| Message ID | 20231220193207.1089469-1-opendmb@gmail.com (mailing list archive) |
|---|---|
| State | Handled Elsewhere |
| Headers | show |
| Series | irqchip/irq-brcmstb-l2: add write memory barrier before exit | expand |
On 12/20/2023 8:32 PM, Doug Berger wrote: > It was observed on Broadcom devices that use GIC v3 architecture > L1 interrupt controllers as the parent of brcmstb-l2 interrupt > controllers that the deactivation of the parent irq could happen > before the brcmstb-l2 deasserted its output. This would lead the > GIC to reactivate the irq only to find that no L2 interrupt was > pending. The result was a spurious interrupt invoking the > handle_bad_irq() with its associated messaging. While this did > not create a functional problem it is a waste of cycles. > > The hazard exists because the memory mapped bus writes to the > brcmstb-l2 registers are buffered and the GIC v3 architecture > uses a very efficient system register write to deactivate the > interrupt. This commit adds a write memory barrier prior to > invoking chained_irq_exit() to introduce a dsb(st) on those > systems to ensure the system register write cannot be executed > until the memory mapped writes are visible to the system. > > Signed-off-by: Doug Berger <opendmb@gmail.com> Acked-by: Florian Fainelli <florian.fainelli@broadcom.com> I would be even keen on slapping a: Fixes: 7f646e92766e ("irqchip: brcmstb-l2: Add Broadcom Set Top Box Level-2 interrupt controller") Thanks Doug!
On 12/21/23 03:14, Florian Fainelli wrote: > > > On 12/20/2023 8:32 PM, Doug Berger wrote: >> It was observed on Broadcom devices that use GIC v3 architecture >> L1 interrupt controllers as the parent of brcmstb-l2 interrupt >> controllers that the deactivation of the parent irq could happen >> before the brcmstb-l2 deasserted its output. This would lead the >> GIC to reactivate the irq only to find that no L2 interrupt was >> pending. The result was a spurious interrupt invoking the >> handle_bad_irq() with its associated messaging. While this did >> not create a functional problem it is a waste of cycles. >> >> The hazard exists because the memory mapped bus writes to the >> brcmstb-l2 registers are buffered and the GIC v3 architecture >> uses a very efficient system register write to deactivate the >> interrupt. This commit adds a write memory barrier prior to >> invoking chained_irq_exit() to introduce a dsb(st) on those >> systems to ensure the system register write cannot be executed >> until the memory mapped writes are visible to the system. >> >> Signed-off-by: Doug Berger <opendmb@gmail.com> > > Acked-by: Florian Fainelli <florian.fainelli@broadcom.com> > > I would be even keen on slapping a: > > Fixes: 7f646e92766e ("irqchip: brcmstb-l2: Add Broadcom Set Top Box > Level-2 interrupt controller") > > Thanks Doug! Thomas, are you able to apply Doug's patch including the Fixes tag, or do you need it to be re-submitted with that tag as well as my Acked-by? Thanks!
diff --git a/drivers/irqchip/irq-brcmstb-l2.c b/drivers/irqchip/irq-brcmstb-l2.c index 5559c943f03f..63aed60dd3f1 100644 --- a/drivers/irqchip/irq-brcmstb-l2.c +++ b/drivers/irqchip/irq-brcmstb-l2.c @@ -2,7 +2,7 @@ /* * Generic Broadcom Set Top Box Level 2 Interrupt controller driver * - * Copyright (C) 2014-2017 Broadcom + * Copyright (C) 2014-2023 Broadcom */ #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt @@ -112,6 +112,9 @@ static void brcmstb_l2_intc_irq_handle(struct irq_desc *desc) generic_handle_domain_irq(b->domain, irq); } while (status); out: + /* Don't ack parent before all device writes are done */ + wmb(); + chained_irq_exit(chip, desc); }
It was observed on Broadcom devices that use GIC v3 architecture L1 interrupt controllers as the parent of brcmstb-l2 interrupt controllers that the deactivation of the parent irq could happen before the brcmstb-l2 deasserted its output. This would lead the GIC to reactivate the irq only to find that no L2 interrupt was pending. The result was a spurious interrupt invoking the handle_bad_irq() with its associated messaging. While this did not create a functional problem it is a waste of cycles. The hazard exists because the memory mapped bus writes to the brcmstb-l2 registers are buffered and the GIC v3 architecture uses a very efficient system register write to deactivate the interrupt. This commit adds a write memory barrier prior to invoking chained_irq_exit() to introduce a dsb(st) on those systems to ensure the system register write cannot be executed until the memory mapped writes are visible to the system. Signed-off-by: Doug Berger <opendmb@gmail.com> --- drivers/irqchip/irq-brcmstb-l2.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)