linux/arch/sparc/kernel/sun4d_smp.c

413 lines
9.5 KiB
C
Raw Normal View History

/* Sparc SS1000/SC2000 SMP support.
*
* Copyright (C) 1998 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
*
* Based on sun4m's smp.c, which is:
* Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
*/
#include <linux/clockchips.h>
#include <linux/interrupt.h>
#include <linux/profile.h>
#include <linux/delay.h>
sched/headers: Move task->mm handling methods to <linux/sched/mm.h> Move the following task->mm helper APIs into a new header file, <linux/sched/mm.h>, to further reduce the size and complexity of <linux/sched.h>. Here are how the APIs are used in various kernel files: # mm_alloc(): arch/arm/mach-rpc/ecard.c fs/exec.c include/linux/sched/mm.h kernel/fork.c # __mmdrop(): arch/arc/include/asm/mmu_context.h include/linux/sched/mm.h kernel/fork.c # mmdrop(): arch/arm/mach-rpc/ecard.c arch/m68k/sun3/mmu_emu.c arch/x86/mm/tlb.c drivers/gpu/drm/amd/amdkfd/kfd_process.c drivers/gpu/drm/i915/i915_gem_userptr.c drivers/infiniband/hw/hfi1/file_ops.c drivers/vfio/vfio_iommu_spapr_tce.c fs/exec.c fs/proc/base.c fs/proc/task_mmu.c fs/proc/task_nommu.c fs/userfaultfd.c include/linux/mmu_notifier.h include/linux/sched/mm.h kernel/fork.c kernel/futex.c kernel/sched/core.c mm/khugepaged.c mm/ksm.c mm/mmu_context.c mm/mmu_notifier.c mm/oom_kill.c virt/kvm/kvm_main.c # mmdrop_async_fn(): include/linux/sched/mm.h # mmdrop_async(): include/linux/sched/mm.h kernel/fork.c # mmget_not_zero(): fs/userfaultfd.c include/linux/sched/mm.h mm/oom_kill.c # mmput(): arch/arc/include/asm/mmu_context.h arch/arc/kernel/troubleshoot.c arch/frv/mm/mmu-context.c arch/powerpc/platforms/cell/spufs/context.c arch/sparc/include/asm/mmu_context_32.h drivers/android/binder.c drivers/gpu/drm/etnaviv/etnaviv_gem.c drivers/gpu/drm/i915/i915_gem_userptr.c drivers/infiniband/core/umem.c drivers/infiniband/core/umem_odp.c drivers/infiniband/core/uverbs_main.c drivers/infiniband/hw/mlx4/main.c drivers/infiniband/hw/mlx5/main.c drivers/infiniband/hw/usnic/usnic_uiom.c drivers/iommu/amd_iommu_v2.c drivers/iommu/intel-svm.c drivers/lguest/lguest_user.c drivers/misc/cxl/fault.c drivers/misc/mic/scif/scif_rma.c drivers/oprofile/buffer_sync.c drivers/vfio/vfio_iommu_type1.c drivers/vhost/vhost.c drivers/xen/gntdev.c fs/exec.c fs/proc/array.c fs/proc/base.c fs/proc/task_mmu.c fs/proc/task_nommu.c fs/userfaultfd.c include/linux/sched/mm.h kernel/cpuset.c kernel/events/core.c kernel/events/uprobes.c kernel/exit.c kernel/fork.c kernel/ptrace.c kernel/sys.c kernel/trace/trace_output.c kernel/tsacct.c mm/memcontrol.c mm/memory.c mm/mempolicy.c mm/migrate.c mm/mmu_notifier.c mm/nommu.c mm/oom_kill.c mm/process_vm_access.c mm/rmap.c mm/swapfile.c mm/util.c virt/kvm/async_pf.c # mmput_async(): include/linux/sched/mm.h kernel/fork.c mm/oom_kill.c # get_task_mm(): arch/arc/kernel/troubleshoot.c arch/powerpc/platforms/cell/spufs/context.c drivers/android/binder.c drivers/gpu/drm/etnaviv/etnaviv_gem.c drivers/infiniband/core/umem.c drivers/infiniband/core/umem_odp.c drivers/infiniband/hw/mlx4/main.c drivers/infiniband/hw/mlx5/main.c drivers/infiniband/hw/usnic/usnic_uiom.c drivers/iommu/amd_iommu_v2.c drivers/iommu/intel-svm.c drivers/lguest/lguest_user.c drivers/misc/cxl/fault.c drivers/misc/mic/scif/scif_rma.c drivers/oprofile/buffer_sync.c drivers/vfio/vfio_iommu_type1.c drivers/vhost/vhost.c drivers/xen/gntdev.c fs/proc/array.c fs/proc/base.c fs/proc/task_mmu.c include/linux/sched/mm.h kernel/cpuset.c kernel/events/core.c kernel/exit.c kernel/fork.c kernel/ptrace.c kernel/sys.c kernel/trace/trace_output.c kernel/tsacct.c mm/memcontrol.c mm/memory.c mm/mempolicy.c mm/migrate.c mm/mmu_notifier.c mm/nommu.c mm/util.c # mm_access(): fs/proc/base.c include/linux/sched/mm.h kernel/fork.c mm/process_vm_access.c # mm_release(): arch/arc/include/asm/mmu_context.h fs/exec.c include/linux/sched/mm.h include/uapi/linux/sched.h kernel/exit.c kernel/fork.c Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: linux-kernel@vger.kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-02-02 02:08:20 +08:00
#include <linux/sched/mm.h>
#include <linux/cpu.h>
#include <asm/cacheflush.h>
#include <asm/switch_to.h>
#include <asm/tlbflush.h>
#include <asm/timer.h>
#include <asm/oplib.h>
#include <asm/sbi.h>
#include <asm/mmu.h>
#include "kernel.h"
#include "irq.h"
#define IRQ_CROSS_CALL 15
static volatile int smp_processors_ready;
static int smp_highest_cpu;
static inline unsigned long sun4d_swap(volatile unsigned long *ptr, unsigned long val)
{
__asm__ __volatile__("swap [%1], %0\n\t" :
"=&r" (val), "=&r" (ptr) :
"0" (val), "1" (ptr));
return val;
}
static void smp4d_ipi_init(void);
static unsigned char cpu_leds[32];
static inline void show_leds(int cpuid)
{
cpuid &= 0x1e;
__asm__ __volatile__ ("stba %0, [%1] %2" : :
"r" ((cpu_leds[cpuid] << 4) | cpu_leds[cpuid+1]),
"r" (ECSR_BASE(cpuid) | BB_LEDS),
"i" (ASI_M_CTL));
}
sparc: delete __cpuinit/__CPUINIT usage from all users The __cpuinit type of throwaway sections might have made sense some time ago when RAM was more constrained, but now the savings do not offset the cost and complications. For example, the fix in commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time") is a good example of the nasty type of bugs that can be created with improper use of the various __init prefixes. After a discussion on LKML[1] it was decided that cpuinit should go the way of devinit and be phased out. Once all the users are gone, we can then finally remove the macros themselves from linux/init.h. Note that some harmless section mismatch warnings may result, since notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c) are flagged as __cpuinit -- so if we remove the __cpuinit from arch specific callers, we will also get section mismatch warnings. As an intermediate step, we intend to turn the linux/init.h cpuinit content into no-ops as early as possible, since that will get rid of these warnings. In any case, they are temporary and harmless. This removes all the arch/sparc uses of the __cpuinit macros from C files and removes __CPUINIT from assembly files. Note that even though arch/sparc/kernel/trampoline_64.S has instances of ".previous" in it, they are all paired off against explicit ".section" directives, and not implicitly paired with __CPUINIT (unlike mips and arm were). [1] https://lkml.org/lkml/2013/5/20/589 Cc: "David S. Miller" <davem@davemloft.net> Cc: sparclinux@vger.kernel.org Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2013-06-18 03:43:14 +08:00
void sun4d_cpu_pre_starting(void *arg)
{
int cpuid = hard_smp_processor_id();
/* Show we are alive */
cpu_leds[cpuid] = 0x6;
show_leds(cpuid);
/* Enable level15 interrupt, disable level14 interrupt for now */
cc_set_imsk((cc_get_imsk() & ~0x8000) | 0x4000);
}
sparc: delete __cpuinit/__CPUINIT usage from all users The __cpuinit type of throwaway sections might have made sense some time ago when RAM was more constrained, but now the savings do not offset the cost and complications. For example, the fix in commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time") is a good example of the nasty type of bugs that can be created with improper use of the various __init prefixes. After a discussion on LKML[1] it was decided that cpuinit should go the way of devinit and be phased out. Once all the users are gone, we can then finally remove the macros themselves from linux/init.h. Note that some harmless section mismatch warnings may result, since notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c) are flagged as __cpuinit -- so if we remove the __cpuinit from arch specific callers, we will also get section mismatch warnings. As an intermediate step, we intend to turn the linux/init.h cpuinit content into no-ops as early as possible, since that will get rid of these warnings. In any case, they are temporary and harmless. This removes all the arch/sparc uses of the __cpuinit macros from C files and removes __CPUINIT from assembly files. Note that even though arch/sparc/kernel/trampoline_64.S has instances of ".previous" in it, they are all paired off against explicit ".section" directives, and not implicitly paired with __CPUINIT (unlike mips and arm were). [1] https://lkml.org/lkml/2013/5/20/589 Cc: "David S. Miller" <davem@davemloft.net> Cc: sparclinux@vger.kernel.org Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2013-06-18 03:43:14 +08:00
void sun4d_cpu_pre_online(void *arg)
{
unsigned long flags;
int cpuid;
cpuid = hard_smp_processor_id();
/* Unblock the master CPU _only_ when the scheduler state
* of all secondary CPUs will be up-to-date, so after
* the SMP initialization the master will be just allowed
* to call the scheduler code.
*/
sun4d_swap((unsigned long *)&cpu_callin_map[cpuid], 1);
local_ops->cache_all();
local_ops->tlb_all();
while ((unsigned long)current_set[cpuid] < PAGE_OFFSET)
barrier();
while (current_set[cpuid]->cpu != cpuid)
barrier();
/* Fix idle thread fields. */
__asm__ __volatile__("ld [%0], %%g6\n\t"
: : "r" (&current_set[cpuid])
: "memory" /* paranoid */);
cpu_leds[cpuid] = 0x9;
show_leds(cpuid);
/* Attach to the address space of init_task. */
mmgrab(&init_mm);
current->active_mm = &init_mm;
local_ops->cache_all();
local_ops->tlb_all();
while (!cpumask_test_cpu(cpuid, &smp_commenced_mask))
barrier();
spin_lock_irqsave(&sun4d_imsk_lock, flags);
cc_set_imsk(cc_get_imsk() & ~0x4000); /* Allow PIL 14 as well */
spin_unlock_irqrestore(&sun4d_imsk_lock, flags);
}
/*
* Cycle through the processors asking the PROM to start each one.
*/
void __init smp4d_boot_cpus(void)
{
smp4d_ipi_init();
if (boot_cpu_id)
current_set[0] = NULL;
local_ops->cache_all();
}
sparc: delete __cpuinit/__CPUINIT usage from all users The __cpuinit type of throwaway sections might have made sense some time ago when RAM was more constrained, but now the savings do not offset the cost and complications. For example, the fix in commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time") is a good example of the nasty type of bugs that can be created with improper use of the various __init prefixes. After a discussion on LKML[1] it was decided that cpuinit should go the way of devinit and be phased out. Once all the users are gone, we can then finally remove the macros themselves from linux/init.h. Note that some harmless section mismatch warnings may result, since notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c) are flagged as __cpuinit -- so if we remove the __cpuinit from arch specific callers, we will also get section mismatch warnings. As an intermediate step, we intend to turn the linux/init.h cpuinit content into no-ops as early as possible, since that will get rid of these warnings. In any case, they are temporary and harmless. This removes all the arch/sparc uses of the __cpuinit macros from C files and removes __CPUINIT from assembly files. Note that even though arch/sparc/kernel/trampoline_64.S has instances of ".previous" in it, they are all paired off against explicit ".section" directives, and not implicitly paired with __CPUINIT (unlike mips and arm were). [1] https://lkml.org/lkml/2013/5/20/589 Cc: "David S. Miller" <davem@davemloft.net> Cc: sparclinux@vger.kernel.org Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2013-06-18 03:43:14 +08:00
int smp4d_boot_one_cpu(int i, struct task_struct *idle)
{
unsigned long *entry = &sun4d_cpu_startup;
int timeout;
int cpu_node;
cpu_find_by_instance(i, &cpu_node, NULL);
current_set[i] = task_thread_info(idle);
/*
* Initialize the contexts table
* Since the call to prom_startcpu() trashes the structure,
* we need to re-initialize it for each cpu
*/
smp_penguin_ctable.which_io = 0;
smp_penguin_ctable.phys_addr = (unsigned int) srmmu_ctx_table_phys;
smp_penguin_ctable.reg_size = 0;
/* whirrr, whirrr, whirrrrrrrrr... */
printk(KERN_INFO "Starting CPU %d at %p\n", i, entry);
local_ops->cache_all();
prom_startcpu(cpu_node,
&smp_penguin_ctable, 0, (char *)entry);
printk(KERN_INFO "prom_startcpu returned :)\n");
/* wheee... it's going... */
for (timeout = 0; timeout < 10000; timeout++) {
if (cpu_callin_map[i])
break;
udelay(200);
}
if (!(cpu_callin_map[i])) {
printk(KERN_ERR "Processor %d is stuck.\n", i);
return -ENODEV;
}
local_ops->cache_all();
return 0;
}
void __init smp4d_smp_done(void)
{
int i, first;
int *prev;
/* setup cpu list for irq rotation */
first = 0;
prev = &first;
for_each_online_cpu(i) {
*prev = i;
prev = &cpu_data(i).next;
}
*prev = first;
local_ops->cache_all();
/* Ok, they are spinning and ready to go. */
smp_processors_ready = 1;
sun4d_distribute_irqs();
}
/* Memory structure giving interrupt handler information about IPI generated */
struct sun4d_ipi_work {
int single;
int msk;
int resched;
};
static DEFINE_PER_CPU_SHARED_ALIGNED(struct sun4d_ipi_work, sun4d_ipi_work);
/* Initialize IPIs on the SUN4D SMP machine */
static void __init smp4d_ipi_init(void)
{
int cpu;
struct sun4d_ipi_work *work;
printk(KERN_INFO "smp4d: setup IPI at IRQ %d\n", SUN4D_IPI_IRQ);
for_each_possible_cpu(cpu) {
work = &per_cpu(sun4d_ipi_work, cpu);
work->single = work->msk = work->resched = 0;
}
}
void sun4d_ipi_interrupt(void)
{
sparc: Replace __get_cpu_var uses __get_cpu_var() is used for multiple purposes in the kernel source. One of them is address calculation via the form &__get_cpu_var(x). This calculates the address for the instance of the percpu variable of the current processor based on an offset. Other use cases are for storing and retrieving data from the current processors percpu area. __get_cpu_var() can be used as an lvalue when writing data or on the right side of an assignment. __get_cpu_var() is defined as : #define __get_cpu_var(var) (*this_cpu_ptr(&(var))) __get_cpu_var() always only does an address determination. However, store and retrieve operations could use a segment prefix (or global register on other platforms) to avoid the address calculation. this_cpu_write() and this_cpu_read() can directly take an offset into a percpu area and use optimized assembly code to read and write per cpu variables. This patch converts __get_cpu_var into either an explicit address calculation using this_cpu_ptr() or into a use of this_cpu operations that use the offset. Thereby address calculations are avoided and less registers are used when code is generated. At the end of the patch set all uses of __get_cpu_var have been removed so the macro is removed too. The patch set includes passes over all arches as well. Once these operations are used throughout then specialized macros can be defined in non -x86 arches as well in order to optimize per cpu access by f.e. using a global register that may be set to the per cpu base. Transformations done to __get_cpu_var() 1. Determine the address of the percpu instance of the current processor. DEFINE_PER_CPU(int, y); int *x = &__get_cpu_var(y); Converts to int *x = this_cpu_ptr(&y); 2. Same as #1 but this time an array structure is involved. DEFINE_PER_CPU(int, y[20]); int *x = __get_cpu_var(y); Converts to int *x = this_cpu_ptr(y); 3. Retrieve the content of the current processors instance of a per cpu variable. DEFINE_PER_CPU(int, y); int x = __get_cpu_var(y) Converts to int x = __this_cpu_read(y); 4. Retrieve the content of a percpu struct DEFINE_PER_CPU(struct mystruct, y); struct mystruct x = __get_cpu_var(y); Converts to memcpy(&x, this_cpu_ptr(&y), sizeof(x)); 5. Assignment to a per cpu variable DEFINE_PER_CPU(int, y) __get_cpu_var(y) = x; Converts to __this_cpu_write(y, x); 6. Increment/Decrement etc of a per cpu variable DEFINE_PER_CPU(int, y); __get_cpu_var(y)++ Converts to __this_cpu_inc(y) Cc: sparclinux@vger.kernel.org Acked-by: David S. Miller <davem@davemloft.net> Signed-off-by: Christoph Lameter <cl@linux.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-08-18 01:30:54 +08:00
struct sun4d_ipi_work *work = this_cpu_ptr(&sun4d_ipi_work);
if (work->single) {
work->single = 0;
smp_call_function_single_interrupt();
}
if (work->msk) {
work->msk = 0;
smp_call_function_interrupt();
}
if (work->resched) {
work->resched = 0;
smp_resched_interrupt();
}
}
/* +-------+-------------+-----------+------------------------------------+
* | bcast | devid | sid | levels mask |
* +-------+-------------+-----------+------------------------------------+
* 31 30 23 22 15 14 0
*/
#define IGEN_MESSAGE(bcast, devid, sid, levels) \
(((bcast) << 31) | ((devid) << 23) | ((sid) << 15) | (levels))
static void sun4d_send_ipi(int cpu, int level)
{
cc_set_igen(IGEN_MESSAGE(0, cpu << 3, 6 + ((level >> 1) & 7), 1 << (level - 1)));
}
static void sun4d_ipi_single(int cpu)
{
struct sun4d_ipi_work *work = &per_cpu(sun4d_ipi_work, cpu);
/* Mark work */
work->single = 1;
/* Generate IRQ on the CPU */
sun4d_send_ipi(cpu, SUN4D_IPI_IRQ);
}
static void sun4d_ipi_mask_one(int cpu)
{
struct sun4d_ipi_work *work = &per_cpu(sun4d_ipi_work, cpu);
/* Mark work */
work->msk = 1;
/* Generate IRQ on the CPU */
sun4d_send_ipi(cpu, SUN4D_IPI_IRQ);
}
static void sun4d_ipi_resched(int cpu)
{
struct sun4d_ipi_work *work = &per_cpu(sun4d_ipi_work, cpu);
/* Mark work */
work->resched = 1;
/* Generate IRQ on the CPU (any IRQ will cause resched) */
sun4d_send_ipi(cpu, SUN4D_IPI_IRQ);
}
static struct smp_funcall {
smpfunc_t func;
unsigned long arg1;
unsigned long arg2;
unsigned long arg3;
unsigned long arg4;
unsigned long arg5;
unsigned char processors_in[NR_CPUS]; /* Set when ipi entered. */
unsigned char processors_out[NR_CPUS]; /* Set when ipi exited. */
} ccall_info __attribute__((aligned(8)));
static DEFINE_SPINLOCK(cross_call_lock);
/* Cross calls must be serialized, at least currently. */
static void sun4d_cross_call(smpfunc_t func, cpumask_t mask, unsigned long arg1,
unsigned long arg2, unsigned long arg3,
unsigned long arg4)
{
if (smp_processors_ready) {
register int high = smp_highest_cpu;
unsigned long flags;
spin_lock_irqsave(&cross_call_lock, flags);
{
/*
* If you make changes here, make sure
* gcc generates proper code...
*/
register smpfunc_t f asm("i0") = func;
register unsigned long a1 asm("i1") = arg1;
register unsigned long a2 asm("i2") = arg2;
register unsigned long a3 asm("i3") = arg3;
register unsigned long a4 asm("i4") = arg4;
register unsigned long a5 asm("i5") = 0;
__asm__ __volatile__(
"std %0, [%6]\n\t"
"std %2, [%6 + 8]\n\t"
"std %4, [%6 + 16]\n\t" : :
"r"(f), "r"(a1), "r"(a2), "r"(a3), "r"(a4), "r"(a5),
"r" (&ccall_info.func));
}
/* Init receive/complete mapping, plus fire the IPI's off. */
{
register int i;
cpumask_clear_cpu(smp_processor_id(), &mask);
cpumask_and(&mask, cpu_online_mask, &mask);
for (i = 0; i <= high; i++) {
if (cpumask_test_cpu(i, &mask)) {
ccall_info.processors_in[i] = 0;
ccall_info.processors_out[i] = 0;
sun4d_send_ipi(i, IRQ_CROSS_CALL);
}
}
}
{
register int i;
i = 0;
do {
if (!cpumask_test_cpu(i, &mask))
continue;
while (!ccall_info.processors_in[i])
barrier();
} while (++i <= high);
i = 0;
do {
if (!cpumask_test_cpu(i, &mask))
continue;
while (!ccall_info.processors_out[i])
barrier();
} while (++i <= high);
}
spin_unlock_irqrestore(&cross_call_lock, flags);
}
}
/* Running cross calls. */
void smp4d_cross_call_irq(void)
{
int i = hard_smp_processor_id();
ccall_info.processors_in[i] = 1;
ccall_info.func(ccall_info.arg1, ccall_info.arg2, ccall_info.arg3,
ccall_info.arg4, ccall_info.arg5);
ccall_info.processors_out[i] = 1;
}
void smp4d_percpu_timer_interrupt(struct pt_regs *regs)
{
struct pt_regs *old_regs;
int cpu = hard_smp_processor_id();
struct clock_event_device *ce;
static int cpu_tick[NR_CPUS];
static char led_mask[] = { 0xe, 0xd, 0xb, 0x7, 0xb, 0xd };
old_regs = set_irq_regs(regs);
bw_get_prof_limit(cpu);
bw_clear_intr_mask(0, 1); /* INTR_TABLE[0] & 1 is Profile IRQ */
cpu_tick[cpu]++;
if (!(cpu_tick[cpu] & 15)) {
if (cpu_tick[cpu] == 0x60)
cpu_tick[cpu] = 0;
cpu_leds[cpu] = led_mask[cpu_tick[cpu] >> 4];
show_leds(cpu);
}
ce = &per_cpu(sparc32_clockevent, cpu);
irq_enter();
ce->event_handler(ce);
irq_exit();
set_irq_regs(old_regs);
}
static const struct sparc32_ipi_ops sun4d_ipi_ops = {
.cross_call = sun4d_cross_call,
.resched = sun4d_ipi_resched,
.single = sun4d_ipi_single,
.mask_one = sun4d_ipi_mask_one,
};
void __init sun4d_init_smp(void)
{
int i;
/* Patch ipi15 trap table */
t_nmi[1] = t_nmi[1] + (linux_trap_ipi15_sun4d - linux_trap_ipi15_sun4m);
sparc32_ipi_ops = &sun4d_ipi_ops;
for (i = 0; i < NR_CPUS; i++) {
ccall_info.processors_in[i] = 1;
ccall_info.processors_out[i] = 1;
}
}