xref: /openbmc/qemu/hw/pci/pcie.c (revision fe7f9b8e)
1 /*
2  * pcie.c
3  *
4  * Copyright (c) 2010 Isaku Yamahata <yamahata at valinux co jp>
5  *                    VA Linux Systems Japan K.K.
6  *
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.
11  *
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.
16  *
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/>.
19  */
20 
21 #include "qemu/osdep.h"
22 #include "qapi/error.h"
23 #include "qemu-common.h"
24 #include "hw/pci/pci_bridge.h"
25 #include "hw/pci/pcie.h"
26 #include "hw/pci/msix.h"
27 #include "hw/pci/msi.h"
28 #include "hw/pci/pci_bus.h"
29 #include "hw/pci/pcie_regs.h"
30 #include "qemu/range.h"
31 
32 //#define DEBUG_PCIE
33 #ifdef DEBUG_PCIE
34 # define PCIE_DPRINTF(fmt, ...)                                         \
35     fprintf(stderr, "%s:%d " fmt, __func__, __LINE__, ## __VA_ARGS__)
36 #else
37 # define PCIE_DPRINTF(fmt, ...) do {} while (0)
38 #endif
39 #define PCIE_DEV_PRINTF(dev, fmt, ...)                                  \
40     PCIE_DPRINTF("%s:%x "fmt, (dev)->name, (dev)->devfn, ## __VA_ARGS__)
41 
42 
43 /***************************************************************************
44  * pci express capability helper functions
45  */
46 
47 static void
48 pcie_cap_v1_fill(PCIDevice *dev, uint8_t port, uint8_t type, uint8_t version)
49 {
50     uint8_t *exp_cap = dev->config + dev->exp.exp_cap;
51     uint8_t *cmask = dev->cmask + dev->exp.exp_cap;
52 
53     /* capability register
54     interrupt message number defaults to 0 */
55     pci_set_word(exp_cap + PCI_EXP_FLAGS,
56                  ((type << PCI_EXP_FLAGS_TYPE_SHIFT) & PCI_EXP_FLAGS_TYPE) |
57                  version);
58 
59     /* device capability register
60      * table 7-12:
61      * roll based error reporting bit must be set by all
62      * Functions conforming to the ECN, PCI Express Base
63      * Specification, Revision 1.1., or subsequent PCI Express Base
64      * Specification revisions.
65      */
66     pci_set_long(exp_cap + PCI_EXP_DEVCAP, PCI_EXP_DEVCAP_RBER);
67 
68     pci_set_long(exp_cap + PCI_EXP_LNKCAP,
69                  (port << PCI_EXP_LNKCAP_PN_SHIFT) |
70                  PCI_EXP_LNKCAP_ASPMS_0S |
71                  PCI_EXP_LNK_MLW_1 |
72                  PCI_EXP_LNK_LS_25);
73 
74     pci_set_word(exp_cap + PCI_EXP_LNKSTA,
75                  PCI_EXP_LNK_MLW_1 | PCI_EXP_LNK_LS_25);
76 
77     if (dev->cap_present & QEMU_PCIE_LNKSTA_DLLLA) {
78         pci_word_test_and_set_mask(exp_cap + PCI_EXP_LNKSTA,
79                                    PCI_EXP_LNKSTA_DLLLA);
80     }
81 
82     /* We changed link status bits over time, and changing them across
83      * migrations is generally fine as hardware changes them too.
84      * Let's not bother checking.
85      */
86     pci_set_word(cmask + PCI_EXP_LNKSTA, 0);
87 }
88 
89 int pcie_cap_init(PCIDevice *dev, uint8_t offset,
90                   uint8_t type, uint8_t port,
91                   Error **errp)
92 {
93     /* PCIe cap v2 init */
94     int pos;
95     uint8_t *exp_cap;
96 
97     assert(pci_is_express(dev));
98 
99     pos = pci_add_capability(dev, PCI_CAP_ID_EXP, offset,
100                              PCI_EXP_VER2_SIZEOF, errp);
101     if (pos < 0) {
102         return pos;
103     }
104     dev->exp.exp_cap = pos;
105     exp_cap = dev->config + pos;
106 
107     /* Filling values common with v1 */
108     pcie_cap_v1_fill(dev, port, type, PCI_EXP_FLAGS_VER2);
109 
110     /* Filling v2 specific values */
111     pci_set_long(exp_cap + PCI_EXP_DEVCAP2,
112                  PCI_EXP_DEVCAP2_EFF | PCI_EXP_DEVCAP2_EETLPP);
113 
114     pci_set_word(dev->wmask + pos + PCI_EXP_DEVCTL2, PCI_EXP_DEVCTL2_EETLPPB);
115 
116     if (dev->cap_present & QEMU_PCIE_EXTCAP_INIT) {
117         /* read-only to behave like a 'NULL' Extended Capability Header */
118         pci_set_long(dev->wmask + PCI_CONFIG_SPACE_SIZE, 0);
119     }
120 
121     return pos;
122 }
123 
124 int pcie_cap_v1_init(PCIDevice *dev, uint8_t offset, uint8_t type,
125                      uint8_t port)
126 {
127     /* PCIe cap v1 init */
128     int pos;
129     Error *local_err = NULL;
130 
131     assert(pci_is_express(dev));
132 
133     pos = pci_add_capability(dev, PCI_CAP_ID_EXP, offset,
134                              PCI_EXP_VER1_SIZEOF, &local_err);
135     if (pos < 0) {
136         error_report_err(local_err);
137         return pos;
138     }
139     dev->exp.exp_cap = pos;
140 
141     pcie_cap_v1_fill(dev, port, type, PCI_EXP_FLAGS_VER1);
142 
143     return pos;
144 }
145 
146 static int
147 pcie_endpoint_cap_common_init(PCIDevice *dev, uint8_t offset, uint8_t cap_size)
148 {
149     uint8_t type = PCI_EXP_TYPE_ENDPOINT;
150     Error *local_err = NULL;
151     int ret;
152 
153     /*
154      * Windows guests will report Code 10, device cannot start, if
155      * a regular Endpoint type is exposed on a root complex.  These
156      * should instead be Root Complex Integrated Endpoints.
157      */
158     if (pci_bus_is_express(pci_get_bus(dev))
159         && pci_bus_is_root(pci_get_bus(dev))) {
160         type = PCI_EXP_TYPE_RC_END;
161     }
162 
163     if (cap_size == PCI_EXP_VER1_SIZEOF) {
164         return pcie_cap_v1_init(dev, offset, type, 0);
165     } else {
166         ret = pcie_cap_init(dev, offset, type, 0, &local_err);
167 
168         if (ret < 0) {
169             error_report_err(local_err);
170         }
171 
172         return ret;
173     }
174 }
175 
176 int pcie_endpoint_cap_init(PCIDevice *dev, uint8_t offset)
177 {
178     return pcie_endpoint_cap_common_init(dev, offset, PCI_EXP_VER2_SIZEOF);
179 }
180 
181 int pcie_endpoint_cap_v1_init(PCIDevice *dev, uint8_t offset)
182 {
183     return pcie_endpoint_cap_common_init(dev, offset, PCI_EXP_VER1_SIZEOF);
184 }
185 
186 void pcie_cap_exit(PCIDevice *dev)
187 {
188     pci_del_capability(dev, PCI_CAP_ID_EXP, PCI_EXP_VER2_SIZEOF);
189 }
190 
191 void pcie_cap_v1_exit(PCIDevice *dev)
192 {
193     pci_del_capability(dev, PCI_CAP_ID_EXP, PCI_EXP_VER1_SIZEOF);
194 }
195 
196 uint8_t pcie_cap_get_type(const PCIDevice *dev)
197 {
198     uint32_t pos = dev->exp.exp_cap;
199     assert(pos > 0);
200     return (pci_get_word(dev->config + pos + PCI_EXP_FLAGS) &
201             PCI_EXP_FLAGS_TYPE) >> PCI_EXP_FLAGS_TYPE_SHIFT;
202 }
203 
204 /* MSI/MSI-X */
205 /* pci express interrupt message number */
206 /* 7.8.2 PCI Express Capabilities Register: Interrupt Message Number */
207 void pcie_cap_flags_set_vector(PCIDevice *dev, uint8_t vector)
208 {
209     uint8_t *exp_cap = dev->config + dev->exp.exp_cap;
210     assert(vector < 32);
211     pci_word_test_and_clear_mask(exp_cap + PCI_EXP_FLAGS, PCI_EXP_FLAGS_IRQ);
212     pci_word_test_and_set_mask(exp_cap + PCI_EXP_FLAGS,
213                                vector << PCI_EXP_FLAGS_IRQ_SHIFT);
214 }
215 
216 uint8_t pcie_cap_flags_get_vector(PCIDevice *dev)
217 {
218     return (pci_get_word(dev->config + dev->exp.exp_cap + PCI_EXP_FLAGS) &
219             PCI_EXP_FLAGS_IRQ) >> PCI_EXP_FLAGS_IRQ_SHIFT;
220 }
221 
222 void pcie_cap_deverr_init(PCIDevice *dev)
223 {
224     uint32_t pos = dev->exp.exp_cap;
225     pci_long_test_and_set_mask(dev->config + pos + PCI_EXP_DEVCAP,
226                                PCI_EXP_DEVCAP_RBER);
227     pci_long_test_and_set_mask(dev->wmask + pos + PCI_EXP_DEVCTL,
228                                PCI_EXP_DEVCTL_CERE | PCI_EXP_DEVCTL_NFERE |
229                                PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE);
230     pci_long_test_and_set_mask(dev->w1cmask + pos + PCI_EXP_DEVSTA,
231                                PCI_EXP_DEVSTA_CED | PCI_EXP_DEVSTA_NFED |
232                                PCI_EXP_DEVSTA_FED | PCI_EXP_DEVSTA_URD);
233 }
234 
235 void pcie_cap_deverr_reset(PCIDevice *dev)
236 {
237     uint8_t *devctl = dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL;
238     pci_long_test_and_clear_mask(devctl,
239                                  PCI_EXP_DEVCTL_CERE | PCI_EXP_DEVCTL_NFERE |
240                                  PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE);
241 }
242 
243 void pcie_cap_lnkctl_init(PCIDevice *dev)
244 {
245     uint32_t pos = dev->exp.exp_cap;
246     pci_long_test_and_set_mask(dev->wmask + pos + PCI_EXP_LNKCTL,
247                                PCI_EXP_LNKCTL_CCC | PCI_EXP_LNKCTL_ES);
248 }
249 
250 void pcie_cap_lnkctl_reset(PCIDevice *dev)
251 {
252     uint8_t *lnkctl = dev->config + dev->exp.exp_cap + PCI_EXP_LNKCTL;
253     pci_long_test_and_clear_mask(lnkctl,
254                                  PCI_EXP_LNKCTL_CCC | PCI_EXP_LNKCTL_ES);
255 }
256 
257 static void hotplug_event_update_event_status(PCIDevice *dev)
258 {
259     uint32_t pos = dev->exp.exp_cap;
260     uint8_t *exp_cap = dev->config + pos;
261     uint16_t sltctl = pci_get_word(exp_cap + PCI_EXP_SLTCTL);
262     uint16_t sltsta = pci_get_word(exp_cap + PCI_EXP_SLTSTA);
263 
264     dev->exp.hpev_notified = (sltctl & PCI_EXP_SLTCTL_HPIE) &&
265         (sltsta & sltctl & PCI_EXP_HP_EV_SUPPORTED);
266 }
267 
268 static void hotplug_event_notify(PCIDevice *dev)
269 {
270     bool prev = dev->exp.hpev_notified;
271 
272     hotplug_event_update_event_status(dev);
273 
274     if (prev == dev->exp.hpev_notified) {
275         return;
276     }
277 
278     /* Note: the logic above does not take into account whether interrupts
279      * are masked. The result is that interrupt will be sent when it is
280      * subsequently unmasked. This appears to be legal: Section 6.7.3.4:
281      * The Port may optionally send an MSI when there are hot-plug events that
282      * occur while interrupt generation is disabled, and interrupt generation is
283      * subsequently enabled. */
284     if (msix_enabled(dev)) {
285         msix_notify(dev, pcie_cap_flags_get_vector(dev));
286     } else if (msi_enabled(dev)) {
287         msi_notify(dev, pcie_cap_flags_get_vector(dev));
288     } else {
289         pci_set_irq(dev, dev->exp.hpev_notified);
290     }
291 }
292 
293 static void hotplug_event_clear(PCIDevice *dev)
294 {
295     hotplug_event_update_event_status(dev);
296     if (!msix_enabled(dev) && !msi_enabled(dev) && !dev->exp.hpev_notified) {
297         pci_irq_deassert(dev);
298     }
299 }
300 
301 /*
302  * A PCI Express Hot-Plug Event has occurred, so update slot status register
303  * and notify OS of the event if necessary.
304  *
305  * 6.7.3 PCI Express Hot-Plug Events
306  * 6.7.3.4 Software Notification of Hot-Plug Events
307  */
308 static void pcie_cap_slot_event(PCIDevice *dev, PCIExpressHotPlugEvent event)
309 {
310     /* Minor optimization: if nothing changed - no event is needed. */
311     if (pci_word_test_and_set_mask(dev->config + dev->exp.exp_cap +
312                                    PCI_EXP_SLTSTA, event)) {
313         return;
314     }
315     hotplug_event_notify(dev);
316 }
317 
318 static void pcie_cap_slot_hotplug_common(PCIDevice *hotplug_dev,
319                                          DeviceState *dev,
320                                          uint8_t **exp_cap, Error **errp)
321 {
322     *exp_cap = hotplug_dev->config + hotplug_dev->exp.exp_cap;
323     uint16_t sltsta = pci_get_word(*exp_cap + PCI_EXP_SLTSTA);
324 
325     PCIE_DEV_PRINTF(PCI_DEVICE(dev), "hotplug state: 0x%x\n", sltsta);
326     if (sltsta & PCI_EXP_SLTSTA_EIS) {
327         /* the slot is electromechanically locked.
328          * This error is propagated up to qdev and then to HMP/QMP.
329          */
330         error_setg_errno(errp, EBUSY, "slot is electromechanically locked");
331     }
332 }
333 
334 void pcie_cap_slot_hotplug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
335                               Error **errp)
336 {
337     uint8_t *exp_cap;
338     PCIDevice *pci_dev = PCI_DEVICE(dev);
339 
340     pcie_cap_slot_hotplug_common(PCI_DEVICE(hotplug_dev), dev, &exp_cap, errp);
341 
342     /* Don't send event when device is enabled during qemu machine creation:
343      * it is present on boot, no hotplug event is necessary. We do send an
344      * event when the device is disabled later. */
345     if (!dev->hotplugged) {
346         pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTSTA,
347                                    PCI_EXP_SLTSTA_PDS);
348         return;
349     }
350 
351     /* To enable multifunction hot-plug, we just ensure the function
352      * 0 added last. When function 0 is added, we set the sltsta and
353      * inform OS via event notification.
354      */
355     if (pci_get_function_0(pci_dev)) {
356         pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTSTA,
357                                    PCI_EXP_SLTSTA_PDS);
358         pcie_cap_slot_event(PCI_DEVICE(hotplug_dev),
359                             PCI_EXP_HP_EV_PDC | PCI_EXP_HP_EV_ABP);
360     }
361 }
362 
363 static void pcie_unplug_device(PCIBus *bus, PCIDevice *dev, void *opaque)
364 {
365     object_unparent(OBJECT(dev));
366 }
367 
368 void pcie_cap_slot_hot_unplug_request_cb(HotplugHandler *hotplug_dev,
369                                          DeviceState *dev, Error **errp)
370 {
371     uint8_t *exp_cap;
372     PCIDevice *pci_dev = PCI_DEVICE(dev);
373     PCIBus *bus = pci_get_bus(pci_dev);
374 
375     pcie_cap_slot_hotplug_common(PCI_DEVICE(hotplug_dev), dev, &exp_cap, errp);
376 
377     /* In case user cancel the operation of multi-function hot-add,
378      * remove the function that is unexposed to guest individually,
379      * without interaction with guest.
380      */
381     if (pci_dev->devfn &&
382         !bus->devices[0]) {
383         pcie_unplug_device(bus, pci_dev, NULL);
384 
385         return;
386     }
387 
388     pcie_cap_slot_push_attention_button(PCI_DEVICE(hotplug_dev));
389 }
390 
391 /* pci express slot for pci express root/downstream port
392    PCI express capability slot registers */
393 void pcie_cap_slot_init(PCIDevice *dev, uint16_t slot)
394 {
395     uint32_t pos = dev->exp.exp_cap;
396 
397     pci_word_test_and_set_mask(dev->config + pos + PCI_EXP_FLAGS,
398                                PCI_EXP_FLAGS_SLOT);
399 
400     pci_long_test_and_clear_mask(dev->config + pos + PCI_EXP_SLTCAP,
401                                  ~PCI_EXP_SLTCAP_PSN);
402     pci_long_test_and_set_mask(dev->config + pos + PCI_EXP_SLTCAP,
403                                (slot << PCI_EXP_SLTCAP_PSN_SHIFT) |
404                                PCI_EXP_SLTCAP_EIP |
405                                PCI_EXP_SLTCAP_HPS |
406                                PCI_EXP_SLTCAP_HPC |
407                                PCI_EXP_SLTCAP_PIP |
408                                PCI_EXP_SLTCAP_AIP |
409                                PCI_EXP_SLTCAP_ABP);
410 
411     if (dev->cap_present & QEMU_PCIE_SLTCAP_PCP) {
412         pci_long_test_and_set_mask(dev->config + pos + PCI_EXP_SLTCAP,
413                                    PCI_EXP_SLTCAP_PCP);
414         pci_word_test_and_clear_mask(dev->config + pos + PCI_EXP_SLTCTL,
415                                      PCI_EXP_SLTCTL_PCC);
416         pci_word_test_and_set_mask(dev->wmask + pos + PCI_EXP_SLTCTL,
417                                    PCI_EXP_SLTCTL_PCC);
418     }
419 
420     pci_word_test_and_clear_mask(dev->config + pos + PCI_EXP_SLTCTL,
421                                  PCI_EXP_SLTCTL_PIC |
422                                  PCI_EXP_SLTCTL_AIC);
423     pci_word_test_and_set_mask(dev->config + pos + PCI_EXP_SLTCTL,
424                                PCI_EXP_SLTCTL_PIC_OFF |
425                                PCI_EXP_SLTCTL_AIC_OFF);
426     pci_word_test_and_set_mask(dev->wmask + pos + PCI_EXP_SLTCTL,
427                                PCI_EXP_SLTCTL_PIC |
428                                PCI_EXP_SLTCTL_AIC |
429                                PCI_EXP_SLTCTL_HPIE |
430                                PCI_EXP_SLTCTL_CCIE |
431                                PCI_EXP_SLTCTL_PDCE |
432                                PCI_EXP_SLTCTL_ABPE);
433     /* Although reading PCI_EXP_SLTCTL_EIC returns always 0,
434      * make the bit writable here in order to detect 1b is written.
435      * pcie_cap_slot_write_config() test-and-clear the bit, so
436      * this bit always returns 0 to the guest.
437      */
438     pci_word_test_and_set_mask(dev->wmask + pos + PCI_EXP_SLTCTL,
439                                PCI_EXP_SLTCTL_EIC);
440 
441     pci_word_test_and_set_mask(dev->w1cmask + pos + PCI_EXP_SLTSTA,
442                                PCI_EXP_HP_EV_SUPPORTED);
443 
444     dev->exp.hpev_notified = false;
445 
446     qbus_set_hotplug_handler(BUS(pci_bridge_get_sec_bus(PCI_BRIDGE(dev))),
447                              DEVICE(dev), NULL);
448 }
449 
450 void pcie_cap_slot_reset(PCIDevice *dev)
451 {
452     uint8_t *exp_cap = dev->config + dev->exp.exp_cap;
453     uint8_t port_type = pcie_cap_get_type(dev);
454 
455     assert(port_type == PCI_EXP_TYPE_DOWNSTREAM ||
456            port_type == PCI_EXP_TYPE_ROOT_PORT);
457 
458     PCIE_DEV_PRINTF(dev, "reset\n");
459 
460     pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTCTL,
461                                  PCI_EXP_SLTCTL_EIC |
462                                  PCI_EXP_SLTCTL_PIC |
463                                  PCI_EXP_SLTCTL_AIC |
464                                  PCI_EXP_SLTCTL_HPIE |
465                                  PCI_EXP_SLTCTL_CCIE |
466                                  PCI_EXP_SLTCTL_PDCE |
467                                  PCI_EXP_SLTCTL_ABPE);
468     pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTCTL,
469                                PCI_EXP_SLTCTL_AIC_OFF);
470 
471     if (dev->cap_present & QEMU_PCIE_SLTCAP_PCP) {
472         /* Downstream ports enforce device number 0. */
473         bool populated = pci_bridge_get_sec_bus(PCI_BRIDGE(dev))->devices[0];
474         uint16_t pic;
475 
476         if (populated) {
477             pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTCTL,
478                                          PCI_EXP_SLTCTL_PCC);
479         } else {
480             pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTCTL,
481                                        PCI_EXP_SLTCTL_PCC);
482         }
483 
484         pic = populated ? PCI_EXP_SLTCTL_PIC_ON : PCI_EXP_SLTCTL_PIC_OFF;
485         pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTCTL, pic);
486     }
487 
488     pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTSTA,
489                                  PCI_EXP_SLTSTA_EIS |/* on reset,
490                                                         the lock is released */
491                                  PCI_EXP_SLTSTA_CC |
492                                  PCI_EXP_SLTSTA_PDC |
493                                  PCI_EXP_SLTSTA_ABP);
494 
495     hotplug_event_update_event_status(dev);
496 }
497 
498 void pcie_cap_slot_write_config(PCIDevice *dev,
499                                 uint32_t addr, uint32_t val, int len)
500 {
501     uint32_t pos = dev->exp.exp_cap;
502     uint8_t *exp_cap = dev->config + pos;
503     uint16_t sltsta = pci_get_word(exp_cap + PCI_EXP_SLTSTA);
504 
505     if (ranges_overlap(addr, len, pos + PCI_EXP_SLTSTA, 2)) {
506         hotplug_event_clear(dev);
507     }
508 
509     if (!ranges_overlap(addr, len, pos + PCI_EXP_SLTCTL, 2)) {
510         return;
511     }
512 
513     if (pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTCTL,
514                                      PCI_EXP_SLTCTL_EIC)) {
515         sltsta ^= PCI_EXP_SLTSTA_EIS; /* toggle PCI_EXP_SLTSTA_EIS bit */
516         pci_set_word(exp_cap + PCI_EXP_SLTSTA, sltsta);
517         PCIE_DEV_PRINTF(dev, "PCI_EXP_SLTCTL_EIC: "
518                         "sltsta -> 0x%02"PRIx16"\n",
519                         sltsta);
520     }
521 
522     /*
523      * If the slot is polulated, power indicator is off and power
524      * controller is off, it is safe to detach the devices.
525      */
526     if ((sltsta & PCI_EXP_SLTSTA_PDS) && (val & PCI_EXP_SLTCTL_PCC) &&
527         ((val & PCI_EXP_SLTCTL_PIC_OFF) == PCI_EXP_SLTCTL_PIC_OFF)) {
528         PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(dev));
529         pci_for_each_device(sec_bus, pci_bus_num(sec_bus),
530                             pcie_unplug_device, NULL);
531 
532         pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTSTA,
533                                      PCI_EXP_SLTSTA_PDS);
534         pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTSTA,
535                                        PCI_EXP_SLTSTA_PDC);
536     }
537 
538     hotplug_event_notify(dev);
539 
540     /*
541      * 6.7.3.2 Command Completed Events
542      *
543      * Software issues a command to a hot-plug capable Downstream Port by
544      * issuing a write transaction that targets any portion of the Port’s Slot
545      * Control register. A single write to the Slot Control register is
546      * considered to be a single command, even if the write affects more than
547      * one field in the Slot Control register. In response to this transaction,
548      * the Port must carry out the requested actions and then set the
549      * associated status field for the command completed event. */
550 
551     /* Real hardware might take a while to complete requested command because
552      * physical movement would be involved like locking the electromechanical
553      * lock.  However in our case, command is completed instantaneously above,
554      * so send a command completion event right now.
555      */
556     pcie_cap_slot_event(dev, PCI_EXP_HP_EV_CCI);
557 }
558 
559 int pcie_cap_slot_post_load(void *opaque, int version_id)
560 {
561     PCIDevice *dev = opaque;
562     hotplug_event_update_event_status(dev);
563     return 0;
564 }
565 
566 void pcie_cap_slot_push_attention_button(PCIDevice *dev)
567 {
568     pcie_cap_slot_event(dev, PCI_EXP_HP_EV_ABP);
569 }
570 
571 /* root control/capabilities/status. PME isn't emulated for now */
572 void pcie_cap_root_init(PCIDevice *dev)
573 {
574     pci_set_word(dev->wmask + dev->exp.exp_cap + PCI_EXP_RTCTL,
575                  PCI_EXP_RTCTL_SECEE | PCI_EXP_RTCTL_SENFEE |
576                  PCI_EXP_RTCTL_SEFEE);
577 }
578 
579 void pcie_cap_root_reset(PCIDevice *dev)
580 {
581     pci_set_word(dev->config + dev->exp.exp_cap + PCI_EXP_RTCTL, 0);
582 }
583 
584 /* function level reset(FLR) */
585 void pcie_cap_flr_init(PCIDevice *dev)
586 {
587     pci_long_test_and_set_mask(dev->config + dev->exp.exp_cap + PCI_EXP_DEVCAP,
588                                PCI_EXP_DEVCAP_FLR);
589 
590     /* Although reading BCR_FLR returns always 0,
591      * the bit is made writable here in order to detect the 1b is written
592      * pcie_cap_flr_write_config() test-and-clear the bit, so
593      * this bit always returns 0 to the guest.
594      */
595     pci_word_test_and_set_mask(dev->wmask + dev->exp.exp_cap + PCI_EXP_DEVCTL,
596                                PCI_EXP_DEVCTL_BCR_FLR);
597 }
598 
599 void pcie_cap_flr_write_config(PCIDevice *dev,
600                                uint32_t addr, uint32_t val, int len)
601 {
602     uint8_t *devctl = dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL;
603     if (pci_get_word(devctl) & PCI_EXP_DEVCTL_BCR_FLR) {
604         /* Clear PCI_EXP_DEVCTL_BCR_FLR after invoking the reset handler
605            so the handler can detect FLR by looking at this bit. */
606         pci_device_reset(dev);
607         pci_word_test_and_clear_mask(devctl, PCI_EXP_DEVCTL_BCR_FLR);
608     }
609 }
610 
611 /* Alternative Routing-ID Interpretation (ARI)
612  * forwarding support for root and downstream ports
613  */
614 void pcie_cap_arifwd_init(PCIDevice *dev)
615 {
616     uint32_t pos = dev->exp.exp_cap;
617     pci_long_test_and_set_mask(dev->config + pos + PCI_EXP_DEVCAP2,
618                                PCI_EXP_DEVCAP2_ARI);
619     pci_long_test_and_set_mask(dev->wmask + pos + PCI_EXP_DEVCTL2,
620                                PCI_EXP_DEVCTL2_ARI);
621 }
622 
623 void pcie_cap_arifwd_reset(PCIDevice *dev)
624 {
625     uint8_t *devctl2 = dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL2;
626     pci_long_test_and_clear_mask(devctl2, PCI_EXP_DEVCTL2_ARI);
627 }
628 
629 bool pcie_cap_is_arifwd_enabled(const PCIDevice *dev)
630 {
631     if (!pci_is_express(dev)) {
632         return false;
633     }
634     if (!dev->exp.exp_cap) {
635         return false;
636     }
637 
638     return pci_get_long(dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL2) &
639         PCI_EXP_DEVCTL2_ARI;
640 }
641 
642 /**************************************************************************
643  * pci express extended capability list management functions
644  * uint16_t ext_cap_id (16 bit)
645  * uint8_t cap_ver (4 bit)
646  * uint16_t cap_offset (12 bit)
647  * uint16_t ext_cap_size
648  */
649 
650 /* Passing a cap_id value > 0xffff will return 0 and put end of list in prev */
651 static uint16_t pcie_find_capability_list(PCIDevice *dev, uint32_t cap_id,
652                                           uint16_t *prev_p)
653 {
654     uint16_t prev = 0;
655     uint16_t next;
656     uint32_t header = pci_get_long(dev->config + PCI_CONFIG_SPACE_SIZE);
657 
658     if (!header) {
659         /* no extended capability */
660         next = 0;
661         goto out;
662     }
663     for (next = PCI_CONFIG_SPACE_SIZE; next;
664          prev = next, next = PCI_EXT_CAP_NEXT(header)) {
665 
666         assert(next >= PCI_CONFIG_SPACE_SIZE);
667         assert(next <= PCIE_CONFIG_SPACE_SIZE - 8);
668 
669         header = pci_get_long(dev->config + next);
670         if (PCI_EXT_CAP_ID(header) == cap_id) {
671             break;
672         }
673     }
674 
675 out:
676     if (prev_p) {
677         *prev_p = prev;
678     }
679     return next;
680 }
681 
682 uint16_t pcie_find_capability(PCIDevice *dev, uint16_t cap_id)
683 {
684     return pcie_find_capability_list(dev, cap_id, NULL);
685 }
686 
687 static void pcie_ext_cap_set_next(PCIDevice *dev, uint16_t pos, uint16_t next)
688 {
689     uint32_t header = pci_get_long(dev->config + pos);
690     assert(!(next & (PCI_EXT_CAP_ALIGN - 1)));
691     header = (header & ~PCI_EXT_CAP_NEXT_MASK) |
692         ((next << PCI_EXT_CAP_NEXT_SHIFT) & PCI_EXT_CAP_NEXT_MASK);
693     pci_set_long(dev->config + pos, header);
694 }
695 
696 /*
697  * Caller must supply valid (offset, size) such that the range wouldn't
698  * overlap with other capability or other registers.
699  * This function doesn't check it.
700  */
701 void pcie_add_capability(PCIDevice *dev,
702                          uint16_t cap_id, uint8_t cap_ver,
703                          uint16_t offset, uint16_t size)
704 {
705     assert(offset >= PCI_CONFIG_SPACE_SIZE);
706     assert(offset < offset + size);
707     assert(offset + size <= PCIE_CONFIG_SPACE_SIZE);
708     assert(size >= 8);
709     assert(pci_is_express(dev));
710 
711     if (offset != PCI_CONFIG_SPACE_SIZE) {
712         uint16_t prev;
713 
714         /*
715          * 0xffffffff is not a valid cap id (it's a 16 bit field). use
716          * internally to find the last capability in the linked list.
717          */
718         pcie_find_capability_list(dev, 0xffffffff, &prev);
719         assert(prev >= PCI_CONFIG_SPACE_SIZE);
720         pcie_ext_cap_set_next(dev, prev, offset);
721     }
722     pci_set_long(dev->config + offset, PCI_EXT_CAP(cap_id, cap_ver, 0));
723 
724     /* Make capability read-only by default */
725     memset(dev->wmask + offset, 0, size);
726     memset(dev->w1cmask + offset, 0, size);
727     /* Check capability by default */
728     memset(dev->cmask + offset, 0xFF, size);
729 }
730 
731 /**************************************************************************
732  * pci express extended capability helper functions
733  */
734 
735 /* ARI */
736 void pcie_ari_init(PCIDevice *dev, uint16_t offset, uint16_t nextfn)
737 {
738     pcie_add_capability(dev, PCI_EXT_CAP_ID_ARI, PCI_ARI_VER,
739                         offset, PCI_ARI_SIZEOF);
740     pci_set_long(dev->config + offset + PCI_ARI_CAP, (nextfn & 0xff) << 8);
741 }
742 
743 void pcie_dev_ser_num_init(PCIDevice *dev, uint16_t offset, uint64_t ser_num)
744 {
745     static const int pci_dsn_ver = 1;
746     static const int pci_dsn_cap = 4;
747 
748     pcie_add_capability(dev, PCI_EXT_CAP_ID_DSN, pci_dsn_ver, offset,
749                         PCI_EXT_CAP_DSN_SIZEOF);
750     pci_set_quad(dev->config + offset + pci_dsn_cap, ser_num);
751 }
752 
753 void pcie_ats_init(PCIDevice *dev, uint16_t offset)
754 {
755     pcie_add_capability(dev, PCI_EXT_CAP_ID_ATS, 0x1,
756                         offset, PCI_EXT_CAP_ATS_SIZEOF);
757 
758     dev->exp.ats_cap = offset;
759 
760     /* Invalidate Queue Depth 0, Page Aligned Request 0 */
761     pci_set_word(dev->config + offset + PCI_ATS_CAP, 0);
762     /* STU 0, Disabled by default */
763     pci_set_word(dev->config + offset + PCI_ATS_CTRL, 0);
764 
765     pci_set_word(dev->wmask + dev->exp.ats_cap + PCI_ATS_CTRL, 0x800f);
766 }
767