Skip to contentSkip to conversation

Add setsockopt08, CVE-2021-22555

PATCH New Richard Palethorpe 4 participants 18 messages Aug 3, 2021 – Aug 6, 2021

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Add setsockopt08, CVE-2021-22555

This is a copy and paste of Nicolai's reproducer. The main difference is that I moved some code around. Of course I also used LTP library features, but essentially it works the same.

There are some hard coded values which I do not like. I guess these could be calculated or varied somehow. However I struggle to understand what the kernel is doing. This perhaps needs more investigation. We could try generalising this test and setsockopt03.

---

runtest/cve | 1 +

 runtest/syscalls                              |   1 +
 .../kernel/syscalls/setsockopt/.gitignore     |   1 +
 .../kernel/syscalls/setsockopt/setsockopt08.c | 151 ++++++++++++++++++
 4 files changed, 154 insertions(+)
 create mode 100644 testcases/kernel/syscalls/setsockopt/setsockopt08.c
runtest/cve +1 -0
index 8aa048a40..60bd22f01 100644
@@ -65,4 +65,5 @@ cve-2020-14416 pty03
 cve-2020-25705 icmp_rate_limit01
 cve-2020-29373 io_uring02
 cve-2021-3444 bpf_prog05
+cve-2021-22555 setsockopt08 -i 100
 cve-2021-26708 vsock01
runtest/syscalls +1 -0
index b379b2d90..2222990fe 100644
@@ -1378,6 +1378,7 @@ setsockopt04 setsockopt04
 setsockopt05 setsockopt05
 setsockopt06 setsockopt06
 setsockopt07 setsockopt07
+setsockopt08 setsockopt08
 
 settimeofday01 settimeofday01
 settimeofday02 settimeofday02
testcases/kernel/syscalls/setsockopt/.gitignore +1 -0
index 1ca5b836b..95a5e43f8 100644
@@ -5,3 +5,4 @@
 /setsockopt05
 /setsockopt06
 /setsockopt07
+/setsockopt08
testcases/kernel/syscalls/setsockopt/setsockopt08.c new file +151 -0

151 lines not on this page · raw message

Signature · 1 line
2.31.1
#1 lore raw

PATCH v2 1/2 Add lapi/ip_tables.h and use it in setsockopt03

---

Oops forgot to add the file below.

V2:
* Add ip_tables.h compatability header

 include/lapi/ip_tables.h                      | 46 +++++++++++++++++++
 .../kernel/syscalls/setsockopt/setsockopt03.c | 39 +---------------
 2 files changed, 48 insertions(+), 37 deletions(-)
 create mode 100644 include/lapi/ip_tables.h
include/lapi/ip_tables.h new file +46 -0
new file mode 100644
index 000000000..e91238ffd
@@ -0,0 +1,46 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#ifndef LAPI_IP_TABLES__
+#define LAPI_IP_TABLES__
+
+#include "config.h"
+
+#include <linux/netfilter_ipv4/ip_tables.h>
+
+#ifndef HAVE_STRUCT_XT_ENTRY_MATCH
+struct xt_entry_match {
+	union {
+		struct {
+			uint16_t match_size;
+			char name[29];
+			uint8_t revision;
+		} user;
+		struct {
+			uint16_t match_size;
+			void *match;
+		} kernel;
+		uint16_t match_size;
+	} u;
+	unsigned char data[0];
+};
+#endif
+
+#ifndef HAVE_STRUCT_XT_ENTRY_TARGET
+struct xt_entry_target {
+	union {
+		struct {
+			uint16_t target_size;
+			char name[29];
+			uint8_t revision;
+		} user;
+		struct {
+			uint16_t target_size;
+			void *target;
+		} kernel;
+		uint16_t target_size;
+	} u;
+	unsigned char data[0];
+};
+#endif
+
+#endif
testcases/kernel/syscalls/setsockopt/setsockopt03.c +2 -37
index 3d49628d6..1434475db 100644
@@ -21,12 +21,13 @@
 #include <netinet/in.h>
 #include <net/if.h>
 #include <limits.h>
-#include <linux/netfilter_ipv4/ip_tables.h>
 
 #include "tst_test.h"
 #include "tst_safe_net.h"
 #include "tst_kernel.h"
 
+#include "lapi/ip_tables.h"
+
 #define TOO_SMALL_OFFSET 74
 #define OFFSET_OVERWRITE 0xFFFF
 #define NEXT_OFFSET (sizeof(struct ipt_entry)		\
@@ -34,42 +35,6 @@
 		     + sizeof(struct xt_entry_target))
 #define PADDING (OFFSET_OVERWRITE - NEXT_OFFSET)
 
-#ifndef HAVE_STRUCT_XT_ENTRY_MATCH
-struct xt_entry_match {
-	union {
-		struct {
-			uint16_t match_size;
-			char name[29];
-			uint8_t revision;
-		} user;
-		struct {
-			uint16_t match_size;
-			void *match;
-		} kernel;
-		uint16_t match_size;
-	} u;
-	unsigned char data[0];
-};
-#endif
-
-#ifndef HAVE_STRUCT_XT_ENTRY_TARGET
-struct xt_entry_target {
-	union {
-		struct {
-			uint16_t target_size;
-			char name[29];
-			uint8_t revision;
-		} user;
-		struct {
-			uint16_t target_size;
-			void *target;
-		} kernel;
-		uint16_t target_size;
-	} u;
-	unsigned char data[0];
-};
-#endif
-
 struct payload {
 	struct ipt_replace repl;
 	struct ipt_entry ent;
Signature · 1 line
2.31.1
#2 lore raw

PATCH v2 2/2 Add setsockopt08, CVE-2021-22555

This is a copy and paste of Nicolai's reproducer. The main difference is that I moved some code around. Of course I also used LTP library features, but essentially it works the same.

There are some hard coded values which I do not like. I guess these could be calculated or varied somehow. However I struggle to understand what the kernel is doing. This perhaps needs more investigation. We could try generalising this test and setsockopt03.

---
 runtest/cve                                   |   1 +
 runtest/syscalls                              |   1 +
 .../kernel/syscalls/setsockopt/.gitignore     |   1 +
 .../kernel/syscalls/setsockopt/setsockopt08.c | 151 ++++++++++++++++++
 4 files changed, 154 insertions(+)
 create mode 100644 testcases/kernel/syscalls/setsockopt/setsockopt08.c
runtest/cve +1 -0
index 8aa048a40..60bd22f01 100644
@@ -65,4 +65,5 @@ cve-2020-14416 pty03
 cve-2020-25705 icmp_rate_limit01
 cve-2020-29373 io_uring02
 cve-2021-3444 bpf_prog05
+cve-2021-22555 setsockopt08 -i 100
 cve-2021-26708 vsock01
runtest/syscalls +1 -0
index b379b2d90..2222990fe 100644
@@ -1378,6 +1378,7 @@ setsockopt04 setsockopt04
 setsockopt05 setsockopt05
 setsockopt06 setsockopt06
 setsockopt07 setsockopt07
+setsockopt08 setsockopt08
 
 settimeofday01 settimeofday01
 settimeofday02 settimeofday02
testcases/kernel/syscalls/setsockopt/.gitignore +1 -0
index 1ca5b836b..95a5e43f8 100644
@@ -5,3 +5,4 @@
 /setsockopt05
 /setsockopt06
 /setsockopt07
+/setsockopt08
testcases/kernel/syscalls/setsockopt/setsockopt08.c new file +151 -0

151 lines not on this page · raw message

Signature · 1 line
2.31.1
#3 lore raw

PATCH v2 2/2 Add setsockopt08, CVE-2021-22555

Hi,
since this vulnerability is in the compat syscall wrappers, you should also copy the arch bits check from setsockopt03 setup().

Quoted 8 lines · Richard Palethorpe via ltp

This is a copy and paste of Nicolai's reproducer. The main difference is that I moved some code around. Of course I also used LTP library features, but essentially it works the same.

There are some hard coded values which I do not like. I guess these could be calculated or varied somehow. However I struggle to understand what the kernel is doing. This perhaps needs more investigation. We could try generalising this test and setsockopt03

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 2/2 Add setsockopt08, CVE-2021-22555

Hi,

Quoted 3 lines · Martin Doucha

Hi,
since this vulnerability is in the compat syscall wrappers, you should also copy the arch bits check from setsockopt03 setup().

+1

Will roll another patch after a delay.

Quoted 10 lines · Richard Palethorpe via ltp
Quoted 8 lines

This is a copy and paste of Nicolai's reproducer. The main difference is that I moved some code around. Of course I also used LTP library features, but essentially it works the same.

There are some hard coded values which I do not like. I guess these could be calculated or varied somehow. However I struggle to understand what the kernel is doing. This perhaps needs more investigation. We could try generalising this test and setsockopt03

Signature · 2 lines
Thank you,
Richard.
#5 lore raw

PATCH v3 1/3 API: TST_EXP_FAIL: Allow passing when errno is not checked

Judging by the existing usage of TST_EXP_FAIL(..., 0, ...) in finit_module02. We want to pass if errno == 0 otherwise the test will not return a result.

This is also less surprising than giving errno == 0 a dual meaning.

This also changes the trailing '\' indentation to tabs. However this is correct and the rest of the file is wrong.

---

V3:
* Add fix for TST_EXP_FAIL which prevented the test from passing on a non buggy system.
* TCONF but continue on non 32-bit compat mode
* Add Fixes trailer

V2:
* Add mising lapi header

 include/tst_test_macros.h | 20 +++++++++-----------
 1 file changed, 9 insertions(+), 11 deletions(-)
include/tst_test_macros.h +9 -11
index 41886fbbc..4c1df58ff 100644
@@ -137,17 +137,15 @@ extern void *TST_RET_PTR;
 			break;                                                 \
 		}                                                              \
 		                                                               \
-		if (ERRNO) {                                                   \
-			if (TST_ERR == ERRNO) {                                \
-				TST_MSG_(TPASS | TTERRNO, " ",                 \
-				         #SCALL, ##__VA_ARGS__);               \
-				TST_PASS = 1;                                  \
-			} else {                                               \
-				TST_MSGP_(TFAIL | TTERRNO, " expected %s",     \
-				          tst_strerrno(ERRNO),                 \
-				          #SCALL, ##__VA_ARGS__);              \
-			}                                                      \
-		}                                                              \
+		if (!ERRNO || TST_ERR == ERRNO) {			\
+			TST_MSG_(TPASS | TTERRNO, " ",			\
+				 #SCALL, ##__VA_ARGS__);		\
+			TST_PASS = 1;					\
+		} else {						\
+			TST_MSGP_(TFAIL | TTERRNO, " expected %s",	\
+				  tst_strerrno(ERRNO),			\
+				  #SCALL, ##__VA_ARGS__);		\
+		}							\
 	} while (0)
 
 #define TST_EXP_FAIL(SCALL, ERRNO, ...) TST_EXP_FAIL_(TST_RET == 0, SCALL, ERRNO, __VA_ARGS__)
Signature · 1 line
2.31.1
#6 lore raw

PATCH v3 2/3 API: Add lapi/ip_tables.h and use it in setsockopt03

---
 include/lapi/ip_tables.h                      | 46 +++++++++++++++++++
 .../kernel/syscalls/setsockopt/setsockopt03.c | 39 +---------------
 2 files changed, 48 insertions(+), 37 deletions(-)
 create mode 100644 include/lapi/ip_tables.h
include/lapi/ip_tables.h new file +46 -0
new file mode 100644
index 000000000..e91238ffd
@@ -0,0 +1,46 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#ifndef LAPI_IP_TABLES__
+#define LAPI_IP_TABLES__
+
+#include "config.h"
+
+#include <linux/netfilter_ipv4/ip_tables.h>
+
+#ifndef HAVE_STRUCT_XT_ENTRY_MATCH
+struct xt_entry_match {
+	union {
+		struct {
+			uint16_t match_size;
+			char name[29];
+			uint8_t revision;
+		} user;
+		struct {
+			uint16_t match_size;
+			void *match;
+		} kernel;
+		uint16_t match_size;
+	} u;
+	unsigned char data[0];
+};
+#endif
+
+#ifndef HAVE_STRUCT_XT_ENTRY_TARGET
+struct xt_entry_target {
+	union {
+		struct {
+			uint16_t target_size;
+			char name[29];
+			uint8_t revision;
+		} user;
+		struct {
+			uint16_t target_size;
+			void *target;
+		} kernel;
+		uint16_t target_size;
+	} u;
+	unsigned char data[0];
+};
+#endif
+
+#endif
testcases/kernel/syscalls/setsockopt/setsockopt03.c +2 -37
index 3d49628d6..1434475db 100644
@@ -21,12 +21,13 @@
 #include <netinet/in.h>
 #include <net/if.h>
 #include <limits.h>
-#include <linux/netfilter_ipv4/ip_tables.h>
 
 #include "tst_test.h"
 #include "tst_safe_net.h"
 #include "tst_kernel.h"
 
+#include "lapi/ip_tables.h"
+
 #define TOO_SMALL_OFFSET 74
 #define OFFSET_OVERWRITE 0xFFFF
 #define NEXT_OFFSET (sizeof(struct ipt_entry)		\
@@ -34,42 +35,6 @@
 		     + sizeof(struct xt_entry_target))
 #define PADDING (OFFSET_OVERWRITE - NEXT_OFFSET)
 
-#ifndef HAVE_STRUCT_XT_ENTRY_MATCH
-struct xt_entry_match {
-	union {
-		struct {
-			uint16_t match_size;
-			char name[29];
-			uint8_t revision;
-		} user;
-		struct {
-			uint16_t match_size;
-			void *match;
-		} kernel;
-		uint16_t match_size;
-	} u;
-	unsigned char data[0];
-};
-#endif
-
-#ifndef HAVE_STRUCT_XT_ENTRY_TARGET
-struct xt_entry_target {
-	union {
-		struct {
-			uint16_t target_size;
-			char name[29];
-			uint8_t revision;
-		} user;
-		struct {
-			uint16_t target_size;
-			void *target;
-		} kernel;
-		uint16_t target_size;
-	} u;
-	unsigned char data[0];
-};
-#endif
-
 struct payload {
 	struct ipt_replace repl;
 	struct ipt_entry ent;
Signature · 1 line
2.31.1
#7 lore raw

PATCH v3 3/3 Add setsockopt08, CVE-2021-22555

In reply to Richard Palethorpe

This is a copy and paste of Nicolai's reproducer. The main difference is that I moved some code around. Of course I also used LTP library features, but essentially it works the same.

There are some hard coded values which I do not like. I guess these could be calculated or varied somehow. However I struggle to understand what the kernel is doing. This perhaps needs more investigation. We could try generalising this test and setsockopt03.

---
 runtest/cve                                   |   1 +
 runtest/syscalls                              |   1 +
 .../kernel/syscalls/setsockopt/.gitignore     |   1 +
 .../kernel/syscalls/setsockopt/setsockopt08.c | 156 ++++++++++++++++++
 4 files changed, 159 insertions(+)
 create mode 100644 testcases/kernel/syscalls/setsockopt/setsockopt08.c
runtest/cve +1 -0
index 8aa048a40..60bd22f01 100644
@@ -65,4 +65,5 @@ cve-2020-14416 pty03
 cve-2020-25705 icmp_rate_limit01
 cve-2020-29373 io_uring02
 cve-2021-3444 bpf_prog05
+cve-2021-22555 setsockopt08 -i 100
 cve-2021-26708 vsock01
runtest/syscalls +1 -0
index b379b2d90..2222990fe 100644
@@ -1378,6 +1378,7 @@ setsockopt04 setsockopt04
 setsockopt05 setsockopt05
 setsockopt06 setsockopt06
 setsockopt07 setsockopt07
+setsockopt08 setsockopt08
 
 settimeofday01 settimeofday01
 settimeofday02 settimeofday02
testcases/kernel/syscalls/setsockopt/.gitignore +1 -0
index 1ca5b836b..95a5e43f8 100644
@@ -5,3 +5,4 @@
 /setsockopt05
 /setsockopt06
 /setsockopt07
+/setsockopt08
testcases/kernel/syscalls/setsockopt/setsockopt08.c new file +156 -0

156 lines not on this page · raw message

Signature · 1 line
2.31.1
#8 lore raw

PATCH v3 1/3 API: TST_EXP_FAIL: Allow passing when errno is not checked

In reply to Richard Palethorpe

Hi!

Quoted 3 lines

Judging by the existing usage of TST_EXP_FAIL(..., 0, ...) in finit_module02. We want to pass if errno == 0 otherwise the test will not return a result.

This is actually not true, we do not pass 0 to TST_EXP_FAIL() in finit_module() at all. The places that are not initialized in the structure are initialized at runtime based on the kernel version. I do not think that we even pass 0 to TST_EXP_FAIL*() as an errno anywhere, but I haven't really checked all the callers.

Quoted 2 lines

This is also less surprising than giving errno == 0 a dual meaning.

But I do agree that the current if (ERRNO) branch is confusing. I would be for dropping the if (ERRNO) and checking the TST_ERR against ERRNO unconditionally.

Also note that the TEST() macro clears errno, so if a syscall fails but does not report any error TST_ERR will end up 0 either way so there is no need for having special handling for 0.

Quoted 55 lines

This also changes the trailing '\' indentation to tabs. However this is correct and the rest of the file is wrong.

---

V3:
* Add fix for TST_EXP_FAIL which prevented the test from passing on a non buggy system.
* TCONF but continue on non 32-bit compat mode
* Add Fixes trailer

V2:
* Add mising lapi header

 include/tst_test_macros.h | 20 +++++++++-----------
 1 file changed, 9 insertions(+), 11 deletions(-)
include/tst_test_macros.h +9 -11
index 41886fbbc..4c1df58ff 100644
@@ -137,17 +137,15 @@ extern void *TST_RET_PTR;
 			break;                                                 \
 		}                                                              \
 		                                                               \
-		if (ERRNO) {                                                   \
-			if (TST_ERR == ERRNO) {                                \
-				TST_MSG_(TPASS | TTERRNO, " ",                 \
-				         #SCALL, ##__VA_ARGS__);               \
-				TST_PASS = 1;                                  \
-			} else {                                               \
-				TST_MSGP_(TFAIL | TTERRNO, " expected %s",     \
-				          tst_strerrno(ERRNO),                 \
-				          #SCALL, ##__VA_ARGS__);              \
-			}                                                      \
-		}                                                              \
+		if (!ERRNO || TST_ERR == ERRNO) {			\
+			TST_MSG_(TPASS | TTERRNO, " ",			\
+				 #SCALL, ##__VA_ARGS__);		\
+			TST_PASS = 1;					\
+		} else {						\
+			TST_MSGP_(TFAIL | TTERRNO, " expected %s",	\
+				  tst_strerrno(ERRNO),			\
+				  #SCALL, ##__VA_ARGS__);		\
+		}							\
 	} while (0)
 
 #define TST_EXP_FAIL(SCALL, ERRNO, ...) TST_EXP_FAIL_(TST_RET == 0, SCALL, ERRNO, __VA_ARGS__)
Signature · 5 lines
2.31.1


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
Signature · 2 lines
Cyril Hrubis
[email protected]
#9 lore raw

PATCH v3 1/3 API: TST_EXP_FAIL: Allow passing when errno is not checked

Hello,

Quoted 14 lines · Cyril Hrubis

Hi!

Quoted 3 lines

Judging by the existing usage of TST_EXP_FAIL(..., 0, ...) in finit_module02. We want to pass if errno == 0 otherwise the test will not return a result.

This is actually not true, we do not pass 0 to TST_EXP_FAIL() in finit_module() at all. The places that are not initialized in the structure are initialized at runtime based on the kernel version. I do not think that we even pass 0 to TST_EXP_FAIL*() as an errno anywhere, but I haven't really checked all the callers.

Quoted 2 lines

This is also less surprising than giving errno == 0 a dual meaning.

Quoted 7 lines

But I do agree that the current if (ERRNO) branch is confusing. I would be for dropping the if (ERRNO) and checking the TST_ERR against ERRNO unconditionally.

Also note that the TEST() macro clears errno, so if a syscall fails but does not report any error TST_ERR will end up 0 either way so there is no need for having special handling for 0.

There is if the errno is set, but is undefined. Like if the resulting errno is platform or config dependent.

In the present case though we can just check for EINVAL. That is what the setsockopt man page indicates. Initially I wasn't sure if EFAULT were equally valid, but it seems not.

I suppose we can fall back to using TEST() if the other case arises.

Signature · 2 lines
Thank you,
Richard.
#10 lore raw

PATCH v4 1/3 API: TST_EXP_FAIL: Allow passing when errno is not set

TEST() sets errno to zero. So if something does not set errno on failure it will be zero. It is normal for functions not to use errno, some standard libraries are the exception.

The original code will not allow for a pass result if errno is not set. So we can just remove the branch to allow that. It's not clear what the original intention was.

This also changes the trailing '\' indentation to tabs. However this is correct and the rest of the file is wrong.

---

V4:
* Check for EINVAL.
* Unconditionally check TST_ERR == ERRNO.
(redundant/irrelevant due to the above change, but we still should do it IMO)

V3:
* Add fix for TST_EXP_FAIL which prevented the test from passing on a non buggy system.
* TCONF but continue on non 32-bit compat mode
* Add Fixes trailer

V2:
* Add mising lapi header

 include/tst_test_macros.h | 20 +++++++++-----------
 1 file changed, 9 insertions(+), 11 deletions(-)
include/tst_test_macros.h +9 -11
index 41886fbbc..cd65d8d0d 100644
@@ -137,17 +137,15 @@ extern void *TST_RET_PTR;
 			break;                                                 \
 		}                                                              \
 		                                                               \
-		if (ERRNO) {                                                   \
-			if (TST_ERR == ERRNO) {                                \
-				TST_MSG_(TPASS | TTERRNO, " ",                 \
-				         #SCALL, ##__VA_ARGS__);               \
-				TST_PASS = 1;                                  \
-			} else {                                               \
-				TST_MSGP_(TFAIL | TTERRNO, " expected %s",     \
-				          tst_strerrno(ERRNO),                 \
-				          #SCALL, ##__VA_ARGS__);              \
-			}                                                      \
-		}                                                              \
+		if (TST_ERR == ERRNO) {					\
+			TST_MSG_(TPASS | TTERRNO, " ",			\
+				 #SCALL, ##__VA_ARGS__);		\
+			TST_PASS = 1;					\
+		} else {						\
+			TST_MSGP_(TFAIL | TTERRNO, " expected %s",	\
+				  tst_strerrno(ERRNO),			\
+				  #SCALL, ##__VA_ARGS__);		\
+		}							\
 	} while (0)
 
 #define TST_EXP_FAIL(SCALL, ERRNO, ...) TST_EXP_FAIL_(TST_RET == 0, SCALL, ERRNO, __VA_ARGS__)
Signature · 1 line
2.31.1
#11 lore raw

PATCH v4 2/3 API: Add lapi/ip_tables.h and use it in setsockopt03

---
 include/lapi/ip_tables.h                      | 46 +++++++++++++++++++
 .../kernel/syscalls/setsockopt/setsockopt03.c | 39 +---------------
 2 files changed, 48 insertions(+), 37 deletions(-)
 create mode 100644 include/lapi/ip_tables.h
include/lapi/ip_tables.h new file +46 -0
new file mode 100644
index 000000000..e91238ffd
@@ -0,0 +1,46 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#ifndef LAPI_IP_TABLES__
+#define LAPI_IP_TABLES__
+
+#include "config.h"
+
+#include <linux/netfilter_ipv4/ip_tables.h>
+
+#ifndef HAVE_STRUCT_XT_ENTRY_MATCH
+struct xt_entry_match {
+	union {
+		struct {
+			uint16_t match_size;
+			char name[29];
+			uint8_t revision;
+		} user;
+		struct {
+			uint16_t match_size;
+			void *match;
+		} kernel;
+		uint16_t match_size;
+	} u;
+	unsigned char data[0];
+};
+#endif
+
+#ifndef HAVE_STRUCT_XT_ENTRY_TARGET
+struct xt_entry_target {
+	union {
+		struct {
+			uint16_t target_size;
+			char name[29];
+			uint8_t revision;
+		} user;
+		struct {
+			uint16_t target_size;
+			void *target;
+		} kernel;
+		uint16_t target_size;
+	} u;
+	unsigned char data[0];
+};
+#endif
+
+#endif
testcases/kernel/syscalls/setsockopt/setsockopt03.c +2 -37
index 3d49628d6..1434475db 100644
@@ -21,12 +21,13 @@
 #include <netinet/in.h>
 #include <net/if.h>
 #include <limits.h>
-#include <linux/netfilter_ipv4/ip_tables.h>
 
 #include "tst_test.h"
 #include "tst_safe_net.h"
 #include "tst_kernel.h"
 
+#include "lapi/ip_tables.h"
+
 #define TOO_SMALL_OFFSET 74
 #define OFFSET_OVERWRITE 0xFFFF
 #define NEXT_OFFSET (sizeof(struct ipt_entry)		\
@@ -34,42 +35,6 @@
 		     + sizeof(struct xt_entry_target))
 #define PADDING (OFFSET_OVERWRITE - NEXT_OFFSET)
 
-#ifndef HAVE_STRUCT_XT_ENTRY_MATCH
-struct xt_entry_match {
-	union {
-		struct {
-			uint16_t match_size;
-			char name[29];
-			uint8_t revision;
-		} user;
-		struct {
-			uint16_t match_size;
-			void *match;
-		} kernel;
-		uint16_t match_size;
-	} u;
-	unsigned char data[0];
-};
-#endif
-
-#ifndef HAVE_STRUCT_XT_ENTRY_TARGET
-struct xt_entry_target {
-	union {
-		struct {
-			uint16_t target_size;
-			char name[29];
-			uint8_t revision;
-		} user;
-		struct {
-			uint16_t target_size;
-			void *target;
-		} kernel;
-		uint16_t target_size;
-	} u;
-	unsigned char data[0];
-};
-#endif
-
 struct payload {
 	struct ipt_replace repl;
 	struct ipt_entry ent;
Signature · 1 line
2.31.1
#12 lore raw

PATCH v4 3/3 Add setsockopt08, CVE-2021-22555

In reply to Richard Palethorpe

This is a copy and paste of Nicolai's reproducer. The main difference is that I moved some code around. Of course I also used LTP library features, but essentially it works the same.

There are some hard coded values which I do not like. I guess these could be calculated or varied somehow. However I struggle to understand what the kernel is doing. This perhaps needs more investigation. We could try generalising this test and setsockopt03.

---
 runtest/cve                                   |   1 +
 runtest/syscalls                              |   1 +
 .../kernel/syscalls/setsockopt/.gitignore     |   1 +
 .../kernel/syscalls/setsockopt/setsockopt08.c | 156 ++++++++++++++++++
 4 files changed, 159 insertions(+)
 create mode 100644 testcases/kernel/syscalls/setsockopt/setsockopt08.c
runtest/cve +1 -0
index 8aa048a40..60bd22f01 100644
@@ -65,4 +65,5 @@ cve-2020-14416 pty03
 cve-2020-25705 icmp_rate_limit01
 cve-2020-29373 io_uring02
 cve-2021-3444 bpf_prog05
+cve-2021-22555 setsockopt08 -i 100
 cve-2021-26708 vsock01
runtest/syscalls +1 -0
index b379b2d90..2222990fe 100644
@@ -1378,6 +1378,7 @@ setsockopt04 setsockopt04
 setsockopt05 setsockopt05
 setsockopt06 setsockopt06
 setsockopt07 setsockopt07
+setsockopt08 setsockopt08
 
 settimeofday01 settimeofday01
 settimeofday02 settimeofday02
testcases/kernel/syscalls/setsockopt/.gitignore +1 -0
index 1ca5b836b..95a5e43f8 100644
@@ -5,3 +5,4 @@
 /setsockopt05
 /setsockopt06
 /setsockopt07
+/setsockopt08
testcases/kernel/syscalls/setsockopt/setsockopt08.c new file +156 -0

156 lines not on this page · raw message

Signature · 1 line
2.31.1
#13 lore raw

PATCH v4 3/3 Add setsockopt08, CVE-2021-22555

Hi!

testcases/kernel/syscalls/setsockopt/setsockopt08.c line 84

Quoted 3 lines
testcases/kernel/syscalls/setsockopt/setsockopt08.c new file +3 -0
+#include "tst_test.h"
+#include "tst_safe_net.h"
+#include "lapi/ip_tables.h"

I've fixed the failure on Centos with missing IFNAMSIZ but this still fails to compile on ubuntu xenial because the tst_safe_net.h pull in netinet/in.h and lapi/ip_tables.h pulls in linux/in.h and on old enough systems these two headers does not like to be included at the same time.

I'm not sure how to fix this, either we drop the include to linux/netfilter_ipv4/ip_tables.h completely or we add a configure check if netinet/in.h and linux/in.h could be included at the same time and ifdef the linux/netfilter_ipv4/ip_tables.h with that check.

Either way both looks like a hack, if anyone has a better idea please suggest it.

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

PATCH v4 3/3 Add setsockopt08, CVE-2021-22555

Hi!

Quoted 9 lines

testcases/kernel/syscalls/setsockopt/setsockopt08.c line 84

Quoted 3 lines
testcases/kernel/syscalls/setsockopt/setsockopt08.c new file +3 -0
+#include "tst_test.h"
+#include "tst_safe_net.h"
+#include "lapi/ip_tables.h"

I've fixed the failure on Centos with missing IFNAMSIZ but this still fails to compile on ubuntu xenial because the tst_safe_net.h pull in netinet/in.h and lapi/ip_tables.h pulls in linux/in.h and on old enough systems these two headers does not like to be included at the same time.

Quoted 7 lines

I'm not sure how to fix this, either we drop the include to linux/netfilter_ipv4/ip_tables.h completely or we add a configure check if netinet/in.h and linux/in.h could be included at the same time and ifdef the linux/netfilter_ipv4/ip_tables.h with that check.

Either way both looks like a hack, if anyone has a better idea please suggest it.

Uff and it looks like setsockopt03 does include the same headers so this could probably be fixed by another shuffle, but I will have to figure out what has to be moved and where.

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

PATCH v4 3/3 Add setsockopt08, CVE-2021-22555

Hello Cyril,

Quoted 15 lines · Cyril Hrubis

Hi!

Quoted 14 lines

testcases/kernel/syscalls/setsockopt/setsockopt08.c line 84

Quoted 3 lines
testcases/kernel/syscalls/setsockopt/setsockopt08.c new file +3 -0
+#include "tst_test.h"
+#include "tst_safe_net.h"
+#include "lapi/ip_tables.h"

I've fixed the failure on Centos with missing IFNAMSIZ but this still fails to compile on ubuntu xenial because the tst_safe_net.h pull in netinet/in.h and lapi/ip_tables.h pulls in linux/in.h and on old enough systems these two headers does not like to be included at the same time.

I'm not sure how to fix this, either we drop the include to linux/netfilter_ipv4/ip_tables.h completely or we add a configure check if netinet/in.h and linux/in.h could be included at the same time and ifdef the linux/netfilter_ipv4/ip_tables.h with that check.

Quoted 6 lines
Quoted 2 lines

Either way both looks like a hack, if anyone has a better idea please suggest it.

Uff and it looks like setsockopt03 does include the same headers so this could probably be fixed by another shuffle, but I will have to figure out what has to be moved and where.

Looking in libc-compat.h we have:

/* Coordinate with glibc netinet/in.h header. */
#if defined(_NETINET_IN_H)

/* GLIBC headers included first so don't define anything
* that would already be defined. */
#define __UAPI_DEF_IN_ADDR 0
...

I suppose we may be able to do something similar. In tst_safe_net.h we could check if _LINUX_IN_H is defined before including netinet/in.h. Then include lapi/ip_tables.h first. Maybe even put

#ifdef _NETINET_IN_H

# error "incompatible system header ..."
#endif

in lapi/ip_tables.h

IDK if this may break some existing tests.

Signature · 2 lines
Thank you,
Richard.
#17 lore raw

PATCH v3 1/3 API: TST_EXP_FAIL: Allow passing when errno is not checked

In reply to Richard Palethorpe

Hi!

Quoted 11 lines
Quoted 10 lines
Quoted 2 lines

This is also less surprising than giving errno == 0 a dual meaning.

But I do agree that the current if (ERRNO) branch is confusing. I would be for dropping the if (ERRNO) and checking the TST_ERR against ERRNO unconditionally.

Also note that the TEST() macro clears errno, so if a syscall fails but does not report any error TST_ERR will end up 0 either way so there is no need for having special handling for 0.

Quoted 6 lines

There is if the errno is set, but is undefined. Like if the resulting errno is platform or config dependent.

In the present case though we can just check for EINVAL. That is what the setsockopt man page indicates. Initially I wasn't sure if EFAULT were equally valid, but it seems not.

There was also an idea to add another version that would take an array of possible errnos if that happens, but I fear that the complexity would be getting out of hand if we keep adding features like that.

Quoted 1 line

I suppose we can fall back to using TEST() if the other case arises.

Yes.

Signature · 2 lines
Cyril Hrubis
[email protected]