On Wed, May 15, 2024 at 10:05:59AM -0700, Palmer Dabbelt wrote:
On Wed, 15 May 2024 07:36:33 PDT (-0700), Conor Dooley wrote:
Palmer, This is the issue I point out to you on the call earlier:
Ya, thanks, I just hit it. Looks like it's a semantic conflict between 21a8f8a0eb35 ("irqchip: Add RISC-V incoming MSI controller early driver") and dc892fb44322 ("riscv: Use IPIs for remote cache/TLB flushes by default"). I think all we need is
diff --git a/drivers/irqchip/irq-riscv-imsic-early.c b/drivers/irqchip/irq-riscv-imsic-early.c index 886418ec06cb..4fbb37074d29 100644 --- a/drivers/irqchip/irq-riscv-imsic-early.c +++ b/drivers/irqchip/irq-riscv-imsic-early.c @@ -49,7 +49,7 @@ static int __init imsic_ipi_domain_init(void) return virq < 0 ? virq : -ENOMEM;
/* Set vIRQ range */
riscv_ipi_set_virq_range(virq, IMSIC_NR_IPI, true);
riscv_ipi_set_virq_range(virq, IMSIC_NR_IPI); /* Announce that IMSIC is providing IPIs */ pr_info("%pfwP: providing IPIs using interrupt %d\n", imsic->fwnode, IMSIC_IPI_ID);as a conflict resolution, which IIUC should happen when Linus merges my next PR. So I'll try and remember to call that out.
Unfortunately it looks like the conflict resolution did not happen, and mainline builds are now affected.
Guenter