diff mbox series

[V10,18/19] locking/qspinlock: Move pv_ops into x86 directory

Message ID 20230802164701.192791-19-guoren@kernel.org (mailing list archive)
State Handled Elsewhere
Headers show
Series riscv: Add Native/Paravirt/CNA qspinlock support | expand

Checks

Context Check Description
conchuod/tree_selection fail Failed to apply to next/pending-fixes, riscv/for-next or riscv/master

Commit Message

Guo Ren Aug. 2, 2023, 4:47 p.m. UTC 
From: Guo Ren <guoren@linux.alibaba.com>

The pv_ops belongs to x86 custom infrastructure and cleans up the
cna_configure_spin_lock_slowpath() with standard code. This is
preparation for riscv support CNA qspoinlock.

Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@kernel.org>
---
 arch/x86/include/asm/qspinlock.h |  3 ++-
 arch/x86/kernel/alternative.c    |  6 +++++-
 kernel/locking/qspinlock_cna.h   | 14 ++++++--------
 3 files changed, 13 insertions(+), 10 deletions(-)

Comments

Waiman Long Aug. 11, 2023, 8:42 p.m. UTC | #1 
On 8/2/23 12:47, guoren@kernel.org wrote:
> From: Guo Ren <guoren@linux.alibaba.com>
>
> The pv_ops belongs to x86 custom infrastructure and cleans up the
> cna_configure_spin_lock_slowpath() with standard code. This is
> preparation for riscv support CNA qspoinlock.

CNA qspinlock has not been merged into mainline yet. I will suggest you 
drop the last 2 patches for now. Of course, you can provide benchmark 
data to boost the case for the inclusion of the CNA qspinlock patch into 
the mainline.

Cheers,
Longman
Guo Ren Aug. 12, 2023, 12:24 a.m. UTC | #2 
On Sat, Aug 12, 2023 at 4:42 AM Waiman Long <longman@redhat.com> wrote:
>
> On 8/2/23 12:47, guoren@kernel.org wrote:
> > From: Guo Ren <guoren@linux.alibaba.com>
> >
> > The pv_ops belongs to x86 custom infrastructure and cleans up the
> > cna_configure_spin_lock_slowpath() with standard code. This is
> > preparation for riscv support CNA qspoinlock.
>
> CNA qspinlock has not been merged into mainline yet. I will suggest you
> drop the last 2 patches for now. Of course, you can provide benchmark
> data to boost the case for the inclusion of the CNA qspinlock patch into
> the mainline.
Yes, my lazy, I would separate paravirt and CNA from this series.

>
> Cheers,
> Longman
>
Waiman Long Aug. 12, 2023, 12:47 a.m. UTC | #3 
On 8/11/23 20:24, Guo Ren wrote:
> On Sat, Aug 12, 2023 at 4:42 AM Waiman Long <longman@redhat.com> wrote:
>> On 8/2/23 12:47, guoren@kernel.org wrote:
>>> From: Guo Ren <guoren@linux.alibaba.com>
>>>
>>> The pv_ops belongs to x86 custom infrastructure and cleans up the
>>> cna_configure_spin_lock_slowpath() with standard code. This is
>>> preparation for riscv support CNA qspoinlock.
>> CNA qspinlock has not been merged into mainline yet. I will suggest you
>> drop the last 2 patches for now. Of course, you can provide benchmark
>> data to boost the case for the inclusion of the CNA qspinlock patch into
>> the mainline.
> Yes, my lazy, I would separate paravirt and CNA from this series.

paravirt is OK, it is just that CNA hasn't been merged yet.

Cheers,
Longman

>
>> Cheers,
>> Longman
>>
>
diff mbox series

Patch

diff --git a/arch/x86/include/asm/qspinlock.h b/arch/x86/include/asm/qspinlock.h
index f48a2a250e57..100adad70bf5 100644
--- a/arch/x86/include/asm/qspinlock.h
+++ b/arch/x86/include/asm/qspinlock.h
@@ -28,7 +28,8 @@  static __always_inline u32 queued_fetch_set_pending_acquire(struct qspinlock *lo
 }
 
 #ifdef CONFIG_NUMA_AWARE_SPINLOCKS
-extern void cna_configure_spin_lock_slowpath(void);
+extern bool cna_configure_spin_lock_slowpath(void);
+extern void __cna_queued_spin_lock_slowpath(struct qspinlock *lock, u32 val);
 #endif
 
 #ifdef CONFIG_PARAVIRT_SPINLOCKS
diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index c36df5aa3ab1..68b7392016c3 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -1538,7 +1538,11 @@  void __init alternative_instructions(void)
 	paravirt_set_cap();
 
 #if defined(CONFIG_NUMA_AWARE_SPINLOCKS)
-	cna_configure_spin_lock_slowpath();
+	if (pv_ops.lock.queued_spin_lock_slowpath == native_queued_spin_lock_slowpath) {
+		if (cna_configure_spin_lock_slowpath())
+			pv_ops.lock.queued_spin_lock_slowpath =
+							__cna_queued_spin_lock_slowpath;
+	}
 #endif
 
 	/*
diff --git a/kernel/locking/qspinlock_cna.h b/kernel/locking/qspinlock_cna.h
index 17d56c739e57..5e297dc687d9 100644
--- a/kernel/locking/qspinlock_cna.h
+++ b/kernel/locking/qspinlock_cna.h
@@ -406,20 +406,18 @@  void __cna_queued_spin_lock_slowpath(struct qspinlock *lock, u32 val);
  * multiple NUMA nodes in native environment, unless the user has
  * overridden this default behavior by setting the numa_spinlock flag.
  */
-void __init cna_configure_spin_lock_slowpath(void)
+bool __init cna_configure_spin_lock_slowpath(void)
 {
 
 	if (numa_spinlock_flag < 0)
-		return;
+		return false;
 
-	if (numa_spinlock_flag == 0 && (nr_node_ids < 2 ||
-		    pv_ops.lock.queued_spin_lock_slowpath !=
-			native_queued_spin_lock_slowpath))
-		return;
+	if (numa_spinlock_flag == 0 && nr_node_ids < 2)
+		return false;
 
 	cna_init_nodes();
 
-	pv_ops.lock.queued_spin_lock_slowpath = __cna_queued_spin_lock_slowpath;
-
 	pr_info("Enabling CNA spinlock\n");
+
+	return true;
 }