Re: [PATCH 2/2] irqchip: ti-sci-intr: Add support for Interrupt Router driver

From: Marc Zyngier
Date: Mon Oct 08 2018 - 09:00:17 EST


Hi Lokesh,

On 08/10/18 10:48, Lokesh Vutla wrote:
Hi Marc,

On 10/6/2018 3:25 PM, Marc Zyngier wrote:
Hi Lokesh,

On Sat, 06 Oct 2018 08:28:12 +0100,
Lokesh Vutla <lokeshvutla@xxxxxx> wrote:

Texas Instruments' K3 generation SoCs has an IP Interrupt Router
that does allows for multiplexing of input interrupts to host
interrupt controller. Interrupt Router inputs are either from a
peripheral or from an Interrupt Aggregator which is another
interrupt controller.

Configuration of the interrupt router registers can only be done by
a system co-processor and the driver needs to send a message to this
co processor over TISCI protocol.

I assume that this co-processor only deals with the routing itself,
and doesn't need to be talked to during interrupt processing, right?

Yes, that's right.



Add support for Interrupt Router driver over TISCI protocol.

Signed-off-by: Lokesh Vutla <lokeshvutla@xxxxxx>
---
MAINTAINERS | 1 +
drivers/irqchip/Kconfig | 11 +
drivers/irqchip/Makefile | 1 +
drivers/irqchip/irq-ti-sci-intr.c | 325 ++++++++++++++++++++++++++++++
4 files changed, 338 insertions(+)
create mode 100644 drivers/irqchip/irq-ti-sci-intr.c

diff --git a/MAINTAINERS b/MAINTAINERS
index a23778b68d74..cf3c834f8cee 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -14626,6 +14626,7 @@ F: Documentation/devicetree/bindings/clock/ti,sci-clk.txt
F: drivers/clk/keystone/sci-clk.c
F: drivers/reset/reset-ti-sci.c
F: Documentation/devicetree/bindings/interrupt-controller/ti,sci-intr.txt
+F: drivers/irqchip/irq-ti-sci-intr.c
THANKO'S RAREMONO AM/FM/SW RADIO RECEIVER USB DRIVER
M: Hans Verkuil <hverkuil@xxxxxxxxx>
diff --git a/drivers/irqchip/Kconfig b/drivers/irqchip/Kconfig
index 96451b581452..9a965fe22043 100644
--- a/drivers/irqchip/Kconfig
+++ b/drivers/irqchip/Kconfig
@@ -374,6 +374,17 @@ config QCOM_PDC
Power Domain Controller driver to manage and configure wakeup
IRQs for Qualcomm Technologies Inc (QTI) mobile chips.
+config TI_SCI_INTR_IRQCHIP
+ tristate "TISCI based Interrupt Router irqchip driver"
+ depends on TI_SCI_PROTOCOL && ARCH_K3
+ select IRQ_DOMAIN
+ select IRQ_DOMAIN_HIERARCHY
+ help
+ This enables the irqchip driver support for K3 Interrupt router
+ over TI System Control Interface available on some new TI's SoCs.
+ If you wish to use interrupt router irq resources managed by the
+ TI System Controller, say Y here. Otherwise, say N.

I don't really see the point of making this user-selectable. If you're
compiling support for a given platform, this platform configuration
fragment should itself select the necessary dependencies for the
system to work as expected. Here, you are leaving the choice to the
user, with a 50% chance of getting a system that doesn't boot...

There are 2 reasons why I made it tristate:
- Not all interrupts go through this irqchip(At least in the AM6 SoC
using this). Most of the legacy peripherals still are directly connected
to GIC
- TI_SCI_PROTOCOL is defined as tristate.

But as you said, these are "legacy" interrupts, and most of the interesting stuff is routed through the system controller. We also try not to have core interrupt controllers as modules. As for having the firmware interface as a module, I wonder what the use-case is.

If you still feel I should not make it user-selectable, I can drop it.

I really wonder what the added value is for the user.



+
endmenu
config SIFIVE_PLIC
diff --git a/drivers/irqchip/Makefile b/drivers/irqchip/Makefile
index b822199445ff..44bf65606d60 100644
--- a/drivers/irqchip/Makefile
+++ b/drivers/irqchip/Makefile
@@ -89,3 +89,4 @@ obj-$(CONFIG_GOLDFISH_PIC) += irq-goldfish-pic.o
obj-$(CONFIG_NDS32) += irq-ativic32.o
obj-$(CONFIG_QCOM_PDC) += qcom-pdc.o
obj-$(CONFIG_SIFIVE_PLIC) += irq-sifive-plic.o
+obj-$(CONFIG_TI_SCI_INTR_IRQCHIP) += irq-ti-sci-intr.o
diff --git a/drivers/irqchip/irq-ti-sci-intr.c b/drivers/irqchip/irq-ti-sci-intr.c
new file mode 100644
index 000000000000..f04fe6da1b09
--- /dev/null
+++ b/drivers/irqchip/irq-ti-sci-intr.c
@@ -0,0 +1,325 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Texas Instruments' K3 Interrupt Router irqchip driver
+ *
+ * Copyright (C) 2018 Texas Instruments Incorporated - http://www.ti.com/
+ * Lokesh Vutla <lokeshvutla@xxxxxx>
+ */
+
+#include <linux/err.h>
+#include <linux/io.h>
+#include <linux/irqchip.h>
+#include <linux/of_platform.h>
+#include <linux/of_address.h>
+#include <linux/of_irq.h>
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/irqdomain.h>
+#include <linux/soc/ti/ti_sci_protocol.h>
+
+#define TI_SCI_DEV_ID_MASK 0xffff
+#define TI_SCI_DEV_ID_SHIFT 16
+#define TI_SCI_IRQ_ID_MASK 0xffff
+#define TI_SCI_IRQ_ID_SHIFT 0
+#define TI_SCI_IS_EVENT_IRQ BIT(31)
+
+#define HWIRQ_TO_DEVID(HWIRQ) (((HWIRQ) >> (TI_SCI_DEV_ID_SHIFT)) & \
+ (TI_SCI_DEV_ID_MASK))
+#define HWIRQ_TO_IRQID(HWIRQ) ((HWIRQ) & (TI_SCI_IRQ_ID_MASK))

nit: s/(HWIRQ)/(hwirq)/g

okay.


+
+/**
+ * struct ti_sci_intr_irq_domain - Structure representing a TISCI based
+ * Interrupt Router IRQ domain.
+ * @sci: Pointer to TISCI handle
+ * @dst_irq: TISCI resource pointer representing destination irq controller.
+ * @dst_id: TISCI device ID of the destination irq controller.
+ */
+struct ti_sci_intr_irq_domain {
+ const struct ti_sci_handle *sci;
+ struct ti_sci_resource *dst_irq;
+ u16 dst_id;
+};
+
+/**
+ * struct ti_sci_intr_irq_desc - Description of an Interrupt Router IRQ
+ * @src_id: TISCI device ID of the IRQ source
+ * @src_index: IRQ source index within the device.
+ * @dst_irq: Destination host IRQ.
+ */
+struct ti_sci_intr_irq_desc {
+ u16 src_id;
+ u16 src_index;
+ u16 dst_irq;
+};

Oh great. So this is reinventing the GICv3 ITS, only for SPIs. :-(

Now, this structure seems completely useless, see below.

+
+static struct irq_chip ti_sci_intr_irq_chip = {
+ .name = "INTR",
+ .irq_eoi = irq_chip_eoi_parent,
+ .irq_mask = irq_chip_mask_parent,
+ .irq_unmask = irq_chip_unmask_parent,
+ .irq_retrigger = irq_chip_retrigger_hierarchy,
+ .irq_set_type = irq_chip_set_type_parent,
+ .irq_set_affinity = irq_chip_set_affinity_parent,
+};
+
+/**
+ * ti_sci_intr_irq_domain_translate() - Retrieve hwirq and type from
+ * IRQ firmware specific handler.
+ * @domain: Pointer to IRQ domain
+ * @fwspec: Pointer to IRQ specific firmware structure
+ * @hwirq: IRQ number identified by hardware
+ * @type: IRQ type
+ *
+ * Return 0 if all went ok else appropriate error.
+ */
+static int ti_sci_intr_irq_domain_translate(struct irq_domain *domain,
+ struct irq_fwspec *fwspec,
+ unsigned long *hwirq,
+ unsigned int *type)
+{
+ if (is_of_node(fwspec->fwnode)) {
+ if (fwspec->param_count != 3)
+ return -EINVAL;
+
+ *hwirq = ((fwspec->param[0] & TI_SCI_DEV_ID_MASK) <<
+ TI_SCI_DEV_ID_SHIFT) |
+ (fwspec->param[1] & TI_SCI_IRQ_ID_MASK);

Maybe it would make sense to have a macro that hides this:

okay.


*hwirq = FWSPEC_TO_HWIRQ(fwspec);

+ *type = fwspec->param[2];
+
+ return 0;
+ }
+
+ return -EINVAL;
+}
+
+static inline void ti_sci_intr_delete_desc(struct ti_sci_intr_irq_domain *intr,
+ struct ti_sci_intr_irq_desc *desc)
+{
+ intr->sci->ops.rm_irq_ops.free_direct_irq(intr->sci, desc->src_id,
+ desc->src_index,
+ intr->dst_id, desc->dst_irq);

This looks horrible. Why doesn't your firmware interface have a helper
functions that hides this? Something like:

ti_sci_free_direct_irq(intr, src_id, src_index, dst_irq);

and you could even add some error checking.

All existing TISCI users follow the same convention, so I did not bother adding
any such wrapper. Will update TISCI with these wrappers and see what firmware
maintainer says.

Frankly, exposing all kind of data structures to the world is a pretty poor form of abstraction, which is what the firmware is supposed to provide.

I'd strongly suggest that include/linux/soc/ti/ti_sci_protocol.h gets cleaned up, and that the whole ti_sci_ops disappears from the that file. Nobody outside of the firmware *implementation* needs to know about its, and it would be much better served by a set of helpers.

Finally, please make the TISCI interrupt management part of this series, so that I can review it as part of the code that uses it.

Thanks,

M.
--
Jazz is not dead. It just smells funny...