[rv_core_ibex] D2S Signoff

[msfschaffner]

Description

Ensure D2S signoff criteria are fulfilled after focus area changes have landed.

[GregAC]

Commits since Earlgrey-ES tapeout

Review hash: 48fd6f0

6766fc8 [rtl, rv_core_ibex] Add alert assertion for lockstep reset count
b402d1c [rv_core_ibex,lint] Update lint waiver with lockstep reset changes
5639924 Revert "[edn] Move prim_edn_req out of prim"
61a237e [util/reggen] reverse order of substruct generation
fc84846 [reggen,hw] Create index parameter for registers windows
ce648ca [ipgen.pwrmgr] Change core files to vlnv naming and label as virtual
b2e29a5 [rv_core_ibex] Update AscentLint waiver for newly added onehot muxes
818f1fd [rv_core_ibex] Connect refgile onehot checker asserts
3b4e36e [edn] Move prim_edn_req out of prim
de31bdf [reggen] Remove the devmode input
963a500 [doc] Minor tweak to md sanitisation code
ca2b62b [dv, testplan] Replace descr by desc for consistency
1ce014f [hw,ibex,rtl] Change fatal alert to for RW1S register
619ce83 [rv_core_ibex, sival] Extend rv_core_ibex testplan with SiVal tests
1b16ca2 [reggen] Add mubi support SWAccess that sets/clears a reg
59f8142 [doc] Moved badges over to using hosted images
7084320 [doc] rv_core_ibex registers and interfaces now use CMDGEN
7688e71 [reggen] Add initial support for version and cip_id hjson fields
fbd888e Revert "[reggen] Add CIP_IDs and bump all major versions"
0ba10b3 [reggen] Add CIP_IDs and bump all major versions
4e615de [rtl/rv_core_ibex] Fix RTL assertion for CDC

Of these commits the following effect RTL:
fc84846 - Generic reggen change
b2e29a5 - Assertion waiver (clk,rst unused by logic, for assertions only)
818f1fd - Add onehot checker assertions
de31bdf - Remove devmode input
1ce014f - Switch software fatal alert to RW1S register
1b16ca2 - prim_subreg changes, no functional change
4e615de - Assertion fix/tweak
6766fc8 - Assertion to check lockstep reset prim_count alert connection
b402d1c - Waiver changes due to lockstep reset changes
61a237e - Reggen changes

Only 1ce014f produces a functional change, which was required for a bug fix
(#19774). The rest have no RTL impact
or are part of wider generic primitive and reggen changes applied across
Earlgrey.

The Ibex core itself is vendored in from the Ibex repository. Changes in that
repository that related to the Ibex RTL also need to be considered. Those
changes that effect RTL files are below:

lowRISC/ibex@27dd6b2e [rtl] Update use of prim_count following port changes
lowRISC/ibex@5a8a1a99 [tracer] Fix reporting of load/store data
lowRISC/ibex@8ec0c6f1 [rtl] Harden lockstep enable against FI
lowRISC/ibex@56413ecf [icache] Disable S&P diffusion layer in memory scrambling
lowRISC/ibex@35bbdb7b [rtl] Fix FI vulnerability in RF
lowRISC/ibex@d097c918 [rtl] Avoid name collision in ibex_pmp.sv
lowRISC/ibex@fe84d64d [verilator] Slight refactor in ibex_tracer to avoid BLKSEQ warning
lowRISC/ibex@bac72d96 [ibex_pmp/lint] Declare functions before using them
lowRISC/ibex@1084ac11 [dv] Add asserts to check alerts for memory integrity failures

Of these changes only 3 actually effect RTL functionality.

• lowRISC/ibex@8ec0c6f1 and lowRISC/ibex@35bbdb7b are security fixes that are
  wanted for PROD
• lowRISC/ibex@56413ecf alters the funtionality of the memory scrambling
  removing the S&P diffusion layer, see details here: RFC: Removing the S&P Layer in Scrambling for Top Earl Grey #20788

Closed Issues

• [chip-test] chip_sw_rv_core_ibex_alerts #21461
  • Sival test related, no RTL impact
• [rv_core_ibex] Vendor in latest Ibex #21128
  • Brings in latest Ibex RTL, seperately considered under Ibex RTL changes
• [rv_core_ibex] Switch single to multi-bit encoding for Lockstep Enable #20779 and [rv_core_ibex] Bit-flip in Register File Muxing #20715
  • Fix for security issues, agreed to be included in PROD, fixed in Ibex commits:
    • lowRISC/ibex@8ec0c6f1
    • lowRISC/ibex@35bbdb7b
• [ip,doc] Fix inaccurate references to reset behavior #20700
  • Documentation related, no RTL impact
• [rv_core_ibex,rtl] Incorrect behavior of SW_FATAL_ERR CSR #19774
  • Incorrect rv_core_ibex SW_FATAL_ERR CSR behaviour, fixed by 1ce014f
• [rv_core_ibex] Consider implementing ECC check across all ways of data RAM in ICache #11403
  • Potential enchancement, concluded disadvantages outweigh potential benefits

Open Issues

• [rv_core_ibex,doc] Typo in documentation #20453
  • Doc typo, no RTL impact
• [rv_core_ibex,RAM] Setup violation from the RAMs #19782
  • Potential timing issues, will be addressed post M2 when we have a new timing report, not a gate to D2
• [rv_core_ibex] Improve V2(S) pass rates #18421
  • Issue tracking test pass rates, not relevant to D2/PROD M2
• [rv_core_ibex/top, rtl] Regularise use of timer interrupt #18093
  • Future enchancement, not planned for PROD
• [chip-test] chip_sw_rv_core_ibex_lockstep_glitch #21460
• [chip-test] chip_sw_rv_core_ibex_alerts #16838
  • SiVal test work, no RTL impact
• [chip, dv, cov] Address low coverage of alert_dump and crashdump signals #15427
  • Potential verification improvement (discussed at end of issue), no RTL impact
• [rv_core_ibex] Consider ICache pinning #14510
  • New feature suggestion, not being included for PROD
• [rv_core_ibex,dv] Integrate the Ibex "core_ibex" tests with dvsim #11794
  • DV flow improvements, no RTL impact
• [rtl/rstmgr] Make cpu and alert info layout known to sw #8945
  • Tooling improvements around crash dumps
• [rv_core_ibex] Suggested Address Translation improvements #20896
  • Suggested improvements to address translation, not being implemented for PROD
• [top_darjeeling] Wrong ePMP reset values #20489
  • Only relevant to darjeeling
• [tlul] How PutFullData vs PutPartialData should be used #1867
  • Future release, not PROD relevant

I am happy for rv_core_ibex to retain its D2(S) status.

[GregAC]

@msfschaffner @andreaskurth @rswarbrick PTAL

[msfschaffner]

LGTM as well. Anything that warrants moving to v1.0.0 or are we happy to leave as-is?
I presume the fix for the alert CSR is actually a bugfix, so it just brings the existing interface into alignment with the spec.

[GregAC]

I think a move to 1.0 makes sense, probably should have happened a while ago.

[GregAC]

Closing now as the switch to v1.0 isn't required for d2

[vogelpi]

From the discussion above it's not fully clear whether Ibex was signed off at D2 or D2S. Since it there was no version increase, no new and or software visible features were added and mostly security fixes were done for M2, I am happy for Ibex to remain at D2S. I am thus changing the title of the issue.