freebsd-nq/sys/mips/nlm/xlp_machdep.c
Ruslan Bukin 84aec472fc Allow the use of soft-interrupts for sending IPIs.
This will be required for SMP support on MIPS Malta platform.

Reviewed by:	adrian
Sponsored by:	DARPA, AFRL
Sponsored by:	HEIF5
Differential Revision:	https://reviews.freebsd.org/D7835
2016-09-08 17:37:13 +00:00

742 lines
17 KiB
C

/*-
* Copyright 2003-2011 Netlogic Microsystems (Netlogic). 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.
*
* THIS SOFTWARE IS PROVIDED BY Netlogic Microsystems ``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 NETLOGIC 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.
*
* NETLOGIC_BSD */
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_ddb.h"
#include "opt_platform.h"
#include <sys/param.h>
#include <sys/bus.h>
#include <sys/conf.h>
#include <sys/rtprio.h>
#include <sys/systm.h>
#include <sys/interrupt.h>
#include <sys/limits.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mutex.h>
#include <sys/random.h>
#include <sys/cons.h> /* cinit() */
#include <sys/kdb.h>
#include <sys/reboot.h>
#include <sys/queue.h>
#include <sys/smp.h>
#include <sys/timetc.h>
#include <vm/vm.h>
#include <vm/vm_page.h>
#include <machine/cpu.h>
#include <machine/cpufunc.h>
#include <machine/cpuinfo.h>
#include <machine/tlb.h>
#include <machine/cpuregs.h>
#include <machine/frame.h>
#include <machine/hwfunc.h>
#include <machine/md_var.h>
#include <machine/asm.h>
#include <machine/pmap.h>
#include <machine/trap.h>
#include <machine/clock.h>
#include <machine/fls64.h>
#include <machine/intr_machdep.h>
#include <machine/smp.h>
#include <mips/nlm/hal/mips-extns.h>
#include <mips/nlm/hal/haldefs.h>
#include <mips/nlm/hal/iomap.h>
#include <mips/nlm/hal/sys.h>
#include <mips/nlm/hal/pic.h>
#include <mips/nlm/hal/uart.h>
#include <mips/nlm/hal/mmu.h>
#include <mips/nlm/hal/bridge.h>
#include <mips/nlm/hal/cpucontrol.h>
#include <mips/nlm/hal/cop2.h>
#include <mips/nlm/clock.h>
#include <mips/nlm/interrupt.h>
#include <mips/nlm/board.h>
#include <mips/nlm/xlp.h>
#include <mips/nlm/msgring.h>
#ifdef FDT
#include <dev/fdt/fdt_common.h>
#include <dev/ofw/openfirm.h>
#endif
/* 4KB static data aread to keep a copy of the bootload env until
the dynamic kenv is setup */
char boot1_env[4096];
uint64_t xlp_cpu_frequency;
uint64_t xlp_io_base = MIPS_PHYS_TO_DIRECT_UNCACHED(XLP_DEFAULT_IO_BASE);
int xlp_ncores;
int xlp_threads_per_core;
uint32_t xlp_hw_thread_mask;
int xlp_cpuid_to_hwtid[MAXCPU];
int xlp_hwtid_to_cpuid[MAXCPU];
uint64_t xlp_pic_base;
static int xlp_mmuval;
extern uint32_t _end;
extern char XLPResetEntry[], XLPResetEntryEnd[];
static void
xlp_setup_core(void)
{
uint64_t reg;
reg = nlm_mfcr(LSU_DEFEATURE);
/* Enable Unaligned and L2HPE */
reg |= (1 << 30) | (1 << 23);
/*
* Experimental : Enable SUE
* Speculative Unmap Enable. Enable speculative L2 cache request for
* unmapped access.
*/
reg |= (1ull << 31);
/* Clear S1RCM - A0 errata */
reg &= ~0xeull;
nlm_mtcr(LSU_DEFEATURE, reg);
reg = nlm_mfcr(SCHED_DEFEATURE);
/* Experimental: Disable BRU accepting ALU ops - A0 errata */
reg |= (1 << 24);
nlm_mtcr(SCHED_DEFEATURE, reg);
}
static void
xlp_setup_mmu(void)
{
uint32_t pagegrain;
if (nlm_threadid() == 0) {
nlm_setup_extended_pagemask(0);
nlm_large_variable_tlb_en(1);
nlm_extended_tlb_en(1);
nlm_mmu_setup(0, 0, 0);
}
/* Enable no-read, no-exec, large-physical-address */
pagegrain = mips_rd_pagegrain();
pagegrain |= (1U << 31) | /* RIE */
(1 << 30) | /* XIE */
(1 << 29); /* ELPA */
mips_wr_pagegrain(pagegrain);
}
static void
xlp_enable_blocks(void)
{
uint64_t sysbase;
int i;
for (i = 0; i < XLP_MAX_NODES; i++) {
if (!nlm_dev_exists(XLP_IO_SYS_OFFSET(i)))
continue;
sysbase = nlm_get_sys_regbase(i);
nlm_sys_enable_block(sysbase, DFS_DEVICE_RSA);
}
}
static void
xlp_parse_mmu_options(void)
{
uint64_t sysbase;
uint32_t cpu_map = xlp_hw_thread_mask;
uint32_t core0_thr_mask, core_thr_mask, cpu_rst_mask;
int i, j, k;
#ifdef SMP
if (cpu_map == 0)
cpu_map = 0xffffffff;
#else /* Uniprocessor! */
if (cpu_map == 0)
cpu_map = 0x1;
else if (cpu_map != 0x1) {
printf("WARNING: Starting uniprocessor kernel on cpumask [0x%lx]!\n"
"WARNING: Other CPUs will be unused.\n", (u_long)cpu_map);
cpu_map = 0x1;
}
#endif
xlp_ncores = 1;
core0_thr_mask = cpu_map & 0xf;
switch (core0_thr_mask) {
case 1:
xlp_threads_per_core = 1;
xlp_mmuval = 0;
break;
case 3:
xlp_threads_per_core = 2;
xlp_mmuval = 2;
break;
case 0xf:
xlp_threads_per_core = 4;
xlp_mmuval = 3;
break;
default:
goto unsupp;
}
/* Try to find the enabled cores from SYS block */
sysbase = nlm_get_sys_regbase(0);
cpu_rst_mask = nlm_read_sys_reg(sysbase, SYS_CPU_RESET) & 0xff;
/* XLP 416 does not report this correctly, fix */
if (nlm_processor_id() == CHIP_PROCESSOR_ID_XLP_416)
cpu_rst_mask = 0xe;
/* Take out cores which do not exist on chip */
for (i = 1; i < XLP_MAX_CORES; i++) {
if ((cpu_rst_mask & (1 << i)) == 0)
cpu_map &= ~(0xfu << (4 * i));
}
/* Verify other cores' CPU masks */
for (i = 1; i < XLP_MAX_CORES; i++) {
core_thr_mask = (cpu_map >> (4 * i)) & 0xf;
if (core_thr_mask == 0)
continue;
if (core_thr_mask != core0_thr_mask)
goto unsupp;
xlp_ncores++;
}
xlp_hw_thread_mask = cpu_map;
/* setup hardware processor id to cpu id mapping */
for (i = 0; i< MAXCPU; i++)
xlp_cpuid_to_hwtid[i] =
xlp_hwtid_to_cpuid[i] = -1;
for (i = 0, k = 0; i < XLP_MAX_CORES; i++) {
if (((cpu_map >> (i * 4)) & 0xf) == 0)
continue;
for (j = 0; j < xlp_threads_per_core; j++) {
xlp_cpuid_to_hwtid[k] = i * 4 + j;
xlp_hwtid_to_cpuid[i * 4 + j] = k;
k++;
}
}
return;
unsupp:
printf("ERROR : Unsupported CPU mask [use 1,2 or 4 threads per core].\n"
"\tcore0 thread mask [%lx], boot cpu mask [%lx].\n",
(u_long)core0_thr_mask, (u_long)cpu_map);
panic("Invalid CPU mask - halting.\n");
return;
}
/* Parse cmd line args as env - copied from ar71xx */
static void
xlp_parse_bootargs(char *cmdline)
{
char *n, *v;
while ((v = strsep(&cmdline, " \n")) != NULL) {
if (*v == '\0')
continue;
if (*v == '-') {
while (*v != '\0') {
v++;
switch (*v) {
case 'a': boothowto |= RB_ASKNAME; break;
case 'd': boothowto |= RB_KDB; break;
case 'g': boothowto |= RB_GDB; break;
case 's': boothowto |= RB_SINGLE; break;
case 'v': boothowto |= RB_VERBOSE; break;
}
}
} else {
n = strsep(&v, "=");
if (v == NULL)
kern_setenv(n, "1");
else
kern_setenv(n, v);
}
}
}
#ifdef FDT
static void
xlp_bootargs_init(__register_t arg)
{
char buf[2048]; /* early stack is big enough */
void *dtbp;
phandle_t chosen;
ihandle_t mask;
dtbp = (void *)(intptr_t)arg;
#if defined(FDT_DTB_STATIC)
/*
* In case the device tree blob was not passed as argument try
* to use the statically embedded one.
*/
if (dtbp == NULL)
dtbp = &fdt_static_dtb;
#endif
if (OF_install(OFW_FDT, 0) == FALSE)
while (1);
if (OF_init((void *)dtbp) != 0)
while (1);
OF_interpret("perform-fixup", 0);
chosen = OF_finddevice("/chosen");
if (OF_getprop(chosen, "cpumask", &mask, sizeof(mask)) != -1) {
xlp_hw_thread_mask = mask;
}
if (OF_getprop(chosen, "bootargs", buf, sizeof(buf)) != -1)
xlp_parse_bootargs(buf);
}
#else
/*
* arg is a pointer to the environment block, the format of the block is
* a=xyz\0b=pqr\0\0
*/
static void
xlp_bootargs_init(__register_t arg)
{
char buf[2048]; /* early stack is big enough */
char *p, *v, *n;
uint32_t mask;
/*
* provide backward compat for passing cpu mask as arg
*/
if (arg & 1) {
xlp_hw_thread_mask = arg;
return;
}
p = (void *)(intptr_t)arg;
while (*p != '\0') {
strlcpy(buf, p, sizeof(buf));
v = buf;
n = strsep(&v, "=");
if (v == NULL)
kern_setenv(n, "1");
else
kern_setenv(n, v);
p += strlen(p) + 1;
}
/* CPU mask can be passed thru env */
if (getenv_uint("cpumask", &mask) != 0)
xlp_hw_thread_mask = mask;
/* command line argument */
v = kern_getenv("bootargs");
if (v != NULL) {
strlcpy(buf, v, sizeof(buf));
xlp_parse_bootargs(buf);
freeenv(v);
}
}
#endif
static void
mips_init(void)
{
init_param1();
init_param2(physmem);
mips_cpu_init();
cpuinfo.cache_coherent_dma = TRUE;
pmap_bootstrap();
mips_proc0_init();
mutex_init();
#ifdef DDB
kdb_init();
if (boothowto & RB_KDB) {
kdb_enter("Boot flags requested debugger", NULL);
}
#endif
}
unsigned int
platform_get_timecount(struct timecounter *tc __unused)
{
uint64_t count = nlm_pic_read_timer(xlp_pic_base, PIC_CLOCK_TIMER);
return (unsigned int)~count;
}
static void
xlp_pic_init(void)
{
struct timecounter pic_timecounter = {
platform_get_timecount, /* get_timecount */
0, /* no poll_pps */
~0U, /* counter_mask */
XLP_IO_CLK, /* frequency */
"XLRPIC", /* name */
2000, /* quality (adjusted in code) */
};
int i;
int maxirt;
xlp_pic_base = nlm_get_pic_regbase(0); /* TOOD: Add other nodes */
maxirt = nlm_read_reg(nlm_get_pic_pcibase(nlm_nodeid()),
XLP_PCI_DEVINFO_REG0);
printf("Initializing PIC...@%jx %d IRTs\n", (uintmax_t)xlp_pic_base,
maxirt);
/* Bind all PIC irqs to cpu 0 */
for (i = 0; i < maxirt; i++)
nlm_pic_write_irt(xlp_pic_base, i, 0, 0, 1, 0,
1, 0, 0x1);
nlm_pic_set_timer(xlp_pic_base, PIC_CLOCK_TIMER, ~0ULL, 0, 0);
platform_timecounter = &pic_timecounter;
}
#if defined(__mips_n32) || defined(__mips_n64) /* PHYSADDR_64_BIT */
#ifdef XLP_SIM
#define XLP_MEM_LIM 0x200000000ULL
#else
#define XLP_MEM_LIM 0x10000000000ULL
#endif
#else
#define XLP_MEM_LIM 0xfffff000UL
#endif
static vm_paddr_t xlp_mem_excl[] = {
0, 0, /* for kernel image region, see xlp_mem_init */
0x0c000000, 0x14000000, /* uboot area, cms queue and other stuff */
0x1fc00000, 0x1fd00000, /* reset vec */
0x1e000000, 0x1e200000, /* poe buffers */
};
static int
mem_exclude_add(vm_paddr_t *avail, vm_paddr_t mstart, vm_paddr_t mend)
{
int i, pos;
pos = 0;
for (i = 0; i < nitems(xlp_mem_excl); i += 2) {
if (mstart > xlp_mem_excl[i + 1])
continue;
if (mstart < xlp_mem_excl[i]) {
avail[pos++] = mstart;
if (mend < xlp_mem_excl[i])
avail[pos++] = mend;
else
avail[pos++] = xlp_mem_excl[i];
}
mstart = xlp_mem_excl[i + 1];
if (mend <= mstart)
break;
}
if (mstart < mend) {
avail[pos++] = mstart;
avail[pos++] = mend;
}
return (pos);
}
static void
xlp_mem_init(void)
{
vm_paddr_t physsz, tmp;
uint64_t bridgebase, base, lim, val;
int i, j, k, n;
/* update kernel image area in exclude regions */
tmp = (vm_paddr_t)MIPS_KSEG0_TO_PHYS(&_end);
tmp = round_page(tmp) + 0x20000; /* round up */
xlp_mem_excl[1] = tmp;
printf("Memory (from DRAM BARs):\n");
bridgebase = nlm_get_bridge_regbase(0); /* TODO: Add other nodes */
physsz = 0;
for (i = 0, j = 0; i < 8; i++) {
val = nlm_read_bridge_reg(bridgebase, BRIDGE_DRAM_BAR(i));
val = (val >> 12) & 0xfffff;
base = val << 20;
val = nlm_read_bridge_reg(bridgebase, BRIDGE_DRAM_LIMIT(i));
val = (val >> 12) & 0xfffff;
if (val == 0) /* BAR not enabled */
continue;
lim = (val + 1) << 20;
printf(" BAR %d: %#jx - %#jx : ", i, (intmax_t)base,
(intmax_t)lim);
if (lim <= base) {
printf("\tskipped - malformed %#jx -> %#jx\n",
(intmax_t)base, (intmax_t)lim);
continue;
} else if (base >= XLP_MEM_LIM) {
printf(" skipped - outside usable limit %#jx.\n",
(intmax_t)XLP_MEM_LIM);
continue;
} else if (lim >= XLP_MEM_LIM) {
lim = XLP_MEM_LIM;
printf(" truncated to %#jx.\n", (intmax_t)XLP_MEM_LIM);
} else
printf(" usable\n");
/* exclude unusable regions from BAR and add rest */
n = mem_exclude_add(&phys_avail[j], base, lim);
for (k = j; k < j + n; k += 2) {
physsz += phys_avail[k + 1] - phys_avail[k];
printf("\tMem[%d]: %#jx - %#jx\n", k/2,
(intmax_t)phys_avail[k], (intmax_t)phys_avail[k+1]);
}
j = k;
}
/* setup final entry with 0 */
phys_avail[j] = phys_avail[j + 1] = 0;
/* copy phys_avail to dump_avail */
for (i = 0; i <= j + 1; i++)
dump_avail[i] = phys_avail[i];
realmem = physmem = btoc(physsz);
}
void
platform_start(__register_t a0 __unused,
__register_t a1 __unused,
__register_t a2 __unused,
__register_t a3 __unused)
{
/* Initialize pcpu stuff */
mips_pcpu0_init();
/* initialize console so that we have printf */
boothowto |= (RB_SERIAL | RB_MULTIPLE); /* Use multiple consoles */
init_static_kenv(boot1_env, sizeof(boot1_env));
xlp_bootargs_init(a0);
/* clockrate used by delay, so initialize it here */
xlp_cpu_frequency = xlp_get_cpu_frequency(0, 0);
cpu_clock = xlp_cpu_frequency / 1000000;
mips_timer_early_init(xlp_cpu_frequency);
/* Init console please */
cninit();
/* Early core init and fixes for errata */
xlp_setup_core();
xlp_parse_mmu_options();
xlp_mem_init();
bcopy(XLPResetEntry, (void *)MIPS_RESET_EXC_VEC,
XLPResetEntryEnd - XLPResetEntry);
#ifdef SMP
/*
* We will enable the other threads in core 0 here
* so that the TLB and cache info is correct when
* mips_init runs
*/
xlp_enable_threads(xlp_mmuval);
#endif
/* setup for the startup core */
xlp_setup_mmu();
xlp_enable_blocks();
/* Read/Guess/setup board information */
nlm_board_info_setup();
/* MIPS generic init */
mips_init();
/*
* XLP specific post initialization
* initialize other on chip stuff
*/
xlp_pic_init();
mips_timer_init_params(xlp_cpu_frequency, 0);
}
void
platform_cpu_init()
{
}
void
platform_reset(void)
{
uint64_t sysbase = nlm_get_sys_regbase(0);
nlm_write_sys_reg(sysbase, SYS_CHIP_RESET, 1);
for( ; ; )
__asm __volatile("wait");
}
#ifdef SMP
/*
* XLP threads are started simultaneously when we enable threads, this will
* ensure that the threads are blocked in platform_init_ap, until they are
* ready to proceed to smp_init_secondary()
*/
static volatile int thr_unblock[4];
int
platform_start_ap(int cpuid)
{
uint32_t coremask, val;
uint64_t sysbase = nlm_get_sys_regbase(0);
int hwtid = xlp_cpuid_to_hwtid[cpuid];
int core, thr;
core = hwtid / 4;
thr = hwtid % 4;
if (thr == 0) {
/* First thread in core, do core wake up */
coremask = 1u << core;
/* Enable core clock */
val = nlm_read_sys_reg(sysbase, SYS_CORE_DFS_DIS_CTRL);
val &= ~coremask;
nlm_write_sys_reg(sysbase, SYS_CORE_DFS_DIS_CTRL, val);
/* Remove CPU Reset */
val = nlm_read_sys_reg(sysbase, SYS_CPU_RESET);
val &= ~coremask & 0xff;
nlm_write_sys_reg(sysbase, SYS_CPU_RESET, val);
if (bootverbose)
printf("Waking up core %d ...", core);
/* Poll for CPU to mark itself coherent */
do {
val = nlm_read_sys_reg(sysbase, SYS_CPU_NONCOHERENT_MODE);
} while ((val & coremask) != 0);
if (bootverbose)
printf("Done\n");
} else {
/* otherwise release the threads stuck in platform_init_ap */
thr_unblock[thr] = 1;
}
return (0);
}
void
platform_init_ap(int cpuid)
{
uint32_t stat;
int thr;
/* The first thread has to setup the MMU and enable other threads */
thr = nlm_threadid();
if (thr == 0) {
xlp_setup_core();
xlp_enable_threads(xlp_mmuval);
} else {
/*
* FIXME busy wait here eats too many cycles, especially
* in the core 0 while bootup
*/
while (thr_unblock[thr] == 0)
__asm__ __volatile__ ("nop;nop;nop;nop");
thr_unblock[thr] = 0;
}
xlp_setup_mmu();
stat = mips_rd_status();
KASSERT((stat & MIPS_SR_INT_IE) == 0,
("Interrupts enabled in %s!", __func__));
stat |= MIPS_SR_COP_2_BIT | MIPS_SR_COP_0_BIT;
mips_wr_status(stat);
nlm_write_c0_eimr(0ull);
xlp_enable_irq(IRQ_IPI);
xlp_enable_irq(IRQ_TIMER);
xlp_enable_irq(IRQ_MSGRING);
return;
}
int
platform_ipi_hardintr_num(void)
{
return (IRQ_IPI);
}
int
platform_ipi_softintr_num(void)
{
return (-1);
}
void
platform_ipi_send(int cpuid)
{
nlm_pic_send_ipi(xlp_pic_base, xlp_cpuid_to_hwtid[cpuid],
platform_ipi_hardintr_num(), 0);
}
void
platform_ipi_clear(void)
{
}
int
platform_processor_id(void)
{
return (xlp_hwtid_to_cpuid[nlm_cpuid()]);
}
void
platform_cpu_mask(cpuset_t *mask)
{
int i, s;
CPU_ZERO(mask);
s = xlp_ncores * xlp_threads_per_core;
for (i = 0; i < s; i++)
CPU_SET(i, mask);
}
struct cpu_group *
platform_smp_topo()
{
return (smp_topo_2level(CG_SHARE_L2, xlp_ncores, CG_SHARE_L1,
xlp_threads_per_core, CG_FLAG_THREAD));
}
#endif