nvidia/interface/nv_uvm_types.h

/*
 * SPDX-FileCopyrightText: Copyright (c) 2014-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 * SPDX-License-Identifier: MIT
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
 * DEALINGS IN THE SOFTWARE.
 */

//
// This file provides common types for both UVM driver and RM's UVM interface.
//

#ifndef _NV_UVM_TYPES_H_
#define _NV_UVM_TYPES_H_

#include "nvtypes.h"
#include "nvstatus.h"
#include "nvgputypes.h"
#include "nvCpuUuid.h"


//
// Default Page Size if left "0" because in RM BIG page size is default & there
// are multiple BIG page sizes in RM. These defines are used as flags to "0"
// should be OK when user is not sure which pagesize allocation it wants
//
#define UVM_PAGE_SIZE_DEFAULT    0x0
#define UVM_PAGE_SIZE_4K         0x1000
#define UVM_PAGE_SIZE_64K        0x10000
#define UVM_PAGE_SIZE_128K       0x20000
#define UVM_PAGE_SIZE_2M         0x200000
#define UVM_PAGE_SIZE_512M       0x20000000

//
// When modifying flags, make sure they are compatible with the mirrored
// PMA_* flags in phys_mem_allocator.h.
//
// Input flags
#define UVM_PMA_ALLOCATE_DONT_EVICT             NVBIT(0)
#define UVM_PMA_ALLOCATE_PINNED                 NVBIT(1)
#define UVM_PMA_ALLOCATE_SPECIFY_MINIMUM_SPEED  NVBIT(2)
#define UVM_PMA_ALLOCATE_SPECIFY_ADDRESS_RANGE  NVBIT(3)
#define UVM_PMA_ALLOCATE_SPECIFY_REGION_ID      NVBIT(4)
#define UVM_PMA_ALLOCATE_PREFER_SLOWEST         NVBIT(5)
#define UVM_PMA_ALLOCATE_CONTIGUOUS             NVBIT(6)
#define UVM_PMA_ALLOCATE_PERSISTENT             NVBIT(7)
#define UVM_PMA_ALLOCATE_PROTECTED_REGION       NVBIT(8)
#define UVM_PMA_ALLOCATE_FORCE_ALIGNMENT        NVBIT(9)
#define UVM_PMA_ALLOCATE_NO_ZERO                NVBIT(10)
#define UVM_PMA_ALLOCATE_TURN_BLACKLIST_OFF     NVBIT(11)
#define UVM_PMA_ALLOCATE_ALLOW_PARTIAL          NVBIT(12)

// Output flags
#define UVM_PMA_ALLOCATE_RESULT_IS_ZERO         NVBIT(0)

// Input flags to pmaFree
#define UVM_PMA_FREE_IS_ZERO                    NVBIT(0)

//
// Indicate that the PMA operation is being done from one of the PMA eviction
// callbacks.
//
// Notably this flag is currently used only by the UVM/RM interface and not
// mirrored in PMA.
//
#define UVM_PMA_CALLED_FROM_PMA_EVICTION        16384

#define UVM_UUID_LEN 16
#define UVM_SW_OBJ_SUBCHANNEL 5

typedef unsigned long long UvmGpuPointer;

//
// The following typedefs serve to explain the resources they point to.
// The actual resources remain RM internal and not exposed.
//
typedef struct uvmGpuSession_tag       *uvmGpuSessionHandle;       // gpuSessionHandle
typedef struct uvmGpuDevice_tag        *uvmGpuDeviceHandle;        // gpuDeviceHandle
typedef struct uvmGpuAddressSpace_tag  *uvmGpuAddressSpaceHandle;  // gpuAddressSpaceHandle
typedef struct uvmGpuTsg_tag           *uvmGpuTsgHandle;           // gpuTsgHandle
typedef struct uvmGpuChannel_tag       *uvmGpuChannelHandle;       // gpuChannelHandle
typedef struct uvmGpuCopyEngine_tag    *uvmGpuCopyEngineHandle;    // gpuObjectHandle

typedef struct UvmGpuMemoryInfo_tag
{
    // Out: Memory layout.
    NvU32 kind;

    // Out: Set to TRUE, if the allocation is in sysmem.
    NvBool sysmem;

    // Out: Set to TRUE, if this allocation is treated as EGM.
    //      sysmem is also TRUE when egm is TRUE.
    NvBool egm;

    // Out: Set to TRUE, if the allocation is a constructed
    //      under a Device or Subdevice.
    //      All permutations of sysmem and deviceDescendant are valid.
    //      !sysmem && !deviceDescendant implies a fabric allocation.
    NvBool deviceDescendant;

    // Out: Page size associated with the phys alloc.
    NvU64 pageSize;

    // Out: Set to TRUE, if the allocation is contiguous.
    NvBool contig;

    // Out: Starting Addr if the allocation is contiguous.
    //      This is only valid if contig is NV_TRUE.
    NvU64 physAddr;

    // Out: Total size of the allocation.
    NvU64 size;

    // Out: Uuid of the GPU to which the allocation belongs.
    //      This is only valid if deviceDescendant is NV_TRUE.
    //      When egm is NV_TRUE, this is also the UUID of the GPU
    //      for which EGM is local.
    //      Note: If the allocation is owned by a device in
    //      an SLI group and the allocation is broadcast
    //      across the SLI group, this UUID will be any one
    //      of the subdevices in the SLI group.
    NvProcessorUuid uuid;
} UvmGpuMemoryInfo;

// Some resources must share the same virtual mappings across channels. A mapped
// resource must be shared by a channel iff:
//
// 1) The channel belongs to a TSG (UvmGpuChannelInstanceInfo::bTsgChannel is
//    NV_TRUE).
//
// 2) The channel is in the same TSG as all other channels sharing that mapping
//    (UvmGpuChannelInstanceInfo::tsgId matches among channels).
//
// 3) The channel is in the same GPU address space as the other channels
//    sharing that mapping.
//
// 4) The resource handle(s) match those of the shared mapping
//    (UvmGpuChannelResourceInfo::resourceDescriptor and
//    UvmGpuChannelResourceInfo::resourceId).
typedef struct UvmGpuChannelResourceInfo_tag
{
    // Out: Ptr to the RM memDesc of the channel resource.
    NvP64 resourceDescriptor;

    // Out: RM ID of the channel resource.
    NvU32 resourceId;

    // Out: Alignment needed for the resource allocation.
    NvU64 alignment;

    // Out: Info about the resource allocation.
    UvmGpuMemoryInfo resourceInfo;
} UvmGpuChannelResourceInfo;

typedef struct UvmGpuPagingChannelInfo_tag
{
    // Pointer to a shadown buffer mirroring the contents of the error notifier
    // for the paging channel
    NvNotification    *shadowErrorNotifier;
} UvmGpuPagingChannelInfo;

typedef enum
{
    UVM_GPU_CHANNEL_ENGINE_TYPE_GR = 1,
    UVM_GPU_CHANNEL_ENGINE_TYPE_CE = 2,
    UVM_GPU_CHANNEL_ENGINE_TYPE_SEC2 = 3,
} UVM_GPU_CHANNEL_ENGINE_TYPE;

#define UVM_GPU_CHANNEL_MAX_RESOURCES   13

typedef struct UvmGpuChannelInstanceInfo_tag
{
    // Out: Starting address of the channel instance.
    NvU64 base;

    // Out: Set to NV_TRUE, if the instance is in sysmem.
    //      Set to NV_FALSE, if the instance is in vidmem.
    NvBool sysmem;

    // Out: Hardware runlist ID.
    NvU32 runlistId;

    // Out: Hardware channel ID.
    NvU32 chId;

    // Out: NV_TRUE if the channel belongs to a subcontext or NV_FALSE if it
    // belongs to a regular context.
    NvBool bInSubctx;

    // Out: ID of the subcontext to which the channel belongs.
    NvU32 subctxId;

    // Out: Whether the channel belongs to a TSG or not
    NvBool bTsgChannel;

    // Out: ID of the TSG to which the channel belongs
    NvU32 tsgId;

    // Out: Maximum number of subcontexts in the TSG to which the channel belongs
    NvU32 tsgMaxSubctxCount;

    // Out: Info of channel resources associated with the channel.
    UvmGpuChannelResourceInfo resourceInfo[UVM_GPU_CHANNEL_MAX_RESOURCES];

    // Out: Number of valid entries in resourceInfo array.
    NvU32 resourceCount;

    // Out: Type of the engine the channel is bound to
    NvU32 channelEngineType;

    // Out: Channel handle to be used in the CLEAR_FAULTED method
    NvU32 clearFaultedToken;

    // Out: Address of the NV_CHRAM_CHANNEL register required to clear the
    // ENG_FAULTED/PBDMA_FAULTED bits after servicing non-replayable faults on
    // Ampere+ GPUs
    volatile NvU32 *pChramChannelRegister;

    // Out: Address of the Runlist PRI Base Register required to ring the
    // doorbell after clearing the faulted bit.
    volatile NvU32 *pRunlistPRIBaseRegister;

    // Out: SMC engine id to which the GR channel is bound, or zero if the GPU
    // does not support SMC or it is a CE channel
    NvU32 smcEngineId;

    // Out: Start of the VEID range assigned to the SMC engine the GR channel
    // is bound to, or zero if the GPU does not support SMC or it is a CE
    // channel
    NvU32 smcEngineVeIdOffset;
} UvmGpuChannelInstanceInfo;

typedef struct UvmGpuChannelResourceBindParams_tag
{
    // In: RM ID of the channel resource.
    NvU32 resourceId;

    // In: Starting VA at which the channel resource is mapped.
    NvU64 resourceVa;
} UvmGpuChannelResourceBindParams;

typedef struct UvmGpuChannelInfo_tag
{
    volatile unsigned *gpGet;
    volatile unsigned *gpPut;
    UvmGpuPointer     *gpFifoEntries;
    unsigned           numGpFifoEntries;
    unsigned           channelClassNum;

    // The errorNotifier is filled out when the channel hits an RC error.
    NvNotification    *errorNotifier;

    NvU32              hwRunlistId;
    NvU32              hwChannelId;

    volatile unsigned *dummyBar1Mapping;

    // These values are filled by nvUvmInterfaceCopyEngineAlloc. The work
    // submission token requires the channel to be bound to a runlist and that
    // happens after CE allocation.
    volatile NvU32    *workSubmissionOffset;

    // To be deprecated. See pWorkSubmissionToken below.
    NvU32              workSubmissionToken;

    //
    // This is the memory location where the most recently updated work
    // submission token for this channel will be written to. After submitting
    // new work and updating GP_PUT with the appropriate fence, the token must
    // be read from this location before writing it to the workSubmissionOffset
    // to kick off the new work.
    //
    volatile NvU32    *pWorkSubmissionToken;

    // GPU VAs of both GPFIFO and GPPUT are needed in Confidential Computing
    // so a channel can be controlled via another channel (SEC2 or WLC/LCIC)
    NvU64             gpFifoGpuVa;
    NvU64             gpPutGpuVa;
    NvU64             gpGetGpuVa;
    // GPU VA of work submission offset is needed in Confidential Computing
    // so CE channels can ring doorbell of other channels as required for
    // WLC/LCIC work submission
    NvU64             workSubmissionOffsetGpuVa;
} UvmGpuChannelInfo;

typedef enum
{
    // This value must be passed by Pascal and pre-Pascal GPUs for those
    // allocations for which a specific location cannot be enforced.
    UVM_BUFFER_LOCATION_DEFAULT = 0,

    UVM_BUFFER_LOCATION_SYS  = 1,
    UVM_BUFFER_LOCATION_VID  = 2,
} UVM_BUFFER_LOCATION;

typedef struct UvmGpuTsgAllocParams_tag
{
    // Interpreted as UVM_GPU_CHANNEL_ENGINE_TYPE
    NvU32 engineType;

    // Index of the engine the TSG is bound to.
    // Ignored if engineType is anything other than
    // UVM_GPU_CHANNEL_ENGINE_TYPE_CE.
    NvU32 engineIndex;
} UvmGpuTsgAllocParams;

typedef struct UvmGpuChannelAllocParams_tag
{
    NvU32 numGpFifoEntries;

    // The next two fields store UVM_BUFFER_LOCATION values
    NvU32 gpFifoLoc;
    NvU32 gpPutLoc;
} UvmGpuChannelAllocParams;

typedef struct UvmGpuPagingChannelAllocParams_tag
{
    // Index of the LCE engine the channel will be bound to, a zero-based offset
    // from NV2080_ENGINE_TYPE_COPY0.
    NvU32 engineIndex;
} UvmGpuPagingChannelAllocParams;

// The max number of Copy Engines supported by a GPU.
// The gpu ops build has a static assert that this is the correct number.
#define UVM_COPY_ENGINE_COUNT_MAX 64

typedef struct
{
    // True if the CE is supported at all
    NvBool supported:1;

    // True if the CE is synchronous with GR
    NvBool grce:1;

    // True if the CE shares physical CEs with any other CE
    //
    // The value returned by RM for this field may change when a GPU is
    // registered with RM for the first time, so UVM needs to query it
    // again each time a GPU is registered.
    NvBool shared:1;

    // True if the CE can give enhanced performance for SYSMEM reads over other CEs
    NvBool sysmemRead:1;

    // True if the CE can give enhanced performance for SYSMEM writes over other CEs
    NvBool sysmemWrite:1;

    // True if the CE can be used for SYSMEM transactions
    NvBool sysmem:1;

    // True if the CE can be used for P2P transactions using NVLINK
    NvBool nvlinkP2p:1;

    // True if the CE can be used for P2P transactions
    NvBool p2p:1;

    // Mask of physical CEs assigned to this LCE
    //
    // The value returned by RM for this field may change when a GPU is
    // registered with RM for the first time, so UVM needs to query it
    // again each time a GPU is registered.
    NvU32 cePceMask;
} UvmGpuCopyEngineCaps;

typedef struct UvmGpuCopyEnginesCaps_tag
{
    // Supported CEs may not be contiguous
    UvmGpuCopyEngineCaps copyEngineCaps[UVM_COPY_ENGINE_COUNT_MAX];
} UvmGpuCopyEnginesCaps;

typedef enum
{
    UVM_LINK_TYPE_NONE,
    UVM_LINK_TYPE_PCIE,
    UVM_LINK_TYPE_NVLINK_1,
    UVM_LINK_TYPE_NVLINK_2,
    UVM_LINK_TYPE_NVLINK_3,
    UVM_LINK_TYPE_NVLINK_4,
    UVM_LINK_TYPE_C2C,
} UVM_LINK_TYPE;

typedef struct UvmGpuCaps_tag
{
    // If numaEnabled is NV_TRUE, then the system address of allocated GPU
    // memory can be converted to struct pages. See
    // UvmGpuInfo::systemMemoryWindowStart.
    NvBool   numaEnabled;
    NvU32    numaNodeId;
} UvmGpuCaps;

typedef struct UvmGpuAddressSpaceInfo_tag
{
    NvU64           bigPageSize;

    NvBool          atsEnabled;

    // Mapped registers that contain the current GPU time
    volatile NvU32  *time0Offset;
    volatile NvU32  *time1Offset;

    // Maximum number of subcontexts supported under this GPU address space
    NvU32           maxSubctxCount;

    NvBool          smcEnabled;

    NvU32           smcSwizzId;

    NvU32           smcGpcCount;
} UvmGpuAddressSpaceInfo;

typedef struct UvmGpuAllocInfo_tag
{
    NvU64   gpuPhysOffset;          // Returns gpuPhysOffset if contiguous requested
    NvU64   pageSize;               // default is RM big page size - 64K or 128 K" else use 4K or 2M
    NvU64   alignment;              // Virtual alignment
    NvBool  bContiguousPhysAlloc;   // Flag to request contiguous physical allocation
    NvBool  bMemGrowsDown;          // Causes RM to reserve physical heap from top of FB
    NvBool  bPersistentVidmem;      // Causes RM to allocate persistent video memory
    NvHandle hPhysHandle;           // Handle for phys allocation either provided or retrieved
    NvBool   bUnprotected;            // Allocation to be made in unprotected memory whenever
                                      // SEV or GPU CC modes are enabled. Ignored otherwise
} UvmGpuAllocInfo;

typedef enum
{
    UVM_VIRT_MODE_NONE = 0,             // Baremetal or passthrough virtualization
    UVM_VIRT_MODE_LEGACY = 1,           // Virtualization without SRIOV support
    UVM_VIRT_MODE_SRIOV_HEAVY = 2,      // Virtualization with SRIOV Heavy configured
    UVM_VIRT_MODE_SRIOV_STANDARD = 3,   // Virtualization with SRIOV Standard configured
    UVM_VIRT_MODE_COUNT = 4,
} UVM_VIRT_MODE;

// !!! The following enums (with UvmRm prefix) are defined and documented in
// mm/uvm/interface/uvm_types.h and must be mirrored. Please refer to that file
// for more details.

// UVM GPU mapping types
typedef enum
{
    UvmRmGpuMappingTypeDefault = 0,
    UvmRmGpuMappingTypeReadWriteAtomic = 1,
    UvmRmGpuMappingTypeReadWrite = 2,
    UvmRmGpuMappingTypeReadOnly = 3,
    UvmRmGpuMappingTypeCount = 4
} UvmRmGpuMappingType;

// UVM GPU caching types
typedef enum
{
    UvmRmGpuCachingTypeDefault = 0,
    UvmRmGpuCachingTypeForceUncached = 1,
    UvmRmGpuCachingTypeForceCached = 2,
    UvmRmGpuCachingTypeCount = 3
} UvmRmGpuCachingType;

// UVM GPU format types
typedef enum {
   UvmRmGpuFormatTypeDefault = 0,
   UvmRmGpuFormatTypeBlockLinear = 1,
   UvmRmGpuFormatTypeCount = 2
} UvmRmGpuFormatType;

// UVM GPU Element bits types
typedef enum {
   UvmRmGpuFormatElementBitsDefault = 0,
   UvmRmGpuFormatElementBits8 = 1,
   UvmRmGpuFormatElementBits16 = 2,
   // Cuda does not support 24-bit width
   UvmRmGpuFormatElementBits32 = 4,
   UvmRmGpuFormatElementBits64 = 5,
   UvmRmGpuFormatElementBits128 = 6,
   UvmRmGpuFormatElementBitsCount = 7
} UvmRmGpuFormatElementBits;

// UVM GPU Compression types
typedef enum {
    UvmRmGpuCompressionTypeDefault = 0,
    UvmRmGpuCompressionTypeEnabledNoPlc = 1,
    UvmRmGpuCompressionTypeCount = 2
} UvmRmGpuCompressionType;

typedef struct UvmGpuExternalMappingInfo_tag
{
    // In: GPU caching ability.
    UvmRmGpuCachingType cachingType;

    // In: Virtual permissions.
    UvmRmGpuMappingType mappingType;

    // In: RM virtual mapping memory format
    UvmRmGpuFormatType formatType;

    // In: RM virtual mapping element bits
    UvmRmGpuFormatElementBits elementBits;

    // In: RM virtual compression type
    UvmRmGpuCompressionType compressionType;

    // In: Size of the buffer to store PTEs (in bytes).
    NvU64 pteBufferSize;

    // In: Page size for mapping
    //     If this field is passed as 0, the page size
    //     of the allocation is used for mapping.
    //     nvUvmInterfaceGetExternalAllocPtes must pass
    //     this field as zero.
    NvU64 mappingPageSize;

    // In: Pointer to a buffer to store PTEs.
    // Out: The interface will fill the buffer with PTEs
    NvU64 *pteBuffer;

    // Out: Number of PTEs filled in to the buffer.
    NvU64 numWrittenPtes;

    // Out: Number of PTEs remaining to be filled
    //      if the buffer is not sufficient to accommodate
    //      requested PTEs.
    NvU64 numRemainingPtes;

    // Out: PTE size (in bytes)
    NvU32 pteSize;
} UvmGpuExternalMappingInfo;

typedef struct UvmGpuP2PCapsParams_tag
{
    // Out: peerId[i] contains gpu[i]'s peer id of gpu[1 - i]. Only defined if
    // the GPUs are direct peers.
    NvU32 peerIds[2];

    // Out: UVM_LINK_TYPE
    NvU32 p2pLink;

    // Out: optimalNvlinkWriteCEs[i] contains gpu[i]'s optimal CE for writing to
    // gpu[1 - i]. The CE indexes are valid only if the GPUs are NVLink peers.
    //
    // The value returned by RM for this field may change when a GPU is
    // registered with RM for the first time, so UVM needs to query it again
    // each time a GPU is registered.
    NvU32 optimalNvlinkWriteCEs[2];

    // Out: Maximum unidirectional bandwidth between the peers in megabytes per
    // second, not taking into account the protocols overhead. The reported
    // bandwidth for indirect peers is zero.
    NvU32 totalLinkLineRateMBps;

    // Out: True if the peers have a indirect link to communicate. On P9
    // systems, this is true if peers are connected to different NPUs that
    // forward the requests between them.
    NvU32 indirectAccess      : 1;
} UvmGpuP2PCapsParams;

// Platform-wide information
typedef struct UvmPlatformInfo_tag
{
    // Out: ATS (Address Translation Services) is supported
    NvBool atsSupported;

    // Out: AMD SEV (Secure Encrypted Virtualization) is enabled
    NvBool sevEnabled;
} UvmPlatformInfo;

typedef struct UvmGpuClientInfo_tag
{
    NvHandle hClient;

    NvHandle hSmcPartRef;
} UvmGpuClientInfo;

typedef enum
{
    UVM_GPU_CONF_COMPUTE_MODE_NONE,
    UVM_GPU_CONF_COMPUTE_MODE_APM,
    UVM_GPU_CONF_COMPUTE_MODE_HCC,
    UVM_GPU_CONF_COMPUTE_MODE_COUNT
} UvmGpuConfComputeMode;

typedef struct UvmGpuConfComputeCaps_tag
{
    // Out: GPU's confidential compute mode
    UvmGpuConfComputeMode mode;
} UvmGpuConfComputeCaps;

#define UVM_GPU_NAME_LENGTH 0x40

typedef struct UvmGpuInfo_tag
{
    // Printable gpu name
    char name[UVM_GPU_NAME_LENGTH];

    // Uuid of this gpu
    NvProcessorUuid uuid;

    // Gpu architecture; NV2080_CTRL_MC_ARCH_INFO_ARCHITECTURE_*
    NvU32 gpuArch;

    // Gpu implementation; NV2080_CTRL_MC_ARCH_INFO_IMPLEMENTATION_*
    NvU32 gpuImplementation;

    // Host (gpfifo) class; *_CHANNEL_GPFIFO_*, e.g. KEPLER_CHANNEL_GPFIFO_A
    NvU32 hostClass;

    // Copy engine (dma) class; *_DMA_COPY_*, e.g. KEPLER_DMA_COPY_A
    NvU32 ceClass;

    // Compute class; *_COMPUTE_*, e.g. KEPLER_COMPUTE_A
    NvU32 computeClass;

    // Set if GPU supports TCC Mode & is in TCC mode.
    NvBool gpuInTcc;

    // Number of subdevices in SLI group.
    NvU32 subdeviceCount;

    // Virtualization mode of this gpu.
    NvU32 virtMode;         // UVM_VIRT_MODE

    // NV_TRUE if this is a simulated/emulated GPU. NV_FALSE, otherwise.
    NvBool isSimulated;

    // Number of GPCs
    // If SMC is enabled, this is the currently configured number of GPCs for
    // the given partition (also see the smcSwizzId field below).
    NvU32 gpcCount;

    // Maximum number of GPCs; NV_SCAL_LITTER_NUM_GPCS
    // This number is independent of the partition configuration, and can be
    // used to conservatively size GPU-global constructs.
    NvU32 maxGpcCount;

    // Number of TPCs
    NvU32 tpcCount;

    // Maximum number of TPCs per GPC
    NvU32 maxTpcPerGpcCount;

    // NV_TRUE if SMC is enabled on this GPU.
    NvBool smcEnabled;

    // SMC partition ID (unique per GPU); note: valid when first looked up in
    // nvUvmInterfaceGetGpuInfo(), but not guaranteed to remain valid.
    // nvUvmInterfaceDeviceCreate() re-verifies the swizzId and fails if it is
    // no longer valid.
    NvU32 smcSwizzId;

    UvmGpuClientInfo smcUserClientInfo;

    // Confidential Compute capabilities of this GPU
    UvmGpuConfComputeCaps gpuConfComputeCaps;

    // UVM_LINK_TYPE
    NvU32 sysmemLink;

    // See UvmGpuP2PCapsParams::totalLinkLineRateMBps
    NvU32 sysmemLinkRateMBps;

    // On coherent systems each GPU maps its memory to a window in the System
    // Physical Address (SPA) space. The following fields describe that window.
    //
    // systemMemoryWindowSize > 0 indicates that the window is valid. meaning
    // that GPU memory can be mapped by the CPU as cache-coherent by adding the
    // GPU address to the window start.
    NvU64 systemMemoryWindowStart;
    NvU64 systemMemoryWindowSize;

    // This tells if the GPU is connected to NVSwitch. On systems with NVSwitch
    // all GPUs are connected to it. If connectedToSwitch is NV_TRUE,
    // nvswitchMemoryWindowStart tells the base address for the GPU in the
    // NVSwitch address space. It is used when creating PTEs of memory mappings
    // to NVSwitch peers.
    NvBool connectedToSwitch;
    NvU64 nvswitchMemoryWindowStart;

    // local EGM properties
    NvBool   egmEnabled;
    NvU8     egmPeerId;
} UvmGpuInfo;

typedef struct UvmGpuFbInfo_tag
{
    // Max physical address that can be allocated by UVM. This excludes internal
    // RM regions that are not registered with PMA either.
    NvU64 maxAllocatableAddress;

    NvU32 heapSize;         // RAM in KB available for user allocations
    NvU32 reservedHeapSize; // RAM in KB reserved for internal RM allocation
    NvBool bZeroFb;         // Zero FB mode enabled.
} UvmGpuFbInfo;

typedef struct UvmGpuEccInfo_tag
{
    unsigned eccMask;
    unsigned eccOffset;
    void    *eccReadLocation;
    NvBool  *eccErrorNotifier;
    NvBool   bEccEnabled;
} UvmGpuEccInfo;

typedef struct UvmPmaAllocationOptions_tag
{
    NvU32 flags;
    NvU32 minimumSpeed;         // valid if flags & UVM_PMA_ALLOCATE_SPECIFY_MININUM_SPEED
    NvU64 physBegin, physEnd;   // valid if flags & UVM_PMA_ALLOCATE_SPECIFY_ADDRESS_RANGE
    NvU32 regionId;             // valid if flags & UVM_PMA_ALLOCATE_SPECIFY_REGION_ID
    NvU64 alignment;            // valid if flags & UVM_PMA_ALLOCATE_FORCE_ALIGNMENT
    NvLength numPagesAllocated; // valid if flags & UVM_PMA_ALLOCATE_ALLOW_PARTIAL

    NvU32 resultFlags;          // valid if the allocation function returns NV_OK
} UvmPmaAllocationOptions;

//
// Mirrored in PMA (PMA_STATS)
//
typedef struct UvmPmaStatistics_tag
{
    volatile NvU64 numPages2m;                // PMA-wide 2MB pages count across all regions
    volatile NvU64 numFreePages64k;           // PMA-wide free 64KB page count across all regions
    volatile NvU64 numFreePages2m;            // PMA-wide free 2MB pages count across all regions
    volatile NvU64 numPages2mProtected;       // PMA-wide 2MB pages count in protected memory
    volatile NvU64 numFreePages64kProtected;  // PMA-wide free 64KB page count in protected memory
    volatile NvU64 numFreePages2mProtected;   // PMA-wide free 2MB pages count in protected memory
} UvmPmaStatistics;

/*******************************************************************************
    uvmEventSuspend
    This function will be called by the GPU driver to signal to UVM that the
    system is about to enter a sleep state.  When it is called, the
    following assumptions/guarantees are valid/made:

      * User channels have been preempted and disabled
      * UVM channels are still running normally and will continue to do
        so until after this function returns control
      * User threads are still running, but can no longer issue system
        system calls to the GPU driver
      * Until exit from this function, UVM is allowed to make full use of
        the GPUs under its control, as well as of the GPU driver

    Upon return from this function, UVM may not access GPUs under its control
    until the GPU driver calls uvmEventResume().  It may still receive
    calls to uvmEventIsrTopHalf() during this time, which it should return
    NV_ERR_NO_INTR_PENDING from.  It will not receive any other calls.
*/
typedef NV_STATUS (*uvmEventSuspend_t) (void);

/*******************************************************************************
    uvmEventResume
    This function will be called by the GPU driver to signal to UVM that the
    system has exited a previously entered sleep state.  When it is called,
    the following assumptions/guarantees are valid/made:

      * UVM is again allowed to make full use of the GPUs under its
        control, as well as of the GPU driver
      * UVM channels are running normally
      * User channels are still preempted and disabled
      * User threads are again running, but still cannot issue system
        calls to the GPU driver, nor submit new work

    Upon return from this function, UVM is expected to be fully functional.
*/
typedef NV_STATUS (*uvmEventResume_t) (void);

/*******************************************************************************
    uvmEventStartDevice
    This function will be called by the GPU driver once it has finished its
    initialization to tell the UVM driver that this GPU has come up.
*/
typedef NV_STATUS (*uvmEventStartDevice_t) (const NvProcessorUuid *pGpuUuidStruct);

/*******************************************************************************
    uvmEventStopDevice
    This function will be called by the GPU driver to let UVM know that a GPU
    is going down.
*/
typedef NV_STATUS (*uvmEventStopDevice_t) (const NvProcessorUuid *pGpuUuidStruct);

/*******************************************************************************
    uvmEventIsrTopHalf_t
    This function will be called by the GPU driver to let UVM know
    that an interrupt has occurred.

    Returns:
        NV_OK if the UVM driver handled the interrupt
        NV_ERR_NO_INTR_PENDING if the interrupt is not for the UVM driver
*/
#if defined (__linux__)
typedef NV_STATUS (*uvmEventIsrTopHalf_t) (const NvProcessorUuid *pGpuUuidStruct);
#else
typedef void (*uvmEventIsrTopHalf_t) (void);
#endif

struct UvmOpsUvmEvents
{
    uvmEventSuspend_t     suspend;
    uvmEventResume_t      resume;
    uvmEventStartDevice_t startDevice;
    uvmEventStopDevice_t  stopDevice;
    uvmEventIsrTopHalf_t  isrTopHalf;
};

#define UVM_CSL_SIGN_AUTH_TAG_SIZE_BYTES 32
#define UVM_CSL_CRYPT_AUTH_TAG_SIZE_BYTES 16

typedef union UvmFaultMetadataPacket_tag
{
    struct {
        NvU8   authTag[UVM_CSL_CRYPT_AUTH_TAG_SIZE_BYTES];
        NvBool valid;
    };
    // padding to 32Bytes
    NvU8 _padding[32];
} UvmFaultMetadataPacket;

// This struct shall not be accessed nor modified directly by UVM as it is
// entirely managed by the RM layer
typedef struct UvmCslContext_tag
{
    struct ccslContext_t *ctx;
    void *nvidia_stack;
} UvmCslContext;

typedef struct UvmGpuFaultInfo_tag
{
    struct
    {
        // Fault buffer GET register mapping.
        //
        // When Confidential Computing is enabled, GET refers to the shadow
        // buffer (see bufferAddress below), and not to the actual HW buffer.
        // In this setup, writes of GET (by UVM) do not result on re-evaluation
        // of any interrupt condition.
        volatile NvU32* pFaultBufferGet;

        // Fault buffer PUT register mapping.
        //
        // When Confidential Computing is enabled, PUT refers to the shadow
        // buffer (see bufferAddress below), and not to the actual HW buffer.
        // In this setup, writes of PUT (by GSP-RM) do not result on
        // re-evaluation of any interrupt condition.
        volatile NvU32* pFaultBufferPut;

        // Note: this variable is deprecated since buffer overflow is not a
        // separate register from future chips.
        volatile NvU32* pFaultBufferInfo;

        // Register mapping used to clear a replayable fault interrupt in
        // Turing+ GPUs.
        volatile NvU32* pPmcIntr;

        // Register mapping used to enable replayable fault interrupts.
        volatile NvU32* pPmcIntrEnSet;

        // Register mapping used to disable replayable fault interrupts.
        volatile NvU32* pPmcIntrEnClear;

        // Register used to enable, or disable, faults on prefetches.
        volatile NvU32* pPrefetchCtrl;

        // Replayable fault interrupt mask identifier.
        NvU32 replayableFaultMask;

        // Fault buffer CPU mapping
        void*  bufferAddress;
        //
        // When Confidential Computing is disabled, the mapping points to the
        // actual HW fault buffer.
        //
        // When Confidential Computing is enabled, the mapping points to a
        // copy of the HW fault buffer. This "shadow buffer" is maintained
        // by GSP-RM.

        // Size, in bytes, of the fault buffer pointed by bufferAddress.
        NvU32  bufferSize;
        // Mapping pointing to the start of the fault buffer metadata containing
        // a 16Byte authentication tag and a valid byte. Always NULL when
        // Confidential Computing is disabled.
        UvmFaultMetadataPacket *bufferMetadata;

        // CSL context used for performing decryption of replayable faults when
        // Confidential Computing is enabled.
        UvmCslContext cslCtx;

        // Indicates whether UVM owns the replayable fault buffer.
        // The value of this field is always NV_TRUE When Confidential Computing
        // is disabled.
        NvBool bUvmOwnsHwFaultBuffer;
    } replayable;
    struct
    {
        // Shadow buffer for non-replayable faults on cpu memory. Resman copies
        // here the non-replayable faults that need to be handled by UVM
        void* shadowBufferAddress;

        // Execution context for the queue associated with the fault buffer
        void* shadowBufferContext;

        // Fault buffer size
        NvU32  bufferSize;

        // Preallocated stack for functions called from the UVM isr top half
        void *isr_sp;

        // Preallocated stack for functions called from the UVM isr bottom half
        void *isr_bh_sp;

        // Used only when Hopper Confidential Compute is enabled
        // Register mappings obtained from RM
        volatile NvU32* pFaultBufferPut;

        // Used only when Hopper Confidential Compute is enabled
        // Cached get index of the non-replayable shadow buffer
        NvU32 shadowBufferGet;

        // See replayable.bufferMetadata
        UvmFaultMetadataPacket  *shadowBufferMetadata;
    } nonReplayable;
    NvHandle faultBufferHandle;
    struct Device *pDevice;
} UvmGpuFaultInfo;

struct Device;

typedef struct UvmGpuPagingChannel_tag
{
    struct gpuDevice  *device;
    NvNotification    *errorNotifier;
    NvHandle          channelHandle;
    NvHandle          errorNotifierHandle;
    void              *pushStreamSp;
    struct Device     *pDevice;
} UvmGpuPagingChannel, *UvmGpuPagingChannelHandle;

typedef struct UvmGpuAccessCntrInfo_tag
{
    // Register mappings obtained from RM
    // pointer to the Get register for the access counter buffer
    volatile NvU32* pAccessCntrBufferGet;
    // pointer to the Put register for the access counter buffer
    volatile NvU32* pAccessCntrBufferPut;
    // pointer to the Full register for the access counter buffer
    volatile NvU32* pAccessCntrBufferFull;
    // pointer to the hub interrupt
    volatile NvU32* pHubIntr;
    // pointer to interrupt enable register
    volatile NvU32* pHubIntrEnSet;
    // pointer to interrupt disable register
    volatile NvU32* pHubIntrEnClear;
    // mask for the access counter buffer
    NvU32 accessCounterMask;
    // access counter buffer cpu mapping and size
    void* bufferAddress;
    NvU32  bufferSize;
    NvHandle accessCntrBufferHandle;
} UvmGpuAccessCntrInfo;

typedef enum
{
    UVM_ACCESS_COUNTER_GRANULARITY_64K = 1,
    UVM_ACCESS_COUNTER_GRANULARITY_2M  = 2,
    UVM_ACCESS_COUNTER_GRANULARITY_16M = 3,
    UVM_ACCESS_COUNTER_GRANULARITY_16G = 4,
} UVM_ACCESS_COUNTER_GRANULARITY;

typedef enum
{
    UVM_ACCESS_COUNTER_USE_LIMIT_NONE = 1,
    UVM_ACCESS_COUNTER_USE_LIMIT_QTR  = 2,
    UVM_ACCESS_COUNTER_USE_LIMIT_HALF = 3,
    UVM_ACCESS_COUNTER_USE_LIMIT_FULL = 4,
} UVM_ACCESS_COUNTER_USE_LIMIT;

typedef struct UvmGpuAccessCntrConfig_tag
{
    NvU32 mimcGranularity;

    NvU32 momcGranularity;

    NvU32 mimcUseLimit;

    NvU32 momcUseLimit;

    NvU32 threshold;
} UvmGpuAccessCntrConfig;

//
// When modifying this enum, make sure they are compatible with the mirrored
// MEMORY_PROTECTION enum in phys_mem_allocator.h.
//
typedef enum UvmPmaGpuMemoryType_tag
{
    UVM_PMA_GPU_MEMORY_TYPE_UNPROTECTED = 0,
    UVM_PMA_GPU_MEMORY_TYPE_PROTECTED   = 1
} UVM_PMA_GPU_MEMORY_TYPE;

typedef UvmGpuChannelInfo gpuChannelInfo;
typedef UvmGpuTsgAllocParams gpuTsgAllocParams;
typedef UvmGpuChannelAllocParams gpuChannelAllocParams;
typedef UvmGpuCaps gpuCaps;
typedef UvmGpuCopyEngineCaps gpuCeCaps;
typedef UvmGpuCopyEnginesCaps gpuCesCaps;
typedef UvmGpuP2PCapsParams getP2PCapsParams;
typedef UvmGpuAddressSpaceInfo gpuAddressSpaceInfo;
typedef UvmGpuAllocInfo gpuAllocInfo;
typedef UvmGpuInfo gpuInfo;
typedef UvmGpuClientInfo gpuClientInfo;
typedef UvmGpuAccessCntrInfo gpuAccessCntrInfo;
typedef UvmGpuAccessCntrConfig gpuAccessCntrConfig;
typedef UvmGpuFaultInfo gpuFaultInfo;
typedef UvmGpuMemoryInfo gpuMemoryInfo;
typedef UvmGpuExternalMappingInfo gpuExternalMappingInfo;
typedef UvmGpuChannelResourceInfo gpuChannelResourceInfo;
typedef UvmGpuChannelInstanceInfo gpuChannelInstanceInfo;
typedef UvmGpuChannelResourceBindParams gpuChannelResourceBindParams;
typedef UvmGpuFbInfo gpuFbInfo;
typedef UvmGpuEccInfo gpuEccInfo;
typedef UvmGpuPagingChannel *gpuPagingChannelHandle;
typedef UvmGpuPagingChannelInfo gpuPagingChannelInfo;
typedef UvmGpuPagingChannelAllocParams gpuPagingChannelAllocParams;
typedef UvmPmaAllocationOptions gpuPmaAllocationOptions;

typedef struct UvmCslIv
{
    NvU8 iv[12];
    NvU8 fresh;
} UvmCslIv;

typedef enum UvmCslOperation
{
    UVM_CSL_OPERATION_ENCRYPT,
    UVM_CSL_OPERATION_DECRYPT
} UvmCslOperation;

#endif // _NV_UVM_TYPES_H_