linux/drivers/xen/events.c

838 lines
19 KiB
C
Raw Normal View History

/*
* Xen event channels
*
* Xen models interrupts with abstract event channels. Because each
* domain gets 1024 event channels, but NR_IRQ is not that large, we
* must dynamically map irqs<->event channels. The event channels
* interface with the rest of the kernel by defining a xen interrupt
* chip. When an event is recieved, it is mapped to an irq and sent
* through the normal interrupt processing path.
*
* There are four kinds of events which can be mapped to an event
* channel:
*
* 1. Inter-domain notifications. This includes all the virtual
* device events, since they're driven by front-ends in another domain
* (typically dom0).
* 2. VIRQs, typically used for timers. These are per-cpu events.
* 3. IPIs.
* 4. Hardware interrupts. Not supported at present.
*
* Jeremy Fitzhardinge <jeremy@xensource.com>, XenSource Inc, 2007
*/
#include <linux/linkage.h>
#include <linux/interrupt.h>
#include <linux/irq.h>
#include <linux/module.h>
#include <linux/string.h>
#include <asm/ptrace.h>
#include <asm/irq.h>
#include <asm/sync_bitops.h>
#include <asm/xen/hypercall.h>
#include <asm/xen/hypervisor.h>
#include <xen/xen-ops.h>
#include <xen/events.h>
#include <xen/interface/xen.h>
#include <xen/interface/event_channel.h>
/*
* This lock protects updates to the following mapping and reference-count
* arrays. The lock does not need to be acquired to read the mapping tables.
*/
static DEFINE_SPINLOCK(irq_mapping_update_lock);
/* IRQ <-> VIRQ mapping. */
static DEFINE_PER_CPU(int, virq_to_irq[NR_VIRQS]) = {[0 ... NR_VIRQS-1] = -1};
/* IRQ <-> IPI mapping */
static DEFINE_PER_CPU(int, ipi_to_irq[XEN_NR_IPIS]) = {[0 ... XEN_NR_IPIS-1] = -1};
/* Packed IRQ information: binding type, sub-type index, and event channel. */
struct packed_irq
{
unsigned short evtchn;
unsigned char index;
unsigned char type;
};
static struct packed_irq irq_info[NR_IRQS];
/* Binding types. */
enum {
IRQT_UNBOUND,
IRQT_PIRQ,
IRQT_VIRQ,
IRQT_IPI,
IRQT_EVTCHN
};
/* Convenient shorthand for packed representation of an unbound IRQ. */
#define IRQ_UNBOUND mk_irq_info(IRQT_UNBOUND, 0, 0)
static int evtchn_to_irq[NR_EVENT_CHANNELS] = {
[0 ... NR_EVENT_CHANNELS-1] = -1
};
static unsigned long cpu_evtchn_mask[NR_CPUS][NR_EVENT_CHANNELS/BITS_PER_LONG];
static u8 cpu_evtchn[NR_EVENT_CHANNELS];
/* Reference counts for bindings to IRQs. */
static int irq_bindcount[NR_IRQS];
/* Xen will never allocate port zero for any purpose. */
#define VALID_EVTCHN(chn) ((chn) != 0)
static struct irq_chip xen_dynamic_chip;
/* Constructor for packed IRQ information. */
static inline struct packed_irq mk_irq_info(u32 type, u32 index, u32 evtchn)
{
return (struct packed_irq) { evtchn, index, type };
}
/*
* Accessors for packed IRQ information.
*/
static inline unsigned int evtchn_from_irq(int irq)
{
return irq_info[irq].evtchn;
}
static inline unsigned int index_from_irq(int irq)
{
return irq_info[irq].index;
}
static inline unsigned int type_from_irq(int irq)
{
return irq_info[irq].type;
}
static inline unsigned long active_evtchns(unsigned int cpu,
struct shared_info *sh,
unsigned int idx)
{
return (sh->evtchn_pending[idx] &
cpu_evtchn_mask[cpu][idx] &
~sh->evtchn_mask[idx]);
}
static void bind_evtchn_to_cpu(unsigned int chn, unsigned int cpu)
{
int irq = evtchn_to_irq[chn];
BUG_ON(irq == -1);
#ifdef CONFIG_SMP
irq_to_desc(irq)->affinity = cpumask_of_cpu(cpu);
#endif
__clear_bit(chn, cpu_evtchn_mask[cpu_evtchn[chn]]);
__set_bit(chn, cpu_evtchn_mask[cpu]);
cpu_evtchn[chn] = cpu;
}
static void init_evtchn_cpu_bindings(void)
{
#ifdef CONFIG_SMP
struct irq_desc *desc;
int i;
/* By default all event channels notify CPU#0. */
for_each_irq_desc(i, desc) {
desc->affinity = cpumask_of_cpu(0);
}
#endif
memset(cpu_evtchn, 0, sizeof(cpu_evtchn));
memset(cpu_evtchn_mask[0], ~0, sizeof(cpu_evtchn_mask[0]));
}
static inline unsigned int cpu_from_evtchn(unsigned int evtchn)
{
return cpu_evtchn[evtchn];
}
static inline void clear_evtchn(int port)
{
struct shared_info *s = HYPERVISOR_shared_info;
sync_clear_bit(port, &s->evtchn_pending[0]);
}
static inline void set_evtchn(int port)
{
struct shared_info *s = HYPERVISOR_shared_info;
sync_set_bit(port, &s->evtchn_pending[0]);
}
static inline int test_evtchn(int port)
{
struct shared_info *s = HYPERVISOR_shared_info;
return sync_test_bit(port, &s->evtchn_pending[0]);
}
/**
* notify_remote_via_irq - send event to remote end of event channel via irq
* @irq: irq of event channel to send event to
*
* Unlike notify_remote_via_evtchn(), this is safe to use across
* save/restore. Notifications on a broken connection are silently
* dropped.
*/
void notify_remote_via_irq(int irq)
{
int evtchn = evtchn_from_irq(irq);
if (VALID_EVTCHN(evtchn))
notify_remote_via_evtchn(evtchn);
}
EXPORT_SYMBOL_GPL(notify_remote_via_irq);
static void mask_evtchn(int port)
{
struct shared_info *s = HYPERVISOR_shared_info;
sync_set_bit(port, &s->evtchn_mask[0]);
}
static void unmask_evtchn(int port)
{
struct shared_info *s = HYPERVISOR_shared_info;
unsigned int cpu = get_cpu();
BUG_ON(!irqs_disabled());
/* Slow path (hypercall) if this is a non-local port. */
if (unlikely(cpu != cpu_from_evtchn(port))) {
struct evtchn_unmask unmask = { .port = port };
(void)HYPERVISOR_event_channel_op(EVTCHNOP_unmask, &unmask);
} else {
struct vcpu_info *vcpu_info = __get_cpu_var(xen_vcpu);
sync_clear_bit(port, &s->evtchn_mask[0]);
/*
* The following is basically the equivalent of
* 'hw_resend_irq'. Just like a real IO-APIC we 'lose
* the interrupt edge' if the channel is masked.
*/
if (sync_test_bit(port, &s->evtchn_pending[0]) &&
!sync_test_and_set_bit(port / BITS_PER_LONG,
&vcpu_info->evtchn_pending_sel))
vcpu_info->evtchn_upcall_pending = 1;
}
put_cpu();
}
static int find_unbound_irq(void)
{
int irq;
struct irq_desc *desc;
/* Only allocate from dynirq range */
for (irq = 0; irq < nr_irqs; irq++)
if (irq_bindcount[irq] == 0)
break;
if (irq == nr_irqs)
panic("No available IRQ to bind to: increase nr_irqs!\n");
desc = irq_to_desc_alloc_cpu(irq, 0);
if (WARN_ON(desc == NULL))
return -1;
return irq;
}
int bind_evtchn_to_irq(unsigned int evtchn)
{
int irq;
spin_lock(&irq_mapping_update_lock);
irq = evtchn_to_irq[evtchn];
if (irq == -1) {
irq = find_unbound_irq();
dynamic_irq_init(irq);
set_irq_chip_and_handler_name(irq, &xen_dynamic_chip,
handle_level_irq, "event");
evtchn_to_irq[evtchn] = irq;
irq_info[irq] = mk_irq_info(IRQT_EVTCHN, 0, evtchn);
}
irq_bindcount[irq]++;
spin_unlock(&irq_mapping_update_lock);
return irq;
}
EXPORT_SYMBOL_GPL(bind_evtchn_to_irq);
static int bind_ipi_to_irq(unsigned int ipi, unsigned int cpu)
{
struct evtchn_bind_ipi bind_ipi;
int evtchn, irq;
spin_lock(&irq_mapping_update_lock);
irq = per_cpu(ipi_to_irq, cpu)[ipi];
if (irq == -1) {
irq = find_unbound_irq();
if (irq < 0)
goto out;
dynamic_irq_init(irq);
set_irq_chip_and_handler_name(irq, &xen_dynamic_chip,
handle_level_irq, "ipi");
bind_ipi.vcpu = cpu;
if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_ipi,
&bind_ipi) != 0)
BUG();
evtchn = bind_ipi.port;
evtchn_to_irq[evtchn] = irq;
irq_info[irq] = mk_irq_info(IRQT_IPI, ipi, evtchn);
per_cpu(ipi_to_irq, cpu)[ipi] = irq;
bind_evtchn_to_cpu(evtchn, cpu);
}
irq_bindcount[irq]++;
out:
spin_unlock(&irq_mapping_update_lock);
return irq;
}
static int bind_virq_to_irq(unsigned int virq, unsigned int cpu)
{
struct evtchn_bind_virq bind_virq;
int evtchn, irq;
spin_lock(&irq_mapping_update_lock);
irq = per_cpu(virq_to_irq, cpu)[virq];
if (irq == -1) {
bind_virq.virq = virq;
bind_virq.vcpu = cpu;
if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
&bind_virq) != 0)
BUG();
evtchn = bind_virq.port;
irq = find_unbound_irq();
dynamic_irq_init(irq);
set_irq_chip_and_handler_name(irq, &xen_dynamic_chip,
handle_level_irq, "virq");
evtchn_to_irq[evtchn] = irq;
irq_info[irq] = mk_irq_info(IRQT_VIRQ, virq, evtchn);
per_cpu(virq_to_irq, cpu)[virq] = irq;
bind_evtchn_to_cpu(evtchn, cpu);
}
irq_bindcount[irq]++;
spin_unlock(&irq_mapping_update_lock);
return irq;
}
static void unbind_from_irq(unsigned int irq)
{
struct evtchn_close close;
int evtchn = evtchn_from_irq(irq);
spin_lock(&irq_mapping_update_lock);
if ((--irq_bindcount[irq] == 0) && VALID_EVTCHN(evtchn)) {
close.port = evtchn;
if (HYPERVISOR_event_channel_op(EVTCHNOP_close, &close) != 0)
BUG();
switch (type_from_irq(irq)) {
case IRQT_VIRQ:
per_cpu(virq_to_irq, cpu_from_evtchn(evtchn))
[index_from_irq(irq)] = -1;
break;
case IRQT_IPI:
per_cpu(ipi_to_irq, cpu_from_evtchn(evtchn))
[index_from_irq(irq)] = -1;
break;
default:
break;
}
/* Closed ports are implicitly re-bound to VCPU0. */
bind_evtchn_to_cpu(evtchn, 0);
evtchn_to_irq[evtchn] = -1;
irq_info[irq] = IRQ_UNBOUND;
dynamic_irq_cleanup(irq);
}
spin_unlock(&irq_mapping_update_lock);
}
int bind_evtchn_to_irqhandler(unsigned int evtchn,
irq_handler_t handler,
unsigned long irqflags,
const char *devname, void *dev_id)
{
unsigned int irq;
int retval;
irq = bind_evtchn_to_irq(evtchn);
retval = request_irq(irq, handler, irqflags, devname, dev_id);
if (retval != 0) {
unbind_from_irq(irq);
return retval;
}
return irq;
}
EXPORT_SYMBOL_GPL(bind_evtchn_to_irqhandler);
int bind_virq_to_irqhandler(unsigned int virq, unsigned int cpu,
irq_handler_t handler,
unsigned long irqflags, const char *devname, void *dev_id)
{
unsigned int irq;
int retval;
irq = bind_virq_to_irq(virq, cpu);
retval = request_irq(irq, handler, irqflags, devname, dev_id);
if (retval != 0) {
unbind_from_irq(irq);
return retval;
}
return irq;
}
EXPORT_SYMBOL_GPL(bind_virq_to_irqhandler);
int bind_ipi_to_irqhandler(enum ipi_vector ipi,
unsigned int cpu,
irq_handler_t handler,
unsigned long irqflags,
const char *devname,
void *dev_id)
{
int irq, retval;
irq = bind_ipi_to_irq(ipi, cpu);
if (irq < 0)
return irq;
retval = request_irq(irq, handler, irqflags, devname, dev_id);
if (retval != 0) {
unbind_from_irq(irq);
return retval;
}
return irq;
}
void unbind_from_irqhandler(unsigned int irq, void *dev_id)
{
free_irq(irq, dev_id);
unbind_from_irq(irq);
}
EXPORT_SYMBOL_GPL(unbind_from_irqhandler);
void xen_send_IPI_one(unsigned int cpu, enum ipi_vector vector)
{
int irq = per_cpu(ipi_to_irq, cpu)[vector];
BUG_ON(irq < 0);
notify_remote_via_irq(irq);
}
irqreturn_t xen_debug_interrupt(int irq, void *dev_id)
{
struct shared_info *sh = HYPERVISOR_shared_info;
int cpu = smp_processor_id();
int i;
unsigned long flags;
static DEFINE_SPINLOCK(debug_lock);
spin_lock_irqsave(&debug_lock, flags);
printk("vcpu %d\n ", cpu);
for_each_online_cpu(i) {
struct vcpu_info *v = per_cpu(xen_vcpu, i);
printk("%d: masked=%d pending=%d event_sel %08lx\n ", i,
(get_irq_regs() && i == cpu) ? xen_irqs_disabled(get_irq_regs()) : v->evtchn_upcall_mask,
v->evtchn_upcall_pending,
v->evtchn_pending_sel);
}
printk("pending:\n ");
for(i = ARRAY_SIZE(sh->evtchn_pending)-1; i >= 0; i--)
printk("%08lx%s", sh->evtchn_pending[i],
i % 8 == 0 ? "\n " : " ");
printk("\nmasks:\n ");
for(i = ARRAY_SIZE(sh->evtchn_mask)-1; i >= 0; i--)
printk("%08lx%s", sh->evtchn_mask[i],
i % 8 == 0 ? "\n " : " ");
printk("\nunmasked:\n ");
for(i = ARRAY_SIZE(sh->evtchn_mask)-1; i >= 0; i--)
printk("%08lx%s", sh->evtchn_pending[i] & ~sh->evtchn_mask[i],
i % 8 == 0 ? "\n " : " ");
printk("\npending list:\n");
for(i = 0; i < NR_EVENT_CHANNELS; i++) {
if (sync_test_bit(i, sh->evtchn_pending)) {
printk(" %d: event %d -> irq %d\n",
cpu_evtchn[i], i,
evtchn_to_irq[i]);
}
}
spin_unlock_irqrestore(&debug_lock, flags);
return IRQ_HANDLED;
}
/*
* Search the CPUs pending events bitmasks. For each one found, map
* the event number to an irq, and feed it into do_IRQ() for
* handling.
*
* Xen uses a two-level bitmap to speed searching. The first level is
* a bitset of words which contain pending event bits. The second
* level is a bitset of pending events themselves.
*/
void xen_evtchn_do_upcall(struct pt_regs *regs)
{
int cpu = get_cpu();
struct shared_info *s = HYPERVISOR_shared_info;
struct vcpu_info *vcpu_info = __get_cpu_var(xen_vcpu);
static DEFINE_PER_CPU(unsigned, nesting_count);
unsigned count;
do {
unsigned long pending_words;
vcpu_info->evtchn_upcall_pending = 0;
if (__get_cpu_var(nesting_count)++)
goto out;
#ifndef CONFIG_X86 /* No need for a barrier -- XCHG is a barrier on x86. */
/* Clear master flag /before/ clearing selector flag. */
wmb();
#endif
pending_words = xchg(&vcpu_info->evtchn_pending_sel, 0);
while (pending_words != 0) {
unsigned long pending_bits;
int word_idx = __ffs(pending_words);
pending_words &= ~(1UL << word_idx);
while ((pending_bits = active_evtchns(cpu, s, word_idx)) != 0) {
int bit_idx = __ffs(pending_bits);
int port = (word_idx * BITS_PER_LONG) + bit_idx;
int irq = evtchn_to_irq[port];
if (irq != -1)
xen_do_IRQ(irq, regs);
}
}
BUG_ON(!irqs_disabled());
count = __get_cpu_var(nesting_count);
__get_cpu_var(nesting_count) = 0;
} while(count != 1);
out:
put_cpu();
}
/* Rebind a new event channel to an existing irq. */
void rebind_evtchn_irq(int evtchn, int irq)
{
/* Make sure the irq is masked, since the new event channel
will also be masked. */
disable_irq(irq);
spin_lock(&irq_mapping_update_lock);
/* After resume the irq<->evtchn mappings are all cleared out */
BUG_ON(evtchn_to_irq[evtchn] != -1);
/* Expect irq to have been bound before,
so the bindcount should be non-0 */
BUG_ON(irq_bindcount[irq] == 0);
evtchn_to_irq[evtchn] = irq;
irq_info[irq] = mk_irq_info(IRQT_EVTCHN, 0, evtchn);
spin_unlock(&irq_mapping_update_lock);
/* new event channels are always bound to cpu 0 */
irq_set_affinity(irq, cpumask_of(0));
/* Unmask the event channel. */
enable_irq(irq);
}
/* Rebind an evtchn so that it gets delivered to a specific cpu */
static void rebind_irq_to_cpu(unsigned irq, unsigned tcpu)
{
struct evtchn_bind_vcpu bind_vcpu;
int evtchn = evtchn_from_irq(irq);
if (!VALID_EVTCHN(evtchn))
return;
/* Send future instances of this interrupt to other vcpu. */
bind_vcpu.port = evtchn;
bind_vcpu.vcpu = tcpu;
/*
* If this fails, it usually just indicates that we're dealing with a
* virq or IPI channel, which don't actually need to be rebound. Ignore
* it, but don't do the xenlinux-level rebind in that case.
*/
if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_vcpu, &bind_vcpu) >= 0)
bind_evtchn_to_cpu(evtchn, tcpu);
}
static void set_affinity_irq(unsigned irq, const struct cpumask *dest)
{
unsigned tcpu = cpumask_first(dest);
rebind_irq_to_cpu(irq, tcpu);
}
int resend_irq_on_evtchn(unsigned int irq)
{
int masked, evtchn = evtchn_from_irq(irq);
struct shared_info *s = HYPERVISOR_shared_info;
if (!VALID_EVTCHN(evtchn))
return 1;
masked = sync_test_and_set_bit(evtchn, s->evtchn_mask);
sync_set_bit(evtchn, s->evtchn_pending);
if (!masked)
unmask_evtchn(evtchn);
return 1;
}
static void enable_dynirq(unsigned int irq)
{
int evtchn = evtchn_from_irq(irq);
if (VALID_EVTCHN(evtchn))
unmask_evtchn(evtchn);
}
static void disable_dynirq(unsigned int irq)
{
int evtchn = evtchn_from_irq(irq);
if (VALID_EVTCHN(evtchn))
mask_evtchn(evtchn);
}
static void ack_dynirq(unsigned int irq)
{
int evtchn = evtchn_from_irq(irq);
move_native_irq(irq);
if (VALID_EVTCHN(evtchn))
clear_evtchn(evtchn);
}
static int retrigger_dynirq(unsigned int irq)
{
int evtchn = evtchn_from_irq(irq);
struct shared_info *sh = HYPERVISOR_shared_info;
int ret = 0;
if (VALID_EVTCHN(evtchn)) {
int masked;
masked = sync_test_and_set_bit(evtchn, sh->evtchn_mask);
sync_set_bit(evtchn, sh->evtchn_pending);
if (!masked)
unmask_evtchn(evtchn);
ret = 1;
}
return ret;
}
static void restore_cpu_virqs(unsigned int cpu)
{
struct evtchn_bind_virq bind_virq;
int virq, irq, evtchn;
for (virq = 0; virq < NR_VIRQS; virq++) {
if ((irq = per_cpu(virq_to_irq, cpu)[virq]) == -1)
continue;
BUG_ON(irq_info[irq].type != IRQT_VIRQ);
BUG_ON(irq_info[irq].index != virq);
/* Get a new binding from Xen. */
bind_virq.virq = virq;
bind_virq.vcpu = cpu;
if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
&bind_virq) != 0)
BUG();
evtchn = bind_virq.port;
/* Record the new mapping. */
evtchn_to_irq[evtchn] = irq;
irq_info[irq] = mk_irq_info(IRQT_VIRQ, virq, evtchn);
bind_evtchn_to_cpu(evtchn, cpu);
/* Ready for use. */
unmask_evtchn(evtchn);
}
}
static void restore_cpu_ipis(unsigned int cpu)
{
struct evtchn_bind_ipi bind_ipi;
int ipi, irq, evtchn;
for (ipi = 0; ipi < XEN_NR_IPIS; ipi++) {
if ((irq = per_cpu(ipi_to_irq, cpu)[ipi]) == -1)
continue;
BUG_ON(irq_info[irq].type != IRQT_IPI);
BUG_ON(irq_info[irq].index != ipi);
/* Get a new binding from Xen. */
bind_ipi.vcpu = cpu;
if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_ipi,
&bind_ipi) != 0)
BUG();
evtchn = bind_ipi.port;
/* Record the new mapping. */
evtchn_to_irq[evtchn] = irq;
irq_info[irq] = mk_irq_info(IRQT_IPI, ipi, evtchn);
bind_evtchn_to_cpu(evtchn, cpu);
/* Ready for use. */
unmask_evtchn(evtchn);
}
}
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>
2008-07-08 03:07:53 +08:00
/* Clear an irq's pending state, in preparation for polling on it */
void xen_clear_irq_pending(int irq)
{
int evtchn = evtchn_from_irq(irq);
if (VALID_EVTCHN(evtchn))
clear_evtchn(evtchn);
}
void xen_set_irq_pending(int irq)
{
int evtchn = evtchn_from_irq(irq);
if (VALID_EVTCHN(evtchn))
set_evtchn(evtchn);
}
bool xen_test_irq_pending(int irq)
{
int evtchn = evtchn_from_irq(irq);
bool ret = false;
if (VALID_EVTCHN(evtchn))
ret = test_evtchn(evtchn);
return ret;
}
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>
2008-07-08 03:07:53 +08:00
/* Poll waiting for an irq to become pending. In the usual case, the
irq will be disabled so it won't deliver an interrupt. */
void xen_poll_irq(int irq)
{
evtchn_port_t evtchn = evtchn_from_irq(irq);
if (VALID_EVTCHN(evtchn)) {
struct sched_poll poll;
poll.nr_ports = 1;
poll.timeout = 0;
set_xen_guest_handle(poll.ports, &evtchn);
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>
2008-07-08 03:07:53 +08:00
if (HYPERVISOR_sched_op(SCHEDOP_poll, &poll) != 0)
BUG();
}
}
void xen_irq_resume(void)
{
unsigned int cpu, irq, evtchn;
init_evtchn_cpu_bindings();
/* New event-channel space is not 'live' yet. */
for (evtchn = 0; evtchn < NR_EVENT_CHANNELS; evtchn++)
mask_evtchn(evtchn);
/* No IRQ <-> event-channel mappings. */
for (irq = 0; irq < nr_irqs; irq++)
irq_info[irq].evtchn = 0; /* zap event-channel binding */
for (evtchn = 0; evtchn < NR_EVENT_CHANNELS; evtchn++)
evtchn_to_irq[evtchn] = -1;
for_each_possible_cpu(cpu) {
restore_cpu_virqs(cpu);
restore_cpu_ipis(cpu);
}
}
static struct irq_chip xen_dynamic_chip __read_mostly = {
.name = "xen-dyn",
.mask = disable_dynirq,
.unmask = enable_dynirq,
.ack = ack_dynirq,
.set_affinity = set_affinity_irq,
.retrigger = retrigger_dynirq,
};
void __init xen_init_IRQ(void)
{
int i;
init_evtchn_cpu_bindings();
/* No event channels are 'live' right now. */
for (i = 0; i < NR_EVENT_CHANNELS; i++)
mask_evtchn(i);
/* Dynamic IRQ space is currently unbound. Zero the refcnts. */
for (i = 0; i < nr_irqs; i++)
irq_bindcount[i] = 0;
irq_ctx_init(smp_processor_id());
}