mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH] tools lib traceevent: Fix double free corruption in error processing code
@ 2015-07-17  7:58 Wang Nan
  2015-07-17 16:17 ` Steven Rostedt
  0 siblings, 1 reply; 2+ messages in thread
From: Wang Nan @ 2015-07-17  7:58 UTC (permalink / raw)
  To: rostedt, hekuang
  Cc: ast, acme, a.p.zijlstra, mingo, namhyung, jolsa, pi3orama, linux-kernel

When kernel introduces new function to 'format' file before traceevent
update correspondingly, a double-free corruption occures if the newly
intorduced function resides in following format:

# cat /sys/kernel/debug/tracing/events/bpf/bpf_output_data/format
...
print fmt: "%s", __print_hex(__get_dynamic_array(buf), __get_dynamic_array_len(buf))
...

(where __get_dynamic_array_len is the new function)

And:
 # perf record -e bpf:bpf_output_data ls
   Warning: [bpf:bpf_output_data] function __get_dynamic_array_len not defined
   Warning: Error: expected type 5 but read 0
 *** Error in `/path/to/perf': double free or corruption (fasttop): 0x0000000001821210 ***
 ======= Backtrace: =========
 /lib64/libc.so.6(+0x6eeef)[0x7f86a4cf6eef]
 /lib64/libc.so.6(+0x78cae)[0x7f86a4d00cae]
 /lib64/libc.so.6(+0x79987)[0x7f86a4d01987]
 /path/to/perf[0x51b0e2]
 /path/to/perf[0x51afc2]
 /path/to/perf[0x51fe67]
 /path/to/perf[0x5200a5]
 /path/to/perf(__pevent_parse_format+0x185)[0x525bc9]
 /path/to/perf[0x525d82]
 /path/to/perf(pevent_parse_format+0x3b)[0x525e15]
 /path/to/perf[0x4ceb61]
 /path/to/perf(perf_evsel__newtp_idx+0x9f)[0x48cabf]
 /path/to/perf(parse_events_add_tracepoint+0x25b)[0x49782b]
 /path/to/perf(parse_events_parse+0x10a3)[0x4c8243]
 /path/to/perf(parse_events+0x75)[0x498ce5]
 /path/to/perf(parse_events_option+0x41)[0x498df1]
 /path/to/perf[0x493e9b]
 /path/to/perf(parse_options+0x215)[0x4957c5]
 /path/to/perf(cmd_record+0x6a)[0x43c4ba]
 /path/to/perf[0x47b2a3]
 /path/to/perf(main+0x5f6)[0x42fed6]
 /lib64/libc.so.6(__libc_start_main+0xf5)[0x7f86a4ca9bd5]
 /path/to/perf[0x430005]
 ======= Memory map: ========
 00400000-006b9000 r-xp 00000000 08:05 23596607                           /path/to/perf
 008b9000-00903000 rw-p 002b9000 08:05 23596607                           /path/to/perf

This is caused by error processing code in process_hex() which frees
arg->hex.field but doesn't set it to NULL. When event_read_print_args()
freeing the hex arg, the dangling pointer will be free again and cause
the above error. process_int_array() has similar problem.

This patch fixes the dangling pointer problem by not setting them until
everything is okay.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Alexei Starovoitov <ast@plumgrid.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: <pi3orama@163.com>
---
 tools/lib/traceevent/event-parse.c | 25 +++++++++++++++++--------
 1 file changed, 17 insertions(+), 8 deletions(-)

diff --git a/tools/lib/traceevent/event-parse.c b/tools/lib/traceevent/event-parse.c
index cc25f05..8d662b5 100644
--- a/tools/lib/traceevent/event-parse.c
+++ b/tools/lib/traceevent/event-parse.c
@@ -2546,19 +2546,23 @@ out_free:
 static enum event_type
 process_hex(struct event_format *event, struct print_arg *arg, char **tok)
 {
+	struct print_arg *field, *size;
+
 	memset(arg, 0, sizeof(*arg));
 	arg->type = PRINT_HEX;
 
-	if (alloc_and_process_delim(event, ",", &arg->hex.field))
+	if (alloc_and_process_delim(event, ",", &field))
 		goto out;
 
-	if (alloc_and_process_delim(event, ")", &arg->hex.size))
+	if (alloc_and_process_delim(event, ")", &size))
 		goto free_field;
 
+	arg->hex.field = field;
+	arg->hex.size = size;
 	return read_token_item(tok);
 
 free_field:
-	free_arg(arg->hex.field);
+	free_arg(field);
 out:
 	*tok = NULL;
 	return EVENT_ERROR;
@@ -2567,24 +2571,29 @@ out:
 static enum event_type
 process_int_array(struct event_format *event, struct print_arg *arg, char **tok)
 {
+	struct print_arg *field, *count, *el_size;
+
 	memset(arg, 0, sizeof(*arg));
 	arg->type = PRINT_INT_ARRAY;
 
-	if (alloc_and_process_delim(event, ",", &arg->int_array.field))
+	if (alloc_and_process_delim(event, ",", &field))
 		goto out;
 
-	if (alloc_and_process_delim(event, ",", &arg->int_array.count))
+	if (alloc_and_process_delim(event, ",", &count))
 		goto free_field;
 
-	if (alloc_and_process_delim(event, ")", &arg->int_array.el_size))
+	if (alloc_and_process_delim(event, ")", &el_size))
 		goto free_size;
 
+	arg->int_array.field = field;
+	arg->int_array.count = count;
+	arg->int_array.el_size = el_size;
 	return read_token_item(tok);
 
 free_size:
-	free_arg(arg->int_array.count);
+	free_arg(count);
 free_field:
-	free_arg(arg->int_array.field);
+	free_arg(field);
 out:
 	*tok = NULL;
 	return EVENT_ERROR;
-- 
1.8.3.4


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [PATCH] tools lib traceevent: Fix double free corruption in error processing code
  2015-07-17  7:58 [PATCH] tools lib traceevent: Fix double free corruption in error processing code Wang Nan
@ 2015-07-17 16:17 ` Steven Rostedt
  0 siblings, 0 replies; 2+ messages in thread
From: Steven Rostedt @ 2015-07-17 16:17 UTC (permalink / raw)
  To: Wang Nan
  Cc: hekuang, ast, acme, a.p.zijlstra, mingo, namhyung, jolsa,
	pi3orama, linux-kernel

On Fri, 17 Jul 2015 07:58:46 +0000
Wang Nan <wangnan0@huawei.com> wrote:

> When kernel introduces new function to 'format' file before traceevent
> update correspondingly, a double-free corruption occures if the newly
> intorduced function resides in following format:
> 
> # cat /sys/kernel/debug/tracing/events/bpf/bpf_output_data/format
> ...
> print fmt: "%s", __print_hex(__get_dynamic_array(buf), __get_dynamic_array_len(buf))
> ...
> 
> (where __get_dynamic_array_len is the new function)
> 
> And:
>  # perf record -e bpf:bpf_output_data ls
>    Warning: [bpf:bpf_output_data] function __get_dynamic_array_len not defined
>    Warning: Error: expected type 5 but read 0
>  *** Error in `/path/to/perf': double free or corruption (fasttop): 0x0000000001821210 ***
>  ======= Backtrace: =========
>  /lib64/libc.so.6(+0x6eeef)[0x7f86a4cf6eef]
>  /lib64/libc.so.6(+0x78cae)[0x7f86a4d00cae]
>  /lib64/libc.so.6(+0x79987)[0x7f86a4d01987]
>  /path/to/perf[0x51b0e2]
>  /path/to/perf[0x51afc2]
>  /path/to/perf[0x51fe67]
>  /path/to/perf[0x5200a5]
>  /path/to/perf(__pevent_parse_format+0x185)[0x525bc9]
>  /path/to/perf[0x525d82]
>  /path/to/perf(pevent_parse_format+0x3b)[0x525e15]
>  /path/to/perf[0x4ceb61]
>  /path/to/perf(perf_evsel__newtp_idx+0x9f)[0x48cabf]
>  /path/to/perf(parse_events_add_tracepoint+0x25b)[0x49782b]
>  /path/to/perf(parse_events_parse+0x10a3)[0x4c8243]
>  /path/to/perf(parse_events+0x75)[0x498ce5]
>  /path/to/perf(parse_events_option+0x41)[0x498df1]
>  /path/to/perf[0x493e9b]
>  /path/to/perf(parse_options+0x215)[0x4957c5]
>  /path/to/perf(cmd_record+0x6a)[0x43c4ba]
>  /path/to/perf[0x47b2a3]
>  /path/to/perf(main+0x5f6)[0x42fed6]
>  /lib64/libc.so.6(__libc_start_main+0xf5)[0x7f86a4ca9bd5]
>  /path/to/perf[0x430005]
>  ======= Memory map: ========
>  00400000-006b9000 r-xp 00000000 08:05 23596607                           /path/to/perf
>  008b9000-00903000 rw-p 002b9000 08:05 23596607                           /path/to/perf
> 
> This is caused by error processing code in process_hex() which frees
> arg->hex.field but doesn't set it to NULL. When event_read_print_args()
> freeing the hex arg, the dangling pointer will be free again and cause
> the above error. process_int_array() has similar problem.
> 
> This patch fixes the dangling pointer problem by not setting them until
> everything is okay.

I actually stumbled over this exact bug in trace-cmd but came up with a
much simpler solution:

-- Steve

>From 1d44d0f3361b70724b63c8aab20992704557e9c0 Mon Sep 17 00:00:00 2001
From: "Steven Rostedt (Red Hat)" <rostedt@goodmis.org>
Date: Fri, 17 Jul 2015 12:07:23 -0400
Subject: [PATCH] tools lib traceevent: Set int_array fields to NULL if freeing from error

Had a bug where on error of parsing __print_array() where the
fields are freed after they were allocated, but since they were not
set to NULL, the freeing of the arg also tried to free the already
freed fields causing a double free.

Fix process_hex() while at it.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
---
 tools/lib/traceevent/event-parse.c |    3 +++
 1 file changed, 3 insertions(+)

Index: linux-trace.git/tools/lib/traceevent/event-parse.c
===================================================================
--- linux-trace.git.orig/tools/lib/traceevent/event-parse.c	2015-07-15 14:49:03.275290348 -0400
+++ linux-trace.git/tools/lib/traceevent/event-parse.c	2015-07-17 12:16:54.535359526 -0400
@@ -2559,6 +2559,7 @@ process_hex(struct event_format *event,
 
 free_field:
 	free_arg(arg->hex.field);
+	arg->hex.field = NULL;
 out:
 	*tok = NULL;
 	return EVENT_ERROR;
@@ -2583,8 +2584,10 @@ process_int_array(struct event_format *e
 
 free_size:
 	free_arg(arg->int_array.count);
+	arg->int_array.count = NULL;
 free_field:
 	free_arg(arg->int_array.field);
+	arg->int_array.field = NULL;
 out:
 	*tok = NULL;
 	return EVENT_ERROR;

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2015-07-17 16:22 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2015-07-17  7:58 [PATCH] tools lib traceevent: Fix double free corruption in error processing code Wang Nan
2015-07-17 16:17 ` Steven Rostedt

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®