提交 77ae798f 编写于 作者: J Jayachandran C 提交者: John Crispin

MIPS: Netlogic: Support for multi-chip configuration

Upto 4 Netlogic XLP SoCs can be connected over ICI links to form a
coherent multi-node system.  Each SoC has its own set of on-chip
devices including PIC.  To support this, add a per SoC stucture and
use it for the PIC and SYS block addresses instead of using global
variables.
Signed-off-by: NJayachandran C <jchandra@broadcom.com>
Patchwork: http://patchwork.linux-mips.org/patch/4469Signed-off-by: NJohn Crispin <blogic@openwrt.org>
上级 2a37b1ae
......@@ -46,10 +46,10 @@
#ifndef __ASSEMBLY__
#include <linux/cpumask.h>
#include <linux/spinlock.h>
#include <asm/irq.h>
struct irq_desc;
extern struct plat_smp_ops nlm_smp_ops;
extern char nlm_reset_entry[], nlm_reset_entry_end[];
void nlm_smp_function_ipi_handler(unsigned int irq, struct irq_desc *desc);
void nlm_smp_resched_ipi_handler(unsigned int irq, struct irq_desc *desc);
void nlm_smp_irq_init(void);
......@@ -70,10 +70,42 @@ nlm_set_nmi_handler(void *handler)
* Misc.
*/
unsigned int nlm_get_cpu_frequency(void);
void nlm_node_init(int node);
extern struct plat_smp_ops nlm_smp_ops;
extern char nlm_reset_entry[], nlm_reset_entry_end[];
extern unsigned long nlm_common_ebase;
extern int nlm_threads_per_core;
extern uint32_t nlm_coremask;
extern unsigned int nlm_threads_per_core;
extern cpumask_t nlm_cpumask;
struct nlm_soc_info {
unsigned long coremask; /* cores enabled on the soc */
unsigned long ebase;
uint64_t irqmask;
uint64_t sysbase; /* only for XLP */
uint64_t picbase;
spinlock_t piclock;
};
#define NLM_CORES_PER_NODE 8
#define NLM_THREADS_PER_CORE 4
#define NLM_CPUS_PER_NODE (NLM_CORES_PER_NODE * NLM_THREADS_PER_CORE)
#define nlm_get_node(i) (&nlm_nodes[i])
#define NLM_NR_NODES 1
#define nlm_current_node() (&nlm_nodes[0])
struct irq_data;
uint64_t nlm_pci_irqmask(int node);
void nlm_set_pic_extra_ack(int node, int irq, void (*xack)(struct irq_data *));
/*
* The NR_IRQs is divided between nodes, each of them has a separate irq space
*/
static inline int nlm_irq_to_xirq(int node, int irq)
{
return node * NR_IRQS / NLM_NR_NODES + irq;
}
extern struct nlm_soc_info nlm_nodes[NLM_NR_NODES];
extern int nlm_cpu_ready[];
#endif
#endif /* _NETLOGIC_COMMON_H_ */
......@@ -73,4 +73,9 @@ static inline int hard_smp_processor_id(void)
return __read_32bit_c0_register($15, 1) & 0x3ff;
}
static inline int nlm_nodeid(void)
{
return (__read_32bit_c0_register($15, 1) >> 5) & 0x3;
}
#endif /*_ASM_NLM_MIPS_EXTS_H */
......@@ -381,7 +381,6 @@ nlm_pic_init_irt(uint64_t base, int irt, int irq, int hwt)
nlm_pic_write_irt_direct(base, irt, 0, 0, 0, irq, hwt);
}
extern uint64_t nlm_pic_base;
int nlm_irq_to_irt(int irq);
int nlm_irt_to_irq(int irt);
......
......@@ -124,6 +124,5 @@
#define nlm_get_sys_pcibase(node) nlm_pcicfg_base(XLP_IO_SYS_OFFSET(node))
#define nlm_get_sys_regbase(node) (nlm_get_sys_pcibase(node) + XLP_IO_PCI_HDRSZ)
extern uint64_t nlm_sys_base;
#endif
#endif
......@@ -258,7 +258,5 @@ nlm_pic_init_irt(uint64_t base, int irt, int irq, int hwt)
nlm_write_reg(base, PIC_IRT_1(irt),
(1 << 30) | (1 << 6) | irq);
}
extern uint64_t nlm_pic_base;
#endif
#endif /* _ASM_NLM_XLR_PIC_H */
......@@ -70,33 +70,34 @@
*/
/* Globals */
static uint64_t nlm_irq_mask;
static DEFINE_SPINLOCK(nlm_pic_lock);
static void xlp_pic_enable(struct irq_data *d)
{
unsigned long flags;
struct nlm_soc_info *nodep;
int irt;
nodep = nlm_current_node();
irt = nlm_irq_to_irt(d->irq);
if (irt == -1)
return;
spin_lock_irqsave(&nlm_pic_lock, flags);
nlm_pic_enable_irt(nlm_pic_base, irt);
spin_unlock_irqrestore(&nlm_pic_lock, flags);
spin_lock_irqsave(&nodep->piclock, flags);
nlm_pic_enable_irt(nodep->picbase, irt);
spin_unlock_irqrestore(&nodep->piclock, flags);
}
static void xlp_pic_disable(struct irq_data *d)
{
struct nlm_soc_info *nodep;
unsigned long flags;
int irt;
nodep = nlm_current_node();
irt = nlm_irq_to_irt(d->irq);
if (irt == -1)
return;
spin_lock_irqsave(&nlm_pic_lock, flags);
nlm_pic_disable_irt(nlm_pic_base, irt);
spin_unlock_irqrestore(&nlm_pic_lock, flags);
spin_lock_irqsave(&nodep->piclock, flags);
nlm_pic_disable_irt(nodep->picbase, irt);
spin_unlock_irqrestore(&nodep->piclock, flags);
}
static void xlp_pic_mask_ack(struct irq_data *d)
......@@ -109,8 +110,10 @@ static void xlp_pic_mask_ack(struct irq_data *d)
static void xlp_pic_unmask(struct irq_data *d)
{
void *hd = irq_data_get_irq_handler_data(d);
struct nlm_soc_info *nodep;
int irt;
nodep = nlm_current_node();
irt = nlm_irq_to_irt(d->irq);
if (irt == -1)
return;
......@@ -120,7 +123,7 @@ static void xlp_pic_unmask(struct irq_data *d)
extra_ack(d);
}
/* Ack is a single write, no need to lock */
nlm_pic_ack(nlm_pic_base, irt);
nlm_pic_ack(nodep->picbase, irt);
}
static struct irq_chip xlp_pic = {
......@@ -177,7 +180,11 @@ struct irq_chip nlm_cpu_intr = {
void __init init_nlm_common_irqs(void)
{
int i, irq, irt;
uint64_t irqmask;
struct nlm_soc_info *nodep;
nodep = nlm_current_node();
irqmask = (1ULL << IRQ_TIMER);
for (i = 0; i < PIC_IRT_FIRST_IRQ; i++)
irq_set_chip_and_handler(i, &nlm_cpu_intr, handle_percpu_irq);
......@@ -189,7 +196,7 @@ void __init init_nlm_common_irqs(void)
nlm_smp_function_ipi_handler);
irq_set_chip_and_handler(IRQ_IPI_SMP_RESCHEDULE, &nlm_cpu_intr,
nlm_smp_resched_ipi_handler);
nlm_irq_mask |=
irqmask |=
((1ULL << IRQ_IPI_SMP_FUNCTION) | (1ULL << IRQ_IPI_SMP_RESCHEDULE));
#endif
......@@ -197,11 +204,11 @@ void __init init_nlm_common_irqs(void)
irt = nlm_irq_to_irt(irq);
if (irt == -1)
continue;
nlm_irq_mask |= (1ULL << irq);
nlm_pic_init_irt(nlm_pic_base, irt, irq, 0);
irqmask |= (1ULL << irq);
nlm_pic_init_irt(nodep->picbase, irt, irq, 0);
}
nlm_irq_mask |= (1ULL << IRQ_TIMER);
nodep->irqmask = irqmask;
}
void __init arch_init_irq(void)
......@@ -209,29 +216,39 @@ void __init arch_init_irq(void)
/* Initialize the irq descriptors */
init_nlm_common_irqs();
write_c0_eimr(nlm_irq_mask);
write_c0_eimr(nlm_current_node()->irqmask);
}
void __cpuinit nlm_smp_irq_init(void)
{
/* set interrupt mask for non-zero cpus */
write_c0_eimr(nlm_irq_mask);
write_c0_eimr(nlm_current_node()->irqmask);
}
asmlinkage void plat_irq_dispatch(void)
{
uint64_t eirr;
int i;
int i, node;
node = nlm_nodeid();
eirr = read_c0_eirr() & read_c0_eimr();
if (eirr & (1 << IRQ_TIMER)) {
do_IRQ(IRQ_TIMER);
return;
}
#ifdef CONFIG_SMP
if (eirr & IRQ_IPI_SMP_FUNCTION) {
do_IRQ(IRQ_IPI_SMP_FUNCTION);
return;
}
if (eirr & IRQ_IPI_SMP_RESCHEDULE) {
do_IRQ(IRQ_IPI_SMP_RESCHEDULE);
return;
}
#endif
i = __ilog2_u64(eirr);
if (i == -1)
return;
do_IRQ(i);
do_IRQ(nlm_irq_to_xirq(node, i));
}
......@@ -59,12 +59,17 @@
void nlm_send_ipi_single(int logical_cpu, unsigned int action)
{
int cpu = cpu_logical_map(logical_cpu);
int cpu, node;
uint64_t picbase;
cpu = cpu_logical_map(logical_cpu);
node = cpu / NLM_CPUS_PER_NODE;
picbase = nlm_get_node(node)->picbase;
if (action & SMP_CALL_FUNCTION)
nlm_pic_send_ipi(nlm_pic_base, cpu, IRQ_IPI_SMP_FUNCTION, 0);
nlm_pic_send_ipi(picbase, cpu, IRQ_IPI_SMP_FUNCTION, 0);
if (action & SMP_RESCHEDULE_YOURSELF)
nlm_pic_send_ipi(nlm_pic_base, cpu, IRQ_IPI_SMP_RESCHEDULE, 0);
nlm_pic_send_ipi(picbase, cpu, IRQ_IPI_SMP_RESCHEDULE, 0);
}
void nlm_send_ipi_mask(const struct cpumask *mask, unsigned int action)
......@@ -96,11 +101,12 @@ void nlm_smp_resched_ipi_handler(unsigned int irq, struct irq_desc *desc)
void nlm_early_init_secondary(int cpu)
{
change_c0_config(CONF_CM_CMASK, 0x3);
write_c0_ebase((uint32_t)nlm_common_ebase);
#ifdef CONFIG_CPU_XLP
if (cpu % 4 == 0)
/* mmu init, once per core */
if (cpu % NLM_THREADS_PER_CORE == 0)
xlp_mmu_init();
#endif
write_c0_ebase(nlm_current_node()->ebase);
}
/*
......@@ -108,7 +114,7 @@ void nlm_early_init_secondary(int cpu)
*/
static void __cpuinit nlm_init_secondary(void)
{
current_cpu_data.core = hard_smp_processor_id() / 4;
current_cpu_data.core = hard_smp_processor_id() / NLM_THREADS_PER_CORE;
nlm_smp_irq_init();
}
......@@ -142,22 +148,22 @@ cpumask_t phys_cpu_present_map;
void nlm_boot_secondary(int logical_cpu, struct task_struct *idle)
{
unsigned long gp = (unsigned long)task_thread_info(idle);
unsigned long sp = (unsigned long)__KSTK_TOS(idle);
int cpu = cpu_logical_map(logical_cpu);
int cpu, node;
nlm_next_sp = sp;
nlm_next_gp = gp;
cpu = cpu_logical_map(logical_cpu);
node = cpu / NLM_CPUS_PER_NODE;
nlm_next_sp = (unsigned long)__KSTK_TOS(idle);
nlm_next_gp = (unsigned long)task_thread_info(idle);
/* barrier */
/* barrier for sp/gp store above */
__sync();
nlm_pic_send_ipi(nlm_pic_base, cpu, 1, 1);
nlm_pic_send_ipi(nlm_get_node(node)->picbase, cpu, 1, 1); /* NMI */
}
void __init nlm_smp_setup(void)
{
unsigned int boot_cpu;
int num_cpus, i;
int num_cpus, i, ncore;
boot_cpu = hard_smp_processor_id();
cpumask_clear(&phys_cpu_present_map);
......@@ -182,11 +188,16 @@ void __init nlm_smp_setup(void)
}
}
/* check with the cores we have worken up */
for (ncore = 0, i = 0; i < NLM_NR_NODES; i++)
ncore += hweight32(nlm_get_node(i)->coremask);
pr_info("Phys CPU present map: %lx, possible map %lx\n",
(unsigned long)cpumask_bits(&phys_cpu_present_map)[0],
(unsigned long)cpumask_bits(cpu_possible_mask)[0]);
pr_info("Detected %i Slave CPU(s)\n", num_cpus);
pr_info("Detected (%dc%dt) %d Slave CPU(s)\n", ncore,
nlm_threads_per_core, num_cpus);
nlm_set_nmi_handler(nlm_boot_secondary_cpus);
}
......@@ -196,7 +207,7 @@ static int nlm_parse_cpumask(cpumask_t *wakeup_mask)
int threadmode, i, j;
core0_thr_mask = 0;
for (i = 0; i < 4; i++)
for (i = 0; i < NLM_THREADS_PER_CORE; i++)
if (cpumask_test_cpu(i, wakeup_mask))
core0_thr_mask |= (1 << i);
switch (core0_thr_mask) {
......@@ -217,9 +228,9 @@ static int nlm_parse_cpumask(cpumask_t *wakeup_mask)
}
/* Verify other cores CPU masks */
for (i = 0; i < NR_CPUS; i += 4) {
for (i = 0; i < NR_CPUS; i += NLM_THREADS_PER_CORE) {
core_thr_mask = 0;
for (j = 0; j < 4; j++)
for (j = 0; j < NLM_THREADS_PER_CORE; j++)
if (cpumask_test_cpu(i + j, wakeup_mask))
core_thr_mask |= (1 << j);
if (core_thr_mask != 0 && core_thr_mask != core0_thr_mask)
......
......@@ -40,23 +40,23 @@
#include <asm/mipsregs.h>
#include <asm/time.h>
#include <asm/netlogic/common.h>
#include <asm/netlogic/haldefs.h>
#include <asm/netlogic/xlp-hal/iomap.h>
#include <asm/netlogic/xlp-hal/xlp.h>
#include <asm/netlogic/xlp-hal/pic.h>
#include <asm/netlogic/xlp-hal/sys.h>
/* These addresses are computed by the nlm_hal_init() */
uint64_t nlm_io_base;
uint64_t nlm_sys_base;
uint64_t nlm_pic_base;
/* Main initialization */
void nlm_hal_init(void)
void nlm_node_init(int node)
{
nlm_io_base = CKSEG1ADDR(XLP_DEFAULT_IO_BASE);
nlm_sys_base = nlm_get_sys_regbase(0); /* node 0 */
nlm_pic_base = nlm_get_pic_regbase(0); /* node 0 */
struct nlm_soc_info *nodep;
nodep = nlm_get_node(node);
nodep->sysbase = nlm_get_sys_regbase(node);
nodep->picbase = nlm_get_pic_regbase(node);
nodep->ebase = read_c0_ebase() & (~((1 << 12) - 1));
spin_lock_init(&nodep->piclock);
}
int nlm_irq_to_irt(int irq)
......@@ -138,14 +138,15 @@ int nlm_irt_to_irq(int irt)
}
}
unsigned int nlm_get_core_frequency(int core)
unsigned int nlm_get_core_frequency(int node, int core)
{
unsigned int pll_divf, pll_divr, dfs_div, ext_div;
unsigned int rstval, dfsval, denom;
uint64_t num;
uint64_t num, sysbase;
rstval = nlm_read_sys_reg(nlm_sys_base, SYS_POWER_ON_RESET_CFG);
dfsval = nlm_read_sys_reg(nlm_sys_base, SYS_CORE_DFS_DIV_VALUE);
sysbase = nlm_get_node(node)->sysbase;
rstval = nlm_read_sys_reg(sysbase, SYS_POWER_ON_RESET_CFG);
dfsval = nlm_read_sys_reg(sysbase, SYS_CORE_DFS_DIV_VALUE);
pll_divf = ((rstval >> 10) & 0x7f) + 1;
pll_divr = ((rstval >> 8) & 0x3) + 1;
ext_div = ((rstval >> 30) & 0x3) + 1;
......@@ -159,5 +160,5 @@ unsigned int nlm_get_core_frequency(int core)
unsigned int nlm_get_cpu_frequency(void)
{
return nlm_get_core_frequency(0);
return nlm_get_core_frequency(0, 0);
}
......@@ -52,17 +52,17 @@
#include <asm/netlogic/xlp-hal/xlp.h>
#include <asm/netlogic/xlp-hal/sys.h>
unsigned long nlm_common_ebase = 0x0;
/* default to uniprocessor */
uint32_t nlm_coremask = 1;
uint64_t nlm_io_base;
struct nlm_soc_info nlm_nodes[NLM_NR_NODES];
cpumask_t nlm_cpumask = CPU_MASK_CPU0;
int nlm_threads_per_core = 1;
unsigned int nlm_threads_per_core;
extern u32 __dtb_start[];
static void nlm_linux_exit(void)
{
nlm_write_sys_reg(nlm_sys_base, SYS_CHIP_RESET, 1);
uint64_t sysbase = nlm_get_node(0)->sysbase;
nlm_write_sys_reg(sysbase, SYS_CHIP_RESET, 1);
for ( ; ; )
cpu_wait();
}
......@@ -110,10 +110,9 @@ void xlp_mmu_init(void)
void __init prom_init(void)
{
nlm_io_base = CKSEG1ADDR(XLP_DEFAULT_IO_BASE);
xlp_mmu_init();
nlm_hal_init();
nlm_common_ebase = read_c0_ebase() & (~((1 << 12) - 1));
nlm_node_init(0);
#ifdef CONFIG_SMP
cpumask_setall(&nlm_cpumask);
......
......@@ -79,32 +79,38 @@ static int xlp_wakeup_core(uint64_t sysbase, int core)
static void xlp_enable_secondary_cores(const cpumask_t *wakeup_mask)
{
uint64_t syspcibase, sysbase;
struct nlm_soc_info *nodep;
uint64_t syspcibase;
uint32_t syscoremask;
int core, n;
int core, n, cpu;
for (n = 0; n < 4; n++) {
for (n = 0; n < NLM_NR_NODES; n++) {
syspcibase = nlm_get_sys_pcibase(n);
if (nlm_read_reg(syspcibase, 0) == 0xffffffff)
break;
/* read cores in reset from SYS and account for boot cpu */
sysbase = nlm_get_sys_regbase(n);
syscoremask = nlm_read_sys_reg(sysbase, SYS_CPU_RESET);
nlm_node_init(n);
nodep = nlm_get_node(n);
syscoremask = nlm_read_sys_reg(nodep->sysbase, SYS_CPU_RESET);
if (n == 0)
syscoremask |= 1;
for (core = 0; core < 8; core++) {
for (core = 0; core < NLM_CORES_PER_NODE; core++) {
/* see if the core exists */
if ((syscoremask & (1 << core)) == 0)
continue;
/* see if at least the first thread is enabled */
if (!cpumask_test_cpu((n * 8 + core) * 4, wakeup_mask))
cpu = (n * NLM_CORES_PER_NODE + core)
* NLM_THREADS_PER_CORE;
if (!cpumask_test_cpu(cpu, wakeup_mask))
continue;
/* wake up the core */
if (!xlp_wakeup_core(sysbase, core))
if (xlp_wakeup_core(nodep->sysbase, core))
nodep->coremask |= 1u << core;
else
pr_err("Failed to enable core %d\n", core);
}
}
......
......@@ -51,14 +51,11 @@
#include <asm/netlogic/xlr/gpio.h>
uint64_t nlm_io_base = DEFAULT_NETLOGIC_IO_BASE;
uint64_t nlm_pic_base;
struct psb_info nlm_prom_info;
unsigned long nlm_common_ebase = 0x0;
/* default to uniprocessor */
uint32_t nlm_coremask = 1;
int nlm_threads_per_core = 1;
unsigned int nlm_threads_per_core = 1;
struct nlm_soc_info nlm_nodes[NLM_NR_NODES];
cpumask_t nlm_cpumask = CPU_MASK_CPU0;
static void __init nlm_early_serial_setup(void)
......@@ -177,6 +174,16 @@ static void prom_add_memory(void)
}
}
static void nlm_init_node(void)
{
struct nlm_soc_info *nodep;
nodep = nlm_current_node();
nodep->picbase = nlm_mmio_base(NETLOGIC_IO_PIC_OFFSET);
nodep->ebase = read_c0_ebase() & (~((1 << 12) - 1));
spin_lock_init(&nodep->piclock);
}
void __init prom_init(void)
{
int i, *argv, *envp; /* passed as 32 bit ptrs */
......@@ -188,11 +195,10 @@ void __init prom_init(void)
prom_infop = (struct psb_info *)(long)(int)fw_arg3;
nlm_prom_info = *prom_infop;
nlm_pic_base = nlm_mmio_base(NETLOGIC_IO_PIC_OFFSET);
nlm_init_node();
nlm_early_serial_setup();
build_arcs_cmdline(argv);
nlm_common_ebase = read_c0_ebase() & (~((1 << 12) - 1));
prom_add_memory();
#ifdef CONFIG_SMP
......
......@@ -33,6 +33,7 @@
*/
#include <linux/init.h>
#include <linux/delay.h>
#include <linux/threads.h>
#include <asm/asm.h>
......@@ -50,18 +51,34 @@
int __cpuinit xlr_wakeup_secondary_cpus(void)
{
unsigned int i, boot_cpu;
struct nlm_soc_info *nodep;
unsigned int i, j, boot_cpu;
/*
* In case of RMI boot, hit with NMI to get the cores
* from bootloader to linux code.
*/
nodep = nlm_get_node(0);
boot_cpu = hard_smp_processor_id();
nlm_set_nmi_handler(nlm_rmiboot_preboot);
for (i = 0; i < NR_CPUS; i++) {
if (i == boot_cpu || !cpumask_test_cpu(i, &nlm_cpumask))
continue;
nlm_pic_send_ipi(nlm_pic_base, i, 1, 1); /* send NMI */
nlm_pic_send_ipi(nodep->picbase, i, 1, 1); /* send NMI */
}
/* Fill up the coremask early */
nodep->coremask = 1;
for (i = 1; i < NLM_CORES_PER_NODE; i++) {
for (j = 1000000; j > 0; j--) {
if (nlm_cpu_ready[i * NLM_THREADS_PER_CORE])
break;
udelay(10);
}
if (j != 0)
nodep->coremask |= (1u << i);
else
pr_err("Failed to wakeup core %d\n", i);
}
return 0;
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册