Skip to contentSkip to conversation

shmctl01: don't use hardcoded index == 0 for SHM_STAT test

PATCH New Jan Stancek 3 participants 6 messages May 20, 2019 – May 21, 2019

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 shmctl01: don't use hardcoded index == 0 for SHM_STAT test

Test fails on SHM_STAT testcase:
shmctl01 5 TFAIL : shmctl01.c:173: shmctl01 call failed
errno = 22 : Invalid argument
shmctl(0, SHM_STAT, 0x601060) = -EINVAL

since following commit:
commit 99db46ea292780cd978d56932d9445b1e8bdafe8
Author: Manfred Spraul <[email protected]>
Date: Tue May 14 15:46:36 2019 -0700
ipc: do cyclic id allocation for the ipc object.

Don't rely on index 0 being always available, but instead use (maximum) index returned by SHM_INFO.

---
 testcases/kernel/syscalls/ipc/shmctl/shmctl01.c | 21 ++++++++++++++++++---
 1 file changed, 18 insertions(+), 3 deletions(-)
testcases/kernel/syscalls/ipc/shmctl/shmctl01.c +18 -3
index 1b46977554d5..52bf23a40fc1 100644
@@ -79,6 +79,7 @@ static void func_info(int ret);
 
 /* Check routine for SHM_STAT */
 static void func_sstat(int ret);
+static void func_sstat_setup(void);
 
 /* Check routine for SHM_LOCK */
 static void func_lock(int ret);
@@ -110,7 +111,7 @@ static struct test_case_t {
 #endif
 	{&shm_id_1, IPC_SET, &buf, func_set, set_setup},
 	{&shm_id_1, IPC_INFO, (struct shmid_ds *) &info, func_info, NULL},
-	{&shm_index, SHM_STAT, &buf, func_sstat, NULL},
+	{&shm_index, SHM_STAT, &buf, func_sstat, func_sstat_setup},
 	{&shm_id_1, SHM_LOCK, NULL, func_lock, NULL},
 	{&shm_id_1, SHM_UNLOCK, NULL, func_unlock, NULL},
 	{&shm_id_1, IPC_RMID, NULL, func_rmid, NULL},
@@ -407,9 +408,23 @@ static void func_info(int ret)
 static void func_sstat(int ret)
 {
 	if (ret >= 0)
-		tst_resm(TPASS, "get correct shared memory id");
+		tst_resm(TPASS, "get correct shared memory id for index: %d",
+			shm_index);
 	else
-		tst_resm(TFAIL, "shared memory id is incorrect");
+		tst_resm(TFAIL, "shared memory id is incorrect, index: %d",
+			shm_index);
+}
+
+static void func_sstat_setup(void)
+{
+	struct shm_info tmp;
+	int ret;
+
+	ret = shmctl(shm_id_1, SHM_INFO, (void *)&tmp);
+	if (ret < 0)
+		tst_resm(TFAIL|TERRNO, "shmctl(SHM_INFO)");
+	else
+		shm_index = ret;
 }
 
 static void func_lock(int ret)
Signature · 1 line
1.8.3.1
#1 lore raw

PATCH 1/1 shmctl01: don't use hardcoded index == 0 for SHM_STAT test

Hi Jan,

Quoted 11 lines

Test fails on SHM_STAT testcase:
shmctl01 5 TFAIL : shmctl01.c:173: shmctl01 call failed
errno = 22 : Invalid argument
shmctl(0, SHM_STAT, 0x601060) = -EINVAL

since following commit:
commit 99db46ea292780cd978d56932d9445b1e8bdafe8
Author: Manfred Spraul <[email protected]>
Date: Tue May 14 15:46:36 2019 -0700
ipc: do cyclic id allocation for the ipc object.

Quoted 4 lines

Don't rely on index 0 being always available, but instead use (maximum) index returned by SHM_INFO.

Good catch + further improvements.

Kind regards,
Petr

#2 lore raw

PATCH 1/1 shmctl01: don't use hardcoded index == 0 for SHM_STAT test

In reply to Jan Stancek

Hi!

testcases/kernel/syscalls/ipc/shmctl/shmctl01.c static void func_info(int ret) · line 425

Quoted 8 lines
testcases/kernel/syscalls/ipc/shmctl/shmctl01.c +8 -0
+static void func_sstat_setup(void)
+{
+	struct shm_info tmp;
+	int ret;
+
+	ret = shmctl(shm_id_1, SHM_INFO, (void *)&tmp);
+	if (ret < 0)
+		tst_resm(TFAIL|TERRNO, "shmctl(SHM_INFO)");

I wonder if it makes sense to continue here if shmctl() has failed, maybe we should do tst_brkm(TBROK ...) instead.

Otherwise it looks good, acked.

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

PATCH 1/1 shmctl01: don't use hardcoded index == 0 for SHM_STAT test

----- Original Message -----

Quoted 12 lines

Hi!

testcases/kernel/syscalls/ipc/shmctl/shmctl01.c static void func_info(int ret) · line 425

Quoted 8 lines
testcases/kernel/syscalls/ipc/shmctl/shmctl01.c +8 -0
+static void func_sstat_setup(void)
+{
+	struct shm_info tmp;
+	int ret;
+
+	ret = shmctl(shm_id_1, SHM_INFO, (void *)&tmp);
+	if (ret < 0)
+		tst_resm(TFAIL|TERRNO, "shmctl(SHM_INFO)");

I wonder if it makes sense to continue here if shmctl() has failed, maybe we should do tst_brkm(TBROK ...) instead.

It is 'setup phase', but it also is kind-of test for shmctl() as well. We don't risk much if we continue, worst case is we get one more fail.

@Petr: Do you have any preference?

Quoted 7 lines

Otherwise it looks good, acked.

--
Cyril Hrubis
[email protected]

#4 lore raw

PATCH 1/1 shmctl01: don't use hardcoded index == 0 for SHM_STAT test

Hi Jan, Cyril,

Quoted 14 lines

----- Original Message -----

Quoted 12 lines

Hi!

testcases/kernel/syscalls/ipc/shmctl/shmctl01.c static void func_info(int ret) · line 425

Quoted 8 lines
testcases/kernel/syscalls/ipc/shmctl/shmctl01.c +8 -0
+static void func_sstat_setup(void)
+{
+	struct shm_info tmp;
+	int ret;
+
+	ret = shmctl(shm_id_1, SHM_INFO, (void *)&tmp);
+	if (ret < 0)
+		tst_resm(TFAIL|TERRNO, "shmctl(SHM_INFO)");

I wonder if it makes sense to continue here if shmctl() has failed, maybe we should do tst_brkm(TBROK ...) instead.

Quoted 4 lines

It is 'setup phase', but it also is kind-of test for shmctl() as well. We don't risk much if we continue, worst case is we get one more fail.

@Petr: Do you have any preference?

Cyril's point makes sense, but I prefer Jan's way.

Kind regards,
Petr

#5 lore raw

PATCH 1/1 shmctl01: don't use hardcoded index == 0 for SHM_STAT test

----- Original Message -----

Quoted 16 lines

Hi Jan, Cyril,

Quoted 13 lines

----- Original Message -----

Quoted 12 lines

Hi!

testcases/kernel/syscalls/ipc/shmctl/shmctl01.c static void func_info(int ret) · line 425

Quoted 8 lines
testcases/kernel/syscalls/ipc/shmctl/shmctl01.c +8 -0
+static void func_sstat_setup(void)
+{
+	struct shm_info tmp;
+	int ret;
+
+	ret = shmctl(shm_id_1, SHM_INFO, (void *)&tmp);
+	if (ret < 0)
+		tst_resm(TFAIL|TERRNO, "shmctl(SHM_INFO)");

I wonder if it makes sense to continue here if shmctl() has failed, maybe we should do tst_brkm(TBROK ...) instead.

Quoted 5 lines
Quoted 4 lines

It is 'setup phase', but it also is kind-of test for shmctl() as well. We don't risk much if we continue, worst case is we get one more fail.

@Petr: Do you have any preference?

Cyril's point makes sense, but I prefer Jan's way.

I pushed TFAIL version.

Thanks for review,
Jan

Quoted 5 lines

Kind regards,
Petr