xref: /qemu/hw/vfio/common.c (revision c3bef3b4)

/*
 * generic functions used by VFIO devices
 *
 * Copyright Red Hat, Inc. 2012
 *
 * Authors:
 *  Alex Williamson <alex.williamson@redhat.com>
 *
 * This work is licensed under the terms of the GNU GPL, version 2.  See
 * the COPYING file in the top-level directory.
 *
 * Based on qemu-kvm device-assignment:
 *  Adapted for KVM by Qumranet.
 *  Copyright (c) 2007, Neocleus, Alex Novik (alex@neocleus.com)
 *  Copyright (c) 2007, Neocleus, Guy Zana (guy@neocleus.com)
 *  Copyright (C) 2008, Qumranet, Amit Shah (amit.shah@qumranet.com)
 *  Copyright (C) 2008, Red Hat, Amit Shah (amit.shah@redhat.com)
 *  Copyright (C) 2008, IBM, Muli Ben-Yehuda (muli@il.ibm.com)
 */

#include "qemu/osdep.h"
#include <sys/ioctl.h>
#ifdef CONFIG_KVM
#include <linux/kvm.h>
#endif
#include <linux/vfio.h>

#include "hw/vfio/vfio-common.h"
#include "hw/vfio/vfio.h"
#include "exec/address-spaces.h"
#include "exec/memory.h"
#include "exec/ram_addr.h"
#include "hw/hw.h"
#include "qemu/error-report.h"
#include "qemu/main-loop.h"
#include "qemu/range.h"
#include "sysemu/kvm.h"
#include "sysemu/reset.h"
#include "sysemu/runstate.h"
#include "trace.h"
#include "qapi/error.h"
#include "migration/migration.h"
#include "migration/misc.h"
#include "migration/blocker.h"
#include "sysemu/tpm.h"

VFIOGroupList vfio_group_list =
    QLIST_HEAD_INITIALIZER(vfio_group_list);
static QLIST_HEAD(, VFIOAddressSpace) vfio_address_spaces =
    QLIST_HEAD_INITIALIZER(vfio_address_spaces);

#ifdef CONFIG_KVM
/*
 * We have a single VFIO pseudo device per KVM VM.  Once created it lives
 * for the life of the VM.  Closing the file descriptor only drops our
 * reference to it and the device's reference to kvm.  Therefore once
 * initialized, this file descriptor is only released on QEMU exit and
 * we'll re-use it should another vfio device be attached before then.
 */
static int vfio_kvm_device_fd = -1;
#endif

/*
 * Common VFIO interrupt disable
 */
void vfio_disable_irqindex(VFIODevice *vbasedev, int index)
{
    struct vfio_irq_set irq_set = {
        .argsz = sizeof(irq_set),
        .flags = VFIO_IRQ_SET_DATA_NONE | VFIO_IRQ_SET_ACTION_TRIGGER,
        .index = index,
        .start = 0,
        .count = 0,
    };

    ioctl(vbasedev->fd, VFIO_DEVICE_SET_IRQS, &irq_set);
}

void vfio_unmask_single_irqindex(VFIODevice *vbasedev, int index)
{
    struct vfio_irq_set irq_set = {
        .argsz = sizeof(irq_set),
        .flags = VFIO_IRQ_SET_DATA_NONE | VFIO_IRQ_SET_ACTION_UNMASK,
        .index = index,
        .start = 0,
        .count = 1,
    };

    ioctl(vbasedev->fd, VFIO_DEVICE_SET_IRQS, &irq_set);
}

void vfio_mask_single_irqindex(VFIODevice *vbasedev, int index)
{
    struct vfio_irq_set irq_set = {
        .argsz = sizeof(irq_set),
        .flags = VFIO_IRQ_SET_DATA_NONE | VFIO_IRQ_SET_ACTION_MASK,
        .index = index,
        .start = 0,
        .count = 1,
    };

    ioctl(vbasedev->fd, VFIO_DEVICE_SET_IRQS, &irq_set);
}

static inline const char *action_to_str(int action)
{
    switch (action) {
    case VFIO_IRQ_SET_ACTION_MASK:
        return "MASK";
    case VFIO_IRQ_SET_ACTION_UNMASK:
        return "UNMASK";
    case VFIO_IRQ_SET_ACTION_TRIGGER:
        return "TRIGGER";
    default:
        return "UNKNOWN ACTION";
    }
}

static const char *index_to_str(VFIODevice *vbasedev, int index)
{
    if (vbasedev->type != VFIO_DEVICE_TYPE_PCI) {
        return NULL;
    }

    switch (index) {
    case VFIO_PCI_INTX_IRQ_INDEX:
        return "INTX";
    case VFIO_PCI_MSI_IRQ_INDEX:
        return "MSI";
    case VFIO_PCI_MSIX_IRQ_INDEX:
        return "MSIX";
    case VFIO_PCI_ERR_IRQ_INDEX:
        return "ERR";
    case VFIO_PCI_REQ_IRQ_INDEX:
        return "REQ";
    default:
        return NULL;
    }
}

static int vfio_ram_block_discard_disable(VFIOContainer *container, bool state)
{
    switch (container->iommu_type) {
    case VFIO_TYPE1v2_IOMMU:
    case VFIO_TYPE1_IOMMU:
        /*
         * We support coordinated discarding of RAM via the RamDiscardManager.
         */
        return ram_block_uncoordinated_discard_disable(state);
    default:
        /*
         * VFIO_SPAPR_TCE_IOMMU most probably works just fine with
         * RamDiscardManager, however, it is completely untested.
         *
         * VFIO_SPAPR_TCE_v2_IOMMU with "DMA memory preregistering" does
         * completely the opposite of managing mapping/pinning dynamically as
         * required by RamDiscardManager. We would have to special-case sections
         * with a RamDiscardManager.
         */
        return ram_block_discard_disable(state);
    }
}

int vfio_set_irq_signaling(VFIODevice *vbasedev, int index, int subindex,
                           int action, int fd, Error **errp)
{
    struct vfio_irq_set *irq_set;
    int argsz, ret = 0;
    const char *name;
    int32_t *pfd;

    argsz = sizeof(*irq_set) + sizeof(*pfd);

    irq_set = g_malloc0(argsz);
    irq_set->argsz = argsz;
    irq_set->flags = VFIO_IRQ_SET_DATA_EVENTFD | action;
    irq_set->index = index;
    irq_set->start = subindex;
    irq_set->count = 1;
    pfd = (int32_t *)&irq_set->data;
    *pfd = fd;

    if (ioctl(vbasedev->fd, VFIO_DEVICE_SET_IRQS, irq_set)) {
        ret = -errno;
    }
    g_free(irq_set);

    if (!ret) {
        return 0;
    }

    error_setg_errno(errp, -ret, "VFIO_DEVICE_SET_IRQS failure");

    name = index_to_str(vbasedev, index);
    if (name) {
        error_prepend(errp, "%s-%d: ", name, subindex);
    } else {
        error_prepend(errp, "index %d-%d: ", index, subindex);
    }
    error_prepend(errp,
                  "Failed to %s %s eventfd signaling for interrupt ",
                  fd < 0 ? "tear down" : "set up", action_to_str(action));
    return ret;
}

/*
 * IO Port/MMIO - Beware of the endians, VFIO is always little endian
 */
void vfio_region_write(void *opaque, hwaddr addr,
                       uint64_t data, unsigned size)
{
    VFIORegion *region = opaque;
    VFIODevice *vbasedev = region->vbasedev;
    union {
        uint8_t byte;
        uint16_t word;
        uint32_t dword;
        uint64_t qword;
    } buf;

    switch (size) {
    case 1:
        buf.byte = data;
        break;
    case 2:
        buf.word = cpu_to_le16(data);
        break;
    case 4:
        buf.dword = cpu_to_le32(data);
        break;
    case 8:
        buf.qword = cpu_to_le64(data);
        break;
    default:
        hw_error("vfio: unsupported write size, %u bytes", size);
        break;
    }

    if (pwrite(vbasedev->fd, &buf, size, region->fd_offset + addr) != size) {
        error_report("%s(%s:region%d+0x%"HWADDR_PRIx", 0x%"PRIx64
                     ",%d) failed: %m",
                     __func__, vbasedev->name, region->nr,
                     addr, data, size);
    }

    trace_vfio_region_write(vbasedev->name, region->nr, addr, data, size);

    /*
     * A read or write to a BAR always signals an INTx EOI.  This will
     * do nothing if not pending (including not in INTx mode).  We assume
     * that a BAR access is in response to an interrupt and that BAR
     * accesses will service the interrupt.  Unfortunately, we don't know
     * which access will service the interrupt, so we're potentially
     * getting quite a few host interrupts per guest interrupt.
     */
    vbasedev->ops->vfio_eoi(vbasedev);
}

uint64_t vfio_region_read(void *opaque,
                          hwaddr addr, unsigned size)
{
    VFIORegion *region = opaque;
    VFIODevice *vbasedev = region->vbasedev;
    union {
        uint8_t byte;
        uint16_t word;
        uint32_t dword;
        uint64_t qword;
    } buf;
    uint64_t data = 0;

    if (pread(vbasedev->fd, &buf, size, region->fd_offset + addr) != size) {
        error_report("%s(%s:region%d+0x%"HWADDR_PRIx", %d) failed: %m",
                     __func__, vbasedev->name, region->nr,
                     addr, size);
        return (uint64_t)-1;
    }
    switch (size) {
    case 1:
        data = buf.byte;
        break;
    case 2:
        data = le16_to_cpu(buf.word);
        break;
    case 4:
        data = le32_to_cpu(buf.dword);
        break;
    case 8:
        data = le64_to_cpu(buf.qword);
        break;
    default:
        hw_error("vfio: unsupported read size, %u bytes", size);
        break;
    }

    trace_vfio_region_read(vbasedev->name, region->nr, addr, size, data);

    /* Same as write above */
    vbasedev->ops->vfio_eoi(vbasedev);

    return data;
}

const MemoryRegionOps vfio_region_ops = {
    .read = vfio_region_read,
    .write = vfio_region_write,
    .endianness = DEVICE_LITTLE_ENDIAN,
    .valid = {
        .min_access_size = 1,
        .max_access_size = 8,
    },
    .impl = {
        .min_access_size = 1,
        .max_access_size = 8,
    },
};

/*
 * Device state interfaces
 */

bool vfio_mig_active(void)
{
    VFIOGroup *group;
    VFIODevice *vbasedev;

    if (QLIST_EMPTY(&vfio_group_list)) {
        return false;
    }

    QLIST_FOREACH(group, &vfio_group_list, next) {
        QLIST_FOREACH(vbasedev, &group->device_list, next) {
            if (vbasedev->migration_blocker) {
                return false;
            }
        }
    }
    return true;
}

static Error *multiple_devices_migration_blocker;

static unsigned int vfio_migratable_device_num(void)
{
    VFIOGroup *group;
    VFIODevice *vbasedev;
    unsigned int device_num = 0;

    QLIST_FOREACH(group, &vfio_group_list, next) {
        QLIST_FOREACH(vbasedev, &group->device_list, next) {
            if (vbasedev->migration) {
                device_num++;
            }
        }
    }

    return device_num;
}

int vfio_block_multiple_devices_migration(Error **errp)
{
    int ret;

    if (multiple_devices_migration_blocker ||
        vfio_migratable_device_num() <= 1) {
        return 0;
    }

    error_setg(&multiple_devices_migration_blocker,
               "Migration is currently not supported with multiple "
               "VFIO devices");
    ret = migrate_add_blocker(multiple_devices_migration_blocker, errp);
    if (ret < 0) {
        error_free(multiple_devices_migration_blocker);
        multiple_devices_migration_blocker = NULL;
    }

    return ret;
}

void vfio_unblock_multiple_devices_migration(void)
{
    if (!multiple_devices_migration_blocker ||
        vfio_migratable_device_num() > 1) {
        return;
    }

    migrate_del_blocker(multiple_devices_migration_blocker);
    error_free(multiple_devices_migration_blocker);
    multiple_devices_migration_blocker = NULL;
}

static bool vfio_devices_all_dirty_tracking(VFIOContainer *container)
{
    VFIOGroup *group;
    VFIODevice *vbasedev;
    MigrationState *ms = migrate_get_current();

    if (!migration_is_setup_or_active(ms->state)) {
        return false;
    }

    QLIST_FOREACH(group, &container->group_list, container_next) {
        QLIST_FOREACH(vbasedev, &group->device_list, next) {
            VFIOMigration *migration = vbasedev->migration;

            if (!migration) {
                return false;
            }

            if (vbasedev->pre_copy_dirty_page_tracking == ON_OFF_AUTO_OFF &&
                migration->device_state == VFIO_DEVICE_STATE_RUNNING) {
                return false;
            }
        }
    }
    return true;
}

/*
 * Check if all VFIO devices are running and migration is active, which is
 * essentially equivalent to the migration being in pre-copy phase.
 */
static bool vfio_devices_all_running_and_mig_active(VFIOContainer *container)
{
    VFIOGroup *group;
    VFIODevice *vbasedev;

    if (!migration_is_active(migrate_get_current())) {
        return false;
    }

    QLIST_FOREACH(group, &container->group_list, container_next) {
        QLIST_FOREACH(vbasedev, &group->device_list, next) {
            VFIOMigration *migration = vbasedev->migration;

            if (!migration) {
                return false;
            }

            if (migration->device_state == VFIO_DEVICE_STATE_RUNNING) {
                continue;
            } else {
                return false;
            }
        }
    }
    return true;
}

static int vfio_dma_unmap_bitmap(VFIOContainer *container,
                                 hwaddr iova, ram_addr_t size,
                                 IOMMUTLBEntry *iotlb)
{
    struct vfio_iommu_type1_dma_unmap *unmap;
    struct vfio_bitmap *bitmap;
    uint64_t pages = REAL_HOST_PAGE_ALIGN(size) / qemu_real_host_page_size();
    int ret;

    unmap = g_malloc0(sizeof(*unmap) + sizeof(*bitmap));

    unmap->argsz = sizeof(*unmap) + sizeof(*bitmap);
    unmap->iova = iova;
    unmap->size = size;
    unmap->flags |= VFIO_DMA_UNMAP_FLAG_GET_DIRTY_BITMAP;
    bitmap = (struct vfio_bitmap *)&unmap->data;

    /*
     * cpu_physical_memory_set_dirty_lebitmap() supports pages in bitmap of
     * qemu_real_host_page_size to mark those dirty. Hence set bitmap_pgsize
     * to qemu_real_host_page_size.
     */

    bitmap->pgsize = qemu_real_host_page_size();
    bitmap->size = ROUND_UP(pages, sizeof(__u64) * BITS_PER_BYTE) /
                   BITS_PER_BYTE;

    if (bitmap->size > container->max_dirty_bitmap_size) {
        error_report("UNMAP: Size of bitmap too big 0x%"PRIx64,
                     (uint64_t)bitmap->size);
        ret = -E2BIG;
        goto unmap_exit;
    }

    bitmap->data = g_try_malloc0(bitmap->size);
    if (!bitmap->data) {
        ret = -ENOMEM;
        goto unmap_exit;
    }

    ret = ioctl(container->fd, VFIO_IOMMU_UNMAP_DMA, unmap);
    if (!ret) {
        cpu_physical_memory_set_dirty_lebitmap((unsigned long *)bitmap->data,
                iotlb->translated_addr, pages);
    } else {
        error_report("VFIO_UNMAP_DMA with DIRTY_BITMAP : %m");
    }

    g_free(bitmap->data);
unmap_exit:
    g_free(unmap);
    return ret;
}

/*
 * DMA - Mapping and unmapping for the "type1" IOMMU interface used on x86
 */
static int vfio_dma_unmap(VFIOContainer *container,
                          hwaddr iova, ram_addr_t size,
                          IOMMUTLBEntry *iotlb)
{
    struct vfio_iommu_type1_dma_unmap unmap = {
        .argsz = sizeof(unmap),
        .flags = 0,
        .iova = iova,
        .size = size,
    };

    if (iotlb && container->dirty_pages_supported &&
        vfio_devices_all_running_and_mig_active(container)) {
        return vfio_dma_unmap_bitmap(container, iova, size, iotlb);
    }

    while (ioctl(container->fd, VFIO_IOMMU_UNMAP_DMA, &unmap)) {
        /*
         * The type1 backend has an off-by-one bug in the kernel (71a7d3d78e3c
         * v4.15) where an overflow in its wrap-around check prevents us from
         * unmapping the last page of the address space.  Test for the error
         * condition and re-try the unmap excluding the last page.  The
         * expectation is that we've never mapped the last page anyway and this
         * unmap request comes via vIOMMU support which also makes it unlikely
         * that this page is used.  This bug was introduced well after type1 v2
         * support was introduced, so we shouldn't need to test for v1.  A fix
         * is queued for kernel v5.0 so this workaround can be removed once
         * affected kernels are sufficiently deprecated.
         */
        if (errno == EINVAL && unmap.size && !(unmap.iova + unmap.size) &&
            container->iommu_type == VFIO_TYPE1v2_IOMMU) {
            trace_vfio_dma_unmap_overflow_workaround();
            unmap.size -= 1ULL << ctz64(container->pgsizes);
            continue;
        }
        error_report("VFIO_UNMAP_DMA failed: %s", strerror(errno));
        return -errno;
    }

    if (iotlb && vfio_devices_all_running_and_mig_active(container)) {
        cpu_physical_memory_set_dirty_range(iotlb->translated_addr, size,
                                            tcg_enabled() ? DIRTY_CLIENTS_ALL :
                                            DIRTY_CLIENTS_NOCODE);
    }

    return 0;
}

static int vfio_dma_map(VFIOContainer *container, hwaddr iova,
                        ram_addr_t size, void *vaddr, bool readonly)
{
    struct vfio_iommu_type1_dma_map map = {
        .argsz = sizeof(map),
        .flags = VFIO_DMA_MAP_FLAG_READ,
        .vaddr = (__u64)(uintptr_t)vaddr,
        .iova = iova,
        .size = size,
    };

    if (!readonly) {
        map.flags |= VFIO_DMA_MAP_FLAG_WRITE;
    }

    /*
     * Try the mapping, if it fails with EBUSY, unmap the region and try
     * again.  This shouldn't be necessary, but we sometimes see it in
     * the VGA ROM space.
     */
    if (ioctl(container->fd, VFIO_IOMMU_MAP_DMA, &map) == 0 ||
        (errno == EBUSY && vfio_dma_unmap(container, iova, size, NULL) == 0 &&
         ioctl(container->fd, VFIO_IOMMU_MAP_DMA, &map) == 0)) {
        return 0;
    }

    error_report("VFIO_MAP_DMA failed: %s", strerror(errno));
    return -errno;
}

static void vfio_host_win_add(VFIOContainer *container,
                              hwaddr min_iova, hwaddr max_iova,
                              uint64_t iova_pgsizes)
{
    VFIOHostDMAWindow *hostwin;

    QLIST_FOREACH(hostwin, &container->hostwin_list, hostwin_next) {
        if (ranges_overlap(hostwin->min_iova,
                           hostwin->max_iova - hostwin->min_iova + 1,
                           min_iova,
                           max_iova - min_iova + 1)) {
            hw_error("%s: Overlapped IOMMU are not enabled", __func__);
        }
    }

    hostwin = g_malloc0(sizeof(*hostwin));

    hostwin->min_iova = min_iova;
    hostwin->max_iova = max_iova;
    hostwin->iova_pgsizes = iova_pgsizes;
    QLIST_INSERT_HEAD(&container->hostwin_list, hostwin, hostwin_next);
}

static int vfio_host_win_del(VFIOContainer *container, hwaddr min_iova,
                             hwaddr max_iova)
{
    VFIOHostDMAWindow *hostwin;

    QLIST_FOREACH(hostwin, &container->hostwin_list, hostwin_next) {
        if (hostwin->min_iova == min_iova && hostwin->max_iova == max_iova) {
            QLIST_REMOVE(hostwin, hostwin_next);
            g_free(hostwin);
            return 0;
        }
    }

    return -1;
}

static bool vfio_listener_skipped_section(MemoryRegionSection *section)
{
    return (!memory_region_is_ram(section->mr) &&
            !memory_region_is_iommu(section->mr)) ||
           memory_region_is_protected(section->mr) ||
           /*
            * Sizing an enabled 64-bit BAR can cause spurious mappings to
            * addresses in the upper part of the 64-bit address space.  These
            * are never accessed by the CPU and beyond the address width of
            * some IOMMU hardware.  TODO: VFIO should tell us the IOMMU width.
            */
           section->offset_within_address_space & (1ULL << 63);
}

/* Called with rcu_read_lock held.  */
static bool vfio_get_xlat_addr(IOMMUTLBEntry *iotlb, void **vaddr,
                               ram_addr_t *ram_addr, bool *read_only)
{
    bool ret, mr_has_discard_manager;

    ret = memory_get_xlat_addr(iotlb, vaddr, ram_addr, read_only,
                               &mr_has_discard_manager);
    if (ret && mr_has_discard_manager) {
        /*
         * Malicious VMs might trigger discarding of IOMMU-mapped memory. The
         * pages will remain pinned inside vfio until unmapped, resulting in a
         * higher memory consumption than expected. If memory would get
         * populated again later, there would be an inconsistency between pages
         * pinned by vfio and pages seen by QEMU. This is the case until
         * unmapped from the IOMMU (e.g., during device reset).
         *
         * With malicious guests, we really only care about pinning more memory
         * than expected. RLIMIT_MEMLOCK set for the user/process can never be
         * exceeded and can be used to mitigate this problem.
         */
        warn_report_once("Using vfio with vIOMMUs and coordinated discarding of"
                         " RAM (e.g., virtio-mem) works, however, malicious"
                         " guests can trigger pinning of more memory than"
                         " intended via an IOMMU. It's possible to mitigate "
                         " by setting/adjusting RLIMIT_MEMLOCK.");
    }
    return ret;
}

static void vfio_iommu_map_notify(IOMMUNotifier *n, IOMMUTLBEntry *iotlb)
{
    VFIOGuestIOMMU *giommu = container_of(n, VFIOGuestIOMMU, n);
    VFIOContainer *container = giommu->container;
    hwaddr iova = iotlb->iova + giommu->iommu_offset;
    void *vaddr;
    int ret;

    trace_vfio_iommu_map_notify(iotlb->perm == IOMMU_NONE ? "UNMAP" : "MAP",
                                iova, iova + iotlb->addr_mask);

    if (iotlb->target_as != &address_space_memory) {
        error_report("Wrong target AS \"%s\", only system memory is allowed",
                     iotlb->target_as->name ? iotlb->target_as->name : "none");
        return;
    }

    rcu_read_lock();

    if ((iotlb->perm & IOMMU_RW) != IOMMU_NONE) {
        bool read_only;

        if (!vfio_get_xlat_addr(iotlb, &vaddr, NULL, &read_only)) {
            goto out;
        }
        /*
         * vaddr is only valid until rcu_read_unlock(). But after
         * vfio_dma_map has set up the mapping the pages will be
         * pinned by the kernel. This makes sure that the RAM backend
         * of vaddr will always be there, even if the memory object is
         * destroyed and its backing memory munmap-ed.
         */
        ret = vfio_dma_map(container, iova,
                           iotlb->addr_mask + 1, vaddr,
                           read_only);
        if (ret) {
            error_report("vfio_dma_map(%p, 0x%"HWADDR_PRIx", "
                         "0x%"HWADDR_PRIx", %p) = %d (%m)",
                         container, iova,
                         iotlb->addr_mask + 1, vaddr, ret);
        }
    } else {
        ret = vfio_dma_unmap(container, iova, iotlb->addr_mask + 1, iotlb);
        if (ret) {
            error_report("vfio_dma_unmap(%p, 0x%"HWADDR_PRIx", "
                         "0x%"HWADDR_PRIx") = %d (%m)",
                         container, iova,
                         iotlb->addr_mask + 1, ret);
        }
    }
out:
    rcu_read_unlock();
}

static void vfio_ram_discard_notify_discard(RamDiscardListener *rdl,
                                            MemoryRegionSection *section)
{
    VFIORamDiscardListener *vrdl = container_of(rdl, VFIORamDiscardListener,
                                                listener);
    const hwaddr size = int128_get64(section->size);
    const hwaddr iova = section->offset_within_address_space;
    int ret;

    /* Unmap with a single call. */
    ret = vfio_dma_unmap(vrdl->container, iova, size , NULL);
    if (ret) {
        error_report("%s: vfio_dma_unmap() failed: %s", __func__,
                     strerror(-ret));
    }
}

static int vfio_ram_discard_notify_populate(RamDiscardListener *rdl,
                                            MemoryRegionSection *section)
{
    VFIORamDiscardListener *vrdl = container_of(rdl, VFIORamDiscardListener,
                                                listener);
    const hwaddr end = section->offset_within_region +
                       int128_get64(section->size);
    hwaddr start, next, iova;
    void *vaddr;
    int ret;

    /*
     * Map in (aligned within memory region) minimum granularity, so we can
     * unmap in minimum granularity later.
     */
    for (start = section->offset_within_region; start < end; start = next) {
        next = ROUND_UP(start + 1, vrdl->granularity);
        next = MIN(next, end);

        iova = start - section->offset_within_region +
               section->offset_within_address_space;
        vaddr = memory_region_get_ram_ptr(section->mr) + start;

        ret = vfio_dma_map(vrdl->container, iova, next - start,
                           vaddr, section->readonly);
        if (ret) {
            /* Rollback */
            vfio_ram_discard_notify_discard(rdl, section);
            return ret;
        }
    }
    return 0;
}

static void vfio_register_ram_discard_listener(VFIOContainer *container,
                                               MemoryRegionSection *section)
{
    RamDiscardManager *rdm = memory_region_get_ram_discard_manager(section->mr);
    VFIORamDiscardListener *vrdl;

    /* Ignore some corner cases not relevant in practice. */
    g_assert(QEMU_IS_ALIGNED(section->offset_within_region, TARGET_PAGE_SIZE));
    g_assert(QEMU_IS_ALIGNED(section->offset_within_address_space,
                             TARGET_PAGE_SIZE));
    g_assert(QEMU_IS_ALIGNED(int128_get64(section->size), TARGET_PAGE_SIZE));

    vrdl = g_new0(VFIORamDiscardListener, 1);
    vrdl->container = container;
    vrdl->mr = section->mr;
    vrdl->offset_within_address_space = section->offset_within_address_space;
    vrdl->size = int128_get64(section->size);
    vrdl->granularity = ram_discard_manager_get_min_granularity(rdm,
                                                                section->mr);

    g_assert(vrdl->granularity && is_power_of_2(vrdl->granularity));
    g_assert(container->pgsizes &&
             vrdl->granularity >= 1ULL << ctz64(container->pgsizes));

    ram_discard_listener_init(&vrdl->listener,
                              vfio_ram_discard_notify_populate,
                              vfio_ram_discard_notify_discard, true);
    ram_discard_manager_register_listener(rdm, &vrdl->listener, section);
    QLIST_INSERT_HEAD(&container->vrdl_list, vrdl, next);

    /*
     * Sanity-check if we have a theoretically problematic setup where we could
     * exceed the maximum number of possible DMA mappings over time. We assume
     * that each mapped section in the same address space as a RamDiscardManager
     * section consumes exactly one DMA mapping, with the exception of
     * RamDiscardManager sections; i.e., we don't expect to have gIOMMU sections
     * in the same address space as RamDiscardManager sections.
     *
     * We assume that each section in the address space consumes one memslot.
     * We take the number of KVM memory slots as a best guess for the maximum
     * number of sections in the address space we could have over time,
     * also consuming DMA mappings.
     */
    if (container->dma_max_mappings) {
        unsigned int vrdl_count = 0, vrdl_mappings = 0, max_memslots = 512;

#ifdef CONFIG_KVM
        if (kvm_enabled()) {
            max_memslots = kvm_get_max_memslots();
        }
#endif

        QLIST_FOREACH(vrdl, &container->vrdl_list, next) {
            hwaddr start, end;

            start = QEMU_ALIGN_DOWN(vrdl->offset_within_address_space,
                                    vrdl->granularity);
            end = ROUND_UP(vrdl->offset_within_address_space + vrdl->size,
                           vrdl->granularity);
            vrdl_mappings += (end - start) / vrdl->granularity;
            vrdl_count++;
        }

        if (vrdl_mappings + max_memslots - vrdl_count >
            container->dma_max_mappings) {
            warn_report("%s: possibly running out of DMA mappings. E.g., try"
                        " increasing the 'block-size' of virtio-mem devies."
                        " Maximum possible DMA mappings: %d, Maximum possible"
                        " memslots: %d", __func__, container->dma_max_mappings,
                        max_memslots);
        }
    }
}

static void vfio_unregister_ram_discard_listener(VFIOContainer *container,
                                                 MemoryRegionSection *section)
{
    RamDiscardManager *rdm = memory_region_get_ram_discard_manager(section->mr);
    VFIORamDiscardListener *vrdl = NULL;

    QLIST_FOREACH(vrdl, &container->vrdl_list, next) {
        if (vrdl->mr == section->mr &&
            vrdl->offset_within_address_space ==
            section->offset_within_address_space) {
            break;
        }
    }

    if (!vrdl) {
        hw_error("vfio: Trying to unregister missing RAM discard listener");
    }

    ram_discard_manager_unregister_listener(rdm, &vrdl->listener);
    QLIST_REMOVE(vrdl, next);
    g_free(vrdl);
}

static bool vfio_known_safe_misalignment(MemoryRegionSection *section)
{
    MemoryRegion *mr = section->mr;

    if (!TPM_IS_CRB(mr->owner)) {
        return false;
    }

    /* this is a known safe misaligned region, just trace for debug purpose */
    trace_vfio_known_safe_misalignment(memory_region_name(mr),
                                       section->offset_within_address_space,
                                       section->offset_within_region,
                                       qemu_real_host_page_size());
    return true;
}

static void vfio_listener_region_add(MemoryListener *listener,
                                     MemoryRegionSection *section)
{
    VFIOContainer *container = container_of(listener, VFIOContainer, listener);
    hwaddr iova, end;
    Int128 llend, llsize;
    void *vaddr;
    int ret;
    VFIOHostDMAWindow *hostwin;
    bool hostwin_found;
    Error *err = NULL;

    if (vfio_listener_skipped_section(section)) {
        trace_vfio_listener_region_add_skip(
                section->offset_within_address_space,
                section->offset_within_address_space +
                int128_get64(int128_sub(section->size, int128_one())));
        return;
    }

    if (unlikely((section->offset_within_address_space &
                  ~qemu_real_host_page_mask()) !=
                 (section->offset_within_region & ~qemu_real_host_page_mask()))) {
        if (!vfio_known_safe_misalignment(section)) {
            error_report("%s received unaligned region %s iova=0x%"PRIx64
                         " offset_within_region=0x%"PRIx64
                         " qemu_real_host_page_size=0x%"PRIxPTR,
                         __func__, memory_region_name(section->mr),
                         section->offset_within_address_space,
                         section->offset_within_region,
                         qemu_real_host_page_size());
        }
        return;
    }

    iova = REAL_HOST_PAGE_ALIGN(section->offset_within_address_space);
    llend = int128_make64(section->offset_within_address_space);
    llend = int128_add(llend, section->size);
    llend = int128_and(llend, int128_exts64(qemu_real_host_page_mask()));

    if (int128_ge(int128_make64(iova), llend)) {
        if (memory_region_is_ram_device(section->mr)) {
            trace_vfio_listener_region_add_no_dma_map(
                memory_region_name(section->mr),
                section->offset_within_address_space,
                int128_getlo(section->size),
                qemu_real_host_page_size());
        }
        return;
    }
    end = int128_get64(int128_sub(llend, int128_one()));

    if (container->iommu_type == VFIO_SPAPR_TCE_v2_IOMMU) {
        hwaddr pgsize = 0;

        /* For now intersections are not allowed, we may relax this later */
        QLIST_FOREACH(hostwin, &container->hostwin_list, hostwin_next) {
            if (ranges_overlap(hostwin->min_iova,
                               hostwin->max_iova - hostwin->min_iova + 1,
                               section->offset_within_address_space,
                               int128_get64(section->size))) {
                error_setg(&err,
                    "region [0x%"PRIx64",0x%"PRIx64"] overlaps with existing"
                    "host DMA window [0x%"PRIx64",0x%"PRIx64"]",
                    section->offset_within_address_space,
                    section->offset_within_address_space +
                        int128_get64(section->size) - 1,
                    hostwin->min_iova, hostwin->max_iova);
                goto fail;
            }
        }

        ret = vfio_spapr_create_window(container, section, &pgsize);
        if (ret) {
            error_setg_errno(&err, -ret, "Failed to create SPAPR window");
            goto fail;
        }

        vfio_host_win_add(container, section->offset_within_address_space,
                          section->offset_within_address_space +
                          int128_get64(section->size) - 1, pgsize);
#ifdef CONFIG_KVM
        if (kvm_enabled()) {
            VFIOGroup *group;
            IOMMUMemoryRegion *iommu_mr = IOMMU_MEMORY_REGION(section->mr);
            struct kvm_vfio_spapr_tce param;
            struct kvm_device_attr attr = {
                .group = KVM_DEV_VFIO_GROUP,
                .attr = KVM_DEV_VFIO_GROUP_SET_SPAPR_TCE,
                .addr = (uint64_t)(unsigned long)&param,
            };

            if (!memory_region_iommu_get_attr(iommu_mr, IOMMU_ATTR_SPAPR_TCE_FD,
                                              &param.tablefd)) {
                QLIST_FOREACH(group, &container->group_list, container_next) {
                    param.groupfd = group->fd;
                    if (ioctl(vfio_kvm_device_fd, KVM_SET_DEVICE_ATTR, &attr)) {
                        error_report("vfio: failed to setup fd %d "
                                     "for a group with fd %d: %s",
                                     param.tablefd, param.groupfd,
                                     strerror(errno));
                        return;
                    }
                    trace_vfio_spapr_group_attach(param.groupfd, param.tablefd);
                }
            }
        }
#endif
    }

    hostwin_found = false;
    QLIST_FOREACH(hostwin, &container->hostwin_list, hostwin_next) {
        if (hostwin->min_iova <= iova && end <= hostwin->max_iova) {
            hostwin_found = true;
            break;
        }
    }

    if (!hostwin_found) {
        error_setg(&err, "Container %p can't map guest IOVA region"
                   " 0x%"HWADDR_PRIx"..0x%"HWADDR_PRIx, container, iova, end);
        goto fail;
    }

    memory_region_ref(section->mr);

    if (memory_region_is_iommu(section->mr)) {
        VFIOGuestIOMMU *giommu;
        IOMMUMemoryRegion *iommu_mr = IOMMU_MEMORY_REGION(section->mr);
        int iommu_idx;

        trace_vfio_listener_region_add_iommu(iova, end);
        /*
         * FIXME: For VFIO iommu types which have KVM acceleration to
         * avoid bouncing all map/unmaps through qemu this way, this
         * would be the right place to wire that up (tell the KVM
         * device emulation the VFIO iommu handles to use).
         */
        giommu = g_malloc0(sizeof(*giommu));
        giommu->iommu_mr = iommu_mr;
        giommu->iommu_offset = section->offset_within_address_space -
                               section->offset_within_region;
        giommu->container = container;
        llend = int128_add(int128_make64(section->offset_within_region),
                           section->size);
        llend = int128_sub(llend, int128_one());
        iommu_idx = memory_region_iommu_attrs_to_index(iommu_mr,
                                                       MEMTXATTRS_UNSPECIFIED);
        iommu_notifier_init(&giommu->n, vfio_iommu_map_notify,
                            IOMMU_NOTIFIER_IOTLB_EVENTS,
                            section->offset_within_region,
                            int128_get64(llend),
                            iommu_idx);

        ret = memory_region_iommu_set_page_size_mask(giommu->iommu_mr,
                                                     container->pgsizes,
                                                     &err);
        if (ret) {
            g_free(giommu);
            goto fail;
        }

        ret = memory_region_register_iommu_notifier(section->mr, &giommu->n,
                                                    &err);
        if (ret) {
            g_free(giommu);
            goto fail;
        }
        QLIST_INSERT_HEAD(&container->giommu_list, giommu, giommu_next);
        memory_region_iommu_replay(giommu->iommu_mr, &giommu->n);

        return;
    }

    /* Here we assume that memory_region_is_ram(section->mr)==true */

    /*
     * For RAM memory regions with a RamDiscardManager, we only want to map the
     * actually populated parts - and update the mapping whenever we're notified
     * about changes.
     */
    if (memory_region_has_ram_discard_manager(section->mr)) {
        vfio_register_ram_discard_listener(container, section);
        return;
    }

    vaddr = memory_region_get_ram_ptr(section->mr) +
            section->offset_within_region +
            (iova - section->offset_within_address_space);

    trace_vfio_listener_region_add_ram(iova, end, vaddr);

    llsize = int128_sub(llend, int128_make64(iova));

    if (memory_region_is_ram_device(section->mr)) {
        hwaddr pgmask = (1ULL << ctz64(hostwin->iova_pgsizes)) - 1;

        if ((iova & pgmask) || (int128_get64(llsize) & pgmask)) {
            trace_vfio_listener_region_add_no_dma_map(
                memory_region_name(section->mr),
                section->offset_within_address_space,
                int128_getlo(section->size),
                pgmask + 1);
            return;
        }
    }

    ret = vfio_dma_map(container, iova, int128_get64(llsize),
                       vaddr, section->readonly);
    if (ret) {
        error_setg(&err, "vfio_dma_map(%p, 0x%"HWADDR_PRIx", "
                   "0x%"HWADDR_PRIx", %p) = %d (%m)",
                   container, iova, int128_get64(llsize), vaddr, ret);
        if (memory_region_is_ram_device(section->mr)) {
            /* Allow unexpected mappings not to be fatal for RAM devices */
            error_report_err(err);
            return;
        }
        goto fail;
    }

    return;

fail:
    if (memory_region_is_ram_device(section->mr)) {
        error_report("failed to vfio_dma_map. pci p2p may not work");
        return;
    }
    /*
     * On the initfn path, store the first error in the container so we
     * can gracefully fail.  Runtime, there's not much we can do other
     * than throw a hardware error.
     */
    if (!container->initialized) {
        if (!container->error) {
            error_propagate_prepend(&container->error, err,
                                    "Region %s: ",
                                    memory_region_name(section->mr));
        } else {
            error_free(err);
        }
    } else {
        error_report_err(err);
        hw_error("vfio: DMA mapping failed, unable to continue");
    }
}

static void vfio_listener_region_del(MemoryListener *listener,
                                     MemoryRegionSection *section)
{
    VFIOContainer *container = container_of(listener, VFIOContainer, listener);
    hwaddr iova, end;
    Int128 llend, llsize;
    int ret;
    bool try_unmap = true;

    if (vfio_listener_skipped_section(section)) {
        trace_vfio_listener_region_del_skip(
                section->offset_within_address_space,
                section->offset_within_address_space +
                int128_get64(int128_sub(section->size, int128_one())));
        return;
    }

    if (unlikely((section->offset_within_address_space &
                  ~qemu_real_host_page_mask()) !=
                 (section->offset_within_region & ~qemu_real_host_page_mask()))) {
        if (!vfio_known_safe_misalignment(section)) {
            error_report("%s received unaligned region %s iova=0x%"PRIx64
                         " offset_within_region=0x%"PRIx64
                         " qemu_real_host_page_size=0x%"PRIxPTR,
                         __func__, memory_region_name(section->mr),
                         section->offset_within_address_space,
                         section->offset_within_region,
                         qemu_real_host_page_size());
        }
        return;
    }

    if (memory_region_is_iommu(section->mr)) {
        VFIOGuestIOMMU *giommu;

        QLIST_FOREACH(giommu, &container->giommu_list, giommu_next) {
            if (MEMORY_REGION(giommu->iommu_mr) == section->mr &&
                giommu->n.start == section->offset_within_region) {
                memory_region_unregister_iommu_notifier(section->mr,
                                                        &giommu->n);
                QLIST_REMOVE(giommu, giommu_next);
                g_free(giommu);
                break;
            }
        }

        /*
         * FIXME: We assume the one big unmap below is adequate to
         * remove any individual page mappings in the IOMMU which
         * might have been copied into VFIO. This works for a page table
         * based IOMMU where a big unmap flattens a large range of IO-PTEs.
         * That may not be true for all IOMMU types.
         */
    }

    iova = REAL_HOST_PAGE_ALIGN(section->offset_within_address_space);
    llend = int128_make64(section->offset_within_address_space);
    llend = int128_add(llend, section->size);
    llend = int128_and(llend, int128_exts64(qemu_real_host_page_mask()));

    if (int128_ge(int128_make64(iova), llend)) {
        return;
    }
    end = int128_get64(int128_sub(llend, int128_one()));

    llsize = int128_sub(llend, int128_make64(iova));

    trace_vfio_listener_region_del(iova, end);

    if (memory_region_is_ram_device(section->mr)) {
        hwaddr pgmask;
        VFIOHostDMAWindow *hostwin;
        bool hostwin_found = false;

        QLIST_FOREACH(hostwin, &container->hostwin_list, hostwin_next) {
            if (hostwin->min_iova <= iova && end <= hostwin->max_iova) {
                hostwin_found = true;
                break;
            }
        }
        assert(hostwin_found); /* or region_add() would have failed */

        pgmask = (1ULL << ctz64(hostwin->iova_pgsizes)) - 1;
        try_unmap = !((iova & pgmask) || (int128_get64(llsize) & pgmask));
    } else if (memory_region_has_ram_discard_manager(section->mr)) {
        vfio_unregister_ram_discard_listener(container, section);
        /* Unregistering will trigger an unmap. */
        try_unmap = false;
    }

    if (try_unmap) {
        if (int128_eq(llsize, int128_2_64())) {
            /* The unmap ioctl doesn't accept a full 64-bit span. */
            llsize = int128_rshift(llsize, 1);
            ret = vfio_dma_unmap(container, iova, int128_get64(llsize), NULL);
            if (ret) {
                error_report("vfio_dma_unmap(%p, 0x%"HWADDR_PRIx", "
                             "0x%"HWADDR_PRIx") = %d (%m)",
                             container, iova, int128_get64(llsize), ret);
            }
            iova += int128_get64(llsize);
        }
        ret = vfio_dma_unmap(container, iova, int128_get64(llsize), NULL);
        if (ret) {
            error_report("vfio_dma_unmap(%p, 0x%"HWADDR_PRIx", "
                         "0x%"HWADDR_PRIx") = %d (%m)",
                         container, iova, int128_get64(llsize), ret);
        }
    }

    memory_region_unref(section->mr);

    if (container->iommu_type == VFIO_SPAPR_TCE_v2_IOMMU) {
        vfio_spapr_remove_window(container,
                                 section->offset_within_address_space);
        if (vfio_host_win_del(container,
                              section->offset_within_address_space,
                              section->offset_within_address_space +
                              int128_get64(section->size) - 1) < 0) {
            hw_error("%s: Cannot delete missing window at %"HWADDR_PRIx,
                     __func__, section->offset_within_address_space);
        }
    }
}

static void vfio_set_dirty_page_tracking(VFIOContainer *container, bool start)
{
    int ret;
    struct vfio_iommu_type1_dirty_bitmap dirty = {
        .argsz = sizeof(dirty),
    };

    if (!container->dirty_pages_supported) {
        return;
    }

    if (start) {
        dirty.flags = VFIO_IOMMU_DIRTY_PAGES_FLAG_START;
    } else {
        dirty.flags = VFIO_IOMMU_DIRTY_PAGES_FLAG_STOP;
    }

    ret = ioctl(container->fd, VFIO_IOMMU_DIRTY_PAGES, &dirty);
    if (ret) {
        error_report("Failed to set dirty tracking flag 0x%x errno: %d",
                     dirty.flags, errno);
    }
}

static void vfio_listener_log_global_start(MemoryListener *listener)
{
    VFIOContainer *container = container_of(listener, VFIOContainer, listener);

    vfio_set_dirty_page_tracking(container, true);
}

static void vfio_listener_log_global_stop(MemoryListener *listener)
{
    VFIOContainer *container = container_of(listener, VFIOContainer, listener);

    vfio_set_dirty_page_tracking(container, false);
}

static int vfio_get_dirty_bitmap(VFIOContainer *container, uint64_t iova,
                                 uint64_t size, ram_addr_t ram_addr)
{
    struct vfio_iommu_type1_dirty_bitmap *dbitmap;
    struct vfio_iommu_type1_dirty_bitmap_get *range;
    uint64_t pages;
    int ret;

    if (!container->dirty_pages_supported) {
        cpu_physical_memory_set_dirty_range(ram_addr, size,
                                            tcg_enabled() ? DIRTY_CLIENTS_ALL :
                                            DIRTY_CLIENTS_NOCODE);
        return 0;
    }

    dbitmap = g_malloc0(sizeof(*dbitmap) + sizeof(*range));

    dbitmap->argsz = sizeof(*dbitmap) + sizeof(*range);
    dbitmap->flags = VFIO_IOMMU_DIRTY_PAGES_FLAG_GET_BITMAP;
    range = (struct vfio_iommu_type1_dirty_bitmap_get *)&dbitmap->data;
    range->iova = iova;
    range->size = size;

    /*
     * cpu_physical_memory_set_dirty_lebitmap() supports pages in bitmap of
     * qemu_real_host_page_size to mark those dirty. Hence set bitmap's pgsize
     * to qemu_real_host_page_size.
     */
    range->bitmap.pgsize = qemu_real_host_page_size();

    pages = REAL_HOST_PAGE_ALIGN(range->size) / qemu_real_host_page_size();
    range->bitmap.size = ROUND_UP(pages, sizeof(__u64) * BITS_PER_BYTE) /
                                         BITS_PER_BYTE;
    range->bitmap.data = g_try_malloc0(range->bitmap.size);
    if (!range->bitmap.data) {
        ret = -ENOMEM;
        goto err_out;
    }

    ret = ioctl(container->fd, VFIO_IOMMU_DIRTY_PAGES, dbitmap);
    if (ret) {
        error_report("Failed to get dirty bitmap for iova: 0x%"PRIx64
                " size: 0x%"PRIx64" err: %d", (uint64_t)range->iova,
                (uint64_t)range->size, errno);
        goto err_out;
    }

    cpu_physical_memory_set_dirty_lebitmap((unsigned long *)range->bitmap.data,
                                            ram_addr, pages);

    trace_vfio_get_dirty_bitmap(container->fd, range->iova, range->size,
                                range->bitmap.size, ram_addr);
err_out:
    g_free(range->bitmap.data);
    g_free(dbitmap);

    return ret;
}

typedef struct {
    IOMMUNotifier n;
    VFIOGuestIOMMU *giommu;
} vfio_giommu_dirty_notifier;

static void vfio_iommu_map_dirty_notify(IOMMUNotifier *n, IOMMUTLBEntry *iotlb)
{
    vfio_giommu_dirty_notifier *gdn = container_of(n,
                                                vfio_giommu_dirty_notifier, n);
    VFIOGuestIOMMU *giommu = gdn->giommu;
    VFIOContainer *container = giommu->container;
    hwaddr iova = iotlb->iova + giommu->iommu_offset;
    ram_addr_t translated_addr;

    trace_vfio_iommu_map_dirty_notify(iova, iova + iotlb->addr_mask);

    if (iotlb->target_as != &address_space_memory) {
        error_report("Wrong target AS \"%s\", only system memory is allowed",
                     iotlb->target_as->name ? iotlb->target_as->name : "none");
        return;
    }

    rcu_read_lock();
    if (vfio_get_xlat_addr(iotlb, NULL, &translated_addr, NULL)) {
        int ret;

        ret = vfio_get_dirty_bitmap(container, iova, iotlb->addr_mask + 1,
                                    translated_addr);
        if (ret) {
            error_report("vfio_iommu_map_dirty_notify(%p, 0x%"HWADDR_PRIx", "
                         "0x%"HWADDR_PRIx") = %d (%m)",
                         container, iova,
                         iotlb->addr_mask + 1, ret);
        }
    }
    rcu_read_unlock();
}

static int vfio_ram_discard_get_dirty_bitmap(MemoryRegionSection *section,
                                             void *opaque)
{
    const hwaddr size = int128_get64(section->size);
    const hwaddr iova = section->offset_within_address_space;
    const ram_addr_t ram_addr = memory_region_get_ram_addr(section->mr) +
                                section->offset_within_region;
    VFIORamDiscardListener *vrdl = opaque;

    /*
     * Sync the whole mapped region (spanning multiple individual mappings)
     * in one go.
     */
    return vfio_get_dirty_bitmap(vrdl->container, iova, size, ram_addr);
}

static int vfio_sync_ram_discard_listener_dirty_bitmap(VFIOContainer *container,
                                                   MemoryRegionSection *section)
{
    RamDiscardManager *rdm = memory_region_get_ram_discard_manager(section->mr);
    VFIORamDiscardListener *vrdl = NULL;

    QLIST_FOREACH(vrdl, &container->vrdl_list, next) {
        if (vrdl->mr == section->mr &&
            vrdl->offset_within_address_space ==
            section->offset_within_address_space) {
            break;
        }
    }

    if (!vrdl) {
        hw_error("vfio: Trying to sync missing RAM discard listener");
    }

    /*
     * We only want/can synchronize the bitmap for actually mapped parts -
     * which correspond to populated parts. Replay all populated parts.
     */
    return ram_discard_manager_replay_populated(rdm, section,
                                              vfio_ram_discard_get_dirty_bitmap,
                                                &vrdl);
}

static int vfio_sync_dirty_bitmap(VFIOContainer *container,
                                  MemoryRegionSection *section)
{
    ram_addr_t ram_addr;

    if (memory_region_is_iommu(section->mr)) {
        VFIOGuestIOMMU *giommu;

        QLIST_FOREACH(giommu, &container->giommu_list, giommu_next) {
            if (MEMORY_REGION(giommu->iommu_mr) == section->mr &&
                giommu->n.start == section->offset_within_region) {
                Int128 llend;
                vfio_giommu_dirty_notifier gdn = { .giommu = giommu };
                int idx = memory_region_iommu_attrs_to_index(giommu->iommu_mr,
                                                       MEMTXATTRS_UNSPECIFIED);

                llend = int128_add(int128_make64(section->offset_within_region),
                                   section->size);
                llend = int128_sub(llend, int128_one());

                iommu_notifier_init(&gdn.n,
                                    vfio_iommu_map_dirty_notify,
                                    IOMMU_NOTIFIER_MAP,
                                    section->offset_within_region,
                                    int128_get64(llend),
                                    idx);
                memory_region_iommu_replay(giommu->iommu_mr, &gdn.n);
                break;
            }
        }
        return 0;
    } else if (memory_region_has_ram_discard_manager(section->mr)) {
        return vfio_sync_ram_discard_listener_dirty_bitmap(container, section);
    }

    ram_addr = memory_region_get_ram_addr(section->mr) +
               section->offset_within_region;

    return vfio_get_dirty_bitmap(container,
                   REAL_HOST_PAGE_ALIGN(section->offset_within_address_space),
                   int128_get64(section->size), ram_addr);
}

static void vfio_listener_log_sync(MemoryListener *listener,
        MemoryRegionSection *section)
{
    VFIOContainer *container = container_of(listener, VFIOContainer, listener);

    if (vfio_listener_skipped_section(section)) {
        return;
    }

    if (vfio_devices_all_dirty_tracking(container)) {
        vfio_sync_dirty_bitmap(container, section);
    }
}

static const MemoryListener vfio_memory_listener = {
    .name = "vfio",
    .region_add = vfio_listener_region_add,
    .region_del = vfio_listener_region_del,
    .log_global_start = vfio_listener_log_global_start,
    .log_global_stop = vfio_listener_log_global_stop,
    .log_sync = vfio_listener_log_sync,
};

static void vfio_listener_release(VFIOContainer *container)
{
    memory_listener_unregister(&container->listener);
    if (container->iommu_type == VFIO_SPAPR_TCE_v2_IOMMU) {
        memory_listener_unregister(&container->prereg_listener);
    }
}

static struct vfio_info_cap_header *
vfio_get_cap(void *ptr, uint32_t cap_offset, uint16_t id)
{
    struct vfio_info_cap_header *hdr;

    for (hdr = ptr + cap_offset; hdr != ptr; hdr = ptr + hdr->next) {
        if (hdr->id == id) {
            return hdr;
        }
    }

    return NULL;
}

struct vfio_info_cap_header *
vfio_get_region_info_cap(struct vfio_region_info *info, uint16_t id)
{
    if (!(info->flags & VFIO_REGION_INFO_FLAG_CAPS)) {
        return NULL;
    }

    return vfio_get_cap((void *)info, info->cap_offset, id);
}

static struct vfio_info_cap_header *
vfio_get_iommu_type1_info_cap(struct vfio_iommu_type1_info *info, uint16_t id)
{
    if (!(info->flags & VFIO_IOMMU_INFO_CAPS)) {
        return NULL;
    }

    return vfio_get_cap((void *)info, info->cap_offset, id);
}

struct vfio_info_cap_header *
vfio_get_device_info_cap(struct vfio_device_info *info, uint16_t id)
{
    if (!(info->flags & VFIO_DEVICE_FLAGS_CAPS)) {
        return NULL;
    }

    return vfio_get_cap((void *)info, info->cap_offset, id);
}

bool vfio_get_info_dma_avail(struct vfio_iommu_type1_info *info,
                             unsigned int *avail)
{
    struct vfio_info_cap_header *hdr;
    struct vfio_iommu_type1_info_dma_avail *cap;

    /* If the capability cannot be found, assume no DMA limiting */
    hdr = vfio_get_iommu_type1_info_cap(info,
                                        VFIO_IOMMU_TYPE1_INFO_DMA_AVAIL);
    if (hdr == NULL) {
        return false;
    }

    if (avail != NULL) {
        cap = (void *) hdr;
        *avail = cap->avail;
    }

    return true;
}

static int vfio_setup_region_sparse_mmaps(VFIORegion *region,
                                          struct vfio_region_info *info)
{
    struct vfio_info_cap_header *hdr;
    struct vfio_region_info_cap_sparse_mmap *sparse;
    int i, j;

    hdr = vfio_get_region_info_cap(info, VFIO_REGION_INFO_CAP_SPARSE_MMAP);
    if (!hdr) {
        return -ENODEV;
    }

    sparse = container_of(hdr, struct vfio_region_info_cap_sparse_mmap, header);

    trace_vfio_region_sparse_mmap_header(region->vbasedev->name,
                                         region->nr, sparse->nr_areas);

    region->mmaps = g_new0(VFIOMmap, sparse->nr_areas);

    for (i = 0, j = 0; i < sparse->nr_areas; i++) {
        if (sparse->areas[i].size) {
            trace_vfio_region_sparse_mmap_entry(i, sparse->areas[i].offset,
                                            sparse->areas[i].offset +
                                            sparse->areas[i].size - 1);
            region->mmaps[j].offset = sparse->areas[i].offset;
            region->mmaps[j].size = sparse->areas[i].size;
            j++;
        }
    }

    region->nr_mmaps = j;
    region->mmaps = g_realloc(region->mmaps, j * sizeof(VFIOMmap));

    return 0;
}

int vfio_region_setup(Object *obj, VFIODevice *vbasedev, VFIORegion *region,
                      int index, const char *name)
{
    struct vfio_region_info *info;
    int ret;

    ret = vfio_get_region_info(vbasedev, index, &info);
    if (ret) {
        return ret;
    }

    region->vbasedev = vbasedev;
    region->flags = info->flags;
    region->size = info->size;
    region->fd_offset = info->offset;
    region->nr = index;

    if (region->size) {
        region->mem = g_new0(MemoryRegion, 1);
        memory_region_init_io(region->mem, obj, &vfio_region_ops,
                              region, name, region->size);

        if (!vbasedev->no_mmap &&
            region->flags & VFIO_REGION_INFO_FLAG_MMAP) {

            ret = vfio_setup_region_sparse_mmaps(region, info);

            if (ret) {
                region->nr_mmaps = 1;
                region->mmaps = g_new0(VFIOMmap, region->nr_mmaps);
                region->mmaps[0].offset = 0;
                region->mmaps[0].size = region->size;
            }
        }
    }

    g_free(info);

    trace_vfio_region_setup(vbasedev->name, index, name,
                            region->flags, region->fd_offset, region->size);
    return 0;
}

static void vfio_subregion_unmap(VFIORegion *region, int index)
{
    trace_vfio_region_unmap(memory_region_name(&region->mmaps[index].mem),
                            region->mmaps[index].offset,
                            region->mmaps[index].offset +
                            region->mmaps[index].size - 1);
    memory_region_del_subregion(region->mem, &region->mmaps[index].mem);
    munmap(region->mmaps[index].mmap, region->mmaps[index].size);
    object_unparent(OBJECT(&region->mmaps[index].mem));
    region->mmaps[index].mmap = NULL;
}

int vfio_region_mmap(VFIORegion *region)
{
    int i, prot = 0;
    char *name;

    if (!region->mem) {
        return 0;
    }

    prot |= region->flags & VFIO_REGION_INFO_FLAG_READ ? PROT_READ : 0;
    prot |= region->flags & VFIO_REGION_INFO_FLAG_WRITE ? PROT_WRITE : 0;

    for (i = 0; i < region->nr_mmaps; i++) {
        region->mmaps[i].mmap = mmap(NULL, region->mmaps[i].size, prot,
                                     MAP_SHARED, region->vbasedev->fd,
                                     region->fd_offset +
                                     region->mmaps[i].offset);
        if (region->mmaps[i].mmap == MAP_FAILED) {
            int ret = -errno;

            trace_vfio_region_mmap_fault(memory_region_name(region->mem), i,
                                         region->fd_offset +
                                         region->mmaps[i].offset,
                                         region->fd_offset +
                                         region->mmaps[i].offset +
                                         region->mmaps[i].size - 1, ret);

            region->mmaps[i].mmap = NULL;

            for (i--; i >= 0; i--) {
                vfio_subregion_unmap(region, i);
            }

            return ret;
        }

        name = g_strdup_printf("%s mmaps[%d]",
                               memory_region_name(region->mem), i);
        memory_region_init_ram_device_ptr(&region->mmaps[i].mem,
                                          memory_region_owner(region->mem),
                                          name, region->mmaps[i].size,
                                          region->mmaps[i].mmap);
        g_free(name);
        memory_region_add_subregion(region->mem, region->mmaps[i].offset,
                                    &region->mmaps[i].mem);

        trace_vfio_region_mmap(memory_region_name(&region->mmaps[i].mem),
                               region->mmaps[i].offset,
                               region->mmaps[i].offset +
                               region->mmaps[i].size - 1);
    }

    return 0;
}

void vfio_region_unmap(VFIORegion *region)
{
    int i;

    if (!region->mem) {
        return;
    }

    for (i = 0; i < region->nr_mmaps; i++) {
        if (region->mmaps[i].mmap) {
            vfio_subregion_unmap(region, i);
        }
    }
}

void vfio_region_exit(VFIORegion *region)
{
    int i;

    if (!region->mem) {
        return;
    }

    for (i = 0; i < region->nr_mmaps; i++) {
        if (region->mmaps[i].mmap) {
            memory_region_del_subregion(region->mem, &region->mmaps[i].mem);
        }
    }

    trace_vfio_region_exit(region->vbasedev->name, region->nr);
}

void vfio_region_finalize(VFIORegion *region)
{
    int i;

    if (!region->mem) {
        return;
    }

    for (i = 0; i < region->nr_mmaps; i++) {
        if (region->mmaps[i].mmap) {
            munmap(region->mmaps[i].mmap, region->mmaps[i].size);
            object_unparent(OBJECT(&region->mmaps[i].mem));
        }
    }

    object_unparent(OBJECT(region->mem));

    g_free(region->mem);
    g_free(region->mmaps);

    trace_vfio_region_finalize(region->vbasedev->name, region->nr);

    region->mem = NULL;
    region->mmaps = NULL;
    region->nr_mmaps = 0;
    region->size = 0;
    region->flags = 0;
    region->nr = 0;
}

void vfio_region_mmaps_set_enabled(VFIORegion *region, bool enabled)
{
    int i;

    if (!region->mem) {
        return;
    }

    for (i = 0; i < region->nr_mmaps; i++) {
        if (region->mmaps[i].mmap) {
            memory_region_set_enabled(&region->mmaps[i].mem, enabled);
        }
    }

    trace_vfio_region_mmaps_set_enabled(memory_region_name(region->mem),
                                        enabled);
}

void vfio_reset_handler(void *opaque)
{
    VFIOGroup *group;
    VFIODevice *vbasedev;

    QLIST_FOREACH(group, &vfio_group_list, next) {
        QLIST_FOREACH(vbasedev, &group->device_list, next) {
            if (vbasedev->dev->realized) {
                vbasedev->ops->vfio_compute_needs_reset(vbasedev);
            }
        }
    }

    QLIST_FOREACH(group, &vfio_group_list, next) {
        QLIST_FOREACH(vbasedev, &group->device_list, next) {
            if (vbasedev->dev->realized && vbasedev->needs_reset) {
                vbasedev->ops->vfio_hot_reset_multi(vbasedev);
            }
        }
    }
}

static void vfio_kvm_device_add_group(VFIOGroup *group)
{
#ifdef CONFIG_KVM
    struct kvm_device_attr attr = {
        .group = KVM_DEV_VFIO_GROUP,
        .attr = KVM_DEV_VFIO_GROUP_ADD,
        .addr = (uint64_t)(unsigned long)&group->fd,
    };

    if (!kvm_enabled()) {
        return;
    }

    if (vfio_kvm_device_fd < 0) {
        struct kvm_create_device cd = {
            .type = KVM_DEV_TYPE_VFIO,
        };

        if (kvm_vm_ioctl(kvm_state, KVM_CREATE_DEVICE, &cd)) {
            error_report("Failed to create KVM VFIO device: %m");
            return;
        }

        vfio_kvm_device_fd = cd.fd;
    }

    if (ioctl(vfio_kvm_device_fd, KVM_SET_DEVICE_ATTR, &attr)) {
        error_report("Failed to add group %d to KVM VFIO device: %m",
                     group->groupid);
    }
#endif
}

static void vfio_kvm_device_del_group(VFIOGroup *group)
{
#ifdef CONFIG_KVM
    struct kvm_device_attr attr = {
        .group = KVM_DEV_VFIO_GROUP,
        .attr = KVM_DEV_VFIO_GROUP_DEL,
        .addr = (uint64_t)(unsigned long)&group->fd,
    };

    if (vfio_kvm_device_fd < 0) {
        return;
    }

    if (ioctl(vfio_kvm_device_fd, KVM_SET_DEVICE_ATTR, &attr)) {
        error_report("Failed to remove group %d from KVM VFIO device: %m",
                     group->groupid);
    }
#endif
}

static VFIOAddressSpace *vfio_get_address_space(AddressSpace *as)
{
    VFIOAddressSpace *space;

    QLIST_FOREACH(space, &vfio_address_spaces, list) {
        if (space->as == as) {
            return space;
        }
    }

    /* No suitable VFIOAddressSpace, create a new one */
    space = g_malloc0(sizeof(*space));
    space->as = as;
    QLIST_INIT(&space->containers);

    QLIST_INSERT_HEAD(&vfio_address_spaces, space, list);

    return space;
}

static void vfio_put_address_space(VFIOAddressSpace *space)
{
    if (QLIST_EMPTY(&space->containers)) {
        QLIST_REMOVE(space, list);
        g_free(space);
    }
}

/*
 * vfio_get_iommu_type - selects the richest iommu_type (v2 first)
 */
static int vfio_get_iommu_type(VFIOContainer *container,
                               Error **errp)
{
    int iommu_types[] = { VFIO_TYPE1v2_IOMMU, VFIO_TYPE1_IOMMU,
                          VFIO_SPAPR_TCE_v2_IOMMU, VFIO_SPAPR_TCE_IOMMU };
    int i;

    for (i = 0; i < ARRAY_SIZE(iommu_types); i++) {
        if (ioctl(container->fd, VFIO_CHECK_EXTENSION, iommu_types[i])) {
            return iommu_types[i];
        }
    }
    error_setg(errp, "No available IOMMU models");
    return -EINVAL;
}

static int vfio_init_container(VFIOContainer *container, int group_fd,
                               Error **errp)
{
    int iommu_type, ret;

    iommu_type = vfio_get_iommu_type(container, errp);
    if (iommu_type < 0) {
        return iommu_type;
    }

    ret = ioctl(group_fd, VFIO_GROUP_SET_CONTAINER, &container->fd);
    if (ret) {
        error_setg_errno(errp, errno, "Failed to set group container");
        return -errno;
    }

    while (ioctl(container->fd, VFIO_SET_IOMMU, iommu_type)) {
        if (iommu_type == VFIO_SPAPR_TCE_v2_IOMMU) {
            /*
             * On sPAPR, despite the IOMMU subdriver always advertises v1 and
             * v2, the running platform may not support v2 and there is no
             * way to guess it until an IOMMU group gets added to the container.
             * So in case it fails with v2, try v1 as a fallback.
             */
            iommu_type = VFIO_SPAPR_TCE_IOMMU;
            continue;
        }
        error_setg_errno(errp, errno, "Failed to set iommu for container");
        return -errno;
    }

    container->iommu_type = iommu_type;
    return 0;
}

static int vfio_get_iommu_info(VFIOContainer *container,
                               struct vfio_iommu_type1_info **info)
{

    size_t argsz = sizeof(struct vfio_iommu_type1_info);

    *info = g_new0(struct vfio_iommu_type1_info, 1);
again:
    (*info)->argsz = argsz;

    if (ioctl(container->fd, VFIO_IOMMU_GET_INFO, *info)) {
        g_free(*info);
        *info = NULL;
        return -errno;
    }

    if (((*info)->argsz > argsz)) {
        argsz = (*info)->argsz;
        *info = g_realloc(*info, argsz);
        goto again;
    }

    return 0;
}

static struct vfio_info_cap_header *
vfio_get_iommu_info_cap(struct vfio_iommu_type1_info *info, uint16_t id)
{
    struct vfio_info_cap_header *hdr;
    void *ptr = info;

    if (!(info->flags & VFIO_IOMMU_INFO_CAPS)) {
        return NULL;
    }

    for (hdr = ptr + info->cap_offset; hdr != ptr; hdr = ptr + hdr->next) {
        if (hdr->id == id) {
            return hdr;
        }
    }

    return NULL;
}

static void vfio_get_iommu_info_migration(VFIOContainer *container,
                                         struct vfio_iommu_type1_info *info)
{
    struct vfio_info_cap_header *hdr;
    struct vfio_iommu_type1_info_cap_migration *cap_mig;

    hdr = vfio_get_iommu_info_cap(info, VFIO_IOMMU_TYPE1_INFO_CAP_MIGRATION);
    if (!hdr) {
        return;
    }

    cap_mig = container_of(hdr, struct vfio_iommu_type1_info_cap_migration,
                            header);

    /*
     * cpu_physical_memory_set_dirty_lebitmap() supports pages in bitmap of
     * qemu_real_host_page_size to mark those dirty.
     */
    if (cap_mig->pgsize_bitmap & qemu_real_host_page_size()) {
        container->dirty_pages_supported = true;
        container->max_dirty_bitmap_size = cap_mig->max_dirty_bitmap_size;
        container->dirty_pgsizes = cap_mig->pgsize_bitmap;
    }
}

static int vfio_connect_container(VFIOGroup *group, AddressSpace *as,
                                  Error **errp)
{
    VFIOContainer *container;
    int ret, fd;
    VFIOAddressSpace *space;

    space = vfio_get_address_space(as);

    /*
     * VFIO is currently incompatible with discarding of RAM insofar as the
     * madvise to purge (zap) the page from QEMU's address space does not
     * interact with the memory API and therefore leaves stale virtual to
     * physical mappings in the IOMMU if the page was previously pinned.  We
     * therefore set discarding broken for each group added to a container,
     * whether the container is used individually or shared.  This provides
     * us with options to allow devices within a group to opt-in and allow
     * discarding, so long as it is done consistently for a group (for instance
     * if the device is an mdev device where it is known that the host vendor
     * driver will never pin pages outside of the working set of the guest
     * driver, which would thus not be discarding candidates).
     *
     * The first opportunity to induce pinning occurs here where we attempt to
     * attach the group to existing containers within the AddressSpace.  If any
     * pages are already zapped from the virtual address space, such as from
     * previous discards, new pinning will cause valid mappings to be
     * re-established.  Likewise, when the overall MemoryListener for a new
     * container is registered, a replay of mappings within the AddressSpace
     * will occur, re-establishing any previously zapped pages as well.
     *
     * Especially virtio-balloon is currently only prevented from discarding
     * new memory, it will not yet set ram_block_discard_set_required() and
     * therefore, neither stops us here or deals with the sudden memory
     * consumption of inflated memory.
     *
     * We do support discarding of memory coordinated via the RamDiscardManager
     * with some IOMMU types. vfio_ram_block_discard_disable() handles the
     * details once we know which type of IOMMU we are using.
     */

    QLIST_FOREACH(container, &space->containers, next) {
        if (!ioctl(group->fd, VFIO_GROUP_SET_CONTAINER, &container->fd)) {
            ret = vfio_ram_block_discard_disable(container, true);
            if (ret) {
                error_setg_errno(errp, -ret,
                                 "Cannot set discarding of RAM broken");
                if (ioctl(group->fd, VFIO_GROUP_UNSET_CONTAINER,
                          &container->fd)) {
                    error_report("vfio: error disconnecting group %d from"
                                 " container", group->groupid);
                }
                return ret;
            }
            group->container = container;
            QLIST_INSERT_HEAD(&container->group_list, group, container_next);
            vfio_kvm_device_add_group(group);
            return 0;
        }
    }

    fd = qemu_open_old("/dev/vfio/vfio", O_RDWR);
    if (fd < 0) {
        error_setg_errno(errp, errno, "failed to open /dev/vfio/vfio");
        ret = -errno;
        goto put_space_exit;
    }

    ret = ioctl(fd, VFIO_GET_API_VERSION);
    if (ret != VFIO_API_VERSION) {
        error_setg(errp, "supported vfio version: %d, "
                   "reported version: %d", VFIO_API_VERSION, ret);
        ret = -EINVAL;
        goto close_fd_exit;
    }

    container = g_malloc0(sizeof(*container));
    container->space = space;
    container->fd = fd;
    container->error = NULL;
    container->dirty_pages_supported = false;
    container->dma_max_mappings = 0;
    QLIST_INIT(&container->giommu_list);
    QLIST_INIT(&container->hostwin_list);
    QLIST_INIT(&container->vrdl_list);

    ret = vfio_init_container(container, group->fd, errp);
    if (ret) {
        goto free_container_exit;
    }

    ret = vfio_ram_block_discard_disable(container, true);
    if (ret) {
        error_setg_errno(errp, -ret, "Cannot set discarding of RAM broken");
        goto free_container_exit;
    }

    switch (container->iommu_type) {
    case VFIO_TYPE1v2_IOMMU:
    case VFIO_TYPE1_IOMMU:
    {
        struct vfio_iommu_type1_info *info;

        ret = vfio_get_iommu_info(container, &info);
        if (ret) {
            error_setg_errno(errp, -ret, "Failed to get VFIO IOMMU info");
            goto enable_discards_exit;
        }

        if (info->flags & VFIO_IOMMU_INFO_PGSIZES) {
            container->pgsizes = info->iova_pgsizes;
        } else {
            container->pgsizes = qemu_real_host_page_size();
        }

        if (!vfio_get_info_dma_avail(info, &container->dma_max_mappings)) {
            container->dma_max_mappings = 65535;
        }
        vfio_get_iommu_info_migration(container, info);
        g_free(info);

        /*
         * FIXME: We should parse VFIO_IOMMU_TYPE1_INFO_CAP_IOVA_RANGE
         * information to get the actual window extent rather than assume
         * a 64-bit IOVA address space.
         */
        vfio_host_win_add(container, 0, (hwaddr)-1, container->pgsizes);

        break;
    }
    case VFIO_SPAPR_TCE_v2_IOMMU:
    case VFIO_SPAPR_TCE_IOMMU:
    {
        struct vfio_iommu_spapr_tce_info info;
        bool v2 = container->iommu_type == VFIO_SPAPR_TCE_v2_IOMMU;

        /*
         * The host kernel code implementing VFIO_IOMMU_DISABLE is called
         * when container fd is closed so we do not call it explicitly
         * in this file.
         */
        if (!v2) {
            ret = ioctl(fd, VFIO_IOMMU_ENABLE);
            if (ret) {
                error_setg_errno(errp, errno, "failed to enable container");
                ret = -errno;
                goto enable_discards_exit;
            }
        } else {
            container->prereg_listener = vfio_prereg_listener;

            memory_listener_register(&container->prereg_listener,
                                     &address_space_memory);
            if (container->error) {
                memory_listener_unregister(&container->prereg_listener);
                ret = -1;
                error_propagate_prepend(errp, container->error,
                    "RAM memory listener initialization failed: ");
                goto enable_discards_exit;
            }
        }

        info.argsz = sizeof(info);
        ret = ioctl(fd, VFIO_IOMMU_SPAPR_TCE_GET_INFO, &info);
        if (ret) {
            error_setg_errno(errp, errno,
                             "VFIO_IOMMU_SPAPR_TCE_GET_INFO failed");
            ret = -errno;
            if (v2) {
                memory_listener_unregister(&container->prereg_listener);
            }
            goto enable_discards_exit;
        }

        if (v2) {
            container->pgsizes = info.ddw.pgsizes;
            /*
             * There is a default window in just created container.
             * To make region_add/del simpler, we better remove this
             * window now and let those iommu_listener callbacks
             * create/remove them when needed.
             */
            ret = vfio_spapr_remove_window(container, info.dma32_window_start);
            if (ret) {
                error_setg_errno(errp, -ret,
                                 "failed to remove existing window");
                goto enable_discards_exit;
            }
        } else {
            /* The default table uses 4K pages */
            container->pgsizes = 0x1000;
            vfio_host_win_add(container, info.dma32_window_start,
                              info.dma32_window_start +
                              info.dma32_window_size - 1,
                              0x1000);
        }
    }
    }

    vfio_kvm_device_add_group(group);

    QLIST_INIT(&container->group_list);
    QLIST_INSERT_HEAD(&space->containers, container, next);

    group->container = container;
    QLIST_INSERT_HEAD(&container->group_list, group, container_next);

    container->listener = vfio_memory_listener;

    memory_listener_register(&container->listener, container->space->as);

    if (container->error) {
        ret = -1;
        error_propagate_prepend(errp, container->error,
            "memory listener initialization failed: ");
        goto listener_release_exit;
    }

    container->initialized = true;

    return 0;
listener_release_exit:
    QLIST_REMOVE(group, container_next);
    QLIST_REMOVE(container, next);
    vfio_kvm_device_del_group(group);
    vfio_listener_release(container);

enable_discards_exit:
    vfio_ram_block_discard_disable(container, false);

free_container_exit:
    g_free(container);

close_fd_exit:
    close(fd);

put_space_exit:
    vfio_put_address_space(space);

    return ret;
}

static void vfio_disconnect_container(VFIOGroup *group)
{
    VFIOContainer *container = group->container;

    QLIST_REMOVE(group, container_next);
    group->container = NULL;

    /*
     * Explicitly release the listener first before unset container,
     * since unset may destroy the backend container if it's the last
     * group.
     */
    if (QLIST_EMPTY(&container->group_list)) {
        vfio_listener_release(container);
    }

    if (ioctl(group->fd, VFIO_GROUP_UNSET_CONTAINER, &container->fd)) {
        error_report("vfio: error disconnecting group %d from container",
                     group->groupid);
    }

    if (QLIST_EMPTY(&container->group_list)) {
        VFIOAddressSpace *space = container->space;
        VFIOGuestIOMMU *giommu, *tmp;
        VFIOHostDMAWindow *hostwin, *next;

        QLIST_REMOVE(container, next);

        QLIST_FOREACH_SAFE(giommu, &container->giommu_list, giommu_next, tmp) {
            memory_region_unregister_iommu_notifier(
                    MEMORY_REGION(giommu->iommu_mr), &giommu->n);
            QLIST_REMOVE(giommu, giommu_next);
            g_free(giommu);
        }

        QLIST_FOREACH_SAFE(hostwin, &container->hostwin_list, hostwin_next,
                           next) {
            QLIST_REMOVE(hostwin, hostwin_next);
            g_free(hostwin);
        }

        trace_vfio_disconnect_container(container->fd);
        close(container->fd);
        g_free(container);

        vfio_put_address_space(space);
    }
}

VFIOGroup *vfio_get_group(int groupid, AddressSpace *as, Error **errp)
{
    VFIOGroup *group;
    char path[32];
    struct vfio_group_status status = { .argsz = sizeof(status) };

    QLIST_FOREACH(group, &vfio_group_list, next) {
        if (group->groupid == groupid) {
            /* Found it.  Now is it already in the right context? */
            if (group->container->space->as == as) {
                return group;
            } else {
                error_setg(errp, "group %d used in multiple address spaces",
                           group->groupid);
                return NULL;
            }
        }
    }

    group = g_malloc0(sizeof(*group));

    snprintf(path, sizeof(path), "/dev/vfio/%d", groupid);
    group->fd = qemu_open_old(path, O_RDWR);
    if (group->fd < 0) {
        error_setg_errno(errp, errno, "failed to open %s", path);
        goto free_group_exit;
    }

    if (ioctl(group->fd, VFIO_GROUP_GET_STATUS, &status)) {
        error_setg_errno(errp, errno, "failed to get group %d status", groupid);
        goto close_fd_exit;
    }

    if (!(status.flags & VFIO_GROUP_FLAGS_VIABLE)) {
        error_setg(errp, "group %d is not viable", groupid);
        error_append_hint(errp,
                          "Please ensure all devices within the iommu_group "
                          "are bound to their vfio bus driver.\n");
        goto close_fd_exit;
    }

    group->groupid = groupid;
    QLIST_INIT(&group->device_list);

    if (vfio_connect_container(group, as, errp)) {
        error_prepend(errp, "failed to setup container for group %d: ",
                      groupid);
        goto close_fd_exit;
    }

    if (QLIST_EMPTY(&vfio_group_list)) {
        qemu_register_reset(vfio_reset_handler, NULL);
    }

    QLIST_INSERT_HEAD(&vfio_group_list, group, next);

    return group;

close_fd_exit:
    close(group->fd);

free_group_exit:
    g_free(group);

    return NULL;
}

void vfio_put_group(VFIOGroup *group)
{
    if (!group || !QLIST_EMPTY(&group->device_list)) {
        return;
    }

    if (!group->ram_block_discard_allowed) {
        vfio_ram_block_discard_disable(group->container, false);
    }
    vfio_kvm_device_del_group(group);
    vfio_disconnect_container(group);
    QLIST_REMOVE(group, next);
    trace_vfio_put_group(group->fd);
    close(group->fd);
    g_free(group);

    if (QLIST_EMPTY(&vfio_group_list)) {
        qemu_unregister_reset(vfio_reset_handler, NULL);
    }
}

int vfio_get_device(VFIOGroup *group, const char *name,
                    VFIODevice *vbasedev, Error **errp)
{
    struct vfio_device_info dev_info = { .argsz = sizeof(dev_info) };
    int ret, fd;

    fd = ioctl(group->fd, VFIO_GROUP_GET_DEVICE_FD, name);
    if (fd < 0) {
        error_setg_errno(errp, errno, "error getting device from group %d",
                         group->groupid);
        error_append_hint(errp,
                      "Verify all devices in group %d are bound to vfio-<bus> "
                      "or pci-stub and not already in use\n", group->groupid);
        return fd;
    }

    ret = ioctl(fd, VFIO_DEVICE_GET_INFO, &dev_info);
    if (ret) {
        error_setg_errno(errp, errno, "error getting device info");
        close(fd);
        return ret;
    }

    /*
     * Set discarding of RAM as not broken for this group if the driver knows
     * the device operates compatibly with discarding.  Setting must be
     * consistent per group, but since compatibility is really only possible
     * with mdev currently, we expect singleton groups.
     */
    if (vbasedev->ram_block_discard_allowed !=
        group->ram_block_discard_allowed) {
        if (!QLIST_EMPTY(&group->device_list)) {
            error_setg(errp, "Inconsistent setting of support for discarding "
                       "RAM (e.g., balloon) within group");
            close(fd);
            return -1;
        }

        if (!group->ram_block_discard_allowed) {
            group->ram_block_discard_allowed = true;
            vfio_ram_block_discard_disable(group->container, false);
        }
    }

    vbasedev->fd = fd;
    vbasedev->group = group;
    QLIST_INSERT_HEAD(&group->device_list, vbasedev, next);

    vbasedev->num_irqs = dev_info.num_irqs;
    vbasedev->num_regions = dev_info.num_regions;
    vbasedev->flags = dev_info.flags;

    trace_vfio_get_device(name, dev_info.flags, dev_info.num_regions,
                          dev_info.num_irqs);

    vbasedev->reset_works = !!(dev_info.flags & VFIO_DEVICE_FLAGS_RESET);
    return 0;
}

void vfio_put_base_device(VFIODevice *vbasedev)
{
    if (!vbasedev->group) {
        return;
    }
    QLIST_REMOVE(vbasedev, next);
    vbasedev->group = NULL;
    trace_vfio_put_base_device(vbasedev->fd);
    close(vbasedev->fd);
}

int vfio_get_region_info(VFIODevice *vbasedev, int index,
                         struct vfio_region_info **info)
{
    size_t argsz = sizeof(struct vfio_region_info);

    *info = g_malloc0(argsz);

    (*info)->index = index;
retry:
    (*info)->argsz = argsz;

    if (ioctl(vbasedev->fd, VFIO_DEVICE_GET_REGION_INFO, *info)) {
        g_free(*info);
        *info = NULL;
        return -errno;
    }

    if ((*info)->argsz > argsz) {
        argsz = (*info)->argsz;
        *info = g_realloc(*info, argsz);

        goto retry;
    }

    return 0;
}

int vfio_get_dev_region_info(VFIODevice *vbasedev, uint32_t type,
                             uint32_t subtype, struct vfio_region_info **info)
{
    int i;

    for (i = 0; i < vbasedev->num_regions; i++) {
        struct vfio_info_cap_header *hdr;
        struct vfio_region_info_cap_type *cap_type;

        if (vfio_get_region_info(vbasedev, i, info)) {
            continue;
        }

        hdr = vfio_get_region_info_cap(*info, VFIO_REGION_INFO_CAP_TYPE);
        if (!hdr) {
            g_free(*info);
            continue;
        }

        cap_type = container_of(hdr, struct vfio_region_info_cap_type, header);

        trace_vfio_get_dev_region(vbasedev->name, i,
                                  cap_type->type, cap_type->subtype);

        if (cap_type->type == type && cap_type->subtype == subtype) {
            return 0;
        }

        g_free(*info);
    }

    *info = NULL;
    return -ENODEV;
}

bool vfio_has_region_cap(VFIODevice *vbasedev, int region, uint16_t cap_type)
{
    struct vfio_region_info *info = NULL;
    bool ret = false;

    if (!vfio_get_region_info(vbasedev, region, &info)) {
        if (vfio_get_region_info_cap(info, cap_type)) {
            ret = true;
        }
        g_free(info);
    }

    return ret;
}

/*
 * Interfaces for IBM EEH (Enhanced Error Handling)
 */
static bool vfio_eeh_container_ok(VFIOContainer *container)
{
    /*
     * As of 2016-03-04 (linux-4.5) the host kernel EEH/VFIO
     * implementation is broken if there are multiple groups in a
     * container.  The hardware works in units of Partitionable
     * Endpoints (== IOMMU groups) and the EEH operations naively
     * iterate across all groups in the container, without any logic
     * to make sure the groups have their state synchronized.  For
     * certain operations (ENABLE) that might be ok, until an error
     * occurs, but for others (GET_STATE) it's clearly broken.
     */

    /*
     * XXX Once fixed kernels exist, test for them here
     */

    if (QLIST_EMPTY(&container->group_list)) {
        return false;
    }

    if (QLIST_NEXT(QLIST_FIRST(&container->group_list), container_next)) {
        return false;
    }

    return true;
}

static int vfio_eeh_container_op(VFIOContainer *container, uint32_t op)
{
    struct vfio_eeh_pe_op pe_op = {
        .argsz = sizeof(pe_op),
        .op = op,
    };
    int ret;

    if (!vfio_eeh_container_ok(container)) {
        error_report("vfio/eeh: EEH_PE_OP 0x%x: "
                     "kernel requires a container with exactly one group", op);
        return -EPERM;
    }

    ret = ioctl(container->fd, VFIO_EEH_PE_OP, &pe_op);
    if (ret < 0) {
        error_report("vfio/eeh: EEH_PE_OP 0x%x failed: %m", op);
        return -errno;
    }

    return ret;
}

static VFIOContainer *vfio_eeh_as_container(AddressSpace *as)
{
    VFIOAddressSpace *space = vfio_get_address_space(as);
    VFIOContainer *container = NULL;

    if (QLIST_EMPTY(&space->containers)) {
        /* No containers to act on */
        goto out;
    }

    container = QLIST_FIRST(&space->containers);

    if (QLIST_NEXT(container, next)) {
        /* We don't yet have logic to synchronize EEH state across
         * multiple containers */
        container = NULL;
        goto out;
    }

out:
    vfio_put_address_space(space);
    return container;
}

bool vfio_eeh_as_ok(AddressSpace *as)
{
    VFIOContainer *container = vfio_eeh_as_container(as);

    return (container != NULL) && vfio_eeh_container_ok(container);
}

int vfio_eeh_as_op(AddressSpace *as, uint32_t op)
{
    VFIOContainer *container = vfio_eeh_as_container(as);

    if (!container) {
        return -ENODEV;
    }
    return vfio_eeh_container_op(container, op);
}