Skip to contentSkip to conversation

PCI: Initial KUnit test fixture for resource assignment

PATCH Superseded Naveen Naidu 2 participants 3 messages Nov 18, 2021 PCI

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

v3 cover PCI: Initial KUnit test fixture for resource assignment

Currently it's hard to deubg issues in the resource assignment code of the PCI because of the long reproduction cycles between the developer trying to fix the code and the user testing it due to the lack of hardware device with the developer [1].

[1]:
https://lore.kernel.org/all/20210621123714.GA3286648@bjorn-Precision-5520/

Bjorn, suggested that it would be really good if we could have a test fixture for debugging/testing resource assignment.

The patch attached along with the cover letter is an attempt to lay the foundation and also have a proof of concept to show that it is possible to have a test fixture to debug the resource assignment code.

Since there are a lot of things which happens during the resource assignement phase, the first version only tests the __pci_read_base() function since that was the most easiest to set up.

Hopefully, in the future patches I'll be able to write more KUnit tests for the other parts responsible during the resource assignment phase and get closer to the goal of having a complete test fixtures :)

Thanks,
Naveen

Changelog
=========

v3:
- Rename init_registers to init_test_registers
v2:
- Add test cases to test resource assignment for Type 1 devices
- Fix a error (a function was not static) found by Kernel Test Robot

Naveen Naidu (1):
[PATCH v3 1/1] PCI: Add KUnit tests for __pci_read_base()

 drivers/pci/Kconfig              |   4 +
 drivers/pci/Makefile             |   3 +
 drivers/pci/pci-read-base-test.c | 803 +++++++++++++++++++++++++++++++
 3 files changed, 810 insertions(+)
 create mode 100644 drivers/pci/pci-read-base-test.c
Signature · 1 line
2.25.1
#1 lore raw

PATCH v3 1/1 PCI: Add KUnit tests for __pci_read_base()

Currently it is hard to debug issues in the resource assignment code due to long reporduction cycles between the developer trying to fix the code and the user testing it due to the lack of hardware device with the developer [1].

[1]:
https://lore.kernel.org/all/20210621123714.GA3286648@bjorn-Precision-5520/

This adds KUnit tests for __pci_read_base() which is only dependent on software structures, so no hardware is needed to run these.

This lays the foundation for test fixtures we can use to reproduce the resource assignment code path of PCI.

Sample output from KUnit Test run:

      # Subtest: __pci_read_base()
      1..3
      # test_pci_read_base_type_0_hdr_approach_1: initializing __pci_read_base() tests
   (null): reg 0x18: [mem 0x4f400000-0x4f400fff]
      ok 1 - test_pci_read_base_type_0_hdr_approach_1
      # test_pci_read_base_type_0_hdr_approach_2: initializing __pci_read_base() tests
   (null): reg 0x18: [mem 0x4f400000-0x4f400fff]
   (null): reg 0x1c: [mem 0xaf400000-0xaf4000ff]
      ok 2 - test_pci_read_base_type_0_hdr_approach_2
      # test_pci_read_base_type_1_hdr: initializing __pci_read_base() tests
   (null): reg 0x10: [mem 0xaf400000-0xaf4000ff]
      ok 3 - test_pci_read_base_type_1_hdr
  # __pci_read_base(): pass:3 fail:0 skip:0 total:3
  # Totals: pass:3 fail:0 skip:0 total:3
  # ok 8 - __pci_read_base()
---
 drivers/pci/Kconfig              |   4 +
 drivers/pci/Makefile             |   3 +
 drivers/pci/pci-read-base-test.c | 803 +++++++++++++++++++++++++++++++
 3 files changed, 810 insertions(+)
 create mode 100644 drivers/pci/pci-read-base-test.c
drivers/pci/Kconfig +4 -0
index 43e615aa12ff..12b3779fb640 100644
@@ -252,6 +252,10 @@ config PCIE_BUS_PEER2PEER
 
 endchoice
 
+config PCI_READ_BASE_KUNIT_TEST
+	tristate "KUnit tests for __pci_read_base() in probe.c"
+	depends on PCI && KUNIT=y
+
 source "drivers/pci/hotplug/Kconfig"
 source "drivers/pci/controller/Kconfig"
 source "drivers/pci/endpoint/Kconfig"
drivers/pci/Makefile +3 -0
index d62c4ac4ae1b..010a903c3d5d 100644
@@ -36,4 +36,7 @@ obj-$(CONFIG_PCI_ENDPOINT)	+= endpoint/
 obj-y				+= controller/
 obj-y				+= switch/
 
+# KUnit test files
+obj-$(CONFIG_PCI_READ_BASE_KUNIT_TEST) += pci-read-base-test.o
+
 subdir-ccflags-$(CONFIG_PCI_DEBUG) := -DDEBUG
drivers/pci/pci-read-base-test.c new file +803 -0

803 lines not on this page · raw message

Signature · 1 line
2.25.1
#2 lore raw

Re: [PATCH v3 1/1] PCI: Add KUnit tests for __pci_read_base()

Quoted 87 lines · Naveen Naidu

Currently it is hard to debug issues in the resource assignment code due to long reporduction cycles between the developer trying to fix the code and the user testing it due to the lack of hardware device with the developer [1].

[1]:
https://lore.kernel.org/all/20210621123714.GA3286648@bjorn-Precision-5520/

This adds KUnit tests for __pci_read_base() which is only dependent on software structures, so no hardware is needed to run these.

This lays the foundation for test fixtures we can use to reproduce the resource assignment code path of PCI.

Sample output from KUnit Test run:

      # Subtest: __pci_read_base()
      1..3
      # test_pci_read_base_type_0_hdr_approach_1: initializing __pci_read_base() tests
   (null): reg 0x18: [mem 0x4f400000-0x4f400fff]
      ok 1 - test_pci_read_base_type_0_hdr_approach_1
      # test_pci_read_base_type_0_hdr_approach_2: initializing __pci_read_base() tests
   (null): reg 0x18: [mem 0x4f400000-0x4f400fff]
   (null): reg 0x1c: [mem 0xaf400000-0xaf4000ff]
      ok 2 - test_pci_read_base_type_0_hdr_approach_2
      # test_pci_read_base_type_1_hdr: initializing __pci_read_base() tests
   (null): reg 0x10: [mem 0xaf400000-0xaf4000ff]
      ok 3 - test_pci_read_base_type_1_hdr
  # __pci_read_base(): pass:3 fail:0 skip:0 total:3
  # Totals: pass:3 fail:0 skip:0 total:3
  # ok 8 - __pci_read_base()
---
 drivers/pci/Kconfig              |   4 +
 drivers/pci/Makefile             |   3 +
 drivers/pci/pci-read-base-test.c | 803 +++++++++++++++++++++++++++++++
 3 files changed, 810 insertions(+)
 create mode 100644 drivers/pci/pci-read-base-test.c
drivers/pci/Kconfig +4 -0
index 43e615aa12ff..12b3779fb640 100644
@@ -252,6 +252,10 @@ config PCIE_BUS_PEER2PEER
 
 endchoice
 
+config PCI_READ_BASE_KUNIT_TEST
+	tristate "KUnit tests for __pci_read_base() in probe.c"
+	depends on PCI && KUNIT=y
+
 source "drivers/pci/hotplug/Kconfig"
 source "drivers/pci/controller/Kconfig"
 source "drivers/pci/endpoint/Kconfig"
drivers/pci/Makefile +3 -0
index d62c4ac4ae1b..010a903c3d5d 100644
@@ -36,4 +36,7 @@ obj-$(CONFIG_PCI_ENDPOINT)	+= endpoint/
 obj-y				+= controller/
 obj-y				+= switch/
 
+# KUnit test files
+obj-$(CONFIG_PCI_READ_BASE_KUNIT_TEST) += pci-read-base-test.o
+
 subdir-ccflags-$(CONFIG_PCI_DEBUG) := -DDEBUG
drivers/pci/pci-read-base-test.c new file +13 -0
new file mode 100644
index 000000000000..df89d50b0321
@@ -0,0 +1,803 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * KUnit tests for __pci_read_base()
+ *
+ * Author: Naveen Naidu <[email protected]>
+ */
+#include <kunit/test.h>
+#include <linux/pci.h>
+#include <linux/math.h>
+#include <linux/errno.h>
+
+#include "pci.h"
+

drivers/pci/pci-read-base-test.c line 23

Quoted 10 lines
drivers/pci/pci-read-base-test.c new file +10 -0
+#define MY_PCI_BUS_NUM 0x011
+#define NUM_32_BITCONFIG_REGISTERS 16
+
+/*
+ * Representation of type 0/1 headers.
+ *
+ * Each element of the array represents one 32 bit register of the
+ * Type 0/1 Header register.
+ */
+u32 config_registers[NUM_32_BITCONFIG_REGISTERS];

static?

drivers/pci/pci-read-base-test.c line 26

Quoted 3 lines
drivers/pci/pci-read-base-test.c new file +3 -0
+
+/* Type of the device you are testing */
+unsigned int type_header_test_case;

static?

Quoted 45 lines
drivers/pci/pci-read-base-test.c new file +45 -0
+
+/* Type 0/1 Header register values */
+struct config_space_bitfield {
+	char name[64];
+	unsigned int offset;
+	unsigned int size; /* In bytes */
+	unsigned int bit_array_index;
+	unsigned int value;
+};
+
+/*
+ * The index value of BARS in the type_0/1_header struct.
+ * Useful for setting the values for test cases.
+ */
+enum type_header_BAR_index {
+	BAR0 = 3,
+	BAR1,
+	BAR2,
+	BAR3,
+	BAR4,
+	BAR5
+};
+
+/* PCI Type 0 Header for Endpoints */
+static struct config_space_bitfield type_0_header[] = {
+	{"Vendor ID",	PCI_VENDOR_ID,		2,	0,	0},
+	{"Device ID",	PCI_DEVICE_ID,		2,	16,	0},
+	{"Command",	PCI_COMMAND,		2,	32,	0},
+	{"BAR 0",	PCI_BASE_ADDRESS_0,	4,	128,	0},
+	{"BAR 1",	PCI_BASE_ADDRESS_1,	4,	160,	0},
+	{"BAR 2",	PCI_BASE_ADDRESS_2,	4,	192,	0},
+	{"BAR 3",	PCI_BASE_ADDRESS_3,	4,	224,	0},
+	{"BAR 4",	PCI_BASE_ADDRESS_4,	4,	256,	0},
+	{"BAR 5",	PCI_BASE_ADDRESS_5,	4,	288,	0}
+};
+
+/* PCI Type 1 Header for Bridges */
+static struct config_space_bitfield type_1_header[] = {
+	{"Vendor ID",	PCI_VENDOR_ID,		2,	0,	0},
+	{"Device ID",	PCI_DEVICE_ID,		2,	16,	0},
+	{"Command",	PCI_COMMAND,		2,	32,	0},
+	{"BAR 0",	PCI_BASE_ADDRESS_0,	4,	128,	0},
+	{"BAR 1",	PCI_BASE_ADDRESS_1,	4,	160,	0}
+};
+

drivers/pci/pci-read-base-test.c line 81

Quoted 10 lines
drivers/pci/pci-read-base-test.c new file +10 -0
+/*
+ * -----------------------------------------------------------------------
+ * Data structures to hold test cases values
+ * -----------------------------------------------------------------------
+ */
+
+/* Used for setting the test values of BAR registers. */
+struct config_BARS {
+	char name[64];
+	unsigned int offset;

u32?

drivers/pci/pci-read-base-test.c line 90

Quoted 9 lines
drivers/pci/pci-read-base-test.c new file +9 -0
+	/*
+	 * Allocated Address size.
+	 *
+	 * Value to return, when all 1's are written to BAR registers.
+	 *
+	 * See (NOTE: How to calculate the allocated address size for the BAR
+	 * registers) comment below.
+	 */
+	unsigned int allocated_size;

u32?