Merge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/l…

…inux/kernel/git/tip/tip

Pull x86 fixes from Ingo Molnar:
 "Misc fixes:

   - fix hotplug bugs
   - fix irq live lock
   - fix various topology handling bugs
   - fix APIC ACK ordering
   - fix PV iopl handling
   - fix speling
   - fix/tweak memcpy_mcsafe() return value
   - fix fbcon bug
   - remove stray prototypes"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/msr: Remove unused native_read_tscp()
  x86/apic: Remove declaration of unused hw_nmi_is_cpu_stuck
  x86/oprofile/nmi: Add missing hotplug FROZEN handling
  x86/hpet: Use proper mask to modify hotplug action
  x86/apic/uv: Fix the hotplug notifier
  x86/apb/timer: Use proper mask to modify hotplug action
  x86/topology: Use total_cpus not nr_cpu_ids for logical packages
  x86/topology: Fix Intel HT disable
  x86/topology: Fix logical package mapping
  x86/irq: Cure live lock in fixup_irqs()
  x86/tsc: Prevent NULL pointer deref in calibrate_delay_is_known()
  x86/apic: Fix suspicious RCU usage in smp_trace_call_function_interrupt()
  x86/iopl: Fix iopl capability check on Xen PV
  x86/iopl/64: Properly context-switch IOPL on Xen PV
  selftests/x86: Add an iopl test
  x86/mm, x86/mce: Fix return type/value for memcpy_mcsafe()
  x86/video: Don't assume all FB devices are PCI devices
  arch/x86/irq: Purge useless handler declarations from hw_irq.h
  x86: Fix misspellings in comments

arch/x86/entry/vdso/vclock_gettime.c

@@ -178,7 +178,7 @@ notrace static cycle_t vread_tsc(void)
 
 	/*
 	 * GCC likes to generate cmov here, but this branch is extremely
-	 * predictable (it's just a funciton of time and the likely is
+	 * predictable (it's just a function of time and the likely is
 	 * very likely) and there's a data dependence, so force GCC
 	 * to generate a branch instead.  I don't barrier() because
 	 * we don't actually need a barrier, and if this function

arch/x86/events/intel/lbr.c

@@ -649,7 +649,7 @@ int intel_pmu_setup_lbr_filter(struct perf_event *event)
 
 /*
  * return the type of control flow change at address "from"
- * intruction is not necessarily a branch (in case of interrupt).
+ * instruction is not necessarily a branch (in case of interrupt).
  *
  * The branch type returned also includes the priv level of the
  * target of the control flow change (X86_BR_USER, X86_BR_KERNEL).

arch/x86/events/perf_event.h

@@ -272,7 +272,7 @@ struct cpu_hw_events {
  * events to select for counter rescheduling.
  *
  * Care must be taken as the rescheduling algorithm is O(n!) which
- * will increase scheduling cycles for an over-commited system
+ * will increase scheduling cycles for an over-committed system
  * dramatically.  The number of such EVENT_CONSTRAINT_OVERLAP() macros
  * and its counter masks must be kept at a minimum.
  */

arch/x86/include/asm/apic.h

@@ -643,8 +643,8 @@ static inline void entering_irq(void)
 
 static inline void entering_ack_irq(void)
 {
-	ack_APIC_irq();
 	entering_irq();
+	ack_APIC_irq();
 }
 
 static inline void ipi_entering_ack_irq(void)

arch/x86/include/asm/ftrace.h

@@ -52,7 +52,7 @@ int ftrace_int3_handler(struct pt_regs *regs);
  * this screws up the trace output when tracing a ia32 task.
  * Instead of reporting bogus syscalls, just do not trace them.
  *
- * If the user realy wants these, then they should use the
+ * If the user really wants these, then they should use the
  * raw syscall tracepoints with filtering.
  */
 #define ARCH_TRACE_IGNORE_COMPAT_SYSCALLS 1

arch/x86/include/asm/hw_irq.h

@@ -141,6 +141,7 @@ struct irq_alloc_info {
 struct irq_cfg {
 	unsigned int		dest_apicid;
 	u8			vector;
+	u8			old_vector;
 };
 
 extern struct irq_cfg *irq_cfg(unsigned int irq);
@@ -168,20 +169,6 @@ extern atomic_t irq_mis_count;
 
 extern void elcr_set_level_irq(unsigned int irq);
 
-/* SMP */
-extern __visible void smp_apic_timer_interrupt(struct pt_regs *);
-extern __visible void smp_spurious_interrupt(struct pt_regs *);
-extern __visible void smp_x86_platform_ipi(struct pt_regs *);
-extern __visible void smp_error_interrupt(struct pt_regs *);
-#ifdef CONFIG_X86_IO_APIC
-extern asmlinkage void smp_irq_move_cleanup_interrupt(void);
-#endif
-#ifdef CONFIG_SMP
-extern __visible void smp_reschedule_interrupt(struct pt_regs *);
-extern __visible void smp_call_function_interrupt(struct pt_regs *);
-extern __visible void smp_call_function_single_interrupt(struct pt_regs *);
-#endif
-
 extern char irq_entries_start[];
 #ifdef CONFIG_TRACING
 #define trace_irq_entries_start irq_entries_start

arch/x86/include/asm/msr.h

@@ -42,14 +42,6 @@ struct saved_msrs {
 	struct saved_msr *array;
 };
 
-static inline unsigned long long native_read_tscp(unsigned int *aux)
-{
-	unsigned long low, high;
-	asm volatile(".byte 0x0f,0x01,0xf9"
-		     : "=a" (low), "=d" (high), "=c" (*aux));
-	return low | ((u64)high << 32);
-}
-
 /*
  * both i386 and x86_64 returns 64-bit value in edx:eax, but gcc's "A"
  * constraint has different meanings. For i386, "A" means exactly

arch/x86/include/asm/rwsem.h

@@ -25,7 +25,7 @@
  * This should be totally fair - if anything is waiting, a process that wants a
  * lock will go to the back of the queue. When the currently active lock is
  * released, if there's a writer at the front of the queue, then that and only
- * that will be woken up; if there's a bunch of consequtive readers at the
+ * that will be woken up; if there's a bunch of consecutive readers at the
  * front, then they'll all be woken up, but no other readers will be.
  */
 

arch/x86/include/asm/string_64.h

@@ -87,9 +87,9 @@ int strcmp(const char *cs, const char *ct);
  *
  * Low level memory copy function that catches machine checks
  *
- * Return true for success, false for fail
+ * Return 0 for success, -EFAULT for fail
  */
-bool memcpy_mcsafe(void *dst, const void *src, size_t cnt);
+int memcpy_mcsafe(void *dst, const void *src, size_t cnt);
 
 #endif /* __KERNEL__ */
 

arch/x86/include/asm/xen/hypervisor.h

@@ -62,4 +62,6 @@ void xen_arch_register_cpu(int num);
 void xen_arch_unregister_cpu(int num);
 #endif
 
+extern void xen_set_iopl_mask(unsigned mask);
+
 #endif /* _ASM_X86_XEN_HYPERVISOR_H */

arch/x86/kernel/acpi/boot.c

@@ -956,7 +956,7 @@ static int __init early_acpi_parse_madt_lapic_addr_ovr(void)
 
 	/*
 	 * Note that the LAPIC address is obtained from the MADT (32-bit value)
-	 * and (optionally) overriden by a LAPIC_ADDR_OVR entry (64-bit value).
+	 * and (optionally) overridden by a LAPIC_ADDR_OVR entry (64-bit value).
 	 */
 
 	count = acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_APIC_OVERRIDE,
@@ -984,7 +984,7 @@ static int __init acpi_parse_madt_lapic_entries(void)
 
 	/*
 	 * Note that the LAPIC address is obtained from the MADT (32-bit value)
-	 * and (optionally) overriden by a LAPIC_ADDR_OVR entry (64-bit value).
+	 * and (optionally) overridden by a LAPIC_ADDR_OVR entry (64-bit value).
 	 */
 
 	count = acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_APIC_OVERRIDE,

arch/x86/kernel/apb_timer.c

@@ -221,7 +221,7 @@ static int apbt_cpuhp_notify(struct notifier_block *n,
 	unsigned long cpu = (unsigned long)hcpu;
 	struct apbt_dev *adev = &per_cpu(cpu_apbt_dev, cpu);
 
-	switch (action & 0xf) {
+	switch (action & ~CPU_TASKS_FROZEN) {
 	case CPU_DEAD:
 		dw_apb_clockevent_pause(adev->timer);
 		if (system_state == SYSTEM_RUNNING) {

arch/x86/kernel/apic/apic.c

@@ -1611,7 +1611,7 @@ void __init enable_IR_x2apic(void)
 	legacy_pic->mask_all();
 	mask_ioapic_entries();
 
-	/* If irq_remapping_prepare() succeded, try to enable it */
+	/* If irq_remapping_prepare() succeeded, try to enable it */
 	if (ir_stat >= 0)
 		ir_stat = try_to_enable_IR();
 	/* ir_stat contains the remap mode or an error code */

arch/x86/kernel/apic/vector.c

@@ -213,6 +213,7 @@ static int __assign_irq_vector(int irq, struct apic_chip_data *d,
 	 */
 	cpumask_and(d->old_domain, d->old_domain, cpu_online_mask);
 	d->move_in_progress = !cpumask_empty(d->old_domain);
+	d->cfg.old_vector = d->move_in_progress ? d->cfg.vector : 0;
 	d->cfg.vector = vector;
 	cpumask_copy(d->domain, vector_cpumask);
 success:
@@ -655,46 +656,97 @@ void irq_complete_move(struct irq_cfg *cfg)
 }
 
 /*
- * Called with @desc->lock held and interrupts disabled.
+ * Called from fixup_irqs() with @desc->lock held and interrupts disabled.
  */
 void irq_force_complete_move(struct irq_desc *desc)
 {
 	struct irq_data *irqdata = irq_desc_get_irq_data(desc);
 	struct apic_chip_data *data = apic_chip_data(irqdata);
 	struct irq_cfg *cfg = data ? &data->cfg : NULL;
+	unsigned int cpu;
 
 	if (!cfg)
 		return;
 
-	__irq_complete_move(cfg, cfg->vector);
-
 	/*
 	 * This is tricky. If the cleanup of @data->old_domain has not been
 	 * done yet, then the following setaffinity call will fail with
 	 * -EBUSY. This can leave the interrupt in a stale state.
 	 *
-	 * The cleanup cannot make progress because we hold @desc->lock. So in
-	 * case @data->old_domain is not yet cleaned up, we need to drop the
-	 * lock and acquire it again. @desc cannot go away, because the
-	 * hotplug code holds the sparse irq lock.
+	 * All CPUs are stuck in stop machine with interrupts disabled so
+	 * calling __irq_complete_move() would be completely pointless.
 	 */
 	raw_spin_lock(&vector_lock);
-	/* Clean out all offline cpus (including ourself) first. */
+	/*
+	 * Clean out all offline cpus (including the outgoing one) from the
+	 * old_domain mask.
+	 */
 	cpumask_and(data->old_domain, data->old_domain, cpu_online_mask);
-	while (!cpumask_empty(data->old_domain)) {
+
+	/*
+	 * If move_in_progress is cleared and the old_domain mask is empty,
+	 * then there is nothing to cleanup. fixup_irqs() will take care of
+	 * the stale vectors on the outgoing cpu.
+	 */
+	if (!data->move_in_progress && cpumask_empty(data->old_domain)) {
 		raw_spin_unlock(&vector_lock);
-		raw_spin_unlock(&desc->lock);
-		cpu_relax();
-		raw_spin_lock(&desc->lock);
+		return;
+	}
+
+	/*
+	 * 1) The interrupt is in move_in_progress state. That means that we
+	 *    have not seen an interrupt since the io_apic was reprogrammed to
+	 *    the new vector.
+	 *
+	 * 2) The interrupt has fired on the new vector, but the cleanup IPIs
+	 *    have not been processed yet.
+	 */
+	if (data->move_in_progress) {
 		/*
-		 * Reevaluate apic_chip_data. It might have been cleared after
-		 * we dropped @desc->lock.
+		 * In theory there is a race:
+		 *
+		 * set_ioapic(new_vector) <-- Interrupt is raised before update
+		 *			      is effective, i.e. it's raised on
+		 *			      the old vector.
+		 *
+		 * So if the target cpu cannot handle that interrupt before
+		 * the old vector is cleaned up, we get a spurious interrupt
+		 * and in the worst case the ioapic irq line becomes stale.
+		 *
+		 * But in case of cpu hotplug this should be a non issue
+		 * because if the affinity update happens right before all
+		 * cpus rendevouz in stop machine, there is no way that the
+		 * interrupt can be blocked on the target cpu because all cpus
+		 * loops first with interrupts enabled in stop machine, so the
+		 * old vector is not yet cleaned up when the interrupt fires.
+		 *
+		 * So the only way to run into this issue is if the delivery
+		 * of the interrupt on the apic/system bus would be delayed
+		 * beyond the point where the target cpu disables interrupts
+		 * in stop machine. I doubt that it can happen, but at least
+		 * there is a theroretical chance. Virtualization might be
+		 * able to expose this, but AFAICT the IOAPIC emulation is not
+		 * as stupid as the real hardware.
+		 *
+		 * Anyway, there is nothing we can do about that at this point
+		 * w/o refactoring the whole fixup_irq() business completely.
+		 * We print at least the irq number and the old vector number,
+		 * so we have the necessary information when a problem in that
+		 * area arises.
 		 */
-		data = apic_chip_data(irqdata);
-		if (!data)
-			return;
-		raw_spin_lock(&vector_lock);
+		pr_warn("IRQ fixup: irq %d move in progress, old vector %d\n",
+			irqdata->irq, cfg->old_vector);
 	}
+	/*
+	 * If old_domain is not empty, then other cpus still have the irq
+	 * descriptor set in their vector array. Clean it up.
+	 */
+	for_each_cpu(cpu, data->old_domain)
+		per_cpu(vector_irq, cpu)[cfg->old_vector] = VECTOR_UNUSED;
+
+	/* Cleanup the left overs of the (half finished) move */
+	cpumask_clear(data->old_domain);
+	data->move_in_progress = 0;
 	raw_spin_unlock(&vector_lock);
 }
 #endif

arch/x86/kernel/apic/x2apic_uv_x.c

@@ -792,7 +792,8 @@ static int uv_scir_cpu_notify(struct notifier_block *self, unsigned long action,
 {
 	long cpu = (long)hcpu;
 
-	switch (action) {
+	switch (action & ~CPU_TASKS_FROZEN) {
+	case CPU_DOWN_FAILED:
 	case CPU_ONLINE:
 		uv_heartbeat_enable(cpu);
 		break;
@@ -860,7 +861,7 @@ int uv_set_vga_state(struct pci_dev *pdev, bool decode,
  */
 void uv_cpu_init(void)
 {
-	/* CPU 0 initilization will be done via uv_system_init. */
+	/* CPU 0 initialization will be done via uv_system_init. */
 	if (!uv_blade_info)
 		return;
 

arch/x86/kernel/apm_32.c

@@ -1088,7 +1088,7 @@ static int apm_get_battery_status(u_short which, u_short *status,
  *	@device: identity of device
  *	@enable: on/off
  *
- *	Activate or deactive power management on either a specific device
+ *	Activate or deactivate power management on either a specific device
  *	or the entire system (%APM_DEVICE_ALL).
  */
 

arch/x86/kernel/cpu/amd.c

@@ -85,7 +85,7 @@ static void init_amd_k5(struct cpuinfo_x86 *c)
 #ifdef CONFIG_X86_32
 /*
  * General Systems BIOSen alias the cpu frequency registers
- * of the Elan at 0x000df000. Unfortuantly, one of the Linux
+ * of the Elan at 0x000df000. Unfortunately, one of the Linux
  * drivers subsequently pokes it, and changes the CPU speed.
  * Workaround : Remove the unneeded alias.
  */

arch/x86/kernel/cpu/common.c

@@ -968,7 +968,7 @@ static void identify_cpu(struct cpuinfo_x86 *c)
 	if (this_cpu->c_identify)
 		this_cpu->c_identify(c);
 
-	/* Clear/Set all flags overriden by options, after probe */
+	/* Clear/Set all flags overridden by options, after probe */
 	for (i = 0; i < NCAPINTS; i++) {
 		c->x86_capability[i] &= ~cpu_caps_cleared[i];
 		c->x86_capability[i] |= cpu_caps_set[i];
@@ -1028,7 +1028,7 @@ static void identify_cpu(struct cpuinfo_x86 *c)
 	setup_pku(c);
 
 	/*
-	 * Clear/Set all flags overriden by options, need do it
+	 * Clear/Set all flags overridden by options, need do it
 	 * before following smp all cpus cap AND.
 	 */
 	for (i = 0; i < NCAPINTS; i++) {

arch/x86/kernel/cpu/mtrr/generic.c

@@ -42,7 +42,7 @@ EXPORT_SYMBOL_GPL(mtrr_state);
  * "BIOS and Kernel Developer's Guide for the AMD Athlon 64 and AMD
  * Opteron Processors" (26094 Rev. 3.30 February 2006), section
  * "13.2.1.2 SYSCFG Register": "The MtrrFixDramModEn bit should be set
- * to 1 during BIOS initalization of the fixed MTRRs, then cleared to
+ * to 1 during BIOS initialization of the fixed MTRRs, then cleared to
  * 0 for operation."
  */
 static inline void k8_check_syscfg_dram_mod_en(void)

arch/x86/kernel/early_printk.c

@@ -287,7 +287,7 @@ static __init void early_pci_serial_init(char *s)
 	}
 
 	/*
-	 * Lastly, initalize the hardware
+	 * Lastly, initialize the hardware
 	 */
 	if (*s) {
 		if (strcmp(s, "nocfg") == 0)

arch/x86/kernel/fpu/regset.c

@@ -8,7 +8,7 @@
 /*
  * The xstateregs_active() routine is the same as the regset_fpregs_active() routine,
  * as the "regset->n" for the xstate regset will be updated based on the feature
- * capabilites supported by the xsave.
+ * capabilities supported by the xsave.
  */
 int regset_fpregs_active(struct task_struct *target, const struct user_regset *regset)
 {

arch/x86/kernel/hpet.c

@@ -717,7 +717,7 @@ static int hpet_cpuhp_notify(struct notifier_block *n,
 	struct hpet_work_struct work;
 	struct hpet_dev *hdev = per_cpu(cpu_hpet_dev, cpu);
 
-	switch (action & 0xf) {
+	switch (action & ~CPU_TASKS_FROZEN) {
 	case CPU_ONLINE:
 		INIT_DELAYED_WORK_ONSTACK(&work.work, hpet_work);
 		init_completion(&work.complete);

arch/x86/kernel/ioport.c

@@ -96,18 +96,24 @@ asmlinkage long sys_ioperm(unsigned long from, unsigned long num, int turn_on)
 SYSCALL_DEFINE1(iopl, unsigned int, level)
 {
 	struct pt_regs *regs = current_pt_regs();
-	unsigned int old = (regs->flags >> 12) & 3;
 	struct thread_struct *t = &current->thread;
 
+	/*
+	 * Careful: the IOPL bits in regs->flags are undefined under Xen PV
+	 * and changing them has no effect.
+	 */
+	unsigned int old = t->iopl >> X86_EFLAGS_IOPL_BIT;
+
 	if (level > 3)
 		return -EINVAL;
 	/* Trying to gain more privileges? */
 	if (level > old) {
 		if (!capable(CAP_SYS_RAWIO))
 			return -EPERM;
 	}
-	regs->flags = (regs->flags & ~X86_EFLAGS_IOPL) | (level << 12);
-	t->iopl = level << 12;
+	regs->flags = (regs->flags & ~X86_EFLAGS_IOPL) |
+		(level << X86_EFLAGS_IOPL_BIT);
+	t->iopl = level << X86_EFLAGS_IOPL_BIT;
 	set_iopl_mask(t->iopl);
 
 	return 0;