This series: 1. makes the behavior of_find_device_by_node(), bus_find_device_by_of_node(), bus_find_device_by_fwnode(), etc., more consistent when provided with a NULL node/handle; 2. adds kunit tests to validate the new NULL-argument behavior; and 3. makes some related improvements and refactoring for the drivers/base/ kunit tests.
This series aims to prevent problems like the ones resolved in commit 5c8418cf4025 ("PCI/pwrctrl: Unregister platform device only if one actually exists").
Changes in v2: * Add Rob's Reviewed-by * CC LKML (oops!) * Keep "devm" and "match" tests in separate suites
Brian Norris (3): drivers: base: Don't match devices with NULL of_node/fwnode/etc drivers: base: test: Enable device model tests with KUNIT_ALL_TESTS drivers: base: test: Add ...find_device_by...(... NULL) tests
drivers/base/core.c | 8 ++--- drivers/base/test/Kconfig | 1 + drivers/base/test/platform-device-test.c | 42 +++++++++++++++++++++++- 3 files changed, 46 insertions(+), 5 deletions(-)
of_find_device_by_node(), bus_find_device_by_of_node(), bus_find_device_by_fwnode(), ..., all produce arbitrary results when provided with a NULL of_node, fwnode, ACPI handle, etc. This is counterintuitive, and the source of a few bugs, such as the one fixed by commit 5c8418cf4025 ("PCI/pwrctrl: Unregister platform device only if one actually exists").
It's hard to imagine a good reason that these device_match_*() APIs should return 'true' for a NULL argument. Augment these to return 0 (false).
Signed-off-by: Brian Norris briannorris@chromium.org Reviewed-by: Rob Herring (Arm) robh@kernel.org ---
Changes in v2: * Add Rob's Reviewed-by
drivers/base/core.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/drivers/base/core.c b/drivers/base/core.c index 94865c9d8adc..2b7b13fc36d7 100644 --- a/drivers/base/core.c +++ b/drivers/base/core.c @@ -5246,13 +5246,13 @@ EXPORT_SYMBOL_GPL(device_match_name);
int device_match_of_node(struct device *dev, const void *np) { - return dev->of_node == np; + return np && dev->of_node == np; } EXPORT_SYMBOL_GPL(device_match_of_node);
int device_match_fwnode(struct device *dev, const void *fwnode) { - return dev_fwnode(dev) == fwnode; + return fwnode && dev_fwnode(dev) == fwnode; } EXPORT_SYMBOL_GPL(device_match_fwnode);
@@ -5264,13 +5264,13 @@ EXPORT_SYMBOL_GPL(device_match_devt);
int device_match_acpi_dev(struct device *dev, const void *adev) { - return ACPI_COMPANION(dev) == adev; + return adev && ACPI_COMPANION(dev) == adev; } EXPORT_SYMBOL(device_match_acpi_dev);
int device_match_acpi_handle(struct device *dev, const void *handle) { - return ACPI_HANDLE(dev) == handle; + return handle && ACPI_HANDLE(dev) == handle; } EXPORT_SYMBOL(device_match_acpi_handle);
On Thu, Dec 12, 2024 at 1:32 AM Brian Norris briannorris@chromium.org wrote:
of_find_device_by_node(), bus_find_device_by_of_node(), bus_find_device_by_fwnode(), ..., all produce arbitrary results when provided with a NULL of_node, fwnode, ACPI handle, etc. This is counterintuitive, and the source of a few bugs, such as the one fixed by commit 5c8418cf4025 ("PCI/pwrctrl: Unregister platform device only if one actually exists").
It's hard to imagine a good reason that these device_match_*() APIs should return 'true' for a NULL argument. Augment these to return 0 (false).
Signed-off-by: Brian Norris briannorris@chromium.org Reviewed-by: Rob Herring (Arm) robh@kernel.org
For the ACPI part
Acked-by: Rafael J. Wysocki rafael.j.wysocki@intel.com
Changes in v2:
- Add Rob's Reviewed-by
drivers/base/core.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/drivers/base/core.c b/drivers/base/core.c index 94865c9d8adc..2b7b13fc36d7 100644 --- a/drivers/base/core.c +++ b/drivers/base/core.c @@ -5246,13 +5246,13 @@ EXPORT_SYMBOL_GPL(device_match_name);
int device_match_of_node(struct device *dev, const void *np) {
return dev->of_node == np;
return np && dev->of_node == np;
} EXPORT_SYMBOL_GPL(device_match_of_node);
int device_match_fwnode(struct device *dev, const void *fwnode) {
return dev_fwnode(dev) == fwnode;
return fwnode && dev_fwnode(dev) == fwnode;
} EXPORT_SYMBOL_GPL(device_match_fwnode);
@@ -5264,13 +5264,13 @@ EXPORT_SYMBOL_GPL(device_match_devt);
int device_match_acpi_dev(struct device *dev, const void *adev) {
return ACPI_COMPANION(dev) == adev;
return adev && ACPI_COMPANION(dev) == adev;
} EXPORT_SYMBOL(device_match_acpi_dev);
int device_match_acpi_handle(struct device *dev, const void *handle) {
return ACPI_HANDLE(dev) == handle;
return handle && ACPI_HANDLE(dev) == handle;
} EXPORT_SYMBOL(device_match_acpi_handle);
-- 2.47.0.338.g60cca15819-goog
Per commit bebe94b53eb7 ("drivers: base: default KUNIT_* fragments to KUNIT_ALL_TESTS"), it seems like we should default to KUNIT_ALL_TESTS.
This enables these platform_device tests for common configurations, such as with: ./tools/testing/kunit/kunit.py run
Signed-off-by: Brian Norris briannorris@chromium.org ---
(no changes since v1)
drivers/base/test/Kconfig | 1 + 1 file changed, 1 insertion(+)
diff --git a/drivers/base/test/Kconfig b/drivers/base/test/Kconfig index 5c7fac80611c..2756870615cc 100644 --- a/drivers/base/test/Kconfig +++ b/drivers/base/test/Kconfig @@ -12,6 +12,7 @@ config TEST_ASYNC_DRIVER_PROBE config DM_KUNIT_TEST tristate "KUnit Tests for the device model" if !KUNIT_ALL_TESTS depends on KUNIT + default KUNIT_ALL_TESTS
config DRIVER_PE_KUNIT_TEST tristate "KUnit Tests for property entry API" if !KUNIT_ALL_TESTS
We recently updated these device_match*() (and therefore, various *find_device_by*()) functions to return a consistent 'false' value when trying to match a NULL handle. Add tests for this.
This provides regression-testing coverage for the sorts of bugs that underly commit 5c8418cf4025 ("PCI/pwrctrl: Unregister platform device only if one actually exists").
Signed-off-by: Brian Norris briannorris@chromium.org ---
Changes in v2: * Keep "devm" and "match" tests in separate suites
drivers/base/test/platform-device-test.c | 42 +++++++++++++++++++++++- 1 file changed, 41 insertions(+), 1 deletion(-)
diff --git a/drivers/base/test/platform-device-test.c b/drivers/base/test/platform-device-test.c index ea05b8785743..c8d4b0a385f2 100644 --- a/drivers/base/test/platform-device-test.c +++ b/drivers/base/test/platform-device-test.c @@ -3,6 +3,8 @@ #include <kunit/resource.h>
#include <linux/device.h> +#include <linux/device/bus.h> +#include <linux/of_platform.h> #include <linux/platform_device.h>
#define DEVICE_NAME "test" @@ -217,7 +219,45 @@ static struct kunit_suite platform_device_devm_test_suite = { .test_cases = platform_device_devm_tests, };
-kunit_test_suite(platform_device_devm_test_suite); +static void platform_device_find_by_null_test(struct kunit *test) +{ + struct platform_device *pdev; + int ret; + + pdev = platform_device_alloc(DEVICE_NAME, PLATFORM_DEVID_NONE); + KUNIT_ASSERT_NOT_ERR_OR_NULL(test, pdev); + + ret = platform_device_add(pdev); + KUNIT_ASSERT_EQ(test, ret, 0); + + KUNIT_EXPECT_PTR_EQ(test, of_find_device_by_node(NULL), NULL); + + KUNIT_EXPECT_PTR_EQ(test, bus_find_device_by_of_node(&platform_bus_type, NULL), NULL); + KUNIT_EXPECT_PTR_EQ(test, bus_find_device_by_fwnode(&platform_bus_type, NULL), NULL); + KUNIT_EXPECT_PTR_EQ(test, bus_find_device_by_acpi_dev(&platform_bus_type, NULL), NULL); + + KUNIT_EXPECT_FALSE(test, device_match_of_node(&pdev->dev, NULL)); + KUNIT_EXPECT_FALSE(test, device_match_fwnode(&pdev->dev, NULL)); + KUNIT_EXPECT_FALSE(test, device_match_acpi_dev(&pdev->dev, NULL)); + KUNIT_EXPECT_FALSE(test, device_match_acpi_handle(&pdev->dev, NULL)); + + platform_device_unregister(pdev); +} + +static struct kunit_case platform_device_match_tests[] = { + KUNIT_CASE(platform_device_find_by_null_test), + {} +}; + +static struct kunit_suite platform_device_match_test_suite = { + .name = "platform-device-match", + .test_cases = platform_device_match_tests, +}; + +kunit_test_suites( + &platform_device_devm_test_suite, + &platform_device_match_test_suite, +);
MODULE_DESCRIPTION("Test module for platform devices"); MODULE_AUTHOR("Maxime Ripard mripard@kernel.org");
Hi,
On Wed, Dec 11, 2024 at 04:31:41PM -0800, Brian Norris wrote:
We recently updated these device_match*() (and therefore, various *find_device_by*()) functions to return a consistent 'false' value when trying to match a NULL handle. Add tests for this.
This provides regression-testing coverage for the sorts of bugs that underly commit 5c8418cf4025 ("PCI/pwrctrl: Unregister platform device only if one actually exists").
Signed-off-by: Brian Norris briannorris@chromium.org
Changes in v2:
- Keep "devm" and "match" tests in separate suites
drivers/base/test/platform-device-test.c | 42 +++++++++++++++++++++++- 1 file changed, 41 insertions(+), 1 deletion(-)
diff --git a/drivers/base/test/platform-device-test.c b/drivers/base/test/platform-device-test.c index ea05b8785743..c8d4b0a385f2 100644 --- a/drivers/base/test/platform-device-test.c +++ b/drivers/base/test/platform-device-test.c @@ -3,6 +3,8 @@ #include <kunit/resource.h> #include <linux/device.h> +#include <linux/device/bus.h> +#include <linux/of_platform.h> #include <linux/platform_device.h> #define DEVICE_NAME "test" @@ -217,7 +219,45 @@ static struct kunit_suite platform_device_devm_test_suite = { .test_cases = platform_device_devm_tests, }; -kunit_test_suite(platform_device_devm_test_suite); +static void platform_device_find_by_null_test(struct kunit *test) +{
- struct platform_device *pdev;
- int ret;
- pdev = platform_device_alloc(DEVICE_NAME, PLATFORM_DEVID_NONE);
- KUNIT_ASSERT_NOT_ERR_OR_NULL(test, pdev);
- ret = platform_device_add(pdev);
- KUNIT_ASSERT_EQ(test, ret, 0);
I *think* you have a bug there: if platform_device_add fails, KUNIT_ASSERT will stop the test execution and thus you will leak the platform_device you just allocated.
You need to call platform_device_put in such a case, but if platform_device_add succeeds then you need to call platform_device_unregister instead.
It would be better to use kunit_platform_device_alloc and kunit_platform_device_add that already deal with this.
The rest looks good to me, once fixed: Reviewed-by: Maxime Ripard mripard@kernel.org
Maxime
Hi Maxime,
On Fri, Dec 13, 2024 at 12:59:57PM +0100, Maxime Ripard wrote:
On Wed, Dec 11, 2024 at 04:31:41PM -0800, Brian Norris wrote:
--- a/drivers/base/test/platform-device-test.c +++ b/drivers/base/test/platform-device-test.c
@@ -217,7 +219,45 @@ static struct kunit_suite platform_device_devm_test_suite = { .test_cases = platform_device_devm_tests, }; -kunit_test_suite(platform_device_devm_test_suite); +static void platform_device_find_by_null_test(struct kunit *test) +{
- struct platform_device *pdev;
- int ret;
- pdev = platform_device_alloc(DEVICE_NAME, PLATFORM_DEVID_NONE);
- KUNIT_ASSERT_NOT_ERR_OR_NULL(test, pdev);
- ret = platform_device_add(pdev);
- KUNIT_ASSERT_EQ(test, ret, 0);
I *think* you have a bug there: if platform_device_add fails, KUNIT_ASSERT will stop the test execution and thus you will leak the platform_device you just allocated.
You need to call platform_device_put in such a case, but if platform_device_add succeeds then you need to call platform_device_unregister instead.
Hehe, well I'm imitating the existing leaks in the other tests in this file, then ;) But admittedly, those are a little more complex, because the unregistration is actually part of the test flow.
It would be better to use kunit_platform_device_alloc and kunit_platform_device_add that already deal with this.
Cool, thanks, I'll use those in v3 for my new test.
The rest looks good to me, once fixed: Reviewed-by: Maxime Ripard mripard@kernel.org
Thanks for the tips and review.
Brian
linux-kselftest-mirror@lists.linaro.org