On Tue, May 09, 2023 at 06:57:19PM +0900, Masahisa Kojima wrote:
On Mon, 8 May 2023 at 18:44, Heinrich Schuchardt xypron.glpk@gmx.de wrote:
On 5/8/23 10:15, Masahisa Kojima wrote:
Hi Heinrich,
On Thu, 27 Apr 2023 at 15:09, Heinrich Schuchardt xypron.glpk@gmx.de wrote:
On 4/10/23 11:07, Masahisa Kojima wrote:
Current FMP->GetImageInfo() always return 0 for the firmware version, user can not identify which firmware version is currently running through the EFI interface.
This commit gets the version information from device tree, then fills the firmware version, lowest supported version in FMP->GetImageInfo().
Now FMP->GetImageInfo() and ESRT have the meaningful version number.
Signed-off-by: Masahisa Kojima masahisa.kojima@linaro.org
Changes in v5:
newly implement a device tree based versioning
.../firmware/firmware-version.txt | 25 ++++++++ lib/efi_loader/efi_firmware.c | 63 +++++++++++++++++-- 2 files changed, 84 insertions(+), 4 deletions(-) create mode 100644 doc/device-tree-bindings/firmware/firmware-version.txt
diff --git a/doc/device-tree-bindings/firmware/firmware-version.txt b/doc/device-tree-bindings/firmware/firmware-version.txt new file mode 100644 index 0000000000..6112de4a1d --- /dev/null +++ b/doc/device-tree-bindings/firmware/firmware-version.txt @@ -0,0 +1,25 @@ +firmware-version bindings +-------------------------------
+Required properties: +- image-type-id : guid for image blob type +- image-index : image index +- fw-version : firmware version +- lowest-supported-version : lowest supported version
+Example:
firmware-version {image1 {image-type-id = "09D7CF52-0720-4710-91D1-08469B7FE9C8";image-index = <1>;fw-version = <5>;lowest-supported-version = <3>;};image2 {image-type-id = "5A7021F5-FEF2-48B4-AABA-832E777418C0";image-index = <2>;fw-version = <10>;lowest-supported-version = <7>;};};diff --git a/lib/efi_loader/efi_firmware.c b/lib/efi_loader/efi_firmware.c index 93e2b01c07..1c6ef468bf 100644 --- a/lib/efi_loader/efi_firmware.c +++ b/lib/efi_loader/efi_firmware.c @@ -102,6 +102,56 @@ efi_status_t EFIAPI efi_firmware_set_package_info_unsupported( return EFI_EXIT(EFI_UNSUPPORTED); }
+/**
- efi_firmware_get_firmware_version - get firmware version from dtb
- @image_index: Image index
- @image_type_id: Image type id
- @fw_version: Pointer to store the version number
- @lsv: Pointer to store the lowest supported version
- Authenticate the capsule if authentication is enabled.
- The image pointer and the image size are updated in case of success.
- */
+void efi_firmware_get_firmware_version(u8 image_index,
efi_guid_t *image_type_id,u32 *fw_version, u32 *lsv)+{
const void *fdt = gd->fdt_blob;const fdt32_t *val;const char *guid_str;int len, offset, index;int parent;parent = fdt_subnode_offset(fdt, 0, "firmware-version");if (parent < 0)return;fdt_for_each_subnode(offset, fdt, parent) {efi_guid_t guid;guid_str = fdt_getprop(fdt, offset, "image-type-id", &len);if (!guid_str)continue;uuid_str_to_bin(guid_str, guid.b, UUID_STR_FORMAT_GUID);val = fdt_getprop(fdt, offset, "image-index", &len);if (!val)continue;index = fdt32_to_cpu(*val);if (!guidcmp(&guid, image_type_id) && index == image_index) {val = fdt_getprop(fdt, offset, "fw-version", &len);if (val)*fw_version = fdt32_to_cpu(*val);val = fdt_getprop(fdt, offset,"lowest-supported-version", &len);if (val)*lsv = fdt32_to_cpu(*val);}}+}
- /**
- efi_fill_image_desc_array - populate image descriptor array
- @image_info_size: Size of @image_info
@@ -148,13 +198,19 @@ static efi_status_t efi_fill_image_desc_array( *package_version_name = NULL; /* not supported */
for (i = 0; i < num_image_type_guids; i++) {Currently we define num_image_type_guids per board in a C file. Using the same line of code once per board makes no sense to me. Please, move the definition of that variable to lib/efi_loader/efi_firmware.c.
Sorry for the late reply.
num_image_type_guids is calculated with "ARRAY_SIZE(fw_images)", fw_images[] array is also defined in each board file, so we can not simply move num_image_type_guids into lib/efi_loader/efi_firmware.c.
Why can't we have
int num_image_type_guids = ARRAY_SIZE(fw_images);
in lib/efi_loader/efi_firmware.c?
At first thought, I thought it was a matter of abstraction.
But there is a compilation error when we expose fw_images[]. fw_images[] array is initialized in each board file, and sizeof() of the external fw_images[] array in lib/efi_loader/efi_firmware.c will cause compilation failure. We need to specify the array size when fw_images is exposed, for example: extern struct efi_fw_image fw_images[2];
but currently there is no method to pre-define the fw_images[] array size, it is board specific.
We can define the macro to indicate the array size or having sentinel in the fw_images[] array, but I think the current
I simply wonder if the value should be embedded in "struct efi_capsule_update_info". struct efi_capsule_update_info { const char *dfu_string; int num_images; <- added struct efi_fw_image *images; }; This is the best place because the value must match not only "images" but also (entries in) "dfu_string".
Even now, efi_fill_image_desc_array() tries to access "fw_images[]" via the exposed update_info variable. Beautiful, isn't it?
One more comment: uefi.rst doesn't mention anything about num_image_type_guids.
-Takahiro Akashi
implementation is simpler, I would like to keep the current implementation. Correct me if I'm wrong.
Thanks, Masahisa Kojima
Best regards
Heinrich
And fw_images[] array is abstracted by struct efi_capsule_update_info, so I think we should not extract the fw_images[] array.
u32 fw_version = 0;u32 lowest_supported_version = 0;These assignments should be in efi_firmware_get_firmware_version.
OK.
image_info[i].image_index = fw_array[i].image_index;Why did we ever introduce the field image_index? Please, eliminate it it as the GUID is always sufficient to identify an image.
This is derived from the UEFI specification. UEFI specification "23.1.2 EFI_FIRMWARE_MANAGEMENT_PROTOCOL.GetImageInfo()" requires ImageIndex and ImageTypeId(guid).
ImageIndex: A unique number identifying the firmware image within the device. The number is between 1 and DescriptorCount.
image_info[i].image_type_id = fw_array[i].image_type_id; image_info[i].image_id = fw_array[i].image_index; image_info[i].image_id_name = fw_array[i].fw_name;
image_info[i].version = 0; /* not supported */
efi_firmware_get_firmware_version(fw_array[i].image_index,&fw_array[i].image_type_id,&fw_version,&lowest_supported_version);This interface makes no sense to me. We expect images with specific GUIDs and should not care about images with other GUIDs that may additionally exist in the capsule.
So you must pass the expected GUID as input variable here.
I don't clearly understand this comment, but the expected GUID is fw_array[i].image_type_id.
image_info[i].version = fw_version; image_info[i].version_name = NULL; /* not supported */Please, add the missing functionality to efi_firmware_get_firmware_version().
Does it mean we need to support version_name? I can add a version_name in dtb.
Please, pass *image_info[i] to efi_firmware_get_firmware_version. That will simplify the code.
OK.
Thank you for your review.
Regards, Masahisa Kojima
Best regards
Heinrich
image_info[i].size = 0; image_info[i].attributes_supported =@@ -168,7 +224,7 @@ static efi_status_t efi_fill_image_desc_array( image_info[0].attributes_setting |= IMAGE_ATTRIBUTE_AUTHENTICATION_REQUIRED;
image_info[i].lowest_supported_image_version = 0;
image_info[i].lowest_supported_image_version = lowest_supported_version; image_info[i].last_attempt_version = 0; image_info[i].last_attempt_status = LAST_ATTEMPT_STATUS_SUCCESS; image_info[i].hardware_instance = 1;@@ -290,7 +346,6 @@ efi_status_t EFIAPI efi_firmware_get_image_info( descriptor_version, descriptor_count, descriptor_size, package_version, package_version_name);
return EFI_EXIT(ret);