[Qemu-devel] [v4 4/6] hw/iommu: AMD IOMMU interrupt remapping

[David Kiarie]

Introduce AMD IOMMU interrupt remapping and hook it onto
the existing interrupt remapping infrastructure

Signed-off-by: David Kiarie <address@hidden>
---
 hw/i386/amd_iommu.c | 240 +++++++++++++++++++++++++++++++++++++++++++++++++++-
 hw/i386/amd_iommu.h |   4 +-
 hw/intc/ioapic.c    |   1 -
 3 files changed, 242 insertions(+), 3 deletions(-)

diff --git a/hw/i386/amd_iommu.c b/hw/i386/amd_iommu.c
index 226fea5..76d3816 100644
--- a/hw/i386/amd_iommu.c
+++ b/hw/i386/amd_iommu.c
@@ -20,6 +20,7 @@
  * Cache implementation inspired by hw/i386/intel_iommu.c
  */
 #include "qemu/osdep.h"
+#include "qemu/error-report.h"
 #include "hw/i386/amd_iommu.h"
 #include "trace.h"
 
@@ -255,6 +256,31 @@ typedef struct QEMU_PACKED {
     uint32_t reserved_5:16;
 } CMDCompletePPR;
 
+typedef union IRTE {
+    struct {
+#ifdef HOST_WORDS_BIGENDIAN
+        uint32_t destination:8;
+        uint32_t rsvd_1:1;
+        uint32_t dm:1;
+        uint32_t rq_eoi:1;
+        uint32_t int_type:3;
+        uint32_t no_fault:1;
+        uint32_t valid:1;
+#else
+        uint32_t valid:1;
+        uint32_t no_fault:1;
+        uint32_t int_type:3;
+        uint32_t rq_eoi:1;
+        uint32_t dm:1;
+        uint32_t rsvd_1:1;
+        uint32_t destination:8;
+#endif
+        uint32_t vector:8;
+        uint32_t rsvd_2:8;
+    } bits;
+    uint32_t data;
+} IRTE;
+
 /* configure MMIO registers at startup/reset */
 static void amdvi_set_quad(AMDVIState *s, hwaddr addr, uint64_t val,
                            uint64_t romask, uint64_t w1cmask)
@@ -641,6 +667,11 @@ static void amdvi_inval_inttable(AMDVIState *s, 
CMDInvalIntrTable *inval)
         amdvi_log_illegalcom_error(s, inval->type, s->cmdbuf + s->cmdbuf_head);
         return;
     }
+
+    if (s->ir_cache) {
+        x86_iommu_iec_notify_all(X86_IOMMU_DEVICE(s), true, 0, 0);
+    }
+
     trace_amdvi_intr_inval();
 }
 
@@ -1203,6 +1234,197 @@ static IOMMUTLBEntry amdvi_translate(MemoryRegion 
*iommu, hwaddr addr,
     return ret;
 }
 
+static inline int amdvi_ir_handle_non_vectored(MSIMessage *src,
+                                               MSIMessage *dst, uint8_t bitpos,
+                                               uint64_t dte)
+{
+    if ((dte & (1UL << bitpos))) {
+        /* passing interrupt enabled */
+        memcpy(dst, src, sizeof(*dst));
+    } else {
+        /* should be target aborted */
+        return -AMDVI_TARGET_ABORT;
+    }
+    return 0;
+}
+
+static int amdvi_remap_ir_intctl(uint64_t dte, IRTE irte,
+                                 MSIMessage *src, MSIMessage *dst)
+{
+    int ret = 0;
+
+    switch ((dte >> AMDVI_DTE_INTCTL_RSHIFT) & 3UL) {
+    case AMDVI_INTCTL_PASS:
+        /* pass */
+        memcpy(dst, src, sizeof(*dst));
+        break;
+    case AMDVI_INTCTL_REMAP:
+        /* remap */
+        if (irte.bits.valid) {
+            /* LOCAL APIC address */
+            dst->address = AMDVI_LOCAL_APIC_ADDR;
+            /* destination mode */
+            dst->address |= ((uint64_t)irte.bits.dm) <<
+                            AMDVI_MSI_ADDR_DM_RSHIFT;
+            /* RH */
+            dst->address |= ((uint64_t)irte.bits.rq_eoi) <<
+                            AMDVI_MSI_ADDR_RH_RSHIFT;
+            /* Destination ID */
+            dst->address |= ((uint64_t)irte.bits.destination) <<
+                            AMDVI_MSI_ADDR_DEST_RSHIFT;
+            /* construct data - vector */
+            dst->data |= irte.bits.vector;
+            /* Interrupt type */
+            dst->data |= ((uint64_t)irte.bits.int_type) <<
+                         AMDVI_MSI_DATA_DM_RSHIFT;
+        } else  {
+            ret = -AMDVI_TARGET_ABORT;
+        }
+        break;
+    case AMDVI_INTCTL_ABORT:
+    case AMDVI_INTCTL_RSVD:
+        ret = -AMDVI_TARGET_ABORT;
+    }
+    return ret;
+}
+
+static int amdvi_irte_get(AMDVIState *s, MSIMessage *src, IRTE *irte,
+                          uint64_t *dte, uint16_t devid)
+{
+    uint64_t irte_root, offset = devid * AMDVI_DEVTAB_ENTRY_SIZE,
+             ir_table_size;
+
+    irte_root = dte[2] & AMDVI_IRTEROOT_MASK;
+    offset = (src->data & AMDVI_IRTE_INDEX_MASK) << 2;
+    ir_table_size = 1UL << (dte[2] & AMDVI_IR_TABLE_SIZE_MASK);
+    /* enforce IR table size */
+    if (offset > (ir_table_size * AMDVI_DEFAULT_IRTE_SIZE)) {
+        trace_amdvi_invalid_irte_entry(offset, ir_table_size);
+        return -AMDVI_TARGET_ABORT;
+    }
+    /* read IRTE */
+    if (dma_memory_read(&address_space_memory, irte_root + offset,
+        irte, sizeof(*irte))) {
+        trace_amdvi_irte_get_fail(irte_root, offset);
+        return -AMDVI_DEV_TAB_HW;
+    }
+    return 0;
+}
+
+static int amdvi_int_remap(X86IOMMUState *iommu, MSIMessage *src,
+                           MSIMessage *dst, uint16_t sid)
+{
+    trace_amdvi_ir_request(src->data, src->address, sid);
+
+    AMDVIState *s = AMD_IOMMU_DEVICE(iommu);
+    int ret = 0;
+    uint64_t dte[4];
+    uint32_t bitpos;
+    IRTE irte;
+
+    amdvi_get_dte(s, sid, dte);
+
+    /* interrupt remapping disabled */
+    if (!(dte[2] & AMDVI_IR_VALID)) {
+        memcpy(dst, src, sizeof(*src));
+        return ret;
+    }
+
+    ret = amdvi_irte_get(s, src, &irte, dte, sid);
+    if (ret < 0) {
+        goto no_remap;
+    }
+    switch (src->data & AMDVI_IR_TYPE_MASK) {
+    case AMDVI_MT_FIXED:
+    case AMDVI_MT_ARBIT:
+        ret = amdvi_remap_ir_intctl(dte[2], irte, src, dst);
+        if (ret < 0) {
+            goto no_remap;
+        } else {
+            s->ir_cache = true;
+            trace_amdvi_ir_remap(dst->data, dst->address, sid);
+            return ret;
+        }
+    /* not handling SMI currently */
+    case AMDVI_MT_SMI:
+        error_report("SMI interrupts not currently handled");
+        goto no_remap;
+    case AMDVI_MT_NMI:
+        bitpos = AMDVI_DTE_NMIPASS_LSHIFT;
+        break;
+    case AMDVI_MT_INIT:
+        bitpos = AMDVI_DTE_INTPASS_LSHIFT;
+        break;
+    case AMDVI_MT_EXTINT:
+        bitpos = AMDVI_DTE_EINTPASS_LSHIFT;
+        break;
+    case AMDVI_MT_LINT1:
+        bitpos = AMDVI_DTE_LINT1PASS_LSHIFT;
+        break;
+    case AMDVI_MT_LINT0:
+        bitpos = AMDVI_DTE_LINT0PASS_LSHIFT;
+    default:
+        goto no_remap;
+    }
+
+    ret = amdvi_ir_handle_non_vectored(src, dst, bitpos, dte[2]);
+    if (ret < 0){
+        goto no_remap;
+    }
+    s->ir_cache = true;
+    trace_amdvi_ir_remap(dst->data, dst->address, sid);
+    return ret;
+no_remap:
+    memcpy(dst, src, sizeof(*src));
+    trace_amdvi_ir_target_abort(dst->data, dst->address, sid);
+    return ret;
+}
+
+static MemTxResult amdvi_ir_read(void *opaque, hwaddr addr,
+                                 uint64_t *data, unsigned size,
+                                 MemTxAttrs attrs)
+{
+    return MEMTX_OK;
+}
+
+static MemTxResult amdvi_ir_write(void *opaque, hwaddr addr, uint64_t val,
+                                  unsigned size, MemTxAttrs attrs)
+{
+    AMDVIAddressSpace *as = opaque;
+    MSIMessage from = { addr + AMDVI_INT_ADDR_FIRST, val }, to = { 0, 0};
+    int ret = 0;
+
+    ret  = amdvi_int_remap(X86_IOMMU_DEVICE(as->iommu_state), &from, &to,
+                           attrs.requester_id);
+
+    if (ret < 0) {
+        trace_amdvi_ir_target_abort(from.data, from.address,
+                                    attrs.requester_id);
+        return MEMTX_ERROR;
+    }
+
+    if(dma_memory_write(&address_space_memory, to.address, &to.data, size)) {
+        trace_amdvi_ir_write_fail(to.address, to.data);
+        return MEMTX_ERROR;
+    }
+
+    return MEMTX_OK;
+}
+
+static const MemoryRegionOps amdvi_ir_ops = {
+    .read_with_attrs = amdvi_ir_read,
+    .write_with_attrs = amdvi_ir_write,
+    .endianness = DEVICE_LITTLE_ENDIAN,
+    .impl = {
+        .min_access_size = 4,
+        .max_access_size = 4,
+    },
+    .valid = {
+        .min_access_size = 4,
+        .max_access_size = 4,
+    }
+};
+
 static AddressSpace *amdvi_host_dma_iommu(PCIBus *bus, void *opaque, int devfn)
 {
     AMDVIState *s = opaque;
@@ -1226,6 +1448,12 @@ static AddressSpace *amdvi_host_dma_iommu(PCIBus *bus, 
void *opaque, int devfn)
 
         memory_region_init_iommu(&iommu_as[devfn]->iommu, OBJECT(s),
                                  &s->iommu_ops, "amd-iommu", UINT64_MAX);
+        memory_region_init_io(&iommu_as[devfn]->iommu_ir, OBJECT(s),
+                              &amdvi_ir_ops, iommu_as[devfn], "amd-iommu-ir",
+                              AMDVI_INT_ADDR_SIZE);
+        memory_region_add_subregion(&iommu_as[devfn]->iommu,
+                                    AMDVI_INT_ADDR_FIRST,
+                                    &iommu_as[devfn]->iommu_ir);
         address_space_init(&iommu_as[devfn]->as, &iommu_as[devfn]->iommu,
                            "amd-iommu");
     }
@@ -1274,6 +1502,7 @@ static void amdvi_init(AMDVIState *s)
     s->enabled = false;
     s->ats_enabled = false;
     s->cmdbuf_enabled = false;
+    s->ir_cache = false;
 
     /* reset MMIO */
     memset(s->mmior, 0, AMDVI_MMIO_SIZE);
@@ -1313,11 +1542,15 @@ static void amdvi_realize(DeviceState *dev, Error **err)
     AMDVIState *s = AMD_IOMMU_DEVICE(dev);
     X86IOMMUState *x86_iommu = X86_IOMMU_DEVICE(dev);
     PCIBus *bus = PC_MACHINE(qdev_get_machine())->bus;
+    PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
     s->iotlb = g_hash_table_new_full(amdvi_uint64_hash,
                                      amdvi_uint64_equal, g_free, g_free);
 
-    /* This device should take care of IOMMU PCI properties */
+    /* AMD IOMMU has Interrupt Remapping on by default */
+    x86_iommu->intr_supported = true;
     x86_iommu->type = TYPE_AMD;
+
+    /* This device should take care of IOMMU PCI properties */
     qdev_set_parent_bus(DEVICE(&s->pci), &bus->qbus);
     object_property_set_bool(OBJECT(&s->pci), true, "realized", err);
     s->capab_offset = pci_add_capability(&s->pci.dev, AMDVI_CAPAB_ID_SEC, 0,
@@ -1329,9 +1562,13 @@ static void amdvi_realize(DeviceState *dev, Error **err)
     memory_region_init_io(&s->mmio, OBJECT(s), &mmio_mem_ops, s, "amdvi-mmio",
                           AMDVI_MMIO_SIZE);
 
+    x86_iommu->ioapic_bdf = PCI_BUILD_BDF(AMDVI_BUS_NUM,
+             AMDVI_SB_IOAPIC_ID);
+
     sysbus_init_mmio(SYS_BUS_DEVICE(s), &s->mmio);
     sysbus_mmio_map(SYS_BUS_DEVICE(s), 0, AMDVI_BASE_ADDR);
     pci_setup_iommu(bus, amdvi_host_dma_iommu, s);
+    pcms->ioapic_as = amdvi_host_dma_iommu(bus, s, AMDVI_SB_IOAPIC_ID);
     s->devid = object_property_get_int(OBJECT(&s->pci), "addr", err);
     msi_init(&s->pci.dev, 0, 1, true, false, err);
     amdvi_init(s);
@@ -1358,6 +1595,7 @@ static void amdvi_class_init(ObjectClass *klass, void* 
data)
     dc->vmsd = &vmstate_amdvi;
     dc->hotpluggable = false;
     dc_class->realize = amdvi_realize;
+    dc_class->int_remap = amdvi_int_remap;
 }
 
 static const TypeInfo amdvi = {
diff --git a/hw/i386/amd_iommu.h b/hw/i386/amd_iommu.h
index 5c4a13b..28b9603 100644
--- a/hw/i386/amd_iommu.h
+++ b/hw/i386/amd_iommu.h
@@ -235,7 +235,7 @@
 
 #define AMDVI_BUS_NUM                  0x0
 /* AMD-Vi specific IOAPIC Device function */
-#define AMDVI_DEVFN_IOAPIC             0xa0
+#define AMDVI_SB_IOAPIC_ID            0xa0
 
 #define AMDVI_LOCAL_APIC_ADDR     0xfee00000
 
@@ -338,6 +338,8 @@ typedef struct AMDVIState {
     uint32_t evtlog_len;         /* event log length             */
     uint32_t evtlog_head;        /* current IOMMU write position */
     uint32_t evtlog_tail;        /* current Software read position */
+    /* whether we have remapped any interrupts and hence IR cache */
+    bool ir_cache;
 
     /* unused for now */
     hwaddr excl_base;            /* base DVA - IOMMU exclusion range */
diff --git a/hw/intc/ioapic.c b/hw/intc/ioapic.c
index f2d4c15..3fefe4a 100644
--- a/hw/intc/ioapic.c
+++ b/hw/intc/ioapic.c
@@ -412,7 +412,6 @@ static void ioapic_machine_done_notify(Notifier *notifier, 
void *data)
             }
             kvm_irqchip_commit_routes(kvm_state);
         }
-
     }
 #endif
 }
-- 
2.1.4