summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
* cr50: add CQ build scriptstabilize-14285.B-cr50_stabVadim Sukhomlinov2021-10-152-21/+14
| | | | | | | | | | | | | | | | | | | CQ for cr50 doesn't cover all targets of interest. Borrow script from /platform/ec and adjust it a bit to cr50 targets. BUG=b:202192433 TEST=./firmware_builder.py --metrics m.pb build ./firmware_builder.py --metrics m.pb test ./firmware_builder.py --metrics m.pb bundle Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I77ab1134fb2b7df1e16314068dac05ba8ef49094 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3224365 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: copy firmware_builder.py from /platform/ecVadim Sukhomlinov2021-10-151-0/+250
| | | | | | | | | | | | | | | | To implement CQ make a copy of firmware_builder.py script into cr50 as a first step. BUG=b:202192433 TEST=N/A Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I9d4354242e49e0e1e4029c773fe291fab92d62df Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3224376 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: update comments authTimeSecret -> authTimeSecretHashVadim Sukhomlinov2021-10-141-2/+3
| | | | | | | | | | | | | | BUG=None TEST=make buildall -j Change-Id: Ief14c16e37842794a60c7d10e068db94083aa359 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3221960 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Howard Yang <hcyang@google.com> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Auto-Submit: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Andrey Pronin <apronin@chromium.org>
* ap_ro_verification: Do not allow over USBVadim Bendebury2021-10-143-7/+13
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | It should not be possible to trigger AP RO verification by sending the vendor command over the USB interface, it should be triggered only when the user enters the existing key sequence, or when coming from the AP as a TPM command. When verification is triggered by the key sequence the verification function is still invoked as a TPM vendor command, because a context switch into TPM task is required to be able to use its large stack. The problem is that when TPM task processes context switching commands, it unconditionally sets the bit indicating that they come from USB. This is required for some CCD commands, but not for the AP RO verification request. Let's introduce a new flag to be able to tell between commands coming from USB and from alternative sources, and allow AP RO verification be triggered by alternative sources but not from USB. BUG=b:202902506 TEST=AP RO verification still could be triggered by key presses and through a TPM command coming from the AP, but is rejected if sent over USB by the host. Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Change-Id: Ib3b5bd9867cc3038d9123b23c7a25886331a179c Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3221776 Reviewed-by: Mary Ruthven <mruthven@chromium.org>
* cr50: add run-time detection of gcc version to support gcc 11.2 LTOVadim Sukhomlinov2021-10-142-1/+12
| | | | | | | | | | | | | | | | | | | | | | | | | With planned upgrade to gcc 11.2 we need to make sure it will not fail cr50 build as it will be a reason to revert. gcc 11.2 gives us 1768 b back after TPM2 LTO, but it changes default linker behavior and produce LTO object file during partial link unless -flinker-output=nolto-rel is used. This option however fail 8.3 build. Also, gcc 11.2 introduced new optimization path IPA modref which doesn't play nice with LTO and partial link used together, causing gcc crash. To overcome this issue add -fno-ipa-modref when gcc 11.2 is used. This results in almost no impact on code size. BUG=none TEST=make BOARD=cr50 tested with gcc 8.3 and gcc 11.2 Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I78a3b3403e84dc4a426dede02b399d9d249ece81 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3218577 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Patrick Georgi <pgeorgi@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: update U2F_Sign command to hash authentication secretVadim Sukhomlinov2021-10-142-7/+19
| | | | | | | | | | | | | | | | | | | | | | | | U2F_Sign is expected to receive pre-hashed version of user credentials so should apply SHA2-256 to the value before key handle verification. Also, enforce user verification with authentication time secret if no user presence (power button press) is requested. test/tpm_test/tpmtest.py updated to apply SHA256 in U2F_Generate to match expected u2fd behavior. BUG=b:172971998 TEST=make BOARD=cr50 CRYPTO_TEST=1 U2F_TEST=1; test/tpm_test/tpmtest.py Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I92d73cd1fc0d962fefe11faeb4ce0ed68c798aa7 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3221264 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Reviewed-by: Howard Yang <hcyang@google.com> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: optimize AES/GCM/App cipher implementationVadim Sukhomlinov2021-10-145-153/+189
| | | | | | | | | | | | | | | | | | | | | | | | | | | 1. Introduced AES register structure and replaced register access with accesses to fields. In many cases it reduce code size and number of instructions. 2. Deduplication between AES implementation and App Cipher which use AES engine with key coming from key ladder. Added internal function dcrypto_aes_process() which applies current AES configuration to aligned data in highly optimized manner, same as previous outer_loop and inner_loop() functions. Overall it saves 322 bytes with gcc 8.3 BUG=none TEST=make BOARD=cr50 CRYPTO_TEST=1; test/tpm_test/tpm_test.py In ccd: cipher [to test app_cipher]; TCG Tests. Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I6551e21e5e8798aa4691cb6ba476d565778cea3d Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3213610 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Auto-Submit: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: cleanup public API for FIPS moduleVadim Sukhomlinov2021-10-146-206/+213
| | | | | | | | | | | | | | | | | | | Move declarations of AES-GCM, AES-CMAC, ECIES, HKDF and few other functions from dcrypto.h into internal.h. Merge tpm2/hkdf.c into dcrypto/hkdf.c. It contains only function used for testing and HKDF itself is only used for CRYPTO_TEST=1. BUG=b:134594373 TEST=make buildall -j; make BOARD=cr50 CRYPTO_TEST=1 Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I56c03ff4e8838871cdb28c0d9946c39754d9e054 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3219576 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Andrey Pronin <apronin@chromium.org>
* cr50: add functionality to support FIPS testing by labVadim Sukhomlinov2021-10-133-4/+14
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | 1) Add test commands to break all KAT tests [fips hmac/drbg/ecdsa/pwct] 2) To support PWCT demo reduced number of attempts to retrieve valid p256 key candidate to 16. Probability of false negative would be less than 2^-4080 (255*16), but will prevent DoS attack if it consistently fails for real reasons. 3) Fixed HMAC KAT test failure (was bound SHA failure earlier). BUG=b:138576604 TEST=make BOARD=cr50 CRYPTO_TEST=1 U2F_TEST=1 In ccd: fips fips hmac fips test - see FIPS error reboot fips drbg fips test - see FIPS error reboot fips ecdsa fips test - see FIPS error reboot fips pwct u2f_test - see NOT PASSED of u2f_generate/u2f_sign Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I0a812075bb2436f5823eff446b725f19974a2a31 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3221770 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Auto-Submit: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Andrey Pronin <apronin@chromium.org>
* ccdblock: add option to tri-state the EC UARTMary Ruthven2021-10-131-6/+30
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Block EC UART output and EC COMM with 'ccdblock TRISTATE_EC enable'. This removes the EC UART output, prevents EC-EFS2, and removes the pulldown from the EC RX signal. This is used for flashing the EC with c2d2. It's only available when CCD is enabled. BUG=b:202464674 TEST=manual # Disconnect servo. Verify EC tx is disabled and there's no pull # down after tristating the EC > ccdblock TRISTATE_EC enable CCD ports blocked: EC IGNORE_SERVO WARNING: enabling UART while servo is connected may damage hardware EC_CR50_COMM TRISTATE_EC > pinmux ... 400600c8: DIOB5 0 IN 400600d0: DIOB6 16 IN GPIO0_GPIO15 ... > ccdstate ... State flags: UARTAP USBEC+TX CCD ports blocked: EC EC_CR50_COMM TRISTATE_EC ... # Check UART TX is re-enabled after disabling TRISTATE_EC > ccdblock TRISTATE_EC disable CCD ports blocked: (none) > ccdstate ... State flags: UARTAP+TX UARTEC+TX I2C SPI USBEC+TX CCD ports blocked: (none) ... > pinmux ... 400600c8: DIOB5 78 IN UART2_TX 400600d0: DIOB6 16 IN GPIO0_GPIO15 ... # Connect servo, so cr50 disables the EC UART on it's own. # Verify the pulldown is removed after tristating the EC. > ccdstate ... State flags: UARTAP UARTEC USBEC+TX CCD ports blocked: (none) > pinmux ... 400600c8: DIOB5 0 IN PD 400600d0: DIOB6 16 IN GPIO0_GPIO15 ... > ccdblock TRISTATE_EC enable CCD ports blocked: EC EC_CR50_COMM TRISTATE_EC > ccdstate ... State flags: UARTAP USBEC+TX CCD ports blocked: EC EC_CR50_COMM TRISTATE_EC # Check PD is removed from DIOB5 > pinmux ... 400600c8: DIOB5 0 IN 400600d0: DIOB6 16 IN GPIO0_GPIO15 ... # Check PD is re-enabled after disabling TRISTATE_EC > ccdblock TRISTATE_EC disable CCD ports blocked: (none) > ccdstate ... State flags: UARTAP UARTEC USBEC+TX CCD ports blocked: (none) > pinmux ... 400600c8: DIOB5 0 IN PD 400600d0: DIOB6 16 IN GPIO0_GPIO15 Change-Id: I7bde996be3914bd5d625ad99e418f9bd2c0f41d1 Signed-off-by: Mary Ruthven <mruthven@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3219760 Reviewed-by: Vadim Bendebury <vbendeb@chromium.org>
* rdd: only enable AP UART if the AP is onMary Ruthven2021-10-131-1/+1
| | | | | | | | | | | | | | | When the AP is off, cr50 doesn't need to enable the AP UART. Having it enabled could cause power leakage if the signals aren't terminated correctly. Check ap_is_on and ap_uart_is_on before enabling the AP UART. BUG=b:136602563 TEST=firmware_Cr50DeviceState Change-Id: I192bdc4caef8cfa0ce6d8a3ca181b16388e35cd4 Signed-off-by: Mary Ruthven <mruthven@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3219756 Reviewed-by: Vadim Bendebury <vbendeb@chromium.org>
* i2cp: remove FE_TPM_I2C_ERROR logMary Ruthven2021-10-131-4/+0
| | | | | | | | | | | | | | FE_TPM_I2C_ERROR events don't mean much and it's not a big deal if cr50 needs to recover the I2C bus. Remove FE_TPM_I2C_ERROR logging from chip/i2cp. BUG=b:146067724 TEST=make buildall -j Change-Id: I2bdce35fc794559c8236b1c14d87fa4372ffafa4 Signed-off-by: Mary Ruthven <mruthven@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3219755 Reviewed-by: Vadim Bendebury <vbendeb@chromium.org>
* Reland "cr50_fuzz: Add fuzzer for u2f commands"Howard Yang2021-10-1314-21/+323
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This is a reland of 3cac98670745fc5ca82a058fab512567f8444759 The structure of u2f command related types are updated before the original CL lands. Update the fuzzer to correctly fuzz the new code, and ignore the profdata generated by fuzzers in .gitignore. Original change's description: > cr50_fuzz: Add fuzzer for u2f commands > > Currently there's only one fuzzer for Pinweaver and one for host > commands in cr50. Add a fuzzer for the u2f commands (generate, sign, > attest) used in the WebAuthn flow to ensure its security. Most regions > of the concerning functions are covered except for pure error code > returns and unreachable regions (currently auth secret is not used in > sign and attest command yet). > > Rename old cr50_fuzz namings to pinweaver_fuzz, since they only cover > Pinweaver commands. > > BUG=b:172367435 > TEST=make buildall -j > TEST=make host-u2f_fuzz && \ > ./build/host/u2f_fuzz/u2f_fuzz.exe -timeout=10 \ > -ignore_ooms=false -ignore_timeouts=false -fork=71; \ > llvm-profdata merge -sparse default.profraw -o default.profdata; \ > llvm-cov show ./build/host/u2f_fuzz/u2f_fuzz.exe \ > -object ./build/host/u2f_fuzz/RO/board/cr50/dcrypto/u2f.o \ > --instr-profile default.profdata \ > board/cr50/dcrypto/u2f.c common/u2f.c > report > > Cq-Depend: chromium:3162473 > Change-Id: I02b820cf03f7b46ccad7c3bc7b82e73ff45217c6 > Signed-off-by: Howard Yang <hcyang@google.com> > Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3162469 > Reviewed-by: Andrey Pronin <apronin@chromium.org> > Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> > Reviewed-by: Leo Lai <cylai@google.com> Bug: b:172367435 Change-Id: I279e20b21a11e0ec957b6a5c3e95bc9a3b9df196 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3217474 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Tested-by: Howard Yang <hcyang@google.com> Commit-Queue: Howard Yang <hcyang@google.com>
* cr50: place .text.fips_checksum immediately after FIPS modulefirmware-chameleon-14280.B-cr50_stabVadim Sukhomlinov2021-10-121-10/+15
| | | | | | | | | | | | | | | | | | | | .text.fips_checksum section was placed after .text which resulted in different address for it. Since address of this section is used in FIPS module it resulted in different digest computed. BUG=b:138578318 TEST=make BOARD=cr50, check map file to ensure .text.fips_checksum is just after FIPS module. Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: Ia054fec9191eac8818dcde139320eddbd7c8085b Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3218580 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Vadim Bendebury <vbendeb@chromium.org> Reviewed-by: Mary Ruthven <mruthven@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Auto-Submit: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* fips: silence fips section fillerVadim Bendebury2021-10-112-8/+15
| | | | | | | | | | | | | | | | | | | | | | Script inserting FIPS checksum into the image uses the dd utility which generates stderr output even when there is no errors. This patch adds code which captures the dd stderr output and prints it out only if there is an actual error. stdout output of the script is suppressed unless make was invoked with V=1. Also made a few modifications as requested by shellcheck. BUG=none TEST=make output does not have extra lines. built and ran a Cr50 image, it reports successful FIPS integrity self check. Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Change-Id: I9121bc5a9a40633b9a3d18ea5766bc1ed274a9c2 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3210946 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: use LTO module for TPM2Vadim Bendebury2021-10-111-20/+16
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | In order to reduce code footprint, do not link TPM2 library modules, instead build TPM2 as a single relocatable object module, a collection of library sources compiled and linked with LTO enabled. BUG=b:65253310 TEST=observed code space reduced by 1428 bytes, the bss_libtpm2 section remained practically unchanged: before: *** 5548 bytes in flash and 5652 bytes in RAM still ... 000104d0 B __bss_libtpm2_start 000155d7 B __bss_libtpm2_end after: *** 6976 bytes in flash and 5652 bytes in RAM still ... 000104d0 B __bss_libtpm2_start 000155d4 B __bss_libtpm2_end Verified that the new Cr50 image allows a Chrome OS to successfully boot and restart. Cq-Depend: chromium:3210050 Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Change-Id: I32335df29a332da115d8af56c157d5ad4189e9b0 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3210510 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Mary Ruthven <mruthven@chromium.org>
* cr50: improve stability and performance of tpmtest on H1 Red boardVadim Sukhomlinov2021-10-112-4/+8
| | | | | | | | | | | | | | | | | | | | | | | | | | Currently tpmtest.py fails to connect to H1 red board first time after it is flashed, which is not convenient. It looks like the workaround is to try to read any TPM register. So, implement a workaround. Frequency of FTDI can be safely increased to 2000Khz. Reduced delay to start transaction from 10ms to 200us which greatly improve overall testing (except for RSA which actually runs noticeable time). Overall time to run tests decreased from 165s to about 120s, but take into account RSA key gen tests alone are about 100s. BUG=none TEST=testtpm.py Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: Ib67b71b36457b33f38135f9cec269dcf35881f54 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3214771 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Vadim Bendebury <vbendeb@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: fix test/tpm_test makefile to build bn_testVadim Sukhomlinov2021-10-092-1/+10
| | | | | | | | | | | | | | | | | | | | | | | test/tpm_test Makefile doesn't build bn_test properly. Need to add handling of difference between Cr50 and Chip/g dcrypto. Also, Cr50 code use __always_inline for some functions which itself requires compiler optimization to turn on. Adding -O2 to CFLAGS. BUG=None TEST=make -C test/tpm_test clean make -C test/tpm_test make -C test/tpm_test clean make -C test/tpm_test CR50=1 Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I08749ebaa930fd4f71d7406ed289bf480b5a8510 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3215057 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Vadim Bendebury <vbendeb@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Auto-Submit: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: fix fuzzingVadim Sukhomlinov2021-10-072-8/+17
| | | | | | | | | | | | | | | | | | | | | | | | | | | | make runfuzztests started to fail once both: https://crrev.com/c/3162473 and https://crrev.com/c/3208916 landed. Clang seems to incorrectly discarding sections it generated for profiling, resulting in: __profc_DCRYPTO_hw_sha256_init' referenced in section .text.compute_hash[compute_hash]' of build/host/cr50_fuzz/libec.a(libec.a.2.o): defined in discarded section `__llvm_prf_cnts[__profd_DCRYPTO_hw_sha256_init]' of build/host/cr50_fuzz/libec.a(libec.a.2.o) __profc_DCRYPTO_hw_sha256_init' referenced in section .text.create_merkle_tree[create_merkle_tree]' of build/host/cr50_fuzz/libec.a(libec.a.2.o): defined in discarded section `__llvm_prf_cnts[__profd_DCRYPTO_hw_sha256_init]' of build/host/cr50_fuzz/libec.a(libec.a.2.o) clang-13: error: linker command failed with exit code 1 (use -v to see invocation) This definition of __always_inline should be useful in other cases, so moving it into common.h. Note, we have to #undef it first, as it is previously defined in system headers. BUG=none TEST=make buildall -j Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I073b38a68fd43a14dbe92063011c95758030b225 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3213113 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: fix make file dependencies for FIPS moduleVadim Sukhomlinov2021-10-072-1/+3
| | | | | | | | | | | | | | | | | | | | | Dependencies for FIPS module were incorrectly set as actual dependency on $(out)/ec_version.h and $(out)/env_config.h instead of order-only dependency, causing rebuild on any change in repository. BUG=b:202225290 TEST=make BOARD=cr50 echo ' ' >> board/cr50/tpm2/rsa.c make BOARD=cr50 should only rebuild tpm2/rsa.c and not dcrypto/* Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I4d2e8e4a2182ddf850ccfad18e9b517f41594d55 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3208539 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Vadim Bendebury <vbendeb@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: change static inline to __always_inlineVadim Sukhomlinov2021-10-072-73/+76
| | | | | | | | | | | | | | | | | | | | | | | gcc 11.2 changes semantic of how 'static inline' works with LTO, which causes either ODR or missing symbol issues during linking when several objects created by LTO. After several experiments with inline extern inline It seems that using __inline __attribute__(always_inline) is most reliable method. BUG=None TEST=make buildall -j Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I4d0e8bed00bbc3b3e580c4c610a2f733f2525973 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3208916 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* Revert "cr50_fuzz: Add fuzzer for u2f commands"Vadim Bendebury2021-10-0712-276/+19
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This reverts commit 3cac98670745fc5ca82a058fab512567f8444759. Reason for revert: This patch breaks building of 'make buildall' and seems to be leaving some generated files in the root directory. Original change's description: > cr50_fuzz: Add fuzzer for u2f commands > > Currently there's only one fuzzer for Pinweaver and one for host > commands in cr50. Add a fuzzer for the u2f commands (generate, sign, > attest) used in the WebAuthn flow to ensure its security. Most regions > of the concerning functions are covered except for pure error code > returns and unreachable regions (currently auth secret is not used in > sign and attest command yet). > > Rename old cr50_fuzz namings to pinweaver_fuzz, since they only cover > Pinweaver commands. > > BUG=b:172367435 > TEST=make buildall -j > TEST=make host-u2f_fuzz && \ > ./build/host/u2f_fuzz/u2f_fuzz.exe -timeout=10 \ > -ignore_ooms=false -ignore_timeouts=false -fork=71; \ > llvm-profdata merge -sparse default.profraw -o default.profdata; \ > llvm-cov show ./build/host/u2f_fuzz/u2f_fuzz.exe \ > -object ./build/host/u2f_fuzz/RO/board/cr50/dcrypto/u2f.o \ > --instr-profile default.profdata \ > board/cr50/dcrypto/u2f.c common/u2f.c > report > > Cq-Depend: chromium:3162473 > Change-Id: I02b820cf03f7b46ccad7c3bc7b82e73ff45217c6 > Signed-off-by: Howard Yang <hcyang@google.com> > Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3162469 > Reviewed-by: Andrey Pronin <apronin@chromium.org> > Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> > Reviewed-by: Leo Lai <cylai@google.com> Bug: b:172367435 Change-Id: Ie844e44e0cd6254553694c23a535f18329cef77d Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3212497 Reviewed-by: Vadim Bendebury <vbendeb@chromium.org> Reviewed-by: Mary Ruthven <mruthven@chromium.org> Tested-by: Vadim Bendebury <vbendeb@chromium.org> Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
* cr50_fuzz: Add fuzzer for u2f commandsstabilize-ambassador-14268.43.B-cr50_stabstabilize-14268.67.B-cr50_stabstabilize-14268.52.B-cr50_stabstabilize-14268.51.B-cr50_stabrelease-R96-14268.B-cr50_stabHoward Yang2021-10-0712-19/+276
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Currently there's only one fuzzer for Pinweaver and one for host commands in cr50. Add a fuzzer for the u2f commands (generate, sign, attest) used in the WebAuthn flow to ensure its security. Most regions of the concerning functions are covered except for pure error code returns and unreachable regions (currently auth secret is not used in sign and attest command yet). Rename old cr50_fuzz namings to pinweaver_fuzz, since they only cover Pinweaver commands. BUG=b:172367435 TEST=make buildall -j TEST=make host-u2f_fuzz && \ ./build/host/u2f_fuzz/u2f_fuzz.exe -timeout=10 \ -ignore_ooms=false -ignore_timeouts=false -fork=71; \ llvm-profdata merge -sparse default.profraw -o default.profdata; \ llvm-cov show ./build/host/u2f_fuzz/u2f_fuzz.exe \ -object ./build/host/u2f_fuzz/RO/board/cr50/dcrypto/u2f.o \ --instr-profile default.profdata \ board/cr50/dcrypto/u2f.c common/u2f.c > report Cq-Depend: chromium:3162473 Change-Id: I02b820cf03f7b46ccad7c3bc7b82e73ff45217c6 Signed-off-by: Howard Yang <hcyang@google.com> Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3162469 Reviewed-by: Andrey Pronin <apronin@chromium.org> Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Leo Lai <cylai@google.com>
* cr50: improve toolchainHoward Yang2021-10-074-31/+41
| | | | | | | | | | | | | | Adjust Makefile toolchain setting so that fuzzers can generate coverage mapping correctly and generate debug symbols for source-based coverage report. Also, update gitignore to ignore local vscode settings. BUG=none TEST=make buildall -j Change-Id: I6d5c720895cbb9119c9266df998aa5cc308c1e61 Signed-off-by: Howard Yang <hcyang@google.com> Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3162473 Reviewed-by: Andrey Pronin <apronin@chromium.org>
* cr50: switch CR50_DEV in FIPS module to branches of CRYPTO_TESTVadim Sukhomlinov2021-10-074-29/+40
| | | | | | | | | | | | | | | | | | | | | | | | Due to limited space available with CR50_DEV=1, move some of crypto related functionality which was under CR50_DEV to branches of CRYPTO_TEST=1, namely: - SELF_TEST=1 to print self-integrity hashes - U2F_VERBOSE=1 to print debug information from U2F key generation. Config options sorted alphabetically in ENV_VARS and in processing order. BUG=None TEST=make BOARD=cr50 CR50_DEV=1 make BOARD=cr50 CRYPTO_TEST=1 SELF_TEST=1 make BOARD=cr50 CRYPTO_TEST=1 U2F_TEST=1 U2F_VERBOSE=1 Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I66485b2d1fff8c0947aaf31c93348a16101f14b7 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3209647 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Vadim Bendebury <vbendeb@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: switch TRNG register access to use struct instead of GREADVadim Sukhomlinov2021-10-072-25/+65
| | | | | | | | | | | | | | | This allows a bit more efficient code generation. BUG=none TEST=make BOARD=cr50 CRYPTO_TEST=1 RND_TEST=1 Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: Ia92116a9aa4ac7d9f77d207205e712c03722dd95 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3210238 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: move bn dcrypto microcode into dcrypto_bn.incVadim Sukhomlinov2021-10-072-1102/+1102
| | | | | | | | | | | | | | | | | To simplify code analysis, move dcrypto's blob into dcrypto_bn.inc similar to p256 code in dcrypto_p256.inc. BUG=none TEST=make BOARD=cr50 CRYPTO_TEST=1; test/tpm_test/tpmtest.py Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I2ae3a0793bd829c15844d55061952a69a412e2e9 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3210226 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Vadim Bendebury <vbendeb@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: move sha512 dcrypto microcode into dcrypto_sha512.incVadim Sukhomlinov2021-10-072-432/+435
| | | | | | | | | | | | | | | | | | To simplify code analysis, move dcrypto's blob into dcrypto_sha512.inc similar to p256 code in dcrypto_p256.inc. In the process fix minor compiler warning on signed/unsigned compare. BUG=none TEST=make BOARD=cr50 CRYPTO_TEST=1; test/tpm_test/tpmtest.py Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I4aac81a3b6fa0c055b83f91575f2d37755845e63 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3210229 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Vadim Bendebury <vbendeb@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: add support for v2 of U2F key handle for WebAuthnstabilize-14267.B-cr50_stabVadim Sukhomlinov2021-10-065-142/+381
| | | | | | | | | | | | | | | | | | | Adding v2 of key handle which drops kh_hmac field and use single authorization code for all relevant fields. BUG=b:172971998 TEST=make BOARD=cr50 CRYPTO_TEST=1 U2F_TEST=1; in ccd: u2f_test - unit tests test/tpm_test/tpmtest.py Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I647ded7a2c157cea91ac48a2ba679def318c1e63 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3199671 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Auto-Submit: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: switch RSA/big numbers to dynamic buffer allocation for RSA 4KVadim Sukhomlinov2021-10-061-13/+17
| | | | | | | | | | | | | | | | | | Several bn_* function still use static buffer allocation. Switch to dynamic allocation to enable support for RSA 3K/4K. BUG=none TEST=make BOARD=cr50 CRYPTO_TEST=1; test/tpm_test/tpmtest.py ../../build/tpm_test/bn_test TCG tests Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I150fa99bde89cc486f7ad945b5a312fe7d787fb0 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3207349 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: update AES-CMAC implementationVadim Sukhomlinov2021-10-063-118/+135
| | | | | | | | | | | | | | | | | | | | | | | Cr50 doesn't use CMAC, it is not even compiled, however during internal review potential issues with branching on key values were spotted. 1) Fix key expansion to be constant time 2) Switch to enum dcrypto_result 3) Test commands updated to be compatible with FIPS build (use .rodata) 4) Clean up computed tag on stack during verification BUG=None TEST=make BOARD=cr50 CRYPTO_TEST=1 CMAC_TEST=1 in ccd: test_cmac 1 2 3 4 test_cmac_ver 1 2 3 4 Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: Iff9b84dd8fb2baed9152f1ee5c40ef8e4198edd3 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3194972 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: update RSA public API to block access on FIPS errorsVadim Sukhomlinov2021-10-0610-188/+233
| | | | | | | | | | | | | | | | | | | | | | | | | | | | 1. Switched RSA public and internal functions to use enum dcrypto_result 2. Added checks for FIPS errors. 3. Updated call sites to properly handle result values. BUG=b:197893750 TEST=make BOARD=cr50 CRYPTO_TEST=1; test/tpm_test/tpm_test.py TCG tests: ---------------------- Test Result Summary ----------------------------- Test executed on: Mon Oct 5 18:26:07 2021 Performed Tests: 248 Passed Tests: 248 Failed Tests: 0 Errors: 0 Warnings: 0 ======================================================================== Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I23d391322e55b541d72388b2a4661991a61dd020 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3207348 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: RSA enhancementsVadim Sukhomlinov2021-10-051-59/+68
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | 1.Implemented dynamic stack allocation of big number buffers to support up to RSA 4K for all public APIs. 2. Internal function switched to use enum dcrypto_result 3. Added check that provided exponent is at least odd number (should be prime / co-prime with N). 4. Saving a bit by reusing zero constant. BUG=none TEST=make BOARD=cr50 CRYPTO_TEST=1; test/tpm_test/tpmtest.py TCG tests ---------------------- Test Result Summary ----------------------------- Test executed on: Mon Oct 4 22:46:07 2021 Performed Tests: 248 Passed Tests: 248 Failed Tests: 0 Errors: 0 Warnings: 0 ======================================================================== Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: Id23ebfdc04132de1f26ee0888b00cacdee2eaf43 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3204566 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Vadim Bendebury <vbendeb@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: Update AES public APIsfactory-ambassador-14265.B-cr50_stabVadim Sukhomlinov2021-10-059-63/+118
| | | | | | | | | | | | | | | | | | | | To support FIPS mode we need to block access to crypto in case of errors. 1) Added check for FIPS errors into DCRYPTO_aes_init() 2) Return codes updated to enum dcrypto_result 3) Call sites updated to check for return codes BUG=b:197893750 TEST=make BOARD=cr50 CRYPTO_TEST=1; test/tpm_test/tpmtest.py Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: Id614cc346fe22537e9208196bf1322221a253b0c Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3194985 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: fix build with older gcc8.3 until gcc 11.2 lands properlyVadim Sukhomlinov2021-10-051-1/+2
| | | | | | | | | | | | | | | | New options for gcc 11.2 doesn't work on gcc 8.3 since gcc 11.2 update was reverted. BUG=None TEST=make buildall Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: Idf965bab903f2700dd01eb028e2a1aa6dc53e101 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3206474 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Vadim Bendebury <vbendeb@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: update ap_ro to new HASH APIVadim Sukhomlinov2021-10-051-35/+22
| | | | | | | | | | | | | | | | | | | | | | https://crrev.com/c/3192137 changed hash & hmac APIs to return error codes on failures (primarily FIPS errors), and added convenience api to set hash mode. https://crrev.com/c/3172256 was using internal hash API which became replaced with a new API with error checking, and should be updated to use a new API. BUG=none TEST=make buildall -j; make BOARD=cr50 Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I33f31a8913d9a36dac451dac4312a482f761e3f7 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3200807 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Vadim Bendebury <vbendeb@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: adjust FIPS module dependency to rebuild on configuration changesVadim Sukhomlinov2021-10-021-2/+4
| | | | | | | | | | | | | | | | FIPS module wasn't rebuilt properly when configuration options were changed. Added proper dependencies to ensure it is rebuilt when needed. BUG=none TEST=make buildall -j Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I543eeeb00d7a72c6b8a936948e62830753954f12 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3200516 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Mary Ruthven <mruthven@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: provide public crypto API for HMAC/HASH with error reporting.Vadim Sukhomlinov2021-10-0225-907/+1259
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | To implement FIPS mode for Cr50 we should be able to block access to crypto functions if errors are detected. Historically all HASH/HMAC functions were declared as void with no return type. 1) Split existing functions into public part (data structs, update and final parts) and internal part - unchecked init functions. 2) Introduced new functions to start SHA / HMAC operation which returns status code and block access to crypto in case of FIPS errors. 3) Dcrypto hash algorithms codes updated to match TPM_ALG_ID to simplify adaptation layer and move checks inside Dcrypto module. 4) Updated all uses of API outside FIPS module to check return code and act accordingly. 5) As a side effect RSA can now support SHA384 & SHA512 for signing, board/host mock ups simplified. BUG=b:197893750 TEST=make buildall -j; make BOARD=cr50 CRYPTO_TEST=1; test/tpm_test/tpm_test.py TCG tests ------------------------------ Test Result Summary --------------------- Test executed on: Tue Sep 28 15:23:35 2021 Performed Tests: 248 Passed Tests: 248 Failed Tests: 0 Errors: 0 Warnings: 0 ======================================================================== Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: Ibbc38703496f417cba693c37d39a82a662c3f7ee Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3192137 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: improve RSA core function to support up to RSA 4KVadim Sukhomlinov2021-10-021-23/+47
| | | | | | | | | | | | | | | | | | | | While all Dcrypto code for RSA supports 4K, few software functions were allocating static buffers on stack and thus limiting what RSA sizes can be supported. Updating bn_modinv_vartime() and bn_probable_prime() to allocate stack space dynamically. This simplifies enabling RSA 3K / 4K support. BUG=none TEST=make BOARD=cr50 CRYPTO_TEST=1; test/tpm_test/tpm_test.py Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I7c410d349e8755d491151152168701ecdd54c04c Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3193510 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* ap_ro_verification: add implementation of the new schemeVadim Bendebury2021-10-023-38/+1215
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The new AP RO verification approach, version 2, places the AP RO verification structures in the AP RO itself, as described below. A new section is included in AP RO FMAP, called RO_GSCVD, which contains all information necessary to verify the AP RO: gsc_verification_data (GVD) structure, GVD signature, the platform key to verify the signature and the root key to verify the platform key. GVD contains an array of AP RO ranges to be verified and the hash of the contents of those ranges. The signature of the GVD is followed by the public platform key, which allows to verify the GVD signature. The platform key in turn is signed by the root key, and the public root key is also present in RO_GSCVD, this allows to validate the platform key. The hash of the public root key is hardcoded in the GSC firmware, this gives the GSC the ability to verify the chain of objects in the RO_GSCVD FMAP area starting with the root key. This implementation supports both old and new AP RO verification schemes as follows. The structure describing data saved in the dedicated GSC flash page has been enhanced to support both old and new AP RO verification schemes. To maintain backwards compatibility the 16 bit header field version is converted into two byte fields, version and type. The new version is 1 and the new verification scheme type is 1, the old verification scheme type value is set to zero. If a V1 structure is found in the H1 hash page, but verification fails, a V2 verification is attempted just in case the RO was reprogrammed and now includes RO_GSCVD area in FMAP. The hash of the root key is saved in the text section named .rodata.root_key_hash. As presented, tests/devkeys/kernel_subkey.vbpubk from the vboot_reference tree is used as the root key. The label of the section where the hash is stored will allow the signer to replace the test key hash with the prod root key hash. Verification process is moved to the TPM task context to satisfy the increased stack requirements of V2 calculations. This provides an additional benefit: verification can be triggered by the AP issuing the vendor command. A CCD capability will be added in a follow up CL to restrict conditions when the AP can invoke verification. BUG=b:199904580 TEST=prepared an AP RO image containing RO_GSCVD and verified that the AP RO can be verified, and that the local cache of the GVD is successfully stored and allows to speed up verification. Experiments have shown that verifying the entire chain starting with the root key take 670 ms. Verifying of approximately 70KB of AP RO takes 200 ms. Verified that V1 AP RO verification is still working as expected, and that V2 can take over in case the AP RO was updated with an RO_GSCVD image, and the V1 check does not pass any more. Signed-off-by: Vadim Bendebury <vbendeb@chromium.org> Change-Id: I1397376cd0394888da2cda16c0126a313f07d426 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3172256 Reviewed-by: Andrey Pronin <apronin@chromium.org> Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: detangle RO and RW build settingsVadim Sukhomlinov2021-10-013-15/+16
| | | | | | | | | | | | | | | | | | | | | | | | | | | Cr50 so far builds RO and RW images as part of build process. With adding FIPS module and moving to board-specific crypto library with different interfaces it become hard to maintain build process as RO sources use crypto, but with different APIs, and changing that crypto is challenging as it is also used by other boards with different crypto APIs. In this CL we enable RW and RO to have independent selection of crypto library and include paths, and don't contaminate include paths with unused things like third_party/cryptoc for RW. BUG=none TEST=make buildall -j make BOARD=cr50 make BOARD=cr50 CRYPTO_TEST=1 Built cr50 images can be flashed and are workable. Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I1b666fbb8193b79f71c885a761436443fd3fca7b Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3200069 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Vadim Bendebury <vbendeb@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* chip/g: fix gcc 11.2 build issuesVadim Sukhomlinov2021-10-014-2/+10
| | | | | | | | | | | | | | | | | | | | | For some reason didn't spot earlier another unnecessary complains of gcc 11.2 in private-cr5x builds with make buildall -j. Made it so BOARD can override settings done by CHIP even though CHIP is loaded later. These settings should apply to both CFLAGS and LDFLAGS due to LTO build. BUG=none TEST=make buildall -j Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I8880c518b23778cccf969909e330e9e2d62b5fae Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3194984 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Vadim Bendebury <vbendeb@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org> Auto-Submit: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: make CONFIG_FW_INCLUDE_RO work properlyVadim Sukhomlinov2021-09-303-2/+11
| | | | | | | | | | | | | | | | | | | | | | | | CONFIG_FW_INCLUDE_RO option controls wherever RO is built. This option didn't work properly and RO was always built. Fix this behavior. However, when RO is not built RW image should include some 16KB stub. This is not added yet, so for Cr50 we still set CONFIG_FW_INCLUDE_RO. Also, corrected behavior of CONFIG_CUSTOMIZED_RO which earlier was not properly taken into account and behavior depended on custom-ro_objs variable state which always added some common files so actual result was that CONFIG_CUSTOMIZED_RO is effectively on for chip/g. BUG=none TEST=make buildall -j; make BOARD=cr50 & flash Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I31599170050b360fad5c61dd1f81844bb315e1d6 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3195319 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Vadim Bendebury <vbendeb@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: refactor HMAC_DRBG to simplify reseeding and initialization logicstabilize-14249.B-cr50_stabVadim Sukhomlinov2021-09-288-199/+208
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | 1) Move DRBG initialization flag inside DRBG context to prevent use of DRBG which is not properly initialized. 2) Add configurable reseed threshold to cover both deterministic key gen and non-deterministic randoms. Simplify reseeding logic, remove similar code snippets. Also, can support NDRBG with reseed threshold equal to 0, which will result in reseeding each time. 3) Adjust parameter names to match NIST SP 800-90A specification. 4) Enforce checking result of hmac_drbg_generate(), update call sites to check for errors. 5) Reseeding in generate function consumes additional data as per NIST SP 800-90Ar1 9.3.1 BUG=b:138577416 TEST=make BOARD=cr50 CRYPTO_TEST=1 DRBG_TEST=1; test/tpm_test/tpm_test.py in ccd: hmac_drbg rand_perf Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I0e780b5c237d7fbc64e8b0e74d12559a1f40f84c Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3183397 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: enhance ECDSA sign to retry if zero r or s produced.Vadim Sukhomlinov2021-09-281-6/+24
| | | | | | | | | | | | | | | | | | ECDSA signing can vary rarely result in zero s or r value due to combination of message, nonce and a private key. Detect such cases and retry with another nonce. BUG=b:134594373 TEST=make BOARD=cr50 CRYPTO_TEST=1; tpm_test; in ccd: dcrypto_ecdsa, u2f_test Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I1378259a0dc0e2e62cf071b779c1115c4257dc73 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3188564 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: add support for upcoming update to gcc 11.2 in coreboot-sdkVadim Sukhomlinov2021-09-241-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Coreboot-sdk is going to be updated to gcc 11.2 soon. Preliminary tests show that cr50 can successfully be compiled with it, however default warnings seems produce to many false positive errors. Disable string-overflow to make code compile. I tried to fix it, but found nothing to fix. Example in dcrypto_bn.c:1323 it complains: if (!rand64(ctx->rnd) board/cr50/dcrypto/dcrypto_bn.c: In function 'dcrypto_modexp_blinded': board/cr50/dcrypto/dcrypto_bn.c:1323:14: warning: 'rand64' accessing 8 bytes in a region of size 0 [-Wstringop-overflow=] 1323 | if (!rand64(ctx->rnd)) | ^ board/cr50/dcrypto/dcrypto_bn.c:1323:14: note: referencing argument 1 of type 'uint32_t *' board/cr50/dcrypto/dcrypto_bn.c:1160:13: note: in a call to function 'rand64' 1160 | static bool rand64(uint32_t dst[2]) | ^ board/cr50/tpm2/ecc.c: In function '_cpri__EccPointMultiply': board/cr50/tpm2/ecc.c:81:25: warning: 'p256_to_bin' accessing 32 bytes in a region of size 1 [-Wstringop-overflow=] 81 | p256_to_bin(&out_x, out->x.b.buffer); | ^ board/cr50/tpm2/ecc.c:81:25: note: referencing argument 2 of type 'uint8_t *' board/cr50/dcrypto/p256.c:119:6: note: in a call to function 'p256_to_bin' 119 | void p256_to_bin(const p256_int *src, uint8_t dst[P256_NBYTES]) Common pattern is when function prototype defines pointer as an array of specified type. Interestingly, that in case of ctx->rnd, rnd is uint32_t rnd[2], so complain is unnecessary. In case of ecc it's hard to explain that there is enough space. BUG=None TEST=make buildall -j Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I6bc071e4b536095535b9766d14600f5cb491f118 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3183334 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: remove unnecessary dependency on trng.hVadim Sukhomlinov2021-09-246-6/+0
| | | | | | | | | | | | | | | | | | | cr50 dcrypto use different API to access TRNG - read_rand() which provides indication wherever reading was successful. Common trng.h is not needed, so remove it. BUG=None TEST=make buildall -j Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: If8525cd51db157fbfa47adbfe11146a617c947ce Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3183468 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Auto-Submit: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: minor fix for -Wsign-compare with RND_TEST=1Vadim Sukhomlinov2021-09-241-1/+1
| | | | | | | | | | | | | | | BUG=None TEST=make BOARD=cr50 CRYPTO_TEST=1 RND_TEST=1 Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I0088006cc58f57d21fa2f0d7ecffd833328cb6ca Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3183338 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Auto-Submit: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: switch HMAC_DRBG to use enum dcrypto_resultVadim Sukhomlinov2021-09-248-68/+57
| | | | | | | | | | | | | | | | Make HMAC_DRBG return codes consistent with other functions. BUG=b:197893750 TEST=make BOARD=cr50 CRYPTO_TEST=1; test/tpm_test/tpmtest.py in ccd: u2f_test, dcrypto_ecdsa, rma_auth Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I9c673a45a250bef32c096f8d8be3152756a64cb7 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3180482 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
* cr50: update TRNG continuous test logic to handle intermittent errorsVadim Sukhomlinov2021-09-242-49/+71
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | TRNG health tests have defined false positive. NIST recommends values in the range of 2^(-20) to 20^(-40) - parameter alpha. We choose 2^(-40), and computed thresholds for 2^(-30) if needed. In case of false positive we will try to read several times and update statistics to see if error is intermittent, skip those values until we either get recovered statistics or will be out of attempts. When out of attempts we declare a persistent error and report it. With this implementation we reduce probability of false positive to 2^(-160). This is in compliance with NIST SP 800-90B, 4.3 point 2: When the health tests fail, the entropy source shall notify the consuming application (e.g., the RBG) of the error condition. The developer may have defined different types of failures (e.g., intermittent and persistent), and the application is allowed to react differently to different types of failures (e.g., by inhibiting output for a short time). The developer is allowed to define different cutoff values to detect intermittent and persistent failures. If so, these values (with corresponding false alarm probabilities) shall be specified in the submission documentation. If the entropy source detects intermittent failures and allows the noise source to return to normal functioning, the designer shall provide evidence that: a) The intermittent failures handled in this way are indeed extremely likely to be intermittent failures; and b) the tests will detect a permanent failure when one occurs, and will ultimately signal an error condition to the consuming application and cease operation. In the case where a persistent failure is detected, the entropy source shall not produce any outputs. BUG=b:134594373 TEST=make BOARD=cr50 CRYPTO_TEST=1; In ccd: rand_perf rand perf (repeat several times, each time 8000 readings from TRNG) fips trng rand perf (should report errors) Signed-off-by: Vadim Sukhomlinov <sukhomlinov@google.com> Change-Id: I9db545c1a1e82e7e091724fab6fe46edebeb0650 Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3182622 Reviewed-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Reviewed-by: Andrey Pronin <apronin@chromium.org> Tested-by: Vadim Sukhomlinov <sukhomlinov@chromium.org> Commit-Queue: Vadim Sukhomlinov <sukhomlinov@chromium.org>
View Lorry Controller Status