2007-07-17 19:37:06 -06:00
|
|
|
/*
|
|
|
|
* Xen SMP support
|
|
|
|
*
|
|
|
|
* This file implements the Xen versions of smp_ops. SMP under Xen is
|
|
|
|
* very straightforward. Bringing a CPU up is simply a matter of
|
|
|
|
* loading its initial context and setting it running.
|
|
|
|
*
|
|
|
|
* IPIs are handled through the Xen event mechanism.
|
|
|
|
*
|
|
|
|
* Because virtual CPUs can be scheduled onto any real CPU, there's no
|
|
|
|
* useful topology information for the kernel to make use of. As a
|
|
|
|
* result, all CPUs are treated as if they're single-core and
|
|
|
|
* single-threaded.
|
|
|
|
*/
|
|
|
|
#include <linux/sched.h>
|
|
|
|
#include <linux/err.h>
|
|
|
|
#include <linux/smp.h>
|
|
|
|
|
|
|
|
#include <asm/paravirt.h>
|
|
|
|
#include <asm/desc.h>
|
|
|
|
#include <asm/pgtable.h>
|
|
|
|
#include <asm/cpu.h>
|
|
|
|
|
|
|
|
#include <xen/interface/xen.h>
|
|
|
|
#include <xen/interface/vcpu.h>
|
|
|
|
|
|
|
|
#include <asm/xen/interface.h>
|
|
|
|
#include <asm/xen/hypercall.h>
|
|
|
|
|
|
|
|
#include <xen/page.h>
|
|
|
|
#include <xen/events.h>
|
|
|
|
|
|
|
|
#include "xen-ops.h"
|
|
|
|
#include "mmu.h"
|
|
|
|
|
2008-05-26 16:31:27 -06:00
|
|
|
cpumask_t xen_cpu_initialized_map;
|
2007-07-17 19:37:06 -06:00
|
|
|
|
2008-06-26 03:21:54 -06:00
|
|
|
static DEFINE_PER_CPU(int, resched_irq);
|
|
|
|
static DEFINE_PER_CPU(int, callfunc_irq);
|
|
|
|
static DEFINE_PER_CPU(int, callfuncsingle_irq);
|
|
|
|
static DEFINE_PER_CPU(int, debug_irq) = -1;
|
2007-07-17 19:37:06 -06:00
|
|
|
|
|
|
|
static irqreturn_t xen_call_function_interrupt(int irq, void *dev_id);
|
2008-06-26 03:21:54 -06:00
|
|
|
static irqreturn_t xen_call_function_single_interrupt(int irq, void *dev_id);
|
2007-07-17 19:37:06 -06:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Reschedule call back. Nothing to do,
|
|
|
|
* all the work is done automatically when
|
|
|
|
* we return from the interrupt.
|
|
|
|
*/
|
|
|
|
static irqreturn_t xen_reschedule_interrupt(int irq, void *dev_id)
|
|
|
|
{
|
2008-05-26 16:31:16 -06:00
|
|
|
#ifdef CONFIG_X86_32
|
|
|
|
__get_cpu_var(irq_stat).irq_resched_count++;
|
|
|
|
#else
|
|
|
|
add_pda(irq_resched_count, 1);
|
|
|
|
#endif
|
|
|
|
|
2007-07-17 19:37:06 -06:00
|
|
|
return IRQ_HANDLED;
|
|
|
|
}
|
|
|
|
|
2008-08-22 04:52:15 -06:00
|
|
|
static __cpuinit void cpu_bringup(void)
|
2007-07-17 19:37:06 -06:00
|
|
|
{
|
|
|
|
int cpu = smp_processor_id();
|
|
|
|
|
|
|
|
cpu_init();
|
2008-08-22 04:52:15 -06:00
|
|
|
touch_softlockup_watchdog();
|
2008-07-08 16:06:43 -06:00
|
|
|
preempt_disable();
|
|
|
|
|
2008-03-17 17:37:17 -06:00
|
|
|
xen_enable_sysenter();
|
2008-07-08 16:07:14 -06:00
|
|
|
xen_enable_syscall();
|
2007-07-17 19:37:06 -06:00
|
|
|
|
2008-07-08 16:06:43 -06:00
|
|
|
cpu = smp_processor_id();
|
|
|
|
smp_store_cpu_info(cpu);
|
|
|
|
cpu_data(cpu).x86_max_cores = 1;
|
|
|
|
set_cpu_sibling_map(cpu);
|
2007-07-17 19:37:06 -06:00
|
|
|
|
|
|
|
xen_setup_cpu_clockevents();
|
|
|
|
|
2008-07-08 16:06:43 -06:00
|
|
|
cpu_set(cpu, cpu_online_map);
|
|
|
|
x86_write_percpu(cpu_state, CPU_ONLINE);
|
|
|
|
wmb();
|
|
|
|
|
2007-07-17 19:37:06 -06:00
|
|
|
/* We can take interrupts now: we're officially "up". */
|
|
|
|
local_irq_enable();
|
|
|
|
|
|
|
|
wmb(); /* make sure everything is out */
|
2008-08-22 04:52:15 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
static __cpuinit void cpu_bringup_and_idle(void)
|
|
|
|
{
|
|
|
|
cpu_bringup();
|
2007-07-17 19:37:06 -06:00
|
|
|
cpu_idle();
|
|
|
|
}
|
|
|
|
|
|
|
|
static int xen_smp_intr_init(unsigned int cpu)
|
|
|
|
{
|
|
|
|
int rc;
|
2008-03-17 17:37:18 -06:00
|
|
|
const char *resched_name, *callfunc_name, *debug_name;
|
2007-07-17 19:37:06 -06:00
|
|
|
|
|
|
|
resched_name = kasprintf(GFP_KERNEL, "resched%d", cpu);
|
|
|
|
rc = bind_ipi_to_irqhandler(XEN_RESCHEDULE_VECTOR,
|
|
|
|
cpu,
|
|
|
|
xen_reschedule_interrupt,
|
|
|
|
IRQF_DISABLED|IRQF_PERCPU|IRQF_NOBALANCING,
|
|
|
|
resched_name,
|
|
|
|
NULL);
|
|
|
|
if (rc < 0)
|
|
|
|
goto fail;
|
|
|
|
per_cpu(resched_irq, cpu) = rc;
|
|
|
|
|
|
|
|
callfunc_name = kasprintf(GFP_KERNEL, "callfunc%d", cpu);
|
|
|
|
rc = bind_ipi_to_irqhandler(XEN_CALL_FUNCTION_VECTOR,
|
|
|
|
cpu,
|
|
|
|
xen_call_function_interrupt,
|
|
|
|
IRQF_DISABLED|IRQF_PERCPU|IRQF_NOBALANCING,
|
|
|
|
callfunc_name,
|
|
|
|
NULL);
|
|
|
|
if (rc < 0)
|
|
|
|
goto fail;
|
|
|
|
per_cpu(callfunc_irq, cpu) = rc;
|
|
|
|
|
2008-03-17 17:37:18 -06:00
|
|
|
debug_name = kasprintf(GFP_KERNEL, "debug%d", cpu);
|
|
|
|
rc = bind_virq_to_irqhandler(VIRQ_DEBUG, cpu, xen_debug_interrupt,
|
|
|
|
IRQF_DISABLED | IRQF_PERCPU | IRQF_NOBALANCING,
|
|
|
|
debug_name, NULL);
|
|
|
|
if (rc < 0)
|
|
|
|
goto fail;
|
|
|
|
per_cpu(debug_irq, cpu) = rc;
|
|
|
|
|
2008-06-26 03:21:54 -06:00
|
|
|
callfunc_name = kasprintf(GFP_KERNEL, "callfuncsingle%d", cpu);
|
|
|
|
rc = bind_ipi_to_irqhandler(XEN_CALL_FUNCTION_SINGLE_VECTOR,
|
|
|
|
cpu,
|
|
|
|
xen_call_function_single_interrupt,
|
|
|
|
IRQF_DISABLED|IRQF_PERCPU|IRQF_NOBALANCING,
|
|
|
|
callfunc_name,
|
|
|
|
NULL);
|
|
|
|
if (rc < 0)
|
|
|
|
goto fail;
|
|
|
|
per_cpu(callfuncsingle_irq, cpu) = rc;
|
|
|
|
|
2007-07-17 19:37:06 -06:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
fail:
|
|
|
|
if (per_cpu(resched_irq, cpu) >= 0)
|
|
|
|
unbind_from_irqhandler(per_cpu(resched_irq, cpu), NULL);
|
|
|
|
if (per_cpu(callfunc_irq, cpu) >= 0)
|
|
|
|
unbind_from_irqhandler(per_cpu(callfunc_irq, cpu), NULL);
|
2008-03-17 17:37:18 -06:00
|
|
|
if (per_cpu(debug_irq, cpu) >= 0)
|
|
|
|
unbind_from_irqhandler(per_cpu(debug_irq, cpu), NULL);
|
2008-06-26 03:21:54 -06:00
|
|
|
if (per_cpu(callfuncsingle_irq, cpu) >= 0)
|
|
|
|
unbind_from_irqhandler(per_cpu(callfuncsingle_irq, cpu), NULL);
|
|
|
|
|
2007-07-17 19:37:06 -06:00
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
2008-07-08 16:06:43 -06:00
|
|
|
static void __init xen_fill_possible_map(void)
|
2007-07-17 19:37:06 -06:00
|
|
|
{
|
|
|
|
int i, rc;
|
|
|
|
|
|
|
|
for (i = 0; i < NR_CPUS; i++) {
|
|
|
|
rc = HYPERVISOR_vcpu_op(VCPUOP_is_up, i, NULL);
|
2008-07-08 16:06:56 -06:00
|
|
|
if (rc >= 0) {
|
|
|
|
num_processors++;
|
2007-07-17 19:37:06 -06:00
|
|
|
cpu_set(i, cpu_possible_map);
|
2008-07-08 16:06:56 -06:00
|
|
|
}
|
2007-07-17 19:37:06 -06:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-07-08 16:06:41 -06:00
|
|
|
static void __init xen_smp_prepare_boot_cpu(void)
|
2007-07-17 19:37:06 -06:00
|
|
|
{
|
|
|
|
BUG_ON(smp_processor_id() != 0);
|
|
|
|
native_smp_prepare_boot_cpu();
|
|
|
|
|
|
|
|
/* We've switched to the "real" per-cpu gdt, so make sure the
|
|
|
|
old memory can be recycled */
|
2008-07-08 16:06:43 -06:00
|
|
|
make_lowmem_page_readwrite(&per_cpu_var(gdt_page));
|
2007-07-17 19:37:07 -06:00
|
|
|
|
|
|
|
xen_setup_vcpu_info_placement();
|
2007-07-17 19:37:06 -06:00
|
|
|
}
|
|
|
|
|
2008-07-08 16:06:41 -06:00
|
|
|
static void __init xen_smp_prepare_cpus(unsigned int max_cpus)
|
2007-07-17 19:37:06 -06:00
|
|
|
{
|
|
|
|
unsigned cpu;
|
|
|
|
|
xen: implement Xen-specific spinlocks
The standard ticket spinlocks are very expensive in a virtual
environment, because their performance depends on Xen's scheduler
giving vcpus time in the order that they're supposed to take the
spinlock.
This implements a Xen-specific spinlock, which should be much more
efficient.
The fast-path is essentially the old Linux-x86 locks, using a single
lock byte. The locker decrements the byte; if the result is 0, then
they have the lock. If the lock is negative, then locker must spin
until the lock is positive again.
When there's contention, the locker spin for 2^16[*] iterations waiting
to get the lock. If it fails to get the lock in that time, it adds
itself to the contention count in the lock and blocks on a per-cpu
event channel.
When unlocking the spinlock, the locker looks to see if there's anyone
blocked waiting for the lock by checking for a non-zero waiter count.
If there's a waiter, it traverses the per-cpu "lock_spinners"
variable, which contains which lock each CPU is waiting on. It picks
one CPU waiting on the lock and sends it an event to wake it up.
This allows efficient fast-path spinlock operation, while allowing
spinning vcpus to give up their processor time while waiting for a
contended lock.
[*] 2^16 iterations is threshold at which 98% locks have been taken
according to Thomas Friebel's Xen Summit talk "Preventing Guests from
Spinning Around". Therefore, we'd expect the lock and unlock slow
paths will only be entered 2% of the time.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Christoph Lameter <clameter@linux-foundation.org>
Cc: Petr Tesarik <ptesarik@suse.cz>
Cc: Virtualization <virtualization@lists.linux-foundation.org>
Cc: Xen devel <xen-devel@lists.xensource.com>
Cc: Thomas Friebel <thomas.friebel@amd.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-07-07 13:07:53 -06:00
|
|
|
xen_init_lock_cpu(0);
|
|
|
|
|
2007-07-17 19:37:06 -06:00
|
|
|
smp_store_cpu_info(0);
|
2008-07-08 16:06:43 -06:00
|
|
|
cpu_data(0).x86_max_cores = 1;
|
2007-07-17 19:37:06 -06:00
|
|
|
set_cpu_sibling_map(0);
|
|
|
|
|
|
|
|
if (xen_smp_intr_init(0))
|
|
|
|
BUG();
|
|
|
|
|
2008-03-27 11:06:03 -06:00
|
|
|
xen_cpu_initialized_map = cpumask_of_cpu(0);
|
2007-07-17 19:37:06 -06:00
|
|
|
|
|
|
|
/* Restrict the possible_map according to max_cpus. */
|
|
|
|
while ((num_possible_cpus() > 1) && (num_possible_cpus() > max_cpus)) {
|
2008-04-19 08:55:17 -06:00
|
|
|
for (cpu = NR_CPUS - 1; !cpu_possible(cpu); cpu--)
|
2007-07-17 19:37:06 -06:00
|
|
|
continue;
|
|
|
|
cpu_clear(cpu, cpu_possible_map);
|
|
|
|
}
|
|
|
|
|
|
|
|
for_each_possible_cpu (cpu) {
|
|
|
|
struct task_struct *idle;
|
|
|
|
|
|
|
|
if (cpu == 0)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
idle = fork_idle(cpu);
|
|
|
|
if (IS_ERR(idle))
|
|
|
|
panic("failed fork for CPU %d", cpu);
|
|
|
|
|
|
|
|
cpu_set(cpu, cpu_present_map);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static __cpuinit int
|
|
|
|
cpu_initialize_context(unsigned int cpu, struct task_struct *idle)
|
|
|
|
{
|
|
|
|
struct vcpu_guest_context *ctxt;
|
2008-07-08 16:06:43 -06:00
|
|
|
struct desc_struct *gdt;
|
2007-07-17 19:37:06 -06:00
|
|
|
|
2008-03-27 11:06:03 -06:00
|
|
|
if (cpu_test_and_set(cpu, xen_cpu_initialized_map))
|
2007-07-17 19:37:06 -06:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
|
|
|
|
if (ctxt == NULL)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2008-07-08 16:06:43 -06:00
|
|
|
gdt = get_cpu_gdt_table(cpu);
|
|
|
|
|
2007-07-17 19:37:06 -06:00
|
|
|
ctxt->flags = VGCF_IN_KERNEL;
|
|
|
|
ctxt->user_regs.ds = __USER_DS;
|
|
|
|
ctxt->user_regs.es = __USER_DS;
|
|
|
|
ctxt->user_regs.ss = __KERNEL_DS;
|
2008-07-08 16:06:43 -06:00
|
|
|
#ifdef CONFIG_X86_32
|
|
|
|
ctxt->user_regs.fs = __KERNEL_PERCPU;
|
|
|
|
#endif
|
2007-07-17 19:37:06 -06:00
|
|
|
ctxt->user_regs.eip = (unsigned long)cpu_bringup_and_idle;
|
|
|
|
ctxt->user_regs.eflags = 0x1000; /* IOPL_RING1 */
|
|
|
|
|
|
|
|
memset(&ctxt->fpu_ctxt, 0, sizeof(ctxt->fpu_ctxt));
|
|
|
|
|
|
|
|
xen_copy_trap_info(ctxt->trap_ctxt);
|
|
|
|
|
|
|
|
ctxt->ldt_ents = 0;
|
|
|
|
|
2008-07-08 16:06:43 -06:00
|
|
|
BUG_ON((unsigned long)gdt & ~PAGE_MASK);
|
|
|
|
make_lowmem_page_readonly(gdt);
|
2007-07-17 19:37:06 -06:00
|
|
|
|
2008-07-08 16:06:43 -06:00
|
|
|
ctxt->gdt_frames[0] = virt_to_mfn(gdt);
|
|
|
|
ctxt->gdt_ents = GDT_ENTRIES;
|
2007-07-17 19:37:06 -06:00
|
|
|
|
|
|
|
ctxt->user_regs.cs = __KERNEL_CS;
|
2008-01-30 05:31:02 -07:00
|
|
|
ctxt->user_regs.esp = idle->thread.sp0 - sizeof(struct pt_regs);
|
2007-07-17 19:37:06 -06:00
|
|
|
|
|
|
|
ctxt->kernel_ss = __KERNEL_DS;
|
2008-01-30 05:31:02 -07:00
|
|
|
ctxt->kernel_sp = idle->thread.sp0;
|
2007-07-17 19:37:06 -06:00
|
|
|
|
2008-07-08 16:06:43 -06:00
|
|
|
#ifdef CONFIG_X86_32
|
2007-07-17 19:37:06 -06:00
|
|
|
ctxt->event_callback_cs = __KERNEL_CS;
|
|
|
|
ctxt->failsafe_callback_cs = __KERNEL_CS;
|
2008-07-08 16:06:43 -06:00
|
|
|
#endif
|
|
|
|
ctxt->event_callback_eip = (unsigned long)xen_hypervisor_callback;
|
2007-07-17 19:37:06 -06:00
|
|
|
ctxt->failsafe_callback_eip = (unsigned long)xen_failsafe_callback;
|
|
|
|
|
|
|
|
per_cpu(xen_cr3, cpu) = __pa(swapper_pg_dir);
|
|
|
|
ctxt->ctrlreg[3] = xen_pfn_to_cr3(virt_to_mfn(swapper_pg_dir));
|
|
|
|
|
|
|
|
if (HYPERVISOR_vcpu_op(VCPUOP_initialise, cpu, ctxt))
|
|
|
|
BUG();
|
|
|
|
|
|
|
|
kfree(ctxt);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-07-08 16:06:41 -06:00
|
|
|
static int __cpuinit xen_cpu_up(unsigned int cpu)
|
2007-07-17 19:37:06 -06:00
|
|
|
{
|
|
|
|
struct task_struct *idle = idle_task(cpu);
|
|
|
|
int rc;
|
|
|
|
|
2008-07-08 16:06:43 -06:00
|
|
|
#ifdef CONFIG_X86_64
|
|
|
|
/* Allocate node local memory for AP pdas */
|
|
|
|
WARN_ON(cpu == 0);
|
|
|
|
if (cpu > 0) {
|
|
|
|
rc = get_local_pda(cpu);
|
|
|
|
if (rc)
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifdef CONFIG_X86_32
|
2007-07-17 19:37:06 -06:00
|
|
|
init_gdt(cpu);
|
|
|
|
per_cpu(current_task, cpu) = idle;
|
|
|
|
irq_ctx_init(cpu);
|
2008-07-08 16:06:43 -06:00
|
|
|
#else
|
|
|
|
cpu_pda(cpu)->pcurrent = idle;
|
|
|
|
clear_tsk_thread_flag(idle, TIF_FORK);
|
|
|
|
#endif
|
2007-07-17 19:37:06 -06:00
|
|
|
xen_setup_timer(cpu);
|
xen: implement Xen-specific spinlocks
The standard ticket spinlocks are very expensive in a virtual
environment, because their performance depends on Xen's scheduler
giving vcpus time in the order that they're supposed to take the
spinlock.
This implements a Xen-specific spinlock, which should be much more
efficient.
The fast-path is essentially the old Linux-x86 locks, using a single
lock byte. The locker decrements the byte; if the result is 0, then
they have the lock. If the lock is negative, then locker must spin
until the lock is positive again.
When there's contention, the locker spin for 2^16[*] iterations waiting
to get the lock. If it fails to get the lock in that time, it adds
itself to the contention count in the lock and blocks on a per-cpu
event channel.
When unlocking the spinlock, the locker looks to see if there's anyone
blocked waiting for the lock by checking for a non-zero waiter count.
If there's a waiter, it traverses the per-cpu "lock_spinners"
variable, which contains which lock each CPU is waiting on. It picks
one CPU waiting on the lock and sends it an event to wake it up.
This allows efficient fast-path spinlock operation, while allowing
spinning vcpus to give up their processor time while waiting for a
contended lock.
[*] 2^16 iterations is threshold at which 98% locks have been taken
according to Thomas Friebel's Xen Summit talk "Preventing Guests from
Spinning Around". Therefore, we'd expect the lock and unlock slow
paths will only be entered 2% of the time.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Christoph Lameter <clameter@linux-foundation.org>
Cc: Petr Tesarik <ptesarik@suse.cz>
Cc: Virtualization <virtualization@lists.linux-foundation.org>
Cc: Xen devel <xen-devel@lists.xensource.com>
Cc: Thomas Friebel <thomas.friebel@amd.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-07-07 13:07:53 -06:00
|
|
|
xen_init_lock_cpu(cpu);
|
2007-07-17 19:37:06 -06:00
|
|
|
|
2008-07-08 16:06:43 -06:00
|
|
|
per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
|
|
|
|
|
2007-07-17 19:37:06 -06:00
|
|
|
/* make sure interrupts start blocked */
|
|
|
|
per_cpu(xen_vcpu, cpu)->evtchn_upcall_mask = 1;
|
|
|
|
|
|
|
|
rc = cpu_initialize_context(cpu, idle);
|
|
|
|
if (rc)
|
|
|
|
return rc;
|
|
|
|
|
|
|
|
if (num_online_cpus() == 1)
|
|
|
|
alternatives_smp_switch(1);
|
|
|
|
|
|
|
|
rc = xen_smp_intr_init(cpu);
|
|
|
|
if (rc)
|
|
|
|
return rc;
|
|
|
|
|
|
|
|
rc = HYPERVISOR_vcpu_op(VCPUOP_up, cpu, NULL);
|
|
|
|
BUG_ON(rc);
|
|
|
|
|
2008-07-08 16:06:43 -06:00
|
|
|
while(per_cpu(cpu_state, cpu) != CPU_ONLINE) {
|
|
|
|
HYPERVISOR_sched_op(SCHEDOP_yield, 0);
|
|
|
|
barrier();
|
|
|
|
}
|
|
|
|
|
2007-07-17 19:37:06 -06:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-07-08 16:06:41 -06:00
|
|
|
static void xen_smp_cpus_done(unsigned int max_cpus)
|
2007-07-17 19:37:06 -06:00
|
|
|
{
|
|
|
|
}
|
|
|
|
|
2008-09-08 06:43:33 -06:00
|
|
|
#ifdef CONFIG_HOTPLUG_CPU
|
2008-09-08 06:43:34 -06:00
|
|
|
static int xen_cpu_disable(void)
|
2008-08-22 04:52:15 -06:00
|
|
|
{
|
|
|
|
unsigned int cpu = smp_processor_id();
|
|
|
|
if (cpu == 0)
|
|
|
|
return -EBUSY;
|
|
|
|
|
|
|
|
cpu_disable_common();
|
|
|
|
|
|
|
|
load_cr3(swapper_pg_dir);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-09-08 06:43:34 -06:00
|
|
|
static void xen_cpu_die(unsigned int cpu)
|
2008-08-22 04:52:15 -06:00
|
|
|
{
|
|
|
|
while (HYPERVISOR_vcpu_op(VCPUOP_is_up, cpu, NULL)) {
|
|
|
|
current->state = TASK_UNINTERRUPTIBLE;
|
|
|
|
schedule_timeout(HZ/10);
|
|
|
|
}
|
|
|
|
unbind_from_irqhandler(per_cpu(resched_irq, cpu), NULL);
|
|
|
|
unbind_from_irqhandler(per_cpu(callfunc_irq, cpu), NULL);
|
|
|
|
unbind_from_irqhandler(per_cpu(debug_irq, cpu), NULL);
|
|
|
|
unbind_from_irqhandler(per_cpu(callfuncsingle_irq, cpu), NULL);
|
|
|
|
xen_uninit_lock_cpu(cpu);
|
|
|
|
xen_teardown_timer(cpu);
|
|
|
|
|
|
|
|
if (num_online_cpus() == 1)
|
|
|
|
alternatives_smp_switch(0);
|
|
|
|
}
|
|
|
|
|
2008-11-22 10:38:04 -07:00
|
|
|
static void __cpuinit xen_play_dead(void) /* used only with CPU_HOTPLUG */
|
2008-08-22 04:52:15 -06:00
|
|
|
{
|
|
|
|
play_dead_common();
|
|
|
|
HYPERVISOR_vcpu_op(VCPUOP_down, smp_processor_id(), NULL);
|
|
|
|
cpu_bringup();
|
|
|
|
}
|
|
|
|
|
2008-09-08 06:43:33 -06:00
|
|
|
#else /* !CONFIG_HOTPLUG_CPU */
|
2008-09-08 06:43:34 -06:00
|
|
|
static int xen_cpu_disable(void)
|
2008-09-08 06:43:33 -06:00
|
|
|
{
|
|
|
|
return -ENOSYS;
|
|
|
|
}
|
|
|
|
|
2008-09-08 06:43:34 -06:00
|
|
|
static void xen_cpu_die(unsigned int cpu)
|
2008-09-08 06:43:33 -06:00
|
|
|
{
|
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
|
2008-09-08 06:43:34 -06:00
|
|
|
static void xen_play_dead(void)
|
2008-09-08 06:43:33 -06:00
|
|
|
{
|
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif
|
2007-07-17 19:37:06 -06:00
|
|
|
static void stop_self(void *v)
|
|
|
|
{
|
|
|
|
int cpu = smp_processor_id();
|
|
|
|
|
|
|
|
/* make sure we're not pinning something down */
|
|
|
|
load_cr3(swapper_pg_dir);
|
|
|
|
/* should set up a minimal gdt */
|
|
|
|
|
|
|
|
HYPERVISOR_vcpu_op(VCPUOP_down, cpu, NULL);
|
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
|
2008-07-08 16:06:41 -06:00
|
|
|
static void xen_smp_send_stop(void)
|
2007-07-17 19:37:06 -06:00
|
|
|
{
|
2008-06-06 03:18:06 -06:00
|
|
|
smp_call_function(stop_self, NULL, 0);
|
2007-07-17 19:37:06 -06:00
|
|
|
}
|
|
|
|
|
2008-07-08 16:06:41 -06:00
|
|
|
static void xen_smp_send_reschedule(int cpu)
|
2007-07-17 19:37:06 -06:00
|
|
|
{
|
|
|
|
xen_send_IPI_one(cpu, XEN_RESCHEDULE_VECTOR);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void xen_send_IPI_mask(cpumask_t mask, enum ipi_vector vector)
|
|
|
|
{
|
|
|
|
unsigned cpu;
|
|
|
|
|
|
|
|
cpus_and(mask, mask, cpu_online_map);
|
|
|
|
|
2008-05-12 13:21:13 -06:00
|
|
|
for_each_cpu_mask_nr(cpu, mask)
|
2007-07-17 19:37:06 -06:00
|
|
|
xen_send_IPI_one(cpu, vector);
|
|
|
|
}
|
|
|
|
|
2008-07-08 16:06:41 -06:00
|
|
|
static void xen_smp_send_call_function_ipi(cpumask_t mask)
|
2008-06-26 03:21:54 -06:00
|
|
|
{
|
|
|
|
int cpu;
|
|
|
|
|
|
|
|
xen_send_IPI_mask(mask, XEN_CALL_FUNCTION_VECTOR);
|
|
|
|
|
|
|
|
/* Make sure other vcpus get a chance to run if they need to. */
|
2008-07-15 16:29:07 -06:00
|
|
|
for_each_cpu_mask_nr(cpu, mask) {
|
2008-06-26 03:21:54 -06:00
|
|
|
if (xen_vcpu_stolen(cpu)) {
|
|
|
|
HYPERVISOR_sched_op(SCHEDOP_yield, 0);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-07-08 16:06:41 -06:00
|
|
|
static void xen_smp_send_call_function_single_ipi(int cpu)
|
2008-06-26 03:21:54 -06:00
|
|
|
{
|
|
|
|
xen_send_IPI_mask(cpumask_of_cpu(cpu), XEN_CALL_FUNCTION_SINGLE_VECTOR);
|
|
|
|
}
|
|
|
|
|
2007-07-17 19:37:06 -06:00
|
|
|
static irqreturn_t xen_call_function_interrupt(int irq, void *dev_id)
|
|
|
|
{
|
|
|
|
irq_enter();
|
2008-06-26 03:21:54 -06:00
|
|
|
generic_smp_call_function_interrupt();
|
2008-07-08 16:06:43 -06:00
|
|
|
#ifdef CONFIG_X86_32
|
2007-10-17 10:04:40 -06:00
|
|
|
__get_cpu_var(irq_stat).irq_call_count++;
|
2008-07-08 16:06:43 -06:00
|
|
|
#else
|
|
|
|
add_pda(irq_call_count, 1);
|
|
|
|
#endif
|
2007-07-17 19:37:06 -06:00
|
|
|
irq_exit();
|
|
|
|
|
|
|
|
return IRQ_HANDLED;
|
|
|
|
}
|
|
|
|
|
2008-06-26 03:21:54 -06:00
|
|
|
static irqreturn_t xen_call_function_single_interrupt(int irq, void *dev_id)
|
2007-07-17 19:37:06 -06:00
|
|
|
{
|
2008-06-26 03:21:54 -06:00
|
|
|
irq_enter();
|
|
|
|
generic_smp_call_function_single_interrupt();
|
2008-07-08 16:06:43 -06:00
|
|
|
#ifdef CONFIG_X86_32
|
2008-06-26 03:21:54 -06:00
|
|
|
__get_cpu_var(irq_stat).irq_call_count++;
|
2008-07-08 16:06:43 -06:00
|
|
|
#else
|
|
|
|
add_pda(irq_call_count, 1);
|
|
|
|
#endif
|
2008-06-26 03:21:54 -06:00
|
|
|
irq_exit();
|
2007-07-17 19:37:06 -06:00
|
|
|
|
2008-06-26 03:21:54 -06:00
|
|
|
return IRQ_HANDLED;
|
2007-07-17 19:37:06 -06:00
|
|
|
}
|
2008-07-08 16:06:41 -06:00
|
|
|
|
|
|
|
static const struct smp_ops xen_smp_ops __initdata = {
|
|
|
|
.smp_prepare_boot_cpu = xen_smp_prepare_boot_cpu,
|
|
|
|
.smp_prepare_cpus = xen_smp_prepare_cpus,
|
|
|
|
.smp_cpus_done = xen_smp_cpus_done,
|
|
|
|
|
2008-08-22 04:52:15 -06:00
|
|
|
.cpu_up = xen_cpu_up,
|
|
|
|
.cpu_die = xen_cpu_die,
|
|
|
|
.cpu_disable = xen_cpu_disable,
|
|
|
|
.play_dead = xen_play_dead,
|
|
|
|
|
2008-07-08 16:06:41 -06:00
|
|
|
.smp_send_stop = xen_smp_send_stop,
|
|
|
|
.smp_send_reschedule = xen_smp_send_reschedule,
|
|
|
|
|
|
|
|
.send_call_func_ipi = xen_smp_send_call_function_ipi,
|
|
|
|
.send_call_func_single_ipi = xen_smp_send_call_function_single_ipi,
|
|
|
|
};
|
|
|
|
|
|
|
|
void __init xen_smp_init(void)
|
|
|
|
{
|
|
|
|
smp_ops = xen_smp_ops;
|
2008-07-08 16:06:43 -06:00
|
|
|
xen_fill_possible_map();
|
xen: implement Xen-specific spinlocks
The standard ticket spinlocks are very expensive in a virtual
environment, because their performance depends on Xen's scheduler
giving vcpus time in the order that they're supposed to take the
spinlock.
This implements a Xen-specific spinlock, which should be much more
efficient.
The fast-path is essentially the old Linux-x86 locks, using a single
lock byte. The locker decrements the byte; if the result is 0, then
they have the lock. If the lock is negative, then locker must spin
until the lock is positive again.
When there's contention, the locker spin for 2^16[*] iterations waiting
to get the lock. If it fails to get the lock in that time, it adds
itself to the contention count in the lock and blocks on a per-cpu
event channel.
When unlocking the spinlock, the locker looks to see if there's anyone
blocked waiting for the lock by checking for a non-zero waiter count.
If there's a waiter, it traverses the per-cpu "lock_spinners"
variable, which contains which lock each CPU is waiting on. It picks
one CPU waiting on the lock and sends it an event to wake it up.
This allows efficient fast-path spinlock operation, while allowing
spinning vcpus to give up their processor time while waiting for a
contended lock.
[*] 2^16 iterations is threshold at which 98% locks have been taken
according to Thomas Friebel's Xen Summit talk "Preventing Guests from
Spinning Around". Therefore, we'd expect the lock and unlock slow
paths will only be entered 2% of the time.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Christoph Lameter <clameter@linux-foundation.org>
Cc: Petr Tesarik <ptesarik@suse.cz>
Cc: Virtualization <virtualization@lists.linux-foundation.org>
Cc: Xen devel <xen-devel@lists.xensource.com>
Cc: Thomas Friebel <thomas.friebel@amd.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-07-07 13:07:53 -06:00
|
|
|
xen_init_spinlocks();
|
2008-07-08 16:06:41 -06:00
|
|
|
}
|