* [PATCH v5 1/2] of: kobj: export of_node_ktype for use by modules @ 2026-07-20 10:47 Xu Yang 2026-07-20 10:47 ` [PATCH v5 2/2] device property: add test cases for fwnode_for_each_child_node() Xu Yang 2026-07-20 11:13 ` [PATCH v5 1/2] of: kobj: export of_node_ktype for use by modules Andy Shevchenko 0 siblings, 2 replies; 5+ messages in thread From: Xu Yang @ 2026-07-20 10:47 UTC (permalink / raw) To: gregkh, rafael, dakr, robh, saravanak, andriy.shevchenko Cc: bartosz.golaszewski, driver-core, linux-kernel, devicetree, imx 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 ^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH v5 2/2] device property: add test cases for fwnode_for_each_child_node() 2026-07-20 10:47 [PATCH v5 1/2] of: kobj: export of_node_ktype for use by modules Xu Yang @ 2026-07-20 10:47 ` Xu Yang 2026-07-20 11:02 ` sashiko-bot 2026-07-20 11:21 ` Andy Shevchenko 2026-07-20 11:13 ` [PATCH v5 1/2] of: kobj: export of_node_ktype for use by modules Andy Shevchenko 1 sibling, 2 replies; 5+ messages in thread From: Xu Yang @ 2026-07-20 10:47 UTC (permalink / raw) To: gregkh, rafael, dakr, robh, saravanak, andriy.shevchenko Cc: bartosz.golaszewski, driver-core, linux-kernel, devicetree, imx 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 ^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH v5 2/2] device property: add test cases for fwnode_for_each_child_node() 2026-07-20 10:47 ` [PATCH v5 2/2] device property: add test cases for fwnode_for_each_child_node() Xu Yang @ 2026-07-20 11:02 ` sashiko-bot 2026-07-20 11:21 ` Andy Shevchenko 1 sibling, 0 replies; 5+ messages in thread From: sashiko-bot @ 2026-07-20 11:02 UTC (permalink / raw) To: Xu Yang; +Cc: Frank.Li, robh, imx, devicetree, conor+dt 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 ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH v5 2/2] device property: add test cases for fwnode_for_each_child_node() 2026-07-20 10:47 ` [PATCH v5 2/2] device property: add test cases for fwnode_for_each_child_node() Xu Yang 2026-07-20 11:02 ` sashiko-bot @ 2026-07-20 11:21 ` Andy Shevchenko 1 sibling, 0 replies; 5+ messages in thread From: Andy Shevchenko @ 2026-07-20 11:21 UTC (permalink / raw) To: Xu Yang Cc: gregkh, rafael, dakr, robh, saravanak, bartosz.golaszewski, driver-core, linux-kernel, devicetree, imx 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 ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH v5 1/2] of: kobj: export of_node_ktype for use by modules 2026-07-20 10:47 [PATCH v5 1/2] of: kobj: export of_node_ktype for use by modules Xu Yang 2026-07-20 10:47 ` [PATCH v5 2/2] device property: add test cases for fwnode_for_each_child_node() Xu Yang @ 2026-07-20 11:13 ` Andy Shevchenko 1 sibling, 0 replies; 5+ messages in thread From: Andy Shevchenko @ 2026-07-20 11:13 UTC (permalink / raw) To: Xu Yang Cc: gregkh, rafael, dakr, robh, saravanak, bartosz.golaszewski, driver-core, linux-kernel, devicetree, imx 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 ^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2026-07-20 11:21 UTC | newest] Thread overview: 5+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2026-07-20 10:47 [PATCH v5 1/2] of: kobj: export of_node_ktype for use by modules Xu Yang 2026-07-20 10:47 ` [PATCH v5 2/2] device property: add test cases for fwnode_for_each_child_node() Xu Yang 2026-07-20 11:02 ` sashiko-bot 2026-07-20 11:21 ` Andy Shevchenko 2026-07-20 11:13 ` [PATCH v5 1/2] of: kobj: export of_node_ktype for use by modules Andy Shevchenko
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox