motorhead/qemu
1
0
Fork 0
mirror of https://github.com/Motorhead1991/qemu.git synced 2025-08-05 16:53:55 -06:00
Code Issues Projects Releases Packages Wiki Activity Actions

spapr: Remove support for NVIDIA V100 GPU with NVLink2

Browse source 
NVLink2 support was removed from the PPC PowerNV platform and VFIO in
Linux 5.13 with commits :

  562d1e207d32 ("powerpc/powernv: remove the nvlink support")
  b392a1989170 ("vfio/pci: remove vfio_pci_nvlink2")

This was 2.5 years ago. Do the same in QEMU with a revert of commit
ec132efaa8 ("spapr: Support NVIDIA V100 GPU with NVLink2"). Some
adjustements are required on the NUMA part.

Cc: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Acked-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Message-ID: <20230918091717.149950-1-clg@kaod.org>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
This commit is contained in:
Cédric Le Goater 2023-09-18 11:17:17 +02:00 • committed by Daniel Henrique Barboza
parent 527b238329
commit 44fa20c928
11 changed files with 14 additions and 708 deletions
Download patch file Download diff file Expand all files Collapse all files

115
hw/vfio/pci-quirks.c
View file

@ -1610,121 +1610,6 @@ static int vfio_add_nv_gpudirect_cap(VFIOPCIDevice *vdev, Error **errp)
return 0;
}
int vfio_pci_nvidia_v100_ram_init(VFIOPCIDevice *vdev, Error **errp)
{
int ret;
void *p;
struct vfio_region_info *nv2reg = NULL;
struct vfio_info_cap_header *hdr;
struct vfio_region_info_cap_nvlink2_ssatgt *cap;
VFIOQuirk *quirk;
ret = vfio_get_dev_region_info(&vdev->vbasedev,
VFIO_REGION_TYPE_PCI_VENDOR_TYPE |
PCI_VENDOR_ID_NVIDIA,
VFIO_REGION_SUBTYPE_NVIDIA_NVLINK2_RAM,
&nv2reg);
if (ret) {
return ret;
}
hdr = vfio_get_region_info_cap(nv2reg, VFIO_REGION_INFO_CAP_NVLINK2_SSATGT);
if (!hdr) {
ret = -ENODEV;
goto free_exit;
}
cap = (void *) hdr;
p = mmap(NULL, nv2reg->size, PROT_READ | PROT_WRITE,
MAP_SHARED, vdev->vbasedev.fd, nv2reg->offset);
if (p == MAP_FAILED) {
ret = -errno;
goto free_exit;
}
quirk = vfio_quirk_alloc(1);
memory_region_init_ram_ptr(&quirk->mem[0], OBJECT(vdev), "nvlink2-mr",
nv2reg->size, p);
QLIST_INSERT_HEAD(&vdev->bars[0].quirks, quirk, next);
object_property_add_uint64_ptr(OBJECT(vdev), "nvlink2-tgt",
(uint64_t *) &cap->tgt,
OBJ_PROP_FLAG_READ);
trace_vfio_pci_nvidia_gpu_setup_quirk(vdev->vbasedev.name, cap->tgt,
nv2reg->size);
free_exit:
g_free(nv2reg);
return ret;
}
int vfio_pci_nvlink2_init(VFIOPCIDevice *vdev, Error **errp)
{
int ret;
void *p;
struct vfio_region_info *atsdreg = NULL;
struct vfio_info_cap_header *hdr;
struct vfio_region_info_cap_nvlink2_ssatgt *captgt;
struct vfio_region_info_cap_nvlink2_lnkspd *capspeed;
VFIOQuirk *quirk;
ret = vfio_get_dev_region_info(&vdev->vbasedev,
VFIO_REGION_TYPE_PCI_VENDOR_TYPE |
PCI_VENDOR_ID_IBM,
VFIO_REGION_SUBTYPE_IBM_NVLINK2_ATSD,
&atsdreg);
if (ret) {
return ret;
}
hdr = vfio_get_region_info_cap(atsdreg,
VFIO_REGION_INFO_CAP_NVLINK2_SSATGT);
if (!hdr) {
ret = -ENODEV;
goto free_exit;
}
captgt = (void *) hdr;
hdr = vfio_get_region_info_cap(atsdreg,
VFIO_REGION_INFO_CAP_NVLINK2_LNKSPD);
if (!hdr) {
ret = -ENODEV;
goto free_exit;
}
capspeed = (void *) hdr;
/* Some NVLink bridges may not have assigned ATSD */
if (atsdreg->size) {
p = mmap(NULL, atsdreg->size, PROT_READ | PROT_WRITE,
MAP_SHARED, vdev->vbasedev.fd, atsdreg->offset);
if (p == MAP_FAILED) {
ret = -errno;
goto free_exit;
}
quirk = vfio_quirk_alloc(1);
memory_region_init_ram_device_ptr(&quirk->mem[0], OBJECT(vdev),
"nvlink2-atsd-mr", atsdreg->size, p);
QLIST_INSERT_HEAD(&vdev->bars[0].quirks, quirk, next);
}
object_property_add_uint64_ptr(OBJECT(vdev), "nvlink2-tgt",
(uint64_t *) &captgt->tgt,
OBJ_PROP_FLAG_READ);
trace_vfio_pci_nvlink2_setup_quirk_ssatgt(vdev->vbasedev.name, captgt->tgt,
atsdreg->size);
object_property_add_uint32_ptr(OBJECT(vdev), "nvlink2-link-speed",
&capspeed->link_speed,
OBJ_PROP_FLAG_READ);
trace_vfio_pci_nvlink2_setup_quirk_lnkspd(vdev->vbasedev.name,
capspeed->link_speed);
free_exit:
g_free(atsdreg);
return ret;
}
/*
* The VMD endpoint provides a real PCIe domain to the guest and the guest
* kernel performs enumeration of the VMD sub-device domain. Guest transactions