Merge remote branch 'kwolf/for-anthony' into staging
[qemu.git] / hw / pcie.c
blob6a113a9327bb1e4759b064c28c4dae490f2db6e2
1 /*
2 * pcie.c
4 * Copyright (c) 2010 Isaku Yamahata <yamahata at valinux co jp>
5 * VA Linux Systems Japan K.K.
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License along
18 * with this program; if not, see <http://www.gnu.org/licenses/>.
21 #include "sysemu.h"
22 #include "range.h"
23 #include "pci_bridge.h"
24 #include "pcie.h"
25 #include "msix.h"
26 #include "msi.h"
27 #include "pci_internals.h"
28 #include "pcie_regs.h"
29 #include "range.h"
31 //#define DEBUG_PCIE
32 #ifdef DEBUG_PCIE
33 # define PCIE_DPRINTF(fmt, ...) \
34 fprintf(stderr, "%s:%d " fmt, __func__, __LINE__, ## __VA_ARGS__)
35 #else
36 # define PCIE_DPRINTF(fmt, ...) do {} while (0)
37 #endif
38 #define PCIE_DEV_PRINTF(dev, fmt, ...) \
39 PCIE_DPRINTF("%s:%x "fmt, (dev)->name, (dev)->devfn, ## __VA_ARGS__)
42 /***************************************************************************
43 * pci express capability helper functions
45 int pcie_cap_init(PCIDevice *dev, uint8_t offset, uint8_t type, uint8_t port)
47 int pos;
48 uint8_t *exp_cap;
50 assert(pci_is_express(dev));
52 pos = pci_add_capability(dev, PCI_CAP_ID_EXP, offset,
53 PCI_EXP_VER2_SIZEOF);
54 if (pos < 0) {
55 return pos;
57 dev->exp.exp_cap = pos;
58 exp_cap = dev->config + pos;
60 /* capability register
61 interrupt message number defaults to 0 */
62 pci_set_word(exp_cap + PCI_EXP_FLAGS,
63 ((type << PCI_EXP_FLAGS_TYPE_SHIFT) & PCI_EXP_FLAGS_TYPE) |
64 PCI_EXP_FLAGS_VER2);
66 /* device capability register
67 * table 7-12:
68 * roll based error reporting bit must be set by all
69 * Functions conforming to the ECN, PCI Express Base
70 * Specification, Revision 1.1., or subsequent PCI Express Base
71 * Specification revisions.
73 pci_set_long(exp_cap + PCI_EXP_DEVCAP, PCI_EXP_DEVCAP_RBER);
75 pci_set_long(exp_cap + PCI_EXP_LNKCAP,
76 (port << PCI_EXP_LNKCAP_PN_SHIFT) |
77 PCI_EXP_LNKCAP_ASPMS_0S |
78 PCI_EXP_LNK_MLW_1 |
79 PCI_EXP_LNK_LS_25);
81 pci_set_word(exp_cap + PCI_EXP_LNKSTA,
82 PCI_EXP_LNK_MLW_1 | PCI_EXP_LNK_LS_25);
84 pci_set_long(exp_cap + PCI_EXP_DEVCAP2,
85 PCI_EXP_DEVCAP2_EFF | PCI_EXP_DEVCAP2_EETLPP);
87 pci_set_word(dev->wmask + pos, PCI_EXP_DEVCTL2_EETLPPB);
88 return pos;
91 void pcie_cap_exit(PCIDevice *dev)
93 pci_del_capability(dev, PCI_CAP_ID_EXP, PCI_EXP_VER2_SIZEOF);
96 uint8_t pcie_cap_get_type(const PCIDevice *dev)
98 uint32_t pos = dev->exp.exp_cap;
99 assert(pos > 0);
100 return (pci_get_word(dev->config + pos + PCI_EXP_FLAGS) &
101 PCI_EXP_FLAGS_TYPE) >> PCI_EXP_FLAGS_TYPE_SHIFT;
104 /* MSI/MSI-X */
105 /* pci express interrupt message number */
106 /* 7.8.2 PCI Express Capabilities Register: Interrupt Message Number */
107 void pcie_cap_flags_set_vector(PCIDevice *dev, uint8_t vector)
109 uint8_t *exp_cap = dev->config + dev->exp.exp_cap;
110 assert(vector < 32);
111 pci_word_test_and_clear_mask(exp_cap + PCI_EXP_FLAGS, PCI_EXP_FLAGS_IRQ);
112 pci_word_test_and_set_mask(exp_cap + PCI_EXP_FLAGS,
113 vector << PCI_EXP_FLAGS_IRQ_SHIFT);
116 uint8_t pcie_cap_flags_get_vector(PCIDevice *dev)
118 return (pci_get_word(dev->config + dev->exp.exp_cap + PCI_EXP_FLAGS) &
119 PCI_EXP_FLAGS_IRQ) >> PCI_EXP_FLAGS_IRQ_SHIFT;
122 void pcie_cap_deverr_init(PCIDevice *dev)
124 uint32_t pos = dev->exp.exp_cap;
125 pci_long_test_and_set_mask(dev->config + pos + PCI_EXP_DEVCAP,
126 PCI_EXP_DEVCAP_RBER);
127 pci_long_test_and_set_mask(dev->wmask + pos + PCI_EXP_DEVCTL,
128 PCI_EXP_DEVCTL_CERE | PCI_EXP_DEVCTL_NFERE |
129 PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE);
130 pci_long_test_and_set_mask(dev->w1cmask + pos + PCI_EXP_DEVSTA,
131 PCI_EXP_DEVSTA_CED | PCI_EXP_DEVSTA_NFED |
132 PCI_EXP_DEVSTA_URD | PCI_EXP_DEVSTA_URD);
135 void pcie_cap_deverr_reset(PCIDevice *dev)
137 uint8_t *devctl = dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL;
138 pci_long_test_and_clear_mask(devctl,
139 PCI_EXP_DEVCTL_CERE | PCI_EXP_DEVCTL_NFERE |
140 PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE);
143 static void hotplug_event_update_event_status(PCIDevice *dev)
145 uint32_t pos = dev->exp.exp_cap;
146 uint8_t *exp_cap = dev->config + pos;
147 uint16_t sltctl = pci_get_word(exp_cap + PCI_EXP_SLTCTL);
148 uint16_t sltsta = pci_get_word(exp_cap + PCI_EXP_SLTSTA);
150 dev->exp.hpev_notified = (sltctl & PCI_EXP_SLTCTL_HPIE) &&
151 (sltsta & sltctl & PCI_EXP_HP_EV_SUPPORTED);
154 static void hotplug_event_notify(PCIDevice *dev)
156 bool prev = dev->exp.hpev_notified;
158 hotplug_event_update_event_status(dev);
160 if (prev == dev->exp.hpev_notified) {
161 return;
164 /* Note: the logic above does not take into account whether interrupts
165 * are masked. The result is that interrupt will be sent when it is
166 * subsequently unmasked. This appears to be legal: Section 6.7.3.4:
167 * The Port may optionally send an MSI when there are hot-plug events that
168 * occur while interrupt generation is disabled, and interrupt generation is
169 * subsequently enabled. */
170 if (msix_enabled(dev)) {
171 msix_notify(dev, pcie_cap_flags_get_vector(dev));
172 } else if (msi_enabled(dev)) {
173 msi_notify(dev, pcie_cap_flags_get_vector(dev));
174 } else {
175 qemu_set_irq(dev->irq[dev->exp.hpev_intx], dev->exp.hpev_notified);
180 * A PCI Express Hot-Plug Event has occured, so update slot status register
181 * and notify OS of the event if necessary.
183 * 6.7.3 PCI Express Hot-Plug Events
184 * 6.7.3.4 Software Notification of Hot-Plug Events
186 static void pcie_cap_slot_event(PCIDevice *dev, PCIExpressHotPlugEvent event)
188 /* Minor optimization: if nothing changed - no event is needed. */
189 if (pci_word_test_and_set_mask(dev->config + dev->exp.exp_cap +
190 PCI_EXP_SLTSTA, event)) {
191 return;
193 hotplug_event_notify(dev);
196 static int pcie_cap_slot_hotplug(DeviceState *qdev,
197 PCIDevice *pci_dev, PCIHotplugState state)
199 PCIDevice *d = DO_UPCAST(PCIDevice, qdev, qdev);
200 uint8_t *exp_cap = d->config + d->exp.exp_cap;
201 uint16_t sltsta = pci_get_word(exp_cap + PCI_EXP_SLTSTA);
203 /* Don't send event when device is enabled during qemu machine creation:
204 * it is present on boot, no hotplug event is necessary. We do send an
205 * event when the device is disabled later. */
206 if (state == PCI_COLDPLUG_ENABLED) {
207 pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTSTA,
208 PCI_EXP_SLTSTA_PDS);
209 return 0;
212 PCIE_DEV_PRINTF(pci_dev, "hotplug state: %d\n", state);
213 if (sltsta & PCI_EXP_SLTSTA_EIS) {
214 /* the slot is electromechanically locked.
215 * This error is propagated up to qdev and then to HMP/QMP.
217 return -EBUSY;
220 /* TODO: multifunction hot-plug.
221 * Right now, only a device of function = 0 is allowed to be
222 * hot plugged/unplugged.
224 assert(PCI_FUNC(pci_dev->devfn) == 0);
226 if (state == PCI_HOTPLUG_ENABLED) {
227 pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTSTA,
228 PCI_EXP_SLTSTA_PDS);
229 pcie_cap_slot_event(d, PCI_EXP_HP_EV_PDC);
230 } else {
231 qdev_free(&pci_dev->qdev);
232 pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTSTA,
233 PCI_EXP_SLTSTA_PDS);
234 pcie_cap_slot_event(d, PCI_EXP_HP_EV_PDC);
236 return 0;
239 /* pci express slot for pci express root/downstream port
240 PCI express capability slot registers */
241 void pcie_cap_slot_init(PCIDevice *dev, uint16_t slot)
243 uint32_t pos = dev->exp.exp_cap;
245 pci_word_test_and_set_mask(dev->config + pos + PCI_EXP_FLAGS,
246 PCI_EXP_FLAGS_SLOT);
248 pci_long_test_and_clear_mask(dev->config + pos + PCI_EXP_SLTCAP,
249 ~PCI_EXP_SLTCAP_PSN);
250 pci_long_test_and_set_mask(dev->config + pos + PCI_EXP_SLTCAP,
251 (slot << PCI_EXP_SLTCAP_PSN_SHIFT) |
252 PCI_EXP_SLTCAP_EIP |
253 PCI_EXP_SLTCAP_HPS |
254 PCI_EXP_SLTCAP_HPC |
255 PCI_EXP_SLTCAP_PIP |
256 PCI_EXP_SLTCAP_AIP |
257 PCI_EXP_SLTCAP_ABP);
259 pci_word_test_and_clear_mask(dev->config + pos + PCI_EXP_SLTCTL,
260 PCI_EXP_SLTCTL_PIC |
261 PCI_EXP_SLTCTL_AIC);
262 pci_word_test_and_set_mask(dev->config + pos + PCI_EXP_SLTCTL,
263 PCI_EXP_SLTCTL_PIC_OFF |
264 PCI_EXP_SLTCTL_AIC_OFF);
265 pci_word_test_and_set_mask(dev->wmask + pos + PCI_EXP_SLTCTL,
266 PCI_EXP_SLTCTL_PIC |
267 PCI_EXP_SLTCTL_AIC |
268 PCI_EXP_SLTCTL_HPIE |
269 PCI_EXP_SLTCTL_CCIE |
270 PCI_EXP_SLTCTL_PDCE |
271 PCI_EXP_SLTCTL_ABPE);
272 /* Although reading PCI_EXP_SLTCTL_EIC returns always 0,
273 * make the bit writable here in order to detect 1b is written.
274 * pcie_cap_slot_write_config() test-and-clear the bit, so
275 * this bit always returns 0 to the guest.
277 pci_word_test_and_set_mask(dev->wmask + pos + PCI_EXP_SLTCTL,
278 PCI_EXP_SLTCTL_EIC);
280 pci_word_test_and_set_mask(dev->w1cmask + pos + PCI_EXP_SLTSTA,
281 PCI_EXP_HP_EV_SUPPORTED);
283 dev->exp.hpev_notified = false;
285 pci_bus_hotplug(pci_bridge_get_sec_bus(DO_UPCAST(PCIBridge, dev, dev)),
286 pcie_cap_slot_hotplug, &dev->qdev);
289 void pcie_cap_slot_reset(PCIDevice *dev)
291 uint8_t *exp_cap = dev->config + dev->exp.exp_cap;
293 PCIE_DEV_PRINTF(dev, "reset\n");
295 pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTCTL,
296 PCI_EXP_SLTCTL_EIC |
297 PCI_EXP_SLTCTL_PIC |
298 PCI_EXP_SLTCTL_AIC |
299 PCI_EXP_SLTCTL_HPIE |
300 PCI_EXP_SLTCTL_CCIE |
301 PCI_EXP_SLTCTL_PDCE |
302 PCI_EXP_SLTCTL_ABPE);
303 pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTCTL,
304 PCI_EXP_SLTCTL_PIC_OFF |
305 PCI_EXP_SLTCTL_AIC_OFF);
307 pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTSTA,
308 PCI_EXP_SLTSTA_EIS |/* on reset,
309 the lock is released */
310 PCI_EXP_SLTSTA_CC |
311 PCI_EXP_SLTSTA_PDC |
312 PCI_EXP_SLTSTA_ABP);
314 hotplug_event_update_event_status(dev);
317 void pcie_cap_slot_write_config(PCIDevice *dev,
318 uint32_t addr, uint32_t val, int len)
320 uint32_t pos = dev->exp.exp_cap;
321 uint8_t *exp_cap = dev->config + pos;
322 uint16_t sltsta = pci_get_word(exp_cap + PCI_EXP_SLTSTA);
324 if (!ranges_overlap(addr, len, pos + PCI_EXP_SLTCTL, 2)) {
325 return;
328 if (pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTCTL,
329 PCI_EXP_SLTCTL_EIC)) {
330 sltsta ^= PCI_EXP_SLTSTA_EIS; /* toggle PCI_EXP_SLTSTA_EIS bit */
331 pci_set_word(exp_cap + PCI_EXP_SLTSTA, sltsta);
332 PCIE_DEV_PRINTF(dev, "PCI_EXP_SLTCTL_EIC: "
333 "sltsta -> 0x%02"PRIx16"\n",
334 sltsta);
337 hotplug_event_notify(dev);
340 * 6.7.3.2 Command Completed Events
342 * Software issues a command to a hot-plug capable Downstream Port by
343 * issuing a write transaction that targets any portion of the Port’s Slot
344 * Control register. A single write to the Slot Control register is
345 * considered to be a single command, even if the write affects more than
346 * one field in the Slot Control register. In response to this transaction,
347 * the Port must carry out the requested actions and then set the
348 * associated status field for the command completed event. */
350 /* Real hardware might take a while to complete requested command because
351 * physical movement would be involved like locking the electromechanical
352 * lock. However in our case, command is completed instantaneously above,
353 * so send a command completion event right now.
355 pcie_cap_slot_event(dev, PCI_EXP_HP_EV_CCI);
358 int pcie_cap_slot_post_load(void *opaque, int version_id)
360 PCIDevice *dev = opaque;
361 hotplug_event_update_event_status(dev);
362 return 0;
365 void pcie_cap_slot_push_attention_button(PCIDevice *dev)
367 pcie_cap_slot_event(dev, PCI_EXP_HP_EV_ABP);
370 /* root control/capabilities/status. PME isn't emulated for now */
371 void pcie_cap_root_init(PCIDevice *dev)
373 pci_set_word(dev->wmask + dev->exp.exp_cap + PCI_EXP_RTCTL,
374 PCI_EXP_RTCTL_SECEE | PCI_EXP_RTCTL_SENFEE |
375 PCI_EXP_RTCTL_SEFEE);
378 void pcie_cap_root_reset(PCIDevice *dev)
380 pci_set_word(dev->config + dev->exp.exp_cap + PCI_EXP_RTCTL, 0);
383 /* function level reset(FLR) */
384 void pcie_cap_flr_init(PCIDevice *dev)
386 pci_long_test_and_set_mask(dev->config + dev->exp.exp_cap + PCI_EXP_DEVCAP,
387 PCI_EXP_DEVCAP_FLR);
389 /* Although reading BCR_FLR returns always 0,
390 * the bit is made writable here in order to detect the 1b is written
391 * pcie_cap_flr_write_config() test-and-clear the bit, so
392 * this bit always returns 0 to the guest.
394 pci_word_test_and_set_mask(dev->wmask + dev->exp.exp_cap + PCI_EXP_DEVCTL,
395 PCI_EXP_DEVCTL_BCR_FLR);
398 void pcie_cap_flr_write_config(PCIDevice *dev,
399 uint32_t addr, uint32_t val, int len)
401 uint8_t *devctl = dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL;
402 if (pci_get_word(devctl) & PCI_EXP_DEVCTL_BCR_FLR) {
403 /* Clear PCI_EXP_DEVCTL_BCR_FLR after invoking the reset handler
404 so the handler can detect FLR by looking at this bit. */
405 pci_device_reset(dev);
406 pci_word_test_and_clear_mask(devctl, PCI_EXP_DEVCTL_BCR_FLR);
410 /* Alternative Routing-ID Interpretation (ARI) */
411 /* ari forwarding support for down stream port */
412 void pcie_cap_ari_init(PCIDevice *dev)
414 uint32_t pos = dev->exp.exp_cap;
415 pci_long_test_and_set_mask(dev->config + pos + PCI_EXP_DEVCAP2,
416 PCI_EXP_DEVCAP2_ARI);
417 pci_long_test_and_set_mask(dev->wmask + pos + PCI_EXP_DEVCTL2,
418 PCI_EXP_DEVCTL2_ARI);
421 void pcie_cap_ari_reset(PCIDevice *dev)
423 uint8_t *devctl2 = dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL2;
424 pci_long_test_and_clear_mask(devctl2, PCI_EXP_DEVCTL2_ARI);
427 bool pcie_cap_is_ari_enabled(const PCIDevice *dev)
429 if (!pci_is_express(dev)) {
430 return false;
432 if (!dev->exp.exp_cap) {
433 return false;
436 return pci_get_long(dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL2) &
437 PCI_EXP_DEVCTL2_ARI;
440 /**************************************************************************
441 * pci express extended capability allocation functions
442 * uint16_t ext_cap_id (16 bit)
443 * uint8_t cap_ver (4 bit)
444 * uint16_t cap_offset (12 bit)
445 * uint16_t ext_cap_size
448 static uint16_t pcie_find_capability_list(PCIDevice *dev, uint16_t cap_id,
449 uint16_t *prev_p)
451 uint16_t prev = 0;
452 uint16_t next;
453 uint32_t header = pci_get_long(dev->config + PCI_CONFIG_SPACE_SIZE);
455 if (!header) {
456 /* no extended capability */
457 next = 0;
458 goto out;
460 for (next = PCI_CONFIG_SPACE_SIZE; next;
461 prev = next, next = PCI_EXT_CAP_NEXT(header)) {
463 assert(next >= PCI_CONFIG_SPACE_SIZE);
464 assert(next <= PCIE_CONFIG_SPACE_SIZE - 8);
466 header = pci_get_long(dev->config + next);
467 if (PCI_EXT_CAP_ID(header) == cap_id) {
468 break;
472 out:
473 if (prev_p) {
474 *prev_p = prev;
476 return next;
479 uint16_t pcie_find_capability(PCIDevice *dev, uint16_t cap_id)
481 return pcie_find_capability_list(dev, cap_id, NULL);
484 static void pcie_ext_cap_set_next(PCIDevice *dev, uint16_t pos, uint16_t next)
486 uint16_t header = pci_get_long(dev->config + pos);
487 assert(!(next & (PCI_EXT_CAP_ALIGN - 1)));
488 header = (header & ~PCI_EXT_CAP_NEXT_MASK) |
489 ((next << PCI_EXT_CAP_NEXT_SHIFT) & PCI_EXT_CAP_NEXT_MASK);
490 pci_set_long(dev->config + pos, header);
494 * caller must supply valid (offset, size) * such that the range shouldn't
495 * overlap with other capability or other registers.
496 * This function doesn't check it.
498 void pcie_add_capability(PCIDevice *dev,
499 uint16_t cap_id, uint8_t cap_ver,
500 uint16_t offset, uint16_t size)
502 uint32_t header;
503 uint16_t next;
505 assert(offset >= PCI_CONFIG_SPACE_SIZE);
506 assert(offset < offset + size);
507 assert(offset + size < PCIE_CONFIG_SPACE_SIZE);
508 assert(size >= 8);
509 assert(pci_is_express(dev));
511 if (offset == PCI_CONFIG_SPACE_SIZE) {
512 header = pci_get_long(dev->config + offset);
513 next = PCI_EXT_CAP_NEXT(header);
514 } else {
515 uint16_t prev;
517 /* 0 is reserved cap id. use internally to find the last capability
518 in the linked list */
519 next = pcie_find_capability_list(dev, 0, &prev);
521 assert(prev >= PCI_CONFIG_SPACE_SIZE);
522 assert(next == 0);
523 pcie_ext_cap_set_next(dev, prev, offset);
525 pci_set_long(dev->config + offset, PCI_EXT_CAP(cap_id, cap_ver, next));
527 /* Make capability read-only by default */
528 memset(dev->wmask + offset, 0, size);
529 memset(dev->w1cmask + offset, 0, size);
530 /* Check capability by default */
531 memset(dev->cmask + offset, 0xFF, size);
534 /**************************************************************************
535 * pci express extended capability helper functions
538 /* ARI */
539 void pcie_ari_init(PCIDevice *dev, uint16_t offset, uint16_t nextfn)
541 pcie_add_capability(dev, PCI_EXT_CAP_ID_ARI, PCI_ARI_VER,
542 offset, PCI_ARI_SIZEOF);
543 pci_set_long(dev->config + offset + PCI_ARI_CAP, PCI_ARI_CAP_NFN(nextfn));