|
From: | Stefan Berger |
Subject: | Re: [PATCH v3 1/4] acpi: Convert build_tpm2() to build_append* API |
Date: | Tue, 2 Jun 2020 10:24:03 -0400 |
User-agent: | Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101 Thunderbird/68.4.1 |
On 6/2/20 9:55 AM, Auger Eric wrote:
Hi Stefan, On 6/2/20 3:30 PM, Stefan Berger wrote:On 6/1/20 5:57 AM, Eric Auger wrote:In preparation of its move to the generic acpi code, let's convert build_tpm2() to use build_append API. This latter now is prefered in place of direct ACPI struct field settings with manual endianness conversion. Signed-off-by: Eric Auger <eric.auger@redhat.com> --- hw/i386/acpi-build.c | 28 +++++++++++++++++++--------- 1 file changed, 19 insertions(+), 9 deletions(-) diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c index b5669d6c65..f0d35d7b17 100644 --- a/hw/i386/acpi-build.c +++ b/hw/i386/acpi-build.c @@ -2298,30 +2298,40 @@ build_tpm_tcpa(GArray *table_data, BIOSLinker *linker, GArray *tcpalog) static void build_tpm2(GArray *table_data, BIOSLinker *linker, GArray *tcpalog) { - Acpi20TPM2 *tpm2_ptr = acpi_data_push(table_data, sizeof *tpm2_ptr); + Acpi20TPM2 *tpm2_ptr = acpi_data_push(table_data, sizeof(AcpiTableHeader));And now you want to build the data structure by pushing fields? I would definitely NOT do this.If I didn't misinterpret things, this was recommended by Drew and Igor as buid_append* API avoids to take care of endianness and this is the API now used in the generic ACPI code. Besides I also think that in that case it does not simplify things but maybe I did that the wrong way? Or maybe I didn't understand your remark?
If that's what they are saying... I would prefer filling out data structures with functions like cpu_to_acpi16() because that seems to be less error prone.
unsigned log_addr_size = sizeof(tpm2_ptr->log_area_start_address); unsigned log_addr_offset = (char *)&tpm2_ptr->log_area_start_address - table_data->data; + uint8_t start_method_params[12] = {}; - tpm2_ptr->platform_class = cpu_to_le16(TPM2_ACPI_CLASS_CLIENT); + /* platform class */ + build_append_int_noprefix(table_data, TPM2_ACPI_CLASS_CLIENT, 2); + /* reserved */ + build_append_int_noprefix(table_data, 0, 2); if (TPM_IS_TIS_ISA(tpm_find())) { - tpm2_ptr->control_area_address = cpu_to_le64(0); - tpm2_ptr->start_method = cpu_to_le32(TPM2_START_METHOD_MMIO); + /* address of control area */ + build_append_int_noprefix(table_data, 0, 8); + /* start method */ + build_append_int_noprefix(table_data, TPM2_START_METHOD_MMIO, 4); } else if (TPM_IS_CRB(tpm_find())) { - tpm2_ptr->control_area_address = cpu_to_le64(TPM_CRB_ADDR_CTRL); - tpm2_ptr->start_method = cpu_to_le32(TPM2_START_METHOD_CRB); + build_append_int_noprefix(table_data, TPM_CRB_ADDR_CTRL, 8); + build_append_int_noprefix(table_data, TPM2_START_METHOD_CRB, 4); } else { g_warn_if_reached(); } - tpm2_ptr->log_area_minimum_length = - cpu_to_le32(TPM_LOG_AREA_MINIMUM_SIZE); + /* platform specific parameters */ + g_array_append_vals(table_data, &start_method_params, 12);
Maybe this should be wrapped in an inline function like build_append_array() or so.
- acpi_data_push(tcpalog, le32_to_cpu(tpm2_ptr->log_area_minimum_length)); + /* log area minimum length */ + build_append_int_noprefix(table_data, TPM_LOG_AREA_MINIMUM_SIZE, 4); + + acpi_data_push(tcpalog, TPM_LOG_AREA_MINIMUM_SIZE);
At this point we have a double-allocation of log memory on x86_64. You'd need the patch I posted to create the TCPA table only for TPM 1.2.
bios_linker_loader_alloc(linker, ACPI_BUILD_TPMLOG_FILE, tcpalog, 1, false); /* log area start address to be filled by Guest linker */ + build_append_int_noprefix(table_data, 0, 8); bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE, log_addr_offset, log_addr_size, ACPI_BUILD_TPMLOG_FILE, 0);
[Prev in Thread] | Current Thread | [Next in Thread] |