[PATCH 2/2] efi_loader: add PE/COFF image measurement
Masahisa Kojima
masahisa.kojima at linaro.org
Tue Apr 27 16:05:47 CEST 2021
Hi Heinrich,
> > +++ b/include/tpm-v2.h
> > @@ -61,6 +61,7 @@ struct udevice;
> > #define EV_S_CRTM_VERSION ((u32)0x00000008)
> > #define EV_CPU_MICROCODE ((u32)0x00000009)
> > #define EV_TABLE_OF_DEVICES ((u32)0x0000000B)
>
> Please, add a comment here that the following values are defined in the
> "TCG EFI Platform Specification".
>
> > +#define EV_EFI_BOOT_SERVICES_APPLICATION ((u32)0x80000003)
>
> Please, add all EV_EFI_* constants.
I have sent v2 patch and added EV_EFI_* constants.
Anyway, EV_EFI_* are defined in the "TCG PC Client Platform Firmware
Profile Specification" and it is better to add remaining events
described in "Table 9 Event" as a separate patch.
Best Regards,
Masahisa
On Thu, 22 Apr 2021 at 17:23, Heinrich Schuchardt <xypron.glpk at gmx.de> wrote:
>
> On 22.04.21 10:09, Ilias Apalodimas wrote:
> >>>> + if (!(active & alg_to_mask(hash_alg)))
> >>>> + continue;
> >>>> + switch (hash_alg) {
> >>>> + case TPM2_ALG_SHA1:
> >>>
> >>> SHA1 is known to be unsafe. Why would we support it?
> >>
> >> Basically I agree with removing SHA1 support.
> >> This efi_tcg2.c implementation aims to support TCG v2, so there is no
> >> reason to keep SHA1.
> >> Anyway, SHA1 is supported in tcg2_create_digest() for the measurement
> >> other than PE/COFF image. Do we also remove SHA1 from
> >> tcg2_create_digest()?
> >>
> >
> > The hardware dictates what kind of SHAxxx you are supposed to add in the
> > EventLog and the PCRs. Why would we remove the functionality? If someone
> > considers SHA1 unsafe, he can just disable it from his hardware and remove it
> > from the active algorithms.
> >
> >
> > Cheers
> > /Ilias
>
>
> The TCG EFI ProtocolSpecification explicitely enumerates the four
> hashing algorithms of Masahisa's patch, see chapter 6.4.3, "Related
> Definitions".
>
> So let's support them.
>
> Best regards
>
> Heinrich
>
> >
> >> For other comments, I will modify the code and send v2 patch.
> >>
> >> Thanks,
> >> Masahisa
> >>
> >>
> >> On Wed, 21 Apr 2021 at 19:57, Heinrich Schuchardt <xypron.glpk at gmx.de> wrote:
> >>>
> >>> On 4/15/21 3:30 PM, Masahisa Kojima wrote:
> >>>> "TCG PC Client Platform Firmware Profile Specification"
> >>>> requires to measure every attempt to load and execute
> >>>> a OS Loader(a UEFI application) into PCR[4].
> >>>> This commit adds the PE/COFF image measurement, extends PCR,
> >>>> and appends measurement into Event Log.
> >>>>
> >>>> Signed-off-by: Masahisa Kojima <masahisa.kojima at linaro.org>
> >>>> ---
> >>>> include/efi_loader.h | 4 +
> >>>> include/efi_tcg2.h | 10 ++
> >>>> include/tpm-v2.h | 1 +
> >>>> lib/efi_loader/efi_image_loader.c | 7 ++
> >>>> lib/efi_loader/efi_tcg2.c | 187 ++++++++++++++++++++++++++++--
> >>>> 5 files changed, 199 insertions(+), 10 deletions(-)
> >>>>
> >>>> diff --git a/include/efi_loader.h b/include/efi_loader.h
> >>>> index de1a496a97..b02bc93c8e 100644
> >>>> --- a/include/efi_loader.h
> >>>> +++ b/include/efi_loader.h
> >>>> @@ -426,6 +426,10 @@ efi_status_t efi_disk_register(void);
> >>>> efi_status_t efi_rng_register(void);
> >>>> /* Called by efi_init_obj_list() to install EFI_TCG2_PROTOCOL */
> >>>> efi_status_t efi_tcg2_register(void);
> >>>> +/* measure the pe-coff image, extend PCR and add Event Log */
> >>>> +efi_status_t tcg2_measure_pe_image(void *efi, u64 efi_size,
> >>>> + struct efi_loaded_image_obj *handle,
> >>>> + struct efi_loaded_image *loaded_image_info);
> >>>> /* Create handles and protocols for the partitions of a block device */
> >>>> int efi_disk_create_partitions(efi_handle_t parent, struct blk_desc *desc,
> >>>> const char *if_typename, int diskid,
> >>>> diff --git a/include/efi_tcg2.h b/include/efi_tcg2.h
> >>>> index 40e241ce31..f8d46c5fd2 100644
> >>>> --- a/include/efi_tcg2.h
> >>>> +++ b/include/efi_tcg2.h
> >>>> @@ -9,6 +9,8 @@
> >>>> #if !defined _EFI_TCG2_PROTOCOL_H_
> >>>> #define _EFI_TCG2_PROTOCOL_H_
> >>>>
> >>>> +#include <efi.h>
> >>>
> >>> This include is already included in efi_api.h.
> >>>
> >>>> +#include <efi_api.h>
> >>>> #include <tpm-v2.h>
> >>>>
> >>>> #define EFI_TCG2_PROTOCOL_GUID \
> >>>> @@ -53,6 +55,14 @@ struct efi_tcg2_event {
> >>>> u8 event[];
> >>>> } __packed;
> >>>>
> >>>> +struct uefi_image_load_event {
> >>>> + efi_physical_addr_t image_location_in_memory;
> >>>> + u64 image_length_in_memory;
> >>>> + u64 image_link_time_address;
> >>>> + u64 length_of_device_path;
> >>>> + struct efi_device_path device_path[];
> >>>
> >>> A device path is not an array of struct efi_device_path. But the first
> >>> element is of this type. So ok.
> >>>
> >>>> +} __packed;
> >>>
> >>> Why should this be __packed? You don't use arrays of this structure and
> >>> it is naturally packed.
> >>>
> >>>> +
> >>>> struct efi_tcg2_boot_service_capability {
> >>>> u8 size;
> >>>> struct efi_tcg2_version structure_version;
> >>>> diff --git a/include/tpm-v2.h b/include/tpm-v2.h
> >>>> index df67a196cf..ab9c04dc0a 100644
> >>>> --- a/include/tpm-v2.h
> >>>> +++ b/include/tpm-v2.h
> >>>> @@ -61,6 +61,7 @@ struct udevice;
> >>>> #define EV_S_CRTM_VERSION ((u32)0x00000008)
> >>>> #define EV_CPU_MICROCODE ((u32)0x00000009)
> >>>> #define EV_TABLE_OF_DEVICES ((u32)0x0000000B)
> >>>
> >>> Please, add a comment here that the following values are defined in the
> >>> "TCG EFI Platform Specification".
> >>>
> >>>> +#define EV_EFI_BOOT_SERVICES_APPLICATION ((u32)0x80000003)
> >>>
> >>> Please, add all EV_EFI_* constants.
> >>>
> >>>>
> >>>> /* TPMS_TAGGED_PROPERTY Structure */
> >>>> struct tpms_tagged_property {
> >>>> diff --git a/lib/efi_loader/efi_image_loader.c b/lib/efi_loader/efi_image_loader.c
> >>>> index 2c35cb5651..b032ec5dd8 100644
> >>>> --- a/lib/efi_loader/efi_image_loader.c
> >>>> +++ b/lib/efi_loader/efi_image_loader.c
> >>>> @@ -829,6 +829,13 @@ efi_status_t efi_load_pe(struct efi_loaded_image_obj *handle,
> >>>> goto err;
> >>>> }
> >>>>
> >>>> +#if CONFIG_IS_ENABLED(EFI_TCG2_PROTOCOL)
> >>>> + /* Measure an PE/COFF image */
> >>>> + if (tcg2_measure_pe_image(efi, efi_size, handle,
> >>>> + loaded_image_info))
> >>>> + log_err("PE image measurement failed\n");
> >>>> +#endif
> >>>> +
> >>>> /* Copy PE headers */
> >>>> memcpy(efi_reloc, efi,
> >>>> sizeof(*dos)
> >>>> diff --git a/lib/efi_loader/efi_tcg2.c b/lib/efi_loader/efi_tcg2.c
> >>>> index ed86a220fb..9fab07605f 100644
> >>>> --- a/lib/efi_loader/efi_tcg2.c
> >>>> +++ b/lib/efi_loader/efi_tcg2.c
> >>>> @@ -13,8 +13,10 @@
> >>>> #include <efi_loader.h>
> >>>> #include <efi_tcg2.h>
> >>>> #include <log.h>
> >>>> +#include <malloc.h>
> >>>> #include <version.h>
> >>>> #include <tpm-v2.h>
> >>>> +#include <u-boot/rsa.h>
> >>>> #include <u-boot/sha1.h>
> >>>> #include <u-boot/sha256.h>
> >>>> #include <u-boot/sha512.h>
> >>>> @@ -709,6 +711,172 @@ out:
> >>>> return EFI_EXIT(ret);
> >>>> }
> >>>>
> >>>> +static efi_status_t tcg2_hash_pe_image(void *efi, u64 efi_size,
> >>>> + struct tpml_digest_values *digest_list)
> >>>> +{
> >>>> + IMAGE_NT_HEADERS32 *nt;
> >>>> + WIN_CERTIFICATE *wincerts = NULL;
> >>>> + size_t wincerts_len;
> >>>> + struct efi_image_regions *regs = NULL;
> >>>> + void *new_efi = NULL;
> >>>> + size_t new_efi_size;
> >>>> + u8 hash[TPM2_SHA512_DIGEST_SIZE];
> >>>> + efi_status_t ret;
> >>>> + u32 active;
> >>>> + int i;
> >>>> +
> >>>> + ret = efi_check_pe(efi, efi_size, (void **)&nt);
> >>> Why are you calling this function? It is already called in efi_load_pe().
> >>>
> >>>> + if (ret != EFI_SUCCESS) {
> >>>> + log_err("Not a valid PE-COFF file\n");
> >>>> + return EFI_INVALID_PARAMETER;
> >>>> + }
> >>>> +
> >>>> + /*
> >>>> + * Size must be 8-byte aligned and the trailing bytes must be
> >>>> + * zero'ed. Otherwise hash value may be incorrect.
> >>>> + */
> >>>> + if (!IS_ALIGNED(efi_size, 8)) {
> >>>> + new_efi_size = ALIGN(efi_size, 8);
> >>>> + new_efi = calloc(new_efi_size, 1);
> >>>> + if (!new_efi)
> >>>> + return EFI_OUT_OF_RESOURCES;
> >>>> + memcpy(new_efi, efi, efi_size);
> >>>> + efi = new_efi;
> >>>> + efi_size = new_efi_size;
> >>>> + }
> >>>> +
> >>>> + if (!efi_image_parse(efi, efi_size, ®s, &wincerts,
> >>>> + &wincerts_len)) {
> >>>> + log_err("Parsing PE executable image failed\n");
> >>>> + ret = EFI_INVALID_PARAMETER;
> >>>> + goto out;
> >>>> + }
> >>>
> >>> Please, don't duplicate code from efi_image_authenticate(). Extract a
> >>> common function instead.
> >>>
> >>>> +
> >>>> + ret = __get_active_pcr_banks(&active);
> >>>> + if (ret != EFI_SUCCESS) {
> >>>> + ret = EFI_DEVICE_ERROR;
> >>>> + goto out;
> >>>> + }
> >>>> +
> >>>> + digest_list->count = 0;
> >>>> + for (i = 0; i < MAX_HASH_COUNT; i++) {
> >>>> + u16 hash_alg = hash_algo_list[i].hash_alg;
> >>>> +
> >>>> + if (!(active & alg_to_mask(hash_alg)))
> >>>> + continue;
> >>>> + switch (hash_alg) {
> >>>> + case TPM2_ALG_SHA1:
> >>>
> >>> SHA1 is known to be unsafe. Why would we support it?
> >>>
> >>>> + hash_calculate("sha1", regs->reg, regs->num, hash);
> >>>> + digest_list->count++;
> >>>
> >>> Why do you repeat this line in every case of the switch statement?
> >>> Please, put it below.
> >>>
> >>>> + break;
> >>>> + case TPM2_ALG_SHA256:
> >>>> + hash_calculate("sha256", regs->reg, regs->num, hash);
> >>>> + digest_list->count++;
> >>>> + break;
> >>>> + case TPM2_ALG_SHA384:
> >>>> + hash_calculate("sha384", regs->reg, regs->num, hash);
> >>>> + digest_list->count++;
> >>>> + break;
> >>>> + case TPM2_ALG_SHA512:
> >>>> + hash_calculate("sha512", regs->reg, regs->num, hash);
> >>>> + digest_list->count++;
> >>>> + break;
> >>>> + default:
> >>>> + EFI_PRINT("Unsupported algorithm %x\n", hash_alg);
> >>>> + return EFI_INVALID_PARAMETER;
> >>>> + }
> >>>> + digest_list->digests[i].hash_alg = hash_alg;
> >>>> + memcpy(&digest_list->digests[i].digest, hash, (u32)alg_to_len(hash_alg));
> >>>> + }
> >>>> +
> >>>> +out:
> >>>> + free(new_efi);
> >>>> + free(regs);
> >>>> +
> >>>> + return ret;
> >>>> +}
> >>>> +
> >>>> +efi_status_t tcg2_measure_pe_image(void *efi, u64 efi_size,
> >>>> + struct efi_loaded_image_obj *handle,
> >>>> + struct efi_loaded_image *loaded_image)
> >>>> +{
> >>>> + struct tpml_digest_values digest_list;
> >>>> + efi_status_t ret;
> >>>> + struct udevice *dev;
> >>>> + u32 pcr_index, event_type, event_size;
> >>>> + struct uefi_image_load_event *image_load_event;
> >>>> + u8 *event;
> >>>
> >>> The variable event is not needed. You can use image_load_event directly.
> >>>
> >>>> + struct efi_device_path *device_path;
> >>>> + u32 device_path_length;
> >>>> + IMAGE_DOS_HEADER *dos;
> >>>> + IMAGE_NT_HEADERS32 *nt;
> >>>> +
> >>>> + ret = platform_get_tpm2_device(&dev);
> >>>> + if (ret != EFI_SUCCESS)
> >>>> + return ret;
> >>>> +
> >>>> + if (handle->image_type == IMAGE_SUBSYSTEM_EFI_APPLICATION) {
> >>>
> >>> We should measure drivers too. Cf.
> >>> EV_EFI_BOOT_SERVICES_DRIVER, EV_EFI_RUNTIME_SERVICES_DRIVER.
> >>>
> >>>> + pcr_index = 4;
> >>>> + event_type = EV_EFI_BOOT_SERVICES_APPLICATION;
> >>>> + } else {
> >>>> + return EFI_UNSUPPORTED;
> >>>> + }
> >>>> +
> >>>> + ret = tcg2_hash_pe_image(efi, efi_size, &digest_list);
> >>>> + if (ret != EFI_SUCCESS)
> >>>> + return ret;
> >>>> +
> >>>> + ret = tcg2_pcr_extend(dev, pcr_index, &digest_list);
> >>>> + if (ret != EFI_SUCCESS)
> >>>> + return ret;
> >>>> +
> >>>> + loaded_image->system_table->boottime->handle_protocol(&handle->header,
> >>>> + &efi_guid_loaded_image_device_path,
> >>>> + (void **)&device_path);
> >>>
> >>> You would have to use EFI_CALL() here.
> >>>
> >>> Please, use efi_search_protocol() instead of all this indirection.
> >>>
> >>> Best regards
> >>>
> >>> Heinrich
> >>>
> >>>> + device_path_length = efi_dp_size(device_path);
> >>>> + if (device_path_length > 0) {
> >>>> + /* add end node size */
> >>>> + device_path_length += sizeof(struct efi_device_path);
> >>>> + }
> >>>> + event_size = sizeof(struct uefi_image_load_event) + device_path_length;
> >>>> + event = malloc(event_size);
> >>>> + if (!event)
> >>>> + return EFI_OUT_OF_RESOURCES;
> >>>> +
> >>>> + image_load_event = (struct uefi_image_load_event *)event;
> >>>> + image_load_event->image_location_in_memory = (efi_physical_addr_t)efi;
> >>>> + image_load_event->image_length_in_memory = efi_size;
> >>>> + image_load_event->length_of_device_path = device_path_length;
> >>>> +
> >>>> + dos = (IMAGE_DOS_HEADER *)efi;
> >>>> + nt = (IMAGE_NT_HEADERS32 *)(efi + dos->e_lfanew);
> >>>> + if (nt->OptionalHeader.Magic == IMAGE_NT_OPTIONAL_HDR64_MAGIC) {
> >>>> + IMAGE_NT_HEADERS64 *nt64 = (IMAGE_NT_HEADERS64 *)nt;
> >>>> +
> >>>> + image_load_event->image_link_time_address =
> >>>> + nt64->OptionalHeader.ImageBase;
> >>>> + } else if (nt->OptionalHeader.Magic == IMAGE_NT_OPTIONAL_HDR32_MAGIC) {
> >>>> + image_load_event->image_link_time_address =
> >>>> + nt->OptionalHeader.ImageBase;
> >>>> + } else {
> >>>> + ret = EFI_INVALID_PARAMETER;
> >>>> + goto out;
> >>>> + }
> >>>> +
> >>>> + if (device_path_length > 0) {
> >>>> + memcpy(image_load_event->device_path, device_path,
> >>>> + device_path_length);
> >>>> + }
> >>>> +
> >>>> + ret = tcg2_agile_log_append(pcr_index, event_type, &digest_list,
> >>>> + event_size, event);
> >>>> +
> >>>> +out:
> >>>> + free(event);
> >>>> +
> >>>> + return ret;
> >>>> +}
> >>>> +
> >>>> /**
> >>>> * efi_tcg2_hash_log_extend_event() - extend and optionally log events
> >>>> *
> >>>> @@ -761,24 +929,23 @@ efi_tcg2_hash_log_extend_event(struct efi_tcg2_protocol *this, u64 flags,
> >>>> /*
> >>>> * if PE_COFF_IMAGE is set we need to make sure the image is not
> >>>> * corrupted, verify it and hash the PE/COFF image in accordance with
> >>>> - * the procedure specified in "Calculating the PE Image Hash"
> >>>> - * section of the "Windows Authenticode Portable Executable Signature
> >>>> + * the procedure specified in "Calculating the PE Image Hash"
> >>>> + * section of the "Windows Authenticode Portable Executable Signature
> >>>> * Format"
> >>>> - * Not supported for now
> >>>> */
> >>>> if (flags & PE_COFF_IMAGE) {
> >>>> - ret = EFI_UNSUPPORTED;
> >>>> - goto out;
> >>>> + ret = tcg2_hash_pe_image((void *)data_to_hash, data_to_hash_len,
> >>>> + &digest_list);
> >>>> + } else {
> >>>> + ret = tcg2_create_digest((u8 *)data_to_hash, data_to_hash_len,
> >>>> + &digest_list);
> >>>> }
> >>>> + if (ret != EFI_SUCCESS)
> >>>> + goto out;
> >>>>
> >>>> pcr_index = efi_tcg_event->header.pcr_index;
> >>>> event_type = efi_tcg_event->header.event_type;
> >>>>
> >>>> - ret = tcg2_create_digest((u8 *)data_to_hash, data_to_hash_len,
> >>>> - &digest_list);
> >>>> - if (ret != EFI_SUCCESS)
> >>>> - goto out;
> >>>> -
> >>>> ret = tcg2_pcr_extend(dev, pcr_index, &digest_list);
> >>>> if (ret != EFI_SUCCESS)
> >>>> goto out;
> >>>>
> >>>
>
More information about the U-Boot
mailing list