Re: [PATCH v9 QEMU 13/15] vfio: Add vfio_listener_log_sync to mark dirty pages

[Alex Williamson <alex.williamson@redhat.com>]

On Tue, 12 Nov 2019 22:35:22 +0530
Kirti Wankhede <kwankhede@nvidia.com> wrote:

> vfio_listener_log_sync gets list of dirty pages from container using
> VFIO_IOMMU_GET_DIRTY_BITMAP ioctl and mark those pages dirty when all
> devices are stopped and saving state.
> Return early for the RAM block section of mapped MMIO region.
> 
> Signed-off-by: Kirti Wankhede <kwankhede@nvidia.com>
> Reviewed-by: Neo Jia <cjia@nvidia.com>
> ---
>  hw/vfio/common.c     | 103 +++++++++++++++++++++++++++++++++++++++++++++++++++
>  hw/vfio/trace-events |   1 +
>  2 files changed, 104 insertions(+)
> 
> diff --git a/hw/vfio/common.c b/hw/vfio/common.c
> index ade9839c28a3..66f1c64bf074 100644
> --- a/hw/vfio/common.c
> +++ b/hw/vfio/common.c
> @@ -29,6 +29,7 @@
>  #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"
> @@ -38,6 +39,7 @@
>  #include "sysemu/reset.h"
>  #include "trace.h"
>  #include "qapi/error.h"
> +#include "migration/migration.h"
>  
>  VFIOGroupList vfio_group_list =
>      QLIST_HEAD_INITIALIZER(vfio_group_list);
> @@ -288,6 +290,28 @@ const MemoryRegionOps vfio_region_ops = {
>  };
>  
>  /*
> + * Device state interfaces
> + */
> +
> +static bool vfio_devices_are_stopped_and_saving(void)
> +{
> +    VFIOGroup *group;
> +    VFIODevice *vbasedev;
> +
> +    QLIST_FOREACH(group, &vfio_group_list, next) {
> +        QLIST_FOREACH(vbasedev, &group->device_list, next) {
> +            if ((vbasedev->device_state & VFIO_DEVICE_STATE_SAVING) &&
> +                !(vbasedev->device_state & VFIO_DEVICE_STATE_RUNNING)) {

(device_state & MASK) == SAVING

> +                continue;

Kind of silly to have a continue rather than just changing the polarity
of the test so that we only branch into the return case.

> +            } else {
> +                return false;
> +            }
> +        }
> +    }
> +    return true;
> +}
> +
> +/*
>   * DMA - Mapping and unmapping for the "type1" IOMMU interface used on x86
>   */
>  static int vfio_dma_unmap(VFIOContainer *container,
> @@ -813,9 +837,88 @@ static void vfio_listener_region_del(MemoryListener *listener,
>      }
>  }
>  
> +static int vfio_get_dirty_bitmap(VFIOContainer *container,
> +                                 MemoryRegionSection *section)
> +{
> +    struct vfio_iommu_type1_dirty_bitmap range;
> +    uint64_t bitmap_size;
> +    int ret;
> +
> +    range.argsz = sizeof(range);
> +
> +    if (memory_region_is_iommu(section->mr)) {
> +        VFIOGuestIOMMU *giommu;
> +        IOMMUTLBEntry iotlb;
> +
> +        QLIST_FOREACH(giommu, &container->giommu_list, giommu_next) {
> +            if (MEMORY_REGION(giommu->iommu) == section->mr &&
> +                giommu->n.start == section->offset_within_region) {
> +                break;
> +            }
> +        }
> +
> +        if (!giommu) {
> +            return -EINVAL;
> +        }
> +
> +        iotlb = address_space_get_iotlb_entry(container->space->as,
> +                       TARGET_PAGE_ALIGN(section->offset_within_address_space),
> +                       true, MEMTXATTRS_UNSPECIFIED);
> +        range.iova = iotlb.iova + giommu->iommu_offset;
> +        range.size = iotlb.addr_mask + 1;
> +    } else {
> +        range.iova = TARGET_PAGE_ALIGN(section->offset_within_address_space);
> +        range.size = int128_get64(section->size);
> +    }
> +
> +    bitmap_size = BITS_TO_LONGS(range.size >> TARGET_PAGE_BITS) *
> +                                                             sizeof(uint64_t);
> +
> +    range.bitmap = g_try_malloc0(bitmap_size);
> +    if (!range.bitmap) {
> +        error_report("%s: Error allocating bitmap buffer of size 0x%lx",
> +                     __func__, bitmap_size);
> +        return -ENOMEM;

We could certainly iterate with a smaller bitmap rather than use a
single ioctl.  This doesn't seem like it scales well as VM memory size
increases.

> +    }
> +
> +    range.bitmap_size = bitmap_size;
> +
> +    ret = ioctl(container->fd, VFIO_IOMMU_GET_DIRTY_BITMAP, &range);
> +
> +    if (!ret) {
> +        cpu_physical_memory_set_dirty_lebitmap((uint64_t *)range.bitmap,
> +                       TARGET_PAGE_ALIGN(section->offset_within_address_space),
> +                       bitmap_size >> TARGET_PAGE_BITS);


Like Yan, I think this is relative to the iova address space and needs
a translation for the vIOMMU case.

> +    } else {
> +        error_report("VFIO_IOMMU_GET_DIRTY_BITMAP: %d %d", ret, errno);
> +    }
> +
> +    trace_vfio_get_dirty_bitmap(container->fd, range.iova, range.size,
> +                                bitmap_size);
> +
> +    g_free(range.bitmap);
> +    return ret;
> +}
> +
> +static void vfio_listerner_log_sync(MemoryListener *listener,
> +        MemoryRegionSection *section)
> +{
> +    VFIOContainer *container = container_of(listener, VFIOContainer, listener);
> +
> +    if (memory_region_is_ram_device(section->mr)) {
> +        return;
> +    }
> +
> +    if (vfio_devices_are_stopped_and_saving()) {

I think a comment is necessary here indicating why we're not
participating in an iterative dirty bitmap sync.  Additionally, how
will the kernel indicate that we can support real dirty tracking?

> +
> +        vfio_get_dirty_bitmap(container, section);
> +    }
> +}
> +
>  static const MemoryListener vfio_memory_listener = {
>      .region_add = vfio_listener_region_add,
>      .region_del = vfio_listener_region_del,
> +    .log_sync = vfio_listerner_log_sync,
>  };
>  
>  static void vfio_listener_release(VFIOContainer *container)
> diff --git a/hw/vfio/trace-events b/hw/vfio/trace-events
> index ac065b559f4e..0dd1f2ffe648 100644
> --- a/hw/vfio/trace-events
> +++ b/hw/vfio/trace-events
> @@ -160,3 +160,4 @@ vfio_save_complete_precopy(char *name) " (%s)"
>  vfio_load_device_config_state(char *name) " (%s)"
>  vfio_load_state(char *name, uint64_t data) " (%s) data 0x%"PRIx64
>  vfio_load_state_device_data(char *name, uint64_t data_offset, uint64_t data_size) " (%s) Offset 0x%"PRIx64" size 0x%"PRIx64
> +vfio_get_dirty_bitmap(int fd, uint64_t iova, uint64_t size, uint64_t bitmap_size) "container fd=%d, iova=0x%"PRIx64" size= 0x%"PRIx64" bitmap_size=0x%"PRIx64