x86, irq: Introduce mechanisms to support dynamically allocate IRQ for IOAPIC

Currently x86 support identity mapping between GSI(IOAPIC pin) and IRQ
number, so continous IRQs at low end are statically allocated to IOAPICs
at boot time. This design causes trouble to support IOAPIC hotplug.

This patch implements basic mechanism to dynamically allocate IRQ on
demand for IOAPIC pins by using irqdomain framework.

It first adds several fields into struct ioapic to support irqdomain.
Then it implements an algorithm to dynamically allocate IRQ number
for IOAPIC pins on demand.

Currently it supports three types of irqdomain:
1) LEGACY: used to support IOAPIC hosting legacy IRQs and building
   identity mapping for legacy IRQs. A speical case, we dynamically
   allocate IRQ number for IOAPIC pin which has GSI number below
   nr_legacy_irqs() but isn't legacy IRQ. This is for backward
   compatibility and avoid regression.
2) STRICT: build identity mapping between GSI and IRQ nubmer.
3) DYNAMIC: dynamically allocate IRQ number for IOAPIC pin on demand.

Legacy(ISA) IRQs is not managed by irqdomain because there may be
multiple pins sharing the same IRQ number and current irqdomain only
supports 1:1 mapping between pins and IRQ.

Signed-off-by: Jiang Liu <jiang.liu@linux.intel.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Grant Likely <grant.likely@linaro.org>
Cc: Rafael J. Wysocki <rjw@rjwysocki.net>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Len Brown <len.brown@intel.com>
Cc: Pavel Machek <pavel@ucw.cz>
Link: http://lkml.kernel.org/r/1402302011-23642-24-git-send-email-jiang.liu@linux.intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

4 files changed, 147 insertions, 41 deletions

diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 9df5c453171..147a7b78810 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -836,6 +836,7 @@ config X86_IO_APIC
 	def_bool y
 	depends on X86_64 || SMP || X86_32_NON_STANDARD || X86_UP_IOAPIC || PCI_MSI
 	select GENERIC_IRQ_LEGACY_ALLOC_HWIRQ
+	select IRQ_DOMAIN
 
 config X86_REROUTE_FOR_BROKEN_BOOT_IRQS
 	bool "Reroute for broken boot IRQs"
diff --git a/arch/x86/include/asm/io_apic.h b/arch/x86/include/asm/io_apic.h
index 978e51fdcb5..64c6e344399 100644
--- a/arch/x86/include/asm/io_apic.h
+++ b/arch/x86/include/asm/io_apic.h
@@ -98,6 +98,8 @@ struct IR_IO_APIC_route_entry {
 #define IOAPIC_AUTO     -1
 #define IOAPIC_EDGE     0
 #define IOAPIC_LEVEL    1
+#define	IOAPIC_MAP_ALLOC		0x1
+#define	IOAPIC_MAP_CHECK		0x2
 
 #ifdef CONFIG_X86_IO_APIC
 
@@ -163,6 +165,21 @@ extern int restore_ioapic_entries(void);
 extern void setup_ioapic_ids_from_mpc(void);
 extern void setup_ioapic_ids_from_mpc_nocheck(void);
 
+enum ioapic_domain_type {
+	IOAPIC_DOMAIN_INVALID,
+	IOAPIC_DOMAIN_LEGACY,
+	IOAPIC_DOMAIN_STRICT,
+	IOAPIC_DOMAIN_DYNAMIC,
+};
+
+struct device_node;
+struct irq_domain_ops;
+struct ioapic_domain_cfg {
+	enum ioapic_domain_type		type;
+	const struct irq_domain_ops	*ops;
+	struct device_node		*dev;
+};
+
 struct mp_ioapic_gsi{
 	u32 gsi_base;
 	u32 gsi_end;
@@ -172,7 +189,7 @@ extern u32 gsi_top;
 extern int mp_find_ioapic(u32 gsi);
 extern int mp_find_ioapic_pin(int ioapic, u32 gsi);
 extern u32 mp_pin_to_gsi(int ioapic, int pin);
-extern int mp_map_gsi_to_irq(u32 gsi);
+extern int mp_map_gsi_to_irq(u32 gsi, unsigned int flags);
 extern void __init mp_register_ioapic(int id, u32 address, u32 gsi_base);
 extern void __init pre_init_apic_IRQ0(void);
 
@@ -215,7 +232,7 @@ static inline void ioapic_insert_resources(void) { }
 #define gsi_top (NR_IRQS_LEGACY)
 static inline int mp_find_ioapic(u32 gsi) { return 0; }
 static inline u32 mp_pin_to_gsi(int ioapic, int pin) { return UINT_MAX; }
-static inline int mp_map_gsi_to_irq(u32 gsi) { return gsi; }
+static inline int mp_map_gsi_to_irq(u32 gsi, unsigned int flags) { return gsi; }
 
 struct io_apic_irq_attr;
 static inline int io_apic_set_pci_routing(struct device *dev, int irq,
diff --git a/arch/x86/kernel/acpi/boot.c b/arch/x86/kernel/acpi/boot.c
index 0cf311c72bc..d6635baf9e3 100644
--- a/arch/x86/kernel/acpi/boot.c
+++ b/arch/x86/kernel/acpi/boot.c
@@ -100,7 +100,7 @@ static u32 isa_irq_to_gsi[NR_IRQS_LEGACY] __read_mostly = {
 
 #define	ACPI_INVALID_GSI		INT_MIN
 
-static int map_gsi_to_irq(unsigned int gsi)
+static int map_gsi_to_irq(unsigned int gsi, unsigned int flags)
 {
 	int i;
 
@@ -108,7 +108,7 @@ static int map_gsi_to_irq(unsigned int gsi)
 		if (isa_irq_to_gsi[i] == gsi)
 			return i;
 
-	return mp_map_gsi_to_irq(gsi);
+	return mp_map_gsi_to_irq(gsi, flags);
 }
 
 /*
@@ -417,7 +417,7 @@ static int mp_register_gsi(struct device *dev, u32 gsi, int trigger,
 	if (acpi_gbl_FADT.sci_interrupt == gsi)
 		return gsi;
 
-	irq = map_gsi_to_irq(gsi);
+	irq = map_gsi_to_irq(gsi, IOAPIC_MAP_ALLOC);
 	if (irq < 0)
 		return irq;
 
@@ -608,7 +608,7 @@ void __init acpi_pic_sci_set_trigger(unsigned int irq, u16 trigger)
 
 int acpi_gsi_to_irq(u32 gsi, unsigned int *irqp)
 {
-	int irq = map_gsi_to_irq(gsi);
+	int irq = map_gsi_to_irq(gsi, IOAPIC_MAP_ALLOC | IOAPIC_MAP_CHECK);
 
 	if (irq >= 0) {
 #ifdef CONFIG_X86_IO_APIC
diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c
index 7fd9f1befe0..51ce80004a7 100644
--- a/arch/x86/kernel/apic/io_apic.c
+++ b/arch/x86/kernel/apic/io_apic.c
@@ -31,6 +31,7 @@
 #include <linux/acpi.h>
 #include <linux/module.h>
 #include <linux/syscore_ops.h>
+#include <linux/irqdomain.h>
 #include <linux/msi.h>
 #include <linux/htirq.h>
 #include <linux/freezer.h>
@@ -83,6 +84,7 @@ int sis_apic_bug = -1;
 
 static DEFINE_RAW_SPINLOCK(ioapic_lock);
 static DEFINE_RAW_SPINLOCK(vector_lock);
+static DEFINE_MUTEX(ioapic_mutex);
 
 static struct ioapic {
 	/*
@@ -97,6 +99,8 @@ static struct ioapic {
 	struct mpc_ioapic mp_config;
 	/* IO APIC gsi routing info */
 	struct mp_ioapic_gsi  gsi_config;
+	struct ioapic_domain_cfg irqdomain_cfg;
+	struct irq_domain *irqdomain;
 	DECLARE_BITMAP(pin_programmed, MP_MAX_IOAPIC_PIN + 1);
 } ioapics[MAX_IO_APICS];
 
@@ -142,6 +146,11 @@ static inline int mp_init_irq_at_boot(int ioapic, int irq)
 	return ioapic == 0 || (irq >= 0 && irq < nr_legacy_irqs());
 }
 
+static inline struct irq_domain *mp_ioapic_irqdomain(int ioapic)
+{
+	return ioapics[ioapic].irqdomain;
+}
+
 int nr_ioapics;
 
 /* The one past the highest gsi number used */
@@ -959,19 +968,79 @@ static int irq_trigger(int idx)
 	return trigger;
 }
 
-int mp_map_gsi_to_irq(u32 gsi)
+static int alloc_irq_from_domain(struct irq_domain *domain, u32 gsi, int pin)
 {
+	int irq = -1;
+	int ioapic = (int)(long)domain->host_data;
+	int type = ioapics[ioapic].irqdomain_cfg.type;
+
+	switch (type) {
+	case IOAPIC_DOMAIN_LEGACY:
+		/*
+		 * Dynamically allocate IRQ number for non-ISA IRQs in the first 16
+		 * GSIs on some weird platforms.
+		 */
+		if (gsi < nr_legacy_irqs())
+			irq = irq_create_mapping(domain, pin);
+		else if (irq_create_strict_mappings(domain, gsi, pin, 1) == 0)
+			irq = gsi;
+		break;
+	case IOAPIC_DOMAIN_STRICT:
+		if (irq_create_strict_mappings(domain, gsi, pin, 1) == 0)
+			irq = gsi;
+		break;
+	case IOAPIC_DOMAIN_DYNAMIC:
+		irq = irq_create_mapping(domain, pin);
+		break;
+	default:
+		WARN(1, "ioapic: unknown irqdomain type %d\n", type);
+		break;
+	}
+
+	return irq > 0 ? irq : -1;
+}
+
+static int mp_map_pin_to_irq(u32 gsi, int idx, int ioapic, int pin,
+			     unsigned int flags)
+{
+	int irq;
+	struct irq_domain *domain = mp_ioapic_irqdomain(ioapic);
+
 	/*
-	 * Provide an identity mapping of gsi == irq except on truly weird
-	 * platforms that have non isa irqs in the first 16 gsis.
+	 * Don't use irqdomain to manage ISA IRQs because there may be
+	 * multiple IOAPIC pins sharing the same ISA IRQ number and
+	 * irqdomain only supports 1:1 mapping between IOAPIC pin and
+	 * IRQ number. A typical IOAPIC has 24 pins, pin 0-15 are used
+	 * for legacy IRQs and pin 16-23 are used for PCI IRQs (PIRQ A-H).
+	 * When ACPI is disabled, only legacy IRQ numbers (IRQ0-15) are
+	 * available, and some BIOSes may use MP Interrupt Source records
+	 * to override IRQ numbers for PIRQs instead of reprogramming
+	 * the interrupt routing logic. Thus there may be multiple pins
+	 * sharing the same legacy IRQ number when ACPI is disabled.
 	 */
-	return gsi >= nr_legacy_irqs() ? gsi : gsi_top + gsi;
+	if (idx >= 0 && test_bit(mp_irqs[idx].srcbus, mp_bus_not_pci))
+		return mp_irqs[idx].srcbusirq;
+
+	if (!domain) {
+		/*
+		 * Provide an identity mapping of gsi == irq except on truly
+		 * weird platforms that have non isa irqs in the first 16 gsis.
+		 */
+		return gsi >= nr_legacy_irqs() ? gsi : gsi_top + gsi;
+	}
+
+	mutex_lock(&ioapic_mutex);
+	irq = irq_find_mapping(domain, pin);
+	if (irq <= 0 && (flags & IOAPIC_MAP_ALLOC))
+		irq = alloc_irq_from_domain(domain, gsi, pin);
+	mutex_unlock(&ioapic_mutex);
+
+	return irq > 0 ? irq : -1;
 }
 
-static int pin_2_irq(int idx, int apic, int pin)
+static int pin_2_irq(int idx, int ioapic, int pin, unsigned int flags)
 {
-	int irq;
-	int bus = mp_irqs[idx].srcbus;
+	u32 gsi = mp_pin_to_gsi(ioapic, pin);
 
 	/*
 	 * Debugging check, we are in big trouble if this message pops up!
@@ -989,7 +1058,7 @@ static int pin_2_irq(int idx, int apic, int pin)
 				apic_printk(APIC_VERBOSE, KERN_DEBUG
 						"disabling PIRQ%d\n", pin-16);
 			} else {
-				irq = pirq_entries[pin-16];
+				int irq = pirq_entries[pin-16];
 				apic_printk(APIC_VERBOSE, KERN_DEBUG
 						"using PIRQ%d -> IRQ %d\n",
 						pin-16, irq);
@@ -999,12 +1068,23 @@ static int pin_2_irq(int idx, int apic, int pin)
 	}
 #endif
 
-	if (test_bit(bus, mp_bus_not_pci))
-		irq = mp_irqs[idx].srcbusirq;
-	else
-		irq = mp_map_gsi_to_irq(mp_pin_to_gsi(apic, pin));
+	return  mp_map_pin_to_irq(gsi, idx, ioapic, pin, flags);
+}
 
-	return irq;
+int mp_map_gsi_to_irq(u32 gsi, unsigned int flags)
+{
+	int ioapic, pin, idx;
+
+	ioapic = mp_find_ioapic(gsi);
+	if (ioapic < 0)
+		return -1;
+
+	pin = mp_find_ioapic_pin(ioapic, gsi);
+	idx = find_irq_entry(ioapic, pin, mp_INT);
+	if ((flags & IOAPIC_MAP_CHECK) && idx < 0)
+		return -1;
+
+	return mp_map_pin_to_irq(gsi, idx, ioapic, pin, flags);
 }
 
 /*
@@ -1014,7 +1094,7 @@ static int pin_2_irq(int idx, int apic, int pin)
 int IO_APIC_get_PCI_irq_vector(int bus, int slot, int pin,
 				struct io_apic_irq_attr *irq_attr)
 {
-	int irq, i, best_guess = -1;
+	int irq, i, best_ioapic = -1, best_idx = -1;
 
 	apic_printk(APIC_DEBUG,
 		    "querying PCI -> IRQ mapping bus:%d, slot:%d, pin:%d.\n",
@@ -1043,30 +1123,37 @@ int IO_APIC_get_PCI_irq_vector(int bus, int slot, int pin,
 			continue;
 
 		/* Skip ISA IRQs */
-		irq = pin_2_irq(i, ioapic_idx, mp_irqs[i].dstirq);
-		if (ioapic_idx == 0 && !IO_APIC_IRQ(irq))
+		irq = pin_2_irq(i, ioapic_idx, mp_irqs[i].dstirq, 0);
+		if (irq > 0 && !IO_APIC_IRQ(irq))
 			continue;
 
 		if (pin == (mp_irqs[i].srcbusirq & 3)) {
-			set_io_apic_irq_attr(irq_attr, ioapic_idx,
-					     mp_irqs[i].dstirq,
-					     irq_trigger(i),
-					     irq_polarity(i));
-			return irq;
+			best_idx = i;
+			best_ioapic = ioapic_idx;
+			goto out;
 		}
+
 		/*
 		 * Use the first all-but-pin matching entry as a
 		 * best-guess fuzzy result for broken mptables.
 		 */
-		if (best_guess < 0) {
-			set_io_apic_irq_attr(irq_attr, ioapic_idx,
-					     mp_irqs[i].dstirq,
-					     irq_trigger(i),
-					     irq_polarity(i));
-			best_guess = irq;
+		if (best_idx < 0) {
+			best_idx = i;
+			best_ioapic = ioapic_idx;
 		}
 	}
-	return best_guess;
+	if (best_idx < 0)
+		return -1;
+
+out:
+	irq = pin_2_irq(best_idx, best_ioapic, mp_irqs[best_idx].dstirq,
+			IOAPIC_MAP_ALLOC);
+	if (irq > 0)
+		set_io_apic_irq_attr(irq_attr, best_ioapic,
+				     mp_irqs[best_idx].dstirq,
+				     irq_trigger(best_idx),
+				     irq_polarity(best_idx));
+	return irq;
 }
 EXPORT_SYMBOL(IO_APIC_get_PCI_irq_vector);
 
@@ -1257,7 +1344,7 @@ static inline int IO_APIC_irq_trigger(int irq)
 
 	for_each_ioapic_pin(apic, pin) {
 		idx = find_irq_entry(apic, pin, mp_INT);
-		if ((idx != -1) && (irq == pin_2_irq(idx, apic, pin)))
+		if ((idx != -1) && (irq == pin_2_irq(idx, apic, pin, 0)))
 			return irq_trigger(idx);
 	}
 	/*
@@ -1383,8 +1470,9 @@ static void __init __io_apic_setup_irqs(unsigned int ioapic_idx)
 		if (io_apic_pin_not_connected(idx, ioapic_idx, pin))
 			continue;
 
-		irq = pin_2_irq(idx, ioapic_idx, pin);
-		if (!mp_init_irq_at_boot(ioapic_idx, irq))
+		irq = pin_2_irq(idx, ioapic_idx, pin,
+				ioapic_idx ? 0 : IOAPIC_MAP_ALLOC);
+		if (irq < 0 || !mp_init_irq_at_boot(ioapic_idx, irq))
 			continue;
 
 		/*
@@ -1434,8 +1522,8 @@ void setup_IO_APIC_irq_extra(u32 gsi)
 	if (idx == -1)
 		return;
 
-	irq = pin_2_irq(idx, ioapic_idx, pin);
-	if (mp_init_irq_at_boot(ioapic_idx, irq))
+	irq = pin_2_irq(idx, ioapic_idx, pin, IOAPIC_MAP_ALLOC);
+	if (irq < 0 || mp_init_irq_at_boot(ioapic_idx, irq))
 		return;
 
 	set_io_apic_irq_attr(&attr, ioapic_idx, pin, irq_trigger(idx),
@@ -3543,8 +3631,8 @@ void __init setup_ioapic_dest(void)
 		if (irq_entry == -1)
 			continue;
 
-		irq = pin_2_irq(irq_entry, ioapic, pin);
-		if (!mp_init_irq_at_boot(ioapic, irq))
+		irq = pin_2_irq(irq_entry, ioapic, pin, 0);
+		if (irq < 0 || !mp_init_irq_at_boot(ioapic, irq))
 			continue;
 
 		idata = irq_get_irq_data(irq);