1 /* 2 * QEMU sPAPR PCI host originated from Uninorth PCI host 3 * 4 * Copyright (c) 2011 Alexey Kardashevskiy, IBM Corporation. 5 * Copyright (C) 2011 David Gibson, IBM Corporation. 6 * 7 * Permission is hereby granted, free of charge, to any person obtaining a copy 8 * of this software and associated documentation files (the "Software"), to deal 9 * in the Software without restriction, including without limitation the rights 10 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 11 * copies of the Software, and to permit persons to whom the Software is 12 * furnished to do so, subject to the following conditions: 13 * 14 * The above copyright notice and this permission notice shall be included in 15 * all copies or substantial portions of the Software. 16 * 17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 22 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 23 * THE SOFTWARE. 24 */ 25 #include "hw/hw.h" 26 #include "hw/pci/pci.h" 27 #include "hw/pci/msi.h" 28 #include "hw/pci/msix.h" 29 #include "hw/pci/pci_host.h" 30 #include "hw/ppc/spapr.h" 31 #include "hw/pci-host/spapr.h" 32 #include "exec/address-spaces.h" 33 #include <libfdt.h> 34 #include "trace.h" 35 #include "qemu/error-report.h" 36 37 #include "hw/pci/pci_bus.h" 38 39 /* Copied from the kernel arch/powerpc/platforms/pseries/msi.c */ 40 #define RTAS_QUERY_FN 0 41 #define RTAS_CHANGE_FN 1 42 #define RTAS_RESET_FN 2 43 #define RTAS_CHANGE_MSI_FN 3 44 #define RTAS_CHANGE_MSIX_FN 4 45 46 /* Interrupt types to return on RTAS_CHANGE_* */ 47 #define RTAS_TYPE_MSI 1 48 #define RTAS_TYPE_MSIX 2 49 50 static sPAPRPHBState *find_phb(sPAPREnvironment *spapr, uint64_t buid) 51 { 52 sPAPRPHBState *sphb; 53 54 QLIST_FOREACH(sphb, &spapr->phbs, list) { 55 if (sphb->buid != buid) { 56 continue; 57 } 58 return sphb; 59 } 60 61 return NULL; 62 } 63 64 static PCIDevice *find_dev(sPAPREnvironment *spapr, uint64_t buid, 65 uint32_t config_addr) 66 { 67 sPAPRPHBState *sphb = find_phb(spapr, buid); 68 PCIHostState *phb = PCI_HOST_BRIDGE(sphb); 69 int bus_num = (config_addr >> 16) & 0xFF; 70 int devfn = (config_addr >> 8) & 0xFF; 71 72 if (!phb) { 73 return NULL; 74 } 75 76 return pci_find_device(phb->bus, bus_num, devfn); 77 } 78 79 static uint32_t rtas_pci_cfgaddr(uint32_t arg) 80 { 81 /* This handles the encoding of extended config space addresses */ 82 return ((arg >> 20) & 0xf00) | (arg & 0xff); 83 } 84 85 static void finish_read_pci_config(sPAPREnvironment *spapr, uint64_t buid, 86 uint32_t addr, uint32_t size, 87 target_ulong rets) 88 { 89 PCIDevice *pci_dev; 90 uint32_t val; 91 92 if ((size != 1) && (size != 2) && (size != 4)) { 93 /* access must be 1, 2 or 4 bytes */ 94 rtas_st(rets, 0, RTAS_OUT_HW_ERROR); 95 return; 96 } 97 98 pci_dev = find_dev(spapr, buid, addr); 99 addr = rtas_pci_cfgaddr(addr); 100 101 if (!pci_dev || (addr % size) || (addr >= pci_config_size(pci_dev))) { 102 /* Access must be to a valid device, within bounds and 103 * naturally aligned */ 104 rtas_st(rets, 0, RTAS_OUT_HW_ERROR); 105 return; 106 } 107 108 val = pci_host_config_read_common(pci_dev, addr, 109 pci_config_size(pci_dev), size); 110 111 rtas_st(rets, 0, RTAS_OUT_SUCCESS); 112 rtas_st(rets, 1, val); 113 } 114 115 static void rtas_ibm_read_pci_config(PowerPCCPU *cpu, sPAPREnvironment *spapr, 116 uint32_t token, uint32_t nargs, 117 target_ulong args, 118 uint32_t nret, target_ulong rets) 119 { 120 uint64_t buid; 121 uint32_t size, addr; 122 123 if ((nargs != 4) || (nret != 2)) { 124 rtas_st(rets, 0, RTAS_OUT_HW_ERROR); 125 return; 126 } 127 128 buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2); 129 size = rtas_ld(args, 3); 130 addr = rtas_ld(args, 0); 131 132 finish_read_pci_config(spapr, buid, addr, size, rets); 133 } 134 135 static void rtas_read_pci_config(PowerPCCPU *cpu, sPAPREnvironment *spapr, 136 uint32_t token, uint32_t nargs, 137 target_ulong args, 138 uint32_t nret, target_ulong rets) 139 { 140 uint32_t size, addr; 141 142 if ((nargs != 2) || (nret != 2)) { 143 rtas_st(rets, 0, RTAS_OUT_HW_ERROR); 144 return; 145 } 146 147 size = rtas_ld(args, 1); 148 addr = rtas_ld(args, 0); 149 150 finish_read_pci_config(spapr, 0, addr, size, rets); 151 } 152 153 static void finish_write_pci_config(sPAPREnvironment *spapr, uint64_t buid, 154 uint32_t addr, uint32_t size, 155 uint32_t val, target_ulong rets) 156 { 157 PCIDevice *pci_dev; 158 159 if ((size != 1) && (size != 2) && (size != 4)) { 160 /* access must be 1, 2 or 4 bytes */ 161 rtas_st(rets, 0, RTAS_OUT_HW_ERROR); 162 return; 163 } 164 165 pci_dev = find_dev(spapr, buid, addr); 166 addr = rtas_pci_cfgaddr(addr); 167 168 if (!pci_dev || (addr % size) || (addr >= pci_config_size(pci_dev))) { 169 /* Access must be to a valid device, within bounds and 170 * naturally aligned */ 171 rtas_st(rets, 0, RTAS_OUT_HW_ERROR); 172 return; 173 } 174 175 pci_host_config_write_common(pci_dev, addr, pci_config_size(pci_dev), 176 val, size); 177 178 rtas_st(rets, 0, RTAS_OUT_SUCCESS); 179 } 180 181 static void rtas_ibm_write_pci_config(PowerPCCPU *cpu, sPAPREnvironment *spapr, 182 uint32_t token, uint32_t nargs, 183 target_ulong args, 184 uint32_t nret, target_ulong rets) 185 { 186 uint64_t buid; 187 uint32_t val, size, addr; 188 189 if ((nargs != 5) || (nret != 1)) { 190 rtas_st(rets, 0, RTAS_OUT_HW_ERROR); 191 return; 192 } 193 194 buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2); 195 val = rtas_ld(args, 4); 196 size = rtas_ld(args, 3); 197 addr = rtas_ld(args, 0); 198 199 finish_write_pci_config(spapr, buid, addr, size, val, rets); 200 } 201 202 static void rtas_write_pci_config(PowerPCCPU *cpu, sPAPREnvironment *spapr, 203 uint32_t token, uint32_t nargs, 204 target_ulong args, 205 uint32_t nret, target_ulong rets) 206 { 207 uint32_t val, size, addr; 208 209 if ((nargs != 3) || (nret != 1)) { 210 rtas_st(rets, 0, RTAS_OUT_HW_ERROR); 211 return; 212 } 213 214 215 val = rtas_ld(args, 2); 216 size = rtas_ld(args, 1); 217 addr = rtas_ld(args, 0); 218 219 finish_write_pci_config(spapr, 0, addr, size, val, rets); 220 } 221 222 /* 223 * Find an entry with config_addr or returns the empty one if not found AND 224 * alloc_new is set. 225 * At the moment the msi_table entries are never released so there is 226 * no point to look till the end of the list if we need to find the free entry. 227 */ 228 static int spapr_msicfg_find(sPAPRPHBState *phb, uint32_t config_addr, 229 bool alloc_new) 230 { 231 int i; 232 233 for (i = 0; i < SPAPR_MSIX_MAX_DEVS; ++i) { 234 if (!phb->msi_table[i].nvec) { 235 break; 236 } 237 if (phb->msi_table[i].config_addr == config_addr) { 238 return i; 239 } 240 } 241 if ((i < SPAPR_MSIX_MAX_DEVS) && alloc_new) { 242 trace_spapr_pci_msi("Allocating new MSI config", i, config_addr); 243 return i; 244 } 245 246 return -1; 247 } 248 249 /* 250 * Set MSI/MSIX message data. 251 * This is required for msi_notify()/msix_notify() which 252 * will write at the addresses via spapr_msi_write(). 253 */ 254 static void spapr_msi_setmsg(PCIDevice *pdev, hwaddr addr, bool msix, 255 unsigned first_irq, unsigned req_num) 256 { 257 unsigned i; 258 MSIMessage msg = { .address = addr, .data = first_irq }; 259 260 if (!msix) { 261 msi_set_message(pdev, msg); 262 trace_spapr_pci_msi_setup(pdev->name, 0, msg.address); 263 return; 264 } 265 266 for (i = 0; i < req_num; ++i, ++msg.data) { 267 msix_set_message(pdev, i, msg); 268 trace_spapr_pci_msi_setup(pdev->name, i, msg.address); 269 } 270 } 271 272 static void rtas_ibm_change_msi(PowerPCCPU *cpu, sPAPREnvironment *spapr, 273 uint32_t token, uint32_t nargs, 274 target_ulong args, uint32_t nret, 275 target_ulong rets) 276 { 277 uint32_t config_addr = rtas_ld(args, 0); 278 uint64_t buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2); 279 unsigned int func = rtas_ld(args, 3); 280 unsigned int req_num = rtas_ld(args, 4); /* 0 == remove all */ 281 unsigned int seq_num = rtas_ld(args, 5); 282 unsigned int ret_intr_type; 283 int ndev, irq; 284 sPAPRPHBState *phb = NULL; 285 PCIDevice *pdev = NULL; 286 287 switch (func) { 288 case RTAS_CHANGE_MSI_FN: 289 case RTAS_CHANGE_FN: 290 ret_intr_type = RTAS_TYPE_MSI; 291 break; 292 case RTAS_CHANGE_MSIX_FN: 293 ret_intr_type = RTAS_TYPE_MSIX; 294 break; 295 default: 296 error_report("rtas_ibm_change_msi(%u) is not implemented", func); 297 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR); 298 return; 299 } 300 301 /* Fins sPAPRPHBState */ 302 phb = find_phb(spapr, buid); 303 if (phb) { 304 pdev = find_dev(spapr, buid, config_addr); 305 } 306 if (!phb || !pdev) { 307 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR); 308 return; 309 } 310 311 /* Releasing MSIs */ 312 if (!req_num) { 313 ndev = spapr_msicfg_find(phb, config_addr, false); 314 if (ndev < 0) { 315 trace_spapr_pci_msi("MSI has not been enabled", -1, config_addr); 316 rtas_st(rets, 0, RTAS_OUT_HW_ERROR); 317 return; 318 } 319 trace_spapr_pci_msi("Released MSIs", ndev, config_addr); 320 rtas_st(rets, 0, RTAS_OUT_SUCCESS); 321 rtas_st(rets, 1, 0); 322 return; 323 } 324 325 /* Enabling MSI */ 326 327 /* Find a device number in the map to add or reuse the existing one */ 328 ndev = spapr_msicfg_find(phb, config_addr, true); 329 if (ndev >= SPAPR_MSIX_MAX_DEVS || ndev < 0) { 330 error_report("No free entry for a new MSI device"); 331 rtas_st(rets, 0, RTAS_OUT_HW_ERROR); 332 return; 333 } 334 trace_spapr_pci_msi("Configuring MSI", ndev, config_addr); 335 336 /* Check if there is an old config and MSI number has not changed */ 337 if (phb->msi_table[ndev].nvec && (req_num != phb->msi_table[ndev].nvec)) { 338 /* Unexpected behaviour */ 339 error_report("Cannot reuse MSI config for device#%d", ndev); 340 rtas_st(rets, 0, RTAS_OUT_HW_ERROR); 341 return; 342 } 343 344 /* There is no cached config, allocate MSIs */ 345 if (!phb->msi_table[ndev].nvec) { 346 irq = spapr_allocate_irq_block(req_num, false, 347 ret_intr_type == RTAS_TYPE_MSI); 348 if (irq < 0) { 349 error_report("Cannot allocate MSIs for device#%d", ndev); 350 rtas_st(rets, 0, RTAS_OUT_HW_ERROR); 351 return; 352 } 353 phb->msi_table[ndev].irq = irq; 354 phb->msi_table[ndev].nvec = req_num; 355 phb->msi_table[ndev].config_addr = config_addr; 356 } 357 358 /* Setup MSI/MSIX vectors in the device (via cfgspace or MSIX BAR) */ 359 spapr_msi_setmsg(pdev, spapr->msi_win_addr, ret_intr_type == RTAS_TYPE_MSIX, 360 phb->msi_table[ndev].irq, req_num); 361 362 rtas_st(rets, 0, RTAS_OUT_SUCCESS); 363 rtas_st(rets, 1, req_num); 364 rtas_st(rets, 2, ++seq_num); 365 rtas_st(rets, 3, ret_intr_type); 366 367 trace_spapr_pci_rtas_ibm_change_msi(func, req_num); 368 } 369 370 static void rtas_ibm_query_interrupt_source_number(PowerPCCPU *cpu, 371 sPAPREnvironment *spapr, 372 uint32_t token, 373 uint32_t nargs, 374 target_ulong args, 375 uint32_t nret, 376 target_ulong rets) 377 { 378 uint32_t config_addr = rtas_ld(args, 0); 379 uint64_t buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2); 380 unsigned int intr_src_num = -1, ioa_intr_num = rtas_ld(args, 3); 381 int ndev; 382 sPAPRPHBState *phb = NULL; 383 384 /* Fins sPAPRPHBState */ 385 phb = find_phb(spapr, buid); 386 if (!phb) { 387 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR); 388 return; 389 } 390 391 /* Find device descriptor and start IRQ */ 392 ndev = spapr_msicfg_find(phb, config_addr, false); 393 if (ndev < 0) { 394 trace_spapr_pci_msi("MSI has not been enabled", -1, config_addr); 395 rtas_st(rets, 0, RTAS_OUT_HW_ERROR); 396 return; 397 } 398 399 intr_src_num = phb->msi_table[ndev].irq + ioa_intr_num; 400 trace_spapr_pci_rtas_ibm_query_interrupt_source_number(ioa_intr_num, 401 intr_src_num); 402 403 rtas_st(rets, 0, RTAS_OUT_SUCCESS); 404 rtas_st(rets, 1, intr_src_num); 405 rtas_st(rets, 2, 1);/* 0 == level; 1 == edge */ 406 } 407 408 static int pci_spapr_swizzle(int slot, int pin) 409 { 410 return (slot + pin) % PCI_NUM_PINS; 411 } 412 413 static int pci_spapr_map_irq(PCIDevice *pci_dev, int irq_num) 414 { 415 /* 416 * Here we need to convert pci_dev + irq_num to some unique value 417 * which is less than number of IRQs on the specific bus (4). We 418 * use standard PCI swizzling, that is (slot number + pin number) 419 * % 4. 420 */ 421 return pci_spapr_swizzle(PCI_SLOT(pci_dev->devfn), irq_num); 422 } 423 424 static void pci_spapr_set_irq(void *opaque, int irq_num, int level) 425 { 426 /* 427 * Here we use the number returned by pci_spapr_map_irq to find a 428 * corresponding qemu_irq. 429 */ 430 sPAPRPHBState *phb = opaque; 431 432 trace_spapr_pci_lsi_set(phb->dtbusname, irq_num, phb->lsi_table[irq_num].irq); 433 qemu_set_irq(spapr_phb_lsi_qirq(phb, irq_num), level); 434 } 435 436 static PCIINTxRoute spapr_route_intx_pin_to_irq(void *opaque, int pin) 437 { 438 sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(opaque); 439 PCIINTxRoute route; 440 441 route.mode = PCI_INTX_ENABLED; 442 route.irq = sphb->lsi_table[pin].irq; 443 444 return route; 445 } 446 447 /* 448 * MSI/MSIX memory region implementation. 449 * The handler handles both MSI and MSIX. 450 * For MSI-X, the vector number is encoded as a part of the address, 451 * data is set to 0. 452 * For MSI, the vector number is encoded in least bits in data. 453 */ 454 static void spapr_msi_write(void *opaque, hwaddr addr, 455 uint64_t data, unsigned size) 456 { 457 uint32_t irq = data; 458 459 trace_spapr_pci_msi_write(addr, data, irq); 460 461 qemu_irq_pulse(xics_get_qirq(spapr->icp, irq)); 462 } 463 464 static const MemoryRegionOps spapr_msi_ops = { 465 /* There is no .read as the read result is undefined by PCI spec */ 466 .read = NULL, 467 .write = spapr_msi_write, 468 .endianness = DEVICE_LITTLE_ENDIAN 469 }; 470 471 void spapr_pci_msi_init(sPAPREnvironment *spapr, hwaddr addr) 472 { 473 uint64_t window_size = 4096; 474 475 /* 476 * As MSI/MSIX interrupts trigger by writing at MSI/MSIX vectors, 477 * we need to allocate some memory to catch those writes coming 478 * from msi_notify()/msix_notify(). 479 * As MSIMessage:addr is going to be the same and MSIMessage:data 480 * is going to be a VIRQ number, 4 bytes of the MSI MR will only 481 * be used. 482 * 483 * For KVM we want to ensure that this memory is a full page so that 484 * our memory slot is of page size granularity. 485 */ 486 #ifdef CONFIG_KVM 487 if (kvm_enabled()) { 488 window_size = getpagesize(); 489 } 490 #endif 491 492 spapr->msi_win_addr = addr; 493 memory_region_init_io(&spapr->msiwindow, NULL, &spapr_msi_ops, spapr, 494 "msi", window_size); 495 memory_region_add_subregion(get_system_memory(), spapr->msi_win_addr, 496 &spapr->msiwindow); 497 } 498 499 /* 500 * PHB PCI device 501 */ 502 static AddressSpace *spapr_pci_dma_iommu(PCIBus *bus, void *opaque, int devfn) 503 { 504 sPAPRPHBState *phb = opaque; 505 506 return &phb->iommu_as; 507 } 508 509 static void spapr_phb_realize(DeviceState *dev, Error **errp) 510 { 511 SysBusDevice *s = SYS_BUS_DEVICE(dev); 512 sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(s); 513 PCIHostState *phb = PCI_HOST_BRIDGE(s); 514 const char *busname; 515 char *namebuf; 516 int i; 517 PCIBus *bus; 518 519 if (sphb->index != -1) { 520 hwaddr windows_base; 521 522 if ((sphb->buid != -1) || (sphb->dma_liobn != -1) 523 || (sphb->mem_win_addr != -1) 524 || (sphb->io_win_addr != -1)) { 525 error_setg(errp, "Either \"index\" or other parameters must" 526 " be specified for PAPR PHB, not both"); 527 return; 528 } 529 530 sphb->buid = SPAPR_PCI_BASE_BUID + sphb->index; 531 sphb->dma_liobn = SPAPR_PCI_BASE_LIOBN + sphb->index; 532 533 windows_base = SPAPR_PCI_WINDOW_BASE 534 + sphb->index * SPAPR_PCI_WINDOW_SPACING; 535 sphb->mem_win_addr = windows_base + SPAPR_PCI_MMIO_WIN_OFF; 536 sphb->io_win_addr = windows_base + SPAPR_PCI_IO_WIN_OFF; 537 } 538 539 if (sphb->buid == -1) { 540 error_setg(errp, "BUID not specified for PHB"); 541 return; 542 } 543 544 if (sphb->dma_liobn == -1) { 545 error_setg(errp, "LIOBN not specified for PHB"); 546 return; 547 } 548 549 if (sphb->mem_win_addr == -1) { 550 error_setg(errp, "Memory window address not specified for PHB"); 551 return; 552 } 553 554 if (sphb->io_win_addr == -1) { 555 error_setg(errp, "IO window address not specified for PHB"); 556 return; 557 } 558 559 if (find_phb(spapr, sphb->buid)) { 560 error_setg(errp, "PCI host bridges must have unique BUIDs"); 561 return; 562 } 563 564 sphb->dtbusname = g_strdup_printf("pci@%" PRIx64, sphb->buid); 565 566 namebuf = alloca(strlen(sphb->dtbusname) + 32); 567 568 /* Initialize memory regions */ 569 sprintf(namebuf, "%s.mmio", sphb->dtbusname); 570 memory_region_init(&sphb->memspace, OBJECT(sphb), namebuf, UINT64_MAX); 571 572 sprintf(namebuf, "%s.mmio-alias", sphb->dtbusname); 573 memory_region_init_alias(&sphb->memwindow, OBJECT(sphb), 574 namebuf, &sphb->memspace, 575 SPAPR_PCI_MEM_WIN_BUS_OFFSET, sphb->mem_win_size); 576 memory_region_add_subregion(get_system_memory(), sphb->mem_win_addr, 577 &sphb->memwindow); 578 579 /* On ppc, we only have MMIO no specific IO space from the CPU 580 * perspective. In theory we ought to be able to embed the PCI IO 581 * memory region direction in the system memory space. However, 582 * if any of the IO BAR subregions use the old_portio mechanism, 583 * that won't be processed properly unless accessed from the 584 * system io address space. This hack to bounce things via 585 * system_io works around the problem until all the users of 586 * old_portion are updated */ 587 sprintf(namebuf, "%s.io", sphb->dtbusname); 588 memory_region_init(&sphb->iospace, OBJECT(sphb), 589 namebuf, SPAPR_PCI_IO_WIN_SIZE); 590 /* FIXME: fix to support multiple PHBs */ 591 memory_region_add_subregion(get_system_io(), 0, &sphb->iospace); 592 593 sprintf(namebuf, "%s.io-alias", sphb->dtbusname); 594 memory_region_init_alias(&sphb->iowindow, OBJECT(sphb), namebuf, 595 get_system_io(), 0, SPAPR_PCI_IO_WIN_SIZE); 596 memory_region_add_subregion(get_system_memory(), sphb->io_win_addr, 597 &sphb->iowindow); 598 /* 599 * Selecting a busname is more complex than you'd think, due to 600 * interacting constraints. If the user has specified an id 601 * explicitly for the phb , then we want to use the qdev default 602 * of naming the bus based on the bridge device (so the user can 603 * then assign devices to it in the way they expect). For the 604 * first / default PCI bus (index=0) we want to use just "pci" 605 * because libvirt expects there to be a bus called, simply, 606 * "pci". Otherwise, we use the same name as in the device tree, 607 * since it's unique by construction, and makes the guest visible 608 * BUID clear. 609 */ 610 if (dev->id) { 611 busname = NULL; 612 } else if (sphb->index == 0) { 613 busname = "pci"; 614 } else { 615 busname = sphb->dtbusname; 616 } 617 bus = pci_register_bus(dev, busname, 618 pci_spapr_set_irq, pci_spapr_map_irq, sphb, 619 &sphb->memspace, &sphb->iospace, 620 PCI_DEVFN(0, 0), PCI_NUM_PINS, TYPE_PCI_BUS); 621 phb->bus = bus; 622 623 sphb->dma_window_start = 0; 624 sphb->dma_window_size = 0x40000000; 625 sphb->tcet = spapr_tce_new_table(dev, sphb->dma_liobn, 626 sphb->dma_window_size); 627 if (!sphb->tcet) { 628 error_setg(errp, "Unable to create TCE table for %s", 629 sphb->dtbusname); 630 return; 631 } 632 address_space_init(&sphb->iommu_as, spapr_tce_get_iommu(sphb->tcet), 633 sphb->dtbusname); 634 635 pci_setup_iommu(bus, spapr_pci_dma_iommu, sphb); 636 637 pci_bus_set_route_irq_fn(bus, spapr_route_intx_pin_to_irq); 638 639 QLIST_INSERT_HEAD(&spapr->phbs, sphb, list); 640 641 /* Initialize the LSI table */ 642 for (i = 0; i < PCI_NUM_PINS; i++) { 643 uint32_t irq; 644 645 irq = spapr_allocate_lsi(0); 646 if (!irq) { 647 error_setg(errp, "spapr_allocate_lsi failed"); 648 return; 649 } 650 651 sphb->lsi_table[i].irq = irq; 652 } 653 } 654 655 static void spapr_phb_reset(DeviceState *qdev) 656 { 657 SysBusDevice *s = SYS_BUS_DEVICE(qdev); 658 sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(s); 659 660 /* Reset the IOMMU state */ 661 device_reset(DEVICE(sphb->tcet)); 662 } 663 664 static Property spapr_phb_properties[] = { 665 DEFINE_PROP_INT32("index", sPAPRPHBState, index, -1), 666 DEFINE_PROP_UINT64("buid", sPAPRPHBState, buid, -1), 667 DEFINE_PROP_UINT32("liobn", sPAPRPHBState, dma_liobn, -1), 668 DEFINE_PROP_UINT64("mem_win_addr", sPAPRPHBState, mem_win_addr, -1), 669 DEFINE_PROP_UINT64("mem_win_size", sPAPRPHBState, mem_win_size, 670 SPAPR_PCI_MMIO_WIN_SIZE), 671 DEFINE_PROP_UINT64("io_win_addr", sPAPRPHBState, io_win_addr, -1), 672 DEFINE_PROP_UINT64("io_win_size", sPAPRPHBState, io_win_size, 673 SPAPR_PCI_IO_WIN_SIZE), 674 DEFINE_PROP_END_OF_LIST(), 675 }; 676 677 static const VMStateDescription vmstate_spapr_pci_lsi = { 678 .name = "spapr_pci/lsi", 679 .version_id = 1, 680 .minimum_version_id = 1, 681 .minimum_version_id_old = 1, 682 .fields = (VMStateField []) { 683 VMSTATE_UINT32_EQUAL(irq, struct spapr_pci_lsi), 684 685 VMSTATE_END_OF_LIST() 686 }, 687 }; 688 689 static const VMStateDescription vmstate_spapr_pci_msi = { 690 .name = "spapr_pci/lsi", 691 .version_id = 1, 692 .minimum_version_id = 1, 693 .minimum_version_id_old = 1, 694 .fields = (VMStateField []) { 695 VMSTATE_UINT32(config_addr, struct spapr_pci_msi), 696 VMSTATE_UINT32(irq, struct spapr_pci_msi), 697 VMSTATE_UINT32(nvec, struct spapr_pci_msi), 698 699 VMSTATE_END_OF_LIST() 700 }, 701 }; 702 703 static const VMStateDescription vmstate_spapr_pci = { 704 .name = "spapr_pci", 705 .version_id = 1, 706 .minimum_version_id = 1, 707 .minimum_version_id_old = 1, 708 .fields = (VMStateField []) { 709 VMSTATE_UINT64_EQUAL(buid, sPAPRPHBState), 710 VMSTATE_UINT32_EQUAL(dma_liobn, sPAPRPHBState), 711 VMSTATE_UINT64_EQUAL(mem_win_addr, sPAPRPHBState), 712 VMSTATE_UINT64_EQUAL(mem_win_size, sPAPRPHBState), 713 VMSTATE_UINT64_EQUAL(io_win_addr, sPAPRPHBState), 714 VMSTATE_UINT64_EQUAL(io_win_size, sPAPRPHBState), 715 VMSTATE_STRUCT_ARRAY(lsi_table, sPAPRPHBState, PCI_NUM_PINS, 0, 716 vmstate_spapr_pci_lsi, struct spapr_pci_lsi), 717 VMSTATE_STRUCT_ARRAY(msi_table, sPAPRPHBState, SPAPR_MSIX_MAX_DEVS, 0, 718 vmstate_spapr_pci_msi, struct spapr_pci_msi), 719 720 VMSTATE_END_OF_LIST() 721 }, 722 }; 723 724 static const char *spapr_phb_root_bus_path(PCIHostState *host_bridge, 725 PCIBus *rootbus) 726 { 727 sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(host_bridge); 728 729 return sphb->dtbusname; 730 } 731 732 static void spapr_phb_class_init(ObjectClass *klass, void *data) 733 { 734 PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(klass); 735 DeviceClass *dc = DEVICE_CLASS(klass); 736 737 hc->root_bus_path = spapr_phb_root_bus_path; 738 dc->realize = spapr_phb_realize; 739 dc->props = spapr_phb_properties; 740 dc->reset = spapr_phb_reset; 741 dc->vmsd = &vmstate_spapr_pci; 742 set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories); 743 dc->cannot_instantiate_with_device_add_yet = false; 744 } 745 746 static const TypeInfo spapr_phb_info = { 747 .name = TYPE_SPAPR_PCI_HOST_BRIDGE, 748 .parent = TYPE_PCI_HOST_BRIDGE, 749 .instance_size = sizeof(sPAPRPHBState), 750 .class_init = spapr_phb_class_init, 751 }; 752 753 PCIHostState *spapr_create_phb(sPAPREnvironment *spapr, int index) 754 { 755 DeviceState *dev; 756 757 dev = qdev_create(NULL, TYPE_SPAPR_PCI_HOST_BRIDGE); 758 qdev_prop_set_uint32(dev, "index", index); 759 qdev_init_nofail(dev); 760 761 return PCI_HOST_BRIDGE(dev); 762 } 763 764 /* Macros to operate with address in OF binding to PCI */ 765 #define b_x(x, p, l) (((x) & ((1<<(l))-1)) << (p)) 766 #define b_n(x) b_x((x), 31, 1) /* 0 if relocatable */ 767 #define b_p(x) b_x((x), 30, 1) /* 1 if prefetchable */ 768 #define b_t(x) b_x((x), 29, 1) /* 1 if the address is aliased */ 769 #define b_ss(x) b_x((x), 24, 2) /* the space code */ 770 #define b_bbbbbbbb(x) b_x((x), 16, 8) /* bus number */ 771 #define b_ddddd(x) b_x((x), 11, 5) /* device number */ 772 #define b_fff(x) b_x((x), 8, 3) /* function number */ 773 #define b_rrrrrrrr(x) b_x((x), 0, 8) /* register number */ 774 775 int spapr_populate_pci_dt(sPAPRPHBState *phb, 776 uint32_t xics_phandle, 777 void *fdt) 778 { 779 int bus_off, i, j; 780 char nodename[256]; 781 uint32_t bus_range[] = { cpu_to_be32(0), cpu_to_be32(0xff) }; 782 struct { 783 uint32_t hi; 784 uint64_t child; 785 uint64_t parent; 786 uint64_t size; 787 } QEMU_PACKED ranges[] = { 788 { 789 cpu_to_be32(b_ss(1)), cpu_to_be64(0), 790 cpu_to_be64(phb->io_win_addr), 791 cpu_to_be64(memory_region_size(&phb->iospace)), 792 }, 793 { 794 cpu_to_be32(b_ss(2)), cpu_to_be64(SPAPR_PCI_MEM_WIN_BUS_OFFSET), 795 cpu_to_be64(phb->mem_win_addr), 796 cpu_to_be64(memory_region_size(&phb->memwindow)), 797 }, 798 }; 799 uint64_t bus_reg[] = { cpu_to_be64(phb->buid), 0 }; 800 uint32_t interrupt_map_mask[] = { 801 cpu_to_be32(b_ddddd(-1)|b_fff(0)), 0x0, 0x0, cpu_to_be32(-1)}; 802 uint32_t interrupt_map[PCI_SLOT_MAX * PCI_NUM_PINS][7]; 803 804 /* Start populating the FDT */ 805 sprintf(nodename, "pci@%" PRIx64, phb->buid); 806 bus_off = fdt_add_subnode(fdt, 0, nodename); 807 if (bus_off < 0) { 808 return bus_off; 809 } 810 811 #define _FDT(exp) \ 812 do { \ 813 int ret = (exp); \ 814 if (ret < 0) { \ 815 return ret; \ 816 } \ 817 } while (0) 818 819 /* Write PHB properties */ 820 _FDT(fdt_setprop_string(fdt, bus_off, "device_type", "pci")); 821 _FDT(fdt_setprop_string(fdt, bus_off, "compatible", "IBM,Logical_PHB")); 822 _FDT(fdt_setprop_cell(fdt, bus_off, "#address-cells", 0x3)); 823 _FDT(fdt_setprop_cell(fdt, bus_off, "#size-cells", 0x2)); 824 _FDT(fdt_setprop_cell(fdt, bus_off, "#interrupt-cells", 0x1)); 825 _FDT(fdt_setprop(fdt, bus_off, "used-by-rtas", NULL, 0)); 826 _FDT(fdt_setprop(fdt, bus_off, "bus-range", &bus_range, sizeof(bus_range))); 827 _FDT(fdt_setprop(fdt, bus_off, "ranges", &ranges, sizeof(ranges))); 828 _FDT(fdt_setprop(fdt, bus_off, "reg", &bus_reg, sizeof(bus_reg))); 829 _FDT(fdt_setprop_cell(fdt, bus_off, "ibm,pci-config-space-type", 0x1)); 830 831 /* Build the interrupt-map, this must matches what is done 832 * in pci_spapr_map_irq 833 */ 834 _FDT(fdt_setprop(fdt, bus_off, "interrupt-map-mask", 835 &interrupt_map_mask, sizeof(interrupt_map_mask))); 836 for (i = 0; i < PCI_SLOT_MAX; i++) { 837 for (j = 0; j < PCI_NUM_PINS; j++) { 838 uint32_t *irqmap = interrupt_map[i*PCI_NUM_PINS + j]; 839 int lsi_num = pci_spapr_swizzle(i, j); 840 841 irqmap[0] = cpu_to_be32(b_ddddd(i)|b_fff(0)); 842 irqmap[1] = 0; 843 irqmap[2] = 0; 844 irqmap[3] = cpu_to_be32(j+1); 845 irqmap[4] = cpu_to_be32(xics_phandle); 846 irqmap[5] = cpu_to_be32(phb->lsi_table[lsi_num].irq); 847 irqmap[6] = cpu_to_be32(0x8); 848 } 849 } 850 /* Write interrupt map */ 851 _FDT(fdt_setprop(fdt, bus_off, "interrupt-map", &interrupt_map, 852 sizeof(interrupt_map))); 853 854 spapr_dma_dt(fdt, bus_off, "ibm,dma-window", 855 phb->dma_liobn, phb->dma_window_start, 856 phb->dma_window_size); 857 858 return 0; 859 } 860 861 void spapr_pci_rtas_init(void) 862 { 863 spapr_rtas_register("read-pci-config", rtas_read_pci_config); 864 spapr_rtas_register("write-pci-config", rtas_write_pci_config); 865 spapr_rtas_register("ibm,read-pci-config", rtas_ibm_read_pci_config); 866 spapr_rtas_register("ibm,write-pci-config", rtas_ibm_write_pci_config); 867 if (msi_supported) { 868 spapr_rtas_register("ibm,query-interrupt-source-number", 869 rtas_ibm_query_interrupt_source_number); 870 spapr_rtas_register("ibm,change-msi", rtas_ibm_change_msi); 871 } 872 } 873 874 static void spapr_pci_register_types(void) 875 { 876 type_register_static(&spapr_phb_info); 877 } 878 879 type_init(spapr_pci_register_types) 880