From fc590c22f9f056ab50190b797f6cacead29f9b75 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Fri, 2 Sep 2016 14:45:19 +0200 Subject: genirq: Robustify handle_percpu_devid_irq() The percpu_devid handler is not robust against spurious interrupts. If a spurious interrupt happens and no action is installed then the handler crashes with a NULL pointer dereference. Add a sanity check for this and log the wreckage once in dmesg. Reported-by: Majun Signed-off-by: Thomas Gleixner Cc: Mark Rutland Cc: Marc Zyngier Cc: guohanjun@huawei.com Cc: dingtianhong@huawei.com Cc: linux-arm-kernel@lists.infradead.org Link: http://lkml.kernel.org/r/alpine.DEB.2.20.1609021436160.5647@nanos --- kernel/irq/chip.c | 18 ++++++++++++++---- 1 file changed, 14 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c index b4c1bc7c9ca2..93c373a8b12b 100644 --- a/kernel/irq/chip.c +++ b/kernel/irq/chip.c @@ -756,7 +756,6 @@ void handle_percpu_devid_irq(struct irq_desc *desc) { struct irq_chip *chip = irq_desc_get_chip(desc); struct irqaction *action = desc->action; - void *dev_id = raw_cpu_ptr(action->percpu_dev_id); unsigned int irq = irq_desc_get_irq(desc); irqreturn_t res; @@ -765,9 +764,20 @@ void handle_percpu_devid_irq(struct irq_desc *desc) if (chip->irq_ack) chip->irq_ack(&desc->irq_data); - trace_irq_handler_entry(irq, action); - res = action->handler(irq, dev_id); - trace_irq_handler_exit(irq, action, res); + if (likely(action)) { + trace_irq_handler_entry(irq, action); + res = action->handler(irq, raw_cpu_ptr(action->percpu_dev_id)); + trace_irq_handler_exit(irq, action, res); + } else { + unsigned int cpu = smp_processor_id(); + bool enabled = cpumask_test_cpu(cpu, desc->percpu_enabled); + + if (enabled) + irq_percpu_disable(desc, cpu); + + pr_err_once("Spurious%s percpu IRQ%u on CPU%u\n", + enabled ? " and unmasked" : "", irq, cpu); + } if (chip->irq_eoi) chip->irq_eoi(&desc->irq_data); -- cgit v1.2.3 From 48e0fba842c7daab80f3351398146368c5504a27 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Fri, 2 Sep 2016 17:30:35 +0200 Subject: genirq: Remove export of irq_map_generic_chip() No module users. Signed-off-by: Thomas Gleixner --- kernel/irq/generic-chip.c | 1 - 1 file changed, 1 deletion(-) (limited to 'kernel') diff --git a/kernel/irq/generic-chip.c b/kernel/irq/generic-chip.c index abd286afbd27..5fbb94b077b3 100644 --- a/kernel/irq/generic-chip.c +++ b/kernel/irq/generic-chip.c @@ -409,7 +409,6 @@ int irq_map_generic_chip(struct irq_domain *d, unsigned int virq, irq_modify_status(virq, dgc->irq_flags_to_clear, dgc->irq_flags_to_set); return 0; } -EXPORT_SYMBOL_GPL(irq_map_generic_chip); struct irq_domain_ops irq_generic_chip_ops = { .map = irq_map_generic_chip, -- cgit v1.2.3 From f0c450eaa364cb77c778f2a46ee2aa3ff464b332 Mon Sep 17 00:00:00 2001 From: Sebastian Frias Date: Mon, 1 Aug 2016 16:27:53 +0200 Subject: genirq/generic_chip: Get rid of code duplication irq_map_generic_chip() contains about the same code as irq_get_domain_generic_chip() except for the return values. Split out the irq_get_domain_generic_chip() implementation so it can be reused. [ tglx: Removed the extra churn in irq_get_domain_generic_chip() callers and massaged changelog ] Signed-off-by: Sebastian Frias Cc: Marc Zyngier Cc: Mason Cc: Jason Cooper Link: http://lkml.kernel.org/r/579F5C69.8070006@laposte.net Signed-off-by: Thomas Gleixner --- kernel/irq/generic-chip.c | 34 +++++++++++++++++++--------------- 1 file changed, 19 insertions(+), 15 deletions(-) (limited to 'kernel') diff --git a/kernel/irq/generic-chip.c b/kernel/irq/generic-chip.c index 5fbb94b077b3..11ad73b39d2e 100644 --- a/kernel/irq/generic-chip.c +++ b/kernel/irq/generic-chip.c @@ -328,6 +328,20 @@ int irq_alloc_domain_generic_chips(struct irq_domain *d, int irqs_per_chip, } EXPORT_SYMBOL_GPL(irq_alloc_domain_generic_chips); +static struct irq_chip_generic * +__irq_get_domain_generic_chip(struct irq_domain *d, unsigned int hw_irq) +{ + struct irq_domain_chip_generic *dgc = d->gc; + int idx; + + if (!dgc) + return ERR_PTR(-ENODEV); + idx = hw_irq / dgc->irqs_per_chip; + if (idx >= dgc->num_chips) + return ERR_PTR(-EINVAL); + return dgc->gc[idx]; +} + /** * irq_get_domain_generic_chip - Get a pointer to the generic chip of a hw_irq * @d: irq domain pointer @@ -336,15 +350,9 @@ EXPORT_SYMBOL_GPL(irq_alloc_domain_generic_chips); struct irq_chip_generic * irq_get_domain_generic_chip(struct irq_domain *d, unsigned int hw_irq) { - struct irq_domain_chip_generic *dgc = d->gc; - int idx; + struct irq_chip_generic *gc = __irq_get_domain_generic_chip(d, hw_irq); - if (!dgc) - return NULL; - idx = hw_irq / dgc->irqs_per_chip; - if (idx >= dgc->num_chips) - return NULL; - return dgc->gc[idx]; + return !IS_ERR(gc) ? gc : NULL; } EXPORT_SYMBOL_GPL(irq_get_domain_generic_chip); @@ -368,13 +376,9 @@ int irq_map_generic_chip(struct irq_domain *d, unsigned int virq, unsigned long flags; int idx; - if (!d->gc) - return -ENODEV; - - idx = hw_irq / dgc->irqs_per_chip; - if (idx >= dgc->num_chips) - return -EINVAL; - gc = dgc->gc[idx]; + gc = __irq_get_domain_generic_chip(d, hw_irq); + if (IS_ERR(gc)) + return PTR_ERR(gc); idx = hw_irq % dgc->irqs_per_chip; -- cgit v1.2.3 From ee26c013cdee0b947e29d6cadfb9ff3341c69ff9 Mon Sep 17 00:00:00 2001 From: Sebastian Frias Date: Mon, 1 Aug 2016 16:27:38 +0200 Subject: genirq/generic_chip: Add irq_unmap callback Without this patch irq_domain_disassociate() cannot properly release the interrupt. In fact, irq_map_generic_chip() checks a bit on 'gc->installed' but said bit is never cleared, only set. Commit 088f40b7b027 ("genirq: Generic chip: Add linear irq domain support") added irq_map_generic_chip() function and also stated "This lacks a removal function for now". This commit provides an implementation of an unmap function that can be called by irq_domain_disassociate(). [ tglx: Made the function static and removed the export as we have neither a prototype nor a modular user. ] Fixes: 088f40b7b027 ("genirq: Generic chip: Add linear irq domain support") Signed-off-by: Sebastian Frias Cc: Marc Zyngier Cc: Mason Cc: Jason Cooper Link: http://lkml.kernel.org/r/579F5C5A.2070507@laposte.net Signed-off-by: Thomas Gleixner --- kernel/irq/generic-chip.c | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+) (limited to 'kernel') diff --git a/kernel/irq/generic-chip.c b/kernel/irq/generic-chip.c index 11ad73b39d2e..a3a392097804 100644 --- a/kernel/irq/generic-chip.c +++ b/kernel/irq/generic-chip.c @@ -414,8 +414,29 @@ int irq_map_generic_chip(struct irq_domain *d, unsigned int virq, return 0; } +static void irq_unmap_generic_chip(struct irq_domain *d, unsigned int virq) +{ + struct irq_data *data = irq_domain_get_irq_data(d, virq); + struct irq_domain_chip_generic *dgc = d->gc; + unsigned int hw_irq = data->hwirq; + struct irq_chip_generic *gc; + int irq_idx; + + gc = irq_get_domain_generic_chip(d, hw_irq); + if (!gc) + return; + + irq_idx = hw_irq % dgc->irqs_per_chip; + + clear_bit(irq_idx, &gc->installed); + irq_domain_set_info(d, virq, hw_irq, &no_irq_chip, NULL, NULL, NULL, + NULL); + +} + struct irq_domain_ops irq_generic_chip_ops = { .map = irq_map_generic_chip, + .unmap = irq_unmap_generic_chip, .xlate = irq_domain_xlate_onetwocell, }; EXPORT_SYMBOL_GPL(irq_generic_chip_ops); -- cgit v1.2.3 From 0c228919e04ddec195402296e7ebf2472ed6caef Mon Sep 17 00:00:00 2001 From: Sebastian Frias Date: Tue, 2 Aug 2016 10:52:45 +0200 Subject: irqdomain: Mask irq type in irq_domain_xlate_onetwocell() According to the xlate() callback definition, the 'out_type' parameter needs to be the "linux irq type". A mask for such bits exists, IRQ_TYPE_SENSE_MASK, which is correctly applied in irq_domain_xlate_twocell() So use it for irq_domain_xlate_onetwocell() as well. Signed-off-by: Sebastian Frias Cc: Grant Likely Cc: Marc Zyngier Cc: Mason Cc: Jason Cooper Link: http://lkml.kernel.org/r/57A05F5D.103@laposte.net Signed-off-by: Thomas Gleixner --- kernel/irq/irqdomain.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/irq/irqdomain.c b/kernel/irq/irqdomain.c index 4752b43662e0..f10cffe8aefb 100644 --- a/kernel/irq/irqdomain.c +++ b/kernel/irq/irqdomain.c @@ -868,7 +868,10 @@ int irq_domain_xlate_onetwocell(struct irq_domain *d, if (WARN_ON(intsize < 1)) return -EINVAL; *out_hwirq = intspec[0]; - *out_type = (intsize > 1) ? intspec[1] : IRQ_TYPE_NONE; + if (intsize > 1) + *out_type = intspec[1] & IRQ_TYPE_SENSE_MASK; + else + *out_type = IRQ_TYPE_NONE; return 0; } EXPORT_SYMBOL_GPL(irq_domain_xlate_onetwocell); -- cgit v1.2.3 From f88eecfe2f22b2790e7527c0aaec14ea175919de Mon Sep 17 00:00:00 2001 From: Sebastian Frias Date: Tue, 16 Aug 2016 16:05:08 +0200 Subject: genirq/generic_chip: Verify irqs_per_chip <= 32 Most (if not all) code here implicitly assumes that the maximum number of IRQs per chip will be 32, and thus uses 'u32' or 'unsigned long' for many tasks (for example "struct irq_data" declares its 'mask' field as 'u32', and "struct irq_chip_generic" declares its 'installed' field as 'unsigned long') However, there is no check to verify that irqs_per_chip is <= 32. Hence, calling irq_alloc_domain_generic_chips() with a bigger value will result in unexpected results. Provide a wrapper with a MAYBE_BUILD_BUG_ON(nrirqs >= 32) to catch such cases. [ tglx: Reduced changelog to the essential information ] Signed-off-by: Sebastian Frias Cc: Marc Zyngier Cc: Mason Cc: Jason Cooper Link: http://lkml.kernel.org/r/57B31D94.5040701@laposte.net Signed-off-by: Thomas Gleixner --- include/linux/irq.h | 18 +++++++++++++----- kernel/irq/generic-chip.c | 16 ++++++++-------- 2 files changed, 21 insertions(+), 13 deletions(-) (limited to 'kernel') diff --git a/include/linux/irq.h b/include/linux/irq.h index b52424eaa0ed..603986741f2c 100644 --- a/include/linux/irq.h +++ b/include/linux/irq.h @@ -916,12 +916,20 @@ void irq_remove_generic_chip(struct irq_chip_generic *gc, u32 msk, unsigned int clr, unsigned int set); struct irq_chip_generic *irq_get_domain_generic_chip(struct irq_domain *d, unsigned int hw_irq); -int irq_alloc_domain_generic_chips(struct irq_domain *d, int irqs_per_chip, - int num_ct, const char *name, - irq_flow_handler_t handler, - unsigned int clr, unsigned int set, - enum irq_gc_flags flags); +int __irq_alloc_domain_generic_chips(struct irq_domain *d, int irqs_per_chip, + int num_ct, const char *name, + irq_flow_handler_t handler, + unsigned int clr, unsigned int set, + enum irq_gc_flags flags); + +#define irq_alloc_domain_generic_chips(d, irqs_per_chip, num_ct, name, \ + handler, clr, set, flags) \ +({ \ + MAYBE_BUILD_BUG_ON(irqs_per_chip > 32); \ + __irq_alloc_domain_generic_chips(d, irqs_per_chip, num_ct, name,\ + handler, clr, set, flags); \ +}) static inline struct irq_chip_type *irq_data_get_chip_type(struct irq_data *d) { diff --git a/kernel/irq/generic-chip.c b/kernel/irq/generic-chip.c index a3a392097804..ee32870079c9 100644 --- a/kernel/irq/generic-chip.c +++ b/kernel/irq/generic-chip.c @@ -260,9 +260,9 @@ irq_gc_init_mask_cache(struct irq_chip_generic *gc, enum irq_gc_flags flags) } /** - * irq_alloc_domain_generic_chip - Allocate generic chips for an irq domain + * __irq_alloc_domain_generic_chip - Allocate generic chips for an irq domain * @d: irq domain for which to allocate chips - * @irqs_per_chip: Number of interrupts each chip handles + * @irqs_per_chip: Number of interrupts each chip handles (max 32) * @num_ct: Number of irq_chip_type instances associated with this * @name: Name of the irq chip * @handler: Default flow handler associated with these chips @@ -270,11 +270,11 @@ irq_gc_init_mask_cache(struct irq_chip_generic *gc, enum irq_gc_flags flags) * @set: IRQ_* bits to set in the mapping function * @gcflags: Generic chip specific setup flags */ -int irq_alloc_domain_generic_chips(struct irq_domain *d, int irqs_per_chip, - int num_ct, const char *name, - irq_flow_handler_t handler, - unsigned int clr, unsigned int set, - enum irq_gc_flags gcflags) +int __irq_alloc_domain_generic_chips(struct irq_domain *d, int irqs_per_chip, + int num_ct, const char *name, + irq_flow_handler_t handler, + unsigned int clr, unsigned int set, + enum irq_gc_flags gcflags) { struct irq_domain_chip_generic *dgc; struct irq_chip_generic *gc; @@ -326,7 +326,7 @@ int irq_alloc_domain_generic_chips(struct irq_domain *d, int irqs_per_chip, d->name = name; return 0; } -EXPORT_SYMBOL_GPL(irq_alloc_domain_generic_chips); +EXPORT_SYMBOL_GPL(__irq_alloc_domain_generic_chips); static struct irq_chip_generic * __irq_get_domain_generic_chip(struct irq_domain *d, unsigned int hw_irq) -- cgit v1.2.3 From 545d5d657b720e9c4dc773265bb7e9d88e34b269 Mon Sep 17 00:00:00 2001 From: Punit Agrawal Date: Tue, 31 May 2016 13:56:48 +0100 Subject: genirq: Update stale comment for __irq_domain_add Commit 1bf4ddc46c5d ("irqdomain: Introduce irq_domain_create_{linear, tree}") introduced the use of fwnode_handle to identify the interrupt controller when calling __irq_domain_add but missed updating the kernel doc parameters for the function. Update this comment. While we are touching this code, also consolidate the declaration and assignment of of_node. Signed-off-by: Punit Agrawal Acked-by: Marc Zygnier Link: http://lkml.kernel.org/r/1464699409-23113-1-git-send-email-punit.agrawal@arm.com Signed-off-by: Thomas Gleixner --- kernel/irq/irqdomain.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/irq/irqdomain.c b/kernel/irq/irqdomain.c index f10cffe8aefb..8c0a0ae43521 100644 --- a/kernel/irq/irqdomain.c +++ b/kernel/irq/irqdomain.c @@ -80,7 +80,7 @@ EXPORT_SYMBOL_GPL(irq_domain_free_fwnode); /** * __irq_domain_add() - Allocate a new irq_domain data structure - * @of_node: optional device-tree node of the interrupt controller + * @fwnode: firmware node for the interrupt controller * @size: Size of linear map; 0 for radix mapping only * @hwirq_max: Maximum number of interrupts supported by controller * @direct_max: Maximum value of direct maps; Use ~0 for no limit; 0 for no @@ -96,10 +96,8 @@ struct irq_domain *__irq_domain_add(struct fwnode_handle *fwnode, int size, const struct irq_domain_ops *ops, void *host_data) { + struct device_node *of_node = to_of_node(fwnode); struct irq_domain *domain; - struct device_node *of_node; - - of_node = to_of_node(fwnode); domain = kzalloc_node(sizeof(*domain) + (sizeof(unsigned int) * size), GFP_KERNEL, of_node_to_nid(of_node)); -- cgit v1.2.3 From 00b992deaa08495ab958da5950c9ebbba27d0ddc Mon Sep 17 00:00:00 2001 From: Alexander Kuleshov Date: Tue, 19 Jul 2016 15:54:08 +0600 Subject: genirq: No need to mask non trigger mode flags before __irq_set_trigger() Some callers of __irq_set_trigger() masks all flags except trigger mode flags. This is unnecessary, ase __irq_set_trigger() already does this before usage of flags. [ tglx: Moved the flag mask and adjusted comment. Removed the hunk in enable_percpu_irq() as it is required there ] Signed-off-by: Alexander Kuleshov Link: http://lkml.kernel.org/r/20160719095408.13778-1-kuleshovmail@gmail.com Signed-off-by: Thomas Gleixner --- kernel/irq/chip.c | 1 - kernel/irq/manage.c | 5 ++--- 2 files changed, 2 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c index 93c373a8b12b..e11e8afcf209 100644 --- a/kernel/irq/chip.c +++ b/kernel/irq/chip.c @@ -76,7 +76,6 @@ int irq_set_irq_type(unsigned int irq, unsigned int type) if (!desc) return -EINVAL; - type &= IRQ_TYPE_SENSE_MASK; ret = __irq_set_trigger(desc, type); irq_put_desc_busunlock(desc, flags); return ret; diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c index 73a2b786b5e9..4908617dee28 100644 --- a/kernel/irq/manage.c +++ b/kernel/irq/manage.c @@ -669,8 +669,6 @@ int __irq_set_trigger(struct irq_desc *desc, unsigned long flags) return 0; } - flags &= IRQ_TYPE_SENSE_MASK; - if (chip->flags & IRQCHIP_SET_TYPE_MASKED) { if (!irqd_irq_masked(&desc->irq_data)) mask_irq(desc); @@ -678,7 +676,8 @@ int __irq_set_trigger(struct irq_desc *desc, unsigned long flags) unmask = 1; } - /* caller masked out all except trigger mode flags */ + /* Mask all flags except trigger mode */ + flags &= IRQ_TYPE_SENSE_MASK; ret = chip->irq_set_type(&desc->irq_data, flags); switch (ret) { -- cgit v1.2.3 From ecb3f394c5dba897d215a5422f1b363e93e2ce4e Mon Sep 17 00:00:00 2001 From: Craig Gallek Date: Tue, 13 Sep 2016 12:14:51 -0400 Subject: genirq: Expose interrupt information through sysfs Information about interrupts is exposed via /proc/interrupts, but the format of that file has changed over kernel versions and differs across architectures. It also has varying column numbers depending on hardware. That all makes it hard for tools to parse. To solve this, expose the information through sysfs so each irq attribute is in a separate file in a consistent, machine parsable way. This feature is only available when both CONFIG_SPARSE_IRQ and CONFIG_SYSFS are enabled. Examples: /sys/kernel/irq/18/actions: i801_smbus,ehci_hcd:usb1,uhci_hcd:usb7 /sys/kernel/irq/18/chip_name: IR-IO-APIC /sys/kernel/irq/18/hwirq: 18 /sys/kernel/irq/18/name: fasteoi /sys/kernel/irq/18/per_cpu_count: 0,0 /sys/kernel/irq/18/type: level /sys/kernel/irq/25/actions: ahci0 /sys/kernel/irq/25/chip_name: IR-PCI-MSI /sys/kernel/irq/25/hwirq: 512000 /sys/kernel/irq/25/name: edge /sys/kernel/irq/25/per_cpu_count: 29036,0 /sys/kernel/irq/25/type: edge [ tglx: Moved kobject_del() under sparse_irq_lock, massaged code comments and changelog ] Signed-off-by: Craig Gallek Cc: David Decotigny Link: http://lkml.kernel.org/r/1473783291-122873-1-git-send-email-kraigatgoog@gmail.com Signed-off-by: Thomas Gleixner --- Documentation/ABI/testing/sysfs-kernel-irq | 53 ++++++++ include/linux/irqdesc.h | 3 + kernel/irq/irqdesc.c | 193 ++++++++++++++++++++++++++++- 3 files changed, 247 insertions(+), 2 deletions(-) create mode 100644 Documentation/ABI/testing/sysfs-kernel-irq (limited to 'kernel') diff --git a/Documentation/ABI/testing/sysfs-kernel-irq b/Documentation/ABI/testing/sysfs-kernel-irq new file mode 100644 index 000000000000..eb074b100986 --- /dev/null +++ b/Documentation/ABI/testing/sysfs-kernel-irq @@ -0,0 +1,53 @@ +What: /sys/kernel/irq +Date: September 2016 +KernelVersion: 4.9 +Contact: Craig Gallek +Description: Directory containing information about the system's IRQs. + Specifically, data from the associated struct irq_desc. + The information here is similar to that in /proc/interrupts + but in a more machine-friendly format. This directory contains + one subdirectory for each Linux IRQ number. + +What: /sys/kernel/irq//actions +Date: September 2016 +KernelVersion: 4.9 +Contact: Craig Gallek +Description: The IRQ action chain. A comma-separated list of zero or more + device names associated with this interrupt. + +What: /sys/kernel/irq//chip_name +Date: September 2016 +KernelVersion: 4.9 +Contact: Craig Gallek +Description: Human-readable chip name supplied by the associated device + driver. + +What: /sys/kernel/irq//hwirq +Date: September 2016 +KernelVersion: 4.9 +Contact: Craig Gallek +Description: When interrupt translation domains are used, this file contains + the underlying hardware IRQ number used for this Linux IRQ. + +What: /sys/kernel/irq//name +Date: September 2016 +KernelVersion: 4.9 +Contact: Craig Gallek +Description: Human-readable flow handler name as defined by the irq chip + driver. + +What: /sys/kernel/irq//per_cpu_count +Date: September 2016 +KernelVersion: 4.9 +Contact: Craig Gallek +Description: The number of times the interrupt has fired since boot. This + is a comma-separated list of counters; one per CPU in CPU id + order. NOTE: This file consistently shows counters for all + CPU ids. This differs from the behavior of /proc/interrupts + which only shows counters for online CPUs. + +What: /sys/kernel/irq//type +Date: September 2016 +KernelVersion: 4.9 +Contact: Craig Gallek +Description: The type of the interrupt. Either the string 'level' or 'edge'. diff --git a/include/linux/irqdesc.h b/include/linux/irqdesc.h index b51beebf9804..c9be57931b58 100644 --- a/include/linux/irqdesc.h +++ b/include/linux/irqdesc.h @@ -2,6 +2,7 @@ #define _LINUX_IRQDESC_H #include +#include /* * Core internal functions to deal with irq descriptors @@ -43,6 +44,7 @@ struct pt_regs; * @force_resume_depth: number of irqactions on a irq descriptor with * IRQF_FORCE_RESUME set * @rcu: rcu head for delayed free + * @kobj: kobject used to represent this struct in sysfs * @dir: /proc/irq/ procfs entry * @name: flow handler name for /proc/interrupts output */ @@ -88,6 +90,7 @@ struct irq_desc { #endif #ifdef CONFIG_SPARSE_IRQ struct rcu_head rcu; + struct kobject kobj; #endif int parent_irq; struct module *owner; diff --git a/kernel/irq/irqdesc.c b/kernel/irq/irqdesc.c index a623b44f2d4b..93b51727abaa 100644 --- a/kernel/irq/irqdesc.c +++ b/kernel/irq/irqdesc.c @@ -15,6 +15,7 @@ #include #include #include +#include #include "internals.h" @@ -123,6 +124,181 @@ static DECLARE_BITMAP(allocated_irqs, IRQ_BITMAP_BITS); #ifdef CONFIG_SPARSE_IRQ +static void irq_kobj_release(struct kobject *kobj); + +#ifdef CONFIG_SYSFS +static struct kobject *irq_kobj_base; + +#define IRQ_ATTR_RO(_name) \ +static struct kobj_attribute _name##_attr = __ATTR_RO(_name) + +static ssize_t per_cpu_count_show(struct kobject *kobj, + struct kobj_attribute *attr, char *buf) +{ + struct irq_desc *desc = container_of(kobj, struct irq_desc, kobj); + int cpu, irq = desc->irq_data.irq; + ssize_t ret = 0; + char *p = ""; + + for_each_possible_cpu(cpu) { + unsigned int c = kstat_irqs_cpu(irq, cpu); + + ret += scnprintf(buf + ret, PAGE_SIZE - ret, "%s%u", p, c); + p = ","; + } + + ret += scnprintf(buf + ret, PAGE_SIZE - ret, "\n"); + return ret; +} +IRQ_ATTR_RO(per_cpu_count); + +static ssize_t chip_name_show(struct kobject *kobj, + struct kobj_attribute *attr, char *buf) +{ + struct irq_desc *desc = container_of(kobj, struct irq_desc, kobj); + ssize_t ret = 0; + + raw_spin_lock_irq(&desc->lock); + if (desc->irq_data.chip && desc->irq_data.chip->name) { + ret = scnprintf(buf, PAGE_SIZE, "%s\n", + desc->irq_data.chip->name); + } + raw_spin_unlock_irq(&desc->lock); + + return ret; +} +IRQ_ATTR_RO(chip_name); + +static ssize_t hwirq_show(struct kobject *kobj, + struct kobj_attribute *attr, char *buf) +{ + struct irq_desc *desc = container_of(kobj, struct irq_desc, kobj); + ssize_t ret = 0; + + raw_spin_lock_irq(&desc->lock); + if (desc->irq_data.domain) + ret = sprintf(buf, "%d\n", (int)desc->irq_data.hwirq); + raw_spin_unlock_irq(&desc->lock); + + return ret; +} +IRQ_ATTR_RO(hwirq); + +static ssize_t type_show(struct kobject *kobj, + struct kobj_attribute *attr, char *buf) +{ + struct irq_desc *desc = container_of(kobj, struct irq_desc, kobj); + ssize_t ret = 0; + + raw_spin_lock_irq(&desc->lock); + ret = sprintf(buf, "%s\n", + irqd_is_level_type(&desc->irq_data) ? "level" : "edge"); + raw_spin_unlock_irq(&desc->lock); + + return ret; + +} +IRQ_ATTR_RO(type); + +static ssize_t name_show(struct kobject *kobj, + struct kobj_attribute *attr, char *buf) +{ + struct irq_desc *desc = container_of(kobj, struct irq_desc, kobj); + ssize_t ret = 0; + + raw_spin_lock_irq(&desc->lock); + if (desc->name) + ret = scnprintf(buf, PAGE_SIZE, "%s\n", desc->name); + raw_spin_unlock_irq(&desc->lock); + + return ret; +} +IRQ_ATTR_RO(name); + +static ssize_t actions_show(struct kobject *kobj, + struct kobj_attribute *attr, char *buf) +{ + struct irq_desc *desc = container_of(kobj, struct irq_desc, kobj); + struct irqaction *action; + ssize_t ret = 0; + char *p = ""; + + raw_spin_lock_irq(&desc->lock); + for (action = desc->action; action != NULL; action = action->next) { + ret += scnprintf(buf + ret, PAGE_SIZE - ret, "%s%s", + p, action->name); + p = ","; + } + raw_spin_unlock_irq(&desc->lock); + + if (ret) + ret += scnprintf(buf + ret, PAGE_SIZE - ret, "\n"); + + return ret; +} +IRQ_ATTR_RO(actions); + +static struct attribute *irq_attrs[] = { + &per_cpu_count_attr.attr, + &chip_name_attr.attr, + &hwirq_attr.attr, + &type_attr.attr, + &name_attr.attr, + &actions_attr.attr, + NULL +}; + +static struct kobj_type irq_kobj_type = { + .release = irq_kobj_release, + .sysfs_ops = &kobj_sysfs_ops, + .default_attrs = irq_attrs, +}; + +static void irq_sysfs_add(int irq, struct irq_desc *desc) +{ + if (irq_kobj_base) { + /* + * Continue even in case of failure as this is nothing + * crucial. + */ + if (kobject_add(&desc->kobj, irq_kobj_base, "%d", irq)) + pr_warn("Failed to add kobject for irq %d\n", irq); + } +} + +static int __init irq_sysfs_init(void) +{ + struct irq_desc *desc; + int irq; + + /* Prevent concurrent irq alloc/free */ + irq_lock_sparse(); + + irq_kobj_base = kobject_create_and_add("irq", kernel_kobj); + if (!irq_kobj_base) { + irq_unlock_sparse(); + return -ENOMEM; + } + + /* Add the already allocated interrupts */ + for_each_irq_desc(irq, desc) + irq_sysfs_add(irq, desc); + irq_unlock_sparse(); + + return 0; +} +postcore_initcall(irq_sysfs_init); + +#else /* !CONFIG_SYSFS */ + +static struct kobj_type irq_kobj_type = { + .release = irq_kobj_release, +}; + +static void irq_sysfs_add(int irq, struct irq_desc *desc) {} + +#endif /* CONFIG_SYSFS */ + static RADIX_TREE(irq_desc_tree, GFP_KERNEL); static void irq_insert_desc(unsigned int irq, struct irq_desc *desc) @@ -187,6 +363,7 @@ static struct irq_desc *alloc_desc(int irq, int node, unsigned int flags, desc_set_defaults(irq, desc, node, affinity, owner); irqd_set(&desc->irq_data, flags); + kobject_init(&desc->kobj, &irq_kobj_type); return desc; @@ -197,15 +374,22 @@ err_desc: return NULL; } -static void delayed_free_desc(struct rcu_head *rhp) +static void irq_kobj_release(struct kobject *kobj) { - struct irq_desc *desc = container_of(rhp, struct irq_desc, rcu); + struct irq_desc *desc = container_of(kobj, struct irq_desc, kobj); free_masks(desc); free_percpu(desc->kstat_irqs); kfree(desc); } +static void delayed_free_desc(struct rcu_head *rhp) +{ + struct irq_desc *desc = container_of(rhp, struct irq_desc, rcu); + + kobject_put(&desc->kobj); +} + static void free_desc(unsigned int irq) { struct irq_desc *desc = irq_to_desc(irq); @@ -217,8 +401,12 @@ static void free_desc(unsigned int irq) * kstat_irq_usr(). Once we deleted the descriptor from the * sparse tree we can free it. Access in proc will fail to * lookup the descriptor. + * + * The sysfs entry must be serialized against a concurrent + * irq_sysfs_init() as well. */ mutex_lock(&sparse_irq_lock); + kobject_del(&desc->kobj); delete_irq_desc(irq); mutex_unlock(&sparse_irq_lock); @@ -261,6 +449,7 @@ static int alloc_descs(unsigned int start, unsigned int cnt, int node, goto err; mutex_lock(&sparse_irq_lock); irq_insert_desc(start + i, desc); + irq_sysfs_add(start + i, desc); mutex_unlock(&sparse_irq_lock); } return start; -- cgit v1.2.3 From 28f4b04143c56135b1ca742fc64b664ed04de6a4 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Wed, 14 Sep 2016 16:18:47 +0200 Subject: genirq/msi: Add cpumask allocation to alloc_msi_entry For irq spreading want to store affinity masks in the msi_entry. Add the infrastructure for it. We allocate an array of cpumasks with an array size of the number of used vectors in the entry, so we can hand in the information per linux interrupt later. As we hand in the number of used vectors, we assign them right away. Convert all the call sites. Signed-off-by: Thomas Gleixner Cc: axboe@fb.com Cc: keith.busch@intel.com Cc: agordeev@redhat.com Cc: linux-block@vger.kernel.org Cc: Christoph Hellwig Link: http://lkml.kernel.org/r/1473862739-15032-2-git-send-email-hch@lst.de --- drivers/base/platform-msi.c | 3 +-- drivers/pci/msi.c | 6 ++---- drivers/staging/fsl-mc/bus/mc-msi.c | 3 +-- include/linux/msi.h | 5 +++-- kernel/irq/msi.c | 26 ++++++++++++++++++++++++-- 5 files changed, 31 insertions(+), 12 deletions(-) (limited to 'kernel') diff --git a/drivers/base/platform-msi.c b/drivers/base/platform-msi.c index 279e53989374..be6a599bc0c1 100644 --- a/drivers/base/platform-msi.c +++ b/drivers/base/platform-msi.c @@ -142,13 +142,12 @@ static int platform_msi_alloc_descs_with_irq(struct device *dev, int virq, } for (i = 0; i < nvec; i++) { - desc = alloc_msi_entry(dev); + desc = alloc_msi_entry(dev, 1, NULL); if (!desc) break; desc->platform.msi_priv_data = data; desc->platform.msi_index = base + i; - desc->nvec_used = 1; desc->irq = virq ? virq + i : 0; list_add_tail(&desc->list, dev_to_msi_list(dev)); diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c index 98f12223c734..0db72ba24003 100644 --- a/drivers/pci/msi.c +++ b/drivers/pci/msi.c @@ -555,7 +555,7 @@ static struct msi_desc *msi_setup_entry(struct pci_dev *dev, int nvec) struct msi_desc *entry; /* MSI Entry Initialization */ - entry = alloc_msi_entry(&dev->dev); + entry = alloc_msi_entry(&dev->dev, nvec, NULL); if (!entry) return NULL; @@ -568,7 +568,6 @@ static struct msi_desc *msi_setup_entry(struct pci_dev *dev, int nvec) entry->msi_attrib.default_irq = dev->irq; /* Save IOAPIC IRQ */ entry->msi_attrib.multi_cap = (control & PCI_MSI_FLAGS_QMASK) >> 1; entry->msi_attrib.multiple = ilog2(__roundup_pow_of_two(nvec)); - entry->nvec_used = nvec; entry->affinity = dev->irq_affinity; if (control & PCI_MSI_FLAGS_64BIT) @@ -693,7 +692,7 @@ static int msix_setup_entries(struct pci_dev *dev, void __iomem *base, mask = cpumask_of(cpu); } - entry = alloc_msi_entry(&dev->dev); + entry = alloc_msi_entry(&dev->dev, 1, NULL); if (!entry) { if (!i) iounmap(base); @@ -711,7 +710,6 @@ static int msix_setup_entries(struct pci_dev *dev, void __iomem *base, entry->msi_attrib.entry_nr = i; entry->msi_attrib.default_irq = dev->irq; entry->mask_base = base; - entry->nvec_used = 1; entry->affinity = mask; list_add_tail(&entry->list, dev_to_msi_list(&dev->dev)); diff --git a/drivers/staging/fsl-mc/bus/mc-msi.c b/drivers/staging/fsl-mc/bus/mc-msi.c index c7be156ae5e0..4fd8e41ef468 100644 --- a/drivers/staging/fsl-mc/bus/mc-msi.c +++ b/drivers/staging/fsl-mc/bus/mc-msi.c @@ -213,7 +213,7 @@ static int fsl_mc_msi_alloc_descs(struct device *dev, unsigned int irq_count) struct msi_desc *msi_desc; for (i = 0; i < irq_count; i++) { - msi_desc = alloc_msi_entry(dev); + msi_desc = alloc_msi_entry(dev, 1, NULL); if (!msi_desc) { dev_err(dev, "Failed to allocate msi entry\n"); error = -ENOMEM; @@ -221,7 +221,6 @@ static int fsl_mc_msi_alloc_descs(struct device *dev, unsigned int irq_count) } msi_desc->fsl_mc.msi_index = i; - msi_desc->nvec_used = 1; INIT_LIST_HEAD(&msi_desc->list); list_add_tail(&msi_desc->list, dev_to_msi_list(dev)); } diff --git a/include/linux/msi.h b/include/linux/msi.h index e8c81fbd5f9c..0db320b7bb15 100644 --- a/include/linux/msi.h +++ b/include/linux/msi.h @@ -68,7 +68,7 @@ struct msi_desc { unsigned int nvec_used; struct device *dev; struct msi_msg msg; - const struct cpumask *affinity; + struct cpumask *affinity; union { /* PCI MSI/X specific data */ @@ -123,7 +123,8 @@ static inline void *msi_desc_to_pci_sysdata(struct msi_desc *desc) } #endif /* CONFIG_PCI_MSI */ -struct msi_desc *alloc_msi_entry(struct device *dev); +struct msi_desc *alloc_msi_entry(struct device *dev, int nvec, + const struct cpumask *affinity); void free_msi_entry(struct msi_desc *entry); void __pci_read_msi_msg(struct msi_desc *entry, struct msi_msg *msg); void __pci_write_msi_msg(struct msi_desc *entry, struct msi_msg *msg); diff --git a/kernel/irq/msi.c b/kernel/irq/msi.c index 19e9dfbe97fa..8a3e872798f3 100644 --- a/kernel/irq/msi.c +++ b/kernel/irq/msi.c @@ -18,20 +18,42 @@ /* Temparory solution for building, will be removed later */ #include -struct msi_desc *alloc_msi_entry(struct device *dev) +/** + * alloc_msi_entry - Allocate an initialize msi_entry + * @dev: Pointer to the device for which this is allocated + * @nvec: The number of vectors used in this entry + * @affinity: Optional pointer to an affinity mask array size of @nvec + * + * If @affinity is not NULL then a an affinity array[@nvec] is allocated + * and the affinity masks from @affinity are copied. + */ +struct msi_desc * +alloc_msi_entry(struct device *dev, int nvec, const struct cpumask *affinity) { - struct msi_desc *desc = kzalloc(sizeof(*desc), GFP_KERNEL); + struct msi_desc *desc; + + desc = kzalloc(sizeof(*desc), GFP_KERNEL); if (!desc) return NULL; INIT_LIST_HEAD(&desc->list); desc->dev = dev; + desc->nvec_used = nvec; + if (affinity) { + desc->affinity = kmemdup(affinity, + nvec * sizeof(*desc->affinity), GFP_KERNEL); + if (!desc->affinity) { + kfree(desc); + return NULL; + } + } return desc; } void free_msi_entry(struct msi_desc *entry) { + kfree(entry->affinity); kfree(entry); } -- cgit v1.2.3 From 34c3d9819fda464be4f1bec59b63353814f76c73 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Wed, 14 Sep 2016 16:18:48 +0200 Subject: genirq/affinity: Provide smarter irq spreading infrastructure The current irq spreading infrastructure is just looking at a cpumask and tries to spread the interrupts over the mask. Thats suboptimal as it does not take numa nodes into account. Change the logic so the interrupts are spread across numa nodes and inside the nodes. If there are more cpus than vectors per node, then we set the affinity to several cpus. If HT siblings are available we take that into account and try to set all siblings to a single vector. Signed-off-by: Thomas Gleixner Cc: Christoph Hellwig Cc: axboe@fb.com Cc: keith.busch@intel.com Cc: agordeev@redhat.com Cc: linux-block@vger.kernel.org Link: http://lkml.kernel.org/r/1473862739-15032-3-git-send-email-hch@lst.de --- include/linux/interrupt.h | 15 +++++ kernel/irq/affinity.c | 149 ++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 164 insertions(+) (limited to 'kernel') diff --git a/include/linux/interrupt.h b/include/linux/interrupt.h index b6683f0ffc9f..4e59d122cad9 100644 --- a/include/linux/interrupt.h +++ b/include/linux/interrupt.h @@ -279,6 +279,8 @@ extern int irq_set_affinity_notifier(unsigned int irq, struct irq_affinity_notify *notify); struct cpumask *irq_create_affinity_mask(unsigned int *nr_vecs); +struct cpumask *irq_create_affinity_masks(const struct cpumask *affinity, int nvec); +int irq_calc_affinity_vectors(const struct cpumask *affinity, int maxvec); #else /* CONFIG_SMP */ @@ -316,6 +318,19 @@ static inline struct cpumask *irq_create_affinity_mask(unsigned int *nr_vecs) *nr_vecs = 1; return NULL; } + +static inline struct cpumask * +irq_create_affinity_masks(const struct cpumask *affinity, int nvec) +{ + return NULL; +} + +static inline int +irq_calc_affinity_vectors(const struct cpumask *affinity, int maxvec) +{ + return maxvec; +} + #endif /* CONFIG_SMP */ /* diff --git a/kernel/irq/affinity.c b/kernel/irq/affinity.c index 32f6cfcff212..7812fecc6e2f 100644 --- a/kernel/irq/affinity.c +++ b/kernel/irq/affinity.c @@ -4,6 +4,155 @@ #include #include +static void irq_spread_init_one(struct cpumask *irqmsk, struct cpumask *nmsk, + int cpus_per_vec) +{ + const struct cpumask *siblmsk; + int cpu, sibl; + + for ( ; cpus_per_vec > 0; ) { + cpu = cpumask_first(nmsk); + + /* Should not happen, but I'm too lazy to think about it */ + if (cpu >= nr_cpu_ids) + return; + + cpumask_clear_cpu(cpu, nmsk); + cpumask_set_cpu(cpu, irqmsk); + cpus_per_vec--; + + /* If the cpu has siblings, use them first */ + siblmsk = topology_sibling_cpumask(cpu); + for (sibl = -1; cpus_per_vec > 0; ) { + sibl = cpumask_next(sibl, siblmsk); + if (sibl >= nr_cpu_ids) + break; + if (!cpumask_test_and_clear_cpu(sibl, nmsk)) + continue; + cpumask_set_cpu(sibl, irqmsk); + cpus_per_vec--; + } + } +} + +static int get_nodes_in_cpumask(const struct cpumask *mask, nodemask_t *nodemsk) +{ + int n, nodes; + + /* Calculate the number of nodes in the supplied affinity mask */ + for (n = 0, nodes = 0; n < num_online_nodes(); n++) { + if (cpumask_intersects(mask, cpumask_of_node(n))) { + node_set(n, *nodemsk); + nodes++; + } + } + return nodes; +} + +/** + * irq_create_affinity_masks - Create affinity masks for multiqueue spreading + * @affinity: The affinity mask to spread. If NULL cpu_online_mask + * is used + * @nvecs: The number of vectors + * + * Returns the masks pointer or NULL if allocation failed. + */ +struct cpumask *irq_create_affinity_masks(const struct cpumask *affinity, + int nvec) +{ + int n, nodes, vecs_per_node, cpus_per_vec, extra_vecs, curvec = 0; + nodemask_t nodemsk = NODE_MASK_NONE; + struct cpumask *masks; + cpumask_var_t nmsk; + + if (!zalloc_cpumask_var(&nmsk, GFP_KERNEL)) + return NULL; + + masks = kzalloc(nvec * sizeof(*masks), GFP_KERNEL); + if (!masks) + goto out; + + /* Stabilize the cpumasks */ + get_online_cpus(); + /* If the supplied affinity mask is NULL, use cpu online mask */ + if (!affinity) + affinity = cpu_online_mask; + + nodes = get_nodes_in_cpumask(affinity, &nodemsk); + + /* + * If the number of nodes in the mask is less than or equal the + * number of vectors we just spread the vectors across the nodes. + */ + if (nvec <= nodes) { + for_each_node_mask(n, nodemsk) { + cpumask_copy(masks + curvec, cpumask_of_node(n)); + if (++curvec == nvec) + break; + } + goto outonl; + } + + /* Spread the vectors per node */ + vecs_per_node = nvec / nodes; + /* Account for rounding errors */ + extra_vecs = nvec - (nodes * vecs_per_node); + + for_each_node_mask(n, nodemsk) { + int ncpus, v, vecs_to_assign = vecs_per_node; + + /* Get the cpus on this node which are in the mask */ + cpumask_and(nmsk, affinity, cpumask_of_node(n)); + + /* Calculate the number of cpus per vector */ + ncpus = cpumask_weight(nmsk); + + for (v = 0; curvec < nvec && v < vecs_to_assign; curvec++, v++) { + cpus_per_vec = ncpus / vecs_to_assign; + + /* Account for extra vectors to compensate rounding errors */ + if (extra_vecs) { + cpus_per_vec++; + if (!--extra_vecs) + vecs_per_node++; + } + irq_spread_init_one(masks + curvec, nmsk, cpus_per_vec); + } + + if (curvec >= nvec) + break; + } + +outonl: + put_online_cpus(); +out: + free_cpumask_var(nmsk); + return masks; +} + +/** + * irq_calc_affinity_vectors - Calculate to optimal number of vectors for a given affinity mask + * @affinity: The affinity mask to spread. If NULL cpu_online_mask + * is used + * @maxvec: The maximum number of vectors available + */ +int irq_calc_affinity_vectors(const struct cpumask *affinity, int maxvec) +{ + int cpus, ret; + + /* Stabilize the cpumasks */ + get_online_cpus(); + /* If the supplied affinity mask is NULL, use cpu online mask */ + if (!affinity) + affinity = cpu_online_mask; + + cpus = cpumask_weight(affinity); + ret = (cpus < maxvec) ? cpus : maxvec; + + put_online_cpus(); + return ret; +} + static int get_first_sibling(unsigned int cpu) { unsigned int ret; -- cgit v1.2.3 From e75eafb9b0395c338230b0eef2cc92ca8d20dee2 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Wed, 14 Sep 2016 16:18:49 +0200 Subject: genirq/msi: Switch to new irq spreading infrastructure Switch MSI over to the new spreading code. If a pci device contains a valid pointer to a cpumask, then this mask is used for spreading otherwise the online cpu mask is used. This allows a driver to restrict the spread to a subset of CPUs, e.g. cpus on a particular node. Signed-off-by: Thomas Gleixner Cc: Christoph Hellwig Cc: axboe@fb.com Cc: keith.busch@intel.com Cc: agordeev@redhat.com Cc: linux-block@vger.kernel.org Link: http://lkml.kernel.org/r/1473862739-15032-4-git-send-email-hch@lst.de Signed-off-by: Thomas Gleixner --- drivers/pci/msi.c | 128 +++++++++++++++++++++++++++++---------------------- kernel/irq/irqdesc.c | 31 ++++++------- 2 files changed, 87 insertions(+), 72 deletions(-) (limited to 'kernel') diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c index 0db72ba24003..06100dde0e86 100644 --- a/drivers/pci/msi.c +++ b/drivers/pci/msi.c @@ -549,15 +549,23 @@ error_attrs: return ret; } -static struct msi_desc *msi_setup_entry(struct pci_dev *dev, int nvec) +static struct msi_desc * +msi_setup_entry(struct pci_dev *dev, int nvec, bool affinity) { - u16 control; + struct cpumask *masks = NULL; struct msi_desc *entry; + u16 control; + + if (affinity) { + masks = irq_create_affinity_masks(dev->irq_affinity, nvec); + if (!masks) + pr_err("Unable to allocate affinity masks, ignoring\n"); + } /* MSI Entry Initialization */ - entry = alloc_msi_entry(&dev->dev, nvec, NULL); + entry = alloc_msi_entry(&dev->dev, nvec, masks); if (!entry) - return NULL; + goto out; pci_read_config_word(dev, dev->msi_cap + PCI_MSI_FLAGS, &control); @@ -568,7 +576,6 @@ static struct msi_desc *msi_setup_entry(struct pci_dev *dev, int nvec) entry->msi_attrib.default_irq = dev->irq; /* Save IOAPIC IRQ */ entry->msi_attrib.multi_cap = (control & PCI_MSI_FLAGS_QMASK) >> 1; entry->msi_attrib.multiple = ilog2(__roundup_pow_of_two(nvec)); - entry->affinity = dev->irq_affinity; if (control & PCI_MSI_FLAGS_64BIT) entry->mask_pos = dev->msi_cap + PCI_MSI_MASK_64; @@ -579,6 +586,8 @@ static struct msi_desc *msi_setup_entry(struct pci_dev *dev, int nvec) if (entry->msi_attrib.maskbit) pci_read_config_dword(dev, entry->mask_pos, &entry->masked); +out: + kfree(masks); return entry; } @@ -607,7 +616,7 @@ static int msi_verify_entries(struct pci_dev *dev) * an error, and a positive return value indicates the number of interrupts * which could have been allocated. */ -static int msi_capability_init(struct pci_dev *dev, int nvec) +static int msi_capability_init(struct pci_dev *dev, int nvec, bool affinity) { struct msi_desc *entry; int ret; @@ -615,7 +624,7 @@ static int msi_capability_init(struct pci_dev *dev, int nvec) pci_msi_set_enable(dev, 0); /* Disable MSI during set up */ - entry = msi_setup_entry(dev, nvec); + entry = msi_setup_entry(dev, nvec, affinity); if (!entry) return -ENOMEM; @@ -678,28 +687,29 @@ static void __iomem *msix_map_region(struct pci_dev *dev, unsigned nr_entries) } static int msix_setup_entries(struct pci_dev *dev, void __iomem *base, - struct msix_entry *entries, int nvec) + struct msix_entry *entries, int nvec, + bool affinity) { - const struct cpumask *mask = NULL; + struct cpumask *curmsk, *masks = NULL; struct msi_desc *entry; - int cpu = -1, i; - - for (i = 0; i < nvec; i++) { - if (dev->irq_affinity) { - cpu = cpumask_next(cpu, dev->irq_affinity); - if (cpu >= nr_cpu_ids) - cpu = cpumask_first(dev->irq_affinity); - mask = cpumask_of(cpu); - } + int ret, i; - entry = alloc_msi_entry(&dev->dev, 1, NULL); + if (affinity) { + masks = irq_create_affinity_masks(dev->irq_affinity, nvec); + if (!masks) + pr_err("Unable to allocate affinity masks, ignoring\n"); + } + + for (i = 0, curmsk = masks; i < nvec; i++) { + entry = alloc_msi_entry(&dev->dev, 1, curmsk); if (!entry) { if (!i) iounmap(base); else free_msi_irqs(dev); /* No enough memory. Don't try again */ - return -ENOMEM; + ret = -ENOMEM; + goto out; } entry->msi_attrib.is_msix = 1; @@ -710,11 +720,14 @@ static int msix_setup_entries(struct pci_dev *dev, void __iomem *base, entry->msi_attrib.entry_nr = i; entry->msi_attrib.default_irq = dev->irq; entry->mask_base = base; - entry->affinity = mask; list_add_tail(&entry->list, dev_to_msi_list(&dev->dev)); + if (masks) + curmsk++; } - + ret = 0; +out: + kfree(masks); return 0; } @@ -743,8 +756,8 @@ static void msix_program_entries(struct pci_dev *dev, * single MSI-X irq. A return of zero indicates the successful setup of * requested MSI-X entries with allocated irqs or non-zero for otherwise. **/ -static int msix_capability_init(struct pci_dev *dev, - struct msix_entry *entries, int nvec) +static int msix_capability_init(struct pci_dev *dev, struct msix_entry *entries, + int nvec, bool affinity) { int ret; u16 control; @@ -759,7 +772,7 @@ static int msix_capability_init(struct pci_dev *dev, if (!base) return -ENOMEM; - ret = msix_setup_entries(dev, base, entries, nvec); + ret = msix_setup_entries(dev, base, entries, nvec, affinity); if (ret) return ret; @@ -939,22 +952,8 @@ int pci_msix_vec_count(struct pci_dev *dev) } EXPORT_SYMBOL(pci_msix_vec_count); -/** - * pci_enable_msix - configure device's MSI-X capability structure - * @dev: pointer to the pci_dev data structure of MSI-X device function - * @entries: pointer to an array of MSI-X entries (optional) - * @nvec: number of MSI-X irqs requested for allocation by device driver - * - * Setup the MSI-X capability structure of device function with the number - * of requested irqs upon its software driver call to request for - * MSI-X mode enabled on its hardware device function. A return of zero - * indicates the successful configuration of MSI-X capability structure - * with new allocated MSI-X irqs. A return of < 0 indicates a failure. - * Or a return of > 0 indicates that driver request is exceeding the number - * of irqs or MSI-X vectors available. Driver should use the returned value to - * re-send its request. - **/ -int pci_enable_msix(struct pci_dev *dev, struct msix_entry *entries, int nvec) +static int __pci_enable_msix(struct pci_dev *dev, struct msix_entry *entries, + int nvec, bool affinity) { int nr_entries; int i, j; @@ -986,7 +985,27 @@ int pci_enable_msix(struct pci_dev *dev, struct msix_entry *entries, int nvec) dev_info(&dev->dev, "can't enable MSI-X (MSI IRQ already assigned)\n"); return -EINVAL; } - return msix_capability_init(dev, entries, nvec); + return msix_capability_init(dev, entries, nvec, affinity); +} + +/** + * pci_enable_msix - configure device's MSI-X capability structure + * @dev: pointer to the pci_dev data structure of MSI-X device function + * @entries: pointer to an array of MSI-X entries (optional) + * @nvec: number of MSI-X irqs requested for allocation by device driver + * + * Setup the MSI-X capability structure of device function with the number + * of requested irqs upon its software driver call to request for + * MSI-X mode enabled on its hardware device function. A return of zero + * indicates the successful configuration of MSI-X capability structure + * with new allocated MSI-X irqs. A return of < 0 indicates a failure. + * Or a return of > 0 indicates that driver request is exceeding the number + * of irqs or MSI-X vectors available. Driver should use the returned value to + * re-send its request. + **/ +int pci_enable_msix(struct pci_dev *dev, struct msix_entry *entries, int nvec) +{ + return __pci_enable_msix(dev, entries, nvec, false); } EXPORT_SYMBOL(pci_enable_msix); @@ -1039,6 +1058,7 @@ EXPORT_SYMBOL(pci_msi_enabled); static int __pci_enable_msi_range(struct pci_dev *dev, int minvec, int maxvec, unsigned int flags) { + bool affinity = flags & PCI_IRQ_AFFINITY; int nvec; int rc; @@ -1067,19 +1087,17 @@ static int __pci_enable_msi_range(struct pci_dev *dev, int minvec, int maxvec, nvec = maxvec; for (;;) { - if (flags & PCI_IRQ_AFFINITY) { - dev->irq_affinity = irq_create_affinity_mask(&nvec); + if (affinity) { + nvec = irq_calc_affinity_vectors(dev->irq_affinity, + nvec); if (nvec < minvec) return -ENOSPC; } - rc = msi_capability_init(dev, nvec); + rc = msi_capability_init(dev, nvec, affinity); if (rc == 0) return nvec; - kfree(dev->irq_affinity); - dev->irq_affinity = NULL; - if (rc < 0) return rc; if (rc < minvec) @@ -1111,26 +1129,24 @@ static int __pci_enable_msix_range(struct pci_dev *dev, struct msix_entry *entries, int minvec, int maxvec, unsigned int flags) { - int nvec = maxvec; - int rc; + bool affinity = flags & PCI_IRQ_AFFINITY; + int rc, nvec = maxvec; if (maxvec < minvec) return -ERANGE; for (;;) { - if (flags & PCI_IRQ_AFFINITY) { - dev->irq_affinity = irq_create_affinity_mask(&nvec); + if (affinity) { + nvec = irq_calc_affinity_vectors(dev->irq_affinity, + nvec); if (nvec < minvec) return -ENOSPC; } - rc = pci_enable_msix(dev, entries, nvec); + rc = __pci_enable_msix(dev, entries, nvec, affinity); if (rc == 0) return nvec; - kfree(dev->irq_affinity); - dev->irq_affinity = NULL; - if (rc < 0) return rc; if (rc < minvec) diff --git a/kernel/irq/irqdesc.c b/kernel/irq/irqdesc.c index a623b44f2d4b..5a5a685aba33 100644 --- a/kernel/irq/irqdesc.c +++ b/kernel/irq/irqdesc.c @@ -236,25 +236,24 @@ static int alloc_descs(unsigned int start, unsigned int cnt, int node, const struct cpumask *mask = NULL; struct irq_desc *desc; unsigned int flags; - int i, cpu = -1; + int i; - if (affinity && cpumask_empty(affinity)) - return -EINVAL; + /* Validate affinity mask(s) */ + if (affinity) { + for (i = 0, mask = affinity; i < cnt; i++, mask++) { + if (cpumask_empty(mask)) + return -EINVAL; + } + } flags = affinity ? IRQD_AFFINITY_MANAGED : 0; + mask = NULL; for (i = 0; i < cnt; i++) { if (affinity) { - cpu = cpumask_next(cpu, affinity); - if (cpu >= nr_cpu_ids) - cpu = cpumask_first(affinity); - node = cpu_to_node(cpu); - - /* - * For single allocations we use the caller provided - * mask otherwise we use the mask of the target cpu - */ - mask = cnt == 1 ? affinity : cpumask_of(cpu); + node = cpu_to_node(cpumask_first(affinity)); + mask = affinity; + affinity++; } desc = alloc_desc(start + i, node, flags, mask, owner); if (!desc) @@ -481,9 +480,9 @@ EXPORT_SYMBOL_GPL(irq_free_descs); * @cnt: Number of consecutive irqs to allocate. * @node: Preferred node on which the irq descriptor should be allocated * @owner: Owning module (can be NULL) - * @affinity: Optional pointer to an affinity mask which hints where the - * irq descriptors should be allocated and which default - * affinities to use + * @affinity: Optional pointer to an affinity mask array of size @cnt which + * hints where the irq descriptors should be allocated and which + * default affinities to use * * Returns the first irq number or error code */ -- cgit v1.2.3 From 44082fd6702fb12020967fd375f8bf6dd7c111bf Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Wed, 14 Sep 2016 16:18:50 +0200 Subject: genirq/affinity: Remove old irq spread infrastructure No more users. Signed-off-by: Thomas Gleixner Cc: Christoph Hellwig Cc: axboe@fb.com Cc: keith.busch@intel.com Cc: agordeev@redhat.com Cc: linux-block@vger.kernel.org Link: http://lkml.kernel.org/r/1473862739-15032-5-git-send-email-hch@lst.de Signed-off-by: Thomas Gleixner --- include/linux/interrupt.h | 7 ------ kernel/irq/affinity.c | 58 ----------------------------------------------- 2 files changed, 65 deletions(-) (limited to 'kernel') diff --git a/include/linux/interrupt.h b/include/linux/interrupt.h index 4e59d122cad9..72f0721f75e7 100644 --- a/include/linux/interrupt.h +++ b/include/linux/interrupt.h @@ -278,7 +278,6 @@ extern int irq_set_affinity_hint(unsigned int irq, const struct cpumask *m); extern int irq_set_affinity_notifier(unsigned int irq, struct irq_affinity_notify *notify); -struct cpumask *irq_create_affinity_mask(unsigned int *nr_vecs); struct cpumask *irq_create_affinity_masks(const struct cpumask *affinity, int nvec); int irq_calc_affinity_vectors(const struct cpumask *affinity, int maxvec); @@ -313,12 +312,6 @@ irq_set_affinity_notifier(unsigned int irq, struct irq_affinity_notify *notify) return 0; } -static inline struct cpumask *irq_create_affinity_mask(unsigned int *nr_vecs) -{ - *nr_vecs = 1; - return NULL; -} - static inline struct cpumask * irq_create_affinity_masks(const struct cpumask *affinity, int nvec) { diff --git a/kernel/irq/affinity.c b/kernel/irq/affinity.c index 7812fecc6e2f..17f51d63da56 100644 --- a/kernel/irq/affinity.c +++ b/kernel/irq/affinity.c @@ -152,61 +152,3 @@ int irq_calc_affinity_vectors(const struct cpumask *affinity, int maxvec) put_online_cpus(); return ret; } - -static int get_first_sibling(unsigned int cpu) -{ - unsigned int ret; - - ret = cpumask_first(topology_sibling_cpumask(cpu)); - if (ret < nr_cpu_ids) - return ret; - return cpu; -} - -/* - * Take a map of online CPUs and the number of available interrupt vectors - * and generate an output cpumask suitable for spreading MSI/MSI-X vectors - * so that they are distributed as good as possible around the CPUs. If - * more vectors than CPUs are available we'll map one to each CPU, - * otherwise we map one to the first sibling of each socket. - * - * If there are more vectors than CPUs we will still only have one bit - * set per CPU, but interrupt code will keep on assigning the vectors from - * the start of the bitmap until we run out of vectors. - */ -struct cpumask *irq_create_affinity_mask(unsigned int *nr_vecs) -{ - struct cpumask *affinity_mask; - unsigned int max_vecs = *nr_vecs; - - if (max_vecs == 1) - return NULL; - - affinity_mask = kzalloc(cpumask_size(), GFP_KERNEL); - if (!affinity_mask) { - *nr_vecs = 1; - return NULL; - } - - get_online_cpus(); - if (max_vecs >= num_online_cpus()) { - cpumask_copy(affinity_mask, cpu_online_mask); - *nr_vecs = num_online_cpus(); - } else { - unsigned int vecs = 0, cpu; - - for_each_online_cpu(cpu) { - if (cpu == get_first_sibling(cpu)) { - cpumask_set_cpu(cpu, affinity_mask); - vecs++; - } - - if (--max_vecs == 0) - break; - } - *nr_vecs = vecs; - } - put_online_cpus(); - - return affinity_mask; -} -- cgit v1.2.3 From b8129a1f6aaaca02d92186acf19ceb545b4b489a Mon Sep 17 00:00:00 2001 From: Wei Yongjun Date: Sun, 25 Sep 2016 15:36:39 +0000 Subject: genirq: Make function __irq_do_set_handler() static Fixes the following sparse warning: kernel/irq/chip.c:786:1: warning: symbol '__irq_do_set_handler' was not declared. Should it be static? Signed-off-by: Wei Yongjun Link: http://lkml.kernel.org/r/1474817799-18676-1-git-send-email-weiyj.lk@gmail.com Signed-off-by: Thomas Gleixner --- kernel/irq/chip.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c index d8dfdc630b7e..be3c34e4f2ac 100644 --- a/kernel/irq/chip.c +++ b/kernel/irq/chip.c @@ -782,7 +782,7 @@ void handle_percpu_devid_irq(struct irq_desc *desc) chip->irq_eoi(&desc->irq_data); } -void +static void __irq_do_set_handler(struct irq_desc *desc, irq_flow_handler_t handle, int is_chained, const char *name) { -- cgit v1.2.3 From 4cd13c21b207e80ddb1144c576500098f2d5f882 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Wed, 31 Aug 2016 10:42:29 -0700 Subject: softirq: Let ksoftirqd do its job A while back, Paolo and Hannes sent an RFC patch adding threaded-able napi poll loop support : (https://patchwork.ozlabs.org/patch/620657/) The problem seems to be that softirqs are very aggressive and are often handled by the current process, even if we are under stress and that ksoftirqd was scheduled, so that innocent threads would have more chance to make progress. This patch makes sure that if ksoftirq is running, we let it perform the softirq work. Jonathan Corbet summarized the issue in https://lwn.net/Articles/687617/ Tested: - NIC receiving traffic handled by CPU 0 - UDP receiver running on CPU 0, using a single UDP socket. - Incoming flood of UDP packets targeting the UDP socket. Before the patch, the UDP receiver could almost never get CPU cycles and could only receive ~2,000 packets per second. After the patch, CPU cycles are split 50/50 between user application and ksoftirqd/0, and we can effectively read ~900,000 packets per second, a huge improvement in DOS situation. (Note that more packets are now dropped by the NIC itself, since the BH handlers get less CPU cycles to drain RX ring buffer) Since the load runs in well identified threads context, an admin can more easily tune process scheduling parameters if needed. Reported-by: Paolo Abeni Reported-by: Hannes Frederic Sowa Signed-off-by: Eric Dumazet Signed-off-by: Peter Zijlstra (Intel) Cc: David Miller Cc: Hannes Frederic Sowa Cc: Jesper Dangaard Brouer Cc: Jonathan Corbet Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Rik van Riel Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1472665349.14381.356.camel@edumazet-glaptop3.roam.corp.google.com Signed-off-by: Ingo Molnar --- kernel/softirq.c | 16 +++++++++++++++- 1 file changed, 15 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/softirq.c b/kernel/softirq.c index 17caf4b63342..8ed90e3a88d6 100644 --- a/kernel/softirq.c +++ b/kernel/softirq.c @@ -77,6 +77,17 @@ static void wakeup_softirqd(void) wake_up_process(tsk); } +/* + * If ksoftirqd is scheduled, we do not want to process pending softirqs + * right now. Let ksoftirqd handle this at its own rate, to get fairness. + */ +static bool ksoftirqd_running(void) +{ + struct task_struct *tsk = __this_cpu_read(ksoftirqd); + + return tsk && (tsk->state == TASK_RUNNING); +} + /* * preempt_count and SOFTIRQ_OFFSET usage: * - preempt_count is changed by SOFTIRQ_OFFSET on entering or leaving @@ -313,7 +324,7 @@ asmlinkage __visible void do_softirq(void) pending = local_softirq_pending(); - if (pending) + if (pending && !ksoftirqd_running()) do_softirq_own_stack(); local_irq_restore(flags); @@ -340,6 +351,9 @@ void irq_enter(void) static inline void invoke_softirq(void) { + if (ksoftirqd_running()) + return; + if (!force_irqthreads) { #ifdef CONFIG_HAVE_IRQ_EXIT_ON_IRQ_STACK /* -- cgit v1.2.3