sparse irq_desc[] array: core kernel and x86 changes

Impact: new feature

Problem on distro kernels: irq_desc[NR_IRQS] takes megabytes of RAM with
NR_CPUS set to large values. The goal is to be able to scale up to much
larger NR_IRQS value without impacting the (important) common case.

To solve this, we generalize irq_desc[NR_IRQS] to an (optional) array of
irq_desc pointers.

When CONFIG_SPARSE_IRQ=y is used, we use kzalloc_node to get irq_desc,
this also makes the IRQ descriptors NUMA-local (to the site that calls
request_irq()).

This gets rid of the irq_cfg[] static array on x86 as well: irq_cfg now
uses desc->chip_data for x86 to store irq_cfg.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

arch/x86/Kconfig

@@ -238,6 +238,16 @@ config X86_HAS_BOOT_CPU_ID
 	def_bool y
 	depends on X86_VOYAGER
 
+config SPARSE_IRQ
+	bool "Support sparse irq numbering"
+	depends on PCI_MSI || HT_IRQ
+	default y
+	help
+	  This enables support for sparse irq, esp for msi/msi-x. You may need
+	  if you have lots of cards supports msi-x installed.
+
+	  If you don't know what to do here, say Y.
+
 config X86_FIND_SMP_CONFIG
 	def_bool y
 	depends on X86_MPPARSE || X86_VOYAGER

arch/x86/include/asm/irq_vectors.h

@@ -102,11 +102,20 @@
 #define invalid_vm86_irq(irq)	((irq) < 3 || (irq) > 15)
 
 #if defined(CONFIG_X86_IO_APIC) && !defined(CONFIG_X86_VOYAGER)
+
+#ifndef CONFIG_SPARSE_IRQ
 # if NR_CPUS < MAX_IO_APICS
 #  define NR_IRQS (NR_VECTORS + (32 * NR_CPUS))
 # else
 #  define NR_IRQS (NR_VECTORS + (32 * MAX_IO_APICS))
 # endif
+#else
+# if (8 * NR_CPUS) > (32 * MAX_IO_APICS)
+#  define NR_IRQS (NR_VECTORS + (8 * NR_CPUS))
+# else
+#  define NR_IRQS (NR_VECTORS + (32 * MAX_IO_APICS))
+# endif
+#endif
 
 #elif defined(CONFIG_X86_VOYAGER)