Messages in this thread Patch in this message | | | Date | Tue, 28 Apr 2009 14:41:12 -0700 | From | Yinghai Lu <> | Subject | [PATCH] x86/irq: use move_irq_desc only when CONFIG_NUMA_IRQ_DESC is set |
| |
add config NUMA_IRQ_DESC, and create_irq_nr(), it will call move_irq_desc.
move_irq_desc will try to move irq_desc to node wanted if allocated one is not correct. (happen on device on different nodes that are using MSI, driver are loaded and unloaded randomly)
[ Impact: make irq_desc go with device on node ]
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
--- arch/x86/Kconfig | 9 +++++++++ arch/x86/kernel/apic/io_apic.c | 18 +++++++++++------- include/linux/irq.h | 11 +++++++++-- kernel/irq/Makefile | 2 +- 4 files changed, 30 insertions(+), 10 deletions(-)
Index: linux-2.6/arch/x86/Kconfig =================================================================== --- linux-2.6.orig/arch/x86/Kconfig +++ linux-2.6/arch/x86/Kconfig @@ -280,6 +280,15 @@ config SPARSE_IRQ If you don't know what to do here, say N. +config NUMA_IRQ_DESC + bool "put irq desc on numa node memory" + depends on SPARSE_IRQ && NUMA + default n + ---help--- + This enables irq_desc to cpu/node that that device is connected. + + If you don't know what to do here, say N. + config X86_MPPARSE bool "Enable MPS table" if ACPI default y Index: linux-2.6/arch/x86/kernel/apic/io_apic.c =================================================================== --- linux-2.6.orig/arch/x86/kernel/apic/io_apic.c +++ linux-2.6/arch/x86/kernel/apic/io_apic.c @@ -246,7 +246,7 @@ int arch_init_chip_data(struct irq_desc return 0; } -/* for move_irq_desc */ +#ifdef CONFIG_NUMA_IRQ_DESC static void init_copy_irq_2_pin(struct irq_cfg *old_cfg, struct irq_cfg *cfg, int node) { @@ -347,7 +347,7 @@ void arch_free_chip_data(struct irq_desc old_desc->chip_data = NULL; } } -/* end for move_irq_desc */ +#endif #else static struct irq_cfg *irq_cfg(unsigned int irq) @@ -3143,6 +3143,10 @@ unsigned int create_irq_nr(unsigned int if (irq_want < nr_irqs_gsi) irq_want = nr_irqs_gsi; +#ifndef CONFIG_NUMA_IRQ_DESC + node = cpu_to_node(boot_cpu_id); +#endif + spin_lock_irqsave(&vector_lock, flags); for (new = irq_want; new < nr_irqs; new++) { desc_new = irq_to_desc_alloc_node(new, node); @@ -3155,11 +3159,7 @@ unsigned int create_irq_nr(unsigned int if (cfg_new->vector != 0) continue; -#ifdef CONFIG_NUMA_IRQ_DESC - /* different node ?*/ - if (desc_new->node != node) - desc = move_irq_desc(desc, node); -#endif + desc_new = move_irq_desc(desc_new, node); if (__assign_irq_vector(new, cfg_new, apic->target_cpus()) == 0) irq = new; @@ -3938,10 +3938,14 @@ int io_apic_set_pci_routing(struct devic return -EINVAL; } +#ifdef CONFIG_NUMA_IRQ_DESC if (dev) node = dev_to_node(dev); else node = cpu_to_node(boot_cpu_id); +#else + node = cpu_to_node(boot_cpu_id); +#endif desc = irq_to_desc_alloc_node(irq, node); if (!desc) { Index: linux-2.6/include/linux/irq.h =================================================================== --- linux-2.6.orig/include/linux/irq.h +++ linux-2.6/include/linux/irq.h @@ -206,10 +206,17 @@ extern void arch_free_chip_data(struct i #ifndef CONFIG_SPARSE_IRQ extern struct irq_desc irq_desc[NR_IRQS]; -#else /* CONFIG_SPARSE_IRQ */ -extern struct irq_desc *move_irq_desc(struct irq_desc *old_desc, int node); #endif /* CONFIG_SPARSE_IRQ */ +#ifdef CONFIG_NUMA_IRQ_DESC +extern struct irq_desc *move_irq_desc(struct irq_desc *old_desc, int node); +#else +static inline struct irq_desc *move_irq_desc(struct irq_desc *desc, int node) +{ + return desc; +} +#endif + extern struct irq_desc *irq_to_desc_alloc_node(unsigned int irq, int node); /* Index: linux-2.6/kernel/irq/Makefile =================================================================== --- linux-2.6.orig/kernel/irq/Makefile +++ linux-2.6/kernel/irq/Makefile @@ -3,5 +3,5 @@ obj-y := handle.o manage.o spurious.o re obj-$(CONFIG_GENERIC_IRQ_PROBE) += autoprobe.o obj-$(CONFIG_PROC_FS) += proc.o obj-$(CONFIG_GENERIC_PENDING_IRQ) += migration.o -obj-$(CONFIG_SPARSE_IRQ) += numa_migrate.o +obj-$(CONFIG_NUMA_IRQ_DESC) += numa_migrate.o obj-$(CONFIG_PM_SLEEP) += pm.o
| |