Skip to contentSkip to conversation

perf ftrace: Support inlined functions and display enhancements in function graph tracer

PATCH Superseded Aaron Tomlin 2 participants 15 messages Oct 14 perf events

v4 of 13 Superseded Latest is v13 →

lore
Share & tools
lore
#0 lore raw

v4 cover perf ftrace: Support inlined functions and display enhancements in function graph tracer

The Linux kernel's function graph tracer operates at the machine instruction level via compiler instrumentation (-fpatchable-function-entry), recording entry and return events for physical function calls. Consequently, functions inlined by the compiler are invisible in the resulting call graph, as no discrete call or return instructions are emitted for them. While technically expected, this omission frequently obscures the logical execution flow when analysing kernel subsystems heavily reliant upon inlining (such as the scheduler, memory management, locking primitives, and RCU).

When the kernel is configured with CONFIG_FUNCTION_GRAPH_RETADDR=y, the funcgraph-retaddr trace option records the caller's return address on each function entry, manifested in the trace stream as a comment (e.g. /* <-wake_up_new_task+0x1d1/0x3e0 */). By interrogating DWARF debug information from the kernel image (vmlinux) utilising libdw, perf ftrace resolves this return address to its inlined callchain and synthesises the intermediate inlined frames directly into the streamed call graph.

This patch series addresses these challenges through five focused commits:

1. Stack frame optimisation in __cmd_ftrace()

Replaces the 4096-byte stack array in __cmd_ftrace() with dynamic heap allocation alongside struct strbuf streaming, reducing the stack frame footprint by over 96% (down to ~160 bytes) whilst ensuring complete line-delimited records, explicit NUL termination, and robust error unwind and EOF handling.

2. Return address comment filtering

Introduces ftrace_parse_retaddr() and ftrace_filter_retaddr() with --filter-retaddr and --graph-opts filter-retaddr to strip caller comments from trace output, accompanied by an initial unit test suite in tools/perf/tests/ftrace.c.

3. Inlined function call graph reconstruction

Introduces --inline and -k/--vmlinux, dynamically synthesising inlined frames marked with /* (inline) */. Correctly matches bare closing braces when funcgraph-tail is disabled, clamps stack depth to prevent indentation drift, and expands the automated test suite to verify inline resolution, comment filtering, and stack saturation recovery.

4. Execution duration suppression

Introduces --graph-opts noduration (accepting noduration=[0|1]) to toggle the kernel's funcgraph-duration tracing option, omitting the duration column when the user is focused purely on call graph topology and inlined hierarchy. Rejects noduration when invoked under 'perf ftrace profile'.

5. Extraneous function filtering in function graph tracer

Introduces userspace call graph filtering when targeted tracing (-G or --graph-funcs) is active. Suppresses transient unrelated entries arriving before filter activation as well as detached callee trees (such as ring-buffer wakeups) originating outside the target hierarchy.

Synthesised inlined functions are rendered with an explicit /* (inline) */ annotation at both entry and exit:

# tracer: function_graph
#
# CPU FUNCTION CALLS
# | | | | |
3) activate_task() {
3) enqueue_task() {
3) psi_enqueue() { /* (inline) */
3) arch_static_branch_jump() { /* (inline) */
3) enqueue_task_fair() {
3) enqueue_hierarchy() { /* (inline) */
3) update_curr();
3) enqueue_entity() {
3) account_entity_enqueue() { /* (inline) */
3) account_llc_enqueue();
3) } /* account_entity_enqueue (inline) */
3) }
3) } /* enqueue_hierarchy (inline) */
3) place_entity() {
3) avg_vruntime();
3) }
3) __enqueue_entity();
3) dl_server_start();
3) }
3) } /* arch_static_branch_jump (inline) */
3) } /* psi_enqueue (inline) */
3) sched_info_enqueue() { /* (inline) */
3) psi_task_change() {
3) psi_write_begin() { /* (inline) */
3) psi_flags_change();
3) } /* psi_write_begin (inline) */
3) psi_group_change() {
3) record_times();
3) }
3) }
3) } /* sched_info_enqueue (inline) */
3) }
3) }

Changes since v3:

- Explicitly rejected the noduration option when invoked under the profile subcommand, since profiling inherently requires function duration to compute timing statistics

- Fixed make_blank_prefix() clobbering the comm-pid column in verbose mode by searching backwards for the preceding "|" delimiter, and returning early when graph_noduration is active

- Fixed filter state not resetting for targeted leaf calls at non-zero call stack depths by checking cs->depth <= cs->target_root_depth and clearing cs->target_root_depth, preventing subsequent non-targeted sibling functions within the parent frame from leaking into trace output

- Expanded unit test coverage to verify comm-pid preservation under verbose and verbose-noduration inline modes, and to verify that targeted leaf calls at depth > 0 properly reset the target filter state.

- Link to v3: https://lore.kernel.org/lkml/[email protected]/

Changes since v2:

- Expanded the series from 4 to 5 patches by adding a dedicated commit ("perf ftrace: Filter extraneous functions in function graph tracer") to enforce strict target callgraph boundaries when -G or --graph-funcs is specified

- Suppressed unrelated top-level functions that leak through the transient tracefs filter initialisation window before RCU synchronisation completes

- Filtered detached/orphan calls from appearing as nested children within the target graph by validating caller return addresses against the active target stack

- Tracked nested orphan scope depth (cs->orphan_depth) to cleanly suppress child calls and closing delimiters of detached subtrees without desynchronising the target stack

- Added automated unit test coverage in tools/perf/tests/ftrace.c (test_graph_target_filter()) validating top-level suppression (both with and without --inline) and detached orphan wakeup suppression

- Link to v2: https://lore.kernel.org/lkml/[email protected]/

Changes since v1:

- Modularised the implementation into a 4-patch bisectable series rather than a monolithic commit

- Resolved potential infinite loop on poll() upon trace_pipe EOF

- Fixed stuck call stack frames when funcgraph-tail is disabled by supporting bare closing delimiters matching LIFO stack frames

- Prevented unbounded indentation growth on deep call trees by clamping inline expansion and synchronising cs->inlined_depth to frames successfully pushed onto cs->stack

- Added dedicated unit tests covering bare closing braces and stack saturation recovery

- Refined option documentation in perf-ftrace.txt

- Link to v1: https://lore.kernel.org/lkml/[email protected]/

Aaron Tomlin (5):
perf ftrace: Optimise __cmd_ftrace() stack frame with strbuf streaming perf ftrace: Support filtering return address comments perf ftrace: Support display of inlined functions in function graph
tracer
perf ftrace: Support omitting execution duration in function graph
tracer
perf ftrace: Filter extraneous functions in function graph tracer

 tools/perf/Documentation/perf-ftrace.txt |  20 +
 tools/perf/builtin-ftrace.c              | 143 +++-
 tools/perf/tests/Build                   |   1 +
 tools/perf/tests/builtin-test.c          |   1 +
 tools/perf/tests/ftrace.c                | 750 +++++++++++++++++++
 tools/perf/tests/tests.h                 |   1 +
 tools/perf/util/Build                    |   1 +
 tools/perf/util/ftrace.c                 | 887 +++++++++++++++++++++++
 tools/perf/util/ftrace.h                 |  17 +
 9 files changed, 1805 insertions(+), 16 deletions(-)
 create mode 100644 tools/perf/tests/ftrace.c
 create mode 100644 tools/perf/util/ftrace.c
Signature · 1 line
2.55.0
#1 lore raw

PATCH v4 1/5 perf ftrace: Optimise __cmd_ftrace() stack frame with strbuf streaming

In __cmd_ftrace(), incoming trace data from trace_pipe was historically buffered in a fixed 4096-byte array allocated directly on the stack frame. Furthermore, reads were written directly to stdout via fwrite() in arbitrary chunk sizes, rather than as discrete, well-delimited trace lines.

Refactor __cmd_ftrace() to employ dynamic heap allocation for the read buffer alongside a dynamic struct strbuf to accumulate streamed characters into whole lines. Pre-allocating the strbuf ensures that line processing incurs zero heap reallocations during steady-state polling, while the stack frame of __cmd_ftrace() is reduced by over 96% (down to ~160 bytes). Robust error and EOF detection on stdout and the trace stream ensures polling terminates cleanly upon pipeline or pager exit.

---
 tools/perf/builtin-ftrace.c | 65 +++++++++++++++++++++++++++++--------
 1 file changed, 52 insertions(+), 13 deletions(-)
tools/perf/builtin-ftrace.c +52 -13
index 6017493ff179..efd5343dc9ad 100644
@@ -41,7 +41,10 @@
 #include "util/units.h"
 #include "util/parse-sublevel-options.h"
 
+#include "util/strbuf.h"
+
 #define DEFAULT_TRACER  "function_graph"
+#define TRACE_BUF_SIZE  4096
 
 static volatile sig_atomic_t workload_exec_errno;
 static volatile sig_atomic_t done;
@@ -738,7 +741,8 @@ static int __cmd_ftrace(struct perf_ftrace *ftrace)
 {
 	char *trace_file;
 	int trace_fd;
-	char buf[4096];
+	char *buf = NULL;
+	struct strbuf linebuf = STRBUF_INIT;
 	struct pollfd pollfd = {
 		.events = POLLIN,
 	};
@@ -767,10 +771,18 @@ static int __cmd_ftrace(struct perf_ftrace *ftrace)
 
 	setup_pager();
 
+	buf = malloc(TRACE_BUF_SIZE);
+	if (!buf) {
+		pr_err("failed to allocate trace buffer\n");
+		goto out_reset;
+	}
+	strbuf_init(&linebuf, 512);
+	strbuf_setlen(&linebuf, 0);
+
 	trace_file = get_tracing_instance_file("trace_pipe");
 	if (!trace_file) {
 		pr_err("failed to open trace_pipe\n");
-		goto out_reset;
+		goto out_free;
 	}
 
 	trace_fd = open(trace_file, O_RDONLY);
@@ -779,7 +791,7 @@ static int __cmd_ftrace(struct perf_ftrace *ftrace)
 
 	if (trace_fd < 0) {
 		pr_err("failed to open trace_pipe\n");
-		goto out_reset;
+		goto out_free;
 	}
 
 	fcntl(trace_fd, F_SETFL, O_NONBLOCK);
@@ -810,20 +822,30 @@ static int __cmd_ftrace(struct perf_ftrace *ftrace)
 			break;
 
 		if (pollfd.revents & POLLIN) {
-			int n = read(trace_fd, buf, sizeof(buf));
-			if (n < 0)
-				break;
-			if (fwrite(buf, n, 1, stdout) != 1)
+			int n = read(trace_fd, buf, TRACE_BUF_SIZE);
+
+			if (n <= 0)
 				break;
+			for (int i = 0; i < n; i++) {
+				if (buf[i] == '\n') {
+					if (fprintf(stdout, "%s\n", linebuf.buf) < 0)
+						break;
+					strbuf_setlen(&linebuf, 0);
+				} else {
+					if (strbuf_addch(&linebuf, buf[i]) < 0)
+						goto out_close_fd;
+				}
+			}
 			/* flush output since stdout is in full buffering mode due to pager */
-			fflush(stdout);
+			if (fflush(stdout) < 0 || ferror(stdout))
+				break;
 		}
 	}
 
 	write_tracing_file("tracing_on", "0");
 
 	if (workload_exec_errno) {
-		const char *emsg = str_error_r(workload_exec_errno, buf, sizeof(buf));
+		const char *emsg = str_error_r(workload_exec_errno, buf, TRACE_BUF_SIZE);
 		/* flush stdout first so below error msg appears at the end. */
 		fflush(stdout);
 		pr_err("workload failed: %s\n", emsg);
@@ -831,16 +853,33 @@ static int __cmd_ftrace(struct perf_ftrace *ftrace)
 	}
 
 	/* read remaining buffer contents */
-	while (true) {
-		int n = read(trace_fd, buf, sizeof(buf));
+	while (!ferror(stdout)) {
+		int n = read(trace_fd, buf, TRACE_BUF_SIZE);
+
 		if (n <= 0)
 			break;
-		if (fwrite(buf, n, 1, stdout) != 1)
-			break;
+		for (int i = 0; i < n; i++) {
+			if (buf[i] == '\n') {
+				if (fprintf(stdout, "%s\n", linebuf.buf) < 0)
+					break;
+				strbuf_setlen(&linebuf, 0);
+			} else {
+				if (strbuf_addch(&linebuf, buf[i]) < 0)
+					goto out_close_fd;
+			}
+		}
+	}
+
+	if (!ferror(stdout) && linebuf.len > 0) {
+		fprintf(stdout, "%s\n", linebuf.buf);
+		strbuf_setlen(&linebuf, 0);
 	}
 
 out_close_fd:
 	close(trace_fd);
+out_free:
+	free(buf);
+	strbuf_release(&linebuf);
 out_reset:
 	exit_tracing_instance();
 out:
Signature · 1 line
2.55.0
#2 lore raw

Re: [PATCH perf-tools-next v4 1/5] perf ftrace: Optimise __cmd_ftrace() stack frame with strbuf streaming

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [Medium] perf: premature termination on EAGAIN during ftrace polling --

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] perf: premature termination on EAGAIN during ftrace polling
--

commit 0da8f6d83b353c321a754b6791e275b195595b54
Author: Aaron Tomlin <[email protected]>

perf ftrace: Optimise __cmd_ftrace() stack frame with strbuf streaming

This commit optimizes __cmd_ftrace() by replacing a fixed stack buffer with dynamic heap allocation and a strbuf to accumulate streamed characters into whole lines. This reduces stack usage significantly and adds error and EOF detection on stdout and the trace stream.

Quoted 4 lines
tools/perf/builtin-ftrace.c +0 -0
index 6017493ff1797..efd5343dc9adc 100644

[ ... ]

tools/perf/builtin-ftrace.c static int __cmd_ftrace(struct perf_ftrace *ftrace) · line 828

Quoted 12 lines
tools/perf/builtin-ftrace.c +3 -4
@@ -810,20 +822,30 @@ static int __cmd_ftrace(struct perf_ftrace *ftrace)
 			break;
 
 		if (pollfd.revents & POLLIN) {
-			int n = read(trace_fd, buf, sizeof(buf));
-			if (n < 0)
-				break;
-			if (fwrite(buf, n, 1, stdout) != 1)
+			int n = read(trace_fd, buf, TRACE_BUF_SIZE);
+
+			if (n <= 0)
 				break;

[Severity: Medium]
Since trace_pipe is opened in O_NONBLOCK mode, can this lead to premature termination of the perf ftrace tracing session?

If a spurious wakeup occurs for poll(), or if a non-fatal signal interrupts the syscall, read() returns -1 with errno == EAGAIN or EINTR. Unconditionally breaking out of the loop here stops tracing instead of retrying the read.

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
#3 lore raw

Re: [PATCH perf-tools-next v4 1/5] perf ftrace: Optimise __cmd_ftrace() stack frame with strbuf streaming

Quoted 8 lines · On Wed, Oct 14, 2026 at 04:26:52PM +0000, [email protected] wrote:

[Severity: Medium]
Since trace_pipe is opened in O_NONBLOCK mode, can this lead to premature termination of the perf ftrace tracing session?

If a spurious wakeup occurs for poll(), or if a non-fatal signal interrupts the syscall, read() returns -1 with errno == EAGAIN or EINTR. Unconditionally breaking out of the loop here stops tracing instead of retrying the read.

Yes. Because trace_fd is in O_NONBLOCK mode, spurious wakeups from poll() can result in read() returning -1 with EAGAIN or EWOULDBLOCK, and non-fatal signals cause read() or poll() to return -1 with EINTR. Unconditionally breaking on (n <= 0) aborted tracing prematurely. I will address this in the next iteration.

Signature · 1 line
Aaron Tomlin
#4 lore raw

PATCH v4 2/5 perf ftrace: Support filtering return address comments

In reply to Aaron Tomlin

When tracing with the Linux kernel's function graph tracer, enabling funcgraph-retaddr records the return address of each function entry as a comment in the trace stream (e.g. /* <-load_elf_phdrs+0x6c/0xb0 */). While valuable for tracing execution origins, these trailing comments can add considerable visual clutter to the call graph, particularly when examining deeply nested call trees or when synthesising higher-level abstractions.

Introduce support for filtering out return address comments from function graph trace output. Provide a standalone --filter-retaddr flag.

Key aspects of this implementation:

1. Return address parsing and filtering

Implement ftrace_filter_retaddr() in tools/perf/util/ftrace.c to locate return address comments and strip them in-place with memmove(3).

Additionally, introduce ftrace_parse_retaddr() to parse caller symbol names and offsets across diverse output styles (standard format, return values, and bracketed addresses), establishing the parsing foundation for subsequent inlined callchain resolution.

2. Command-line interface

Expose --filter-retaddr in tools/perf/builtin-ftrace.c, updating the documentation in tools/perf/Documentation/perf-ftrace.txt accordingly.

3. Automated verification

Add a dedicated unit test suite ("Ftrace return address processing") in tools/perf/tests/ftrace.c to validate comment parsing and string filtering across all supported formats.

Assisted-by Antigravity:gemini-3.1-pro
---
 tools/perf/Documentation/perf-ftrace.txt |  4 ++
 tools/perf/builtin-ftrace.c              |  8 +++
 tools/perf/tests/Build                   |  1 +
 tools/perf/tests/builtin-test.c          |  1 +
 tools/perf/tests/ftrace.c                | 80 ++++++++++++++++++++++
 tools/perf/tests/tests.h                 |  1 +
 tools/perf/util/Build                    |  1 +
 tools/perf/util/ftrace.c                 | 87 ++++++++++++++++++++++++
 tools/perf/util/ftrace.h                 |  5 ++
 9 files changed, 188 insertions(+)
 create mode 100644 tools/perf/tests/ftrace.c
 create mode 100644 tools/perf/util/ftrace.c
tools/perf/Documentation/perf-ftrace.txt +4 -0
index 3f3808e513fe..e8105083e67c 100644
@@ -134,6 +134,10 @@ OPTIONS for 'perf ftrace trace'
 	  - depth=<n>    - Set max depth for function graph tracer to follow.
 	  - tail         - Print function name at the end.
 
+--filter-retaddr::
+	Filter out function return address comments (`/* <-caller+offset */`)
+	from function_graph output.
+
 
 OPTIONS for 'perf ftrace latency'
 ---------------------------------
tools/perf/builtin-ftrace.c +8 -0
index efd5343dc9ad..b8ab76912a0e 100644
@@ -828,6 +828,8 @@ static int __cmd_ftrace(struct perf_ftrace *ftrace)
 				break;
 			for (int i = 0; i < n; i++) {
 				if (buf[i] == '\n') {
+					if (ftrace->filter_retaddr)
+						ftrace_filter_retaddr(linebuf.buf);
 					if (fprintf(stdout, "%s\n", linebuf.buf) < 0)
 						break;
 					strbuf_setlen(&linebuf, 0);
@@ -860,6 +862,8 @@ static int __cmd_ftrace(struct perf_ftrace *ftrace)
 			break;
 		for (int i = 0; i < n; i++) {
 			if (buf[i] == '\n') {
+				if (ftrace->filter_retaddr)
+					ftrace_filter_retaddr(linebuf.buf);
 				if (fprintf(stdout, "%s\n", linebuf.buf) < 0)
 					break;
 				strbuf_setlen(&linebuf, 0);
@@ -871,6 +875,8 @@ static int __cmd_ftrace(struct perf_ftrace *ftrace)
 	}
 
 	if (!ferror(stdout) && linebuf.len > 0) {
+		if (ftrace->filter_retaddr)
+			ftrace_filter_retaddr(linebuf.buf);
 		fprintf(stdout, "%s\n", linebuf.buf);
 		strbuf_setlen(&linebuf, 0);
 	}
@@ -1836,6 +1842,8 @@ int cmd_ftrace(int argc, const char **argv)
 		     "Size of per cpu buffer, needs to use a B, K, M or G suffix.", parse_buffer_size),
 	OPT_BOOLEAN(0, "inherit", &ftrace.inherit,
 		    "Trace children processes"),
+	OPT_BOOLEAN_SET(0, "filter-retaddr", &ftrace.filter_retaddr, &ftrace.filter_retaddr_set,
+			"Filter out funcgraph return address comments"),
 	OPT_INTEGER('D', "delay", &ftrace.target.initial_delay,
 		    "Number of milliseconds to wait before starting tracing after program start"),
 	OPT_PARENT(common_options),
tools/perf/tests/Build +1 -0
index 395f2866094a..136138932fcb 100644
@@ -72,6 +72,7 @@ perf-test-y += hwmon_pmu.o
 perf-test-y += tool_pmu.o
 perf-test-y += subcmd-help.o
 perf-test-y += kallsyms-split.o
+perf-test-y += ftrace.o
 
 ifeq ($(SRCARCH),$(filter $(SRCARCH),x86 arm arm64 powerpc riscv))
 perf-test-$(CONFIG_DWARF_UNWIND) += dwarf-unwind.o
tools/perf/tests/builtin-test.c +1 -0
index d2f594921e25..9663f48afb92 100644
@@ -137,6 +137,7 @@ static struct test_suite *generic_tests[] = {
 	&suite__perf_hooks,
 	&suite__unit_number__scnprint,
 	&suite__mem2node,
+	&suite__ftrace,
 	&suite__time_utils,
 	&suite__jit_write_elf,
 	&suite__pfm,
tools/perf/tests/ftrace.c new file +80 -0

80 lines not on this page · raw message

tools/perf/tests/tests.h +1 -0
index 9c96f33483d1..1aec32e094c8 100644
@@ -163,6 +163,7 @@ DECLARE_SUITE(perf_hooks);
 DECLARE_SUITE(unit_number__scnprint);
 DECLARE_SUITE(mem2node);
 DECLARE_SUITE(maps);
+DECLARE_SUITE(ftrace);
 DECLARE_SUITE(time_utils);
 DECLARE_SUITE(jit_write_elf);
 DECLARE_SUITE(api_io);
tools/perf/util/Build +1 -0
index 2c1f880c4c47..5b082d09006d 100644
@@ -169,6 +169,7 @@ perf-util-y += list_sort.o
 perf-util-y += mutex.o
 perf-util-y += sharded_mutex.o
 perf-util-y += intel-tpebs.o
+perf-util-y += ftrace.o
 
 perf-util-$(CONFIG_PERF_BPF_SKEL) += bpf_counter.o
 perf-util-$(CONFIG_PERF_BPF_SKEL) += bpf_counter_cgroup.o
tools/perf/util/ftrace.c new file +87 -0

87 lines not on this page · raw message

tools/perf/util/ftrace.h +5 -0
index 950f2efafad2..50216d3fef64 100644
@@ -39,6 +39,8 @@ struct perf_ftrace {
 	int			graph_verbose;
 	int			graph_thresh;
 	int			graph_tail;
+	bool			filter_retaddr;
+	bool			filter_retaddr_set;
 };
 
 struct filter_entry {
@@ -93,4 +95,7 @@ perf_ftrace__latency_cleanup_bpf(struct perf_ftrace *ftrace __maybe_unused)
 
 #endif  /* HAVE_BPF_SKEL */
 
+bool ftrace_parse_retaddr(const char *str, char *sym_name, size_t sym_len, u64 *offset);
+void ftrace_filter_retaddr(char *str);
+
 #endif  /* __PERF_FTRACE_H__ */
Signature · 1 line
2.55.0
#5 lore raw

Re: [PATCH perf-tools-next v4 2/5] perf ftrace: Support filtering return address comments

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [Low] perf: build failure due to unused variable 'endptr' --

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Low] perf: build failure due to unused variable 'endptr'
--

commit 8db97766d362752b2eec14b113b39acd9e702f2a
Author: Aaron Tomlin <[email protected]>

perf ftrace: Support filtering return address comments

This commit introduces support for filtering out return address comments from function graph trace output in perf ftrace. It adds the --filter-retaddr flag and implements the parsing and filtering logic.

Quoted 5 lines
tools/perf/util/ftrace.c new file +0 -0
new file mode 100644
index 0000000000000..344d36293f5ca

[ ... ]

Quoted 7 lines
tools/perf/util/ftrace.c new file +7 -0
+bool ftrace_parse_retaddr(const char *str, char *sym_name, size_t sym_len, u64 *offset)
+{
+	const char *p = strstr(str, "<-");
+	const char *plus;
+	const char *bracket;
+	char *endptr;
+	size_t name_len;

[ ... ]

tools/perf/util/ftrace.c line 57

Quoted 4 lines
tools/perf/util/ftrace.c new file +4 -0
+	memcpy(sym_name, p, name_len);
+	sym_name[name_len] = '\0';
+
+	*offset = strtoull(plus + 1, &endptr, 16);

[Severity: Low]
Will this cause a build failure due to an unused variable?

The tools/perf build system enables -Werror by default, which elevates -Wunused-but-set-variable to a build failure. The local variable endptr is written to by strtoull() but is never subsequently read or evaluated.

Could we pass NULL to strtoull() instead to resolve this and avoid breaking the build?

Quoted 2 lines
tools/perf/util/ftrace.c new file +2 -0
+	return true;
+}
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=2
#6 lore raw

Re: [PATCH perf-tools-next v4 2/5] perf ftrace: Support filtering return address comments

Quoted 9 lines · On Wed, Oct 14, 2026 at 04:24:14PM +0000, [email protected] wrote:

[Severity: Low]
Will this cause a build failure due to an unused variable?

The tools/perf build system enables -Werror by default, which elevates -Wunused-but-set-variable to a build failure. The local variable endptr is written to by strtoull() but is never subsequently read or evaluated.

Could we pass NULL to strtoull() instead to resolve this and avoid breaking the build?

Indeed. This will be resolved in the next iteration.

Signature · 1 line
Aaron Tomlin
#7 lore raw

PATCH v4 3/5 perf ftrace: Support display of inlined functions in function graph tracer

In reply to Aaron Tomlin

The Linux kernel's function graph tracer operates at the machine instruction level via compiler instrumentation (-fpatchable-function-entry), recording entry and return events for physical function calls. Consequently, functions inlined by the compiler are invisible in the resulting call graph, as no discrete call or return instructions are emitted for them. While technically expected, this omission frequently obscures the logical execution flow when analysing kernel subsystems heavily reliant upon inlining.

Address this by introducing the --inline option to perf ftrace for the function_graph tracer.

When the kernel is configured with CONFIG_FUNCTION_GRAPH_RETADDR=y, the funcgraph-retaddr trace option records the caller's return address on each function entry, manifested in the trace stream as a comment (e.g. /* <-irq_work_queue_on+0x81/0xa0 */). By interrogating DWARF debug information from the kernel image (vmlinux) utilising libdw, perf ftrace resolves this return address to its inlined callchain and synthesises the intermediate inlined frames directly into the streamed call graph.

Key aspects of this implementation:

1. Inlined frame presentation

Synthesised inlined functions are rendered with an explicit /* (inline) */ annotation at both entry and exit:

# CPU DURATION FUNCTION CALLS
# | | | | | | |
0) | wake_up_new_task() {
0) | activate_task() { /* (inline) */
0) 1.234 us | enqueue_task();
0) | } /* activate_task (inline) */
0) 2.567 us | }

This affords immediate visual clarity to the reader upon entering an inlined scope, without necessitating a search for the closing delimiter.

2. Call graph reconstruction and dynamic stack tracking

Maintains a per-CPU stack tracking real and inlined frames, synthesising entry and exit delimiters to accurately reconstruct the execution hierarchy. Inlined frames are structured strictly root-to-leaf and validated against the active stack to prevent un-traced callers from attaching orphan inlines. Bare closing braces without trailing function names (the default when funcgraph-tail is disabled) are correctly matched against preceding frames on the stack. Context column alignment, including task comm-pid delimiters in verbose mode, is preserved when generating prefixes for synthesised inline frames. Furthermore, stack depth saturation guards ensure that inlined frame depth tracking remains clamped and indentation cannot drift even under deep call trees exceeding FTRACE_MAX_STACK.

3. Option handling and diagnostics

Adds -k or --vmlinux to designate an explicit debug image, and provides clear diagnostic guidance should --inline be invoked on a kernel lacking CONFIG_FUNCTION_GRAPH_RETADDR. Automatically enables return address comment filtering unless explicitly requested.

4. Automated verification

Expands the unit test suite in tools/perf/tests/ftrace.c to validate DWARF inline symbol resolution, comment filtering toggles, bare closing braces handling, and stack saturation recovery.

Assisted-by Antigravity:gemini-3.1-pro
---
 tools/perf/Documentation/perf-ftrace.txt |  17 +-
 tools/perf/builtin-ftrace.c              |  58 ++-
 tools/perf/tests/ftrace.c                | 343 +++++++++++-
 tools/perf/util/ftrace.c                 | 632 ++++++++++++++++++++++-
 tools/perf/util/ftrace.h                 |  11 +
 5 files changed, 1043 insertions(+), 18 deletions(-)
tools/perf/Documentation/perf-ftrace.txt +16 -1
index e8105083e67c..445c8ccc22d2 100644
@@ -134,9 +134,24 @@ OPTIONS for 'perf ftrace trace'
 	  - depth=<n>    - Set max depth for function graph tracer to follow.
 	  - tail         - Print function name at the end.
 
+--inline::
+	Show inlined functions in function_graph tracer. This requires
+	kernel support for the funcgraph-retaddr option and a vmlinux with
+	debug symbols. Inlined functions are displayed in the call graph
+	hierarchy, and their inlined status is indicated with `/* (inline) */`
+	on entry and exit, e.g. `func() { /* (inline) */` and `} /* func (inline) */`.
+	By default, raw return address comments (`/* <-caller+offset */`)
+	are hidden in the output; specify `--graph-opts retaddr` to show them.
+
 --filter-retaddr::
 	Filter out function return address comments (`/* <-caller+offset */`)
-	from function_graph output.
+	from function_graph output. This is enabled by default when `--inline`
+	is used; use `--no-filter-retaddr` to disable it.
+
+-k::
+--vmlinux=::
+	Path to the vmlinux file containing debug symbols for resolving
+	inlined functions.
 
 
 OPTIONS for 'perf ftrace latency'
tools/perf/builtin-ftrace.c +45 -13

107 lines not on this page · raw message

tools/perf/tests/ftrace.c +342 -1

361 lines not on this page · raw message

tools/perf/util/ftrace.c +629 -3

645 lines not on this page · raw message

tools/perf/util/ftrace.h +11 -0
index 50216d3fef64..4140902d9de6 100644
@@ -41,6 +41,8 @@ struct perf_ftrace {
 	int			graph_tail;
 	bool			filter_retaddr;
 	bool			filter_retaddr_set;
+	bool			use_inline;
+	struct machine		*machine;
 };
 
 struct filter_entry {
@@ -95,7 +97,16 @@ perf_ftrace__latency_cleanup_bpf(struct perf_ftrace *ftrace __maybe_unused)
 
 #endif  /* HAVE_BPF_SKEL */
 
+struct machine;
+
 bool ftrace_parse_retaddr(const char *str, char *sym_name, size_t sym_len, u64 *offset);
 void ftrace_filter_retaddr(char *str);
+int ftrace_resolve_inlines(struct machine *machine,
+			   const char *sym_name, u64 offset,
+			   const char **inlined_names, int max_inlines);
+int ftrace_process_fgraph_line(struct perf_ftrace *ftrace, const char *line, FILE *out);
+void ftrace_cpu_states__clear(void);
+int perf_ftrace__setup_inlines(struct perf_ftrace *ftrace);
+void perf_ftrace__cleanup_inlines(struct perf_ftrace *ftrace);
 
 #endif  /* __PERF_FTRACE_H__ */
Signature · 1 line
2.55.0
#8 lore raw

Re: [PATCH perf-tools-next v4 3/5] perf ftrace: Support display of inlined functions in function graph tracer

Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider: - [Medium] perf: buffer overflow in make_blank_prefix() due to unclamped loop bounds - [Medium] perf…

Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider:
- [Medium] perf: buffer overflow in make_blank_prefix() due to unclamped loop bounds
- [Medium] perf ftrace: parser stack desynchronization on FTRACE_MAX_STACK saturation
- [Low] perf/ftrace: stale real frames prevent inactive inlines from being popped --

commit 74f9ce28f357fc14d2ef1925d4b8504c439b4901
Author: Aaron Tomlin <[email protected]>

perf ftrace: Support display of inlined functions in function graph tracer

This commit introduces the --inline option to perf ftrace for the function_graph tracer. It leverages DWARF debug information from the kernel image to resolve return addresses to their inlined callchains and synthesizes intermediate frames directly into the streamed call graph.

tools/perf/util/ftrace.c +0 -0

[ ... ]

Quoted 14 lines
(quoted diff) +14 -0
+static void make_blank_prefix(const struct perf_ftrace *ftrace, const char *line,
+			      const char *pipe_char, char *buf, size_t size)
+{
+	size_t len = pipe_char - line + 1;
+	const char *start_delim = NULL;
+	char *c;
+
+	if (len >= size)
+		len = size - 1;
+
+	memcpy(buf, line, len);
+	buf[len] = '\0';
+
+	if (*pipe_char == '|') {

[ ... ]

Review

Quoted 3 lines
(quoted diff) +3 -0
+		if (start_delim && start_delim < pipe_char) {
+			for (c = buf + (start_delim - line) + 1; c < buf + (pipe_char - line); c++)
+				*c = ' ';

[Severity: Medium]
Does this code overflow the buf array?

The buf array is populated from a fixed 128-byte array (blank_prefix) in the caller. The earlier memcpy is safely clamped to size - 1, but this loop bound uses the unclamped distance pipe_char - line.

If a trace line has an exceptionally long prefix (such as from a long comm string or malformed input), could pipe_char - line exceed the buffer size and cause out-of-bounds writes on the stack?

Quoted 3 lines
(quoted diff) +3 -0
+		}
+	}
+}

[ ... ]

Quoted 2 lines
(quoted diff) +2 -0
+int ftrace_process_fgraph_line(struct perf_ftrace *ftrace, const char *line, FILE *out)
+{

[ ... ]

Quoted 7 lines
(quoted diff) +7 -0
+	if (is_exit) {
+		char exit_name[128];
+
+		extract_exit_func_name(func_start, exit_name, sizeof(exit_name));
+
+		/* Find matching frame on stack */
+		match_idx = -1;

Review

Quoted 10 lines
(quoted diff) +10 -0
+		for (i = cs->depth - 1; i >= 0; i--) {
+			if (!cs->stack[i].inlined) {
+				if (exit_name[0] == '\0' ||
+				    (cs->stack[i].name &&
+				     strcmp(cs->stack[i].name, exit_name) == 0)) {
+					match_idx = i;
+					break;
+				}
+			}
+		}

[Severity: Medium]
Can this cause parser stack desynchronization on deep call stacks?

If a trace contains more than FTRACE_MAX_STACK (256) nested frames, the push logic correctly drops them to prevent overflowing the stack tracking. However, when these untracked frames exit, their trace lines often contain bare closing braces, which results in exit_name[0] == '\0'.

Will this unconditionally match the topmost valid non-inlined frame on the stack and pop it prematurely, permanently desynchronizing the stack state and breaking inline resolution for subsequent callers?

[ ... ]

Quoted 7 lines
(quoted diff) +7 -0
+	/* Close inlined frames that are no longer active */
+	while (active_inlines > common && cs->depth > 0) {
+		if (cs->stack[cs->depth - 1].inlined) {
+			struct ftrace_stack_frame *top = &cs->stack[cs->depth - 1];
+			int exit_indent = top->indent;
+
+			if (cs->inlined_depth > 0)

Review

Quoted 10 lines
(quoted diff) +10 -0
+				cs->inlined_depth--;
+			fprintf(out, "%s%*s} /* %s (inline) */\n",
+				blank_prefix, exit_indent, "", top->name);
+			free(top->name);
+			cs->depth--;
+			active_inlines--;
+		} else {
+			break;
+		}
+	}

[Severity: Low]
Can stale real frames prevent inactive inlined frames from being popped?

If a function exit event is dropped (for example, due to a ring buffer overwrite), the corresponding real frame remains on the stack. When the next entry event is processed and we attempt to close inactive inlined frames above the caller, the loop breaks immediately upon encountering this stale real frame.

Does this leave the inactive inlines stuck on the stack, corrupting the nesting and indentation of subsequent call graph output?

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=3
#9 lore raw

Re: [PATCH perf-tools-next v4 3/5] perf ftrace: Support display of inlined functions in function graph tracer

Quoted 10 lines · On Wed, Oct 14, 2026 at 04:27:39PM +0000, [email protected] wrote:

[Severity: Medium]
Does this code overflow the buf array?

The buf array is populated from a fixed 128-byte array (blank_prefix) in the caller. The earlier memcpy is safely clamped to size - 1, but this loop bound uses the unclamped distance pipe_char - line.

If a trace line has an exceptionally long prefix (such as from a long comm string or malformed input), could pipe_char - line exceed the buffer size and cause out-of-bounds writes on the stack?

The buffer blanking loop indeed used the unclamped offset (pipe_char - line). If a trace record prefix exceeded the 128-byte destination buffer, buf + (pipe_char - line) would point past the buffer boundary and overwrite stack memory with spaces.

This will be resolved in the next iteration.

Quoted 11 lines

[Severity: Medium]
Can this cause parser stack desynchronization on deep call stacks?

If a trace contains more than FTRACE_MAX_STACK (256) nested frames, the push logic correctly drops them to prevent overflowing the stack tracking. However, when these untracked frames exit, their trace lines often contain bare closing braces, which results in exit_name[0] == '\0'.

Will this unconditionally match the topmost valid non-inlined frame on the stack and pop it prematurely, permanently desynchronizing the stack state and breaking inline resolution for subsequent callers?

I will address this in the next iteration by introducing an overflow_depth counter in struct ftrace_cpu_state. Entries beyond FTRACE_MAX_STACK increment overflow_depth, and exits decrement overflow_depth without popping the tracked stack. Only once overflow_depth drops to zero are tracked stack frames popped. I have also expanded test_stack_saturation() in tools/perf/tests/ftrace.c to verify that overflow exits do not prematurely evict tracked frames.

Quoted 11 lines

[Severity: Low]
Can stale real frames prevent inactive inlined frames from being popped?

If a function exit event is dropped (for example, due to a ring buffer overwrite), the corresponding real frame remains on the stack. When the next entry event is processed and we attempt to close inactive inlined frames above the caller, the loop breaks immediately upon encountering this stale real frame.

Does this leave the inactive inlines stuck on the stack, corrupting the nesting and indentation of subsequent call graph output?

Breaking on non-inlined frames during entry unwinding indeed caused stale callee frames (left over from dropped exit events) to block the closure of inactive inlined frames underneath them. This left inactive inlines stranded on the stack and broke subsequent indentation.

I will address this in the next iteration.

Signature · 1 line
Aaron Tomlin
#10 lore raw

PATCH v4 4/5 perf ftrace: Support omitting execution duration in function graph tracer

In reply to Aaron Tomlin

The Linux kernel's function graph tracer displays the execution duration of completed functions by default via the funcgraph-duration trace option. Whilst informative, in scenarios where the user is solely interested in the call graph topology, call flow sequence, or inlined hierarchy, the duration column can add visual clutter and widen each trace line.

Introduce support for omitting function duration by exposing the noduration sub-option under --graph-opts. Consistent with the existing nosleep-time and noirqs sub-options, it accepts an optional value, so noduration=0 explicitly retains the duration column; when specified more than once, the last occurrence takes effect. When enabled, perf ftrace configures the kernel's tracing instance option funcgraph-duration to 0, and restores it to 1 upon resetting tracing options.

Because the profile subcommand fundamentally requires function execution duration to compute timing statistics and call counts, combining noduration with perf ftrace profile is rejected with an explicit error.

---
 tools/perf/Documentation/perf-ftrace.txt |  1 +
 tools/perf/builtin-ftrace.c              | 34 ++++++++++++-
 tools/perf/tests/ftrace.c                | 64 ++++++++++++++++++++++--
 tools/perf/util/ftrace.c                 | 18 +++++--
 tools/perf/util/ftrace.h                 |  1 +
 5 files changed, 109 insertions(+), 9 deletions(-)
tools/perf/Documentation/perf-ftrace.txt +1 -0
index 445c8ccc22d2..a50f3bae86b2 100644
@@ -129,6 +129,7 @@ OPTIONS for 'perf ftrace trace'
 	  - retaddr      - Show function return address.
 	  - nosleep-time - Measure on-CPU time only for function_graph tracer.
 	  - noirqs       - Ignore functions that happen inside interrupt.
+	  - noduration   - Do not print function duration.
 	  - verbose      - Show process names, PIDs, timestamps, etc.
 	  - thresh=<n>   - Setup trace duration threshold in microseconds.
 	  - depth=<n>    - Set max depth for function graph tracer to follow.
tools/perf/builtin-ftrace.c +33 -1
index 2707d3d62bca..1c86e0a711bd 100644
@@ -299,6 +299,7 @@ static void reset_tracing_options(struct perf_ftrace *ftrace __maybe_unused)
 	write_tracing_option_file("func_stack_trace", "0");
 	write_tracing_option_file("sleep-time", "1");
 	write_tracing_option_file("funcgraph-irqs", "1");
+	write_tracing_option_file("funcgraph-duration", "1");
 	write_tracing_option_file("funcgraph-proc", "0");
 	write_tracing_option_file("funcgraph-abstime", "0");
 	write_tracing_option_file("funcgraph-tail", "0");
@@ -598,6 +599,26 @@ static int set_tracing_funcgraph_irqs(struct perf_ftrace *ftrace)
 	return 0;
 }
 
+/**
+ * set_tracing_funcgraph_duration - Set funcgraph-duration tracing option
+ * @ftrace: Perf ftrace context
+ *
+ * Configure the kernel's funcgraph-duration tracing option based on the
+ * ftrace->graph_noduration flag.
+ *
+ * Return: 0 on success, or a negative error code on failure.
+ */
+static int set_tracing_funcgraph_duration(struct perf_ftrace *ftrace)
+{
+	if (!ftrace->graph_noduration)
+		return 0;
+
+	if (write_tracing_option_file("funcgraph-duration", "0") < 0)
+		return -1;
+
+	return 0;
+}
+
 static int set_tracing_funcgraph_verbose(struct perf_ftrace *ftrace)
 {
 	if (!ftrace->graph_verbose)
@@ -707,6 +728,11 @@ static int set_tracing_options(struct perf_ftrace *ftrace)
 		return -1;
 	}
 
+	if (set_tracing_funcgraph_duration(ftrace) < 0) {
+		pr_err("failed to set tracing option funcgraph-duration\n");
+		return -1;
+	}
+
 	if (set_tracing_funcgraph_verbose(ftrace) < 0) {
 		pr_err("failed to set tracing option funcgraph-proc/funcgraph-abstime\n");
 		return -1;
@@ -1760,6 +1786,7 @@ static int parse_graph_tracer_opts(const struct option *opt,
 		{ .name = "retaddr",		.value_ptr = &ftrace->graph_retaddr },
 		{ .name = "nosleep-time",	.value_ptr = &ftrace->graph_nosleep_time },
 		{ .name = "noirqs",		.value_ptr = &ftrace->graph_noirqs },
+		{ .name = "noduration",		.value_ptr = &ftrace->graph_noduration },
 		{ .name = "verbose",		.value_ptr = &ftrace->graph_verbose },
 		{ .name = "thresh",		.value_ptr = &ftrace->graph_thresh },
 		{ .name = "depth",		.value_ptr = &ftrace->graph_depth },
@@ -1849,7 +1876,7 @@ int cmd_ftrace(int argc, const char **argv)
 	OPT_CALLBACK('g', "nograph-funcs", &ftrace.nograph_funcs, "func",
 		     "Set nograph filter on given functions", parse_filter_func),
 	OPT_CALLBACK(0, "graph-opts", &ftrace, "options",
-		     "Graph tracer options, available options: args,retval,retval-hex,retaddr,nosleep-time,noirqs,verbose,thresh=<n>,depth=<n>",
+		     "Graph tracer options, available options: args,retval,retval-hex,retaddr,nosleep-time,noirqs,noduration,verbose,thresh=<n>,depth=<n>,tail",
 		     parse_graph_tracer_opts),
 	OPT_CALLBACK('m', "buffer-size", &ftrace.percpu_buffer_size, "size",
 		     "Size of per cpu buffer, needs to use a B, K, M or G suffix.", parse_buffer_size),
@@ -2056,6 +2083,11 @@ int cmd_ftrace(int argc, const char **argv)
 		cmd_func = __cmd_latency;
 		break;
 	case PERF_FTRACE_PROFILE:
+		if (ftrace.graph_noduration) {
+			pr_err("Error: 'noduration' option cannot be used with profile\n");
+			ret = -EINVAL;
+			goto out_delete_filters;
+		}
 		cmd_func = __cmd_profile;
 		break;
 	case PERF_FTRACE_NONE:
tools/perf/tests/ftrace.c +60 -4

99 lines not on this page · raw message

tools/perf/util/ftrace.c +14 -4
index 8a7f8f6eead1..0602e2cc1793 100644
@@ -292,6 +292,9 @@ static void make_blank_prefix(const struct perf_ftrace *ftrace, const char *line
 	memcpy(buf, line, len);
 	buf[len] = '\0';
 
+	if (ftrace->graph_noduration)
+		return;
+
 	if (*pipe_char == '|') {
 		if (ftrace->graph_verbose) {
 			/* In verbose mode, duration starts after the comm-pid '|' */
@@ -326,10 +329,12 @@ static void make_blank_prefix(const struct perf_ftrace *ftrace, const char *line
  *   [ abstime |]  CPU)  [ comm-pid |]  [ duration |]  func
  *
  * The abstime and comm-pid columns are only present with --graph-opts
- * verbose.
+ * verbose, whereas the duration column, and thus its '|' delimiter, is
+ * omitted with --graph-opts noduration.
  *
  * Return: A pointer to the last character of the context columns (the final
- * '|' delimiter). Returns NULL if @line is not a trace record.
+ * '|' delimiter, or the CPU ')' when no delimiter follows it). Returns NULL
+ * if @line is not a trace record (e.g. a header or a context switch marker).
  */
 static const char *find_func_column(const struct perf_ftrace *ftrace,
 				    const char *line, int *cpu)
@@ -357,9 +362,14 @@ static const char *find_func_column(const struct perf_ftrace *ftrace,
 			return NULL;
 	}
 
-	p = strchr(p + 1, '|');
-	if (!p)
+	if (!ftrace->graph_noduration) {
+		p = strchr(p + 1, '|');
+		if (!p)
+			return NULL;
+	} else if (strstr(p, " => ")) {
+		/* Context switch marker: " CPU) comm-pid => comm-pid" */
 		return NULL;
+	}
 
 	*cpu = (int)val;
 	return p;
tools/perf/util/ftrace.h +1 -0
index 4140902d9de6..d202a09d112f 100644
@@ -36,6 +36,7 @@ struct perf_ftrace {
 	int			graph_retaddr;
 	int			graph_nosleep_time;
 	int			graph_noirqs;
+	int			graph_noduration;
 	int			graph_verbose;
 	int			graph_thresh;
 	int			graph_tail;
Signature · 1 line
2.55.0
#11 lore raw

Re: [PATCH perf-tools-next v4 4/5] perf ftrace: Support omitting execution duration in function graph tracer

Sashiko has reviewed this patch and found no issues. It looks great!
Quoted 16 lines

The Linux kernel's function graph tracer displays the execution duration of completed functions by default via the funcgraph-duration trace option. Whilst informative, in scenarios where the user is solely interested in the call graph topology, call flow sequence, or inlined hierarchy, the duration column can add visual clutter and widen each trace line.

Introduce support for omitting function duration by exposing the noduration sub-option under --graph-opts. Consistent with the existing nosleep-time and noirqs sub-options, it accepts an optional value, so noduration=0 explicitly retains the duration column; when specified more than once, the last occurrence takes effect. When enabled, perf ftrace configures the kernel's tracing instance option funcgraph-duration to 0, and restores it to 1 upon resetting tracing options.

Quoted 5 lines

Because the profile subcommand fundamentally requires function execution duration to compute timing statistics and call counts, combining noduration with perf ftrace profile is rejected with an explicit error.

Sashiko has reviewed this patch and found no issues. It looks great!

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=4
#12 lore raw

PATCH v4 5/5 perf ftrace: Filter extraneous functions in function graph tracer

In reply to Aaron Tomlin

When tracing with the function graph tracer using a targeted filter (-G or --graph-funcs), users occasionally observe unrelated functions in the trace output:

1. Top-level functions (e.g. most notably finish_task_switch.isra.0()) appearing before or alongside the requested target function. This occurs because clearing and rewriting set_graph_function creates a transient window where remote CPUs still observe an empty filter (tracing all functions) before RCU synchronisation completes, coupled with perf ftrace sleeping in poll() on trace_pipe right as tracing starts.

2. Detached/orphan calls inside the target callgraph (e.g. housekeeping_any_cpu() and arch_irq_work_raise()) appearing as child frames under the target function. These routines are invoked by rb_wakeups() to notify perf's poll() on trace_pipe whenever an event is committed, and execute while the target function is on the CPU.

Resolve both issues deterministically in userspace:

1. Track active target callgraph state per CPU in tools/perf/util/ftrace.c. Any entry not matching the requested -G targets is suppressed along with its nested children until its scope exits. When a target function is matched (at root or nested depths), record its root depth; for leaf target calls, immediately reset the filter state upon completion to ensure subsequent sibling functions remain suppressed.

2. Within an active target callgraph, validate that child entries descend from an active frame on the target stack using their return addresses (CONFIG_FUNCTION_GRAPH_RETADDR). Any detached calls originating outside the target hierarchy (such as ring-buffer wakeups) are suppressed along with their subtrees.

For example:

# tracer: function_graph
#
# CPU FUNCTION CALLS
# | | | | |
3) activate_task() {
3) enqueue_task() {
3) psi_enqueue() { /* (inline) */
3) arch_static_branch_jump() { /* (inline) */
3) enqueue_task_fair() {
3) enqueue_hierarchy() { /* (inline) */
3) update_curr();
3) enqueue_entity() {
3) account_entity_enqueue() { /* (inline) */
3) account_llc_enqueue();
3) } /* account_entity_enqueue (inline) */
3) }
3) } /* enqueue_hierarchy (inline) */
3) place_entity() {
3) avg_vruntime();
3) }
3) __enqueue_entity();
3) dl_server_start();
3) }
3) } /* arch_static_branch_jump (inline) */
3) } /* psi_enqueue (inline) */
3) sched_info_enqueue() { /* (inline) */
3) psi_task_change() {
3) psi_write_begin() { /* (inline) */
3) psi_flags_change();
3) } /* psi_write_begin (inline) */
3) psi_group_change() {
3) record_times();
3) }
3) }
3) } /* sched_info_enqueue (inline) */
3) }
3) }

---
 tools/perf/tests/ftrace.c | 273 ++++++++++++++++++++++++++++++++++++++
 tools/perf/util/ftrace.c  | 176 +++++++++++++++++++++++-
 2 files changed, 443 insertions(+), 6 deletions(-)
tools/perf/tests/ftrace.c +273 -0
index 24b5bd4a1e03..1ae81b3efc2e 100644
@@ -447,6 +447,275 @@ static int test_column_layouts(void)
 	return TEST_OK;
 }
 
+static int test_graph_target_filter(void)
+{
+	static const char line_idle[] = " 7)   3.000 us    |  sched_idle_set_state();";
+	static const char line_sw_entry[] = " 6)               |  finish_task_switch.isra.0() {";
+	static const char line_sw_child[] = " 6)   1.000 us    |    _raw_spin_unlock();";
+	static const char line_sw_exit[] = " 6)   2.000 us    |  }";
+	static const char line_tgt_entry[] = " 2)               |  activate_task() {";
+	static const char line_tgt_child[] = " 2)   1.500 us    |    enqueue_task();";
+	static const char line_tgt_exit[] = " 2)   2.500 us    |  }";
+	struct perf_ftrace ftrace;
+	struct filter_entry *entry;
+	struct filter_entry *pos, *tmp;
+	char *out_buf = NULL;
+	size_t out_len = 0;
+	FILE *out;
+	int ret = TEST_OK;
+
+	memset(&ftrace, 0, sizeof(ftrace));
+	INIT_LIST_HEAD(&ftrace.graph_funcs);
+
+	entry = malloc(sizeof(*entry) + strlen("activate_task") + 1);
+	if (!entry)
+		return TEST_FAIL;
+	strcpy(entry->name, "activate_task");
+	list_add_tail(&entry->list, &ftrace.graph_funcs);
+
+	/* 1. Test without --inline */
+	ftrace.use_inline = false;
+	out = open_memstream(&out_buf, &out_len);
+	if (!out) {
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+
+	ftrace_process_fgraph_line(&ftrace, line_idle, out);
+	ftrace_process_fgraph_line(&ftrace, line_sw_entry, out);
+	ftrace_process_fgraph_line(&ftrace, line_sw_child, out);
+	ftrace_process_fgraph_line(&ftrace, line_sw_exit, out);
+	ftrace_process_fgraph_line(&ftrace, line_tgt_entry, out);
+	ftrace_process_fgraph_line(&ftrace, line_tgt_child, out);
+	ftrace_process_fgraph_line(&ftrace, line_tgt_exit, out);
+
+	fclose(out);
+	ftrace_cpu_states__clear();
+
+	if (strstr(out_buf, "sched_idle_set_state")) {
+		pr_debug("FAILED: sched_idle_set_state not suppressed\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+	if (strstr(out_buf, "finish_task_switch")) {
+		pr_debug("FAILED: finish_task_switch not suppressed\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+	if (strstr(out_buf, "_raw_spin_unlock")) {
+		pr_debug("FAILED: _raw_spin_unlock not suppressed\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+	if (!strstr(out_buf, "activate_task() {")) {
+		pr_debug("FAILED: activate_task entry missing\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+	if (!strstr(out_buf, "enqueue_task();")) {
+		pr_debug("FAILED: enqueue_task child missing\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+
+	free(out_buf);
+	out_buf = NULL;
+
+	/* 2. Test with --inline */
+	ftrace.use_inline = true;
+	out = open_memstream(&out_buf, &out_len);
+	if (!out) {
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+
+	ftrace_process_fgraph_line(&ftrace, line_idle, out);
+	ftrace_process_fgraph_line(&ftrace, line_sw_entry, out);
+	ftrace_process_fgraph_line(&ftrace, line_sw_child, out);
+	ftrace_process_fgraph_line(&ftrace, line_sw_exit, out);
+	ftrace_process_fgraph_line(&ftrace, line_tgt_entry, out);
+	ftrace_process_fgraph_line(&ftrace, line_tgt_child, out);
+	ftrace_process_fgraph_line(&ftrace, line_tgt_exit, out);
+
+	fclose(out);
+	ftrace_cpu_states__clear();
+
+	pr_debug("Target filter output with --inline:\n%s\n", out_buf);
+
+	if (strstr(out_buf, "finish_task_switch")) {
+		pr_debug("FAILED: finish_task_switch not suppressed with --inline\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+	if (!strstr(out_buf, "activate_task() {")) {
+		pr_debug("FAILED: activate_task entry missing with --inline\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+
+	free(out_buf);
+	out_buf = NULL;
+
+	/* 3. Test suppressing detached orphan calls within target graph */
+	ftrace.use_inline = true;
+	ftrace.filter_retaddr = 1;
+	out = open_memstream(&out_buf, &out_len);
+	if (!out) {
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+
+	ftrace_process_fgraph_line(&ftrace, line_tgt_entry, out);
+	/* Leaf orphan call from ring_buffer_unlock_commit */
+	ftrace_process_fgraph_line(&ftrace,
+		" 2)               |    housekeeping_any_cpu(); /* <-ring_buffer_unlock_commit+0x1a0/0x1f0 */",
+		out);
+	/* Non-leaf orphan call from irq_work_queue_on */
+	ftrace_process_fgraph_line(&ftrace,
+		" 2)               |    arch_irq_work_raise() { /* <-irq_work_queue_on+0x81/0xa0 */",
+		out);
+	ftrace_process_fgraph_line(&ftrace,
+		" 2)               |      apic_wait_icr_idle() { /* (inline) */",
+		out);
+	ftrace_process_fgraph_line(&ftrace,
+		" 2)               |        x2apic_send_IPI_self();",
+		out);
+	ftrace_process_fgraph_line(&ftrace,
+		" 2)               |      } /* apic_wait_icr_idle (inline) */",
+		out);
+	ftrace_process_fgraph_line(&ftrace,
+		" 2)   1.000 us    |    }",
+		out);
+	/* Genuine child call from activate_task */
+	ftrace_process_fgraph_line(&ftrace,
+		" 2)               |    enqueue_task() { /* <-activate_task+0x1d1/0x3e0 */",
+		out);
+	ftrace_process_fgraph_line(&ftrace,
+		" 2)   1.500 us    |    }",
+		out);
+	ftrace_process_fgraph_line(&ftrace, line_tgt_exit, out);
+
+	fclose(out);
+	ftrace_cpu_states__clear();
+
+	pr_debug("Orphan call filter output:\n%s\n", out_buf);
+
+	if (strstr(out_buf, "housekeeping_any_cpu")) {
+		pr_debug("FAILED: housekeeping_any_cpu orphan not suppressed\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+	if (strstr(out_buf, "arch_irq_work_raise")) {
+		pr_debug("FAILED: arch_irq_work_raise orphan not suppressed\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+	if (strstr(out_buf, "x2apic_send_IPI_self")) {
+		pr_debug("FAILED: x2apic_send_IPI_self not suppressed\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+	if (!strstr(out_buf, "enqueue_task() {")) {
+		pr_debug("FAILED: genuine enqueue_task missing\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+
+	free(out_buf);
+	out_buf = NULL;
+
+	/* 4. Test targeted leaf call at depth > 0 resetting filter */
+	ftrace.use_inline = false;
+	list_for_each_entry_safe(pos, tmp, &ftrace.graph_funcs, list) {
+		list_del(&pos->list);
+		free(pos);
+	}
+	entry = malloc(sizeof(*entry) + strlen("enqueue_task") + 1);
+	if (!entry) {
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+	strcpy(entry->name, "enqueue_task");
+	list_add_tail(&entry->list, &ftrace.graph_funcs);
+
+	out = open_memstream(&out_buf, &out_len);
+	if (!out) {
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+
+	ftrace_process_fgraph_line(&ftrace, line_sw_entry, out);
+	ftrace_process_fgraph_line(&ftrace, line_tgt_child, out);
+	ftrace_process_fgraph_line(&ftrace, line_sw_child, out);
+	ftrace_process_fgraph_line(&ftrace, line_sw_exit, out);
+
+	fclose(out);
+	ftrace_cpu_states__clear();
+
+	if (strstr(out_buf, "finish_task_switch")) {
+		pr_debug("FAILED: finish_task_switch not suppressed\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+	if (!strstr(out_buf, "enqueue_task();")) {
+		pr_debug("FAILED: targeted leaf enqueue_task missing\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+	if (strstr(out_buf, "_raw_spin_unlock")) {
+		pr_debug("FAILED: sibling _raw_spin_unlock not suppressed after leaf target\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+
+	free(out_buf);
+	out_buf = NULL;
+
+	/* 5. Test targeted leaf call at depth > 0 with inlines resetting filter */
+	ftrace.use_inline = true;
+	out = open_memstream(&out_buf, &out_len);
+	if (!out) {
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+
+	ftrace_process_fgraph_line(&ftrace, line_sw_entry, out);
+	ftrace_process_fgraph_line(&ftrace, line_tgt_child, out);
+	ftrace_process_fgraph_line(&ftrace, line_sw_child, out);
+	ftrace_process_fgraph_line(&ftrace, line_sw_exit, out);
+
+	fclose(out);
+	ftrace_cpu_states__clear();
+
+	if (strstr(out_buf, "finish_task_switch")) {
+		pr_debug("FAILED: finish_task_switch not suppressed with inlines\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+	if (!strstr(out_buf, "enqueue_task();")) {
+		pr_debug("FAILED: targeted leaf enqueue_task missing with inlines\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+	if (strstr(out_buf, "_raw_spin_unlock")) {
+		pr_debug("FAILED: sibling _raw_spin_unlock leaked with inlines\n");
+		ret = TEST_FAIL;
+		goto out_free_filter;
+	}
+
+	free(out_buf);
+	out_buf = NULL;
+
+out_free_filter:
+	if (out_buf)
+		free(out_buf);
+	list_for_each_entry_safe(pos, tmp, &ftrace.graph_funcs, list) {
+		list_del(&pos->list);
+		free(pos);
+	}
+	return ret;
+}
+
 static int test__ftrace(struct test_suite *test __maybe_unused, int subtest __maybe_unused)
 {
 	int ret;
@@ -467,6 +736,10 @@ static int test__ftrace(struct test_suite *test __maybe_unused, int subtest __ma
 	if (ret != TEST_OK)
 		return ret;
 
+	ret = test_graph_target_filter();
+	if (ret != TEST_OK)
+		return ret;
+
 	ret = test_inline_processing();
 	if (ret != TEST_OK)
 		return ret;
tools/perf/util/ftrace.c +170 -6

231 lines not on this page · raw message

Signature · 1 line
2.55.0
#13 lore raw

Re: [PATCH perf-tools-next v4 5/5] perf ftrace: Filter extraneous functions in function graph tracer

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider: - [Medium] perf/ftrace: filter state leak on leaf functions with inlines - [Medium] perf: stack buffer…

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [Medium] perf/ftrace: filter state leak on leaf functions with inlines
- [Medium] perf: stack buffer overflow in make_blank_prefix()
--

commit 5c319821fcd67517723413eff78ea02ad9a706a1
Author: Aaron Tomlin <[email protected]>

perf ftrace: Filter extraneous functions in function graph tracer

This commit introduces a mechanism to track active callgraph state per CPU in the perf ftrace function graph tracer. It suppresses unrelated top-level functions and detached orphan calls from the trace output when using a targeted filter.

Quoted 3 lines
tools/perf/util/ftrace.c +0 -0

[ ... ]

tools/perf/util/ftrace.c int ftrace_process_fgraph_line(struct perf_ftrace *ftrace, const char *line, FIL · line 487

Quoted 11 lines
tools/perf/util/ftrace.c +2 -1
@@ -443,8 +479,9 @@ int ftrace_process_fgraph_line(struct perf_ftrace *ftrace, const char *line, FIL
 	bool has_open_brace;
 	bool has_semicolon;
 	bool has_retaddr = false;
+	bool has_filter = ftrace_has_graph_funcs(ftrace);
 
-	if (!ftrace->use_inline)
+	if (!has_filter && !ftrace->use_inline)
 		return print_line(ftrace, line, out);
 
 	pipe_char = find_func_column(ftrace, line, &cpu);

[Severity: Medium]
Shortly after this in ftrace_process_fgraph_line(), make_blank_prefix() is called and provided with the 128-byte blank_prefix array from the stack:

    if (start_delim && start_delim < pipe_char) {
        for (c = buf + (start_delim - line) + 1; c < buf + (pipe_char - line); c++)
            *c = ' ';
    }

Can this loop in make_blank_prefix() write out of bounds?

The loop blindly writes space characters up to buf + (pipe_char - line). If the trace line prefix before pipe_char exceeds 128 characters, it appears this will write past the end of the blank_prefix stack buffer without clamping to the size of the destination.

[ ... ]

Quoted 9 lines
tools/perf/util/ftrace.c +5 -0
@@ -549,6 +691,21 @@ int ftrace_process_fgraph_line(struct perf_ftrace *ftrace, const char *line, FIL
 		}
 	}
 
+	/*
+	 * When inside a target graph filter, child functions must descend
+	 * from an active frame on the target stack. Suppress detached
+	 * calls originating outside the target hierarchy.
+	 */

tools/perf/util/ftrace.c int ftrace_process_fgraph_line(struct perf_ftrace *ftrace, const char *line, FIL · line 708

Quoted 10 lines
tools/perf/util/ftrace.c +10 -0
+	if (has_filter && cs->in_target && cs->depth > cs->target_root_depth) {
+		if (has_retaddr && caller_idx < cs->target_root_depth) {
+			if (has_open_brace && !has_semicolon)
+				cs->orphan_depth = 1;
+			for (i = 0; i < num_inlines; i++)
+				free((void *)inlined_names[i]);
+			return 0;
+		}
+	}
+

[Severity: Medium]
Because cs->depth is inflated by the inline frames for a leaf target function, the filter state check later on fails and leaves cs->in_target set to true.

When that happens, does this block cause subsequent sibling functions to incorrectly return early, entirely skipping the logic that would normally close the stale inline frames?

[ ... ]

Quoted 4 lines
tools/perf/util/ftrace.c +0 -0
@@ -659,6 +816,13 @@ int ftrace_process_fgraph_line(struct perf_ftrace *ftrace, const char *line, FIL
 		}
 	}
 

tools/perf/util/ftrace.c int ftrace_process_fgraph_line(struct perf_ftrace *ftrace, const char *line, FIL · line 828

Quoted 10 lines
tools/perf/util/ftrace.c +7 -0
+	if (has_filter && cs->in_target && cs->depth <= cs->target_root_depth &&
+	    has_semicolon && !has_open_brace) {
+		cs->in_target = false;
+		cs->target_root_depth = 0;
+		cs->orphan_depth = 0;
+	}
+
 	for (i = 0; i < num_inlines; i++)
 		free((void *)inlined_names[i]);
 	return 0;

[Severity: Medium]
Does this reset condition fail for leaf functions when ftrace->use_inline is true?

If a targeted leaf function resolves to inline frames, ftrace_process_fgraph_line() pushes the inline frames onto cs->stack and increments cs->depth. Because cs->depth is inflated by num_inlines, the check against cs->target_root_depth will be false, leaving in_target permanently true for leaf calls with inlines.

Could this leak the filter state, causing subsequent sibling functions to be incorrectly evaluated as being inside the target graph, leading to malformed or missing trace lines?

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=5
#14 lore raw

Re: [PATCH perf-tools-next v4 5/5] perf ftrace: Filter extraneous functions in function graph tracer

Quoted 9 lines · On Wed, Oct 14, 2026 at 04:26:31PM +0000, [email protected] wrote:

[Severity: Medium]
Shortly after this in ftrace_process_fgraph_line(), make_blank_prefix() is called and provided with the 128-byte blank_prefix array from the stack:

    if (start_delim && start_delim < pipe_char) {
        for (c = buf + (start_delim - line) + 1; c < buf + (pipe_char - line); c++)
            *c = ' ';
    }
Quoted 6 lines

Can this loop in make_blank_prefix() write out of bounds?

The loop blindly writes space characters up to buf + (pipe_char - line). If the trace line prefix before pipe_char exceeds 128 characters, it appears this will write past the end of the blank_prefix stack buffer without clamping to the size of the destination.

Well make_blank_prefix() was already updated with explicit destination bounds clamping. Specifically, len is clamped to (size - 1), start_delim is checked against (line + len), and the end pointer for blanking is capped:

        char *end = buf + (pipe_char - line);

        if (end > buf + len)
            end = buf + len;
        for (c = buf + (start_delim - line) + 1; c < end; c++)
            *c = ' ';

This ensures writes cannot exceed the destination array boundary even if prefix context columns exceed 128 bytes.

Quoted 18 lines

[Severity: Medium]
Because cs->depth is inflated by the inline frames for a leaf target function, the filter state check later on fails and leaves cs->in_target set to true.

When that happens, does this block cause subsequent sibling functions to incorrectly return early, entirely skipping the logic that would normally close the stale inline frames?
[ ... ]
[Severity: Medium]
Does this reset condition fail for leaf functions when ftrace->use_inline is true?

If a targeted leaf function resolves to inline frames, ftrace_process_fgraph_line() pushes the inline frames onto cs->stack and increments cs->depth. Because cs->depth is inflated by num_inlines, the check against cs->target_root_depth will be false, leaving in_target permanently true for leaf calls with inlines.

Quoted 3 lines

Could this leak the filter state, causing subsequent sibling functions to be incorrectly evaluated as being inside the target graph, leading to malformed or missing trace lines?

Yes. When a targeted function is a leaf call (semicolon with no open brace) and has synthesised caller inlines, cs->depth is inflated by num_inlines, causing the check cs->depth <= cs->target_root_depth to evaluate to false. in_target remained true, and subsequent non-targeted sibling functions outside the target hierarchy were treated as detached orphan calls and dropped early, bypassing inline frame unwinding and leaving unclosed frames on cs->stack.

This will be resolved in the next iteration.

Signature · 1 line
Aaron Tomlin