Skip to contentSkip to conversation

perf: Add support for memory region/range reporting

PATCH Superseded Thomas Falcon 3 participants 18 messages Aug 11 – Aug 20 perf events

v4 of 12 Superseded Latest is v12 →

lore
Share & tools
lore
#0 lore raw

v4 cover perf: Add support for memory region/range reporting

Add support for two memory-related reporting features in the perf tool:

1. Memory region reporting in perf-c2c and perf-script subcommands.
2. Memory range data in perf.data file header and perf-c2c subcommand.

Memory region reporting was introduced as part of support for the Off-module Response facility (OMR) [1], which provides a new data source encoding supporting "up to 8 fine-grained memory regions in addition to the cache region, offering more detailed insights into memory access regions."

Memory Range support was introduced with the addition of the ACPI Memory Range and Region Mapping (MRRM) [2] table. It provides a base address plus a length value for each memory range, as well as NUMA node and local and remote "region ID's" so that "platform firmware can indicate the type of memory for each range."[3]

Include a change allowing PERF_MEM_LVLNUM_L0 to be printed in perf-mem output as well.

[1]: https://lore.kernel.org/all/[email protected]/
[2]: https://lore.kernel.org/lkml/[email protected]/
[3]: MRRM definition allow for future expansion for the OS to assign these region IDs.

v4:
-- perf-script: gate the Region field on the HEADER_MEMORY_RANGES feature
bit via perf_session instead of a global flag, with a pipe-mode fallback
-- perf-c2c: fix output_str allocation error handling
-- perf header: Fix typo in region id bounds checking which made region 255
invalid

v3:
-- included missing ff-size check in process_memory_ranges
-- make memory region reporting in perf c2c-and perf-script conditional
on feature bit, requiring some patch reordering (Namhyung Kim)
-- used open/openat to read memory range sysfs files in memory_range__read()
(Namhyung Kim)
-- Removed path name and file name buffers in memory_range__read(),
instead pass path name from as a parameter from memory_range__parse() (Namhyung Kim)
-- use calloc instead of zalloc in memory_range__parse() and
process_memory_ranges() (Namhyung Kim)
-- Removed redundant check for existance of memory range directory in
memory_range__read() (Namhyung Kim)
-- Included output examples of memory ranges in perf header and perf-c2c
in commit message (Namhyung Kim)

v2:
-- Added missing newline in WARN_ONCE message in c2c_he__set_mem_region() -- increased buffers passed to perf_script__meminfo_scnprintf() from 150
to 200 bytes to avoid silent truncation
-- Added check for NULL return of sysfs__mountpoint() when parsing
memory ranges in sysfs
-- increased MAX_MEMORY_RANGES sanity check from 64 to 256 based on
ACPI MRRM table implementation in Linux kernel
-- added comment for MAX_MEMORY_RANGES to clarify that it is a sanity check
for malformed perf.data files
-- removed a line of code was removed from util/env.h but was added back in
v1 due to bad rebase

Dapeng Mi (3):
perf mem: Add support for printing PERF_MEM_LVLNUM_L0 perf tools: Show memory region in perf-c2c subcommand perf tools: Show memory region in perf-script subcommand

Thomas Falcon (3):
perf mem: Fix size tracking for mem_lvl's in
perf_script__meminfo_scnprintf()
perf header: Support memory ranges
perf c2c: print memory region data with stdio output

 tools/perf/Documentation/perf-record.txt      |   2 +-
 .../Documentation/perf.data-file-format.txt   |  13 ++
 tools/perf/builtin-c2c.c                      | 123 ++++++++++-
 tools/perf/builtin-inject.c                   |   1 +
 tools/perf/builtin-script.c                   |  13 +-
 tools/perf/util/bpf-filter.l                  |   1 +
 tools/perf/util/env.c                         |   1 +
 tools/perf/util/env.h                         |  10 +
 tools/perf/util/header.c                      | 203 ++++++++++++++++++
 tools/perf/util/header.h                      |   1 +
 tools/perf/util/mem-events.c                  |  93 +++++++-
 tools/perf/util/mem-events.h                  |   5 +-
 .../scripting-engines/trace-event-python.c    |   5 +-
 13 files changed, 449 insertions(+), 22 deletions(-)
Signature · 1 line
2.55.0
#1 lore raw

PATCH v4 1/6 perf mem: Fix size tracking for mem_lvl's in perf_script__meminfo_scnprintf()

When printing memory info in perf script, the entire size of the buffer is passed to perf_mem__lvl_scnprintf() instead of the remaining size. Pass the remaining buffer size instead.

This issue was detected by sashiko during an internal code review.

Assisted-by Sashiko:gemini-3.1-pro-preview
Fixes fdefc3750e847 ("perf mem: Print memory operation type")
---
 tools/perf/util/mem-events.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/mem-events.c +1 -1
index 0b49fce251fc..4e490f9cd348 100644
@@ -610,7 +610,7 @@ int perf_script__meminfo_scnprintf(char *out, size_t sz, const struct mem_info *
 	i += scnprintf(out, sz, "|OP ");
 	i += perf_mem__op_scnprintf(out + i, sz - i, mem_info);
 	i += scnprintf(out + i, sz - i, "|LVL ");
-	i += perf_mem__lvl_scnprintf(out + i, sz, mem_info);
+	i += perf_mem__lvl_scnprintf(out + i, sz - i, mem_info);
 	i += scnprintf(out + i, sz - i, "|SNP ");
 	i += perf_mem__snp_scnprintf(out + i, sz - i, mem_info);
 	i += scnprintf(out + i, sz - i, "|TLB ");
Signature · 1 line
2.55.0
reviewer Reviewed-by on v4 1/6
#2 lore raw

Re: [PATCH v4 1/6] perf mem: Fix size tracking for mem_lvl's in perf_script__meminfo_scnprintf()

Quoted 11 lines · Thomas Falcon

When printing memory info in perf script, the entire size of the buffer is passed to perf_mem__lvl_scnprintf() instead of the remaining size. Pass the remaining buffer size instead.

This issue was detected by sashiko during an internal code review.

Assisted-by Sashiko:gemini-3.1-pro-preview
Fixes fdefc3750e847 ("perf mem: Print memory operation type")

Thanks,
Ian

Quoted 20 lines
---
 tools/perf/util/mem-events.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/mem-events.c +1 -2
index 0b49fce251fc..4e490f9cd348 100644
@@ -610,7 +610,7 @@ int perf_script__meminfo_scnprintf(char *out, size_t sz, const struct mem_info *
        i += scnprintf(out, sz, "|OP ");
        i += perf_mem__op_scnprintf(out + i, sz - i, mem_info);
        i += scnprintf(out + i, sz - i, "|LVL ");
-       i += perf_mem__lvl_scnprintf(out + i, sz, mem_info);
+       i += perf_mem__lvl_scnprintf(out + i, sz - i, mem_info);
        i += scnprintf(out + i, sz - i, "|SNP ");
        i += perf_mem__snp_scnprintf(out + i, sz - i, mem_info);
        i += scnprintf(out + i, sz - i, "|TLB ");
--

2.55.0

#3 lore raw

PATCH v4 2/6 perf mem: Add support for printing PERF_MEM_LVLNUM_L0

In reply to Thomas Falcon

From: Dapeng Mi <[email protected]>

Add support for printing PERF_MEM_LVLNUM_L0 in perf mem report.

The L0 cache is newly added a small piece of cache which is the closest, lowest-latency memory cache tied directly to the execution pipeline.

The table "Table 9-4. Data Source Field Encodings for Panther Cove and Coyote Cove Microarchitectures" in the ISE doc chapter "9.2.1 Panther Cove and Coyote Cove Microarchitectures Memory Auxiliary Field Layout"[1] indicates the code "01H" means the "L0 Hit - Minimal latency core cache hit. This request was satisfied by the L0 data cache."

[1]: https://www.intel.com/content/www/us/en/content-details/922690/intel-architecture-instruction-set-extensions-programming-reference.html

Assisted-by Sashiko:gemini-3.1-pro-preview

---
v3: Included more information about the L0 cache in the commit message.

---
 tools/perf/Documentation/perf-record.txt | 2 +-
 tools/perf/util/bpf-filter.l             | 1 +
 tools/perf/util/mem-events.c             | 5 +++++
 tools/perf/util/mem-events.h             | 1 +
 4 files changed, 8 insertions(+), 1 deletion(-)
tools/perf/Documentation/perf-record.txt +1 -1
index 178f483140ed..b54032efe41c 100644
@@ -212,7 +212,7 @@ OPTIONS
 	The <value> can be one of:
 	  <number> (for any term)
 	  na, load, store, pfetch, exec (for mem_op)
-	  l1, l2, l3, l4, cxl, io, any_cache, lfb, ram, pmem (for mem_lvl)
+	  l0, l1, l2, l3, l4, cxl, io, any_cache, lfb, ram, pmem (for mem_lvl)
 	  na, none, hit, miss, hitm, fwd, peer (for mem_snoop)
 	  remote (for mem_remote)
 	  na, locked (for mem_locked)
tools/perf/util/bpf-filter.l +1 -0
index 6aa65ade3385..1be9df6550fc 100644
@@ -131,6 +131,7 @@ store		{ return constant(PERF_MEM_OP_STORE); }
 pfetch		{ return constant(PERF_MEM_OP_PFETCH); }
 exec		{ return constant(PERF_MEM_OP_EXEC); }
 
+l0		{ return constant(PERF_MEM_LVLNUM_L0); }
 l1		{ return constant(PERF_MEM_LVLNUM_L1); }
 l2		{ return constant(PERF_MEM_LVLNUM_L2); }
 l3		{ return constant(PERF_MEM_LVLNUM_L3); }
tools/perf/util/mem-events.c +5 -0
index 4e490f9cd348..4fd48fd20055 100644
@@ -391,6 +391,7 @@ static const char * const mem_lvlnum[] = {
 	[PERF_MEM_LVLNUM_L4] = "L4",
 	[PERF_MEM_LVLNUM_L2_MHB] = "L2 MHB",
 	[PERF_MEM_LVLNUM_MSC] = "Memory-side Cache",
+	[PERF_MEM_LVLNUM_L0] = "L0",
 	[PERF_MEM_LVLNUM_UNC] = "Uncached",
 	[PERF_MEM_LVLNUM_CXL] = "CXL",
 	[PERF_MEM_LVLNUM_IO] = "I/O",
@@ -831,6 +832,8 @@ int mem_stat_index(const enum mem_stat_type mst, const u64 val)
 		}
 	case PERF_MEM_STAT_CACHE:
 		switch (src.mem_lvl_num) {
+		case PERF_MEM_LVLNUM_L0:
+			return MEM_STAT_CACHE_L0;
 		case PERF_MEM_LVLNUM_L1:
 			return MEM_STAT_CACHE_L1;
 		case PERF_MEM_LVLNUM_L2:
@@ -915,6 +918,8 @@ const char *mem_stat_name(const enum mem_stat_type mst, const int idx)
 		}
 	case PERF_MEM_STAT_CACHE:
 		switch (idx) {
+		case MEM_STAT_CACHE_L0:
+			return "L0";
 		case MEM_STAT_CACHE_L1:
 			return "L1";
 		case MEM_STAT_CACHE_L2:
tools/perf/util/mem-events.h +1 -0
index 5b98076904b0..daa22748f9fe 100644
@@ -109,6 +109,7 @@ enum mem_stat_op {
 };
 
 enum mem_stat_cache {
+	MEM_STAT_CACHE_L0,
 	MEM_STAT_CACHE_L1,
 	MEM_STAT_CACHE_L2,
 	MEM_STAT_CACHE_L3,
Signature · 1 line
2.55.0
reviewer Reviewed-by on v4 2/6
#4 lore raw

Re: [PATCH v4 2/6] perf mem: Add support for printing PERF_MEM_LVLNUM_L0

Quoted 16 lines · Thomas Falcon

From: Dapeng Mi <[email protected]>

Add support for printing PERF_MEM_LVLNUM_L0 in perf mem report.

The L0 cache is newly added a small piece of cache which is the closest, lowest-latency memory cache tied directly to the execution pipeline.

The table "Table 9-4. Data Source Field Encodings for Panther Cove and Coyote Cove Microarchitectures" in the ISE doc chapter "9.2.1 Panther Cove and Coyote Cove Microarchitectures Memory Auxiliary Field Layout"[1] indicates the code "01H" means the "L0 Hit - Minimal latency core cache hit. This request was satisfied by the L0 data cache."

Quoted 6 lines

[1]: https://www.intel.com/content/www/us/en/content-details/922690/intel-architecture-instruction-set-extensions-programming-reference.html

Assisted-by Sashiko:gemini-3.1-pro-preview

Thanks,
Ian

Quoted 79 lines

---
v3: Included more information about the L0 cache in the commit message.

---
 tools/perf/Documentation/perf-record.txt | 2 +-
 tools/perf/util/bpf-filter.l             | 1 +
 tools/perf/util/mem-events.c             | 5 +++++
 tools/perf/util/mem-events.h             | 1 +
 4 files changed, 8 insertions(+), 1 deletion(-)
tools/perf/Documentation/perf-record.txt +1 -1
index 178f483140ed..b54032efe41c 100644
@@ -212,7 +212,7 @@ OPTIONS
        The <value> can be one of:
          <number> (for any term)
          na, load, store, pfetch, exec (for mem_op)
-         l1, l2, l3, l4, cxl, io, any_cache, lfb, ram, pmem (for mem_lvl)
+         l0, l1, l2, l3, l4, cxl, io, any_cache, lfb, ram, pmem (for mem_lvl)
          na, none, hit, miss, hitm, fwd, peer (for mem_snoop)
          remote (for mem_remote)
          na, locked (for mem_locked)
tools/perf/util/bpf-filter.l +1 -0
index 6aa65ade3385..1be9df6550fc 100644
@@ -131,6 +131,7 @@ store               { return constant(PERF_MEM_OP_STORE); }
 pfetch         { return constant(PERF_MEM_OP_PFETCH); }
 exec           { return constant(PERF_MEM_OP_EXEC); }
 
+l0             { return constant(PERF_MEM_LVLNUM_L0); }
 l1             { return constant(PERF_MEM_LVLNUM_L1); }
 l2             { return constant(PERF_MEM_LVLNUM_L2); }
 l3             { return constant(PERF_MEM_LVLNUM_L3); }
tools/perf/util/mem-events.c +5 -0
index 4e490f9cd348..4fd48fd20055 100644
@@ -391,6 +391,7 @@ static const char * const mem_lvlnum[] = {
        [PERF_MEM_LVLNUM_L4] = "L4",
        [PERF_MEM_LVLNUM_L2_MHB] = "L2 MHB",
        [PERF_MEM_LVLNUM_MSC] = "Memory-side Cache",
+       [PERF_MEM_LVLNUM_L0] = "L0",
        [PERF_MEM_LVLNUM_UNC] = "Uncached",
        [PERF_MEM_LVLNUM_CXL] = "CXL",
        [PERF_MEM_LVLNUM_IO] = "I/O",
@@ -831,6 +832,8 @@ int mem_stat_index(const enum mem_stat_type mst, const u64 val)
                }
        case PERF_MEM_STAT_CACHE:
                switch (src.mem_lvl_num) {
+               case PERF_MEM_LVLNUM_L0:
+                       return MEM_STAT_CACHE_L0;
                case PERF_MEM_LVLNUM_L1:
                        return MEM_STAT_CACHE_L1;
                case PERF_MEM_LVLNUM_L2:
@@ -915,6 +918,8 @@ const char *mem_stat_name(const enum mem_stat_type mst, const int idx)
                }
        case PERF_MEM_STAT_CACHE:
                switch (idx) {
+               case MEM_STAT_CACHE_L0:
+                       return "L0";
                case MEM_STAT_CACHE_L1:
                        return "L1";
                case MEM_STAT_CACHE_L2:
tools/perf/util/mem-events.h +1 -1
index 5b98076904b0..daa22748f9fe 100644
@@ -109,6 +109,7 @@ enum mem_stat_op {
 };
 
 enum mem_stat_cache {
+       MEM_STAT_CACHE_L0,
        MEM_STAT_CACHE_L1,
        MEM_STAT_CACHE_L2,
        MEM_STAT_CACHE_L3,
--

2.55.0

#5 lore raw

PATCH v4 3/6 perf header: Support memory ranges

In reply to Thomas Falcon

Memory ranges were created to track different types of memory, such as persistent, high bandwidth, or CXL-attached, that may be present on a system for performance monitoring and resource control purposes.

Memory range data are parsed from the ACPI MRRM table and exposed to userspace tools via sysfs [1]. Memory range data is read from:

/sys/firmware/acpi/memory_ranges/rangeX

With the following attributes:

	u64 base;
	u64 length;
	int node;
	u8  local_region_id;
	u8  remote_region_id;

Read memory range data from sysfs if present and save it in the header of the perf data file under a new feature bit, HEADER_MEMORY_RANGES (35). Memory range data can be viewed with the --header or --header-only options of perf-report and perf-script.

Example output:

 # memory ranges (nr 5):
 # range0: [0x0000000000000000-0x00000000bfffffff], node = 0, local_region_id = 0, remote_region_id = 255
 # range1: [0x0000000100000000-0x000000203fffffff], node = 0, local_region_id = 0, remote_region_id = 255
 # range2: [0x0000008000000000-0x0000027fffffffff], node = -2, local_region_id = 1, remote_region_id = 255
 # range3: [0x0000028000000000-0x0000047fffffffff], node = -2, local_region_id = 1, remote_region_id = 255
 # range4: [0x0000048000000000-0x000004ffffffffff], node = -2, local_region_id = 1, remote_region_id = 255

[1]: https://lore.kernel.org/lkml/[email protected]/

Assisted-by Sashiko:gemini-3.1-pro-preview
Assisted-by GitHub-Copilot:claude-opus-4-8

---
v4:
-- Fixed bounds checking for local and remote region IDs in
memory_range__read(), which made region "255" invalid

v3:
-- Added missing ff->size check in process_memory_ranges()
-- used open/openat to read memory range sysfs files in memory_range__read() -- Removed path name and file name buffers in memory_range__read(),
instead pass path name from as a parameter from memory_range__parse()
-- use calloc instead of zalloc in memory_range__parse() and
process_memory_ranges()
-- Removed redundant check for existance of memory range directory in
memory_range__read()
-- Included examples of memory range output in commit message

v2:
-- Added check for NULL return of sysfs__mountpoint() when parsing
memory ranges in sysfs
-- increased MAX_MEMORY_RANGES sanity check from 64 to 256 based on
ACPI MRRM table implementation in Linux kernel
-- added comment for MAX_MEMORY_RANGES to clarify that it is a sanity check
for malformed perf.data files
-- removed a line of code was removed from util/env.h but was added back in
v1 due to bad rebase
---

 .../Documentation/perf.data-file-format.txt   |  13 ++
 tools/perf/builtin-inject.c                   |   1 +
 tools/perf/util/env.c                         |   1 +
 tools/perf/util/env.h                         |  10 +
 tools/perf/util/header.c                      | 203 ++++++++++++++++++
 tools/perf/util/header.h                      |   1 +
 6 files changed, 229 insertions(+)
tools/perf/Documentation/perf.data-file-format.txt +13 -0
index b90cba9168f8..1f8e7d8e9ae2 100644
@@ -478,6 +478,19 @@ The size of the cacheline in bytes. Format:
 
 	unsigned int cln_size;
 
+	HEADER_MEMORY_RANGES = 35,
+
+List of memory ranges. The format of the data is as below.
+
+u32 nr_memory_ranges;
+struct memory_range {
+	u64 base;
+	u64 length;
+	int node;
+	u8 local_region_id;
+	u8 remote_region_id;
+}; [nr_memory_ranges]
+
 	other bits are reserved and should be ignored for now
 	HEADER_FEAT_BITS	= 256,
 
tools/perf/builtin-inject.c +1 -0
index 70bbfad5653e..169617a884d3 100644
@@ -2306,6 +2306,7 @@ static bool keep_feat(struct perf_inject *inject, int feat)
 	case HEADER_PMU_CAPS:
 	case HEADER_CPU_DOMAIN_INFO:
 	case HEADER_CLN_SIZE:
+	case HEADER_MEMORY_RANGES:
 		return true;
 	/* Information that can be updated */
 	case HEADER_BUILD_ID:
tools/perf/util/env.c +1 -0
index c0e2b9d5f0b2..50993b0dedc5 100644
@@ -297,6 +297,7 @@ void perf_env__exit(struct perf_env *env)
 		zfree(&env->pmu_caps[i].pmu_name);
 	}
 	zfree(&env->pmu_caps);
+	zfree(&env->memory_ranges);
 	free_cpu_domain_info(env->cpu_domain, env->schedstat_version, env->nr_cpus_avail);
 }
 
tools/perf/util/env.h +10 -0
index 7acca39b42ff..23d35425efe3 100644
@@ -68,6 +68,14 @@ struct cpu_domain_map {
 	struct domain_info	**domains;
 };
 
+struct memory_range {
+	u64 base;
+	u64 length;
+	int node;
+	u8 local_region_id;
+	u8 remote_region_id;
+};
+
 struct perf_env {
 	char			*hostname;
 	char			*os_release;
@@ -100,6 +108,7 @@ struct perf_env {
 	int			nr_cpu_pmu_caps;
 	int			nr_hybrid_nodes;
 	int			nr_pmus_with_caps;
+	int			nr_memory_ranges;
 	char			*cmdline;
 	const char		**cmdline_argv;
 	char			*sibling_cores;
@@ -122,6 +131,7 @@ struct perf_env {
 	unsigned long long	 memory_bsize;
 	struct hybrid_node	*hybrid_nodes;
 	struct pmu_caps		*pmu_caps;
+	struct memory_range	*memory_ranges;
 #ifdef HAVE_LIBBPF_SUPPORT
 	/*
 	 * bpf_info_lock protects bpf rbtrees. This is needed because the
tools/perf/util/header.c +203 -0

233 lines not on this page · raw message

tools/perf/util/header.h +1 -0
index 5e03f884b7cc..765037762758 100644
@@ -56,6 +56,7 @@ enum {
 	HEADER_CPU_DOMAIN_INFO,
 	HEADER_E_MACHINE,
 	HEADER_CLN_SIZE,
+	HEADER_MEMORY_RANGES,
 	HEADER_LAST_FEATURE,
 	HEADER_FEAT_BITS	= 256,
 };
Signature · 1 line
2.55.0
reviewer
#6 lore raw

Re: [PATCH v4 3/6] perf header: Support memory ranges

Quoted 32 lines · Thomas Falcon

Memory ranges were created to track different types of memory, such as persistent, high bandwidth, or CXL-attached, that may be present on a system for performance monitoring and resource control purposes.

Memory range data are parsed from the ACPI MRRM table and exposed to userspace tools via sysfs [1]. Memory range data is read from:

/sys/firmware/acpi/memory_ranges/rangeX

With the following attributes:

        u64 base;
        u64 length;
        int node;
        u8  local_region_id;
        u8  remote_region_id;

Read memory range data from sysfs if present and save it in the header of the perf data file under a new feature bit, HEADER_MEMORY_RANGES (35). Memory range data can be viewed with the --header or --header-only options of perf-report and perf-script.

Example output:

 # memory ranges (nr 5):
 # range0: [0x0000000000000000-0x00000000bfffffff], node = 0, local_region_id = 0, remote_region_id = 255
 # range1: [0x0000000100000000-0x000000203fffffff], node = 0, local_region_id = 0, remote_region_id = 255
 # range2: [0x0000008000000000-0x0000027fffffffff], node = -2, local_region_id = 1, remote_region_id = 255
 # range3: [0x0000028000000000-0x0000047fffffffff], node = -2, local_region_id = 1, remote_region_id = 255
 # range4: [0x0000048000000000-0x000004ffffffffff], node = -2, local_region_id = 1, remote_region_id = 255
Quoted 5 lines

[1]: https://lore.kernel.org/lkml/[email protected]/

Assisted-by Sashiko:gemini-3.1-pro-preview
Assisted-by GitHub-Copilot:claude-opus-4-8

Looks good to me. Did you do pipe mode testing? Is there any additional testing coverage we can add?

Thanks,
Ian

Quoted 375 lines

---
v4:
-- Fixed bounds checking for local and remote region IDs in
memory_range__read(), which made region "255" invalid

v3:
-- Added missing ff->size check in process_memory_ranges()
-- used open/openat to read memory range sysfs files in memory_range__read() -- Removed path name and file name buffers in memory_range__read(),
instead pass path name from as a parameter from memory_range__parse()
-- use calloc instead of zalloc in memory_range__parse() and
process_memory_ranges()
-- Removed redundant check for existance of memory range directory in
memory_range__read()
-- Included examples of memory range output in commit message

v2:
-- Added check for NULL return of sysfs__mountpoint() when parsing
memory ranges in sysfs
-- increased MAX_MEMORY_RANGES sanity check from 64 to 256 based on
ACPI MRRM table implementation in Linux kernel
-- added comment for MAX_MEMORY_RANGES to clarify that it is a sanity check
for malformed perf.data files
-- removed a line of code was removed from util/env.h but was added back in
v1 due to bad rebase
---

 .../Documentation/perf.data-file-format.txt   |  13 ++
 tools/perf/builtin-inject.c                   |   1 +
 tools/perf/util/env.c                         |   1 +
 tools/perf/util/env.h                         |  10 +
 tools/perf/util/header.c                      | 203 ++++++++++++++++++
 tools/perf/util/header.h                      |   1 +
 6 files changed, 229 insertions(+)
tools/perf/Documentation/perf.data-file-format.txt +13 -0
index b90cba9168f8..1f8e7d8e9ae2 100644
@@ -478,6 +478,19 @@ The size of the cacheline in bytes. Format:
 
        unsigned int cln_size;
 
+       HEADER_MEMORY_RANGES = 35,
+
+List of memory ranges. The format of the data is as below.
+
+u32 nr_memory_ranges;
+struct memory_range {
+       u64 base;
+       u64 length;
+       int node;
+       u8 local_region_id;
+       u8 remote_region_id;
+}; [nr_memory_ranges]
+
        other bits are reserved and should be ignored for now
        HEADER_FEAT_BITS        = 256,
 
tools/perf/builtin-inject.c +1 -0
index 70bbfad5653e..169617a884d3 100644
@@ -2306,6 +2306,7 @@ static bool keep_feat(struct perf_inject *inject, int feat)
        case HEADER_PMU_CAPS:
        case HEADER_CPU_DOMAIN_INFO:
        case HEADER_CLN_SIZE:
+       case HEADER_MEMORY_RANGES:
                return true;
        /* Information that can be updated */
        case HEADER_BUILD_ID:
tools/perf/util/env.c +1 -0
index c0e2b9d5f0b2..50993b0dedc5 100644
@@ -297,6 +297,7 @@ void perf_env__exit(struct perf_env *env)
                zfree(&env->pmu_caps[i].pmu_name);
        }
        zfree(&env->pmu_caps);
+       zfree(&env->memory_ranges);
        free_cpu_domain_info(env->cpu_domain, env->schedstat_version, env->nr_cpus_avail);
 }
 
tools/perf/util/env.h +10 -0
index 7acca39b42ff..23d35425efe3 100644
@@ -68,6 +68,14 @@ struct cpu_domain_map {
        struct domain_info      **domains;
 };
 
+struct memory_range {
+       u64 base;
+       u64 length;
+       int node;
+       u8 local_region_id;
+       u8 remote_region_id;
+};
+
 struct perf_env {
        char                    *hostname;
        char                    *os_release;
@@ -100,6 +108,7 @@ struct perf_env {
        int                     nr_cpu_pmu_caps;
        int                     nr_hybrid_nodes;
        int                     nr_pmus_with_caps;
+       int                     nr_memory_ranges;
        char                    *cmdline;
        const char              **cmdline_argv;
        char                    *sibling_cores;
@@ -122,6 +131,7 @@ struct perf_env {
        unsigned long long       memory_bsize;
        struct hybrid_node      *hybrid_nodes;
        struct pmu_caps         *pmu_caps;
+       struct memory_range     *memory_ranges;
 #ifdef HAVE_LIBBPF_SUPPORT
        /*
         * bpf_info_lock protects bpf rbtrees. This is needed because the
tools/perf/util/header.c +203 -0
index e90e541f546b..1dbc83a2d798 100644
@@ -92,6 +92,7 @@
 #define MAX_PMU_CAPS           512
 #define MAX_PMU_MAPPINGS       4096
 #define MAX_SCHED_DOMAINS      64
+#define MAX_MEMORY_RANGES      256
 
 /*
  * magic2 = "PERFILE2"
@@ -1891,6 +1892,132 @@ static int write_cpu_domain_info(struct feat_fd *ff,
        return ret;
 }
 
+static int memory_range__read(struct memory_range *range, const char *path)
+{
+       char buf[32];
+       ssize_t n;
+       int dfd, fd, tmp, ret = -1;
+
+       dfd = open(path, O_RDONLY | O_DIRECTORY);
+       if (dfd < 0)
+               return -1;
+
+#define _R(name, dst, conv)                    \
+       fd = openat(dfd, name, O_RDONLY);       \
+       if (fd < 0)                             \
+               goto out;                       \
+       n = read(fd, buf, sizeof(buf) - 1);     \
+       close(fd);                              \
+       if (n <= 0)                             \
+               goto out;                       \
+       buf[n] = '\0';                          \
+       dst = conv(buf, NULL, 0);
+
+       _R("base", range->base, strtoull);
+       _R("length", range->length, strtoull);
+       _R("node", range->node, strtol);
+       _R("local_region_id", tmp, strtol);
+       if (tmp < 0 || tmp > UINT8_MAX)
+               goto out;
+       range->local_region_id = tmp;
+       _R("remote_region_id", tmp, strtol);
+       if (tmp < 0 || tmp > UINT8_MAX)
+               goto out;
+       range->remote_region_id = tmp;
+#undef _R
+
+       ret = 0;
+out:
+       close(dfd);
+       return ret;
+}
+
+static int memory_range__parse(struct memory_range **ranges)
+{
+       const char *sysfs = sysfs__mountpoint();
+       int i, err, nr_memory_ranges = 0;
+       char path[PATH_MAX];
+       struct stat st;
+
+       if (!sysfs)
+               return 0;
+
+       scnprintf(path, PATH_MAX, "%s/firmware/acpi/memory_ranges", sysfs);
+       if (stat(path, &st))
+               return 0;
+
+       while (1) {
+               scnprintf(path, PATH_MAX,
+                         "%s/firmware/acpi/memory_ranges/range%d",
+                         sysfs, nr_memory_ranges);
+               if (stat(path, &st))
+                       break;
+
+               nr_memory_ranges++;
+       }
+
+       if (nr_memory_ranges == 0)
+               return 0;
+
+       *ranges = calloc(nr_memory_ranges, sizeof(struct memory_range));
+       if (!(*ranges))
+               return -ENOMEM;
+
+       for (i = 0; i < nr_memory_ranges; i++) {
+               struct memory_range range;
+
+               scnprintf(path, PATH_MAX,
+                         "%s/firmware/acpi/memory_ranges/range%d", sysfs, i);
+               err = memory_range__read(&range, path);
+               if (err < 0)
+                       goto out_error;
+
+               (*ranges)[i] = range;
+       }
+
+       return nr_memory_ranges;
+
+out_error:
+       zfree(ranges);
+       return -1;
+}
+
+static int write_memory_ranges(struct feat_fd *ff,
+                        struct evlist *evlist __maybe_unused)
+{
+       struct memory_range *ranges = NULL;
+       int nr_memory_ranges = 0, ret;
+
+       nr_memory_ranges = memory_range__parse(&ranges);
+       if (nr_memory_ranges < 0)
+               return nr_memory_ranges;
+
+       ret = do_write(ff, &nr_memory_ranges, sizeof(nr_memory_ranges));
+       if (ret < 0)
+               goto out;
+
+       for (int i = 0; i < nr_memory_ranges; i++) {
+               ret = do_write(ff, &ranges[i].base, sizeof(u64));
+               if (ret < 0)
+                       goto out;
+               ret = do_write(ff, &ranges[i].length, sizeof(u64));
+               if (ret < 0)
+                       goto out;
+               ret = do_write(ff, &ranges[i].node, sizeof(u32));
+               if (ret < 0)
+                       goto out;
+               ret = do_write(ff, &ranges[i].local_region_id, sizeof(u8));
+               if (ret < 0)
+                       goto out;
+               ret = do_write(ff, &ranges[i].remote_region_id, sizeof(u8));
+               if (ret < 0)
+                       goto out;
+       }
+out:
+       zfree(&ranges);
+       return ret;
+}
+
 static void print_hostname(struct feat_fd *ff, FILE *fp)
 {
        fprintf(fp, "# hostname : %s\n", ff->ph->env.hostname);
@@ -2628,6 +2755,23 @@ static void print_cpu_domain_info(struct feat_fd *ff, FILE *fp)
        }
 }
 
+static void print_memory_ranges(struct feat_fd *ff, FILE *fp)
+{
+       struct memory_range *ranges = ff->ph->env.memory_ranges;
+       int nr_memory_ranges = ff->ph->env.nr_memory_ranges;
+       int i;
+
+       fprintf(fp, "# memory ranges (nr %d):\n", nr_memory_ranges);
+
+       for (i = 0; i < nr_memory_ranges; i++) {
+               fprintf(fp, "# range%u: [0x%016" PRIx64 "-0x%016" PRIx64,
+                       i, ranges[i].base, ranges[i].base + ranges[i].length - 1);
+               fprintf(fp, "], node = %d, local_region_id = %d, remote_region_id = %d\n",
+                       ranges[i].node, ranges[i].local_region_id,
+                       ranges[i].remote_region_id);
+       }
+}
+
 static int __event_process_build_id(struct perf_record_header_build_id *bev,
                                    char *filename,
                                    struct perf_session *session)
@@ -4201,6 +4345,64 @@ static int process_cpu_domain_info(struct feat_fd *ff, void *data __maybe_unused
        return ret;
 }
 
+static int process_memory_ranges(struct feat_fd *ff, void *data __maybe_unused)
+{
+       struct perf_env *env = &ff->ph->env;
+       struct memory_range *ranges, *r;
+       u32 nr_memory_ranges, i;
+
+       if (do_read_u32(ff, &nr_memory_ranges))
+               return -1;
+
+       if (!nr_memory_ranges) {
+               pr_debug("memory ranges not available\n");
+               return 0;
+       }
+
+       /* According to version 1.1 of the ACPI MRRM table, the maximum
+        * number of memory regions can be at most 255. Do a sanity check
+        * here to guard against a malformed perf.data file.
+        */
+       if (nr_memory_ranges >= MAX_MEMORY_RANGES) {
+               pr_err("Invalid memory_ranges: nr_memory_ranges (%u) > %u\n",
+                      nr_memory_ranges, MAX_MEMORY_RANGES);
+               return -1;
+       }
+
+       if (ff->size < sizeof(u32) + nr_memory_ranges * (2 * sizeof(u64) + sizeof(u32) + 2 * sizeof(u8))) {
+               pr_err("Invalid HEADER_MEMORY_RANGES: section too small (%zu) for %u range entries\n",
+                      ff->size, nr_memory_ranges);
+               return -1;
+       }
+
+       ranges = calloc(nr_memory_ranges, sizeof(*ranges));
+       if (!ranges)
+               return -1;
+
+       for (i = 0; i < nr_memory_ranges; i++) {
+               r = &ranges[i];
+
+               if (do_read_u64(ff, &r->base))
+                       goto error;
+               if (do_read_u64(ff, &r->length))
+                       goto error;
+               if (do_read_u32(ff, (u32 *) &r->node))
+                       goto error;
+               if (__do_read(ff, &r->local_region_id, sizeof(u8)))
+                       goto error;
+               if (__do_read(ff, &r->remote_region_id, sizeof(u8)))
+                       goto error;
+       }
+
+       env->memory_ranges = ranges;
+       env->nr_memory_ranges = nr_memory_ranges;
+
+       return 0;
+error:
+       zfree(&ranges);
+       return -1;
+}
+
 #define FEAT_OPR(n, func, __full_only) \
        [HEADER_##n] = {                                        \
                .name       = __stringify(n),                   \
@@ -4265,6 +4467,7 @@ const struct perf_header_feature_ops feat_ops[HEADER_LAST_FEATURE] = {
        FEAT_OPR(CPU_DOMAIN_INFO,       cpu_domain_info,        true),
        FEAT_OPR(E_MACHINE,     e_machine,      false),
        FEAT_OPR(CLN_SIZE,      cln_size,       false),
+       FEAT_OPR(MEMORY_RANGES, memory_ranges,  false),
 };
 
 struct header_print_data {
tools/perf/util/header.h +1 -1
index 5e03f884b7cc..765037762758 100644
@@ -56,6 +56,7 @@ enum {
        HEADER_CPU_DOMAIN_INFO,
        HEADER_E_MACHINE,
        HEADER_CLN_SIZE,
+       HEADER_MEMORY_RANGES,
        HEADER_LAST_FEATURE,
        HEADER_FEAT_BITS        = 256,
 };
--

2.55.0

#7 lore raw

Re: [PATCH v4 3/6] perf header: Support memory ranges

Quoted 46 lines · Ian Rogers
Quoted 37 lines · Thomas Falcon

Memory ranges were created to track different types of memory, such as persistent, high bandwidth, or CXL-attached, that may be present on a system for performance monitoring and resource control purposes.

Memory range data are parsed from the ACPI MRRM table and exposed to userspace tools via sysfs [1]. Memory range data is read from:

        /sys/firmware/acpi/memory_ranges/rangeX

With the following attributes:

        u64 base;
        u64 length;
        int node;
        u8  local_region_id;
        u8  remote_region_id;

Read memory range data from sysfs if present and save it in the header of the perf data file under a new feature bit, HEADER_MEMORY_RANGES (35). Memory range data can be viewed with the --header or --header-only options of perf-report and perf-script.

Example output:

 # memory ranges (nr 5):
 # range0: [0x0000000000000000-0x00000000bfffffff], node = 0,
local_region_id = 0, remote_region_id = 255
 # range1: [0x0000000100000000-0x000000203fffffff], node = 0,
local_region_id = 0, remote_region_id = 255
 # range2: [0x0000008000000000-0x0000027fffffffff], node = -2,
local_region_id = 1, remote_region_id = 255
 # range3: [0x0000028000000000-0x0000047fffffffff], node = -2,
local_region_id = 1, remote_region_id = 255
 # range4: [0x0000048000000000-0x000004ffffffffff], node = -2,
local_region_id = 1, remote_region_id = 255
Quoted 6 lines

[1]:
https://lore.kernel.org/lkml/[email protected]/

Assisted-by Sashiko:gemini-3.1-pro-preview
Assisted-by GitHub-Copilot:claude-opus-4-8
Quoted 6 lines

Looks good to me. Did you do pipe mode testing? Is there any additional testing coverage we can add?

Thanks,
Ian

Hi Ian, thanks for reviewing,

Yes, here is some example output in pipe mode:

./perf mem record -o - -a sleep 5 | ./perf script -i - --header

...

# memory ranges (nr 5):
# range0: [0x0000000000000000-0x00000000bfffffff], node = 0, local_region_id = 0, remote_region_id = 255
# range1: [0x0000000100000000-0x000000203fffffff], node = 0, local_region_id = 0, remote_region_id = 255
# range2: [0x0000008000000000-0x0000027fffffffff], node = -2, local_region_id = 1, remote_region_id = 255
# range3: [0x0000028000000000-0x0000047fffffffff], node = -2, local_region_id = 1, remote_region_id = 255
# range4: [0x0000048000000000-0x000004ffffffffff], node = -2, local_region_id = 1, remote_region_id = 255
            perf   12280 [007] 82283.852958:         63 cpu/mem-loads,ldlat=30/P: ff40225e7b1acfb8     10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region N/A               4             302              17 ffffffff9eafe901 perf_event_ctx_lock_nested.isra.0+0x31>
         swapper       0 [000] 82283.858694:         63 cpu/mem-loads,ldlat=30/P: ff40225e7d0b31c0    810668100842 |OP LOAD|LVL L3 hit|SNP None|TLB L1 or L2 hit|LCK No|BLK  N/A|Region Local-shared-cache              60              62               0 ffffffff9e858b3f update_sg_lb_stats+0x13f>
         swapper       0 [000] 82283.878498:         63 cpu/mem-loads,ldlat=30/P: ff40225e7d9b3f98   1810668800842 |OP LOAD|LVL L3 hit|SNP HitM|TLB L1 or L2 hit|LCK No|BLK  N/A|Region Other-non-shared-cache             434             435               0 ffffffff9e863e80 _nohz_idle_balance.i>
       perf-exec   12283 [023] 82283.882372:         63 cpu/mem-loads,ldlat=30/P: ffe698a847e86408   2011a68101042 |OP LOAD|LVL RAM hit|SNP None|TLB L1 or L2 hit|LCK No|BLK  N/A|Region Mem-0                             164             165               0 ffffffff9eb97335 zap_present_ptes.con>

I think we could test that perf c2c/script show the "Region..." field and perf c2c report --stdio and perf report/script --header print the memory range table, if supported.

Thanks,
Tom

Quoted 410 lines
Quoted 410 lines

---
v4:
-- Fixed bounds checking for local and remote region IDs in memory_range__read(), which made region "255" invalid

v3:
-- Added missing ff->size check in process_memory_ranges() -- used open/openat to read memory range sysfs files in memory_range__read()
-- Removed path name and file name buffers in memory_range__read(), instead pass path name from as a parameter from memory_range__parse()
-- use calloc instead of zalloc in memory_range__parse() and process_memory_ranges()
-- Removed redundant check for existance of memory range directory in
   memory_range__read()
-- Included examples of memory range output in commit message

v2:
-- Added check for NULL return of sysfs__mountpoint() when parsing memory ranges in sysfs
-- increased MAX_MEMORY_RANGES sanity check from 64 to 256 based on ACPI MRRM table implementation in Linux kernel
-- added comment for MAX_MEMORY_RANGES to clarify that it is a sanity check
   for malformed perf.data files
-- removed a line of code was removed from util/env.h but was added back in
   v1 due to bad rebase
---

 .../Documentation/perf.data-file-format.txt   |  13 ++ tools/perf/builtin-inject.c                   |   1 + tools/perf/util/env.c                         |   1 + tools/perf/util/env.h                         |  10 + tools/perf/util/header.c                      | 203
++++++++++++++++++
 tools/perf/util/header.h                      |   1 +

 6 files changed, 229 insertions(+)
tools/perf/Documentation/perf.data-file-format.txt +13 -0
index b90cba9168f8..1f8e7d8e9ae2 100644
@@ -478,6 +478,19 @@ The size of the cacheline in bytes. Format:
 
         unsigned int cln_size;
 
+       HEADER_MEMORY_RANGES = 35,
+
+List of memory ranges. The format of the data is as below.
+
+u32 nr_memory_ranges;
+struct memory_range {
+       u64 base;
+       u64 length;
+       int node;
+       u8 local_region_id;
+       u8 remote_region_id;
+}; [nr_memory_ranges]
+

        other bits are reserved and should be ignored for now HEADER_FEAT_BITS        = 256,

tools/perf/builtin- renamed from tools/perf/builtin-inject.c +0 -0

inject.c
index 70bbfad5653e..169617a884d3 100644

tools/perf/builtin-inject.c +0 -0
@@ -2306,6 +2306,7 @@ static bool keep_feat(struct perf_inject
 *inject, int feat)

        case HEADER_PMU_CAPS:
        case HEADER_CPU_DOMAIN_INFO:
        case HEADER_CLN_SIZE:
+       case HEADER_MEMORY_RANGES:
                return true;
        /* Information that can be updated */
        case HEADER_BUILD_ID:

tools/perf/util/env.c +1 -0
index c0e2b9d5f0b2..50993b0dedc5 100644
@@ -297,6 +297,7 @@ void perf_env__exit(struct perf_env *env)
                 zfree(&env->pmu_caps[i].pmu_name);
         }
         zfree(&env->pmu_caps);
+       zfree(&env->memory_ranges);

        free_cpu_domain_info(env->cpu_domain, env-

Quoted 1 line

schedstat_version, env->nr_cpus_avail);

 }

tools/perf/util/env.h +10 -0
index 7acca39b42ff..23d35425efe3 100644
@@ -68,6 +68,14 @@ struct cpu_domain_map {
         struct domain_info      **domains;
  };
 
+struct memory_range {
+       u64 base;
+       u64 length;
+       int node;
+       u8 local_region_id;
+       u8 remote_region_id;
+};
+
  struct perf_env {
         char                    *hostname;
         char                    *os_release;
@@ -100,6 +108,7 @@ struct perf_env {
         int                     nr_cpu_pmu_caps;
         int                     nr_hybrid_nodes;
         int                     nr_pmus_with_caps;
+       int                     nr_memory_ranges;
         char                    *cmdline;
         const char              **cmdline_argv;
         char                    *sibling_cores;
@@ -122,6 +131,7 @@ struct perf_env {
         unsigned long long       memory_bsize;
         struct hybrid_node      *hybrid_nodes;
         struct pmu_caps         *pmu_caps;
+       struct memory_range     *memory_ranges;
  #ifdef HAVE_LIBBPF_SUPPORT
         /*
          * bpf_info_lock protects bpf rbtrees. This is needed

because the

tools/perf/util/header.c +69 -0
index e90e541f546b..1dbc83a2d798 100644
@@ -92,6 +92,7 @@
  #define MAX_PMU_CAPS           512
  #define MAX_PMU_MAPPINGS       4096
  #define MAX_SCHED_DOMAINS      64
+#define MAX_MEMORY_RANGES      256
 
  /*
   * magic2 = "PERFILE2"
@@ -1891,6 +1892,132 @@ static int write_cpu_domain_info(struct
 feat_fd *ff,
         return ret;
  }
 
+static int memory_range__read(struct memory_range *range, const
 char *path)
+{
+       char buf[32];
+       ssize_t n;
+       int dfd, fd, tmp, ret = -1;
+
+       dfd = open(path, O_RDONLY | O_DIRECTORY);
+       if (dfd < 0)
+               return -1;
+
+#define _R(name, dst, conv)                    \
+       fd = openat(dfd, name, O_RDONLY);       \
+       if (fd < 0)                             \
+               goto out;                       \
+       n = read(fd, buf, sizeof(buf) - 1);     \
+       close(fd);                              \
+       if (n <= 0)                             \
+               goto out;                       \
+       buf[n] = '\0';                          \
+       dst = conv(buf, NULL, 0);
+
+       _R("base", range->base, strtoull);
+       _R("length", range->length, strtoull);
+       _R("node", range->node, strtol);
+       _R("local_region_id", tmp, strtol);
+       if (tmp < 0 || tmp > UINT8_MAX)
+               goto out;
+       range->local_region_id = tmp;
+       _R("remote_region_id", tmp, strtol);
+       if (tmp < 0 || tmp > UINT8_MAX)
+               goto out;
+       range->remote_region_id = tmp;
+#undef _R
+
+       ret = 0;
+out:
+       close(dfd);
+       return ret;
+}
+
+static int memory_range__parse(struct memory_range **ranges)
+{
+       const char *sysfs = sysfs__mountpoint();
+       int i, err, nr_memory_ranges = 0;
+       char path[PATH_MAX];
+       struct stat st;
+
+       if (!sysfs)
+               return 0;
+
+       scnprintf(path, PATH_MAX, "%s/firmware/acpi/memory_ranges",
 sysfs);
+       if (stat(path, &st))
+               return 0;
+
+       while (1) {
+               scnprintf(path, PATH_MAX,
+                         "%s/firmware/acpi/memory_ranges/range%d",
+                         sysfs, nr_memory_ranges);
+               if (stat(path, &st))
+                       break;
+
+               nr_memory_ranges++;
+       }
+
+       if (nr_memory_ranges == 0)
+               return 0;
+
+       *ranges = calloc(nr_memory_ranges, sizeof(struct

memory_range));
+       if (!(*ranges))
+               return -ENOMEM;

tools/perf/util/header.c +6 -0
+
+       for (i = 0; i < nr_memory_ranges; i++) {
+               struct memory_range range;
+
+               scnprintf(path, PATH_MAX,
+                         "%s/firmware/acpi/memory_ranges/range%d",

sysfs, i);
+               err = memory_range__read(&range, path); +               if (err < 0)
+                       goto out_error;

tools/perf/util/header.c +22 -0
+
+               (*ranges)[i] = range;
+       }
+
+       return nr_memory_ranges;
+
+out_error:
+       zfree(ranges);
+       return -1;
+}
+
+static int write_memory_ranges(struct feat_fd *ff,
+                        struct evlist *evlist __maybe_unused)
+{
+       struct memory_range *ranges = NULL;
+       int nr_memory_ranges = 0, ret;
+
+       nr_memory_ranges = memory_range__parse(&ranges);
+       if (nr_memory_ranges < 0)
+               return nr_memory_ranges;
+
+       ret = do_write(ff, &nr_memory_ranges,

sizeof(nr_memory_ranges));
+       if (ret < 0)
+               goto out;
+
+       for (int i = 0; i < nr_memory_ranges; i++) {
+               ret = do_write(ff, &ranges[i].base, sizeof(u64)); +               if (ret < 0)
+                       goto out;
+               ret = do_write(ff, &ranges[i].length, sizeof(u64)); +               if (ret < 0)
+                       goto out;
+               ret = do_write(ff, &ranges[i].node, sizeof(u32)); +               if (ret < 0)
+                       goto out;
+               ret = do_write(ff, &ranges[i].local_region_id, sizeof(u8));
+               if (ret < 0)
+                       goto out;
+               ret = do_write(ff, &ranges[i].remote_region_id, sizeof(u8));
+               if (ret < 0)
+                       goto out;
+       }

tools/perf/util/header.c +5 -0
+out:
+       zfree(&ranges);
+       return ret;
+}
+

 static void print_hostname(struct feat_fd *ff, FILE *fp) {
        fprintf(fp, "# hostname : %s\n", ff->ph->env.hostname);

tools/perf/util/header.c +11 -0
@@ -2628,6 +2755,23 @@ static void print_cpu_domain_info(struct
 feat_fd *ff, FILE *fp)
         }
  }
 
+static void print_memory_ranges(struct feat_fd *ff, FILE *fp)
+{
+       struct memory_range *ranges = ff->ph->env.memory_ranges;
+       int nr_memory_ranges = ff->ph->env.nr_memory_ranges;
+       int i;
+
+       fprintf(fp, "# memory ranges (nr %d):\n",
 nr_memory_ranges);
+
+       for (i = 0; i < nr_memory_ranges; i++) {
+               fprintf(fp, "# range%u: [0x%016" PRIx64 "-0x%016"
 PRIx64,
+                       i, ranges[i].base, ranges[i].base +

ranges[i].length - 1);
+               fprintf(fp, "], node = %d, local_region_id = %d, remote_region_id = %d\n",
+                       ranges[i].node, ranges[i].local_region_id, +                       ranges[i].remote_region_id);
+       }

tools/perf/util/header.c +2 -0
+}
+

 static int __event_process_build_id(struct perf_record_header_build_id *bev,
                                    char *filename, struct perf_session *session)

tools/perf/util/header.c +0 -0
@@ -4201,6 +4345,64 @@ static int process_cpu_domain_info(struct

feat_fd *ff, void *data __maybe_unused
        return ret;
 }

+static int process_memory_ranges(struct feat_fd *ff, void *data __maybe_unused)

tools/perf/util/header.c +14 -0
+{
+       struct perf_env *env = &ff->ph->env;
+       struct memory_range *ranges, *r;
+       u32 nr_memory_ranges, i;
+
+       if (do_read_u32(ff, &nr_memory_ranges))
+               return -1;
+
+       if (!nr_memory_ranges) {
+               pr_debug("memory ranges not available\n");
+               return 0;
+       }
+
+       /* According to version 1.1 of the ACPI MRRM table, the

maximum
+        * number of memory regions can be at most 255. Do a sanity check
+        * here to guard against a malformed perf.data file. +        */
+       if (nr_memory_ranges >= MAX_MEMORY_RANGES) {
+               pr_err("Invalid memory_ranges: nr_memory_ranges (%u) > %u\n",
+                      nr_memory_ranges, MAX_MEMORY_RANGES); +               return -1;
+       }
+
+       if (ff->size < sizeof(u32) + nr_memory_ranges * (2 * sizeof(u64) + sizeof(u32) + 2 * sizeof(u8))) {
+               pr_err("Invalid HEADER_MEMORY_RANGES: section too small (%zu) for %u range entries\n",
+                      ff->size, nr_memory_ranges);
+               return -1;
+       }

tools/perf/util/header.c +17 -0
+
+       ranges = calloc(nr_memory_ranges, sizeof(*ranges));
+       if (!ranges)
+               return -1;
+
+       for (i = 0; i < nr_memory_ranges; i++) {
+               r = &ranges[i];
+
+               if (do_read_u64(ff, &r->base))
+                       goto error;
+               if (do_read_u64(ff, &r->length))
+                       goto error;
+               if (do_read_u32(ff, (u32 *) &r->node))
+                       goto error;
+               if (__do_read(ff, &r->local_region_id, sizeof(u8)))
+                       goto error;
+               if (__do_read(ff, &r->remote_region_id,

sizeof(u8)))
+                       goto error;
+       }

tools/perf/util/header.c +10 -0
+
+       env->memory_ranges = ranges;
+       env->nr_memory_ranges = nr_memory_ranges;
+
+       return 0;
+error:
+       zfree(&ranges);
+       return -1;
+}
+

 #define FEAT_OPR(n, func, __full_only) \
        [HEADER_##n] = {                                        \
                .name       = __stringify(n),                   \

tools/perf/util/header.c +1 -0
@@ -4265,6 +4467,7 @@ const struct perf_header_feature_ops
 feat_ops[HEADER_LAST_FEATURE] = {
         FEAT_OPR(CPU_DOMAIN_INFO,       cpu_domain_info,       
 true),
         FEAT_OPR(E_MACHINE,     e_machine,      false),
         FEAT_OPR(CLN_SIZE,      cln_size,       false),
+       FEAT_OPR(MEMORY_RANGES, memory_ranges,  false),
  };

 struct header_print_data {

tools/perf/util/header.h +1 -1
index 5e03f884b7cc..765037762758 100644
@@ -56,6 +56,7 @@ enum {
         HEADER_CPU_DOMAIN_INFO,
         HEADER_E_MACHINE,
         HEADER_CLN_SIZE,
+       HEADER_MEMORY_RANGES,
         HEADER_LAST_FEATURE,
         HEADER_FEAT_BITS        = 256,
  };
--

2.55.0

reviewer Reviewed-by
#8 lore raw

Re: [PATCH v4 3/6] perf header: Support memory ranges

Quoted 73 lines · Thomas
Quoted 46 lines · Ian Rogers
Quoted 37 lines · Thomas Falcon

Memory ranges were created to track different types of memory, such as persistent, high bandwidth, or CXL-attached, that may be present on a system for performance monitoring and resource control purposes.

Memory range data are parsed from the ACPI MRRM table and exposed to userspace tools via sysfs [1]. Memory range data is read from:

/sys/firmware/acpi/memory_ranges/rangeX

With the following attributes:

        u64 base;
        u64 length;
        int node;
        u8  local_region_id;
        u8  remote_region_id;

Read memory range data from sysfs if present and save it in the header of the perf data file under a new feature bit, HEADER_MEMORY_RANGES (35). Memory range data can be viewed with the --header or --header-only options of perf-report and perf-script.

Example output:

 # memory ranges (nr 5):
 # range0: [0x0000000000000000-0x00000000bfffffff], node = 0,
local_region_id = 0, remote_region_id = 255
 # range1: [0x0000000100000000-0x000000203fffffff], node = 0,
local_region_id = 0, remote_region_id = 255
 # range2: [0x0000008000000000-0x0000027fffffffff], node = -2,
local_region_id = 1, remote_region_id = 255
 # range3: [0x0000028000000000-0x0000047fffffffff], node = -2,
local_region_id = 1, remote_region_id = 255
 # range4: [0x0000048000000000-0x000004ffffffffff], node = -2,
local_region_id = 1, remote_region_id = 255
Quoted 6 lines

[1]:
https://lore.kernel.org/lkml/[email protected]/

Assisted-by Sashiko:gemini-3.1-pro-preview
Assisted-by GitHub-Copilot:claude-opus-4-8
Quoted 6 lines

Looks good to me. Did you do pipe mode testing? Is there any additional testing coverage we can add?

Thanks,
Ian

Hi Ian, thanks for reviewing,

Yes, here is some example output in pipe mode:

./perf mem record -o - -a sleep 5 | ./perf script -i - --header

...

# memory ranges (nr 5):
# range0: [0x0000000000000000-0x00000000bfffffff], node = 0, local_region_id = 0, remote_region_id = 255
# range1: [0x0000000100000000-0x000000203fffffff], node = 0, local_region_id = 0, remote_region_id = 255
# range2: [0x0000008000000000-0x0000027fffffffff], node = -2, local_region_id = 1, remote_region_id = 255
# range3: [0x0000028000000000-0x0000047fffffffff], node = -2, local_region_id = 1, remote_region_id = 255
# range4: [0x0000048000000000-0x000004ffffffffff], node = -2, local_region_id = 1, remote_region_id = 255
            perf   12280 [007] 82283.852958:         63 cpu/mem-loads,ldlat=30/P: ff40225e7b1acfb8     10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region N/A               4             302              17 ffffffff9eafe901 perf_event_ctx_lock_nested.isra.0+0x31>
         swapper       0 [000] 82283.858694:         63 cpu/mem-loads,ldlat=30/P: ff40225e7d0b31c0    810668100842 |OP LOAD|LVL L3 hit|SNP None|TLB L1 or L2 hit|LCK No|BLK  N/A|Region Local-shared-cache              60              62               0 ffffffff9e858b3f update_sg_lb_stats+0x13f>
         swapper       0 [000] 82283.878498:         63 cpu/mem-loads,ldlat=30/P: ff40225e7d9b3f98   1810668800842 |OP LOAD|LVL L3 hit|SNP HitM|TLB L1 or L2 hit|LCK No|BLK  N/A|Region Other-non-shared-cache             434             435               0 ffffffff9e863e80 _nohz_idle_balance.i>
       perf-exec   12283 [023] 82283.882372:         63 cpu/mem-loads,ldlat=30/P: ffe698a847e86408   2011a68101042 |OP LOAD|LVL RAM hit|SNP None|TLB L1 or L2 hit|LCK No|BLK  N/A|Region Mem-0                             164             165               0 ffffffff9eb97335 zap_present_ptes.con>
Quoted 1 line

I think we could test that perf c2c/script show the "Region..." field and perf c2c report --stdio and perf report/script --header print the memory range table, if supported.

Thanks Thomas!

Ian

Quoted 415 lines

Thanks,
Tom

Quoted 410 lines
Quoted 410 lines

---
v4:
-- Fixed bounds checking for local and remote region IDs in
memory_range__read(), which made region "255" invalid

v3:
-- Added missing ff->size check in process_memory_ranges() -- used open/openat to read memory range sysfs files in memory_range__read()
-- Removed path name and file name buffers in memory_range__read(),
instead pass path name from as a parameter from
memory_range__parse()
-- use calloc instead of zalloc in memory_range__parse() and
process_memory_ranges()
-- Removed redundant check for existance of memory range directory in
memory_range__read()
-- Included examples of memory range output in commit message

v2:
-- Added check for NULL return of sysfs__mountpoint() when parsing
memory ranges in sysfs
-- increased MAX_MEMORY_RANGES sanity check from 64 to 256 based on
ACPI MRRM table implementation in Linux kernel
-- added comment for MAX_MEMORY_RANGES to clarify that it is a sanity check
for malformed perf.data files
-- removed a line of code was removed from util/env.h but was added back in
v1 due to bad rebase
---

 .../Documentation/perf.data-file-format.txt   |  13 ++
 tools/perf/builtin-inject.c                   |   1 +
 tools/perf/util/env.c                         |   1 +
 tools/perf/util/env.h                         |  10 +
 tools/perf/util/header.c                      | 203
++++++++++++++++++
 tools/perf/util/header.h                      |   1 +
 6 files changed, 229 insertions(+)
tools/perf/Documentation/perf.data-file-format.txt +13 -0
index b90cba9168f8..1f8e7d8e9ae2 100644
@@ -478,6 +478,19 @@ The size of the cacheline in bytes. Format:
 
        unsigned int cln_size;
 
+       HEADER_MEMORY_RANGES = 35,
+
+List of memory ranges. The format of the data is as below.
+
+u32 nr_memory_ranges;
+struct memory_range {
+       u64 base;
+       u64 length;
+       int node;
+       u8 local_region_id;
+       u8 remote_region_id;
+}; [nr_memory_ranges]
+
        other bits are reserved and should be ignored for now
        HEADER_FEAT_BITS        = 256,
 
tools/perf/builtin- renamed from tools/perf/builtin-inject.c +0 -0

inject.c
index 70bbfad5653e..169617a884d3 100644

tools/perf/builtin-inject.c +1 -0
@@ -2306,6 +2306,7 @@ static bool keep_feat(struct perf_inject
 *inject, int feat)
        case HEADER_PMU_CAPS:
        case HEADER_CPU_DOMAIN_INFO:
        case HEADER_CLN_SIZE:
+       case HEADER_MEMORY_RANGES:
                return true;
        /* Information that can be updated */

case HEADER_BUILD_ID:

tools/perf/util/env.c +1 -0
index c0e2b9d5f0b2..50993b0dedc5 100644
@@ -297,6 +297,7 @@ void perf_env__exit(struct perf_env *env)
                zfree(&env->pmu_caps[i].pmu_name);
        }
        zfree(&env->pmu_caps);
+       zfree(&env->memory_ranges);
        free_cpu_domain_info(env->cpu_domain, env-
 >schedstat_version, env->nr_cpus_avail);
 }
tools/perf/util/env.h +10 -0
index 7acca39b42ff..23d35425efe3 100644
@@ -68,6 +68,14 @@ struct cpu_domain_map {
        struct domain_info      **domains;
 };
 
+struct memory_range {
+       u64 base;
+       u64 length;
+       int node;
+       u8 local_region_id;
+       u8 remote_region_id;
+};
+
 struct perf_env {
        char                    *hostname;
        char                    *os_release;
@@ -100,6 +108,7 @@ struct perf_env {
        int                     nr_cpu_pmu_caps;
        int                     nr_hybrid_nodes;
        int                     nr_pmus_with_caps;
+       int                     nr_memory_ranges;
        char                    *cmdline;
        const char              **cmdline_argv;
        char                    *sibling_cores;
@@ -122,6 +131,7 @@ struct perf_env {
        unsigned long long       memory_bsize;
        struct hybrid_node      *hybrid_nodes;
        struct pmu_caps         *pmu_caps;
+       struct memory_range     *memory_ranges;
 #ifdef HAVE_LIBBPF_SUPPORT
        /*
         * bpf_info_lock protects bpf rbtrees. This is needed

because the

tools/perf/util/header.c +69 -0
index e90e541f546b..1dbc83a2d798 100644
@@ -92,6 +92,7 @@
 #define MAX_PMU_CAPS           512
 #define MAX_PMU_MAPPINGS       4096
 #define MAX_SCHED_DOMAINS      64
+#define MAX_MEMORY_RANGES      256
 
 /*
  * magic2 = "PERFILE2"
@@ -1891,6 +1892,132 @@ static int write_cpu_domain_info(struct
 feat_fd *ff,
        return ret;
 }
 
+static int memory_range__read(struct memory_range *range, const
 char *path)
+{
+       char buf[32];
+       ssize_t n;
+       int dfd, fd, tmp, ret = -1;
+
+       dfd = open(path, O_RDONLY | O_DIRECTORY);
+       if (dfd < 0)
+               return -1;
+
+#define _R(name, dst, conv)                    \
+       fd = openat(dfd, name, O_RDONLY);       \
+       if (fd < 0)                             \
+               goto out;                       \
+       n = read(fd, buf, sizeof(buf) - 1);     \
+       close(fd);                              \
+       if (n <= 0)                             \
+               goto out;                       \
+       buf[n] = '\0';                          \
+       dst = conv(buf, NULL, 0);
+
+       _R("base", range->base, strtoull);
+       _R("length", range->length, strtoull);
+       _R("node", range->node, strtol);
+       _R("local_region_id", tmp, strtol);
+       if (tmp < 0 || tmp > UINT8_MAX)
+               goto out;
+       range->local_region_id = tmp;
+       _R("remote_region_id", tmp, strtol);
+       if (tmp < 0 || tmp > UINT8_MAX)
+               goto out;
+       range->remote_region_id = tmp;
+#undef _R
+
+       ret = 0;
+out:
+       close(dfd);
+       return ret;
+}
+
+static int memory_range__parse(struct memory_range **ranges)
+{
+       const char *sysfs = sysfs__mountpoint();
+       int i, err, nr_memory_ranges = 0;
+       char path[PATH_MAX];
+       struct stat st;
+
+       if (!sysfs)
+               return 0;
+
+       scnprintf(path, PATH_MAX, "%s/firmware/acpi/memory_ranges",
 sysfs);
+       if (stat(path, &st))
+               return 0;
+
+       while (1) {
+               scnprintf(path, PATH_MAX,
+                         "%s/firmware/acpi/memory_ranges/range%d",
+                         sysfs, nr_memory_ranges);
+               if (stat(path, &st))
+                       break;
+
+               nr_memory_ranges++;
+       }
+
+       if (nr_memory_ranges == 0)
+               return 0;
+
+       *ranges = calloc(nr_memory_ranges, sizeof(struct

memory_range));

tools/perf/util/header.c +8 -0
+       if (!(*ranges))
+               return -ENOMEM;
+
+       for (i = 0; i < nr_memory_ranges; i++) {
+               struct memory_range range;
+
+               scnprintf(path, PATH_MAX,
+                         "%s/firmware/acpi/memory_ranges/range%d",

sysfs, i);

tools/perf/util/header.c +25 -0
+               err = memory_range__read(&range, path);
+               if (err < 0)
+                       goto out_error;
+
+               (*ranges)[i] = range;
+       }
+
+       return nr_memory_ranges;
+
+out_error:
+       zfree(ranges);
+       return -1;
+}
+
+static int write_memory_ranges(struct feat_fd *ff,
+                        struct evlist *evlist __maybe_unused)
+{
+       struct memory_range *ranges = NULL;
+       int nr_memory_ranges = 0, ret;
+
+       nr_memory_ranges = memory_range__parse(&ranges);
+       if (nr_memory_ranges < 0)
+               return nr_memory_ranges;
+
+       ret = do_write(ff, &nr_memory_ranges,

sizeof(nr_memory_ranges));

tools/perf/util/header.c +14 -0
+       if (ret < 0)
+               goto out;
+
+       for (int i = 0; i < nr_memory_ranges; i++) {
+               ret = do_write(ff, &ranges[i].base, sizeof(u64));
+               if (ret < 0)
+                       goto out;
+               ret = do_write(ff, &ranges[i].length, sizeof(u64));
+               if (ret < 0)
+                       goto out;
+               ret = do_write(ff, &ranges[i].node, sizeof(u32));
+               if (ret < 0)
+                       goto out;
+               ret = do_write(ff, &ranges[i].local_region_id,

sizeof(u8));

tools/perf/util/header.c +3 -0
+               if (ret < 0)
+                       goto out;
+               ret = do_write(ff, &ranges[i].remote_region_id,

sizeof(u8));

tools/perf/util/header.c +19 -0
+               if (ret < 0)
+                       goto out;
+       }
+out:
+       zfree(&ranges);
+       return ret;
+}
+
 static void print_hostname(struct feat_fd *ff, FILE *fp)
 {
        fprintf(fp, "# hostname : %s\n", ff->ph->env.hostname);
@@ -2628,6 +2755,23 @@ static void print_cpu_domain_info(struct
 feat_fd *ff, FILE *fp)
        }
 }
 
+static void print_memory_ranges(struct feat_fd *ff, FILE *fp)
+{
+       struct memory_range *ranges = ff->ph->env.memory_ranges;
+       int nr_memory_ranges = ff->ph->env.nr_memory_ranges;
+       int i;
+
+       fprintf(fp, "# memory ranges (nr %d):\n",
 nr_memory_ranges);
+
+       for (i = 0; i < nr_memory_ranges; i++) {
+               fprintf(fp, "# range%u: [0x%016" PRIx64 "-0x%016"
 PRIx64,
+                       i, ranges[i].base, ranges[i].base +

ranges[i].length - 1);
+ fprintf(fp, "], node = %d, local_region_id = %d,
remote_region_id = %d\n",

tools/perf/util/header.c +5 -0
+                       ranges[i].node, ranges[i].local_region_id,
+                       ranges[i].remote_region_id);
+       }
+}
+
 static int __event_process_build_id(struct

perf_record_header_build_id *bev,
char *filename, struct perf_session *session)

tools/perf/util/header.c +0 -0
@@ -4201,6 +4345,64 @@ static int process_cpu_domain_info(struct

feat_fd *ff, void *data __maybe_unused

        return ret;

}

+static int process_memory_ranges(struct feat_fd *ff, void *data __maybe_unused)

tools/perf/util/header.c +14 -0
+{
+       struct perf_env *env = &ff->ph->env;
+       struct memory_range *ranges, *r;
+       u32 nr_memory_ranges, i;
+
+       if (do_read_u32(ff, &nr_memory_ranges))
+               return -1;
+
+       if (!nr_memory_ranges) {
+               pr_debug("memory ranges not available\n");
+               return 0;
+       }
+
+       /* According to version 1.1 of the ACPI MRRM table, the

maximum
+ * number of memory regions can be at most 255. Do a sanity
check

tools/perf/util/header.c +4 -0
+        * here to guard against a malformed perf.data file.
+        */
+       if (nr_memory_ranges >= MAX_MEMORY_RANGES) {
+               pr_err("Invalid memory_ranges: nr_memory_ranges

(%u) > %u\n",

tools/perf/util/header.c +5 -0
+                      nr_memory_ranges, MAX_MEMORY_RANGES);
+               return -1;
+       }
+
+       if (ff->size < sizeof(u32) + nr_memory_ranges * (2 *

sizeof(u64) + sizeof(u32) + 2 * sizeof(u8))) {
+ pr_err("Invalid HEADER_MEMORY_RANGES: section too
small (%zu) for %u range entries\n",

tools/perf/util/header.c +20 -0
+                      ff->size, nr_memory_ranges);
+               return -1;
+       }
+
+       ranges = calloc(nr_memory_ranges, sizeof(*ranges));
+       if (!ranges)
+               return -1;
+
+       for (i = 0; i < nr_memory_ranges; i++) {
+               r = &ranges[i];
+
+               if (do_read_u64(ff, &r->base))
+                       goto error;
+               if (do_read_u64(ff, &r->length))
+                       goto error;
+               if (do_read_u32(ff, (u32 *) &r->node))
+                       goto error;
+               if (__do_read(ff, &r->local_region_id, sizeof(u8)))
+                       goto error;
+               if (__do_read(ff, &r->remote_region_id,

sizeof(u8)))

tools/perf/util/header.c +13 -0
+                       goto error;
+       }
+
+       env->memory_ranges = ranges;
+       env->nr_memory_ranges = nr_memory_ranges;
+
+       return 0;
+error:
+       zfree(&ranges);
+       return -1;
+}
+
 #define FEAT_OPR(n, func, __full_only) \
        [HEADER_##n] = {                                        \
                .name       = __stringify(n),                   \
@@ -4265,6 +4467,7 @@ const struct perf_header_feature_ops
 feat_ops[HEADER_LAST_FEATURE] = {
        FEAT_OPR(CPU_DOMAIN_INFO,       cpu_domain_info,
 true),
        FEAT_OPR(E_MACHINE,     e_machine,      false),
        FEAT_OPR(CLN_SIZE,      cln_size,       false),
+       FEAT_OPR(MEMORY_RANGES, memory_ranges,  false),
 };

struct header_print_data {

tools/perf/util/header.h +1 -1
index 5e03f884b7cc..765037762758 100644
@@ -56,6 +56,7 @@ enum {
        HEADER_CPU_DOMAIN_INFO,
        HEADER_E_MACHINE,
        HEADER_CLN_SIZE,
+       HEADER_MEMORY_RANGES,
        HEADER_LAST_FEATURE,
        HEADER_FEAT_BITS        = 256,
 };
--

2.55.0

#9 lore raw

PATCH v4 4/6 perf tools: Show memory region in perf-c2c subcommand

In reply to Thomas Falcon

From: Dapeng Mi <[email protected]>

Add memory region field to the cacheline list view to help users identify the memory region to which the cacheline belongs. The memory region field was included with the introduction of support for the Off-module Response facility (OMR) [1] in Intel's Diamond Rapids and Nova Lake architectures.

An example of the new perf c2c output including the memory region field is shown below:

Shared Data Cache Line Table (112 entries, sorted on Total HITMs)

       --------------- Cacheline --------------      Tot  ------- Load Hitm -------    Total    Total    Total
Index             Address  Region  Node  PA cnt     Hitm    Total  LclHitm  RmtHitm  records    Loads   Stores
    0  0xff1b7032a1e8c5c0  N/A        1      47    2.25%       50       44        6     1024     1023        1
    1  0xff1b6ff3255bb880  N/A        0      11    1.62%       36       34        2       96       93        4
    2  0xff1b70328e3b9880  N/A        1      13    1.49%       33       33        0      100       91        9
    3  0xff1b70328b023800  N/A        1       1    1.40%       31       14       17       52       48        4
    4  0xff1b70328e3b9c00  N/A        1       1    1.31%       29       27        2       67       33       34
    5  0xff1b70328b0237c0  N/A        1       1    1.26%       28       10       18      154      150        4
    6  0xff1b6ff3255bbc00  N/A        0       1    1.13%       25       25        0       48       25       23
    7  0xff1b6ff3255bba40  N/A        0       1    0.99%       22       22        0       46       23       23
    8  0xff3ab9ba50255c80  N/A      N/A       0    0.77%       17       15        2       35       35        0
    9  0xff3ab9ba503e3040  N/A      N/A       0    0.77%       17       11        6       37       37        1
   10  0xff1b703289e88f40  N/A        1      33    0.72%       16        9        7       69       63        6
   11  0xff1b70328e3b9a40  N/A        1       1    0.68%       15       15        0       43       17       26
   12  0xff1b7032c9fd6a40  N/A        1      15    0.68%       15       15        0       57       54        4
   13  0xff1b7032a1e8c980  N/A        1      27    0.54%       12       11        1      761      761        0
   14  0xff1b70727ffd57c0  N/A        1       7    0.54%       12       12        0      219      218        1
   15  0xffffffffaefe2380  N/A        1       1    0.50%       11        8        3       14       14        0

Note: DMR simics does not support memory regions. Since the output is captured on SPR, the memory region field shows "N/A" for all cachelines.

[1]: https://lore.kernel.org/all/[email protected]/

Assisted-by Sashiko:gemini-3.1-pro-preview
Assisted-by GitHub-Copilot:claude-opus-4-8

---
v4: correctly handle output_str memory allocation failure

v3: make memory region reporting conditional on feature bit

---
 tools/perf/builtin-c2c.c | 93 +++++++++++++++++++++++++++++++++++-----
 1 file changed, 83 insertions(+), 10 deletions(-)
tools/perf/builtin-c2c.c +83 -10
index c9584dbedf77..8cd90524fe49 100644
@@ -74,6 +74,7 @@ struct c2c_hist_entry {
 	unsigned long		*nodeset;
 	struct c2c_stats	*node_stats;
 	unsigned int		 cacheline_idx;
+	unsigned int		 mem_region;
 
 	struct compute_stats	 cstats;
 
@@ -281,6 +282,18 @@ static void c2c_he__set_node(struct c2c_hist_entry *c2c_he,
 	}
 }
 
+static void c2c_he__set_mem_region(struct c2c_hist_entry *c2c_he,
+				   unsigned int mem_region)
+{
+	if (WARN_ONCE(mem_region > PERF_MEM_REGION_MEM7,
+		      "WARNING: invalid memory region ID\n"))
+		return;
+
+	/* Update mem_region only if it really accesses memory */
+	if (mem_region >= PERF_MEM_REGION_MMIO)
+		c2c_he->mem_region = mem_region;
+}
+
 static void compute_stats(struct c2c_hist_entry *c2c_he,
 			  struct c2c_stats *stats,
 			  u64 weight)
@@ -339,6 +352,7 @@ static int process_sample_event(const struct perf_tool *tool __maybe_unused,
 	struct addr_location al;
 	struct mem_info *mi = NULL;
 	struct callchain_cursor *cursor;
+	unsigned int mem_region;
 	int ret;
 
 	addr_location__init(&al);
@@ -366,6 +380,7 @@ static int process_sample_event(const struct perf_tool *tool __maybe_unused,
 	}
 
 	c2c_decode_stats(&stats, mi);
+	mem_region = mem_info__data_src(mi)->mem_region;
 
 	he = hists__add_entry_ops(&c2c_hists->hists, &c2c_entry_ops,
 				  &al, NULL, NULL, mi, NULL,
@@ -382,6 +397,7 @@ static int process_sample_event(const struct perf_tool *tool __maybe_unused,
 	c2c_he__set_cpu(c2c_he, sample);
 	c2c_he__set_node(c2c_he, sample);
 	c2c_he__set_evsel(c2c_he, evsel);
+	c2c_he__set_mem_region(c2c_he, mem_region);
 
 	hists__inc_nr_samples(&c2c_hists->hists, he->filtered);
 
@@ -435,6 +451,7 @@ static int process_sample_event(const struct perf_tool *tool __maybe_unused,
 		c2c_he__set_cpu(c2c_he, sample);
 		c2c_he__set_node(c2c_he, sample);
 		c2c_he__set_evsel(c2c_he, evsel);
+		c2c_he__set_mem_region(c2c_he, mem_region);
 
 		hists__inc_nr_samples(&c2c_hists->hists, he->filtered);
 		ret = hist_entry__append_callchain(he, sample);
@@ -603,6 +620,30 @@ dcacheline_node_count(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
 	return scnprintf(hpp->buf, hpp->size, "%*lu", width, c2c_he->paddr_cnt);
 }
 
+static int
+dcacheline_node_mem_region(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
+			   struct hist_entry *he)
+{
+	int width = c2c_width(fmt, hpp, he->hists);
+	struct c2c_hist_entry *c2c_he;
+	unsigned int mem_region;
+	char buf[20];
+
+	c2c_he = container_of(he, struct c2c_hist_entry, he);
+	mem_region = c2c_he->mem_region;
+
+	if (mem_region == PERF_MEM_REGION_NA)
+		scnprintf(buf, sizeof(buf),  "N/A");
+	/* mem_region could only be >= PERF_MEM_REGION_MMIO */
+	else if (mem_region == PERF_MEM_REGION_MMIO)
+		scnprintf(buf, sizeof(buf), "MMIO");
+	else
+		scnprintf(buf, sizeof(buf), "0x%x",
+			  mem_region - PERF_MEM_REGION_MEM0);
+
+	return scnprintf(hpp->buf, hpp->size, "%*s", width, buf);
+}
+
 static int offset_entry(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
 			struct hist_entry *he)
 {
@@ -1425,7 +1466,7 @@ cl_idx_empty_entry(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
 	}
 
 static struct c2c_dimension dim_dcacheline = {
-	.header		= HEADER_SPAN("--- Cacheline ----", "Address", 2),
+	.header		= HEADER_SPAN("--- Cacheline ----", "Address", 3),
 	.name		= "dcacheline",
 	.cmp		= dcacheline_cmp,
 	.entry		= dcacheline_entry,
@@ -1440,6 +1481,14 @@ static struct c2c_dimension dim_dcacheline_node = {
 	.width		= 4,
 };
 
+static struct c2c_dimension dim_dcacheline_mem_region = {
+	.header		= HEADER_LOW("Region"),
+	.name		= "dcacheline_mem_region",
+	.cmp		= empty_cmp,
+	.entry		= dcacheline_node_mem_region,
+	.width		= 6,
+};
+
 static struct c2c_dimension dim_dcacheline_count = {
 	.header		= HEADER_LOW("PA cnt"),
 	.name		= "dcacheline_count",
@@ -1871,6 +1920,7 @@ static struct c2c_dimension dim_dcacheline_num_empty = {
 
 static struct c2c_dimension *dimensions[] = {
 	&dim_dcacheline,
+	&dim_dcacheline_mem_region,
 	&dim_dcacheline_node,
 	&dim_dcacheline_count,
 	&dim_offset,
@@ -2898,8 +2948,9 @@ static int ui_quirks(void)
 	/* Fix the zero line for dcacheline column. */
 	buf = fill_line(chk_double_cl ? "Double-Cacheline" : "Cacheline",
 				dim_dcacheline.width +
+				dim_dcacheline_mem_region.width +
 				dim_dcacheline_node.width +
-				dim_dcacheline_count.width + 4);
+				dim_dcacheline_count.width + 6);
 	if (!buf)
 		return -ENOMEM;
 
@@ -3151,8 +3202,10 @@ static int perf_c2c__report(int argc, const char **argv)
 	OPT_END()
 	};
 	int err = 0;
-	const char *output_str, *sort_str = NULL;
+	const char *sort_str = NULL;
+	char *output_str = NULL;
 	struct perf_env *env;
+	bool has_mem_regions;
 
 	annotation_options__init();
 
@@ -3316,9 +3369,14 @@ static int perf_c2c__report(int argc, const char **argv)
 		goto out_mem2node;
 	}
 
-	if (c2c.display != DISPLAY_SNP_PEER)
-		output_str = "cl_idx,"
+	has_mem_regions = perf_header__has_feat(&session->header,
+						HEADER_MEMORY_RANGES);
+
+	if (c2c.display != DISPLAY_SNP_PEER) {
+		if (asprintf(&output_str,
+			     "cl_idx,"
 			     "dcacheline,"
+			     "%s"
 			     "dcacheline_node,"
 			     "dcacheline_count,"
 			     "percent_costly_snoop,"
@@ -3330,10 +3388,17 @@ static int perf_c2c__report(int argc, const char **argv)
 			     "ld_fbhit,ld_l1hit,ld_l2hit,"
 			     "ld_lclhit,lcl_hitm,"
 			     "ld_rmthit,rmt_hitm,"
-			     "dram_lcl,dram_rmt";
-	else
-		output_str = "cl_idx,"
+			     "dram_lcl,dram_rmt",
+			     has_mem_regions ?
+			     "dcacheline_mem_region," : "") < 0) {
+			err = -ENOMEM;
+			goto out_mem2node;
+		}
+	} else {
+		if (asprintf(&output_str,
+			     "cl_idx,"
 			     "dcacheline,"
+			     "%s"
 			     "dcacheline_node,"
 			     "dcacheline_count,"
 			     "percent_costly_snoop,"
@@ -3345,7 +3410,13 @@ static int perf_c2c__report(int argc, const char **argv)
 			     "ld_fbhit,ld_l1hit,ld_l2hit,"
 			     "ld_lclhit,lcl_hitm,"
 			     "ld_rmthit,rmt_hitm,"
-			     "dram_lcl,dram_rmt";
+			     "dram_lcl,dram_rmt",
+			     has_mem_regions ?
+			     "dcacheline_mem_region," : "") < 0) {
+			err = -ENOMEM;
+			goto out_mem2node;
+		}
+	}
 
 	if (c2c.display == DISPLAY_TOT_HITM)
 		sort_str = "tot_hitm";
@@ -3368,11 +3439,13 @@ static int perf_c2c__report(int argc, const char **argv)
 
 	if (ui_quirks()) {
 		pr_err("failed to setup UI\n");
-		goto out_mem2node;
+		goto out_str;
 	}
 
 	perf_c2c_display(session);
 
+out_str:
+	free(output_str);
 out_mem2node:
 	mem2node__exit(&c2c.mem2node);
 out_session:
Signature · 1 line
2.55.0
reviewer Reviewed-by on v4 4/6
#10 lore raw

Re: [PATCH v4 4/6] perf tools: Show memory region in perf-c2c subcommand

Quoted 35 lines · Thomas Falcon

From: Dapeng Mi <[email protected]>

Add memory region field to the cacheline list view to help users identify the memory region to which the cacheline belongs. The memory region field was included with the introduction of support for the Off-module Response facility (OMR) [1] in Intel's Diamond Rapids and Nova Lake architectures.

An example of the new perf c2c output including the memory region field is shown below:

Shared Data Cache Line Table (112 entries, sorted on Total HITMs)

       --------------- Cacheline --------------      Tot  ------- Load Hitm -------    Total    Total    Total
Index             Address  Region  Node  PA cnt     Hitm    Total  LclHitm  RmtHitm  records    Loads   Stores
    0  0xff1b7032a1e8c5c0  N/A        1      47    2.25%       50       44        6     1024     1023        1
    1  0xff1b6ff3255bb880  N/A        0      11    1.62%       36       34        2       96       93        4
    2  0xff1b70328e3b9880  N/A        1      13    1.49%       33       33        0      100       91        9
    3  0xff1b70328b023800  N/A        1       1    1.40%       31       14       17       52       48        4
    4  0xff1b70328e3b9c00  N/A        1       1    1.31%       29       27        2       67       33       34
    5  0xff1b70328b0237c0  N/A        1       1    1.26%       28       10       18      154      150        4
    6  0xff1b6ff3255bbc00  N/A        0       1    1.13%       25       25        0       48       25       23
    7  0xff1b6ff3255bba40  N/A        0       1    0.99%       22       22        0       46       23       23
    8  0xff3ab9ba50255c80  N/A      N/A       0    0.77%       17       15        2       35       35        0
    9  0xff3ab9ba503e3040  N/A      N/A       0    0.77%       17       11        6       37       37        1
   10  0xff1b703289e88f40  N/A        1      33    0.72%       16        9        7       69       63        6
   11  0xff1b70328e3b9a40  N/A        1       1    0.68%       15       15        0       43       17       26
   12  0xff1b7032c9fd6a40  N/A        1      15    0.68%       15       15        0       57       54        4
   13  0xff1b7032a1e8c980  N/A        1      27    0.54%       12       11        1      761      761        0
   14  0xff1b70727ffd57c0  N/A        1       7    0.54%       12       12        0      219      218        1
   15  0xffffffffaefe2380  N/A        1       1    0.50%       11        8        3       14       14        0

Note: DMR simics does not support memory regions. Since the output is captured on SPR, the memory region field shows "N/A" for all cachelines.

Quoted 7 lines

[1]: https://lore.kernel.org/all/[email protected]/

Assisted-by Sashiko:gemini-3.1-pro-preview
Assisted-by GitHub-Copilot:claude-opus-4-8

Thanks,
Ian

Quoted 228 lines

---
v4: correctly handle output_str memory allocation failure

v3: make memory region reporting conditional on feature bit

---
 tools/perf/builtin-c2c.c | 93 +++++++++++++++++++++++++++++++++++-----
 1 file changed, 83 insertions(+), 10 deletions(-)
tools/perf/builtin-c2c.c +83 -11
index c9584dbedf77..8cd90524fe49 100644
@@ -74,6 +74,7 @@ struct c2c_hist_entry {
        unsigned long           *nodeset;
        struct c2c_stats        *node_stats;
        unsigned int             cacheline_idx;
+       unsigned int             mem_region;
 
        struct compute_stats     cstats;
 
@@ -281,6 +282,18 @@ static void c2c_he__set_node(struct c2c_hist_entry *c2c_he,
        }
 }
 
+static void c2c_he__set_mem_region(struct c2c_hist_entry *c2c_he,
+                                  unsigned int mem_region)
+{
+       if (WARN_ONCE(mem_region > PERF_MEM_REGION_MEM7,
+                     "WARNING: invalid memory region ID\n"))
+               return;
+
+       /* Update mem_region only if it really accesses memory */
+       if (mem_region >= PERF_MEM_REGION_MMIO)
+               c2c_he->mem_region = mem_region;
+}
+
 static void compute_stats(struct c2c_hist_entry *c2c_he,
                          struct c2c_stats *stats,
                          u64 weight)
@@ -339,6 +352,7 @@ static int process_sample_event(const struct perf_tool *tool __maybe_unused,
        struct addr_location al;
        struct mem_info *mi = NULL;
        struct callchain_cursor *cursor;
+       unsigned int mem_region;
        int ret;
 
        addr_location__init(&al);
@@ -366,6 +380,7 @@ static int process_sample_event(const struct perf_tool *tool __maybe_unused,
        }
 
        c2c_decode_stats(&stats, mi);
+       mem_region = mem_info__data_src(mi)->mem_region;
 
        he = hists__add_entry_ops(&c2c_hists->hists, &c2c_entry_ops,
                                  &al, NULL, NULL, mi, NULL,
@@ -382,6 +397,7 @@ static int process_sample_event(const struct perf_tool *tool __maybe_unused,
        c2c_he__set_cpu(c2c_he, sample);
        c2c_he__set_node(c2c_he, sample);
        c2c_he__set_evsel(c2c_he, evsel);
+       c2c_he__set_mem_region(c2c_he, mem_region);
 
        hists__inc_nr_samples(&c2c_hists->hists, he->filtered);
 
@@ -435,6 +451,7 @@ static int process_sample_event(const struct perf_tool *tool __maybe_unused,
                c2c_he__set_cpu(c2c_he, sample);
                c2c_he__set_node(c2c_he, sample);
                c2c_he__set_evsel(c2c_he, evsel);
+               c2c_he__set_mem_region(c2c_he, mem_region);
 
                hists__inc_nr_samples(&c2c_hists->hists, he->filtered);
                ret = hist_entry__append_callchain(he, sample);
@@ -603,6 +620,30 @@ dcacheline_node_count(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
        return scnprintf(hpp->buf, hpp->size, "%*lu", width, c2c_he->paddr_cnt);
 }
 
+static int
+dcacheline_node_mem_region(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
+                          struct hist_entry *he)
+{
+       int width = c2c_width(fmt, hpp, he->hists);
+       struct c2c_hist_entry *c2c_he;
+       unsigned int mem_region;
+       char buf[20];
+
+       c2c_he = container_of(he, struct c2c_hist_entry, he);
+       mem_region = c2c_he->mem_region;
+
+       if (mem_region == PERF_MEM_REGION_NA)
+               scnprintf(buf, sizeof(buf),  "N/A");
+       /* mem_region could only be >= PERF_MEM_REGION_MMIO */
+       else if (mem_region == PERF_MEM_REGION_MMIO)
+               scnprintf(buf, sizeof(buf), "MMIO");
+       else
+               scnprintf(buf, sizeof(buf), "0x%x",
+                         mem_region - PERF_MEM_REGION_MEM0);
+
+       return scnprintf(hpp->buf, hpp->size, "%*s", width, buf);
+}
+
 static int offset_entry(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
                        struct hist_entry *he)
 {
@@ -1425,7 +1466,7 @@ cl_idx_empty_entry(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
        }
 
 static struct c2c_dimension dim_dcacheline = {
-       .header         = HEADER_SPAN("--- Cacheline ----", "Address", 2),
+       .header         = HEADER_SPAN("--- Cacheline ----", "Address", 3),
        .name           = "dcacheline",
        .cmp            = dcacheline_cmp,
        .entry          = dcacheline_entry,
@@ -1440,6 +1481,14 @@ static struct c2c_dimension dim_dcacheline_node = {
        .width          = 4,
 };
 
+static struct c2c_dimension dim_dcacheline_mem_region = {
+       .header         = HEADER_LOW("Region"),
+       .name           = "dcacheline_mem_region",
+       .cmp            = empty_cmp,
+       .entry          = dcacheline_node_mem_region,
+       .width          = 6,
+};
+
 static struct c2c_dimension dim_dcacheline_count = {
        .header         = HEADER_LOW("PA cnt"),
        .name           = "dcacheline_count",
@@ -1871,6 +1920,7 @@ static struct c2c_dimension dim_dcacheline_num_empty = {
 
 static struct c2c_dimension *dimensions[] = {
        &dim_dcacheline,
+       &dim_dcacheline_mem_region,
        &dim_dcacheline_node,
        &dim_dcacheline_count,
        &dim_offset,
@@ -2898,8 +2948,9 @@ static int ui_quirks(void)
        /* Fix the zero line for dcacheline column. */
        buf = fill_line(chk_double_cl ? "Double-Cacheline" : "Cacheline",
                                dim_dcacheline.width +
+                               dim_dcacheline_mem_region.width +
                                dim_dcacheline_node.width +
-                               dim_dcacheline_count.width + 4);
+                               dim_dcacheline_count.width + 6);
        if (!buf)
                return -ENOMEM;
 
@@ -3151,8 +3202,10 @@ static int perf_c2c__report(int argc, const char **argv)
        OPT_END()
        };
        int err = 0;
-       const char *output_str, *sort_str = NULL;
+       const char *sort_str = NULL;
+       char *output_str = NULL;
        struct perf_env *env;
+       bool has_mem_regions;
 
        annotation_options__init();
 
@@ -3316,9 +3369,14 @@ static int perf_c2c__report(int argc, const char **argv)
                goto out_mem2node;
        }
 
-       if (c2c.display != DISPLAY_SNP_PEER)
-               output_str = "cl_idx,"
+       has_mem_regions = perf_header__has_feat(&session->header,
+                                               HEADER_MEMORY_RANGES);
+
+       if (c2c.display != DISPLAY_SNP_PEER) {
+               if (asprintf(&output_str,
+                            "cl_idx,"
                             "dcacheline,"
+                            "%s"
                             "dcacheline_node,"
                             "dcacheline_count,"
                             "percent_costly_snoop,"
@@ -3330,10 +3388,17 @@ static int perf_c2c__report(int argc, const char **argv)
                             "ld_fbhit,ld_l1hit,ld_l2hit,"
                             "ld_lclhit,lcl_hitm,"
                             "ld_rmthit,rmt_hitm,"
-                            "dram_lcl,dram_rmt";
-       else
-               output_str = "cl_idx,"
+                            "dram_lcl,dram_rmt",
+                            has_mem_regions ?
+                            "dcacheline_mem_region," : "") < 0) {
+                       err = -ENOMEM;
+                       goto out_mem2node;
+               }
+       } else {
+               if (asprintf(&output_str,
+                            "cl_idx,"
                             "dcacheline,"
+                            "%s"
                             "dcacheline_node,"
                             "dcacheline_count,"
                             "percent_costly_snoop,"
@@ -3345,7 +3410,13 @@ static int perf_c2c__report(int argc, const char **argv)
                             "ld_fbhit,ld_l1hit,ld_l2hit,"
                             "ld_lclhit,lcl_hitm,"
                             "ld_rmthit,rmt_hitm,"
-                            "dram_lcl,dram_rmt";
+                            "dram_lcl,dram_rmt",
+                            has_mem_regions ?
+                            "dcacheline_mem_region," : "") < 0) {
+                       err = -ENOMEM;
+                       goto out_mem2node;
+               }
+       }
 
        if (c2c.display == DISPLAY_TOT_HITM)
                sort_str = "tot_hitm";
@@ -3368,11 +3439,13 @@ static int perf_c2c__report(int argc, const char **argv)
 
        if (ui_quirks()) {
                pr_err("failed to setup UI\n");
-               goto out_mem2node;
+               goto out_str;
        }
 
        perf_c2c_display(session);
 
+out_str:
+       free(output_str);
 out_mem2node:
        mem2node__exit(&c2c.mem2node);
 out_session:
--

2.55.0

maintainer
#11 lore raw

Re: [PATCH v4 4/6] perf tools: Show memory region in perf-c2c subcommand

Hello,

Quoted 154 lines · Ian Rogers
Quoted 35 lines · Thomas Falcon

From: Dapeng Mi <[email protected]>

Add memory region field to the cacheline list view to help users identify the memory region to which the cacheline belongs. The memory region field was included with the introduction of support for the Off-module Response facility (OMR) [1] in Intel's Diamond Rapids and Nova Lake architectures.

An example of the new perf c2c output including the memory region field is shown below:

Shared Data Cache Line Table (112 entries, sorted on Total HITMs)

       --------------- Cacheline --------------      Tot  ------- Load Hitm -------    Total    Total    Total
Index             Address  Region  Node  PA cnt     Hitm    Total  LclHitm  RmtHitm  records    Loads   Stores
    0  0xff1b7032a1e8c5c0  N/A        1      47    2.25%       50       44        6     1024     1023        1
    1  0xff1b6ff3255bb880  N/A        0      11    1.62%       36       34        2       96       93        4
    2  0xff1b70328e3b9880  N/A        1      13    1.49%       33       33        0      100       91        9
    3  0xff1b70328b023800  N/A        1       1    1.40%       31       14       17       52       48        4
    4  0xff1b70328e3b9c00  N/A        1       1    1.31%       29       27        2       67       33       34
    5  0xff1b70328b0237c0  N/A        1       1    1.26%       28       10       18      154      150        4
    6  0xff1b6ff3255bbc00  N/A        0       1    1.13%       25       25        0       48       25       23
    7  0xff1b6ff3255bba40  N/A        0       1    0.99%       22       22        0       46       23       23
    8  0xff3ab9ba50255c80  N/A      N/A       0    0.77%       17       15        2       35       35        0
    9  0xff3ab9ba503e3040  N/A      N/A       0    0.77%       17       11        6       37       37        1
   10  0xff1b703289e88f40  N/A        1      33    0.72%       16        9        7       69       63        6
   11  0xff1b70328e3b9a40  N/A        1       1    0.68%       15       15        0       43       17       26
   12  0xff1b7032c9fd6a40  N/A        1      15    0.68%       15       15        0       57       54        4
   13  0xff1b7032a1e8c980  N/A        1      27    0.54%       12       11        1      761      761        0
   14  0xff1b70727ffd57c0  N/A        1       7    0.54%       12       12        0      219      218        1
   15  0xffffffffaefe2380  N/A        1       1    0.50%       11        8        3       14       14        0

Note: DMR simics does not support memory regions. Since the output is captured on SPR, the memory region field shows "N/A" for all cachelines.

Quoted 7 lines

[1]: https://lore.kernel.org/all/[email protected]/

Assisted-by Sashiko:gemini-3.1-pro-preview
Assisted-by GitHub-Copilot:claude-opus-4-8

Thanks,
Ian

Quoted 105 lines

---
v4: correctly handle output_str memory allocation failure

v3: make memory region reporting conditional on feature bit

---
 tools/perf/builtin-c2c.c | 93 +++++++++++++++++++++++++++++++++++-----
 1 file changed, 83 insertions(+), 10 deletions(-)
tools/perf/builtin-c2c.c +41 -0
index c9584dbedf77..8cd90524fe49 100644
@@ -74,6 +74,7 @@ struct c2c_hist_entry {
        unsigned long           *nodeset;
        struct c2c_stats        *node_stats;
        unsigned int             cacheline_idx;
+       unsigned int             mem_region;
 
        struct compute_stats     cstats;
 
@@ -281,6 +282,18 @@ static void c2c_he__set_node(struct c2c_hist_entry *c2c_he,
        }
 }
 
+static void c2c_he__set_mem_region(struct c2c_hist_entry *c2c_he,
+                                  unsigned int mem_region)
+{
+       if (WARN_ONCE(mem_region > PERF_MEM_REGION_MEM7,
+                     "WARNING: invalid memory region ID\n"))
+               return;
+
+       /* Update mem_region only if it really accesses memory */
+       if (mem_region >= PERF_MEM_REGION_MMIO)
+               c2c_he->mem_region = mem_region;
+}
+
 static void compute_stats(struct c2c_hist_entry *c2c_he,
                          struct c2c_stats *stats,
                          u64 weight)
@@ -339,6 +352,7 @@ static int process_sample_event(const struct perf_tool *tool __maybe_unused,
        struct addr_location al;
        struct mem_info *mi = NULL;
        struct callchain_cursor *cursor;
+       unsigned int mem_region;
        int ret;
 
        addr_location__init(&al);
@@ -366,6 +380,7 @@ static int process_sample_event(const struct perf_tool *tool __maybe_unused,
        }
 
        c2c_decode_stats(&stats, mi);
+       mem_region = mem_info__data_src(mi)->mem_region;
 
        he = hists__add_entry_ops(&c2c_hists->hists, &c2c_entry_ops,
                                  &al, NULL, NULL, mi, NULL,
@@ -382,6 +397,7 @@ static int process_sample_event(const struct perf_tool *tool __maybe_unused,
        c2c_he__set_cpu(c2c_he, sample);
        c2c_he__set_node(c2c_he, sample);
        c2c_he__set_evsel(c2c_he, evsel);
+       c2c_he__set_mem_region(c2c_he, mem_region);
 
        hists__inc_nr_samples(&c2c_hists->hists, he->filtered);
 
@@ -435,6 +451,7 @@ static int process_sample_event(const struct perf_tool *tool __maybe_unused,
                c2c_he__set_cpu(c2c_he, sample);
                c2c_he__set_node(c2c_he, sample);
                c2c_he__set_evsel(c2c_he, evsel);
+               c2c_he__set_mem_region(c2c_he, mem_region);
 
                hists__inc_nr_samples(&c2c_hists->hists, he->filtered);
                ret = hist_entry__append_callchain(he, sample);
@@ -603,6 +620,30 @@ dcacheline_node_count(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
        return scnprintf(hpp->buf, hpp->size, "%*lu", width, c2c_he->paddr_cnt);
 }
 
+static int
+dcacheline_node_mem_region(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
+                          struct hist_entry *he)
+{
+       int width = c2c_width(fmt, hpp, he->hists);
+       struct c2c_hist_entry *c2c_he;
+       unsigned int mem_region;
+       char buf[20];
+
+       c2c_he = container_of(he, struct c2c_hist_entry, he);
+       mem_region = c2c_he->mem_region;
+
+       if (mem_region == PERF_MEM_REGION_NA)
+               scnprintf(buf, sizeof(buf),  "N/A");
+       /* mem_region could only be >= PERF_MEM_REGION_MMIO */
+       else if (mem_region == PERF_MEM_REGION_MMIO)
+               scnprintf(buf, sizeof(buf), "MMIO");
+       else
+               scnprintf(buf, sizeof(buf), "0x%x",
+                         mem_region - PERF_MEM_REGION_MEM0);
+
+       return scnprintf(hpp->buf, hpp->size, "%*s", width, buf);
+}
+
 static int offset_entry(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
                        struct hist_entry *he)
 {
@@ -1425,7 +1466,7 @@ cl_idx_empty_entry(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
        }
 
Quoted 3 lines

tools/perf/builtin-c2c.c cl_idx_empty_entry(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, · line 1469

Quoted 3 lines
tools/perf/builtin-c2c.c +1 -1
 static struct c2c_dimension dim_dcacheline = {
-       .header         = HEADER_SPAN("--- Cacheline ----", "Address", 2),
+       .header         = HEADER_SPAN("--- Cacheline ----", "Address", 3),

Doesn't this also need to be handled dynamically depending on the availability of memory regions?

Thanks,
Namhyung

Quoted 120 lines
Quoted 120 lines
tools/perf/builtin-c2c.c +41 -10
        .name           = "dcacheline",
        .cmp            = dcacheline_cmp,
        .entry          = dcacheline_entry,
@@ -1440,6 +1481,14 @@ static struct c2c_dimension dim_dcacheline_node = {
        .width          = 4,
 };
 
+static struct c2c_dimension dim_dcacheline_mem_region = {
+       .header         = HEADER_LOW("Region"),
+       .name           = "dcacheline_mem_region",
+       .cmp            = empty_cmp,
+       .entry          = dcacheline_node_mem_region,
+       .width          = 6,
+};
+
 static struct c2c_dimension dim_dcacheline_count = {
        .header         = HEADER_LOW("PA cnt"),
        .name           = "dcacheline_count",
@@ -1871,6 +1920,7 @@ static struct c2c_dimension dim_dcacheline_num_empty = {
 
 static struct c2c_dimension *dimensions[] = {
        &dim_dcacheline,
+       &dim_dcacheline_mem_region,
        &dim_dcacheline_node,
        &dim_dcacheline_count,
        &dim_offset,
@@ -2898,8 +2948,9 @@ static int ui_quirks(void)
        /* Fix the zero line for dcacheline column. */
        buf = fill_line(chk_double_cl ? "Double-Cacheline" : "Cacheline",
                                dim_dcacheline.width +
+                               dim_dcacheline_mem_region.width +
                                dim_dcacheline_node.width +
-                               dim_dcacheline_count.width + 4);
+                               dim_dcacheline_count.width + 6);
        if (!buf)
                return -ENOMEM;
 
@@ -3151,8 +3202,10 @@ static int perf_c2c__report(int argc, const char **argv)
        OPT_END()
        };
        int err = 0;
-       const char *output_str, *sort_str = NULL;
+       const char *sort_str = NULL;
+       char *output_str = NULL;
        struct perf_env *env;
+       bool has_mem_regions;
 
        annotation_options__init();
 
@@ -3316,9 +3369,14 @@ static int perf_c2c__report(int argc, const char **argv)
                goto out_mem2node;
        }
 
-       if (c2c.display != DISPLAY_SNP_PEER)
-               output_str = "cl_idx,"
+       has_mem_regions = perf_header__has_feat(&session->header,
+                                               HEADER_MEMORY_RANGES);
+
+       if (c2c.display != DISPLAY_SNP_PEER) {
+               if (asprintf(&output_str,
+                            "cl_idx,"
                             "dcacheline,"
+                            "%s"
                             "dcacheline_node,"
                             "dcacheline_count,"
                             "percent_costly_snoop,"
@@ -3330,10 +3388,17 @@ static int perf_c2c__report(int argc, const char **argv)
                             "ld_fbhit,ld_l1hit,ld_l2hit,"
                             "ld_lclhit,lcl_hitm,"
                             "ld_rmthit,rmt_hitm,"
-                            "dram_lcl,dram_rmt";
-       else
-               output_str = "cl_idx,"
+                            "dram_lcl,dram_rmt",
+                            has_mem_regions ?
+                            "dcacheline_mem_region," : "") < 0) {
+                       err = -ENOMEM;
+                       goto out_mem2node;
+               }
+       } else {
+               if (asprintf(&output_str,
+                            "cl_idx,"
                             "dcacheline,"
+                            "%s"
                             "dcacheline_node,"
                             "dcacheline_count,"
                             "percent_costly_snoop,"
@@ -3345,7 +3410,13 @@ static int perf_c2c__report(int argc, const char **argv)
                             "ld_fbhit,ld_l1hit,ld_l2hit,"
                             "ld_lclhit,lcl_hitm,"
                             "ld_rmthit,rmt_hitm,"
-                            "dram_lcl,dram_rmt";
+                            "dram_lcl,dram_rmt",
+                            has_mem_regions ?
+                            "dcacheline_mem_region," : "") < 0) {
+                       err = -ENOMEM;
+                       goto out_mem2node;
+               }
+       }
 
        if (c2c.display == DISPLAY_TOT_HITM)
                sort_str = "tot_hitm";
@@ -3368,11 +3439,13 @@ static int perf_c2c__report(int argc, const char **argv)
 
        if (ui_quirks()) {
                pr_err("failed to setup UI\n");
-               goto out_mem2node;
+               goto out_str;
        }
 
        perf_c2c_display(session);
 
+out_str:
+       free(output_str);
 out_mem2node:
        mem2node__exit(&c2c.mem2node);
 out_session:
--

2.55.0

#12 lore raw

Re: [PATCH v4 4/6] perf tools: Show memory region in perf-c2c subcommand

Quoted 198 lines · Namhyung Kim

Hello,

Quoted 195 lines · Ian Rogers
Quoted 62 lines · Thomas Falcon

From: Dapeng Mi <[email protected]>

Add memory region field to the cacheline list view to help users identify the memory region to which the cacheline belongs. The memory
region field was included with the introduction of support for the
Off-module Response facility (OMR) [1] in Intel's Diamond Rapids and
Nova Lake architectures.

An example of the new perf c2c output including the memory region field is shown below:

Shared Data Cache Line Table     (112 entries, sorted on Total HITMs)
       --------------- Cacheline --------------      Tot  ------- Load Hitm -------    Total    Total    Total
Index             Address  Region  Node  PA cnt     Hitm Total  LclHitm  RmtHitm  records    Loads   Stores
    0  0xff1b7032a1e8c5c0  N/A        1      47    2.25% 50       44        6     1024     1023        1
    1  0xff1b6ff3255bb880  N/A        0      11    1.62% 36       34        2       96       93        4
    2  0xff1b70328e3b9880  N/A        1      13    1.49% 33       33        0      100       91        9
    3  0xff1b70328b023800  N/A        1       1    1.40% 31       14       17       52       48        4
    4  0xff1b70328e3b9c00  N/A        1       1    1.31% 29       27        2       67       33       34
    5  0xff1b70328b0237c0  N/A        1       1    1.26% 28       10       18      154      150        4
    6  0xff1b6ff3255bbc00  N/A        0       1    1.13% 25       25        0       48       25       23
    7  0xff1b6ff3255bba40  N/A        0       1    0.99% 22       22        0       46       23       23
    8  0xff3ab9ba50255c80  N/A      N/A       0    0.77% 17       15        2       35       35        0
    9  0xff3ab9ba503e3040  N/A      N/A       0    0.77% 17       11        6       37       37        1
   10  0xff1b703289e88f40  N/A        1      33    0.72% 16        9        7       69       63        6
   11  0xff1b70328e3b9a40  N/A        1       1    0.68% 15       15        0       43       17       26
   12  0xff1b7032c9fd6a40  N/A        1      15    0.68% 15       15        0       57       54        4
   13  0xff1b7032a1e8c980  N/A        1      27    0.54% 12       11        1      761      761        0
   14  0xff1b70727ffd57c0  N/A        1       7    0.54% 12       12        0      219      218        1
   15  0xffffffffaefe2380  N/A        1       1    0.50% 11        8        3       14       14        0

Note: DMR simics does not support memory regions. Since the output is
captured on SPR, the memory region field shows "N/A" for all cachelines.

[1]:
https://lore.kernel.org/all/[email protected]/

Quoted 6 lines
Assisted-by Sashiko:gemini-3.1-pro-preview
Assisted-by GitHub-Copilot:claude-opus-4-8

Link:
https://lore.kernel.org/all/[email protected]/

Thanks,
Ian

Quoted 119 lines

---
v4: correctly handle output_str memory allocation failure

v3: make memory region reporting conditional on feature bit
---
 tools/perf/builtin-c2c.c | 93
+++++++++++++++++++++++++++++++++++-----

 1 file changed, 83 insertions(+), 10 deletions(-)
tools/perf/builtin-c2c.c +13 -0
index c9584dbedf77..8cd90524fe49 100644
@@ -74,6 +74,7 @@ struct c2c_hist_entry {
         unsigned long           *nodeset;
         struct c2c_stats        *node_stats;
         unsigned int             cacheline_idx;
+       unsigned int             mem_region;
 
         struct compute_stats     cstats;
 
@@ -281,6 +282,18 @@ static void c2c_he__set_node(struct
 c2c_hist_entry *c2c_he,
         }
  }
 
+static void c2c_he__set_mem_region(struct c2c_hist_entry
 *c2c_he,
+                                  unsigned int mem_region)
+{
+       if (WARN_ONCE(mem_region > PERF_MEM_REGION_MEM7,
+                     "WARNING: invalid memory region ID\n"))
+               return;
+
+       /* Update mem_region only if it really accesses memory */
+       if (mem_region >= PERF_MEM_REGION_MMIO)
+               c2c_he->mem_region = mem_region;
+}
+
  static void compute_stats(struct c2c_hist_entry *c2c_he,

                          struct c2c_stats *stats, u64 weight)

tools/perf/builtin-c2c.c +1 -0
@@ -339,6 +352,7 @@ static int process_sample_event(const struct
 perf_tool *tool __maybe_unused,
         struct addr_location al;
         struct mem_info *mi = NULL;
         struct callchain_cursor *cursor;
+       unsigned int mem_region;
         int ret;
 

        addr_location__init(&al);

tools/perf/builtin-c2c.c +1 -0
@@ -366,6 +380,7 @@ static int process_sample_event(const struct
 perf_tool *tool __maybe_unused,
         }
 
         c2c_decode_stats(&stats, mi);
+       mem_region = mem_info__data_src(mi)->mem_region;
 
         he = hists__add_entry_ops(&c2c_hists->hists,

&c2c_entry_ops,
                                  &al, NULL, NULL, mi, NULL,

tools/perf/builtin-c2c.c +2 -0
@@ -382,6 +397,7 @@ static int process_sample_event(const struct
 perf_tool *tool __maybe_unused,
         c2c_he__set_cpu(c2c_he, sample);
         c2c_he__set_node(c2c_he, sample);
         c2c_he__set_evsel(c2c_he, evsel);
+       c2c_he__set_mem_region(c2c_he, mem_region);
 
         hists__inc_nr_samples(&c2c_hists->hists, he->filtered);
@@ -435,6 +451,7 @@ static int process_sample_event(const struct
 perf_tool *tool __maybe_unused,
                 c2c_he__set_cpu(c2c_he, sample);
                 c2c_he__set_node(c2c_he, sample);
                 c2c_he__set_evsel(c2c_he, evsel);
+               c2c_he__set_mem_region(c2c_he, mem_region);
 

                hists__inc_nr_samples(&c2c_hists->hists, he-

Quoted 1 line

filtered);

                ret = hist_entry__append_callchain(he, sample);

tools/perf/builtin-c2c.c +24 -0
@@ -603,6 +620,30 @@ dcacheline_node_count(struct perf_hpp_fmt
 *fmt, struct perf_hpp *hpp,
         return scnprintf(hpp->buf, hpp->size, "%*lu", width,
 c2c_he->paddr_cnt);
  }
 
+static int
+dcacheline_node_mem_region(struct perf_hpp_fmt *fmt, struct
 perf_hpp *hpp,
+                          struct hist_entry *he)
+{
+       int width = c2c_width(fmt, hpp, he->hists);
+       struct c2c_hist_entry *c2c_he;
+       unsigned int mem_region;
+       char buf[20];
+
+       c2c_he = container_of(he, struct c2c_hist_entry, he);
+       mem_region = c2c_he->mem_region;
+
+       if (mem_region == PERF_MEM_REGION_NA)
+               scnprintf(buf, sizeof(buf),  "N/A");
+       /* mem_region could only be >= PERF_MEM_REGION_MMIO */
+       else if (mem_region == PERF_MEM_REGION_MMIO)
+               scnprintf(buf, sizeof(buf), "MMIO");
+       else
+               scnprintf(buf, sizeof(buf), "0x%x",
+                         mem_region - PERF_MEM_REGION_MEM0);
+
+       return scnprintf(hpp->buf, hpp->size, "%*s", width, buf);
+}
+

 static int offset_entry(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
                        struct hist_entry *he)
 {

tools/perf/builtin-c2c.c +0 -0
@@ -1425,7 +1466,7 @@ cl_idx_empty_entry(struct perf_hpp_fmt
 *fmt, struct perf_hpp *hpp,
         }
 
Quoted 8 lines
Quoted 5 lines

tools/perf/builtin-c2c.c cl_idx_empty_entry(struct perf_hpp_fmt · line 1472

Quoted 5 lines
tools/perf/builtin-c2c.c +1 -1
  static struct c2c_dimension dim_dcacheline = {
-       .header         = HEADER_SPAN("--- Cacheline ----",
 "Address", 2),
+       .header         = HEADER_SPAN("--- Cacheline ----",
 "Address", 3),

Doesn't this also need to be handled dynamically depending on the availability of memory regions?

Thanks, I will update it to handle memory regions dynamically.

Quoted 28 lines

Thanks,
Namhyung

Quoted 23 lines
Quoted 23 lines

        .name           = "dcacheline",
        .cmp            = dcacheline_cmp,
        .entry          = dcacheline_entry,

tools/perf/builtin-c2c.c +8 -0
@@ -1440,6 +1481,14 @@ static struct c2c_dimension
 dim_dcacheline_node = {
         .width          = 4,
  };
 
+static struct c2c_dimension dim_dcacheline_mem_region = {
+       .header         = HEADER_LOW("Region"),
+       .name           = "dcacheline_mem_region",
+
        .cmp            = empty_cmp,
+       .entry          = dcacheline_node_mem_region,
+       .width          = 6,
+};
+
  static struct c2c_dimension dim_dcacheline_count = {

        .header         = HEADER_LOW("PA cnt"),
        .name           = "dcacheline_count",

tools/perf/builtin-c2c.c +0 -0
@@ -1871,6 +1920,7 @@ static struct c2c_dimension
 dim_dcacheline_num_empty = {
 
Quoted 15 lines
Quoted 7 lines
Quoted 7 lines
tools/perf/builtin-c2c.c +1 -0
  static struct c2c_dimension *dimensions[] = {
         &dim_dcacheline,
+       &dim_dcacheline_mem_region,
         &dim_dcacheline_node,
         &dim_dcacheline_count,

        &dim_offset,

tools/perf/builtin-c2c.c +0 -0
@@ -2898,8 +2948,9 @@ static int ui_quirks(void)
Quoted 8 lines
Quoted 8 lines
tools/perf/builtin-c2c.c +0 -0
         /* Fix the zero line for dcacheline column. */

        buf = fill_line(chk_double_cl ? "Double-Cacheline" :
"Cacheline",
                                dim_dcacheline.width + +                               dim_dcacheline_mem_region.width + dim_dcacheline_node.width + -                               dim_dcacheline_count.width + 4); +                               dim_dcacheline_count.width + 6);

Do you think this needs to be handled dynamically too? Thanks,
Tom

Quoted 94 lines
Quoted 94 lines
Quoted 94 lines

        if (!buf)
                return -ENOMEM;

tools/perf/builtin-c2c.c +9 -3
@@ -3151,8 +3202,10 @@ static int perf_c2c__report(int argc,
 const char **argv)
         OPT_END()
         };
         int err = 0;
-       const char *output_str, *sort_str = NULL;
+       const char *sort_str = NULL;
+       char *output_str = NULL;
         struct perf_env *env;
+       bool has_mem_regions;
 
         annotation_options__init();
@@ -3316,9 +3369,14 @@ static int perf_c2c__report(int argc,
 const char **argv)
                 goto out_mem2node;
         }
 
-       if (c2c.display != DISPLAY_SNP_PEER)
-               output_str = "cl_idx,"
+       has_mem_regions = perf_header__has_feat(&session->header,
+                                              
 HEADER_MEMORY_RANGES);
+
+       if (c2c.display != DISPLAY_SNP_PEER) {
+               if (asprintf(&output_str,
+                            "cl_idx,"

                             "dcacheline,"
+                            "%s" "dcacheline_node," "dcacheline_count," "percent_costly_snoop,"

tools/perf/builtin-c2c.c +0 -0
@@ -3330,10 +3388,17 @@ static int perf_c2c__report(int argc,
 const char **argv)

                             "ld_fbhit,ld_l1hit,ld_l2hit," "ld_lclhit,lcl_hitm," "ld_rmthit,rmt_hitm,"
-                            "dram_lcl,dram_rmt";
-       else
-               output_str = "cl_idx,"
+                            "dram_lcl,dram_rmt",
+                            has_mem_regions ?
+                            "dcacheline_mem_region," : "") < 0) {
+                       err = -ENOMEM;
+                       goto out_mem2node;
+               }
+       } else {
+               if (asprintf(&output_str,
+                            "cl_idx,"
                             "dcacheline,"
+                            "%s"
                             "dcacheline_node," "dcacheline_count," "percent_costly_snoop,"

tools/perf/builtin-c2c.c +0 -0
@@ -3345,7 +3410,13 @@ static int perf_c2c__report(int argc,
 const char **argv)

                             "ld_fbhit,ld_l1hit,ld_l2hit," "ld_lclhit,lcl_hitm," "ld_rmthit,rmt_hitm,"
-                            "dram_lcl,dram_rmt";
+                            "dram_lcl,dram_rmt",
+                            has_mem_regions ?
+                            "dcacheline_mem_region," : "") < 0) {
+                       err = -ENOMEM;
+                       goto out_mem2node;
+               }
+       }

        if (c2c.display == DISPLAY_TOT_HITM)
                sort_str = "tot_hitm";

tools/perf/builtin-c2c.c +3 -1
@@ -3368,11 +3439,13 @@ static int perf_c2c__report(int argc,
 const char **argv)
 
         if (ui_quirks()) {
                 pr_err("failed to setup UI\n");
-               goto out_mem2node;
+               goto out_str;
         }
 
         perf_c2c_display(session);
 
+out_str:
+       free(output_str);

 out_mem2node:
        mem2node__exit(&c2c.mem2node);
 out_session:
--
2.55.0

maintainer
#13 lore raw

Re: [PATCH v4 4/6] perf tools: Show memory region in perf-c2c subcommand

Quoted 5 lines · Thomas
Quoted 4 lines · Namhyung Kim

Hello,

Quoted 1 line · Ian Rogers

On Tue, Aug 11, 2026 at 10:34 AM Thomas Falcon

Quoted 10 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
tools/perf/builtin-c2c.c +0 -0
@@ -2898,8 +2948,9 @@ static int ui_quirks(void)
         /* Fix the zero line for dcacheline column. */
         buf = fill_line(chk_double_cl ? "Double-Cacheline" : "Cacheline",

                                dim_dcacheline.width + +                               dim_dcacheline_mem_region.width + dim_dcacheline_node.width + -                               dim_dcacheline_count.width + 4); +                               dim_dcacheline_count.width + 6);

Do you think this needs to be handled dynamically too?

Ideally it should not change anything when memory region is not available.

Thanks,
Namhyung

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

        if (!buf)
                return -ENOMEM;

#14 lore raw

PATCH v4 5/6 perf tools: Show memory region in perf-script subcommand

In reply to Thomas Falcon

From: Dapeng Mi <[email protected]>

Show the memory region in perf-script subcommand. Memory region is found in the mem_region field of the memory information data source. This field was included with the introduction of support for the Off-module Response facility (OMR) [1] in Intel's Diamond Rapids and Nova Lake Architectures.

An example of perf-script output with the new memory region field is shown below:

random      10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region  N/A                           7f32ae64a2ec
random    2411a68201042 |OP LOAD|LVL RAM hit|SNP Hit|TLB L1 or L2 hit|LCK No|BLK  N/A|Region  Mem-1                          561de7f8910a
random      10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region  N/A                           7f32ae64a2d0
random      20e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  Data|Region  N/A                          7f32ae64a2d0
random      20e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  Data|Region  N/A                          7f32ae64a2d0
random      10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region  N/A                           7f32ae64a2ec
random      10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region  N/A                           7f32ae64a2d0
random      10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region  N/A                           7f32ae64a2d0
random    2411a68201042 |OP LOAD|LVL RAM hit|SNP Hit|TLB L1 or L2 hit|LCK No|BLK  N/A|Region  Mem-1                          561de7f8910a
random      10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region  N/A                           7f32ae64a2ec
random      10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region  N/A                           7f32ae64a2d0

[1]: https://lore.kernel.org/all/[email protected]/

Assisted-by Sashiko:gemini-3.1-pro-preview
Assisted-by GitHub-Copilot:claude-opus-4-8

---
v4: Drop the global show-region flag; pass perf_session to
perf_script__meminfo_scnprintf() and gate the Region field on the feature bit, with a pipe-mode fallback

v3: make memory region reporting conditional on feature bit

---
 tools/perf/builtin-script.c                   | 13 +--
 tools/perf/util/mem-events.c                  | 86 ++++++++++++++++++-
 tools/perf/util/mem-events.h                  |  4 +-
 .../scripting-engines/trace-event-python.c    |  5 +-
 4 files changed, 98 insertions(+), 10 deletions(-)
tools/perf/builtin-script.c +7 -6
index f91d8b1fbd01..410ccad56f99 100644
@@ -2078,11 +2078,12 @@ static int evlist__max_name_len(struct evlist *evlist)
 	return max;
 }
 
-static int data_src__fprintf(u64 data_src, FILE *fp)
+static int data_src__fprintf(struct perf_session *session,
+			     u64 data_src, FILE *fp)
 {
 	struct mem_info *mi = mem_info__new();
-	char decode[100];
-	char out[100];
+	char decode[200];
+	char out[200];
 	static int maxlen;
 	int len;
 
@@ -2090,10 +2091,10 @@ static int data_src__fprintf(u64 data_src, FILE *fp)
 		return -ENOMEM;
 
 	mem_info__data_src(mi)->val = data_src;
-	perf_script__meminfo_scnprintf(decode, 100, mi);
+	perf_script__meminfo_scnprintf(decode, 200, mi, session);
 	mem_info__put(mi);
 
-	len = scnprintf(out, 100, "%16" PRIx64 " %s", data_src, decode);
+	len = scnprintf(out, 200, "%16" PRIx64 " %s", data_src, decode);
 	if (maxlen < len)
 		maxlen = len;
 
@@ -2487,7 +2488,7 @@ static void process_event(struct perf_script *script,
 		perf_sample__fprintf_addr(sample, thread, evsel, fp);
 
 	if (PRINT_FIELD(DATA_SRC))
-		data_src__fprintf(sample->data_src, fp);
+		data_src__fprintf(evsel__session(evsel), sample->data_src, fp);
 
 	if (PRINT_FIELD(WEIGHT))
 		fprintf(fp, "%16" PRIu64, sample->weight);
tools/perf/util/mem-events.c +85 -1

99 lines not on this page · raw message

tools/perf/util/mem-events.h +3 -1
index daa22748f9fe..4ebb8109fc3c 100644
@@ -4,6 +4,7 @@
 
 #include <stdbool.h>
 #include <linux/types.h>
+#include "session.h"
 
 struct perf_mem_event {
 	bool		supported;
@@ -47,7 +48,8 @@ int perf_mem__snp_scnprintf(char *out, size_t sz, const struct mem_info *mem_inf
 int perf_mem__lck_scnprintf(char *out, size_t sz, const struct mem_info *mem_info);
 int perf_mem__blk_scnprintf(char *out, size_t sz, const struct mem_info *mem_info);
 
-int perf_script__meminfo_scnprintf(char *bf, size_t size, const struct mem_info *mem_info);
+int perf_script__meminfo_scnprintf(char *bf, size_t size, const struct mem_info *mem_info,
+				   struct perf_session *session);
 
 struct c2c_stats {
 	u32	nr_entries;
tools/perf/util/scripting-engines/trace-event-python.c +3 -2
index 8f832ae316ca..e359fe529586 100644
@@ -694,8 +694,9 @@ static void set_sample_read_in_dict(PyObject *dict_sample, struct perf_sample *s
 static void set_sample_datasrc_in_dict(PyObject *dict,
 				      struct perf_sample *sample)
 {
+	struct perf_session *session = evsel__session(sample->evsel);
 	struct mem_info *mi = mem_info__new();
-	char decode[100];
+	char decode[200];
 
 	if (!mi)
 		Py_FatalError("couldn't create mem-info");
@@ -704,7 +705,7 @@ static void set_sample_datasrc_in_dict(PyObject *dict,
 			PyLong_FromUnsignedLongLong(sample->data_src));
 
 	mem_info__data_src(mi)->val = sample->data_src;
-	perf_script__meminfo_scnprintf(decode, 100, mi);
+	perf_script__meminfo_scnprintf(decode, 200, mi, session);
 	mem_info__put(mi);
 
 	pydict_set_item_string_decref(dict, "datasrc_decode",
Signature · 1 line
2.55.0
reviewer Reviewed-by on v4 5/6
#15 lore raw

Re: [PATCH v4 5/6] perf tools: Show memory region in perf-script subcommand

Quoted 24 lines · Thomas Falcon

From: Dapeng Mi <[email protected]>

Show the memory region in perf-script subcommand. Memory region is found in the mem_region field of the memory information data source. This field was included with the introduction of support for the Off-module Response facility (OMR) [1] in Intel's Diamond Rapids and Nova Lake Architectures.

An example of perf-script output with the new memory region field is shown below:

random      10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region  N/A                           7f32ae64a2ec
random    2411a68201042 |OP LOAD|LVL RAM hit|SNP Hit|TLB L1 or L2 hit|LCK No|BLK  N/A|Region  Mem-1                          561de7f8910a
random      10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region  N/A                           7f32ae64a2d0
random      20e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  Data|Region  N/A                          7f32ae64a2d0
random      20e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  Data|Region  N/A                          7f32ae64a2d0
random      10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region  N/A                           7f32ae64a2ec
random      10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region  N/A                           7f32ae64a2d0
random      10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region  N/A                           7f32ae64a2d0
random    2411a68201042 |OP LOAD|LVL RAM hit|SNP Hit|TLB L1 or L2 hit|LCK No|BLK  N/A|Region  Mem-1                          561de7f8910a
random      10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region  N/A                           7f32ae64a2ec
random      10e6a100042 |OP LOAD|LVL L0 hit|SNP None|TLB L1 or L2 hit|LCK Yes|BLK  N/A|Region  N/A                           7f32ae64a2d0
Quoted 7 lines

[1]: https://lore.kernel.org/all/[email protected]/

Assisted-by Sashiko:gemini-3.1-pro-preview
Assisted-by GitHub-Copilot:claude-opus-4-8

Thanks,
Ian

Quoted 209 lines

---
v4: Drop the global show-region flag; pass perf_session to
perf_script__meminfo_scnprintf() and gate the Region field on the feature bit, with a pipe-mode fallback

v3: make memory region reporting conditional on feature bit

---
 tools/perf/builtin-script.c                   | 13 +--
 tools/perf/util/mem-events.c                  | 86 ++++++++++++++++++-
 tools/perf/util/mem-events.h                  |  4 +-
 .../scripting-engines/trace-event-python.c    |  5 +-
 4 files changed, 98 insertions(+), 10 deletions(-)
tools/perf/builtin-script.c +7 -6
index f91d8b1fbd01..410ccad56f99 100644
@@ -2078,11 +2078,12 @@ static int evlist__max_name_len(struct evlist *evlist)
        return max;
 }
 
-static int data_src__fprintf(u64 data_src, FILE *fp)
+static int data_src__fprintf(struct perf_session *session,
+                            u64 data_src, FILE *fp)
 {
        struct mem_info *mi = mem_info__new();
-       char decode[100];
-       char out[100];
+       char decode[200];
+       char out[200];
        static int maxlen;
        int len;
 
@@ -2090,10 +2091,10 @@ static int data_src__fprintf(u64 data_src, FILE *fp)
                return -ENOMEM;
 
        mem_info__data_src(mi)->val = data_src;
-       perf_script__meminfo_scnprintf(decode, 100, mi);
+       perf_script__meminfo_scnprintf(decode, 200, mi, session);
        mem_info__put(mi);
 
-       len = scnprintf(out, 100, "%16" PRIx64 " %s", data_src, decode);
+       len = scnprintf(out, 200, "%16" PRIx64 " %s", data_src, decode);
        if (maxlen < len)
                maxlen = len;
 
@@ -2487,7 +2488,7 @@ static void process_event(struct perf_script *script,
                perf_sample__fprintf_addr(sample, thread, evsel, fp);
 
        if (PRINT_FIELD(DATA_SRC))
-               data_src__fprintf(sample->data_src, fp);
+               data_src__fprintf(evsel__session(evsel), sample->data_src, fp);
 
        if (PRINT_FIELD(WEIGHT))
                fprintf(fp, "%16" PRIu64, sample->weight);
tools/perf/util/mem-events.c +85 -1
index 4fd48fd20055..8ce4996cad8d 100644
@@ -604,8 +604,77 @@ int perf_mem__blk_scnprintf(char *out, size_t sz, const struct mem_info *mem_inf
        return l;
 }
 
-int perf_script__meminfo_scnprintf(char *out, size_t sz, const struct mem_info *mem_info)
+static int perf_mem__region_scnprintf(char *out, size_t sz, const struct mem_info *mem_info)
 {
+       size_t l = 0;
+       u64 mem = PERF_MEM_REGION_NA;
+
+       sz -= 1; /* -1 for null termination */
+       out[0] = '\0';
+
+       if (mem_info)
+               mem = mem_info__const_data_src(mem_info)->mem_region;
+
+       switch (mem) {
+       case PERF_MEM_REGION_NA:
+       case PERF_MEM_REGION_RSVD:
+               l += scnprintf(out + l, sz - l, "N/A");
+               break;
+       case PERF_MEM_REGION_L_SHARE:
+               l += scnprintf(out + l, sz - l, "Local-shared-cache");
+               break;
+       case PERF_MEM_REGION_L_NON_SHARE:
+               l += scnprintf(out + l, sz - l, "Local-non-shared-cache");
+               break;
+       case PERF_MEM_REGION_O_IO:
+               l += scnprintf(out + l, sz - l, "Other-IO");
+               break;
+       case PERF_MEM_REGION_O_SHARE:
+               l += scnprintf(out + l, sz - l, "Other-shared-cache");
+               break;
+       case PERF_MEM_REGION_O_NON_SHARE:
+               l += scnprintf(out + l, sz - l, "Other-non-shared-cache");
+               break;
+       case PERF_MEM_REGION_MMIO:
+               l += scnprintf(out + l, sz - l, "MMIO");
+               break;
+       case PERF_MEM_REGION_MEM0:
+               l += scnprintf(out + l, sz - l, "Mem-0");
+               break;
+       case PERF_MEM_REGION_MEM1:
+               l += scnprintf(out + l, sz - l, "Mem-1");
+               break;
+       case PERF_MEM_REGION_MEM2:
+               l += scnprintf(out + l, sz - l, "Mem-2");
+               break;
+       case PERF_MEM_REGION_MEM3:
+               l += scnprintf(out + l, sz - l, "Mem-3");
+               break;
+       case PERF_MEM_REGION_MEM4:
+               l += scnprintf(out + l, sz - l, "Mem-4");
+               break;
+       case PERF_MEM_REGION_MEM5:
+               l += scnprintf(out + l, sz - l, "Mem-5");
+               break;
+       case PERF_MEM_REGION_MEM6:
+               l += scnprintf(out + l, sz - l, "Mem-6");
+               break;
+       case PERF_MEM_REGION_MEM7:
+               l += scnprintf(out + l, sz - l, "Mem-7");
+               break;
+       default:
+               l += scnprintf(out + l, sz - l, "N/A");
+               break;
+       }
+
+       return l;
+}
+
+int perf_script__meminfo_scnprintf(char *out, size_t sz,
+                                  const struct mem_info *mem_info,
+                                  struct perf_session *session)
+{
+       struct perf_env *env;
        int i = 0;
 
        i += scnprintf(out, sz, "|OP ");
@@ -620,6 +689,21 @@ int perf_script__meminfo_scnprintf(char *out, size_t sz, const struct mem_info *
        i += perf_mem__lck_scnprintf(out + i, sz - i, mem_info);
        i += scnprintf(out + i, sz - i, "|BLK ");
        i += perf_mem__blk_scnprintf(out + i, sz - i, mem_info);
+       if (session) {
+               /*
+                * In case the feature bits are not available, as in
+                * pipe mode, fallback to checking for the existence of
+                * memory ranges
+                */
+               env = perf_session__env(session);
+               if ((env && session->data->is_pipe && env->nr_memory_ranges) ||
+                   perf_header__has_feat(&session->header,
+                                         HEADER_MEMORY_RANGES)) {
+                       i += scnprintf(out + i, sz - i, "|Region ");
+                       i += perf_mem__region_scnprintf(out + i, sz - i,
+                                                       mem_info);
+               }
+       }
 
        return i;
 }
tools/perf/util/mem-events.h +3 -1
index daa22748f9fe..4ebb8109fc3c 100644
@@ -4,6 +4,7 @@
 
 #include <stdbool.h>
 #include <linux/types.h>
+#include "session.h"
 
 struct perf_mem_event {
        bool            supported;
@@ -47,7 +48,8 @@ int perf_mem__snp_scnprintf(char *out, size_t sz, const struct mem_info *mem_inf
 int perf_mem__lck_scnprintf(char *out, size_t sz, const struct mem_info *mem_info);
 int perf_mem__blk_scnprintf(char *out, size_t sz, const struct mem_info *mem_info);
 
-int perf_script__meminfo_scnprintf(char *bf, size_t size, const struct mem_info *mem_info);
+int perf_script__meminfo_scnprintf(char *bf, size_t size, const struct mem_info *mem_info,
+                                  struct perf_session *session);
 
 struct c2c_stats {
        u32     nr_entries;
tools/perf/util/scripting-engines/trace-event-python.c +3 -3
index 8f832ae316ca..e359fe529586 100644
@@ -694,8 +694,9 @@ static void set_sample_read_in_dict(PyObject *dict_sample, struct perf_sample *s
 static void set_sample_datasrc_in_dict(PyObject *dict,
                                      struct perf_sample *sample)
 {
+       struct perf_session *session = evsel__session(sample->evsel);
        struct mem_info *mi = mem_info__new();
-       char decode[100];
+       char decode[200];
 
        if (!mi)
                Py_FatalError("couldn't create mem-info");
@@ -704,7 +705,7 @@ static void set_sample_datasrc_in_dict(PyObject *dict,
                        PyLong_FromUnsignedLongLong(sample->data_src));
 
        mem_info__data_src(mi)->val = sample->data_src;
-       perf_script__meminfo_scnprintf(decode, 100, mi);
+       perf_script__meminfo_scnprintf(decode, 200, mi, session);
        mem_info__put(mi);
 
        pydict_set_item_string_decref(dict, "datasrc_decode",
--

2.55.0

#16 lore raw

PATCH v4 6/6 perf c2c: print memory region data with stdio output

In reply to Thomas Falcon

Print memory range data in perf-c2c subcommand. Only available in stdio mode currently. If memory ranges are not supported or present, print nothing.

Example output of perf c2c report --stdio:

=================================================
Memory Ranges
=================================================
Range 0: [0x0000000000000000-0x00000000bfffffff] Node 0, local region id 0, remote region id 255 Range 1: [0x0000000100000000-0x000000203fffffff] Node 0, local region id 0, remote region id 255 Range 2: [0x0000008000000000-0x0000027fffffffff] Node -2, local region id 1, remote region id 255 Range 3: [0x0000028000000000-0x0000047fffffffff] Node -2, local region id 1, remote region id 255 Range 4: [0x0000048000000000-0x000004ffffffffff] Node -2, local region id 1, remote region id 255

Assisted-by Sashiko:gemini-3.1-pro-preview

---
v3: Included examples of memory range output in commit message

---
 tools/perf/builtin-c2c.c | 30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)
tools/perf/builtin-c2c.c +30 -0
index 8cd90524fe49..7739023ced41 100644
@@ -2635,6 +2635,34 @@ static void print_c2c_info(FILE *out, struct perf_session *session)
 	fprintf(out, "  Cacheline data grouping           : %s\n", c2c.cl_sort);
 }
 
+
+static void print_memory_ranges_info(FILE *out, struct perf_session *session)
+{
+	struct perf_env *env = perf_session__env(session);
+	int nr_ranges = 0;
+
+	if (!perf_header__has_feat(&session->header, HEADER_MEMORY_RANGES))
+		return;
+	nr_ranges = env->nr_memory_ranges;
+	if (nr_ranges == 0) {
+		pr_debug("No memory ranges found, skipping\n");
+		return;
+	}
+
+	fprintf(out, "\n");
+	fprintf(out, "=================================================\n");
+	fprintf(out, "                  Memory Ranges                  \n");
+	fprintf(out, "=================================================\n");
+
+	for (int i = 0; i < nr_ranges; i++) {
+		struct memory_range *r = &env->memory_ranges[i];
+
+		fprintf(out, "Range %d: [0x%016" PRIx64 "-0x%016" PRIx64 "] Node %d, local region id %u, remote region id %u\n",
+			i, r->base, r->base + r->length - 1, r->node,
+			r->local_region_id, r->remote_region_id);
+	}
+}
+
 static void perf_c2c__hists_fprintf(FILE *out, struct perf_session *session)
 {
 	setup_pager();
@@ -2648,6 +2676,8 @@ static void perf_c2c__hists_fprintf(FILE *out, struct perf_session *session)
 	if (c2c.stats_only)
 		return;
 
+	print_memory_ranges_info(out, session);
+
 	fprintf(out, "\n");
 	fprintf(out, "=================================================\n");
 	fprintf(out, "           Shared Data Cache Line Table          \n");
Signature · 1 line
2.55.0
reviewer Reviewed-by on v4 6/6
#17 lore raw

Re: [PATCH v4 6/6] perf c2c: print memory region data with stdio output

Quoted 16 lines · Thomas Falcon

Print memory range data in perf-c2c subcommand. Only available in stdio mode currently. If memory ranges are not supported or present, print nothing.

Example output of perf c2c report --stdio:

=================================================
Memory Ranges
=================================================
Range 0: [0x0000000000000000-0x00000000bfffffff] Node 0, local region id 0, remote region id 255 Range 1: [0x0000000100000000-0x000000203fffffff] Node 0, local region id 0, remote region id 255 Range 2: [0x0000008000000000-0x0000027fffffffff] Node -2, local region id 1, remote region id 255 Range 3: [0x0000028000000000-0x0000047fffffffff] Node -2, local region id 1, remote region id 255 Range 4: [0x0000048000000000-0x000004ffffffffff] Node -2, local region id 1, remote region id 255

Quoted 2 lines
Assisted-by Sashiko:gemini-3.1-pro-preview

I wonder about opportunities around this information and what is in /proc/iomem as exposed in the mem-phys-addr.py script:
https://web.git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/scripts/python/mem-phys-addr.py?h=tmp.perf-tools-next

Thanks,
Ian

Quoted 57 lines

---
v3: Included examples of memory range output in commit message

---
 tools/perf/builtin-c2c.c | 30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)
tools/perf/builtin-c2c.c +30 -1
index 8cd90524fe49..7739023ced41 100644
@@ -2635,6 +2635,34 @@ static void print_c2c_info(FILE *out, struct perf_session *session)
        fprintf(out, "  Cacheline data grouping           : %s\n", c2c.cl_sort);
 }
 
+
+static void print_memory_ranges_info(FILE *out, struct perf_session *session)
+{
+       struct perf_env *env = perf_session__env(session);
+       int nr_ranges = 0;
+
+       if (!perf_header__has_feat(&session->header, HEADER_MEMORY_RANGES))
+               return;
+       nr_ranges = env->nr_memory_ranges;
+       if (nr_ranges == 0) {
+               pr_debug("No memory ranges found, skipping\n");
+               return;
+       }
+
+       fprintf(out, "\n");
+       fprintf(out, "=================================================\n");
+       fprintf(out, "                  Memory Ranges                  \n");
+       fprintf(out, "=================================================\n");
+
+       for (int i = 0; i < nr_ranges; i++) {
+               struct memory_range *r = &env->memory_ranges[i];
+
+               fprintf(out, "Range %d: [0x%016" PRIx64 "-0x%016" PRIx64 "] Node %d, local region id %u, remote region id %u\n",
+                       i, r->base, r->base + r->length - 1, r->node,
+                       r->local_region_id, r->remote_region_id);
+       }
+}
+
 static void perf_c2c__hists_fprintf(FILE *out, struct perf_session *session)
 {
        setup_pager();
@@ -2648,6 +2676,8 @@ static void perf_c2c__hists_fprintf(FILE *out, struct perf_session *session)
        if (c2c.stats_only)
                return;
 
+       print_memory_ranges_info(out, session);
+
        fprintf(out, "\n");
        fprintf(out, "=================================================\n");
        fprintf(out, "           Shared Data Cache Line Table          \n");
--

2.55.0