Re: [RFC PATCH v2 4/4] arch/x86: L1D flush, optimize the context switch

[Singh, Balbir]

On Wed, 2020-03-25 at 18:11 +1100, Balbir Singh wrote:
> Use a static branch/jump label to optimize the code. Right now
> we don't ref count the users, but that could be done if needed
> in the future.
> 
> Signed-off-by: Balbir Singh <sblbir@xxxxxxxxxx>
> 

I sent an older version of the patch, here is the updated version

Balbir Singh


From fa80e9691202183a2879a013e497f221b22305a6 Mon Sep 17 00:00:00 2001
From: Balbir Singh <sblbir@xxxxxxxxxx>
Date: Wed, 25 Mar 2020 16:41:18 +1100
Subject: [RFC PATCH v2 4/4] arch/x86: L1D flush, optimize the context switch

Use a static branch/jump label to optimize the code. Right now
we don't ref count the users, but that could be done if needed
in the future.

Signed-off-by: Balbir Singh <sblbir@xxxxxxxxxx>
---
 arch/x86/include/asm/nospec-branch.h |  2 ++
 arch/x86/mm/tlb.c                    | 52 +++++++++++++++++-----------
 2 files changed, 34 insertions(+), 20 deletions(-)

diff --git a/arch/x86/include/asm/nospec-branch.h b/arch/x86/include/asm/nospec-branch.h
index 07e95dcb40ad..371e28cea1f4 100644
--- a/arch/x86/include/asm/nospec-branch.h
+++ b/arch/x86/include/asm/nospec-branch.h
@@ -310,6 +310,8 @@ DECLARE_STATIC_KEY_FALSE(switch_mm_always_ibpb);
 DECLARE_STATIC_KEY_FALSE(mds_user_clear);
 DECLARE_STATIC_KEY_FALSE(mds_idle_clear);
 
+DECLARE_STATIC_KEY_FALSE(switch_mm_l1d_flush);
+
 #include <asm/segment.h>
 
 /**
diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c
index 22f96c5f74f0..8f272e5921ce 100644
--- a/arch/x86/mm/tlb.c
+++ b/arch/x86/mm/tlb.c
@@ -155,6 +155,11 @@ EXPORT_SYMBOL_GPL(leave_mm);
 static void *l1d_flush_pages;
 static DEFINE_MUTEX(l1d_flush_mutex);
 
+/* Flush L1D on switch_mm() */
+DEFINE_STATIC_KEY_FALSE(switch_mm_l1d_flush);
+EXPORT_SYMBOL_GPL(switch_mm_l1d_flush);
+
+
 int enable_l1d_flush_for_task(struct task_struct *tsk)
 {
 	struct page *page;
@@ -170,6 +175,11 @@ int enable_l1d_flush_for_task(struct task_struct *tsk)
 			l1d_flush_pages = alloc_l1d_flush_pages();
 			if (!l1d_flush_pages)
 				return -ENOMEM;
+			/*
+			 * We could do more accurate ref counting
+			 * if needed
+			 */
+			static_branch_enable(&switch_mm_l1d_flush);
 		}
 		mutex_unlock(&l1d_flush_mutex);
 	}
@@ -195,29 +205,31 @@ static void l1d_flush(struct mm_struct *next, struct task_struct *tsk)
 {
 	struct mm_struct *real_prev = this_cpu_read(cpu_tlbstate.loaded_mm);
 
-	/*
-	 * If we are not really switching mm's, we can just return
-	 */
-	if (real_prev == next)
-		return;
+	if (static_branch_unlikely(&switch_mm_l1d_flush)) {
+		/*
+		 * If we are not really switching mm's, we can just return
+		 */
+		if (real_prev == next)
+			return;
 
-	/*
-	 * Do we need flushing for by the previous task
-	 */
-	if (this_cpu_read(cpu_tlbstate.last_user_mm_l1d_flush) != 0) {
-		if (!flush_l1d_cache_hw())
-			flush_l1d_cache_sw(l1d_flush_pages);
-		this_cpu_write(cpu_tlbstate.last_user_mm_l1d_flush, 0);
-		/* Make sure we clear the values before we set it again */
-		barrier();
-	}
+		/*
+		 * Do we need flushing for by the previous task
+		 */
+		if (this_cpu_read(cpu_tlbstate.last_user_mm_l1d_flush) != 0) {
+			if (!flush_l1d_cache_hw())
+				flush_l1d_cache_sw(l1d_flush_pages);
+			this_cpu_write(cpu_tlbstate.last_user_mm_l1d_flush, 0);
+			/* Make sure we clear the values before we set it again */
+			barrier();
+		}
 
-	if (tsk == NULL)
-		return;
+		if (tsk == NULL)
+			return;
 
-	/* We don't need stringent checks as we opt-in/opt-out */
-	if (test_ti_thread_flag(&tsk->thread_info, TIF_L1D_FLUSH))
-		this_cpu_write(cpu_tlbstate.last_user_mm_l1d_flush, 1);
+		/* We don't need stringent checks as we opt-in/opt-out */
+		if (test_ti_thread_flag(&tsk->thread_info, TIF_L1D_FLUSH))
+			this_cpu_write(cpu_tlbstate.last_user_mm_l1d_flush, 1);
+	}
 }
 
 void switch_mm(struct mm_struct *prev, struct mm_struct *next,
-- 
2.17.1