| Message ID | YnkfWFzvusFFktSt@linutronix.de (mailing list archive) |
|---|---|
| State | RFC |
| Delegated to: | Johannes Berg |
| Headers | show |
| Series | [RFC] genirq: Provide generic_handle_domain_irq_safe(). | expand |
On Mon, May 09, 2022 at 04:04:08PM +0200, Sebastian Andrzej Siewior wrote: > The problem with generic_handle_domain_irq() is that with `threadirqs' > it will trigger "WARN_ON_ONCE(!in_hardirq())". Now silenced by: https://git.kernel.org/linus/792ea6a074ae > +int generic_handle_domain_irq_safe(struct irq_domain *domain, unsigned int hwirq) > +{ > + unsigned long flags; > + int ret; > + > + local_irq_save(flags); > + ret = handle_irq_desc(irq_resolve_mapping(domain, hwirq)); > + local_irq_restore(flags); > + return ret; > +} > +EXPORT_SYMBOL_GPL(generic_handle_domain_irq_safe); AFAICS you don't need to disable hardirqs at least for the "threadirqs" case because irq_forced_thread_fn() already does that. > drivers/bcma/driver_gpio.c | 2 +- > drivers/gpio/gpio-mlxbf2.c | 6 ++---- > drivers/pinctrl/pinctrl-amd.c | 2 +- > drivers/platform/x86/intel/int0002_vgpio.c | 3 +-- > drivers/ssb/driver_gpio.c | 6 ++++-- From a quick look, the proper solution for all of those drivers is probably to just add IRQF_NO_THREAD and be done with it. Thanks, Lukas
On 2022-05-16 12:18:14 [+0200], Lukas Wunner wrote: > On Mon, May 09, 2022 at 04:04:08PM +0200, Sebastian Andrzej Siewior wrote: > > The problem with generic_handle_domain_irq() is that with `threadirqs' > > it will trigger "WARN_ON_ONCE(!in_hardirq())". > > Now silenced by: > https://git.kernel.org/linus/792ea6a074ae > > > > +int generic_handle_domain_irq_safe(struct irq_domain *domain, unsigned int hwirq) > > +{ > > + unsigned long flags; > > + int ret; > > + > > + local_irq_save(flags); > > + ret = handle_irq_desc(irq_resolve_mapping(domain, hwirq)); > > + local_irq_restore(flags); > > + return ret; > > +} > > +EXPORT_SYMBOL_GPL(generic_handle_domain_irq_safe); > > AFAICS you don't need to disable hardirqs at least for the "threadirqs" > case because irq_forced_thread_fn() already does that. PREEMPT_RT does not disable interrupts. Also completions in softirq won't disable interrupts. > > > drivers/bcma/driver_gpio.c | 2 +- > > drivers/gpio/gpio-mlxbf2.c | 6 ++---- > > drivers/pinctrl/pinctrl-amd.c | 2 +- > > drivers/platform/x86/intel/int0002_vgpio.c | 3 +-- > > drivers/ssb/driver_gpio.c | 6 ++++-- > > From a quick look, the proper solution for all of those drivers is > probably to just add IRQF_NO_THREAD and be done with it. I think I mentioned that part in the commit description: IRQF_NO_THREAD must be specified by all handlers of a shared interrupt. It is an option for the handler that owns an interrupt exclusive. > Thanks, > > Lukas Sebastian
diff --git a/drivers/bcma/driver_gpio.c b/drivers/bcma/driver_gpio.c index 1e74ec1c7f231..ed2730a21e7c4 100644 --- a/drivers/bcma/driver_gpio.c +++ b/drivers/bcma/driver_gpio.c @@ -113,7 +113,7 @@ static irqreturn_t bcma_gpio_irq_handler(int irq, void *dev_id) return IRQ_NONE; for_each_set_bit(gpio, &irqs, gc->ngpio) - generic_handle_irq(irq_find_mapping(gc->irq.domain, gpio)); + generic_handle_domain_irq_safe(gc->irq.domain, gpio); bcma_chipco_gpio_polarity(cc, irqs, val & irqs); return IRQ_HANDLED; diff --git a/drivers/gpio/gpio-mlxbf2.c b/drivers/gpio/gpio-mlxbf2.c index 3d89912a05b87..d4916f32fee73 100644 --- a/drivers/gpio/gpio-mlxbf2.c +++ b/drivers/gpio/gpio-mlxbf2.c @@ -273,10 +273,8 @@ static irqreturn_t mlxbf2_gpio_irq_handler(int irq, void *ptr) pending = readl(gs->gpio_io + YU_GPIO_CAUSE_OR_CAUSE_EVTEN0); writel(pending, gs->gpio_io + YU_GPIO_CAUSE_OR_CLRCAUSE); - for_each_set_bit(level, &pending, gc->ngpio) { - int gpio_irq = irq_find_mapping(gc->irq.domain, level); - generic_handle_irq(gpio_irq); - } + for_each_set_bit(level, &pending, gc->ngpio) + generic_handle_domain_irq_safe(gc->irq.domain, level); return IRQ_RETVAL(pending); } diff --git a/drivers/pinctrl/pinctrl-amd.c b/drivers/pinctrl/pinctrl-amd.c index 1a7d686494ffb..ce6fa6a76d1f6 100644 --- a/drivers/pinctrl/pinctrl-amd.c +++ b/drivers/pinctrl/pinctrl-amd.c @@ -638,7 +638,7 @@ static bool do_amd_gpio_irq_handler(int irq, void *dev_id) if (!(regval & PIN_IRQ_PENDING) || !(regval & BIT(INTERRUPT_MASK_OFF))) continue; - generic_handle_domain_irq(gc->irq.domain, irqnr + i); + generic_handle_domain_irq_safe(gc->irq.domain, irqnr + i); /* Clear interrupt. * We must read the pin register again, in case the diff --git a/drivers/platform/x86/intel/int0002_vgpio.c b/drivers/platform/x86/intel/int0002_vgpio.c index 617dbf98980ec..97cfbc520a02c 100644 --- a/drivers/platform/x86/intel/int0002_vgpio.c +++ b/drivers/platform/x86/intel/int0002_vgpio.c @@ -125,8 +125,7 @@ static irqreturn_t int0002_irq(int irq, void *data) if (!(gpe_sts_reg & GPE0A_PME_B0_STS_BIT)) return IRQ_NONE; - generic_handle_irq(irq_find_mapping(chip->irq.domain, - GPE0A_PME_B0_VIRT_GPIO_PIN)); + generic_handle_domain_irq_safe(chip->irq.domain, GPE0A_PME_B0_VIRT_GPIO_PIN); pm_wakeup_hard_event(chip->parent); diff --git a/drivers/ssb/driver_gpio.c b/drivers/ssb/driver_gpio.c index 2de3896489c84..897cb8db5084f 100644 --- a/drivers/ssb/driver_gpio.c +++ b/drivers/ssb/driver_gpio.c @@ -132,7 +132,8 @@ static irqreturn_t ssb_gpio_irq_chipco_handler(int irq, void *dev_id) return IRQ_NONE; for_each_set_bit(gpio, &irqs, bus->gpio.ngpio) - generic_handle_irq(ssb_gpio_to_irq(&bus->gpio, gpio)); + generic_handle_domain_irq_safe(bus->irq_domain, gpio); + ssb_chipco_gpio_polarity(chipco, irqs, val & irqs); return IRQ_HANDLED; @@ -330,7 +331,8 @@ static irqreturn_t ssb_gpio_irq_extif_handler(int irq, void *dev_id) return IRQ_NONE; for_each_set_bit(gpio, &irqs, bus->gpio.ngpio) - generic_handle_irq(ssb_gpio_to_irq(&bus->gpio, gpio)); + generic_handle_domain_irq_safe(bus->irq_domain, gpio); + ssb_extif_gpio_polarity(extif, irqs, val & irqs); return IRQ_HANDLED; diff --git a/include/linux/irqdesc.h b/include/linux/irqdesc.h index a77584593f7d1..98253955e2ae7 100644 --- a/include/linux/irqdesc.h +++ b/include/linux/irqdesc.h @@ -169,6 +169,7 @@ int generic_handle_irq_safe(unsigned int irq); * conversion failed. */ int generic_handle_domain_irq(struct irq_domain *domain, unsigned int hwirq); +int generic_handle_domain_irq_safe(struct irq_domain *domain, unsigned int hwirq); int generic_handle_domain_nmi(struct irq_domain *domain, unsigned int hwirq); #endif diff --git a/kernel/irq/irqdesc.c b/kernel/irq/irqdesc.c index 0099b87dd8530..48c34d47255cc 100644 --- a/kernel/irq/irqdesc.c +++ b/kernel/irq/irqdesc.c @@ -706,6 +706,30 @@ int generic_handle_domain_irq(struct irq_domain *domain, unsigned int hwirq) } EXPORT_SYMBOL_GPL(generic_handle_domain_irq); + /** + * generic_handle_irq_safe - Invoke the handler for a HW irq belonging + * to a domain from any context. + * @domain: The domain where to perform the lookup + * @hwirq: The HW irq number to convert to a logical one + * + * Returns: 0 on success, a negative value on error. + * + * This function can be called from any context (IRQ or process context). It + * will report an error if not invoked from IRQ context and the irq has been + * marked to enforce IRQ-context only. + */ +int generic_handle_domain_irq_safe(struct irq_domain *domain, unsigned int hwirq) +{ + unsigned long flags; + int ret; + + local_irq_save(flags); + ret = handle_irq_desc(irq_resolve_mapping(domain, hwirq)); + local_irq_restore(flags); + return ret; +} +EXPORT_SYMBOL_GPL(generic_handle_domain_irq_safe); + /** * generic_handle_domain_nmi - Invoke the handler for a HW nmi belonging * to a domain.
Provide generic_handle_domain_irq_safe() which can used from any context. This similar to commit 509853f9e1e7b ("genirq: Provide generic_handle_irq_safe()") but this time for the irq-domains interface. It has been reported for the amd-pinctrl driver via bugzilla https://bugzilla.kernel.org/show_bug.cgi?id=215954 I looked around and added a few users so it is not just one user API :) Instead of generic_handle_irq(irq_find_mapping)) one can use generic_handle_domain_irq(). The problem with generic_handle_domain_irq() is that with `threadirqs' it will trigger "WARN_ON_ONCE(!in_hardirq())". That interrupt handler can't be marked non-threaded because it is a shared handler (it is marked as such and I can't tell the interrupt can be really shared on the system). Ignoring the just mentioned warning, on PREEMPT_RT the threaded handler is invoked with enabled interrupts leading other problems. Do we do this? Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> --- drivers/bcma/driver_gpio.c | 2 +- drivers/gpio/gpio-mlxbf2.c | 6 ++---- drivers/pinctrl/pinctrl-amd.c | 2 +- drivers/platform/x86/intel/int0002_vgpio.c | 3 +-- drivers/ssb/driver_gpio.c | 6 ++++-- include/linux/irqdesc.h | 1 + kernel/irq/irqdesc.c | 24 ++++++++++++++++++++++ 7 files changed, 34 insertions(+), 10 deletions(-)