Skip to contentSkip to conversation

perf c2c hardening

PATCH Superseded Arnaldo Carvalho de Melo 2 participants 5 messages Aug 4 perf events

v7 of 8 Superseded Latest is v8 →

lore
Share & tools
lore

Arnaldo Carvalho de Melo

author maintainer
#0 lore raw

v7 cover perf c2c hardening

Hi,

Fixes for 'perf c2c' found by the sashiko-bot and Opencode:mimo-v2.5-free AI reviewers, both using Chris Mason's review-prompts:

Patch 1 fixes three silent failure modes in hpp_list__parse().

Patch 2 fixes a format list leak: when c2c_hists__[re]init() fails partway through, entries registered via perf_hpp_list__column_register() and perf_hpp_list__register_sort_field() are left on the hpp_list.

Please consider merging,

- Arnaldo

Changes since v6:

- c2c_hists__reinit() has the same leak: it clears the list then calls hpp_list__parse(), which can register formats before failing, and neither the function nor its callers clean up on failure. Add the same cleanup-on-failure pattern to close this gap.

Changes since v5:

- The second v5 patch didn't got thru, resending the same contents as v5.

Changes since v4:

- Replace %m with str_error_r(-ret) as suggested by Sashiko.

Changes since v3:

- Use 'goto out' to break out from both the switch and the for loop in __hpp_list__parse() as noticed by sashiko.

Changes since v2:

- Added the string.h and stdlib.h missing headers.
- Addressed sashiko comment on PARSE_LIST, turning it into a function and handling all errors returned from _fn().

Changes since v1:

- Addressed sashiko-bot [Medium] finding: the early exits added by v1 skip perf_hpp__setup_output_field().
- Dismissed sashiko-bot [Low] finding (missing string.h/stdlib.h):
strdup() and free() were already used in this function before this patch; string2.h (included at the top of the file) pulls in string.h. No new library calls were introduced.
- Subject updated to reflect the additional caller fixes.

This series was developed with AI assistance (Claude:Sonnet 4.6, Opencode:mimo-v2.5-free).

Arnaldo Carvalho de Melo (2):
perf c2c: Fix error masking, OOM, and unchecked caller errors in

    hpp_list__parse()

perf c2c: Clean up registered formats on c2c_hists__init() and
c2c_hists__reinit() failure

 tools/perf/builtin-c2c.c | 105 +++++++++++++++++++++++++++++----------
 tools/perf/util/sort.c   |  38 ++++++++------
 2 files changed, 100 insertions(+), 43 deletions(-)
Signature · 1 line
2.55.0
author maintainer
#1 lore raw

PATCH v7 1/2 perf c2c: Fix error masking, OOM, and unchecked caller errors in hpp_list__parse()

From: Arnaldo Carvalho de Melo <[email protected]>

hpp_list__parse() has three bugs:

1. The PARSE_LIST macro resets ret = 0 at the start of each invocation, so an error from output parsing is silently overwritten when the sort parsing block runs. The function returns success with partially initialized state.

2. When the caller passes a non-NULL output_ or sort_ string, but strdup() returns NULL due to OOM, NULL is passed to PARSE_LIST which treats it as empty input (the "if (!_list) break" branch). No error is returned.

3. When the called _fn function fails and returns something other than -ESRCH or -EINVAL (-ENOMEM, for instance) it was not bailing out of the strtok loop.

Fix them by checking strdup() return values before proceeding and adding a cleanup label so that ret from each PARSE_LIST call is checked before the next runs, preserving the first error.

The early exits now skip perf_hpp__setup_output_field(), which means c2c_hists__reinit() can return a non-zero value in cases that previously always succeeded silently. Both callers discarded its return:
resort_cl_cb() continued into hists__collapse_resort() on a broken list, and perf_c2c__report() proceeded with uninitialised hists. Fix the full
chain: check and propagate the error in resort_cl_cb() -- hists__iterate_cb() already stops iteration and returns the callback error -- and check both c2c_hists__reinit() and hists__iterate_cb() in perf_c2c__report().

Also turn PARSE_LIST into a function, using a switch to catch other errors, converting the called functions to return an appropriate errno instead of -1 on failure.

Fixes 2d388bd0c9d3 ("perf c2c report: Add stdio output support")
Cc Jiri Olsa <[email protected]>
Assisted-by Claude:claude-sonnet-4.6
---
 tools/perf/builtin-c2c.c | 85 ++++++++++++++++++++++++++++------------
 tools/perf/util/sort.c   | 38 ++++++++++--------
 2 files changed, 82 insertions(+), 41 deletions(-)
tools/perf/builtin-c2c.c +60 -25
index c9584dbedf77afe8..160b82694d391c50 100644
@@ -12,11 +12,14 @@
  */
 #include <errno.h>
 #include <inttypes.h>
+#include <stdlib.h>
+#include <string.h>
 
 #include <asm/bug.h>
 #include <linux/compiler.h>
 #include <linux/err.h>
 #include <linux/kernel.h>
+#include <linux/string.h>
 #include <linux/stringify.h>
 #include <linux/zalloc.h>
 #include <sys/param.h>
@@ -2063,26 +2066,38 @@ static int c2c_hists__init_sort(struct perf_hpp_list *hpp_list, char *name, stru
 	return 0;
 }
 
-#define PARSE_LIST(_list, _fn)							\
-	do {									\
-		char *tmp, *tok;						\
-		ret = 0;							\
-										\
-		if (!_list)							\
-			break;							\
-										\
-		for (tok = strtok_r((char *)_list, ", ", &tmp);			\
-				tok; tok = strtok_r(NULL, ", ", &tmp)) {	\
-			ret = _fn(hpp_list, tok, env);				\
-			if (ret == -EINVAL) {					\
-				pr_err("Invalid --fields key: `%s'", tok);	\
-				break;						\
-			} else if (ret == -ESRCH) {				\
-				pr_err("Unknown --fields key: `%s'", tok);	\
-				break;						\
-			}							\
-		}								\
-	} while (0)
+static int __hpp_list__parse(struct perf_hpp_list *hpp_list, char *_list, struct perf_env *env,
+			     int (*_fn)(struct perf_hpp_list *hpp_list, char *name, struct perf_env *env))
+{
+	char *tmp, *tok;
+	int ret = 0;
+
+	if (!_list)
+		return 0;
+
+	for (tok = strtok_r(_list, ", ", &tmp); tok; tok = strtok_r(NULL, ", ", &tmp)) {
+		ret = _fn(hpp_list, tok, env);
+		switch (ret) {
+		case 0:
+			continue;
+		case -EINVAL:
+			pr_err("Invalid --fields key: `%s'", tok);
+			goto out;
+		case -ESRCH:
+			pr_err("Unknown --fields key: `%s'", tok);
+			goto out;
+		default: {
+			char buf[STRERR_BUFSIZE];
+
+			pr_err("%s for --fields key: `%s'",
+			       str_error_r(-ret, buf, sizeof(buf)), tok);
+			goto out;
+		}
+		}
+	}
+out:
+	return ret;
+}
 
 static int hpp_list__parse(struct perf_hpp_list *hpp_list,
 			   const char *output_,
@@ -2093,8 +2108,18 @@ static int hpp_list__parse(struct perf_hpp_list *hpp_list,
 	char *sort   = sort_   ? strdup(sort_) : NULL;
 	int ret;
 
-	PARSE_LIST(output, c2c_hists__init_output);
-	PARSE_LIST(sort,   c2c_hists__init_sort);
+	/* strdup() returns NULL on OOM, don't silently treat as empty */
+	if ((output_ && !output) || (sort_ && !sort)) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	ret = __hpp_list__parse(hpp_list, output, env, c2c_hists__init_output);
+	if (ret)
+		goto out;
+	ret = __hpp_list__parse(hpp_list, sort, env, c2c_hists__init_sort);
+	if (ret)
+		goto out;
 
 	/* copy sort keys to output fields */
 	perf_hpp__setup_output_field(hpp_list);
@@ -2111,6 +2136,7 @@ static int hpp_list__parse(struct perf_hpp_list *hpp_list,
 	perf_hpp__append_sort_keys(&hists->list);
 #endif
 
+out:
 	free(output);
 	free(sort);
 	return ret;
@@ -2281,6 +2307,7 @@ static int resort_cl_cb(struct hist_entry *he, void *arg)
 	struct c2c_hist_entry *c2c_he;
 	struct c2c_hists *c2c_hists;
 	bool display = he__display(he, &c2c.shared_clines_stats);
+	int ret;
 
 	c2c_he = container_of(he, struct c2c_hist_entry, he);
 	c2c_hists = c2c_he->hists;
@@ -2291,7 +2318,9 @@ static int resort_cl_cb(struct hist_entry *he, void *arg)
 		c2c_he->cacheline_idx = idx++;
 		calc_width(c2c_he);
 
-		c2c_hists__reinit(c2c_hists, c2c.cl_output, c2c.cl_resort, env);
+		ret = c2c_hists__reinit(c2c_hists, c2c.cl_output, c2c.cl_resort, env);
+		if (ret)
+			return ret;
 
 		hists__collapse_resort(&c2c_hists->hists, NULL);
 		hists__output_resort_cb(&c2c_hists->hists, NULL, filter_cb);
@@ -3356,13 +3385,19 @@ static int perf_c2c__report(int argc, const char **argv)
 	else if (c2c.display == DISPLAY_SNP_PEER)
 		sort_str = "tot_peer";
 
-	c2c_hists__reinit(&c2c.hists, output_str, sort_str, perf_session__env(session));
+	err = c2c_hists__reinit(&c2c.hists, output_str, sort_str, perf_session__env(session));
+	if (err) {
+		pr_err("Failed to reinitialize hists\n");
+		goto out_mem2node;
+	}
 
 	ui_progress__init(&prog, c2c.hists.hists.nr_entries, "Sorting...");
 
 	hists__collapse_resort(&c2c.hists.hists, NULL);
 	hists__output_resort_cb(&c2c.hists.hists, &prog, resort_shared_cl_cb);
-	hists__iterate_cb(&c2c.hists.hists, resort_cl_cb, perf_session__env(session));
+	err = hists__iterate_cb(&c2c.hists.hists, resort_cl_cb, perf_session__env(session));
+	if (err)
+		goto out_mem2node;
 
 	ui_progress__finish();
 
tools/perf/util/sort.c +22 -16
index dcf9189786f8aeba..2f59e37089ef06a2 100644
@@ -3105,7 +3105,7 @@ static int __sort_dimension__add_hpp_sort(struct sort_dimension *sd,
 	struct hpp_sort_entry *hse = __sort_dimension__alloc_hpp(sd, level);
 
 	if (hse == NULL)
-		return -1;
+		return -ENOMEM;
 
 	perf_hpp_list__register_sort_field(list, &hse->hpp);
 	return 0;
@@ -3118,7 +3118,7 @@ static int __sort_dimension__add_hpp_output(struct sort_dimension *sd,
 	struct hpp_sort_entry *hse = __sort_dimension__alloc_hpp(sd, level);
 
 	if (hse == NULL)
-		return -1;
+		return -ENOMEM;
 
 	perf_hpp_list__column_register(list, &hse->hpp);
 	return 0;
@@ -3742,14 +3742,18 @@ static int __sort_dimension__add(struct sort_dimension *sd,
 				 struct perf_hpp_list *list,
 				 int level)
 {
+	int ret;
+
 	if (sd->taken)
 		return 0;
 
-	if (__sort_dimension__add_hpp_sort(sd, list, level) < 0)
-		return -1;
+	ret = __sort_dimension__add_hpp_sort(sd, list, level);
+	if (ret < 0)
+		return ret;
 
-	if (__sort_dimension__update(sd, list) < 0)
-		return -1;
+	ret = __sort_dimension__update(sd, list);
+	if (ret < 0)
+		return ret;
 
 	sd->taken = 1;
 
@@ -3767,7 +3771,7 @@ static int __hpp_dimension__add(struct hpp_dimension *hd,
 
 	fmt = __hpp_dimension__alloc_hpp(hd, level);
 	if (!fmt)
-		return -1;
+		return -ENOMEM;
 
 	hd->taken = 1;
 	hd->was_taken = 1;
@@ -3779,14 +3783,18 @@ static int __sort_dimension__add_output(struct perf_hpp_list *list,
 					struct sort_dimension *sd,
 					int level)
 {
+	int ret;
+
 	if (sd->taken)
 		return 0;
 
-	if (__sort_dimension__add_hpp_output(sd, list, level) < 0)
-		return -1;
+	ret = __sort_dimension__add_hpp_output(sd, list, level);
+	if (ret < 0)
+		return ret;
 
-	if (__sort_dimension__update(sd, list) < 0)
-		return -1;
+	ret = __sort_dimension__update(sd, list);
+	if (ret < 0)
+		return ret;
 
 	sd->taken = 1;
 	return 0;
@@ -3803,7 +3811,7 @@ static int __hpp_dimension__add_output(struct perf_hpp_list *list,
 
 	fmt = __hpp_dimension__alloc_hpp(hd, level);
 	if (!fmt)
-		return -1;
+		return -ENOMEM;
 
 	hd->taken = 1;
 	perf_hpp_list__column_register(list, fmt);
@@ -3869,8 +3877,7 @@ int sort_dimension__add(struct perf_hpp_list *list, const char *tok,
 				    strlen(tok)))
 			return -EINVAL;
 
-		__sort_dimension__add(sd, list, level);
-		return 0;
+		return __sort_dimension__add(sd, list, level);
 	}
 
 	for (i = 0; i < ARRAY_SIZE(memory_sort_dimensions); i++) {
@@ -3882,8 +3889,7 @@ int sort_dimension__add(struct perf_hpp_list *list, const char *tok,
 		if (sort__mode != SORT_MODE__MEMORY)
 			return -EINVAL;
 
-		__sort_dimension__add(sd, list, level);
-		return 0;
+		return __sort_dimension__add(sd, list, level);
 	}
 
 	for (i = 0; i < ARRAY_SIZE(hpp_sort_dimensions); i++) {
Signature · 1 line
2.55.0
#2 lore raw

Re: [PATCH 1/2] perf c2c: Fix error masking, OOM, and unchecked caller errors in hpp_list__parse()

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [Medium] Silent error masking and inconsistent propagation for -ENOMEM during sort and output token…

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Silent error masking and inconsistent propagation for -ENOMEM during sort and output token parsing in setup_sort_list() and setup_output_list(). --

--- Patch [1]: [PATCH 1/2] perf c2c: Fix error masking, OOM, and unchecked caller errors in hpp_list__parse() --- commit c1a9e4a1694a477755b6abff539aa40b1fed31e1
Author: Arnaldo Carvalho de Melo <[email protected]>

perf c2c: Fix error masking, OOM, and unchecked caller errors in hpp_list__parse()

This commit fixes three bugs in hpp_list__parse() by checking strdup() return values and adding a cleanup label to preserve early errors. It also converts PARSE_LIST to a function, changing called functions to return appropriate error codes like -ENOMEM instead of -1 on failure.

Quoted 4 lines
tools/perf/util/sort.c +0 -0
index 005e7d85dc4a5..616a583e3d4d6 100644

[ ... ]

tools/perf/util/sort.c int sort_dimension__add(struct perf_hpp_list *list, const char *tok, · line 3883

Quoted 10 lines
tools/perf/util/sort.c +1 -2
@@ -3869,8 +3877,7 @@ int sort_dimension__add(struct perf_hpp_list *list, const char *tok,
 				    strlen(tok)))
 			return -EINVAL;
 
-		__sort_dimension__add(sd, list, level);
-		return 0;
+		return __sort_dimension__add(sd, list, level);
 	}
 
 	for (i = 0; i < ARRAY_SIZE(memory_sort_dimensions); i++) {

[Severity: Medium]
Since sort_dimension__add() can now return -ENOMEM, could this introduce a regression where the callers in setup_sort_list() and setup_output_list() fail to handle the new return value?

Currently, setup_sort_list() only checks for -EINVAL and -ESRCH:

tools/perf/util/sort.c:setup_sort_list() {
    ...
		ret = sort_dimension__add(list, tok, evlist, env, level);
		if (ret == -EINVAL) {
			...
			break;
		} else if (ret == -ESRCH) {
			...
			break;
		}
    ...
}

If sort_dimension__add() returns -ENOMEM, the loop will not break and ret will be silently overwritten on the next iteration.

A similar issue appears to exist in setup_output_list() which ignores -ENOMEM from output_field_add().

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

PATCH v7 2/2 perf c2c: Clean up registered formats on c2c_hists__init() and c2c_hists__reinit() failure

In reply to Arnaldo Carvalho de Melo

From: Arnaldo Carvalho de Melo <[email protected]>

When c2c_hists__init() or c2c_hists__reinit() calls hpp_list__parse() and it fails partway through, format structures registered via perf_hpp_list__column_register() and perf_hpp_list__register_sort_field() are left on the hpp_list.

In c2c_hists__init(), the c2c_he__alloc_hists() caller handles this with perf_hpp__reset_output_field(), but perf_c2c_report() does not, leaking the partially registered entries.

In c2c_hists__reinit(), neither perf_c2c_report() nor resort_cl_cb() clean up on failure.

Fix by adding cleanup inside both functions themselves, so all callers are protected.

Fixes 78b275437873 ("perf c2c report: Add sample processing")
Cc Jiri Olsa <[email protected]>
Assisted-by Claude:claude-opus-4.6
Assisted-by Opencode:mimo-v2.5-free
---
 tools/perf/builtin-c2c.c | 20 ++++++++++++++++++--
 1 file changed, 18 insertions(+), 2 deletions(-)
tools/perf/builtin-c2c.c +18 -2
index 160b82694d391c50..1856cccbee27818b 100644
@@ -2147,6 +2147,8 @@ static int c2c_hists__init(struct c2c_hists *hists,
 			   int nr_header_lines,
 			   struct perf_env *env)
 {
+	int ret;
+
 	__hists__init(&hists->hists, &hists->list);
 
 	/*
@@ -2159,7 +2161,13 @@ static int c2c_hists__init(struct c2c_hists *hists,
 	/* Overload number of header lines.*/
 	hists->list.nr_header_lines = nr_header_lines;
 
-	return hpp_list__parse(&hists->list, /*output=*/NULL, sort, env);
+	ret = hpp_list__parse(&hists->list, /*output=*/NULL, sort, env);
+
+	/* Unregister any formats added before the failure point */
+	if (ret)
+		perf_hpp__reset_output_field(&hists->list);
+
+	return ret;
 }
 
 static int c2c_hists__reinit(struct c2c_hists *c2c_hists,
@@ -2167,8 +2175,16 @@ static int c2c_hists__reinit(struct c2c_hists *c2c_hists,
 			     const char *sort,
 			     struct perf_env *env)
 {
+	int ret;
+
 	perf_hpp__reset_output_field(&c2c_hists->list);
-	return hpp_list__parse(&c2c_hists->list, output, sort, env);
+	ret = hpp_list__parse(&c2c_hists->list, output, sort, env);
+
+	/* Unregister any formats added before the failure point */
+	if (ret)
+		perf_hpp__reset_output_field(&c2c_hists->list);
+
+	return ret;
 }
 
 #define DISPLAY_LINE_LIMIT  0.001
Signature · 1 line
2.55.0
#4 lore raw

Re: [PATCH 2/2] perf c2c: Clean up registered formats on c2c_hists__init() and c2c_hists__reinit() failure

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Pre-existing issues:
- [Medium] print_pareto() calls hpp_list__parse() on a local stack variable but never cleans it up, leaking dynamically allocated formats.
- [Medium] kvm_hists__init() and kvm_hists__reinit() suffer from identical partial-registration leaks when format parsing fails. --

--- Patch [2]: [PATCH 2/2] perf c2c: Clean up registered formats on c2c_hists__init() and c2c_hists__reinit() failure --- commit 92165e1ae89bef5baeedb3dd8908e10cf564714e
Author: Arnaldo Carvalho de Melo <[email protected]>

perf c2c: Clean up registered formats on c2c_hists__init() and c2c_hists__reinit() failure

When c2c_hists__init() or c2c_hists__reinit() calls hpp_list__parse() and it fails partway through, format structures registered are left on the hpp_list. This commit adds cleanup inside both functions to protect all callers from leaking partially registered entries.

Quoted 4 lines
tools/perf/builtin-c2c.c +0 -0
index 160b82694d391..1856cccbee278 100644

[ ... ]

Quoted 6 lines
tools/perf/builtin-c2c.c +2 -0
@@ -2167,8 +2175,16 @@ static int c2c_hists__reinit(struct c2c_hists *c2c_hists,
 			     const char *sort,
 			     struct perf_env *env)
 {
+	int ret;
+

tools/perf/builtin-c2c.c static int c2c_hists__reinit(struct c2c_hists *c2c_hists, · line 2188

Quoted 10 lines
tools/perf/builtin-c2c.c +7 -1
 	perf_hpp__reset_output_field(&c2c_hists->list);
-	return hpp_list__parse(&c2c_hists->list, output, sort, env);
+	ret = hpp_list__parse(&c2c_hists->list, output, sort, env);
+
+	/* Unregister any formats added before the failure point */
+	if (ret)
+		perf_hpp__reset_output_field(&c2c_hists->list);
+
+	return ret;
 }

[Severity: Medium]
This isn't a bug introduced by this patch, but I noticed that print_pareto() in the same file calls hpp_list__parse() on a local stack variable:

tools/perf/builtin-c2c.c:print_pareto() {
    ...
    ret = hpp_list__parse(&hpp_list, cl_output, /*evlist=*/NULL, env);
    ...
}

Because print_pareto() returns without ever calling perf_hpp__reset_output_field(), does it leak all allocated c2c_fmt structures on both the success and failure paths?

[Severity: Medium]
This is a pre-existing issue, but kvm_hists__init() and kvm_hists__reinit() in tools/perf/builtin-kvm.c seem to suffer from identical partial-registration leaks when format parsing fails:

tools/perf/builtin-kvm.c:kvm_hists__init() {
    ...
    return kvm_hpp_list__parse(&kvm_hists.list, NULL, "ev_name");
}
tools/perf/builtin-kvm.c:kvm_hists__reinit() {
    ...
    return kvm_hpp_list__parse(&kvm_hists.list, output, sort);
}

If parsing fails partway through these functions, will partially allocated formats remain in the global list without cleanup, replicating the exact leak pattern fixed here in c2c?

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=2