[PATCH AUTOSEL 5.15 15/41] perf/x86/rapl: fix AMD event handling

[PATCH AUTOSEL 5.15 15/41] perf/x86/rapl: fix AMD event handling

[Sasha Levin]

From: Stephane Eranian <eranian@google.com>

[ Upstream commit 0036fb00a756a2f6e360d44e2e3d2200a8afbc9b ]

The RAPL events exposed under /sys/devices/power/events should only reflect
what the underlying hardware actually support. This is how it works on Intel
RAPL and Intel core/uncore PMUs in general.
But on AMD, this was not the case. All possible RAPL events were advertised.

This is what it showed on an AMD Fam17h:
$ ls /sys/devices/power/events/
energy-cores        energy-gpu          energy-pkg          energy-psys
energy-ram          energy-cores.scale  energy-gpu.scale    energy-pkg.scale
energy-psys.scale   energy-ram.scale    energy-cores.unit   energy-gpu.unit
energy-pkg.unit     energy-psys.unit    energy-ram.unit

Yet, on AMD Fam17h, only energy-pkg is supported.

This patch fixes the problem. Given the way perf_msr_probe() works, the
amd_rapl_msrs[] table has to have all entries filled out and in particular
the group field, otherwise perf_msr_probe() defaults to making the event
visible.

With the patch applied, the kernel now only shows was is actually supported:

$ ls /sys/devices/power/events/
energy-pkg  energy-pkg.scale  energy-pkg.unit

The patch also uses the RAPL_MSR_MASK because only the 32-bits LSB of the
RAPL counters are relevant when reading power consumption.

Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20220105185659.643355-1-eranian@google.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 arch/x86/events/rapl.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/arch/x86/events/rapl.c b/arch/x86/events/rapl.c
index 85feafacc445d..77e3a47af5ad5 100644
--- a/arch/x86/events/rapl.c
+++ b/arch/x86/events/rapl.c
@@ -536,11 +536,14 @@ static struct perf_msr intel_rapl_spr_msrs[] = {
  * - perf_msr_probe(PERF_RAPL_MAX)
  * - want to use same event codes across both architectures
  */
-static struct perf_msr amd_rapl_msrs[PERF_RAPL_MAX] = {
-	[PERF_RAPL_PKG]  = { MSR_AMD_PKG_ENERGY_STATUS,  &rapl_events_pkg_group,   test_msr },
+static struct perf_msr amd_rapl_msrs[] = {
+	[PERF_RAPL_PP0]  = { 0, &rapl_events_cores_group, 0, false, 0 },
+	[PERF_RAPL_PKG]  = { MSR_AMD_PKG_ENERGY_STATUS,  &rapl_events_pkg_group,   test_msr, false, RAPL_MSR_MASK },
+	[PERF_RAPL_RAM]  = { 0, &rapl_events_ram_group,   0, false, 0 },
+	[PERF_RAPL_PP1]  = { 0, &rapl_events_gpu_group,   0, false, 0 },
+	[PERF_RAPL_PSYS] = { 0, &rapl_events_psys_group,  0, false, 0 },
 };
 
-
 static int rapl_cpu_offline(unsigned int cpu)
 {
 	struct rapl_pmu *pmu = cpu_to_rapl_pmu(cpu);
-- 
2.34.1

[PATCH AUTOSEL 5.15 16/41] x86/perf: Avoid warning for Arch LBR without XSAVE

[Sasha Levin]

From: Andi Kleen <ak@linux.intel.com>

[ Upstream commit 8c16dc047b5dd8f7b3bf4584fa75733ea0dde7dc ]

Some hypervisors support Arch LBR, but without the LBR XSAVE support.
The current Arch LBR init code prints a warning when the xsave size (0) is
unexpected. Avoid printing the warning for the "no LBR XSAVE" case.

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20211215204029.150686-1-ak@linux.intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 arch/x86/events/intel/lbr.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/arch/x86/events/intel/lbr.c b/arch/x86/events/intel/lbr.c
index 9e6d6eaeb4cb6..f455dd93f9219 100644
--- a/arch/x86/events/intel/lbr.c
+++ b/arch/x86/events/intel/lbr.c
@@ -1734,6 +1734,9 @@ static bool is_arch_lbr_xsave_available(void)
 	 * Check the LBR state with the corresponding software structure.
 	 * Disable LBR XSAVES support if the size doesn't match.
 	 */
+	if (xfeature_size(XFEATURE_LBR) == 0)
+		return false;
+
 	if (WARN_ON(xfeature_size(XFEATURE_LBR) != get_lbr_state_size()))
 		return false;
 
-- 
2.34.1

[PATCH AUTOSEL 5.15 37/41] perf: Always wake the parent event

[Sasha Levin]

From: James Clark <james.clark@arm.com>

[ Upstream commit 961c39121759ad09a89598ec4ccdd34ae0468a19 ]

When using per-process mode and event inheritance is set to true,
forked processes will create a new perf events via inherit_event() ->
perf_event_alloc(). But these events will not have ring buffers
assigned to them. Any call to wakeup will be dropped if it's called on
an event with no ring buffer assigned because that's the object that
holds the wakeup list.

If the child event is disabled due to a call to
perf_aux_output_begin() or perf_aux_output_end(), the wakeup is
dropped leaving userspace hanging forever on the poll.

Normally the event is explicitly re-enabled by userspace after it
wakes up to read the aux data, but in this case it does not get woken
up so the event remains disabled.

This can be reproduced when using Arm SPE and 'stress' which forks once
before running the workload. By looking at the list of aux buffers read,
it's apparent that they stop after the fork:

  perf record -e arm_spe// -vvv -- stress -c 1

With this patch applied they continue to be printed. This behaviour
doesn't happen when using systemwide or per-cpu mode.

Reported-by: Ruben Ayrapetyan <Ruben.Ayrapetyan@arm.com>
Signed-off-by: James Clark <james.clark@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20211206113840.130802-2-james.clark@arm.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 kernel/events/core.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index 0fe6a65bbd58f..d3eed25daeb4c 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -5934,6 +5934,8 @@ static void ring_buffer_attach(struct perf_event *event,
 	struct perf_buffer *old_rb = NULL;
 	unsigned long flags;
 
+	WARN_ON_ONCE(event->parent);
+
 	if (event->rb) {
 		/*
 		 * Should be impossible, we set this when removing
@@ -5991,6 +5993,9 @@ static void ring_buffer_wakeup(struct perf_event *event)
 {
 	struct perf_buffer *rb;
 
+	if (event->parent)
+		event = event->parent;
+
 	rcu_read_lock();
 	rb = rcu_dereference(event->rb);
 	if (rb) {
@@ -6004,6 +6009,9 @@ struct perf_buffer *ring_buffer_get(struct perf_event *event)
 {
 	struct perf_buffer *rb;
 
+	if (event->parent)
+		event = event->parent;
+
 	rcu_read_lock();
 	rb = rcu_dereference(event->rb);
 	if (rb) {
@@ -6703,7 +6711,7 @@ static unsigned long perf_prepare_sample_aux(struct perf_event *event,
 	if (WARN_ON_ONCE(READ_ONCE(sampler->oncpu) != smp_processor_id()))
 		goto out;
 
-	rb = ring_buffer_get(sampler->parent ? sampler->parent : sampler);
+	rb = ring_buffer_get(sampler);
 	if (!rb)
 		goto out;
 
@@ -6769,7 +6777,7 @@ static void perf_aux_sample_output(struct perf_event *event,
 	if (WARN_ON_ONCE(!sampler || !data->aux_size))
 		return;
 
-	rb = ring_buffer_get(sampler->parent ? sampler->parent : sampler);
+	rb = ring_buffer_get(sampler);
 	if (!rb)
 		return;
 
-- 
2.34.1