Skip to contentSkip to conversation

pidns05: Use tst_getpid()

PATCH In review Martin Doucha 3 participants 6 messages Sep 20, 2023 – Sep 25, 2023

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 pidns05: Use tst_getpid()

Older glibc versions cache getpid() results. Calling clone() through the direct syscall interface (bypassing glibc) with the CLONE_NEWPID argument will result in stale getpid() cache and the test will fail. Use tst_getpid() instead to get the correct PID value.

---
 testcases/kernel/containers/pidns/pidns05.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
testcases/kernel/containers/pidns/pidns05.c +2 -2
index ec9ce9bf2..0e7739aaa 100644
@@ -28,7 +28,7 @@ static void child_func(int *level)
 {
 	pid_t cpid, ppid;
 
-	cpid = getpid();
+	cpid = tst_getpid();
 	ppid = getppid();
 
 	TST_EXP_EQ_LI(cpid, 1);
@@ -55,7 +55,7 @@ static int find_cinit_pids(pid_t *pids)
 	int next = 0;
 	pid_t parentpid, pgid, pgid2;
 
-	parentpid = getpid();
+	parentpid = tst_getpid();
 	pgid = SAFE_GETPGID(parentpid);
 
 	for (pid = 2; pid < pid_max; pid++) {
Signature · 5 lines
2.42.0


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by on v1 1/1
#1 lore raw

Hi Martin,

Obviously candidate for pre-release fix.

Kind regards,
Petr

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by on v1 1/1
#2 lore raw
In reply to Martin Doucha

Hi!

Quoted 12 lines
testcases/kernel/containers/pidns/pidns05.c +1 -1
@@ -28,7 +28,7 @@ static void child_func(int *level)
 {
 	pid_t cpid, ppid;
 
-	cpid = getpid();
+	cpid = tst_getpid();
 	ppid = getppid();
 
 	TST_EXP_EQ_LI(cpid, 1);

testcases/kernel/containers/pidns/pidns05.c static int find_cinit_pids(pid_t *pids) · line 58

Quoted 5 lines
testcases/kernel/containers/pidns/pidns05.c +1 -1
@@ -55,7 +55,7 @@ static int find_cinit_pids(pid_t *pids)
 	int next = 0;
 	pid_t parentpid, pgid, pgid2;
 
-	parentpid = getpid();
+	parentpid = tst_getpid();

I suppose that this one is not strictly required, but it does not harm to consistently use tst_getpid()

Signature · 5 lines
Cyril Hrubis
[email protected]

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#3 lore raw

Hi all,

Quoted 16 lines

Hi!

Quoted 14 lines
testcases/kernel/containers/pidns/pidns05.c +1 -1
@@ -28,7 +28,7 @@ static void child_func(int *level)
 {
 	pid_t cpid, ppid;
 
-	cpid = getpid();
+	cpid = tst_getpid();
 	ppid = getppid();
 
 	TST_EXP_EQ_LI(cpid, 1);
@@ -55,7 +55,7 @@ static int find_cinit_pids(pid_t *pids)
 	int next = 0;
 	pid_t parentpid, pgid, pgid2;
Quoted 5 lines

testcases/kernel/containers/pidns/pidns05.c static int find_cinit_pids(pid_t *pids) · line 58

Quoted 2 lines
testcases/kernel/containers/pidns/pidns05.c +1 -1
-	parentpid = getpid();
+	parentpid = tst_getpid();

I suppose that this one is not strictly required, but it does not harm to consistently use tst_getpid()

Well, Martin found that some of SLES versions (e.g. some SLES versions based on kernel 4.12 or older) started to fail after rewrite:

8c201e55f ("Remove libclone dependency from pidns05 test")

tst_test.c:1684: TINFO: LTP version: 20230516
tst_test.c:1570: TINFO: Timeout per run is 0h 00m 30s pidns05.c:34: TFAIL: cpid (2075) != 1 (1)
pidns05.c:35: TPASS: ppid == 0 (0)
pidns05.c:34: TFAIL: cpid (2075) != 1 (1)
pidns05.c:35: TPASS: ppid == 0 (0)
pidns05.c:34: TFAIL: cpid (2075) != 1 (1)
pidns05.c:35: TPASS: ppid == 0 (0)
pidns05.c:34: TFAIL: cpid (2075) != 1 (1)
pidns05.c:35: TPASS: ppid == 0 (0)
pidns05.c:34: TFAIL: cpid (2075) != 1 (1)
pidns05.c:35: TPASS: ppid == 0 (0)
pidns05.c:34: TFAIL: cpid (2075) != 1 (1)
pidns05.c:35: TPASS: ppid == 0 (0)
pidns05.c:97: TPASS: find_cinit_pids(pids) returned 6 pidns05.c:115: TPASS: No children left after sending SIGKILL to the first child

Therefore it was needed.

Kind regards,
Petr

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
#4 lore raw

Hi!

Quoted 2 lines

Well, Martin found that some of SLES versions (e.g. some SLES versions based on kernel 4.12 or older) started to fail after rewrite:

I mean that the second instance of getpid() is not called from cloned() process, the first one has to be replaced for sure.

Signature · 5 lines
Cyril Hrubis
[email protected]

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#5 lore raw
Quoted 6 lines

Hi!

Quoted 2 lines

Well, Martin found that some of SLES versions (e.g. some SLES versions based on kernel 4.12 or older) started to fail after rewrite:

I mean that the second instance of getpid() is not called from cloned() process, the first one has to be replaced for sure.

Ah, right. But be consistent with tst_getpid() looks to me slightly better.

Kind regards,
Petr

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp