diff options
Diffstat (limited to 'hw/vfio/pci.c')
| -rw-r--r-- | hw/vfio/pci.c | 67 |
1 files changed, 65 insertions, 2 deletions
diff --git a/hw/vfio/pci.c b/hw/vfio/pci.c index ec9a854..bf27a39 100644 --- a/hw/vfio/pci.c +++ b/hw/vfio/pci.c @@ -2066,6 +2066,54 @@ static int vfio_add_std_cap(VFIOPCIDevice *vdev, uint8_t pos, Error **errp) return 0; } +static int vfio_setup_rebar_ecap(VFIOPCIDevice *vdev, uint16_t pos) +{ + uint32_t ctrl; + int i, nbar; + + ctrl = pci_get_long(vdev->pdev.config + pos + PCI_REBAR_CTRL); + nbar = (ctrl & PCI_REBAR_CTRL_NBAR_MASK) >> PCI_REBAR_CTRL_NBAR_SHIFT; + + for (i = 0; i < nbar; i++) { + uint32_t cap; + int size; + + ctrl = pci_get_long(vdev->pdev.config + pos + PCI_REBAR_CTRL + (i * 8)); + size = (ctrl & PCI_REBAR_CTRL_BAR_SIZE) >> PCI_REBAR_CTRL_BAR_SHIFT; + + /* The cap register reports sizes 1MB to 128TB, with 4 reserved bits */ + cap = size <= 27 ? 1U << (size + 4) : 0; + + /* + * The PCIe spec (v6.0.1, 7.8.6) requires HW to support at least one + * size in the range 1MB to 512GB. We intend to mask all sizes except + * the one currently enabled in the size field, therefore if it's + * outside the range, hide the whole capability as this virtualization + * trick won't work. If >512GB resizable BARs start to appear, we + * might need an opt-in or reservation scheme in the kernel. + */ + if (!(cap & PCI_REBAR_CAP_SIZES)) { + return -EINVAL; + } + + /* Hide all sizes reported in the ctrl reg per above requirement. */ + ctrl &= (PCI_REBAR_CTRL_BAR_SIZE | + PCI_REBAR_CTRL_NBAR_MASK | + PCI_REBAR_CTRL_BAR_IDX); + + /* + * The BAR size field is RW, however we've mangled the capability + * register such that we only report a single size, ie. the current + * BAR size. A write of an unsupported value is undefined, therefore + * the register field is essentially RO. + */ + vfio_add_emulated_long(vdev, pos + PCI_REBAR_CAP + (i * 8), cap, ~0); + vfio_add_emulated_long(vdev, pos + PCI_REBAR_CTRL + (i * 8), ctrl, ~0); + } + + return 0; +} + static void vfio_add_ext_cap(VFIOPCIDevice *vdev) { PCIDevice *pdev = &vdev->pdev; @@ -2139,9 +2187,13 @@ static void vfio_add_ext_cap(VFIOPCIDevice *vdev) case 0: /* kernel masked capability */ case PCI_EXT_CAP_ID_SRIOV: /* Read-only VF BARs confuse OVMF */ case PCI_EXT_CAP_ID_ARI: /* XXX Needs next function virtualization */ - case PCI_EXT_CAP_ID_REBAR: /* Can't expose read-only */ trace_vfio_add_ext_cap_dropped(vdev->vbasedev.name, cap_id, next); break; + case PCI_EXT_CAP_ID_REBAR: + if (!vfio_setup_rebar_ecap(vdev, next)) { + pcie_add_capability(pdev, cap_id, cap_ver, next, size); + } + break; default: pcie_add_capability(pdev, cap_id, cap_ver, next, size); } @@ -2856,6 +2908,8 @@ static void vfio_realize(PCIDevice *pdev, Error **errp) int groupid; int i, ret; bool is_mdev; + char uuid[UUID_FMT_LEN]; + char *name; if (!vbasedev->sysfsdev) { if (!(~vdev->host.domain || ~vdev->host.bus || @@ -2936,7 +2990,15 @@ static void vfio_realize(PCIDevice *pdev, Error **errp) goto error; } - ret = vfio_get_device(group, vbasedev->name, vbasedev, errp); + if (!qemu_uuid_is_null(&vdev->vf_token)) { + qemu_uuid_unparse(&vdev->vf_token, uuid); + name = g_strdup_printf("%s vf_token=%s", vbasedev->name, uuid); + } else { + name = vbasedev->name; + } + + ret = vfio_get_device(group, name, vbasedev, errp); + g_free(name); if (ret) { vfio_put_group(group); goto error; @@ -3268,6 +3330,7 @@ static void vfio_instance_init(Object *obj) static Property vfio_pci_dev_properties[] = { DEFINE_PROP_PCI_HOST_DEVADDR("host", VFIOPCIDevice, host), + DEFINE_PROP_UUID_NODEFAULT("vf-token", VFIOPCIDevice, vf_token), DEFINE_PROP_STRING("sysfsdev", VFIOPCIDevice, vbasedev.sysfsdev), DEFINE_PROP_ON_OFF_AUTO("x-pre-copy-dirty-page-tracking", VFIOPCIDevice, vbasedev.pre_copy_dirty_page_tracking, |