kvm: Drop CONFIG_KVM_PARA
[qemu/stefanha.git] / hw / pcie.c
blob39607bf31a8ce0ffa8e45eb6ecf424d4025953f9
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 "qemu-common.h"
22 #include "pci_bridge.h"
23 #include "pcie.h"
24 #include "msix.h"
25 #include "msi.h"
26 #include "pci_internals.h"
27 #include "pcie_regs.h"
28 #include "range.h"
30 //#define DEBUG_PCIE
31 #ifdef DEBUG_PCIE
32 # define PCIE_DPRINTF(fmt, ...) \
33 fprintf(stderr, "%s:%d " fmt, __func__, __LINE__, ## __VA_ARGS__)
34 #else
35 # define PCIE_DPRINTF(fmt, ...) do {} while (0)
36 #endif
37 #define PCIE_DEV_PRINTF(dev, fmt, ...) \
38 PCIE_DPRINTF("%s:%x "fmt, (dev)->name, (dev)->devfn, ## __VA_ARGS__)
41 /***************************************************************************
42 * pci express capability helper functions
44 int pcie_cap_init(PCIDevice *dev, uint8_t offset, uint8_t type, uint8_t port)
46 int pos;
47 uint8_t *exp_cap;
49 assert(pci_is_express(dev));
51 pos = pci_add_capability(dev, PCI_CAP_ID_EXP, offset,
52 PCI_EXP_VER2_SIZEOF);
53 if (pos < 0) {
54 return pos;
56 dev->exp.exp_cap = pos;
57 exp_cap = dev->config + pos;
59 /* capability register
60 interrupt message number defaults to 0 */
61 pci_set_word(exp_cap + PCI_EXP_FLAGS,
62 ((type << PCI_EXP_FLAGS_TYPE_SHIFT) & PCI_EXP_FLAGS_TYPE) |
63 PCI_EXP_FLAGS_VER2);
65 /* device capability register
66 * table 7-12:
67 * roll based error reporting bit must be set by all
68 * Functions conforming to the ECN, PCI Express Base
69 * Specification, Revision 1.1., or subsequent PCI Express Base
70 * Specification revisions.
72 pci_set_long(exp_cap + PCI_EXP_DEVCAP, PCI_EXP_DEVCAP_RBER);
74 pci_set_long(exp_cap + PCI_EXP_LNKCAP,
75 (port << PCI_EXP_LNKCAP_PN_SHIFT) |
76 PCI_EXP_LNKCAP_ASPMS_0S |
77 PCI_EXP_LNK_MLW_1 |
78 PCI_EXP_LNK_LS_25);
80 pci_set_word(exp_cap + PCI_EXP_LNKSTA,
81 PCI_EXP_LNK_MLW_1 | PCI_EXP_LNK_LS_25);
83 pci_set_long(exp_cap + PCI_EXP_DEVCAP2,
84 PCI_EXP_DEVCAP2_EFF | PCI_EXP_DEVCAP2_EETLPP);
86 pci_set_word(dev->wmask + pos, PCI_EXP_DEVCTL2_EETLPPB);
87 return pos;
90 void pcie_cap_exit(PCIDevice *dev)
92 pci_del_capability(dev, PCI_CAP_ID_EXP, PCI_EXP_VER2_SIZEOF);
95 uint8_t pcie_cap_get_type(const PCIDevice *dev)
97 uint32_t pos = dev->exp.exp_cap;
98 assert(pos > 0);
99 return (pci_get_word(dev->config + pos + PCI_EXP_FLAGS) &
100 PCI_EXP_FLAGS_TYPE) >> PCI_EXP_FLAGS_TYPE_SHIFT;
103 /* MSI/MSI-X */
104 /* pci express interrupt message number */
105 /* 7.8.2 PCI Express Capabilities Register: Interrupt Message Number */
106 void pcie_cap_flags_set_vector(PCIDevice *dev, uint8_t vector)
108 uint8_t *exp_cap = dev->config + dev->exp.exp_cap;
109 assert(vector < 32);
110 pci_word_test_and_clear_mask(exp_cap + PCI_EXP_FLAGS, PCI_EXP_FLAGS_IRQ);
111 pci_word_test_and_set_mask(exp_cap + PCI_EXP_FLAGS,
112 vector << PCI_EXP_FLAGS_IRQ_SHIFT);
115 uint8_t pcie_cap_flags_get_vector(PCIDevice *dev)
117 return (pci_get_word(dev->config + dev->exp.exp_cap + PCI_EXP_FLAGS) &
118 PCI_EXP_FLAGS_IRQ) >> PCI_EXP_FLAGS_IRQ_SHIFT;
121 void pcie_cap_deverr_init(PCIDevice *dev)
123 uint32_t pos = dev->exp.exp_cap;
124 pci_long_test_and_set_mask(dev->config + pos + PCI_EXP_DEVCAP,
125 PCI_EXP_DEVCAP_RBER);
126 pci_long_test_and_set_mask(dev->wmask + pos + PCI_EXP_DEVCTL,
127 PCI_EXP_DEVCTL_CERE | PCI_EXP_DEVCTL_NFERE |
128 PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE);
129 pci_long_test_and_set_mask(dev->w1cmask + pos + PCI_EXP_DEVSTA,
130 PCI_EXP_DEVSTA_CED | PCI_EXP_DEVSTA_NFED |
131 PCI_EXP_DEVSTA_URD | PCI_EXP_DEVSTA_URD);
134 void pcie_cap_deverr_reset(PCIDevice *dev)
136 uint8_t *devctl = dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL;
137 pci_long_test_and_clear_mask(devctl,
138 PCI_EXP_DEVCTL_CERE | PCI_EXP_DEVCTL_NFERE |
139 PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE);
142 static void hotplug_event_update_event_status(PCIDevice *dev)
144 uint32_t pos = dev->exp.exp_cap;
145 uint8_t *exp_cap = dev->config + pos;
146 uint16_t sltctl = pci_get_word(exp_cap + PCI_EXP_SLTCTL);
147 uint16_t sltsta = pci_get_word(exp_cap + PCI_EXP_SLTSTA);
149 dev->exp.hpev_notified = (sltctl & PCI_EXP_SLTCTL_HPIE) &&
150 (sltsta & sltctl & PCI_EXP_HP_EV_SUPPORTED);
153 static void hotplug_event_notify(PCIDevice *dev)
155 bool prev = dev->exp.hpev_notified;
157 hotplug_event_update_event_status(dev);
159 if (prev == dev->exp.hpev_notified) {
160 return;
163 /* Note: the logic above does not take into account whether interrupts
164 * are masked. The result is that interrupt will be sent when it is
165 * subsequently unmasked. This appears to be legal: Section 6.7.3.4:
166 * The Port may optionally send an MSI when there are hot-plug events that
167 * occur while interrupt generation is disabled, and interrupt generation is
168 * subsequently enabled. */
169 if (msix_enabled(dev)) {
170 msix_notify(dev, pcie_cap_flags_get_vector(dev));
171 } else if (msi_enabled(dev)) {
172 msi_notify(dev, pcie_cap_flags_get_vector(dev));
173 } else {
174 qemu_set_irq(dev->irq[dev->exp.hpev_intx], dev->exp.hpev_notified);
179 * A PCI Express Hot-Plug Event has occurred, so update slot status register
180 * and notify OS of the event if necessary.
182 * 6.7.3 PCI Express Hot-Plug Events
183 * 6.7.3.4 Software Notification of Hot-Plug Events
185 static void pcie_cap_slot_event(PCIDevice *dev, PCIExpressHotPlugEvent event)
187 /* Minor optimization: if nothing changed - no event is needed. */
188 if (pci_word_test_and_set_mask(dev->config + dev->exp.exp_cap +
189 PCI_EXP_SLTSTA, event)) {
190 return;
192 hotplug_event_notify(dev);
195 static int pcie_cap_slot_hotplug(DeviceState *qdev,
196 PCIDevice *pci_dev, PCIHotplugState state)
198 PCIDevice *d = DO_UPCAST(PCIDevice, qdev, qdev);
199 uint8_t *exp_cap = d->config + d->exp.exp_cap;
200 uint16_t sltsta = pci_get_word(exp_cap + PCI_EXP_SLTSTA);
202 /* Don't send event when device is enabled during qemu machine creation:
203 * it is present on boot, no hotplug event is necessary. We do send an
204 * event when the device is disabled later. */
205 if (state == PCI_COLDPLUG_ENABLED) {
206 pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTSTA,
207 PCI_EXP_SLTSTA_PDS);
208 return 0;
211 PCIE_DEV_PRINTF(pci_dev, "hotplug state: %d\n", state);
212 if (sltsta & PCI_EXP_SLTSTA_EIS) {
213 /* the slot is electromechanically locked.
214 * This error is propagated up to qdev and then to HMP/QMP.
216 return -EBUSY;
219 /* TODO: multifunction hot-plug.
220 * Right now, only a device of function = 0 is allowed to be
221 * hot plugged/unplugged.
223 assert(PCI_FUNC(pci_dev->devfn) == 0);
225 if (state == PCI_HOTPLUG_ENABLED) {
226 pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTSTA,
227 PCI_EXP_SLTSTA_PDS);
228 pcie_cap_slot_event(d, PCI_EXP_HP_EV_PDC);
229 } else {
230 qdev_free(&pci_dev->qdev);
231 pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTSTA,
232 PCI_EXP_SLTSTA_PDS);
233 pcie_cap_slot_event(d, PCI_EXP_HP_EV_PDC);
235 return 0;
238 /* pci express slot for pci express root/downstream port
239 PCI express capability slot registers */
240 void pcie_cap_slot_init(PCIDevice *dev, uint16_t slot)
242 uint32_t pos = dev->exp.exp_cap;
244 pci_word_test_and_set_mask(dev->config + pos + PCI_EXP_FLAGS,
245 PCI_EXP_FLAGS_SLOT);
247 pci_long_test_and_clear_mask(dev->config + pos + PCI_EXP_SLTCAP,
248 ~PCI_EXP_SLTCAP_PSN);
249 pci_long_test_and_set_mask(dev->config + pos + PCI_EXP_SLTCAP,
250 (slot << PCI_EXP_SLTCAP_PSN_SHIFT) |
251 PCI_EXP_SLTCAP_EIP |
252 PCI_EXP_SLTCAP_HPS |
253 PCI_EXP_SLTCAP_HPC |
254 PCI_EXP_SLTCAP_PIP |
255 PCI_EXP_SLTCAP_AIP |
256 PCI_EXP_SLTCAP_ABP);
258 pci_word_test_and_clear_mask(dev->config + pos + PCI_EXP_SLTCTL,
259 PCI_EXP_SLTCTL_PIC |
260 PCI_EXP_SLTCTL_AIC);
261 pci_word_test_and_set_mask(dev->config + pos + PCI_EXP_SLTCTL,
262 PCI_EXP_SLTCTL_PIC_OFF |
263 PCI_EXP_SLTCTL_AIC_OFF);
264 pci_word_test_and_set_mask(dev->wmask + pos + PCI_EXP_SLTCTL,
265 PCI_EXP_SLTCTL_PIC |
266 PCI_EXP_SLTCTL_AIC |
267 PCI_EXP_SLTCTL_HPIE |
268 PCI_EXP_SLTCTL_CCIE |
269 PCI_EXP_SLTCTL_PDCE |
270 PCI_EXP_SLTCTL_ABPE);
271 /* Although reading PCI_EXP_SLTCTL_EIC returns always 0,
272 * make the bit writable here in order to detect 1b is written.
273 * pcie_cap_slot_write_config() test-and-clear the bit, so
274 * this bit always returns 0 to the guest.
276 pci_word_test_and_set_mask(dev->wmask + pos + PCI_EXP_SLTCTL,
277 PCI_EXP_SLTCTL_EIC);
279 pci_word_test_and_set_mask(dev->w1cmask + pos + PCI_EXP_SLTSTA,
280 PCI_EXP_HP_EV_SUPPORTED);
282 dev->exp.hpev_notified = false;
284 pci_bus_hotplug(pci_bridge_get_sec_bus(DO_UPCAST(PCIBridge, dev, dev)),
285 pcie_cap_slot_hotplug, &dev->qdev);
288 void pcie_cap_slot_reset(PCIDevice *dev)
290 uint8_t *exp_cap = dev->config + dev->exp.exp_cap;
292 PCIE_DEV_PRINTF(dev, "reset\n");
294 pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTCTL,
295 PCI_EXP_SLTCTL_EIC |
296 PCI_EXP_SLTCTL_PIC |
297 PCI_EXP_SLTCTL_AIC |
298 PCI_EXP_SLTCTL_HPIE |
299 PCI_EXP_SLTCTL_CCIE |
300 PCI_EXP_SLTCTL_PDCE |
301 PCI_EXP_SLTCTL_ABPE);
302 pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTCTL,
303 PCI_EXP_SLTCTL_PIC_OFF |
304 PCI_EXP_SLTCTL_AIC_OFF);
306 pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTSTA,
307 PCI_EXP_SLTSTA_EIS |/* on reset,
308 the lock is released */
309 PCI_EXP_SLTSTA_CC |
310 PCI_EXP_SLTSTA_PDC |
311 PCI_EXP_SLTSTA_ABP);
313 hotplug_event_update_event_status(dev);
316 void pcie_cap_slot_write_config(PCIDevice *dev,
317 uint32_t addr, uint32_t val, int len)
319 uint32_t pos = dev->exp.exp_cap;
320 uint8_t *exp_cap = dev->config + pos;
321 uint16_t sltsta = pci_get_word(exp_cap + PCI_EXP_SLTSTA);
323 if (!ranges_overlap(addr, len, pos + PCI_EXP_SLTCTL, 2)) {
324 return;
327 if (pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTCTL,
328 PCI_EXP_SLTCTL_EIC)) {
329 sltsta ^= PCI_EXP_SLTSTA_EIS; /* toggle PCI_EXP_SLTSTA_EIS bit */
330 pci_set_word(exp_cap + PCI_EXP_SLTSTA, sltsta);
331 PCIE_DEV_PRINTF(dev, "PCI_EXP_SLTCTL_EIC: "
332 "sltsta -> 0x%02"PRIx16"\n",
333 sltsta);
336 hotplug_event_notify(dev);
339 * 6.7.3.2 Command Completed Events
341 * Software issues a command to a hot-plug capable Downstream Port by
342 * issuing a write transaction that targets any portion of the Port’s Slot
343 * Control register. A single write to the Slot Control register is
344 * considered to be a single command, even if the write affects more than
345 * one field in the Slot Control register. In response to this transaction,
346 * the Port must carry out the requested actions and then set the
347 * associated status field for the command completed event. */
349 /* Real hardware might take a while to complete requested command because
350 * physical movement would be involved like locking the electromechanical
351 * lock. However in our case, command is completed instantaneously above,
352 * so send a command completion event right now.
354 pcie_cap_slot_event(dev, PCI_EXP_HP_EV_CCI);
357 int pcie_cap_slot_post_load(void *opaque, int version_id)
359 PCIDevice *dev = opaque;
360 hotplug_event_update_event_status(dev);
361 return 0;
364 void pcie_cap_slot_push_attention_button(PCIDevice *dev)
366 pcie_cap_slot_event(dev, PCI_EXP_HP_EV_ABP);
369 /* root control/capabilities/status. PME isn't emulated for now */
370 void pcie_cap_root_init(PCIDevice *dev)
372 pci_set_word(dev->wmask + dev->exp.exp_cap + PCI_EXP_RTCTL,
373 PCI_EXP_RTCTL_SECEE | PCI_EXP_RTCTL_SENFEE |
374 PCI_EXP_RTCTL_SEFEE);
377 void pcie_cap_root_reset(PCIDevice *dev)
379 pci_set_word(dev->config + dev->exp.exp_cap + PCI_EXP_RTCTL, 0);
382 /* function level reset(FLR) */
383 void pcie_cap_flr_init(PCIDevice *dev)
385 pci_long_test_and_set_mask(dev->config + dev->exp.exp_cap + PCI_EXP_DEVCAP,
386 PCI_EXP_DEVCAP_FLR);
388 /* Although reading BCR_FLR returns always 0,
389 * the bit is made writable here in order to detect the 1b is written
390 * pcie_cap_flr_write_config() test-and-clear the bit, so
391 * this bit always returns 0 to the guest.
393 pci_word_test_and_set_mask(dev->wmask + dev->exp.exp_cap + PCI_EXP_DEVCTL,
394 PCI_EXP_DEVCTL_BCR_FLR);
397 void pcie_cap_flr_write_config(PCIDevice *dev,
398 uint32_t addr, uint32_t val, int len)
400 uint8_t *devctl = dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL;
401 if (pci_get_word(devctl) & PCI_EXP_DEVCTL_BCR_FLR) {
402 /* Clear PCI_EXP_DEVCTL_BCR_FLR after invoking the reset handler
403 so the handler can detect FLR by looking at this bit. */
404 pci_device_reset(dev);
405 pci_word_test_and_clear_mask(devctl, PCI_EXP_DEVCTL_BCR_FLR);
409 /* Alternative Routing-ID Interpretation (ARI) */
410 /* ari forwarding support for down stream port */
411 void pcie_cap_ari_init(PCIDevice *dev)
413 uint32_t pos = dev->exp.exp_cap;
414 pci_long_test_and_set_mask(dev->config + pos + PCI_EXP_DEVCAP2,
415 PCI_EXP_DEVCAP2_ARI);
416 pci_long_test_and_set_mask(dev->wmask + pos + PCI_EXP_DEVCTL2,
417 PCI_EXP_DEVCTL2_ARI);
420 void pcie_cap_ari_reset(PCIDevice *dev)
422 uint8_t *devctl2 = dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL2;
423 pci_long_test_and_clear_mask(devctl2, PCI_EXP_DEVCTL2_ARI);
426 bool pcie_cap_is_ari_enabled(const PCIDevice *dev)
428 if (!pci_is_express(dev)) {
429 return false;
431 if (!dev->exp.exp_cap) {
432 return false;
435 return pci_get_long(dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL2) &
436 PCI_EXP_DEVCTL2_ARI;
439 /**************************************************************************
440 * pci express extended capability allocation functions
441 * uint16_t ext_cap_id (16 bit)
442 * uint8_t cap_ver (4 bit)
443 * uint16_t cap_offset (12 bit)
444 * uint16_t ext_cap_size
447 static uint16_t pcie_find_capability_list(PCIDevice *dev, uint16_t cap_id,
448 uint16_t *prev_p)
450 uint16_t prev = 0;
451 uint16_t next;
452 uint32_t header = pci_get_long(dev->config + PCI_CONFIG_SPACE_SIZE);
454 if (!header) {
455 /* no extended capability */
456 next = 0;
457 goto out;
459 for (next = PCI_CONFIG_SPACE_SIZE; next;
460 prev = next, next = PCI_EXT_CAP_NEXT(header)) {
462 assert(next >= PCI_CONFIG_SPACE_SIZE);
463 assert(next <= PCIE_CONFIG_SPACE_SIZE - 8);
465 header = pci_get_long(dev->config + next);
466 if (PCI_EXT_CAP_ID(header) == cap_id) {
467 break;
471 out:
472 if (prev_p) {
473 *prev_p = prev;
475 return next;
478 uint16_t pcie_find_capability(PCIDevice *dev, uint16_t cap_id)
480 return pcie_find_capability_list(dev, cap_id, NULL);
483 static void pcie_ext_cap_set_next(PCIDevice *dev, uint16_t pos, uint16_t next)
485 uint16_t header = pci_get_long(dev->config + pos);
486 assert(!(next & (PCI_EXT_CAP_ALIGN - 1)));
487 header = (header & ~PCI_EXT_CAP_NEXT_MASK) |
488 ((next << PCI_EXT_CAP_NEXT_SHIFT) & PCI_EXT_CAP_NEXT_MASK);
489 pci_set_long(dev->config + pos, header);
493 * caller must supply valid (offset, size) * such that the range shouldn't
494 * overlap with other capability or other registers.
495 * This function doesn't check it.
497 void pcie_add_capability(PCIDevice *dev,
498 uint16_t cap_id, uint8_t cap_ver,
499 uint16_t offset, uint16_t size)
501 uint32_t header;
502 uint16_t next;
504 assert(offset >= PCI_CONFIG_SPACE_SIZE);
505 assert(offset < offset + size);
506 assert(offset + size < PCIE_CONFIG_SPACE_SIZE);
507 assert(size >= 8);
508 assert(pci_is_express(dev));
510 if (offset == PCI_CONFIG_SPACE_SIZE) {
511 header = pci_get_long(dev->config + offset);
512 next = PCI_EXT_CAP_NEXT(header);
513 } else {
514 uint16_t prev;
516 /* 0 is reserved cap id. use internally to find the last capability
517 in the linked list */
518 next = pcie_find_capability_list(dev, 0, &prev);
520 assert(prev >= PCI_CONFIG_SPACE_SIZE);
521 assert(next == 0);
522 pcie_ext_cap_set_next(dev, prev, offset);
524 pci_set_long(dev->config + offset, PCI_EXT_CAP(cap_id, cap_ver, next));
526 /* Make capability read-only by default */
527 memset(dev->wmask + offset, 0, size);
528 memset(dev->w1cmask + offset, 0, size);
529 /* Check capability by default */
530 memset(dev->cmask + offset, 0xFF, size);
533 /**************************************************************************
534 * pci express extended capability helper functions
537 /* ARI */
538 void pcie_ari_init(PCIDevice *dev, uint16_t offset, uint16_t nextfn)
540 pcie_add_capability(dev, PCI_EXT_CAP_ID_ARI, PCI_ARI_VER,
541 offset, PCI_ARI_SIZEOF);
542 pci_set_long(dev->config + offset + PCI_ARI_CAP, PCI_ARI_CAP_NFN(nextfn));