]> git.proxmox.com Git - mirror_qemu.git/blob - hw/ppc/spapr_pci.c
projects / mirror_qemu.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge remote-tracking branch 'remotes/stefanha/tags/tracing-pull-request' into staging
[mirror_qemu.git] / hw / ppc / spapr_pci.c
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 "qemu/osdep.h"
26 #include "qapi/error.h"
27 #include "qemu-common.h"
28 #include "cpu.h"
29 #include "hw/hw.h"
30 #include "hw/sysbus.h"
31 #include "hw/pci/pci.h"
32 #include "hw/pci/msi.h"
33 #include "hw/pci/msix.h"
34 #include "hw/pci/pci_host.h"
35 #include "hw/ppc/spapr.h"
36 #include "hw/pci-host/spapr.h"
37 #include "exec/address-spaces.h"
38 #include "exec/ram_addr.h"
39 #include <libfdt.h>
40 #include "trace.h"
41 #include "qemu/error-report.h"
42 #include "qapi/qmp/qerror.h"
43 #include "hw/ppc/fdt.h"
44 #include "hw/pci/pci_bridge.h"
45 #include "hw/pci/pci_bus.h"
46 #include "hw/pci/pci_ids.h"
47 #include "hw/ppc/spapr_drc.h"
48 #include "sysemu/device_tree.h"
49 #include "sysemu/kvm.h"
50 #include "sysemu/hostmem.h"
51 #include "sysemu/numa.h"
52
53 /* Copied from the kernel arch/powerpc/platforms/pseries/msi.c */
54 #define RTAS_QUERY_FN 0
55 #define RTAS_CHANGE_FN 1
56 #define RTAS_RESET_FN 2
57 #define RTAS_CHANGE_MSI_FN 3
58 #define RTAS_CHANGE_MSIX_FN 4
59
60 /* Interrupt types to return on RTAS_CHANGE_* */
61 #define RTAS_TYPE_MSI 1
62 #define RTAS_TYPE_MSIX 2
63
64 sPAPRPHBState *spapr_pci_find_phb(sPAPRMachineState *spapr, uint64_t buid)
65 {
66 sPAPRPHBState *sphb;
67
68 QLIST_FOREACH(sphb, &spapr->phbs, list) {
69 if (sphb->buid != buid) {
70 continue;
71 }
72 return sphb;
73 }
74
75 return NULL;
76 }
77
78 PCIDevice *spapr_pci_find_dev(sPAPRMachineState *spapr, uint64_t buid,
79 uint32_t config_addr)
80 {
81 sPAPRPHBState *sphb = spapr_pci_find_phb(spapr, buid);
82 PCIHostState *phb = PCI_HOST_BRIDGE(sphb);
83 int bus_num = (config_addr >> 16) & 0xFF;
84 int devfn = (config_addr >> 8) & 0xFF;
85
86 if (!phb) {
87 return NULL;
88 }
89
90 return pci_find_device(phb->bus, bus_num, devfn);
91 }
92
93 static uint32_t rtas_pci_cfgaddr(uint32_t arg)
94 {
95 /* This handles the encoding of extended config space addresses */
96 return ((arg >> 20) & 0xf00) | (arg & 0xff);
97 }
98
99 static void finish_read_pci_config(sPAPRMachineState *spapr, uint64_t buid,
100 uint32_t addr, uint32_t size,
101 target_ulong rets)
102 {
103 PCIDevice *pci_dev;
104 uint32_t val;
105
106 if ((size != 1) && (size != 2) && (size != 4)) {
107 /* access must be 1, 2 or 4 bytes */
108 rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
109 return;
110 }
111
112 pci_dev = spapr_pci_find_dev(spapr, buid, addr);
113 addr = rtas_pci_cfgaddr(addr);
114
115 if (!pci_dev || (addr % size) || (addr >= pci_config_size(pci_dev))) {
116 /* Access must be to a valid device, within bounds and
117 * naturally aligned */
118 rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
119 return;
120 }
121
122 val = pci_host_config_read_common(pci_dev, addr,
123 pci_config_size(pci_dev), size);
124
125 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
126 rtas_st(rets, 1, val);
127 }
128
129 static void rtas_ibm_read_pci_config(PowerPCCPU *cpu, sPAPRMachineState *spapr,
130 uint32_t token, uint32_t nargs,
131 target_ulong args,
132 uint32_t nret, target_ulong rets)
133 {
134 uint64_t buid;
135 uint32_t size, addr;
136
137 if ((nargs != 4) || (nret != 2)) {
138 rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
139 return;
140 }
141
142 buid = rtas_ldq(args, 1);
143 size = rtas_ld(args, 3);
144 addr = rtas_ld(args, 0);
145
146 finish_read_pci_config(spapr, buid, addr, size, rets);
147 }
148
149 static void rtas_read_pci_config(PowerPCCPU *cpu, sPAPRMachineState *spapr,
150 uint32_t token, uint32_t nargs,
151 target_ulong args,
152 uint32_t nret, target_ulong rets)
153 {
154 uint32_t size, addr;
155
156 if ((nargs != 2) || (nret != 2)) {
157 rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
158 return;
159 }
160
161 size = rtas_ld(args, 1);
162 addr = rtas_ld(args, 0);
163
164 finish_read_pci_config(spapr, 0, addr, size, rets);
165 }
166
167 static void finish_write_pci_config(sPAPRMachineState *spapr, uint64_t buid,
168 uint32_t addr, uint32_t size,
169 uint32_t val, target_ulong rets)
170 {
171 PCIDevice *pci_dev;
172
173 if ((size != 1) && (size != 2) && (size != 4)) {
174 /* access must be 1, 2 or 4 bytes */
175 rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
176 return;
177 }
178
179 pci_dev = spapr_pci_find_dev(spapr, buid, addr);
180 addr = rtas_pci_cfgaddr(addr);
181
182 if (!pci_dev || (addr % size) || (addr >= pci_config_size(pci_dev))) {
183 /* Access must be to a valid device, within bounds and
184 * naturally aligned */
185 rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
186 return;
187 }
188
189 pci_host_config_write_common(pci_dev, addr, pci_config_size(pci_dev),
190 val, size);
191
192 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
193 }
194
195 static void rtas_ibm_write_pci_config(PowerPCCPU *cpu, sPAPRMachineState *spapr,
196 uint32_t token, uint32_t nargs,
197 target_ulong args,
198 uint32_t nret, target_ulong rets)
199 {
200 uint64_t buid;
201 uint32_t val, size, addr;
202
203 if ((nargs != 5) || (nret != 1)) {
204 rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
205 return;
206 }
207
208 buid = rtas_ldq(args, 1);
209 val = rtas_ld(args, 4);
210 size = rtas_ld(args, 3);
211 addr = rtas_ld(args, 0);
212
213 finish_write_pci_config(spapr, buid, addr, size, val, rets);
214 }
215
216 static void rtas_write_pci_config(PowerPCCPU *cpu, sPAPRMachineState *spapr,
217 uint32_t token, uint32_t nargs,
218 target_ulong args,
219 uint32_t nret, target_ulong rets)
220 {
221 uint32_t val, size, addr;
222
223 if ((nargs != 3) || (nret != 1)) {
224 rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
225 return;
226 }
227
228
229 val = rtas_ld(args, 2);
230 size = rtas_ld(args, 1);
231 addr = rtas_ld(args, 0);
232
233 finish_write_pci_config(spapr, 0, addr, size, val, rets);
234 }
235
236 /*
237 * Set MSI/MSIX message data.
238 * This is required for msi_notify()/msix_notify() which
239 * will write at the addresses via spapr_msi_write().
240 *
241 * If hwaddr == 0, all entries will have .data == first_irq i.e.
242 * table will be reset.
243 */
244 static void spapr_msi_setmsg(PCIDevice *pdev, hwaddr addr, bool msix,
245 unsigned first_irq, unsigned req_num)
246 {
247 unsigned i;
248 MSIMessage msg = { .address = addr, .data = first_irq };
249
250 if (!msix) {
251 msi_set_message(pdev, msg);
252 trace_spapr_pci_msi_setup(pdev->name, 0, msg.address);
253 return;
254 }
255
256 for (i = 0; i < req_num; ++i) {
257 msix_set_message(pdev, i, msg);
258 trace_spapr_pci_msi_setup(pdev->name, i, msg.address);
259 if (addr) {
260 ++msg.data;
261 }
262 }
263 }
264
265 static void rtas_ibm_change_msi(PowerPCCPU *cpu, sPAPRMachineState *spapr,
266 uint32_t token, uint32_t nargs,
267 target_ulong args, uint32_t nret,
268 target_ulong rets)
269 {
270 uint32_t config_addr = rtas_ld(args, 0);
271 uint64_t buid = rtas_ldq(args, 1);
272 unsigned int func = rtas_ld(args, 3);
273 unsigned int req_num = rtas_ld(args, 4); /* 0 == remove all */
274 unsigned int seq_num = rtas_ld(args, 5);
275 unsigned int ret_intr_type;
276 unsigned int irq, max_irqs = 0;
277 sPAPRPHBState *phb = NULL;
278 PCIDevice *pdev = NULL;
279 spapr_pci_msi *msi;
280 int *config_addr_key;
281 Error *err = NULL;
282
283 /* Fins sPAPRPHBState */
284 phb = spapr_pci_find_phb(spapr, buid);
285 if (phb) {
286 pdev = spapr_pci_find_dev(spapr, buid, config_addr);
287 }
288 if (!phb || !pdev) {
289 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
290 return;
291 }
292
293 switch (func) {
294 case RTAS_CHANGE_FN:
295 if (msi_present(pdev)) {
296 ret_intr_type = RTAS_TYPE_MSI;
297 } else if (msix_present(pdev)) {
298 ret_intr_type = RTAS_TYPE_MSIX;
299 } else {
300 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
301 return;
302 }
303 break;
304 case RTAS_CHANGE_MSI_FN:
305 if (msi_present(pdev)) {
306 ret_intr_type = RTAS_TYPE_MSI;
307 } else {
308 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
309 return;
310 }
311 break;
312 case RTAS_CHANGE_MSIX_FN:
313 if (msix_present(pdev)) {
314 ret_intr_type = RTAS_TYPE_MSIX;
315 } else {
316 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
317 return;
318 }
319 break;
320 default:
321 error_report("rtas_ibm_change_msi(%u) is not implemented", func);
322 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
323 return;
324 }
325
326 msi = (spapr_pci_msi *) g_hash_table_lookup(phb->msi, &config_addr);
327
328 /* Releasing MSIs */
329 if (!req_num) {
330 if (!msi) {
331 trace_spapr_pci_msi("Releasing wrong config", config_addr);
332 rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
333 return;
334 }
335
336 spapr_irq_free(spapr, msi->first_irq, msi->num);
337 if (msi_present(pdev)) {
338 spapr_msi_setmsg(pdev, 0, false, 0, 0);
339 }
340 if (msix_present(pdev)) {
341 spapr_msi_setmsg(pdev, 0, true, 0, 0);
342 }
343 g_hash_table_remove(phb->msi, &config_addr);
344
345 trace_spapr_pci_msi("Released MSIs", config_addr);
346 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
347 rtas_st(rets, 1, 0);
348 return;
349 }
350
351 /* Enabling MSI */
352
353 /* Check if the device supports as many IRQs as requested */
354 if (ret_intr_type == RTAS_TYPE_MSI) {
355 max_irqs = msi_nr_vectors_allocated(pdev);
356 } else if (ret_intr_type == RTAS_TYPE_MSIX) {
357 max_irqs = pdev->msix_entries_nr;
358 }
359 if (!max_irqs) {
360 error_report("Requested interrupt type %d is not enabled for device %x",
361 ret_intr_type, config_addr);
362 rtas_st(rets, 0, -1); /* Hardware error */
363 return;
364 }
365 /* Correct the number if the guest asked for too many */
366 if (req_num > max_irqs) {
367 trace_spapr_pci_msi_retry(config_addr, req_num, max_irqs);
368 req_num = max_irqs;
369 irq = 0; /* to avoid misleading trace */
370 goto out;
371 }
372
373 /* Allocate MSIs */
374 irq = spapr_irq_alloc_block(spapr, req_num, false,
375 ret_intr_type == RTAS_TYPE_MSI, &err);
376 if (err) {
377 error_reportf_err(err, "Can't allocate MSIs for device %x: ",
378 config_addr);
379 rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
380 return;
381 }
382
383 /* Release previous MSIs */
384 if (msi) {
385 spapr_irq_free(spapr, msi->first_irq, msi->num);
386 g_hash_table_remove(phb->msi, &config_addr);
387 }
388
389 /* Setup MSI/MSIX vectors in the device (via cfgspace or MSIX BAR) */
390 spapr_msi_setmsg(pdev, SPAPR_PCI_MSI_WINDOW, ret_intr_type == RTAS_TYPE_MSIX,
391 irq, req_num);
392
393 /* Add MSI device to cache */
394 msi = g_new(spapr_pci_msi, 1);
395 msi->first_irq = irq;
396 msi->num = req_num;
397 config_addr_key = g_new(int, 1);
398 *config_addr_key = config_addr;
399 g_hash_table_insert(phb->msi, config_addr_key, msi);
400
401 out:
402 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
403 rtas_st(rets, 1, req_num);
404 rtas_st(rets, 2, ++seq_num);
405 if (nret > 3) {
406 rtas_st(rets, 3, ret_intr_type);
407 }
408
409 trace_spapr_pci_rtas_ibm_change_msi(config_addr, func, req_num, irq);
410 }
411
412 static void rtas_ibm_query_interrupt_source_number(PowerPCCPU *cpu,
413 sPAPRMachineState *spapr,
414 uint32_t token,
415 uint32_t nargs,
416 target_ulong args,
417 uint32_t nret,
418 target_ulong rets)
419 {
420 uint32_t config_addr = rtas_ld(args, 0);
421 uint64_t buid = rtas_ldq(args, 1);
422 unsigned int intr_src_num = -1, ioa_intr_num = rtas_ld(args, 3);
423 sPAPRPHBState *phb = NULL;
424 PCIDevice *pdev = NULL;
425 spapr_pci_msi *msi;
426
427 /* Find sPAPRPHBState */
428 phb = spapr_pci_find_phb(spapr, buid);
429 if (phb) {
430 pdev = spapr_pci_find_dev(spapr, buid, config_addr);
431 }
432 if (!phb || !pdev) {
433 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
434 return;
435 }
436
437 /* Find device descriptor and start IRQ */
438 msi = (spapr_pci_msi *) g_hash_table_lookup(phb->msi, &config_addr);
439 if (!msi || !msi->first_irq || !msi->num || (ioa_intr_num >= msi->num)) {
440 trace_spapr_pci_msi("Failed to return vector", config_addr);
441 rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
442 return;
443 }
444 intr_src_num = msi->first_irq + ioa_intr_num;
445 trace_spapr_pci_rtas_ibm_query_interrupt_source_number(ioa_intr_num,
446 intr_src_num);
447
448 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
449 rtas_st(rets, 1, intr_src_num);
450 rtas_st(rets, 2, 1);/* 0 == level; 1 == edge */
451 }
452
453 static void rtas_ibm_set_eeh_option(PowerPCCPU *cpu,
454 sPAPRMachineState *spapr,
455 uint32_t token, uint32_t nargs,
456 target_ulong args, uint32_t nret,
457 target_ulong rets)
458 {
459 sPAPRPHBState *sphb;
460 uint32_t addr, option;
461 uint64_t buid;
462 int ret;
463
464 if ((nargs != 4) || (nret != 1)) {
465 goto param_error_exit;
466 }
467
468 buid = rtas_ldq(args, 1);
469 addr = rtas_ld(args, 0);
470 option = rtas_ld(args, 3);
471
472 sphb = spapr_pci_find_phb(spapr, buid);
473 if (!sphb) {
474 goto param_error_exit;
475 }
476
477 if (!spapr_phb_eeh_available(sphb)) {
478 goto param_error_exit;
479 }
480
481 ret = spapr_phb_vfio_eeh_set_option(sphb, addr, option);
482 rtas_st(rets, 0, ret);
483 return;
484
485 param_error_exit:
486 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
487 }
488
489 static void rtas_ibm_get_config_addr_info2(PowerPCCPU *cpu,
490 sPAPRMachineState *spapr,
491 uint32_t token, uint32_t nargs,
492 target_ulong args, uint32_t nret,
493 target_ulong rets)
494 {
495 sPAPRPHBState *sphb;
496 PCIDevice *pdev;
497 uint32_t addr, option;
498 uint64_t buid;
499
500 if ((nargs != 4) || (nret != 2)) {
501 goto param_error_exit;
502 }
503
504 buid = rtas_ldq(args, 1);
505 sphb = spapr_pci_find_phb(spapr, buid);
506 if (!sphb) {
507 goto param_error_exit;
508 }
509
510 if (!spapr_phb_eeh_available(sphb)) {
511 goto param_error_exit;
512 }
513
514 /*
515 * We always have PE address of form "00BB0001". "BB"
516 * represents the bus number of PE's primary bus.
517 */
518 option = rtas_ld(args, 3);
519 switch (option) {
520 case RTAS_GET_PE_ADDR:
521 addr = rtas_ld(args, 0);
522 pdev = spapr_pci_find_dev(spapr, buid, addr);
523 if (!pdev) {
524 goto param_error_exit;
525 }
526
527 rtas_st(rets, 1, (pci_bus_num(pci_get_bus(pdev)) << 16) + 1);
528 break;
529 case RTAS_GET_PE_MODE:
530 rtas_st(rets, 1, RTAS_PE_MODE_SHARED);
531 break;
532 default:
533 goto param_error_exit;
534 }
535
536 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
537 return;
538
539 param_error_exit:
540 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
541 }
542
543 static void rtas_ibm_read_slot_reset_state2(PowerPCCPU *cpu,
544 sPAPRMachineState *spapr,
545 uint32_t token, uint32_t nargs,
546 target_ulong args, uint32_t nret,
547 target_ulong rets)
548 {
549 sPAPRPHBState *sphb;
550 uint64_t buid;
551 int state, ret;
552
553 if ((nargs != 3) || (nret != 4 && nret != 5)) {
554 goto param_error_exit;
555 }
556
557 buid = rtas_ldq(args, 1);
558 sphb = spapr_pci_find_phb(spapr, buid);
559 if (!sphb) {
560 goto param_error_exit;
561 }
562
563 if (!spapr_phb_eeh_available(sphb)) {
564 goto param_error_exit;
565 }
566
567 ret = spapr_phb_vfio_eeh_get_state(sphb, &state);
568 rtas_st(rets, 0, ret);
569 if (ret != RTAS_OUT_SUCCESS) {
570 return;
571 }
572
573 rtas_st(rets, 1, state);
574 rtas_st(rets, 2, RTAS_EEH_SUPPORT);
575 rtas_st(rets, 3, RTAS_EEH_PE_UNAVAIL_INFO);
576 if (nret >= 5) {
577 rtas_st(rets, 4, RTAS_EEH_PE_RECOVER_INFO);
578 }
579 return;
580
581 param_error_exit:
582 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
583 }
584
585 static void rtas_ibm_set_slot_reset(PowerPCCPU *cpu,
586 sPAPRMachineState *spapr,
587 uint32_t token, uint32_t nargs,
588 target_ulong args, uint32_t nret,
589 target_ulong rets)
590 {
591 sPAPRPHBState *sphb;
592 uint32_t option;
593 uint64_t buid;
594 int ret;
595
596 if ((nargs != 4) || (nret != 1)) {
597 goto param_error_exit;
598 }
599
600 buid = rtas_ldq(args, 1);
601 option = rtas_ld(args, 3);
602 sphb = spapr_pci_find_phb(spapr, buid);
603 if (!sphb) {
604 goto param_error_exit;
605 }
606
607 if (!spapr_phb_eeh_available(sphb)) {
608 goto param_error_exit;
609 }
610
611 ret = spapr_phb_vfio_eeh_reset(sphb, option);
612 rtas_st(rets, 0, ret);
613 return;
614
615 param_error_exit:
616 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
617 }
618
619 static void rtas_ibm_configure_pe(PowerPCCPU *cpu,
620 sPAPRMachineState *spapr,
621 uint32_t token, uint32_t nargs,
622 target_ulong args, uint32_t nret,
623 target_ulong rets)
624 {
625 sPAPRPHBState *sphb;
626 uint64_t buid;
627 int ret;
628
629 if ((nargs != 3) || (nret != 1)) {
630 goto param_error_exit;
631 }
632
633 buid = rtas_ldq(args, 1);
634 sphb = spapr_pci_find_phb(spapr, buid);
635 if (!sphb) {
636 goto param_error_exit;
637 }
638
639 if (!spapr_phb_eeh_available(sphb)) {
640 goto param_error_exit;
641 }
642
643 ret = spapr_phb_vfio_eeh_configure(sphb);
644 rtas_st(rets, 0, ret);
645 return;
646
647 param_error_exit:
648 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
649 }
650
651 /* To support it later */
652 static void rtas_ibm_slot_error_detail(PowerPCCPU *cpu,
653 sPAPRMachineState *spapr,
654 uint32_t token, uint32_t nargs,
655 target_ulong args, uint32_t nret,
656 target_ulong rets)
657 {
658 sPAPRPHBState *sphb;
659 int option;
660 uint64_t buid;
661
662 if ((nargs != 8) || (nret != 1)) {
663 goto param_error_exit;
664 }
665
666 buid = rtas_ldq(args, 1);
667 sphb = spapr_pci_find_phb(spapr, buid);
668 if (!sphb) {
669 goto param_error_exit;
670 }
671
672 if (!spapr_phb_eeh_available(sphb)) {
673 goto param_error_exit;
674 }
675
676 option = rtas_ld(args, 7);
677 switch (option) {
678 case RTAS_SLOT_TEMP_ERR_LOG:
679 case RTAS_SLOT_PERM_ERR_LOG:
680 break;
681 default:
682 goto param_error_exit;
683 }
684
685 /* We don't have error log yet */
686 rtas_st(rets, 0, RTAS_OUT_NO_ERRORS_FOUND);
687 return;
688
689 param_error_exit:
690 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
691 }
692
693 static int pci_spapr_swizzle(int slot, int pin)
694 {
695 return (slot + pin) % PCI_NUM_PINS;
696 }
697
698 static int pci_spapr_map_irq(PCIDevice *pci_dev, int irq_num)
699 {
700 /*
701 * Here we need to convert pci_dev + irq_num to some unique value
702 * which is less than number of IRQs on the specific bus (4). We
703 * use standard PCI swizzling, that is (slot number + pin number)
704 * % 4.
705 */
706 return pci_spapr_swizzle(PCI_SLOT(pci_dev->devfn), irq_num);
707 }
708
709 static void pci_spapr_set_irq(void *opaque, int irq_num, int level)
710 {
711 /*
712 * Here we use the number returned by pci_spapr_map_irq to find a
713 * corresponding qemu_irq.
714 */
715 sPAPRPHBState *phb = opaque;
716
717 trace_spapr_pci_lsi_set(phb->dtbusname, irq_num, phb->lsi_table[irq_num].irq);
718 qemu_set_irq(spapr_phb_lsi_qirq(phb, irq_num), level);
719 }
720
721 static PCIINTxRoute spapr_route_intx_pin_to_irq(void *opaque, int pin)
722 {
723 sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(opaque);
724 PCIINTxRoute route;
725
726 route.mode = PCI_INTX_ENABLED;
727 route.irq = sphb->lsi_table[pin].irq;
728
729 return route;
730 }
731
732 /*
733 * MSI/MSIX memory region implementation.
734 * The handler handles both MSI and MSIX.
735 * The vector number is encoded in least bits in data.
736 */
737 static void spapr_msi_write(void *opaque, hwaddr addr,
738 uint64_t data, unsigned size)
739 {
740 sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
741 uint32_t irq = data;
742
743 trace_spapr_pci_msi_write(addr, data, irq);
744
745 qemu_irq_pulse(spapr_qirq(spapr, irq));
746 }
747
748 static const MemoryRegionOps spapr_msi_ops = {
749 /* There is no .read as the read result is undefined by PCI spec */
750 .read = NULL,
751 .write = spapr_msi_write,
752 .endianness = DEVICE_LITTLE_ENDIAN
753 };
754
755 /*
756 * PHB PCI device
757 */
758 static AddressSpace *spapr_pci_dma_iommu(PCIBus *bus, void *opaque, int devfn)
759 {
760 sPAPRPHBState *phb = opaque;
761
762 return &phb->iommu_as;
763 }
764
765 static char *spapr_phb_vfio_get_loc_code(sPAPRPHBState *sphb, PCIDevice *pdev)
766 {
767 char *path = NULL, *buf = NULL, *host = NULL;
768
769 /* Get the PCI VFIO host id */
770 host = object_property_get_str(OBJECT(pdev), "host", NULL);
771 if (!host) {
772 goto err_out;
773 }
774
775 /* Construct the path of the file that will give us the DT location */
776 path = g_strdup_printf("/sys/bus/pci/devices/%s/devspec", host);
777 g_free(host);
778 if (!g_file_get_contents(path, &buf, NULL, NULL)) {
779 goto err_out;
780 }
781 g_free(path);
782
783 /* Construct and read from host device tree the loc-code */
784 path = g_strdup_printf("/proc/device-tree%s/ibm,loc-code", buf);
785 g_free(buf);
786 if (!g_file_get_contents(path, &buf, NULL, NULL)) {
787 goto err_out;
788 }
789 return buf;
790
791 err_out:
792 g_free(path);
793 return NULL;
794 }
795
796 static char *spapr_phb_get_loc_code(sPAPRPHBState *sphb, PCIDevice *pdev)
797 {
798 char *buf;
799 const char *devtype = "qemu";
800 uint32_t busnr = pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(pdev))));
801
802 if (object_dynamic_cast(OBJECT(pdev), "vfio-pci")) {
803 buf = spapr_phb_vfio_get_loc_code(sphb, pdev);
804 if (buf) {
805 return buf;
806 }
807 devtype = "vfio";
808 }
809 /*
810 * For emulated devices and VFIO-failure case, make up
811 * the loc-code.
812 */
813 buf = g_strdup_printf("%s_%s:%04x:%02x:%02x.%x",
814 devtype, pdev->name, sphb->index, busnr,
815 PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
816 return buf;
817 }
818
819 /* Macros to operate with address in OF binding to PCI */
820 #define b_x(x, p, l) (((x) & ((1<<(l))-1)) << (p))
821 #define b_n(x) b_x((x), 31, 1) /* 0 if relocatable */
822 #define b_p(x) b_x((x), 30, 1) /* 1 if prefetchable */
823 #define b_t(x) b_x((x), 29, 1) /* 1 if the address is aliased */
824 #define b_ss(x) b_x((x), 24, 2) /* the space code */
825 #define b_bbbbbbbb(x) b_x((x), 16, 8) /* bus number */
826 #define b_ddddd(x) b_x((x), 11, 5) /* device number */
827 #define b_fff(x) b_x((x), 8, 3) /* function number */
828 #define b_rrrrrrrr(x) b_x((x), 0, 8) /* register number */
829
830 /* for 'reg'/'assigned-addresses' OF properties */
831 #define RESOURCE_CELLS_SIZE 2
832 #define RESOURCE_CELLS_ADDRESS 3
833
834 typedef struct ResourceFields {
835 uint32_t phys_hi;
836 uint32_t phys_mid;
837 uint32_t phys_lo;
838 uint32_t size_hi;
839 uint32_t size_lo;
840 } QEMU_PACKED ResourceFields;
841
842 typedef struct ResourceProps {
843 ResourceFields reg[8];
844 ResourceFields assigned[7];
845 uint32_t reg_len;
846 uint32_t assigned_len;
847 } ResourceProps;
848
849 /* fill in the 'reg'/'assigned-resources' OF properties for
850 * a PCI device. 'reg' describes resource requirements for a
851 * device's IO/MEM regions, 'assigned-addresses' describes the
852 * actual resource assignments.
853 *
854 * the properties are arrays of ('phys-addr', 'size') pairs describing
855 * the addressable regions of the PCI device, where 'phys-addr' is a
856 * RESOURCE_CELLS_ADDRESS-tuple of 32-bit integers corresponding to
857 * (phys.hi, phys.mid, phys.lo), and 'size' is a
858 * RESOURCE_CELLS_SIZE-tuple corresponding to (size.hi, size.lo).
859 *
860 * phys.hi = 0xYYXXXXZZ, where:
861 * 0xYY = npt000ss
862 * ||| |
863 * ||| +-- space code
864 * ||| |
865 * ||| + 00 if configuration space
866 * ||| + 01 if IO region,
867 * ||| + 10 if 32-bit MEM region
868 * ||| + 11 if 64-bit MEM region
869 * |||
870 * ||+------ for non-relocatable IO: 1 if aliased
871 * || for relocatable IO: 1 if below 64KB
872 * || for MEM: 1 if below 1MB
873 * |+------- 1 if region is prefetchable
874 * +-------- 1 if region is non-relocatable
875 * 0xXXXX = bbbbbbbb dddddfff, encoding bus, slot, and function
876 * bits respectively
877 * 0xZZ = rrrrrrrr, the register number of the BAR corresponding
878 * to the region
879 *
880 * phys.mid and phys.lo correspond respectively to the hi/lo portions
881 * of the actual address of the region.
882 *
883 * how the phys-addr/size values are used differ slightly between
884 * 'reg' and 'assigned-addresses' properties. namely, 'reg' has
885 * an additional description for the config space region of the
886 * device, and in the case of QEMU has n=0 and phys.mid=phys.lo=0
887 * to describe the region as relocatable, with an address-mapping
888 * that corresponds directly to the PHB's address space for the
889 * resource. 'assigned-addresses' always has n=1 set with an absolute
890 * address assigned for the resource. in general, 'assigned-addresses'
891 * won't be populated, since addresses for PCI devices are generally
892 * unmapped initially and left to the guest to assign.
893 *
894 * note also that addresses defined in these properties are, at least
895 * for PAPR guests, relative to the PHBs IO/MEM windows, and
896 * correspond directly to the addresses in the BARs.
897 *
898 * in accordance with PCI Bus Binding to Open Firmware,
899 * IEEE Std 1275-1994, section 4.1.1, as implemented by PAPR+ v2.7,
900 * Appendix C.
901 */
902 static void populate_resource_props(PCIDevice *d, ResourceProps *rp)
903 {
904 int bus_num = pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(d))));
905 uint32_t dev_id = (b_bbbbbbbb(bus_num) |
906 b_ddddd(PCI_SLOT(d->devfn)) |
907 b_fff(PCI_FUNC(d->devfn)));
908 ResourceFields *reg, *assigned;
909 int i, reg_idx = 0, assigned_idx = 0;
910
911 /* config space region */
912 reg = &rp->reg[reg_idx++];
913 reg->phys_hi = cpu_to_be32(dev_id);
914 reg->phys_mid = 0;
915 reg->phys_lo = 0;
916 reg->size_hi = 0;
917 reg->size_lo = 0;
918
919 for (i = 0; i < PCI_NUM_REGIONS; i++) {
920 if (!d->io_regions[i].size) {
921 continue;
922 }
923
924 reg = &rp->reg[reg_idx++];
925
926 reg->phys_hi = cpu_to_be32(dev_id | b_rrrrrrrr(pci_bar(d, i)));
927 if (d->io_regions[i].type & PCI_BASE_ADDRESS_SPACE_IO) {
928 reg->phys_hi |= cpu_to_be32(b_ss(1));
929 } else if (d->io_regions[i].type & PCI_BASE_ADDRESS_MEM_TYPE_64) {
930 reg->phys_hi |= cpu_to_be32(b_ss(3));
931 } else {
932 reg->phys_hi |= cpu_to_be32(b_ss(2));
933 }
934 reg->phys_mid = 0;
935 reg->phys_lo = 0;
936 reg->size_hi = cpu_to_be32(d->io_regions[i].size >> 32);
937 reg->size_lo = cpu_to_be32(d->io_regions[i].size);
938
939 if (d->io_regions[i].addr == PCI_BAR_UNMAPPED) {
940 continue;
941 }
942
943 assigned = &rp->assigned[assigned_idx++];
944 assigned->phys_hi = cpu_to_be32(reg->phys_hi | b_n(1));
945 assigned->phys_mid = cpu_to_be32(d->io_regions[i].addr >> 32);
946 assigned->phys_lo = cpu_to_be32(d->io_regions[i].addr);
947 assigned->size_hi = reg->size_hi;
948 assigned->size_lo = reg->size_lo;
949 }
950
951 rp->reg_len = reg_idx * sizeof(ResourceFields);
952 rp->assigned_len = assigned_idx * sizeof(ResourceFields);
953 }
954
955 typedef struct PCIClass PCIClass;
956 typedef struct PCISubClass PCISubClass;
957 typedef struct PCIIFace PCIIFace;
958
959 struct PCIIFace {
960 int iface;
961 const char *name;
962 };
963
964 struct PCISubClass {
965 int subclass;
966 const char *name;
967 const PCIIFace *iface;
968 };
969
970 struct PCIClass {
971 const char *name;
972 const PCISubClass *subc;
973 };
974
975 static const PCISubClass undef_subclass[] = {
976 { PCI_CLASS_NOT_DEFINED_VGA, "display", NULL },
977 { 0xFF, NULL, NULL },
978 };
979
980 static const PCISubClass mass_subclass[] = {
981 { PCI_CLASS_STORAGE_SCSI, "scsi", NULL },
982 { PCI_CLASS_STORAGE_IDE, "ide", NULL },
983 { PCI_CLASS_STORAGE_FLOPPY, "fdc", NULL },
984 { PCI_CLASS_STORAGE_IPI, "ipi", NULL },
985 { PCI_CLASS_STORAGE_RAID, "raid", NULL },
986 { PCI_CLASS_STORAGE_ATA, "ata", NULL },
987 { PCI_CLASS_STORAGE_SATA, "sata", NULL },
988 { PCI_CLASS_STORAGE_SAS, "sas", NULL },
989 { 0xFF, NULL, NULL },
990 };
991
992 static const PCISubClass net_subclass[] = {
993 { PCI_CLASS_NETWORK_ETHERNET, "ethernet", NULL },
994 { PCI_CLASS_NETWORK_TOKEN_RING, "token-ring", NULL },
995 { PCI_CLASS_NETWORK_FDDI, "fddi", NULL },
996 { PCI_CLASS_NETWORK_ATM, "atm", NULL },
997 { PCI_CLASS_NETWORK_ISDN, "isdn", NULL },
998 { PCI_CLASS_NETWORK_WORLDFIP, "worldfip", NULL },
999 { PCI_CLASS_NETWORK_PICMG214, "picmg", NULL },
1000 { 0xFF, NULL, NULL },
1001 };
1002
1003 static const PCISubClass displ_subclass[] = {
1004 { PCI_CLASS_DISPLAY_VGA, "vga", NULL },
1005 { PCI_CLASS_DISPLAY_XGA, "xga", NULL },
1006 { PCI_CLASS_DISPLAY_3D, "3d-controller", NULL },
1007 { 0xFF, NULL, NULL },
1008 };
1009
1010 static const PCISubClass media_subclass[] = {
1011 { PCI_CLASS_MULTIMEDIA_VIDEO, "video", NULL },
1012 { PCI_CLASS_MULTIMEDIA_AUDIO, "sound", NULL },
1013 { PCI_CLASS_MULTIMEDIA_PHONE, "telephony", NULL },
1014 { 0xFF, NULL, NULL },
1015 };
1016
1017 static const PCISubClass mem_subclass[] = {
1018 { PCI_CLASS_MEMORY_RAM, "memory", NULL },
1019 { PCI_CLASS_MEMORY_FLASH, "flash", NULL },
1020 { 0xFF, NULL, NULL },
1021 };
1022
1023 static const PCISubClass bridg_subclass[] = {
1024 { PCI_CLASS_BRIDGE_HOST, "host", NULL },
1025 { PCI_CLASS_BRIDGE_ISA, "isa", NULL },
1026 { PCI_CLASS_BRIDGE_EISA, "eisa", NULL },
1027 { PCI_CLASS_BRIDGE_MC, "mca", NULL },
1028 { PCI_CLASS_BRIDGE_PCI, "pci", NULL },
1029 { PCI_CLASS_BRIDGE_PCMCIA, "pcmcia", NULL },
1030 { PCI_CLASS_BRIDGE_NUBUS, "nubus", NULL },
1031 { PCI_CLASS_BRIDGE_CARDBUS, "cardbus", NULL },
1032 { PCI_CLASS_BRIDGE_RACEWAY, "raceway", NULL },
1033 { PCI_CLASS_BRIDGE_PCI_SEMITP, "semi-transparent-pci", NULL },
1034 { PCI_CLASS_BRIDGE_IB_PCI, "infiniband", NULL },
1035 { 0xFF, NULL, NULL },
1036 };
1037
1038 static const PCISubClass comm_subclass[] = {
1039 { PCI_CLASS_COMMUNICATION_SERIAL, "serial", NULL },
1040 { PCI_CLASS_COMMUNICATION_PARALLEL, "parallel", NULL },
1041 { PCI_CLASS_COMMUNICATION_MULTISERIAL, "multiport-serial", NULL },
1042 { PCI_CLASS_COMMUNICATION_MODEM, "modem", NULL },
1043 { PCI_CLASS_COMMUNICATION_GPIB, "gpib", NULL },
1044 { PCI_CLASS_COMMUNICATION_SC, "smart-card", NULL },
1045 { 0xFF, NULL, NULL, },
1046 };
1047
1048 static const PCIIFace pic_iface[] = {
1049 { PCI_CLASS_SYSTEM_PIC_IOAPIC, "io-apic" },
1050 { PCI_CLASS_SYSTEM_PIC_IOXAPIC, "io-xapic" },
1051 { 0xFF, NULL },
1052 };
1053
1054 static const PCISubClass sys_subclass[] = {
1055 { PCI_CLASS_SYSTEM_PIC, "interrupt-controller", pic_iface },
1056 { PCI_CLASS_SYSTEM_DMA, "dma-controller", NULL },
1057 { PCI_CLASS_SYSTEM_TIMER, "timer", NULL },
1058 { PCI_CLASS_SYSTEM_RTC, "rtc", NULL },
1059 { PCI_CLASS_SYSTEM_PCI_HOTPLUG, "hot-plug-controller", NULL },
1060 { PCI_CLASS_SYSTEM_SDHCI, "sd-host-controller", NULL },
1061 { 0xFF, NULL, NULL },
1062 };
1063
1064 static const PCISubClass inp_subclass[] = {
1065 { PCI_CLASS_INPUT_KEYBOARD, "keyboard", NULL },
1066 { PCI_CLASS_INPUT_PEN, "pen", NULL },
1067 { PCI_CLASS_INPUT_MOUSE, "mouse", NULL },
1068 { PCI_CLASS_INPUT_SCANNER, "scanner", NULL },
1069 { PCI_CLASS_INPUT_GAMEPORT, "gameport", NULL },
1070 { 0xFF, NULL, NULL },
1071 };
1072
1073 static const PCISubClass dock_subclass[] = {
1074 { PCI_CLASS_DOCKING_GENERIC, "dock", NULL },
1075 { 0xFF, NULL, NULL },
1076 };
1077
1078 static const PCISubClass cpu_subclass[] = {
1079 { PCI_CLASS_PROCESSOR_PENTIUM, "pentium", NULL },
1080 { PCI_CLASS_PROCESSOR_POWERPC, "powerpc", NULL },
1081 { PCI_CLASS_PROCESSOR_MIPS, "mips", NULL },
1082 { PCI_CLASS_PROCESSOR_CO, "co-processor", NULL },
1083 { 0xFF, NULL, NULL },
1084 };
1085
1086 static const PCIIFace usb_iface[] = {
1087 { PCI_CLASS_SERIAL_USB_UHCI, "usb-uhci" },
1088 { PCI_CLASS_SERIAL_USB_OHCI, "usb-ohci", },
1089 { PCI_CLASS_SERIAL_USB_EHCI, "usb-ehci" },
1090 { PCI_CLASS_SERIAL_USB_XHCI, "usb-xhci" },
1091 { PCI_CLASS_SERIAL_USB_UNKNOWN, "usb-unknown" },
1092 { PCI_CLASS_SERIAL_USB_DEVICE, "usb-device" },
1093 { 0xFF, NULL },
1094 };
1095
1096 static const PCISubClass ser_subclass[] = {
1097 { PCI_CLASS_SERIAL_FIREWIRE, "firewire", NULL },
1098 { PCI_CLASS_SERIAL_ACCESS, "access-bus", NULL },
1099 { PCI_CLASS_SERIAL_SSA, "ssa", NULL },
1100 { PCI_CLASS_SERIAL_USB, "usb", usb_iface },
1101 { PCI_CLASS_SERIAL_FIBER, "fibre-channel", NULL },
1102 { PCI_CLASS_SERIAL_SMBUS, "smb", NULL },
1103 { PCI_CLASS_SERIAL_IB, "infiniband", NULL },
1104 { PCI_CLASS_SERIAL_IPMI, "ipmi", NULL },
1105 { PCI_CLASS_SERIAL_SERCOS, "sercos", NULL },
1106 { PCI_CLASS_SERIAL_CANBUS, "canbus", NULL },
1107 { 0xFF, NULL, NULL },
1108 };
1109
1110 static const PCISubClass wrl_subclass[] = {
1111 { PCI_CLASS_WIRELESS_IRDA, "irda", NULL },
1112 { PCI_CLASS_WIRELESS_CIR, "consumer-ir", NULL },
1113 { PCI_CLASS_WIRELESS_RF_CONTROLLER, "rf-controller", NULL },
1114 { PCI_CLASS_WIRELESS_BLUETOOTH, "bluetooth", NULL },
1115 { PCI_CLASS_WIRELESS_BROADBAND, "broadband", NULL },
1116 { 0xFF, NULL, NULL },
1117 };
1118
1119 static const PCISubClass sat_subclass[] = {
1120 { PCI_CLASS_SATELLITE_TV, "satellite-tv", NULL },
1121 { PCI_CLASS_SATELLITE_AUDIO, "satellite-audio", NULL },
1122 { PCI_CLASS_SATELLITE_VOICE, "satellite-voice", NULL },
1123 { PCI_CLASS_SATELLITE_DATA, "satellite-data", NULL },
1124 { 0xFF, NULL, NULL },
1125 };
1126
1127 static const PCISubClass crypt_subclass[] = {
1128 { PCI_CLASS_CRYPT_NETWORK, "network-encryption", NULL },
1129 { PCI_CLASS_CRYPT_ENTERTAINMENT,
1130 "entertainment-encryption", NULL },
1131 { 0xFF, NULL, NULL },
1132 };
1133
1134 static const PCISubClass spc_subclass[] = {
1135 { PCI_CLASS_SP_DPIO, "dpio", NULL },
1136 { PCI_CLASS_SP_PERF, "counter", NULL },
1137 { PCI_CLASS_SP_SYNCH, "measurement", NULL },
1138 { PCI_CLASS_SP_MANAGEMENT, "management-card", NULL },
1139 { 0xFF, NULL, NULL },
1140 };
1141
1142 static const PCIClass pci_classes[] = {
1143 { "legacy-device", undef_subclass },
1144 { "mass-storage", mass_subclass },
1145 { "network", net_subclass },
1146 { "display", displ_subclass, },
1147 { "multimedia-device", media_subclass },
1148 { "memory-controller", mem_subclass },
1149 { "unknown-bridge", bridg_subclass },
1150 { "communication-controller", comm_subclass},
1151 { "system-peripheral", sys_subclass },
1152 { "input-controller", inp_subclass },
1153 { "docking-station", dock_subclass },
1154 { "cpu", cpu_subclass },
1155 { "serial-bus", ser_subclass },
1156 { "wireless-controller", wrl_subclass },
1157 { "intelligent-io", NULL },
1158 { "satellite-device", sat_subclass },
1159 { "encryption", crypt_subclass },
1160 { "data-processing-controller", spc_subclass },
1161 };
1162
1163 static const char *pci_find_device_name(uint8_t class, uint8_t subclass,
1164 uint8_t iface)
1165 {
1166 const PCIClass *pclass;
1167 const PCISubClass *psubclass;
1168 const PCIIFace *piface;
1169 const char *name;
1170
1171 if (class >= ARRAY_SIZE(pci_classes)) {
1172 return "pci";
1173 }
1174
1175 pclass = pci_classes + class;
1176 name = pclass->name;
1177
1178 if (pclass->subc == NULL) {
1179 return name;
1180 }
1181
1182 psubclass = pclass->subc;
1183 while ((psubclass->subclass & 0xff) != 0xff) {
1184 if ((psubclass->subclass & 0xff) == subclass) {
1185 name = psubclass->name;
1186 break;
1187 }
1188 psubclass++;
1189 }
1190
1191 piface = psubclass->iface;
1192 if (piface == NULL) {
1193 return name;
1194 }
1195 while ((piface->iface & 0xff) != 0xff) {
1196 if ((piface->iface & 0xff) == iface) {
1197 name = piface->name;
1198 break;
1199 }
1200 piface++;
1201 }
1202
1203 return name;
1204 }
1205
1206 static gchar *pci_get_node_name(PCIDevice *dev)
1207 {
1208 int slot = PCI_SLOT(dev->devfn);
1209 int func = PCI_FUNC(dev->devfn);
1210 uint32_t ccode = pci_default_read_config(dev, PCI_CLASS_PROG, 3);
1211 const char *name;
1212
1213 name = pci_find_device_name((ccode >> 16) & 0xff, (ccode >> 8) & 0xff,
1214 ccode & 0xff);
1215
1216 if (func != 0) {
1217 return g_strdup_printf("%s@%x,%x", name, slot, func);
1218 } else {
1219 return g_strdup_printf("%s@%x", name, slot);
1220 }
1221 }
1222
1223 static uint32_t spapr_phb_get_pci_drc_index(sPAPRPHBState *phb,
1224 PCIDevice *pdev);
1225
1226 static void spapr_populate_pci_child_dt(PCIDevice *dev, void *fdt, int offset,
1227 sPAPRPHBState *sphb)
1228 {
1229 ResourceProps rp;
1230 bool is_bridge = false;
1231 int pci_status;
1232 char *buf = NULL;
1233 uint32_t drc_index = spapr_phb_get_pci_drc_index(sphb, dev);
1234 uint32_t ccode = pci_default_read_config(dev, PCI_CLASS_PROG, 3);
1235 uint32_t max_msi, max_msix;
1236
1237 if (pci_default_read_config(dev, PCI_HEADER_TYPE, 1) ==
1238 PCI_HEADER_TYPE_BRIDGE) {
1239 is_bridge = true;
1240 }
1241
1242 /* in accordance with PAPR+ v2.7 13.6.3, Table 181 */
1243 _FDT(fdt_setprop_cell(fdt, offset, "vendor-id",
1244 pci_default_read_config(dev, PCI_VENDOR_ID, 2)));
1245 _FDT(fdt_setprop_cell(fdt, offset, "device-id",
1246 pci_default_read_config(dev, PCI_DEVICE_ID, 2)));
1247 _FDT(fdt_setprop_cell(fdt, offset, "revision-id",
1248 pci_default_read_config(dev, PCI_REVISION_ID, 1)));
1249 _FDT(fdt_setprop_cell(fdt, offset, "class-code", ccode));
1250 if (pci_default_read_config(dev, PCI_INTERRUPT_PIN, 1)) {
1251 _FDT(fdt_setprop_cell(fdt, offset, "interrupts",
1252 pci_default_read_config(dev, PCI_INTERRUPT_PIN, 1)));
1253 }
1254
1255 if (!is_bridge) {
1256 _FDT(fdt_setprop_cell(fdt, offset, "min-grant",
1257 pci_default_read_config(dev, PCI_MIN_GNT, 1)));
1258 _FDT(fdt_setprop_cell(fdt, offset, "max-latency",
1259 pci_default_read_config(dev, PCI_MAX_LAT, 1)));
1260 }
1261
1262 if (pci_default_read_config(dev, PCI_SUBSYSTEM_ID, 2)) {
1263 _FDT(fdt_setprop_cell(fdt, offset, "subsystem-id",
1264 pci_default_read_config(dev, PCI_SUBSYSTEM_ID, 2)));
1265 }
1266
1267 if (pci_default_read_config(dev, PCI_SUBSYSTEM_VENDOR_ID, 2)) {
1268 _FDT(fdt_setprop_cell(fdt, offset, "subsystem-vendor-id",
1269 pci_default_read_config(dev, PCI_SUBSYSTEM_VENDOR_ID, 2)));
1270 }
1271
1272 _FDT(fdt_setprop_cell(fdt, offset, "cache-line-size",
1273 pci_default_read_config(dev, PCI_CACHE_LINE_SIZE, 1)));
1274
1275 /* the following fdt cells are masked off the pci status register */
1276 pci_status = pci_default_read_config(dev, PCI_STATUS, 2);
1277 _FDT(fdt_setprop_cell(fdt, offset, "devsel-speed",
1278 PCI_STATUS_DEVSEL_MASK & pci_status));
1279
1280 if (pci_status & PCI_STATUS_FAST_BACK) {
1281 _FDT(fdt_setprop(fdt, offset, "fast-back-to-back", NULL, 0));
1282 }
1283 if (pci_status & PCI_STATUS_66MHZ) {
1284 _FDT(fdt_setprop(fdt, offset, "66mhz-capable", NULL, 0));
1285 }
1286 if (pci_status & PCI_STATUS_UDF) {
1287 _FDT(fdt_setprop(fdt, offset, "udf-supported", NULL, 0));
1288 }
1289
1290 _FDT(fdt_setprop_string(fdt, offset, "name",
1291 pci_find_device_name((ccode >> 16) & 0xff,
1292 (ccode >> 8) & 0xff,
1293 ccode & 0xff)));
1294
1295 buf = spapr_phb_get_loc_code(sphb, dev);
1296 _FDT(fdt_setprop_string(fdt, offset, "ibm,loc-code", buf));
1297 g_free(buf);
1298
1299 if (drc_index) {
1300 _FDT(fdt_setprop_cell(fdt, offset, "ibm,my-drc-index", drc_index));
1301 }
1302
1303 _FDT(fdt_setprop_cell(fdt, offset, "#address-cells",
1304 RESOURCE_CELLS_ADDRESS));
1305 _FDT(fdt_setprop_cell(fdt, offset, "#size-cells",
1306 RESOURCE_CELLS_SIZE));
1307
1308 if (msi_present(dev)) {
1309 max_msi = msi_nr_vectors_allocated(dev);
1310 if (max_msi) {
1311 _FDT(fdt_setprop_cell(fdt, offset, "ibm,req#msi", max_msi));
1312 }
1313 }
1314 if (msix_present(dev)) {
1315 max_msix = dev->msix_entries_nr;
1316 if (max_msix) {
1317 _FDT(fdt_setprop_cell(fdt, offset, "ibm,req#msi-x", max_msix));
1318 }
1319 }
1320
1321 populate_resource_props(dev, &rp);
1322 _FDT(fdt_setprop(fdt, offset, "reg", (uint8_t *)rp.reg, rp.reg_len));
1323 _FDT(fdt_setprop(fdt, offset, "assigned-addresses",
1324 (uint8_t *)rp.assigned, rp.assigned_len));
1325
1326 if (sphb->pcie_ecs && pci_is_express(dev)) {
1327 _FDT(fdt_setprop_cell(fdt, offset, "ibm,pci-config-space-type", 0x1));
1328 }
1329 }
1330
1331 /* create OF node for pci device and required OF DT properties */
1332 static int spapr_create_pci_child_dt(sPAPRPHBState *phb, PCIDevice *dev,
1333 void *fdt, int node_offset)
1334 {
1335 int offset;
1336 gchar *nodename;
1337
1338 nodename = pci_get_node_name(dev);
1339 _FDT(offset = fdt_add_subnode(fdt, node_offset, nodename));
1340 g_free(nodename);
1341
1342 spapr_populate_pci_child_dt(dev, fdt, offset, phb);
1343
1344 return offset;
1345 }
1346
1347 /* Callback to be called during DRC release. */
1348 void spapr_phb_remove_pci_device_cb(DeviceState *dev)
1349 {
1350 /* some version guests do not wait for completion of a device
1351 * cleanup (generally done asynchronously by the kernel) before
1352 * signaling to QEMU that the device is safe, but instead sleep
1353 * for some 'safe' period of time. unfortunately on a busy host
1354 * this sleep isn't guaranteed to be long enough, resulting in
1355 * bad things like IRQ lines being left asserted during final
1356 * device removal. to deal with this we call reset just prior
1357 * to finalizing the device, which will put the device back into
1358 * an 'idle' state, as the device cleanup code expects.
1359 */
1360 pci_device_reset(PCI_DEVICE(dev));
1361 object_unparent(OBJECT(dev));
1362 }
1363
1364 static sPAPRDRConnector *spapr_phb_get_pci_func_drc(sPAPRPHBState *phb,
1365 uint32_t busnr,
1366 int32_t devfn)
1367 {
1368 return spapr_drc_by_id(TYPE_SPAPR_DRC_PCI,
1369 (phb->index << 16) | (busnr << 8) | devfn);
1370 }
1371
1372 static sPAPRDRConnector *spapr_phb_get_pci_drc(sPAPRPHBState *phb,
1373 PCIDevice *pdev)
1374 {
1375 uint32_t busnr = pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(pdev))));
1376 return spapr_phb_get_pci_func_drc(phb, busnr, pdev->devfn);
1377 }
1378
1379 static uint32_t spapr_phb_get_pci_drc_index(sPAPRPHBState *phb,
1380 PCIDevice *pdev)
1381 {
1382 sPAPRDRConnector *drc = spapr_phb_get_pci_drc(phb, pdev);
1383
1384 if (!drc) {
1385 return 0;
1386 }
1387
1388 return spapr_drc_index(drc);
1389 }
1390
1391 static void spapr_pci_plug(HotplugHandler *plug_handler,
1392 DeviceState *plugged_dev, Error **errp)
1393 {
1394 sPAPRPHBState *phb = SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler));
1395 PCIDevice *pdev = PCI_DEVICE(plugged_dev);
1396 sPAPRDRConnector *drc = spapr_phb_get_pci_drc(phb, pdev);
1397 Error *local_err = NULL;
1398 PCIBus *bus = PCI_BUS(qdev_get_parent_bus(DEVICE(pdev)));
1399 uint32_t slotnr = PCI_SLOT(pdev->devfn);
1400 void *fdt = NULL;
1401 int fdt_start_offset, fdt_size;
1402
1403 /* if DR is disabled we don't need to do anything in the case of
1404 * hotplug or coldplug callbacks
1405 */
1406 if (!phb->dr_enabled) {
1407 /* if this is a hotplug operation initiated by the user
1408 * we need to let them know it's not enabled
1409 */
1410 if (plugged_dev->hotplugged) {
1411 error_setg(&local_err, QERR_BUS_NO_HOTPLUG,
1412 object_get_typename(OBJECT(phb)));
1413 }
1414 goto out;
1415 }
1416
1417 g_assert(drc);
1418
1419 /* Following the QEMU convention used for PCIe multifunction
1420 * hotplug, we do not allow functions to be hotplugged to a
1421 * slot that already has function 0 present
1422 */
1423 if (plugged_dev->hotplugged && bus->devices[PCI_DEVFN(slotnr, 0)] &&
1424 PCI_FUNC(pdev->devfn) != 0) {
1425 error_setg(&local_err, "PCI: slot %d function 0 already ocuppied by %s,"
1426 " additional functions can no longer be exposed to guest.",
1427 slotnr, bus->devices[PCI_DEVFN(slotnr, 0)]->name);
1428 goto out;
1429 }
1430
1431 fdt = create_device_tree(&fdt_size);
1432 fdt_start_offset = spapr_create_pci_child_dt(phb, pdev, fdt, 0);
1433
1434 spapr_drc_attach(drc, DEVICE(pdev), fdt, fdt_start_offset, &local_err);
1435 if (local_err) {
1436 goto out;
1437 }
1438
1439 /* If this is function 0, signal hotplug for all the device functions.
1440 * Otherwise defer sending the hotplug event.
1441 */
1442 if (!spapr_drc_hotplugged(plugged_dev)) {
1443 spapr_drc_reset(drc);
1444 } else if (PCI_FUNC(pdev->devfn) == 0) {
1445 int i;
1446
1447 for (i = 0; i < 8; i++) {
1448 sPAPRDRConnector *func_drc;
1449 sPAPRDRConnectorClass *func_drck;
1450 sPAPRDREntitySense state;
1451
1452 func_drc = spapr_phb_get_pci_func_drc(phb, pci_bus_num(bus),
1453 PCI_DEVFN(slotnr, i));
1454 func_drck = SPAPR_DR_CONNECTOR_GET_CLASS(func_drc);
1455 state = func_drck->dr_entity_sense(func_drc);
1456
1457 if (state == SPAPR_DR_ENTITY_SENSE_PRESENT) {
1458 spapr_hotplug_req_add_by_index(func_drc);
1459 }
1460 }
1461 }
1462
1463 out:
1464 if (local_err) {
1465 error_propagate(errp, local_err);
1466 g_free(fdt);
1467 }
1468 }
1469
1470 static void spapr_pci_unplug_request(HotplugHandler *plug_handler,
1471 DeviceState *plugged_dev, Error **errp)
1472 {
1473 sPAPRPHBState *phb = SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler));
1474 PCIDevice *pdev = PCI_DEVICE(plugged_dev);
1475 sPAPRDRConnector *drc = spapr_phb_get_pci_drc(phb, pdev);
1476
1477 if (!phb->dr_enabled) {
1478 error_setg(errp, QERR_BUS_NO_HOTPLUG,
1479 object_get_typename(OBJECT(phb)));
1480 return;
1481 }
1482
1483 g_assert(drc);
1484 g_assert(drc->dev == plugged_dev);
1485
1486 if (!spapr_drc_unplug_requested(drc)) {
1487 PCIBus *bus = PCI_BUS(qdev_get_parent_bus(DEVICE(pdev)));
1488 uint32_t slotnr = PCI_SLOT(pdev->devfn);
1489 sPAPRDRConnector *func_drc;
1490 sPAPRDRConnectorClass *func_drck;
1491 sPAPRDREntitySense state;
1492 int i;
1493
1494 /* ensure any other present functions are pending unplug */
1495 if (PCI_FUNC(pdev->devfn) == 0) {
1496 for (i = 1; i < 8; i++) {
1497 func_drc = spapr_phb_get_pci_func_drc(phb, pci_bus_num(bus),
1498 PCI_DEVFN(slotnr, i));
1499 func_drck = SPAPR_DR_CONNECTOR_GET_CLASS(func_drc);
1500 state = func_drck->dr_entity_sense(func_drc);
1501 if (state == SPAPR_DR_ENTITY_SENSE_PRESENT
1502 && !spapr_drc_unplug_requested(func_drc)) {
1503 error_setg(errp,
1504 "PCI: slot %d, function %d still present. "
1505 "Must unplug all non-0 functions first.",
1506 slotnr, i);
1507 return;
1508 }
1509 }
1510 }
1511
1512 spapr_drc_detach(drc);
1513
1514 /* if this isn't func 0, defer unplug event. otherwise signal removal
1515 * for all present functions
1516 */
1517 if (PCI_FUNC(pdev->devfn) == 0) {
1518 for (i = 7; i >= 0; i--) {
1519 func_drc = spapr_phb_get_pci_func_drc(phb, pci_bus_num(bus),
1520 PCI_DEVFN(slotnr, i));
1521 func_drck = SPAPR_DR_CONNECTOR_GET_CLASS(func_drc);
1522 state = func_drck->dr_entity_sense(func_drc);
1523 if (state == SPAPR_DR_ENTITY_SENSE_PRESENT) {
1524 spapr_hotplug_req_remove_by_index(func_drc);
1525 }
1526 }
1527 }
1528 }
1529 }
1530
1531 static void spapr_phb_realize(DeviceState *dev, Error **errp)
1532 {
1533 /* We don't use SPAPR_MACHINE() in order to exit gracefully if the user
1534 * tries to add a sPAPR PHB to a non-pseries machine.
1535 */
1536 sPAPRMachineState *spapr =
1537 (sPAPRMachineState *) object_dynamic_cast(qdev_get_machine(),
1538 TYPE_SPAPR_MACHINE);
1539 SysBusDevice *s = SYS_BUS_DEVICE(dev);
1540 sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(s);
1541 PCIHostState *phb = PCI_HOST_BRIDGE(s);
1542 char *namebuf;
1543 int i;
1544 PCIBus *bus;
1545 uint64_t msi_window_size = 4096;
1546 sPAPRTCETable *tcet;
1547 const unsigned windows_supported =
1548 sphb->ddw_enabled ? SPAPR_PCI_DMA_MAX_WINDOWS : 1;
1549
1550 if (!spapr) {
1551 error_setg(errp, TYPE_SPAPR_PCI_HOST_BRIDGE " needs a pseries machine");
1552 return;
1553 }
1554
1555 if (sphb->index != (uint32_t)-1) {
1556 sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
1557 Error *local_err = NULL;
1558
1559 smc->phb_placement(spapr, sphb->index,
1560 &sphb->buid, &sphb->io_win_addr,
1561 &sphb->mem_win_addr, &sphb->mem64_win_addr,
1562 windows_supported, sphb->dma_liobn, &local_err);
1563 if (local_err) {
1564 error_propagate(errp, local_err);
1565 return;
1566 }
1567 } else {
1568 error_setg(errp, "\"index\" for PAPR PHB is mandatory");
1569 return;
1570 }
1571
1572 if (sphb->mem64_win_size != 0) {
1573 if (sphb->mem_win_size > SPAPR_PCI_MEM32_WIN_SIZE) {
1574 error_setg(errp, "32-bit memory window of size 0x%"HWADDR_PRIx
1575 " (max 2 GiB)", sphb->mem_win_size);
1576 return;
1577 }
1578
1579 /* 64-bit window defaults to identity mapping */
1580 sphb->mem64_win_pciaddr = sphb->mem64_win_addr;
1581 } else if (sphb->mem_win_size > SPAPR_PCI_MEM32_WIN_SIZE) {
1582 /*
1583 * For compatibility with old configuration, if no 64-bit MMIO
1584 * window is specified, but the ordinary (32-bit) memory
1585 * window is specified as > 2GiB, we treat it as a 2GiB 32-bit
1586 * window, with a 64-bit MMIO window following on immediately
1587 * afterwards
1588 */
1589 sphb->mem64_win_size = sphb->mem_win_size - SPAPR_PCI_MEM32_WIN_SIZE;
1590 sphb->mem64_win_addr = sphb->mem_win_addr + SPAPR_PCI_MEM32_WIN_SIZE;
1591 sphb->mem64_win_pciaddr =
1592 SPAPR_PCI_MEM_WIN_BUS_OFFSET + SPAPR_PCI_MEM32_WIN_SIZE;
1593 sphb->mem_win_size = SPAPR_PCI_MEM32_WIN_SIZE;
1594 }
1595
1596 if (spapr_pci_find_phb(spapr, sphb->buid)) {
1597 error_setg(errp, "PCI host bridges must have unique BUIDs");
1598 return;
1599 }
1600
1601 if (sphb->numa_node != -1 &&
1602 (sphb->numa_node >= MAX_NODES || !numa_info[sphb->numa_node].present)) {
1603 error_setg(errp, "Invalid NUMA node ID for PCI host bridge");
1604 return;
1605 }
1606
1607 sphb->dtbusname = g_strdup_printf("pci@%" PRIx64, sphb->buid);
1608
1609 /* Initialize memory regions */
1610 namebuf = g_strdup_printf("%s.mmio", sphb->dtbusname);
1611 memory_region_init(&sphb->memspace, OBJECT(sphb), namebuf, UINT64_MAX);
1612 g_free(namebuf);
1613
1614 namebuf = g_strdup_printf("%s.mmio32-alias", sphb->dtbusname);
1615 memory_region_init_alias(&sphb->mem32window, OBJECT(sphb),
1616 namebuf, &sphb->memspace,
1617 SPAPR_PCI_MEM_WIN_BUS_OFFSET, sphb->mem_win_size);
1618 g_free(namebuf);
1619 memory_region_add_subregion(get_system_memory(), sphb->mem_win_addr,
1620 &sphb->mem32window);
1621
1622 if (sphb->mem64_win_size != 0) {
1623 namebuf = g_strdup_printf("%s.mmio64-alias", sphb->dtbusname);
1624 memory_region_init_alias(&sphb->mem64window, OBJECT(sphb),
1625 namebuf, &sphb->memspace,
1626 sphb->mem64_win_pciaddr, sphb->mem64_win_size);
1627 g_free(namebuf);
1628
1629 memory_region_add_subregion(get_system_memory(),
1630 sphb->mem64_win_addr,
1631 &sphb->mem64window);
1632 }
1633
1634 /* Initialize IO regions */
1635 namebuf = g_strdup_printf("%s.io", sphb->dtbusname);
1636 memory_region_init(&sphb->iospace, OBJECT(sphb),
1637 namebuf, SPAPR_PCI_IO_WIN_SIZE);
1638 g_free(namebuf);
1639
1640 namebuf = g_strdup_printf("%s.io-alias", sphb->dtbusname);
1641 memory_region_init_alias(&sphb->iowindow, OBJECT(sphb), namebuf,
1642 &sphb->iospace, 0, SPAPR_PCI_IO_WIN_SIZE);
1643 g_free(namebuf);
1644 memory_region_add_subregion(get_system_memory(), sphb->io_win_addr,
1645 &sphb->iowindow);
1646
1647 bus = pci_register_root_bus(dev, NULL,
1648 pci_spapr_set_irq, pci_spapr_map_irq, sphb,
1649 &sphb->memspace, &sphb->iospace,
1650 PCI_DEVFN(0, 0), PCI_NUM_PINS, TYPE_PCI_BUS);
1651 phb->bus = bus;
1652 qbus_set_hotplug_handler(BUS(phb->bus), DEVICE(sphb), NULL);
1653
1654 /*
1655 * Initialize PHB address space.
1656 * By default there will be at least one subregion for default
1657 * 32bit DMA window.
1658 * Later the guest might want to create another DMA window
1659 * which will become another memory subregion.
1660 */
1661 namebuf = g_strdup_printf("%s.iommu-root", sphb->dtbusname);
1662 memory_region_init(&sphb->iommu_root, OBJECT(sphb),
1663 namebuf, UINT64_MAX);
1664 g_free(namebuf);
1665 address_space_init(&sphb->iommu_as, &sphb->iommu_root,
1666 sphb->dtbusname);
1667
1668 /*
1669 * As MSI/MSIX interrupts trigger by writing at MSI/MSIX vectors,
1670 * we need to allocate some memory to catch those writes coming
1671 * from msi_notify()/msix_notify().
1672 * As MSIMessage:addr is going to be the same and MSIMessage:data
1673 * is going to be a VIRQ number, 4 bytes of the MSI MR will only
1674 * be used.
1675 *
1676 * For KVM we want to ensure that this memory is a full page so that
1677 * our memory slot is of page size granularity.
1678 */
1679 #ifdef CONFIG_KVM
1680 if (kvm_enabled()) {
1681 msi_window_size = getpagesize();
1682 }
1683 #endif
1684
1685 memory_region_init_io(&sphb->msiwindow, OBJECT(sphb), &spapr_msi_ops, spapr,
1686 "msi", msi_window_size);
1687 memory_region_add_subregion(&sphb->iommu_root, SPAPR_PCI_MSI_WINDOW,
1688 &sphb->msiwindow);
1689
1690 pci_setup_iommu(bus, spapr_pci_dma_iommu, sphb);
1691
1692 pci_bus_set_route_irq_fn(bus, spapr_route_intx_pin_to_irq);
1693
1694 QLIST_INSERT_HEAD(&spapr->phbs, sphb, list);
1695
1696 /* Initialize the LSI table */
1697 for (i = 0; i < PCI_NUM_PINS; i++) {
1698 uint32_t irq;
1699 Error *local_err = NULL;
1700
1701 irq = spapr_irq_alloc_block(spapr, 1, true, false, &local_err);
1702 if (local_err) {
1703 error_propagate(errp, local_err);
1704 error_prepend(errp, "can't allocate LSIs: ");
1705 return;
1706 }
1707
1708 sphb->lsi_table[i].irq = irq;
1709 }
1710
1711 /* allocate connectors for child PCI devices */
1712 if (sphb->dr_enabled) {
1713 for (i = 0; i < PCI_SLOT_MAX * 8; i++) {
1714 spapr_dr_connector_new(OBJECT(phb), TYPE_SPAPR_DRC_PCI,
1715 (sphb->index << 16) | i);
1716 }
1717 }
1718
1719 /* DMA setup */
1720 if (((sphb->page_size_mask & qemu_getrampagesize()) == 0)
1721 && kvm_enabled()) {
1722 warn_report("System page size 0x%lx is not enabled in page_size_mask "
1723 "(0x%"PRIx64"). Performance may be slow",
1724 qemu_getrampagesize(), sphb->page_size_mask);
1725 }
1726
1727 for (i = 0; i < windows_supported; ++i) {
1728 tcet = spapr_tce_new_table(DEVICE(sphb), sphb->dma_liobn[i]);
1729 if (!tcet) {
1730 error_setg(errp, "Creating window#%d failed for %s",
1731 i, sphb->dtbusname);
1732 return;
1733 }
1734 memory_region_add_subregion(&sphb->iommu_root, 0,
1735 spapr_tce_get_iommu(tcet));
1736 }
1737
1738 sphb->msi = g_hash_table_new_full(g_int_hash, g_int_equal, g_free, g_free);
1739 }
1740
1741 static int spapr_phb_children_reset(Object *child, void *opaque)
1742 {
1743 DeviceState *dev = (DeviceState *) object_dynamic_cast(child, TYPE_DEVICE);
1744
1745 if (dev) {
1746 device_reset(dev);
1747 }
1748
1749 return 0;
1750 }
1751
1752 void spapr_phb_dma_reset(sPAPRPHBState *sphb)
1753 {
1754 int i;
1755 sPAPRTCETable *tcet;
1756
1757 for (i = 0; i < SPAPR_PCI_DMA_MAX_WINDOWS; ++i) {
1758 tcet = spapr_tce_find_by_liobn(sphb->dma_liobn[i]);
1759
1760 if (tcet && tcet->nb_table) {
1761 spapr_tce_table_disable(tcet);
1762 }
1763 }
1764
1765 /* Register default 32bit DMA window */
1766 tcet = spapr_tce_find_by_liobn(sphb->dma_liobn[0]);
1767 spapr_tce_table_enable(tcet, SPAPR_TCE_PAGE_SHIFT, sphb->dma_win_addr,
1768 sphb->dma_win_size >> SPAPR_TCE_PAGE_SHIFT);
1769 }
1770
1771 static void spapr_phb_reset(DeviceState *qdev)
1772 {
1773 sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(qdev);
1774
1775 spapr_phb_dma_reset(sphb);
1776
1777 /* Reset the IOMMU state */
1778 object_child_foreach(OBJECT(qdev), spapr_phb_children_reset, NULL);
1779
1780 if (spapr_phb_eeh_available(SPAPR_PCI_HOST_BRIDGE(qdev))) {
1781 spapr_phb_vfio_reset(qdev);
1782 }
1783 }
1784
1785 static Property spapr_phb_properties[] = {
1786 DEFINE_PROP_UINT32("index", sPAPRPHBState, index, -1),
1787 DEFINE_PROP_UINT64("mem_win_size", sPAPRPHBState, mem_win_size,
1788 SPAPR_PCI_MEM32_WIN_SIZE),
1789 DEFINE_PROP_UINT64("mem64_win_size", sPAPRPHBState, mem64_win_size,
1790 SPAPR_PCI_MEM64_WIN_SIZE),
1791 DEFINE_PROP_UINT64("io_win_size", sPAPRPHBState, io_win_size,
1792 SPAPR_PCI_IO_WIN_SIZE),
1793 DEFINE_PROP_BOOL("dynamic-reconfiguration", sPAPRPHBState, dr_enabled,
1794 true),
1795 /* Default DMA window is 0..1GB */
1796 DEFINE_PROP_UINT64("dma_win_addr", sPAPRPHBState, dma_win_addr, 0),
1797 DEFINE_PROP_UINT64("dma_win_size", sPAPRPHBState, dma_win_size, 0x40000000),
1798 DEFINE_PROP_UINT64("dma64_win_addr", sPAPRPHBState, dma64_win_addr,
1799 0x800000000000000ULL),
1800 DEFINE_PROP_BOOL("ddw", sPAPRPHBState, ddw_enabled, true),
1801 DEFINE_PROP_UINT64("pgsz", sPAPRPHBState, page_size_mask,
1802 (1ULL << 12) | (1ULL << 16)),
1803 DEFINE_PROP_UINT32("numa_node", sPAPRPHBState, numa_node, -1),
1804 DEFINE_PROP_BOOL("pre-2.8-migration", sPAPRPHBState,
1805 pre_2_8_migration, false),
1806 DEFINE_PROP_BOOL("pcie-extended-configuration-space", sPAPRPHBState,
1807 pcie_ecs, true),
1808 DEFINE_PROP_END_OF_LIST(),
1809 };
1810
1811 static const VMStateDescription vmstate_spapr_pci_lsi = {
1812 .name = "spapr_pci/lsi",
1813 .version_id = 1,
1814 .minimum_version_id = 1,
1815 .fields = (VMStateField[]) {
1816 VMSTATE_UINT32_EQUAL(irq, struct spapr_pci_lsi, NULL),
1817
1818 VMSTATE_END_OF_LIST()
1819 },
1820 };
1821
1822 static const VMStateDescription vmstate_spapr_pci_msi = {
1823 .name = "spapr_pci/msi",
1824 .version_id = 1,
1825 .minimum_version_id = 1,
1826 .fields = (VMStateField []) {
1827 VMSTATE_UINT32(key, spapr_pci_msi_mig),
1828 VMSTATE_UINT32(value.first_irq, spapr_pci_msi_mig),
1829 VMSTATE_UINT32(value.num, spapr_pci_msi_mig),
1830 VMSTATE_END_OF_LIST()
1831 },
1832 };
1833
1834 static int spapr_pci_pre_save(void *opaque)
1835 {
1836 sPAPRPHBState *sphb = opaque;
1837 GHashTableIter iter;
1838 gpointer key, value;
1839 int i;
1840
1841 if (sphb->pre_2_8_migration) {
1842 sphb->mig_liobn = sphb->dma_liobn[0];
1843 sphb->mig_mem_win_addr = sphb->mem_win_addr;
1844 sphb->mig_mem_win_size = sphb->mem_win_size;
1845 sphb->mig_io_win_addr = sphb->io_win_addr;
1846 sphb->mig_io_win_size = sphb->io_win_size;
1847
1848 if ((sphb->mem64_win_size != 0)
1849 && (sphb->mem64_win_addr
1850 == (sphb->mem_win_addr + sphb->mem_win_size))) {
1851 sphb->mig_mem_win_size += sphb->mem64_win_size;
1852 }
1853 }
1854
1855 g_free(sphb->msi_devs);
1856 sphb->msi_devs = NULL;
1857 sphb->msi_devs_num = g_hash_table_size(sphb->msi);
1858 if (!sphb->msi_devs_num) {
1859 return 0;
1860 }
1861 sphb->msi_devs = g_malloc(sphb->msi_devs_num * sizeof(spapr_pci_msi_mig));
1862
1863 g_hash_table_iter_init(&iter, sphb->msi);
1864 for (i = 0; g_hash_table_iter_next(&iter, &key, &value); ++i) {
1865 sphb->msi_devs[i].key = *(uint32_t *) key;
1866 sphb->msi_devs[i].value = *(spapr_pci_msi *) value;
1867 }
1868
1869 return 0;
1870 }
1871
1872 static int spapr_pci_post_load(void *opaque, int version_id)
1873 {
1874 sPAPRPHBState *sphb = opaque;
1875 gpointer key, value;
1876 int i;
1877
1878 for (i = 0; i < sphb->msi_devs_num; ++i) {
1879 key = g_memdup(&sphb->msi_devs[i].key,
1880 sizeof(sphb->msi_devs[i].key));
1881 value = g_memdup(&sphb->msi_devs[i].value,
1882 sizeof(sphb->msi_devs[i].value));
1883 g_hash_table_insert(sphb->msi, key, value);
1884 }
1885 g_free(sphb->msi_devs);
1886 sphb->msi_devs = NULL;
1887 sphb->msi_devs_num = 0;
1888
1889 return 0;
1890 }
1891
1892 static bool pre_2_8_migration(void *opaque, int version_id)
1893 {
1894 sPAPRPHBState *sphb = opaque;
1895
1896 return sphb->pre_2_8_migration;
1897 }
1898
1899 static const VMStateDescription vmstate_spapr_pci = {
1900 .name = "spapr_pci",
1901 .version_id = 2,
1902 .minimum_version_id = 2,
1903 .pre_save = spapr_pci_pre_save,
1904 .post_load = spapr_pci_post_load,
1905 .fields = (VMStateField[]) {
1906 VMSTATE_UINT64_EQUAL(buid, sPAPRPHBState, NULL),
1907 VMSTATE_UINT32_TEST(mig_liobn, sPAPRPHBState, pre_2_8_migration),
1908 VMSTATE_UINT64_TEST(mig_mem_win_addr, sPAPRPHBState, pre_2_8_migration),
1909 VMSTATE_UINT64_TEST(mig_mem_win_size, sPAPRPHBState, pre_2_8_migration),
1910 VMSTATE_UINT64_TEST(mig_io_win_addr, sPAPRPHBState, pre_2_8_migration),
1911 VMSTATE_UINT64_TEST(mig_io_win_size, sPAPRPHBState, pre_2_8_migration),
1912 VMSTATE_STRUCT_ARRAY(lsi_table, sPAPRPHBState, PCI_NUM_PINS, 0,
1913 vmstate_spapr_pci_lsi, struct spapr_pci_lsi),
1914 VMSTATE_INT32(msi_devs_num, sPAPRPHBState),
1915 VMSTATE_STRUCT_VARRAY_ALLOC(msi_devs, sPAPRPHBState, msi_devs_num, 0,
1916 vmstate_spapr_pci_msi, spapr_pci_msi_mig),
1917 VMSTATE_END_OF_LIST()
1918 },
1919 };
1920
1921 static const char *spapr_phb_root_bus_path(PCIHostState *host_bridge,
1922 PCIBus *rootbus)
1923 {
1924 sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(host_bridge);
1925
1926 return sphb->dtbusname;
1927 }
1928
1929 static void spapr_phb_class_init(ObjectClass *klass, void *data)
1930 {
1931 PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(klass);
1932 DeviceClass *dc = DEVICE_CLASS(klass);
1933 HotplugHandlerClass *hp = HOTPLUG_HANDLER_CLASS(klass);
1934
1935 hc->root_bus_path = spapr_phb_root_bus_path;
1936 dc->realize = spapr_phb_realize;
1937 dc->props = spapr_phb_properties;
1938 dc->reset = spapr_phb_reset;
1939 dc->vmsd = &vmstate_spapr_pci;
1940 /* Supported by TYPE_SPAPR_MACHINE */
1941 dc->user_creatable = true;
1942 set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
1943 hp->plug = spapr_pci_plug;
1944 hp->unplug_request = spapr_pci_unplug_request;
1945 }
1946
1947 static const TypeInfo spapr_phb_info = {
1948 .name = TYPE_SPAPR_PCI_HOST_BRIDGE,
1949 .parent = TYPE_PCI_HOST_BRIDGE,
1950 .instance_size = sizeof(sPAPRPHBState),
1951 .class_init = spapr_phb_class_init,
1952 .interfaces = (InterfaceInfo[]) {
1953 { TYPE_HOTPLUG_HANDLER },
1954 { }
1955 }
1956 };
1957
1958 PCIHostState *spapr_create_phb(sPAPRMachineState *spapr, int index)
1959 {
1960 DeviceState *dev;
1961
1962 dev = qdev_create(NULL, TYPE_SPAPR_PCI_HOST_BRIDGE);
1963 qdev_prop_set_uint32(dev, "index", index);
1964 qdev_init_nofail(dev);
1965
1966 return PCI_HOST_BRIDGE(dev);
1967 }
1968
1969 typedef struct sPAPRFDT {
1970 void *fdt;
1971 int node_off;
1972 sPAPRPHBState *sphb;
1973 } sPAPRFDT;
1974
1975 static void spapr_populate_pci_devices_dt(PCIBus *bus, PCIDevice *pdev,
1976 void *opaque)
1977 {
1978 PCIBus *sec_bus;
1979 sPAPRFDT *p = opaque;
1980 int offset;
1981 sPAPRFDT s_fdt;
1982
1983 offset = spapr_create_pci_child_dt(p->sphb, pdev, p->fdt, p->node_off);
1984 if (!offset) {
1985 error_report("Failed to create pci child device tree node");
1986 return;
1987 }
1988
1989 if ((pci_default_read_config(pdev, PCI_HEADER_TYPE, 1) !=
1990 PCI_HEADER_TYPE_BRIDGE)) {
1991 return;
1992 }
1993
1994 sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
1995 if (!sec_bus) {
1996 return;
1997 }
1998
1999 s_fdt.fdt = p->fdt;
2000 s_fdt.node_off = offset;
2001 s_fdt.sphb = p->sphb;
2002 pci_for_each_device_reverse(sec_bus, pci_bus_num(sec_bus),
2003 spapr_populate_pci_devices_dt,
2004 &s_fdt);
2005 }
2006
2007 static void spapr_phb_pci_enumerate_bridge(PCIBus *bus, PCIDevice *pdev,
2008 void *opaque)
2009 {
2010 unsigned int *bus_no = opaque;
2011 unsigned int primary = *bus_no;
2012 unsigned int subordinate = 0xff;
2013 PCIBus *sec_bus = NULL;
2014
2015 if ((pci_default_read_config(pdev, PCI_HEADER_TYPE, 1) !=
2016 PCI_HEADER_TYPE_BRIDGE)) {
2017 return;
2018 }
2019
2020 (*bus_no)++;
2021 pci_default_write_config(pdev, PCI_PRIMARY_BUS, primary, 1);
2022 pci_default_write_config(pdev, PCI_SECONDARY_BUS, *bus_no, 1);
2023 pci_default_write_config(pdev, PCI_SUBORDINATE_BUS, *bus_no, 1);
2024
2025 sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
2026 if (!sec_bus) {
2027 return;
2028 }
2029
2030 pci_default_write_config(pdev, PCI_SUBORDINATE_BUS, subordinate, 1);
2031 pci_for_each_device(sec_bus, pci_bus_num(sec_bus),
2032 spapr_phb_pci_enumerate_bridge, bus_no);
2033 pci_default_write_config(pdev, PCI_SUBORDINATE_BUS, *bus_no, 1);
2034 }
2035
2036 static void spapr_phb_pci_enumerate(sPAPRPHBState *phb)
2037 {
2038 PCIBus *bus = PCI_HOST_BRIDGE(phb)->bus;
2039 unsigned int bus_no = 0;
2040
2041 pci_for_each_device(bus, pci_bus_num(bus),
2042 spapr_phb_pci_enumerate_bridge,
2043 &bus_no);
2044
2045 }
2046
2047 int spapr_populate_pci_dt(sPAPRPHBState *phb,
2048 uint32_t xics_phandle,
2049 void *fdt)
2050 {
2051 int bus_off, i, j, ret;
2052 gchar *nodename;
2053 uint32_t bus_range[] = { cpu_to_be32(0), cpu_to_be32(0xff) };
2054 struct {
2055 uint32_t hi;
2056 uint64_t child;
2057 uint64_t parent;
2058 uint64_t size;
2059 } QEMU_PACKED ranges[] = {
2060 {
2061 cpu_to_be32(b_ss(1)), cpu_to_be64(0),
2062 cpu_to_be64(phb->io_win_addr),
2063 cpu_to_be64(memory_region_size(&phb->iospace)),
2064 },
2065 {
2066 cpu_to_be32(b_ss(2)), cpu_to_be64(SPAPR_PCI_MEM_WIN_BUS_OFFSET),
2067 cpu_to_be64(phb->mem_win_addr),
2068 cpu_to_be64(phb->mem_win_size),
2069 },
2070 {
2071 cpu_to_be32(b_ss(3)), cpu_to_be64(phb->mem64_win_pciaddr),
2072 cpu_to_be64(phb->mem64_win_addr),
2073 cpu_to_be64(phb->mem64_win_size),
2074 },
2075 };
2076 const unsigned sizeof_ranges =
2077 (phb->mem64_win_size ? 3 : 2) * sizeof(ranges[0]);
2078 uint64_t bus_reg[] = { cpu_to_be64(phb->buid), 0 };
2079 uint32_t interrupt_map_mask[] = {
2080 cpu_to_be32(b_ddddd(-1)|b_fff(0)), 0x0, 0x0, cpu_to_be32(-1)};
2081 uint32_t interrupt_map[PCI_SLOT_MAX * PCI_NUM_PINS][7];
2082 uint32_t ddw_applicable[] = {
2083 cpu_to_be32(RTAS_IBM_QUERY_PE_DMA_WINDOW),
2084 cpu_to_be32(RTAS_IBM_CREATE_PE_DMA_WINDOW),
2085 cpu_to_be32(RTAS_IBM_REMOVE_PE_DMA_WINDOW)
2086 };
2087 uint32_t ddw_extensions[] = {
2088 cpu_to_be32(1),
2089 cpu_to_be32(RTAS_IBM_RESET_PE_DMA_WINDOW)
2090 };
2091 uint32_t associativity[] = {cpu_to_be32(0x4),
2092 cpu_to_be32(0x0),
2093 cpu_to_be32(0x0),
2094 cpu_to_be32(0x0),
2095 cpu_to_be32(phb->numa_node)};
2096 sPAPRTCETable *tcet;
2097 PCIBus *bus = PCI_HOST_BRIDGE(phb)->bus;
2098 sPAPRFDT s_fdt;
2099
2100 /* Start populating the FDT */
2101 nodename = g_strdup_printf("pci@%" PRIx64, phb->buid);
2102 _FDT(bus_off = fdt_add_subnode(fdt, 0, nodename));
2103 g_free(nodename);
2104
2105 /* Write PHB properties */
2106 _FDT(fdt_setprop_string(fdt, bus_off, "device_type", "pci"));
2107 _FDT(fdt_setprop_string(fdt, bus_off, "compatible", "IBM,Logical_PHB"));
2108 _FDT(fdt_setprop_cell(fdt, bus_off, "#address-cells", 0x3));
2109 _FDT(fdt_setprop_cell(fdt, bus_off, "#size-cells", 0x2));
2110 _FDT(fdt_setprop_cell(fdt, bus_off, "#interrupt-cells", 0x1));
2111 _FDT(fdt_setprop(fdt, bus_off, "used-by-rtas", NULL, 0));
2112 _FDT(fdt_setprop(fdt, bus_off, "bus-range", &bus_range, sizeof(bus_range)));
2113 _FDT(fdt_setprop(fdt, bus_off, "ranges", &ranges, sizeof_ranges));
2114 _FDT(fdt_setprop(fdt, bus_off, "reg", &bus_reg, sizeof(bus_reg)));
2115 _FDT(fdt_setprop_cell(fdt, bus_off, "ibm,pci-config-space-type", 0x1));
2116 _FDT(fdt_setprop_cell(fdt, bus_off, "ibm,pe-total-#msi", XICS_IRQS_SPAPR));
2117
2118 /* Dynamic DMA window */
2119 if (phb->ddw_enabled) {
2120 _FDT(fdt_setprop(fdt, bus_off, "ibm,ddw-applicable", &ddw_applicable,
2121 sizeof(ddw_applicable)));
2122 _FDT(fdt_setprop(fdt, bus_off, "ibm,ddw-extensions",
2123 &ddw_extensions, sizeof(ddw_extensions)));
2124 }
2125
2126 /* Advertise NUMA via ibm,associativity */
2127 if (phb->numa_node != -1) {
2128 _FDT(fdt_setprop(fdt, bus_off, "ibm,associativity", associativity,
2129 sizeof(associativity)));
2130 }
2131
2132 /* Build the interrupt-map, this must matches what is done
2133 * in pci_spapr_map_irq
2134 */
2135 _FDT(fdt_setprop(fdt, bus_off, "interrupt-map-mask",
2136 &interrupt_map_mask, sizeof(interrupt_map_mask)));
2137 for (i = 0; i < PCI_SLOT_MAX; i++) {
2138 for (j = 0; j < PCI_NUM_PINS; j++) {
2139 uint32_t *irqmap = interrupt_map[i*PCI_NUM_PINS + j];
2140 int lsi_num = pci_spapr_swizzle(i, j);
2141
2142 irqmap[0] = cpu_to_be32(b_ddddd(i)|b_fff(0));
2143 irqmap[1] = 0;
2144 irqmap[2] = 0;
2145 irqmap[3] = cpu_to_be32(j+1);
2146 irqmap[4] = cpu_to_be32(xics_phandle);
2147 spapr_dt_xics_irq(&irqmap[5], phb->lsi_table[lsi_num].irq, true);
2148 }
2149 }
2150 /* Write interrupt map */
2151 _FDT(fdt_setprop(fdt, bus_off, "interrupt-map", &interrupt_map,
2152 sizeof(interrupt_map)));
2153
2154 tcet = spapr_tce_find_by_liobn(phb->dma_liobn[0]);
2155 if (!tcet) {
2156 return -1;
2157 }
2158 spapr_dma_dt(fdt, bus_off, "ibm,dma-window",
2159 tcet->liobn, tcet->bus_offset,
2160 tcet->nb_table << tcet->page_shift);
2161
2162 /* Walk the bridges and program the bus numbers*/
2163 spapr_phb_pci_enumerate(phb);
2164 _FDT(fdt_setprop_cell(fdt, bus_off, "qemu,phb-enumerated", 0x1));
2165
2166 /* Populate tree nodes with PCI devices attached */
2167 s_fdt.fdt = fdt;
2168 s_fdt.node_off = bus_off;
2169 s_fdt.sphb = phb;
2170 pci_for_each_device_reverse(bus, pci_bus_num(bus),
2171 spapr_populate_pci_devices_dt,
2172 &s_fdt);
2173
2174 ret = spapr_drc_populate_dt(fdt, bus_off, OBJECT(phb),
2175 SPAPR_DR_CONNECTOR_TYPE_PCI);
2176 if (ret) {
2177 return ret;
2178 }
2179
2180 return 0;
2181 }
2182
2183 void spapr_pci_rtas_init(void)
2184 {
2185 spapr_rtas_register(RTAS_READ_PCI_CONFIG, "read-pci-config",
2186 rtas_read_pci_config);
2187 spapr_rtas_register(RTAS_WRITE_PCI_CONFIG, "write-pci-config",
2188 rtas_write_pci_config);
2189 spapr_rtas_register(RTAS_IBM_READ_PCI_CONFIG, "ibm,read-pci-config",
2190 rtas_ibm_read_pci_config);
2191 spapr_rtas_register(RTAS_IBM_WRITE_PCI_CONFIG, "ibm,write-pci-config",
2192 rtas_ibm_write_pci_config);
2193 if (msi_nonbroken) {
2194 spapr_rtas_register(RTAS_IBM_QUERY_INTERRUPT_SOURCE_NUMBER,
2195 "ibm,query-interrupt-source-number",
2196 rtas_ibm_query_interrupt_source_number);
2197 spapr_rtas_register(RTAS_IBM_CHANGE_MSI, "ibm,change-msi",
2198 rtas_ibm_change_msi);
2199 }
2200
2201 spapr_rtas_register(RTAS_IBM_SET_EEH_OPTION,
2202 "ibm,set-eeh-option",
2203 rtas_ibm_set_eeh_option);
2204 spapr_rtas_register(RTAS_IBM_GET_CONFIG_ADDR_INFO2,
2205 "ibm,get-config-addr-info2",
2206 rtas_ibm_get_config_addr_info2);
2207 spapr_rtas_register(RTAS_IBM_READ_SLOT_RESET_STATE2,
2208 "ibm,read-slot-reset-state2",
2209 rtas_ibm_read_slot_reset_state2);
2210 spapr_rtas_register(RTAS_IBM_SET_SLOT_RESET,
2211 "ibm,set-slot-reset",
2212 rtas_ibm_set_slot_reset);
2213 spapr_rtas_register(RTAS_IBM_CONFIGURE_PE,
2214 "ibm,configure-pe",
2215 rtas_ibm_configure_pe);
2216 spapr_rtas_register(RTAS_IBM_SLOT_ERROR_DETAIL,
2217 "ibm,slot-error-detail",
2218 rtas_ibm_slot_error_detail);
2219 }
2220
2221 static void spapr_pci_register_types(void)
2222 {
2223 type_register_static(&spapr_phb_info);
2224 }
2225
2226 type_init(spapr_pci_register_types)
2227
2228 static int spapr_switch_one_vga(DeviceState *dev, void *opaque)
2229 {
2230 bool be = *(bool *)opaque;
2231
2232 if (object_dynamic_cast(OBJECT(dev), "VGA")
2233 || object_dynamic_cast(OBJECT(dev), "secondary-vga")) {
2234 object_property_set_bool(OBJECT(dev), be, "big-endian-framebuffer",
2235 &error_abort);
2236 }
2237 return 0;
2238 }
2239
2240 void spapr_pci_switch_vga(bool big_endian)
2241 {
2242 sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
2243 sPAPRPHBState *sphb;
2244
2245 /*
2246 * For backward compatibility with existing guests, we switch
2247 * the endianness of the VGA controller when changing the guest
2248 * interrupt mode
2249 */
2250 QLIST_FOREACH(sphb, &spapr->phbs, list) {
2251 BusState *bus = &PCI_HOST_BRIDGE(sphb)->bus->qbus;
2252 qbus_walk_children(bus, spapr_switch_one_vga, NULL, NULL, NULL,
2253 &big_endian);
2254 }
2255 }
QEMU mirror