[PATCH v2 0/6] perf arm_spe: Dump IMPDEF events

[PATCH v2 0/6] perf arm_spe: Dump IMPDEF events

[James Clark]

In the Arm SPE raw data dump, IMPDEF events aren't printed. Add the
ability to add names for some known events or print the raw event number
for unknown events.

For example:

  $ perf report -D

  ... ARM SPE data: size 0xd000 bytes
  00000000: b0 18 c6 32 80 00 80 ff a0  PC 0xff80008032c618 el1 ns=1
  00000009: 64 e7 42 00 00              CONTEXT 0x42e7 el1
  0000000e: 00 00 00 00 00              PAD
  00000013: 49 00                       LD GP-REG
  00000015: 52 16 10               	EV RETIRED L1D-ACCESS TLB-ACCESS

On N1 the event line becomes:

  00000015: 52 16 10                    EV RETIRED L1D-ACCESS TLB-ACCESS LATE-PREFETCH

Or on other cores it becomes:

  00000015: 52 16 10                    EV RETIRED L1D-ACCESS TLB-ACCESS IMPDEF:12

Signed-off-by: James Clark <james.clark@linaro.org>
---
Changes in v2:
- Put MIDR in arm_spe_pkt (Leo)
- Use for_each_set_bit() (Leo)
- Use BIT_ULL() to fix 32bit builds (Ian)
- Don't call strtol() with NULL (Ian)
- Link to v1: https://lore.kernel.org/r/20260401-james-spe-impdef-decode-v1-0-ad0d372c220c@linaro.org

---
James Clark (6):
      perf arm_spe: Make a function to get the MIDR
      perf arm_spe: Handle missing CPU IDs
      perf arm_spe: Store MIDR in arm_spe_pkt
      perf arm_spe: Turn event name mappings into an array
      perf arm_spe: Decode Arm N1 IMPDEF events
      perf arm_spe: Print remaining IMPDEF event numbers

 tools/perf/util/arm-spe-decoder/Build              |   2 +
 tools/perf/util/arm-spe-decoder/arm-spe-decoder.c  |  17 ++-
 tools/perf/util/arm-spe-decoder/arm-spe-decoder.h  |   3 +-
 .../util/arm-spe-decoder/arm-spe-pkt-decoder.c     | 144 ++++++++++++++-------
 .../util/arm-spe-decoder/arm-spe-pkt-decoder.h     |   5 +-
 tools/perf/util/arm-spe.c                          |  62 ++++++---
 6 files changed, 158 insertions(+), 75 deletions(-)
---
base-commit: 74e2dbe7be5037a5e5eed6bc1ad562747ac88566
change-id: 20260331-james-spe-impdef-decode-d944f4fdcff7

Best regards,
-- 
James Clark <james.clark@linaro.org>

[PATCH v2 1/6] perf arm_spe: Make a function to get the MIDR

[James Clark]

We'll need the MIDR to dump IMPDEF events in the next commits so extract
a function for it.

No functional changes intended.

Reviewed-by: Leo Yan <leo.yan@arm.com>
Signed-off-by: James Clark <james.clark@linaro.org>
---
 tools/perf/util/arm-spe.c | 36 ++++++++++++++++++++++--------------
 1 file changed, 22 insertions(+), 14 deletions(-)

diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
index 70dd9bee47c7..7447b000f9cd 100644
--- a/tools/perf/util/arm-spe.c
+++ b/tools/perf/util/arm-spe.c
@@ -958,14 +958,9 @@ static void arm_spe__synth_memory_level(struct arm_spe_queue *speq,
 	}
 }
 
-static void arm_spe__synth_ds(struct arm_spe_queue *speq,
-			      const struct arm_spe_record *record,
-			      union perf_mem_data_src *data_src)
+static int arm_spe__get_midr(struct arm_spe *spe, int cpu, u64 *midr)
 {
-	struct arm_spe *spe = speq->spe;
-	u64 *metadata = NULL;
-	u64 midr;
-	unsigned int i;
+	u64 *metadata;
 
 	/* Metadata version 1 assumes all CPUs are the same (old behavior) */
 	if (spe->metadata_ver == 1) {
@@ -973,15 +968,28 @@ static void arm_spe__synth_ds(struct arm_spe_queue *speq,
 
 		pr_warning_once("Old SPE metadata, re-record to improve decode accuracy\n");
 		cpuid = perf_env__cpuid(perf_session__env(spe->session));
-		midr = strtol(cpuid, NULL, 16);
-	} else {
-		metadata = arm_spe__get_metadata_by_cpu(spe, speq->cpu);
-		if (!metadata)
-			return;
-
-		midr = metadata[ARM_SPE_CPU_MIDR];
+		*midr = strtol(cpuid, NULL, 16);
+		return 0;
 	}
 
+	metadata = arm_spe__get_metadata_by_cpu(spe, cpu);
+	if (!metadata)
+		return -EINVAL;
+
+	*midr = metadata[ARM_SPE_CPU_MIDR];
+	return 0;
+}
+
+static void arm_spe__synth_ds(struct arm_spe_queue *speq,
+			      const struct arm_spe_record *record,
+			      union perf_mem_data_src *data_src)
+{
+	u64 midr;
+	unsigned int i;
+
+	if (arm_spe__get_midr(speq->spe, speq->cpu, &midr))
+		return;
+
 	for (i = 0; i < ARRAY_SIZE(data_source_handles); i++) {
 		if (is_midr_in_range_list(midr, data_source_handles[i].midr_ranges)) {
 			return data_source_handles[i].ds_synth(record, data_src);

-- 
2.34.1

[PATCH v2 2/6] perf arm_spe: Handle missing CPU IDs

[James Clark]

Don't call strtol() with a null pointer to avoid undefined behavior.

I'm not sure of the exact scenario for missing CPU IDs but I don't think
it happens in practice. SPE decoding can continue without them with
reduced functionality, but print an error message anyway.

Signed-off-by: James Clark <james.clark@linaro.org>
---
 tools/perf/util/arm-spe.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
index 7447b000f9cd..fc11f32e4911 100644
--- a/tools/perf/util/arm-spe.c
+++ b/tools/perf/util/arm-spe.c
@@ -968,16 +968,23 @@ static int arm_spe__get_midr(struct arm_spe *spe, int cpu, u64 *midr)
 
 		pr_warning_once("Old SPE metadata, re-record to improve decode accuracy\n");
 		cpuid = perf_env__cpuid(perf_session__env(spe->session));
+		if (!cpuid)
+			goto err;
+
 		*midr = strtol(cpuid, NULL, 16);
 		return 0;
 	}
 
 	metadata = arm_spe__get_metadata_by_cpu(spe, cpu);
 	if (!metadata)
-		return -EINVAL;
+		goto err;
 
 	*midr = metadata[ARM_SPE_CPU_MIDR];
 	return 0;
+
+err:
+	pr_err("Failed to get MIDR for CPU %d\n", cpu);
+	return -EINVAL;
 }
 
 static void arm_spe__synth_ds(struct arm_spe_queue *speq,

-- 
2.34.1

[PATCH v2 3/6] perf arm_spe: Store MIDR in arm_spe_pkt

[James Clark]

The MIDR will affect printing of arm_spe_pkts, so store a copy of it
there. Technically it's constant for each decoder, but there is no
decoder when doing a raw dump, so it has to be stored in every packet.
It will only be used in raw dump mode and not in normal decoding for
now, but to avoid any surprises, set MIDR properly on the decoder too.

Having both the MIDR and the arm_spe_pkt (which has a copy of it) in the
decoder seemed a bit weird, so remove arm_spe_pkt from the decoder. The
packet is only short lived anyway so probably shouldn't have been there
in the first place.

Signed-off-by: James Clark <james.clark@linaro.org>
---
 tools/perf/util/arm-spe-decoder/arm-spe-decoder.c   | 17 ++++++++++-------
 tools/perf/util/arm-spe-decoder/arm-spe-decoder.h   |  3 +--
 .../perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c |  3 ++-
 .../perf/util/arm-spe-decoder/arm-spe-pkt-decoder.h |  3 ++-
 tools/perf/util/arm-spe.c                           | 21 +++++++++++++++------
 5 files changed, 30 insertions(+), 17 deletions(-)

diff --git a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c
index 9e02b2bdd117..7a3a4815fd37 100644
--- a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c
+++ b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c
@@ -120,7 +120,8 @@ static int arm_spe_get_data(struct arm_spe_decoder *decoder)
 	return decoder->len;
 }
 
-static int arm_spe_get_next_packet(struct arm_spe_decoder *decoder)
+static int arm_spe_get_next_packet(struct arm_spe_decoder *decoder,
+				  struct arm_spe_pkt *packet)
 {
 	int ret;
 
@@ -134,7 +135,8 @@ static int arm_spe_get_next_packet(struct arm_spe_decoder *decoder)
 		}
 
 		ret = arm_spe_get_packet(decoder->buf, decoder->len,
-					 &decoder->packet);
+					 packet, decoder->midr);
+
 		if (ret <= 0) {
 			/* Move forward for 1 byte */
 			decoder->buf += 1;
@@ -144,7 +146,7 @@ static int arm_spe_get_next_packet(struct arm_spe_decoder *decoder)
 
 		decoder->buf += ret;
 		decoder->len -= ret;
-	} while (decoder->packet.type == ARM_SPE_PAD);
+	} while (packet->type == ARM_SPE_PAD);
 
 	return 1;
 }
@@ -154,19 +156,20 @@ static int arm_spe_read_record(struct arm_spe_decoder *decoder)
 	int err;
 	int idx;
 	u64 payload, ip;
+	struct arm_spe_pkt packet;
 
 	memset(&decoder->record, 0x0, sizeof(decoder->record));
 	decoder->record.context_id = (u64)-1;
 
 	while (1) {
-		err = arm_spe_get_next_packet(decoder);
+		err = arm_spe_get_next_packet(decoder, &packet);
 		if (err <= 0)
 			return err;
 
-		idx = decoder->packet.index;
-		payload = decoder->packet.payload;
+		idx = packet.index;
+		payload = packet.payload;
 
-		switch (decoder->packet.type) {
+		switch (packet.type) {
 		case ARM_SPE_TIMESTAMP:
 			decoder->record.timestamp = payload;
 			return 1;
diff --git a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h
index 3310e05122f0..0cbcb501edc9 100644
--- a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h
+++ b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h
@@ -147,8 +147,7 @@ struct arm_spe_decoder {
 
 	const unsigned char *buf;
 	size_t len;
-
-	struct arm_spe_pkt packet;
+	u64 midr;
 };
 
 struct arm_spe_decoder *arm_spe_decoder_new(struct arm_spe_params *params);
diff --git a/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c b/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c
index 5769ba2f4140..718022aecec3 100644
--- a/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c
+++ b/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c
@@ -222,11 +222,12 @@ static int arm_spe_do_get_packet(const unsigned char *buf, size_t len,
 }
 
 int arm_spe_get_packet(const unsigned char *buf, size_t len,
-		       struct arm_spe_pkt *packet)
+		       struct arm_spe_pkt *packet, u64 midr)
 {
 	int ret;
 
 	ret = arm_spe_do_get_packet(buf, len, packet);
+	packet->midr = midr;
 	/* put multiple consecutive PADs on the same line, up to
 	 * the fixed-width output format of 16 bytes per line.
 	 */
diff --git a/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.h b/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.h
index adf4cde320aa..a457821f3bcc 100644
--- a/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.h
+++ b/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.h
@@ -35,6 +35,7 @@ struct arm_spe_pkt {
 	enum arm_spe_pkt_type	type;
 	unsigned char		index;
 	uint64_t		payload;
+	uint64_t		midr;
 };
 
 /* Short header (HEADER0) and extended header (HEADER1) */
@@ -184,7 +185,7 @@ enum arm_spe_events {
 const char *arm_spe_pkt_name(enum arm_spe_pkt_type);
 
 int arm_spe_get_packet(const unsigned char *buf, size_t len,
-		       struct arm_spe_pkt *packet);
+		       struct arm_spe_pkt *packet, u64 midr);
 
 int arm_spe_pkt_desc(const struct arm_spe_pkt *packet, char *buf, size_t len);
 #endif
diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
index fc11f32e4911..7fb33fe27693 100644
--- a/tools/perf/util/arm-spe.c
+++ b/tools/perf/util/arm-spe.c
@@ -134,8 +134,10 @@ struct data_source_handle {
 		.ds_synth = arm_spe__synth_##func,	\
 	}
 
+static int arm_spe__get_midr(struct arm_spe *spe, int cpu, u64 *midr);
+
 static void arm_spe_dump(struct arm_spe *spe __maybe_unused,
-			 unsigned char *buf, size_t len)
+			 unsigned char *buf, size_t len, u64 midr)
 {
 	struct arm_spe_pkt packet;
 	size_t pos = 0;
@@ -148,7 +150,8 @@ static void arm_spe_dump(struct arm_spe *spe __maybe_unused,
 		      len);
 
 	while (len) {
-		ret = arm_spe_get_packet(buf, len, &packet);
+		ret = arm_spe_get_packet(buf, len, &packet, midr);
+
 		if (ret > 0)
 			pkt_len = ret;
 		else
@@ -174,10 +177,10 @@ static void arm_spe_dump(struct arm_spe *spe __maybe_unused,
 }
 
 static void arm_spe_dump_event(struct arm_spe *spe, unsigned char *buf,
-			       size_t len)
+			       size_t len, u64 midr)
 {
 	printf(".\n");
-	arm_spe_dump(spe, buf, len);
+	arm_spe_dump(spe, buf, len, midr);
 }
 
 static int arm_spe_get_trace(struct arm_spe_buffer *b, void *data)
@@ -302,8 +305,10 @@ static void arm_spe_set_pid_tid_cpu(struct arm_spe *spe,
 
 	if (speq->thread) {
 		speq->pid = thread__pid(speq->thread);
-		if (queue->cpu == -1)
+		if (queue->cpu == -1) {
 			speq->cpu = thread__cpu(speq->thread);
+			arm_spe__get_midr(spe, speq->cpu, &speq->decoder->midr);
+		}
 	}
 }
 
@@ -1234,6 +1239,7 @@ static int arm_spe__setup_queue(struct arm_spe *spe,
 
 	if (queue->cpu != -1)
 		speq->cpu = queue->cpu;
+	arm_spe__get_midr(spe, queue->cpu, &speq->decoder->midr);
 
 	if (!speq->on_heap) {
 		int ret;
@@ -1476,8 +1482,11 @@ static int arm_spe_process_auxtrace_event(struct perf_session *session,
 		/* Dump here now we have copied a piped trace out of the pipe */
 		if (dump_trace) {
 			if (auxtrace_buffer__get_data(buffer, fd)) {
+				u64 midr = 0;
+
+				arm_spe__get_midr(spe, buffer->cpu.cpu, &midr);
 				arm_spe_dump_event(spe, buffer->data,
-						buffer->size);
+						   buffer->size, midr);
 				auxtrace_buffer__put_data(buffer);
 			}
 		}

-- 
2.34.1

[PATCH v2 4/6] perf arm_spe: Turn event name mappings into an array

[James Clark]

This is so we can have a single function that prints events and can be
used with multiple mappings from different CPUs. Remove any bit that was
printed so that later we can print out the remaining unknown impdef
bits.

No functional changes intended.

Reviewed-by: Leo Yan <leo.yan@arm.com>
Signed-off-by: James Clark <james.clark@linaro.org>
---
 .../util/arm-spe-decoder/arm-spe-pkt-decoder.c     | 88 +++++++++++-----------
 1 file changed, 43 insertions(+), 45 deletions(-)

diff --git a/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c b/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c
index 718022aecec3..67ca356100e5 100644
--- a/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c
+++ b/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c
@@ -277,6 +277,48 @@ static int arm_spe_pkt_out_string(int *err, char **buf_p, size_t *blen,
 	return ret;
 }
 
+struct ev_string {
+	u8 event;
+	const char *desc;
+};
+
+static const struct ev_string common_ev_strings[] = {
+	{ .event = EV_EXCEPTION_GEN, .desc = "EXCEPTION-GEN" },
+	{ .event = EV_RETIRED, .desc = "RETIRED" },
+	{ .event = EV_L1D_ACCESS, .desc = "L1D-ACCESS" },
+	{ .event = EV_L1D_REFILL, .desc = "L1D-REFILL" },
+	{ .event = EV_TLB_ACCESS, .desc = "TLB-ACCESS" },
+	{ .event = EV_TLB_WALK, .desc = "TLB-REFILL" },
+	{ .event = EV_NOT_TAKEN, .desc = "NOT-TAKEN" },
+	{ .event = EV_MISPRED, .desc = "MISPRED" },
+	{ .event = EV_LLC_ACCESS, .desc = "LLC-ACCESS" },
+	{ .event = EV_LLC_MISS, .desc = "LLC-REFILL" },
+	{ .event = EV_REMOTE_ACCESS, .desc = "REMOTE-ACCESS" },
+	{ .event = EV_ALIGNMENT, .desc = "ALIGNMENT" },
+	{ .event = EV_TRANSACTIONAL, .desc = "TXN" },
+	{ .event = EV_PARTIAL_PREDICATE, .desc = "SVE-PARTIAL-PRED" },
+	{ .event = EV_EMPTY_PREDICATE, .desc = "SVE-EMPTY-PRED" },
+	{ .event = EV_L2D_ACCESS, .desc = "L2D-ACCESS" },
+	{ .event = EV_L2D_MISS, .desc = "L2D-MISS" },
+	{ .event = EV_CACHE_DATA_MODIFIED, .desc = "HITM" },
+	{ .event = EV_RECENTLY_FETCHED, .desc = "LFB" },
+	{ .event = EV_DATA_SNOOPED, .desc = "SNOOPED" },
+	{ .event = EV_STREAMING_SVE_MODE, .desc = "STREAMING-SVE" },
+	{ .event = EV_SMCU, .desc = "SMCU" },
+	{ .event = 0, .desc = NULL },
+};
+
+static u64 print_event_list(int *err, char **buf, size_t *buf_len,
+			    const struct ev_string *ev_strings, u64 payload)
+{
+	for (const struct ev_string *ev = ev_strings; ev->desc != NULL; ev++) {
+		if (payload & BIT_ULL(ev->event))
+			arm_spe_pkt_out_string(err, buf, buf_len, " %s", ev->desc);
+		payload &= ~BIT_ULL(ev->event);
+	}
+	return payload;
+}
+
 static int arm_spe_pkt_desc_event(const struct arm_spe_pkt *packet,
 				  char *buf, size_t buf_len)
 {
@@ -284,51 +326,7 @@ static int arm_spe_pkt_desc_event(const struct arm_spe_pkt *packet,
 	int err = 0;
 
 	arm_spe_pkt_out_string(&err, &buf, &buf_len, "EV");
-
-	if (payload & BIT(EV_EXCEPTION_GEN))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " EXCEPTION-GEN");
-	if (payload & BIT(EV_RETIRED))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " RETIRED");
-	if (payload & BIT(EV_L1D_ACCESS))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " L1D-ACCESS");
-	if (payload & BIT(EV_L1D_REFILL))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " L1D-REFILL");
-	if (payload & BIT(EV_TLB_ACCESS))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " TLB-ACCESS");
-	if (payload & BIT(EV_TLB_WALK))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " TLB-REFILL");
-	if (payload & BIT(EV_NOT_TAKEN))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " NOT-TAKEN");
-	if (payload & BIT(EV_MISPRED))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " MISPRED");
-	if (payload & BIT(EV_LLC_ACCESS))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " LLC-ACCESS");
-	if (payload & BIT(EV_LLC_MISS))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " LLC-REFILL");
-	if (payload & BIT(EV_REMOTE_ACCESS))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " REMOTE-ACCESS");
-	if (payload & BIT(EV_ALIGNMENT))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " ALIGNMENT");
-	if (payload & BIT(EV_TRANSACTIONAL))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " TXN");
-	if (payload & BIT(EV_PARTIAL_PREDICATE))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " SVE-PARTIAL-PRED");
-	if (payload & BIT(EV_EMPTY_PREDICATE))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " SVE-EMPTY-PRED");
-	if (payload & BIT(EV_L2D_ACCESS))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " L2D-ACCESS");
-	if (payload & BIT(EV_L2D_MISS))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " L2D-MISS");
-	if (payload & BIT(EV_CACHE_DATA_MODIFIED))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " HITM");
-	if (payload & BIT(EV_RECENTLY_FETCHED))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " LFB");
-	if (payload & BIT(EV_DATA_SNOOPED))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " SNOOPED");
-	if (payload & BIT(EV_STREAMING_SVE_MODE))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " STREAMING-SVE");
-	if (payload & BIT(EV_SMCU))
-		arm_spe_pkt_out_string(&err, &buf, &buf_len, " SMCU");
+	print_event_list(&err, &buf, &buf_len, common_ev_strings, payload);
 
 	return err;
 }

-- 
2.34.1

[PATCH v2 5/6] perf arm_spe: Decode Arm N1 IMPDEF events

[James Clark]

From the TRM [1], N1 has one IMPDEF event which isn't covered by the
common list. Add a framework so that more cores can be added in the
future and that the N1 IMPDEF event can be decoded. Also increase the
size of the buffer because we're adding more strings and if it gets
truncated it falls back to a hex dump only.

[1]: https://developer.arm.com/documentation/100616/0401/Statistical-Profiling-Extension/implementation-defined-features-of-SPE
Suggested-by: Al Grant <al.grant@arm.com>
Signed-off-by: James Clark <james.clark@linaro.org>
---
 tools/perf/util/arm-spe-decoder/Build              |  2 ++
 .../util/arm-spe-decoder/arm-spe-pkt-decoder.c     | 39 +++++++++++++++++++++-
 .../util/arm-spe-decoder/arm-spe-pkt-decoder.h     |  2 +-
 3 files changed, 41 insertions(+), 2 deletions(-)

diff --git a/tools/perf/util/arm-spe-decoder/Build b/tools/perf/util/arm-spe-decoder/Build
index ab500e0efe24..97a298d1e279 100644
--- a/tools/perf/util/arm-spe-decoder/Build
+++ b/tools/perf/util/arm-spe-decoder/Build
@@ -1 +1,3 @@
 perf-util-y += arm-spe-pkt-decoder.o arm-spe-decoder.o
+
+CFLAGS_arm-spe-pkt-decoder.o += -I$(srctree)/tools/arch/arm64/include/ -I$(OUTPUT)arch/arm64/include/generated/
diff --git a/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c b/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c
index 67ca356100e5..b74f887a48f2 100644
--- a/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c
+++ b/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c
@@ -15,6 +15,8 @@
 
 #include "arm-spe-pkt-decoder.h"
 
+#include "../../arm64/include/asm/cputype.h"
+
 static const char * const arm_spe_packet_name[] = {
 	[ARM_SPE_PAD]		= "PAD",
 	[ARM_SPE_END]		= "END",
@@ -308,6 +310,11 @@ static const struct ev_string common_ev_strings[] = {
 	{ .event = 0, .desc = NULL },
 };
 
+static const struct ev_string n1_event_strings[] = {
+	{ .event = 12, .desc = "LATE-PREFETCH" },
+	{ .event = 0, .desc = NULL },
+};
+
 static u64 print_event_list(int *err, char **buf, size_t *buf_len,
 			    const struct ev_string *ev_strings, u64 payload)
 {
@@ -319,6 +326,26 @@ static u64 print_event_list(int *err, char **buf, size_t *buf_len,
 	return payload;
 }
 
+struct event_print_handle {
+	const struct midr_range *midr_ranges;
+	const struct ev_string *ev_strings;
+};
+
+#define EV_PRINT(range, strings)			\
+	{					\
+		.midr_ranges = range,		\
+		.ev_strings = strings,	\
+	}
+
+static const struct midr_range n1_event_encoding_cpus[] = {
+	MIDR_ALL_VERSIONS(MIDR_NEOVERSE_N1),
+	{},
+};
+
+static const struct event_print_handle event_print_handles[] = {
+	EV_PRINT(n1_event_encoding_cpus, n1_event_strings),
+};
+
 static int arm_spe_pkt_desc_event(const struct arm_spe_pkt *packet,
 				  char *buf, size_t buf_len)
 {
@@ -326,7 +353,17 @@ static int arm_spe_pkt_desc_event(const struct arm_spe_pkt *packet,
 	int err = 0;
 
 	arm_spe_pkt_out_string(&err, &buf, &buf_len, "EV");
-	print_event_list(&err, &buf, &buf_len, common_ev_strings, payload);
+	payload = print_event_list(&err, &buf, &buf_len, common_ev_strings,
+				   payload);
+
+	/* Try to decode IMPDEF bits for known CPUs */
+	for (unsigned int i = 0; i < ARRAY_SIZE(event_print_handles); i++) {
+		if (is_midr_in_range_list(packet->midr,
+					  event_print_handles[i].midr_ranges))
+			payload = print_event_list(&err, &buf, &buf_len,
+						   event_print_handles[i].ev_strings,
+						   payload);
+	}
 
 	return err;
 }
diff --git a/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.h b/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.h
index a457821f3bcc..a3300bec4990 100644
--- a/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.h
+++ b/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.h
@@ -11,7 +11,7 @@
 #include <stddef.h>
 #include <stdint.h>
 
-#define ARM_SPE_PKT_DESC_MAX		256
+#define ARM_SPE_PKT_DESC_MAX		512
 
 #define ARM_SPE_NEED_MORE_BYTES		-1
 #define ARM_SPE_BAD_PACKET		-2

-- 
2.34.1

[PATCH v2 6/6] perf arm_spe: Print remaining IMPDEF event numbers

[James Clark]

Any IMPDEF events not printed out from a known core's IMPDEF list or for
a completely unknown core will still not be shown to the user. Fix this
by printing the remaining bits as comma separated raw numbers, e.g.
"IMPDEF:1,2,3,4".

Suggested-by: Al Grant <al.grant@arm.com>
Signed-off-by: James Clark <james.clark@linaro.org>
---
 tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c b/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c
index b74f887a48f2..c65b22a2179c 100644
--- a/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c
+++ b/tools/perf/util/arm-spe-decoder/arm-spe-pkt-decoder.c
@@ -8,6 +8,7 @@
 #include <string.h>
 #include <endian.h>
 #include <byteswap.h>
+#include <linux/bitmap.h>
 #include <linux/bitops.h>
 #include <stdarg.h>
 #include <linux/kernel.h>
@@ -365,6 +366,23 @@ static int arm_spe_pkt_desc_event(const struct arm_spe_pkt *packet,
 						   payload);
 	}
 
+	/*
+	 * Print remaining IMPDEF bits that weren't printed above as raw
+	 * "IMPDEF:1,2,3,4" etc.
+	 */
+	if (payload) {
+		int i;
+
+		arm_spe_pkt_out_string(&err, &buf, &buf_len, " IMPDEF:");
+		for_each_set_bit(i, &payload, 64) {
+			const char *sep = payload & (payload - 1) ? "," : "";
+
+			arm_spe_pkt_out_string(&err, &buf, &buf_len, "%d%s", i,
+					       sep);
+			payload &= ~BIT_ULL(i);
+		}
+	}
+
 	return err;
 }
 

-- 
2.34.1

Re: [PATCH v2 2/6] perf arm_spe: Handle missing CPU IDs

[Leo Yan]

On Tue, Apr 07, 2026 at 03:05:16PM +0100, James Clark wrote:
> Don't call strtol() with a null pointer to avoid undefined behavior.
> 
> I'm not sure of the exact scenario for missing CPU IDs but I don't think
> it happens in practice. SPE decoding can continue without them with
> reduced functionality, but print an error message anyway.
> 
> Signed-off-by: James Clark <james.clark@linaro.org>

Reviewed-by: Leo Yan <leo.yan@arm.com>

> ---
>  tools/perf/util/arm-spe.c | 9 ++++++++-
>  1 file changed, 8 insertions(+), 1 deletion(-)
> 
> diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
> index 7447b000f9cd..fc11f32e4911 100644
> --- a/tools/perf/util/arm-spe.c
> +++ b/tools/perf/util/arm-spe.c
> @@ -968,16 +968,23 @@ static int arm_spe__get_midr(struct arm_spe *spe, int cpu, u64 *midr)
>  
>  		pr_warning_once("Old SPE metadata, re-record to improve decode accuracy\n");
>  		cpuid = perf_env__cpuid(perf_session__env(spe->session));
> +		if (!cpuid)
> +			goto err;
> +
>  		*midr = strtol(cpuid, NULL, 16);
>  		return 0;
>  	}
>  
>  	metadata = arm_spe__get_metadata_by_cpu(spe, cpu);
>  	if (!metadata)
> -		return -EINVAL;
> +		goto err;
>  
>  	*midr = metadata[ARM_SPE_CPU_MIDR];
>  	return 0;
> +
> +err:
> +	pr_err("Failed to get MIDR for CPU %d\n", cpu);
> +	return -EINVAL;
>  }
>  
>  static void arm_spe__synth_ds(struct arm_spe_queue *speq,
> 
> -- 
> 2.34.1
> 
> 

Re: [PATCH v2 3/6] perf arm_spe: Store MIDR in arm_spe_pkt

[Leo Yan]

On Tue, Apr 07, 2026 at 03:05:17PM +0100, James Clark wrote:
> The MIDR will affect printing of arm_spe_pkts, so store a copy of it
> there. Technically it's constant for each decoder, but there is no
> decoder when doing a raw dump, so it has to be stored in every packet.
> It will only be used in raw dump mode and not in normal decoding for
> now, but to avoid any surprises, set MIDR properly on the decoder too.
> 
> Having both the MIDR and the arm_spe_pkt (which has a copy of it) in the
> decoder seemed a bit weird, so remove arm_spe_pkt from the decoder. The
> packet is only short lived anyway so probably shouldn't have been there
> in the first place.
> 
> Signed-off-by: James Clark <james.clark@linaro.org>

Reviewed-by: Leo Yan <leo.yan@arm.com>

Re: [PATCH v2 5/6] perf arm_spe: Decode Arm N1 IMPDEF events

[Leo Yan]

On Tue, Apr 07, 2026 at 03:05:19PM +0100, James Clark wrote:
> From the TRM [1], N1 has one IMPDEF event which isn't covered by the
> common list. Add a framework so that more cores can be added in the
> future and that the N1 IMPDEF event can be decoded. Also increase the
> size of the buffer because we're adding more strings and if it gets
> truncated it falls back to a hex dump only.
> 
> [1]: https://developer.arm.com/documentation/100616/0401/Statistical-Profiling-Extension/implementation-defined-features-of-SPE
> Suggested-by: Al Grant <al.grant@arm.com>
> Signed-off-by: James Clark <james.clark@linaro.org>

Reviewed-by: Leo Yan <leo.yan@arm.com>

Re: [PATCH v2 6/6] perf arm_spe: Print remaining IMPDEF event numbers

[Leo Yan]

On Tue, Apr 07, 2026 at 03:05:20PM +0100, James Clark wrote:
> Any IMPDEF events not printed out from a known core's IMPDEF list or for
> a completely unknown core will still not be shown to the user. Fix this
> by printing the remaining bits as comma separated raw numbers, e.g.
> "IMPDEF:1,2,3,4".
> 
> Suggested-by: Al Grant <al.grant@arm.com>
> Signed-off-by: James Clark <james.clark@linaro.org>

Reviewed-by: Leo Yan <leo.yan@arm.com>