl \
--cc=alan@lxorguk.ukuu.org.uk \
--cc=avi@redhat.com \
--cc=chris.mason@oracle.com \
--cc=chrisw@sous-sol.org \
--cc=drepper@gmail.com \
--cc=dvhltc@us.ibm.com \
--cc=eric.dumazet@gmail.com \
--cc=ghaskins@novell.com \
--cc=john.cooper@third-harmonic.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=peterz@infradead.org \
--cc=pmorreale@novell.com \
--cc=riel@redhat.com \
--cc=rostedt@goodmis.org \
--cc=sdietrich@novell.com \
/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
all inboxes | Powered by JetHome®