Execute FIPS Self Test in Runtime #684

rusty1968 · 2023-08-24T18:34:14Z

This commit enables Runtime to run FIPS Self Tests as follows : KATs, verification of firmware image in ICCM, and ROM integrity test.

Light refactor to decouple KAT crate from Rom dependency.
Light refactor to decouple Image verification crate from ROM dependency.
Ability for RT to defer execution of self test (basic job control).

Runtime copies image from ICCM to mailbox SRAM , so it needs to fake a send a transaction and hold the mailbox lock while the image is being validated,. It defers the actual self test execution until the mailbox lock can be acquired. SoC must "poll" for self test completion.

SoC must send a FIPS self test command one time to schedule the execution.
Caliptra holds the mailbox lock while the self test is in progress.
Failure of FIPS self test is handled as fatal error.
Upon successful self-test completion the mailbox lock is released and SoC will receive a positive acknowledgement when FIPS Self test command is re-issued.

runtime/src/fips.rs

builder/src/lib.rs

kat/src/lib.rs

runtime/src/fips.rs

jhand2 · 2023-08-29T20:26:41Z

runtime/src/lib.rs

+    if let SelfTestStatus::InProgress(execute) = drivers.self_test_status {
+        if drivers.mbox.lock() == false {
+            match execute(drivers) {
+                Ok(_) => drivers.self_test_status = SelfTestStatus::Done,
+                Err(e) => caliptra_drivers::report_fw_error_non_fatal(e.into()),
+            }
+        }
+    }


Is there a specific reason SelfTest can't be blocking?

We need to “reply with data” to conform with new FIPS API (response contains header + checksum), but the mailbox SRAM would already be mutated with the image.

The alternative approach is to make the image-verification digest generic, using the sha384 peripheral instead of sha384acc with the mailbox.

If we want to abandon calling back into ROM forever, perhaps just using the sha384 peripheral is better...

I think we should keep this run-self-test mailbox API as is. If a future version of FIPS ever needs us to test the sha384acc peripheral with something other than empty input data we'll need a mechanism to modify the mailbox.

runtime/src/lib.rs

When callers from separate crates call a large generic function like verify_lms_signature_cfi(), rustc 1.70 may build multiple versions (depending on optimizer heuristics), even when all the generic parameters are identical. This is bad, as it can bloat the binary and the second copy violates the FIPS requirements that the same machine code be used for the KAT as the actual implementation. To defend against it, we provide a non-generic function that production firmware should call instead.

error/src/lib.rs

korran · 2023-09-05T21:56:24Z

test/tests/smoke_test.rs

+        .mailbox_execute(u32::from(CommandId::SELF_TEST), payload.as_bytes())
+        .unwrap_err();
+
+    hw.step_until_boot_status(


I don't think it is realistic to expect callers to do this. Also, I'm not sure the boot status register is really intended to be used once we've fully booted.

Perhaps the mailbox response from the SELF_TEST command should explain to the user that the test is currently in progress.

Maybe it would be simpler if this were two separate commands CommandId::SELF_TEST_START and CommandID:SELF_TEST_GET_RESULT?

korran · 2023-09-05T21:57:53Z

runtime/src/lib.rs

@@ -256,7 +291,18 @@ fn handle_command(drivers: &mut Drivers) -> CaliptraResult<MboxStatusE> {
        #[cfg(feature = "test_only_commands")]
        CommandId::TEST_ONLY_HMAC384_VERIFY => HmacVerifyCmd::execute(drivers, cmd_bytes),
        CommandId::VERSION => FipsVersionCmd::execute(drivers),
-        CommandId::SELF_TEST => FipsSelfTestCmd::execute(drivers),
+        #[cfg(feature = "fips_self_test")]
+        CommandId::SELF_TEST => match drivers.self_test_status {


I don't think you should overload the SELF_TEST command for the "start" and "get result" phases. The response in both cases is an empty success result, which isn't very informative. See my comments in smoke_test.rs.

jhand2 · 2023-09-05T23:24:36Z

runtime/src/lib.rs

+        #[cfg(feature = "fips_self_test")]
+        CommandId::SELF_TEST => match drivers.self_test_status {
+            SelfTestStatus::Idle => {
+                drivers.self_test_status = SelfTestStatus::InProgress(fips_self_test_cmd::execute);


Does this need to take a function? Or could the handler code in goto_idle just call fips_self_test_cmd::execute directly? I feel like the current code is a bit hard to follow.

runtime/src/lib.rs

When callers from separate crates call a large generic function like verify_lms_signature_cfi(), rustc 1.70 may build multiple versions (depending on optimizer heuristics), even when all the generic parameters are identical. This is bad, as it can bloat the binary and the second copy violates the FIPS requirements that the same machine code be used for the KAT as the actual implementation. To defend against it, we provide a non-generic function that production firmware should call instead.

runtime/src/fips.rs

rusty1968 · 2023-09-07T23:43:38Z

@mhatrevi , ajisaxena : please give feedback on changes in crates owned by msft.

) * Decouple dependency of Kat execution from Rom * Decouple verification environment from ROM dependency * Move fips-self-test test to test crate * Use conditional compilation to build caliptra-runtime as ROM image * Postpone self test execution * Fix 1600 byte ROM size regression in #684 When callers from separate crates call a large generic function like verify_lms_signature_cfi(), rustc 1.70 may build multiple versions (depending on optimizer heuristics), even when all the generic parameters are identical. This is bad, as it can bloat the binary and the second copy violates the FIPS requirements that the same machine code be used for the KAT as the actual implementation. To defend against it, we provide a non-generic function that production firmware should call instead. * Rename goto_idle to enter_idle * Fix typo * Renamed SELF_TEST * START/GET_PROGRESS commands * kats executes before image verification * Remove persistent_data from venv * Implement FIPS commands in ROM VERSION SELF_TEST SHUTDOWN * Fix unassigned error codes * Address feedback - Remove mut from fips create_slice - Use copy_bytes_to_mbox in write_response - Restore idevid tests mistakenly removed - Add bounds checking on fmc and rt sizes * Rom itegrity test call from Runtime. * Fixes for gate --------- Co-authored-by: Anthony Rocha <anthony.rocha@amd.com> Co-authored-by: Kor Nielsen <kor@google.com>

FerralCoder · 2023-09-11T15:18:27Z

changes were included with PR #693

rusty1968 force-pushed the antrocha/fips-invoke-rom-integrity branch from 4f3d0b2 to b8553bf Compare August 24, 2023 18:53

FerralCoder reviewed Aug 24, 2023

View reviewed changes

runtime/src/fips.rs Show resolved Hide resolved

rusty1968 force-pushed the antrocha/fips-invoke-rom-integrity branch 2 times, most recently from 3ad0673 to 1e2c0ce Compare August 27, 2023 04:35

rusty1968 changed the title ~~Execute ROM integrity tests from Runtime~~ Execute FIPS Self Test in Runtime Aug 29, 2023

jhand2 reviewed Aug 29, 2023

View reviewed changes

rusty1968 force-pushed the antrocha/fips-invoke-rom-integrity branch 3 times, most recently from 8bd794f to 6734392 Compare August 29, 2023 22:28

rusty1968 marked this pull request as ready for review August 29, 2023 22:43

rusty1968 requested review from fdamato, bluegate010, mhatrevi, ajisaxena, vsonims, korran and JohnTraverAmd as code owners August 29, 2023 22:43

FerralCoder previously approved these changes Aug 31, 2023

View reviewed changes

korran reviewed Sep 5, 2023

View reviewed changes

error/src/lib.rs Outdated Show resolved Hide resolved

korran reviewed Sep 5, 2023

View reviewed changes

jhand2 reviewed Sep 5, 2023

View reviewed changes

rusty1968 dismissed FerralCoder’s stale review via dbd83b8 September 7, 2023 17:39

rusty1968 force-pushed the antrocha/fips-invoke-rom-integrity branch from 8c49288 to dbd83b8 Compare September 7, 2023 17:39

rusty1968 force-pushed the antrocha/fips-invoke-rom-integrity branch from dbd83b8 to 1a6b7ff Compare September 7, 2023 17:53

rusty1968 and others added 10 commits September 7, 2023 13:02

Decouple dependency of Kat execution from Rom

ad2d8c8

Decouple verification environment from ROM dependency

ac04a2e

Move fips-self-test test to test crate

e8a0249

Use conditional compilation to build caliptra-runtime as ROM image

beb9289

Postpone self test execution

2b5b5ab

Rename goto_idle to enter_idle

95c631d

Fix typo

b391327

Renamed SELF_TEST

b50b422

START/GET_PROGRESS commands

bcf5b40

rusty1968 force-pushed the antrocha/fips-invoke-rom-integrity branch from fe40e89 to bcf5b40 Compare September 7, 2023 21:36

mhatrevi requested changes Sep 7, 2023

View reviewed changes

runtime/src/fips.rs Show resolved Hide resolved

kats executes before image verification

5f573c9

mhatrevi approved these changes Sep 8, 2023

View reviewed changes

FerralCoder closed this Sep 11, 2023

rusty1968 deleted the antrocha/fips-invoke-rom-integrity branch October 20, 2024 23:48

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Execute FIPS Self Test in Runtime #684

Execute FIPS Self Test in Runtime #684

rusty1968 commented Aug 24, 2023 •

edited

Loading

jhand2 Aug 29, 2023

rusty1968 Aug 29, 2023

korran Aug 30, 2023

korran Sep 5, 2023

korran Sep 5, 2023

rusty1968 Sep 7, 2023

korran Sep 5, 2023

rusty1968 Sep 7, 2023

jhand2 Sep 5, 2023

rusty1968 commented Sep 7, 2023

FerralCoder commented Sep 11, 2023 •

edited

Loading

Execute FIPS Self Test in Runtime #684

Execute FIPS Self Test in Runtime #684

Conversation

rusty1968 commented Aug 24, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rusty1968 commented Sep 7, 2023

FerralCoder commented Sep 11, 2023 • edited Loading

rusty1968 commented Aug 24, 2023 •

edited

Loading

FerralCoder commented Sep 11, 2023 •

edited

Loading