From 7f7ace0cda64c99599c23785f8979a072e118058 Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Sat, 10 Jan 2009 21:58:08 -0800 Subject: cpumask: update irq_desc to use cpumask_var_t Impact: reduce memory usage, use new cpumask API. Replace the affinity and pending_masks with cpumask_var_t's. This adds to the significant size reduction done with the SPARSE_IRQS changes. The added functions (init_alloc_desc_masks & init_copy_desc_masks) are in the include file so they can be inlined (and optimized out for the !CONFIG_CPUMASKS_OFFSTACK case.) [Naming chosen to be consistent with the other init*irq functions, as well as the backwards arg declaration of "from, to" instead of the more common "to, from" standard.] Includes a slight change to the declaration of struct irq_desc to embed the pending_mask within ifdef(CONFIG_SMP) to be consistent with other references, and some small changes to Xen. Tested: sparse/non-sparse/cpumask_offstack/non-cpumask_offstack/nonuma/nosmp on x86_64 Signed-off-by: Mike Travis Cc: Chris Wright Cc: Jeremy Fitzhardinge Cc: KOSAKI Motohiro Cc: Venkatesh Pallipadi Cc: virtualization@lists.osdl.org Cc: xen-devel@lists.xensource.com Cc: Yinghai Lu --- include/linux/irq.h | 81 +++++++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 78 insertions(+), 3 deletions(-) (limited to 'include') diff --git a/include/linux/irq.h b/include/linux/irq.h index f899b502f18..fa27210f1df 100644 --- a/include/linux/irq.h +++ b/include/linux/irq.h @@ -182,11 +182,11 @@ struct irq_desc { unsigned int irqs_unhandled; spinlock_t lock; #ifdef CONFIG_SMP - cpumask_t affinity; + cpumask_var_t affinity; unsigned int cpu; -#endif #ifdef CONFIG_GENERIC_PENDING_IRQ - cpumask_t pending_mask; + cpumask_var_t pending_mask; +#endif #endif #ifdef CONFIG_PROC_FS struct proc_dir_entry *dir; @@ -422,4 +422,79 @@ extern int set_irq_msi(unsigned int irq, struct msi_desc *entry); #endif /* !CONFIG_S390 */ +#ifdef CONFIG_SMP +/** + * init_alloc_desc_masks - allocate cpumasks for irq_desc + * @desc: pointer to irq_desc struct + * @boot: true if need bootmem + * + * Allocates affinity and pending_mask cpumask if required. + * Returns true if successful (or not required). + * Side effect: affinity has all bits set, pending_mask has all bits clear. + */ +static inline bool init_alloc_desc_masks(struct irq_desc *desc, int node, + bool boot) +{ + if (boot) { + alloc_bootmem_cpumask_var(&desc->affinity); + cpumask_setall(desc->affinity); + +#ifdef CONFIG_GENERIC_PENDING_IRQ + alloc_bootmem_cpumask_var(&desc->pending_mask); + cpumask_clear(desc->pending_mask); +#endif + return true; + } + + if (!alloc_cpumask_var_node(&desc->affinity, GFP_ATOMIC, node)) + return false; + cpumask_setall(desc->affinity); + +#ifdef CONFIG_GENERIC_PENDING_IRQ + if (!alloc_cpumask_var_node(&desc->pending_mask, GFP_ATOMIC, node)) { + free_cpumask_var(desc->affinity); + return false; + } + cpumask_clear(desc->pending_mask); +#endif + return true; +} + +/** + * init_copy_desc_masks - copy cpumasks for irq_desc + * @old_desc: pointer to old irq_desc struct + * @new_desc: pointer to new irq_desc struct + * + * Insures affinity and pending_masks are copied to new irq_desc. + * If !CONFIG_CPUMASKS_OFFSTACK the cpumasks are embedded in the + * irq_desc struct so the copy is redundant. + */ + +static inline void init_copy_desc_masks(struct irq_desc *old_desc, + struct irq_desc *new_desc) +{ +#ifdef CONFIG_CPUMASKS_OFFSTACK + cpumask_copy(new_desc->affinity, old_desc->affinity); + +#ifdef CONFIG_GENERIC_PENDING_IRQ + cpumask_copy(new_desc->pending_mask, old_desc->pending_mask); +#endif +#endif +} + +#else /* !CONFIG_SMP */ + +static inline bool init_alloc_desc_masks(struct irq_desc *desc, int node, + bool boot) +{ + return true; +} + +static inline void init_copy_desc_masks(struct irq_desc *old_desc, + struct irq_desc *new_desc) +{ +} + +#endif /* CONFIG_SMP */ + #endif /* _LINUX_IRQ_H */ -- cgit v1.2.3-18-g5258 From fbd59a8d1f7cf325fdb6828659f1fb76631e87b3 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Sat, 10 Jan 2009 21:58:08 -0800 Subject: cpumask: Use topology_core_cpumask()/topology_thread_cpumask() Impact: reduce stack usage, use new cpumask API. This actually uses topology_core_cpumask() and topology_thread_cpumask(), removing the only users of topology_core_siblings() and topology_thread_siblings() Signed-off-by: Rusty Russell Signed-off-by: Mike Travis Cc: linux-net-drivers@solarflare.com --- include/linux/topology.h | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'include') diff --git a/include/linux/topology.h b/include/linux/topology.h index e632d29f054..a16b9e06f2e 100644 --- a/include/linux/topology.h +++ b/include/linux/topology.h @@ -193,5 +193,11 @@ int arch_update_cpu_topology(void); #ifndef topology_core_siblings #define topology_core_siblings(cpu) cpumask_of_cpu(cpu) #endif +#ifndef topology_thread_cpumask +#define topology_thread_cpumask(cpu) cpumask_of(cpu) +#endif +#ifndef topology_core_cpumask +#define topology_core_cpumask(cpu) cpumask_of(cpu) +#endif #endif /* _LINUX_TOPOLOGY_H */ -- cgit v1.2.3-18-g5258 From 802bf931f2688ad125b73db597ce63cc842fb27a Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Sat, 10 Jan 2009 21:58:09 -0800 Subject: cpumask: fix bug in use cpumask_var_t in irq_desc Impact: fix bug where new irq_desc uses old cpumask pointers which are freed. As Yinghai pointed out, init_copy_one_irq_desc() copies the old desc to the new desc overwriting the cpumask pointers. Since the old_desc and the cpumask pointers are freed, then memory corruption will occur if these old pointers are used. Move the allocation of these pointers to after the copy. Signed-off-by: Mike Travis Cc: Yinghai Lu --- include/linux/irq.h | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'include') diff --git a/include/linux/irq.h b/include/linux/irq.h index fa27210f1df..27a67536511 100644 --- a/include/linux/irq.h +++ b/include/linux/irq.h @@ -426,15 +426,18 @@ extern int set_irq_msi(unsigned int irq, struct msi_desc *entry); /** * init_alloc_desc_masks - allocate cpumasks for irq_desc * @desc: pointer to irq_desc struct + * @cpu: cpu which will be handling the cpumasks * @boot: true if need bootmem * * Allocates affinity and pending_mask cpumask if required. * Returns true if successful (or not required). * Side effect: affinity has all bits set, pending_mask has all bits clear. */ -static inline bool init_alloc_desc_masks(struct irq_desc *desc, int node, +static inline bool init_alloc_desc_masks(struct irq_desc *desc, int cpu, bool boot) { + int node; + if (boot) { alloc_bootmem_cpumask_var(&desc->affinity); cpumask_setall(desc->affinity); @@ -446,6 +449,8 @@ static inline bool init_alloc_desc_masks(struct irq_desc *desc, int node, return true; } + node = cpu_to_node(cpu); + if (!alloc_cpumask_var_node(&desc->affinity, GFP_ATOMIC, node)) return false; cpumask_setall(desc->affinity); @@ -484,7 +489,7 @@ static inline void init_copy_desc_masks(struct irq_desc *old_desc, #else /* !CONFIG_SMP */ -static inline bool init_alloc_desc_masks(struct irq_desc *desc, int node, +static inline bool init_alloc_desc_masks(struct irq_desc *desc, int cpu, bool boot) { return true; -- cgit v1.2.3-18-g5258 From 9332fccdedf8e09448f3b69b624211ae879f6c45 Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Sat, 10 Jan 2009 22:24:07 -0800 Subject: irq: initialize nr_irqs based on nr_cpu_ids Impact: Reduce memory usage. This is the second half of the changes to make the irq_desc_ptrs be variable sized based on nr_cpu_ids. This is done by adding a new "max_nr_irqs" macro to irq_vectors.h (and a dummy in irqnr.h) to return a max NR_IRQS value based on NR_CPUS or nr_cpu_ids. This necessitated moving the define of MAX_IO_APICS to a separate file (asm/apicnum.h) so it could be included without the baggage of the other asm/apicdef.h declarations. Signed-off-by: Mike Travis --- include/linux/irqnr.h | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'include') diff --git a/include/linux/irqnr.h b/include/linux/irqnr.h index 86af92e9e84..de66e4e1040 100644 --- a/include/linux/irqnr.h +++ b/include/linux/irqnr.h @@ -20,11 +20,18 @@ # define for_each_irq_desc_reverse(irq, desc) \ for (irq = nr_irqs - 1; irq >= 0; irq--) + #else /* CONFIG_GENERIC_HARDIRQS */ +#include /* need possible max_nr_irqs() */ + extern int nr_irqs; extern struct irq_desc *irq_to_desc(unsigned int irq); +# ifndef max_nr_irqs +# define max_nr_irqs(nr_cpus) NR_IRQS +# endif + # define for_each_irq_desc(irq, desc) \ for (irq = 0, desc = irq_to_desc(irq); irq < nr_irqs; \ irq++, desc = irq_to_desc(irq)) \ -- cgit v1.2.3-18-g5258 From 92296c6d6e908c35fca287a21af27be814af9c75 Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Sun, 11 Jan 2009 09:22:58 -0800 Subject: cpumask, irq: non-x86 build failures Ingo Molnar wrote: > All non-x86 architectures fail to build: > > In file included from /home/mingo/tip/include/linux/random.h:11, > from /home/mingo/tip/include/linux/stackprotector.h:6, > from /home/mingo/tip/init/main.c:17: > /home/mingo/tip/include/linux/irqnr.h:26:63: error: asm/irq_vectors.h: No such file or directory Do not include asm/irq_vectors.h in generic code - it's not available on all architectures. Signed-off-by: Ingo Molnar --- include/linux/irqnr.h | 6 ------ 1 file changed, 6 deletions(-) (limited to 'include') diff --git a/include/linux/irqnr.h b/include/linux/irqnr.h index de66e4e1040..887477bc2ab 100644 --- a/include/linux/irqnr.h +++ b/include/linux/irqnr.h @@ -23,15 +23,9 @@ #else /* CONFIG_GENERIC_HARDIRQS */ -#include /* need possible max_nr_irqs() */ - extern int nr_irqs; extern struct irq_desc *irq_to_desc(unsigned int irq); -# ifndef max_nr_irqs -# define max_nr_irqs(nr_cpus) NR_IRQS -# endif - # define for_each_irq_desc(irq, desc) \ for (irq = 0, desc = irq_to_desc(irq); irq < nr_irqs; \ irq++, desc = irq_to_desc(irq)) \ -- cgit v1.2.3-18-g5258 From 4a046d1754ee6ebb6f399696805ed61ea0444d4c Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Mon, 12 Jan 2009 17:39:24 -0800 Subject: x86: arch_probe_nr_irqs Impact: save RAM with large NR_CPUS, get smaller nr_irqs Signed-off-by: Yinghai Lu Signed-off-by: Mike Travis --- include/linux/interrupt.h | 1 + 1 file changed, 1 insertion(+) (limited to 'include') diff --git a/include/linux/interrupt.h b/include/linux/interrupt.h index 9127f6b51a3..472f11765f6 100644 --- a/include/linux/interrupt.h +++ b/include/linux/interrupt.h @@ -467,6 +467,7 @@ int show_interrupts(struct seq_file *p, void *v); struct irq_desc; extern int early_irq_init(void); +extern int arch_probe_nr_irqs(void); extern int arch_early_irq_init(void); extern int arch_init_chip_data(struct irq_desc *desc, int cpu); -- cgit v1.2.3-18-g5258 From c8399943bdb70fef78798b97f975506ecc99e039 Mon Sep 17 00:00:00 2001 From: Andi Kleen Date: Mon, 12 Jan 2009 23:01:15 +0100 Subject: x86, generic: mark complex bitops.h inlines as __always_inline Impact: reduce kernel image size Hugh Dickins noticed that older gcc versions when the kernel is built for code size didn't inline some of the bitops. Mark all complex x86 bitops that have more than a single asm statement or two as always inline to avoid this problem. Probably should be done for other architectures too. Ingo then found a better fix that only requires a single line change, but it unfortunately only works on gcc 4.3. On older gccs the original patch still makes a ~0.3% defconfig difference with CONFIG_OPTIMIZE_INLINING=y. With gcc 4.1 and a defconfig like build: 6116998 1138540 883788 8139326 7c323e vmlinux-oi-with-patch 6137043 1138540 883788 8159371 7c808b vmlinux-optimize-inlining ~20k / 0.3% difference. Signed-off-by: Andi Kleen Signed-off-by: Andrew Morton Signed-off-by: Ingo Molnar --- include/asm-generic/bitops/__ffs.h | 2 +- include/asm-generic/bitops/__fls.h | 2 +- include/asm-generic/bitops/fls.h | 2 +- include/asm-generic/bitops/fls64.h | 4 ++-- 4 files changed, 5 insertions(+), 5 deletions(-) (limited to 'include') diff --git a/include/asm-generic/bitops/__ffs.h b/include/asm-generic/bitops/__ffs.h index 9a3274aecf8..937d7c43557 100644 --- a/include/asm-generic/bitops/__ffs.h +++ b/include/asm-generic/bitops/__ffs.h @@ -9,7 +9,7 @@ * * Undefined if no bit exists, so code should check against 0 first. */ -static inline unsigned long __ffs(unsigned long word) +static __always_inline unsigned long __ffs(unsigned long word) { int num = 0; diff --git a/include/asm-generic/bitops/__fls.h b/include/asm-generic/bitops/__fls.h index be24465403d..a60a7ccb678 100644 --- a/include/asm-generic/bitops/__fls.h +++ b/include/asm-generic/bitops/__fls.h @@ -9,7 +9,7 @@ * * Undefined if no set bit exists, so code should check against 0 first. */ -static inline unsigned long __fls(unsigned long word) +static __always_inline unsigned long __fls(unsigned long word) { int num = BITS_PER_LONG - 1; diff --git a/include/asm-generic/bitops/fls.h b/include/asm-generic/bitops/fls.h index 850859bc506..0576d1f42f4 100644 --- a/include/asm-generic/bitops/fls.h +++ b/include/asm-generic/bitops/fls.h @@ -9,7 +9,7 @@ * Note fls(0) = 0, fls(1) = 1, fls(0x80000000) = 32. */ -static inline int fls(int x) +static __always_inline int fls(int x) { int r = 32; diff --git a/include/asm-generic/bitops/fls64.h b/include/asm-generic/bitops/fls64.h index 86d403f8b25..b097cf8444e 100644 --- a/include/asm-generic/bitops/fls64.h +++ b/include/asm-generic/bitops/fls64.h @@ -15,7 +15,7 @@ * at position 64. */ #if BITS_PER_LONG == 32 -static inline int fls64(__u64 x) +static __always_inline int fls64(__u64 x) { __u32 h = x >> 32; if (h) @@ -23,7 +23,7 @@ static inline int fls64(__u64 x) return fls(x); } #elif BITS_PER_LONG == 64 -static inline int fls64(__u64 x) +static __always_inline int fls64(__u64 x) { if (x == 0) return 0; -- cgit v1.2.3-18-g5258