[PATCH v4 21/26] irqchip/gic-v5: Add GICv5 IRS/SPI support

Lorenzo Pieralisi lpieralisi at kernel.org
Thu May 29 01:38:05 PDT 2025


On Wed, May 28, 2025 at 05:03:18PM +0100, Jonathan Cameron wrote:
> On Tue, 13 May 2025 19:48:14 +0200
> Lorenzo Pieralisi <lpieralisi at kernel.org> wrote:
> 
> > The GICv5 Interrupt Routing Service (IRS) component implements
> > interrupt management and routing in the GICv5 architecture.
> > 
> > A GICv5 system comprises one or more IRSes, that together
> > handle the interrupt routing and state for the system.
> > 
> > An IRS supports Shared Peripheral Interrupts (SPIs), that are
> > interrupt sources directly connected to the IRS; they do not
> > rely on memory for storage. The number of supported SPIs is
> > fixed for a given implementation and can be probed through IRS
> > IDR registers.
> > 
> > SPI interrupt state and routing are managed through GICv5
> > instructions.
> > 
> > Each core (PE in GICv5 terms) in a GICv5 system is identified with
> > an Interrupt AFFinity ID (IAFFID).
> > 
> > An IRS manages a set of cores that are connected to it.
> > 
> > Firmware provides a topology description that the driver uses
> > to detect to which IRS a CPU (ie an IAFFID) is associated with.
> > 
> > Use probeable information and firmware description to initialize
> > the IRSes and implement GICv5 IRS SPIs support through an
> > SPI-specific IRQ domain.
> > 
> > The GICv5 IRS driver:
> > 
> > - Probes IRSes in the system to detect SPI ranges
> > - Associates an IRS with a set of cores connected to it
> > - Adds an IRQchip structure for SPI handling
> > 
> > SPIs priority is set to a value corresponding to the lowest
> > permissible priority in the system (taking into account the
> > implemented priority bits of the IRS and CPU interface).
> > 
> > Since all IRQs are set to the same priority value, the value
> > itself does not matter as long as it is a valid one.
> > 
> > Co-developed-by: Sascha Bischoff <sascha.bischoff at arm.com>
> > Signed-off-by: Sascha Bischoff <sascha.bischoff at arm.com>
> > Co-developed-by: Timothy Hayes <timothy.hayes at arm.com>
> > Signed-off-by: Timothy Hayes <timothy.hayes at arm.com>
> > Signed-off-by: Lorenzo Pieralisi <lpieralisi at kernel.org>
> > Cc: Will Deacon <will at kernel.org>
> > Cc: Thomas Gleixner <tglx at linutronix.de>
> > Cc: Catalin Marinas <catalin.marinas at arm.com>
> > Cc: Marc Zyngier <maz at kernel.org>
> 
> A few comments inline. Mostly around of cpu phandle parsing.
> +CC Ali as there is a comment on his SMT DT patch set inline.
> 
> 
> > ---
> >  arch/arm64/include/asm/sysreg.h    |  36 +++
> >  drivers/irqchip/Makefile           |   2 +-
> >  drivers/irqchip/irq-gic-v5-irs.c   | 433 +++++++++++++++++++++++++++++++++++++
> >  drivers/irqchip/irq-gic-v5.c       | 341 +++++++++++++++++++++++++++--
> >  include/linux/irqchip/arm-gic-v5.h | 130 +++++++++++
> >  5 files changed, 920 insertions(+), 22 deletions(-)
> > 
> > diff --git a/arch/arm64/include/asm/sysreg.h b/arch/arm64/include/asm/sysreg.h
> > index 9d28d408f9c6df24526dd8ecbf3c7d920246b22d..fbac3b6f056ae6fafd64457600d45808e4904ae3 100644
> > --- a/arch/arm64/include/asm/sysreg.h
> > +++ b/arch/arm64/include/asm/sysreg.h
> > @@ -1082,14 +1082,50 @@
> >  /*
> >   * Definitions for GICv5 instructions
> >   */
> 
> >  
> > +/* Shift and mask definitions for GIC CDAFF */
> 
> Similar comment. Mask definitions seems more accurate to me.

Fixed.

> > +/* Shift and mask definitions for GIC CDDIS */
> > +#define GICV5_GIC_CDDIS_TYPE_MASK	GENMASK_ULL(31, 29)
> > +#define GICV5_GIC_CDDIS_TYPE(r)		FIELD_GET(GICV5_GIC_CDDIS_TYPE_MASK, r)
> > +#define GICV5_GIC_CDDIS_ID_MASK		GENMASK_ULL(23, 0)
> > +#define GICV5_GIC_CDDIS_ID(r)		FIELD_GET(GICV5_GIC_CDDIS_ID_MASK, r)
> 
> Similar to earlier comment. I'm not sure the shortened forms are worth the bother.

Same as previous replies.

> > diff --git a/drivers/irqchip/irq-gic-v5-irs.c b/drivers/irqchip/irq-gic-v5-irs.c
> > new file mode 100644
> > index 0000000000000000000000000000000000000000..8c448487b909c7d3b4e1f95a5bc02b741ecc40b3
> > --- /dev/null
> > +++ b/drivers/irqchip/irq-gic-v5-irs.c
> > @@ -0,0 +1,433 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +/*
> > + * Copyright (C) 2024-2025 ARM Limited, All Rights Reserved.
> > + */
> > +
> > +#define pr_fmt(fmt)	"GICv5 IRS: " fmt
> > +
> > +#include <linux/of.h>
> > +#include <linux/of_address.h>
> > +
> > +#include <linux/irqchip.h>
> > +#include <linux/irqchip/arm-gic-v5.h>
> 
> 
> 
> > +int gicv5_spi_irq_set_type(struct irq_data *d, unsigned int type)
> > +{
> > +	struct gicv5_irs_chip_data *irs_data = d->chip_data;
> > +	u32 selr, cfgr;
> > +	bool level;
> > +	int ret;
> > +
> > +	switch (type) {
> > +	case IRQ_TYPE_EDGE_RISING:
> > +	case IRQ_TYPE_EDGE_FALLING:
> > +		level = false;
> > +		break;
> > +	case IRQ_TYPE_LEVEL_HIGH:
> > +	case IRQ_TYPE_LEVEL_LOW:
> > +		level = true;
> > +		break;
> > +	default:
> > +		return -EINVAL;
> > +	}
> > +
> > +	guard(raw_spinlock)(&irs_data->spi_config_lock);
> > +
> > +	selr = FIELD_PREP(GICV5_IRS_SPI_SELR_ID, d->hwirq);
> > +	irs_writel_relaxed(irs_data, selr, GICV5_IRS_SPI_SELR);
> > +	ret = gicv5_irs_wait_for_spi_op(irs_data);
> > +	if (ret)
> > +		return ret;
> > +
> > +	cfgr = FIELD_PREP(GICV5_IRS_SPI_CFGR_TM, level);
> > +
> > +	irs_writel_relaxed(irs_data, cfgr, GICV5_IRS_SPI_CFGR);
> > +	ret = gicv5_irs_wait_for_spi_op(irs_data);
> > +	if (ret)
> > +		return ret;
> > +
> > +	return 0;
> 	return gicv5_irq_wait_for_spi_op()
> 
> unless more code turns up here in a later patch.

Ok.

> 
> > +}
> 
> > +
> > +static void __init gicv5_irs_init_bases(struct gicv5_irs_chip_data *irs_data,
> > +					void __iomem *irs_base,
> > +					struct fwnode_handle *handle)
> > +{
> > +	struct device_node *np = to_of_node(handle);
> > +	u32 cr0, cr1;
> > +
> > +	irs_data->fwnode = handle;
> > +	irs_data->irs_base = irs_base;
> > +
> > +	if (of_property_read_bool(np, "dma-noncoherent")) {
> > +		/*
> > +		 * A non-coherent IRS implies that some cache levels cannot be
> > +		 * used coherently by the cores and GIC. Our only option is to mark
> > +		 * memory attributes for the GIC as non-cacheable; by default,
> > +		 * non-cacheable memory attributes imply outer-shareable
> > +		 * shareability, the value written into IRS_CR1_SH is ignored.
> > +		 */
> > +		cr1 = FIELD_PREP(GICV5_IRS_CR1_VPED_WA, GICV5_NO_WRITE_ALLOC)	|
> As per earlier comments is this less clear as:
> 		cr1 = FIELD_PREP(GICV5_IRS_CR1_VPED_WA, 0)	|
> 
> To me, seems fine but up to you.
> 
> > +			FIELD_PREP(GICV5_IRS_CR1_VPED_RA, GICV5_NO_READ_ALLOC)	|
> > +			FIELD_PREP(GICV5_IRS_CR1_VMD_WA, GICV5_NO_WRITE_ALLOC)	|
> > +			FIELD_PREP(GICV5_IRS_CR1_VMD_RA, GICV5_NO_READ_ALLOC)	|
> > +			FIELD_PREP(GICV5_IRS_CR1_VPET_RA, GICV5_NO_READ_ALLOC)	|
> > +			FIELD_PREP(GICV5_IRS_CR1_VMT_RA, GICV5_NO_READ_ALLOC)	|
> > +			FIELD_PREP(GICV5_IRS_CR1_IST_WA, GICV5_NO_WRITE_ALLOC)	|
> > +			FIELD_PREP(GICV5_IRS_CR1_IST_RA, GICV5_NO_READ_ALLOC)	|
> > +			FIELD_PREP(GICV5_IRS_CR1_IC, GICV5_NON_CACHE)		|
> > +			FIELD_PREP(GICV5_IRS_CR1_OC, GICV5_NON_CACHE);
> > +			irs_data->flags |= IRS_FLAGS_NON_COHERENT;
> > +	} else {
> > +		cr1 = FIELD_PREP(GICV5_IRS_CR1_VPED_WA, GICV5_WRITE_ALLOC)	|
> > +			FIELD_PREP(GICV5_IRS_CR1_VPED_RA, GICV5_READ_ALLOC)	|
> > +			FIELD_PREP(GICV5_IRS_CR1_VMD_WA, GICV5_WRITE_ALLOC)	|
> > +			FIELD_PREP(GICV5_IRS_CR1_VMD_RA, GICV5_READ_ALLOC)	|
> > +			FIELD_PREP(GICV5_IRS_CR1_VPET_RA, GICV5_READ_ALLOC)	|
> > +			FIELD_PREP(GICV5_IRS_CR1_VMT_RA, GICV5_READ_ALLOC)	|
> > +			FIELD_PREP(GICV5_IRS_CR1_IST_WA, GICV5_WRITE_ALLOC)	|
> > +			FIELD_PREP(GICV5_IRS_CR1_IST_RA, GICV5_READ_ALLOC)	|
> > +			FIELD_PREP(GICV5_IRS_CR1_IC, GICV5_WB_CACHE)		|
> > +			FIELD_PREP(GICV5_IRS_CR1_OC, GICV5_WB_CACHE)		|
> > +			FIELD_PREP(GICV5_IRS_CR1_SH, GICV5_INNER_SHARE);
> > +	}
> > +
> > +	irs_writel_relaxed(irs_data, cr1, GICV5_IRS_CR1);
> > +
> > +	cr0 = FIELD_PREP(GICV5_IRS_CR0_IRSEN, 0x1);
> > +	irs_writel_relaxed(irs_data, cr0, GICV5_IRS_CR0);
> > +	gicv5_irs_wait_for_idle(irs_data);
> > +}
> > +
> > +static int __init gicv5_irs_of_init_affinity(struct device_node *node,
> > +					     struct gicv5_irs_chip_data *irs_data,
> > +					     u8 iaffid_bits)
> > +{
> > +	/*
> > +	 * Detect IAFFID<->CPU mappings from the device tree and
> > +	 * record IRS<->CPU topology information.
> > +	 */
> > +	u16 iaffid_mask = GENMASK(iaffid_bits - 1, 0);
> > +	u16 *iaffids __free(kfree) = NULL;
> 
> See comments in cleanup.h.  Linus has been fairly clear he doesn't like
> separating the constructor and destructor like this - just declare
> iaffids where you construct it.

Right.

> > +	int ret, i, ncpus, niaffids;
> > +
> 
> > +	ncpus = of_property_count_elems_of_size(node, "cpus", sizeof(u32));
> 
> cpus is a phandle? I think this is going to run into current discussion
> on what phandles to CPUs on an SMT system look like (Rob Herring and Mark
> Rutland have different views)
> https://lore.kernel.org/linux-arm-kernel/20250512080715.82-1-alireza.sanaee@huawei.com/

I will make sure to steer clear of that then ;-), whatever the outcome
the current "cpus" bindings should continue to work as-is, right ?

> Anyhow this doesn't look right to me.
> I think it should be of_count_phandle_with_args() 

Aren't they equivalent in functionality if of_count_phandle_with_args()
cells_name == NULL ?

I will update the code but if the functionality provided is not the
same there is kernel code to fix (it is an example, there are others):

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/irqchip/irq-apple-aic.c?h=v6.15#n903

> Potentially with cpu-cells as the argument depending on how that
> thread goes.
> 
> > +	if (ncpus < 0)
> > +		return -EINVAL;
> > +
> > +	niaffids = of_property_count_elems_of_size(node, "arm,iaffids",
> > +						   sizeof(u16));
> > +	if (niaffids != ncpus)
> > +		return -EINVAL;
> > +
> 	u16 *iaffids __free(kfree) = kcalloc(niaffids, sizeof(*iaffids),
> 					     GFP_KERNEL);

Maybe I should rewrite this in Rust :)

> > +	iaffids = kcalloc(niaffids, sizeof(*iaffids), GFP_KERNEL);
> > +	if (!iaffids)
> > +		return -ENOMEM;
> > +
> > +	ret = of_property_read_u16_array(node, "arm,iaffids", iaffids, niaffids);
> > +	if (ret)
> > +		return ret;
> > +
> > +	for (i = 0; i < ncpus; i++) {
> > +		struct device_node *cpu_node;
> > +		u32 cpu_phandle;
> > +		int cpu;
> > +
> > +		if (of_property_read_u32_index(node, "cpus", i, &cpu_phandle))
> > +			continue;
> > +
> > +		cpu_node = of_find_node_by_phandle(cpu_phandle);
> 
> 		cpu_node = of_parse_phandle(node, "cpus", i);
> 
> not work here?

I think it would.

>  
> > +		if (WARN_ON(!cpu_node))
> > +			continue;
> > +
> > +		cpu = of_cpu_node_to_id(cpu_node);
> 
> If this is all you want then Ali's series gives you a helper
> 
> 		cpu = of_cpu_phandle_to_id(node, &cpu_node, i);
> 
> Though even better to have a helper that allows
> 		cpu = of_cpu_phandle_to_id(node, NULL, i); and handles
> the node put as internally.
> 
> Ali, any reason we can't do that?  Seems to be a fairly common
> pattern.
> 
> 
>  
> > +		of_node_put(cpu_node);
> > +		if (WARN_ON(cpu < 0))
> > +			continue;
> > +
> > +		if (iaffids[i] & ~iaffid_mask) {
> > +			pr_warn("CPU %d iaffid 0x%x exceeds IRS iaffid bits\n",
> > +				cpu, iaffids[i]);
> > +			continue;
> > +		}
> > +
> > +		per_cpu(cpu_iaffid, cpu).iaffid = iaffids[i];
> > +		per_cpu(cpu_iaffid, cpu).valid = true;
> > +
> > +		/* We also know that the CPU is connected to this IRS */
> > +		per_cpu(per_cpu_irs_data, cpu) = irs_data;
> > +	}
> > +
> > +	return ret;
> > +}
> 
> > diff --git a/drivers/irqchip/irq-gic-v5.c b/drivers/irqchip/irq-gic-v5.c
> > index a50982e5d98816d88e4fca37cc0ac31684fb6c76..e58ff345dbfaf840b17ad63c4fdb6c227137cf4b 100644
> > --- a/drivers/irqchip/irq-gic-v5.c
> > +++ b/drivers/irqchip/irq-gic-v5.c
> >
> > +
> > +static int gicv5_spi_irq_set_irqchip_state(struct irq_data *d,
> > +					   enum irqchip_irq_state which,
> > +					   bool val)
> > +{
> 
> Similar to previous, I'd call the state parameter state rather than val.

Right.

> > diff --git a/include/linux/irqchip/arm-gic-v5.h b/include/linux/irqchip/arm-gic-v5.h
> > index 4ff0ba64d9840c3844671f7850bb3d81ba2eb1b6..187af307de9170d9569898cb1e50de376a38bd0a 100644
> > --- a/include/linux/irqchip/arm-gic-v5.h
> > +++ b/include/linux/irqchip/arm-gic-v5.h
> > @@ -5,6 +5,8 @@
> >  #ifndef __LINUX_IRQCHIP_ARM_GIC_V5_H
> >  #define __LINUX_IRQCHIP_ARM_GIC_V5_H
> 
> >  
> > +#define GICV5_NO_READ_ALLOC		0b0
> > +#define GICV5_READ_ALLOC		0b1
> > +#define GICV5_NO_WRITE_ALLOC		0b0
> > +#define GICV5_WRITE_ALLOC		0b1
> 
> Given these are being written to fields called _RA and _WA
> so the defines provide value over 0 and 1 in appropriate places?
> Maybe just about. Anyhow, your code so on this up to you.
> 
> > +
> > +#define GICV5_NON_CACHE			0b00
> > +#define GICV5_WB_CACHE			0b01
> > +#define GICV5_WT_CACHE			0b10
> > +
> > +#define GICV5_NON_SHARE			0b00
> > +#define GICV5_OUTER_SHARE		0b10
> > +#define GICV5_INNER_SHARE		0b11
> > +
> > +#define GICV5_IRS_IDR1			0x0004
> > +#define GICV5_IRS_IDR2			0x0008
> > +#define GICV5_IRS_IDR5			0x0014
> > +#define GICV5_IRS_IDR6			0x0018
> > +#define GICV5_IRS_IDR7			0x001c
> > +#define GICV5_IRS_CR0			0x0080
> > +#define GICV5_IRS_CR1			0x0084
> > +#define GICV5_IRS_SPI_SELR		0x0108
> > +#define GICV5_IRS_SPI_CFGR		0x0114
> > +#define GICV5_IRS_SPI_STATUSR		0x0118
> > +#define GICV5_IRS_PE_SELR		0x0140
> > +#define GICV5_IRS_PE_STATUSR		0x0144
> > +#define GICV5_IRS_PE_CR0		0x0148
> 
> Blank line here as this is end of register offsets.

Yep, fixed it.

Thanks for having a look !
Lorenzo

> > +#define GICV5_IRS_IDR1_PRIORITY_BITS	GENMASK(22, 20)
> > +#define GICV5_IRS_IDR1_IAFFID_BITS	GENMASK(19, 16)
> 
> 
> 



More information about the linux-arm-kernel mailing list