Skip to contentSkip to conversation

Add test for CVE 2017-10661

PATCH Superseded Martin Doucha 3 participants 10 messages Feb 21, 2020 – Mar 3, 2020

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 Add test for CVE 2017-10661

---

Sorry about the tmp expression in v1, the comma at the end was a typo.

Changes since v1:
- typo fixes
- code style fixes
- move test from cve/ to kernel/syscall/timerfd/
- add test to cve runfile

 runtest/cve                                   |   1 +
 testcases/kernel/syscalls/timerfd/Makefile    |   3 +-
 .../syscalls/timerfd/timerfd_settime02.c      | 110 ++++++++++++++++++
 3 files changed, 113 insertions(+), 1 deletion(-)
 create mode 100644 testcases/kernel/syscalls/timerfd/timerfd_settime02.c
runtest/cve +1 -0
index 57cf66075..36c5ba1d6 100644
@@ -19,6 +19,7 @@ cve-2017-2671 cve-2017-2671
 cve-2017-6951 request_key05
 cve-2017-7308 setsockopt02
 cve-2017-7472 keyctl04
+cve-2017-10661 timerfd_settime02
 cve-2017-12192 keyctl07
 cve-2017-12193 add_key04
 cve-2017-15274 add_key02
testcases/kernel/syscalls/timerfd/Makefile +2 -1
index 3691a8c36..10b3862ea 100644
@@ -20,6 +20,7 @@ top_srcdir		?= ../../../..
 
 include $(top_srcdir)/include/mk/testcases.mk
 
-LDLIBS			+= -lpthread -lrt
+timerfd_settime02:	CFLAGS	+= -pthread
+timerfd_settime02:	LDLIBS	+= -pthread -lrt
 
 include $(top_srcdir)/include/mk/generic_leaf_target.mk
testcases/kernel/syscalls/timerfd/timerfd_settime02.c new file +110 -0

110 lines not on this page · raw message

Signature · 1 line
2.25.0
#1 lore raw

PATCH v2 1/1 Add test for CVE 2017-10661

Hi Martin,

Quoted 1 line
Quoted 5 lines

Changes since v1:
- typo fixes
- code style fixes
- move test from cve/ to kernel/syscall/timerfd/
- add test to cve runfile

Thanks for fixing all Richie's comments.

testcases/kernel/syscalls/timerfd/Makefile top_srcdir ?= ../../../.. · line 24

Quoted 8 lines

+++ b/testcases/kernel/syscalls/timerfd/Makefile

testcases/kernel/syscalls/timerfd/Makefile +2 -1
@@ -20,6 +20,7 @@ top_srcdir		?= ../../../..
 
 include $(top_srcdir)/include/mk/testcases.mk
 
-LDLIBS			+= -lpthread -lrt
+timerfd_settime02:	CFLAGS	+= -pthread
+timerfd_settime02:	LDLIBS	+= -pthread -lrt

Interesting, I thought -pthread wouldn't be needed to be in LDLIBS (enough to have it in CFLAGS), but old distros (CentOS 6 in our travis: gcc 4.4.7, glibc 2.12).

...

Quoted 1 line
testcases/kernel/syscalls/timerfd/timerfd_settime02.c +0 -0

...

testcases/kernel/syscalls/timerfd/timerfd_settime02.c line 38

Quoted 6 lines
testcases/kernel/syscalls/timerfd/timerfd_settime02.c +6 -0
+static void setup(void)
+{
+	int ttype;
+
+	tst_taint_init(TST_TAINT_W | TST_TAINT_D);
+	errno = 0;

I guess reset errno shouldn't be needed, but sure it's safe.

testcases/kernel/syscalls/timerfd/timerfd_settime02.c line 44

Quoted 6 lines
testcases/kernel/syscalls/timerfd/timerfd_settime02.c +6 -0
+	fd = timerfd_create(CLOCK_REALTIME, 0);
+
+	if (fd < 0) {
+		ttype = (errno == ENOTSUP ? TCONF : TBROK);
+		tst_brk(ttype | TERRNO, "Cannot create timer");
+	}

It looks like we'd benefit from SAFE_TIMERFD_CREATE().
+ I like you take care about ENOTSUP.

...

testcases/kernel/syscalls/timerfd/timerfd_settime02.c line 81

Quoted 9 lines
testcases/kernel/syscalls/timerfd/timerfd_settime02.c +9 -0
+static void run(void)
+{
+	tst_fzsync_pair_reset(&fzsync_pair, thread_run);
+
+	while (tst_fzsync_run_a(&fzsync_pair)) {
+		TEST(punch_clock(TFD_TIMER_ABSTIME | TFD_TIMER_CANCEL_ON_SET));
+
+		if (TST_RET == -1)
+			tst_res(TBROK | TTERRNO, TIMERFD_FLAGS " failed");

Now this needs to be tst_res(TFAIL (person who merges this will do).

testcases/kernel/syscalls/timerfd/timerfd_settime02.c line 85

Quoted 4 lines
testcases/kernel/syscalls/timerfd/timerfd_settime02.c +4 -0
+
+		if (TST_RET != 0)
+			tst_res(TBROK | TTERRNO, "Invalid " TIMERFD_FLAGS
+				" return value");

The same here.

Kind regards,
Petr

#2 lore raw

PATCH v2 1/1 Add test for CVE 2017-10661

Quoted 15 lines

Hi Martin,

Quoted 1 line
Quoted 5 lines

Changes since v1:
- typo fixes
- code style fixes
- move test from cve/ to kernel/syscall/timerfd/
- add test to cve runfile

Thanks for fixing all Richie's comments.

Quoted 2 lines

+++ b/testcases/kernel/syscalls/timerfd/Makefile

(quoted diff) +0 -0
@@ -20,6 +20,7 @@ top_srcdir		?= ../../../..
Quoted 8 lines

testcases/kernel/syscalls/timerfd/Makefile top_srcdir ?= ../../../.. · line 24

Quoted 5 lines
testcases/kernel/syscalls/timerfd/Makefile +2 -1
 include $(top_srcdir)/include/mk/testcases.mk
 
-LDLIBS			+= -lpthread -lrt
+timerfd_settime02:	CFLAGS	+= -pthread
+timerfd_settime02:	LDLIBS	+= -pthread -lrt

Interesting, I thought -pthread wouldn't be needed to be in LDLIBS (enough to have it in CFLAGS), but old distros (CentOS 6 in our travis: gcc 4.4.7, glibc 2.12).

OK: it can be just this:
timerfd_settime02: CFLAGS += -pthread
timerfd_settime02: LDLIBS += -lrt

because failure in CentOS 6 is different:
timerfd01.c:36: undefined reference to `clock_gettime'
https://api.travis-ci.org/v3/job/657694167/log.txt

Obviously we need autotools check for it (trivial, I can do it before merge).

Kind regards,
Petr

#3 lore raw

PATCH v2 1/1 Add test for CVE 2017-10661

Quoted 5 lines · Petr Vorel

because failure in CentOS 6 is different:
timerfd01.c:36: undefined reference to `clock_gettime'
https://api.travis-ci.org/v3/job/657694167/log.txt

Obviously we need autotools check for it (trivial, I can do it before merge).

Note that this is a compile error in completely different test program.

Signature · 7 lines
Martin Doucha   [email protected]
QA Engineer for Software Maintenance
SUSE LINUX, s.r.o.
CORSO IIa
Krizikova 148/34
186 00 Prague 8
Czech Republic
#4 lore raw

PATCH v2 1/1 Add test for CVE 2017-10661

In reply to Petr Vorel

Hi!

Quoted 16 lines
Quoted 11 lines

testcases/kernel/syscalls/timerfd/Makefile top_srcdir ?= ../../../.. · line 24

Quoted 8 lines

+++ b/testcases/kernel/syscalls/timerfd/Makefile

testcases/kernel/syscalls/timerfd/Makefile +2 -1
@@ -20,6 +20,7 @@ top_srcdir		?= ../../../..
 
 include $(top_srcdir)/include/mk/testcases.mk
 
-LDLIBS			+= -lpthread -lrt
+timerfd_settime02:	CFLAGS	+= -pthread
+timerfd_settime02:	LDLIBS	+= -pthread -lrt

Interesting, I thought -pthread wouldn't be needed to be in LDLIBS (enough to have it in CFLAGS), but old distros (CentOS 6 in our travis: gcc 4.4.7, glibc 2.12).

OK: it can be just this:
timerfd_settime02: CFLAGS += -pthread
timerfd_settime02: LDLIBS += -lrt

Quoted 5 lines

because failure in CentOS 6 is different:
timerfd01.c:36: undefined reference to `clock_gettime'
https://api.travis-ci.org/v3/job/657694167/log.txt

Obviously we need autotools check for it (trivial, I can do it before merge).

-lrt is required in LDLIBS for older libc that had all the posix realtime outside of the glibc.so, and it's harmless on newer glibc so for the time being we just need -lrt in LDLIBS for quite a lot of testcases...

Signature · 2 lines
Cyril Hrubis
[email protected]
#5 lore raw

PATCH v2 1/1 Add test for CVE 2017-10661

Quoted 8 lines · Cyril Hrubis
Quoted 8 lines

OK: it can be just this:
timerfd_settime02: CFLAGS += -pthread
timerfd_settime02: LDLIBS += -lrt

because failure in CentOS 6 is different:
timerfd01.c:36: undefined reference to `clock_gettime'
https://api.travis-ci.org/v3/job/657694167/log.txt

Quoted 7 lines
Quoted 1 line

Obviously we need autotools check for it (trivial, I can do it before merge).

-lrt is required in LDLIBS for older libc that had all the posix realtime outside of the glibc.so, and it's harmless on newer glibc so for the time being we just need -lrt in LDLIBS for quite a lot of testcases...

Ah, right. I broke the Makefile when I deleted the unconditional LDLIBS line... The -lpthread was useless but I didn't realize the -lrt was needed.

I'll resubmit with all fixes mentioned so far then.

Signature · 7 lines
Martin Doucha   [email protected]
QA Engineer for Software Maintenance
SUSE LINUX, s.r.o.
CORSO IIa
Krizikova 148/34
186 00 Prague 8
Czech Republic
#6 lore raw

PATCH v2 1/1 Add test for CVE 2017-10661

Hi Martin,

Quoted 4 lines

Ah, right. I broke the Makefile when I deleted the unconditional LDLIBS line... The -lpthread was useless but I didn't realize the -lrt was needed.

I'll resubmit with all fixes mentioned so far then.

OK, thanks for doing new version.

Kind regards,
Petr

#7 lore raw

PATCH v2 1/1 Add test for CVE 2017-10661

In reply to Cyril Hrubis

Hi,

Quoted 15 lines
Quoted 14 lines
Quoted 6 lines

testcases/kernel/syscalls/timerfd/Makefile top_srcdir ?= ../../../.. · line 24

Quoted 3 lines
testcases/kernel/syscalls/timerfd/Makefile +2 -1
-LDLIBS			+= -lpthread -lrt
+timerfd_settime02:	CFLAGS	+= -pthread
+timerfd_settime02:	LDLIBS	+= -pthread -lrt

Interesting, I thought -pthread wouldn't be needed to be in LDLIBS (enough to have it in CFLAGS), but old distros (CentOS 6 in our travis: gcc 4.4.7, glibc 2.12).

OK: it can be just this:
timerfd_settime02: CFLAGS += -pthread
timerfd_settime02: LDLIBS += -lrt

because failure in CentOS 6 is different:
timerfd01.c:36: undefined reference to `clock_gettime'
https://api.travis-ci.org/v3/job/657694167/log.txt

Quoted 6 lines
Quoted 1 line

Obviously we need autotools check for it (trivial, I can do it before merge).

-lrt is required in LDLIBS for older libc that had all the posix realtime outside of the glibc.so, and it's harmless on newer glibc so for the time being we just need -lrt in LDLIBS for quite a lot of testcases...

Thanks, that's documented. I was mentioning -pthread.

Kind regards,
Petr

#8 lore raw

PATCH v2 1/1 Add test for CVE 2017-10661

In reply to Cyril Hrubis

Hi,

Quoted 4 lines

-lrt is required in LDLIBS for older libc that had all the posix realtime outside of the glibc.so, and it's harmless on newer glibc so for the time being we just need -lrt in LDLIBS for quite a lot of testcases...

I mean I'd put in the Makefile:

CFLAGS += -pthread
LDLIBS += -lrt

Kind regards,
Petr

#9 lore raw

PATCH v2 1/1 Add test for CVE 2017-10661

In reply to Petr Vorel

(Resending because I forgot to CC the LTP mailing list in the original message to Petr Vorel.)

Quoted 6 lines · Petr Vorel

testcases/kernel/syscalls/timerfd/Makefile top_srcdir ?= ../../../.. · line 24

Quoted 3 lines
testcases/kernel/syscalls/timerfd/Makefile +2 -1
-LDLIBS			+= -lpthread -lrt
+timerfd_settime02:	CFLAGS	+= -pthread
+timerfd_settime02:	LDLIBS	+= -pthread -lrt

Interesting, I thought -pthread wouldn't be needed to be in LDLIBS (enough to have it in CFLAGS), but old distros (CentOS 6 in our travis: gcc 4.4.7, glibc 2.12).

pthread_create() man page say "compile AND link with -pthread". ?\_(?)_/?

Quoted 7 lines

testcases/kernel/syscalls/timerfd/timerfd_settime02.c line 81

Quoted 5 lines
testcases/kernel/syscalls/timerfd/timerfd_settime02.c new file +5 -0
+	while (tst_fzsync_run_a(&fzsync_pair)) {
+		TEST(punch_clock(TFD_TIMER_ABSTIME | TFD_TIMER_CANCEL_ON_SET));
+
+		if (TST_RET == -1)
+			tst_res(TBROK | TTERRNO, TIMERFD_FLAGS " failed");

Now this needs to be tst_res(TFAIL (person who merges this will do).

Quoted 5 lines

testcases/kernel/syscalls/timerfd/timerfd_settime02.c line 85

Quoted 4 lines
testcases/kernel/syscalls/timerfd/timerfd_settime02.c new file +4 -0
+
+		if (TST_RET != 0)
+			tst_res(TBROK | TTERRNO, "Invalid " TIMERFD_FLAGS
+				" return value");

The same here.

No, both should be tst_brk(). There's no point continuing the test if setting the CANCEL_ON_SET flag fails. Sorry about the typo, I need more coffee.

Signature · 7 lines
Martin Doucha   [email protected]
QA Engineer for Software Maintenance
SUSE LINUX, s.r.o.
CORSO IIa
Krizikova 148/34
186 00 Prague 8
Czech Republic