From f6d87f4bd259cf33e092cd1a8fde05f291c47af1 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Fri, 7 Nov 2008 13:18:30 +0100 Subject: genirq: keep affinities set from userspace across free/request_irq() Impact: preserve user-modified affinities on interrupts Kumar Galak noticed that commit 18404756765c713a0be4eb1082920c04822ce588 (genirq: Expose default irq affinity mask (take 3)) overrides an already set affinity setting across a free / request_irq(). Happens e.g. with ifdown/ifup of a network device. Change the logic to mark the affinities as set and keep them intact. This also fixes the unlocked access to irq_desc in irq_select_affinity() when called from irq_affinity_proc_write() Signed-off-by: Thomas Gleixner Signed-off-by: Ingo Molnar --- kernel/irq/manage.c | 58 ++++++++++++++++++++++++++++++++++++++++++++--------- 1 file changed, 48 insertions(+), 10 deletions(-) (limited to 'kernel/irq/manage.c') diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c index c498a1b8c621..634a2a955104 100644 --- a/kernel/irq/manage.c +++ b/kernel/irq/manage.c @@ -82,24 +82,27 @@ int irq_can_set_affinity(unsigned int irq) int irq_set_affinity(unsigned int irq, cpumask_t cpumask) { struct irq_desc *desc = irq_to_desc(irq); + unsigned long flags; if (!desc->chip->set_affinity) return -EINVAL; + spin_lock_irqsave(&desc->lock, flags); + #ifdef CONFIG_GENERIC_PENDING_IRQ if (desc->status & IRQ_MOVE_PCNTXT || desc->status & IRQ_DISABLED) { - unsigned long flags; - - spin_lock_irqsave(&desc->lock, flags); desc->affinity = cpumask; desc->chip->set_affinity(irq, cpumask); - spin_unlock_irqrestore(&desc->lock, flags); - } else - set_pending_irq(irq, cpumask); + } else { + desc->status |= IRQ_MOVE_PENDING; + desc->pending_mask = cpumask; + } #else desc->affinity = cpumask; desc->chip->set_affinity(irq, cpumask); #endif + desc->status |= IRQ_AFFINITY_SET; + spin_unlock_irqrestore(&desc->lock, flags); return 0; } @@ -107,24 +110,59 @@ int irq_set_affinity(unsigned int irq, cpumask_t cpumask) /* * Generic version of the affinity autoselector. */ -int irq_select_affinity(unsigned int irq) +int do_irq_select_affinity(unsigned int irq, struct irq_desc *desc) { cpumask_t mask; - struct irq_desc *desc; if (!irq_can_set_affinity(irq)) return 0; cpus_and(mask, cpu_online_map, irq_default_affinity); - desc = irq_to_desc(irq); + /* + * Preserve an userspace affinity setup, but make sure that + * one of the targets is online. + */ + if (desc->status & IRQ_AFFINITY_SET) { + if (cpus_intersects(desc->affinity, cpu_online_map)) + mask = desc->affinity; + else + desc->status &= ~IRQ_AFFINITY_SET; + } + desc->affinity = mask; desc->chip->set_affinity(irq, mask); return 0; } +#else +static inline int do_irq_select_affinity(unsigned int irq, struct irq_desc *d) +{ + return irq_select_affinity(irq); +} #endif +/* + * Called when affinity is set via /proc/irq + */ +int irq_select_affinity_usr(unsigned int irq) +{ + struct irq_desc *desc = irq_to_desc(irq); + unsigned long flags; + int ret; + + spin_lock_irqsave(&desc->lock, flags); + ret = do_irq_select_affinity(irq, desc); + spin_unlock_irqrestore(&desc->lock, flags); + + return ret; +} + +#else +static inline int do_select_irq_affinity(int irq, struct irq_desc *desc) +{ + return 0; +} #endif /** @@ -446,7 +484,7 @@ __setup_irq(unsigned int irq, struct irq_desc * desc, struct irqaction *new) desc->depth = 1; /* Set default affinity mask once everything is setup */ - irq_select_affinity(irq); + do_irq_select_affinity(irq, desc); } else if ((new->flags & IRQF_TRIGGER_MASK) && (new->flags & IRQF_TRIGGER_MASK) -- cgit v1.2.3-70-g09d2 From 612e3684c1b7752d2890510e4f90115fd1eb2afb Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Fri, 7 Nov 2008 13:58:46 +0100 Subject: genirq: fix the affinity setting in setup_irq The affinity setting in setup irq is called before the NO_BALANCING flag is checked and might therefore override affinity settings from the calling code with the default setting. Move the NO_BALANCING flag check before the call to the affinity setting. Signed-off-by: Thomas Gleixner Signed-off-by: Ingo Molnar --- kernel/irq/manage.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'kernel/irq/manage.c') diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c index 634a2a955104..948a22a2c013 100644 --- a/kernel/irq/manage.c +++ b/kernel/irq/manage.c @@ -123,7 +123,7 @@ int do_irq_select_affinity(unsigned int irq, struct irq_desc *desc) * Preserve an userspace affinity setup, but make sure that * one of the targets is online. */ - if (desc->status & IRQ_AFFINITY_SET) { + if (desc->status & (IRQ_AFFINITY_SET | IRQ_NO_BALANCING)) { if (cpus_intersects(desc->affinity, cpu_online_map)) mask = desc->affinity; else @@ -483,6 +483,10 @@ __setup_irq(unsigned int irq, struct irq_desc * desc, struct irqaction *new) /* Undo nested disables: */ desc->depth = 1; + /* Exclude IRQ from balancing if requested */ + if (new->flags & IRQF_NOBALANCING) + desc->status |= IRQ_NO_BALANCING; + /* Set default affinity mask once everything is setup */ do_irq_select_affinity(irq, desc); @@ -497,10 +501,6 @@ __setup_irq(unsigned int irq, struct irq_desc * desc, struct irqaction *new) *p = new; - /* Exclude IRQ from balancing */ - if (new->flags & IRQF_NOBALANCING) - desc->status |= IRQ_NO_BALANCING; - /* Reset broken irq detection when installing new handler */ desc->irq_count = 0; desc->irqs_unhandled = 0; -- cgit v1.2.3-70-g09d2 From f131e2436ddbac2527bb2d6297a823aae4b024f8 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Sat, 8 Nov 2008 09:57:40 +0100 Subject: irq: fix typo Impact: build fix fix build failure on UP. Signed-off-by: Ingo Molnar --- kernel/irq/manage.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel/irq/manage.c') diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c index 948a22a2c013..435861284e4c 100644 --- a/kernel/irq/manage.c +++ b/kernel/irq/manage.c @@ -159,7 +159,7 @@ int irq_select_affinity_usr(unsigned int irq) } #else -static inline int do_select_irq_affinity(int irq, struct irq_desc *desc) +static inline int do_irq_select_affinity(int irq, struct irq_desc *desc) { return 0; } -- cgit v1.2.3-70-g09d2 From 3ff68a6a106c362a6811d3e51bced58e6fc87de7 Mon Sep 17 00:00:00 2001 From: Mark Nelson Date: Thu, 13 Nov 2008 21:37:41 +1100 Subject: genirq: __irq_set_trigger: change pr_warning to pr_debug Commit 0c5d1eb77a8be917b638344a22afe1398236482b (genirq: record trigger type) caused powerpc platforms that had no set_type() function in their struct irq_chip to spew out warnings about "No set_type function for IRQ...". This warning isn't necessarily justified though because the generic powerpc platform code calls set_irq_type() (which in turn calls __irq_set_trigger) with information from the device tree to establish the interrupt mappings, regardless of whether the PIC can actually set a type. A platform's irq_chip might not have a set_type function for a variety of reasons, for example: the platform may have the type essentially hard-coded, or as in the case for Cell interrupts are just messages past around that have no real concept of type, or the platform could even have a virtual PIC as on the PS3. Signed-off-by: Mark Nelson Signed-off-by: Ingo Molnar --- kernel/irq/manage.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel/irq/manage.c') diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c index 435861284e4c..801addda3c43 100644 --- a/kernel/irq/manage.c +++ b/kernel/irq/manage.c @@ -365,7 +365,7 @@ int __irq_set_trigger(struct irq_desc *desc, unsigned int irq, * IRQF_TRIGGER_* but the PIC does not support multiple * flow-types? */ - pr_warning("No set_type function for IRQ %d (%s)\n", irq, + pr_debug("No set_type function for IRQ %d (%s)\n", irq, chip ? (chip->name ? : "unknown") : "unknown"); return 0; } -- cgit v1.2.3-70-g09d2 From f2b662da8d6bd44673537f3f64220afefdca369f Mon Sep 17 00:00:00 2001 From: David Brownell Date: Mon, 1 Dec 2008 14:31:38 -0800 Subject: genirq: record IRQ_LEVEL in irq_desc[] Impact: fix __irq_set_trigger() for IRQ_LEVEL When recording the irq trigger type, let's also make sure that IRQ_LEVEL gets set correctly. Signed-off-by: David Brownell Signed-off-by: Andrew Morton Acked-by: Benjamin Herrenschmidt Signed-off-by: Ingo Molnar --- kernel/irq/chip.c | 1 + kernel/irq/manage.c | 15 +++++++++------ 2 files changed, 10 insertions(+), 6 deletions(-) (limited to 'kernel/irq/manage.c') diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c index 10b5092e9bfe..7765d4c80c30 100644 --- a/kernel/irq/chip.c +++ b/kernel/irq/chip.c @@ -124,6 +124,7 @@ int set_irq_type(unsigned int irq, unsigned int type) return -ENODEV; } + type &= IRQ_TYPE_SENSE_MASK; if (type == IRQ_TYPE_NONE) return 0; diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c index 801addda3c43..46953a06f4a8 100644 --- a/kernel/irq/manage.c +++ b/kernel/irq/manage.c @@ -370,16 +370,18 @@ int __irq_set_trigger(struct irq_desc *desc, unsigned int irq, return 0; } - ret = chip->set_type(irq, flags & IRQF_TRIGGER_MASK); + /* caller masked out all except trigger mode flags */ + ret = chip->set_type(irq, flags); if (ret) pr_err("setting trigger mode %d for irq %u failed (%pF)\n", - (int)(flags & IRQF_TRIGGER_MASK), - irq, chip->set_type); + (int)flags, irq, chip->set_type); else { + if (flags & (IRQ_TYPE_LEVEL_LOW | IRQ_TYPE_LEVEL_HIGH)) + flags |= IRQ_LEVEL; /* note that IRQF_TRIGGER_MASK == IRQ_TYPE_SENSE_MASK */ - desc->status &= ~IRQ_TYPE_SENSE_MASK; - desc->status |= flags & IRQ_TYPE_SENSE_MASK; + desc->status &= ~(IRQ_LEVEL | IRQ_TYPE_SENSE_MASK); + desc->status |= flags; } return ret; @@ -459,7 +461,8 @@ __setup_irq(unsigned int irq, struct irq_desc * desc, struct irqaction *new) /* Setup the type (level, edge polarity) if configured: */ if (new->flags & IRQF_TRIGGER_MASK) { - ret = __irq_set_trigger(desc, irq, new->flags); + ret = __irq_set_trigger(desc, irq, + new->flags & IRQF_TRIGGER_MASK); if (ret) { spin_unlock_irqrestore(&desc->lock, flags); -- cgit v1.2.3-70-g09d2 From 470c66239ef0336429b35345f3f615d47341e13b Mon Sep 17 00:00:00 2001 From: David Brownell Date: Mon, 1 Dec 2008 14:31:37 -0800 Subject: genirq: warn when IRQF_DISABLED may be ignored Impact: emit new warning We periodically waste time tracking down problems from the genirq framework not respecting IRQF_DISABLED for some shared IRQ cases. Linus views this as "will not fix", but we're still left with the bugs caused by this misbehavior. This patch adds a nag message in request_irq(), so that drivers can fix their IRQ handlers to avoid this problem. Note that developers will never see the relevant bugs when they run with LOCKDEP, so it's no wonder these bugs are hard to find. (That also means LOCKDEP is overlooking some IRQ-related bugs involving IRQ handlers that don't set IRQF_DISABLED...) Signed-off-by: David Brownell Signed-off-by: Andrew Morton Acked-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- kernel/irq/manage.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'kernel/irq/manage.c') diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c index c498a1b8c621..7fd891c3a33d 100644 --- a/kernel/irq/manage.c +++ b/kernel/irq/manage.c @@ -635,6 +635,18 @@ int request_irq(unsigned int irq, irq_handler_t handler, struct irq_desc *desc; int retval; + /* + * handle_IRQ_event() always ignores IRQF_DISABLED except for + * the _first_ irqaction (sigh). That can cause oopsing, but + * the behavior is classified as "will not fix" so we need to + * start nudging drivers away from using that idiom. + */ + if ((irqflags & (IRQF_SHARED|IRQF_DISABLED)) + == (IRQF_SHARED|IRQF_DISABLED)) + pr_warning("IRQ %d/%s: IRQF_DISABLED is not " + "guaranteed on shared IRQs\n", + irq, devname); + #ifdef CONFIG_LOCKDEP /* * Lockdep wants atomic interrupt handlers: -- cgit v1.2.3-70-g09d2