s390x/virtio-ccw: wire up irq routing and irqfds

Make use of the new s390 adapter irq routing support to enable real
in-kernel irqfds for virtio-ccw with adapter interrupts.

Note that s390 doesn't provide the common KVM_CAP_IRQCHIP capability, but
rather needs KVM_CAP_S390_IRQCHIP to be enabled. This is to ensure backward
compatibility.

Reviewed-by: Thomas Huth <thuth@linux.vnet.ibm.com>
Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>

hw/intc/s390_flic.c

@@ -52,11 +52,32 @@ static int qemu_s390_register_io_adapter(S390FLICState *fs, uint32_t id,
     return 0;
 }
 
+static int qemu_s390_io_adapter_map(S390FLICState *fs, uint32_t id,
+                                    uint64_t map_addr, bool do_map)
+{
+    /* nothing to do */
+    return 0;
+}
+
+static int qemu_s390_add_adapter_routes(S390FLICState *fs,
+                                        AdapterRoutes *routes)
+{
+    return -ENOSYS;
+}
+
+static void qemu_s390_release_adapter_routes(S390FLICState *fs,
+                                             AdapterRoutes *routes)
+{
+}
+
 static void qemu_s390_flic_class_init(ObjectClass *oc, void *data)
 {
     S390FLICStateClass *fsc = S390_FLIC_COMMON_CLASS(oc);
 
     fsc->register_io_adapter = qemu_s390_register_io_adapter;
+    fsc->io_adapter_map = qemu_s390_io_adapter_map;
+    fsc->add_adapter_routes = qemu_s390_add_adapter_routes;
+    fsc->release_adapter_routes = qemu_s390_release_adapter_routes;
 }
 
 static const TypeInfo qemu_s390_flic_info = {

hw/intc/s390_flic_kvm.c

@@ -16,6 +16,7 @@
 #include "sysemu/kvm.h"
 #include "migration/qemu-file.h"
 #include "hw/s390x/s390_flic.h"
+#include "hw/s390x/adapter.h"
 #include "trace.h"
 
 #define FLIC_SAVE_INITIAL_SIZE getpagesize()
@@ -178,6 +179,68 @@ static int kvm_s390_register_io_adapter(S390FLICState *fs, uint32_t id,
     return ret;
 }
 
+static int kvm_s390_io_adapter_map(S390FLICState *fs, uint32_t id,
+                                   uint64_t map_addr, bool do_map)
+{
+    struct kvm_s390_io_adapter_req req = {
+        .id = id,
+        .type = do_map ? KVM_S390_IO_ADAPTER_MAP : KVM_S390_IO_ADAPTER_UNMAP,
+        .addr = map_addr,
+    };
+    struct kvm_device_attr attr = {
+        .group = KVM_DEV_FLIC_ADAPTER_MODIFY,
+        .addr = (uint64_t)&req,
+    };
+    KVMS390FLICState *flic = KVM_S390_FLIC(fs);
+    int r;
+
+    if (!kvm_check_extension(kvm_state, KVM_CAP_IRQ_ROUTING)) {
+        return -ENOSYS;
+    }
+
+    r = ioctl(flic->fd, KVM_SET_DEVICE_ATTR, &attr);
+    return r ? -errno : 0;
+}
+
+static int kvm_s390_add_adapter_routes(S390FLICState *fs,
+                                       AdapterRoutes *routes)
+{
+    int ret, i;
+    uint64_t ind_offset = routes->adapter.ind_offset;
+
+    for (i = 0; i < routes->num_routes; i++) {
+        ret = kvm_irqchip_add_adapter_route(kvm_state, &routes->adapter);
+        if (ret < 0) {
+            goto out_undo;
+        }
+        routes->gsi[i] = ret;
+        routes->adapter.ind_offset++;
+    }
+    /* Restore passed-in structure to original state. */
+    routes->adapter.ind_offset = ind_offset;
+    return 0;
+out_undo:
+    while (--i >= 0) {
+        kvm_irqchip_release_virq(kvm_state, routes->gsi[i]);
+        routes->gsi[i] = -1;
+    }
+    routes->adapter.ind_offset = ind_offset;
+    return ret;
+}
+
+static void kvm_s390_release_adapter_routes(S390FLICState *fs,
+                                            AdapterRoutes *routes)
+{
+    int i;
+
+    for (i = 0; i < routes->num_routes; i++) {
+        if (routes->gsi[i] >= 0) {
+            kvm_irqchip_release_virq(kvm_state, routes->gsi[i]);
+            routes->gsi[i] = -1;
+        }
+    }
+}
+
 /**
  * kvm_flic_save - Save pending floating interrupts
  * @f: QEMUFile containing migration state
@@ -337,6 +400,9 @@ static void kvm_s390_flic_class_init(ObjectClass *oc, void *data)
     dc->unrealize = kvm_s390_flic_unrealize;
     dc->reset = kvm_s390_flic_reset;
     fsc->register_io_adapter = kvm_s390_register_io_adapter;
+    fsc->io_adapter_map = kvm_s390_io_adapter_map;
+    fsc->add_adapter_routes = kvm_s390_add_adapter_routes;
+    fsc->release_adapter_routes = kvm_s390_release_adapter_routes;
 }
 
 static const TypeInfo kvm_s390_flic_info = {

hw/s390x/virtio-ccw.c

@@ -21,6 +21,8 @@
 #include "hw/sysbus.h"
 #include "qemu/bitops.h"
 #include "hw/virtio/virtio-bus.h"
+#include "hw/s390x/adapter.h"
+#include "hw/s390x/s390_flic.h"
 
 #include "ioinst.h"
 #include "css.h"
@@ -48,7 +50,16 @@ static IndAddr *get_indicator(hwaddr ind_addr, int len)
     return indicator;
 }
 
-static void release_indicator(IndAddr *indicator)
+static int s390_io_adapter_map(AdapterInfo *adapter, uint64_t map_addr,
+                               bool do_map)
+{
+    S390FLICState *fs = s390_get_flic();
+    S390FLICStateClass *fsc = S390_FLIC_COMMON_GET_CLASS(fs);
+
+    return fsc->io_adapter_map(fs, adapter->adapter_id, map_addr, do_map);
+}
+
+static void release_indicator(AdapterInfo *adapter, IndAddr *indicator)
 {
     assert(indicator->refcnt > 0);
     indicator->refcnt--;
@@ -56,9 +67,31 @@ static void release_indicator(IndAddr *indicator)
         return;
     }
     QTAILQ_REMOVE(&indicator_addresses, indicator, sibling);
+    if (indicator->map) {
+        s390_io_adapter_map(adapter, indicator->map, false);
+    }
     g_free(indicator);
 }
 
+static int map_indicator(AdapterInfo *adapter, IndAddr *indicator)
+{
+    int ret;
+
+    if (indicator->map) {
+        return 0; /* already mapped is not an error */
+    }
+    indicator->map = indicator->addr;
+    ret = s390_io_adapter_map(adapter, indicator->map, true);
+    if ((ret != 0) && (ret != -ENOSYS)) {
+        goto out_err;
+    }
+    return 0;
+
+out_err:
+    indicator->map = 0;
+    return ret;
+}
+
 static void virtio_ccw_bus_new(VirtioBusState *bus, size_t bus_size,
                                VirtioCcwDevice *dev);
 
@@ -554,11 +587,12 @@ static int virtio_ccw_cb(SubchDev *sch, CCW1 ccw)
                 dev->indicators = get_indicator(thinint->device_indicator,
                                                 thinint->ind_bit / 8 + 1);
                 dev->thinint_isc = thinint->isc;
-                dev->ind_bit = thinint->ind_bit;
+                dev->routes.adapter.ind_offset = thinint->ind_bit;
+                dev->routes.adapter.summary_offset = 7;
                 cpu_physical_memory_unmap(thinint, hw_len, 0, hw_len);
                 ret = css_register_io_adapter(CSS_IO_ADAPTER_VIRTIO,
                                               dev->thinint_isc, true, false,
-                                              &dev->adapter_id);
+                                              &dev->routes.adapter.adapter_id);
                 assert(ret == 0);
                 sch->thinint_active = ((dev->indicators != NULL) &&
                                        (dev->summary_indicator != NULL));
@@ -732,7 +766,7 @@ static int virtio_ccw_exit(VirtioCcwDevice *dev)
         g_free(sch);
     }
     if (dev->indicators) {
-        release_indicator(dev->indicators);
+        release_indicator(&dev->routes.adapter, dev->indicators);
         dev->indicators = NULL;
     }
     return 0;
@@ -991,9 +1025,11 @@ static void virtio_ccw_notify(DeviceState *d, uint16_t vector)
              * ind_bit indicates the start of the indicators in a big
              * endian notation.
              */
+            uint64_t ind_bit = dev->routes.adapter.ind_offset;
+
             virtio_set_ind_atomic(sch, dev->indicators->addr +
-                                  (dev->ind_bit + vector) / 8,
-                                  0x80 >> ((dev->ind_bit + vector) % 8));
+                                  (ind_bit + vector) / 8,
+                                  0x80 >> ((ind_bit + vector) % 8));
             if (!virtio_set_ind_atomic(sch, dev->summary_indicator->addr,
                                        0x01)) {
                 css_adapter_interrupt(dev->thinint_isc);
@@ -1033,15 +1069,15 @@ static void virtio_ccw_reset(DeviceState *d)
     virtio_reset(vdev);
     css_reset_sch(dev->sch);
     if (dev->indicators) {
-        release_indicator(dev->indicators);
+        release_indicator(&dev->routes.adapter, dev->indicators);
         dev->indicators = NULL;
     }
     if (dev->indicators2) {
-        release_indicator(dev->indicators2);
+        release_indicator(&dev->routes.adapter, dev->indicators2);
         dev->indicators2 = NULL;
     }
     if (dev->summary_indicator) {
-        release_indicator(dev->summary_indicator);
+        release_indicator(&dev->routes.adapter, dev->summary_indicator);
         dev->summary_indicator = NULL;
     }
 }
@@ -1077,6 +1113,79 @@ static int virtio_ccw_set_host_notifier(DeviceState *d, int n, bool assign)
     return virtio_ccw_set_guest2host_notifier(dev, n, assign, false);
 }
 
+static int virtio_ccw_get_mappings(VirtioCcwDevice *dev)
+{
+    int r;
+
+    if (!dev->sch->thinint_active) {
+        return -EINVAL;
+    }
+
+    r = map_indicator(&dev->routes.adapter, dev->summary_indicator);
+    if (r) {
+        return r;
+    }
+    r = map_indicator(&dev->routes.adapter, dev->indicators);
+    if (r) {
+        return r;
+    }
+    dev->routes.adapter.summary_addr = dev->summary_indicator->map;
+    dev->routes.adapter.ind_addr = dev->indicators->map;
+
+    return 0;
+}
+
+static int virtio_ccw_setup_irqroutes(VirtioCcwDevice *dev, int nvqs)
+{
+    int i;
+    VirtIODevice *vdev = virtio_bus_get_device(&dev->bus);
+    int ret;
+    S390FLICState *fs = s390_get_flic();
+    S390FLICStateClass *fsc = S390_FLIC_COMMON_GET_CLASS(fs);
+
+    ret = virtio_ccw_get_mappings(dev);
+    if (ret) {
+        return ret;
+    }
+    for (i = 0; i < nvqs; i++) {
+        if (!virtio_queue_get_num(vdev, i)) {
+            break;
+        }
+    }
+    dev->routes.num_routes = i;
+    return fsc->add_adapter_routes(fs, &dev->routes);
+}
+
+static void virtio_ccw_release_irqroutes(VirtioCcwDevice *dev, int nvqs)
+{
+    S390FLICState *fs = s390_get_flic();
+    S390FLICStateClass *fsc = S390_FLIC_COMMON_GET_CLASS(fs);
+
+    fsc->release_adapter_routes(fs, &dev->routes);
+}
+
+static int virtio_ccw_add_irqfd(VirtioCcwDevice *dev, int n)
+{
+    VirtIODevice *vdev = virtio_bus_get_device(&dev->bus);
+    VirtQueue *vq = virtio_get_queue(vdev, n);
+    EventNotifier *notifier = virtio_queue_get_guest_notifier(vq);
+
+    return kvm_irqchip_add_irqfd_notifier(kvm_state, notifier, NULL,
+                                          dev->routes.gsi[n]);
+}
+
+static void virtio_ccw_remove_irqfd(VirtioCcwDevice *dev, int n)
+{
+    VirtIODevice *vdev = virtio_bus_get_device(&dev->bus);
+    VirtQueue *vq = virtio_get_queue(vdev, n);
+    EventNotifier *notifier = virtio_queue_get_guest_notifier(vq);
+    int ret;
+
+    ret = kvm_irqchip_remove_irqfd_notifier(kvm_state, notifier,
+                                            dev->routes.gsi[n]);
+    assert(ret == 0);
+}
+
 static int virtio_ccw_set_guest_notifier(VirtioCcwDevice *dev, int n,
                                          bool assign, bool with_irqfd)
 {
@@ -1092,11 +1201,17 @@ static int virtio_ccw_set_guest_notifier(VirtioCcwDevice *dev, int n,
             return r;
         }
         virtio_queue_set_guest_notifier_fd_handler(vq, true, with_irqfd);
-        /* We do not support irqfd for classic I/O interrupts, because the
-         * classic interrupts are intermixed with the subchannel status, that
-         * is queried with test subchannel. We want to use vhost, though.
-         * Lets make sure to have vhost running and wire up the irq fd to
-         * land in qemu (and only the irq fd) in this code.
+        if (with_irqfd) {
+            r = virtio_ccw_add_irqfd(dev, n);
+            if (r) {
+                virtio_queue_set_guest_notifier_fd_handler(vq, false,
+                                                           with_irqfd);
+                return r;
+            }
+        }
+        /*
+         * We do not support individual masking for channel devices, so we
+         * need to manually trigger any guest masking callbacks here.
          */
         if (k->guest_notifier_mask) {
             k->guest_notifier_mask(vdev, n, false);
@@ -1110,6 +1225,9 @@ static int virtio_ccw_set_guest_notifier(VirtioCcwDevice *dev, int n,
         if (k->guest_notifier_mask) {
             k->guest_notifier_mask(vdev, n, true);
         }
+        if (with_irqfd) {
+            virtio_ccw_remove_irqfd(dev, n);
+        }
         virtio_queue_set_guest_notifier_fd_handler(vq, false, with_irqfd);
         event_notifier_cleanup(notifier);
     }
@@ -1121,24 +1239,39 @@ static int virtio_ccw_set_guest_notifiers(DeviceState *d, int nvqs,
 {
     VirtioCcwDevice *dev = VIRTIO_CCW_DEVICE(d);
     VirtIODevice *vdev = virtio_bus_get_device(&dev->bus);
+    bool with_irqfd = dev->sch->thinint_active && kvm_irqfds_enabled();
     int r, n;
 
+    if (with_irqfd && assigned) {
+        /* irq routes need to be set up before assigning irqfds */
+        r = virtio_ccw_setup_irqroutes(dev, nvqs);
+        if (r < 0) {
+            goto irqroute_error;
+        }
+    }
     for (n = 0; n < nvqs; n++) {
         if (!virtio_queue_get_num(vdev, n)) {
             break;
         }
-        /* false -> true, as soon as irqfd works */
-        r = virtio_ccw_set_guest_notifier(dev, n, assigned, false);
+        r = virtio_ccw_set_guest_notifier(dev, n, assigned, with_irqfd);
         if (r < 0) {
             goto assign_error;
         }
     }
+    if (with_irqfd && !assigned) {
+        /* release irq routes after irqfds have been released */
+        virtio_ccw_release_irqroutes(dev, nvqs);
+    }
     return 0;
 
 assign_error:
     while (--n >= 0) {
         virtio_ccw_set_guest_notifier(dev, n, !assigned, false);
     }
+irqroute_error:
+    if (with_irqfd && assigned) {
+        virtio_ccw_release_irqroutes(dev, nvqs);
+    }
     return r;
 }
 

hw/s390x/virtio-ccw.h

@@ -22,6 +22,7 @@
 #include <hw/virtio/virtio-balloon.h>
 #include <hw/virtio/virtio-rng.h>
 #include <hw/virtio/virtio-bus.h>
+#include <hw/s390x/s390_flic.h>
 
 #define VIRTUAL_CSSID 0xfe
 
@@ -77,6 +78,7 @@ typedef struct VirtIOCCWDeviceClass {
 
 typedef struct IndAddr {
     hwaddr addr;
+    uint64_t map;
     unsigned long refcnt;
     int len;
     QTAILQ_ENTRY(IndAddr) sibling;
@@ -92,7 +94,7 @@ struct VirtioCcwDevice {
     bool ioeventfd_disabled;
     uint32_t flags;
     uint8_t thinint_isc;
-    uint32_t adapter_id;
+    AdapterRoutes routes;
     /* Guest provided values: */
     IndAddr *indicators;
     IndAddr *indicators2;

include/hw/s390x/adapter.h

@@ -0,0 +1,23 @@
+/*
+ * s390 adapter definitions
+ *
+ * Copyright 2013,2014 IBM Corp.
+ * Author(s): Cornelia Huck <cornelia.huck@de.ibm.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or (at
+ * your option) any later version. See the COPYING file in the top-level
+ * directory.
+ */
+
+#ifndef S390X_ADAPTER_H
+#define S390X_ADAPTER_H
+
+struct AdapterInfo {
+    uint64_t ind_addr;
+    uint64_t summary_addr;
+    uint64_t ind_offset;
+    uint32_t summary_offset;
+    uint32_t adapter_id;
+};
+
+#endif

include/hw/s390x/s390_flic.h

@@ -14,6 +14,14 @@
 #define __HW_S390_FLIC_H
 
 #include "hw/sysbus.h"
+#include "hw/s390x/adapter.h"
+#include "hw/virtio/virtio.h"
+
+typedef struct AdapterRoutes {
+    AdapterInfo adapter;
+    int num_routes;
+    int gsi[VIRTIO_PCI_QUEUE_MAX];
+} AdapterRoutes;
 
 #define TYPE_S390_FLIC_COMMON "s390-flic"
 #define S390_FLIC_COMMON(obj) \
@@ -34,6 +42,10 @@ typedef struct S390FLICStateClass {
 
     int (*register_io_adapter)(S390FLICState *fs, uint32_t id, uint8_t isc,
                                bool swap, bool maskable);
+    int (*io_adapter_map)(S390FLICState *fs, uint32_t id, uint64_t map_addr,
+                          bool do_map);
+    int (*add_adapter_routes)(S390FLICState *fs, AdapterRoutes *routes);
+    void (*release_adapter_routes)(S390FLICState *fs, AdapterRoutes *routes);
 } S390FLICStateClass;
 
 #define TYPE_KVM_S390_FLIC "s390-flic-kvm"

include/qemu/typedefs.h

@@ -74,5 +74,6 @@ typedef struct SHPCDevice SHPCDevice;
 typedef struct FWCfgState FWCfgState;
 typedef struct PcGuestInfo PcGuestInfo;
 typedef struct Range Range;
+typedef struct AdapterInfo AdapterInfo;
 
 #endif /* QEMU_TYPEDEFS_H */

include/sysemu/kvm.h

@@ -363,6 +363,8 @@ int kvm_irqchip_add_msi_route(KVMState *s, MSIMessage msg);
 int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg);
 void kvm_irqchip_release_virq(KVMState *s, int virq);
 
+int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter);
+
 int kvm_irqchip_add_irqfd_notifier(KVMState *s, EventNotifier *n,
                                    EventNotifier *rn, int virq);
 int kvm_irqchip_remove_irqfd_notifier(KVMState *s, EventNotifier *n, int virq);

kvm-all.c

@@ -27,6 +27,7 @@
 #include "sysemu/sysemu.h"
 #include "hw/hw.h"
 #include "hw/pci/msi.h"
+#include "hw/s390x/adapter.h"
 #include "exec/gdbstub.h"
 #include "sysemu/kvm.h"
 #include "qemu/bswap.h"
@@ -1236,6 +1237,35 @@ static int kvm_irqchip_assign_irqfd(KVMState *s, int fd, int rfd, int virq,
     return kvm_vm_ioctl(s, KVM_IRQFD, &irqfd);
 }
 
+int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter)
+{
+    struct kvm_irq_routing_entry kroute;
+    int virq;
+
+    if (!kvm_gsi_routing_enabled()) {
+        return -ENOSYS;
+    }
+
+    virq = kvm_irqchip_get_virq(s);
+    if (virq < 0) {
+        return virq;
+    }
+
+    kroute.gsi = virq;
+    kroute.type = KVM_IRQ_ROUTING_S390_ADAPTER;
+    kroute.flags = 0;
+    kroute.u.adapter.summary_addr = adapter->summary_addr;
+    kroute.u.adapter.ind_addr = adapter->ind_addr;
+    kroute.u.adapter.summary_offset = adapter->summary_offset;
+    kroute.u.adapter.ind_offset = adapter->ind_offset;
+    kroute.u.adapter.adapter_id = adapter->adapter_id;
+
+    kvm_add_routing_entry(s, &kroute);
+    kvm_irqchip_commit_routes(s);
+
+    return virq;
+}
+
 #else /* !KVM_CAP_IRQ_ROUTING */
 
 void kvm_init_irq_routing(KVMState *s)
@@ -1256,6 +1286,11 @@ int kvm_irqchip_add_msi_route(KVMState *s, MSIMessage msg)
     return -ENOSYS;
 }
 
+int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter)
+{
+    return -ENOSYS;
+}
+
 static int kvm_irqchip_assign_irqfd(KVMState *s, int fd, int virq, bool assign)
 {
     abort();
@@ -1285,7 +1320,8 @@ static int kvm_irqchip_create(KVMState *s)
     int ret;
 
     if (!qemu_opt_get_bool(qemu_get_machine_opts(), "kernel_irqchip", true) ||
-        !kvm_check_extension(s, KVM_CAP_IRQCHIP)) {
+        (!kvm_check_extension(s, KVM_CAP_IRQCHIP) &&
+         (kvm_vm_enable_cap(s, KVM_CAP_S390_IRQCHIP, 0) < 0))) {
         return 0;
     }
 

kvm-stub.c

@@ -136,6 +136,11 @@ int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg)
     return -ENOSYS;
 }
 
+int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter)
+{
+    return -ENOSYS;
+}
+
 int kvm_irqchip_add_irqfd_notifier(KVMState *s, EventNotifier *n,
                                    EventNotifier *rn, int virq)
 {

target-s390x/kvm.c

@@ -911,6 +911,16 @@ void kvm_s390_enable_css_support(S390CPU *cpu)
 
 void kvm_arch_init_irq_routing(KVMState *s)
 {
+    /*
+     * Note that while irqchip capabilities generally imply that cpustates
+     * are handled in-kernel, it is not true for s390 (yet); therefore, we
+     * have to override the common code kvm_halt_in_kernel_allowed setting.
+     */
+    if (kvm_check_extension(s, KVM_CAP_IRQ_ROUTING)) {
+        kvm_irqfds_allowed = true;
+        kvm_gsi_routing_allowed = true;
+        kvm_halt_in_kernel_allowed = false;
+    }
 }
 
 int kvm_s390_assign_subch_ioeventfd(EventNotifier *notifier, uint32_t sch,