Skip to contentSkip to conversation

perf parse-events: Fix parsing of >30kb event strings

PATCH In review Ian Rogers 2 participants 4 messages Sep 5, 2025 – Oct 4, 2025 perf events

lore
Share & tools
lore

Ian Rogers

author reviewer
#0 lore raw

PATCH 1/1 perf parse-events: Fix parsing of >30kb event strings

Metrics may generate many particularly uncore event references. The resulting event string may then be >32kb. The parse events lex is using "%option reject" which stores backtracking state in a buffer sized at roughtly 30kb. If the event string is larger than this then a buffer overflow and typically a crash happens.

The need for "%option reject" was for BPF events which were removed in commit 3d6dfae88917 ("perf parse-events: Remove BPF event support"). As "%option reject" is both a memory and performance cost let's remove it and fix the parsing case for event strings being over ~30kb.

Whilst cleaning up "%option reject" make the header files accurately reflect functions used in the code and tidy up not requiring yywrap.

Measuring on the "PMU JSON event tests" a modest reduction of 0.41% user time and 0.27% max resident size was observed. More importantly this change fixes parsing large metrics and event strings.

---
 tools/perf/util/parse-events.l | 17 +++--------------
 1 file changed, 3 insertions(+), 14 deletions(-)
tools/perf/util/parse-events.l +3 -14
index 2034590eb789..1eaa8dbc26d8 100644
@@ -5,16 +5,14 @@
 %option stack
 %option bison-locations
 %option yylineno
-%option reject
+%option noyywrap
 
 %{
 #include <errno.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <unistd.h>
+#include <stdlib.h>
+#include <stdio.h>
 #include "parse-events.h"
 #include "parse-events-bison.h"
-#include "evsel.h"
 
 char *parse_events_get_text(yyscan_t yyscanner);
 YYSTYPE *parse_events_get_lval(yyscan_t yyscanner);
@@ -222,10 +220,6 @@ do {							\
 	yycolumn += yyleng;				\
 } while (0);
 
-#define USER_REJECT		\
-	yycolumn -= yyleng;	\
-	REJECT
-
 %}
 
 %x mem
@@ -423,8 +417,3 @@ r{num_raw_hex}		{ return str(yyscanner, PE_RAW); }
 .			{ }
 
 %%
-
-int parse_events_wrap(void *scanner __maybe_unused)
-{
-	return 1;
-}
Signature · 1 line
2.51.0.355.g5224444f11-goog
author reviewer
#1 lore raw
Quoted 13 lines · Ian Rogers

Metrics may generate many particularly uncore event references. The resulting event string may then be >32kb. The parse events lex is using "%option reject" which stores backtracking state in a buffer sized at roughtly 30kb. If the event string is larger than this then a buffer overflow and typically a crash happens.

The need for "%option reject" was for BPF events which were removed in commit 3d6dfae88917 ("perf parse-events: Remove BPF event support"). As "%option reject" is both a memory and performance cost let's remove it and fix the parsing case for event strings being over ~30kb.

Quoted 8 lines

Whilst cleaning up "%option reject" make the header files accurately reflect functions used in the code and tidy up not requiring yywrap.

Measuring on the "PMU JSON event tests" a modest reduction of 0.41% user time and 0.27% max resident size was observed. More importantly this change fixes parsing large metrics and event strings.

Ping. I think this may have gotten lost in noise about things like hardware json and the python metrics. It is a small change, bug fix and performance win. It can land independently of anything else. PTAL.

Thanks,
Ian

Quoted 51 lines
---
 tools/perf/util/parse-events.l | 17 +++--------------
 1 file changed, 3 insertions(+), 14 deletions(-)
tools/perf/util/parse-events.l +3 -15
index 2034590eb789..1eaa8dbc26d8 100644
@@ -5,16 +5,14 @@
 %option stack
 %option bison-locations
 %option yylineno
-%option reject
+%option noyywrap
 
 %{
 #include <errno.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <unistd.h>
+#include <stdlib.h>
+#include <stdio.h>
 #include "parse-events.h"
 #include "parse-events-bison.h"
-#include "evsel.h"
 
 char *parse_events_get_text(yyscan_t yyscanner);
 YYSTYPE *parse_events_get_lval(yyscan_t yyscanner);
@@ -222,10 +220,6 @@ do {                                                       \
        yycolumn += yyleng;                             \
 } while (0);
 
-#define USER_REJECT            \
-       yycolumn -= yyleng;     \
-       REJECT
-
 %}
 
 %x mem
@@ -423,8 +417,3 @@ r{num_raw_hex}              { return str(yyscanner, PE_RAW); }
 .                      { }
 
 %%
-
-int parse_events_wrap(void *scanner __maybe_unused)
-{
-       return 1;
-}
--

2.51.0.355.g5224444f11-goog

author reviewer
#2 lore raw
Quoted 22 lines · Ian Rogers
Quoted 20 lines · Ian Rogers

Metrics may generate many particularly uncore event references. The resulting event string may then be >32kb. The parse events lex is using "%option reject" which stores backtracking state in a buffer sized at roughtly 30kb. If the event string is larger than this then a buffer overflow and typically a crash happens.

The need for "%option reject" was for BPF events which were removed in commit 3d6dfae88917 ("perf parse-events: Remove BPF event support"). As "%option reject" is both a memory and performance cost let's remove it and fix the parsing case for event strings being over ~30kb.

Whilst cleaning up "%option reject" make the header files accurately reflect functions used in the code and tidy up not requiring yywrap.

Measuring on the "PMU JSON event tests" a modest reduction of 0.41% user time and 0.27% max resident size was observed. More importantly this change fixes parsing large metrics and event strings.

Quoted 5 lines
Quoted 1 line

Ping. I think this may have gotten lost in noise about things like hardware json and the python metrics. It is a small change, bug fix and performance win. It can land independently of anything else. PTAL.

It'd be nice to land this for the v6.18 pull. I don't think it has any visible implications and just makes stuff better.

Thanks,
Ian

Quoted 54 lines

Thanks,
Ian

Quoted 51 lines
---
 tools/perf/util/parse-events.l | 17 +++--------------
 1 file changed, 3 insertions(+), 14 deletions(-)
tools/perf/util/parse-events.l +3 -15
index 2034590eb789..1eaa8dbc26d8 100644
@@ -5,16 +5,14 @@
 %option stack
 %option bison-locations
 %option yylineno
-%option reject
+%option noyywrap
 
 %{
 #include <errno.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <unistd.h>
+#include <stdlib.h>
+#include <stdio.h>
 #include "parse-events.h"
 #include "parse-events-bison.h"
-#include "evsel.h"
 
 char *parse_events_get_text(yyscan_t yyscanner);
 YYSTYPE *parse_events_get_lval(yyscan_t yyscanner);
@@ -222,10 +220,6 @@ do {                                                       \
        yycolumn += yyleng;                             \
 } while (0);
 
-#define USER_REJECT            \
-       yycolumn -= yyleng;     \
-       REJECT
-
 %}
 
 %x mem
@@ -423,8 +417,3 @@ r{num_raw_hex}              { return str(yyscanner, PE_RAW); }
 .                      { }
 
 %%
-
-int parse_events_wrap(void *scanner __maybe_unused)
-{
-       return 1;
-}
--

2.51.0.355.g5224444f11-goog

#3 lore raw
Quoted 9 lines · Ian Rogers
Quoted 7 lines · Ian Rogers
Quoted 6 lines · Ian Rogers

Whilst cleaning up "%option reject" make the header files accurately reflect functions used in the code and tidy up not requiring yywrap.

Measuring on the "PMU JSON event tests" a modest reduction of 0.41% user time and 0.27% max resident size was observed. More importantly this change fixes parsing large metrics and event strings.

Quoted 6 lines
Quoted 3 lines

Ping. I think this may have gotten lost in noise about things like hardware json and the python metrics. It is a small change, bug fix and performance win. It can land independently of anything else. PTAL.

It'd be nice to land this for the v6.18 pull. I don't think it has any visible implications and just makes stuff better.

Applied and now testing on the container build set.

- Arnaldo