SRAM_CTRL DV document

Goals

  • DV
    • Verify all SRAM_CTRL IP features by running dynamic simulations with a SV/UVM based testbench
    • Develop and run all tests based on the testplan below towards closing code and functional coverage on the IP and all of its sub-modules
  • FPV
    • Verify TileLink device protocol compliance with an SVA based testbench

Current status

Design features

For detailed information on SRAM_CTRL design features, please see the SRAM_CTRL HWIP technical specification.

Testbench architecture

SRAM_CTRL testbench has been constructed based on the CIP testbench architecture. Note that there are 2 separate TLUL interfaces exposed to the rest of the OpenTitan system - one for CSR accesses, and one for accessing memory directly. This is because the “full” DUT consists of the actual SRAM memory controller (which contains the CSR file) connected to a scrambling RAM primitive, with a TLUL adapter module to convert TL requests on the memory TL interface into SRAM format for the RAM primitive.

Block diagram

Block diagram

Top level testbench

Top level testbench is located at hw/ip/sram_ctrl/dv/tb/tb.sv. It instantiates the SRAM_CTRL DUT module hw/ip/sram_ctrl/rtl/sram_ctrl.sv. In addition, it instantiates the following interfaces, connects them to the DUT and sets their handle into uvm_config_db:

Common DV utility components

The following utilities provide generic helper tasks and functions to perform activities that are common across the project:

Compile-time configurations

Two compile-time configurations are tested:

  • sram_ctrl_main - this configuration uses a 15-bit wide address space to emulate the main SRAM used in the full design
  • sram_ctrl_ret - this configuration uses a 10-bit-wide address space to emulate the retention SRAM used in the full design

A macro-define SRAM_ADDR_WIDTH is defined as a build option in hw/ip/sram_ctrl/dv/sram_ctrl_base_sim_cfg.hjson, which is used to set the correct compile-time settings for each configuration.

Global types & methods

All common types and methods defined at the package level can be found in sram_ctrl_env_pkg. Some of them in use are:

// number of bits in a full KDI transactions
parameter int KDI_DATA_SIZE = 1 + otp_ctrl_pkg::SramKeyWidth + otp_ctrl_pkg::SramNonceWidth;

// a completed KDI transaction takes 4 cycles to be fully synchronized into the
// SRAM clock domain
parameter int KDI_PROPAGATION_CYCLES = 4;

// a LC escalation request needs 3 cycles to be fully propagated through the DUT
parameter int LC_ESCALATION_PROPAGATION_CYCLES = 3;

typedef enum bit {
  SramCtrlRenewScrKey = 0,
  SramCtrlInit        = 1
} sram_ctrl_e;

typedef enum bit [1:0] {
  SramCtrlError           = 0,
  SramCtrlEscalated       = 1,
  SramCtrlScrKeyValid     = 2,
  SramCtrlScrKeySeedValid = 3
} sram_ctrl_status_e;

TL_agent

SRAM_CTRL testbench instantiates (already handled in CIP base env) tl_agent which provides the ability to drive and independently monitor random traffic via TL host interface into SRAM_CTRL device.

Alert agents

The SRAM_CTRL testbench instantiates 2 alert agents for:

  • fatal_intg_error - signals a transmission integrity error
  • fatal_parity_error - signals an internal parity error in the memory primitive

The alert agents provide the ability to independently drive and monitor both alert handshakes.

SRAM_CTRL_LC_IF

The SRAM_CTRL IP has a simple sideband interface to the LC_CTRL to receive escalation requests. This interface initializes the escalation bus and utilizes a simple task to drive escalation requests.

SRAM_CTRL_EXEC_IF

The SRAM_CTRL IP has an interface to enable instruction execution from SRAM - allowing the Icache to fetch instruction data from the SRAM for the CPU. This interface contains the necessary Lifecycle and OTP structs to enable and disable this functionality.

UVM RAL Model

The SRAM_CTRL RAL model is created with the ralgen FuseSoC generator script automatically when the simulation is at the build stage.

It can be created manually by invoking regtool:

Reference models

The SRAM_CTRL testbench uses a SystemVerilog golden model of its address and data scrambling to ensure that internal encryption and decryption are performed correctly. This golden model, sram_scrambler_pkg is tightly integrated with the mem_bkdr_util for ease of use such that we can choose to enable encryption on any backdoor access.

Stimulus strategy

Test sequences

All test sequences reside in hw/ip/sram_ctrl/dv/env/seq_lib. The sram_ctrl_base_vseq virtual sequence is extended from cip_base_vseq and serves as a starting point. All test sequences are extended from sram_ctrl_base_vseq. It provides commonly used handles, variables, functions and tasks that the test sequences can simple use / call. Some of the most commonly used tasks / functions are as follows:

  • do_rand_ops: This task sends an arbitrary number of random TL transactions to the memory primitive.
  • do_stress_ops: This task sends an arbitrary number of random TL transactions to the same word in memory, to stress the memory’s data forwarding functionalities.

Functional coverage

To ensure high quality constrained random stimulus, it is necessary to develop a functional coverage model. The following covergroups have been developed to prove that the test intent has been adequately met:

  • cg1:
  • cg2:

Self-checking strategy

Scoreboard

The sram_ctrl_scoreboard is primarily used for end to end checking. It creates the following analysis ports to retrieve the data monitored by corresponding interface agents:

  • tl_a_chan_fifo: TL address channel for CSR accesses
  • tl_d_chan_fifo: TL data channel for CSR accesses
  • alert_fifos: Alert handshakes
  • sram_tl_a_chan_fifo: TL address channel for memory accesses
  • sram_tl_d_chan_fifo: TL data channel for memory accesses
  • kdi_fifo: For key refresh operations from OTP_CTRL

The following internal data structure is used to store information about each requested memory access:

typedef struct {
  // 1 for writes, 0 for reads
  bit we;

  // TLUL address
  bit [TL_AW-1:0] addr;

  // Contains either the requested write data or the read response data
  bit [TL_DW-1:0] data;

  // used to mask read and write data to the requested granularity
  bit [TL_DBW-1:0] mask;

  // Tag the memory transaction with the appropriate key and nonce,
  // so that we can keep track even if a new key is requested.
  otp_ctrl_pkg::sram_key_t key;
  otp_ctrl_pkg::sram_nonce_t nonce;

} sram_trans_t;

All CSR accesses made to the SRAM_CTRL register file are tracked and predicted by the scoreboard.

Verifying memory accesses is somewhat involved, and makes heavy use of the mem_bkdr_util.

At a high level, whenever a TL access is made to memory, all relevant information is stored in an sram_trans_t struct.

The scrambling RAM primitive contains a data holding register, which is used to forward unwritten write data in case a read transaction immediately follows a write transaction. Note that the pending write data is held no matter if the back-to-back read transaction accesses the same address in memory or not. This means that every read transaction will return data in 1 cycle, but a write request may not actually modify the memory until many cycles later when any pending read request has been processed.

This data forwarding is modeled in the scoreboard such that we check all memory transactions in the exact same order that they are actually executed by the RAM primitive.

At transactions complete, we use the memory backdoor interface to read the word in memory at the transaction’s address and compare it against the requested write data or the read response data, using abyte-mask corresponding to the granularity of the memory access.

We use a specific set of methods in the memory backdoor interface that utilize the memory scrambling golden model in order to provide the testbench with the correct unencrypted data for these comparisons.

For example, consider a memory access that writes 1 to address 0x0. Internally, the scrambling RAM primitive will map address 0x0 to some other address, say scrambled address 0x4, and will also encrypt the data. In the scoreboard we detect this transaction on the TLUL bus and use the memory backdoor interface and scrambling golden model to perform a backdoor read to logical address 0x0. This will access the correct scrambled location in memory, decrypt the data there, and return it to the scoreboard. This makes it easier to diagnose issues relating to incorrect addresses, and to data encryption or decryption.

The testbench will also sporadically trigger the DUT to issue a request for a new key and nonce from the OTP controller. After receiving the fresh key and nonce from the KDI agent, those new values will then be used for all encryption/decryption accesses using the memory backdoor interface.

If a lifecycle escalation request is issued during SRAM operation, the scoreboard will detect it and then check that no further memory requests are accepted. The testbench will then issue a system-level reset, as a lifecycle escalation is a terminal state.

To check that the SRAM’s executable configurations are functioning correct is relatively more straightforward. First, it’s important to note that all incoming TL memory transactions are tagged as either InstrType or DataType, indicating whether the transaction is fetching an instruction word or a data word. If the scoreboard detects that an incoming transaction is tagged as InstrType, we first check that the SRAM is properly configured in executable mode to accept InstrType transactions. If the DUT is configured correctly the scoreboard will let the memory transaction finish being checked, otherwise the scoreboard will discard that transaction - this approach makes it easy to check whether the design is behaving correctly as well.

Assertions

  • TLUL assertions: The tb/sram_ctrl_bind.sv binds the tlul_assert assertions to the IP to ensure TileLink interface protocol compliance.
  • Unknown checks on DUT outputs: The RTL has assertions to ensure all outputs are initialized to known values after coming out of reset.

Building and running tests

We are using our in-house developed regression tool for building and running our tests and regressions. Please take a look at the link for detailed information on the usage, capabilities, features and known issues. Here’s how to run a smoke test:

$ $REPO_TOP/util/dvsim/dvsim.py $REPO_TOP/hw/ip/sram_ctrl/dv/sram_ctrl_sim_cfg.hjson -i sram_ctrl_smoke

Testplan

Testpoints

Milestone Name Tests Description
V1 smoke _smoke

This test performs basic SRAM initialization procedure and tests basic memory function:

  • Initialize SRAM memory to zero
  • Perform some random memory operations, verify that they all succeed with an all-zero key and nonce
  • Request a new scrambling key from the OTP interface and verify that:
    • A valid key is received
    • The key seed used by OTP is valid
  • Perform a number of random memory accesses to the SRAM, verify that all accesses were executed correctly using the mem_bkdr_util
V1 csr_hw_reset sram_ctrl_csr_hw_reset

Verify the reset values as indicated in the RAL specification.

  • Write all CSRs with a random value.
  • Apply reset to the DUT as well as the RAL model.
  • Read each CSR and compare it against the reset value. it is mandatory to replicate this test for each reset that affects all or a subset of the CSRs.
  • It is mandatory to run this test for all available interfaces the CSRs are accessible from.
  • Shuffle the list of CSRs first to remove the effect of ordering.
V1 csr_rw sram_ctrl_csr_rw

Verify accessibility of CSRs as indicated in the RAL specification.

  • Loop through each CSR to write it with a random value.
  • Read the CSR back and check for correctness while adhering to its access policies.
  • It is mandatory to run this test for all available interfaces the CSRs are accessible from.
  • Shuffle the list of CSRs first to remove the effect of ordering.
V1 csr_bit_bash sram_ctrl_csr_bit_bash

Verify no aliasing within individual bits of a CSR.

  • Walk a 1 through each CSR by flipping 1 bit at a time.
  • Read the CSR back and check for correctness while adhering to its access policies.
  • This verify that writing a specific bit within the CSR did not affect any of the other bits.
  • It is mandatory to run this test for all available interfaces the CSRs are accessible from.
  • Shuffle the list of CSRs first to remove the effect of ordering.
V1 csr_aliasing sram_ctrl_csr_aliasing

Verify no aliasing within the CSR address space.

  • Loop through each CSR to write it with a random value
  • Shuffle and read ALL CSRs back.
  • All CSRs except for the one that was written in this iteration should read back the previous value.
  • The CSR that was written in this iteration is checked for correctness while adhering to its access policies.
  • It is mandatory to run this test for all available interfaces the CSRs are accessible from.
  • Shuffle the list of CSRs first to remove the effect of ordering.
V1 csr_mem_rw_with_rand_resetsram_ctrl_csr_mem_rw_with_rand_reset

Verify random reset during CSR/memory access.

  • Run csr_rw sequence to randomly access CSRs
  • If memory exists, run mem_partial_access in parallel with csr_rw
  • Randomly issue reset and then use hw_reset sequence to check all CSRs are reset to default value
  • It is mandatory to run this test for all available interfaces the CSRs are accessible from.
V2 multiple_keys _multiple_keys

In this test we request multiple scrambling keys from OTP and verify that the memory scrambling is performed correctly even with multiple seeds. Perform the following steps:

  • Initialize the memory to zero
  • Perform some random memory operations, verify that they succeed with an all-zero key and nonce
  • Repeat the following steps a number of times:
    • Get a scrambling key from the OTP interface
    • Perform a number of random memory accesses to the SRAM
  • Verify that all memory access succeed even if the scrambling key changes at arbitrary intervals
V2 stress_pipeline _stress_pipeline

This test is the same as the multiple_keys_test but we now do a series of back-to-back memory accesses at each random address in order to create read/write conflicts and stress the encryption pipeline.

V2 bijection _bijection

In this test we iterate through each address in the SRAM memory. For each address write the current address to the SRAM.

After this is done, read every address and check that the stored data is equivalent to the current address.

This will verify that the SRAM encryption mechanism is actually bijective, and will not cause any address collisions.

e.g. if the encryption scheme causes addresses 0x1 and 0x2 to collide and we write 0x1 and 0x2 respectively, we will see a return value of 0x2 when we read from 0x1, instead of the expected 0x1.

This process will be repeated for a number of new key seeds.

V2 mem_tl_errors _mem_tl_errors

This test will reuse the common tl_access_tests to run TLUL error sequences on the SRAM TLUL interface to verify that erroneous TLUL transactions are handled correctly.

V2 access_during_key_req _access_during_key_req

This test is the same as the multiple_keys test, except we make sure to sequence some memory transactions while a key request to OTP is still pending. Verify that these transactions are completely ignored by the memory.

TODO: Behavior might change in future to throw an error instead of ignore, should be reflected in TB.

V2 lc_escalation _lc_escalation

This test is the same as the multiple_keys test, except we now randomly assert the lifecycle escalation signal. Upon sending an escalation request, we verify that the DUT has properly latched it, and all scrambling state has been reset. In this state, we perform some memory accesses, they should all be blocked and not go through. We then issue a reset to the SRAM to get it out of the terminal state, and issue a couple of memory accesses just to make sure everything is still in working order.

V2 executable _executable

This test is intended to test the "execute from SRAM" feature, in which TLUL memory transactions tagged with the InstrType value in the user bits are allowed to be handled by the SRAM memory.

This behavior is enabled by either setting the exec CSR to 1 or by driving a second lifecycle input to On - both of these are muxed between with a otp_en_sram_ifetch_i input from the OTP controller.

If this functionality is disabled, any memory transaction NOT tagged as DataType should error out, however DataType transactions should be successful when the SRAM is configured to be executable.

V2 alert_test sram_ctrl_alert_test

Verify common alert_test CSR that allows SW to mock-inject alert requests.

  • Enable a random set of alert requests by writing random value to alert_test CSR.
  • Check each alert_tx.alert_p pin to verify that only the requested alerts are triggered.
  • During alert_handshakes, write alert_test CSR again to verify that: If alert_test writes to current ongoing alert handshake, the alert_test request will be ignored. If alert_test writes to current idle alert handshake, a new alert_handshake should be triggered.
  • Wait for the alert handshakes to finish and verify alert_tx.alert_p pins all sets back to 0.
  • Repeat the above steps a bunch of times.
V2 tl_d_oob_addr_access sram_ctrl_tl_errors

Access out of bounds address and verify correctness of response / behavior

V2 tl_d_illegal_access sram_ctrl_tl_errors

Drive unsupported requests via TL interface and verify correctness of response / behavior. Below error cases are tested bases on the [TLUL spec]({{< relref "hw/ip/tlul/doc/_index.md#explicit-error-cases" >}})

  • TL-UL protocol error cases
    • invalid opcode
    • some mask bits not set when opcode is PutFullData
    • mask does not match the transfer size, e.g. a_address = 0x00, a_size = 0, a_mask = 'b0010
    • mask and address misaligned, e.g. a_address = 0x01, a_mask = 'b0001
    • address and size aren't aligned, e.g. a_address = 0x01, a_size != 0
    • size is greater than 2
  • OpenTitan defined error cases
    • access unmapped address, expect d_error = 1 when devmode_i == 1
    • write a CSR with unaligned address, e.g. a_address[1:0] != 0
    • write a CSR less than its width, e.g. when CSR is 2 bytes wide, only write 1 byte
    • write a memory with a_mask != '1 when it doesn't support partial accesses
    • read a WO (write-only) memory
    • write a RO (read-only) memory
V2 tl_d_outstanding_access sram_ctrl_csr_hw_reset
sram_ctrl_csr_rw
sram_ctrl_csr_aliasing
sram_ctrl_same_csr_outstanding

Drive back-to-back requests without waiting for response to ensure there is one transaction outstanding within the TL device. Also, verify one outstanding when back- to-back accesses are made to the same address.

V2 tl_d_partial_access sram_ctrl_csr_hw_reset
sram_ctrl_csr_rw
sram_ctrl_csr_aliasing
sram_ctrl_same_csr_outstanding

Access CSR with one or more bytes of data. For read, expect to return all word value of the CSR. For write, enabling bytes should cover all CSR valid fields.

V3 parity _parity

TODO - to be changed into an ECC test

This test is the same as the multiple_keys test, except we randomly inject a parity error into the memory (TODO: figure out how exactly to do this). Verify that the SRAM reports the error and the faulty address correctly, and that the alert is sent out properly. We then perform some memory accesses and verify that none of them go through. This error is terminal, so like the lc_escalation test, issue a reset and then perform some memory accesses to make sure everything comes back online correctly.

V3 tl_intg_err sram_ctrl_tl_intg_err

Verify that the data integrity check violation generates an alert.

Randomly inject errors on the control, data, or the ECC bits during CSR accesses. Verify that triggers the correct fatal alert.

Covergroups

Name Description
access_during_key_req_cg

Covers that SRAM handles memory accesses during key requests.

b2b_access_types_cg

Covers that any combination of access types (R/R, R/W, W/R, W/W) can be present in b2b transaction scenarios.

executable_cg

Covers the various important scenarios that can enable SRAM executability.

key_seed_valid_cg

Covers SRAM receiving a key from OTP in Off/On states, with both valid and invalid key seeds.

lc_escalation_rst_cg

Covers de-assertion of LC escalation input before/after/during a reset trigger to get SRAM out of terminal state.

raw_hazard_cg

Covers that the scrambling RAM sees address collisions on RAW hazards (indicating that b2b RAW access has been made to the same memory line).

subword_access_cg

Covers that all possible types of subword accesses (both reads and writes) have been performed.

tl_errors_cg

Cover the following error cases on TL-UL bus:

  • TL-UL protocol error cases.
  • OpenTitan defined error cases, refer to testpoint tl_d_illegal_access.
tl_intg_err_cg

Cover all kinds of integrity errors (command, data or both) and cover number of error bits on each integrity check.

tl_intg_err_mem_subword_cg

Cover the kinds of integrity errors with byte enabled write on memory.

Some memories store the integrity values. When there is a subword write, design re-calculate the integrity with full word data and update integrity in the memory. This coverage ensures that memory byte write has been issued and the related design logic has been verfied.