[PATCH BlueZ 0/6] [v4] Fix bugs found by SVACE static analisys tool

[PATCH BlueZ 0/6] [v4] Fix bugs found by SVACE static analisys tool

[Ildar Kamaletdinov]

This patch set includes few fixes that was found by Linux Verification Center
(linuxtesting.org) with the SVACE static analysis tool.

I have manually filtered out non-relevant and false positive problems and only
procedeed with bugs that currently lead to some errors/vulnerabilities or may
lead to them in some specific conditions.

Changelog:
[v4] one patch was dropped due to overchecking, seems that it is not required
[v3] one fix wasn't staged, sorry, one more fix after CI checks
[v2] some minor style fixes after CI check.
[v1] initial version.

Ildar Kamaletdinov (6):
  monitor: Fix out-of-bound read in print_le_states
  tools: Fix buffer overflow in hciattach_tialt.c
  tools: Fix signed integer overflow in btsnoop.c
  tools: Limit width of fields in sscanf
  device: Limit width of fields in sscanf
  gatt: Fix double free and freed memory dereference

 monitor/packet.c        |  7 ++++---
 src/device.c            | 14 +++++++-------
 src/gatt-database.c     |  4 ++++
 tools/btmgmt.c          |  2 +-
 tools/btsnoop.c         |  2 +-
 tools/hciattach_tialt.c |  3 ++-
 tools/hex2hcd.c         |  2 +-
 7 files changed, 20 insertions(+), 14 deletions(-)

-- 
2.35.1

[PATCH BlueZ 1/6] monitor: Fix out-of-bound read in print_le_states

[Ildar Kamaletdinov]

Accessing le_states_desc_table array with value 15 can cause
out-of-bound read because current size of array is 14.

Currently this cannot lead to any problems becase we do no have such
state in le_states_comb_table but this could be changed in future and
raise described problem.

Found by Linux Verification Center (linuxtesting.org) with the SVACE
static analysis tool.
---
 monitor/packet.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/monitor/packet.c b/monitor/packet.c
index b7431b57d..1f04063d3 100644
--- a/monitor/packet.c
+++ b/monitor/packet.c
@@ -2816,7 +2816,8 @@ static const struct {
 static void print_le_states(const uint8_t *states_array)
 {
 	uint64_t mask, states = 0;
-	int i, n;
+	int i = 0;
+	size_t n = 0;
 
 	for (i = 0; i < 8; i++)
 		states |= ((uint64_t) states_array[i]) << (i * 8);
@@ -2828,12 +2829,12 @@ static void print_le_states(const uint8_t *states_array)
 	for (i = 0; le_states_comb_table[i].states; i++) {
 		uint64_t val = (((uint64_t) 1) << le_states_comb_table[i].bit);
 		const char *str[3] = { NULL, };
-		int num = 0;
+		size_t num = 0;
 
 		if (!(states & val))
 			continue;
 
-		for (n = 0; n < 16; n++) {
+		for (n = 0; n < ARRAY_SIZE(le_states_desc_table); n++) {
 			if (le_states_comb_table[i].states & (1 << n))
 				str[num++] = le_states_desc_table[n].str;
 		}
-- 
2.35.1

RE: Fix bugs found by SVACE static analisys tool

[bluez.test.bot]

[-- Attachment #1: Type: text/plain, Size: 998 bytes --]

This is automated email and please do not reply to this email!

Dear submitter,

Thank you for submitting the patches to the linux bluetooth mailing list.
This is a CI test results with your patch series:
PW Link:https://patchwork.kernel.org/project/bluetooth/list/?series=628188

---Test result---

Test Summary:
CheckPatch                    PASS      8.20 seconds
GitLint                       PASS      5.62 seconds
Prep - Setup ELL              PASS      39.64 seconds
Build - Prep                  PASS      0.69 seconds
Build - Configure             PASS      7.87 seconds
Build - Make                  PASS      1330.57 seconds
Make Check                    PASS      11.64 seconds
Make Check w/Valgrind         PASS      404.30 seconds
Make Distcheck                PASS      209.94 seconds
Build w/ext ELL - Configure   PASS      7.88 seconds
Build w/ext ELL - Make        PASS      1278.53 seconds
Incremental Build with patchesPASS      8064.66 seconds



---
Regards,
Linux Bluetooth

[PATCH BlueZ 2/6] tools: Fix buffer overflow in hciattach_tialt.c

[Ildar Kamaletdinov]

Array 'c_brf_chip' of size 8 could be accessed by index > 7. We should
limit array access like in previous check at line 221.

Found by Linux Verification Center (linuxtesting.org) with the SVACE
static analysis tool.
---
 tools/hciattach_tialt.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/tools/hciattach_tialt.c b/tools/hciattach_tialt.c
index 520b383a1..4f7fd42a3 100644
--- a/tools/hciattach_tialt.c
+++ b/tools/hciattach_tialt.c
@@ -221,7 +221,8 @@ int texasalt_init(int fd, int speed, struct termios *ti)
 				((brf_chip > 7) ? "unknown" : c_brf_chip[brf_chip]),
 				brf_chip);
 
-		sprintf(fw, "/etc/firmware/%s.bin", c_brf_chip[brf_chip]);
+		sprintf(fw, "/etc/firmware/%s.bin",
+			(brf_chip > 7) ? "unknown" : c_brf_chip[brf_chip]);
 		texas_load_firmware(fd, fw);
 
 		texas_change_speed(fd, speed);
-- 
2.35.1

[PATCH BlueZ 3/6] tools: Fix signed integer overflow in btsnoop.c

[Ildar Kamaletdinov]

If malformed packet is proceed with zero 'size' field we will face with
wrong behaviour of write() call. Value 'toread - 1' gives wrong sign
for value 'written' (-1) in write() call. To prevent this we should
check that 'toread' is not equal to zero.

Found by Linux Verification Center (linuxtesting.org) with the SVACE
static analysis tool.
---
 tools/btsnoop.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/btsnoop.c b/tools/btsnoop.c
index 738027dfc..a0d6cf356 100644
--- a/tools/btsnoop.c
+++ b/tools/btsnoop.c
@@ -193,7 +193,7 @@ next_packet:
 	flags = be32toh(input_pkt[select_input].flags);
 
 	len = read(input_fd[select_input], buf, toread);
-	if (len < 0 || len != (ssize_t) toread) {
+	if (toread == 0 || len < 0 || len != (ssize_t) toread) {
 		close(input_fd[select_input]);
 		input_fd[select_input] = -1;
 		goto next_packet;
-- 
2.35.1

[PATCH BlueZ 4/6] tools: Limit width of fields in sscanf

[Ildar Kamaletdinov]

In tools/btmgmt.c and tools/hex2hcd.c few sscanf does not limit width
of fields. This could lead to static overflow and stack corruption.

Found by Linux Verification Center (linuxtesting.org) with the SVACE
static analysis tool.
---
 tools/btmgmt.c  | 2 +-
 tools/hex2hcd.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/tools/btmgmt.c b/tools/btmgmt.c
index 42ef9acef..8f63f12ba 100644
--- a/tools/btmgmt.c
+++ b/tools/btmgmt.c
@@ -5164,7 +5164,7 @@ static bool str2pattern(struct mgmt_adv_pattern *pattern, const char *str)
 	char pattern_str[62] = { 0 };
 	char tmp;
 
-	if (sscanf(str, "%2hhx%n:%2hhx%n:%s", &pattern->ad_type, &type_len,
+	if (sscanf(str, "%2hhx%n:%2hhx%n:%61s", &pattern->ad_type, &type_len,
 			&pattern->offset, &offset_end_pos, pattern_str) != 3)
 		return false;
 
diff --git a/tools/hex2hcd.c b/tools/hex2hcd.c
index 674d62744..e6dca5a81 100644
--- a/tools/hex2hcd.c
+++ b/tools/hex2hcd.c
@@ -248,7 +248,7 @@ static void ver_parse_file(const char *pathname)
 
 	memset(ver, 0, sizeof(*ver));
 
-	if (sscanf(pathname, "%[A-Z0-9]_%3c.%3c.%3c.%4c.%4c.hex",
+	if (sscanf(pathname, "%19[A-Z0-9]_%3c.%3c.%3c.%4c.%4c.hex",
 					ver->name, ver->major, ver->minor,
 					ver->build, dummy1, dummy2) != 6) {
 		printf("\t/* failed to parse %s */\n", pathname);
-- 
2.35.1

[PATCH BlueZ 5/6] device: Limit width of fields in sscanf

[Ildar Kamaletdinov]

In src/device.c few sscanf does not limit width of uuid field. This
could lead to static overflow and stack corruption.

Found by Linux Verification Center (linuxtesting.org) with the SVACE
static analysis tool.
---
 src/device.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/src/device.c b/src/device.c
index 381faf91c..8dc12d026 100644
--- a/src/device.c
+++ b/src/device.c
@@ -3790,8 +3790,8 @@ static int load_desc(char *handle, char *value,
 		return -EIO;
 
 	/* Check if there is any value stored, otherwise it is just the UUID */
-	if (sscanf(value, "%04hx:%s", &val, uuid_str) != 2) {
-		if (sscanf(value, "%s", uuid_str) != 1)
+	if (sscanf(value, "%04hx:%36s", &val, uuid_str) != 2) {
+		if (sscanf(value, "%36s", uuid_str) != 1)
 			return -EIO;
 		val = 0;
 	}
@@ -3840,9 +3840,9 @@ static int load_chrc(char *handle, char *value,
 		return -EIO;
 
 	/* Check if there is any value stored */
-	if (sscanf(value, GATT_CHARAC_UUID_STR ":%04hx:%02hx:%32s:%s",
+	if (sscanf(value, GATT_CHARAC_UUID_STR ":%04hx:%02hx:%32s:%36s",
 			&value_handle, &properties, val_str, uuid_str) != 4) {
-		if (sscanf(value, GATT_CHARAC_UUID_STR ":%04hx:%02hx:%s",
+		if (sscanf(value, GATT_CHARAC_UUID_STR ":%04hx:%02hx:%36s",
 				&value_handle, &properties, uuid_str) != 3)
 			return -EIO;
 		val_len = 0;
@@ -3884,8 +3884,8 @@ static int load_incl(struct gatt_db *db, char *handle, char *value,
 	if (sscanf(handle, "%04hx", &start) != 1)
 		return -EIO;
 
-	if (sscanf(value, GATT_INCLUDE_UUID_STR ":%04hx:%04hx:%s", &start, &end,
-								uuid_str) != 3)
+	if (sscanf(value, GATT_INCLUDE_UUID_STR ":%04hx:%04hx:%36s", &start,
+							&end, uuid_str) != 3)
 		return -EIO;
 
 	/* Log debug message. */
@@ -3918,7 +3918,7 @@ static int load_service(struct gatt_db *db, char *handle, char *value)
 	if (sscanf(handle, "%04hx", &start) != 1)
 		return -EIO;
 
-	if (sscanf(value, "%[^:]:%04hx:%s", type, &end, uuid_str) != 3)
+	if (sscanf(value, "%[^:]:%04hx:%36s", type, &end, uuid_str) != 3)
 		return -EIO;
 
 	if (g_str_equal(type, GATT_PRIM_SVC_UUID_STR))
-- 
2.35.1

[PATCH BlueZ 6/6] gatt: Fix double free and freed memory dereference

[Ildar Kamaletdinov]

If device is no longer exists or not paired when notifications send it
is possible to get double free and dereference of already freed memory.

To avoid this we need to recheck the state of device after sending
notification.

Found by Linux Verification Center (linuxtesting.org) with the SVACE
static analysis tool.
---
 src/gatt-database.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/src/gatt-database.c b/src/gatt-database.c
index d6c94058c..d32f616a9 100644
--- a/src/gatt-database.c
+++ b/src/gatt-database.c
@@ -3877,6 +3877,10 @@ void btd_gatt_database_server_connected(struct btd_gatt_database *database,
 
 	send_notification_to_device(state, state->pending);
 
+	state = find_device_state(database, &bdaddr, bdaddr_type);
+	if (!state || !state->pending)
+		return;
+
 	free(state->pending->value);
 	free(state->pending);
 	state->pending = NULL;
-- 
2.35.1

Re: [PATCH BlueZ 0/6] [v4] Fix bugs found by SVACE static analisys tool

[patchwork-bot+bluetooth]

Hello:

This series was applied to bluetooth/bluez.git (master)
by Luiz Augusto von Dentz <luiz.von.dentz@intel.com>:

On Fri, 1 Apr 2022 15:16:41 +0300 you wrote:
> This patch set includes few fixes that was found by Linux Verification Center
> (linuxtesting.org) with the SVACE static analysis tool.
> 
> I have manually filtered out non-relevant and false positive problems and only
> procedeed with bugs that currently lead to some errors/vulnerabilities or may
> lead to them in some specific conditions.
> 
> [...]

Here is the summary with links:
  - [BlueZ,1/6] monitor: Fix out-of-bound read in print_le_states
    https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=7fdfb67284a2
  - [BlueZ,2/6] tools: Fix buffer overflow in hciattach_tialt.c
    https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=d328abaa1715
  - [BlueZ,3/6] tools: Fix signed integer overflow in btsnoop.c
    https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=497a0b220dbd
  - [BlueZ,4/6] tools: Limit width of fields in sscanf
    https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=1e664b9838cd
  - [BlueZ,5/6] device: Limit width of fields in sscanf
    https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=4afbef7790ce
  - [BlueZ,6/6] gatt: Fix double free and freed memory dereference
    https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=3627eddea130

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html