[Patch V3 19/37] x86, irq: introduce mechanisms to support dynamically allocate IRQ for IOAPIC

[Jiang Liu <jiang.liu@linux.intel.com>]

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 as below:
1) 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.
2) Build identity mapping for GSIs between NR_IRQS_LEGACY and
   arch_dynirq_lower_bound(0). This is typically used to support legacy
   IRQs and simple platforms.
3) Dynamically allocate IRQs for GSIs above arch_dynirq_lower_bound(0).
   This may be used to support big system and IOAPIC hotplug.
4) Dynamically allocate IRQs for non-ISA IRQs below NR_IRQS_LEGACY
   if platform doens't request identity mapping. This is to support
   some really weired platforms.
5) Build identity mapping for GSIs below NR_IRQS_LEGACY if platform requests
   identity mapping. This is to support MID and CE41000 etc.

Function arch_dynirq_lower_bound(0) will be enhanced in coming patch
to enable dynamic IRQ allocation for IOAPIC. To ease our life,
arch_dynirq_lower_bound(0) must be greater than or equal to
NR_IRQS_LEGACY, otherwise it may break backward compatibilities.

Signed-off-by: Jiang Liu <jiang.liu@linux.intel.com>
---
 arch/x86/Kconfig               |    1 +
 arch/x86/include/asm/io_apic.h |   11 +++-
 arch/x86/kernel/acpi/boot.c    |    8 +--
 arch/x86/kernel/apic/io_apic.c |  142 ++++++++++++++++++++++++++++++----------
 4 files changed, 120 insertions(+), 42 deletions(-)

diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 47247708c9eb..ebdb1193821b 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -832,6 +832,7 @@ 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 978e51fdcb59..4778d129f225 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
 
@@ -130,6 +132,9 @@ extern int noioapicquirk;
 /* -1 if "noapic" boot option passed */
 extern int noioapicreroute;
 
+/* Build identity mapping for non-ISA IRQs below NR_IRQS_LEGACY */
+extern int ioapic_identity_map;
+
 /*
  * If we use the IO-APIC for IRQ routing, disable automatic
  * assignment of PCI IRQ's.
@@ -169,10 +174,12 @@ struct mp_ioapic_gsi{
 };
 extern u32 gsi_top;
 
+typedef struct irq_domain *(*ioapic_create_domain_fn)(int idx, void *arg);
+
 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 +222,7 @@ 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 f740d2766c42..9bbdd685df64 100644
--- a/arch/x86/kernel/acpi/boot.c
+++ b/arch/x86/kernel/acpi/boot.c
@@ -99,7 +99,7 @@ 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;
 
@@ -107,7 +107,7 @@ 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);
 }
 
 /*
@@ -483,7 +483,7 @@ 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
@@ -1030,7 +1030,7 @@ 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;
 
diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c
index cd3a886431b8..b16e74133686 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 +99,9 @@ static struct ioapic {
 	struct mpc_ioapic mp_config;
 	/* IO APIC gsi routing info */
 	struct mp_ioapic_gsi  gsi_config;
+	struct irq_domain *irqdomain;
+	ioapic_create_domain_fn irqdomain_cb;
+	void *irqdomain_arg;
 	DECLARE_BITMAP(pin_programmed, MP_MAX_IOAPIC_PIN + 1);
 } ioapics[MAX_IO_APICS];
 
@@ -135,6 +140,11 @@ static inline int mp_init_irq_at_boot(int ioapic, int irq)
 	return ioapic == 0 || (irq >= 0 && irq < NR_IRQS_LEGACY);
 }
 
+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 */
@@ -153,6 +163,7 @@ int mp_bus_id_to_type[MAX_MP_BUSSES];
 DECLARE_BITMAP(mp_bus_not_pci, MAX_MP_BUSSES);
 
 int skip_ioapic_setup;
+int ioapic_identity_map;
 
 /**
  * disable_ioapic_support() - disables ioapic support at runtime
@@ -963,19 +974,75 @@ 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;
+
+	if (gsi >= arch_dynirq_lower_bound(0)) {
+		irq = irq_create_mapping(domain, pin);
+	} else if (gsi < NR_IRQS_LEGACY) {
+		if (!ioapic_identity_map)
+			irq = irq_create_mapping(domain, pin);
+		else if (irq_domain_associate(domain, gsi, pin) == 0)
+			irq = gsi;
+	} else if (irq_create_strict_mappings(domain, gsi, pin, 1) == 0) {
+		irq = gsi;
+	}
+
+	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.
 	 */
-	return gsi >= NR_IRQS_LEGACY ? 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_IRQS_LEGACY ? 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)
+int mp_map_gsi_to_irq(u32 gsi, unsigned int flags)
 {
-	int irq;
-	int bus = mp_irqs[idx].srcbus;
+	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);
+}
+
+static int pin_2_irq(int idx, int ioapic, int pin, unsigned int flags)
+{
+	u32 gsi = mp_pin_to_gsi(ioapic, pin);
 
 	/*
 	 * Debugging check, we are in big trouble if this message pops up!
@@ -993,7 +1060,7 @@ 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);
@@ -1003,12 +1070,7 @@ 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 irq;
+	return  mp_map_pin_to_irq(gsi, idx, ioapic, pin, flags);
 }
 
 /*
@@ -1018,7 +1080,7 @@ 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",
@@ -1047,30 +1109,37 @@ 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);
 
@@ -1261,7 +1330,7 @@ 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);
 	}
 	/*
@@ -1387,8 +1456,9 @@ 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;
 
 		/*
@@ -1438,8 +1508,8 @@ 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),
@@ -3554,8 +3624,8 @@ 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);
-- 
1.7.10.4