Merge branch 'x86-apic-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull x86 apic updates from Ingo Molnar:
 "The main changes in this cycle were:

   - introduce optimized single IPI sending methods on modern APICs
     (Linus Torvalds, Thomas Gleixner)

   - kexec/crash APIC handling fixes and enhancements (Hidehiro Kawai)

   - extend lapic vector saving/restoring to the CMCI (MCE) vector as
     well (Juergen Gross)

   - various fixes and enhancements (Jake Oshins, Len Brown)"

* 'x86-apic-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (23 commits)
  x86/irq: Export functions to allow MSI domains in modules
  Documentation: Document kernel.panic_on_io_nmi sysctl
  x86/nmi: Save regs in crash dump on external NMI
  x86/apic: Introduce apic_extnmi command line parameter
  kexec: Fix race between panic() and crash_kexec()
  panic, x86: Allow CPUs to save registers even if looping in NMI context
  panic, x86: Fix re-entrance problem due to panic on NMI
  x86/apic: Fix the saving and restoring of lapic vectors during suspend/resume
  x86/smpboot: Re-enable init_udelay=0 by default on modern CPUs
  x86/smp: Remove single IPI wrapper
  x86/apic: Use default send single IPI wrapper
  x86/apic: Provide default send single IPI wrapper
  x86/apic: Implement single IPI for apic_noop
  x86/apic: Wire up single IPI for apic_numachip
  x86/apic: Wire up single IPI for x2apic_uv
  x86/apic: Implement single IPI for x2apic_phys
  x86/apic: Wire up single IPI for bigsmp_apic
  x86/apic: Remove pointless indirections from bigsmp_apic
  x86/apic: Wire up single IPI for apic_physflat
  x86/apic: Remove pointless indirections from apic_physflat
  ...

Documentation/kernel-parameters.txt

@@ -472,6 +472,15 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
 			Change the amount of debugging information output
 			when initialising the APIC and IO-APIC components.
 
+	apic_extnmi=	[APIC,X86] External NMI delivery setting
+			Format: { bsp (default) | all | none }
+			bsp:  External NMI is delivered only to CPU 0
+			all:  External NMIs are broadcast to all CPUs as a
+			      backup of CPU 0
+			none: External NMI is masked for all CPUs. This is
+			      useful so that a dump capture kernel won't be
+			      shot down by NMI
+
 	autoconf=	[IPV6]
 			See Documentation/networking/ipv6.txt.
 

Documentation/sysctl/kernel.txt

@@ -551,6 +551,21 @@ the recommended setting is 60.
 
 ==============================================================
 
+panic_on_io_nmi:
+
+Controls the kernel's behavior when a CPU receives an NMI caused by
+an IO error.
+
+0: try to continue operation (default)
+
+1: panic immediately. The IO error triggered an NMI. This indicates a
+   serious system condition which could result in IO data corruption.
+   Rather than continuing, panicking might be a better choice. Some
+   servers issue this sort of NMI when the dump button is pushed,
+   and you can use this option to take a crash dump.
+
+==============================================================
+
 panic_on_oops:
 
 Controls the kernel's behaviour when an oops or BUG is encountered.

arch/x86/include/asm/apic.h

@@ -23,6 +23,11 @@
 #define APIC_VERBOSE 1
 #define APIC_DEBUG   2
 
+/* Macros for apic_extnmi which controls external NMI masking */
+#define APIC_EXTNMI_BSP		0 /* Default */
+#define APIC_EXTNMI_ALL		1
+#define APIC_EXTNMI_NONE	2
+
 /*
  * Define the default level of output to be very little
  * This can be turned up by using apic=verbose for more
@@ -303,6 +308,7 @@ struct apic {
 				      unsigned int *apicid);
 
 	/* ipi */
+	void (*send_IPI)(int cpu, int vector);
 	void (*send_IPI_mask)(const struct cpumask *mask, int vector);
 	void (*send_IPI_mask_allbutself)(const struct cpumask *mask,
 					 int vector);

arch/x86/include/asm/ipi.h

@@ -119,6 +119,8 @@ static inline void
 	native_apic_mem_write(APIC_ICR, cfg);
 }
 
+extern void default_send_IPI_single(int cpu, int vector);
+extern void default_send_IPI_single_phys(int cpu, int vector);
 extern void default_send_IPI_mask_sequence_phys(const struct cpumask *mask,
 						 int vector);
 extern void default_send_IPI_mask_allbutself_phys(const struct cpumask *mask,

arch/x86/include/asm/msi.h

 #ifndef _ASM_X86_MSI_H
 #define _ASM_X86_MSI_H
 #include <asm/hw_irq.h>
+#include <asm/irqdomain.h>
 
 typedef struct irq_alloc_info msi_alloc_info_t;
 
+int pci_msi_prepare(struct irq_domain *domain, struct device *dev, int nvec,
+		    msi_alloc_info_t *arg);
+
+void pci_msi_set_desc(msi_alloc_info_t *arg, struct msi_desc *desc);
+
 #endif /* _ASM_X86_MSI_H */

arch/x86/include/asm/reboot.h

@@ -25,5 +25,6 @@ void __noreturn machine_real_restart(unsigned int type);
 
 typedef void (*nmi_shootdown_cb)(int, struct pt_regs*);
 void nmi_shootdown_cpus(nmi_shootdown_cb callback);
+void run_crash_ipi_callback(struct pt_regs *regs);
 
 #endif /* _ASM_X86_REBOOT_H */

arch/x86/kernel/apic/apic.c

@@ -81,6 +81,12 @@ physid_mask_t phys_cpu_present_map;
  */
 static unsigned int disabled_cpu_apicid __read_mostly = BAD_APICID;
 
+/*
+ * This variable controls which CPUs receive external NMIs.  By default,
+ * external NMIs are delivered only to the BSP.
+ */
+static int apic_extnmi = APIC_EXTNMI_BSP;
+
 /*
  * Map cpu index to physical APIC ID
  */
@@ -1161,6 +1167,8 @@ void __init init_bsp_APIC(void)
 	value = APIC_DM_NMI;
 	if (!lapic_is_integrated())		/* 82489DX */
 		value |= APIC_LVT_LEVEL_TRIGGER;
+	if (apic_extnmi == APIC_EXTNMI_NONE)
+		value |= APIC_LVT_MASKED;
 	apic_write(APIC_LVT1, value);
 }
 
@@ -1378,9 +1386,11 @@ void setup_local_APIC(void)
 	apic_write(APIC_LVT0, value);
 
 	/*
-	 * only the BP should see the LINT1 NMI signal, obviously.
+	 * Only the BSP sees the LINT1 NMI signal by default. This can be
+	 * modified by apic_extnmi= boot option.
 	 */
-	if (!cpu)
+	if ((!cpu && apic_extnmi != APIC_EXTNMI_NONE) ||
+	    apic_extnmi == APIC_EXTNMI_ALL)
 		value = APIC_DM_NMI;
 	else
 		value = APIC_DM_NMI | APIC_LVT_MASKED;
@@ -2270,6 +2280,7 @@ static struct {
 	unsigned int apic_tmict;
 	unsigned int apic_tdcr;
 	unsigned int apic_thmr;
+	unsigned int apic_cmci;
 } apic_pm_state;
 
 static int lapic_suspend(void)
@@ -2299,6 +2310,10 @@ static int lapic_suspend(void)
 	if (maxlvt >= 5)
 		apic_pm_state.apic_thmr = apic_read(APIC_LVTTHMR);
 #endif
+#ifdef CONFIG_X86_MCE_INTEL
+	if (maxlvt >= 6)
+		apic_pm_state.apic_cmci = apic_read(APIC_LVTCMCI);
+#endif
 
 	local_irq_save(flags);
 	disable_local_APIC();
@@ -2355,9 +2370,13 @@ static void lapic_resume(void)
 	apic_write(APIC_SPIV, apic_pm_state.apic_spiv);
 	apic_write(APIC_LVT0, apic_pm_state.apic_lvt0);
 	apic_write(APIC_LVT1, apic_pm_state.apic_lvt1);
-#if defined(CONFIG_X86_MCE_INTEL)
+#ifdef CONFIG_X86_THERMAL_VECTOR
 	if (maxlvt >= 5)
 		apic_write(APIC_LVTTHMR, apic_pm_state.apic_thmr);
+#endif
+#ifdef CONFIG_X86_MCE_INTEL
+	if (maxlvt >= 6)
+		apic_write(APIC_LVTCMCI, apic_pm_state.apic_cmci);
 #endif
 	if (maxlvt >= 4)
 		apic_write(APIC_LVTPC, apic_pm_state.apic_lvtpc);
@@ -2548,3 +2567,23 @@ static int __init apic_set_disabled_cpu_apicid(char *arg)
 	return 0;
 }
 early_param("disable_cpu_apicid", apic_set_disabled_cpu_apicid);
+
+static int __init apic_set_extnmi(char *arg)
+{
+	if (!arg)
+		return -EINVAL;
+
+	if (!strncmp("all", arg, 3))
+		apic_extnmi = APIC_EXTNMI_ALL;
+	else if (!strncmp("none", arg, 4))
+		apic_extnmi = APIC_EXTNMI_NONE;
+	else if (!strncmp("bsp", arg, 3))
+		apic_extnmi = APIC_EXTNMI_BSP;
+	else {
+		pr_warn("Unknown external NMI delivery mode `%s' ignored\n", arg);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+early_param("apic_extnmi", apic_set_extnmi);

arch/x86/kernel/apic/apic_flat_64.c

@@ -185,6 +185,7 @@ static struct apic apic_flat =  {
 
 	.cpu_mask_to_apicid_and		= flat_cpu_mask_to_apicid_and,
 
+	.send_IPI			= default_send_IPI_single,
 	.send_IPI_mask			= flat_send_IPI_mask,
 	.send_IPI_mask_allbutself	= flat_send_IPI_mask_allbutself,
 	.send_IPI_allbutself		= flat_send_IPI_allbutself,
@@ -230,17 +231,6 @@ static int physflat_acpi_madt_oem_check(char *oem_id, char *oem_table_id)
 	return 0;
 }
 
-static void physflat_send_IPI_mask(const struct cpumask *cpumask, int vector)
-{
-	default_send_IPI_mask_sequence_phys(cpumask, vector);
-}
-
-static void physflat_send_IPI_mask_allbutself(const struct cpumask *cpumask,
-					      int vector)
-{
-	default_send_IPI_mask_allbutself_phys(cpumask, vector);
-}
-
 static void physflat_send_IPI_allbutself(int vector)
 {
 	default_send_IPI_mask_allbutself_phys(cpu_online_mask, vector);
@@ -248,7 +238,7 @@ static void physflat_send_IPI_allbutself(int vector)
 
 static void physflat_send_IPI_all(int vector)
 {
-	physflat_send_IPI_mask(cpu_online_mask, vector);
+	default_send_IPI_mask_sequence_phys(cpu_online_mask, vector);
 }
 
 static int physflat_probe(void)
@@ -292,8 +282,9 @@ static struct apic apic_physflat =  {
 
 	.cpu_mask_to_apicid_and		= default_cpu_mask_to_apicid_and,
 
-	.send_IPI_mask			= physflat_send_IPI_mask,
-	.send_IPI_mask_allbutself	= physflat_send_IPI_mask_allbutself,
+	.send_IPI			= default_send_IPI_single_phys,
+	.send_IPI_mask			= default_send_IPI_mask_sequence_phys,
+	.send_IPI_mask_allbutself	= default_send_IPI_mask_allbutself_phys,
 	.send_IPI_allbutself		= physflat_send_IPI_allbutself,
 	.send_IPI_all			= physflat_send_IPI_all,
 	.send_IPI_self			= apic_send_IPI_self,

arch/x86/kernel/apic/apic_noop.c

@@ -30,6 +30,7 @@
 #include <asm/e820.h>
 
 static void noop_init_apic_ldr(void) { }
+static void noop_send_IPI(int cpu, int vector) { }
 static void noop_send_IPI_mask(const struct cpumask *cpumask, int vector) { }
 static void noop_send_IPI_mask_allbutself(const struct cpumask *cpumask, int vector) { }
 static void noop_send_IPI_allbutself(int vector) { }
@@ -144,6 +145,7 @@ struct apic apic_noop = {
 
 	.cpu_mask_to_apicid_and		= flat_cpu_mask_to_apicid_and,
 
+	.send_IPI			= noop_send_IPI,
 	.send_IPI_mask			= noop_send_IPI_mask,
 	.send_IPI_mask_allbutself	= noop_send_IPI_mask_allbutself,
 	.send_IPI_allbutself		= noop_send_IPI_allbutself,

arch/x86/kernel/apic/apic_numachip.c

@@ -273,6 +273,7 @@ static const struct apic apic_numachip1 __refconst = {
 
 	.cpu_mask_to_apicid_and		= default_cpu_mask_to_apicid_and,
 
+	.send_IPI			= numachip_send_IPI_one,
 	.send_IPI_mask			= numachip_send_IPI_mask,
 	.send_IPI_mask_allbutself	= numachip_send_IPI_mask_allbutself,
 	.send_IPI_allbutself		= numachip_send_IPI_allbutself,
@@ -324,6 +325,7 @@ static const struct apic apic_numachip2 __refconst = {
 
 	.cpu_mask_to_apicid_and		= default_cpu_mask_to_apicid_and,
 
+	.send_IPI			= numachip_send_IPI_one,
 	.send_IPI_mask			= numachip_send_IPI_mask,
 	.send_IPI_mask_allbutself	= numachip_send_IPI_mask_allbutself,
 	.send_IPI_allbutself		= numachip_send_IPI_allbutself,

arch/x86/kernel/apic/bigsmp_32.c

@@ -96,11 +96,6 @@ static int bigsmp_phys_pkg_id(int cpuid_apic, int index_msb)
 	return cpuid_apic >> index_msb;
 }
 
-static inline void bigsmp_send_IPI_mask(const struct cpumask *mask, int vector)
-{
-	default_send_IPI_mask_sequence_phys(mask, vector);
-}
-
 static void bigsmp_send_IPI_allbutself(int vector)
 {
 	default_send_IPI_mask_allbutself_phys(cpu_online_mask, vector);
@@ -108,7 +103,7 @@ static void bigsmp_send_IPI_allbutself(int vector)
 
 static void bigsmp_send_IPI_all(int vector)
 {
-	bigsmp_send_IPI_mask(cpu_online_mask, vector);
+	default_send_IPI_mask_sequence_phys(cpu_online_mask, vector);
 }
 
 static int dmi_bigsmp; /* can be set by dmi scanners */
@@ -180,7 +175,8 @@ static struct apic apic_bigsmp = {
 
 	.cpu_mask_to_apicid_and		= default_cpu_mask_to_apicid_and,
 
-	.send_IPI_mask			= bigsmp_send_IPI_mask,
+	.send_IPI			= default_send_IPI_single_phys,
+	.send_IPI_mask			= default_send_IPI_mask_sequence_phys,
 	.send_IPI_mask_allbutself	= NULL,
 	.send_IPI_allbutself		= bigsmp_send_IPI_allbutself,
 	.send_IPI_all			= bigsmp_send_IPI_all,

arch/x86/kernel/apic/ipi.c

@@ -18,6 +18,16 @@
 #include <asm/proto.h>
 #include <asm/ipi.h>
 
+void default_send_IPI_single_phys(int cpu, int vector)
+{
+	unsigned long flags;
+
+	local_irq_save(flags);
+	__default_send_IPI_dest_field(per_cpu(x86_cpu_to_apicid, cpu),
+				      vector, APIC_DEST_PHYSICAL);
+	local_irq_restore(flags);
+}
+
 void default_send_IPI_mask_sequence_phys(const struct cpumask *mask, int vector)
 {
 	unsigned long query_cpu;
@@ -55,6 +65,14 @@ void default_send_IPI_mask_allbutself_phys(const struct cpumask *mask,
 	local_irq_restore(flags);
 }
 
+/*
+ * Helper function for APICs which insist on cpumasks
+ */
+void default_send_IPI_single(int cpu, int vector)
+{
+	apic->send_IPI_mask(cpumask_of(cpu), vector);
+}
+
 #ifdef CONFIG_X86_32
 
 void default_send_IPI_mask_sequence_logical(const struct cpumask *mask,

arch/x86/kernel/apic/msi.c

@@ -96,8 +96,8 @@ static irq_hw_number_t pci_msi_get_hwirq(struct msi_domain_info *info,
 	return arg->msi_hwirq;
 }
 
-static int pci_msi_prepare(struct irq_domain *domain, struct device *dev,
-			   int nvec, msi_alloc_info_t *arg)
+int pci_msi_prepare(struct irq_domain *domain, struct device *dev, int nvec,
+		    msi_alloc_info_t *arg)
 {
 	struct pci_dev *pdev = to_pci_dev(dev);
 	struct msi_desc *desc = first_pci_msi_entry(pdev);
@@ -113,11 +113,13 @@ static int pci_msi_prepare(struct irq_domain *domain, struct device *dev,
 
 	return 0;
 }
+EXPORT_SYMBOL_GPL(pci_msi_prepare);
 
-static void pci_msi_set_desc(msi_alloc_info_t *arg, struct msi_desc *desc)
+void pci_msi_set_desc(msi_alloc_info_t *arg, struct msi_desc *desc)
 {
 	arg->msi_hwirq = pci_msi_domain_calc_hwirq(arg->msi_dev, desc);
 }
+EXPORT_SYMBOL_GPL(pci_msi_set_desc);
 
 static struct msi_domain_ops pci_msi_domain_ops = {
 	.get_hwirq	= pci_msi_get_hwirq,

arch/x86/kernel/apic/probe_32.c

@@ -105,6 +105,7 @@ static struct apic apic_default = {
 
 	.cpu_mask_to_apicid_and		= flat_cpu_mask_to_apicid_and,
 
+	.send_IPI			= default_send_IPI_single,
 	.send_IPI_mask			= default_send_IPI_mask_logical,
 	.send_IPI_mask_allbutself	= default_send_IPI_mask_allbutself_logical,
 	.send_IPI_allbutself		= default_send_IPI_allbutself,

arch/x86/kernel/apic/vector.c

@@ -29,6 +29,7 @@ struct apic_chip_data {
 };
 
 struct irq_domain *x86_vector_domain;
+EXPORT_SYMBOL_GPL(x86_vector_domain);
 static DEFINE_RAW_SPINLOCK(vector_lock);
 static cpumask_var_t vector_cpumask;
 static struct irq_chip lapic_controller;
@@ -66,6 +67,7 @@ struct irq_cfg *irqd_cfg(struct irq_data *irq_data)
 
 	return data ? &data->cfg : NULL;
 }
+EXPORT_SYMBOL_GPL(irqd_cfg);
 
 struct irq_cfg *irq_cfg(unsigned int irq)
 {

arch/x86/kernel/apic/x2apic_cluster.c

@@ -23,6 +23,14 @@ static inline u32 x2apic_cluster(int cpu)
 	return per_cpu(x86_cpu_to_logical_apicid, cpu) >> 16;
 }
 
+static void x2apic_send_IPI(int cpu, int vector)
+{
+	u32 dest = per_cpu(x86_cpu_to_logical_apicid, cpu);
+
+	x2apic_wrmsr_fence();
+	__x2apic_send_IPI_dest(dest, vector, APIC_DEST_LOGICAL);
+}
+
 static void
 __x2apic_send_IPI_mask(const struct cpumask *mask, int vector, int apic_dest)
 {
@@ -266,6 +274,7 @@ static struct apic apic_x2apic_cluster = {
 
 	.cpu_mask_to_apicid_and		= x2apic_cpu_mask_to_apicid_and,
 
+	.send_IPI			= x2apic_send_IPI,
 	.send_IPI_mask			= x2apic_send_IPI_mask,
 	.send_IPI_mask_allbutself	= x2apic_send_IPI_mask_allbutself,
 	.send_IPI_allbutself		= x2apic_send_IPI_allbutself,

arch/x86/kernel/apic/x2apic_phys.c

@@ -36,6 +36,14 @@ static int x2apic_acpi_madt_oem_check(char *oem_id, char *oem_table_id)
 	return x2apic_enabled() && (x2apic_phys || x2apic_fadt_phys());
 }
 
+static void x2apic_send_IPI(int cpu, int vector)
+{
+	u32 dest = per_cpu(x86_cpu_to_apicid, cpu);
+
+	x2apic_wrmsr_fence();
+	__x2apic_send_IPI_dest(dest, vector, APIC_DEST_PHYSICAL);
+}
+
 static void
 __x2apic_send_IPI_mask(const struct cpumask *mask, int vector, int apic_dest)
 {
@@ -122,6 +130,7 @@ static struct apic apic_x2apic_phys = {
 
 	.cpu_mask_to_apicid_and		= default_cpu_mask_to_apicid_and,
 
+	.send_IPI			= x2apic_send_IPI,
 	.send_IPI_mask			= x2apic_send_IPI_mask,
 	.send_IPI_mask_allbutself	= x2apic_send_IPI_mask_allbutself,
 	.send_IPI_allbutself		= x2apic_send_IPI_allbutself,

arch/x86/kernel/apic/x2apic_uv_x.c

@@ -406,6 +406,7 @@ static struct apic __refdata apic_x2apic_uv_x = {
 
 	.cpu_mask_to_apicid_and		= uv_cpu_mask_to_apicid_and,
 
+	.send_IPI			= uv_send_IPI_one,
 	.send_IPI_mask			= uv_send_IPI_mask,
 	.send_IPI_mask_allbutself	= uv_send_IPI_mask_allbutself,
 	.send_IPI_allbutself		= uv_send_IPI_allbutself,

arch/x86/kernel/nmi.c

@@ -29,6 +29,7 @@
 #include <asm/mach_traps.h>
 #include <asm/nmi.h>
 #include <asm/x86_init.h>
+#include <asm/reboot.h>
 
 #define CREATE_TRACE_POINTS
 #include <trace/events/nmi.h>
@@ -231,7 +232,7 @@ pci_serr_error(unsigned char reason, struct pt_regs *regs)
 #endif
 
 	if (panic_on_unrecovered_nmi)
-		panic("NMI: Not continuing");
+		nmi_panic(regs, "NMI: Not continuing");
 
 	pr_emerg("Dazed and confused, but trying to continue\n");
 
@@ -255,8 +256,16 @@ io_check_error(unsigned char reason, struct pt_regs *regs)
 		 reason, smp_processor_id());
 	show_regs(regs);
 
-	if (panic_on_io_nmi)
-		panic("NMI IOCK error: Not continuing");
+	if (panic_on_io_nmi) {
+		nmi_panic(regs, "NMI IOCK error: Not continuing");
+
+		/*
+		 * If we end up here, it means we have received an NMI while
+		 * processing panic(). Simply return without delaying and
+		 * re-enabling NMIs.
+		 */
+		return;
+	}
 
 	/* Re-enable the IOCK line, wait for a few seconds */
 	reason = (reason & NMI_REASON_CLEAR_MASK) | NMI_REASON_CLEAR_IOCHK;
@@ -297,7 +306,7 @@ unknown_nmi_error(unsigned char reason, struct pt_regs *regs)
 
 	pr_emerg("Do you have a strange power saving mode enabled?\n");
 	if (unknown_nmi_panic || panic_on_unrecovered_nmi)
-		panic("NMI: Not continuing");
+		nmi_panic(regs, "NMI: Not continuing");
 
 	pr_emerg("Dazed and confused, but trying to continue\n");
 }
@@ -348,8 +357,19 @@ static void default_do_nmi(struct pt_regs *regs)
 		return;
 	}
 
-	/* Non-CPU-specific NMI: NMI sources can be processed on any CPU */
-	raw_spin_lock(&nmi_reason_lock);
+	/*
+	 * Non-CPU-specific NMI: NMI sources can be processed on any CPU.
+	 *
+	 * Another CPU may be processing panic routines while holding
+	 * nmi_reason_lock. Check if the CPU issued the IPI for crash dumping,
+	 * and if so, call its callback directly.  If there is no CPU preparing
+	 * crash dump, we simply loop here.
+	 */
+	while (!raw_spin_trylock(&nmi_reason_lock)) {
+		run_crash_ipi_callback(regs);
+		cpu_relax();
+	}
+
 	reason = x86_platform.get_nmi_reason();
 
 	if (reason & NMI_REASON_MASK) {

arch/x86/kernel/reboot.c

@@ -718,6 +718,7 @@ static int crashing_cpu;
 static nmi_shootdown_cb shootdown_callback;
 
 static atomic_t waiting_for_crash_ipi;
+static int crash_ipi_issued;
 
 static int crash_nmi_callback(unsigned int val, struct pt_regs *regs)
 {
@@ -780,6 +781,9 @@ void nmi_shootdown_cpus(nmi_shootdown_cb callback)
 
 	smp_send_nmi_allbutself();
 
+	/* Kick CPUs looping in NMI context. */
+	WRITE_ONCE(crash_ipi_issued, 1);
+
 	msecs = 1000; /* Wait at most a second for the other cpus to stop */
 	while ((atomic_read(&waiting_for_crash_ipi) > 0) && msecs) {
 		mdelay(1);
@@ -788,9 +792,35 @@ void nmi_shootdown_cpus(nmi_shootdown_cb callback)
 
 	/* Leave the nmi callback set */
 }
+
+/*
+ * Check if the crash dumping IPI got issued and if so, call its callback
+ * directly. This function is used when we have already been in NMI handler.
+ * It doesn't return.
+ */
+void run_crash_ipi_callback(struct pt_regs *regs)
+{
+	if (crash_ipi_issued)
+		crash_nmi_callback(0, regs);
+}
+
+/* Override the weak function in kernel/panic.c */
+void nmi_panic_self_stop(struct pt_regs *regs)
+{
+	while (1) {
+		/* If no CPU is preparing crash dump, we simply loop here. */
+		run_crash_ipi_callback(regs);
+		cpu_relax();
+	}
+}
+
 #else /* !CONFIG_SMP */
 void nmi_shootdown_cpus(nmi_shootdown_cb callback)
 {
 	/* No other CPUs to shoot down */
 }
+
+void run_crash_ipi_callback(struct pt_regs *regs)
+{
+}
 #endif

arch/x86/kernel/smp.c

@@ -125,12 +125,12 @@ static void native_smp_send_reschedule(int cpu)
 		WARN_ON(1);
 		return;
 	}
-	apic->send_IPI_mask(cpumask_of(cpu), RESCHEDULE_VECTOR);
+	apic->send_IPI(cpu, RESCHEDULE_VECTOR);
 }
 
 void native_send_call_func_single_ipi(int cpu)
 {
-	apic->send_IPI_mask(cpumask_of(cpu), CALL_FUNCTION_SINGLE_VECTOR);
+	apic->send_IPI(cpu, CALL_FUNCTION_SINGLE_VECTOR);
 }
 
 void native_send_call_func_ipi(const struct cpumask *mask)

include/linux/kernel.h

@@ -255,6 +255,7 @@ extern long (*panic_blink)(int state);
 __printf(1, 2)
 void panic(const char *fmt, ...)
 	__noreturn __cold;
+void nmi_panic_self_stop(struct pt_regs *);
 extern void oops_enter(void);
 extern void oops_exit(void);
 void print_oops_end_marker(void);
@@ -445,6 +446,33 @@ extern int sysctl_panic_on_stackoverflow;
 
 extern bool crash_kexec_post_notifiers;
 
+/*
+ * panic_cpu is used for synchronizing panic() and crash_kexec() execution. It
+ * holds a CPU number which is executing panic() currently. A value of
+ * PANIC_CPU_INVALID means no CPU has entered panic() or crash_kexec().
+ */
+extern atomic_t panic_cpu;
+#define PANIC_CPU_INVALID	-1
+
+/*
+ * A variant of panic() called from NMI context. We return if we've already
+ * panicked on this CPU. If another CPU already panicked, loop in
+ * nmi_panic_self_stop() which can provide architecture dependent code such
+ * as saving register state for crash dump.
+ */
+#define nmi_panic(regs, fmt, ...)					\
+do {									\
+	int old_cpu, cpu;						\
+									\
+	cpu = raw_smp_processor_id();					\
+	old_cpu = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, cpu);	\
+									\
+	if (old_cpu == PANIC_CPU_INVALID)				\
+		panic(fmt, ##__VA_ARGS__);				\
+	else if (old_cpu != cpu)					\
+		nmi_panic_self_stop(regs);				\
+} while (0)
+
 /*
  * Only to be used by arch init code. If the user over-wrote the default
  * CONFIG_PANIC_TIMEOUT, honor it.

include/linux/kexec.h

@@ -237,6 +237,7 @@ extern int kexec_purgatory_get_set_symbol(struct kimage *image,
 					  unsigned int size, bool get_value);
 extern void *kexec_purgatory_get_symbol_addr(struct kimage *image,
 					     const char *name);
+extern void __crash_kexec(struct pt_regs *);
 extern void crash_kexec(struct pt_regs *);
 int kexec_should_crash(struct task_struct *);
 void crash_save_cpu(struct pt_regs *regs, int cpu);
@@ -332,6 +333,7 @@ int __weak arch_kexec_apply_relocations(const Elf_Ehdr *ehdr, Elf_Shdr *sechdrs,
 #else /* !CONFIG_KEXEC_CORE */
 struct pt_regs;
 struct task_struct;
+static inline void __crash_kexec(struct pt_regs *regs) { }
 static inline void crash_kexec(struct pt_regs *regs) { }
 static inline int kexec_should_crash(struct task_struct *p) { return 0; }
 #define kexec_in_progress false

kernel/kexec_core.c

@@ -853,7 +853,12 @@ struct kimage *kexec_image;
 struct kimage *kexec_crash_image;
 int kexec_load_disabled;
 
-void crash_kexec(struct pt_regs *regs)
+/*
+ * No panic_cpu check version of crash_kexec().  This function is called
+ * only when panic_cpu holds the current CPU number; this is the only CPU
+ * which processes crash_kexec routines.
+ */
+void __crash_kexec(struct pt_regs *regs)
 {
 	/* Take the kexec_mutex here to prevent sys_kexec_load
 	 * running on one cpu from replacing the crash kernel
@@ -876,6 +881,29 @@ void crash_kexec(struct pt_regs *regs)
 	}
 }
 
+void crash_kexec(struct pt_regs *regs)
+{
+	int old_cpu, this_cpu;
+
+	/*
+	 * Only one CPU is allowed to execute the crash_kexec() code as with
+	 * panic().  Otherwise parallel calls of panic() and crash_kexec()
+	 * may stop each other.  To exclude them, we use panic_cpu here too.
+	 */
+	this_cpu = raw_smp_processor_id();
+	old_cpu = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, this_cpu);
+	if (old_cpu == PANIC_CPU_INVALID) {
+		/* This is the 1st CPU which comes here, so go ahead. */
+		__crash_kexec(regs);
+
+		/*
+		 * Reset panic_cpu to allow another panic()/crash_kexec()
+		 * call.
+		 */
+		atomic_set(&panic_cpu, PANIC_CPU_INVALID);
+	}
+}
+
 size_t crash_get_memory_size(void)
 {
 	size_t size = 0;

kernel/panic.c

@@ -61,6 +61,17 @@ void __weak panic_smp_self_stop(void)
 		cpu_relax();
 }
 
+/*
+ * Stop ourselves in NMI context if another CPU has already panicked. Arch code
+ * may override this to prepare for crash dumping, e.g. save regs info.
+ */
+void __weak nmi_panic_self_stop(struct pt_regs *regs)
+{
+	panic_smp_self_stop();
+}
+
+atomic_t panic_cpu = ATOMIC_INIT(PANIC_CPU_INVALID);
+
 /**
  *	panic - halt the system
  *	@fmt: The text string to print
@@ -71,17 +82,17 @@ void __weak panic_smp_self_stop(void)
  */
 void panic(const char *fmt, ...)
 {
-	static DEFINE_SPINLOCK(panic_lock);
 	static char buf[1024];
 	va_list args;
 	long i, i_next = 0;
 	int state = 0;
+	int old_cpu, this_cpu;
 
 	/*
 	 * Disable local interrupts. This will prevent panic_smp_self_stop
 	 * from deadlocking the first cpu that invokes the panic, since
 	 * there is nothing to prevent an interrupt handler (that runs
-	 * after the panic_lock is acquired) from invoking panic again.
+	 * after setting panic_cpu) from invoking panic() again.
 	 */
 	local_irq_disable();
 
@@ -94,8 +105,16 @@ void panic(const char *fmt, ...)
 	 * multiple parallel invocations of panic, all other CPUs either
 	 * stop themself or will wait until they are stopped by the 1st CPU
 	 * with smp_send_stop().
+	 *
+	 * `old_cpu == PANIC_CPU_INVALID' means this is the 1st CPU which
+	 * comes here, so go ahead.
+	 * `old_cpu == this_cpu' means we came from nmi_panic() which sets
+	 * panic_cpu to this CPU.  In this case, this is also the 1st CPU.
 	 */
-	if (!spin_trylock(&panic_lock))
+	this_cpu = raw_smp_processor_id();
+	old_cpu  = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, this_cpu);
+
+	if (old_cpu != PANIC_CPU_INVALID && old_cpu != this_cpu)
 		panic_smp_self_stop();
 
 	console_verbose();
@@ -117,9 +136,11 @@ void panic(const char *fmt, ...)
 	 * everything else.
 	 * If we want to run this after calling panic_notifiers, pass
 	 * the "crash_kexec_post_notifiers" option to the kernel.
+	 *
+	 * Bypass the panic_cpu check and call __crash_kexec directly.
 	 */
 	if (!crash_kexec_post_notifiers)
-		crash_kexec(NULL);
+		__crash_kexec(NULL);
 
 	/*
 	 * Note smp_send_stop is the usual smp shutdown function, which
@@ -142,9 +163,11 @@ void panic(const char *fmt, ...)
 	 * panic_notifiers and dumping kmsg before kdump.
 	 * Note: since some panic_notifiers can make crashed kernel
 	 * more unstable, it can increase risks of the kdump failure too.
+	 *
+	 * Bypass the panic_cpu check and call __crash_kexec directly.
 	 */
 	if (crash_kexec_post_notifiers)
-		crash_kexec(NULL);
+		__crash_kexec(NULL);
 
 	bust_spinlocks(0);
 

kernel/watchdog.c

@@ -351,7 +351,7 @@ static void watchdog_overflow_callback(struct perf_event *event,
 			trigger_allbutself_cpu_backtrace();
 
 		if (hardlockup_panic)
-			panic("Hard LOCKUP");
+			nmi_panic(regs, "Hard LOCKUP");
 
 		__this_cpu_write(hard_watchdog_warn, true);
 		return;