coreboot-kgpe-d16/src/lib/tpm2_marshaling.c

550 lines
14 KiB
C
Raw Normal View History

/*
* Copyright 2016 The Chromium OS Authors. All rights reserved.
* Use of this source code is governed by a BSD-style license that can be
* found in the LICENSE file.
*/
#include <commonlib/endian.h>
#include <console/console.h>
#include <stdlib.h>
#include <string.h>
#include "tpm2_marshaling.h"
static uint16_t tpm_tag; /* Depends on the command type. */
/*
* Each unmarshaling function receives a pointer to the buffer pointer and a
* pointer to the size of data still in the buffer. The function extracts data
* from the buffer and adjusts both buffer pointer and remaining data size.
*
* Should there be not enough data in the buffer to unmarshal the required
* object, the remaining data size is set to -1 to indicate the error. The
* remaining data size is expected to be set to zero once the last data item
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
* has been extracted from the receive buffer.
*/
static uint16_t unmarshal_u16(void **buffer, int *buffer_space)
{
uint16_t value;
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
if (*buffer_space < 0)
return 0;
if (*buffer_space < sizeof(value)) {
*buffer_space = -1; /* Indicate a failure. */
return 0;
}
value = read_be16(*buffer);
*buffer = (void *) ((uintptr_t) (*buffer) + sizeof(value));
*buffer_space -= sizeof(value);
return value;
}
static uint16_t unmarshal_u32(void **buffer, int *buffer_space)
{
uint32_t value;
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
if (*buffer_space < 0)
return 0;
if (*buffer_space < sizeof(value)) {
*buffer_space = -1; /* Indicate a failure. */
return 0;
}
value = read_be32(*buffer);
*buffer = (void *) ((uintptr_t) (*buffer) + sizeof(value));
*buffer_space -= sizeof(value);
return value;
}
static uint8_t unmarshal_u8(void **buffer, int *buffer_space)
{
uint8_t value;
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
if (*buffer_space < 0)
return 0;
if (*buffer_space < sizeof(value)) {
*buffer_space = -1; /* Indicate a failure. */
return 0;
}
value = ((uint8_t *)(*buffer))[0];
*buffer = (void *) ((uintptr_t) (*buffer) + sizeof(value));
*buffer_space -= sizeof(value);
return value;
}
#define unmarshal_TPM_CAP(a, b) unmarshal_u32(a, b)
#define unmarshal_TPM_CC(a, b) unmarshal_u32(a, b)
#define unmarshal_TPM_PT(a, b) unmarshal_u32(a, b)
#define unmarshal_TPM_HANDLE(a, b) unmarshal_u32(a, b)
/*
* Each marshaling function receives a pointer to the buffer to marshal into,
* a pointer to the data item to be marshaled, and a pointer to the remaining
* room in the buffer.
*/
/*
* Marshaling an arbitrary blob requires its size in addition to common
* parameter set.
*/
static void marshal_blob(void **buffer, void *blob,
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
size_t blob_size, size_t *buffer_space)
{
if (*buffer_space < blob_size) {
*buffer_space = -1;
return;
}
memcpy(*buffer, blob, blob_size);
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
*buffer_space -= blob_size;
*buffer = (void *)((uintptr_t)(*buffer) + blob_size);
}
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
static void marshal_u8(void **buffer, uint8_t value, size_t *buffer_space)
{
uint8_t *bp = *buffer;
if (*buffer_space < sizeof(value)) {
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
*buffer_space = 0;
return;
}
*bp++ = value;
*buffer = bp;
*buffer_space -= sizeof(value);
}
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
static void marshal_u16(void **buffer, uint16_t value, size_t *buffer_space)
{
if (*buffer_space < sizeof(value)) {
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
*buffer_space = 0;
return;
}
write_be16(*buffer, value);
*buffer = (void *)((uintptr_t)(*buffer) + sizeof(value));
*buffer_space -= sizeof(value);
}
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
static void marshal_u32(void **buffer, uint32_t value, size_t *buffer_space)
{
if (*buffer_space < sizeof(value)) {
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
*buffer_space = 0;
return;
}
write_be32(*buffer, value);
*buffer = (void *)((uintptr_t)(*buffer) + sizeof(value));
*buffer_space -= sizeof(value);
}
#define marshal_TPM_HANDLE(a, b, c) marshal_u32(a, b, c)
#define marshal_TPMI_RH_NV_AUTH(a, b, c) marshal_TPM_HANDLE(a, b, c)
#define marshal_TPMI_RH_NV_INDEX(a, b, c) marshal_TPM_HANDLE(a, b, c)
#define marshal_TPMI_SH_AUTH_SESSION(a, b, c) marshal_TPM_HANDLE(a, b, c)
#define marshal_TPMI_ALG_HASH(a, b, c) marshal_u16(a, b, c)
static void marshal_startup(void **buffer,
struct tpm2_startup *cmd_body,
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
size_t *buffer_space)
{
marshal_u16(buffer, cmd_body->startup_type, buffer_space);
}
static void marshal_get_capability(void **buffer,
struct tpm2_get_capability *cmd_body,
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
size_t *buffer_space)
{
marshal_u32(buffer, cmd_body->capability, buffer_space);
marshal_u32(buffer, cmd_body->property, buffer_space);
marshal_u32(buffer, cmd_body->propertyCount, buffer_space);
}
static void marshal_TPM2B(void **buffer,
TPM2B *data,
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
size_t *buffer_space)
{
size_t total_size = data->size + sizeof(data->size);
if (total_size > *buffer_space) {
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
*buffer_space = 0;
return;
}
marshal_u16(buffer, data->size, buffer_space);
memcpy(*buffer, data->buffer, data->size);
*buffer = ((uint8_t *)(*buffer)) + data->size;
*buffer_space -= data->size;
}
static void marshal_TPMA_NV(void **buffer,
TPMA_NV *nv,
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
size_t *buffer_space)
{
marshal_u32(buffer, *((uint32_t *)nv), buffer_space);
}
static void marshal_TPMS_NV_PUBLIC(void **buffer,
TPMS_NV_PUBLIC *nvpub,
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
size_t *buffer_space)
{
marshal_TPM_HANDLE(buffer, nvpub->nvIndex, buffer_space);
marshal_TPMI_ALG_HASH(buffer, nvpub->nameAlg, buffer_space);
marshal_TPMA_NV(buffer, &nvpub->attributes, buffer_space);
marshal_TPM2B(buffer, &nvpub->authPolicy.b, buffer_space);
marshal_u16(buffer, nvpub->dataSize, buffer_space);
}
static void marshal_session_header(void **buffer,
struct tpm2_session_header *session_header,
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
size_t *buffer_space)
{
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
size_t base_size;
void *size_location = *buffer;
/* Skip room for the session header size. */
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
if (*buffer_space < sizeof(uint32_t)) {
*buffer_space = 0;
return;
}
*buffer_space -= sizeof(uint32_t);
*buffer = (void *)(((uintptr_t) *buffer) + sizeof(uint32_t));
base_size = *buffer_space;
marshal_u32(buffer, session_header->session_handle, buffer_space);
marshal_u16(buffer, session_header->nonce_size, buffer_space);
marshal_blob(buffer, session_header->nonce,
session_header->nonce_size, buffer_space);
marshal_u8(buffer, session_header->session_attrs, buffer_space);
marshal_u16(buffer, session_header->auth_size, buffer_space);
marshal_blob(buffer, session_header->auth,
session_header->auth_size, buffer_space);
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
if (!*buffer_space)
return; /* The structure did not fit. */
/* Paste in the session size. */
marshal_u32(&size_location, base_size - *buffer_space, &base_size);
}
static void marshal_nv_define_space(void **buffer,
struct tpm2_nv_define_space_cmd *nvd_in,
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
size_t *buffer_space)
{
void *size_location;
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
size_t base_size;
size_t sizeof_nv_public_size = sizeof(uint16_t);
struct tpm2_session_header session_header;
marshal_TPM_HANDLE(buffer, TPM_RH_PLATFORM, buffer_space);
memset(&session_header, 0, sizeof(session_header));
session_header.session_handle = TPM_RS_PW;
marshal_session_header(buffer, &session_header, buffer_space);
tpm_tag = TPM_ST_SESSIONS;
marshal_TPM2B(buffer, &nvd_in->auth.b, buffer_space);
/* This is where the TPMS_NV_PUBLIC size will be stored. */
size_location = *buffer;
/* Allocate room for the size. */
*buffer = ((uint8_t *)(*buffer)) + sizeof(uint16_t);
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
if (*buffer_space < sizeof_nv_public_size) {
*buffer_space = 0;
return;
}
*buffer_space -= sizeof_nv_public_size;
base_size = *buffer_space;
marshal_TPMS_NV_PUBLIC(buffer, &nvd_in->publicInfo, buffer_space);
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
if (!*buffer_space)
return;
base_size = base_size - *buffer_space;
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
marshal_u16(&size_location, base_size, &sizeof_nv_public_size);
}
static void marshal_nv_write(void **buffer,
struct tpm2_nv_write_cmd *command_body,
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
size_t *buffer_space)
{
struct tpm2_session_header session_header;
marshal_TPM_HANDLE(buffer, TPM_RH_PLATFORM, buffer_space);
marshal_TPM_HANDLE(buffer, command_body->nvIndex, buffer_space);
memset(&session_header, 0, sizeof(session_header));
session_header.session_handle = TPM_RS_PW;
marshal_session_header(buffer, &session_header, buffer_space);
tpm_tag = TPM_ST_SESSIONS;
marshal_TPM2B(buffer, &command_body->data.b, buffer_space);
marshal_u16(buffer, command_body->offset, buffer_space);
}
static void marshal_nv_read(void **buffer,
struct tpm2_nv_read_cmd *command_body,
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
size_t *buffer_space)
{
struct tpm2_session_header session_header;
marshal_TPM_HANDLE(buffer, TPM_RH_PLATFORM, buffer_space);
marshal_TPM_HANDLE(buffer, command_body->nvIndex, buffer_space);
memset(&session_header, 0, sizeof(session_header));
session_header.session_handle = TPM_RS_PW;
marshal_session_header(buffer, &session_header, buffer_space);
tpm_tag = TPM_ST_SESSIONS;
marshal_u16(buffer, command_body->size, buffer_space);
marshal_u16(buffer, command_body->offset, buffer_space);
}
static void marshal_selftest(void **buffer,
struct tpm2_self_test *command_body,
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
size_t *buffer_space)
{
marshal_u8(buffer, command_body->yes_no, buffer_space);
}
int tpm_marshal_command(TPM_CC command, void *tpm_command_body,
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
void *buffer, size_t buffer_size)
{
void *cmd_body = (uint8_t *)buffer + sizeof(struct tpm_header);
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
size_t max_body_size = buffer_size - sizeof(struct tpm_header);
size_t body_size = max_body_size;
/* Will be modified when marshaling some commands. */
tpm_tag = TPM_ST_NO_SESSIONS;
switch (command) {
case TPM2_Startup:
marshal_startup(&cmd_body, tpm_command_body, &body_size);
break;
case TPM2_GetCapability:
marshal_get_capability(&cmd_body, tpm_command_body,
&body_size);
break;
case TPM2_NV_Read:
marshal_nv_read(&cmd_body, tpm_command_body, &body_size);
break;
case TPM2_NV_DefineSpace:
marshal_nv_define_space(&cmd_body,
tpm_command_body, &body_size);
break;
case TPM2_NV_Write:
marshal_nv_write(&cmd_body, tpm_command_body, &body_size);
break;
case TPM2_SelfTest:
marshal_selftest(&cmd_body, tpm_command_body, &body_size);
break;
default:
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
body_size = 0;
printk(BIOS_INFO, "%s:%d:Request to marshal unsupported command %#x\n",
__FILE__, __LINE__, command);
}
if (body_size > 0) {
size_t marshaled_size;
size_t header_room = sizeof(struct tpm_header);
/* See how much room was taken by marshaling. */
marshaled_size = max_body_size - body_size;
/* Total size includes the header size. */
marshaled_size += sizeof(struct tpm_header);
marshal_u16(&buffer, tpm_tag, &header_room);
marshal_u32(&buffer, marshaled_size, &header_room);
marshal_u32(&buffer, command, &header_room);
return marshaled_size;
}
return -1;
}
static void unmarshal_get_capability(void **buffer, int *size,
struct get_cap_response *gcr)
{
int i;
gcr->more_data = unmarshal_u8(buffer, size);
gcr->cd.capability = unmarshal_TPM_CAP(buffer, size);
if (*size < 0)
return;
switch (gcr->cd.capability) {
case TPM_CAP_TPM_PROPERTIES:
gcr->cd.data.tpmProperties.count =
unmarshal_u32(buffer, size);
if (*size < 0)
return;
if (gcr->cd.data.tpmProperties.count > ARRAY_SIZE
(gcr->cd.data.tpmProperties.tpmProperty)) {
printk(BIOS_INFO, "%s:%s:%d - %d - too many properties\n",
__FILE__, __func__, __LINE__,
gcr->cd.data.tpmProperties.count);
*size = -1;
return;
}
for (i = 0; i < gcr->cd.data.tpmProperties.count; i++) {
TPMS_TAGGED_PROPERTY *pp;
pp = gcr->cd.data.tpmProperties.tpmProperty + i;
pp->property = unmarshal_TPM_PT(buffer, size);
pp->value = unmarshal_u32(buffer, size);
}
break;
default:
printk(BIOS_ERR,
"%s:%d - unable to unmarshal capability response",
__func__, __LINE__);
printk(BIOS_ERR, " for %d\n", gcr->cd.capability);
*size = -1;
break;
}
}
static void unmarshal_TPM2B_MAX_NV_BUFFER(void **buffer,
int *size,
TPM2B_MAX_NV_BUFFER *nv_buffer)
{
nv_buffer->t.size = unmarshal_u16(buffer, size);
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
if ((*size < 0) || (nv_buffer->t.size > *size)) {
printk(BIOS_ERR, "%s:%d - "
"size mismatch: expected %d, remaining %d\n",
__func__, __LINE__, nv_buffer->t.size, *size);
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
*size = -1;
return;
}
nv_buffer->t.buffer = *buffer;
*buffer = ((uint8_t *)(*buffer)) + nv_buffer->t.size;
*size -= nv_buffer->t.size;
}
static void unmarshal_nv_read(void **buffer, int *size,
struct nv_read_response *nvr)
{
/* Total size of the parameter field. */
nvr->params_size = unmarshal_u32(buffer, size);
unmarshal_TPM2B_MAX_NV_BUFFER(buffer, size, &nvr->buffer);
if (nvr->params_size !=
(nvr->buffer.t.size + sizeof(nvr->buffer.t.size))) {
printk(BIOS_ERR,
"%s:%d - parameter/buffer %d/%d size mismatch",
__func__, __LINE__, nvr->params_size,
nvr->buffer.t.size);
return;
}
if (*size < 0)
return;
/*
* Let's ignore the authorisation section. It should be 5 bytes total,
* just confirm that this is the case and report any discrepancy.
*/
if (*size != 5)
printk(BIOS_ERR,
"%s:%d - unexpected authorisation seciton size %d\n",
__func__, __LINE__, *size);
*buffer = ((uint8_t *)(*buffer)) + *size;
*size = 0;
}
struct tpm2_response *tpm_unmarshal_response(TPM_CC command,
void *response_body,
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
size_t in_size)
{
static struct tpm2_response tpm2_resp;
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
/*
* Should be 0 when done, positive and negaitive values indicate
* unmarshaling errors.
*/
int cr_size = in_size;
tpm2: avoid comparison between signed and unsigned ints The marshaling/unmarshaling code is using integer values to represent room left in the buffer, to be able to communicate three conditions: positive number means there is room left in the buffer, zero means that the exact amount of data in the buffer was unmarshaled and negative value means that the result of the operation did not fit into the buffer. The implementation is wrong though, as it compares directly signed and unsigned values, which is illegal, as signed values get promoted to unsigned by the compiler. This patch changes the marshaling code to use size_t for the size, and use zero as marshaling failure indication - after all the buffer where the data is marshaled to should definitely be large enough, and it is reasonable to expect at least some room left in it after marshaling. The unmarshaling situation is different: we sure want to communicate errors to the caller, but do not want to propagate error return values through multiple layers. This patch keeps the size value in int, but checks if it is negative separately, before comparing with positive values. BRANCH=none BUG=chrome-os-partner:50645 TEST=with the rest of the patches applied kevin successfully boots up. Change-Id: Ibfbd1b351e35e37c8925a78d095e4e8492805bad Signed-off-by: Martin Roth <martinroth@chromium.org> Original-Commit-Id: b1e862c2a650fa5f6cb25a01fe61e848a696cf17 Original-Change-Id: Ie7552b333afaff9a1234c948caf9d9a64447b2e1 Original-Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/358772 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/15610 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2016-07-07 19:52:46 +02:00
if ((cr_size < 0) || (in_size < sizeof(struct tpm_header)))
return NULL;
tpm2_resp.hdr.tpm_tag = unmarshal_u16(&response_body, &cr_size);
tpm2_resp.hdr.tpm_size = unmarshal_u32(&response_body, &cr_size);
tpm2_resp.hdr.tpm_code = unmarshal_TPM_CC(&response_body, &cr_size);
if (!cr_size) {
if (tpm2_resp.hdr.tpm_size != sizeof(tpm2_resp.hdr))
printk(BIOS_ERR,
"%s: size mismatch in response to command %#x\n",
__func__, command);
return &tpm2_resp;
}
switch (command) {
case TPM2_Startup:
break;
case TPM2_GetCapability:
unmarshal_get_capability(&response_body, &cr_size,
&tpm2_resp.gc);
break;
case TPM2_NV_Read:
unmarshal_nv_read(&response_body, &cr_size,
&tpm2_resp.nvr);
break;
case TPM2_NV_DefineSpace:
case TPM2_NV_Write:
/* Session data included in response can be safely ignored. */
cr_size = 0;
break;
default:
{
int i;
printk(BIOS_INFO, "%s:%d:"
"Request to unmarshal unexpected command %#x,"
" code %#x",
__func__, __LINE__, command,
tpm2_resp.hdr.tpm_code);
for (i = 0; i < cr_size; i++) {
if (!(i % 16))
printk(BIOS_INFO, "\n");
printk(BIOS_INFO, "%2.2x ",
((uint8_t *)response_body)[i]);
}
}
printk(BIOS_INFO, "\n");
return NULL;
}
if (cr_size) {
printk(BIOS_INFO,
"%s:%d got %d bytes back in response to %#x,"
" failed to parse (%d)\n",
__func__, __LINE__, tpm2_resp.hdr.tpm_size,
command, cr_size);
return NULL;
}
/* The entire message have been parsed. */
return &tpm2_resp;
}