1 /*
2  * SPDX-FileCopyrightText: Copyright (c) 2023-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
3  * SPDX-License-Identifier: MIT
4  *
5  * Permission is hereby granted, free of charge, to any person obtaining a
6  * copy of this software and associated documentation files (the "Software"),
7  * to deal in the Software without restriction, including without limitation
8  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
9  * and/or sell copies of the Software, and to permit persons to whom the
10  * Software is furnished to do so, subject to the following conditions:
11  *
12  * The above copyright notice and this permission notice shall be included in
13  * all copies or substantial portions of the Software.
14  *
15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
21  * DEALINGS IN THE SOFTWARE.
22  */
23 
24 #ifndef _CHANNEL_UTILS_H_
25 #define _CHANNEL_UTILS_H_
26 
27 #include "core/core.h"
28 #include "gpu/gpu.h"
29 #include "gpu/mem_mgr/mem_mgr.h"
30 #include "gpu/ce/kernel_ce.h"
31 #include "gpu/bus/kern_bus.h"
32 #include "core/prelude.h"
33 #include "rmapi/rs_utils.h"
34 #include "nvos.h"
35 
36 #include "class/cl906f.h"
37 #include "class/cl906f.h"
38 #include "class/cl906fsw.h"
39 
40 #include "class/clb0b5.h"   // MAXWELL_DMA_COPY_A
41 #include "class/clc0b5.h"   // PASCAL_DMA_COPY_A
42 #include "class/clc1b5.h"   // PASCAL_DMA_COPY_B
43 #include "class/clc3b5.h"   // VOLTA_DMA_COPY_A
44 #include "class/clc5b5.h"   // TURING_DMA_COPY_A
45 #include "class/clc6b5.h"   // AMPERE_DMA_COPY_A
46 #include "class/clc7b5.h"   // AMPERE_DMA_COPY_B
47 
48 #include "class/clc8b5.h"   // HOPPER_DMA_COPY_A
49 
50 #include "class/clc86f.h"   // HOPPER_CHANNEL_GPFIFO_A
51 
52 #include "gpu/conf_compute/ccsl.h"
53 
54 #include "nvctassert.h"
55 #include "vgpu/vgpu_guest_pma_scrubber.h"
56 
57 #define RM_SUBCHANNEL 0x0
58 
59 #define NV_PUSH_METHOD(OpType, SubCh, Method, Count)                    \
60     (DRF_DEF(906F, _DMA, _SEC_OP, OpType) |                             \
61      DRF_NUM(906F, _DMA, _METHOD_ADDRESS, (Method) >> 2) |              \
62      DRF_NUM(906F, _DMA, _METHOD_SUBCHANNEL, (SubCh)) |                 \
63      DRF_NUM(906F, _DMA, _METHOD_COUNT, (Count)))
64 
65 #define _NV_ASSERT_CONTIGUOUS_METHOD(a1, a2) NV_ASSERT((a2) - (a1) == 4)
66 
67 #define NV_PUSH_DATA(Data) MEM_WR32(pPtr++, (Data))
68 
69 #define _NV_PUSH_INC_1U(SubCh, a1, d1, Count)                           \
70     do                                                                  \
71     {                                                                   \
72         NV_PUSH_DATA(NV_PUSH_METHOD(_INC_METHOD, SubCh, a1, Count));    \
73         NV_PUSH_DATA(d1);                                               \
74     } while (0)
75 
76 #define NV_PUSH_INC_1U(SubCh, a1, d1)                                   \
77     do                                                                  \
78     {                                                                   \
79         _NV_PUSH_INC_1U (SubCh, a1, d1, 1);                             \
80     } while (0)
81 
82 #define NV_PUSH_INC_2U(SubCh, a1, d1, a2, d2)                           \
83     do                                                                  \
84     {                                                                   \
85         _NV_ASSERT_CONTIGUOUS_METHOD(a1, a2);                           \
86         _NV_PUSH_INC_1U(SubCh, a1, d1, 2);                              \
87         NV_PUSH_DATA(d2);                                               \
88     } while (0)
89 
90 #define NV_PUSH_INC_3U(SubCh, a1, d1, a2, d2, a3, d3)                   \
91     do                                                                  \
92     {                                                                   \
93         _NV_ASSERT_CONTIGUOUS_METHOD(a1, a2);                           \
94         _NV_ASSERT_CONTIGUOUS_METHOD(a2, a3);                           \
95         _NV_PUSH_INC_1U(SubCh, a1, d1, 3);                              \
96         NV_PUSH_DATA(d2);                                               \
97         NV_PUSH_DATA(d3);                                               \
98     } while (0)
99 
100 #define NV_PUSH_INC_4U(SubCh, a1, d1, a2, d2, a3, d3, a4, d4)           \
101     do                                                                  \
102     {                                                                   \
103         _NV_ASSERT_CONTIGUOUS_METHOD(a1, a2);                           \
104         _NV_ASSERT_CONTIGUOUS_METHOD(a2, a3);                           \
105         _NV_ASSERT_CONTIGUOUS_METHOD(a3, a4);                           \
106         _NV_PUSH_INC_1U(SubCh, a1, d1, 4);                              \
107         NV_PUSH_DATA(d2);                                               \
108         NV_PUSH_DATA(d3);                                               \
109         NV_PUSH_DATA(d4);                                               \
110     } while (0)
111 
112 #define READ_CHANNEL_PAYLOAD_SEMA(channel)  channelReadChannelMemdesc(channel, channel->finishPayloadOffset)
113 #define READ_CHANNEL_PB_SEMA(channel)       channelReadChannelMemdesc(channel, channel->semaOffset)
114 
115 //
116 // This struct contains parameters needed to send a pushbuffer for a CE
117 // operation. This interface only supports contiguous operations.
118 //
119 typedef struct
120 {
121     NvBool bCeMemcopy;   // Whether this is a CE memcopy;
122                          // If set to false, this will be a memset operation
123     NvU64 dstAddr;       // Physical address of the source address
124     NvU64 srcAddr;       // Physical address of the source address; only valid for memcopy
125     NvU32 size;
126     NvU32 pattern;       // Fixed pattern to memset to. Only valid for memset
127     NvU32 payload;       // Payload value used to release semaphore
128     NvU64 clientSemaAddr;
129     NV_ADDRESS_SPACE dstAddressSpace;
130     NV_ADDRESS_SPACE srcAddressSpace;
131     NvU32 dstCpuCacheAttrib;
132     NvU32 srcCpuCacheAttrib;
133 
134     NvBool bSecureCopy; // The copy encrypts/decrypts protected memory
135     NvBool bEncrypt; // encrypt/decrypt
136     NvU64 authTagAddr;
137     NvU64 encryptIvAddr;
138 
139 } CHANNEL_PB_INFO;
140 
141 NV_STATUS channelSetupIDs(OBJCHANNEL *pChannel, OBJGPU *pGpu, NvBool bUseVasForCeCopy, NvBool bMIGInUse);
142 NV_STATUS channelAllocSubdevice(OBJGPU *pGpu, OBJCHANNEL *pChannel);
143 void channelSetupChannelBufferSizes(OBJCHANNEL *pChannel);
144 NvU32 channelReadChannelMemdesc(OBJCHANNEL *pChannel, NvU32 offset);
145 
146 // Needed for pushbuffer management
147 NV_STATUS channelWaitForFreeEntry(OBJCHANNEL *pChannel, NvU32 *pPutIndex);
148 NV_STATUS channelFillGpFifo(OBJCHANNEL *pChannel, NvU32 putIndex, NvU32 methodsLength);
149 NvU32 channelFillCePb(OBJCHANNEL *pChannel, NvU32 putIndex, NvBool bPipelined,
150                       NvBool bInsertFinishPayload, CHANNEL_PB_INFO *pChannelPbInfo);
151 NvU32 channelFillPbFastScrub(OBJCHANNEL *pChannel, NvU32 putIndex, NvBool bPipelined,
152                     NvBool bInsertFinishPayload, CHANNEL_PB_INFO *pChannelPbInfo);
153 
154 NV_STATUS channelFillSec2Pb(OBJCHANNEL *pChannel, NvU32 putIndex, NvBool bInsertFinishPayload,
155                             CHANNEL_PB_INFO *pChannelPbInfo, CCSL_CONTEXT *pCcslCtx,
156                             MEMORY_DESCRIPTOR *pScrubMemDesc, MEMORY_DESCRIPTOR *pSemaMemDesc,
157                             NvU64 scrubMthdAuthTagBufGpuVA, NvU32 scrubAuthTagBufIndex,
158                             NvU64 semaMthdAuthTagBufGpuVA, NvU32 semaAuthTagBufIndex, NvU32* methodLength);
159 
160 // Needed for work tracking
161 NV_STATUS channelWaitForFinishPayload(OBJCHANNEL *pChannel, NvU64 targetPayload);
162 NvU64 channelGetFinishPayload(OBJCHANNEL *pChannel);
163 
164 void channelServiceScrubberInterrupts(OBJCHANNEL *pChannel);
165 
166 #endif // _CHANNEL_UTILS_H_
167