spapr/xive: add KVM support

This introduces a set of helpers when KVM is in use, which create the
KVM XIVE device, initialize the interrupt sources at a KVM level and
connect the interrupt presenters to the vCPU.

They also handle the initialization of the TIMA and the source ESB
memory regions of the controller. These have a different type under
KVM. They are 'ram device' memory mappings, similarly to VFIO, exposed
to the guest and the associated VMAs on the host are populated
dynamically with the appropriate pages using a fault handler.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Message-Id: <20190513084245.25755-3-clg@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

6 files changed, 308 insertions, 10 deletions

diff --git a/hw/intc/Makefile.objs b/hw/intc/Makefile.objs
index df712c3e6c..03019b9a03 100644
--- a/hw/intc/Makefile.objs
+++ b/hw/intc/Makefile.objs
@@ -39,6 +39,7 @@ obj-$(CONFIG_XICS_SPAPR) += xics_spapr.o
 obj-$(CONFIG_XICS_KVM) += xics_kvm.o
 obj-$(CONFIG_XIVE) += xive.o
 obj-$(CONFIG_XIVE_SPAPR) += spapr_xive.o
+obj-$(CONFIG_XIVE_KVM) += spapr_xive_kvm.o
 obj-$(CONFIG_POWERNV) += xics_pnv.o pnv_xive.o
 obj-$(CONFIG_ALLWINNER_A10_PIC) += allwinner-a10-pic.o
 obj-$(CONFIG_S390_FLIC) += s390_flic.o
diff --git a/hw/intc/spapr_xive.c b/hw/intc/spapr_xive.c
index 62e13ac353..27632683e6 100644
--- a/hw/intc/spapr_xive.c
+++ b/hw/intc/spapr_xive.c
@@ -174,7 +174,7 @@ void spapr_xive_pic_print_info(SpaprXive *xive, Monitor *mon)
     }
 }
 
-static void spapr_xive_map_mmio(SpaprXive *xive)
+void spapr_xive_map_mmio(SpaprXive *xive)
 {
     sysbus_mmio_map(SYS_BUS_DEVICE(xive), 0, xive->vc_base);
     sysbus_mmio_map(SYS_BUS_DEVICE(xive), 1, xive->end_base);
@@ -251,6 +251,9 @@ static void spapr_xive_instance_init(Object *obj)
     object_initialize_child(obj, "end_source", &xive->end_source,
                             sizeof(xive->end_source), TYPE_XIVE_END_SOURCE,
                             &error_abort, NULL);
+
+    /* Not connected to the KVM XIVE device */
+    xive->fd = -1;
 }
 
 static void spapr_xive_realize(DeviceState *dev, Error **errp)
@@ -259,6 +262,7 @@ static void spapr_xive_realize(DeviceState *dev, Error **errp)
     XiveSource *xsrc = &xive->source;
     XiveENDSource *end_xsrc = &xive->end_source;
     Error *local_err = NULL;
+    MachineState *machine = MACHINE(qdev_get_machine());
 
     if (!xive->nr_irqs) {
         error_setg(errp, "Number of interrupt needs to be greater 0");
@@ -305,6 +309,32 @@ static void spapr_xive_realize(DeviceState *dev, Error **errp)
     xive->eat = g_new0(XiveEAS, xive->nr_irqs);
     xive->endt = g_new0(XiveEND, xive->nr_ends);
 
+    xive->nodename = g_strdup_printf("interrupt-controller@%" PRIx64,
+                           xive->tm_base + XIVE_TM_USER_PAGE * (1 << TM_SHIFT));
+
+    qemu_register_reset(spapr_xive_reset, dev);
+
+    if (kvm_enabled() && machine_kernel_irqchip_allowed(machine)) {
+        kvmppc_xive_connect(xive, &local_err);
+        if (local_err && machine_kernel_irqchip_required(machine)) {
+            error_prepend(&local_err,
+                          "kernel_irqchip requested but unavailable: ");
+            error_propagate(errp, local_err);
+            return;
+        }
+
+        if (!local_err) {
+            return;
+        }
+
+        /*
+         * We failed to initialize the XIVE KVM device, fallback to
+         * emulated mode
+         */
+        error_prepend(&local_err, "kernel_irqchip allowed but unavailable: ");
+        warn_report_err(local_err);
+    }
+
     /* TIMA initialization */
     memory_region_init_io(&xive->tm_mmio, OBJECT(xive), &xive_tm_ops, xive,
                           "xive.tima", 4ull << TM_SHIFT);
@@ -316,11 +346,6 @@ static void spapr_xive_realize(DeviceState *dev, Error **errp)
 
     /* Map all regions */
     spapr_xive_map_mmio(xive);
-
-    xive->nodename = g_strdup_printf("interrupt-controller@%" PRIx64,
-                           xive->tm_base + XIVE_TM_USER_PAGE * (1 << TM_SHIFT));
-
-    qemu_register_reset(spapr_xive_reset, dev);
 }
 
 static int spapr_xive_get_eas(XiveRouter *xrtr, uint8_t eas_blk,
@@ -495,6 +520,17 @@ bool spapr_xive_irq_claim(SpaprXive *xive, uint32_t lisn, bool lsi)
     if (lsi) {
         xive_source_irq_set_lsi(xsrc, lisn);
     }
+
+    if (kvm_irqchip_in_kernel()) {
+        Error *local_err = NULL;
+
+        kvmppc_xive_source_reset_one(xsrc, lisn, &local_err);
+        if (local_err) {
+            error_report_err(local_err);
+            return false;
+        }
+    }
+
     return true;
 }
 
diff --git a/hw/intc/spapr_xive_kvm.c b/hw/intc/spapr_xive_kvm.c
new file mode 100644
index 0000000000..7d9e771e8a
--- /dev/null
+++ b/hw/intc/spapr_xive_kvm.c
@@ -0,0 +1,237 @@
+/*
+ * QEMU PowerPC sPAPR XIVE interrupt controller model
+ *
+ * Copyright (c) 2017-2019, IBM Corporation.
+ *
+ * This code is licensed under the GPL version 2 or later. See the
+ * COPYING file in the top-level directory.
+ */
+
+#include "qemu/osdep.h"
+#include "qemu/log.h"
+#include "qemu/error-report.h"
+#include "qapi/error.h"
+#include "target/ppc/cpu.h"
+#include "sysemu/cpus.h"
+#include "sysemu/kvm.h"
+#include "hw/ppc/spapr.h"
+#include "hw/ppc/spapr_xive.h"
+#include "hw/ppc/xive.h"
+#include "kvm_ppc.h"
+
+#include <sys/ioctl.h>
+
+/*
+ * Helpers for CPU hotplug
+ *
+ * TODO: make a common KVMEnabledCPU layer for XICS and XIVE
+ */
+typedef struct KVMEnabledCPU {
+    unsigned long vcpu_id;
+    QLIST_ENTRY(KVMEnabledCPU) node;
+} KVMEnabledCPU;
+
+static QLIST_HEAD(, KVMEnabledCPU)
+    kvm_enabled_cpus = QLIST_HEAD_INITIALIZER(&kvm_enabled_cpus);
+
+static bool kvm_cpu_is_enabled(CPUState *cs)
+{
+    KVMEnabledCPU *enabled_cpu;
+    unsigned long vcpu_id = kvm_arch_vcpu_id(cs);
+
+    QLIST_FOREACH(enabled_cpu, &kvm_enabled_cpus, node) {
+        if (enabled_cpu->vcpu_id == vcpu_id) {
+            return true;
+        }
+    }
+    return false;
+}
+
+static void kvm_cpu_enable(CPUState *cs)
+{
+    KVMEnabledCPU *enabled_cpu;
+    unsigned long vcpu_id = kvm_arch_vcpu_id(cs);
+
+    enabled_cpu = g_malloc(sizeof(*enabled_cpu));
+    enabled_cpu->vcpu_id = vcpu_id;
+    QLIST_INSERT_HEAD(&kvm_enabled_cpus, enabled_cpu, node);
+}
+
+/*
+ * XIVE Thread Interrupt Management context (KVM)
+ */
+
+void kvmppc_xive_cpu_connect(XiveTCTX *tctx, Error **errp)
+{
+    SpaprXive *xive = SPAPR_MACHINE(qdev_get_machine())->xive;
+    unsigned long vcpu_id;
+    int ret;
+
+    /* Check if CPU was hot unplugged and replugged. */
+    if (kvm_cpu_is_enabled(tctx->cs)) {
+        return;
+    }
+
+    vcpu_id = kvm_arch_vcpu_id(tctx->cs);
+
+    ret = kvm_vcpu_enable_cap(tctx->cs, KVM_CAP_PPC_IRQ_XIVE, 0, xive->fd,
+                              vcpu_id, 0);
+    if (ret < 0) {
+        error_setg(errp, "XIVE: unable to connect CPU%ld to KVM device: %s",
+                   vcpu_id, strerror(errno));
+        return;
+    }
+
+    kvm_cpu_enable(tctx->cs);
+}
+
+/*
+ * XIVE Interrupt Source (KVM)
+ */
+
+/*
+ * At reset, the interrupt sources are simply created and MASKED. We
+ * only need to inform the KVM XIVE device about their type: LSI or
+ * MSI.
+ */
+void kvmppc_xive_source_reset_one(XiveSource *xsrc, int srcno, Error **errp)
+{
+    SpaprXive *xive = SPAPR_XIVE(xsrc->xive);
+    uint64_t state = 0;
+
+    if (xive_source_irq_is_lsi(xsrc, srcno)) {
+        state |= KVM_XIVE_LEVEL_SENSITIVE;
+        if (xsrc->status[srcno] & XIVE_STATUS_ASSERTED) {
+            state |= KVM_XIVE_LEVEL_ASSERTED;
+        }
+    }
+
+    kvm_device_access(xive->fd, KVM_DEV_XIVE_GRP_SOURCE, srcno, &state,
+                      true, errp);
+}
+
+void kvmppc_xive_source_reset(XiveSource *xsrc, Error **errp)
+{
+    int i;
+
+    for (i = 0; i < xsrc->nr_irqs; i++) {
+        Error *local_err = NULL;
+
+        kvmppc_xive_source_reset_one(xsrc, i, &local_err);
+        if (local_err) {
+            error_propagate(errp, local_err);
+            return;
+        }
+    }
+}
+
+void kvmppc_xive_source_set_irq(void *opaque, int srcno, int val)
+{
+    XiveSource *xsrc = opaque;
+    struct kvm_irq_level args;
+    int rc;
+
+    args.irq = srcno;
+    if (!xive_source_irq_is_lsi(xsrc, srcno)) {
+        if (!val) {
+            return;
+        }
+        args.level = KVM_INTERRUPT_SET;
+    } else {
+        if (val) {
+            xsrc->status[srcno] |= XIVE_STATUS_ASSERTED;
+            args.level = KVM_INTERRUPT_SET_LEVEL;
+        } else {
+            xsrc->status[srcno] &= ~XIVE_STATUS_ASSERTED;
+            args.level = KVM_INTERRUPT_UNSET;
+        }
+    }
+    rc = kvm_vm_ioctl(kvm_state, KVM_IRQ_LINE, &args);
+    if (rc < 0) {
+        error_report("XIVE: kvm_irq_line() failed : %s", strerror(errno));
+    }
+}
+
+/*
+ * sPAPR XIVE interrupt controller (KVM)
+ */
+
+static void *kvmppc_xive_mmap(SpaprXive *xive, int pgoff, size_t len,
+                              Error **errp)
+{
+    void *addr;
+    uint32_t page_shift = 16; /* TODO: fix page_shift */
+
+    addr = mmap(NULL, len, PROT_WRITE | PROT_READ, MAP_SHARED, xive->fd,
+                pgoff << page_shift);
+    if (addr == MAP_FAILED) {
+        error_setg_errno(errp, errno, "XIVE: unable to set memory mapping");
+        return NULL;
+    }
+
+    return addr;
+}
+
+/*
+ * All the XIVE memory regions are now backed by mappings from the KVM
+ * XIVE device.
+ */
+void kvmppc_xive_connect(SpaprXive *xive, Error **errp)
+{
+    XiveSource *xsrc = &xive->source;
+    XiveENDSource *end_xsrc = &xive->end_source;
+    Error *local_err = NULL;
+    size_t esb_len = (1ull << xsrc->esb_shift) * xsrc->nr_irqs;
+    size_t tima_len = 4ull << TM_SHIFT;
+
+    if (!kvmppc_has_cap_xive()) {
+        error_setg(errp, "IRQ_XIVE capability must be present for KVM");
+        return;
+    }
+
+    /* First, create the KVM XIVE device */
+    xive->fd = kvm_create_device(kvm_state, KVM_DEV_TYPE_XIVE, false);
+    if (xive->fd < 0) {
+        error_setg_errno(errp, -xive->fd, "XIVE: error creating KVM device");
+        return;
+    }
+
+    /*
+     * 1. Source ESB pages - KVM mapping
+     */
+    xsrc->esb_mmap = kvmppc_xive_mmap(xive, KVM_XIVE_ESB_PAGE_OFFSET, esb_len,
+                                      &local_err);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        return;
+    }
+
+    memory_region_init_ram_device_ptr(&xsrc->esb_mmio, OBJECT(xsrc),
+                                      "xive.esb", esb_len, xsrc->esb_mmap);
+    sysbus_init_mmio(SYS_BUS_DEVICE(xive), &xsrc->esb_mmio);
+
+    /*
+     * 2. END ESB pages (No KVM support yet)
+     */
+    sysbus_init_mmio(SYS_BUS_DEVICE(xive), &end_xsrc->esb_mmio);
+
+    /*
+     * 3. TIMA pages - KVM mapping
+     */
+    xive->tm_mmap = kvmppc_xive_mmap(xive, KVM_XIVE_TIMA_PAGE_OFFSET, tima_len,
+                                     &local_err);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        return;
+    }
+    memory_region_init_ram_device_ptr(&xive->tm_mmio, OBJECT(xive),
+                                      "xive.tima", tima_len, xive->tm_mmap);
+    sysbus_init_mmio(SYS_BUS_DEVICE(xive), &xive->tm_mmio);
+
+    kvm_kernel_irqchip = true;
+    kvm_msi_via_irqfd_allowed = true;
+    kvm_gsi_direct_mapping = true;
+
+    /* Map all regions */
+    spapr_xive_map_mmio(xive);
+}
diff --git a/hw/intc/xive.c b/hw/intc/xive.c
index dcf2fcd108..78047adb11 100644
--- a/hw/intc/xive.c
+++ b/hw/intc/xive.c
@@ -555,6 +555,15 @@ static void xive_tctx_realize(DeviceState *dev, Error **errp)
         return;
     }
 
+    /* Connect the presenter to the VCPU (required for CPU hotplug) */
+    if (kvm_irqchip_in_kernel()) {
+        kvmppc_xive_cpu_connect(tctx, &local_err);
+        if (local_err) {
+            error_propagate(errp, local_err);
+            return;
+        }
+    }
+
     qemu_register_reset(xive_tctx_reset, dev);
 }
 
@@ -957,6 +966,10 @@ static void xive_source_reset(void *dev)
 
     /* PQs are initialized to 0b01 (Q=1) which corresponds to "ints off" */
     memset(xsrc->status, XIVE_ESB_OFF, xsrc->nr_irqs);
+
+    if (kvm_irqchip_in_kernel()) {
+        kvmppc_xive_source_reset(xsrc, &error_fatal);
+    }
 }
 
 static void xive_source_realize(DeviceState *dev, Error **errp)
@@ -990,9 +1003,11 @@ static void xive_source_realize(DeviceState *dev, Error **errp)
     xsrc->status = g_malloc0(xsrc->nr_irqs);
     xsrc->lsi_map = bitmap_new(xsrc->nr_irqs);
 
-    memory_region_init_io(&xsrc->esb_mmio, OBJECT(xsrc),
-                          &xive_source_esb_ops, xsrc, "xive.esb",
-                          (1ull << xsrc->esb_shift) * xsrc->nr_irqs);
+    if (!kvm_irqchip_in_kernel()) {
+        memory_region_init_io(&xsrc->esb_mmio, OBJECT(xsrc),
+                              &xive_source_esb_ops, xsrc, "xive.esb",
+                              (1ull << xsrc->esb_shift) * xsrc->nr_irqs);
+    }
 
     qemu_register_reset(xive_source_reset, dev);
 }
diff --git a/hw/ppc/Kconfig b/hw/ppc/Kconfig
index a3465155f0..f927ec9c74 100644
--- a/hw/ppc/Kconfig
+++ b/hw/ppc/Kconfig
@@ -122,3 +122,8 @@ config XIVE_SPAPR
     default y
     depends on PSERIES
     select XIVE
+
+config XIVE_KVM
+    bool
+    default y
+    depends on XIVE_SPAPR && KVM
diff --git a/hw/ppc/spapr_irq.c b/hw/ppc/spapr_irq.c
index b1f79ea9de..5c4a44855d 100644
--- a/hw/ppc/spapr_irq.c
+++ b/hw/ppc/spapr_irq.c
@@ -372,7 +372,11 @@ static void spapr_irq_set_irq_xive(void *opaque, int srcno, int val)
 {
     SpaprMachineState *spapr = opaque;
 
-    xive_source_set_irq(&spapr->xive->source, srcno, val);
+    if (kvm_irqchip_in_kernel()) {
+        kvmppc_xive_source_set_irq(&spapr->xive->source, srcno, val);
+    } else {
+        xive_source_set_irq(&spapr->xive->source, srcno, val);
+    }
 }
 
 static const char *spapr_irq_get_nodename_xive(SpaprMachineState *spapr)