[PATCH V7] LoongArch: Add AVEC irqchip support
From: Tianyang Zhang
Date: Fri Jul 26 2024 - 06:25:25 EST
Introduce the advanced extended interrupt controllers (AVECINTC). This
feature will allow each core to have 256 independent interrupt vectors
and MSI interrupts can be independently routed to any vector on any CPU.
The whole topology of irqchips in LoongArch machines looks like this if
AVECINTC is supported:
+-----+ +-----------------------+ +-------+
| IPI | --> | CPUINTC | <-- | Timer |
+-----+ +-----------------------+ +-------+
^ ^ ^
| | |
+---------+ +----------+ +---------+ +-------+
| EIOINTC | | AVECINTC | | LIOINTC | <-- | UARTs |
+---------+ +----------+ +---------+ +-------+
^ ^
| |
+---------+ +---------+
| PCH-PIC | | PCH-MSI |
+---------+ +---------+
^ ^ ^
| | |
+---------+ +---------+ +---------+
| Devices | | PCH-LPC | | Devices |
+---------+ +---------+ +---------+
^
|
+---------+
| Devices |
+---------+
Co-developed-by: Jianmin Lv <lvjianmin@xxxxxxxxxxx>
Signed-off-by: Jianmin Lv <lvjianmin@xxxxxxxxxxx>
Co-developed-by: Liupu Wang <wangliupu@xxxxxxxxxxx>
Signed-off-by: Liupu Wang <wangliupu@xxxxxxxxxxx>
Co-developed-by: Huacai Chen <chenhuacai@xxxxxxxxxxx>
Signed-off-by: Huacai Chen <chenhuacai@xxxxxxxxxxx>
Signed-off-by: Tianyang Zhang <zhangtianyang@xxxxxxxxxxx>
---
.../arch/loongarch/irq-chip-model.rst | 32 ++
.../zh_CN/arch/loongarch/irq-chip-model.rst | 32 ++
arch/loongarch/Kconfig | 1 +
arch/loongarch/include/asm/cpu-features.h | 1 +
arch/loongarch/include/asm/cpu.h | 2 +
arch/loongarch/include/asm/hardirq.h | 3 +-
arch/loongarch/include/asm/irq.h | 25 +-
arch/loongarch/include/asm/loongarch.h | 18 +-
arch/loongarch/include/asm/smp.h | 2 +
arch/loongarch/kernel/cpu-probe.c | 3 +-
arch/loongarch/kernel/irq.c | 14 +-
arch/loongarch/kernel/paravirt.c | 5 +
arch/loongarch/kernel/smp.c | 6 +
drivers/irqchip/Makefile | 2 +-
drivers/irqchip/irq-loongarch-avec.c | 447 ++++++++++++++++++
drivers/irqchip/irq-loongarch-cpu.c | 5 +-
drivers/irqchip/irq-loongson-eiointc.c | 7 +-
drivers/irqchip/irq-loongson-pch-msi.c | 41 +-
include/linux/cpuhotplug.h | 3 +-
19 files changed, 632 insertions(+), 17 deletions(-)
create mode 100644 drivers/irqchip/irq-loongarch-avec.c
Changes log:
V0->V1:
1.Modified some formats and declarations
2.Removed kmalloc/kfree when adding affinity related data to pending_list,
and used moving tag to replace the original behavior
3.Adjusted the process that enables AVEC interrupts, now it is at the end of all processes
4.Removed CPUHP related callbacks, now irq_matrix_online/irq_matrix_offline is completed in start_secondary/loongson_cpu_disable
5.Adjusted compatibility issues for CONFIG_ACPI
6.About question:
> irr = csr_read64(LOONGARCH_CSR_IRR0 + vector / 64);
> should be good enough, no?
csr_read64 was built-in as __csrrd_d, it doesn't seem to support variables as parameters
>>>>
drivers/irqchip/irq-loongarch-avec.c: In function ‘complete_irq_moving’:
./arch/loongarch/include/asm/loongarch.h:164:25: error: invalid argument to built-in function
164 | #define csr_read64(reg) __csrrd_d(reg)
| ^~~~~~~~~
drivers/irqchip/irq-loongarch-avec.c:170:23: note: in expansion of macro ‘csr_read64’
170 | irr = csr_read64(LOONGARCH_CSR_IRR_BASE + vector / VECTORS_PER_REG);
| ^~~~~~~~~~
>>>>
So we have temporarily retained the previous implementation.
V1->V2:
Fixed up coding style. Made on/offline functions void
Added compatibility when CONFIG_SMP is turned off
V2->V3:
Squash two patches into one
V3->V4:
Update NR_IRQS
Update Register's name
Fixed up coding style
V4->V5:
Retain feature CPUCFG1_MSGINT
Fixed up coding style
Delete the test code introduced by V4, and now msi msg address still uses the 32-bit address
V5->V6:
Fix definition of NR_IRQS
Define arch_probe_nr_irqs()
Handle all avecintc interrupts in one dispatch
Use cpuhotplug callbacks instead of direct call to avec_online_cpu()/avec_offline_cpu()
Rename {SMP,ACTION}_CLEAR_VECT to {SMP,ACTION}_CLEAR_VECTOR
Use avecintc_ prefix instead of loongarch_avec_ to keep consistancy
V6->V7:
Fixed compatibility issue with cpuhp_setup_state_nocalls when CONFIG_SMP is turned off
Rename avecintc_online/offline_cpu as avecintc_cpu_online/offline
Use pch_msi_handle[0] as default value of get_pch_msi_handle
Rework commit-message
diff --git a/Documentation/arch/loongarch/irq-chip-model.rst b/Documentation/arch/loongarch/irq-chip-model.rst
index 7988f41923639..6dd48256e39f7 100644
--- a/Documentation/arch/loongarch/irq-chip-model.rst
+++ b/Documentation/arch/loongarch/irq-chip-model.rst
@@ -85,6 +85,38 @@ to CPUINTC directly::
| Devices |
+---------+
+Advanced Extended IRQ model
+===========================
+
+In this model, IPI (Inter-Processor Interrupt) and CPU Local Timer interrupt go
+to CPUINTC directly, CPU UARTS interrupts go to LIOINTC, PCH-MSI interrupts go
+to AVECINTC, and then go to CPUINTC directly, while all other devices interrupts
+go to PCH-PIC/PCH-LPC and gathered by EIOINTC, and then go to CPUINTC directly::
+
+ +-----+ +-----------------------+ +-------+
+ | IPI | --> | CPUINTC | <-- | Timer |
+ +-----+ +-----------------------+ +-------+
+ ^ ^ ^
+ | | |
+ +---------+ +----------+ +---------+ +-------+
+ | EIOINTC | | AVECINTC | | LIOINTC | <-- | UARTs |
+ +---------+ +----------+ +---------+ +-------+
+ ^ ^
+ | |
+ +---------+ +---------+
+ | PCH-PIC | | PCH-MSI |
+ +---------+ +---------+
+ ^ ^ ^
+ | | |
+ +---------+ +---------+ +---------+
+ | Devices | | PCH-LPC | | Devices |
+ +---------+ +---------+ +---------+
+ ^
+ |
+ +---------+
+ | Devices |
+ +---------+
+
ACPI-related definitions
========================
diff --git a/Documentation/translations/zh_CN/arch/loongarch/irq-chip-model.rst b/Documentation/translations/zh_CN/arch/loongarch/irq-chip-model.rst
index f1e9ab18206c3..472761938682c 100644
--- a/Documentation/translations/zh_CN/arch/loongarch/irq-chip-model.rst
+++ b/Documentation/translations/zh_CN/arch/loongarch/irq-chip-model.rst
@@ -87,6 +87,38 @@ PCH-LPC/PCH-MSI，然后被EIOINTC统一收集，再直接到达CPUINTC::
| Devices |
+---------+
+高级扩展IRQ模型
+===============
+
+在这种模型里面，IPI（Inter-Processor Interrupt）和CPU本地时钟中断直接发送到CPUINTC，
+CPU串口（UARTs）中断发送到LIOINTC，PCH-MSI中断发送到AVECINTC，而后通过AVECINTC直接
+送达CPUINTC，而其他所有设备的中断则分别发送到所连接的PCH-PIC/PCH-LPC，然后由EIOINTC
+统一收集，再直接到达CPUINTC::
+
+ +-----+ +-----------------------+ +-------+
+ | IPI | --> | CPUINTC | <-- | Timer |
+ +-----+ +-----------------------+ +-------+
+ ^ ^ ^
+ | | |
+ +---------+ +----------+ +---------+ +-------+
+ | EIOINTC | | AVECINTC | | LIOINTC | <-- | UARTs |
+ +---------+ +----------+ +---------+ +-------+
+ ^ ^
+ | |
+ +---------+ +---------+
+ | PCH-PIC | | PCH-MSI |
+ +---------+ +---------+
+ ^ ^ ^
+ | | |
+ +---------+ +---------+ +---------+
+ | Devices | | PCH-LPC | | Devices |
+ +---------+ +---------+ +---------+
+ ^
+ |
+ +---------+
+ | Devices |
+ +---------+
+
ACPI相关的定义
==============
diff --git a/arch/loongarch/Kconfig b/arch/loongarch/Kconfig
index ebdb7156560c7..e30641fa80706 100644
--- a/arch/loongarch/Kconfig
+++ b/arch/loongarch/Kconfig
@@ -85,6 +85,7 @@ config LOONGARCH
select GENERIC_ENTRY
select GENERIC_GETTIMEOFDAY
select GENERIC_IOREMAP if !ARCH_IOREMAP
+ select GENERIC_IRQ_MATRIX_ALLOCATOR
select GENERIC_IRQ_MULTI_HANDLER
select GENERIC_IRQ_PROBE
select GENERIC_IRQ_SHOW
diff --git a/arch/loongarch/include/asm/cpu-features.h b/arch/loongarch/include/asm/cpu-features.h
index 2eafe6a6aca81..16a716f88a5ca 100644
--- a/arch/loongarch/include/asm/cpu-features.h
+++ b/arch/loongarch/include/asm/cpu-features.h
@@ -65,5 +65,6 @@
#define cpu_has_guestid cpu_opt(LOONGARCH_CPU_GUESTID)
#define cpu_has_hypervisor cpu_opt(LOONGARCH_CPU_HYPERVISOR)
#define cpu_has_ptw cpu_opt(LOONGARCH_CPU_PTW)
+#define cpu_has_avecint cpu_opt(LOONGARCH_CPU_AVECINT)
#endif /* __ASM_CPU_FEATURES_H */
diff --git a/arch/loongarch/include/asm/cpu.h b/arch/loongarch/include/asm/cpu.h
index 48b9f7168bcca..843f9c4ec9807 100644
--- a/arch/loongarch/include/asm/cpu.h
+++ b/arch/loongarch/include/asm/cpu.h
@@ -99,6 +99,7 @@ enum cpu_type_enum {
#define CPU_FEATURE_GUESTID 24 /* CPU has GuestID feature */
#define CPU_FEATURE_HYPERVISOR 25 /* CPU has hypervisor (running in VM) */
#define CPU_FEATURE_PTW 26 /* CPU has hardware page table walker */
+#define CPU_FEATURE_AVECINT 27 /* CPU has avec interrupt */
#define LOONGARCH_CPU_CPUCFG BIT_ULL(CPU_FEATURE_CPUCFG)
#define LOONGARCH_CPU_LAM BIT_ULL(CPU_FEATURE_LAM)
@@ -127,5 +128,6 @@ enum cpu_type_enum {
#define LOONGARCH_CPU_GUESTID BIT_ULL(CPU_FEATURE_GUESTID)
#define LOONGARCH_CPU_HYPERVISOR BIT_ULL(CPU_FEATURE_HYPERVISOR)
#define LOONGARCH_CPU_PTW BIT_ULL(CPU_FEATURE_PTW)
+#define LOONGARCH_CPU_AVECINT BIT_ULL(CPU_FEATURE_AVECINT)
#endif /* _ASM_CPU_H */
diff --git a/arch/loongarch/include/asm/hardirq.h b/arch/loongarch/include/asm/hardirq.h
index 1d7feb7195157..10da8d6961cb0 100644
--- a/arch/loongarch/include/asm/hardirq.h
+++ b/arch/loongarch/include/asm/hardirq.h
@@ -12,12 +12,13 @@
extern void ack_bad_irq(unsigned int irq);
#define ack_bad_irq ack_bad_irq
-#define NR_IPI 3
+#define NR_IPI 4
enum ipi_msg_type {
IPI_RESCHEDULE,
IPI_CALL_FUNCTION,
IPI_IRQ_WORK,
+ IPI_CLEAR_VECTOR,
};
typedef struct {
diff --git a/arch/loongarch/include/asm/irq.h b/arch/loongarch/include/asm/irq.h
index 480418bc5071a..1b255bf8168f0 100644
--- a/arch/loongarch/include/asm/irq.h
+++ b/arch/loongarch/include/asm/irq.h
@@ -39,11 +39,22 @@ void spurious_interrupt(void);
#define NR_IRQS_LEGACY 16
+/*
+ * 256 Vectors Mapping for AVECINTC:
+ *
+ * 0 - 15: Mapping classic IPs, e.g. IP0-12.
+ * 16 - 255: Mapping vectors for external IRQ.
+ *
+ */
+#define NR_VECTORS 256
+#define NR_LEGACY_VECTORS 16
+#define IRQ_MATRIX_BITS NR_VECTORS
+
#define arch_trigger_cpumask_backtrace arch_trigger_cpumask_backtrace
void arch_trigger_cpumask_backtrace(const struct cpumask *mask, int exclude_cpu);
#define MAX_IO_PICS 2
-#define NR_IRQS (64 + (256 * MAX_IO_PICS))
+#define NR_IRQS (64 + NR_VECTORS * (NR_CPUS + MAX_IO_PICS))
struct acpi_vector_group {
int node;
@@ -65,7 +76,7 @@ extern struct acpi_vector_group msi_group[MAX_IO_PICS];
#define LOONGSON_LPC_LAST_IRQ (LOONGSON_LPC_IRQ_BASE + 15)
#define LOONGSON_CPU_IRQ_BASE 16
-#define LOONGSON_CPU_LAST_IRQ (LOONGSON_CPU_IRQ_BASE + 14)
+#define LOONGSON_CPU_LAST_IRQ (LOONGSON_CPU_IRQ_BASE + 15)
#define LOONGSON_PCH_IRQ_BASE 64
#define LOONGSON_PCH_ACPI_IRQ (LOONGSON_PCH_IRQ_BASE + 47)
@@ -92,15 +103,21 @@ int liointc_acpi_init(struct irq_domain *parent,
struct acpi_madt_lio_pic *acpi_liointc);
int eiointc_acpi_init(struct irq_domain *parent,
struct acpi_madt_eio_pic *acpi_eiointc);
+int avecintc_acpi_init(struct irq_domain *parent);
+
+void complete_irq_moving(void);
int htvec_acpi_init(struct irq_domain *parent,
struct acpi_madt_ht_pic *acpi_htvec);
int pch_lpc_acpi_init(struct irq_domain *parent,
struct acpi_madt_lpc_pic *acpi_pchlpc);
-int pch_msi_acpi_init(struct irq_domain *parent,
- struct acpi_madt_msi_pic *acpi_pchmsi);
int pch_pic_acpi_init(struct irq_domain *parent,
struct acpi_madt_bio_pic *acpi_pchpic);
+int pch_msi_acpi_init(struct irq_domain *parent,
+ struct acpi_madt_msi_pic *acpi_pchmsi);
+int pch_msi_acpi_init_v2(struct irq_domain *parent,
+ struct acpi_madt_msi_pic *acpi_pchmsi);
+
int find_pch_pic(u32 gsi);
struct fwnode_handle *get_pch_msi_handle(int pci_segment);
diff --git a/arch/loongarch/include/asm/loongarch.h b/arch/loongarch/include/asm/loongarch.h
index 04a78010fc725..70834a47257de 100644
--- a/arch/loongarch/include/asm/loongarch.h
+++ b/arch/loongarch/include/asm/loongarch.h
@@ -253,8 +253,8 @@
#define CSR_ESTAT_EXC_WIDTH 6
#define CSR_ESTAT_EXC (_ULCAST_(0x3f) << CSR_ESTAT_EXC_SHIFT)
#define CSR_ESTAT_IS_SHIFT 0
-#define CSR_ESTAT_IS_WIDTH 14
-#define CSR_ESTAT_IS (_ULCAST_(0x3fff) << CSR_ESTAT_IS_SHIFT)
+#define CSR_ESTAT_IS_WIDTH 15
+#define CSR_ESTAT_IS (_ULCAST_(0x7fff) << CSR_ESTAT_IS_SHIFT)
#define LOONGARCH_CSR_ERA 0x6 /* ERA */
@@ -649,6 +649,13 @@
#define LOONGARCH_CSR_CTAG 0x98 /* TagLo + TagHi */
+#define LOONGARCH_CSR_ISR0 0xa0
+#define LOONGARCH_CSR_ISR1 0xa1
+#define LOONGARCH_CSR_ISR2 0xa2
+#define LOONGARCH_CSR_ISR3 0xa3
+
+#define LOONGARCH_CSR_IRR 0xa4
+
#define LOONGARCH_CSR_PRID 0xc0
/* Shadow MCSR : 0xc0 ~ 0xff */
@@ -1011,7 +1018,7 @@
/*
* CSR_ECFG IM
*/
-#define ECFG0_IM 0x00001fff
+#define ECFG0_IM 0x00005fff
#define ECFGB_SIP0 0
#define ECFGF_SIP0 (_ULCAST_(1) << ECFGB_SIP0)
#define ECFGB_SIP1 1
@@ -1054,6 +1061,7 @@
#define IOCSRF_EIODECODE BIT_ULL(9)
#define IOCSRF_FLATMODE BIT_ULL(10)
#define IOCSRF_VM BIT_ULL(11)
+#define IOCSRF_AVEC BIT_ULL(15)
#define LOONGARCH_IOCSR_VENDOR 0x10
@@ -1065,6 +1073,7 @@
#define IOCSR_MISC_FUNC_SOFT_INT BIT_ULL(10)
#define IOCSR_MISC_FUNC_TIMER_RESET BIT_ULL(21)
#define IOCSR_MISC_FUNC_EXT_IOI_EN BIT_ULL(48)
+#define IOCSR_MISC_FUNC_AVEC_EN BIT_ULL(51)
#define LOONGARCH_IOCSR_CPUTEMP 0x428
@@ -1387,9 +1396,10 @@ __BUILD_CSR_OP(tlbidx)
#define INT_TI 11 /* Timer */
#define INT_IPI 12
#define INT_NMI 13
+#define INT_AVEC 14
/* ExcCodes corresponding to interrupts */
-#define EXCCODE_INT_NUM (INT_NMI + 1)
+#define EXCCODE_INT_NUM (INT_AVEC + 1)
#define EXCCODE_INT_START 64
#define EXCCODE_INT_END (EXCCODE_INT_START + EXCCODE_INT_NUM - 1)
diff --git a/arch/loongarch/include/asm/smp.h b/arch/loongarch/include/asm/smp.h
index 50db503f44e3c..3383c9d24e942 100644
--- a/arch/loongarch/include/asm/smp.h
+++ b/arch/loongarch/include/asm/smp.h
@@ -70,10 +70,12 @@ extern int __cpu_logical_map[NR_CPUS];
#define ACTION_RESCHEDULE 1
#define ACTION_CALL_FUNCTION 2
#define ACTION_IRQ_WORK 3
+#define ACTION_CLEAR_VECTOR 4
#define SMP_BOOT_CPU BIT(ACTION_BOOT_CPU)
#define SMP_RESCHEDULE BIT(ACTION_RESCHEDULE)
#define SMP_CALL_FUNCTION BIT(ACTION_CALL_FUNCTION)
#define SMP_IRQ_WORK BIT(ACTION_IRQ_WORK)
+#define SMP_CLEAR_VECTOR BIT(ACTION_CLEAR_VECTOR)
struct secondary_data {
unsigned long stack;
diff --git a/arch/loongarch/kernel/cpu-probe.c b/arch/loongarch/kernel/cpu-probe.c
index 55320813ee081..14f0449f54520 100644
--- a/arch/loongarch/kernel/cpu-probe.c
+++ b/arch/loongarch/kernel/cpu-probe.c
@@ -106,7 +106,6 @@ static void cpu_probe_common(struct cpuinfo_loongarch *c)
elf_hwcap |= HWCAP_LOONGARCH_CRC32;
}
-
config = read_cpucfg(LOONGARCH_CPUCFG2);
if (config & CPUCFG2_LAM) {
c->options |= LOONGARCH_CPU_LAM;
@@ -174,6 +173,8 @@ static void cpu_probe_common(struct cpuinfo_loongarch *c)
c->options |= LOONGARCH_CPU_FLATMODE;
if (config & IOCSRF_EIODECODE)
c->options |= LOONGARCH_CPU_EIODECODE;
+ if (config & IOCSRF_AVEC)
+ c->options |= LOONGARCH_CPU_AVECINT;
if (config & IOCSRF_VM)
c->options |= LOONGARCH_CPU_HYPERVISOR;
diff --git a/arch/loongarch/kernel/irq.c b/arch/loongarch/kernel/irq.c
index f4991c03514f4..1311546a7b4ee 100644
--- a/arch/loongarch/kernel/irq.c
+++ b/arch/loongarch/kernel/irq.c
@@ -87,6 +87,18 @@ static void __init init_vec_parent_group(void)
acpi_table_parse(ACPI_SIG_MCFG, early_pci_mcfg_parse);
}
+int __init arch_probe_nr_irqs(void)
+{
+ int nr_io_pics = bitmap_weight(loongson_sysconf.cores_io_master, NR_CPUS);
+
+ if (!cpu_has_avecint)
+ nr_irqs = (64 + NR_VECTORS * nr_io_pics);
+ else
+ nr_irqs = (64 + NR_VECTORS * (nr_cpu_ids + nr_io_pics));
+
+ return NR_IRQS_LEGACY;
+}
+
void __init init_IRQ(void)
{
int i;
@@ -102,7 +114,7 @@ void __init init_IRQ(void)
mp_ops.init_ipi();
#endif
- for (i = 0; i < NR_IRQS; i++)
+ for (i = 0; i < nr_irqs; i++)
irq_set_noprobe(i);
for_each_possible_cpu(i) {
diff --git a/arch/loongarch/kernel/paravirt.c b/arch/loongarch/kernel/paravirt.c
index 9c9b75b76f62f..4d736a4e488dd 100644
--- a/arch/loongarch/kernel/paravirt.c
+++ b/arch/loongarch/kernel/paravirt.c
@@ -134,6 +134,11 @@ static irqreturn_t pv_ipi_interrupt(int irq, void *dev)
info->ipi_irqs[IPI_IRQ_WORK]++;
}
+ if (action & SMP_CLEAR_VECTOR) {
+ complete_irq_moving();
+ info->ipi_irqs[IPI_CLEAR_VECTOR]++;
+ }
+
return IRQ_HANDLED;
}
diff --git a/arch/loongarch/kernel/smp.c b/arch/loongarch/kernel/smp.c
index ca405ab86aaef..4adbbef3450ac 100644
--- a/arch/loongarch/kernel/smp.c
+++ b/arch/loongarch/kernel/smp.c
@@ -72,6 +72,7 @@ static const char *ipi_types[NR_IPI] __tracepoint_string = {
[IPI_RESCHEDULE] = "Rescheduling interrupts",
[IPI_CALL_FUNCTION] = "Function call interrupts",
[IPI_IRQ_WORK] = "IRQ work interrupts",
+ [IPI_CLEAR_VECTOR] = "Clear vector interrupts",
};
void show_ipi_list(struct seq_file *p, int prec)
@@ -248,6 +249,11 @@ static irqreturn_t loongson_ipi_interrupt(int irq, void *dev)
per_cpu(irq_stat, cpu).ipi_irqs[IPI_IRQ_WORK]++;
}
+ if (action & SMP_CLEAR_VECTOR) {
+ complete_irq_moving();
+ per_cpu(irq_stat, cpu).ipi_irqs[IPI_CLEAR_VECTOR]++;
+ }
+
return IRQ_HANDLED;
}
diff --git a/drivers/irqchip/Makefile b/drivers/irqchip/Makefile
index 15635812b2d66..e3679ec2b9f76 100644
--- a/drivers/irqchip/Makefile
+++ b/drivers/irqchip/Makefile
@@ -110,7 +110,7 @@ obj-$(CONFIG_LS1X_IRQ) += irq-ls1x.o
obj-$(CONFIG_TI_SCI_INTR_IRQCHIP) += irq-ti-sci-intr.o
obj-$(CONFIG_TI_SCI_INTA_IRQCHIP) += irq-ti-sci-inta.o
obj-$(CONFIG_TI_PRUSS_INTC) += irq-pruss-intc.o
-obj-$(CONFIG_IRQ_LOONGARCH_CPU) += irq-loongarch-cpu.o
+obj-$(CONFIG_IRQ_LOONGARCH_CPU) += irq-loongarch-cpu.o irq-loongarch-avec.o
obj-$(CONFIG_LOONGSON_LIOINTC) += irq-loongson-liointc.o
obj-$(CONFIG_LOONGSON_EIOINTC) += irq-loongson-eiointc.o
obj-$(CONFIG_LOONGSON_HTPIC) += irq-loongson-htpic.o
diff --git a/drivers/irqchip/irq-loongarch-avec.c b/drivers/irqchip/irq-loongarch-avec.c
new file mode 100644
index 0000000000000..0033aafeb1ecf
--- /dev/null
+++ b/drivers/irqchip/irq-loongarch-avec.c
@@ -0,0 +1,447 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2020-2024 Loongson Technologies, Inc.
+ */
+
+#include <linux/cpuhotplug.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/irqchip.h>
+#include <linux/irqchip/chained_irq.h>
+#include <linux/irqdomain.h>
+#include <linux/kernel.h>
+#include <linux/msi.h>
+#include <linux/radix-tree.h>
+#include <linux/spinlock.h>
+
+#include <asm/loongarch.h>
+#include <asm/setup.h>
+
+#define VECTORS_PER_REG 64
+#define IRR_VECTOR_MASK 0xffUL
+#define IRR_INVALID_MASK 0x80000000UL
+#define AVEC_MSG_OFFSET 0x100000
+
+static phys_addr_t msi_base_addr;
+
+#ifdef CONFIG_SMP
+struct pending_list {
+ struct list_head head;
+};
+
+static struct cpumask intersect_mask;
+static DEFINE_PER_CPU(struct pending_list, pending_list);
+#endif
+
+static DEFINE_PER_CPU(struct irq_desc * [NR_VECTORS], irq_map);
+
+struct avecintc_chip {
+ struct fwnode_handle *fwnode;
+ struct irq_domain *domain;
+ struct irq_matrix *vector_matrix;
+ raw_spinlock_t lock;
+};
+
+static struct avecintc_chip loongarch_avec;
+
+struct avecintc_data {
+ struct list_head entry;
+ unsigned int cpu;
+ unsigned int vec;
+ unsigned int prev_cpu;
+ unsigned int prev_vec;
+ unsigned int moving : 1,
+ managed : 1;
+};
+
+static inline void avecintc_ack_irq(struct irq_data *d)
+{
+}
+
+static inline void avecintc_mask_irq(struct irq_data *d)
+{
+}
+
+static inline void avecintc_unmask_irq(struct irq_data *d)
+{
+}
+
+#ifdef CONFIG_SMP
+static inline void pending_list_init(int cpu)
+{
+ struct pending_list *plist = per_cpu_ptr(&pending_list, cpu);
+
+ INIT_LIST_HEAD(&plist->head);
+}
+
+static void avecintc_sync(struct avecintc_data *adata)
+{
+ struct pending_list *plist;
+
+ if (cpu_online(adata->prev_cpu)) {
+ plist = per_cpu_ptr(&pending_list, adata->prev_cpu);
+ list_add_tail(&adata->entry, &plist->head);
+ adata->moving = 1;
+ mp_ops.send_ipi_single(adata->prev_cpu, ACTION_CLEAR_VECTOR);
+ }
+}
+
+static int avecintc_set_affinity(struct irq_data *data, const struct cpumask *dest,
+ bool force)
+{
+ unsigned int cpu, ret, vector;
+ unsigned long flags;
+ struct avecintc_data *adata;
+
+ raw_spin_lock_irqsave(&loongarch_avec.lock, flags);
+ adata = irq_data_get_irq_chip_data(data);
+
+ if (adata->vec && cpu_online(adata->cpu) && cpumask_test_cpu(adata->cpu, dest)) {
+ raw_spin_unlock_irqrestore(&loongarch_avec.lock, flags);
+ return 0;
+ }
+
+ if (adata->moving)
+ return -EBUSY;
+
+ cpumask_and(&intersect_mask, dest, cpu_online_mask);
+
+ ret = irq_matrix_alloc(loongarch_avec.vector_matrix, &intersect_mask, false, &cpu);
+ if (ret < 0) {
+ raw_spin_unlock_irqrestore(&loongarch_avec.lock, flags);
+ return ret;
+ }
+
+ vector = ret;
+ adata->cpu = cpu;
+ adata->vec = vector;
+ per_cpu_ptr(irq_map, adata->cpu)[adata->vec] = irq_data_to_desc(data);
+ avecintc_sync(adata);
+
+ raw_spin_unlock_irqrestore(&loongarch_avec.lock, flags);
+ irq_data_update_effective_affinity(data, cpumask_of(cpu));
+
+ return IRQ_SET_MASK_OK;
+}
+
+static int avecintc_cpu_online(unsigned int cpu)
+{
+ unsigned long flags;
+
+ if (!loongarch_avec.vector_matrix)
+ return 0;
+
+ raw_spin_lock_irqsave(&loongarch_avec.lock, flags);
+
+ irq_matrix_online(loongarch_avec.vector_matrix);
+
+ pending_list_init(cpu);
+
+ raw_spin_unlock_irqrestore(&loongarch_avec.lock, flags);
+
+ return 0;
+}
+
+static int avecintc_cpu_offline(unsigned int cpu)
+{
+ unsigned long flags;
+ struct pending_list *plist = per_cpu_ptr(&pending_list, cpu);
+
+ if (!loongarch_avec.vector_matrix)
+ return 0;
+
+ raw_spin_lock_irqsave(&loongarch_avec.lock, flags);
+ if (list_empty(&plist->head))
+ irq_matrix_offline(loongarch_avec.vector_matrix);
+ else
+ pr_warn("CPU#%d advanced vector is busy\n", cpu);
+ raw_spin_unlock_irqrestore(&loongarch_avec.lock, flags);
+
+ return 0;
+}
+
+void complete_irq_moving(void)
+{
+ struct pending_list *plist = this_cpu_ptr(&pending_list);
+ struct avecintc_data *adata, *tdata;
+ int cpu, vector, bias;
+ uint64_t isr;
+
+ raw_spin_lock(&loongarch_avec.lock);
+
+ list_for_each_entry_safe(adata, tdata, &plist->head, entry) {
+ cpu = adata->prev_cpu;
+ vector = adata->prev_vec;
+ bias = vector / VECTORS_PER_REG;
+ switch (bias) {
+ case 0:
+ isr = csr_read64(LOONGARCH_CSR_ISR0);
+ case 1:
+ isr = csr_read64(LOONGARCH_CSR_ISR1);
+ case 2:
+ isr = csr_read64(LOONGARCH_CSR_ISR2);
+ case 3:
+ isr = csr_read64(LOONGARCH_CSR_ISR3);
+ }
+
+ if (isr & (1UL << (vector % VECTORS_PER_REG))) {
+ mp_ops.send_ipi_single(cpu, ACTION_CLEAR_VECTOR);
+ continue;
+ }
+ list_del(&adata->entry);
+ irq_matrix_free(loongarch_avec.vector_matrix, cpu, vector, adata->managed);
+ this_cpu_write(irq_map[vector], NULL);
+ adata->prev_cpu = adata->cpu;
+ adata->prev_vec = adata->vec;
+ adata->moving = 0;
+ }
+
+ raw_spin_unlock(&loongarch_avec.lock);
+}
+#endif
+
+static void avecintc_compose_msi_msg(struct irq_data *d, struct msi_msg *msg)
+{
+ struct avecintc_data *adata;
+
+ adata = irq_data_get_irq_chip_data(d);
+
+ msg->address_hi = 0x0;
+ msg->address_lo = (msi_base_addr | (adata->vec & 0xff) << 4) |
+ ((cpu_logical_map(adata->cpu & 0xffff)) << 12);
+ msg->data = 0x0;
+}
+
+static struct irq_chip avec_irq_controller = {
+ .name = "AVECINTC",
+ .irq_ack = avecintc_ack_irq,
+ .irq_mask = avecintc_mask_irq,
+ .irq_unmask = avecintc_unmask_irq,
+#ifdef CONFIG_SMP
+ .irq_set_affinity = avecintc_set_affinity,
+#endif
+ .irq_compose_msi_msg = avecintc_compose_msi_msg,
+};
+
+static void avecintc_irq_dispatch(struct irq_desc *desc)
+{
+ unsigned long vector;
+ struct irq_chip *chip = irq_desc_get_chip(desc);
+ struct irq_desc *d;
+
+ chained_irq_enter(chip, desc);
+
+ while (true) {
+ vector = csr_read64(LOONGARCH_CSR_IRR);
+ if (vector & IRR_INVALID_MASK)
+ break;
+
+ vector &= IRR_VECTOR_MASK;
+
+ d = this_cpu_read(irq_map[vector]);
+ if (d)
+ generic_handle_irq_desc(d);
+ else {
+ spurious_interrupt();
+ pr_warn("Unexpected IRQ occurs on CPU#%d [vector %ld]\n", smp_processor_id(), vector);
+ }
+ }
+
+ chained_irq_exit(chip, desc);
+}
+
+static int avecintc_domain_alloc(struct irq_domain *domain,
+ unsigned int virq, unsigned int nr_irqs, void *arg)
+{
+ unsigned int cpu, i, ret;
+ unsigned long flags;
+ struct irq_data *irqd;
+ struct avecintc_data *adata;
+
+ raw_spin_lock_irqsave(&loongarch_avec.lock, flags);
+ for (i = 0; i < nr_irqs; i++) {
+ irqd = irq_domain_get_irq_data(domain, virq + i);
+ adata = kzalloc(sizeof(*adata), GFP_KERNEL);
+ if (!adata) {
+ raw_spin_unlock_irqrestore(&loongarch_avec.lock, flags);
+ return -ENOMEM;
+ }
+
+ ret = irq_matrix_alloc(loongarch_avec.vector_matrix, cpu_online_mask, false, &cpu);
+ if (ret < 0) {
+ raw_spin_unlock_irqrestore(&loongarch_avec.lock, flags);
+ return ret;
+ }
+
+ adata->prev_cpu = adata->cpu = cpu;
+ adata->prev_vec = adata->vec = ret;
+ adata->managed = irqd_affinity_is_managed(irqd);
+ irq_domain_set_info(domain, virq + i, virq + i, &avec_irq_controller,
+ adata, handle_edge_irq, NULL, NULL);
+ adata->moving = 0;
+ irqd_set_single_target(irqd);
+ irqd_set_affinity_on_activate(irqd);
+
+ per_cpu_ptr(irq_map, adata->cpu)[adata->vec] = irq_data_to_desc(irqd);
+ }
+ raw_spin_unlock_irqrestore(&loongarch_avec.lock, flags);
+
+ return 0;
+}
+
+static void clear_free_vector(struct irq_data *irqd)
+{
+ bool managed = irqd_affinity_is_managed(irqd);
+ struct avecintc_data *adata = irq_data_get_irq_chip_data(irqd);
+
+ per_cpu(irq_map, adata->cpu)[adata->vec] = NULL;
+ irq_matrix_free(loongarch_avec.vector_matrix, adata->cpu, adata->vec, managed);
+ adata->cpu = 0;
+ adata->vec = 0;
+
+#ifdef CONFIG_SMP
+ if (!adata->moving)
+ return;
+
+ per_cpu(irq_map, adata->prev_cpu)[adata->prev_vec] = NULL;
+ irq_matrix_free(loongarch_avec.vector_matrix,
+ adata->prev_cpu, adata->prev_vec, adata->managed);
+ adata->moving = 0;
+ adata->prev_vec = 0;
+ adata->prev_cpu = 0;
+ list_del_init(&adata->entry);
+#endif
+}
+
+static void avecintc_domain_free(struct irq_domain *domain,
+ unsigned int virq, unsigned int nr_irqs)
+{
+ unsigned int i;
+ unsigned long flags;
+ struct irq_data *d;
+
+ raw_spin_lock_irqsave(&loongarch_avec.lock, flags);
+ for (i = 0; i < nr_irqs; i++) {
+ d = irq_domain_get_irq_data(domain, virq + i);
+ if (d) {
+ clear_free_vector(d);
+ irq_domain_reset_irq_data(d);
+
+ }
+ }
+
+ raw_spin_unlock_irqrestore(&loongarch_avec.lock, flags);
+}
+
+static const struct irq_domain_ops avecintc_domain_ops = {
+ .alloc = avecintc_domain_alloc,
+ .free = avecintc_domain_free,
+};
+
+static int __init irq_matrix_init(void)
+{
+ int i;
+
+ loongarch_avec.vector_matrix = irq_alloc_matrix(NR_VECTORS, 0, NR_VECTORS - 1);
+ if (!loongarch_avec.vector_matrix)
+ return -ENOMEM;
+
+ for (i = 0; i < NR_LEGACY_VECTORS; i++)
+ irq_matrix_assign_system(loongarch_avec.vector_matrix, i, false);
+
+ irq_matrix_online(loongarch_avec.vector_matrix);
+
+ return 0;
+}
+
+static int __init avecintc_init(struct irq_domain *parent)
+{
+ int ret, parent_irq;
+ unsigned long value;
+
+ raw_spin_lock_init(&loongarch_avec.lock);
+
+ loongarch_avec.fwnode = irq_domain_alloc_named_fwnode("AVECINTC");
+ if (!loongarch_avec.fwnode) {
+ pr_err("Unable to allocate domain handle\n");
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ loongarch_avec.domain = irq_domain_create_tree(loongarch_avec.fwnode,
+ &avecintc_domain_ops, NULL);
+ if (!loongarch_avec.domain) {
+ pr_err("Unable to create IRQ domain\n");
+ ret = -ENOMEM;
+ goto out_free_handle;
+ }
+
+ parent_irq = irq_create_mapping(parent, INT_AVEC);
+ if (!parent_irq) {
+ pr_err("Failed to mapping hwirq\n");
+ ret = -EINVAL;
+ goto out_remove_domain;
+ }
+ irq_set_chained_handler_and_data(parent_irq, avecintc_irq_dispatch, NULL);
+
+ ret = irq_matrix_init();
+ if (ret < 0) {
+ pr_err("Failed to init irq matrix\n");
+ goto out_free_matrix;
+ }
+#ifdef CONFIG_SMP
+ pending_list_init(0);
+ ret = cpuhp_setup_state_nocalls(CPUHP_AP_IRQ_AVECINTC_STARTING,
+ "irqchip/loongarch/avecintc:starting",
+ avecintc_cpu_online, avecintc_cpu_offline);
+#endif
+ value = iocsr_read64(LOONGARCH_IOCSR_MISC_FUNC);
+ value |= IOCSR_MISC_FUNC_AVEC_EN;
+ iocsr_write64(value, LOONGARCH_IOCSR_MISC_FUNC);
+
+ return ret;
+
+out_free_matrix:
+ kfree(loongarch_avec.vector_matrix);
+out_remove_domain:
+ irq_domain_remove(loongarch_avec.domain);
+out_free_handle:
+ irq_domain_free_fwnode(loongarch_avec.fwnode);
+out:
+ return ret;
+}
+
+static int __init pch_msi_parse_madt(union acpi_subtable_headers *header,
+ const unsigned long end)
+{
+ struct acpi_madt_msi_pic *pchmsi_entry = (struct acpi_madt_msi_pic *)header;
+
+ msi_base_addr = pchmsi_entry->msg_address - AVEC_MSG_OFFSET;
+
+ return pch_msi_acpi_init_v2(loongarch_avec.domain, pchmsi_entry);
+}
+
+static inline int __init acpi_cascade_irqdomain_init(void)
+{
+ return acpi_table_parse_madt(ACPI_MADT_TYPE_MSI_PIC, pch_msi_parse_madt, 1);
+}
+
+int __init avecintc_acpi_init(struct irq_domain *parent)
+{
+ int ret;
+
+ ret = avecintc_init(parent);
+ if (ret < 0) {
+ pr_err("Failed to init IRQ domain\n");
+ return ret;
+ }
+
+ ret = acpi_cascade_irqdomain_init();
+ if (ret < 0) {
+ pr_err("Failed to init cascade IRQ domain\n");
+ return ret;
+ }
+
+ return ret;
+}
diff --git a/drivers/irqchip/irq-loongarch-cpu.c b/drivers/irqchip/irq-loongarch-cpu.c
index b35903a06902f..808d241c481fb 100644
--- a/drivers/irqchip/irq-loongarch-cpu.c
+++ b/drivers/irqchip/irq-loongarch-cpu.c
@@ -140,7 +140,10 @@ static int __init acpi_cascade_irqdomain_init(void)
if (r < 0)
return r;
- return 0;
+ if (cpu_has_avecint)
+ r = avecintc_acpi_init(irq_domain);
+
+ return r;
}
static int __init cpuintc_acpi_init(union acpi_subtable_headers *header,
diff --git a/drivers/irqchip/irq-loongson-eiointc.c b/drivers/irqchip/irq-loongson-eiointc.c
index b1f2080be2beb..895d15b96669e 100644
--- a/drivers/irqchip/irq-loongson-eiointc.c
+++ b/drivers/irqchip/irq-loongson-eiointc.c
@@ -360,6 +360,9 @@ static int __init acpi_cascade_irqdomain_init(void)
if (r < 0)
return r;
+ if (cpu_has_avecint)
+ return 0;
+
r = acpi_table_parse_madt(ACPI_MADT_TYPE_MSI_PIC, pch_msi_parse_madt, 1);
if (r < 0)
return r;
@@ -396,8 +399,8 @@ static int __init eiointc_init(struct eiointc_priv *priv, int parent_irq,
if (nr_pics == 1) {
register_syscore_ops(&eiointc_syscore_ops);
- cpuhp_setup_state_nocalls(CPUHP_AP_IRQ_LOONGARCH_STARTING,
- "irqchip/loongarch/intc:starting",
+ cpuhp_setup_state_nocalls(CPUHP_AP_IRQ_EIOINTC_STARTING,
+ "irqchip/loongarch/eiointc:starting",
eiointc_router_init, NULL);
}
diff --git a/drivers/irqchip/irq-loongson-pch-msi.c b/drivers/irqchip/irq-loongson-pch-msi.c
index dd4d699170f4e..e4c22c239628e 100644
--- a/drivers/irqchip/irq-loongson-pch-msi.c
+++ b/drivers/irqchip/irq-loongson-pch-msi.c
@@ -268,11 +268,14 @@ struct fwnode_handle *get_pch_msi_handle(int pci_segment)
{
int i;
+ if (cpu_has_avecint)
+ return pch_msi_handle[0];
+
for (i = 0; i < MAX_IO_PICS; i++) {
if (msi_group[i].pci_segment == pci_segment)
return pch_msi_handle[i];
}
- return NULL;
+ return pch_msi_handle[0];
}
int __init pch_msi_acpi_init(struct irq_domain *parent,
@@ -289,4 +292,40 @@ int __init pch_msi_acpi_init(struct irq_domain *parent,
return ret;
}
+
+static struct irq_chip pch_msi_irq_chip_v2 = {
+ .name = "PCH PCI MSI",
+ .irq_ack = irq_chip_ack_parent,
+};
+
+static struct msi_domain_info pch_msi_domain_info_v2 = {
+ .flags = MSI_FLAG_USE_DEF_DOM_OPS | MSI_FLAG_USE_DEF_CHIP_OPS |
+ MSI_FLAG_MULTI_PCI_MSI | MSI_FLAG_PCI_MSIX,
+ .chip = &pch_msi_irq_chip_v2,
+};
+
+int __init pch_msi_acpi_init_v2(struct irq_domain *parent,
+ struct acpi_madt_msi_pic *acpi_pchmsi)
+{
+ struct irq_domain *msi_domain;
+
+ if (pch_msi_handle[0])
+ return 0;
+
+ pch_msi_handle[0] = irq_domain_alloc_fwnode(&acpi_pchmsi->msg_address);
+ if (!pch_msi_handle[0]) {
+ pr_err("Unable to allocate domain handle\n");
+ return -ENOMEM;
+ }
+
+ msi_domain = pci_msi_create_irq_domain(pch_msi_handle[0],
+ &pch_msi_domain_info_v2, parent);
+ if (!msi_domain) {
+ pr_err("Failed to create PCI MSI domain\n");
+ kfree(pch_msi_handle[0]);
+ return -ENOMEM;
+ }
+
+ return 0;
+}
#endif
diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h
index 89f5c34ce4df9..287605bd5d4a7 100644
--- a/include/linux/cpuhotplug.h
+++ b/include/linux/cpuhotplug.h
@@ -144,7 +144,8 @@ enum cpuhp_state {
CPUHP_AP_IRQ_ARMADA_XP_STARTING,
CPUHP_AP_IRQ_BCM2836_STARTING,
CPUHP_AP_IRQ_MIPS_GIC_STARTING,
- CPUHP_AP_IRQ_LOONGARCH_STARTING,
+ CPUHP_AP_IRQ_EIOINTC_STARTING,
+ CPUHP_AP_IRQ_AVECINTC_STARTING,
CPUHP_AP_IRQ_SIFIVE_PLIC_STARTING,
CPUHP_AP_IRQ_RISCV_IMSIC_STARTING,
CPUHP_AP_ARM_MVEBU_COHERENCY,