On 10/18/18 9:55 AM, Alex Williamson wrote:
On Thu, 18 Oct 2018 11:31:33 +1100
Alexey Kardashevskiy <aik@xxxxxxxxx> wrote:
On 18/10/2018 08:52, Alex Williamson wrote:
On Wed, 17 Oct 2018 12:19:20 +1100
Alexey Kardashevskiy <aik@xxxxxxxxx> wrote:
On 17/10/2018 06:08, Alex Williamson wrote:
On Mon, 15 Oct 2018 20:42:33 +1100
Alexey Kardashevskiy <aik@xxxxxxxxx> wrote:
POWER9 Witherspoon machines come with 4 or 6 V100 GPUs which are not
pluggable PCIe devices but implement PCIe links for config space and MMIO.
In addition to that the GPUs are interconnected to each other and also
have direct links to the P9 CPU. The links are NVLink2 and provide direct
access to the system RAM for GPUs via NPU (an NVLink2 "proxy" on P9 chip).
These systems also support ATS (address translation services) which is
a part of the NVLink2 prototol. Such GPUs also share on-board RAM
(16GB in tested config) to the system via the same NVLink2 so a CPU has
cache-coherent access to a GPU RAM.
This exports GPU RAM to the userspace as a new PCI region. This
preregisters the new memory as device memory as it might be used for DMA.
This inserts pfns from the fault handler as the GPU memory is not onlined
until the NVIDIA driver is loaded and trained the links so doing this
earlier produces low level errors which we fence in the firmware so
it does not hurt the host system but still better to avoid.
This exports ATSD (Address Translation Shootdown) register of NPU which
allows the guest to invalidate TLB. The register conviniently occupies
a single 64k page. Since NPU maps the GPU memory, it has a "tgt" property
(which is an abbreviated host system bus address). This exports the "tgt"
as a capability so the guest can program it into the GPU so the GPU can
know how to route DMA trafic.
I'm not really following what "tgt" is and why it's needed. Is the GPU
memory here different than the GPU RAM region above? Why does the user
need the host system bus address of this "tgt" thing? Are we not able
to relocate it in guest physical address space, does this shootdown
only work in the host physical address space and therefore we need this
offset? Please explain, I'm confused.
This "tgt" is made of:
- "memory select" (bits 45, 46)
- "group select" (bits 43, 44)
- "chip select" (bit 42)
- chip internal address (bits 0..41)
These are internal to GPU and this is where GPU RAM is mapped into the
GPU's real space, this fits 46 bits.
On POWER9 CPU the bits are different and higher so the same memory is
mapped higher on P9 CPU. Just because we can map it higher, I guess.
So it is not exactly the address but this provides the exact physical
location of the memory.
We have a group of 3 interconnected GPUs, they got their own
memory/group/chip numbers. The GPUs use ATS service to translate
userspace to physical (host or guest) addresses. Now a GPU needs to know
which specific link to use for a specific physical address, in other
words what this physical address belongs to - a CPU or one of GPUs. This
is when "tgt" is used by the GPU hardware.
Clear as mud ;)
/me is sad. I hope Piotr explained it better...
It's starting to be a bit more clear, and Piotr anticipated the
security questions I was mulling over.
Great.
So tgt, provided by the npu2 capability of the ATSD
region of the NPU tells the GPU (a completely separate device) how to
route it its own RAM via its NVLink interface? How can one tgt
indicate the routing for multiple interfaces?
This NVLink DMA is using direct host physical addresses (no IOMMU, no
filtering) which come from ATS. So unless we tell the GPU its own
address range on the host CPU, it will route trafic via CPU. And the
driver can also discover the NVLink topology and tell each GPU physical
addresses of peer GPUs.
I think this is a key point too, the tgt seems to only identify the
routing for the GPU local RAM, but the guest driver is able to
conglomerate this information so each GPU knows how to get directly to
the other GPUs.
Yes.
A GPU could run all the DMA trafic via the system bus indeed, just not
as fast.
I am also struggling here and adding an Nvidia person in cc: (I should
have done that when I posted the patches, my bad) to correct when/if I
am wrong.
For ATS to work, the nest MMU (an NVIDIA block in a P9 CPU) needs to
know LPID (a logical partition ID or a KVM guest hardware ID in other
words) and PID (a memory context ID of an userspace process, not to be
confused with a linux pid). This assigns a GPU to LPID in the NPU and
this is why this adds a listener for KVM on an IOMMU group. A PID comes
via NVLink from a GPU and NPU uses a PID wildcard to pass it through.
This requires coherent memory and ATSD to be available on the host as
the GPU vendor only supports configurations with both features enabled
and other configurations are known not to work. Because of this and
because of the ways the features are advertised to the host system
(which is a device tree with very platform specific properties),
this requires enabled POWERNV platform.
This hardcodes the NVLink2 support for specific vendor and device IDs
as there is no reliable way of knowing about coherent memory and ATS
support. The GPU has an unique vendor PCIe capability 0x23 but it was
confirmed that it does not provide required information (and it is still
undisclosed what it actually does).
Signed-off-by: Alexey Kardashevskiy <aik@xxxxxxxxx>
---
drivers/vfio/pci/Makefile | 1 +
drivers/vfio/pci/vfio_pci_private.h | 2 +
include/uapi/linux/vfio.h | 18 ++
drivers/vfio/pci/vfio_pci.c | 37 +++-
drivers/vfio/pci/vfio_pci_nvlink2.c | 409 ++++++++++++++++++++++++++++++++++++
drivers/vfio/pci/Kconfig | 4 +
6 files changed, 469 insertions(+), 2 deletions(-)
create mode 100644 drivers/vfio/pci/vfio_pci_nvlink2.c
diff --git a/drivers/vfio/pci/Makefile b/drivers/vfio/pci/Makefile
index 76d8ec0..9662c06 100644
--- a/drivers/vfio/pci/Makefile
+++ b/drivers/vfio/pci/Makefile
@@ -1,5 +1,6 @@
vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
+vfio-pci-$(CONFIG_VFIO_PCI_NVLINK2) += vfio_pci_nvlink2.o
obj-$(CONFIG_VFIO_PCI) += vfio-pci.o
diff --git a/drivers/vfio/pci/vfio_pci_private.h b/drivers/vfio/pci/vfio_pci_private.h
index 93c1738..7639241 100644
--- a/drivers/vfio/pci/vfio_pci_private.h
+++ b/drivers/vfio/pci/vfio_pci_private.h
@@ -163,4 +163,6 @@ static inline int vfio_pci_igd_init(struct vfio_pci_device *vdev)
return -ENODEV;
}
#endif
+extern int vfio_pci_nvdia_v100_nvlink2_init(struct vfio_pci_device *vdev);
+extern int vfio_pci_ibm_npu2_init(struct vfio_pci_device *vdev);
#endif /* VFIO_PCI_PRIVATE_H */
diff --git a/include/uapi/linux/vfio.h b/include/uapi/linux/vfio.h
index f378b98..9e9a8d3 100644
--- a/include/uapi/linux/vfio.h
+++ b/include/uapi/linux/vfio.h
@@ -303,6 +303,12 @@ struct vfio_region_info_cap_type {
#define VFIO_REGION_SUBTYPE_INTEL_IGD_HOST_CFG (2)
#define VFIO_REGION_SUBTYPE_INTEL_IGD_LPC_CFG (3)
+/* NVIDIA GPU NVlink2 RAM */
+#define VFIO_REGION_SUBTYPE_NVIDIA_NVLINK2_RAM (1)
+
+/* IBM NPU NVlink2 ATSD */
+#define VFIO_REGION_SUBTYPE_IBM_NVLINK2_ATSD (1)
+
Please include some of the description in the commitlog here for
reference. Also please be explicit that these are vendor defined
regions and note the numerical vendor ID associated with them.
These are PCI region subtypes which are not from any PCI spec and which
we define as we like, are not they all "vendor"?
No, it's not entirely obvious that they're vendor regions, I had to
look at the usage later in the patch. See linux-next where Gerd has
added an EDID region which is not a vendor region. You make use of
these by masking VFIO_REGION_TYPE_PCI_VENDOR_TYPE into the type, which
indicates they are a PCI vendor type region, which means they are
associated to a specific vendor. That vendor should be explicitly,
numerically, indicated (some hardware vendors own multiple PCI vendor
IDs or may acquire more).
Ah, I see now. ok.
/*
* The MSIX mappable capability informs that MSIX data of a BAR can be mmapped
* which allows direct access to non-MSIX registers which happened to be within
@@ -313,6 +319,18 @@ struct vfio_region_info_cap_type {
*/
#define VFIO_REGION_INFO_CAP_MSIX_MAPPABLE 3
+/*
+ * Capability with compressed real address (aka SSA - small system address)
+ * where GPU RAM is mapped on a system bus. Used by a GPU for DMA routing.
+ */
+#define VFIO_REGION_INFO_CAP_NPU2 4
+
+struct vfio_region_info_cap_npu2 {
+ struct vfio_info_cap_header header;
+ __u64 tgt;
+ /* size is defined in VFIO_REGION_SUBTYPE_NVIDIA_NVLINK2_RAM */
But this is a capability for the IBM_NVLINK2_ATSD? What is the
relevance of this comment? Is this capability relevant to the RAM or
ATSD?
It is relevant to NPU (NVLink host bus adapter of POWER9) which maps the
GPU RAM to the system bus and acts as a proxy to nestMMU (NVIDIA's unit
in POWER9 CPU) for ATS/ATSD services so it is a property of NPU. But
then one might ask "wait, here is the address, where is the size then",
hence the comment...
So the tgt field within the npu2 capability of the ATSD region on the
NPU device describes the GPU internal address of the GPU RAM which is
described by the NVLINK2 RAM region on the GPU device... *twitch* What
business does the NPU have exposing this piece of information and how
is it related to the ATSD region/register?
NPU is the source of the information on the host. ATSD is just another
feature of NPU.
Is this tgt base used in
the process of doing address translation shootdowns?
No, only routing.
+};
+
/**
* VFIO_DEVICE_GET_IRQ_INFO - _IOWR(VFIO_TYPE, VFIO_BASE + 9,
* struct vfio_irq_info)
diff --git a/drivers/vfio/pci/vfio_pci.c b/drivers/vfio/pci/vfio_pci.c
index 4a3b93e..e9afd43 100644
--- a/drivers/vfio/pci/vfio_pci.c
+++ b/drivers/vfio/pci/vfio_pci.c
@@ -224,6 +224,16 @@ static bool vfio_pci_nointx(struct pci_dev *pdev)
return false;
}
+int __weak vfio_pci_nvdia_v100_nvlink2_init(struct vfio_pci_device *vdev)
+{
+ return -ENODEV;
+}
+
+int __weak vfio_pci_ibm_npu2_init(struct vfio_pci_device *vdev)
+{
+ return -ENODEV;
+}
+
static int vfio_pci_enable(struct vfio_pci_device *vdev)
{
struct pci_dev *pdev = vdev->pdev;
@@ -302,14 +312,37 @@ static int vfio_pci_enable(struct vfio_pci_device *vdev)
if (ret) {
dev_warn(&vdev->pdev->dev,
"Failed to setup Intel IGD regions\n");
- vfio_pci_disable(vdev);
- return ret;
+ goto disable_exit;
+ }
+ }
+
+ if (pdev->vendor == PCI_VENDOR_ID_NVIDIA &&
+ pdev->device == 0x1db1) {
+ ret = vfio_pci_nvdia_v100_nvlink2_init(vdev);
+ if (ret) {
+ dev_warn(&vdev->pdev->dev,
+ "Failed to setup NVIDIA NV2 RAM region\n");
+ goto disable_exit;
+ }
+ }
This device ID is not unique to POWER9 Witherspoon systems, I see your
comment in the commitlog, but this is clearly going to generate a
dev_warn and failure on an x86 system with the same hardware. Perhaps
this could be masked off with IS_ENABLED(CONFIG_VFIO_PCI_NVLINK2) like
the IGD code above this chunk does?
Right, will fix.
+
+ if (pdev->vendor == PCI_VENDOR_ID_IBM &&
+ pdev->device == 0x04ea) {
+ ret = vfio_pci_ibm_npu2_init(vdev);
+ if (ret) {
+ dev_warn(&vdev->pdev->dev,
+ "Failed to setup NVIDIA NV2 ATSD region\n");
+ goto disable_exit;
}
So the NPU is also actually owned by vfio-pci and assigned to the VM?
Yes. On a running system it looks like:
0007:00:00.0 Bridge: IBM Device 04ea (rev 01)
0007:00:00.1 Bridge: IBM Device 04ea (rev 01)
0007:00:01.0 Bridge: IBM Device 04ea (rev 01)
0007:00:01.1 Bridge: IBM Device 04ea (rev 01)
0007:00:02.0 Bridge: IBM Device 04ea (rev 01)
0007:00:02.1 Bridge: IBM Device 04ea (rev 01)
0035:00:00.0 PCI bridge: IBM Device 04c1
0035:01:00.0 PCI bridge: PLX Technology, Inc. Device 8725 (rev ca)
0035:02:04.0 PCI bridge: PLX Technology, Inc. Device 8725 (rev ca)
0035:02:05.0 PCI bridge: PLX Technology, Inc. Device 8725 (rev ca)
0035:02:0d.0 PCI bridge: PLX Technology, Inc. Device 8725 (rev ca)
0035:03:00.0 3D controller: NVIDIA Corporation GV100GL [Tesla V100 SXM2]
(rev a1
0035:04:00.0 3D controller: NVIDIA Corporation GV100GL [Tesla V100 SXM2]
(rev a1)
0035:05:00.0 3D controller: NVIDIA Corporation GV100GL [Tesla V100 SXM2]
(rev a1)
One "IBM Device" bridge represents one NVLink2, i.e. a piece of NPU.
They all and 3 GPUs go to the same IOMMU group and get passed through to
a guest.
The entire NPU does not have representation via sysfs as a whole though.
So the NPU is a bridge, but it uses a normal header type so vfio-pci
will bind to it?
An NPU is a NVLink bridge, it is not PCI in any sense. We (the host
powerpc firmware known as "skiboot" or "opal") have chosen to emulate a
virtual bridge per 1 NVLink on the firmware level. So for each physical
NPU there are 6 virtual bridges. So the NVIDIA driver does not need to
know much about NPUs.
And the ATSD register that we need on it is not
accessible through these PCI representations of the sub-pieces of the
NPU? Thanks,
No, only via the device tree. The skiboot puts the ATSD register address
to the PHB's DT property called 'ibm,mmio-atsd' of these virtual bridges.
Ok, so the NPU is essential a virtual device already, mostly just a
stub. But it seems that each NPU is associated to a specific GPU, how
is that association done? In the use case here it seems like it's just
a vehicle to provide this ibm,mmio-atsd property to guest DT and the tgt
routing information to the GPU. So if both of those were attached to
the GPU, there'd be no purpose in assigning the NPU other than it's in
the same IOMMU group with a type 0 header, so something needs to be
done with it. If it's a virtual device, perhaps it could have a type 1
header so vfio wouldn't care about it, then we would only assign the
GPU with these extra properties, which seems easier for management
tools and users. If the guest driver needs a visible NPU device, QEMU
could possibly emulate one to make the GPU association work
automatically. Maybe this isn't really a problem, but I wonder if
you've looked up the management stack to see what tools need to know to
assign these NPU devices and whether specific configurations are
required to make the NPU to GPU association work. Thanks,
I'm not that familiar with how this was originally set up, but note that
Alexey is just making it work exactly like baremetal does. The baremetal
GPU driver works as-is in the VM and expects the same properties in the
device-tree. Obviously it doesn't have to be that way, but there is
value in keeping it identical.
Another probably bigger point is that the NPU device also implements the
nvlink HW interface and is required for actually training and
maintaining the link up. The driver in the guest trains the links by
programming both the GPU end and the NPU end of each link so the NPU
device needs to be exposed to the guest.
Thanks,
Piotr
Alex
