neorv32/sw/example/processor_check/main.c
Stephan 2b3712acbd
Merge branch 'main' into sysinfo_ptr
Signed-off-by: Stephan <stnolting@gmail.com>
2023-02-27 19:14:19 +01:00

1741 lines
56 KiB
C

// #################################################################################################
// # << NEORV32 - Processor Test Program >> #
// # ********************************************************************************************* #
// # BSD 3-Clause License #
// # #
// # Copyright (c) 2023, Stephan Nolting. All rights reserved. #
// # #
// # Redistribution and use in source and binary forms, with or without modification, are #
// # permitted provided that the following conditions are met: #
// # #
// # 1. Redistributions of source code must retain the above copyright notice, this list of #
// # conditions and the following disclaimer. #
// # #
// # 2. Redistributions in binary form must reproduce the above copyright notice, this list of #
// # conditions and the following disclaimer in the documentation and/or other materials #
// # provided with the distribution. #
// # #
// # 3. Neither the name of the copyright holder nor the names of its contributors may be used to #
// # endorse or promote products derived from this software without specific prior written #
// # permission. #
// # #
// # THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS #
// # OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF #
// # MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE #
// # COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, #
// # EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE #
// # GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED #
// # AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING #
// # NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED #
// # OF THE POSSIBILITY OF SUCH DAMAGE. #
// # ********************************************************************************************* #
// # The NEORV32 Processor - https://github.com/stnolting/neorv32 (c) Stephan Nolting #
// #################################################################################################
/**********************************************************************//**
* @file processor_check/main.c
* @author Stephan Nolting
* @brief CPU/Processor test program.
**************************************************************************/
#include <neorv32.h>
#include <string.h>
/**********************************************************************//**
* @name User configuration
**************************************************************************/
/**@{*/
/** UART BAUD rate */
#define BAUD_RATE (19200)
//** Reachable unaligned address */
#define ADDR_UNALIGNED_1 (0x00000001)
//** Reachable unaligned address */
#define ADDR_UNALIGNED_2 (0x00000002)
//** Unreachable word-aligned address */
#define ADDR_UNREACHABLE (IO_BASE_ADDRESS-4)
//**Read-only word-aligned address */
#define ADDR_READONLY ((uint32_t)&NEORV32_SYSINFO->CLK)
//** external memory base address */
#define EXT_MEM_BASE (0xF0000000)
/**@}*/
/**********************************************************************//**
* @name UART print macros
**************************************************************************/
/**@{*/
//** for simulation only! */
#ifdef SUPPRESS_OPTIONAL_UART_PRINT
//** print standard output to UART0 */
#define PRINT_STANDARD(...)
//** print critical output to UART1 */
#define PRINT_CRITICAL(...) neorv32_uart1_printf(__VA_ARGS__)
#else
//** print standard output to UART0 */
#define PRINT_STANDARD(...) neorv32_uart0_printf(__VA_ARGS__)
//** print critical output to UART0 */
#define PRINT_CRITICAL(...) neorv32_uart0_printf(__VA_ARGS__)
#endif
/**@}*/
// Prototypes
void sim_irq_trigger(uint32_t sel);
void global_trap_handler(void);
void xirq_trap_handler0(void);
void xirq_trap_handler1(void);
void test_ok(void);
void test_fail(void);
// MCAUSE value that will be NEVER set by the hardware
const uint32_t mcause_never_c = 0x80000000U; // = reserved
// Global variables (also test initialization of global vars here)
/// Global counter for failing tests
int cnt_fail = 0;
/// Global counter for successful tests
int cnt_ok = 0;
/// Global counter for total number of tests
int cnt_test = 0;
/// Global number of available HPMs
uint32_t num_hpm_cnts_global = 0;
/// XIRQ trap handler acknowledge
uint32_t xirq_trap_handler_ack = 0;
/// Variable to test store accesses
volatile uint32_t store_access_addr[2];
/// Variable to test PMP
volatile uint32_t pmp_access_addr;
/// Number of implemented PMP regions
uint32_t pmp_num_regions;
/**********************************************************************//**
* High-level CPU/processor test program.
*
* @note Applications has to be compiler with <USER_FLAGS+=-DRUN_CPUTEST>
* @warning This test is intended for simulation only.
* @warning This test requires all optional extensions/modules to be enabled.
*
* @return 0 if execution was successful
**************************************************************************/
int main() {
uint32_t tmp_a, tmp_b;
uint8_t id;
// disable machine-mode interrupts
neorv32_cpu_csr_clr(CSR_MSTATUS, 1 << CSR_MSTATUS_MIE);
// setup UARTs at default baud rate, no parity bits, no HW flow control
neorv32_uart0_setup(BAUD_RATE, PARITY_NONE, FLOW_CONTROL_NONE);
NEORV32_UART1.CTRL = NEORV32_UART0.CTRL; // copy configuration to initialize UART1
#ifdef SUPPRESS_OPTIONAL_UART_PRINT
neorv32_uart0_disable(); // do not generate any UART0 output
#endif
// Disable processor_check compilation by default
#ifndef RUN_CHECK
#warning processor_check HAS NOT BEEN COMPILED! Use >>make USER_FLAGS+=-DRUN_CHECK clean_all exe<< to compile it.
// inform the user if you are actually executing this
PRINT_CRITICAL("ERROR! processor_check has not been compiled. Use >>make USER_FLAGS+=-DRUN_CHECK clean_all exe<< to compile it.\n");
return 1;
#endif
// setup RTE
// -----------------------------------------------
neorv32_rte_setup(); // this will install a full-detailed debug handler for ALL traps
int install_err = 0;
// initialize ALL provided trap handler (overriding the default debug handlers)
for (id=0; id<NEORV32_RTE_NUM_TRAPS; id++) {
install_err += neorv32_rte_handler_install(id, global_trap_handler);
}
if (install_err) {
PRINT_CRITICAL("RTE fail!\n");
return 1;
}
// check available hardware extensions and compare with compiler flags
neorv32_rte_check_isa(0); // silent = 0 -> show message if isa mismatch
// intro
PRINT_STANDARD("\n<< PROCESSOR CHECK >>\n");
// prepare (performance) counters
neorv32_cpu_csr_write(CSR_MCOUNTINHIBIT, 0); // enable counter auto increment (ALL counters)
neorv32_cpu_csr_write(CSR_MCOUNTEREN, 7); // allow access from user-mode code to standard counters only
// set CMP of machine system timer MTIME to max to prevent an IRQ
neorv32_mtime_set_timecmp(-1);
neorv32_mtime_set_time(0);
// get number of implemented PMP regions
pmp_num_regions = neorv32_cpu_pmp_get_num_regions();
// fancy intro
// -----------------------------------------------
// show NEORV32 ASCII logo
neorv32_rte_print_logo();
// show project credits
neorv32_rte_print_credits();
// show full hardware configuration report
neorv32_rte_print_hw_config();
// **********************************************************************************************
// Run CPU and SoC tests
// **********************************************************************************************
// tests intro
PRINT_STANDARD("\nStarting tests...\n\n");
// clear testbench IRQ triggers
sim_irq_trigger(0);
// clear all interrupts, enable only where needed
neorv32_cpu_csr_write(CSR_MIE, 0);
neorv32_cpu_csr_write(CSR_MIP, 0);
// enable machine-mode interrupts
neorv32_cpu_csr_set(CSR_MSTATUS, 1 << CSR_MSTATUS_MIE);
// ----------------------------------------------------------
// Setup PMP for tests
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] Initial PMP setup ", cnt_test);
// check if PMP is already locked
tmp_a = neorv32_cpu_csr_read(CSR_PMPCFG0);
if ((tmp_a & ((1 << PMPCFG_L) << 0*8)) ||
(tmp_a & ((1 << PMPCFG_L) << 1*8)) ||
(tmp_a & ((1 << PMPCFG_L) << 2*8)) ||
(tmp_a & ((1 << PMPCFG_L) << 3*8))) {
PRINT_CRITICAL("\nERROR! PMP locked!\n");
return 1;
}
if (pmp_num_regions >= 4) { // sufficient regions for tests
cnt_test++;
// full access for M & U mode
// use entries 2 & 3 so we can use entries 0 & 1 later on for higher-prioritized configurations
tmp_a = neorv32_cpu_pmp_configure_region(0, 0x00000000, (PMP_OFF << PMPCFG_A_LSB));
tmp_a = neorv32_cpu_pmp_configure_region(1, 0x00000000, (PMP_OFF << PMPCFG_A_LSB));
tmp_a = neorv32_cpu_pmp_configure_region(2, 0x00000000, (PMP_OFF << PMPCFG_A_LSB));
tmp_a += neorv32_cpu_pmp_configure_region(3, 0xFFFFFFFF, (PMP_TOR << PMPCFG_A_LSB) | (1 << PMPCFG_L) | (1 << PMPCFG_R) | (1 << PMPCFG_W) | (1 << PMPCFG_X));
if ((neorv32_cpu_csr_read(CSR_MCAUSE) == mcause_never_c) && (tmp_a == 0)) {
test_ok();
}
else {
test_fail();
}
}
else if ((pmp_num_regions > 0) && (pmp_num_regions < 4)) {
PRINT_CRITICAL("\nERROR! Insufficient PMP regions!\n");
return 1;
}
// ----------------------------------------------------------
// Test fence instructions
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] FENCE(.I) ", cnt_test);
cnt_test++;
asm volatile ("fence");
if (neorv32_cpu_csr_read(CSR_MXISA) & (1<<CSR_MXISA_ZIFENCEI)) {
asm volatile ("fence.i");
}
if (neorv32_cpu_csr_read(CSR_MCAUSE) == mcause_never_c) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Test performance counter: setup as many events and counters as possible
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] Setup HPM events ", cnt_test);
num_hpm_cnts_global = neorv32_cpu_hpm_get_num_counters();
if (num_hpm_cnts_global != 0) {
cnt_test++;
neorv32_cpu_csr_write(CSR_MHPMCOUNTER3, 0); neorv32_cpu_csr_write(CSR_MHPMEVENT3, 1 << HPMCNT_EVENT_CIR);
neorv32_cpu_csr_write(CSR_MHPMCOUNTER4, 0); neorv32_cpu_csr_write(CSR_MHPMEVENT4, 1 << HPMCNT_EVENT_WAIT_IF);
neorv32_cpu_csr_write(CSR_MHPMCOUNTER5, 0); neorv32_cpu_csr_write(CSR_MHPMEVENT5, 1 << HPMCNT_EVENT_WAIT_II);
neorv32_cpu_csr_write(CSR_MHPMCOUNTER6, 0); neorv32_cpu_csr_write(CSR_MHPMEVENT6, 1 << HPMCNT_EVENT_WAIT_MC);
neorv32_cpu_csr_write(CSR_MHPMCOUNTER7, 0); neorv32_cpu_csr_write(CSR_MHPMEVENT7, 1 << HPMCNT_EVENT_LOAD);
neorv32_cpu_csr_write(CSR_MHPMCOUNTER8, 0); neorv32_cpu_csr_write(CSR_MHPMEVENT8, 1 << HPMCNT_EVENT_STORE);
neorv32_cpu_csr_write(CSR_MHPMCOUNTER9, 0); neorv32_cpu_csr_write(CSR_MHPMEVENT9, 1 << HPMCNT_EVENT_WAIT_LS);
neorv32_cpu_csr_write(CSR_MHPMCOUNTER10, 0); neorv32_cpu_csr_write(CSR_MHPMEVENT10, 1 << HPMCNT_EVENT_JUMP);
neorv32_cpu_csr_write(CSR_MHPMCOUNTER11, 0); neorv32_cpu_csr_write(CSR_MHPMEVENT11, 1 << HPMCNT_EVENT_BRANCH);
neorv32_cpu_csr_write(CSR_MHPMCOUNTER12, 0); neorv32_cpu_csr_write(CSR_MHPMEVENT12, 1 << HPMCNT_EVENT_TBRANCH);
neorv32_cpu_csr_write(CSR_MHPMCOUNTER13, 0); neorv32_cpu_csr_write(CSR_MHPMEVENT13, 1 << HPMCNT_EVENT_TRAP);
neorv32_cpu_csr_write(CSR_MHPMCOUNTER14, 0); neorv32_cpu_csr_write(CSR_MHPMEVENT14, 1 << HPMCNT_EVENT_ILLEGAL);
neorv32_cpu_csr_write(CSR_MCOUNTINHIBIT, 0); // enable all counters
// make sure there was no exception
if (neorv32_cpu_csr_read(CSR_MCAUSE) == mcause_never_c) {
test_ok();
}
else {
test_fail();
}
}
else {
PRINT_STANDARD("[skipped, n.a.]\n");
}
// ----------------------------------------------------------
// Test standard RISC-V performance counter [m]cycle[h]
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] cycle counter ", cnt_test);
cnt_test++;
// make sure counter is enabled
asm volatile ("csrci %[addr], %[imm]" : : [addr] "i" (CSR_MCOUNTINHIBIT), [imm] "i" (1<<CSR_MCOUNTINHIBIT_CY));
// prepare overflow
neorv32_cpu_set_mcycle(0x00000000FFFFFFFFULL);
asm volatile ("nop"); // counter LOW should overflow here
// get current cycle counter HIGH
tmp_a = neorv32_cpu_csr_read(CSR_MCYCLEH);
// make sure cycle counter high has incremented and there was no exception during access
if ((tmp_a == 1) && (neorv32_cpu_csr_read(CSR_MCAUSE) == mcause_never_c)) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Test standard RISC-V performance counter [m]instret[h]
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] instret counter ", cnt_test);
cnt_test++;
// make sure counter is enabled
asm volatile ("csrci %[addr], %[imm]" : : [addr] "i" (CSR_MCOUNTINHIBIT), [imm] "i" (1<<CSR_MCOUNTINHIBIT_IR));
// prepare overflow
neorv32_cpu_set_minstret(0x00000000FFFFFFFFULL);
asm volatile ("nop"); // counter LOW should overflow here
// get instruction counter HIGH
tmp_a = neorv32_cpu_csr_read(CSR_INSTRETH);
// make sure instruction counter high has incremented and there was no exception during access
if ((tmp_a == 1) && (neorv32_cpu_csr_read(CSR_MCAUSE) == mcause_never_c)) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Test mcountinhibt: inhibit auto-inc of [m]cycle
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] mcountinhibt.cy CSR ", cnt_test);
cnt_test++;
// inhibit [m]cycle CSR
tmp_a = neorv32_cpu_csr_read(CSR_MCOUNTINHIBIT);
tmp_a |= (1<<CSR_MCOUNTINHIBIT_CY); // inhibit cycle counter auto-increment
neorv32_cpu_csr_write(CSR_MCOUNTINHIBIT, tmp_a);
// get current cycle counter
tmp_a = neorv32_cpu_csr_read(CSR_CYCLE);
// wait some time to have a nice "increment" (there should be NO increment at all!)
asm volatile ("nop");
tmp_b = neorv32_cpu_csr_read(CSR_CYCLE);
// make sure instruction counter has NOT incremented and there was no exception during access
if ((tmp_a == tmp_b) && (tmp_a != 0) && (neorv32_cpu_csr_read(CSR_MCAUSE) == mcause_never_c)) {
test_ok();
}
else {
test_fail();
}
// re-enable [m]cycle CSR
tmp_a = neorv32_cpu_csr_read(CSR_MCOUNTINHIBIT);
tmp_a &= ~(1<<CSR_MCOUNTINHIBIT_CY); // clear inhibit of cycle counter auto-increment
neorv32_cpu_csr_write(CSR_MCOUNTINHIBIT, tmp_a);
// ----------------------------------------------------------
// Execute MRET in U-mode (has to trap!)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] MRET in U-mode ", cnt_test);
cnt_test++;
// switch to user mode (hart will be back in MACHINE mode when trap handler returns)
neorv32_cpu_goto_user_mode();
{
asm volatile ("mret");
}
if (neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_I_ILLEGAL) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// External memory interface test
// (and iCache block-/word-wise error check)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] Ext. memory access (@0x%x) ", cnt_test, (uint32_t)EXT_MEM_BASE);
if (NEORV32_SYSINFO->SOC & (1 << SYSINFO_SOC_MEM_EXT)) {
cnt_test++;
// clear scratch CSR
neorv32_cpu_csr_write(CSR_MSCRATCH, 0);
// setup test program in external memory
neorv32_cpu_store_unsigned_word((uint32_t)EXT_MEM_BASE+0, 0x3407D073); // csrwi mscratch, 15
neorv32_cpu_store_unsigned_word((uint32_t)EXT_MEM_BASE+4, 0x00008067); // ret (32-bit)
// execute program
asm volatile ("fence.i"); // flush i-cache
tmp_a = (uint32_t)EXT_MEM_BASE; // call the dummy sub program
asm volatile ("jalr ra, %[input_i]" : : [input_i] "r" (tmp_a));
if ((neorv32_cpu_csr_read(CSR_MCAUSE) == mcause_never_c) && // make sure there was no exception
(neorv32_cpu_csr_read(CSR_MSCRATCH) == 15)) { // make sure the program was executed in the right way
test_ok();
}
else {
test_fail();
}
}
else {
PRINT_STANDARD("[skipped, n.a.]\n");
}
// ----------------------------------------------------------
// Illegal CSR access
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] Illegal CSR ", cnt_test);
cnt_test++;
tmp_a = neorv32_cpu_csr_read(CSR_DSCRATCH0); // only accessible in debug mode
if (neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_I_ILLEGAL) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Write-access to read-only CSR
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] Read-only CSR write ", cnt_test);
cnt_test++;
neorv32_cpu_csr_write(CSR_CYCLE, 0); // cycle CSR is read-only
if (neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_I_ILLEGAL) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// No "real" CSR write access (because rs1 = r0)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] Read-only CSR 'no-write' (rs1=x0) access ", cnt_test);
cnt_test++;
// cycle CSR is read-only, but no actual write is performed because rs1=r0
// -> should cause no exception
asm volatile ("csrrs zero, cycle, zero");
if (neorv32_cpu_csr_read(CSR_MCAUSE) == mcause_never_c) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Unaligned instruction address
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] I_ALG (instr. align) EXC ", cnt_test);
// skip if C-mode is implemented
if ((neorv32_cpu_csr_read(CSR_MISA) & (1<<CSR_MISA_C)) == 0) {
cnt_test++;
// call unaligned address
((void (*)(void))ADDR_UNALIGNED_2)();
asm volatile ("nop");
if (neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_I_MISALIGNED) {
test_ok();
}
else {
test_fail();
}
}
else {
PRINT_STANDARD("[skipped, n.a. with C-ext]\n");
}
// ----------------------------------------------------------
// Instruction access fault
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] I_ACC (instr. bus access) EXC ", cnt_test);
cnt_test++;
// put two "ret" instructions to the beginning of the external memory module
neorv32_cpu_store_unsigned_word((uint32_t)EXT_MEM_BASE+0, 0x00008067); // exception handler hack will see this instruction as exception source
neorv32_cpu_store_unsigned_word((uint32_t)EXT_MEM_BASE+4, 0x00008067); // and will try to resume execution here
// jump to beginning of external memory minus 4 bytes
// this will cause an instruction access fault as there is no module responding to the fetch request
// the exception handler will try to resume at the instruction 4 bytes ahead, which is the "ret" we just created
asm volatile ("fence.i"); // flush i-cache
tmp_a = ((uint32_t)EXT_MEM_BASE) - 4;
asm volatile ("jalr ra, %[input_i]" : : [input_i] "r" (tmp_a));
if ((neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_I_ACCESS) && // correct exception cause
(neorv32_cpu_csr_read(CSR_MTVAL) == tmp_a)) { // correct trap value (address of instruction that caused ifetch error)
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Illegal instruction
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] I_ILL (illegal instr.) EXC ", cnt_test);
cnt_test++;
// clear mstatus.mie and set mstatus.mpie
neorv32_cpu_csr_clr(CSR_MSTATUS, 1 << CSR_MSTATUS_MIE);
neorv32_cpu_csr_set(CSR_MSTATUS, 1 << CSR_MSTATUS_MPIE);
// illegal 32-bit instruction (MRET with incorrect opcode)
asm volatile (".align 4 \n"
".word 0x3020007f");
// make sure this has caused an illegal exception
if ((neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_I_ILLEGAL) && // illegal instruction exception
((neorv32_cpu_csr_read(CSR_MSTATUS) & (1 << CSR_MSTATUS_MIE)) == 0)) { // MIE should still be cleared
test_ok();
}
else {
test_fail();
}
// reenable machine-mode interrupts
neorv32_cpu_csr_set(CSR_MSTATUS, 1 << CSR_MSTATUS_MIE);
// ----------------------------------------------------------
// Illegal compressed instruction
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] CI_ILL (illegal compr. instr.) EXC ", cnt_test);
// skip if C-mode is not implemented
if ((neorv32_cpu_csr_read(CSR_MISA) & (1<<CSR_MISA_C))) {
cnt_test++;
// illegal 16-bit instruction (official UNIMP instruction)
asm volatile (".align 2 \n"
".half 0x0001 \n" // NOP
".half 0x0000"); // UNIMP
if (neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_I_ILLEGAL) {
test_ok();
}
else {
test_fail();
}
}
else {
PRINT_STANDARD("[skipped, n.a. with C-ext]\n");
}
// ----------------------------------------------------------
// Breakpoint instruction
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] BREAK EXC ", cnt_test);
// skip on real hardware since ebreak will make problems when running this test program via gdb
if (NEORV32_SYSINFO->SOC & (1<<SYSINFO_SOC_IS_SIM)) {
cnt_test++;
asm volatile ("ebreak");
if (neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_BREAKPOINT) {
test_ok();
}
else {
test_fail();
}
}
else {
PRINT_STANDARD("[skipped]\n");
}
// ----------------------------------------------------------
// Unaligned load address
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] L_ALG (load align) EXC ", cnt_test);
cnt_test++;
// load from unaligned address
asm volatile ("li %[da], 0xcafe1230 \n" // initialize destination register with known value
"lw %[da], 0(%[ad]) " // must not update destination register to to exception
: [da] "=r" (tmp_b) : [ad] "r" (ADDR_UNALIGNED_1));
if ((neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_L_MISALIGNED) &&
(neorv32_cpu_csr_read(CSR_MTVAL) == ADDR_UNALIGNED_1) &&
(tmp_b == 0xcafe1230)) { // make sure dest. reg is not updated
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Load access fault
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] L_ACC (load access) EXC ", cnt_test);
cnt_test++;
tmp_a = (1 << BUSKEEPER_ERR_FLAG) | (1 << BUSKEEPER_ERR_TYPE);
// load from unreachable aligned address
asm volatile ("li %[da], 0xcafe1230 \n" // initialize destination register with known value
"lw %[da], 0(%[ad]) " // must not update destination register to to exception
: [da] "=r" (tmp_b) : [ad] "r" (ADDR_UNREACHABLE));
if ((neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_L_ACCESS) && // load bus access error exception
(neorv32_cpu_csr_read(CSR_MTVAL) == ADDR_UNREACHABLE) &&
(tmp_b == 0xcafe1230) && // make sure dest. reg is not updated
(NEORV32_BUSKEEPER->CTRL = tmp_a)) { // buskeeper: error flag + timeout error
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Unaligned store address
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] S_ALG (store align) EXC ", cnt_test);
cnt_test++;
// initialize test variable
store_access_addr[0] = 0x11223344;
store_access_addr[1] = 0x55667788;
tmp_a = (uint32_t)(&store_access_addr[0]);
tmp_a += 2; // make word-unaligned
// store to unaligned address
neorv32_cpu_store_unsigned_word(tmp_a, 0);
if ((neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_S_MISALIGNED) &&
(neorv32_cpu_csr_read(CSR_MTVAL) == tmp_a) &&
(store_access_addr[0] == 0x11223344) &&
(store_access_addr[1] == 0x55667788)) { // make sure memory was not altered
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Store access fault
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] S_ACC (store access) EXC ", cnt_test);
cnt_test++;
tmp_a = (1 << BUSKEEPER_ERR_FLAG) | (0 << BUSKEEPER_ERR_TYPE);
// store to unreachable aligned address
neorv32_cpu_store_unsigned_word(ADDR_READONLY, 0);
if ((neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_S_ACCESS) && // store bus access error exception
(neorv32_cpu_csr_read(CSR_MTVAL) == ADDR_READONLY) &&
(NEORV32_BUSKEEPER->CTRL == tmp_a)) { // buskeeper: error flag + device error
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Environment call from M-mode
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] ENVCALL M EXC ", cnt_test);
cnt_test++;
asm volatile ("ecall");
if (neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_MENV_CALL) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Environment call from U-mode
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] ENVCALL U EXC ", cnt_test);
cnt_test++;
// switch to user mode (hart will be back in MACHINE mode when trap handler returns)
neorv32_cpu_goto_user_mode();
{
asm volatile ("ecall");
}
if (neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_UENV_CALL) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Machine timer interrupt (MTIME)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] MTI (MTIME) IRQ ", cnt_test);
cnt_test++;
// configure MTIME (and check overflow from low word to high word)
neorv32_mtime_set_timecmp(0x0000000100000000ULL);
neorv32_mtime_set_time( 0x00000000FFFFFFFEULL);
// enable interrupt
neorv32_cpu_csr_write(CSR_MIE, 1 << CSR_MIE_MTIE);
// wait some time for the IRQ to trigger and arrive the CPU
asm volatile ("nop");
asm volatile ("nop");
neorv32_cpu_csr_write(CSR_MIE, 0);
if (neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_MTI) {
test_ok();
}
else {
test_fail();
}
// no more MTIME interrupts
neorv32_mtime_set_timecmp(-1);
// ----------------------------------------------------------
// Machine software interrupt (MSI) via testbench
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] MSI (sim) IRQ ", cnt_test);
cnt_test++;
// enable interrupt
neorv32_cpu_csr_write(CSR_MIE, 1 << CSR_MIE_MSIE);
// trigger IRQ
sim_irq_trigger(1 << CSR_MIE_MSIE);
// wait some time for the IRQ to arrive the CPU
asm volatile ("nop");
asm volatile ("nop");
neorv32_cpu_csr_write(CSR_MIE, 0);
sim_irq_trigger(0);
if (neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_MSI) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Machine external interrupt (MEI) via testbench
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] MEI (sim) IRQ ", cnt_test);
cnt_test++;
// enable interrupt
neorv32_cpu_csr_write(CSR_MIE, 1 << CSR_MIE_MEIE);
// trigger IRQ
sim_irq_trigger(1 << CSR_MIE_MEIE);
// wait some time for the IRQ to arrive the CPU
asm volatile ("nop");
asm volatile ("nop");
neorv32_cpu_csr_write(CSR_MIE, 0);
sim_irq_trigger(0);
if (neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_MEI) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Permanent IRQ (make sure interrupted program advances)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] Permanent IRQ (MTIME) ", cnt_test);
cnt_test++;
// fire MTIME IRQ
neorv32_cpu_csr_write(CSR_MIE, 1 << CSR_MIE_MTIE);
neorv32_mtime_set_timecmp(0); // force interrupt
int test_cnt = 0;
while(test_cnt < 2) {
test_cnt++;
}
neorv32_cpu_csr_write(CSR_MIE, 0);
if (test_cnt == 2) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Test pending interrupt
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] Pending IRQ (MTIME) ", cnt_test);
cnt_test++;
// disable all interrupt setting
neorv32_cpu_csr_write(CSR_MIE, 0);
// fire MTIME IRQ
neorv32_mtime_set_timecmp(0); // force interrupt
// wait some time for the IRQ to arrive the CPU
asm volatile ("nop");
asm volatile ("nop");
uint32_t was_pending = neorv32_cpu_csr_read(CSR_MIP) & (1 << CSR_MIP_MTIP); // should be pending now
// clear pending MTI
neorv32_mtime_set_timecmp(-1);
uint32_t is_pending = neorv32_cpu_csr_read(CSR_MIP) & (1 << CSR_MIP_MTIP); // should NOT be pending anymore
if ((was_pending != 0) && (is_pending == 0)) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Fast interrupt channel 0 (WDT)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] FIRQ0 (WDT) ", cnt_test);
cnt_test++;
// enable fast interrupt
neorv32_cpu_csr_write(CSR_MIE, 1 << WDT_FIRQ_ENABLE);
// configure WDT:
// timeout = 1*4096 cycles, no lock, disable in debug mode, enable in sleep mode
neorv32_wdt_setup(1, 0, 0, 1);
// wait in sleep mode for WDT interrupt
asm volatile ("wfi");
neorv32_cpu_csr_write(CSR_MIE, 0);
NEORV32_WDT->CTRL = 0;
if (neorv32_cpu_csr_read(CSR_MCAUSE) == WDT_TRAP_CODE) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Fast interrupt channel 1 (CFS)
// ----------------------------------------------------------
PRINT_STANDARD("[%i] FIRQ1 (CFS) ", cnt_test);
PRINT_STANDARD("[skipped, n.a.]\n");
// ----------------------------------------------------------
// Fast interrupt channel 2 (UART0.RX)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] FIRQ2 (UART0.RX) ", cnt_test);
cnt_test++;
// wait for UART0 to finish transmitting
while(neorv32_uart0_tx_busy());
// backup current UART0 configuration
tmp_a = NEORV32_UART0.CTRL;
// make sure UART is enabled
NEORV32_UART0.CTRL |= (1 << UART_CTRL_EN);
// make sure sim mode is disabled
NEORV32_UART0.CTRL &= ~(1 << UART_CTRL_SIM_MODE);
// enable fast interrupt
neorv32_cpu_csr_write(CSR_MIE, 1 << UART0_RX_FIRQ_ENABLE);
// trigger UART0 RX IRQ
neorv32_uart0_putc(0);
// wait for UART0 to finish transmitting
while(neorv32_uart0_tx_busy());
// wait some time for the IRQ to arrive the CPU
asm volatile ("nop");
asm volatile ("nop");
neorv32_cpu_csr_write(CSR_MIE, 0);
// restore original configuration
NEORV32_UART0.CTRL = tmp_a;
if (neorv32_cpu_csr_read(CSR_MCAUSE) == UART0_RX_TRAP_CODE) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Fast interrupt channel 3 (UART0.TX)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] FIRQ3 (UART0.TX) ", cnt_test);
cnt_test++;
// wait for UART0 to finish transmitting
while(neorv32_uart0_tx_busy());
// backup current UART0 configuration
tmp_a = NEORV32_UART0.CTRL;
// make sure UART is enabled
NEORV32_UART0.CTRL |= (1 << UART_CTRL_EN);
// make sure sim mode is disabled
NEORV32_UART0.CTRL &= ~(1 << UART_CTRL_SIM_MODE);
// UART0 TX interrupt enable
neorv32_cpu_csr_write(CSR_MIE, 1 << UART0_TX_FIRQ_ENABLE);
// trigger UART0 TX IRQ
neorv32_uart0_putc(0);
// wait for UART to finish transmitting
while(neorv32_uart0_tx_busy());
// wait some time for the IRQ to arrive the CPU
asm volatile ("nop");
asm volatile ("nop");
neorv32_cpu_csr_write(CSR_MIE, 0);
// restore original configuration
NEORV32_UART0.CTRL = tmp_a;
if (neorv32_cpu_csr_read(CSR_MCAUSE) == UART0_TX_TRAP_CODE) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Fast interrupt channel 4 (UART1.RX)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] FIRQ4 (UART1.RX) ", cnt_test);
cnt_test++;
// backup current UART1 configuration
tmp_a = NEORV32_UART1.CTRL;
// make sure UART is enabled
NEORV32_UART1.CTRL |= (1 << UART_CTRL_EN);
// make sure sim mode is disabled
NEORV32_UART1.CTRL &= ~(1 << UART_CTRL_SIM_MODE);
// UART1 RX interrupt enable
neorv32_cpu_csr_write(CSR_MIE, 1 << UART1_RX_FIRQ_ENABLE);
// trigger UART1 RX IRQ
neorv32_uart1_putc(0);
// wait for UART1 to finish transmitting
while(neorv32_uart1_tx_busy());
// wait some time for the IRQ to arrive the CPU
asm volatile ("nop");
asm volatile ("nop");
neorv32_cpu_csr_write(CSR_MIE, 0);
// restore original configuration
NEORV32_UART1.CTRL = tmp_a;
if (neorv32_cpu_csr_read(CSR_MCAUSE) == UART1_RX_TRAP_CODE) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Fast interrupt channel 5 (UART1.TX)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] FIRQ5 (UART1.TX) ", cnt_test);
cnt_test++;
// backup current UART1 configuration
tmp_a = NEORV32_UART1.CTRL;
// make sure UART is enabled
NEORV32_UART1.CTRL |= (1 << UART_CTRL_EN);
// make sure sim mode is disabled
NEORV32_UART1.CTRL &= ~(1 << UART_CTRL_SIM_MODE);
// UART1 RX interrupt enable
neorv32_cpu_csr_write(CSR_MIE, 1 << UART1_TX_FIRQ_ENABLE);
// trigger UART1 TX IRQ
neorv32_uart1_putc(0);
// wait for UART1 to finish transmitting
while(neorv32_uart1_tx_busy());
// wait some time for the IRQ to arrive the CPU
asm volatile ("nop");
asm volatile ("nop");
neorv32_cpu_csr_write(CSR_MIE, 0);
// restore original configuration
NEORV32_UART1.CTRL = tmp_a;
if (neorv32_cpu_csr_read(CSR_MCAUSE) == UART1_TX_TRAP_CODE) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Fast interrupt channel 6 (SPI)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] FIRQ6 (SPI) ", cnt_test);
cnt_test++;
// configure SPI
neorv32_spi_setup(CLK_PRSC_2, 0, 0, 0, 0, 0); // IRQ when SPI PHY becomes idle
// enable fast interrupt
neorv32_cpu_csr_write(CSR_MIE, 1 << SPI_FIRQ_ENABLE);
// trigger SPI IRQ
neorv32_spi_trans(0); // blocking
// wait some time for the IRQ to arrive the CPU
asm volatile ("nop");
asm volatile ("nop");
neorv32_cpu_csr_write(CSR_MIE, 0);
if (neorv32_cpu_csr_read(CSR_MCAUSE) == SPI_TRAP_CODE) {
test_ok();
}
else {
test_fail();
}
// disable SPI
neorv32_spi_disable();
// ----------------------------------------------------------
// Fast interrupt channel 7 (TWI)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] FIRQ7 (TWI) ", cnt_test);
cnt_test++;
// configure TWI, fastest clock, no clock stretching
neorv32_twi_setup(CLK_PRSC_2, 0, 0);
// enable TWI FIRQ
neorv32_cpu_csr_write(CSR_MIE, 1 << TWI_FIRQ_ENABLE);
// trigger TWI IRQ
neorv32_twi_start_trans(0xA5);
// wait some time for the IRQ to arrive the CPU
asm volatile ("nop");
asm volatile ("nop");
neorv32_cpu_csr_write(CSR_MIE, 0);
if (neorv32_cpu_csr_read(CSR_MCAUSE) == TWI_TRAP_CODE) {
test_ok();
}
else {
test_fail();
}
// disable TWI
neorv32_twi_disable();
// ----------------------------------------------------------
// Fast interrupt channel 8 (XIRQ)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] FIRQ8 (XIRQ) ", cnt_test);
cnt_test++;
int xirq_err_cnt = 0;
xirq_trap_handler_ack = 0;
xirq_err_cnt += neorv32_xirq_setup(); // initialize XIRQ
xirq_err_cnt += neorv32_xirq_install(0, xirq_trap_handler0); // install XIRQ IRQ handler channel 0
xirq_err_cnt += neorv32_xirq_install(1, xirq_trap_handler1); // install XIRQ IRQ handler channel 1
// enable XIRQ FIRQ
neorv32_cpu_csr_write(CSR_MIE, 1 << XIRQ_FIRQ_ENABLE);
// trigger XIRQ channel 1 and 0
neorv32_gpio_port_set(3);
// wait for IRQs to arrive CPU
asm volatile ("nop");
asm volatile ("nop");
neorv32_cpu_csr_write(CSR_MIE, 0);
if ((neorv32_cpu_csr_read(CSR_MCAUSE) == XIRQ_TRAP_CODE) && // FIRQ8 IRQ
(xirq_err_cnt == 0) && // no errors during XIRQ configuration
(xirq_trap_handler_ack == 4)) { // XIRQ channel handler 0 executed before handler 1
test_ok();
}
else {
test_fail();
}
NEORV32_XIRQ->IER = 0;
NEORV32_XIRQ->IPR = -1;
// ----------------------------------------------------------
// Fast interrupt channel 9 (NEOLED)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] FIRQ9 (NEOLED) ", cnt_test);
cnt_test++;
// enable fast interrupt
neorv32_cpu_csr_write(CSR_MIE, 1 << NEOLED_FIRQ_ENABLE);
// configure NEOLED
neorv32_neoled_setup(CLK_PRSC_2, 0, 0, 0);
// send dummy data
neorv32_neoled_write_nonblocking(0);
// wait some time for the IRQ to arrive the CPU
asm volatile ("nop");
asm volatile ("nop");
neorv32_cpu_csr_write(CSR_MIE, 0);
if (neorv32_cpu_csr_read(CSR_MCAUSE) == NEOLED_TRAP_CODE) {
test_ok();
}
else {
test_fail();
}
// no more NEOLED interrupts
neorv32_neoled_disable();
// ----------------------------------------------------------
// Fast interrupt channel 10 (reserved)
// ----------------------------------------------------------
PRINT_STANDARD("[%i] FIRQ10 [skipped, n.a.]\n", cnt_test);
// ----------------------------------------------------------
// Fast interrupt channel 11 (SDI)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] FIRQ11 (SDI) ", cnt_test);
cnt_test++;
// configure and enable SDI + SPI
neorv32_sdi_setup(1 << SDI_CTRL_IRQ_RX_AVAIL);
neorv32_spi_setup(CLK_PRSC_4, 0, 0, 0, 0, 0); // IRQ when SPI PHY becomes idle
// enable fast interrupt
neorv32_cpu_csr_write(CSR_MIE, 1 << SDI_FIRQ_ENABLE);
// write test data to SDI
neorv32_sdi_rx_clear();
neorv32_sdi_put(0xab);
// trigger SDI IRQ by sending data via SPI
neorv32_spi_cs_en(7); // select SDI
tmp_a = neorv32_spi_trans(0x83);
neorv32_spi_cs_dis();
// wait some time for the IRQ to arrive the CPU
asm volatile ("nop");
asm volatile ("nop");
neorv32_cpu_csr_write(CSR_MIE, 0);
if ((neorv32_cpu_csr_read(CSR_MCAUSE) == SDI_TRAP_CODE) && // correct trap code
(neorv32_sdi_get_nonblocking() == 0x83) && // correct SDI read data
((tmp_a & 0xff) == 0xab)) { // correct SPI read data
test_ok();
}
else {
test_fail();
}
// disable SDI + SPI
neorv32_sdi_disable();
neorv32_spi_disable();
// ----------------------------------------------------------
// Fast interrupt channel 12 (GPTMR)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] FIRQ12 (GPTMR) ", cnt_test);
cnt_test++;
// enable GPTMR FIRQ
neorv32_cpu_csr_write(CSR_MIE, 1 << GPTMR_FIRQ_ENABLE);
// configure timer IRQ for one-shot mode after CLK_PRSC_2*2=4 clock cycles
neorv32_gptmr_setup(CLK_PRSC_2, 0, 2);
// wait some time for the IRQ to arrive the CPU
asm volatile ("nop");
asm volatile ("nop");
neorv32_cpu_csr_write(CSR_MIE, 0);
// check if IRQ
if (neorv32_cpu_csr_read(CSR_MCAUSE) == GPTMR_TRAP_CODE) {
test_ok();
}
else {
test_fail();
}
// disable GPTMR
neorv32_gptmr_disable();
// ----------------------------------------------------------
// Fast interrupt channel 13 (ONEWIRE)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] FIRQ13 (ONEWIRE) ", cnt_test);
cnt_test++;
// enable ONEWIRE FIRQ
neorv32_cpu_csr_write(CSR_MIE, 1 << ONEWIRE_FIRQ_ENABLE);
// configure interface for minimal timing
neorv32_onewire_setup(200); // t_base = 200ns
// read single bit from bus
neorv32_onewire_read_bit_blocking();
// wait some time for the IRQ to arrive the CPU
asm volatile ("nop");
asm volatile ("nop");
neorv32_cpu_csr_write(CSR_MIE, 0);
// check if IRQ
if (neorv32_cpu_csr_read(CSR_MCAUSE) == ONEWIRE_TRAP_CODE) {
test_ok();
}
else {
test_fail();
}
// disable ONEWIRE
neorv32_onewire_disable();
// ----------------------------------------------------------
// Fast interrupt channel 14..15 (reserved)
// ----------------------------------------------------------
PRINT_STANDARD("[%i] FIRQ14..15 [skipped, n.a.]\n", cnt_test);
// ----------------------------------------------------------
// Test WFI ("sleep") instruction (executed in user mode), wakeup via MTIME
// mstatus.mie is cleared before to check if machine-mode IRQ still trigger in user-mode
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] user-mode WFI (wake-up via MTIME) ", cnt_test);
cnt_test++;
// program wake-up timer
neorv32_mtime_set_timecmp(neorv32_mtime_get_time() + 500);
// enable mtime interrupt
neorv32_cpu_csr_write(CSR_MIE, 1 << CSR_MIE_MTIE);
// clear mstatus.TW to allow execution of WFI also in user-mode
// clear mstatus.MIE and mstatus.MPIE to check if IRQ can still trigger in User-mode
neorv32_cpu_csr_clr(CSR_MSTATUS, (1<<CSR_MSTATUS_TW) | (1<<CSR_MSTATUS_MIE) | (1<<CSR_MSTATUS_MPIE));
// put CPU into sleep mode (from user mode)
neorv32_cpu_goto_user_mode();
{
asm volatile ("wfi");
}
neorv32_cpu_csr_write(CSR_MIE, 0);
if (neorv32_cpu_csr_read(CSR_MCAUSE) != TRAP_CODE_MTI) {
test_fail();
}
else {
test_ok();
}
// ----------------------------------------------------------
// Test unallowed WFI ("sleep") instruction (executed in user mode)
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] WFI (not allowed in u-mode) ", cnt_test);
cnt_test++;
// set mstatus.TW to disallow execution of WFI in user-mode
neorv32_cpu_csr_set(CSR_MSTATUS, 1 << CSR_MSTATUS_TW);
// put CPU into sleep mode (from user mode)
neorv32_cpu_goto_user_mode();
{
asm volatile ("wfi"); // this has to fail
}
if (neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_I_ILLEGAL) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Test invalid CSR access in user mode
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] Invalid CSR access from U-mode ", cnt_test);
cnt_test++;
// switch to user mode (hart will be back in MACHINE mode when trap handler returns)
neorv32_cpu_goto_user_mode();
{
// access to misa not allowed for user-level programs
tmp_a = neorv32_cpu_csr_read(CSR_MISA);
}
if (neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_I_ILLEGAL) {
test_ok();
}
else {
test_fail();
}
// ----------------------------------------------------------
// Test RTE debug trap handler
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] RTE debug trap handler ", cnt_test);
cnt_test++;
// uninstall custom handler and use default RTE debug handler
neorv32_rte_handler_uninstall(RTE_TRAP_I_ILLEGAL);
// trigger illegal instruction exception
neorv32_cpu_csr_read(0xfff); // CSR not available
PRINT_STANDARD(" ");
if (neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_I_ILLEGAL) {
test_ok();
}
else {
test_fail();
}
// restore original handler
neorv32_rte_handler_install(RTE_TRAP_I_ILLEGAL, global_trap_handler);
// ----------------------------------------------------------
// Test physical memory protection
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] PMP:\n", cnt_test);
// check if PMP is implemented
if (pmp_num_regions >= 4) {
// Create PMP protected region
// ---------------------------------------------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
cnt_test++;
pmp_access_addr = 0xcafe1234; // initialize
tmp_a = (uint32_t)(&pmp_access_addr); // base address of protected region
// configure new region (with highest priority)
int pmp_res = 0;
// base
PRINT_STANDARD(" Setup PMP(0) OFF [-,-,-,-] @ 0x%x\n", tmp_a);
pmp_res += neorv32_cpu_pmp_configure_region(0, tmp_a, 0);
// bound
PRINT_STANDARD(" Setup PMP(1) TOR [!L,!X,!W,R] @ 0x%x ", tmp_a+4);
pmp_res += neorv32_cpu_pmp_configure_region(1, tmp_a+4, (PMP_TOR << PMPCFG_A_LSB) | (1 << PMPCFG_R)); // read-only
if ((pmp_res == 0) && (neorv32_cpu_csr_read(CSR_MCAUSE) == mcause_never_c)) {
test_ok();
}
else {
test_fail();
}
// ------ LOAD from U-mode: should succeed ------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] PMP: U-mode read (SUCCEED) ", cnt_test);
cnt_test++;
// switch to user mode (hart will be back in MACHINE mode when trap handler returns)
neorv32_cpu_goto_user_mode();
{
tmp_b = 0;
tmp_b = neorv32_cpu_load_unsigned_word((uint32_t)(&pmp_access_addr));
}
asm volatile ("ecall"); // switch back to machine mode
if (tmp_b == 0xcafe1234) {
test_ok();
}
else {
test_fail();
}
// ------ STORE from U-mode: should fail ------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] PMP: U-mode write (FAIL) ", cnt_test);
cnt_test++;
// switch to user mode (hart will be back in MACHINE mode when trap handler returns)
neorv32_cpu_goto_user_mode();
{
neorv32_cpu_store_unsigned_word((uint32_t)(&pmp_access_addr), 0); // store access -> should fail
}
asm volatile ("ecall"); // switch back to machine mode
if (pmp_access_addr == 0xcafe1234) {
test_ok();
}
else {
test_fail();
}
// ------ STORE from M mode using U mode permissions: should fail ------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] PMP: M-mode (U-mode permissions) write (FAIL) ", cnt_test);
cnt_test++;
// make M-mode load/store accesses use U-mode rights
neorv32_cpu_csr_set(CSR_MSTATUS, 1 << CSR_MSTATUS_MPRV); // set MPRV: M uses U permissions for load/stores
neorv32_cpu_csr_clr(CSR_MSTATUS, 3 << CSR_MSTATUS_MPP_L); // clear MPP: use U as effective privilege mode
neorv32_cpu_store_unsigned_word((uint32_t)(&pmp_access_addr), 0); // store access -> should fail
neorv32_cpu_csr_clr(CSR_MSTATUS, 1 << CSR_MSTATUS_MPRV);
if ((neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_S_ACCESS) && (pmp_access_addr == 0xcafe1234)) {
test_ok();
}
else {
test_fail();
}
// ------ STORE from M mode with LOCKED: should fail ------
neorv32_cpu_csr_write(CSR_MCAUSE, mcause_never_c);
PRINT_STANDARD("[%i] PMP: M-mode (LOCKED) write (FAIL) ", cnt_test);
cnt_test++;
// set lock bit
tmp_a = neorv32_cpu_csr_read(CSR_PMPCFG0);
tmp_a |= (1 << PMPCFG_L) << 8; // set lock bit in entry 1
neorv32_cpu_csr_write(CSR_PMPCFG0, tmp_a);
neorv32_cpu_store_unsigned_word((uint32_t)(&pmp_access_addr), 0); // store access -> should fail
if ((neorv32_cpu_csr_read(CSR_MCAUSE) == TRAP_CODE_S_ACCESS) && (pmp_access_addr == 0xcafe1234)) {
test_ok();
}
else {
test_fail();
}
}
else {
PRINT_STANDARD("[skipped, n.a.]\n");
}
// ----------------------------------------------------------
// HPM reports
// ----------------------------------------------------------
neorv32_cpu_csr_write(CSR_MCOUNTINHIBIT, -1); // stop all HPM counters
PRINT_STANDARD("\n\n--<< HPM.low (%u) >>--\n", num_hpm_cnts_global);
PRINT_STANDARD("#00 Instr.: %u\n", (uint32_t)neorv32_cpu_csr_read(CSR_INSTRET));
// HPM #01 does not exist
PRINT_STANDARD("#02 Clocks: %u\n", (uint32_t)neorv32_cpu_csr_read(CSR_CYCLE));
PRINT_STANDARD("#03 C-instr.: %u\n", (uint32_t)neorv32_cpu_csr_read(CSR_MHPMCOUNTER3));
PRINT_STANDARD("#04 IF wait: %u\n", (uint32_t)neorv32_cpu_csr_read(CSR_MHPMCOUNTER4));
PRINT_STANDARD("#05 II wait: %u\n", (uint32_t)neorv32_cpu_csr_read(CSR_MHPMCOUNTER5));
PRINT_STANDARD("#06 ALU wait: %u\n", (uint32_t)neorv32_cpu_csr_read(CSR_MHPMCOUNTER6));
PRINT_STANDARD("#07 M loads: %u\n", (uint32_t)neorv32_cpu_csr_read(CSR_MHPMCOUNTER7));
PRINT_STANDARD("#08 M stores: %u\n", (uint32_t)neorv32_cpu_csr_read(CSR_MHPMCOUNTER8));
PRINT_STANDARD("#09 M wait: %u\n", (uint32_t)neorv32_cpu_csr_read(CSR_MHPMCOUNTER9));
PRINT_STANDARD("#10 Jumps: %u\n", (uint32_t)neorv32_cpu_csr_read(CSR_MHPMCOUNTER10));
PRINT_STANDARD("#11 Branch.: %u\n", (uint32_t)neorv32_cpu_csr_read(CSR_MHPMCOUNTER11));
PRINT_STANDARD("#12 > taken: %u\n", (uint32_t)neorv32_cpu_csr_read(CSR_MHPMCOUNTER12));
PRINT_STANDARD("#13 EXCs: %u\n", (uint32_t)neorv32_cpu_csr_read(CSR_MHPMCOUNTER13));
PRINT_STANDARD("#14 Illegals: %u\n", (uint32_t)neorv32_cpu_csr_read(CSR_MHPMCOUNTER14));
// ----------------------------------------------------------
// Final test reports
// ----------------------------------------------------------
PRINT_CRITICAL("\n\nTest results:\nPASS: %i/%i\nFAIL: %i/%i\n\n", cnt_ok, cnt_test, cnt_fail, cnt_test);
// final result
if (cnt_fail == 0) {
PRINT_STANDARD("%c[1m[PROCESSOR TEST COMPLETED SUCCESSFULLY!]%c[0m\n", 27, 27);
}
else {
PRINT_STANDARD("%c[1m[PROCESSOR TEST FAILED!]%c[0m\n", 27, 27);
}
return (int)cnt_fail; // return error counter for after-main handler
}
/**********************************************************************//**
* Simulation-based function to set/clear CPU interrupts (MSI, MEI).
*
* @param[in] sel IRQ select mask (bit positions according to #NEORV32_CSR_MIE_enum).
**************************************************************************/
void sim_irq_trigger(uint32_t sel) {
*((volatile uint32_t*) (0xFF000000)) = sel;
}
/**********************************************************************//**
* Trap handler for ALL exceptions/interrupts.
**************************************************************************/
void global_trap_handler(void) {
uint32_t cause = neorv32_cpu_csr_read(CSR_MCAUSE);
// clear pending FIRQ
if (cause & (1<<31)) {
neorv32_cpu_csr_write(CSR_MIP, ~(1 << (cause & 0xf)));
}
// hack: make "instruction access fault" exception resumable as we *exactly* know how to handle it in this case
if (cause == TRAP_CODE_I_ACCESS) {
neorv32_cpu_csr_write(CSR_MEPC, neorv32_cpu_csr_read(CSR_MEPC) + 4);
}
// hack: always come back in MACHINE MODE
neorv32_cpu_csr_set(CSR_MSTATUS, (1<<CSR_MSTATUS_MPP_H) | (1<<CSR_MSTATUS_MPP_L));
}
/**********************************************************************//**
* XIRQ handler channel 0.
**************************************************************************/
void xirq_trap_handler0(void) {
xirq_trap_handler_ack += 2;
}
/**********************************************************************//**
* XIRQ handler channel 1.
**************************************************************************/
void xirq_trap_handler1(void) {
xirq_trap_handler_ack *= 2;
}
/**********************************************************************//**
* Test results helper function: Shows "[ok]" and increments global cnt_ok
**************************************************************************/
void test_ok(void) {
PRINT_STANDARD("%c[1m[ok]%c[0m\n", 27, 27);
cnt_ok++;
}
/**********************************************************************//**
* Test results helper function: Shows "[FAIL]" and increments global cnt_fail
**************************************************************************/
void test_fail(void) {
PRINT_CRITICAL("%c[1m[fail(%u)]%c[0m\n", 27, cnt_test, 27);
cnt_fail++;
}
/**********************************************************************//**
* "after-main" handler that is executed after the application's
* main function returns (called by crt0.S start-up code): Output minimal
* test report to physical UART
**************************************************************************/
void __neorv32_crt0_after_main(int32_t return_code) {
// make sure sim mode is disabled and UARTs are actually enabled
NEORV32_UART0.CTRL |= (1 << UART_CTRL_EN);
NEORV32_UART0.CTRL &= ~(1 << UART_CTRL_SIM_MODE);
NEORV32_UART1.CTRL = NEORV32_UART0.CTRL;
// minimal result report
PRINT_CRITICAL("%u/%u\n", (uint32_t)return_code, (uint32_t)cnt_test);
}