From: Suresh Siddha <suresh.b.siddha@intel.com>
To: Nick Piggin <npiggin@suse.de>
Cc: Ingo Molnar <mingo@elte.hu>,
Peter Zijlstra <a.p.zijlstra@chello.nl>,
Oleg Nesterov <oleg@redhat.com>,
Jens Axboe <jens.axboe@oracle.com>,
Linus Torvalds <torvalds@linux-foundation.org>,
"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>,
Rusty Russell <rusty@rustcorp.com.au>,
Steven Rostedt <rostedt@goodmis.org>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
"linux-arch@vger.kernel.org" <linux-arch@vger.kernel.org>,
"steiner@sgi.com" <steiner@sgi.com>
Subject: Re: smp.c && barriers (Was: [PATCH 1/4] generic-smp: remove single ipi fallback for smp_call_function_many())
Date: Tue, 17 Mar 2009 10:16:54 -0800 [thread overview]
Message-ID: <1237313814.27006.203.camel@localhost.localdomain> (raw)
In-Reply-To: <20090225033230.GA4611@wotan.suse.de>
On Tue, 2009-02-24 at 19:32 -0800, Nick Piggin wrote:
> Just if I may add something -- I would probably slightly prefer if
> you explicitly used an sfence or other serializing instruction rather
> than smp_mb(). Maybe call it wrmsr_fence() or something.
I was planning to send a cleanup patch doing this, but looks like Ingo
never took the original patch. Here it is appended. Ingo, please
consider.
thanks,
suresh
---
From: Suresh Siddha <suresh.b.siddha@intel.com>
Subject: x86: Add x2apic_wrmsr_fence() to x2apic flush tlb paths
uncached MMIO accesses for xapic are inherently serializing and hence
we don't need explicit barriers for xapic IPI paths.
x2apic MSR writes/reads don't have serializing semantics and hence need
a serializing instruction or mfence, to make all the previous memory
stores globally visisble before the x2apic msr write for IPI.
Add x2apic_wrmsr_fence() in flush tlb path to x2apic specific paths.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
---
diff --git a/arch/x86/include/asm/apic.h b/arch/x86/include/asm/apic.h
index 394d177..cfc0871 100644
--- a/arch/x86/include/asm/apic.h
+++ b/arch/x86/include/asm/apic.h
@@ -108,6 +108,16 @@ extern void native_apic_icr_write(u32 low, u32 id);
extern u64 native_apic_icr_read(void);
#ifdef CONFIG_X86_X2APIC
+/*
+ * Make previous memory operations globally visible before
+ * sending the IPI through x2apic wrmsr. We need a serializing instruction or
+ * mfence for this.
+ */
+static inline void x2apic_wrmsr_fence(void)
+{
+ asm volatile("mfence":::"memory");
+}
+
static inline void native_apic_msr_write(u32 reg, u32 v)
{
if (reg == APIC_DFR || reg == APIC_ID || reg == APIC_LDR ||
diff --git a/arch/x86/kernel/apic/x2apic_cluster.c b/arch/x86/kernel/apic/x2apic_cluster.c
index 8fb87b6..4a903e2 100644
--- a/arch/x86/kernel/apic/x2apic_cluster.c
+++ b/arch/x86/kernel/apic/x2apic_cluster.c
@@ -57,6 +57,8 @@ static void x2apic_send_IPI_mask(const struct cpumask *mask, int vector)
unsigned long query_cpu;
unsigned long flags;
+ x2apic_wrmsr_fence();
+
local_irq_save(flags);
for_each_cpu(query_cpu, mask) {
__x2apic_send_IPI_dest(
@@ -73,6 +75,8 @@ static void
unsigned long query_cpu;
unsigned long flags;
+ x2apic_wrmsr_fence();
+
local_irq_save(flags);
for_each_cpu(query_cpu, mask) {
if (query_cpu == this_cpu)
@@ -90,6 +94,8 @@ static void x2apic_send_IPI_allbutself(int vector)
unsigned long query_cpu;
unsigned long flags;
+ x2apic_wrmsr_fence();
+
local_irq_save(flags);
for_each_online_cpu(query_cpu) {
if (query_cpu == this_cpu)
diff --git a/arch/x86/kernel/apic/x2apic_phys.c b/arch/x86/kernel/apic/x2apic_phys.c
index 23625b9..a284359 100644
--- a/arch/x86/kernel/apic/x2apic_phys.c
+++ b/arch/x86/kernel/apic/x2apic_phys.c
@@ -58,6 +58,8 @@ static void x2apic_send_IPI_mask(const struct cpumask *mask, int vector)
unsigned long query_cpu;
unsigned long flags;
+ x2apic_wrmsr_fence();
+
local_irq_save(flags);
for_each_cpu(query_cpu, mask) {
__x2apic_send_IPI_dest(per_cpu(x86_cpu_to_apicid, query_cpu),
@@ -73,6 +75,8 @@ static void
unsigned long query_cpu;
unsigned long flags;
+ x2apic_wrmsr_fence();
+
local_irq_save(flags);
for_each_cpu(query_cpu, mask) {
if (query_cpu != this_cpu)
@@ -89,6 +93,8 @@ static void x2apic_send_IPI_allbutself(int vector)
unsigned long query_cpu;
unsigned long flags;
+ x2apic_wrmsr_fence();
+
local_irq_save(flags);
for_each_online_cpu(query_cpu) {
if (query_cpu == this_cpu)
diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c
index a654d59..821e970 100644
--- a/arch/x86/mm/tlb.c
+++ b/arch/x86/mm/tlb.c
@@ -187,11 +187,6 @@ static void flush_tlb_others_ipi(const struct cpumask *cpumask,
cpumask, cpumask_of(smp_processor_id()));
/*
- * Make the above memory operations globally visible before
- * sending the IPI.
- */
- smp_mb();
- /*
* We have to send the IPI only to
* CPUs affected.
*/
next prev parent reply other threads:[~2009-03-17 18:17 UTC|newest]
Thread overview: 101+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-02-16 16:38 [PATCH 0/4] generic smp helpers vs kmalloc Peter Zijlstra
2009-02-16 16:38 ` [PATCH 1/4] generic-smp: remove single ipi fallback for smp_call_function_many() Peter Zijlstra
2009-02-16 19:10 ` Oleg Nesterov
2009-02-16 19:41 ` Peter Zijlstra
2009-02-16 20:30 ` Oleg Nesterov
2009-02-16 20:55 ` Peter Zijlstra
2009-02-16 21:22 ` Oleg Nesterov
2009-02-17 12:25 ` Oleg Nesterov
2009-02-16 20:49 ` Q: smp.c && barriers (Was: [PATCH 1/4] generic-smp: remove single ipi fallback for smp_call_function_many()) Oleg Nesterov
2009-02-16 21:03 ` Peter Zijlstra
2009-02-16 21:32 ` Oleg Nesterov
2009-02-16 21:45 ` Peter Zijlstra
2009-02-16 22:02 ` Oleg Nesterov
2009-02-16 22:24 ` Peter Zijlstra
2009-02-16 23:19 ` Oleg Nesterov
2009-02-17 9:29 ` Peter Zijlstra
2009-02-17 10:11 ` Nick Piggin
2009-02-17 10:27 ` Peter Zijlstra
2009-02-17 10:39 ` Nick Piggin
2009-02-17 11:26 ` Nick Piggin
2009-02-17 11:48 ` Peter Zijlstra
2009-02-17 15:51 ` Paul E. McKenney
2009-02-18 2:15 ` Suresh Siddha
2009-02-18 2:40 ` Paul E. McKenney
2009-02-17 19:28 ` Q: " Oleg Nesterov
2009-02-17 21:32 ` Paul E. McKenney
2009-02-17 21:45 ` Oleg Nesterov
2009-02-17 22:39 ` Paul E. McKenney
2009-02-18 13:52 ` Nick Piggin
2009-02-18 16:09 ` Linus Torvalds
2009-02-18 16:21 ` Ingo Molnar
2009-02-18 16:33 ` Linus Torvalds
2009-02-18 16:58 ` Ingo Molnar
2009-02-18 17:05 ` Ingo Molnar
2009-02-18 17:10 ` Ingo Molnar
2009-02-18 17:17 ` Linus Torvalds
2009-02-18 17:23 ` Ingo Molnar
2009-02-18 17:14 ` Linus Torvalds
2009-02-18 17:47 ` Ingo Molnar
2009-02-18 18:33 ` Suresh Siddha
2009-02-18 16:37 ` Gleb Natapov
2009-02-19 0:12 ` Nick Piggin
2009-02-19 6:47 ` Benjamin Herrenschmidt
2009-02-19 13:11 ` Nick Piggin
2009-02-19 15:06 ` Ingo Molnar
2009-02-19 21:49 ` Benjamin Herrenschmidt
2009-02-18 2:21 ` Suresh Siddha
2009-02-18 13:59 ` Nick Piggin
2009-02-18 16:19 ` Linus Torvalds
2009-02-18 16:23 ` Ingo Molnar
2009-02-18 18:43 ` Suresh Siddha
2009-02-18 19:17 ` Ingo Molnar
2009-02-18 23:55 ` Suresh Siddha
2009-02-19 12:20 ` Ingo Molnar
2009-02-19 12:29 ` Nick Piggin
2009-02-19 12:45 ` Ingo Molnar
2009-02-19 22:00 ` Suresh Siddha
2009-02-20 10:56 ` Ingo Molnar
2009-02-20 18:56 ` Suresh Siddha
2009-02-20 19:40 ` Ingo Molnar
2009-02-20 23:28 ` Jack Steiner
2009-02-25 3:32 ` Nick Piggin
2009-02-25 12:47 ` Ingo Molnar
2009-02-25 18:25 ` Luck, Tony
2009-03-17 18:16 ` Suresh Siddha [this message]
2009-03-18 8:51 ` [tip:x86/x2apic] x86: add x2apic_wrmsr_fence() to x2apic flush tlb paths Suresh Siddha
2009-02-17 12:40 ` Q: smp.c && barriers (Was: [PATCH 1/4] generic-smp: remove single ipi fallback for smp_call_function_many()) Peter Zijlstra
2009-02-17 15:43 ` Paul E. McKenney
2009-02-17 15:40 ` [PATCH] generic-smp: remove kmalloc() Peter Zijlstra
2009-02-17 17:21 ` Oleg Nesterov
2009-02-17 17:40 ` Peter Zijlstra
2009-02-17 17:46 ` Peter Zijlstra
2009-02-17 18:30 ` Oleg Nesterov
2009-02-17 19:29 ` [PATCH -v4] generic-ipi: " Peter Zijlstra
2009-02-17 20:02 ` Oleg Nesterov
2009-02-17 20:11 ` Peter Zijlstra
2009-02-17 20:16 ` Peter Zijlstra
2009-02-17 20:44 ` Oleg Nesterov
2009-02-17 20:49 ` Peter Zijlstra
2009-02-17 22:09 ` Oleg Nesterov
2009-02-17 22:15 ` Peter Zijlstra
2009-02-17 21:30 ` Paul E. McKenney
2009-02-17 21:38 ` Peter Zijlstra
2009-02-16 16:38 ` [PATCH 2/4] generic-smp: remove kmalloc usage Peter Zijlstra
2009-02-17 0:40 ` Linus Torvalds
2009-02-17 8:24 ` Peter Zijlstra
2009-02-17 9:43 ` Ingo Molnar
2009-02-17 9:49 ` Peter Zijlstra
2009-02-17 10:56 ` Ingo Molnar
2009-02-18 4:50 ` Rusty Russell
2009-02-18 16:05 ` Ingo Molnar
2009-02-19 0:00 ` Jeremy Fitzhardinge
2009-02-19 12:21 ` Ingo Molnar
2009-02-19 4:31 ` Rusty Russell
2009-02-19 9:10 ` Peter Zijlstra
2009-02-19 11:04 ` Jens Axboe
2009-02-19 16:52 ` Linus Torvalds
2009-02-17 15:44 ` Linus Torvalds
2009-02-16 16:38 ` [PATCH 3/4] generic-smp: properly allocate the cpumasks Peter Zijlstra
2009-02-16 23:17 ` Rusty Russell
2009-02-16 16:38 ` [PATCH 4/4] generic-smp: clean up some of the csd->flags fiddling Peter Zijlstra
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=1237313814.27006.203.camel@localhost.localdomain \
--to=suresh.b.siddha@intel.com \
--cc=a.p.zijlstra@chello.nl \
--cc=jens.axboe@oracle.com \
--cc=linux-arch@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=npiggin@suse.de \
--cc=oleg@redhat.com \
--cc=paulmck@linux.vnet.ibm.com \
--cc=rostedt@goodmis.org \
--cc=rusty@rustcorp.com.au \
--cc=steiner@sgi.com \
--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).