[edn] V2S Signoff

[andreaskurth]

For reference, these are previous signoff issues:

• EDN D2S #20982
• EDN V1 #21017

[vogelpi]

Commits since Earlgrey-PROD.M2

For Commits since Earlgrey-ES tapeout, see #20982 (comment)

git rev-parse --short HEAD bddd787

git log 9ddf276..HEAD --oneline hw/ip/edn
04a79b6 [csrng/dv] Increase number of seeds for edn_alert test
8eebd19 [csrng/dv] Rework edn_alert test
b30e5b3 [edn] Hold csrng_req_valid until valid/ready handshake completes
-> RTL bug fixes to respect valdi/ready handshake protocol in error case and fix error signaling from CSRNG (see #23526)
3289508 [edn/dv] Add transition to reject entropy state to alert test
-> Close FSM coverage holes, see #22352.
975eadb [edn/rtl] Fix bugs for SW and HW cmd status interface
-> RTL bug fixes flagged by DV commit above
7c92d49 [edn/rtl] Fix Verilator lint errors and warnings
0c98679 [edn/dv] Fix FSM coverage holes
-> Close FSM coverage holes, see #22352.
53868ed [edn/rtl] Make Error main SM state terminal again
-> RTL bug fix
bfed776 [edn/dv] Fix some edge cases for edn_disable_auto_req_mode_vseq
6450330 [edn/dv] This commit fixes the deadlock in edn_disable_auto_req_mode
-> Fix edn_disable_auto_req_mode sequence, see #20818
7f84f7c [csrng/rtl] Add reseed interval status error
-> Add new error response to EDN
6e8108c [edn/doc] Switch back to EDN's registers.md
1d12ff7 [csrng/rtl] Remove zeroize state after uni error from ctr drbg cmd
-> Removing an error condition inside CSRNG which was impossible to reach, aligning error response to EDN
dcaf19e [edn/dv] Correct prediction of sw_cmd_reg_rdy for auto mode Instantiate
903b319 [csrng/rtl] Add command out of sequence cmd_sts error
-> Add new error response to EDN
375c7bd [edn/dv] fix broken private CI tests
b389e28 [edn/dv] Remove leftover of removed output FIFO in edn_err test
00fc69d [edn] Harden FIFO instantiations
-> Use hardened prim_sync_fifo instantiations inside EDN, this standardized countermeasures is verified via auto generated edn_sec_cm test
9a6b449 [edn,rtl] Fix assertions
042e0b9 [csrng/rtl] Change and assign cmd_sts signal
-> Alignment of error response to EDN
8295903 [bazel,autogen_hjson] Split C and rust header generation
30d7e78 (tag: Earlgrey-PROD-M2-RC5, tag: Earlgrey-PROD-M2) Add the project name to the copyright header
f4c2bb9 Remove trailing whitespaces
73c4e33 [edn] Increase version number, revert verification stage to V1

Issues closed since Earlgrey-PROD.M2

For issues closed since Earlgrey-ES tapeout, see #20982 (comment)

• [csrng] Define how CSRNG status response should operate #16516
  Reworking error status responses between EDN and CSRNG, led to various PRs, the EDN side of things was mostly handled as part of M2 already, i.e. ,is part of the D2S signoff ([edn] D2S Signoff #20982).
• [csrng] Don't handle Generate, Reseed, or Update commands on uninstantiated interfaces #19293
  A change to CSRNG related to error status handling, required alignment of EDN.
• [csrng] SW instance instantiate command hangs after entropy complex initialization #19568
  Fixed by a a PR in M2 ([csrng/rtl] Propagate csrng errors to edn through cmd sts signal #21981), software workaround later removed in [crypto-lib/csrng] Remove wait for idle in csrng_send_app_cmd #22263. Confirmed that the issue is resolved.
• [edn/dv] edn_disable_auto_req_mode test sequence failing #20818
  Debugged and fixed a test sequence failure.
• [edn] D2S Signoff #20982
• [edn] Consider modifying FIFO instantiations #21155
  Hardening FIFO instantiations inside EDN, done with [edn] Harden FIFO instantiations #22599
• [edn] Investigate FSM coverage #22352
  FSM coverage holes fixed with [edn] Improve FSM coverage #23092 and [edn/dv] Add transition to reject entropy state to alert test #23120.
• [fpv] Inspect formal FPV results and get them back up to 100% pass rate #22385
  EDN FPV results improved with [edn,rtl] Fix indices in assertion #22587 .
• [cryptolib] Correct flag0 values for CSRNG and EDN configurations #22524
  A cryptolib bug that got resolved.
• [csrng/edn] Consider making FIPS flag software defined #22732
  Resolved with [csrng] Add option to force FIPS/CC compliance flag #23324 who makes the FIPS/CC compliance flag software defined to enable KATs on OTBN RND.
• [entropy_complex] Debug hang issues #22753
  Lots of debugging, see [entropy_complex] Debug hang issues #22753 (comment) for a detailed report. Discovered a cryptolib bug (fixed with [cryptolib] Fix CSRNG driver #23166), tightened a csrng_edn_concurrency TLT ([chip-level/sival] Tighten csrng_edn_concurrency test #23277) and increase number of seeds run in nightlies ([chip-test/sival] Optimize csrng_edn_concurrency test and run more seeds in nightlies #23314). Not a single test failure was recorded since Apr 15 for this TLT.
• [edn] Investigate why there are many EDN requests after ROM_EXT #22819
  These requests turned out to be due to keymgr continuously requesting lots of entropy. Fixed by RTL changes for PROD and somewhat alleviated for ES using configuration changes, see [edn] Investigate why there are many EDN requests after ROM_EXT #22819 (comment) for details.
• [keymgr] Reduce reseeding and PRNG activity during Disabled and Invalid states #22997
  Follow-up issue of the above issue where keymgr was still consuming lots of entropy in disabled and invalid states. Fixed by [keymgr] Don't update and reseed PRNG in Disabled/Invalid state forever #23101.
• [edn] valid/ready protocol violations in case of ack errors and disablement #23526
  EDN / CSRNG violated the valid/ready handshake protocol between them in the case of errors. EDN side fixed by [edn] Hold csrng_req_valid until valid/ready handshake completes #23607.

Currently open issues

• [top, edn] Future enhancement to edn interface to various blocks #7667
  FutureRelease
• [hw/csrng] Integrity protection for KV in CSRNG #10738
  FutureRelease
• [req/ack] Examine req/ack interfaces between IP blocks and enhance their security #11172
  It was concluded that the EDN interfaces aren't an issue.
• [cryptolib] continuous mode config incl. health checks for entropy_src #14364
  Cryptolib
• [dv/sec_cm] Add assertions to check if local escalation can trigger fatal alerts #16269
  Tracked for V3 / M5.
• [dv] More targetted FI testing for hardened FSMs and coverage #16572
  Tracked for M5.
• [dv,fpv] Automatically add FSM asserts #16593
  Tracked for M5.
• [top_darjeeling] RNG integration #17109
  Relevant for ENTROPY_SRC and Darjeeling / other top levels only.
• [keymgr, kmac] Entropy error code SW guideline #17785
  Cryptolib
• [edn/dv] Improve testing and coverage of endpoint requests during fatal alerts #19027
  Tracked for V3 / M5.
• [edn/dv] Randomly enable regwen in as many tests as possible #19028
  Tracked for V3 / M5.
• [edn] FIFO error does not cause local escalation #19789
  Fixed by aligning EDN, CSRNG and ENTROPY_SRC with [entropy_src] Locally escalate upon internal FIFO errors and FSM errors #23560 and [csrng] Locally escalate upon FSM errors #23561
• [bazel, dvsim] Convert dvsim-executed tests to the new rules #19797
  EDN-related tests already taken care of.
• [chip-test] chip_sw_csrng_edn_cmd #19846
  SiVal for CSRNG
• [sival] prod_all_tests #20194
  SiVal
• [edn/dv] Add checks on the alert signal #21216
  Coverage hole, tracked for V3 / M5. Related to #$16269.
• [chip-test] Incomplete SiVal Stage 1 and 2 tests #21222
  SiVal,
• [sival,triag] Some SiVal ROM_EXT are broken #21706
  Broken SiVal tests, entropy complex related ones got fixed see [sival] Mark fixed entropy complex tests as unbroken in ROM_EXT #23035
• [prodc-integration] Normal Sleep Power Consumption High #22241
  Debugged power consumption due to high entropy consumption of keymgr, fixed the RTL, issue moved to Earlgrey-SiVal for re-evaluation of production silicon
• [entropy_src] Power management functionality and tests #22293
  Mostly questions for which answers can be found in the documentation.
• [edn, doc, rom] EDN documentation requires updates #22366
  Documentation updates to be implemented when there is time.
• [edn/sw] Return the whole sts code in dif_edn_get_status #22465
  A ROM_EXT software issue.
• [edn] V2S Signoff #22468
• [edn] D3 Signoff #22634
• [edn] V3 Signoff #22667
• [entropy_src] Integration and top-level testing extensions #22705
  Possible TLT extensions for the entropy complex, to be moved to M5.
• [entropy_src,csrng,edn] Features for next version / future releases of entropy complex #22850
  FutureRelease

Coverage report from 2024-06-11

After fixing the FSM coverage holes, all coverage metrics are again above the V2(S) threshold. All tests have a pass rate above 90% (there is one single test failure in the whole regression run due to a test time out).

Summary

Since we signed EDN off at D2S and V1 as part of M2, most work went into the other blocks of the entropy complex (ENTROPY_SRC and in particular CSRNG). But the coverage holes and test failures preventing from signing off EDN again at V2S previously have now been fixed.

Confidence in the block is high. There are no open RTL issues. I thus suggest to sign off EDN at V2S. WDYT? @andreaskurth @h-filali ?

[moidx]

Discussed moving as P1 to M5 due to RTL risk analysis performed during last issue triage.

[vogelpi]

Update: I've now updated the signoff issue to reflect that the RTL issue found in the last minute (#23526 ) has also been fixed.

[h-filali]

I double checked for all my PR's if there are some open issues and couldn't find any. I also double checked all the issues I've been working on and they are all mentioned here.

I also double checked the last nightly regression and can confirm that everything is above 90%.
The nightly that ran on 2024.06.10_06.28.44 had ~82% progress for V2 but the errors that went away in the most recent nightly align exactly with the changes made by @vogelpi described in #23526 and #23607.

I double checked if I could find any commits for EDN that are missing but I couldn't find any.

One of the open issues in the list above is now closed (#19789).

To me this looks good. Nicely done @vogelpi ^^

[vogelpi]

Thanks for your feedback @h-filali . I've now filed the PR (#23700) to close this issue and move EDN v.2.0.0 to V2S.