Skip to contentSkip to conversation

Make _FORTIFY_SOURCE defines dependent on the feature

PATCH Applied Ian Rogers 3 participants 6 messages Sep 25, 2019 – Oct 7, 2019 perf events

lore
Share & tools
lore

Ian Rogers

author reviewer
#0 lore raw

PATCH 1/2 Make _FORTIFY_SOURCE defines dependent on the feature

Unconditionally defining _FORTIFY_SOURCE can break tools that don't work with it, such as memory sanitizers:
https://github.com/google/sanitizers/wiki/AddressSanitizer#faq

---
 tools/lib/subcmd/Makefile | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
tools/lib/subcmd/Makefile +7 -1
index ed61fb3a46c0..5b2cd5e58df0 100644
@@ -20,7 +20,13 @@ MAKEFLAGS += --no-print-directory
 LIBFILE = $(OUTPUT)libsubcmd.a
 
 CFLAGS := $(EXTRA_WARNINGS) $(EXTRA_CFLAGS)
-CFLAGS += -ggdb3 -Wall -Wextra -std=gnu99 -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=2 -fPIC
+CFLAGS += -ggdb3 -Wall -Wextra -std=gnu99 -fPIC
+
+ifeq ($(DEBUG),0)
+  ifeq ($(feature-fortify-source), 1)
+    CFLAGS += -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=2
+  endif
+endif
 
 ifeq ($(CC_NO_CLANG), 0)
   CFLAGS += -O3
Signature · 1 line
2.23.0.351.gc4317032e6-goog
author reviewer
#1 lore raw

PATCH 2/2 Avoid raising segv using an obvious null dereference

An optimized build such as:
make -C tools/perf CLANG=1 CC=clang EXTRA_CFLAGS="-O3
will turn the dereference operation into a ud2 instruction, raising a SIGILL rather than a SIGSEGV. Use raise(..) for correctness and clarity.

Similar issues were addressed in Numfor Mbiziwo-Tiapo's patch:
https://lkml.org/lkml/2019/7/8/1234

---
 tools/perf/tests/perf-hooks.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
tools/perf/tests/perf-hooks.c +1 -2
index dbc27199c65e..dd865e0bea12 100644
@@ -19,12 +19,11 @@ static void sigsegv_handler(int sig __maybe_unused)
 static void the_hook(void *_hook_flags)
 {
 	int *hook_flags = _hook_flags;
-	int *p = NULL;
 
 	*hook_flags = 1234;
 
 	/* Generate a segfault, test perf_hooks__recover */
-	*p = 0;
+	raise(SIGSEGV);
 }
 
 int test__perf_hooks(struct test *test __maybe_unused, int subtest __maybe_unused)
Signature · 1 line
2.23.0.351.gc4317032e6-goog
#2 lore raw

Re: [PATCH 2/2] Avoid raising segv using an obvious null dereference

Em Wed, Sep 25, 2019 at 12:59:24PM -0700, Ian Rogers escreveu:

Quoted 7 lines

An optimized build such as:
make -C tools/perf CLANG=1 CC=clang EXTRA_CFLAGS="-O3
will turn the dereference operation into a ud2 instruction, raising a SIGILL rather than a SIGSEGV. Use raise(..) for correctness and clarity.

Similar issues were addressed in Numfor Mbiziwo-Tiapo's patch:
https://lkml.org/lkml/2019/7/8/1234

Added:

Cc Numfor Mbiziwo-Tiapo <[email protected]>

And:

Cc Wang Nan <[email protected]>
Fixes a074865e60ed ("perf tools: Introduce perf hooks")

And:

Committer testing:

Before:

[root@quaco ~]# perf test hooks
55: perf hooks : Ok
[root@quaco ~]# perf test -v hooks
55: perf hooks :
--- start ---
test child forked, pid 17092
SIGSEGV is observed as expected, try to recover. Fatal error (SEGFAULT) in perf hook 'test' test child finished with 0
---- end ----
perf hooks: Ok
[root@quaco ~]#

After:

[root@quaco ~]# perf test hooks
55: perf hooks : Ok
[root@quaco ~]# perf test -v hooks
55: perf hooks :
--- start ---
test child forked, pid 17909
SIGSEGV is observed as expected, try to recover. Fatal error (SEGFAULT) in perf hook 'test' test child finished with 0
---- end ----
perf hooks: Ok
[root@quaco ~]#

----------------------------------------------------------------------------

Thanks, applied.

- Arnaldo

Quoted 25 lines
---
 tools/perf/tests/perf-hooks.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
tools/perf/tests/perf-hooks.c +1 -2
index dbc27199c65e..dd865e0bea12 100644
@@ -19,12 +19,11 @@ static void sigsegv_handler(int sig __maybe_unused)
 static void the_hook(void *_hook_flags)
 {
 	int *hook_flags = _hook_flags;
-	int *p = NULL;
 
 	*hook_flags = 1234;
 
 	/* Generate a segfault, test perf_hooks__recover */
-	*p = 0;
+	raise(SIGSEGV);
 }
 
 int test__perf_hooks(struct test *test __maybe_unused, int subtest __maybe_unused)
Signature · 1 line
2.23.0.351.gc4317032e6-goog
Signature · 1 line
- Arnaldo
merge bot Applied
#3 lore raw

[tip: perf/urgent] perf tests: Avoid raising SEGV using an obvious NULL dereference

In reply to Ian Rogers
The following commit has been merged into the perf/urgent branch of tip:

The following commit has been merged into the perf/urgent branch of tip:

Commit-ID:     e3e2cf3d5b1fe800b032e14c0fdcd9a6fb20cf3b
Gitweb:        https://git.kernel.org/tip/e3e2cf3d5b1fe800b032e14c0fdcd9a6fb20cf3b
Author:        Ian Rogers <[email protected]>
AuthorDate:    Wed, 25 Sep 2019 12:59:24 -07:00
Committer:     Arnaldo Carvalho de Melo <[email protected]>
CommitterDate: Fri, 27 Sep 2019 09:26:14 -03:00

perf tests: Avoid raising SEGV using an obvious NULL dereference

An optimized build such as:

make -C tools/perf CLANG=1 CC=clang EXTRA_CFLAGS="-O3

will turn the dereference operation into a ud2 instruction, raising a SIGILL rather than a SIGSEGV. Use raise(..) for correctness and clarity.

Similar issues were addressed in Numfor Mbiziwo-Tiapo's patch:

https://lkml.org/lkml/2019/7/8/1234

Committer testing:

Before:

[root@quaco ~]# perf test hooks
55: perf hooks : Ok
[root@quaco ~]# perf test -v hooks
55: perf hooks :
--- start ---
test child forked, pid 17092
SIGSEGV is observed as expected, try to recover. Fatal error (SEGFAULT) in perf hook 'test' test child finished with 0
---- end ----
perf hooks: Ok
[root@quaco ~]#

After:

[root@quaco ~]# perf test hooks
55: perf hooks : Ok
[root@quaco ~]# perf test -v hooks
55: perf hooks :
--- start ---
test child forked, pid 17909
SIGSEGV is observed as expected, try to recover. Fatal error (SEGFAULT) in perf hook 'test' test child finished with 0
---- end ----
perf hooks: Ok
[root@quaco ~]#

Fixes a074865e60ed ("perf tools: Introduce perf hooks")
Cc Alexander Shishkin <[email protected]>
Cc Andi Kleen <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
Cc Wang Nan <[email protected]>
---
 tools/perf/tests/perf-hooks.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
tools/perf/tests/perf-hooks.c +1 -2
index dbc2719..dd865e0 100644
@@ -19,12 +19,11 @@ static void sigsegv_handler(int sig __maybe_unused)
 static void the_hook(void *_hook_flags)
 {
 	int *hook_flags = _hook_flags;
-	int *p = NULL;
 
 	*hook_flags = 1234;
 
 	/* Generate a segfault, test perf_hooks__recover */
-	*p = 0;
+	raise(SIGSEGV);
 }
 
 int test__perf_hooks(struct test *test __maybe_unused, int subtest __maybe_unused)
#4 lore raw
In reply to Ian Rogers

Em Wed, Sep 25, 2019 at 12:59:23PM -0700, Ian Rogers escreveu:

Quoted 3 lines

Unconditionally defining _FORTIFY_SOURCE can break tools that don't work with it, such as memory sanitizers:
https://github.com/google/sanitizers/wiki/AddressSanitizer#faq

Thanks, and added this:

Cc Josh Poimboeuf <[email protected]>
Fixes 4b6ab94eabe4 ("perf subcmd: Create subcmd library")
Quoted 26 lines
---
 tools/lib/subcmd/Makefile | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
tools/lib/subcmd/Makefile +7 -1
index ed61fb3a46c0..5b2cd5e58df0 100644
@@ -20,7 +20,13 @@ MAKEFLAGS += --no-print-directory
 LIBFILE = $(OUTPUT)libsubcmd.a
 
 CFLAGS := $(EXTRA_WARNINGS) $(EXTRA_CFLAGS)
-CFLAGS += -ggdb3 -Wall -Wextra -std=gnu99 -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=2 -fPIC
+CFLAGS += -ggdb3 -Wall -Wextra -std=gnu99 -fPIC
+
+ifeq ($(DEBUG),0)
+  ifeq ($(feature-fortify-source), 1)
+    CFLAGS += -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=2
+  endif
+endif
 
 ifeq ($(CC_NO_CLANG), 0)
   CFLAGS += -O3
Signature · 1 line
2.23.0.351.gc4317032e6-goog
Signature · 1 line
- Arnaldo
merge bot Applied
#5 lore raw

[tip: perf/urgent] libsubcmd: Make _FORTIFY_SOURCE defines dependent on the feature

In reply to Ian Rogers
The following commit has been merged into the perf/urgent branch of tip:

The following commit has been merged into the perf/urgent branch of tip:

Commit-ID:     4b0b2b096da9d296e0e5668cdfba8613bd6f5bc8
Gitweb:        https://git.kernel.org/tip/4b0b2b096da9d296e0e5668cdfba8613bd6f5bc8
Author:        Ian Rogers <[email protected]>
AuthorDate:    Wed, 25 Sep 2019 12:59:23 -07:00
Committer:     Arnaldo Carvalho de Melo <[email protected]>
CommitterDate: Fri, 27 Sep 2019 09:26:14 -03:00

libsubcmd: Make _FORTIFY_SOURCE defines dependent on the feature

Unconditionally defining _FORTIFY_SOURCE can break tools that don't work with it, such as memory sanitizers:

https://github.com/google/sanitizers/wiki/AddressSanitizer#faq

Fixes 4b6ab94eabe4 ("perf subcmd: Create subcmd library")
Cc Alexander Shishkin <[email protected]>
Cc Andi Kleen <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Josh Poimboeuf <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
---
 tools/lib/subcmd/Makefile | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
tools/lib/subcmd/Makefile +7 -1
index ed61fb3..5b2cd5e 100644
@@ -20,7 +20,13 @@ MAKEFLAGS += --no-print-directory
 LIBFILE = $(OUTPUT)libsubcmd.a
 
 CFLAGS := $(EXTRA_WARNINGS) $(EXTRA_CFLAGS)
-CFLAGS += -ggdb3 -Wall -Wextra -std=gnu99 -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=2 -fPIC
+CFLAGS += -ggdb3 -Wall -Wextra -std=gnu99 -fPIC
+
+ifeq ($(DEBUG),0)
+  ifeq ($(feature-fortify-source), 1)
+    CFLAGS += -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=2
+  endif
+endif
 
 ifeq ($(CC_NO_CLANG), 0)
   CFLAGS += -O3