[PATCH 0/4] x86: Add and use support for clflushopt

[PATCH 0/4] x86: Add and use support for clflushopt

[Ross Zwisler]

This patch series adds support for the new clflushopt instruction and then
uses it in a few appropriate places in x86 specific code.  It does this using
the alternatives mechanism, so any platforms without support for clflushopt
will continue to use clflush instead.

clflushopt was announced in the document "Intel Architecture Instruction Set
Extensions Programming Reference" with Ref # 319433-018.

http://download-software.intel.com/sites/default/files/managed/50/1a/319433-018.pdf

clflushopt has the same flushing behavior as clflush, but has more relaxed
ordering.  clflushopt must be explicitly ordered by sfence or mfence.

The inline assembly for clflushopt was implemented using %P so that the
generated addresses will always be absolute instead of sometimes being RIP
relative.  This is necessary for the alternatives code to behave correctly.

Ross Zwisler (4):
  x86: Add support for clflushopt
  x86: Use clflushopt in clflush_cache_range
  x86: Use clflushopt in drm_clflush_page
  x86: Use clflushopt in drm_clflush_virt_range

 arch/x86/include/asm/cpufeature.h    |    1 +
 arch/x86/include/asm/special_insns.h |    8 ++++++++
 arch/x86/mm/pageattr.c               |    8 ++++----
 drivers/gpu/drm/drm_cache.c          |   10 ++++++++--
 4 files changed, 21 insertions(+), 6 deletions(-)

-- 
1.7.10.4

[PATCH 1/4] x86: Add support for clflushopt

[Ross Zwisler]

Add support for the new clflushopt instruction.  This instruction was
announced in the document "Intel Architecture Instruction Set Extensions
Programming Reference" with Ref # 319433-018.

http://download-software.intel.com/sites/default/files/managed/50/1a/319433-018.pdf

Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: H Peter Anvin <h.peter.anvin@intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
---
 arch/x86/include/asm/cpufeature.h    |    1 +
 arch/x86/include/asm/special_insns.h |    8 ++++++++
 2 files changed, 9 insertions(+)

diff --git a/arch/x86/include/asm/cpufeature.h b/arch/x86/include/asm/cpufeature.h
index 89270b4..bfad1ad 100644
--- a/arch/x86/include/asm/cpufeature.h
+++ b/arch/x86/include/asm/cpufeature.h
@@ -219,6 +219,7 @@
 #define X86_FEATURE_RDSEED	(9*32+18) /* The RDSEED instruction */
 #define X86_FEATURE_ADX		(9*32+19) /* The ADCX and ADOX instructions */
 #define X86_FEATURE_SMAP	(9*32+20) /* Supervisor Mode Access Prevention */
+#define X86_FEATURE_CLFLSHOPT   (9*32+23) /* "clflushopt" CLFLUSHOPT instruction */
 
 /*
  * BUG word(s)
diff --git a/arch/x86/include/asm/special_insns.h b/arch/x86/include/asm/special_insns.h
index 645cad2..617389a 100644
--- a/arch/x86/include/asm/special_insns.h
+++ b/arch/x86/include/asm/special_insns.h
@@ -191,6 +191,14 @@ static inline void clflush(volatile void *__p)
 	asm volatile("clflush %0" : "+m" (*(volatile char __force *)__p));
 }
 
+static inline void clflushopt(volatile void *__p)
+{
+	alternative_io(".byte " __stringify(NOP_DS_PREFIX) "; clflush %P0",
+		       ".byte 0x66; clflush %P0",
+		       X86_FEATURE_CLFLSHOPT,
+		       "+m" (*(volatile char __force *)__p));
+}
+
 #define nop() asm volatile ("nop")
 
 
-- 
1.7.10.4

[tip:x86/cpufeature] x86: Add support for the clflushopt instruction

[tip-bot for Ross Zwisler]

Commit-ID:  171699f7630c92d0a928f83e5fb3aeabe35398c0
Gitweb:     http://git.kernel.org/tip/171699f7630c92d0a928f83e5fb3aeabe35398c0
Author:     Ross Zwisler <ross.zwisler@linux.intel.com>
AuthorDate: Wed, 26 Feb 2014 12:06:49 -0700
Committer:  H. Peter Anvin <hpa@linux.intel.com>
CommitDate: Thu, 27 Feb 2014 08:23:28 -0800

x86: Add support for the clflushopt instruction

Add support for the new clflushopt instruction.  This instruction was
announced in the document "Intel Architecture Instruction Set Extensions
Programming Reference" with Ref # 319433-018.

http://download-software.intel.com/sites/default/files/managed/50/1a/319433-018.pdf

[ hpa: changed the feature flag to simply X86_FEATURE_CLFLUSHOPT - if
  that is what we want to report in /proc/cpuinfo anyway... ]

Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Link: http://lkml.kernel.org/r/1393441612-19729-2-git-send-email-ross.zwisler@linux.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
---
 arch/x86/include/asm/cpufeature.h    | 1 +
 arch/x86/include/asm/special_insns.h | 8 ++++++++
 2 files changed, 9 insertions(+)

diff --git a/arch/x86/include/asm/cpufeature.h b/arch/x86/include/asm/cpufeature.h
index 5f12968..bc507d7 100644
--- a/arch/x86/include/asm/cpufeature.h
+++ b/arch/x86/include/asm/cpufeature.h
@@ -221,6 +221,7 @@
 #define X86_FEATURE_RDSEED	(9*32+18) /* The RDSEED instruction */
 #define X86_FEATURE_ADX		(9*32+19) /* The ADCX and ADOX instructions */
 #define X86_FEATURE_SMAP	(9*32+20) /* Supervisor Mode Access Prevention */
+#define X86_FEATURE_CLFLUSHOPT	(9*32+23) /* CLFLUSHOPT instruction */
 #define X86_FEATURE_AVX512PF	(9*32+26) /* AVX-512 Prefetch */
 #define X86_FEATURE_AVX512ER	(9*32+27) /* AVX-512 Exponential and Reciprocal */
 #define X86_FEATURE_AVX512CD	(9*32+28) /* AVX-512 Conflict Detection */
diff --git a/arch/x86/include/asm/special_insns.h b/arch/x86/include/asm/special_insns.h
index 645cad2..e820c08 100644
--- a/arch/x86/include/asm/special_insns.h
+++ b/arch/x86/include/asm/special_insns.h
@@ -191,6 +191,14 @@ static inline void clflush(volatile void *__p)
 	asm volatile("clflush %0" : "+m" (*(volatile char __force *)__p));
 }
 
+static inline void clflushopt(volatile void *__p)
+{
+	alternative_io(".byte " __stringify(NOP_DS_PREFIX) "; clflush %P0",
+		       ".byte 0x66; clflush %P0",
+		       X86_FEATURE_CLFLUSHOPT,
+		       "+m" (*(volatile char __force *)__p));
+}
+
 #define nop() asm volatile ("nop")
 
 

[PATCH 2/4] x86: Use clflushopt in clflush_cache_range

[Ross Zwisler]

If clflushopt is available on the system, use it instead of clflush in
clflush_cache_range.

Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: H Peter Anvin <h.peter.anvin@intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
---
 arch/x86/mm/pageattr.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c
index bb32480..11d500a 100644
--- a/arch/x86/mm/pageattr.c
+++ b/arch/x86/mm/pageattr.c
@@ -125,8 +125,8 @@ within(unsigned long addr, unsigned long start, unsigned long end)
  * @vaddr:	virtual start address
  * @size:	number of bytes to flush
  *
- * clflush is an unordered instruction which needs fencing with mfence
- * to avoid ordering issues.
+ * clflushopt is an unordered instruction which needs fencing with mfence or
+ * sfence to avoid ordering issues.
  */
 void clflush_cache_range(void *vaddr, unsigned int size)
 {
@@ -135,11 +135,11 @@ void clflush_cache_range(void *vaddr, unsigned int size)
 	mb();
 
 	for (; vaddr < vend; vaddr += boot_cpu_data.x86_clflush_size)
-		clflush(vaddr);
+		clflushopt(vaddr);
 	/*
 	 * Flush any possible final partial cacheline:
 	 */
-	clflush(vend);
+	clflushopt(vend);
 
 	mb();
 }
-- 
1.7.10.4

[tip:x86/cpu] x86: Use clflushopt in clflush_cache_range

[tip-bot for Ross Zwisler]

Commit-ID:  8b80fd8b45458a7d0b726d454c51c5219e0fb2ee
Gitweb:     http://git.kernel.org/tip/8b80fd8b45458a7d0b726d454c51c5219e0fb2ee
Author:     Ross Zwisler <ross.zwisler@linux.intel.com>
AuthorDate: Wed, 26 Feb 2014 12:06:50 -0700
Committer:  H. Peter Anvin <hpa@linux.intel.com>
CommitDate: Thu, 27 Feb 2014 08:26:10 -0800

x86: Use clflushopt in clflush_cache_range

If clflushopt is available on the system, use it instead of clflush in
clflush_cache_range.

Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Link: http://lkml.kernel.org/r/1393441612-19729-3-git-send-email-ross.zwisler@linux.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
---
 arch/x86/mm/pageattr.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c
index b3b19f4..b1c4672 100644
--- a/arch/x86/mm/pageattr.c
+++ b/arch/x86/mm/pageattr.c
@@ -126,8 +126,8 @@ within(unsigned long addr, unsigned long start, unsigned long end)
  * @vaddr:	virtual start address
  * @size:	number of bytes to flush
  *
- * clflush is an unordered instruction which needs fencing with mfence
- * to avoid ordering issues.
+ * clflushopt is an unordered instruction which needs fencing with mfence or
+ * sfence to avoid ordering issues.
  */
 void clflush_cache_range(void *vaddr, unsigned int size)
 {
@@ -136,11 +136,11 @@ void clflush_cache_range(void *vaddr, unsigned int size)
 	mb();
 
 	for (; vaddr < vend; vaddr += boot_cpu_data.x86_clflush_size)
-		clflush(vaddr);
+		clflushopt(vaddr);
 	/*
 	 * Flush any possible final partial cacheline:
 	 */
-	clflush(vend);
+	clflushopt(vend);
 
 	mb();
 }

[PATCH 3/4] x86: Use clflushopt in drm_clflush_page

[Ross Zwisler]

If clflushopt is available on the system, use it instead of clflush in
drm_clflush_page.

Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: H Peter Anvin <h.peter.anvin@intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
---
 drivers/gpu/drm/drm_cache.c |    8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/drm_cache.c b/drivers/gpu/drm/drm_cache.c
index bb8f580..c518fb6 100644
--- a/drivers/gpu/drm/drm_cache.c
+++ b/drivers/gpu/drm/drm_cache.c
@@ -32,6 +32,12 @@
 #include <drm/drmP.h>
 
 #if defined(CONFIG_X86)
+
+/*
+ * clflushopt is an unordered instruction which needs fencing with mfence or
+ * sfence to avoid ordering issues.  For drm_clflush_page this fencing happens
+ * in the caller.
+ */
 static void
 drm_clflush_page(struct page *page)
 {
@@ -44,7 +50,7 @@ drm_clflush_page(struct page *page)
 
 	page_virtual = kmap_atomic(page);
 	for (i = 0; i < PAGE_SIZE; i += size)
-		clflush(page_virtual + i);
+		clflushopt(page_virtual + i);
 	kunmap_atomic(page_virtual);
 }
 
-- 
1.7.10.4

[tip:x86/cpu] x86: Use clflushopt in drm_clflush_page

[tip-bot for Ross Zwisler]

Commit-ID:  2a0c772f1967565d5b71a98012e74d5a7bddc1c4
Gitweb:     http://git.kernel.org/tip/2a0c772f1967565d5b71a98012e74d5a7bddc1c4
Author:     Ross Zwisler <ross.zwisler@linux.intel.com>
AuthorDate: Wed, 26 Feb 2014 12:06:51 -0700
Committer:  H. Peter Anvin <hpa@linux.intel.com>
CommitDate: Thu, 27 Feb 2014 08:26:20 -0800

x86: Use clflushopt in drm_clflush_page

If clflushopt is available on the system, use it instead of clflush in
drm_clflush_page.

Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Link: http://lkml.kernel.org/r/1393441612-19729-4-git-send-email-ross.zwisler@linux.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
---
 drivers/gpu/drm/drm_cache.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/drm_cache.c b/drivers/gpu/drm/drm_cache.c
index bb8f580..c518fb6 100644
--- a/drivers/gpu/drm/drm_cache.c
+++ b/drivers/gpu/drm/drm_cache.c
@@ -32,6 +32,12 @@
 #include <drm/drmP.h>
 
 #if defined(CONFIG_X86)
+
+/*
+ * clflushopt is an unordered instruction which needs fencing with mfence or
+ * sfence to avoid ordering issues.  For drm_clflush_page this fencing happens
+ * in the caller.
+ */
 static void
 drm_clflush_page(struct page *page)
 {
@@ -44,7 +50,7 @@ drm_clflush_page(struct page *page)
 
 	page_virtual = kmap_atomic(page);
 	for (i = 0; i < PAGE_SIZE; i += size)
-		clflush(page_virtual + i);
+		clflushopt(page_virtual + i);
 	kunmap_atomic(page_virtual);
 }
 

[PATCH 4/4] x86: Use clflushopt in drm_clflush_virt_range

[Ross Zwisler]

If clflushopt is available on the system, use it instead of clflush in
drm_clflush_virt_range.

Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: H Peter Anvin <h.peter.anvin@intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
---
 drivers/gpu/drm/drm_cache.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/drm_cache.c b/drivers/gpu/drm/drm_cache.c
index c518fb6..534cb89 100644
--- a/drivers/gpu/drm/drm_cache.c
+++ b/drivers/gpu/drm/drm_cache.c
@@ -139,7 +139,7 @@ drm_clflush_virt_range(char *addr, unsigned long length)
 		mb();
 		for (; addr < end; addr += boot_cpu_data.x86_clflush_size)
 			clflush(addr);
-		clflush(end - 1);
+		clflushopt(end - 1);
 		mb();
 		return;
 	}
-- 
1.7.10.4

[tip:x86/cpu] x86: Use clflushopt in drm_clflush_virt_range

[tip-bot for Ross Zwisler]

Commit-ID:  2a0788dc9bc46789ec98aea0f30c6fb420196b12
Gitweb:     http://git.kernel.org/tip/2a0788dc9bc46789ec98aea0f30c6fb420196b12
Author:     Ross Zwisler <ross.zwisler@linux.intel.com>
AuthorDate: Wed, 26 Feb 2014 12:06:52 -0700
Committer:  H. Peter Anvin <hpa@linux.intel.com>
CommitDate: Thu, 27 Feb 2014 08:26:31 -0800

x86: Use clflushopt in drm_clflush_virt_range

If clflushopt is available on the system, use it instead of clflush in
drm_clflush_virt_range.

Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Link: http://lkml.kernel.org/r/1393441612-19729-5-git-send-email-ross.zwisler@linux.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
---
 drivers/gpu/drm/drm_cache.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/drm_cache.c b/drivers/gpu/drm/drm_cache.c
index c518fb6..534cb89 100644
--- a/drivers/gpu/drm/drm_cache.c
+++ b/drivers/gpu/drm/drm_cache.c
@@ -139,7 +139,7 @@ drm_clflush_virt_range(char *addr, unsigned long length)
 		mb();
 		for (; addr < end; addr += boot_cpu_data.x86_clflush_size)
 			clflush(addr);
-		clflush(end - 1);
+		clflushopt(end - 1);
 		mb();
 		return;
 	}