target/xtensa: add MX interrupt controller

MX interrupt controller is a collection of the following devices
accessible through the external registers interface:
- interrupt distributor can route each external IRQ line to the
  corresponding external IRQ pin of selected subset of connected xtensa
  cores. It has per-CPU and per-IRQ enable signals and per-IRQ software
  assert signals;
- IPI controller has 16 per-CPU IPI signals that may be routed to a
  combination of 3 designated external IRQ pins of connected xtensa
  cores;
- cache coherecy register controls core L1 cache participation in the
  SMP cluster cache coherency protocol;
- runstall register lets BSP core stall and unstall AP cores.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>

hw/xtensa/Makefile.objs

@@ -1,3 +1,4 @@
+obj-y += mx_pic.o
 obj-y += pic_cpu.o
 obj-y += sim.o
 obj-y += xtensa_memory.o

hw/xtensa/mx_pic.c

@@ -0,0 +1,354 @@
+/*
+ * Copyright (c) 2013 - 2019, Max Filippov, Open Source and Linux Lab.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in the
+ *       documentation and/or other materials provided with the distribution.
+ *     * Neither the name of the Open Source and Linux Lab nor the
+ *       names of its contributors may be used to endorse or promote products
+ *       derived from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
+ * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+ * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+ * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include "qemu/osdep.h"
+#include "hw/hw.h"
+#include "hw/xtensa/mx_pic.h"
+#include "qemu/log.h"
+
+#define MX_MAX_CPU 32
+#define MX_MAX_IRQ 32
+
+#define MIROUT 0x0
+#define MIPICAUSE 0x100
+#define MIPISET 0x140
+#define MIENG 0x180
+#define MIENGSET 0x184
+#define MIASG 0x188
+#define MIASGSET 0x18c
+#define MIPIPART 0x190
+#define SYSCFGID 0x1a0
+#define MPSCORE 0x200
+#define CCON 0x220
+
+struct XtensaMxPic {
+    unsigned n_cpu;
+    unsigned n_irq;
+
+    uint32_t ext_irq_state;
+    uint32_t mieng;
+    uint32_t miasg;
+    uint32_t mirout[MX_MAX_IRQ];
+    uint32_t mipipart;
+    uint32_t runstall;
+
+    qemu_irq *irq_inputs;
+    struct XtensaMxPicCpu {
+        XtensaMxPic *mx;
+        qemu_irq *irq;
+        qemu_irq runstall;
+        uint32_t mipicause;
+        uint32_t mirout_cache;
+        uint32_t irq_state_cache;
+        uint32_t ccon;
+        MemoryRegion reg;
+    } cpu[MX_MAX_CPU];
+};
+
+static uint64_t xtensa_mx_pic_ext_reg_read(void *opaque, hwaddr offset,
+                                           unsigned size)
+{
+    struct XtensaMxPicCpu *mx_cpu = opaque;
+    struct XtensaMxPic *mx = mx_cpu->mx;
+
+    if (offset < MIROUT + MX_MAX_IRQ) {
+        return mx->mirout[offset - MIROUT];
+    } else if (offset >= MIPICAUSE && offset < MIPICAUSE + MX_MAX_CPU) {
+        return mx->cpu[offset - MIPICAUSE].mipicause;
+    } else {
+        switch (offset) {
+        case MIENG:
+            return mx->mieng;
+
+        case MIASG:
+            return mx->miasg;
+
+        case MIPIPART:
+            return mx->mipipart;
+
+        case SYSCFGID:
+            return ((mx->n_cpu - 1) << 18) | (mx_cpu - mx->cpu);
+
+        case MPSCORE:
+            return mx->runstall;
+
+        case CCON:
+            return mx_cpu->ccon;
+
+        default:
+            qemu_log_mask(LOG_GUEST_ERROR,
+                          "unknown RER in MX PIC range: 0x%08x\n",
+                          (uint32_t)offset);
+            return 0;
+        }
+    }
+}
+
+static uint32_t xtensa_mx_pic_get_ipi_for_cpu(const XtensaMxPic *mx,
+                                              unsigned cpu)
+{
+    uint32_t mipicause = mx->cpu[cpu].mipicause;
+    uint32_t mipipart = mx->mipipart;
+
+    return (((mipicause & 1) << (mipipart & 3)) |
+            ((mipicause & 0x000e) != 0) << ((mipipart >> 2) & 3) |
+            ((mipicause & 0x00f0) != 0) << ((mipipart >> 4) & 3) |
+            ((mipicause & 0xff00) != 0) << ((mipipart >> 6) & 3)) & 0x7;
+}
+
+static uint32_t xtensa_mx_pic_get_ext_irq_for_cpu(const XtensaMxPic *mx,
+                                                  unsigned cpu)
+{
+    return ((((mx->ext_irq_state & mx->mieng) | mx->miasg) &
+             mx->cpu[cpu].mirout_cache) << 2) |
+        xtensa_mx_pic_get_ipi_for_cpu(mx, cpu);
+}
+
+static void xtensa_mx_pic_update_cpu(XtensaMxPic *mx, unsigned cpu)
+{
+    uint32_t irq = xtensa_mx_pic_get_ext_irq_for_cpu(mx, cpu);
+    uint32_t changed_irq = mx->cpu[cpu].irq_state_cache ^ irq;
+    unsigned i;
+
+    qemu_log_mask(CPU_LOG_INT, "%s: CPU %d, irq: %08x, changed_irq: %08x\n",
+                  __func__, cpu, irq, changed_irq);
+    mx->cpu[cpu].irq_state_cache = irq;
+    for (i = 0; changed_irq; ++i) {
+        uint32_t mask = 1u << i;
+
+        if (changed_irq & mask) {
+            changed_irq ^= mask;
+            qemu_set_irq(mx->cpu[cpu].irq[i], irq & mask);
+        }
+    }
+}
+
+static void xtensa_mx_pic_update_all(XtensaMxPic *mx)
+{
+    unsigned cpu;
+
+    for (cpu = 0; cpu < mx->n_cpu; ++cpu) {
+        xtensa_mx_pic_update_cpu(mx, cpu);
+    }
+}
+
+static void xtensa_mx_pic_ext_reg_write(void *opaque, hwaddr offset,
+                                        uint64_t v, unsigned size)
+{
+    struct XtensaMxPicCpu *mx_cpu = opaque;
+    struct XtensaMxPic *mx = mx_cpu->mx;
+    unsigned cpu;
+
+    if (offset < MIROUT + mx->n_irq) {
+        mx->mirout[offset - MIROUT] = v;
+        for (cpu = 0; cpu < mx->n_cpu; ++cpu) {
+            uint32_t mask = 1u << (offset - MIROUT);
+
+            if (!(mx->cpu[cpu].mirout_cache & mask) != !(v & (1u << cpu))) {
+                mx->cpu[cpu].mirout_cache ^= mask;
+                xtensa_mx_pic_update_cpu(mx, cpu);
+            }
+        }
+    } else if (offset >= MIPICAUSE && offset < MIPICAUSE + mx->n_cpu) {
+        cpu = offset - MIPICAUSE;
+        mx->cpu[cpu].mipicause &= ~v;
+        xtensa_mx_pic_update_cpu(mx, cpu);
+    } else if (offset >= MIPISET && offset < MIPISET + 16) {
+        for (cpu = 0; cpu < mx->n_cpu; ++cpu) {
+            if (v & (1u << cpu)) {
+                mx->cpu[cpu].mipicause |= 1u << (offset - MIPISET);
+                xtensa_mx_pic_update_cpu(mx, cpu);
+            }
+        }
+    } else {
+        uint32_t change = 0;
+        uint32_t oldv, newv;
+        const char *name = "???";
+
+        switch (offset) {
+        case MIENG:
+            change = mx->mieng & v;
+            oldv = mx->mieng;
+            mx->mieng &= ~v;
+            newv = mx->mieng;
+            name = "MIENG";
+            break;
+
+        case MIENGSET:
+            change = ~mx->mieng & v;
+            oldv = mx->mieng;
+            mx->mieng |= v;
+            newv = mx->mieng;
+            name = "MIENG";
+            break;
+
+        case MIASG:
+            change = mx->miasg & v;
+            oldv = mx->miasg;
+            mx->miasg &= ~v;
+            newv = mx->miasg;
+            name = "MIASG";
+            break;
+
+        case MIASGSET:
+            change = ~mx->miasg & v;
+            oldv = mx->miasg;
+            mx->miasg |= v;
+            newv = mx->miasg;
+            name = "MIASG";
+            break;
+
+        case MIPIPART:
+            change = mx->mipipart ^ v;
+            oldv = mx->mipipart;
+            mx->mipipart = v;
+            newv = mx->mipipart;
+            name = "MIPIPART";
+            break;
+
+        case MPSCORE:
+            change = mx->runstall ^ v;
+            oldv = mx->runstall;
+            mx->runstall = v;
+            newv = mx->runstall;
+            name = "RUNSTALL";
+            for (cpu = 0; cpu < mx->n_cpu; ++cpu) {
+                if (change & (1u << cpu)) {
+                    qemu_set_irq(mx->cpu[cpu].runstall, v & (1u << cpu));
+                }
+            }
+            break;
+
+        case CCON:
+            mx_cpu->ccon = v & 0x1;
+            break;
+
+        default:
+            qemu_log_mask(LOG_GUEST_ERROR,
+                          "unknown WER in MX PIC range: 0x%08x = 0x%08x\n",
+                          (uint32_t)offset, (uint32_t)v);
+            break;
+        }
+        if (change) {
+            qemu_log_mask(CPU_LOG_INT,
+                          "%s: %s changed by CPU %d: %08x -> %08x\n",
+                          __func__, name, (int)(mx_cpu - mx->cpu),
+                          oldv, newv);
+            xtensa_mx_pic_update_all(mx);
+        }
+    }
+}
+
+static const MemoryRegionOps xtensa_mx_pic_ops = {
+    .read = xtensa_mx_pic_ext_reg_read,
+    .write = xtensa_mx_pic_ext_reg_write,
+    .endianness = DEVICE_NATIVE_ENDIAN,
+    .valid = {
+        .unaligned = true,
+    },
+};
+
+MemoryRegion *xtensa_mx_pic_register_cpu(XtensaMxPic *mx,
+                                         qemu_irq *irq,
+                                         qemu_irq runstall)
+{
+    struct XtensaMxPicCpu *mx_cpu = mx->cpu + mx->n_cpu;
+
+    mx_cpu->mx = mx;
+    mx_cpu->irq = irq;
+    mx_cpu->runstall = runstall;
+
+    memory_region_init_io(&mx_cpu->reg, NULL, &xtensa_mx_pic_ops, mx_cpu,
+                          "mx_pic", 0x280);
+
+    ++mx->n_cpu;
+    return &mx_cpu->reg;
+}
+
+static void xtensa_mx_pic_set_irq(void *opaque, int irq, int active)
+{
+    XtensaMxPic *mx = opaque;
+
+    if (irq < mx->n_irq) {
+        uint32_t old_irq_state = mx->ext_irq_state;
+
+        if (active) {
+            mx->ext_irq_state |= 1u << irq;
+        } else {
+            mx->ext_irq_state &= ~(1u << irq);
+        }
+        if (old_irq_state != mx->ext_irq_state) {
+            qemu_log_mask(CPU_LOG_INT,
+                          "%s: IRQ %d, active: %d, ext_irq_state: %08x -> %08x\n",
+                          __func__, irq, active,
+                          old_irq_state, mx->ext_irq_state);
+            xtensa_mx_pic_update_all(mx);
+        }
+    } else {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: IRQ %d out of range\n",
+                      __func__, irq);
+    }
+}
+
+XtensaMxPic *xtensa_mx_pic_init(unsigned n_irq)
+{
+    XtensaMxPic *mx = calloc(1, sizeof(XtensaMxPic));
+
+    mx->n_irq = n_irq + 1;
+    mx->irq_inputs = qemu_allocate_irqs(xtensa_mx_pic_set_irq, mx,
+                                        mx->n_irq);
+    return mx;
+}
+
+void xtensa_mx_pic_reset(void *opaque)
+{
+    XtensaMxPic *mx = opaque;
+    unsigned i;
+
+    mx->ext_irq_state = 0;
+    mx->mieng = mx->n_irq < 32 ? (1u << mx->n_irq) - 1 : ~0u;
+    mx->miasg = 0;
+    mx->mipipart = 0;
+    for (i = 0; i < mx->n_irq; ++i) {
+        mx->mirout[i] = 1;
+    }
+    for (i = 0; i < mx->n_cpu; ++i) {
+        mx->cpu[i].mipicause = 0;
+        mx->cpu[i].mirout_cache = i ? 0 : mx->mieng;
+        mx->cpu[i].irq_state_cache = 0;
+        mx->cpu[i].ccon = 0;
+    }
+    mx->runstall = (1u << mx->n_cpu) - 2;
+    for (i = 0; i < mx->n_cpu; ++i) {
+        qemu_set_irq(mx->cpu[i].runstall, i > 0);
+    }
+}
+
+qemu_irq *xtensa_mx_pic_get_extints(XtensaMxPic *mx)
+{
+    return mx->irq_inputs + 1;
+}