Skip to contentSkip to conversation

fanotify25: fix test failure on kernel with CONFIG_MODULES=n

PATCH In review Amir Goldstein 5 participants 11 messages Mar 2 – Mar 3

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 fanotify25: fix test failure on kernel with CONFIG_MODULES=n

On a kernel built without modules loading support the test fails:
fanotify25.c:61: TBROK: write(4,0x55a523b99160,31) failed: EINVAL (22)

Use another symbol which is less likely to be missing.

---
 testcases/kernel/syscalls/fanotify/fanotify25.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
testcases/kernel/syscalls/fanotify/fanotify25.c +4 -4
index c982f3225..9b03fd9ab 100644
@@ -24,10 +24,10 @@ static const struct traceconfig {
 	const char *filename;
 	const char *wdata;
 } trace_cmds[] = {
-	{EVENTS_SYSFILE, "p:ltp_load_module_0 load_module"},
-	{MNTPOINT "/events/kprobes/ltp_load_module_0/enable", "1"},
-	{MNTPOINT "/events/kprobes/ltp_load_module_0/enable", "0"},
-	{EVENTS_SYSFILE, "-:ltp_load_module_0"},
+	{EVENTS_SYSFILE, "p:ltp_load_script_0 load_script"},
+	{MNTPOINT "/events/kprobes/ltp_load_script_0/enable", "1"},
+	{MNTPOINT "/events/kprobes/ltp_load_script_0/enable", "0"},
+	{EVENTS_SYSFILE, "-:ltp_load_script_0"},
 	{}
 };
 
Signature · 5 lines
2.53.0


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

Hi!

Quoted 11 lines
---
 testcases/kernel/syscalls/fanotify/fanotify25.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
testcases/kernel/syscalls/fanotify/fanotify25.c +0 -0
index c982f3225..9b03fd9ab 100644
@@ -24,10 +24,10 @@ static const struct traceconfig {
 	const char *filename;

testcases/kernel/syscalls/fanotify/fanotify25.c static const struct traceconfig { · line 30

Quoted 10 lines
testcases/kernel/syscalls/fanotify/fanotify25.c +4 -4
 	const char *wdata;
 } trace_cmds[] = {
-	{EVENTS_SYSFILE, "p:ltp_load_module_0 load_module"},
-	{MNTPOINT "/events/kprobes/ltp_load_module_0/enable", "1"},
-	{MNTPOINT "/events/kprobes/ltp_load_module_0/enable", "0"},
-	{EVENTS_SYSFILE, "-:ltp_load_module_0"},
+	{EVENTS_SYSFILE, "p:ltp_load_script_0 load_script"},
+	{MNTPOINT "/events/kprobes/ltp_load_script_0/enable", "1"},
+	{MNTPOINT "/events/kprobes/ltp_load_script_0/enable", "0"},
+	{EVENTS_SYSFILE, "-:ltp_load_script_0"},

Doesn't this depend on CONFIG_BINFMT_SCRIPT? How likely is that this is turned off?

If possible I would like to avoid whack-a-mole game trying to find something that is commonly enabled.

Signature · 5 lines
Cyril Hrubis
[email protected]

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#2 lore raw
Quoted 18 lines · Cyril Hrubis

Hi!

Quoted 16 lines
---
 testcases/kernel/syscalls/fanotify/fanotify25.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
testcases/kernel/syscalls/fanotify/fanotify25.c +0 -3
index c982f3225..9b03fd9ab 100644
@@ -24,10 +24,10 @@ static const struct traceconfig {
      const char *filename;
      const char *wdata;
 } trace_cmds[] = {
-     {EVENTS_SYSFILE, "p:ltp_load_module_0 load_module"},
-     {MNTPOINT "/events/kprobes/ltp_load_module_0/enable", "1"},
-     {MNTPOINT "/events/kprobes/ltp_load_module_0/enable", "0"},
Quoted 8 lines

testcases/kernel/syscalls/fanotify/fanotify25.c static const struct traceconfig { · line 30

Quoted 5 lines
testcases/kernel/syscalls/fanotify/fanotify25.c +4 -1
-     {EVENTS_SYSFILE, "-:ltp_load_module_0"},
+     {EVENTS_SYSFILE, "p:ltp_load_script_0 load_script"},
+     {MNTPOINT "/events/kprobes/ltp_load_script_0/enable", "1"},
+     {MNTPOINT "/events/kprobes/ltp_load_script_0/enable", "0"},
+     {EVENTS_SYSFILE, "-:ltp_load_script_0"},

Doesn't this depend on CONFIG_BINFMT_SCRIPT? How likely is that this is turned off?

CONFIG_MODULES=n is a common config for testing I don't think this is the case for CONFIG_BINFMT_SCRIPT

Quoted 3 lines

If possible I would like to avoid whack-a-mole game trying to find something that is commonly enabled.

Feel free to choose any other symbol.

Thanks,
Amir.

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by
#3 lore raw
Quoted 28 lines · Cyril Hrubis
Quoted 18 lines

Hi!

Quoted 16 lines
---
 testcases/kernel/syscalls/fanotify/fanotify25.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
testcases/kernel/syscalls/fanotify/fanotify25.c +0 -3
index c982f3225..9b03fd9ab 100644
@@ -24,10 +24,10 @@ static const struct traceconfig {
      const char *filename;
      const char *wdata;
 } trace_cmds[] = {
-     {EVENTS_SYSFILE, "p:ltp_load_module_0 load_module"},
-     {MNTPOINT "/events/kprobes/ltp_load_module_0/enable", "1"},
-     {MNTPOINT "/events/kprobes/ltp_load_module_0/enable", "0"},
Quoted 8 lines

testcases/kernel/syscalls/fanotify/fanotify25.c static const struct traceconfig { · line 30

Quoted 5 lines
testcases/kernel/syscalls/fanotify/fanotify25.c +4 -1
-     {EVENTS_SYSFILE, "-:ltp_load_module_0"},
+     {EVENTS_SYSFILE, "p:ltp_load_script_0 load_script"},
+     {MNTPOINT "/events/kprobes/ltp_load_script_0/enable", "1"},
+     {MNTPOINT "/events/kprobes/ltp_load_script_0/enable", "0"},
+     {EVENTS_SYSFILE, "-:ltp_load_script_0"},

Doesn't this depend on CONFIG_BINFMT_SCRIPT? How likely is that this is turned off?

Quoted 2 lines

CONFIG_MODULES=n is a common config for testing I don't think this is the case for CONFIG_BINFMT_SCRIPT

Yes, it's default y and just briefly checking it's in Tumbleweed and Debian testing => LGTM.

Kind regards,
Petr

Quoted 7 lines
Quoted 2 lines

If possible I would like to avoid whack-a-mole game trying to find something that is commonly enabled.

Feel free to choose any other symbol.

Thanks,
Amir.

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
#4 lore raw
In reply to Amir Goldstein
Quoted 5 lines · Amir Goldstein
Quoted 2 lines · Cyril Hrubis

If possible I would like to avoid whack-a-mole game trying to find something that is commonly enabled.

Feel free to choose any other symbol.

Hi,
how about oom_kill_process()? That one's unconditionally available and shouldn't get called very often in a reasonable test environment either.

I'll prepare a patch and check that it still triggers the expected failures on broken kernels.

Signature · 10 lines
Martin Doucha   [email protected]
SW Quality Engineer
SUSE LINUX, s.r.o.
CORSO IIa
Krizikova 148/34
186 00 Prague 8
Czech Republic

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#5 lore raw
Quoted 11 lines · Martin Doucha
Quoted 5 lines · Amir Goldstein
Quoted 2 lines · Cyril Hrubis

If possible I would like to avoid whack-a-mole game trying to find something that is commonly enabled.

Feel free to choose any other symbol.

Hi,
how about oom_kill_process()? That one's unconditionally available and shouldn't get called very often in a reasonable test environment either.

Why does it matter how much the function gets called? Isn't that irrelevant for the test?

Quoted 3 lines

I'll prepare a patch and check that it still triggers the expected failures on broken kernels.

I really think we are over-thinking this.

The problem with load_module was that some test setup configs do not have it - that's all.

Thanks,
Amir.

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
#6 lore raw
Quoted 8 lines · Amir Goldstein
Quoted 4 lines · Martin Doucha

Hi,
how about oom_kill_process()? That one's unconditionally available and shouldn't get called very often in a reasonable test environment either.

Why does it matter how much the function gets called? Isn't that irrelevant for the test?

I've written the test to test a specific fanotify regression. Tracing a syscall that gets called too often could increase system load and change timing of the race which I'm trying to trigger with it.

Signature · 10 lines
Martin Doucha   [email protected]
SW Quality Engineer
SUSE LINUX, s.r.o.
CORSO IIa
Krizikova 148/34
186 00 Prague 8
Czech Republic

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#7 lore raw
Quoted 8 lines · Martin Doucha
Quoted 6 lines · Amir Goldstein
Quoted 4 lines · Martin Doucha

Hi,
how about oom_kill_process()? That one's unconditionally available and shouldn't get called very often in a reasonable test environment either.

Quoted 7 lines
Quoted 2 lines

Why does it matter how much the function gets called? Isn't that irrelevant for the test?

I've written the test to test a specific fanotify regression. Tracing a syscall that gets called too often could increase system load and change timing of the race which I'm trying to trigger with it.

Which fanotify regression?
As a reviewer, I got no indication in the commit message,
Test description, nor in LTP tst_test tags that this is a test for a regression.

I have no objection whatsoever to using any other symbol available in most of the common builds.
Alternatively, the test could test for the desired configuration.

But if the test is meant to track a regression, then *please* document the regression in the test description and the fix commit if such fix exists in LTP tst_test tags.

Thanks,
Amir.

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
#8 lore raw
Quoted 10 lines · Amir Goldstein
Quoted 4 lines · Martin Doucha

I've written the test to test a specific fanotify regression. Tracing a syscall that gets called too often could increase system load and change timing of the race which I'm trying to trigger with it.

Which fanotify regression?
As a reviewer, I got no indication in the commit message,
Test description, nor in LTP tst_test tags that this is a test for a regression.

Quoted 7 lines

I have no objection whatsoever to using any other symbol available in most of the common builds.
Alternatively, the test could test for the desired configuration.

But if the test is meant to track a regression, then *please* document the regression in the test description and the fix commit if such fix exists in LTP tst_test tags.

The regression in question was not in upstream but a broken backport in a custom kernel branch. It was fixed by reverting the backported patch. So while the test will be useful for detecting similar regressions in upstream, there is nothing I can publicly document about it.

Signature · 10 lines
Martin Doucha   [email protected]
SW Quality Engineer
SUSE LINUX, s.r.o.
CORSO IIa
Krizikova 148/34
186 00 Prague 8
Czech Republic

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#9 lore raw
Quoted 16 lines · Martin Doucha
Quoted 14 lines · Amir Goldstein
Quoted 4 lines · Martin Doucha

I've written the test to test a specific fanotify regression. Tracing a syscall that gets called too often could increase system load and change timing of the race which I'm trying to trigger with it.

Which fanotify regression?
As a reviewer, I got no indication in the commit message,
Test description, nor in LTP tst_test tags that this is a test for a regression.

I have no objection whatsoever to using any other symbol available in most of the common builds.
Alternatively, the test could test for the desired configuration.

Quoted 9 lines
Quoted 3 lines

But if the test is meant to track a regression, then *please* document the regression in the test description and the fix commit if such fix exists in LTP tst_test tags.

The regression in question was not in upstream but a broken backport in a custom kernel branch. It was fixed by reverting the backported patch. So while the test will be useful for detecting similar regressions in upstream, there is nothing I can publicly document about it.

Fair enough, but when choosing the function to trace, please document the reason for choosing this function and document the fact the expectation that this event will not be triggered during the test or will not be triggered too frequently, whichever is the case.

Thanks,
Amir.

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by on v1 1/1
#10 lore raw
In reply to Amir Goldstein
Quoted 6 lines · Amir Goldstein

On a kernel built without modules loading support the test fails:
fanotify25.c:61: TBROK: write(4,0x55a523b99160,31) failed: EINVAL (22)

Use another symbol which is less likely to be missing.

Looks good. Feel free to add:

Honza

Quoted 26 lines
---
 testcases/kernel/syscalls/fanotify/fanotify25.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
testcases/kernel/syscalls/fanotify/fanotify25.c +4 -4
index c982f3225..9b03fd9ab 100644
@@ -24,10 +24,10 @@ static const struct traceconfig {
 	const char *filename;
 	const char *wdata;
 } trace_cmds[] = {
-	{EVENTS_SYSFILE, "p:ltp_load_module_0 load_module"},
-	{MNTPOINT "/events/kprobes/ltp_load_module_0/enable", "1"},
-	{MNTPOINT "/events/kprobes/ltp_load_module_0/enable", "0"},
-	{EVENTS_SYSFILE, "-:ltp_load_module_0"},
+	{EVENTS_SYSFILE, "p:ltp_load_script_0 load_script"},
+	{MNTPOINT "/events/kprobes/ltp_load_script_0/enable", "1"},
+	{MNTPOINT "/events/kprobes/ltp_load_script_0/enable", "0"},
+	{EVENTS_SYSFILE, "-:ltp_load_script_0"},
 	{}
 };
 
Signature · 1 line
2.53.0
Signature · 5 lines
Jan Kara <[email protected]>
SUSE Labs, CR

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp