[Xen-changelog] [xen-unstable] SMP support for Xen PM.

[Xen patchbot-unstable <patchbot-unstable@xxxxxxxxxxxxxxxxxxx>]

# HG changeset patch
# User kfraser@xxxxxxxxxxxxxxxxxxxxx
# Date 1184847813 -3600
# Node ID 858b9bc8d0e6245ba361cff097547ed8a1729e29
# Parent  0d291a7c8c1f4a77a01c49f2e8cc90271ed19d37
SMP support for Xen PM.

Add SMP sleep support to Xen. One new utility is created
to allow vcpu continue previous running after migration
to new processor.

Signed-off-by: Kevin Tian <kevin.tian@xxxxxxxxx>
Signed-off-by: Keir Fraser <keir@xxxxxxxxxxxxx>
---
 xen/arch/x86/acpi/power.c          |   44 +++++++++++++++++++++++++++------
 xen/arch/x86/cpu/intel_cacheinfo.c |    4 +--
 xen/arch/x86/domain.c              |   49 +++++++++++++++++++++++++++++++++++++
 xen/include/asm-x86/domain.h       |    8 +++++-
 xen/include/asm-x86/smp.h          |    2 +
 5 files changed, 96 insertions(+), 11 deletions(-)

diff -r 0d291a7c8c1f -r 858b9bc8d0e6 xen/arch/x86/acpi/power.c
--- a/xen/arch/x86/acpi/power.c Thu Jul 19 12:53:32 2007 +0100
+++ b/xen/arch/x86/acpi/power.c Thu Jul 19 13:23:33 2007 +0100
@@ -119,19 +119,25 @@ static int enter_state(u32 state)
     if (state <= ACPI_STATE_S0 || state > ACPI_S_STATES_MAX)
         return -EINVAL;
 
-    /* Sync lazy state on ths cpu */
     __sync_lazy_execstate();
     pmprintk(XENLOG_INFO, "Flush lazy state\n");
 
     if (!spin_trylock(&pm_lock))
         return -EBUSY;
     
-    freeze_domains();
-
-    hvm_cpu_down();
-
     pmprintk(XENLOG_INFO, "PM: Preparing system for %s sleep\n",
         acpi_states[state]);
+
+    freeze_domains();
+
+    disable_nonboot_cpus();
+    if (num_online_cpus() != 1)
+    {
+        error = -EBUSY;
+        goto Enable_cpu;
+    }
+
+    hvm_cpu_down();
 
     acpi_sleep_prepare(state);
 
@@ -169,16 +175,31 @@ static int enter_state(u32 state)
     if ( !hvm_cpu_up() )
         BUG();
 
+ Enable_cpu:
+    enable_nonboot_cpus();
+
     thaw_domains();
     spin_unlock(&pm_lock);
     return error;
 }
 
+static void acpi_power_off(void)
+{
+    pmprintk(XENLOG_INFO, "%s called\n", __FUNCTION__);
+    local_irq_disable();
+    /* Some SMP machines only can poweroff in boot CPU */
+    acpi_enter_sleep_state(ACPI_STATE_S5);
+}
+
+static long enter_state_helper(void *data)
+{
+    struct acpi_sleep_info *sinfo = (struct acpi_sleep_info *)data;
+    return enter_state(sinfo->sleep_state);
+}
+
 /*
  * Dom0 issues this hypercall in place of writing pm1a_cnt. Xen then
  * takes over the control and put the system into sleep state really.
- * Also video flags and mode are passed here, in case user may use
- * "acpi_sleep=***" for video resume.
  *
  * Guest may issue a two-phases write to PM1x_CNT, to work
  * around poorly implemented hardware. It's better to keep
@@ -216,7 +237,14 @@ int acpi_enter_sleep(struct xenpf_enter_
     acpi_sinfo.pm1b_cnt_val = sleep->pm1b_cnt_val;
     acpi_sinfo.sleep_state = sleep->sleep_state;
 
-    return enter_state(acpi_sinfo.sleep_state);
+    /* ACPI power-off method. */
+    if ( acpi_sinfo.sleep_state == ACPI_STATE_S5 )
+    {
+        for ( ; ; )
+            acpi_power_off();
+    }
+
+    return continue_hypercall_on_cpu(0, enter_state_helper, &acpi_sinfo);
 }
 
 static int acpi_get_wake_status(void)
diff -r 0d291a7c8c1f -r 858b9bc8d0e6 xen/arch/x86/cpu/intel_cacheinfo.c
--- a/xen/arch/x86/cpu/intel_cacheinfo.c        Thu Jul 19 12:53:32 2007 +0100
+++ b/xen/arch/x86/cpu/intel_cacheinfo.c        Thu Jul 19 13:23:33 2007 +0100
@@ -17,7 +17,7 @@ struct _cache_table
 };
 
 /* all the cache descriptor types we care about (no TLB or trace cache 
entries) */
-static struct _cache_table cache_table[] __initdata =
+static struct _cache_table cache_table[] __devinitdata =
 {
        { 0x06, LVL_1_INST, 8 },        /* 4-way set assoc, 32 byte line size */
        { 0x08, LVL_1_INST, 16 },       /* 4-way set assoc, 32 byte line size */
@@ -60,7 +60,7 @@ static struct _cache_table cache_table[]
        { 0x00, 0, 0}
 };
 
-unsigned int __init init_intel_cacheinfo(struct cpuinfo_x86 *c)
+unsigned int __devinit init_intel_cacheinfo(struct cpuinfo_x86 *c)
 {
        unsigned int trace = 0, l1i = 0, l1d = 0, l2 = 0, l3 = 0; /* Cache 
sizes */
 
diff -r 0d291a7c8c1f -r 858b9bc8d0e6 xen/arch/x86/domain.c
--- a/xen/arch/x86/domain.c     Thu Jul 19 12:53:32 2007 +0100
+++ b/xen/arch/x86/domain.c     Thu Jul 19 13:23:33 2007 +0100
@@ -82,6 +82,7 @@ static void play_dead(void)
     __cpu_disable();
     /* This must be done before dead CPU ack */
     cpu_exit_clear();
+    hvm_cpu_down();
     wbinvd();
     mb();
     /* Ack it */
@@ -1361,6 +1362,54 @@ void sync_vcpu_execstate(struct vcpu *v)
 
     /* Other cpus call __sync_lazy_execstate from flush ipi handler. */
     flush_tlb_mask(v->vcpu_dirty_cpumask);
+}
+
+struct migrate_info {
+    long (*func)(void *data);
+    void *data;
+    void (*saved_schedule_tail)(struct vcpu *);
+    cpumask_t saved_affinity;
+};
+
+static void continue_hypercall_on_cpu_helper(struct vcpu *v)
+{
+    struct cpu_user_regs *regs = guest_cpu_user_regs();
+    struct migrate_info *info = v->arch.continue_info;
+
+    regs->eax = info->func(info->data);
+
+    v->arch.schedule_tail = info->saved_schedule_tail;
+    v->cpu_affinity = info->saved_affinity;
+
+    xfree(info);
+    v->arch.continue_info = NULL;
+
+    vcpu_set_affinity(v, &v->cpu_affinity);
+    schedule_tail(v);
+}
+
+int continue_hypercall_on_cpu(int cpu, long (*func)(void *data), void *data)
+{
+    struct vcpu *v = current;
+    struct migrate_info *info;
+    cpumask_t mask = cpumask_of_cpu(cpu);
+
+    info = xmalloc(struct migrate_info);
+    if ( info == NULL )
+        return -ENOMEM;
+
+    info->func = func;
+    info->data = data;
+    info->saved_schedule_tail = v->arch.schedule_tail;
+    v->arch.schedule_tail = continue_hypercall_on_cpu_helper;
+
+    info->saved_affinity = v->cpu_affinity;
+    v->arch.continue_info = info;
+
+    vcpu_set_affinity(v, &mask);
+    schedule_tail(v);
+
+    return 0;
 }
 
 #define next_arg(fmt, args) ({                                              \
diff -r 0d291a7c8c1f -r 858b9bc8d0e6 xen/include/asm-x86/domain.h
--- a/xen/include/asm-x86/domain.h      Thu Jul 19 12:53:32 2007 +0100
+++ b/xen/include/asm-x86/domain.h      Thu Jul 19 13:23:33 2007 +0100
@@ -268,6 +268,9 @@ struct arch_vcpu
     void (*ctxt_switch_from) (struct vcpu *);
     void (*ctxt_switch_to) (struct vcpu *);
 
+    /* Record information required to continue execution after migration */
+    void *continue_info;
+
     /* Bounce information for propagating an exception to guest OS. */
     struct trap_bounce trap_bounce;
 
@@ -311,9 +314,12 @@ struct arch_vcpu
     unsigned long vcpu_info_mfn;
 } __cacheline_aligned;
 
-/* shorthands to improve code legibility */
+/* Shorthands to improve code legibility. */
 #define hvm_vmx         hvm_vcpu.u.vmx
 #define hvm_svm         hvm_vcpu.u.svm
+
+/* Continue the current hypercall via func(data) on specified cpu. */
+int continue_hypercall_on_cpu(int cpu, long (*func)(void *data), void *data);
 
 #endif /* __ASM_DOMAIN_H__ */
 
diff -r 0d291a7c8c1f -r 858b9bc8d0e6 xen/include/asm-x86/smp.h
--- a/xen/include/asm-x86/smp.h Thu Jul 19 12:53:32 2007 +0100
+++ b/xen/include/asm-x86/smp.h Thu Jul 19 13:23:33 2007 +0100
@@ -66,6 +66,8 @@ extern void enable_nonboot_cpus(void);
 extern void enable_nonboot_cpus(void);
 #else
 static inline int cpu_is_offline(int cpu) {return 0;}
+static inline void disable_nonboot_cpus(void) {}
+static inline void enable_nonboot_cpus(void) {}
 #endif
 
 /*

_______________________________________________
Xen-changelog mailing list
Xen-changelog@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-changelog