[PATCH v3 0/6] powerpc/64s: interrupt speedups

[PATCH v3 0/6] powerpc/64s: interrupt speedups

[Nicholas Piggin]

Here's a few stragglers. The first patch was submitted already but had
some bugs with unrecoverable exceptions on HPT (current->blah being
accessed before MSR[RI] was enabled). Those should be fixed now.

The others are generally for helping asynch interrupts, which are a bit
harder to measure well but important for IO and IPIs.

After this series, the SPR accesses of the interrupt handlers for radix
are becoming pretty optimal except for PPR which we could improve on,
and virt CPU accounting which is very costly -- we might disable that
by default unless someone comes up with a good reason to keep it.

Since v1:
- Compile fixes for 64e.
- Fixed a SOFT_MASK_DEBUG false positive.
- Improve function name and comments explaining why patch 2 does not
  need to hard enable when PMU is enabled via sysfs.

Since v2:
- Split first patch into patch 1 and 2, improve on the changelogs.
- More compile fixes.
- Fixed several review comments from Daniel.
- Added patch 5.

Thanks,
Nick

Nicholas Piggin (6):
  powerpc/64/interrupt: make normal synchronous interrupts enable
    MSR[EE] if possible
  powerpc/64s/interrupt: handle MSR EE and RI in interrupt entry wrapper
  powerpc/64s/perf: add power_pmu_wants_prompt_pmi to say whether perf
    wants PMIs to be soft-NMI
  powerpc/64s/interrupt: Don't enable MSR[EE] in irq handlers unless
    perf is in use
  powerpc/64/interrupt: reduce expensive debug tests
  powerpc/64s/interrupt: avoid saving CFAR in some asynchronous
    interrupts

 arch/powerpc/include/asm/hw_irq.h    |  59 +++++++++++++---
 arch/powerpc/include/asm/interrupt.h |  58 ++++++++++++---
 arch/powerpc/kernel/dbell.c          |   3 +-
 arch/powerpc/kernel/exceptions-64s.S | 101 ++++++++++++++++++---------
 arch/powerpc/kernel/fpu.S            |   5 ++
 arch/powerpc/kernel/irq.c            |   3 +-
 arch/powerpc/kernel/time.c           |  31 ++++----
 arch/powerpc/kernel/vector.S         |  10 +++
 arch/powerpc/perf/core-book3s.c      |  31 ++++++++
 9 files changed, 232 insertions(+), 69 deletions(-)

-- 
2.23.0

[PATCH v3 1/6] powerpc/64/interrupt: make normal synchronous interrupts enable MSR[EE] if possible

[Nicholas Piggin]

Make synchronous interrupt handler entry wrappers enable MSR[EE] if
MSR[EE] was enabled in the interrupted context. IRQs are soft-disabled
at this point so there is no change to high level code, but it's a
masked interrupt could fire.

This is a performance disadvantage for interrupts which do not later
call interrupt_cond_local_irq_enable(), because an an additional mtmsrd
or wrtee instruction is executed. However the important synchronous
interrupts (e.g., page fault) do enable interrupts, so the performance
disadvantage is mostly avoided.

In the next patch, MSR[RI] enabling can be combined with MSR[EE]
enabling, which mitigates the performance drop for the former and gives
a performance advanage for the latter interrupts, on 64s machines. 64e
is coming along for the ride for now to avoid divergences with 64s in
this tricky code.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
 arch/powerpc/include/asm/interrupt.h | 19 ++++++++++++++++++-
 1 file changed, 18 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/include/asm/interrupt.h b/arch/powerpc/include/asm/interrupt.h
index b76ab848aa0d..3802390d8eea 100644
--- a/arch/powerpc/include/asm/interrupt.h
+++ b/arch/powerpc/include/asm/interrupt.h
@@ -150,7 +150,20 @@ static inline void interrupt_enter_prepare(struct pt_regs *regs, struct interrup
 #ifdef CONFIG_PPC64
 	if (irq_soft_mask_set_return(IRQS_ALL_DISABLED) == IRQS_ENABLED)
 		trace_hardirqs_off();
-	local_paca->irq_happened |= PACA_IRQ_HARD_DIS;
+
+	/*
+	 * If the interrupt was taken with HARD_DIS clear, then enable MSR[EE].
+	 * Asynchronous interrupts get here with HARD_DIS set (see below), so
+	 * this enables MSR[EE] for synchronous interrupts. IRQs remain
+	 * soft-masked. The interrupt handler may later call
+	 * interrupt_cond_local_irq_enable() to achieve a regular process
+	 * context.
+	 */
+	if (!(local_paca->irq_happened & PACA_IRQ_HARD_DIS)) {
+		if (IS_ENABLED(CONFIG_PPC_IRQ_SOFT_MASK_DEBUG))
+			BUG_ON(!(regs->msr & MSR_EE));
+		__hard_irq_enable();
+	}
 
 	if (user_mode(regs)) {
 		CT_WARN_ON(ct_state() != CONTEXT_USER);
@@ -200,6 +213,10 @@ static inline void interrupt_exit_prepare(struct pt_regs *regs, struct interrupt
 
 static inline void interrupt_async_enter_prepare(struct pt_regs *regs, struct interrupt_state *state)
 {
+#ifdef CONFIG_PPC64
+	/* Ensure interrupt_enter_prepare does not enable MSR[EE] */
+	local_paca->irq_happened |= PACA_IRQ_HARD_DIS;
+#endif
 #ifdef CONFIG_PPC_BOOK3S_64
 	if (cpu_has_feature(CPU_FTR_CTRL) &&
 	    !test_thread_local_flags(_TLF_RUNLATCH))
-- 
2.23.0

[PATCH v3 2/6] powerpc/64s/interrupt: handle MSR EE and RI in interrupt entry wrapper

[Nicholas Piggin]

The mtmsrd to enable MSR[RI] can be combined with the mtmsrd to enable
MSR[EE] in interrupt entry code, for those interrupts which enable EE.
This helps performance of important synchronous interrupts (e.g., page
faults).

This is similar to what commit dd152f70bdc1 ("powerpc/64s: system call
avoid setting MSR[RI] until we set MSR[EE]") does for system calls.

Do this by enabling EE and RI together at the beginning of the entry
wrapper if PACA_IRQ_HARD_DIS is clear, and only enabling RI if it is
set.

Asynchronous interrupts set PACA_IRQ_HARD_DIS, but synchronous ones
leave it unchanged, so by default they always get EE=1 unless they have
interrupted a caller that is hard disabled. When the sync interrupt
later calls interrupt_cond_local_irq_enable(), it will not require
another mtmsrd because MSR[EE] was already enabled here.

This avoids one mtmsrd L=1 for synchronous interrupts on 64s, which
saves about 20 cycles on POWER9. And for kernel-mode interrupts, both
synchronous and asynchronous, this saves an additional 40 cycles due to
the mtmsrd being moved ahead of mfspr SPRN_AMR, which prevents a SPR
scoreboard stall.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
 arch/powerpc/include/asm/interrupt.h | 27 +++++++++++++++++---
 arch/powerpc/kernel/exceptions-64s.S | 38 +++-------------------------
 arch/powerpc/kernel/fpu.S            |  5 ++++
 arch/powerpc/kernel/vector.S         | 10 ++++++++
 4 files changed, 42 insertions(+), 38 deletions(-)

diff --git a/arch/powerpc/include/asm/interrupt.h b/arch/powerpc/include/asm/interrupt.h
index 3802390d8eea..e178d143671a 100644
--- a/arch/powerpc/include/asm/interrupt.h
+++ b/arch/powerpc/include/asm/interrupt.h
@@ -148,8 +148,14 @@ static inline void interrupt_enter_prepare(struct pt_regs *regs, struct interrup
 #endif
 
 #ifdef CONFIG_PPC64
-	if (irq_soft_mask_set_return(IRQS_ALL_DISABLED) == IRQS_ENABLED)
-		trace_hardirqs_off();
+	bool trace_enable = false;
+
+	if (IS_ENABLED(CONFIG_TRACE_IRQFLAGS)) {
+		if (irq_soft_mask_set_return(IRQS_ALL_DISABLED) == IRQS_ENABLED)
+			trace_enable = true;
+	} else {
+		irq_soft_mask_set(IRQS_ALL_DISABLED);
+	}
 
 	/*
 	 * If the interrupt was taken with HARD_DIS clear, then enable MSR[EE].
@@ -163,8 +169,14 @@ static inline void interrupt_enter_prepare(struct pt_regs *regs, struct interrup
 		if (IS_ENABLED(CONFIG_PPC_IRQ_SOFT_MASK_DEBUG))
 			BUG_ON(!(regs->msr & MSR_EE));
 		__hard_irq_enable();
+	} else {
+		__hard_RI_enable();
 	}
 
+	/* Do this when RI=1 because it can cause SLB faults */
+	if (trace_enable)
+		trace_hardirqs_off();
+
 	if (user_mode(regs)) {
 		CT_WARN_ON(ct_state() != CONTEXT_USER);
 		user_exit_irqoff();
@@ -217,13 +229,16 @@ static inline void interrupt_async_enter_prepare(struct pt_regs *regs, struct in
 	/* Ensure interrupt_enter_prepare does not enable MSR[EE] */
 	local_paca->irq_happened |= PACA_IRQ_HARD_DIS;
 #endif
+	interrupt_enter_prepare(regs, state);
 #ifdef CONFIG_PPC_BOOK3S_64
+	/*
+	 * RI=1 is set by interrupt_enter_prepare, so this thread flags access
+	 * has to come afterward (it can cause SLB faults).
+	 */
 	if (cpu_has_feature(CPU_FTR_CTRL) &&
 	    !test_thread_local_flags(_TLF_RUNLATCH))
 		__ppc64_runlatch_on();
 #endif
-
-	interrupt_enter_prepare(regs, state);
 	irq_enter();
 }
 
@@ -293,6 +308,8 @@ static inline void interrupt_nmi_enter_prepare(struct pt_regs *regs, struct inte
 		regs->softe = IRQS_ALL_DISABLED;
 	}
 
+	__hard_RI_enable();
+
 	/* Don't do any per-CPU operations until interrupt state is fixed */
 
 	if (nmi_disables_ftrace(regs)) {
@@ -390,6 +407,8 @@ interrupt_handler long func(struct pt_regs *regs)			\
 {									\
 	long ret;							\
 									\
+	__hard_RI_enable();						\
+									\
 	ret = ____##func (regs);					\
 									\
 	return ret;							\
diff --git a/arch/powerpc/kernel/exceptions-64s.S b/arch/powerpc/kernel/exceptions-64s.S
index 37859e62a8dc..4dcc76206f8e 100644
--- a/arch/powerpc/kernel/exceptions-64s.S
+++ b/arch/powerpc/kernel/exceptions-64s.S
@@ -113,7 +113,6 @@ name:
 #define IISIDE		.L_IISIDE_\name\()	/* Uses SRR0/1 not DAR/DSISR */
 #define IDAR		.L_IDAR_\name\()	/* Uses DAR (or SRR0) */
 #define IDSISR		.L_IDSISR_\name\()	/* Uses DSISR (or SRR1) */
-#define ISET_RI		.L_ISET_RI_\name\()	/* Run common code w/ MSR[RI]=1 */
 #define IBRANCH_TO_COMMON	.L_IBRANCH_TO_COMMON_\name\() /* ENTRY branch to common */
 #define IREALMODE_COMMON	.L_IREALMODE_COMMON_\name\() /* Common runs in realmode */
 #define IMASK		.L_IMASK_\name\()	/* IRQ soft-mask bit */
@@ -157,9 +156,6 @@ do_define_int n
 	.ifndef IDSISR
 		IDSISR=0
 	.endif
-	.ifndef ISET_RI
-		ISET_RI=1
-	.endif
 	.ifndef IBRANCH_TO_COMMON
 		IBRANCH_TO_COMMON=1
 	.endif
@@ -512,11 +508,6 @@ DEFINE_FIXED_SYMBOL(\name\()_common_real)
 	stb	r10,PACASRR_VALID(r13)
 	.endif
 
-	.if ISET_RI
-	li	r10,MSR_RI
-	mtmsrd	r10,1			/* Set MSR_RI */
-	.endif
-
 	.if ISTACK
 	.if IKUAP
 	kuap_save_amr_and_lock r9, r10, cr1, cr0
@@ -902,11 +893,6 @@ INT_DEFINE_BEGIN(system_reset)
 	IVEC=0x100
 	IAREA=PACA_EXNMI
 	IVIRT=0 /* no virt entry point */
-	/*
-	 * MSR_RI is not enabled, because PACA_EXNMI and nmi stack is
-	 * being used, so a nested NMI exception would corrupt it.
-	 */
-	ISET_RI=0
 	ISTACK=0
 	IKVM_REAL=1
 INT_DEFINE_END(system_reset)
@@ -979,16 +965,14 @@ TRAMP_REAL_BEGIN(system_reset_fwnmi)
 EXC_COMMON_BEGIN(system_reset_common)
 	__GEN_COMMON_ENTRY system_reset
 	/*
-	 * Increment paca->in_nmi then enable MSR_RI. SLB or MCE will be able
-	 * to recover, but nested NMI will notice in_nmi and not recover
-	 * because of the use of the NMI stack. in_nmi reentrancy is tested in
-	 * system_reset_exception.
+	 * Increment paca->in_nmi. When the interrupt entry wrapper later
+	 * enable MSR_RI, then SLB or MCE will be able to recover, but a nested
+	 * NMI will notice in_nmi and not recover because of the use of the NMI
+	 * stack. in_nmi reentrancy is tested in system_reset_exception.
 	 */
 	lhz	r10,PACA_IN_NMI(r13)
 	addi	r10,r10,1
 	sth	r10,PACA_IN_NMI(r13)
-	li	r10,MSR_RI
-	mtmsrd 	r10,1
 
 	mr	r10,r1
 	ld	r1,PACA_NMI_EMERG_SP(r13)
@@ -1062,12 +1046,6 @@ INT_DEFINE_BEGIN(machine_check_early)
 	IAREA=PACA_EXMC
 	IVIRT=0 /* no virt entry point */
 	IREALMODE_COMMON=1
-	/*
-	 * MSR_RI is not enabled, because PACA_EXMC is being used, so a
-	 * nested machine check corrupts it. machine_check_common enables
-	 * MSR_RI.
-	 */
-	ISET_RI=0
 	ISTACK=0
 	IDAR=1
 	IDSISR=1
@@ -1078,7 +1056,6 @@ INT_DEFINE_BEGIN(machine_check)
 	IVEC=0x200
 	IAREA=PACA_EXMC
 	IVIRT=0 /* no virt entry point */
-	ISET_RI=0
 	IDAR=1
 	IDSISR=1
 	IKVM_REAL=1
@@ -1148,9 +1125,6 @@ EXC_COMMON_BEGIN(machine_check_early_common)
 BEGIN_FTR_SECTION
 	bl	enable_machine_check
 END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
-	li	r10,MSR_RI
-	mtmsrd	r10,1
-
 	addi	r3,r1,STACK_FRAME_OVERHEAD
 	bl	machine_check_early
 	std	r3,RESULT(r1)	/* Save result */
@@ -1238,10 +1212,6 @@ EXC_COMMON_BEGIN(machine_check_common)
 	 * save area: PACA_EXMC instead of PACA_EXGEN.
 	 */
 	GEN_COMMON machine_check
-
-	/* Enable MSR_RI when finished with PACA_EXMC */
-	li	r10,MSR_RI
-	mtmsrd 	r10,1
 	addi	r3,r1,STACK_FRAME_OVERHEAD
 	bl	machine_check_exception
 	b	interrupt_return_srr
diff --git a/arch/powerpc/kernel/fpu.S b/arch/powerpc/kernel/fpu.S
index ba4afe3b5a9c..f71f2bbd4de6 100644
--- a/arch/powerpc/kernel/fpu.S
+++ b/arch/powerpc/kernel/fpu.S
@@ -81,7 +81,12 @@ EXPORT_SYMBOL(store_fp_state)
  */
 _GLOBAL(load_up_fpu)
 	mfmsr	r5
+#ifdef CONFIG_PPC_BOOK3S_64
+	/* interrupt doesn't set MSR[RI] and HPT can fault on current access */
+	ori	r5,r5,MSR_FP|MSR_RI
+#else
 	ori	r5,r5,MSR_FP
+#endif
 #ifdef CONFIG_VSX
 BEGIN_FTR_SECTION
 	oris	r5,r5,MSR_VSX@h
diff --git a/arch/powerpc/kernel/vector.S b/arch/powerpc/kernel/vector.S
index ba03eedfdcd8..5cc24d8cce94 100644
--- a/arch/powerpc/kernel/vector.S
+++ b/arch/powerpc/kernel/vector.S
@@ -47,6 +47,10 @@ EXPORT_SYMBOL(store_vr_state)
  */
 _GLOBAL(load_up_altivec)
 	mfmsr	r5			/* grab the current MSR */
+#ifdef CONFIG_PPC_BOOK3S_64
+	/* interrupt doesn't set MSR[RI] and HPT can fault on current access */
+	ori	r5,r5,MSR_RI
+#endif
 	oris	r5,r5,MSR_VEC@h
 	MTMSRD(r5)			/* enable use of AltiVec now */
 	isync
@@ -126,6 +130,12 @@ _GLOBAL(load_up_vsx)
 	andis.	r5,r12,MSR_VEC@h
 	beql+	load_up_altivec		/* skip if already loaded */
 
+#ifdef CONFIG_PPC_BOOK3S_64
+	/* interrupt doesn't set MSR[RI] and HPT can fault on current access */
+	li	r5,MSR_RI
+	mtmsrd	r5,1
+#endif
+
 	ld	r4,PACACURRENT(r13)
 	addi	r4,r4,THREAD		/* Get THREAD */
 	li	r6,1
-- 
2.23.0

[PATCH v3 3/6] powerpc/64s/perf: add power_pmu_wants_prompt_pmi to say whether perf wants PMIs to be soft-NMI

[Nicholas Piggin]

Interrupt code enables MSR[EE] in some irq handlers while keeping local
irqs disabled via soft-mask, allowing PMI interrupts to be taken as
soft-NMI to improve profiling of irq handlers.

When perf is not enabled, there is no point to doing this, it's
additional overhead. So provide a function that can say if PMIs should
be taken promptly if possible.

Cc: Madhavan Srinivasan <maddy@linux.ibm.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
 arch/powerpc/include/asm/hw_irq.h |  2 ++
 arch/powerpc/perf/core-book3s.c   | 31 +++++++++++++++++++++++++++++++
 2 files changed, 33 insertions(+)

diff --git a/arch/powerpc/include/asm/hw_irq.h b/arch/powerpc/include/asm/hw_irq.h
index 21cc571ea9c2..b987822e552e 100644
--- a/arch/powerpc/include/asm/hw_irq.h
+++ b/arch/powerpc/include/asm/hw_irq.h
@@ -306,6 +306,8 @@ static inline bool lazy_irq_pending_nocheck(void)
 	return __lazy_irq_pending(local_paca->irq_happened);
 }
 
+bool power_pmu_wants_prompt_pmi(void);
+
 /*
  * This is called by asynchronous interrupts to conditionally
  * re-enable hard interrupts after having cleared the source
diff --git a/arch/powerpc/perf/core-book3s.c b/arch/powerpc/perf/core-book3s.c
index 73e62e9b179b..773d07d68b6d 100644
--- a/arch/powerpc/perf/core-book3s.c
+++ b/arch/powerpc/perf/core-book3s.c
@@ -17,6 +17,7 @@
 #include <asm/firmware.h>
 #include <asm/ptrace.h>
 #include <asm/code-patching.h>
+#include <asm/hw_irq.h>
 #include <asm/interrupt.h>
 
 #ifdef CONFIG_PPC64
@@ -2381,6 +2382,36 @@ static void perf_event_interrupt(struct pt_regs *regs)
 	perf_sample_event_took(sched_clock() - start_clock);
 }
 
+/*
+ * If the perf subsystem wants performance monitor interrupts as soon as
+ * possible (e.g., to sample the instruction address and stack chain),
+ * this should return true. The IRQ masking code can then enable MSR[EE]
+ * in some places (e.g., interrupt handlers) that allows PMI interrupts
+ * though to improve accuracy of profiles, at the cost of some performance.
+ *
+ * The PMU counters can be enabled by other means (e.g., sysfs raw SPR
+ * access), but in that case there is no need for prompt PMI handling.
+ *
+ * This currently returns true if any perf counter is being used. It
+ * could possibly return false if only events are being counted rather than
+ * samples being taken, but for now this is good enough.
+ */
+bool power_pmu_wants_prompt_pmi(void)
+{
+	struct cpu_hw_events *cpuhw;
+
+	/*
+	 * This could simply test local_paca->pmcregs_in_use if that were not
+	 * under ifdef KVM.
+	 */
+
+	if (!ppmu)
+		return false;
+
+	cpuhw = this_cpu_ptr(&cpu_hw_events);
+	return cpuhw->n_events;
+}
+
 static int power_pmu_prepare_cpu(unsigned int cpu)
 {
 	struct cpu_hw_events *cpuhw = &per_cpu(cpu_hw_events, cpu);
-- 
2.23.0

[PATCH v3 4/6] powerpc/64s/interrupt: Don't enable MSR[EE] in irq handlers unless perf is in use

[Nicholas Piggin]

Enabling MSR[EE] in interrupt handlers while interrupts are still soft
masked allows PMIs to profile interrupt handlers to some degree, beyond
what SIAR latching allows.

When perf is not being used, this is almost useless work. It requires an
extra mtmsrd in the irq handler, and it also opens the door to masked
interrupts hitting and requiring replay, which is more expensive than
just taking them directly. This effect can be noticable in high IRQ
workloads.

Avoid enabling MSR[EE] unless perf is currently in use. This saves about
60 cycles (or 8%) on a simple decrementer interrupt microbenchmark.
Replayed interrupts drop from 1.4% of all interrupts taken, to 0.003%.

This does prevent the soft-nmi interrupt being taken in these handlers,
but that's not too reliable anyway. The SMP watchdog will continue to be
the reliable way to catch lockups.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
 arch/powerpc/include/asm/hw_irq.h | 57 +++++++++++++++++++++++++------
 arch/powerpc/kernel/dbell.c       |  3 +-
 arch/powerpc/kernel/irq.c         |  3 +-
 arch/powerpc/kernel/time.c        | 31 +++++++++--------
 4 files changed, 67 insertions(+), 27 deletions(-)

diff --git a/arch/powerpc/include/asm/hw_irq.h b/arch/powerpc/include/asm/hw_irq.h
index b987822e552e..55e3fa44f280 100644
--- a/arch/powerpc/include/asm/hw_irq.h
+++ b/arch/powerpc/include/asm/hw_irq.h
@@ -309,17 +309,54 @@ static inline bool lazy_irq_pending_nocheck(void)
 bool power_pmu_wants_prompt_pmi(void);
 
 /*
- * This is called by asynchronous interrupts to conditionally
- * re-enable hard interrupts after having cleared the source
- * of the interrupt. They are kept disabled if there is a different
- * soft-masked interrupt pending that requires hard masking.
+ * This is called by asynchronous interrupts to check whether to
+ * conditionally re-enable hard interrupts after having cleared
+ * the source of the interrupt. They are kept disabled if there
+ * is a different soft-masked interrupt pending that requires hard
+ * masking.
  */
-static inline void may_hard_irq_enable(void)
+static inline bool should_hard_irq_enable(void)
 {
-	if (!(get_paca()->irq_happened & PACA_IRQ_MUST_HARD_MASK)) {
-		get_paca()->irq_happened &= ~PACA_IRQ_HARD_DIS;
-		__hard_irq_enable();
-	}
+#ifdef CONFIG_PPC_IRQ_SOFT_MASK_DEBUG
+	WARN_ON(irq_soft_mask_return() == IRQS_ENABLED);
+	WARN_ON(mfmsr() & MSR_EE);
+#endif
+#ifdef CONFIG_PERF_EVENTS
+	/*
+	 * If the PMU is not running, there is not much reason to enable
+	 * MSR[EE] in irq handlers because any interrupts would just be
+	 * soft-masked.
+	 *
+	 * TODO: Add test for 64e
+	 */
+	if (IS_ENABLED(CONFIG_PPC_BOOK3S_64) && !power_pmu_wants_prompt_pmi())
+		return false;
+
+	if (get_paca()->irq_happened & PACA_IRQ_MUST_HARD_MASK)
+		return false;
+
+	return true;
+#else
+	return false;
+#endif
+}
+
+/*
+ * Do the hard enabling, only call this if should_hard_irq_enable is true.
+ */
+static inline void do_hard_irq_enable(void)
+{
+#ifdef CONFIG_PPC_IRQ_SOFT_MASK_DEBUG
+	WARN_ON(irq_soft_mask_return() == IRQS_ENABLED);
+	WARN_ON(get_paca()->irq_happened & PACA_IRQ_MUST_HARD_MASK);
+	WARN_ON(mfmsr() & MSR_EE);
+#endif
+	/*
+	 * This allows PMI interrupts (and watchdog soft-NMIs) through.
+	 * There is no other reason to enable this way.
+	 */
+	get_paca()->irq_happened &= ~PACA_IRQ_HARD_DIS;
+	__hard_irq_enable();
 }
 
 static inline bool arch_irq_disabled_regs(struct pt_regs *regs)
@@ -400,7 +437,7 @@ static inline bool arch_irq_disabled_regs(struct pt_regs *regs)
 	return !(regs->msr & MSR_EE);
 }
 
-static inline bool may_hard_irq_enable(void)
+static inline bool should_hard_irq_enable(void)
 {
 	return false;
 }
diff --git a/arch/powerpc/kernel/dbell.c b/arch/powerpc/kernel/dbell.c
index 5545c9cd17c1..f55c6fb34a3a 100644
--- a/arch/powerpc/kernel/dbell.c
+++ b/arch/powerpc/kernel/dbell.c
@@ -27,7 +27,8 @@ DEFINE_INTERRUPT_HANDLER_ASYNC(doorbell_exception)
 
 	ppc_msgsync();
 
-	may_hard_irq_enable();
+	if (should_hard_irq_enable())
+		do_hard_irq_enable();
 
 	kvmppc_clear_host_ipi(smp_processor_id());
 	__this_cpu_inc(irq_stat.doorbell_irqs);
diff --git a/arch/powerpc/kernel/irq.c b/arch/powerpc/kernel/irq.c
index 551b653228c4..f658aa22a21e 100644
--- a/arch/powerpc/kernel/irq.c
+++ b/arch/powerpc/kernel/irq.c
@@ -739,7 +739,8 @@ void __do_irq(struct pt_regs *regs)
 	irq = ppc_md.get_irq();
 
 	/* We can hard enable interrupts now to allow perf interrupts */
-	may_hard_irq_enable();
+	if (should_hard_irq_enable())
+		do_hard_irq_enable();
 
 	/* And finally process it */
 	if (unlikely(!irq))
diff --git a/arch/powerpc/kernel/time.c b/arch/powerpc/kernel/time.c
index 934d8ae66cc6..c3c663ff7c48 100644
--- a/arch/powerpc/kernel/time.c
+++ b/arch/powerpc/kernel/time.c
@@ -566,22 +566,23 @@ DEFINE_INTERRUPT_HANDLER_ASYNC(timer_interrupt)
 		return;
 	}
 
-	/* Ensure a positive value is written to the decrementer, or else
-	 * some CPUs will continue to take decrementer exceptions. When the
-	 * PPC_WATCHDOG (decrementer based) is configured, keep this at most
-	 * 31 bits, which is about 4 seconds on most systems, which gives
-	 * the watchdog a chance of catching timer interrupt hard lockups.
-	 */
-	if (IS_ENABLED(CONFIG_PPC_WATCHDOG))
-		set_dec(0x7fffffff);
-	else
-		set_dec(decrementer_max);
-
-	/* Conditionally hard-enable interrupts now that the DEC has been
-	 * bumped to its maximum value
-	 */
-	may_hard_irq_enable();
+	/* Conditionally hard-enable interrupts. */
+	if (should_hard_irq_enable()) {
+		/*
+		 * Ensure a positive value is written to the decrementer, or
+		 * else some CPUs will continue to take decrementer exceptions.
+		 * When the PPC_WATCHDOG (decrementer based) is configured,
+		 * keep this at most 31 bits, which is about 4 seconds on most
+		 * systems, which gives the watchdog a chance of catching timer
+		 * interrupt hard lockups.
+		 */
+		if (IS_ENABLED(CONFIG_PPC_WATCHDOG))
+			set_dec(0x7fffffff);
+		else
+			set_dec(decrementer_max);
 
+		do_hard_irq_enable();
+	}
 
 #if defined(CONFIG_PPC32) && defined(CONFIG_PPC_PMAC)
 	if (atomic_read(&ppc_n_lost_interrupts) != 0)
-- 
2.23.0

[PATCH v3 5/6] powerpc/64/interrupt: reduce expensive debug tests

[Nicholas Piggin]

Move the assertions requiring restart table searches under
CONFIG_PPC_IRQ_SOFT_MASK_DEBUG.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
 arch/powerpc/include/asm/interrupt.h | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

diff --git a/arch/powerpc/include/asm/interrupt.h b/arch/powerpc/include/asm/interrupt.h
index e178d143671a..0e84e99af37b 100644
--- a/arch/powerpc/include/asm/interrupt.h
+++ b/arch/powerpc/include/asm/interrupt.h
@@ -97,6 +97,11 @@ static inline void srr_regs_clobbered(void)
 	local_paca->hsrr_valid = 0;
 }
 #else
+static inline unsigned long search_kernel_restart_table(unsigned long addr)
+{
+	return 0;
+}
+
 static inline bool is_implicit_soft_masked(struct pt_regs *regs)
 {
 	return false;
@@ -190,13 +195,14 @@ static inline void interrupt_enter_prepare(struct pt_regs *regs, struct interrup
 		 */
 		if (TRAP(regs) != INTERRUPT_PROGRAM) {
 			CT_WARN_ON(ct_state() != CONTEXT_KERNEL);
-			BUG_ON(is_implicit_soft_masked(regs));
+			if (IS_ENABLED(CONFIG_PPC_IRQ_SOFT_MASK_DEBUG))
+				BUG_ON(is_implicit_soft_masked(regs));
 		}
-#ifdef CONFIG_PPC_BOOK3S
+
 		/* Move this under a debugging check */
-		if (arch_irq_disabled_regs(regs))
+		if (IS_ENABLED(CONFIG_PPC_IRQ_SOFT_MASK_DEBUG) &&
+				arch_irq_disabled_regs(regs))
 			BUG_ON(search_kernel_restart_table(regs->nip));
-#endif
 	}
 	if (IS_ENABLED(CONFIG_PPC_IRQ_SOFT_MASK_DEBUG))
 		BUG_ON(!arch_irq_disabled_regs(regs) && !(regs->msr & MSR_EE));
-- 
2.23.0

[PATCH v3 6/6] powerpc/64s/interrupt: avoid saving CFAR in some asynchronous interrupts

[Nicholas Piggin]

Reading the CFAR register is quite costly (~20 cycles on POWER9). It is
a good idea to have for most synchronous interrupts, but for async ones
it is much less important.

Doorbell, external, and decrementer interrupts are the important
asynchronous ones. HV interrupts can't skip CFAR if KVM HV is possible,
because it might be a guest exit that requires CFAR preserved. But the
important pseries interrupts can avoid loading CFAR.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
 arch/powerpc/kernel/exceptions-64s.S | 63 ++++++++++++++++++++++++++++
 1 file changed, 63 insertions(+)

diff --git a/arch/powerpc/kernel/exceptions-64s.S b/arch/powerpc/kernel/exceptions-64s.S
index 4dcc76206f8e..eb3af5abdd37 100644
--- a/arch/powerpc/kernel/exceptions-64s.S
+++ b/arch/powerpc/kernel/exceptions-64s.S
@@ -111,6 +111,8 @@ name:
 #define IAREA		.L_IAREA_\name\()	/* PACA save area */
 #define IVIRT		.L_IVIRT_\name\()	/* Has virt mode entry point */
 #define IISIDE		.L_IISIDE_\name\()	/* Uses SRR0/1 not DAR/DSISR */
+#define ICFAR		.L_ICFAR_\name\()	/* Uses CFAR */
+#define ICFAR_IF_HVMODE	.L_ICFAR_IF_HVMODE_\name\() /* Uses CFAR if HV */
 #define IDAR		.L_IDAR_\name\()	/* Uses DAR (or SRR0) */
 #define IDSISR		.L_IDSISR_\name\()	/* Uses DSISR (or SRR1) */
 #define IBRANCH_TO_COMMON	.L_IBRANCH_TO_COMMON_\name\() /* ENTRY branch to common */
@@ -150,6 +152,12 @@ do_define_int n
 	.ifndef IISIDE
 		IISIDE=0
 	.endif
+	.ifndef ICFAR
+		ICFAR=1
+	.endif
+	.ifndef ICFAR_IF_HVMODE
+		ICFAR_IF_HVMODE=0
+	.endif
 	.ifndef IDAR
 		IDAR=0
 	.endif
@@ -287,9 +295,21 @@ BEGIN_FTR_SECTION
 END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
 	HMT_MEDIUM
 	std	r10,IAREA+EX_R10(r13)		/* save r10 - r12 */
+	.if ICFAR
 BEGIN_FTR_SECTION
 	mfspr	r10,SPRN_CFAR
 END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
+	.elseif ICFAR_IF_HVMODE
+BEGIN_FTR_SECTION
+  BEGIN_FTR_SECTION_NESTED(69)
+	mfspr	r10,SPRN_CFAR
+  END_FTR_SECTION_NESTED(CPU_FTR_CFAR, CPU_FTR_CFAR, 69)
+FTR_SECTION_ELSE
+  BEGIN_FTR_SECTION_NESTED(69)
+	li	r10,0
+  END_FTR_SECTION_NESTED(CPU_FTR_CFAR, CPU_FTR_CFAR, 69)
+ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
+	.endif
 	.if \ool
 	.if !\virt
 	b	tramp_real_\name
@@ -305,9 +325,11 @@ END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
 BEGIN_FTR_SECTION
 	std	r9,IAREA+EX_PPR(r13)
 END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
+	.if ICFAR || ICFAR_IF_HVMODE
 BEGIN_FTR_SECTION
 	std	r10,IAREA+EX_CFAR(r13)
 END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
+	.endif
 	INTERRUPT_TO_KERNEL
 	mfctr	r10
 	std	r10,IAREA+EX_CTR(r13)
@@ -559,7 +581,11 @@ END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
 	.endif
 
 BEGIN_FTR_SECTION
+	.if ICFAR || ICFAR_IF_HVMODE
 	ld	r10,IAREA+EX_CFAR(r13)
+	.else
+	li	r10,0
+	.endif
 	std	r10,ORIG_GPR3(r1)
 END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
 	ld	r10,IAREA+EX_CTR(r13)
@@ -1502,6 +1528,12 @@ ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
  *
  * If soft masked, the masked handler will note the pending interrupt for
  * replay, and clear MSR[EE] in the interrupted context.
+ *
+ * CFAR is not required because this is an asynchronous interrupt that in
+ * general won't have much bearing on the state of the CPU, with the possible
+ * exception of crash/debug IPIs, but those are generally moving to use SRESET
+ * IPIs. Unless this is an HV interrupt and KVM HV is possible, in which case
+ * it may be exiting the guest and need CFAR to be saved.
  */
 INT_DEFINE_BEGIN(hardware_interrupt)
 	IVEC=0x500
@@ -1509,6 +1541,10 @@ INT_DEFINE_BEGIN(hardware_interrupt)
 	IMASK=IRQS_DISABLED
 	IKVM_REAL=1
 	IKVM_VIRT=1
+	ICFAR=0
+#ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
+	ICFAR_IF_HVMODE=1
+#endif
 INT_DEFINE_END(hardware_interrupt)
 
 EXC_REAL_BEGIN(hardware_interrupt, 0x500, 0x100)
@@ -1727,6 +1763,10 @@ END_FTR_SECTION_IFSET(CPU_FTR_TM)
  * If PPC_WATCHDOG is configured, the soft masked handler will actually set
  * things back up to run soft_nmi_interrupt as a regular interrupt handler
  * on the emergency stack.
+ *
+ * CFAR is not required because this is asynchronous (see hardware_interrupt).
+ * A watchdog interrupt may like to have CFAR, but usually the interesting
+ * branch is long gone by that point (e.g., infinite loop).
  */
 INT_DEFINE_BEGIN(decrementer)
 	IVEC=0x900
@@ -1734,6 +1774,7 @@ INT_DEFINE_BEGIN(decrementer)
 #ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
 	IKVM_REAL=1
 #endif
+	ICFAR=0
 INT_DEFINE_END(decrementer)
 
 EXC_REAL_BEGIN(decrementer, 0x900, 0x80)
@@ -1809,6 +1850,8 @@ EXC_COMMON_BEGIN(hdecrementer_common)
  * If soft masked, the masked handler will note the pending interrupt for
  * replay, leaving MSR[EE] enabled in the interrupted context because the
  * doorbells are edge triggered.
+ *
+ * CFAR is not required, similarly to hardware_interrupt.
  */
 INT_DEFINE_BEGIN(doorbell_super)
 	IVEC=0xa00
@@ -1816,6 +1859,7 @@ INT_DEFINE_BEGIN(doorbell_super)
 #ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
 	IKVM_REAL=1
 #endif
+	ICFAR=0
 INT_DEFINE_END(doorbell_super)
 
 EXC_REAL_BEGIN(doorbell_super, 0xa00, 0x100)
@@ -1867,6 +1911,7 @@ INT_DEFINE_BEGIN(system_call)
 	IVEC=0xc00
 	IKVM_REAL=1
 	IKVM_VIRT=1
+	ICFAR=0
 INT_DEFINE_END(system_call)
 
 .macro SYSTEM_CALL virt
@@ -2165,6 +2210,11 @@ EXC_COMMON_BEGIN(hmi_exception_common)
  * Interrupt 0xe80 - Directed Hypervisor Doorbell Interrupt.
  * This is an asynchronous interrupt in response to a msgsnd doorbell.
  * Similar to the 0xa00 doorbell but for host rather than guest.
+ *
+ * CFAR is not required (similar to doorbell_interrupt), unless KVM HV
+ * is enabled, in which case it may be a guest exit. Most PowerNV kernels
+ * include KVM support so it would be nice if this could be dynamically
+ * patched out if KVM was not currently running any guests.
  */
 INT_DEFINE_BEGIN(h_doorbell)
 	IVEC=0xe80
@@ -2172,6 +2222,9 @@ INT_DEFINE_BEGIN(h_doorbell)
 	IMASK=IRQS_DISABLED
 	IKVM_REAL=1
 	IKVM_VIRT=1
+#ifndef CONFIG_KVM_BOOK3S_HV_POSSIBLE
+	ICFAR=0
+#endif
 INT_DEFINE_END(h_doorbell)
 
 EXC_REAL_BEGIN(h_doorbell, 0xe80, 0x20)
@@ -2195,6 +2248,9 @@ EXC_COMMON_BEGIN(h_doorbell_common)
  * Interrupt 0xea0 - Hypervisor Virtualization Interrupt.
  * This is an asynchronous interrupt in response to an "external exception".
  * Similar to 0x500 but for host only.
+ *
+ * Like h_doorbell, CFAR is only required for KVM HV because this can be
+ * a guest exit.
  */
 INT_DEFINE_BEGIN(h_virt_irq)
 	IVEC=0xea0
@@ -2202,6 +2258,9 @@ INT_DEFINE_BEGIN(h_virt_irq)
 	IMASK=IRQS_DISABLED
 	IKVM_REAL=1
 	IKVM_VIRT=1
+#ifndef CONFIG_KVM_BOOK3S_HV_POSSIBLE
+	ICFAR=0
+#endif
 INT_DEFINE_END(h_virt_irq)
 
 EXC_REAL_BEGIN(h_virt_irq, 0xea0, 0x20)
@@ -2238,6 +2297,8 @@ EXC_VIRT_NONE(0x4ee0, 0x20)
  *
  * If soft masked, the masked handler will note the pending interrupt for
  * replay, and clear MSR[EE] in the interrupted context.
+ *
+ * CFAR is not used by perf interrupts so not required.
  */
 INT_DEFINE_BEGIN(performance_monitor)
 	IVEC=0xf00
@@ -2245,6 +2306,7 @@ INT_DEFINE_BEGIN(performance_monitor)
 #ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
 	IKVM_REAL=1
 #endif
+	ICFAR=0
 INT_DEFINE_END(performance_monitor)
 
 EXC_REAL_BEGIN(performance_monitor, 0xf00, 0x20)
@@ -2669,6 +2731,7 @@ EXC_VIRT_NONE(0x5800, 0x100)
 INT_DEFINE_BEGIN(soft_nmi)
 	IVEC=0x900
 	ISTACK=0
+	ICFAR=0
 INT_DEFINE_END(soft_nmi)
 
 /*
-- 
2.23.0

Re: [PATCH v3 1/6] powerpc/64/interrupt: make normal synchronous interrupts enable MSR[EE] if possible

[Christophe Leroy]

Le 22/09/2021 à 16:54, Nicholas Piggin a écrit :
> Make synchronous interrupt handler entry wrappers enable MSR[EE] if
> MSR[EE] was enabled in the interrupted context. IRQs are soft-disabled
> at this point so there is no change to high level code, but it's a
> masked interrupt could fire.
> 
> This is a performance disadvantage for interrupts which do not later
> call interrupt_cond_local_irq_enable(), because an an additional mtmsrd
> or wrtee instruction is executed. However the important synchronous
> interrupts (e.g., page fault) do enable interrupts, so the performance
> disadvantage is mostly avoided.
> 
> In the next patch, MSR[RI] enabling can be combined with MSR[EE]
> enabling, which mitigates the performance drop for the former and gives
> a performance advanage for the latter interrupts, on 64s machines. 64e
> is coming along for the ride for now to avoid divergences with 64s in
> this tricky code.
> 
> Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
> ---
>   arch/powerpc/include/asm/interrupt.h | 19 ++++++++++++++++++-
>   1 file changed, 18 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/powerpc/include/asm/interrupt.h b/arch/powerpc/include/asm/interrupt.h
> index b76ab848aa0d..3802390d8eea 100644
> --- a/arch/powerpc/include/asm/interrupt.h
> +++ b/arch/powerpc/include/asm/interrupt.h
> @@ -150,7 +150,20 @@ static inline void interrupt_enter_prepare(struct pt_regs *regs, struct interrup
>   #ifdef CONFIG_PPC64
>   	if (irq_soft_mask_set_return(IRQS_ALL_DISABLED) == IRQS_ENABLED)
>   		trace_hardirqs_off();
> -	local_paca->irq_happened |= PACA_IRQ_HARD_DIS;
> +
> +	/*
> +	 * If the interrupt was taken with HARD_DIS clear, then enable MSR[EE].
> +	 * Asynchronous interrupts get here with HARD_DIS set (see below), so
> +	 * this enables MSR[EE] for synchronous interrupts. IRQs remain
> +	 * soft-masked. The interrupt handler may later call
> +	 * interrupt_cond_local_irq_enable() to achieve a regular process
> +	 * context.
> +	 */
> +	if (!(local_paca->irq_happened & PACA_IRQ_HARD_DIS)) {
> +		if (IS_ENABLED(CONFIG_PPC_IRQ_SOFT_MASK_DEBUG))
> +			BUG_ON(!(regs->msr & MSR_EE));

Could be:
	BUG_ON(IS_ENABLED(CONFIG_PPC_IRQ_SOFT_MASK_DEBUG) && !(regs->msr & 
MSR_EE));

> +		__hard_irq_enable();
> +	}
>   
>   	if (user_mode(regs)) {
>   		CT_WARN_ON(ct_state() != CONTEXT_USER);
> @@ -200,6 +213,10 @@ static inline void interrupt_exit_prepare(struct pt_regs *regs, struct interrupt
>   
>   static inline void interrupt_async_enter_prepare(struct pt_regs *regs, struct interrupt_state *state)
>   {
> +#ifdef CONFIG_PPC64
> +	/* Ensure interrupt_enter_prepare does not enable MSR[EE] */
> +	local_paca->irq_happened |= PACA_IRQ_HARD_DIS;
> +#endif
>   #ifdef CONFIG_PPC_BOOK3S_64
>   	if (cpu_has_feature(CPU_FTR_CTRL) &&
>   	    !test_thread_local_flags(_TLF_RUNLATCH))
> 

Re: [PATCH v3 5/6] powerpc/64/interrupt: reduce expensive debug tests

[Christophe Leroy]

Le 22/09/2021 à 16:54, Nicholas Piggin a écrit :
> Move the assertions requiring restart table searches under
> CONFIG_PPC_IRQ_SOFT_MASK_DEBUG.
> 
> Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
> ---
>   arch/powerpc/include/asm/interrupt.h | 14 ++++++++++----
>   1 file changed, 10 insertions(+), 4 deletions(-)
> 
> diff --git a/arch/powerpc/include/asm/interrupt.h b/arch/powerpc/include/asm/interrupt.h
> index e178d143671a..0e84e99af37b 100644
> --- a/arch/powerpc/include/asm/interrupt.h
> +++ b/arch/powerpc/include/asm/interrupt.h
> @@ -97,6 +97,11 @@ static inline void srr_regs_clobbered(void)
>   	local_paca->hsrr_valid = 0;
>   }
>   #else
> +static inline unsigned long search_kernel_restart_table(unsigned long addr)
> +{
> +	return 0;
> +}
> +

Not sure you need that. Moving the 64s prototype out of the #ifdef would 
do it as well.


>   static inline bool is_implicit_soft_masked(struct pt_regs *regs)
>   {
>   	return false;
> @@ -190,13 +195,14 @@ static inline void interrupt_enter_prepare(struct pt_regs *regs, struct interrup
>   		 */
>   		if (TRAP(regs) != INTERRUPT_PROGRAM) {
>   			CT_WARN_ON(ct_state() != CONTEXT_KERNEL);
> -			BUG_ON(is_implicit_soft_masked(regs));
> +			if (IS_ENABLED(CONFIG_PPC_IRQ_SOFT_MASK_DEBUG))
> +				BUG_ON(is_implicit_soft_masked(regs));
>   		}
> -#ifdef CONFIG_PPC_BOOK3S
> +
>   		/* Move this under a debugging check */
> -		if (arch_irq_disabled_regs(regs))
> +		if (IS_ENABLED(CONFIG_PPC_IRQ_SOFT_MASK_DEBUG) &&
> +				arch_irq_disabled_regs(regs))
>   			BUG_ON(search_kernel_restart_table(regs->nip));
> -#endif
>   	}
>   	if (IS_ENABLED(CONFIG_PPC_IRQ_SOFT_MASK_DEBUG))
>   		BUG_ON(!arch_irq_disabled_regs(regs) && !(regs->msr & MSR_EE));
> 

Re: [PATCH v3 0/6] powerpc/64s: interrupt speedups

[Michael Ellerman]

On Thu, 23 Sep 2021 00:54:46 +1000, Nicholas Piggin wrote:
> Here's a few stragglers. The first patch was submitted already but had
> some bugs with unrecoverable exceptions on HPT (current->blah being
> accessed before MSR[RI] was enabled). Those should be fixed now.
> 
> The others are generally for helping asynch interrupts, which are a bit
> harder to measure well but important for IO and IPIs.
> 
> [...]

Applied to powerpc/next.

[1/6] powerpc/64/interrupt: make normal synchronous interrupts enable MSR[EE] if possible
      https://git.kernel.org/powerpc/c/4423eb5ae32ec613af3fceee2fe84234e417ee55
[2/6] powerpc/64s/interrupt: handle MSR EE and RI in interrupt entry wrapper
      https://git.kernel.org/powerpc/c/ff0b0d6e1a7bc202241a9b1e28d1da4b744e0312
[3/6] powerpc/64s/perf: add power_pmu_wants_prompt_pmi to say whether perf wants PMIs to be soft-NMI
      https://git.kernel.org/powerpc/c/5a7745b96f43c69f9b4875bcf516a0341acbc3fb
[4/6] powerpc/64s/interrupt: Don't enable MSR[EE] in irq handlers unless perf is in use
      https://git.kernel.org/powerpc/c/0faf20a1ad1647c0fc0f5a367c71e5e84deaf899
[5/6] powerpc/64/interrupt: reduce expensive debug tests
      https://git.kernel.org/powerpc/c/ecb1057c0f9a0f3f052294de6cc2eb43ecf7547b
[6/6] powerpc/64s/interrupt: avoid saving CFAR in some asynchronous interrupts
      https://git.kernel.org/powerpc/c/af47d79b041deccc31e0dddc6310a654c13d04b6

cheers