[PATCH v5 1/2] of: kobj: export of_node_ktype for use by modules

[PATCH v5 1/2] of: kobj: export of_node_ktype for use by modules

[Xu Yang]

From: Xu Yang <xu.yang_2@nxp.com>

of_node_init() is a static inline that references of_node_ktype when
CONFIG_OF_KOBJ=y. Any module that calls of_node_init() will therefore
have an unresolved reference to of_node_ktype at load time, because
the symbol is defined in drivers/of/kobj.c but was never exported.

This causes a modpost build error when CONFIG_OF_KOBJ=y and
CONFIG_DRIVER_PE_KUNIT_TEST=m:

  ERROR: modpost: "of_node_ktype"
  [drivers/base/test/property-entry-test.ko] undefined!

Add EXPORT_SYMBOL_GPL(of_node_ktype) so that GPL-licensed modules
such as the KUnit property-entry test can call of_node_init() without
hitting this linker error.

Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202607181651.RnUuV8n6-lkp@intel.com/
Signed-off-by: Xu Yang <xu.yang_2@nxp.com>

---
Changes in v5:
 - new one
---
 drivers/of/kobj.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/of/kobj.c b/drivers/of/kobj.c
index 1bb61a2c3399..eacc53324d62 100644
--- a/drivers/of/kobj.c
+++ b/drivers/of/kobj.c
@@ -27,6 +27,7 @@ static void of_node_release(struct kobject *kobj)
 const struct kobj_type of_node_ktype = {
 	.release = of_node_release,
 };
+EXPORT_SYMBOL_GPL(of_node_ktype);
 
 static ssize_t of_node_property_read(struct file *filp, struct kobject *kobj,
 				const struct bin_attribute *bin_attr, char *buf,
-- 
2.34.1

[PATCH v5 2/2] device property: add test cases for fwnode_for_each_child_node()

[Xu Yang]

From: Xu Yang <xu.yang_2@nxp.com>

Add test cases for fwnode_for_each_child_node() API.

Test command:
$ ./tools/testing/kunit/kunit.py run property-entry

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Xu Yang <xu.yang_2@nxp.com>

---
Changes in v5:
 - add patch#1 to fix "of_node_ktype undefined error"
---
 drivers/base/test/Kconfig               |   1 +
 drivers/base/test/property-entry-test.c | 136 ++++++++++++++++++++++++
 2 files changed, 137 insertions(+)

diff --git a/drivers/base/test/Kconfig b/drivers/base/test/Kconfig
index 1ecf0791241a..542ce07530a1 100644
--- a/drivers/base/test/Kconfig
+++ b/drivers/base/test/Kconfig
@@ -17,6 +17,7 @@ config DM_KUNIT_TEST
 config DRIVER_PE_KUNIT_TEST
 	tristate "KUnit Tests for property entry API" if !KUNIT_ALL_TESTS
 	depends on KUNIT
+	select OF
 	default KUNIT_ALL_TESTS
 
 config DRIVER_SWNODE_KUNIT_TEST
diff --git a/drivers/base/test/property-entry-test.c b/drivers/base/test/property-entry-test.c
index a8657eb06f94..a3d1caf28cc7 100644
--- a/drivers/base/test/property-entry-test.c
+++ b/drivers/base/test/property-entry-test.c
@@ -4,6 +4,8 @@
 // Copyright 2019 Google LLC.
 
 #include <kunit/test.h>
+
+#include <linux/of.h>
 #include <linux/property.h>
 #include <linux/types.h>
 
@@ -489,6 +491,139 @@ static void pe_test_reference(struct kunit *test)
 	software_node_unregister_node_group(group);
 }
 
+static struct fwnode_handle *create_device_node(struct kunit *test,
+						const char *name,
+						const char *full_name,
+						struct device_node *parent)
+{
+	struct device_node *node;
+
+	node = kunit_kzalloc(test, sizeof(*node), GFP_KERNEL);
+	KUNIT_ASSERT_NOT_ERR_OR_NULL(test, node);
+
+	node->name = kunit_kstrdup(test, name, GFP_KERNEL);
+	node->full_name = kunit_kstrdup(test, full_name, GFP_KERNEL);
+
+	if (parent) {
+		node->sibling = parent->child;
+		/* set the node as the first child of the parent */
+		parent->child = node;
+		node->parent = parent;
+	}
+
+	of_node_init(node);
+	return of_fwnode_handle(node);
+}
+
+/* Verifies that fwnode_for_each_child_node() can output correct children */
+static void pe_test_child_iteration(struct kunit *test)
+{
+	struct fwnode_handle *of_node, *of_node1;
+	struct fwnode_handle *sw_node, *sw_node1;
+	struct fwnode_handle *child;
+	int error, i, num;
+
+	static const struct software_node node = { .name = "sw" };
+	static const struct software_node node1 = { .name = "sw-1", .parent = &node};
+	static const struct software_node node2 = { .name = "sw-2", .parent = &node};
+	static const struct software_node node3 = { .name = "sw-3", .parent = &node};
+	static const struct software_node *group[] = { &node, &node1, &node2, &node3, NULL };
+
+	static const char * const of_child_array[] = { "of-1", "of-2", "of-3" };
+	static const char * const sw_child_array[] = { "sw-1", "sw-2", "sw-3" };
+	static const char * const of_sw_child_array[] = { "of-1", "of-2", "of-3",
+							  "sw-1", "sw-2", "sw-3" };
+	static const char * const sw_of_child_array[] = { "sw-1", "sw-2", "sw-3",
+							  "of-1", "of-2", "of-3" };
+
+	/* 1. Test OF node child iteration */
+
+	of_node = create_device_node(test, "of", "of", NULL);
+	create_device_node(test, "of", "of-3", to_of_node(of_node));
+	create_device_node(test, "of", "of-2", to_of_node(of_node));
+	of_node1 = create_device_node(test, "of", "of-1", to_of_node(of_node));
+
+	i = 0;
+	num = ARRAY_SIZE(of_child_array);
+	fwnode_for_each_child_node(of_node, child) {
+		KUNIT_ASSERT_LT(test, i, num);
+		KUNIT_EXPECT_STREQ(test, of_child_array[i++], fwnode_get_name(child));
+	}
+	KUNIT_EXPECT_PTR_EQ(test, child, NULL);
+
+	/* 2. Test SW node child iteration */
+
+	error = software_node_register_node_group(group);
+	KUNIT_ASSERT_EQ(test, error, 0);
+
+	sw_node = software_node_fwnode(&node);
+
+	i = 0;
+	num = ARRAY_SIZE(sw_child_array);
+	fwnode_for_each_child_node(sw_node, child) {
+		KUNIT_ASSERT_LT(test, i, num);
+		KUNIT_EXPECT_STREQ(test, sw_child_array[i++], fwnode_get_name(child));
+	}
+	KUNIT_EXPECT_PTR_EQ(test, child, NULL);
+
+	/* 3. Test OF (primary) + SW (secondary) node child iteration */
+
+	of_node->secondary = sw_node;
+	sw_node->secondary = ERR_PTR(-ENODEV);
+
+	i = 0;
+	num = ARRAY_SIZE(of_sw_child_array);
+	fwnode_for_each_child_node(of_node, child) {
+		KUNIT_ASSERT_LT(test, i, num);
+		KUNIT_EXPECT_STREQ(test, of_sw_child_array[i++], fwnode_get_name(child));
+	}
+	KUNIT_EXPECT_PTR_EQ(test, child, NULL);
+
+	/* 4. Test SW (primary) + OF (secondary) node child iteration */
+
+	sw_node->secondary = of_node;
+	of_node->secondary = ERR_PTR(-ENODEV);
+
+	i = 0;
+	num = ARRAY_SIZE(sw_of_child_array);
+	fwnode_for_each_child_node(sw_node, child) {
+		KUNIT_ASSERT_LT(test, i, num);
+		KUNIT_EXPECT_STREQ(test, sw_of_child_array[i++], fwnode_get_name(child));
+	}
+	KUNIT_EXPECT_PTR_EQ(test, child, NULL);
+
+	/* 5. Test OF (primary) + SW (secondary, but no children) node child iteration */
+
+	sw_node1 = software_node_fwnode(&node1);
+	of_node->secondary = sw_node1;
+	sw_node->secondary = ERR_PTR(-ENODEV);
+
+	i = 0;
+	num = ARRAY_SIZE(of_child_array);
+	fwnode_for_each_child_node(of_node, child) {
+		KUNIT_ASSERT_LT(test, i, num);
+		KUNIT_EXPECT_STREQ(test, of_child_array[i++], fwnode_get_name(child));
+	}
+	KUNIT_EXPECT_PTR_EQ(test, child, NULL);
+
+	/* 6. Test SW (primary) + OF (secondary, but no children) node child iteration */
+
+	sw_node->secondary = of_node1;
+	of_node->secondary = ERR_PTR(-ENODEV);
+
+	i = 0;
+	num = ARRAY_SIZE(sw_child_array);
+	fwnode_for_each_child_node(sw_node, child) {
+		KUNIT_ASSERT_LT(test, i, num);
+		KUNIT_EXPECT_STREQ(test, sw_child_array[i++], fwnode_get_name(child));
+	}
+	KUNIT_EXPECT_PTR_EQ(test, child, NULL);
+
+	of_node->secondary = NULL;
+	sw_node->secondary = NULL;
+	software_node_unregister_node_group(group);
+}
+
 static struct kunit_case property_entry_test_cases[] = {
 	KUNIT_CASE(pe_test_uints),
 	KUNIT_CASE(pe_test_uint_arrays),
@@ -497,6 +632,7 @@ static struct kunit_case property_entry_test_cases[] = {
 	KUNIT_CASE(pe_test_move_inline_u8),
 	KUNIT_CASE(pe_test_move_inline_str),
 	KUNIT_CASE(pe_test_reference),
+	KUNIT_CASE(pe_test_child_iteration),
 	{ }
 };
 
-- 
2.34.1

Re: [PATCH v5 2/2] device property: add test cases for fwnode_for_each_child_node()

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

New issues:
- [Medium] The KUnit test fails to use `kunit_add_action()` for global state cleanup, permanently leaking software nodes on test aborts.

Pre-existing issues:
- [High] The newly added tests expose pre-existing infinite loop and double-put bugs in `fwnode_get_next_child_node()` when transitioning between primary OF nodes and secondary SW nodes.
--

--- Patch [2]: [PATCH v5 2/2] device property: add test cases for fwnode_for_each_child_node() ---
commit 6f128f7ffab2a2787ad6820b967d94a17fcbcbc2
Author: Xu Yang <xu.yang_2@nxp.com>
device property: add test cases for fwnode_for_each_child_node()

Add test cases for fwnode_for_each_child_node() API.

> diff --git a/drivers/base/test/property-entry-test.c b/drivers/base/test/property-entry-test.c
> index a8657eb06f94e..a3d1caf28cc7d 100644
> --- a/drivers/base/test/property-entry-test.c
> +++ b/drivers/base/test/property-entry-test.c
[ ... ]
> @@ -489,6 +491,139 @@ static void pe_test_reference(struct kunit *test)
[ ... ]
> +	/* 2. Test SW node child iteration */
> +
> +	error = software_node_register_node_group(group);
> +	KUNIT_ASSERT_EQ(test, error, 0);
> +
> +	sw_node = software_node_fwnode(&node);
> +
> +	i = 0;
> +	num = ARRAY_SIZE(sw_child_array);
> +	fwnode_for_each_child_node(sw_node, child) {
> +		KUNIT_ASSERT_LT(test, i, num);

[Severity: Medium]
If KUNIT_ASSERT_LT() fails, it will immediately abort the test execution.
Does this bypass the manual software_node_unregister_node_group() call at
the end of the function?

To prevent leaking the software nodes and causing sysfs conflicts on
subsequent test runs, should kunit_add_action() be used to register the
cleanup immediately after registration?

> +		KUNIT_EXPECT_STREQ(test, sw_child_array[i++], fwnode_get_name(child));
> +	}
> +	KUNIT_EXPECT_PTR_EQ(test, child, NULL);
> +
> +	/* 3. Test OF (primary) + SW (secondary) node child iteration */
> +
> +	of_node->secondary = sw_node;
> +	sw_node->secondary = ERR_PTR(-ENODEV);
> +
> +	i = 0;
> +	num = ARRAY_SIZE(of_sw_child_array);
> +	fwnode_for_each_child_node(of_node, child) {
> +		KUNIT_ASSERT_LT(test, i, num);

[Severity: High]
This isn't a bug introduced by this patch, but these new tests appear to
expose a pre-existing infinite loop and double-put bug in the core fwnode
iterator.

When fwnode_for_each_child_node() transitions from primary OF children to
secondary SW children, of_fwnode_get_next_child_node() drops the reference
on the last OF child and returns NULL. The core iterator then passes this
same child pointer to the secondary node's iterator:

drivers/base/property.c:fwnode_get_next_child_node() {
    ...
    next = fwnode_call_ptr_op(fwnode, get_next_child_node, child);
    if (next)
        return next;
    return fwnode_call_ptr_op(fwnode->secondary, get_next_child_node, child);
}

Does this cause the secondary iterator (software_node_get_next_child) to
unconditionally drop the reference a second time?

Furthermore, when iterating over the secondary children, the returned SW node
is eventually passed back to the primary iterator. Since to_of_node()
evaluates to NULL for SW nodes, of_get_next_available_child() restarts from
the first OF child:

drivers/of/property.c:of_fwnode_get_next_child_node() {
    ...
    return of_fwnode_handle(of_get_next_available_child(to_of_node(fwnode),
                                                        to_of_node(child)));
}

Will this restart cause an infinite loop in the test runner when testing
mixed primary and secondary firmware nodes?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260720104730.1285552-1-xu.yang_2@oss.nxp.com?part=2

Re: [PATCH v5 1/2] of: kobj: export of_node_ktype for use by modules

[Andy Shevchenko]

On Mon, Jul 20, 2026 at 06:47:29PM +0800, Xu Yang wrote:

> of_node_init() is a static inline that references of_node_ktype when
> CONFIG_OF_KOBJ=y. Any module that calls of_node_init() will therefore
> have an unresolved reference to of_node_ktype at load time, because
> the symbol is defined in drivers/of/kobj.c but was never exported.
> 
> This causes a modpost build error when CONFIG_OF_KOBJ=y and
> CONFIG_DRIVER_PE_KUNIT_TEST=m:
> 
>   ERROR: modpost: "of_node_ktype"
>   [drivers/base/test/property-entry-test.ko] undefined!
> 
> Add EXPORT_SYMBOL_GPL(of_node_ktype) so that GPL-licensed modules
> such as the KUnit property-entry test can call of_node_init() without
> hitting this linker error.

Use special one for KUnits: EXPORT_SYMBOL_IF_KUNIT().

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v5 2/2] device property: add test cases for fwnode_for_each_child_node()

[Andy Shevchenko]

On Mon, Jul 20, 2026 at 06:47:30PM +0800, Xu Yang wrote:

> Add test cases for fwnode_for_each_child_node() API.

> Test command:
> $ ./tools/testing/kunit/kunit.py run property-entry


> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

Nope, I haven't Signed-off it here, it's your new submission where you are
the author. I only made a few cosmetic changes, if you want to give a credit
you can mention that in the comment block (SoB+Co-developed-by here seems
like an overkill).

...

You have a series of two patches but I haven't seen the cover letter.
I use this script [1] that helps me not to forget this.

[1]: https://github.com/andy-shev/home-bin-tools/blob/master/ge2maintainer.sh

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v5 1/2] of: kobj: export of_node_ktype for use by modules

[Xu Yang]

On Mon, Jul 20, 2026 at 02:13:36PM +0300, Andy Shevchenko wrote:
> On Mon, Jul 20, 2026 at 06:47:29PM +0800, Xu Yang wrote:
> 
> > of_node_init() is a static inline that references of_node_ktype when
> > CONFIG_OF_KOBJ=y. Any module that calls of_node_init() will therefore
> > have an unresolved reference to of_node_ktype at load time, because
> > the symbol is defined in drivers/of/kobj.c but was never exported.
> > 
> > This causes a modpost build error when CONFIG_OF_KOBJ=y and
> > CONFIG_DRIVER_PE_KUNIT_TEST=m:
> > 
> >   ERROR: modpost: "of_node_ktype"
> >   [drivers/base/test/property-entry-test.ko] undefined!
> > 
> > Add EXPORT_SYMBOL_GPL(of_node_ktype) so that GPL-licensed modules
> > such as the KUnit property-entry test can call of_node_init() without
> > hitting this linker error.
> 
> Use special one for KUnits: EXPORT_SYMBOL_IF_KUNIT().

OK. Thanks for the suggestion!

Thanks,
Xu Yang

Re: [PATCH v5 2/2] device property: add test cases for fwnode_for_each_child_node()

[Xu Yang]

On Mon, Jul 20, 2026 at 02:21:11PM +0300, Andy Shevchenko wrote:
> On Mon, Jul 20, 2026 at 06:47:30PM +0800, Xu Yang wrote:
> 
> > Add test cases for fwnode_for_each_child_node() API.
> 
> > Test command:
> > $ ./tools/testing/kunit/kunit.py run property-entry
> 
> 
> > Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> 
> Nope, I haven't Signed-off it here, it's your new submission where you are
> the author. I only made a few cosmetic changes, if you want to give a credit
> you can mention that in the comment block (SoB+Co-developed-by here seems
> like an overkill).

Ah.. I just picked the patch from lore without modification. Let me remove
it in v2.

> 
> ...
> 
> You have a series of two patches but I haven't seen the cover letter.
> I use this script [1] that helps me not to forget this.
> 
> [1]: https://github.com/andy-shev/home-bin-tools/blob/master/ge2maintainer.sh

Thanks for the reminder and the script :)

Thanks,
Xu Yang