In 'NOFENTRY_ARGS' test case for syntax check, any offset X of `vfs_read+X` except function entry offset (0) fits the criterion, even if that offset is not at instruction boundary, as the parser comes before probing. But with "ENDBR64" instruction on x86, offset 4 is treated as function entry. So, X can't be 4 as well. Thus, 8 was used as offset for the test case. On 64-bit powerpc though, any offset <= 16 can be considered function entry depending on build configuration (see arch_kprobe_on_func_entry() for implementation details). So, use `vfs_read+20` to accommodate that scenario too.
Suggested-by: Masami Hiramatsu mhiramat@kernel.org Signed-off-by: Hari Bathini hbathini@linux.ibm.com ---
Changes in v2: * Use 20 as offset for all arches.
.../selftests/ftrace/test.d/kprobe/kprobe_syntax_errors.tc | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tools/testing/selftests/ftrace/test.d/kprobe/kprobe_syntax_errors.tc b/tools/testing/selftests/ftrace/test.d/kprobe/kprobe_syntax_errors.tc index a16c6a6f6055..8f1c58f0c239 100644 --- a/tools/testing/selftests/ftrace/test.d/kprobe/kprobe_syntax_errors.tc +++ b/tools/testing/selftests/ftrace/test.d/kprobe/kprobe_syntax_errors.tc @@ -111,7 +111,7 @@ check_error 'p vfs_read $arg* ^$arg*' # DOUBLE_ARGS if !grep -q 'kernel return probes support:' README; then check_error 'r vfs_read ^$arg*' # NOFENTRY_ARGS fi -check_error 'p vfs_read+8 ^$arg*' # NOFENTRY_ARGS +check_error 'p vfs_read+20 ^$arg*' # NOFENTRY_ARGS check_error 'p vfs_read ^hoge' # NO_BTFARG check_error 'p kfree ^$arg10' # NO_BTFARG (exceed the number of parameters) check_error 'r kfree ^$retval' # NO_RETVAL
On Sat, 30 Nov 2024 01:56:21 +0530 Hari Bathini hbathini@linux.ibm.com wrote:
In 'NOFENTRY_ARGS' test case for syntax check, any offset X of `vfs_read+X` except function entry offset (0) fits the criterion, even if that offset is not at instruction boundary, as the parser comes before probing. But with "ENDBR64" instruction on x86, offset 4 is treated as function entry. So, X can't be 4 as well. Thus, 8 was used as offset for the test case. On 64-bit powerpc though, any offset <= 16 can be considered function entry depending on build configuration (see arch_kprobe_on_func_entry() for implementation details). So, use `vfs_read+20` to accommodate that scenario too.
Suggested-by: Masami Hiramatsu mhiramat@kernel.org Signed-off-by: Hari Bathini hbathini@linux.ibm.com
Acked-by: Steven Rostedt (Google) rostedt@goodmis.org
Shuah,
Can you take this through your tree?
Thanks,
-- Steve
Changes in v2:
- Use 20 as offset for all arches.
.../selftests/ftrace/test.d/kprobe/kprobe_syntax_errors.tc | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tools/testing/selftests/ftrace/test.d/kprobe/kprobe_syntax_errors.tc b/tools/testing/selftests/ftrace/test.d/kprobe/kprobe_syntax_errors.tc index a16c6a6f6055..8f1c58f0c239 100644 --- a/tools/testing/selftests/ftrace/test.d/kprobe/kprobe_syntax_errors.tc +++ b/tools/testing/selftests/ftrace/test.d/kprobe/kprobe_syntax_errors.tc @@ -111,7 +111,7 @@ check_error 'p vfs_read $arg* ^$arg*' # DOUBLE_ARGS if !grep -q 'kernel return probes support:' README; then check_error 'r vfs_read ^$arg*' # NOFENTRY_ARGS fi -check_error 'p vfs_read+8 ^$arg*' # NOFENTRY_ARGS +check_error 'p vfs_read+20 ^$arg*' # NOFENTRY_ARGS check_error 'p vfs_read ^hoge' # NO_BTFARG check_error 'p kfree ^$arg10' # NO_BTFARG (exceed the number of parameters) check_error 'r kfree ^$retval' # NO_RETVAL
On 12/2/24 12:41, Steven Rostedt wrote:
On Sat, 30 Nov 2024 01:56:21 +0530 Hari Bathini hbathini@linux.ibm.com wrote:
In 'NOFENTRY_ARGS' test case for syntax check, any offset X of `vfs_read+X` except function entry offset (0) fits the criterion, even if that offset is not at instruction boundary, as the parser comes before probing. But with "ENDBR64" instruction on x86, offset 4 is treated as function entry. So, X can't be 4 as well. Thus, 8 was used as offset for the test case. On 64-bit powerpc though, any offset <= 16 can be considered function entry depending on build configuration (see arch_kprobe_on_func_entry() for implementation details). So, use `vfs_read+20` to accommodate that scenario too.
Suggested-by: Masami Hiramatsu mhiramat@kernel.org Signed-off-by: Hari Bathini hbathini@linux.ibm.com
Acked-by: Steven Rostedt (Google) rostedt@goodmis.org
Shuah,
Can you take this through your tree?
Yes I can take it. I do have question about whether this is a fix - sounds like it is from the change log.
Clearly stating that it is a fix will help so it can be picked up for stables.
thanks, -- Shuah
On Tue, 3 Dec 2024 18:01:06 -0700 Shuah Khan skhan@linuxfoundation.org wrote:
On 12/2/24 12:41, Steven Rostedt wrote:
On Sat, 30 Nov 2024 01:56:21 +0530 Hari Bathini hbathini@linux.ibm.com wrote:
In 'NOFENTRY_ARGS' test case for syntax check, any offset X of `vfs_read+X` except function entry offset (0) fits the criterion, even if that offset is not at instruction boundary, as the parser comes before probing. But with "ENDBR64" instruction on x86, offset 4 is treated as function entry. So, X can't be 4 as well. Thus, 8 was used as offset for the test case. On 64-bit powerpc though, any offset <= 16 can be considered function entry depending on build configuration (see arch_kprobe_on_func_entry() for implementation details). So, use `vfs_read+20` to accommodate that scenario too.
Suggested-by: Masami Hiramatsu mhiramat@kernel.org Signed-off-by: Hari Bathini hbathini@linux.ibm.com
Acked-by: Steven Rostedt (Google) rostedt@goodmis.org
Shuah,
Can you take this through your tree?
Yes I can take it. I do have question about whether this is a fix - sounds like it is from the change log.
Clearly stating that it is a fix will help so it can be picked up for stables.
I would say it's a fix, as the test currently fails in certain scenarios for powerpc.
You can add:
Fixes: 4231f30fcc34a ("selftests/ftrace: Add BTF arguments test cases")
-- Steve
On 12/3/24 18:20, Steven Rostedt wrote:
On Tue, 3 Dec 2024 18:01:06 -0700 Shuah Khan skhan@linuxfoundation.org wrote:
On 12/2/24 12:41, Steven Rostedt wrote:
On Sat, 30 Nov 2024 01:56:21 +0530 Hari Bathini hbathini@linux.ibm.com wrote:
In 'NOFENTRY_ARGS' test case for syntax check, any offset X of `vfs_read+X` except function entry offset (0) fits the criterion, even if that offset is not at instruction boundary, as the parser comes before probing. But with "ENDBR64" instruction on x86, offset 4 is treated as function entry. So, X can't be 4 as well. Thus, 8 was used as offset for the test case. On 64-bit powerpc though, any offset <= 16 can be considered function entry depending on build configuration (see arch_kprobe_on_func_entry() for implementation details). So, use `vfs_read+20` to accommodate that scenario too.
Suggested-by: Masami Hiramatsu mhiramat@kernel.org Signed-off-by: Hari Bathini hbathini@linux.ibm.com
Acked-by: Steven Rostedt (Google) rostedt@goodmis.org
Shuah,
Can you take this through your tree?
Yes I can take it. I do have question about whether this is a fix - sounds like it is from the change log.
Clearly stating that it is a fix will help so it can be picked up for stables.
I would say it's a fix, as the test currently fails in certain scenarios for powerpc.
You can add:
Fixes: 4231f30fcc34a ("selftests/ftrace: Add BTF arguments test cases")
I applied this to linux-kselftest fixes - will send it up for rc2 or rc3
thanks, -- Shuah
On Thu, 5 Dec 2024 09:06:43 -0700 Shuah Khan skhan@linuxfoundation.org wrote:
I applied this to linux-kselftest fixes - will send it up for rc2 or rc3
Thanks Shuah,
-- Steve
linux-kselftest-mirror@lists.linaro.org
-
Hari Bathini -
Shuah Khan -
Steven Rostedt