mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Waiman Long <longman@redhat.com>
To: Peter Zijlstra <peterz@infradead.org>,
	Ingo Molnar <mingo@redhat.com>, Will Deacon <will.deacon@arm.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Borislav Petkov <bp@alien8.de>, Arnd Bergmann <arnd@arndb.de>
Cc: linux-kernel@vger.kernel.org, x86@kernel.org,
	linux-arch@vger.kernel.org, Nicholas Piggin <npiggin@gmail.com>,
	Davidlohr Bueso <dave@stgolabs.net>,
	Waiman Long <longman@redhat.com>
Subject: [PATCH v2 1/5] x86/smp: Add saturated +1/+2 1-byte cpu numbers
Date: Thu, 16 Jul 2020 15:29:23 -0400	[thread overview]
Message-ID: <20200716192927.12944-2-longman@redhat.com> (raw)
In-Reply-To: <20200716192927.12944-1-longman@redhat.com>

Both qspinlock and qrwlock use one whole byte to store the binary
lock/unlock state. We can actually store more information in the
lock byte like an encoded lock holder cpu number to aid debugging
and crash dump analysis.

To make that possible, a saturated +1 and +2 1-byte per-cpu cpu numbers
are added. The qrwlock can use the +1 number for the lock holding writer
and the qspinlock can use the +2 number for the lock holder.

The new per-cpu numbers are placed right after the commonly used
cpu_number (smp_processor_id()) which has more 1700 references in the
kernel. Therefore these new cpu numbers are very likely to be located
in the same hot cacheline as cpu_number.  As these numbers are before
the unsigned long this_cpu_off, no additional percpu space will be
consumed in x86-64.

Signed-off-by: Waiman Long <longman@redhat.com>
---
 arch/x86/include/asm/spinlock.h |  5 +++++
 arch/x86/kernel/setup_percpu.c  | 11 +++++++++++
 2 files changed, 16 insertions(+)

diff --git a/arch/x86/include/asm/spinlock.h b/arch/x86/include/asm/spinlock.h
index 5b6bc7016c22..319fa58caa9b 100644
--- a/arch/x86/include/asm/spinlock.h
+++ b/arch/x86/include/asm/spinlock.h
@@ -10,6 +10,11 @@
 #include <asm/paravirt.h>
 #include <asm/bitops.h>
 
+DECLARE_PER_CPU_READ_MOSTLY(u8, cpu_number_sadd1);
+DECLARE_PER_CPU_READ_MOSTLY(u8, cpu_number_sadd2);
+#define __cpu_number_sadd1	this_cpu_read(cpu_number_sadd1)
+#define __cpu_number_sadd2	this_cpu_read(cpu_number_sadd2)
+
 /*
  * Your basic SMP spinlocks, allowing only a single CPU anywhere
  *
diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c
index fd945ce78554..859c5b950d08 100644
--- a/arch/x86/kernel/setup_percpu.c
+++ b/arch/x86/kernel/setup_percpu.c
@@ -26,6 +26,14 @@
 DEFINE_PER_CPU_READ_MOSTLY(int, cpu_number);
 EXPORT_PER_CPU_SYMBOL(cpu_number);
 
+/*
+ * Saturated +1 and +2 1-byte cpu numbers
+ */
+DEFINE_PER_CPU_READ_MOSTLY(u8, cpu_number_sadd1); /* +1 saturated cpu# */
+DEFINE_PER_CPU_READ_MOSTLY(u8, cpu_number_sadd2); /* +2 saturated cpu# */
+EXPORT_PER_CPU_SYMBOL(cpu_number_sadd1);
+EXPORT_PER_CPU_SYMBOL(cpu_number_sadd2);
+
 #ifdef CONFIG_X86_64
 #define BOOT_PERCPU_OFFSET ((unsigned long)__per_cpu_load)
 #else
@@ -223,6 +231,9 @@ void __init setup_per_cpu_areas(void)
 		per_cpu_offset(cpu) = delta + pcpu_unit_offsets[cpu];
 		per_cpu(this_cpu_off, cpu) = per_cpu_offset(cpu);
 		per_cpu(cpu_number, cpu) = cpu;
+		per_cpu(cpu_number_sadd1, cpu) = (cpu + 1 < 0x100) ? cpu + 1 : 0xff;
+		per_cpu(cpu_number_sadd2, cpu) = (cpu + 2 < 0x100) ? cpu + 2 : 0xff;
+
 		setup_percpu_segment(cpu);
 		setup_stack_canary_segment(cpu);
 		/*
-- 
2.18.1


  reply	other threads:[~2020-07-16 19:31 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-07-16 19:29 [PATCH v2 0/5] x86, locking/qspinlock: Allow lock to store lock holder cpu number Waiman Long
2020-07-16 19:29 ` Waiman Long [this message]
2020-07-16 19:29 ` [PATCH v2 2/5] locking/pvqspinlock: Make pvqsinlock code easier to read Waiman Long
2020-07-16 19:29 ` [PATCH v2 3/5] locking/qspinlock: Pass lock value as function argument Waiman Long
2020-07-16 19:29 ` [PATCH v2 4/5] locking/qspinlock: Make qspinhlock store lock holder cpu number Waiman Long
2020-07-17  7:39   ` [locking/qspinlock] 45877ea393: BUG:spinlock_already_unlocked_on_CPU kernel test robot
2020-07-19 23:48     ` Waiman Long
2020-07-16 19:29 ` [PATCH v2 5/5] locking/qrwlock: Make qrwlock store writer cpu number Waiman Long

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=20200716192927.12944-2-longman@redhat.com \
    --to=longman@redhat.com \
    --cc=arnd@arndb.de \
    --cc=bp@alien8.de \
    --cc=dave@stgolabs.net \
    --cc=linux-arch@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@redhat.com \
    --cc=npiggin@gmail.com \
    --cc=peterz@infradead.org \
    --cc=tglx@linutronix.de \
    --cc=will.deacon@arm.com \
    --cc=x86@kernel.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

Powered by JetHome