Re: [PATCH v3 11/15] vfio: add region info cache

["Cédric Le Goater" <clg@redhat.com>]

On 5/7/25 17:20, John Levon wrote:
> Instead of requesting region information on demand with
> VFIO_DEVICE_GET_REGION_INFO, maintain a cache: this will become
> necessary for performance for vfio-user, where this call becomes a
> message over the control socket, so is of higher overhead than the
> traditional path.
> 
> We will also need it to generalize region accesses, as that means we
> can't use ->config_offset for configuration space accesses, but must
> look up the region offset (if relevant) each time.
> 
> Originally-by: John Johnson <john.g.johnson@oracle.com>
> Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
> Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
> Signed-off-by: John Levon <john.levon@nutanix.com>


Reviewed-by: Cédric Le Goater <clg@redhat.com>

Thanks,

C.


> ---
>   include/hw/vfio/vfio-device.h |  1 +
>   hw/vfio/ccw.c                 |  5 -----
>   hw/vfio/device.c              | 25 +++++++++++++++++++++----
>   hw/vfio/igd.c                 | 10 +++++-----
>   hw/vfio/pci.c                 |  6 +++---
>   hw/vfio/region.c              |  2 +-
>   6 files changed, 31 insertions(+), 18 deletions(-)
> 
> diff --git a/include/hw/vfio/vfio-device.h b/include/hw/vfio/vfio-device.h
> index 7e1e81e76b..4fff3dcee3 100644
> --- a/include/hw/vfio/vfio-device.h
> +++ b/include/hw/vfio/vfio-device.h
> @@ -83,6 +83,7 @@ typedef struct VFIODevice {
>       IOMMUFDBackend *iommufd;
>       VFIOIOASHwpt *hwpt;
>       QLIST_ENTRY(VFIODevice) hwpt_next;
> +    struct vfio_region_info **reginfo;
>   } VFIODevice;
>   
>   struct VFIODeviceOps {
> diff --git a/hw/vfio/ccw.c b/hw/vfio/ccw.c
> index ab3fabf991..cea9d6e005 100644
> --- a/hw/vfio/ccw.c
> +++ b/hw/vfio/ccw.c
> @@ -504,7 +504,6 @@ static bool vfio_ccw_get_region(VFIOCCWDevice *vcdev, Error **errp)
>   
>       vcdev->io_region_offset = info->offset;
>       vcdev->io_region = g_malloc0(info->size);
> -    g_free(info);
>   
>       /* check for the optional async command region */
>       ret = vfio_device_get_region_info_type(vdev, VFIO_REGION_TYPE_CCW,
> @@ -517,7 +516,6 @@ static bool vfio_ccw_get_region(VFIOCCWDevice *vcdev, Error **errp)
>           }
>           vcdev->async_cmd_region_offset = info->offset;
>           vcdev->async_cmd_region = g_malloc0(info->size);
> -        g_free(info);
>       }
>   
>       ret = vfio_device_get_region_info_type(vdev, VFIO_REGION_TYPE_CCW,
> @@ -530,7 +528,6 @@ static bool vfio_ccw_get_region(VFIOCCWDevice *vcdev, Error **errp)
>           }
>           vcdev->schib_region_offset = info->offset;
>           vcdev->schib_region = g_malloc(info->size);
> -        g_free(info);
>       }
>   
>       ret = vfio_device_get_region_info_type(vdev, VFIO_REGION_TYPE_CCW,
> @@ -544,7 +541,6 @@ static bool vfio_ccw_get_region(VFIOCCWDevice *vcdev, Error **errp)
>           }
>           vcdev->crw_region_offset = info->offset;
>           vcdev->crw_region = g_malloc(info->size);
> -        g_free(info);
>       }
>   
>       return true;
> @@ -554,7 +550,6 @@ out_err:
>       g_free(vcdev->schib_region);
>       g_free(vcdev->async_cmd_region);
>       g_free(vcdev->io_region);
> -    g_free(info);
>       return false;
>   }
>   
> diff --git a/hw/vfio/device.c b/hw/vfio/device.c
> index 40a196bfb9..77b0675abe 100644
> --- a/hw/vfio/device.c
> +++ b/hw/vfio/device.c
> @@ -202,6 +202,12 @@ int vfio_device_get_region_info(VFIODevice *vbasedev, int index,
>       size_t argsz = sizeof(struct vfio_region_info);
>       int ret;
>   
> +    /* check cache */
> +    if (vbasedev->reginfo[index] != NULL) {
> +        *info = vbasedev->reginfo[index];
> +        return 0;
> +    }
> +
>       *info = g_malloc0(argsz);
>   
>       (*info)->index = index;
> @@ -222,6 +228,9 @@ retry:
>           goto retry;
>       }
>   
> +    /* fill cache */
> +    vbasedev->reginfo[index] = *info;
> +
>       return 0;
>   }
>   
> @@ -240,7 +249,6 @@ int vfio_device_get_region_info_type(VFIODevice *vbasedev, uint32_t type,
>   
>           hdr = vfio_get_region_info_cap(*info, VFIO_REGION_INFO_CAP_TYPE);
>           if (!hdr) {
> -            g_free(*info);
>               continue;
>           }
>   
> @@ -252,8 +260,6 @@ int vfio_device_get_region_info_type(VFIODevice *vbasedev, uint32_t type,
>           if (cap_type->type == type && cap_type->subtype == subtype) {
>               return 0;
>           }
> -
> -        g_free(*info);
>       }
>   
>       *info = NULL;
> @@ -262,7 +268,7 @@ int vfio_device_get_region_info_type(VFIODevice *vbasedev, uint32_t type,
>   
>   bool vfio_device_has_region_cap(VFIODevice *vbasedev, int region, uint16_t cap_type)
>   {
> -    g_autofree struct vfio_region_info *info = NULL;
> +    struct vfio_region_info *info = NULL;
>       bool ret = false;
>   
>       if (!vfio_device_get_region_info(vbasedev, region, &info)) {
> @@ -435,10 +441,21 @@ void vfio_device_prepare(VFIODevice *vbasedev, VFIOContainerBase *bcontainer,
>       QLIST_INSERT_HEAD(&bcontainer->device_list, vbasedev, container_next);
>   
>       QLIST_INSERT_HEAD(&vfio_device_list, vbasedev, global_next);
> +
> +    vbasedev->reginfo = g_new0(struct vfio_region_info *,
> +                               vbasedev->num_regions);
>   }
>   
>   void vfio_device_unprepare(VFIODevice *vbasedev)
>   {
> +    int i;
> +
> +    for (i = 0; i < vbasedev->num_regions; i++) {
> +        g_free(vbasedev->reginfo[i]);
> +    }
> +    g_free(vbasedev->reginfo);
> +    vbasedev->reginfo = NULL;
> +
>       QLIST_REMOVE(vbasedev, container_next);
>       QLIST_REMOVE(vbasedev, global_next);
>       vbasedev->bcontainer = NULL;
> diff --git a/hw/vfio/igd.c b/hw/vfio/igd.c
> index 3ee1a73b57..e7952d15a0 100644
> --- a/hw/vfio/igd.c
> +++ b/hw/vfio/igd.c
> @@ -349,8 +349,8 @@ static int vfio_pci_igd_lpc_init(VFIOPCIDevice *vdev,
>   
>   static bool vfio_pci_igd_setup_lpc_bridge(VFIOPCIDevice *vdev, Error **errp)
>   {
> -    g_autofree struct vfio_region_info *host = NULL;
> -    g_autofree struct vfio_region_info *lpc = NULL;
> +    struct vfio_region_info *host = NULL;
> +    struct vfio_region_info *lpc = NULL;
>       PCIDevice *lpc_bridge;
>       int ret;
>   
> @@ -510,7 +510,7 @@ void vfio_probe_igd_bar0_quirk(VFIOPCIDevice *vdev, int nr)
>   
>   static bool vfio_pci_igd_config_quirk(VFIOPCIDevice *vdev, Error **errp)
>   {
> -    g_autofree struct vfio_region_info *opregion = NULL;
> +    struct vfio_region_info *opregion = NULL;
>       int ret, gen;
>       uint64_t gms_size = 0;
>       uint64_t *bdsm_size;
> @@ -551,7 +551,7 @@ static bool vfio_pci_igd_config_quirk(VFIOPCIDevice *vdev, Error **errp)
>            * - OpRegion
>            * - Same LPC bridge and Host bridge VID/DID/SVID/SSID as host
>            */
> -        g_autofree struct vfio_region_info *rom = NULL;
> +        struct vfio_region_info *rom = NULL;
>   
>           legacy_mode_enabled = true;
>           info_report("IGD legacy mode enabled, "
> @@ -681,7 +681,7 @@ error:
>    */
>   static bool vfio_pci_kvmgt_config_quirk(VFIOPCIDevice *vdev, Error **errp)
>   {
> -    g_autofree struct vfio_region_info *opregion = NULL;
> +    struct vfio_region_info *opregion = NULL;
>       int gen;
>   
>       if (!vfio_pci_is(vdev, PCI_VENDOR_ID_INTEL, PCI_ANY_ID) ||
> diff --git a/hw/vfio/pci.c b/hw/vfio/pci.c
> index da2ffc9bf3..9136cf52c8 100644
> --- a/hw/vfio/pci.c
> +++ b/hw/vfio/pci.c
> @@ -883,8 +883,8 @@ static void vfio_update_msi(VFIOPCIDevice *vdev)
>   
>   static void vfio_pci_load_rom(VFIOPCIDevice *vdev)
>   {
> -    g_autofree struct vfio_region_info *reg_info = NULL;
>       VFIODevice *vbasedev = &vdev->vbasedev;
> +    struct vfio_region_info *reg_info = NULL;
>       uint64_t size;
>       off_t off = 0;
>       ssize_t bytes;
> @@ -2710,7 +2710,7 @@ static VFIODeviceOps vfio_pci_ops = {
>   bool vfio_populate_vga(VFIOPCIDevice *vdev, Error **errp)
>   {
>       VFIODevice *vbasedev = &vdev->vbasedev;
> -    g_autofree struct vfio_region_info *reg_info = NULL;
> +    struct vfio_region_info *reg_info = NULL;
>       int ret;
>   
>       ret = vfio_device_get_region_info(vbasedev, VFIO_PCI_VGA_REGION_INDEX, &reg_info);
> @@ -2775,7 +2775,7 @@ bool vfio_populate_vga(VFIOPCIDevice *vdev, Error **errp)
>   static bool vfio_populate_device(VFIOPCIDevice *vdev, Error **errp)
>   {
>       VFIODevice *vbasedev = &vdev->vbasedev;
> -    g_autofree struct vfio_region_info *reg_info = NULL;
> +    struct vfio_region_info *reg_info = NULL;
>       struct vfio_irq_info irq_info;
>       int i, ret = -1;
>   
> diff --git a/hw/vfio/region.c b/hw/vfio/region.c
> index 04bf9eb098..ef2630cac3 100644
> --- a/hw/vfio/region.c
> +++ b/hw/vfio/region.c
> @@ -182,7 +182,7 @@ static int vfio_setup_region_sparse_mmaps(VFIORegion *region,
>   int vfio_region_setup(Object *obj, VFIODevice *vbasedev, VFIORegion *region,
>                         int index, const char *name)
>   {
> -    g_autofree struct vfio_region_info *info = NULL;
> +    struct vfio_region_info *info = NULL;
>       int ret;
>   
>       ret = vfio_device_get_region_info(vbasedev, index, &info);