diff mbox series

[v3,4/5] x86/mce: Move storm handling to core.

Message ID 20230317172042.117201-5-tony.luck@intel.com (mailing list archive)
State New, archived
Headers show
Series Handle corrected machine check interrupt storms | expand

Commit Message

Tony Luck March 17, 2023, 5:20 p.m. UTC
From: Smita Koralahalli <Smita.KoralahalliChannabasappa@amd.com>

AMD's storm handling for threshold interrupts is similar to Intel's CMCI
storm handling. Hence, make the storm handling code common by moving to
core and removing the vendor exclusivity.

On the contrary, setting different thresholds to reduce rate of interrupts
in IA32_MCi_CTL2 register is kept Intel intact as the storm handling for
AMD slightly differs where in it handles the storms by turning off the
interrupts.

No functional changes.

[Tony: Same as Smita's original, plus changes rolled in from prior patches]

Signed-off-by: Smita Koralahalli <Smita.KoralahalliChannabasappa@amd.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
---
 arch/x86/kernel/cpu/mce/internal.h | 18 ++++++
 arch/x86/kernel/cpu/mce/core.c     | 81 ++++++++++++++++++++++++++
 arch/x86/kernel/cpu/mce/intel.c    | 93 +-----------------------------
 3 files changed, 100 insertions(+), 92 deletions(-)

Comments

Yazen Ghannam March 23, 2023, 3:27 p.m. UTC | #1
On Fri, Mar 17, 2023 at 10:20:41AM -0700, Tony Luck wrote:
> From: Smita Koralahalli <Smita.KoralahalliChannabasappa@amd.com>
> 
> AMD's storm handling for threshold interrupts is similar to Intel's CMCI
> storm handling. Hence, make the storm handling code common by moving to
> core and removing the vendor exclusivity.
> 
> On the contrary, setting different thresholds to reduce rate of interrupts
> in IA32_MCi_CTL2 register is kept Intel intact as the storm handling for
> AMD slightly differs where in it handles the storms by turning off the
> interrupts.
> 
> No functional changes.
> 
> [Tony: Same as Smita's original, plus changes rolled in from prior patches]
> 
> Signed-off-by: Smita Koralahalli <Smita.KoralahalliChannabasappa@amd.com>
> Signed-off-by: Tony Luck <tony.luck@intel.com>
> ---

Can this patch and the previous two be squashed together?

Like so?
 Patch 1: Remove old code.
 Patch 2: Add new common and Intel-specific code.
 Patch 3: Add AMD-specific code.

Thanks,
Yazen
Tony Luck March 23, 2023, 6:10 p.m. UTC | #2
> Can this patch and the previous two be squashed together?
>
> Like so?
>  Patch 1: Remove old code.
>  Patch 2: Add new common and Intel-specific code.
>  Patch 3: Add AMD-specific code.

Yazen,

Those three patches could be merged ... but they already seem big:

0002: 3 files changed, 158 insertions(+), 11 deletions(-)
0003: 3 files changed, 22 insertions(+), 2 deletions(-)
0004: 3 files changed, 100 insertions(+), 92 deletions(-)

Lumping them together wouldn't be the sum of those but would be worse (IMHO)

-Tony
Tony Luck March 23, 2023, 8:26 p.m. UTC | #3
Yazen,

I folded the fixes for the issues you pointed to in patch 3/5 into the series and rebased to v6.3-rc3

Resulting tree pushed here:

git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux.git     try-storm-on6-3

Builds, boots, and passes my storm tests here.

How is testing going on the AMD side of this series?

-Tony
Yazen Ghannam March 24, 2023, 8:44 p.m. UTC | #4
On Thu, Mar 23, 2023 at 08:26:02PM +0000, Luck, Tony wrote:
> Yazen,
> 
> I folded the fixes for the issues you pointed to in patch 3/5 into the series and rebased to v6.3-rc3
> 
> Resulting tree pushed here:
> 
> git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux.git     try-storm-on6-3
> 
> Builds, boots, and passes my storm tests here.
> 
> How is testing going on the AMD side of this series?
>

Thanks Tony. I'll try to test by the middle of next week. Sorry for the delay,
I just got back from work travel.

I think the code looks good though.

Thanks,
Yazen
Yazen Ghannam March 29, 2023, 3:26 p.m. UTC | #5
On Thu, Mar 23, 2023 at 08:26:02PM +0000, Luck, Tony wrote:
> Yazen,
> 
> I folded the fixes for the issues you pointed to in patch 3/5 into the series and rebased to v6.3-rc3
> 
> Resulting tree pushed here:
> 
> git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux.git     try-storm-on6-3
> 
> Builds, boots, and passes my storm tests here.
> 
> How is testing going on the AMD side of this series?
>

Hi Tony,

Builds, boots, and passes my tests here too.

Reviewed-by: Yazen Ghannam <yazen.ghannam@amd.com>
Tested-by: Yazen Ghannam <yazen.ghannam@amd.com>

Thanks!

-Yazen
Tony Luck April 3, 2023, 7:03 p.m. UTC | #6
> Hi Tony,
>
> Builds, boots, and passes my tests here too.
>
> Reviewed-by: Yazen Ghannam <yazen.ghannam@amd.com>
> Tested-by: Yazen Ghannam <yazen.ghannam@amd.com>

Yazen,

Thanks. I'll add your tags and post a v4 to the mailing list for Boris to review.

-Tony
diff mbox series

Patch

diff --git a/arch/x86/kernel/cpu/mce/internal.h b/arch/x86/kernel/cpu/mce/internal.h
index f37816b4d4cf..9b2c54f30fb9 100644
--- a/arch/x86/kernel/cpu/mce/internal.h
+++ b/arch/x86/kernel/cpu/mce/internal.h
@@ -60,6 +60,24 @@  static inline bool intel_filter_mce(struct mce *m) { return false; }
 
 void mce_timer_kick(bool storm);
 void mce_handle_storm(int bank, bool on);
+void cmci_storm_begin(int bank);
+void cmci_storm_end(int bank);
+
+DECLARE_PER_CPU(int, stormy_bank_count);
+DECLARE_PER_CPU(u64 [MAX_NR_BANKS], bank_history);
+DECLARE_PER_CPU(bool [MAX_NR_BANKS], bank_storm);
+DECLARE_PER_CPU(unsigned long [MAX_NR_BANKS], bank_time_stamp);
+
+/*
+ * How many errors within the history buffer mark the start of a storm
+ */
+#define STORM_BEGIN_THRESHOLD	5
+
+/*
+ * How many polls of machine check bank without an error before declaring
+ * the storm is over
+ */
+#define STORM_END_POLL_THRESHOLD	30
 
 #ifdef CONFIG_ACPI_APEI
 int apei_write_mce(struct mce *m);
diff --git a/arch/x86/kernel/cpu/mce/core.c b/arch/x86/kernel/cpu/mce/core.c
index f4d2a7ba29f7..d27daa199523 100644
--- a/arch/x86/kernel/cpu/mce/core.c
+++ b/arch/x86/kernel/cpu/mce/core.c
@@ -613,6 +613,87 @@  static struct notifier_block mce_default_nb = {
 	.priority	= MCE_PRIO_LOWEST,
 };
 
+/*
+ * CMCI storm tracking state
+ *	stormy_bank_count: per-cpu count of MC banks in storm state
+ *	bank_history: bitmask tracking of corrected errors seen in each bank
+ *	bank_time_stamp: last time (in jiffies) that each bank was polled
+ */
+DEFINE_PER_CPU(int, stormy_bank_count);
+DEFINE_PER_CPU(u64 [MAX_NR_BANKS], bank_history);
+DEFINE_PER_CPU(bool [MAX_NR_BANKS], bank_storm);
+DEFINE_PER_CPU(unsigned long [MAX_NR_BANKS], bank_time_stamp);
+
+void cmci_storm_begin(int bank)
+{
+	__set_bit(bank, this_cpu_ptr(mce_poll_banks));
+	this_cpu_write(bank_storm[bank], true);
+
+	/*
+	 * If this is the first bank on this CPU to enter storm mode
+	 * start polling
+	 */
+	if (this_cpu_inc_return(stormy_bank_count) == 1)
+		mce_timer_kick(true);
+}
+
+void cmci_storm_end(int bank)
+{
+	__clear_bit(bank, this_cpu_ptr(mce_poll_banks));
+	this_cpu_write(bank_history[bank], 0ull);
+	this_cpu_write(bank_storm[bank], false);
+
+	/* If no banks left in storm mode, stop polling */
+	if (!this_cpu_dec_return(stormy_bank_count))
+		mce_timer_kick(false);
+}
+
+void track_cmci_storm(int bank, u64 status)
+{
+	unsigned long now = jiffies, delta;
+	unsigned int shift = 1;
+	u64 history;
+
+	/*
+	 * When a bank is in storm mode it is polled once per second and
+	 * the history mask will record about the last minute of poll results.
+	 * If it is not in storm mode, then the bank is only checked when
+	 * there is a CMCI interrupt. Check how long it has been since
+	 * this bank was last checked, and adjust the amount of "shift"
+	 * to apply to history.
+	 */
+	if (!this_cpu_read(bank_storm[bank])) {
+		delta = now - this_cpu_read(bank_time_stamp[bank]);
+		shift = (delta + HZ) / HZ;
+	}
+
+	/* If has been a long time since the last poll, clear history */
+	if (shift >= 64)
+		history = 0;
+	else
+		history = this_cpu_read(bank_history[bank]) << shift;
+	this_cpu_write(bank_time_stamp[bank], now);
+
+	/* History keeps track of corrected errors. VAL=1 && UC=0 */
+	if ((status & (MCI_STATUS_VAL | MCI_STATUS_UC)) == MCI_STATUS_VAL)
+		history |= 1;
+	this_cpu_write(bank_history[bank], history);
+
+	if (this_cpu_read(bank_storm[bank])) {
+		if (history & GENMASK_ULL(STORM_END_POLL_THRESHOLD - 1, 0))
+			return;
+		pr_notice("CPU%d BANK%d CMCI storm subsided\n", smp_processor_id(), bank);
+		mce_handle_storm(bank, true);
+		cmci_storm_end(bank);
+	} else {
+		if (hweight64(history) < STORM_BEGIN_THRESHOLD)
+			return;
+		pr_notice("CPU%d BANK%d CMCI storm detected\n", smp_processor_id(), bank);
+		mce_handle_storm(bank, false);
+		cmci_storm_begin(bank);
+	}
+}
+
 /*
  * Read ADDR and MISC registers.
  */
diff --git a/arch/x86/kernel/cpu/mce/intel.c b/arch/x86/kernel/cpu/mce/intel.c
index 4238b73c2143..6cc9aa97c092 100644
--- a/arch/x86/kernel/cpu/mce/intel.c
+++ b/arch/x86/kernel/cpu/mce/intel.c
@@ -47,17 +47,7 @@  static DEFINE_PER_CPU(mce_banks_t, mce_banks_owned);
  */
 static DEFINE_RAW_SPINLOCK(cmci_discover_lock);
 
-/*
- * CMCI storm tracking state
- *	stormy_bank_count: per-cpu count of MC banks in storm state
- *	bank_history: bitmask tracking of corrected errors seen in each bank
- *	bank_time_stamp: last time (in jiffies) that each bank was polled
- *	cmci_threshold: MCi_CTL2 threshold for each bank when there is no storm
- */
-static DEFINE_PER_CPU(int, stormy_bank_count);
-static DEFINE_PER_CPU(u64 [MAX_NR_BANKS], bank_history);
-static DEFINE_PER_CPU(bool [MAX_NR_BANKS], bank_storm);
-static DEFINE_PER_CPU(unsigned long [MAX_NR_BANKS], bank_time_stamp);
+/* MCi_CTL2 threshold for each bank when there is no storm */
 static int cmci_threshold[MAX_NR_BANKS];
 
 /* Linux non-storm CMCI threshold (may be overridden by BIOS */
@@ -70,17 +60,6 @@  static int cmci_threshold[MAX_NR_BANKS];
  */
 #define CMCI_STORM_THRESHOLD	32749
 
-/*
- * How many errors within the history buffer mark the start of a storm
- */
-#define STORM_BEGIN_THRESHOLD	5
-
-/*
- * How many polls of machine check bank without an error before declaring
- * the storm is over
- */
-#define STORM_END_POLL_THRESHOLD	30
-
 static int cmci_supported(int *banks)
 {
 	u64 cap;
@@ -160,76 +139,6 @@  void mce_intel_handle_storm(int bank, bool on)
 		cmci_set_threshold(bank, CMCI_STORM_THRESHOLD);
 }
 
-static void cmci_storm_begin(int bank)
-{
-	__set_bit(bank, this_cpu_ptr(mce_poll_banks));
-	this_cpu_write(bank_storm[bank], true);
-
-	/*
-	 * If this is the first bank on this CPU to enter storm mode
-	 * start polling
-	 */
-	if (this_cpu_inc_return(stormy_bank_count) == 1)
-		mce_timer_kick(true);
-}
-
-static void cmci_storm_end(int bank)
-{
-	__clear_bit(bank, this_cpu_ptr(mce_poll_banks));
-	this_cpu_write(bank_history[bank], 0ull);
-	this_cpu_write(bank_storm[bank], false);
-
-	/* If no banks left in storm mode, stop polling */
-	if (!this_cpu_dec_return(stormy_bank_count))
-		mce_timer_kick(false);
-}
-
-void track_cmci_storm(int bank, u64 status)
-{
-	unsigned long now = jiffies, delta;
-	unsigned int shift = 1;
-	u64 history;
-
-	/*
-	 * When a bank is in storm mode it is polled once per second and
-	 * the history mask will record about the last minute of poll results.
-	 * If it is not in storm mode, then the bank is only checked when
-	 * there is a CMCI interrupt. Check how long it has been since
-	 * this bank was last checked, and adjust the amount of "shift"
-	 * to apply to history.
-	 */
-	if (!this_cpu_read(bank_storm[bank])) {
-		delta = now - this_cpu_read(bank_time_stamp[bank]);
-		shift = (delta + HZ) / HZ;
-	}
-
-	/* If has been a long time since the last poll, clear history */
-	if (shift >= 64)
-		history = 0;
-	else
-		history = this_cpu_read(bank_history[bank]) << shift;
-	this_cpu_write(bank_time_stamp[bank], now);
-
-	/* History keeps track of corrected errors. VAL=1 && UC=0 */
-	if ((status & (MCI_STATUS_VAL | MCI_STATUS_UC)) == MCI_STATUS_VAL)
-		history |= 1;
-	this_cpu_write(bank_history[bank], history);
-
-	if (this_cpu_read(bank_storm[bank])) {
-		if (history & GENMASK_ULL(STORM_END_POLL_THRESHOLD - 1, 0))
-			return;
-		pr_notice("CPU%d BANK%d CMCI storm subsided\n", smp_processor_id(), bank);
-		mce_handle_storm(bank, true);
-		cmci_storm_end(bank);
-	} else {
-		if (hweight64(history) < STORM_BEGIN_THRESHOLD)
-			return;
-		pr_notice("CPU%d BANK%d CMCI storm detected\n", smp_processor_id(), bank);
-		mce_handle_storm(bank, false);
-		cmci_storm_begin(bank);
-	}
-}
-
 /*
  * The interrupt handler. This is called on every event.
  * Just call the poller directly to log any events.