On 10/28/24 17:52, Tom Rini wrote:
On Mon, Oct 28, 2024 at 05:47:08PM +0100, Heinrich Schuchardt wrote:
On 10/28/24 16:17, Tom Rini wrote:
On Mon, Oct 28, 2024 at 06:59:10AM +0100, Heinrich Schuchardt wrote:
On 10/22/24 14:00, Simon Glass wrote:
Add a simple app to use for testing. This is intended to do whatever it
needs to for testing purposes. For now it just prints a message and
exits boot services.
There was a considerable amount of discussion about whether it is OK to
call exit-boot-services and then return to U-Boot. This is not normally
done in a real application, since exit-boot-services is used to
completely disconnect from U-Boot. However, since this is a test, we
need to check the results of running the app, so returning is necessary.
It works correctly and it provides a nice model of how to test the EFI
code in a simple way.
Signed-off-by: Simon Glass <[email protected]>
---
(no changes since v7)
Changes in v7:
- Update commit message
lib/efi_loader/Kconfig | 10 ++++++
lib/efi_loader/Makefile | 1 +
lib/efi_loader/testapp.c | 68 ++++++++++++++++++++++++++++++++++++++++
3 files changed, 79 insertions(+)
create mode 100644 lib/efi_loader/testapp.c
diff --git a/lib/efi_loader/Kconfig b/lib/efi_loader/Kconfig
index 69b2c9360d8..6ced29da719 100644
--- a/lib/efi_loader/Kconfig
+++ b/lib/efi_loader/Kconfig
@@ -565,6 +565,16 @@ config BOOTEFI_HELLO_COMPILE
No additional space will be required in the resulting U-Boot binary
when this option is enabled.
+config BOOTEFI_TESTAPP_COMPILE
+ bool "Compile an EFI test app for testing"
+ default y
+ help
+ This compiles an app designed for testing. It is packed into an image
+ by the test.py testing frame in the setup_efi_image() function.
+
+ No additional space will be required in the resulting U-Boot binary
+ when this option is enabled.
+
endif
source "lib/efi/Kconfig"
diff --git a/lib/efi_loader/Makefile b/lib/efi_loader/Makefile
index 00d18966f9e..87131ab911d 100644
--- a/lib/efi_loader/Makefile
+++ b/lib/efi_loader/Makefile
@@ -20,6 +20,7 @@ apps-$(CONFIG_EFI_LOAD_FILE2_INITRD) += initrddump
ifeq ($(CONFIG_GENERATE_ACPI_TABLE),)
apps-y += dtbdump
endif
+apps-$(CONFIG_BOOTEFI_TESTAPP_COMPILE) += testapp
obj-$(CONFIG_CMD_BOOTEFI_HELLO) += helloworld_efi.o
obj-$(CONFIG_EFI_BOOTMGR) += efi_bootmgr.o
diff --git a/lib/efi_loader/testapp.c b/lib/efi_loader/testapp.c
new file mode 100644
index 00000000000..feb444c92e9
--- /dev/null
+++ b/lib/efi_loader/testapp.c
@@ -0,0 +1,68 @@
+// SPDX-License-Identifier: GPL-2.0+
This is not a valid SPDX identifier. Please, use GPL-2.0-or-later.
We should switch to the proper one here, but there are numerous examples
of this today.
+/*
+ * Hello world EFI application
+ *
+ * Copyright 2024 Google LLC
+ * Written by Simon Glass <[email protected]>
+ *
+ * This test program is used to test the invocation of an EFI application.
+ * It writes a few messages to the console and then exits boot services
+ */
+
+#include <efi_api.h>
+
+static const efi_guid_t loaded_image_guid = EFI_LOADED_IMAGE_PROTOCOL_GUID;
+
+static struct efi_system_table *systable;
+static struct efi_boot_services *boottime;
+static struct efi_simple_text_output_protocol *con_out;
+
+/**
+ * efi_main() - entry point of the EFI application.
+ *
+ * @handle: handle of the loaded image
+ * @systab: system table
+ * Return: status code
+ */
+efi_status_t EFIAPI efi_main(efi_handle_t handle,
+ struct efi_system_table *systab)
+{
+ struct efi_loaded_image *loaded_image;
+ efi_status_t ret;
+ efi_uintn_t map_size;
+ efi_uintn_t map_key;
+ efi_uintn_t desc_size;
+ u32 desc_version;
+
+ systable = systab;
+ boottime = systable->boottime;
+ con_out = systable->con_out;
+
+ /* Get the loaded image protocol */
+ ret = boottime->open_protocol(handle, &loaded_image_guid,
+ (void **)&loaded_image, NULL, NULL,
+ EFI_OPEN_PROTOCOL_GET_PROTOCOL);
+ if (ret != EFI_SUCCESS) {
+ con_out->output_string
+ (con_out, u"Cannot open loaded image protocol\r\n");
+ goto out;
+ }
+
+ /* UEFI requires CR LF */
+ con_out->output_string(con_out, u"U-Boot test app for EFI_LOADER\r\n");
+
+out:
+ map_size = 0;
+ ret = boottime->get_memory_map(&map_size, NULL, &map_key, &desc_size,
+ &desc_version);
+ con_out->output_string(con_out, u"Exiting boot sevices\n");
%s/sevices/services/g
+
+ /* exit boot services so that this part of U-Boot can be tested */
+ boottime->exit_boot_services(handle, map_key);
+
+ /* now exit for real */
+ ret = boottime->exit(handle, ret, 0, NULL);
As written before boot services are not available after
ExitBootServices(). Please, call the ResetSystem() service.
Right, but the point is to not do that and instead test what happens. I
think Ilias had said we need to make some big loud print on console that
you must reset the system for it to be usable afterwards, would that
be enough?
There is no point in testing what happens as this call is not allowable.
Do we really need to zero out the bootservices table for people who do
not read the spec to find out the hard way?
Then perhaps this is a case where "test causes U-Boot to reset, check
for that as the result" is what the test needs to do, as it's not an
arbitrary reset?
Yes, we should check that a reset occurs after ResetSystem(). Currently
we are only testing that the EFI watchdog resets the system in
test_efi_selftest_watchdog_reboot().
As on all non-sandbox systems all devices are removed at
ExitBootServices() this is the only way forward to write a test that
works on all UEFI supporting U-Boot systems.
Best regards
Heinrich
