[PATCH 0/3] Provide mechanism to disable select RDT features & use for Skylake

[PATCH 0/3] Provide mechanism to disable select RDT features & use for Skylake

[Luck, Tony]

From: Tony Luck <tony.luck@intel.com>

Some of the RDT features are affected by errata on Skylake. Make sure
unsuspecting users do not accidentally use these.

To do this we add a CPU model quirk mechanism that disables the
features affected by the errata by default. But also add a command
line mechanism for users to turn them back on again for cases
where the use case still works despite the issue.

First patch establishes the CPU quirk mechanism and uses it for
the existing Haswell probe code.

Second sets up the command line arguments.

Last adds the code for Skylake model to disable features
affected by errata.

Tony Luck (3):
  x86/intel_rdt: Move special case code for Haswell to a quirk function
  x86/intel_rdt: Add command line options for resource director
    technology
  x86/intel_rdt: Turn off most RDT features on Skylake

 Documentation/admin-guide/kernel-parameters.rst |   1 +
 Documentation/admin-guide/kernel-parameters.txt |   6 +
 arch/x86/kernel/cpu/intel_rdt.c                 | 147 +++++++++++++++++++-----
 3 files changed, 123 insertions(+), 31 deletions(-)

-- 
2.11.0

[PATCH 1/3] x86/intel_rdt: Move special case code for Haswell to a quirk function

[Luck, Tony]

From: Tony Luck <tony.luck@intel.com>

No functional change, but lay the ground work for other per-model
quirks.

Signed-off-by: Tony Luck <tony.luck@intel.com>
---
 arch/x86/kernel/cpu/intel_rdt.c | 52 ++++++++++++++++++++++-------------------
 1 file changed, 28 insertions(+), 24 deletions(-)

diff --git a/arch/x86/kernel/cpu/intel_rdt.c b/arch/x86/kernel/cpu/intel_rdt.c
index 6935c8ecad7f..25514cd454b3 100644
--- a/arch/x86/kernel/cpu/intel_rdt.c
+++ b/arch/x86/kernel/cpu/intel_rdt.c
@@ -172,34 +172,28 @@ static unsigned int cbm_idx(struct rdt_resource *r, unsigned int closid)
  * is always 20 on hsw server parts. The minimum cache bitmask length
  * allowed for HSW server is always 2 bits. Hardcode all of them.
  */
-static inline bool cache_alloc_hsw_probe(void)
+static inline void cache_alloc_hsw_probe(void)
 {
-	if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL &&
-	    boot_cpu_data.x86 == 6 &&
-	    boot_cpu_data.x86_model == INTEL_FAM6_HASWELL_X) {
-		struct rdt_resource *r  = &rdt_resources_all[RDT_RESOURCE_L3];
-		u32 l, h, max_cbm = BIT_MASK(20) - 1;
+	struct rdt_resource *r  = &rdt_resources_all[RDT_RESOURCE_L3];
+	u32 l, h, max_cbm = BIT_MASK(20) - 1;
 
-		if (wrmsr_safe(IA32_L3_CBM_BASE, max_cbm, 0))
-			return false;
-		rdmsr(IA32_L3_CBM_BASE, l, h);
-
-		/* If all the bits were set in MSR, return success */
-		if (l != max_cbm)
-			return false;
+	if (wrmsr_safe(IA32_L3_CBM_BASE, max_cbm, 0))
+		return;
+	rdmsr(IA32_L3_CBM_BASE, l, h);
 
-		r->num_closid = 4;
-		r->default_ctrl = max_cbm;
-		r->cache.cbm_len = 20;
-		r->cache.shareable_bits = 0xc0000;
-		r->cache.min_cbm_bits = 2;
-		r->alloc_capable = true;
-		r->alloc_enabled = true;
+	/* If all the bits were set in MSR, return success */
+	if (l != max_cbm)
+		return;
 
-		return true;
-	}
+	r->num_closid = 4;
+	r->default_ctrl = max_cbm;
+	r->cache.cbm_len = 20;
+	r->cache.shareable_bits = 0xc0000;
+	r->cache.min_cbm_bits = 2;
+	r->alloc_capable = true;
+	r->alloc_enabled = true;
 
-	return false;
+	rdt_alloc_capable = true;
 }
 
 /*
@@ -647,7 +641,7 @@ static __init bool get_rdt_alloc_resources(void)
 {
 	bool ret = false;
 
-	if (cache_alloc_hsw_probe())
+	if (rdt_alloc_capable)
 		return true;
 
 	if (!boot_cpu_has(X86_FEATURE_RDT_A))
@@ -689,8 +683,18 @@ static __init bool get_rdt_mon_resources(void)
 	return !rdt_get_mon_l3_config(&rdt_resources_all[RDT_RESOURCE_L3]);
 }
 
+static __init void rdt_quirks(void)
+{
+	switch (boot_cpu_data.x86_model) {
+	case INTEL_FAM6_HASWELL_X:
+		cache_alloc_hsw_probe();
+		break;
+	}
+}
+
 static __init bool get_rdt_resources(void)
 {
+	rdt_quirks();
 	rdt_alloc_capable = get_rdt_alloc_resources();
 	rdt_mon_capable = get_rdt_mon_resources();
 
-- 
2.11.0

[PATCH 2/3] x86/intel_rdt: Add command line options for resource director technology

[Luck, Tony]

From: Tony Luck <tony.luck@intel.com>

Command line options allow us to ignore features that we don't want.
Also we can re-enable options that have been disabled on a platform
(so long as the underlying h/w actually supports the option).

Signed-off-by: Tony Luck <tony.luck@intel.com>
---
 Documentation/admin-guide/kernel-parameters.rst |  1 +
 Documentation/admin-guide/kernel-parameters.txt |  6 ++
 arch/x86/kernel/cpu/intel_rdt.c                 | 94 ++++++++++++++++++++++---
 3 files changed, 93 insertions(+), 8 deletions(-)

diff --git a/Documentation/admin-guide/kernel-parameters.rst b/Documentation/admin-guide/kernel-parameters.rst
index d76ab3907e2b..b2598cc9834c 100644
--- a/Documentation/admin-guide/kernel-parameters.rst
+++ b/Documentation/admin-guide/kernel-parameters.rst
@@ -138,6 +138,7 @@ parameter is applicable::
 	PPT	Parallel port support is enabled.
 	PS2	Appropriate PS/2 support is enabled.
 	RAM	RAM disk support is enabled.
+	RDT	Intel Resource Director Technology.
 	S390	S390 architecture is enabled.
 	SCSI	Appropriate SCSI support is enabled.
 			A lot of drivers have their options described inside
diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index d9c171ce4190..ef52ae4ed6e8 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -3598,6 +3598,12 @@
 			Run specified binary instead of /init from the ramdisk,
 			used for early userspace startup. See initrd.
 
+	rdt=		[HW,X86,RDT]
+			Turn on/off individual RDT features. List is:
+			cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, mba.
+			E.g. to turn on cmt and turn off mba use:
+				rdt=cmt,!mba
+
 	reboot=		[KNL]
 			Format (x86 or x86_64):
 				[w[arm] | c[old] | h[ard] | s[oft] | g[pio]] \
diff --git a/arch/x86/kernel/cpu/intel_rdt.c b/arch/x86/kernel/cpu/intel_rdt.c
index 25514cd454b3..65222d8fd5ec 100644
--- a/arch/x86/kernel/cpu/intel_rdt.c
+++ b/arch/x86/kernel/cpu/intel_rdt.c
@@ -637,6 +637,83 @@ static __init void rdt_init_padding(void)
 	}
 }
 
+enum {
+	RDT_FLAG_CMT,
+	RDT_FLAG_MBM_TOTAL,
+	RDT_FLAG_MBM_LOCAL,
+	RDT_FLAG_L3_CAT,
+	RDT_FLAG_L3_CDP,
+	RDT_FLAG_L2_CAT,
+	RDT_FLAG_MBA,
+};
+
+#define RDT_OPT(idx, n, f)	\
+[idx] = {			\
+	.name = n,		\
+	.flag = f		\
+}
+
+static struct rdt_options {
+	char	*name;
+	int	flag;
+	bool	force_off, force_on;
+} rdt_options[] = {
+	RDT_OPT(RDT_FLAG_CMT, "cmt", X86_FEATURE_CQM_OCCUP_LLC),
+	RDT_OPT(RDT_FLAG_MBM_TOTAL, "mbmtotal", X86_FEATURE_CQM_MBM_TOTAL),
+	RDT_OPT(RDT_FLAG_MBM_LOCAL, "mbmlocal", X86_FEATURE_CQM_MBM_LOCAL),
+	RDT_OPT(RDT_FLAG_L3_CAT, "l3cat", X86_FEATURE_CAT_L3),
+	RDT_OPT(RDT_FLAG_L3_CDP, "l3cdp", X86_FEATURE_CDP_L3),
+	RDT_OPT(RDT_FLAG_L2_CAT, "l2cat", X86_FEATURE_CAT_L2),
+	RDT_OPT(RDT_FLAG_MBA, "mba", X86_FEATURE_MBA),
+};
+#define NUM_RDT_OPTIONS ARRAY_SIZE(rdt_options)
+
+static int __init set_rdt_options(char *str)
+{
+	struct rdt_options *o;
+	char *tok;
+	bool force_off;
+
+	if (*str == '=')
+		str++;
+	while ((tok = strsep(&str, ",")) != NULL) {
+		force_off = *tok == '!';
+		if (force_off)
+			tok++;
+		for (o = rdt_options; o < &rdt_options[NUM_RDT_OPTIONS]; o++) {
+			if (strcmp(tok, o->name) == 0) {
+				if (force_off)
+					o->force_off = true;
+				else
+					o->force_on = true;
+				break;
+			}
+		}
+	}
+	return 1;
+}
+__setup("rdt", set_rdt_options);
+
+static int rdt_cpu_has(int flag)
+{
+	int ret = boot_cpu_has(flag);
+	struct rdt_options *o;
+
+	if (!ret)
+		return ret;
+
+	for (o = rdt_options; o < &rdt_options[NUM_RDT_OPTIONS]; o++) {
+		if (flag == o->flag) {
+			if (o->force_off)
+				ret = 0;
+			if (o->force_on)
+				ret = 1;
+			break;
+		}
+	}
+	return ret;
+}
+
 static __init bool get_rdt_alloc_resources(void)
 {
 	bool ret = false;
@@ -647,21 +724,21 @@ static __init bool get_rdt_alloc_resources(void)
 	if (!boot_cpu_has(X86_FEATURE_RDT_A))
 		return false;
 
-	if (boot_cpu_has(X86_FEATURE_CAT_L3)) {
+	if (rdt_cpu_has(X86_FEATURE_CAT_L3)) {
 		rdt_get_cache_alloc_cfg(1, &rdt_resources_all[RDT_RESOURCE_L3]);
-		if (boot_cpu_has(X86_FEATURE_CDP_L3)) {
+		if (rdt_cpu_has(X86_FEATURE_CDP_L3)) {
 			rdt_get_cdp_l3_config(RDT_RESOURCE_L3DATA);
 			rdt_get_cdp_l3_config(RDT_RESOURCE_L3CODE);
 		}
 		ret = true;
 	}
-	if (boot_cpu_has(X86_FEATURE_CAT_L2)) {
+	if (rdt_cpu_has(X86_FEATURE_CAT_L2)) {
 		/* CPUID 0x10.2 fields are same format at 0x10.1 */
 		rdt_get_cache_alloc_cfg(2, &rdt_resources_all[RDT_RESOURCE_L2]);
 		ret = true;
 	}
 
-	if (boot_cpu_has(X86_FEATURE_MBA)) {
+	if (rdt_cpu_has(X86_FEATURE_MBA)) {
 		if (rdt_get_mem_config(&rdt_resources_all[RDT_RESOURCE_MBA]))
 			ret = true;
 	}
@@ -670,11 +747,11 @@ static __init bool get_rdt_alloc_resources(void)
 
 static __init bool get_rdt_mon_resources(void)
 {
-	if (boot_cpu_has(X86_FEATURE_CQM_OCCUP_LLC))
+	if (rdt_cpu_has(X86_FEATURE_CQM_OCCUP_LLC))
 		rdt_mon_features |= (1 << QOS_L3_OCCUP_EVENT_ID);
-	if (boot_cpu_has(X86_FEATURE_CQM_MBM_TOTAL))
+	if (rdt_cpu_has(X86_FEATURE_CQM_MBM_TOTAL))
 		rdt_mon_features |= (1 << QOS_L3_MBM_TOTAL_EVENT_ID);
-	if (boot_cpu_has(X86_FEATURE_CQM_MBM_LOCAL))
+	if (rdt_cpu_has(X86_FEATURE_CQM_MBM_LOCAL))
 		rdt_mon_features |= (1 << QOS_L3_MBM_LOCAL_EVENT_ID);
 
 	if (!rdt_mon_features)
@@ -687,7 +764,8 @@ static __init void rdt_quirks(void)
 {
 	switch (boot_cpu_data.x86_model) {
 	case INTEL_FAM6_HASWELL_X:
-		cache_alloc_hsw_probe();
+		if (!rdt_options[RDT_FLAG_L3_CAT].force_off)
+			cache_alloc_hsw_probe();
 		break;
 	}
 }
-- 
2.11.0

[PATCH 3/3] x86/intel_rdt: Turn off most RDT features on Skylake

[Luck, Tony]

From: Tony Luck <tony.luck@intel.com>

Errata list is included in this document:
https://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/6th-gen-x-series-spec-update.pdf
with more details in:
https://www.intel.com/content/www/us/en/processors/xeon/scalable/xeon-scalable-spec-update.html

But the tl;dr summary (using tags from first of the documents) is:
SKZ4  MBM does not accurately track write bandwidth
SKZ17 CMT counters may not count accurately
SKZ18 CAT may not restrict cacheline allocation under certain conditions
SKZ19 MBM counters may undercount

Disable all these features on Skylake models. Users who understand the
errata may re-enable using boot command line options.

Signed-off-by: Tony Luck <tony.luck@intel.com>
---
 arch/x86/kernel/cpu/intel_rdt.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/arch/x86/kernel/cpu/intel_rdt.c b/arch/x86/kernel/cpu/intel_rdt.c
index 65222d8fd5ec..a55b74ca0973 100644
--- a/arch/x86/kernel/cpu/intel_rdt.c
+++ b/arch/x86/kernel/cpu/intel_rdt.c
@@ -767,6 +767,9 @@ static __init void rdt_quirks(void)
 		if (!rdt_options[RDT_FLAG_L3_CAT].force_off)
 			cache_alloc_hsw_probe();
 		break;
+	case INTEL_FAM6_SKYLAKE_X:
+		if (boot_cpu_data.x86_mask <= 4)
+			set_rdt_options("!cmt,!mbmtotal,!mbmlocal,!l3cat");
 	}
 }
 
-- 
2.11.0

[tip:x86/cache] x86/intel_rdt: Move special case code for Haswell to a quirk function

[tip-bot for Tony Luck]

Commit-ID:  0576113a387e0c8a5d9e24b4cd62605d1c9c0db8
Gitweb:     http://git.kernel.org/tip/0576113a387e0c8a5d9e24b4cd62605d1c9c0db8
Author:     Tony Luck <tony.luck@intel.com>
AuthorDate: Thu, 24 Aug 2017 09:26:50 -0700
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Fri, 25 Aug 2017 22:00:44 +0200

x86/intel_rdt: Move special case code for Haswell to a quirk function

No functional change, but lay the ground work for other per-model
quirks.

Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Fenghua" <fenghua.yu@intel.com>
Cc: Ravi V" <ravi.v.shankar@intel.com>
Cc: "Peter Zijlstra" <peterz@infradead.org>
Cc: "Stephane Eranian" <eranian@google.com>
Cc: "Andi Kleen" <ak@linux.intel.com>
Cc: "David Carrillo-Cisneros" <davidcc@google.com>
Cc: Vikas Shivappa <vikas.shivappa@linux.intel.com>
Link: http://lkml.kernel.org/r/f195a83751b5f8b1d8a78bd3c1914300c8fa3142.1503512900.git.tony.luck@intel.com

---
 arch/x86/kernel/cpu/intel_rdt.c | 52 ++++++++++++++++++++++-------------------
 1 file changed, 28 insertions(+), 24 deletions(-)

diff --git a/arch/x86/kernel/cpu/intel_rdt.c b/arch/x86/kernel/cpu/intel_rdt.c
index 6935c8e..25514cd 100644
--- a/arch/x86/kernel/cpu/intel_rdt.c
+++ b/arch/x86/kernel/cpu/intel_rdt.c
@@ -172,34 +172,28 @@ static unsigned int cbm_idx(struct rdt_resource *r, unsigned int closid)
  * is always 20 on hsw server parts. The minimum cache bitmask length
  * allowed for HSW server is always 2 bits. Hardcode all of them.
  */
-static inline bool cache_alloc_hsw_probe(void)
+static inline void cache_alloc_hsw_probe(void)
 {
-	if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL &&
-	    boot_cpu_data.x86 == 6 &&
-	    boot_cpu_data.x86_model == INTEL_FAM6_HASWELL_X) {
-		struct rdt_resource *r  = &rdt_resources_all[RDT_RESOURCE_L3];
-		u32 l, h, max_cbm = BIT_MASK(20) - 1;
+	struct rdt_resource *r  = &rdt_resources_all[RDT_RESOURCE_L3];
+	u32 l, h, max_cbm = BIT_MASK(20) - 1;
 
-		if (wrmsr_safe(IA32_L3_CBM_BASE, max_cbm, 0))
-			return false;
-		rdmsr(IA32_L3_CBM_BASE, l, h);
-
-		/* If all the bits were set in MSR, return success */
-		if (l != max_cbm)
-			return false;
+	if (wrmsr_safe(IA32_L3_CBM_BASE, max_cbm, 0))
+		return;
+	rdmsr(IA32_L3_CBM_BASE, l, h);
 
-		r->num_closid = 4;
-		r->default_ctrl = max_cbm;
-		r->cache.cbm_len = 20;
-		r->cache.shareable_bits = 0xc0000;
-		r->cache.min_cbm_bits = 2;
-		r->alloc_capable = true;
-		r->alloc_enabled = true;
+	/* If all the bits were set in MSR, return success */
+	if (l != max_cbm)
+		return;
 
-		return true;
-	}
+	r->num_closid = 4;
+	r->default_ctrl = max_cbm;
+	r->cache.cbm_len = 20;
+	r->cache.shareable_bits = 0xc0000;
+	r->cache.min_cbm_bits = 2;
+	r->alloc_capable = true;
+	r->alloc_enabled = true;
 
-	return false;
+	rdt_alloc_capable = true;
 }
 
 /*
@@ -647,7 +641,7 @@ static __init bool get_rdt_alloc_resources(void)
 {
 	bool ret = false;
 
-	if (cache_alloc_hsw_probe())
+	if (rdt_alloc_capable)
 		return true;
 
 	if (!boot_cpu_has(X86_FEATURE_RDT_A))
@@ -689,8 +683,18 @@ static __init bool get_rdt_mon_resources(void)
 	return !rdt_get_mon_l3_config(&rdt_resources_all[RDT_RESOURCE_L3]);
 }
 
+static __init void rdt_quirks(void)
+{
+	switch (boot_cpu_data.x86_model) {
+	case INTEL_FAM6_HASWELL_X:
+		cache_alloc_hsw_probe();
+		break;
+	}
+}
+
 static __init bool get_rdt_resources(void)
 {
+	rdt_quirks();
 	rdt_alloc_capable = get_rdt_alloc_resources();
 	rdt_mon_capable = get_rdt_mon_resources();
 

[tip:x86/cache] x86/intel_rdt: Add command line options for resource director technology

[tip-bot for Tony Luck]

Commit-ID:  1d9807fc64c131a83a96917f2b2da1c9b00cf127
Gitweb:     http://git.kernel.org/tip/1d9807fc64c131a83a96917f2b2da1c9b00cf127
Author:     Tony Luck <tony.luck@intel.com>
AuthorDate: Thu, 24 Aug 2017 09:26:51 -0700
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Fri, 25 Aug 2017 22:00:45 +0200

x86/intel_rdt: Add command line options for resource director technology

Command line options allow us to ignore features that we don't want.
Also we can re-enable options that have been disabled on a platform
(so long as the underlying h/w actually supports the option).

[ tglx: Marked the option array __initdata and the helper function __init ]

Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Fenghua" <fenghua.yu@intel.com>
Cc: Ravi V" <ravi.v.shankar@intel.com>
Cc: "Peter Zijlstra" <peterz@infradead.org>
Cc: "Stephane Eranian" <eranian@google.com>
Cc: "Andi Kleen" <ak@linux.intel.com>
Cc: "David Carrillo-Cisneros" <davidcc@google.com>
Cc: Vikas Shivappa <vikas.shivappa@linux.intel.com>
Link: http://lkml.kernel.org/r/0c37b0d4dbc30977a3c1cee08b66420f83662694.1503512900.git.tony.luck@intel.com

---
 Documentation/admin-guide/kernel-parameters.rst |  1 +
 Documentation/admin-guide/kernel-parameters.txt |  6 ++
 arch/x86/kernel/cpu/intel_rdt.c                 | 96 ++++++++++++++++++++++---
 3 files changed, 95 insertions(+), 8 deletions(-)

diff --git a/Documentation/admin-guide/kernel-parameters.rst b/Documentation/admin-guide/kernel-parameters.rst
index d76ab39..b2598cc 100644
--- a/Documentation/admin-guide/kernel-parameters.rst
+++ b/Documentation/admin-guide/kernel-parameters.rst
@@ -138,6 +138,7 @@ parameter is applicable::
 	PPT	Parallel port support is enabled.
 	PS2	Appropriate PS/2 support is enabled.
 	RAM	RAM disk support is enabled.
+	RDT	Intel Resource Director Technology.
 	S390	S390 architecture is enabled.
 	SCSI	Appropriate SCSI support is enabled.
 			A lot of drivers have their options described inside
diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index d9c171c..ef52ae4 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -3598,6 +3598,12 @@
 			Run specified binary instead of /init from the ramdisk,
 			used for early userspace startup. See initrd.
 
+	rdt=		[HW,X86,RDT]
+			Turn on/off individual RDT features. List is:
+			cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, mba.
+			E.g. to turn on cmt and turn off mba use:
+				rdt=cmt,!mba
+
 	reboot=		[KNL]
 			Format (x86 or x86_64):
 				[w[arm] | c[old] | h[ard] | s[oft] | g[pio]] \
diff --git a/arch/x86/kernel/cpu/intel_rdt.c b/arch/x86/kernel/cpu/intel_rdt.c
index 25514cd..b641622 100644
--- a/arch/x86/kernel/cpu/intel_rdt.c
+++ b/arch/x86/kernel/cpu/intel_rdt.c
@@ -637,6 +637,85 @@ static __init void rdt_init_padding(void)
 	}
 }
 
+enum {
+	RDT_FLAG_CMT,
+	RDT_FLAG_MBM_TOTAL,
+	RDT_FLAG_MBM_LOCAL,
+	RDT_FLAG_L3_CAT,
+	RDT_FLAG_L3_CDP,
+	RDT_FLAG_L2_CAT,
+	RDT_FLAG_MBA,
+};
+
+#define RDT_OPT(idx, n, f)	\
+[idx] = {			\
+	.name = n,		\
+	.flag = f		\
+}
+
+struct rdt_options {
+	char	*name;
+	int	flag;
+	bool	force_off, force_on;
+};
+
+static struct rdt_options rdt_options[]  __initdata = {
+	RDT_OPT(RDT_FLAG_CMT,	    "cmt",	X86_FEATURE_CQM_OCCUP_LLC),
+	RDT_OPT(RDT_FLAG_MBM_TOTAL, "mbmtotal", X86_FEATURE_CQM_MBM_TOTAL),
+	RDT_OPT(RDT_FLAG_MBM_LOCAL, "mbmlocal", X86_FEATURE_CQM_MBM_LOCAL),
+	RDT_OPT(RDT_FLAG_L3_CAT,    "l3cat",	X86_FEATURE_CAT_L3),
+	RDT_OPT(RDT_FLAG_L3_CDP,    "l3cdp",	X86_FEATURE_CDP_L3),
+	RDT_OPT(RDT_FLAG_L2_CAT,    "l2cat",	X86_FEATURE_CAT_L2),
+	RDT_OPT(RDT_FLAG_MBA,	    "mba",	X86_FEATURE_MBA),
+};
+#define NUM_RDT_OPTIONS ARRAY_SIZE(rdt_options)
+
+static int __init set_rdt_options(char *str)
+{
+	struct rdt_options *o;
+	bool force_off;
+	char *tok;
+
+	if (*str == '=')
+		str++;
+	while ((tok = strsep(&str, ",")) != NULL) {
+		force_off = *tok == '!';
+		if (force_off)
+			tok++;
+		for (o = rdt_options; o < &rdt_options[NUM_RDT_OPTIONS]; o++) {
+			if (strcmp(tok, o->name) == 0) {
+				if (force_off)
+					o->force_off = true;
+				else
+					o->force_on = true;
+				break;
+			}
+		}
+	}
+	return 1;
+}
+__setup("rdt", set_rdt_options);
+
+static bool __init rdt_cpu_has(int flag)
+{
+	bool ret = boot_cpu_has(flag);
+	struct rdt_options *o;
+
+	if (!ret)
+		return ret;
+
+	for (o = rdt_options; o < &rdt_options[NUM_RDT_OPTIONS]; o++) {
+		if (flag == o->flag) {
+			if (o->force_off)
+				ret = false;
+			if (o->force_on)
+				ret = true;
+			break;
+		}
+	}
+	return ret;
+}
+
 static __init bool get_rdt_alloc_resources(void)
 {
 	bool ret = false;
@@ -647,21 +726,21 @@ static __init bool get_rdt_alloc_resources(void)
 	if (!boot_cpu_has(X86_FEATURE_RDT_A))
 		return false;
 
-	if (boot_cpu_has(X86_FEATURE_CAT_L3)) {
+	if (rdt_cpu_has(X86_FEATURE_CAT_L3)) {
 		rdt_get_cache_alloc_cfg(1, &rdt_resources_all[RDT_RESOURCE_L3]);
-		if (boot_cpu_has(X86_FEATURE_CDP_L3)) {
+		if (rdt_cpu_has(X86_FEATURE_CDP_L3)) {
 			rdt_get_cdp_l3_config(RDT_RESOURCE_L3DATA);
 			rdt_get_cdp_l3_config(RDT_RESOURCE_L3CODE);
 		}
 		ret = true;
 	}
-	if (boot_cpu_has(X86_FEATURE_CAT_L2)) {
+	if (rdt_cpu_has(X86_FEATURE_CAT_L2)) {
 		/* CPUID 0x10.2 fields are same format at 0x10.1 */
 		rdt_get_cache_alloc_cfg(2, &rdt_resources_all[RDT_RESOURCE_L2]);
 		ret = true;
 	}
 
-	if (boot_cpu_has(X86_FEATURE_MBA)) {
+	if (rdt_cpu_has(X86_FEATURE_MBA)) {
 		if (rdt_get_mem_config(&rdt_resources_all[RDT_RESOURCE_MBA]))
 			ret = true;
 	}
@@ -670,11 +749,11 @@ static __init bool get_rdt_alloc_resources(void)
 
 static __init bool get_rdt_mon_resources(void)
 {
-	if (boot_cpu_has(X86_FEATURE_CQM_OCCUP_LLC))
+	if (rdt_cpu_has(X86_FEATURE_CQM_OCCUP_LLC))
 		rdt_mon_features |= (1 << QOS_L3_OCCUP_EVENT_ID);
-	if (boot_cpu_has(X86_FEATURE_CQM_MBM_TOTAL))
+	if (rdt_cpu_has(X86_FEATURE_CQM_MBM_TOTAL))
 		rdt_mon_features |= (1 << QOS_L3_MBM_TOTAL_EVENT_ID);
-	if (boot_cpu_has(X86_FEATURE_CQM_MBM_LOCAL))
+	if (rdt_cpu_has(X86_FEATURE_CQM_MBM_LOCAL))
 		rdt_mon_features |= (1 << QOS_L3_MBM_LOCAL_EVENT_ID);
 
 	if (!rdt_mon_features)
@@ -687,7 +766,8 @@ static __init void rdt_quirks(void)
 {
 	switch (boot_cpu_data.x86_model) {
 	case INTEL_FAM6_HASWELL_X:
-		cache_alloc_hsw_probe();
+		if (!rdt_options[RDT_FLAG_L3_CAT].force_off)
+			cache_alloc_hsw_probe();
 		break;
 	}
 }

[tip:x86/cache] x86/intel_rdt: Turn off most RDT features on Skylake

[tip-bot for Tony Luck]

Commit-ID:  d56593eb5eda8f593db92927059697bbf89bc4b3
Gitweb:     http://git.kernel.org/tip/d56593eb5eda8f593db92927059697bbf89bc4b3
Author:     Tony Luck <tony.luck@intel.com>
AuthorDate: Thu, 24 Aug 2017 09:26:52 -0700
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Fri, 25 Aug 2017 22:00:45 +0200

x86/intel_rdt: Turn off most RDT features on Skylake

Errata list is included in this document:
https://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/6th-gen-x-series-spec-update.pdf
with more details in:
https://www.intel.com/content/www/us/en/processors/xeon/scalable/xeon-scalable-spec-update.html

But the tl;dr summary (using tags from first of the documents) is:
SKZ4  MBM does not accurately track write bandwidth
SKZ17 CMT counters may not count accurately
SKZ18 CAT may not restrict cacheline allocation under certain conditions
SKZ19 MBM counters may undercount

Disable all these features on Skylake models. Users who understand the
errata may re-enable using boot command line options.

Signed-off-by: Tony Luck <tony.luck@intel.com>
Cc: Fenghua" <fenghua.yu@intel.com>
Cc: Ravi V" <ravi.v.shankar@intel.com>
Cc: "Peter Zijlstra" <peterz@infradead.org>
Cc: "Stephane Eranian" <eranian@google.com>
Cc: "Andi Kleen" <ak@linux.intel.com>
Cc: "David Carrillo-Cisneros" <davidcc@google.com>
Cc: Vikas Shivappa <vikas.shivappa@linux.intel.com>
Link: http://lkml.kernel.org/r/3aea0a3bae219062c812668bd9b7b8f1a25003ba.1503512900.git.tony.luck@intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

---
 arch/x86/kernel/cpu/intel_rdt.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/arch/x86/kernel/cpu/intel_rdt.c b/arch/x86/kernel/cpu/intel_rdt.c
index b641622..cd5fc61 100644
--- a/arch/x86/kernel/cpu/intel_rdt.c
+++ b/arch/x86/kernel/cpu/intel_rdt.c
@@ -769,6 +769,9 @@ static __init void rdt_quirks(void)
 		if (!rdt_options[RDT_FLAG_L3_CAT].force_off)
 			cache_alloc_hsw_probe();
 		break;
+	case INTEL_FAM6_SKYLAKE_X:
+		if (boot_cpu_data.x86_mask <= 4)
+			set_rdt_options("!cmt,!mbmtotal,!mbmlocal,!l3cat");
 	}
 }