1 /* 2 * Copyright 2019 Red Hat Inc. 3 * 4 * Permission is hereby granted, free of charge, to any person obtaining a 5 * copy of this software and associated documentation files (the "Software"), 6 * to deal in the Software without restriction, including without limitation 7 * the rights to use, copy, modify, merge, publish, distribute, sublicense, 8 * and/or sell copies of the Software, and to permit persons to whom the 9 * Software is furnished to do so, subject to the following conditions: 10 * 11 * The above copyright notice and this permission notice shall be included in 12 * all copies or substantial portions of the Software. 13 * 14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 17 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR 18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, 19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR 20 * OTHER DEALINGS IN THE SOFTWARE. 21 */ 22 #include "gf100.h" 23 #include "ctxgf100.h" 24 25 #include <core/firmware.h> 26 #include <subdev/gsp.h> 27 #include <subdev/acr.h> 28 #include <subdev/timer.h> 29 #include <subdev/vfn.h> 30 31 #include <nvfw/flcn.h> 32 33 #include <nvif/class.h> 34 35 static void 36 ga102_gr_zbc_clear_color(struct gf100_gr *gr, int zbc) 37 { 38 struct nvkm_device *device = gr->base.engine.subdev.device; 39 u32 invalid[] = { 0, 0, 0, 0 }, *color; 40 41 if (gr->zbc_color[zbc].format) 42 color = gr->zbc_color[zbc].l2; 43 else 44 color = invalid; 45 46 nvkm_mask(device, 0x41bcb4, 0x0000001f, zbc); 47 nvkm_wr32(device, 0x41bcec, color[0]); 48 nvkm_wr32(device, 0x41bcf0, color[1]); 49 nvkm_wr32(device, 0x41bcf4, color[2]); 50 nvkm_wr32(device, 0x41bcf8, color[3]); 51 } 52 53 static const struct gf100_gr_func_zbc 54 ga102_gr_zbc = { 55 .clear_color = ga102_gr_zbc_clear_color, 56 .clear_depth = gp100_gr_zbc_clear_depth, 57 .stencil_get = gp102_gr_zbc_stencil_get, 58 .clear_stencil = gp102_gr_zbc_clear_stencil, 59 }; 60 61 static void 62 ga102_gr_gpccs_reset(struct gf100_gr *gr) 63 { 64 struct nvkm_device *device = gr->base.engine.subdev.device; 65 66 nvkm_wr32(device, 0x41a610, 0x00000000); 67 nvkm_msec(device, 1, NVKM_DELAY); 68 nvkm_wr32(device, 0x41a610, 0x00000001); 69 } 70 71 static const struct nvkm_acr_lsf_func 72 ga102_gr_gpccs_acr = { 73 .flags = NVKM_ACR_LSF_FORCE_PRIV_LOAD, 74 .bl_entry = 0x3400, 75 .bld_size = sizeof(struct flcn_bl_dmem_desc_v2), 76 .bld_write = gp108_gr_acr_bld_write, 77 .bld_patch = gp108_gr_acr_bld_patch, 78 }; 79 80 static void 81 ga102_gr_fecs_reset(struct gf100_gr *gr) 82 { 83 struct nvkm_device *device = gr->base.engine.subdev.device; 84 85 nvkm_wr32(device, 0x409614, 0x00000010); 86 nvkm_wr32(device, 0x41a614, 0x00000020); 87 nvkm_usec(device, 10, NVKM_DELAY); 88 nvkm_wr32(device, 0x409614, 0x00000110); 89 nvkm_wr32(device, 0x41a614, 0x00000a20); 90 nvkm_usec(device, 10, NVKM_DELAY); 91 nvkm_rd32(device, 0x409614); 92 nvkm_rd32(device, 0x41a614); 93 } 94 95 static const struct nvkm_acr_lsf_func 96 ga102_gr_fecs_acr = { 97 .bl_entry = 0x7e00, 98 .bld_size = sizeof(struct flcn_bl_dmem_desc_v2), 99 .bld_write = gp108_gr_acr_bld_write, 100 .bld_patch = gp108_gr_acr_bld_patch, 101 }; 102 103 static void 104 ga102_gr_init_rop_exceptions(struct gf100_gr *gr) 105 { 106 struct nvkm_device *device = gr->base.engine.subdev.device; 107 108 nvkm_wr32(device, 0x41bcbc, 0x40000000); 109 nvkm_wr32(device, 0x41bc38, 0x40000000); 110 nvkm_wr32(device, 0x41ac94, nvkm_rd32(device, 0x502c94)); 111 } 112 113 static void 114 ga102_gr_init_40a790(struct gf100_gr *gr) 115 { 116 nvkm_wr32(gr->base.engine.subdev.device, 0x40a790, 0xc0000000); 117 } 118 119 static void 120 ga102_gr_init_gpc_mmu(struct gf100_gr *gr) 121 { 122 struct nvkm_device *device = gr->base.engine.subdev.device; 123 124 nvkm_wr32(device, 0x418880, nvkm_rd32(device, 0x100c80) & 0xf8001fff); 125 nvkm_wr32(device, 0x418894, 0x00000000); 126 127 nvkm_wr32(device, 0x4188b4, nvkm_rd32(device, 0x100cc8)); 128 nvkm_wr32(device, 0x4188b8, nvkm_rd32(device, 0x100ccc)); 129 nvkm_wr32(device, 0x4188b0, nvkm_rd32(device, 0x100cc4)); 130 } 131 132 static struct nvkm_intr * 133 ga102_gr_oneinit_intr(struct gf100_gr *gr, enum nvkm_intr_type *pvector) 134 { 135 struct nvkm_device *device = gr->base.engine.subdev.device; 136 137 *pvector = nvkm_rd32(device, 0x400154) & 0x00000fff; 138 return &device->vfn->intr; 139 } 140 141 static int 142 ga102_gr_nonstall(struct gf100_gr *gr) 143 { 144 return nvkm_rd32(gr->base.engine.subdev.device, 0x400160) & 0x00000fff; 145 } 146 147 static const struct gf100_gr_func 148 ga102_gr = { 149 .nonstall = ga102_gr_nonstall, 150 .oneinit_intr = ga102_gr_oneinit_intr, 151 .oneinit_tiles = gm200_gr_oneinit_tiles, 152 .oneinit_sm_id = gv100_gr_oneinit_sm_id, 153 .init = gf100_gr_init, 154 .init_419bd8 = gv100_gr_init_419bd8, 155 .init_gpc_mmu = ga102_gr_init_gpc_mmu, 156 .init_vsc_stream_master = gk104_gr_init_vsc_stream_master, 157 .init_zcull = tu102_gr_init_zcull, 158 .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, 159 .init_swdx_pes_mask = gp102_gr_init_swdx_pes_mask, 160 .init_fs = tu102_gr_init_fs, 161 .init_fecs_exceptions = tu102_gr_init_fecs_exceptions, 162 .init_40a790 = ga102_gr_init_40a790, 163 .init_ds_hww_esr_2 = gm200_gr_init_ds_hww_esr_2, 164 .init_sked_hww_esr = gk104_gr_init_sked_hww_esr, 165 .init_ppc_exceptions = gk104_gr_init_ppc_exceptions, 166 .init_504430 = gv100_gr_init_504430, 167 .init_shader_exceptions = gv100_gr_init_shader_exceptions, 168 .init_rop_exceptions = ga102_gr_init_rop_exceptions, 169 .init_4188a4 = gv100_gr_init_4188a4, 170 .trap_mp = gv100_gr_trap_mp, 171 .fecs.reset = ga102_gr_fecs_reset, 172 .gpccs.reset = ga102_gr_gpccs_reset, 173 .rops = gm200_gr_rops, 174 .gpc_nr = 7, 175 .tpc_nr = 6, 176 .ppc_nr = 3, 177 .grctx = &ga102_grctx, 178 .zbc = &ga102_gr_zbc, 179 .sclass = { 180 { -1, -1, FERMI_TWOD_A }, 181 { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, 182 { -1, -1, AMPERE_B, &gf100_fermi }, 183 { -1, -1, AMPERE_COMPUTE_B }, 184 {} 185 } 186 }; 187 188 MODULE_FIRMWARE("nvidia/ga102/gr/fecs_bl.bin"); 189 MODULE_FIRMWARE("nvidia/ga102/gr/fecs_sig.bin"); 190 MODULE_FIRMWARE("nvidia/ga102/gr/gpccs_bl.bin"); 191 MODULE_FIRMWARE("nvidia/ga102/gr/gpccs_sig.bin"); 192 MODULE_FIRMWARE("nvidia/ga102/gr/NET_img.bin"); 193 194 MODULE_FIRMWARE("nvidia/ga103/gr/fecs_bl.bin"); 195 MODULE_FIRMWARE("nvidia/ga103/gr/fecs_sig.bin"); 196 MODULE_FIRMWARE("nvidia/ga103/gr/gpccs_bl.bin"); 197 MODULE_FIRMWARE("nvidia/ga103/gr/gpccs_sig.bin"); 198 MODULE_FIRMWARE("nvidia/ga103/gr/NET_img.bin"); 199 200 MODULE_FIRMWARE("nvidia/ga104/gr/fecs_bl.bin"); 201 MODULE_FIRMWARE("nvidia/ga104/gr/fecs_sig.bin"); 202 MODULE_FIRMWARE("nvidia/ga104/gr/gpccs_bl.bin"); 203 MODULE_FIRMWARE("nvidia/ga104/gr/gpccs_sig.bin"); 204 MODULE_FIRMWARE("nvidia/ga104/gr/NET_img.bin"); 205 206 MODULE_FIRMWARE("nvidia/ga106/gr/fecs_bl.bin"); 207 MODULE_FIRMWARE("nvidia/ga106/gr/fecs_sig.bin"); 208 MODULE_FIRMWARE("nvidia/ga106/gr/gpccs_bl.bin"); 209 MODULE_FIRMWARE("nvidia/ga106/gr/gpccs_sig.bin"); 210 MODULE_FIRMWARE("nvidia/ga106/gr/NET_img.bin"); 211 212 MODULE_FIRMWARE("nvidia/ga107/gr/fecs_bl.bin"); 213 MODULE_FIRMWARE("nvidia/ga107/gr/fecs_sig.bin"); 214 MODULE_FIRMWARE("nvidia/ga107/gr/gpccs_bl.bin"); 215 MODULE_FIRMWARE("nvidia/ga107/gr/gpccs_sig.bin"); 216 MODULE_FIRMWARE("nvidia/ga107/gr/NET_img.bin"); 217 218 struct netlist_region { 219 u32 region_id; 220 u32 data_size; 221 u32 data_offset; 222 }; 223 224 struct netlist_image_header { 225 u32 version; 226 u32 regions; 227 }; 228 229 struct netlist_image { 230 struct netlist_image_header header; 231 struct netlist_region regions[]; 232 }; 233 234 struct netlist_av64 { 235 u32 addr; 236 u32 data_hi; 237 u32 data_lo; 238 }; 239 240 static int 241 ga102_gr_av64_to_init(struct nvkm_blob *blob, struct gf100_gr_pack **ppack) 242 { 243 struct gf100_gr_init *init; 244 struct gf100_gr_pack *pack; 245 int nent; 246 int i; 247 248 nent = (blob->size / sizeof(struct netlist_av64)); 249 pack = vzalloc((sizeof(*pack) * 2) + (sizeof(*init) * (nent + 1))); 250 if (!pack) 251 return -ENOMEM; 252 253 init = (void *)(pack + 2); 254 pack[0].init = init; 255 pack[0].type = 64; 256 257 for (i = 0; i < nent; i++) { 258 struct gf100_gr_init *ent = &init[i]; 259 struct netlist_av64 *av = &((struct netlist_av64 *)blob->data)[i]; 260 261 ent->addr = av->addr; 262 ent->data = ((u64)av->data_hi << 32) | av->data_lo; 263 ent->count = 1; 264 ent->pitch = 1; 265 } 266 267 *ppack = pack; 268 return 0; 269 } 270 271 static int 272 ga102_gr_load(struct gf100_gr *gr, int ver, const struct gf100_gr_fwif *fwif) 273 { 274 struct nvkm_subdev *subdev = &gr->base.engine.subdev; 275 const struct firmware *fw; 276 const struct netlist_image *net; 277 const struct netlist_region *fecs_inst = NULL; 278 const struct netlist_region *fecs_data = NULL; 279 const struct netlist_region *gpccs_inst = NULL; 280 const struct netlist_region *gpccs_data = NULL; 281 int ret, i; 282 283 ret = nvkm_firmware_get(subdev, "gr/NET_img", 0, &fw); 284 if (ret) 285 return ret; 286 287 net = (const void *)fw->data; 288 nvkm_debug(subdev, "netlist version %d, %d regions\n", 289 net->header.version, net->header.regions); 290 291 for (i = 0; i < net->header.regions; i++) { 292 const struct netlist_region *reg = &net->regions[i]; 293 struct nvkm_blob blob = { 294 .data = (void *)fw->data + reg->data_offset, 295 .size = reg->data_size, 296 }; 297 298 nvkm_debug(subdev, "\t%2d: %08x %08x\n", 299 reg->region_id, reg->data_offset, reg->data_size); 300 301 switch (reg->region_id) { 302 case 0: fecs_data = reg; break; 303 case 1: fecs_inst = reg; break; 304 case 2: gpccs_data = reg; break; 305 case 3: gpccs_inst = reg; break; 306 case 4: gk20a_gr_av_to_init(&blob, &gr->bundle); break; 307 case 5: gk20a_gr_aiv_to_init(&blob, &gr->sw_ctx); break; 308 case 7: gk20a_gr_av_to_method(&blob, &gr->method); break; 309 case 28: tu102_gr_av_to_init_veid(&blob, &gr->bundle_veid); break; 310 case 34: ga102_gr_av64_to_init(&blob, &gr->bundle64); break; 311 case 48: gk20a_gr_av_to_init(&blob, &gr->sw_nonctx1); break; 312 case 49: gk20a_gr_av_to_init(&blob, &gr->sw_nonctx2); break; 313 case 50: gk20a_gr_av_to_init(&blob, &gr->sw_nonctx3); break; 314 case 51: gk20a_gr_av_to_init(&blob, &gr->sw_nonctx4); break; 315 default: 316 break; 317 } 318 } 319 320 ret = nvkm_acr_lsfw_load_bl_sig_net(subdev, &gr->fecs.falcon, NVKM_ACR_LSF_FECS, 321 "gr/fecs_", ver, fwif->fecs, 322 fw->data + fecs_inst->data_offset, 323 fecs_inst->data_size, 324 fw->data + fecs_data->data_offset, 325 fecs_data->data_size); 326 if (ret) 327 return ret; 328 329 ret = nvkm_acr_lsfw_load_bl_sig_net(subdev, &gr->gpccs.falcon, NVKM_ACR_LSF_GPCCS, 330 "gr/gpccs_", ver, fwif->gpccs, 331 fw->data + gpccs_inst->data_offset, 332 gpccs_inst->data_size, 333 fw->data + gpccs_data->data_offset, 334 gpccs_data->data_size); 335 if (ret) 336 return ret; 337 338 gr->firmware = true; 339 340 nvkm_firmware_put(fw); 341 return 0; 342 } 343 344 static const struct gf100_gr_fwif 345 ga102_gr_fwif[] = { 346 { 0, ga102_gr_load, &ga102_gr, &ga102_gr_fecs_acr, &ga102_gr_gpccs_acr }, 347 { -1, gm200_gr_nofw }, 348 {} 349 }; 350 351 int 352 ga102_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) 353 { 354 if (nvkm_gsp_rm(device->gsp)) 355 return r535_gr_new(&ga102_gr, device, type, inst, pgr); 356 357 return gf100_gr_new_(ga102_gr_fwif, device, type, inst, pgr); 358 } 359