Reduce 'd' debug key's global impact

On large systems, dumping state may cause time management to get
stalled for so long a period that it wouldn't recover. Therefore alter
the state dumping logic to alternatively block each CPU as it prints
rather than one CPU for a very long time (using the alternative key
handling toggle introduced with an earlier patch).

Further, instead of using on_selected_cpus(), which is unsafe when
the dumping happens from a hardware interrupt, introduce and use a
dedicated IPI sending function (which each architecture can implement
to its liking)

Finally, don't print useless data (e.g. the hypervisor context of the
interrupt that is used for triggering the printing, but isn't part of
the context that's actually interesting).

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>

xen/arch/ia64/linux-xen/smp.c

@@ -94,6 +94,7 @@ static volatile struct call_data_struct *call_data;
 
 #define IPI_CALL_FUNC		0
 #define IPI_CPU_STOP		1
+#define IPI_STATE_DUMP		2
 
 /* This needs to be cacheline aligned because it is written to by *other* CPUs.  */
 static DEFINE_PER_CPU(u64, ipi_operation) ____cacheline_aligned;
@@ -202,6 +203,10 @@ handle_IPI (int irq, void *dev_id, struct pt_regs *regs)
 				stop_this_cpu();
 				break;
 
+			      case IPI_STATE_DUMP:
+				dump_execstate(regs);
+				break;
+
 			      default:
 				printk(KERN_CRIT "Unknown IPI on CPU %d: %lu\n", this_cpu, which);
 				break;
@@ -479,6 +484,12 @@ smp_send_stop (void)
 	send_IPI_allbutself(IPI_CPU_STOP);
 }
 
+void
+smp_send_state_dump (unsigned int cpu)
+{
+	send_IPI_single(cpu, IPI_STATE_DUMP);
+}
+
 int __init
 setup_profiling_timer (unsigned int multiplier)
 {

xen/arch/x86/apic.c

@@ -1233,8 +1233,17 @@ fastcall void smp_apic_timer_interrupt(struct cpu_user_regs * regs)
     set_irq_regs(old_regs);
 }
 
+static DEFINE_PER_CPU(bool_t, state_dump_pending);
+
+void smp_send_state_dump(unsigned int cpu)
+{
+    /* We overload the spurious interrupt handler to handle the dump. */
+    per_cpu(state_dump_pending, cpu) = 1;
+    send_IPI_mask(cpumask_of(cpu), SPURIOUS_APIC_VECTOR);
+}
+
 /*
- * This interrupt should _never_ happen with our APIC/SMP architecture
+ * Spurious interrupts should _never_ happen with our APIC/SMP architecture.
  */
 fastcall void smp_spurious_interrupt(struct cpu_user_regs *regs)
 {
@@ -1242,18 +1251,27 @@ fastcall void smp_spurious_interrupt(struct cpu_user_regs *regs)
     struct cpu_user_regs *old_regs = set_irq_regs(regs);
 
     irq_enter();
+
     /*
-     * Check if this really is a spurious interrupt and ACK it
-     * if it is a vectored one.  Just in case...
-     * Spurious interrupts should not be ACKed.
+     * Check if this is a vectored interrupt (most likely, as this is probably
+     * a request to dump local CPU state). Vectored interrupts are ACKed;
+     * spurious interrupts are not.
      */
     v = apic_read(APIC_ISR + ((SPURIOUS_APIC_VECTOR & ~0x1f) >> 1));
-    if (v & (1 << (SPURIOUS_APIC_VECTOR & 0x1f)))
+    if (v & (1 << (SPURIOUS_APIC_VECTOR & 0x1f))) {
         ack_APIC_irq();
+        if (this_cpu(state_dump_pending)) {
+            this_cpu(state_dump_pending) = 0;
+            dump_execstate(regs);
+            goto out;
+        }
+    }
 
     /* see sw-dev-man vol 3, chapter 7.4.13.5 */
-    printk(KERN_INFO "spurious APIC interrupt on CPU#%d, should never happen.\n",
-           smp_processor_id());
+    printk(KERN_INFO "spurious APIC interrupt on CPU#%d, should "
+           "never happen.\n", smp_processor_id());
+
+ out:
     irq_exit();
     set_irq_regs(old_regs);
 }

xen/common/keyhandler.c

@@ -71,14 +71,40 @@ static struct keyhandler show_handlers_keyhandler = {
     .desc = "show this message"
 };
 
-static void __dump_execstate(void *unused)
+static cpumask_t dump_execstate_mask;
+
+void dump_execstate(struct cpu_user_regs *regs)
 {
-    dump_execution_state();
-    printk("*** Dumping CPU%d guest state: ***\n", smp_processor_id());
-    if ( is_idle_vcpu(current) )
-        printk("No guest context (CPU is idle).\n");
-    else
+    unsigned int cpu = smp_processor_id();
+
+    if ( !guest_mode(regs) )
+    {
+        printk("*** Dumping CPU%u host state: ***\n", cpu);
+        show_execution_state(regs);
+    }
+
+    if ( !is_idle_vcpu(current) )
+    {
+        printk("*** Dumping CPU%u guest state (d%d:v%d): ***\n",
+               smp_processor_id(), current->domain->domain_id,
+               current->vcpu_id);
         show_execution_state(guest_cpu_user_regs());
+        printk("\n");
+    }
+
+    cpu_clear(cpu, dump_execstate_mask);
+    if ( !alt_key_handling )
+        return;
+
+    cpu = cycle_cpu(cpu, dump_execstate_mask);
+    if ( cpu < NR_CPUS )
+    {
+        smp_send_state_dump(cpu);
+        return;
+    }
+
+    console_end_sync();
+    watchdog_enable();
 }
 
 static void dump_registers(unsigned char key, struct cpu_user_regs *regs)
@@ -89,22 +115,25 @@ static void dump_registers(unsigned char key, struct cpu_user_regs *regs)
     watchdog_disable();
     console_start_sync();
 
-    printk("'%c' pressed -> dumping registers\n", key);
+    printk("'%c' pressed -> dumping registers\n\n", key);
+
+    dump_execstate_mask = cpu_online_map;
 
     /* Get local execution state out immediately, in case we get stuck. */
-    printk("\n*** Dumping CPU%d host state: ***\n", smp_processor_id());
-    __dump_execstate(NULL);
+    dump_execstate(regs);
 
-    for_each_online_cpu ( cpu )
+    /* Alt. handling: remaining CPUs are dumped asynchronously one-by-one. */
+    if ( alt_key_handling )
+        return;
+
+    /* Normal handling: synchronously dump the remaining CPUs' states. */
+    for_each_cpu_mask ( cpu, dump_execstate_mask )
     {
-        if ( cpu == smp_processor_id() )
-            continue;
-        printk("\n*** Dumping CPU%d host state: ***\n", cpu);
-        on_selected_cpus(cpumask_of(cpu), __dump_execstate, NULL, 1);
+        smp_send_state_dump(cpu);
+        while ( cpu_isset(cpu, dump_execstate_mask) )
+            cpu_relax();
     }
 
-    printk("\n");
-
     console_end_sync();
     watchdog_enable();
 }

xen/include/asm-ia64/linux-xen/asm/ptrace.h

@@ -280,7 +280,7 @@ struct switch_stack {
 # define ia64_task_regs(t)		(((struct pt_regs *) ((char *) (t) + IA64_STK_OFFSET)) - 1)
 # define ia64_psr(regs)			((struct ia64_psr *) &(regs)->cr_ipsr)
 #ifdef XEN
-# define guest_mode(regs)		(ia64_psr(regs)->cpl != 0)
+# define guest_mode(regs)		(ia64_psr(regs)->cpl && !ia64_psr(regs)->vm)
 # define guest_kernel_mode(regs)	(ia64_psr(regs)->cpl == CONFIG_CPL0_EMUL)
 # define vmx_guest_kernel_mode(regs)	(ia64_psr(regs)->cpl == 0)
 # define regs_increment_iip(regs)					\

xen/include/xen/lib.h

@@ -111,4 +111,7 @@ extern int tainted;
 extern char *print_tainted(char *str);
 extern void add_taint(unsigned);
 
+struct cpu_user_regs;
+void dump_execstate(struct cpu_user_regs *);
+
 #endif /* __LIB_H__ */

xen/include/xen/smp.h

@@ -13,6 +13,8 @@ extern void smp_send_event_check_mask(const cpumask_t *mask);
 #define smp_send_event_check_cpu(cpu) \
     smp_send_event_check_mask(cpumask_of(cpu))
 
+extern void smp_send_state_dump(unsigned int cpu);
+
 /*
  * Prepare machine for booting other CPUs.
  */