[PATCH 0/5] x86/platform/UV: Update TSC support

[PATCH 0/5] x86/platform/UV: Update TSC support

[mike.travis]

[Patches merged against 4.14.0-rc4]

The UV BIOS goes to considerable effort to get the TSC synchronization
accurate across the entire system.  Included in that are multiple chassis
that can have 32+ sockets.  The architecture does support an external
high resolution clock to aid in maintaining this synchronization.

The resulting TSC accuracy set by the UV system BIOS is much better
than the generic kernel TSC ADJUST functions.  This is important for
applications that read the TSC values directly for accessing data bases.

*   Patch 1 disables an assumption made by the kernel tsc sync functions
    that Socket 0 in the system should all have a TSC ADJUST value of
    zero.  This is not correct when the chassis are reset asynchronously
    to each other so which TSC's should be zero is not predictable.

*   Patch 2 prevents the kernel from attempting to fix the TSC if the
    system BIOS has determined that the TSC is not stable.  This prevents
    a slew of useless warning messages.

*   Patch 3 eliminates another avalanche of warning messages from older
    BIOS that did not have the TSC ADJUST MSR (ex. >3000 msgs in a 32
    socket Skylake system).  It now notes this with a single warning
    message and then moves on with fixing them.

*   Patch 4 puts in a facility to disable ART if the art to tsc conversion
    factor is not constant for all sockets.

*   Patch 5 puts a new check in the UV system init to check this new BIOS
    flag that displays the result of the TSC synchronization phase.  Three
    possible states are available, "sync is valid", "sync is unstable", or
    TSC sync state is unavailable in this BIOS.  In the later case, the
    UV kernel init reverts to prior assumptions about TSC sync state.

-- 

[PATCH 1/5] x86/kernel: Add option that TSC on Socket 0 being non-zero is valid

[mike.travis]

[-- Attachment #1: add-base-nonzero --]
[-- Type: text/plain, Size: 3484 bytes --]

Add a flag to indicate and process that TSC counters are on chassis
that reset at different times during system startup.  Therefore which
TSC ADJUST values should be zero is not predictable.

Signed-off-by: Mike Travis <mike.travis@hpe.com>
Reviewed-by: Dimitri Sivanich <dimitri.sivanich@hpe.com>
Reviewed-by: Russ Anderson <russ.anderson@hpe.com>
Reviewed-by: Andrew Banman <andrew.abanman@hpe.com>
Reviewed-by: Peter Zijlstra <peterz@infradead.org>
---
v2: changed to assume TSC != 0 valid on all systems.
v3: changed to put back flag for above with a clearer name.
v4: Patches merged against 4.14.0-rc4
---
 arch/x86/include/asm/tsc.h |    2 ++
 arch/x86/kernel/tsc_sync.c |   39 +++++++++++++++++++++++++++++++++++----
 2 files changed, 37 insertions(+), 4 deletions(-)

--- linux.orig/arch/x86/include/asm/tsc.h
+++ linux/arch/x86/include/asm/tsc.h
@@ -35,11 +35,13 @@ extern void tsc_init(void);
 extern void mark_tsc_unstable(char *reason);
 extern int unsynchronized_tsc(void);
 extern int check_tsc_unstable(void);
+extern void mark_tsc_async_resets(char *reason);
 extern unsigned long native_calibrate_cpu(void);
 extern unsigned long native_calibrate_tsc(void);
 extern unsigned long long native_sched_clock_from_tsc(u64 tsc);
 
 extern int tsc_clocksource_reliable;
+extern bool tsc_async_resets;
 
 /*
  * Boot-time check whether the TSCs are synchronized across
--- linux.orig/arch/x86/kernel/tsc_sync.c
+++ linux/arch/x86/kernel/tsc_sync.c
@@ -30,6 +30,20 @@ struct tsc_adjust {
 
 static DEFINE_PER_CPU(struct tsc_adjust, tsc_adjust);
 
+/*
+ * TSC's on different sockets may be reset asynchronously.
+ * This may cause the TSC ADJUST value on socket 0 to be NOT 0.
+ */
+bool __read_mostly tsc_async_resets;
+
+void mark_tsc_async_resets(char *reason)
+{
+	if (tsc_async_resets)
+		return;
+	tsc_async_resets = true;
+	pr_info("tsc: Marking TSC async resets true due to %s\n", reason);
+}
+
 void tsc_verify_tsc_adjust(bool resume)
 {
 	struct tsc_adjust *adj = this_cpu_ptr(&tsc_adjust);
@@ -71,12 +85,22 @@ static void tsc_sanitize_first_cpu(struc
 	 * non zero. We don't do that on non boot cpus because physical
 	 * hotplug should have set the ADJUST register to a value > 0 so
 	 * the TSC is in sync with the already running cpus.
+	 *
+	 * Also don't force the ADJUST value to zero if that is a valid value
+	 * for socket 0 as determined by the system arch.  This is required
+	 * when multiple sockets are reset asynchronously with each other
+	 * and socket 0 may not have an TSC ADJUST value of 0.
 	 */
 	if (bootcpu && bootval != 0) {
-		pr_warn(FW_BUG "TSC ADJUST: CPU%u: %lld force to 0\n", cpu,
-			bootval);
-		wrmsrl(MSR_IA32_TSC_ADJUST, 0);
-		bootval = 0;
+		if (likely(!tsc_async_resets)) {
+			pr_warn(FW_BUG "TSC ADJUST: CPU%u: %lld force to 0\n",
+				cpu, bootval);
+			wrmsrl(MSR_IA32_TSC_ADJUST, 0);
+			bootval = 0;
+		} else {
+			pr_info("TSC ADJUST: CPU%u: %lld NOT forced to 0\n",
+				cpu, bootval);
+		}
 	}
 	cur->adjusted = bootval;
 }
@@ -118,6 +142,13 @@ bool tsc_store_and_check_tsc_adjust(bool
 	cur->warned = false;
 
 	/*
+	 * If a non-zero TSC value for socket 0 may be valid then the default
+	 * adjusted value cannot assumed to be zero either.
+	 */
+	if (tsc_async_resets)
+		cur->adjusted = bootval;
+
+	/*
 	 * Check whether this CPU is the first in a package to come up. In
 	 * this case do not check the boot value against another package
 	 * because the new package might have been physically hotplugged,

-- 

[tip:x86/timers] x86/tsc: Add option that TSC on Socket 0 being non-zero is valid

[tip-bot for mike.travis@hpe.com]

Commit-ID:  341102c3ef29c33611586072363cf9982a8bdb77
Gitweb:     https://git.kernel.org/tip/341102c3ef29c33611586072363cf9982a8bdb77
Author:     mike.travis@hpe.com <mike.travis@hpe.com>
AuthorDate: Thu, 12 Oct 2017 11:32:02 -0500
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Mon, 16 Oct 2017 22:50:36 +0200

x86/tsc: Add option that TSC on Socket 0 being non-zero is valid

Add a flag to indicate and process that TSC counters are on chassis
that reset at different times during system startup.  Therefore which
TSC ADJUST values should be zero is not predictable.

Signed-off-by: Mike Travis <mike.travis@hpe.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Dimitri Sivanich <dimitri.sivanich@hpe.com>
Reviewed-by: Russ Anderson <russ.anderson@hpe.com>
Reviewed-by: Andrew Banman <andrew.abanman@hpe.com>
Reviewed-by: Peter Zijlstra <peterz@infradead.org>
Cc: Prarit Bhargava <prarit@redhat.com>
Cc: Andrew Banman <andrew.banman@hpe.com>
Cc: Bin Gao <bin.gao@linux.intel.com>
Link: https://lkml.kernel.org/r/20171012163201.944370012@stormcage.americas.sgi.com

---
 arch/x86/include/asm/tsc.h |  2 ++
 arch/x86/kernel/tsc_sync.c | 39 +++++++++++++++++++++++++++++++++++----
 2 files changed, 37 insertions(+), 4 deletions(-)

diff --git a/arch/x86/include/asm/tsc.h b/arch/x86/include/asm/tsc.h
index d0509c7..79125f3 100644
--- a/arch/x86/include/asm/tsc.h
+++ b/arch/x86/include/asm/tsc.h
@@ -36,11 +36,13 @@ extern void tsc_init(void);
 extern void mark_tsc_unstable(char *reason);
 extern int unsynchronized_tsc(void);
 extern int check_tsc_unstable(void);
+extern void mark_tsc_async_resets(char *reason);
 extern unsigned long native_calibrate_cpu(void);
 extern unsigned long native_calibrate_tsc(void);
 extern unsigned long long native_sched_clock_from_tsc(u64 tsc);
 
 extern int tsc_clocksource_reliable;
+extern bool tsc_async_resets;
 
 /*
  * Boot-time check whether the TSCs are synchronized across
diff --git a/arch/x86/kernel/tsc_sync.c b/arch/x86/kernel/tsc_sync.c
index 7842371..3873dcd 100644
--- a/arch/x86/kernel/tsc_sync.c
+++ b/arch/x86/kernel/tsc_sync.c
@@ -30,6 +30,20 @@ struct tsc_adjust {
 
 static DEFINE_PER_CPU(struct tsc_adjust, tsc_adjust);
 
+/*
+ * TSC's on different sockets may be reset asynchronously.
+ * This may cause the TSC ADJUST value on socket 0 to be NOT 0.
+ */
+bool __read_mostly tsc_async_resets;
+
+void mark_tsc_async_resets(char *reason)
+{
+	if (tsc_async_resets)
+		return;
+	tsc_async_resets = true;
+	pr_info("tsc: Marking TSC async resets true due to %s\n", reason);
+}
+
 void tsc_verify_tsc_adjust(bool resume)
 {
 	struct tsc_adjust *adj = this_cpu_ptr(&tsc_adjust);
@@ -71,12 +85,22 @@ static void tsc_sanitize_first_cpu(struct tsc_adjust *cur, s64 bootval,
 	 * non zero. We don't do that on non boot cpus because physical
 	 * hotplug should have set the ADJUST register to a value > 0 so
 	 * the TSC is in sync with the already running cpus.
+	 *
+	 * Also don't force the ADJUST value to zero if that is a valid value
+	 * for socket 0 as determined by the system arch.  This is required
+	 * when multiple sockets are reset asynchronously with each other
+	 * and socket 0 may not have an TSC ADJUST value of 0.
 	 */
 	if (bootcpu && bootval != 0) {
-		pr_warn(FW_BUG "TSC ADJUST: CPU%u: %lld force to 0\n", cpu,
-			bootval);
-		wrmsrl(MSR_IA32_TSC_ADJUST, 0);
-		bootval = 0;
+		if (likely(!tsc_async_resets)) {
+			pr_warn(FW_BUG "TSC ADJUST: CPU%u: %lld force to 0\n",
+				cpu, bootval);
+			wrmsrl(MSR_IA32_TSC_ADJUST, 0);
+			bootval = 0;
+		} else {
+			pr_info("TSC ADJUST: CPU%u: %lld NOT forced to 0\n",
+				cpu, bootval);
+		}
 	}
 	cur->adjusted = bootval;
 }
@@ -118,6 +142,13 @@ bool tsc_store_and_check_tsc_adjust(bool bootcpu)
 	cur->warned = false;
 
 	/*
+	 * If a non-zero TSC value for socket 0 may be valid then the default
+	 * adjusted value cannot assumed to be zero either.
+	 */
+	if (tsc_async_resets)
+		cur->adjusted = bootval;
+
+	/*
 	 * Check whether this CPU is the first in a package to come up. In
 	 * this case do not check the boot value against another package
 	 * because the new package might have been physically hotplugged,

[PATCH 2/5] x86/kernel: Skip TSC test and error messages if already unstable

[mike.travis]

[-- Attachment #1: unstable-tsc-disable-check --]
[-- Type: text/plain, Size: 1270 bytes --]

If the TSC has already been determined to be unstable, then checking
TSC ADJUST values is a waste of time and generates unnecessary error
messages.

Signed-off-by: Mike Travis <mike.travis@hpe.com>
Reviewed-by: Dimitri Sivanich <dimitri.sivanich@hpe.com>
Reviewed-by: Russ Anderson <russ.anderson@hpe.com>
Reviewed-by: Peter Zijlstra <peterz@infradead.org>
---
v2: Add check_tsc_unstable to !CONFIG_SMP case.
    Patches merged against 4.14.0-rc4
---
 arch/x86/kernel/tsc_sync.c |    8 ++++++++
 1 file changed, 8 insertions(+)

--- linux.orig/arch/x86/kernel/tsc_sync.c
+++ linux/arch/x86/kernel/tsc_sync.c
@@ -52,6 +52,10 @@ void tsc_verify_tsc_adjust(bool resume)
 	if (!boot_cpu_has(X86_FEATURE_TSC_ADJUST))
 		return;
 
+	/* Skip unnecessary error messages if TSC already unstable */
+	if (check_tsc_unstable())
+		return;
+
 	/* Rate limit the MSR check */
 	if (!resume && time_before(jiffies, adj->nextcheck))
 		return;
@@ -114,6 +118,10 @@ bool __init tsc_store_and_check_tsc_adju
 	if (!boot_cpu_has(X86_FEATURE_TSC_ADJUST))
 		return false;
 
+	/* Skip unnecessary error messages if TSC already unstable */
+	if (check_tsc_unstable())
+		return false;
+
 	rdmsrl(MSR_IA32_TSC_ADJUST, bootval);
 	cur->bootval = bootval;
 	cur->nextcheck = jiffies + HZ;

-- 

[tip:x86/timers] x86/tsc: Skip TSC test and error messages if already unstable

[tip-bot for mike.travis@hpe.com]

Commit-ID:  9514ececa52e9f1436e7682e98c852d1338b699f
Gitweb:     https://git.kernel.org/tip/9514ececa52e9f1436e7682e98c852d1338b699f
Author:     mike.travis@hpe.com <mike.travis@hpe.com>
AuthorDate: Thu, 12 Oct 2017 11:32:03 -0500
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Mon, 16 Oct 2017 22:50:36 +0200

x86/tsc: Skip TSC test and error messages if already unstable

If the TSC has already been determined to be unstable, then checking
TSC ADJUST values is a waste of time and generates unnecessary error
messages.

Signed-off-by: Mike Travis <mike.travis@hpe.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Dimitri Sivanich <dimitri.sivanich@hpe.com>
Reviewed-by: Russ Anderson <russ.anderson@hpe.com>
Reviewed-by: Peter Zijlstra <peterz@infradead.org>
Cc: Prarit Bhargava <prarit@redhat.com>
Cc: Andrew Banman <andrew.banman@hpe.com>
Cc: Bin Gao <bin.gao@linux.intel.com>
Link: https://lkml.kernel.org/r/20171012163202.060777495@stormcage.americas.sgi.com

---
 arch/x86/kernel/tsc_sync.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/arch/x86/kernel/tsc_sync.c b/arch/x86/kernel/tsc_sync.c
index 3873dcd..3bdb983 100644
--- a/arch/x86/kernel/tsc_sync.c
+++ b/arch/x86/kernel/tsc_sync.c
@@ -52,6 +52,10 @@ void tsc_verify_tsc_adjust(bool resume)
 	if (!boot_cpu_has(X86_FEATURE_TSC_ADJUST))
 		return;
 
+	/* Skip unnecessary error messages if TSC already unstable */
+	if (check_tsc_unstable())
+		return;
+
 	/* Rate limit the MSR check */
 	if (!resume && time_before(jiffies, adj->nextcheck))
 		return;
@@ -114,6 +118,10 @@ bool __init tsc_store_and_check_tsc_adjust(bool bootcpu)
 	if (!boot_cpu_has(X86_FEATURE_TSC_ADJUST))
 		return false;
 
+	/* Skip unnecessary error messages if TSC already unstable */
+	if (check_tsc_unstable())
+		return false;
+
 	rdmsrl(MSR_IA32_TSC_ADJUST, bootval);
 	cur->bootval = bootval;
 	cur->nextcheck = jiffies + HZ;

[PATCH 3/5] x86/kernel: Drastically reduce the number of firmware bug warnings

[mike.travis]

[-- Attachment #1: quiet-excessive-warnings --]
[-- Type: text/plain, Size: 1756 bytes --]

Prior to the TSC ADJUST MSR being available, the method to set TSC's in
sync with each other naturally caused a small skew between cpu threads.
This was NOT a firmware bug at the time so introducing a whole avalanche
of alarming warning messages might cause unnecessary concern and customer
complaints. (Example: >3000 msgs in a 32 socket Skylake system.)

Simply report the warning condition, if possible do the necessary fixes,
and move on.

Signed-off-by: Mike Travis <mike.travis@hpe.com>
Reviewed-by: Dimitri Sivanich <dimitri.sivanich@hpe.com>
Reviewed-by: Russ Anderson <russ.anderson@hpe.com>
Reviewed-by: Peter Zijlstra <peterz@infradead.org>
---
v2: Patches merged against 4.14.0-rc4
---
 arch/x86/kernel/tsc_sync.c |    9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

--- linux.orig/arch/x86/kernel/tsc_sync.c
+++ linux/arch/x86/kernel/tsc_sync.c
@@ -177,10 +177,9 @@ bool tsc_store_and_check_tsc_adjust(bool
 	 * Compare the boot value and complain if it differs in the
 	 * package.
 	 */
-	if (bootval != ref->bootval) {
-		pr_warn(FW_BUG "TSC ADJUST differs: Reference CPU%u: %lld CPU%u: %lld\n",
-			refcpu, ref->bootval, cpu, bootval);
-	}
+	if (bootval != ref->bootval)
+		printk_once(FW_BUG "TSC ADJUST differs within socket(s), fixing all errors\n");
+
 	/*
 	 * The TSC_ADJUST values in a package must be the same. If the boot
 	 * value on this newly upcoming CPU differs from the adjustment
@@ -188,8 +187,6 @@ bool tsc_store_and_check_tsc_adjust(bool
 	 * adjusted value.
 	 */
 	if (bootval != ref->adjusted) {
-		pr_warn("TSC ADJUST synchronize: Reference CPU%u: %lld CPU%u: %lld\n",
-			refcpu, ref->adjusted, cpu, bootval);
 		cur->adjusted = ref->adjusted;
 		wrmsrl(MSR_IA32_TSC_ADJUST, ref->adjusted);
 	}

-- 

[tip:x86/timers] x86/tsc: Drastically reduce the number of firmware bug warnings

[tip-bot for mike.travis@hpe.com]

Commit-ID:  41e7864ab5ce4ec36e89a9f55d8d9dfe19b0392c
Gitweb:     https://git.kernel.org/tip/41e7864ab5ce4ec36e89a9f55d8d9dfe19b0392c
Author:     mike.travis@hpe.com <mike.travis@hpe.com>
AuthorDate: Thu, 12 Oct 2017 11:32:04 -0500
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Mon, 16 Oct 2017 22:50:36 +0200

x86/tsc: Drastically reduce the number of firmware bug warnings

Prior to the TSC ADJUST MSR being available, the method to set TSC's in
sync with each other naturally caused a small skew between cpu threads.
This was NOT a firmware bug at the time so introducing a whole avalanche
of alarming warning messages might cause unnecessary concern and customer
complaints. (Example: >3000 msgs in a 32 socket Skylake system.)

Simply report the warning condition, if possible do the necessary fixes,
and move on.

Signed-off-by: Mike Travis <mike.travis@hpe.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Dimitri Sivanich <dimitri.sivanich@hpe.com>
Reviewed-by: Russ Anderson <russ.anderson@hpe.com>
Reviewed-by: Peter Zijlstra <peterz@infradead.org>
Cc: Prarit Bhargava <prarit@redhat.com>
Cc: Andrew Banman <andrew.banman@hpe.com>
Cc: Bin Gao <bin.gao@linux.intel.com>
Link: https://lkml.kernel.org/r/20171012163202.175062400@stormcage.americas.sgi.com

---
 arch/x86/kernel/tsc_sync.c | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/arch/x86/kernel/tsc_sync.c b/arch/x86/kernel/tsc_sync.c
index 3bdb983..26ba053 100644
--- a/arch/x86/kernel/tsc_sync.c
+++ b/arch/x86/kernel/tsc_sync.c
@@ -177,10 +177,9 @@ bool tsc_store_and_check_tsc_adjust(bool bootcpu)
 	 * Compare the boot value and complain if it differs in the
 	 * package.
 	 */
-	if (bootval != ref->bootval) {
-		pr_warn(FW_BUG "TSC ADJUST differs: Reference CPU%u: %lld CPU%u: %lld\n",
-			refcpu, ref->bootval, cpu, bootval);
-	}
+	if (bootval != ref->bootval)
+		printk_once(FW_BUG "TSC ADJUST differs within socket(s), fixing all errors\n");
+
 	/*
 	 * The TSC_ADJUST values in a package must be the same. If the boot
 	 * value on this newly upcoming CPU differs from the adjustment
@@ -188,8 +187,6 @@ bool tsc_store_and_check_tsc_adjust(bool bootcpu)
 	 * adjusted value.
 	 */
 	if (bootval != ref->adjusted) {
-		pr_warn("TSC ADJUST synchronize: Reference CPU%u: %lld CPU%u: %lld\n",
-			refcpu, ref->adjusted, cpu, bootval);
 		cur->adjusted = ref->adjusted;
 		wrmsrl(MSR_IA32_TSC_ADJUST, ref->adjusted);
 	}

[PATCH 4/5] x86/kernel: Provide a means to disable TSC ART

[mike.travis]

[-- Attachment #1: disable-detect_art --]
[-- Type: text/plain, Size: 1221 bytes --]

On systems where multiple chassis are reset asynchronously, and thus
the TSC counters are started asynchronously, the offset needed to
convert to TSC to ART would be different.  Disable ART in that case
and rely on the TSC counters to supply the accurate time.

Signed-off-by: Mike Travis <mike.travis@hpe.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
---
v2: changed to use common "tsc_async_resets" flag.
v3: Patches merged against 4.14.0-rc4
---
 arch/x86/kernel/tsc.c |    8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

--- linux.orig/arch/x86/kernel/tsc.c
+++ linux/arch/x86/kernel/tsc.c
@@ -966,10 +966,14 @@ static void detect_art(void)
 	if (boot_cpu_data.cpuid_level < ART_CPUID_LEAF)
 		return;
 
-	/* Don't enable ART in a VM, non-stop TSC and TSC_ADJUST required */
+	/*
+	 * Don't enable ART in a VM, non-stop TSC and TSC_ADJUST required,
+	 * and the TSC counter resets must not occur asynchronously.
+	 */
 	if (boot_cpu_has(X86_FEATURE_HYPERVISOR) ||
 	    !boot_cpu_has(X86_FEATURE_NONSTOP_TSC) ||
-	    !boot_cpu_has(X86_FEATURE_TSC_ADJUST))
+	    !boot_cpu_has(X86_FEATURE_TSC_ADJUST) ||
+	    tsc_async_resets)
 		return;
 
 	cpuid(ART_CPUID_LEAF, &art_to_tsc_denominator,

-- 

[tip:x86/timers] x86/tsc: Provide a means to disable TSC ART

[tip-bot for mike.travis@hpe.com]

Commit-ID:  6c66350d0a482892793b888b07c1177fc6d4b344
Gitweb:     https://git.kernel.org/tip/6c66350d0a482892793b888b07c1177fc6d4b344
Author:     mike.travis@hpe.com <mike.travis@hpe.com>
AuthorDate: Thu, 12 Oct 2017 11:32:05 -0500
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Mon, 16 Oct 2017 22:50:37 +0200

x86/tsc: Provide a means to disable TSC ART

On systems where multiple chassis are reset asynchronously, and thus
the TSC counters are started asynchronously, the offset needed to
convert to TSC to ART would be different.  Disable ART in that case
and rely on the TSC counters to supply the accurate time.

Signed-off-by: Mike Travis <mike.travis@hpe.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Prarit Bhargava <prarit@redhat.com>
Cc: Dimitri Sivanich <dimitri.sivanich@hpe.com>
Cc: Russ Anderson <russ.anderson@hpe.com>
Cc: Andrew Banman <andrew.banman@hpe.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Bin Gao <bin.gao@linux.intel.com>
Link: https://lkml.kernel.org/r/20171012163202.289397994@stormcage.americas.sgi.com

---
 arch/x86/kernel/tsc.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
index 896dbe3..f1326c0 100644
--- a/arch/x86/kernel/tsc.c
+++ b/arch/x86/kernel/tsc.c
@@ -962,10 +962,14 @@ static void detect_art(void)
 	if (boot_cpu_data.cpuid_level < ART_CPUID_LEAF)
 		return;
 
-	/* Don't enable ART in a VM, non-stop TSC and TSC_ADJUST required */
+	/*
+	 * Don't enable ART in a VM, non-stop TSC and TSC_ADJUST required,
+	 * and the TSC counter resets must not occur asynchronously.
+	 */
 	if (boot_cpu_has(X86_FEATURE_HYPERVISOR) ||
 	    !boot_cpu_has(X86_FEATURE_NONSTOP_TSC) ||
-	    !boot_cpu_has(X86_FEATURE_TSC_ADJUST))
+	    !boot_cpu_has(X86_FEATURE_TSC_ADJUST) ||
+	    tsc_async_resets)
 		return;
 
 	cpuid(ART_CPUID_LEAF, &art_to_tsc_denominator,

[PATCH 5/5] x86/platform/UV: Add check of TSC state set by UV BIOS

[mike.travis]

[-- Attachment #1: uv-set-tsc-valid-state --]
[-- Type: text/plain, Size: 3977 bytes --]

Insert a check early in UV system startup that checks whether BIOS was
able to obtain satisfactory TSC Sync stability.  If not, it usually
is caused by an error in the external TSC clock generation source.
In this case the best fallback is to use the builtin hardware RTC
as the kernel will not be able to set an accurate TSC sync either.

Signed-off-by: Mike Travis <mike.travis@hpe.com>
Reviewed-by: Dimitri Sivanich <dimitri.sivanich@hpe.com>
Reviewed-by: Russ Anderson <russ.anderson@hpe.com>
Reviewed-by: Andrew Banman <andrew.abanman@hpe.com>
---
v2: Patches merged against 4.14.0-rc4
---
 arch/x86/include/asm/uv/uv_hub.h   |   23 +++++++++++++++----
 arch/x86/kernel/apic/x2apic_uv_x.c |   43 +++++++++++++++++++++++++++++++++++++
 2 files changed, 61 insertions(+), 5 deletions(-)

--- linux.orig/arch/x86/include/asm/uv/uv_hub.h
+++ linux/arch/x86/include/asm/uv/uv_hub.h
@@ -776,23 +776,36 @@ static inline int uv_num_possible_blades
 extern void uv_nmi_setup(void);
 extern void uv_nmi_setup_hubless(void);
 
+/* BIOS/Kernel flags exchange MMR */
+#define UVH_BIOS_KERNEL_MMR		UVH_SCRATCH5
+#define UVH_BIOS_KERNEL_MMR_ALIAS	UVH_SCRATCH5_ALIAS
+#define UVH_BIOS_KERNEL_MMR_ALIAS_2	UVH_SCRATCH5_ALIAS_2
+
+/* TSC sync valid, set by BIOS */
+#define UVH_TSC_SYNC_MMR	UVH_BIOS_KERNEL_MMR
+#define UVH_TSC_SYNC_SHIFT	10
+#define UVH_TSC_SYNC_SHIFT_UV2K	16	/* UV2/3k have different bits */
+#define UVH_TSC_SYNC_MASK	3	/* 0011 */
+#define UVH_TSC_SYNC_VALID	3	/* 0011 */
+#define UVH_TSC_SYNC_INVALID	2	/* 0010 */
+
 /* BMC sets a bit this MMR non-zero before sending an NMI */
-#define UVH_NMI_MMR		UVH_SCRATCH5
-#define UVH_NMI_MMR_CLEAR	UVH_SCRATCH5_ALIAS
+#define UVH_NMI_MMR		UVH_BIOS_KERNEL_MMR
+#define UVH_NMI_MMR_CLEAR	UVH_BIOS_KERNEL_MMR_ALIAS
 #define UVH_NMI_MMR_SHIFT	63
-#define	UVH_NMI_MMR_TYPE	"SCRATCH5"
+#define UVH_NMI_MMR_TYPE	"SCRATCH5"
 
 /* Newer SMM NMI handler, not present in all systems */
 #define UVH_NMI_MMRX		UVH_EVENT_OCCURRED0
 #define UVH_NMI_MMRX_CLEAR	UVH_EVENT_OCCURRED0_ALIAS
 #define UVH_NMI_MMRX_SHIFT	UVH_EVENT_OCCURRED0_EXTIO_INT0_SHFT
-#define	UVH_NMI_MMRX_TYPE	"EXTIO_INT0"
+#define UVH_NMI_MMRX_TYPE	"EXTIO_INT0"
 
 /* Non-zero indicates newer SMM NMI handler present */
 #define UVH_NMI_MMRX_SUPPORTED	UVH_EXTIO_INT0_BROADCAST
 
 /* Indicates to BIOS that we want to use the newer SMM NMI handler */
-#define UVH_NMI_MMRX_REQ	UVH_SCRATCH5_ALIAS_2
+#define UVH_NMI_MMRX_REQ	UVH_BIOS_KERNEL_MMR_ALIAS_2
 #define UVH_NMI_MMRX_REQ_SHIFT	62
 
 struct uv_hub_nmi_s {
--- linux.orig/arch/x86/kernel/apic/x2apic_uv_x.c
+++ linux/arch/x86/kernel/apic/x2apic_uv_x.c
@@ -154,6 +154,48 @@ static int __init early_get_pnodeid(void
 	return pnode;
 }
 
+static void uv_tsc_check_sync(void)
+{
+	u64 mmr;
+	int sync_state;
+	int mmr_shift;
+	char *state;
+	bool valid;
+
+	/* Accommodate different UV arch BIOSes */
+	mmr = uv_early_read_mmr(UVH_TSC_SYNC_MMR);
+	mmr_shift =
+		is_uv1_hub() ? 0 :
+		is_uv2_hub() ? UVH_TSC_SYNC_SHIFT_UV2K : UVH_TSC_SYNC_SHIFT;
+	if (mmr_shift)
+		sync_state = (mmr >> mmr_shift) & UVH_TSC_SYNC_MASK;
+	else
+		sync_state = 0;
+
+	switch (sync_state) {
+	case UVH_TSC_SYNC_VALID:
+		state = "in sync";
+		valid = true;
+		break;
+
+	case UVH_TSC_SYNC_INVALID:
+		state = "unstable";
+		valid = false;
+		break;
+	default:
+		state = "unknown: assuming valid";
+		valid = true;
+		break;
+	}
+	pr_info("UV: TSC sync state from BIOS:0%d(%s)\n", sync_state, state);
+
+	/* Mark flag that says TSC != 0 is valid for socket 0 */
+	if (valid)
+		mark_tsc_async_resets("UV BIOS");
+	else
+		mark_tsc_unstable("UV BIOS");
+}
+
 /* [Copied from arch/x86/kernel/cpu/topology.c:detect_extended_topology()] */
 
 #define SMT_LEVEL			0	/* Leaf 0xb SMT level */
@@ -288,6 +330,7 @@ static int __init uv_acpi_madt_oem_check
 	}
 
 	pr_info("UV: OEM IDs %s/%s, System/HUB Types %d/%d, uv_apic %d\n", oem_id, oem_table_id, uv_system_type, uv_min_hub_revision_id, uv_apic);
+	uv_tsc_check_sync();
 
 	return uv_apic;
 

-- 

[tip:x86/timers] x86/platform/UV: Add check of TSC state set by UV BIOS

[tip-bot for mike.travis@hpe.com]

Commit-ID:  97d21003df3e7504c899b1701546f18ff475966f
Gitweb:     https://git.kernel.org/tip/97d21003df3e7504c899b1701546f18ff475966f
Author:     mike.travis@hpe.com <mike.travis@hpe.com>
AuthorDate: Thu, 12 Oct 2017 11:32:06 -0500
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Mon, 16 Oct 2017 22:50:37 +0200

x86/platform/UV: Add check of TSC state set by UV BIOS

Insert a check early in UV system startup that checks whether BIOS was
able to obtain satisfactory TSC Sync stability.  If not, it usually
is caused by an error in the external TSC clock generation source.
In this case the best fallback is to use the builtin hardware RTC
as the kernel will not be able to set an accurate TSC sync either.

Signed-off-by: Mike Travis <mike.travis@hpe.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Dimitri Sivanich <dimitri.sivanich@hpe.com>
Reviewed-by: Russ Anderson <russ.anderson@hpe.com>
Reviewed-by: Andrew Banman <andrew.abanman@hpe.com>
Cc: Prarit Bhargava <prarit@redhat.com>
Cc: Andrew Banman <andrew.banman@hpe.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Bin Gao <bin.gao@linux.intel.com>
Link: https://lkml.kernel.org/r/20171012163202.406294490@stormcage.americas.sgi.com

---
 arch/x86/include/asm/uv/uv_hub.h   | 23 +++++++++++++++-----
 arch/x86/kernel/apic/x2apic_uv_x.c | 43 ++++++++++++++++++++++++++++++++++++++
 2 files changed, 61 insertions(+), 5 deletions(-)

diff --git a/arch/x86/include/asm/uv/uv_hub.h b/arch/x86/include/asm/uv/uv_hub.h
index 9cffb44..036e26d 100644
--- a/arch/x86/include/asm/uv/uv_hub.h
+++ b/arch/x86/include/asm/uv/uv_hub.h
@@ -776,23 +776,36 @@ static inline int uv_num_possible_blades(void)
 extern void uv_nmi_setup(void);
 extern void uv_nmi_setup_hubless(void);
 
+/* BIOS/Kernel flags exchange MMR */
+#define UVH_BIOS_KERNEL_MMR		UVH_SCRATCH5
+#define UVH_BIOS_KERNEL_MMR_ALIAS	UVH_SCRATCH5_ALIAS
+#define UVH_BIOS_KERNEL_MMR_ALIAS_2	UVH_SCRATCH5_ALIAS_2
+
+/* TSC sync valid, set by BIOS */
+#define UVH_TSC_SYNC_MMR	UVH_BIOS_KERNEL_MMR
+#define UVH_TSC_SYNC_SHIFT	10
+#define UVH_TSC_SYNC_SHIFT_UV2K	16	/* UV2/3k have different bits */
+#define UVH_TSC_SYNC_MASK	3	/* 0011 */
+#define UVH_TSC_SYNC_VALID	3	/* 0011 */
+#define UVH_TSC_SYNC_INVALID	2	/* 0010 */
+
 /* BMC sets a bit this MMR non-zero before sending an NMI */
-#define UVH_NMI_MMR		UVH_SCRATCH5
-#define UVH_NMI_MMR_CLEAR	UVH_SCRATCH5_ALIAS
+#define UVH_NMI_MMR		UVH_BIOS_KERNEL_MMR
+#define UVH_NMI_MMR_CLEAR	UVH_BIOS_KERNEL_MMR_ALIAS
 #define UVH_NMI_MMR_SHIFT	63
-#define	UVH_NMI_MMR_TYPE	"SCRATCH5"
+#define UVH_NMI_MMR_TYPE	"SCRATCH5"
 
 /* Newer SMM NMI handler, not present in all systems */
 #define UVH_NMI_MMRX		UVH_EVENT_OCCURRED0
 #define UVH_NMI_MMRX_CLEAR	UVH_EVENT_OCCURRED0_ALIAS
 #define UVH_NMI_MMRX_SHIFT	UVH_EVENT_OCCURRED0_EXTIO_INT0_SHFT
-#define	UVH_NMI_MMRX_TYPE	"EXTIO_INT0"
+#define UVH_NMI_MMRX_TYPE	"EXTIO_INT0"
 
 /* Non-zero indicates newer SMM NMI handler present */
 #define UVH_NMI_MMRX_SUPPORTED	UVH_EXTIO_INT0_BROADCAST
 
 /* Indicates to BIOS that we want to use the newer SMM NMI handler */
-#define UVH_NMI_MMRX_REQ	UVH_SCRATCH5_ALIAS_2
+#define UVH_NMI_MMRX_REQ	UVH_BIOS_KERNEL_MMR_ALIAS_2
 #define UVH_NMI_MMRX_REQ_SHIFT	62
 
 struct uv_hub_nmi_s {
diff --git a/arch/x86/kernel/apic/x2apic_uv_x.c b/arch/x86/kernel/apic/x2apic_uv_x.c
index 0d57bb9..4408254 100644
--- a/arch/x86/kernel/apic/x2apic_uv_x.c
+++ b/arch/x86/kernel/apic/x2apic_uv_x.c
@@ -154,6 +154,48 @@ static int __init early_get_pnodeid(void)
 	return pnode;
 }
 
+static void uv_tsc_check_sync(void)
+{
+	u64 mmr;
+	int sync_state;
+	int mmr_shift;
+	char *state;
+	bool valid;
+
+	/* Accommodate different UV arch BIOSes */
+	mmr = uv_early_read_mmr(UVH_TSC_SYNC_MMR);
+	mmr_shift =
+		is_uv1_hub() ? 0 :
+		is_uv2_hub() ? UVH_TSC_SYNC_SHIFT_UV2K : UVH_TSC_SYNC_SHIFT;
+	if (mmr_shift)
+		sync_state = (mmr >> mmr_shift) & UVH_TSC_SYNC_MASK;
+	else
+		sync_state = 0;
+
+	switch (sync_state) {
+	case UVH_TSC_SYNC_VALID:
+		state = "in sync";
+		valid = true;
+		break;
+
+	case UVH_TSC_SYNC_INVALID:
+		state = "unstable";
+		valid = false;
+		break;
+	default:
+		state = "unknown: assuming valid";
+		valid = true;
+		break;
+	}
+	pr_info("UV: TSC sync state from BIOS:0%d(%s)\n", sync_state, state);
+
+	/* Mark flag that says TSC != 0 is valid for socket 0 */
+	if (valid)
+		mark_tsc_async_resets("UV BIOS");
+	else
+		mark_tsc_unstable("UV BIOS");
+}
+
 /* [Copied from arch/x86/kernel/cpu/topology.c:detect_extended_topology()] */
 
 #define SMT_LEVEL			0	/* Leaf 0xb SMT level */
@@ -288,6 +330,7 @@ static int __init uv_acpi_madt_oem_check(char *oem_id, char *oem_table_id)
 	}
 
 	pr_info("UV: OEM IDs %s/%s, System/HUB Types %d/%d, uv_apic %d\n", oem_id, oem_table_id, uv_system_type, uv_min_hub_revision_id, uv_apic);
+	uv_tsc_check_sync();
 
 	return uv_apic;