xref: /openbmc/qemu/hw/pci/pcie.c (revision 869e9aec)
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, uint8_t type, uint8_t port)
90 {
91     /* PCIe cap v2 init */
92     int pos;
93     uint8_t *exp_cap;
94 
95     assert(pci_is_express(dev));
96 
97     pos = pci_add_capability(dev, PCI_CAP_ID_EXP, offset, PCI_EXP_VER2_SIZEOF);
98     if (pos < 0) {
99         return pos;
100     }
101     dev->exp.exp_cap = pos;
102     exp_cap = dev->config + pos;
103 
104     /* Filling values common with v1 */
105     pcie_cap_v1_fill(dev, port, type, PCI_EXP_FLAGS_VER2);
106 
107     /* Filling v2 specific values */
108     pci_set_long(exp_cap + PCI_EXP_DEVCAP2,
109                  PCI_EXP_DEVCAP2_EFF | PCI_EXP_DEVCAP2_EETLPP);
110 
111     pci_set_word(dev->wmask + pos + PCI_EXP_DEVCTL2, PCI_EXP_DEVCTL2_EETLPPB);
112     return pos;
113 }
114 
115 int pcie_cap_v1_init(PCIDevice *dev, uint8_t offset, uint8_t type,
116                      uint8_t port)
117 {
118     /* PCIe cap v1 init */
119     int pos;
120 
121     assert(pci_is_express(dev));
122 
123     pos = pci_add_capability(dev, PCI_CAP_ID_EXP, offset, PCI_EXP_VER1_SIZEOF);
124     if (pos < 0) {
125         return pos;
126     }
127     dev->exp.exp_cap = pos;
128 
129     pcie_cap_v1_fill(dev, port, type, PCI_EXP_FLAGS_VER1);
130 
131     return pos;
132 }
133 
134 static int
135 pcie_endpoint_cap_common_init(PCIDevice *dev, uint8_t offset, uint8_t cap_size)
136 {
137     uint8_t type = PCI_EXP_TYPE_ENDPOINT;
138 
139     /*
140      * Windows guests will report Code 10, device cannot start, if
141      * a regular Endpoint type is exposed on a root complex.  These
142      * should instead be Root Complex Integrated Endpoints.
143      */
144     if (pci_bus_is_express(dev->bus) && pci_bus_is_root(dev->bus)) {
145         type = PCI_EXP_TYPE_RC_END;
146     }
147 
148     return (cap_size == PCI_EXP_VER1_SIZEOF)
149         ? pcie_cap_v1_init(dev, offset, type, 0)
150         : pcie_cap_init(dev, offset, type, 0);
151 }
152 
153 int pcie_endpoint_cap_init(PCIDevice *dev, uint8_t offset)
154 {
155     return pcie_endpoint_cap_common_init(dev, offset, PCI_EXP_VER2_SIZEOF);
156 }
157 
158 int pcie_endpoint_cap_v1_init(PCIDevice *dev, uint8_t offset)
159 {
160     return pcie_endpoint_cap_common_init(dev, offset, PCI_EXP_VER1_SIZEOF);
161 }
162 
163 void pcie_cap_exit(PCIDevice *dev)
164 {
165     pci_del_capability(dev, PCI_CAP_ID_EXP, PCI_EXP_VER2_SIZEOF);
166 }
167 
168 void pcie_cap_v1_exit(PCIDevice *dev)
169 {
170     pci_del_capability(dev, PCI_CAP_ID_EXP, PCI_EXP_VER1_SIZEOF);
171 }
172 
173 uint8_t pcie_cap_get_type(const PCIDevice *dev)
174 {
175     uint32_t pos = dev->exp.exp_cap;
176     assert(pos > 0);
177     return (pci_get_word(dev->config + pos + PCI_EXP_FLAGS) &
178             PCI_EXP_FLAGS_TYPE) >> PCI_EXP_FLAGS_TYPE_SHIFT;
179 }
180 
181 /* MSI/MSI-X */
182 /* pci express interrupt message number */
183 /* 7.8.2 PCI Express Capabilities Register: Interrupt Message Number */
184 void pcie_cap_flags_set_vector(PCIDevice *dev, uint8_t vector)
185 {
186     uint8_t *exp_cap = dev->config + dev->exp.exp_cap;
187     assert(vector < 32);
188     pci_word_test_and_clear_mask(exp_cap + PCI_EXP_FLAGS, PCI_EXP_FLAGS_IRQ);
189     pci_word_test_and_set_mask(exp_cap + PCI_EXP_FLAGS,
190                                vector << PCI_EXP_FLAGS_IRQ_SHIFT);
191 }
192 
193 uint8_t pcie_cap_flags_get_vector(PCIDevice *dev)
194 {
195     return (pci_get_word(dev->config + dev->exp.exp_cap + PCI_EXP_FLAGS) &
196             PCI_EXP_FLAGS_IRQ) >> PCI_EXP_FLAGS_IRQ_SHIFT;
197 }
198 
199 void pcie_cap_deverr_init(PCIDevice *dev)
200 {
201     uint32_t pos = dev->exp.exp_cap;
202     pci_long_test_and_set_mask(dev->config + pos + PCI_EXP_DEVCAP,
203                                PCI_EXP_DEVCAP_RBER);
204     pci_long_test_and_set_mask(dev->wmask + pos + PCI_EXP_DEVCTL,
205                                PCI_EXP_DEVCTL_CERE | PCI_EXP_DEVCTL_NFERE |
206                                PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE);
207     pci_long_test_and_set_mask(dev->w1cmask + pos + PCI_EXP_DEVSTA,
208                                PCI_EXP_DEVSTA_CED | PCI_EXP_DEVSTA_NFED |
209                                PCI_EXP_DEVSTA_FED | PCI_EXP_DEVSTA_URD);
210 }
211 
212 void pcie_cap_deverr_reset(PCIDevice *dev)
213 {
214     uint8_t *devctl = dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL;
215     pci_long_test_and_clear_mask(devctl,
216                                  PCI_EXP_DEVCTL_CERE | PCI_EXP_DEVCTL_NFERE |
217                                  PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE);
218 }
219 
220 static void hotplug_event_update_event_status(PCIDevice *dev)
221 {
222     uint32_t pos = dev->exp.exp_cap;
223     uint8_t *exp_cap = dev->config + pos;
224     uint16_t sltctl = pci_get_word(exp_cap + PCI_EXP_SLTCTL);
225     uint16_t sltsta = pci_get_word(exp_cap + PCI_EXP_SLTSTA);
226 
227     dev->exp.hpev_notified = (sltctl & PCI_EXP_SLTCTL_HPIE) &&
228         (sltsta & sltctl & PCI_EXP_HP_EV_SUPPORTED);
229 }
230 
231 static void hotplug_event_notify(PCIDevice *dev)
232 {
233     bool prev = dev->exp.hpev_notified;
234 
235     hotplug_event_update_event_status(dev);
236 
237     if (prev == dev->exp.hpev_notified) {
238         return;
239     }
240 
241     /* Note: the logic above does not take into account whether interrupts
242      * are masked. The result is that interrupt will be sent when it is
243      * subsequently unmasked. This appears to be legal: Section 6.7.3.4:
244      * The Port may optionally send an MSI when there are hot-plug events that
245      * occur while interrupt generation is disabled, and interrupt generation is
246      * subsequently enabled. */
247     if (msix_enabled(dev)) {
248         msix_notify(dev, pcie_cap_flags_get_vector(dev));
249     } else if (msi_enabled(dev)) {
250         msi_notify(dev, pcie_cap_flags_get_vector(dev));
251     } else {
252         pci_set_irq(dev, dev->exp.hpev_notified);
253     }
254 }
255 
256 static void hotplug_event_clear(PCIDevice *dev)
257 {
258     hotplug_event_update_event_status(dev);
259     if (!msix_enabled(dev) && !msi_enabled(dev) && !dev->exp.hpev_notified) {
260         pci_irq_deassert(dev);
261     }
262 }
263 
264 /*
265  * A PCI Express Hot-Plug Event has occurred, so update slot status register
266  * and notify OS of the event if necessary.
267  *
268  * 6.7.3 PCI Express Hot-Plug Events
269  * 6.7.3.4 Software Notification of Hot-Plug Events
270  */
271 static void pcie_cap_slot_event(PCIDevice *dev, PCIExpressHotPlugEvent event)
272 {
273     /* Minor optimization: if nothing changed - no event is needed. */
274     if (pci_word_test_and_set_mask(dev->config + dev->exp.exp_cap +
275                                    PCI_EXP_SLTSTA, event)) {
276         return;
277     }
278     hotplug_event_notify(dev);
279 }
280 
281 static void pcie_cap_slot_hotplug_common(PCIDevice *hotplug_dev,
282                                          DeviceState *dev,
283                                          uint8_t **exp_cap, Error **errp)
284 {
285     *exp_cap = hotplug_dev->config + hotplug_dev->exp.exp_cap;
286     uint16_t sltsta = pci_get_word(*exp_cap + PCI_EXP_SLTSTA);
287 
288     PCIE_DEV_PRINTF(PCI_DEVICE(dev), "hotplug state: 0x%x\n", sltsta);
289     if (sltsta & PCI_EXP_SLTSTA_EIS) {
290         /* the slot is electromechanically locked.
291          * This error is propagated up to qdev and then to HMP/QMP.
292          */
293         error_setg_errno(errp, EBUSY, "slot is electromechanically locked");
294     }
295 }
296 
297 void pcie_cap_slot_hotplug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
298                               Error **errp)
299 {
300     uint8_t *exp_cap;
301     PCIDevice *pci_dev = PCI_DEVICE(dev);
302 
303     pcie_cap_slot_hotplug_common(PCI_DEVICE(hotplug_dev), dev, &exp_cap, errp);
304 
305     /* Don't send event when device is enabled during qemu machine creation:
306      * it is present on boot, no hotplug event is necessary. We do send an
307      * event when the device is disabled later. */
308     if (!dev->hotplugged) {
309         pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTSTA,
310                                    PCI_EXP_SLTSTA_PDS);
311         return;
312     }
313 
314     /* To enable multifunction hot-plug, we just ensure the function
315      * 0 added last. When function 0 is added, we set the sltsta and
316      * inform OS via event notification.
317      */
318     if (pci_get_function_0(pci_dev)) {
319         pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTSTA,
320                                    PCI_EXP_SLTSTA_PDS);
321         pcie_cap_slot_event(PCI_DEVICE(hotplug_dev),
322                             PCI_EXP_HP_EV_PDC | PCI_EXP_HP_EV_ABP);
323     }
324 }
325 
326 static void pcie_unplug_device(PCIBus *bus, PCIDevice *dev, void *opaque)
327 {
328     object_unparent(OBJECT(dev));
329 }
330 
331 void pcie_cap_slot_hot_unplug_request_cb(HotplugHandler *hotplug_dev,
332                                          DeviceState *dev, Error **errp)
333 {
334     uint8_t *exp_cap;
335     PCIDevice *pci_dev = PCI_DEVICE(dev);
336     PCIBus *bus = pci_dev->bus;
337 
338     pcie_cap_slot_hotplug_common(PCI_DEVICE(hotplug_dev), dev, &exp_cap, errp);
339 
340     /* In case user cancel the operation of multi-function hot-add,
341      * remove the function that is unexposed to guest individually,
342      * without interaction with guest.
343      */
344     if (pci_dev->devfn &&
345         !bus->devices[0]) {
346         pcie_unplug_device(bus, pci_dev, NULL);
347 
348         return;
349     }
350 
351     pcie_cap_slot_push_attention_button(PCI_DEVICE(hotplug_dev));
352 }
353 
354 /* pci express slot for pci express root/downstream port
355    PCI express capability slot registers */
356 void pcie_cap_slot_init(PCIDevice *dev, uint16_t slot)
357 {
358     uint32_t pos = dev->exp.exp_cap;
359 
360     pci_word_test_and_set_mask(dev->config + pos + PCI_EXP_FLAGS,
361                                PCI_EXP_FLAGS_SLOT);
362 
363     pci_long_test_and_clear_mask(dev->config + pos + PCI_EXP_SLTCAP,
364                                  ~PCI_EXP_SLTCAP_PSN);
365     pci_long_test_and_set_mask(dev->config + pos + PCI_EXP_SLTCAP,
366                                (slot << PCI_EXP_SLTCAP_PSN_SHIFT) |
367                                PCI_EXP_SLTCAP_EIP |
368                                PCI_EXP_SLTCAP_HPS |
369                                PCI_EXP_SLTCAP_HPC |
370                                PCI_EXP_SLTCAP_PIP |
371                                PCI_EXP_SLTCAP_AIP |
372                                PCI_EXP_SLTCAP_ABP);
373 
374     if (dev->cap_present & QEMU_PCIE_SLTCAP_PCP) {
375         pci_long_test_and_set_mask(dev->config + pos + PCI_EXP_SLTCAP,
376                                    PCI_EXP_SLTCAP_PCP);
377         pci_word_test_and_clear_mask(dev->config + pos + PCI_EXP_SLTCTL,
378                                      PCI_EXP_SLTCTL_PCC);
379         pci_word_test_and_set_mask(dev->wmask + pos + PCI_EXP_SLTCTL,
380                                    PCI_EXP_SLTCTL_PCC);
381     }
382 
383     pci_word_test_and_clear_mask(dev->config + pos + PCI_EXP_SLTCTL,
384                                  PCI_EXP_SLTCTL_PIC |
385                                  PCI_EXP_SLTCTL_AIC);
386     pci_word_test_and_set_mask(dev->config + pos + PCI_EXP_SLTCTL,
387                                PCI_EXP_SLTCTL_PIC_OFF |
388                                PCI_EXP_SLTCTL_AIC_OFF);
389     pci_word_test_and_set_mask(dev->wmask + pos + PCI_EXP_SLTCTL,
390                                PCI_EXP_SLTCTL_PIC |
391                                PCI_EXP_SLTCTL_AIC |
392                                PCI_EXP_SLTCTL_HPIE |
393                                PCI_EXP_SLTCTL_CCIE |
394                                PCI_EXP_SLTCTL_PDCE |
395                                PCI_EXP_SLTCTL_ABPE);
396     /* Although reading PCI_EXP_SLTCTL_EIC returns always 0,
397      * make the bit writable here in order to detect 1b is written.
398      * pcie_cap_slot_write_config() test-and-clear the bit, so
399      * this bit always returns 0 to the guest.
400      */
401     pci_word_test_and_set_mask(dev->wmask + pos + PCI_EXP_SLTCTL,
402                                PCI_EXP_SLTCTL_EIC);
403 
404     pci_word_test_and_set_mask(dev->w1cmask + pos + PCI_EXP_SLTSTA,
405                                PCI_EXP_HP_EV_SUPPORTED);
406 
407     dev->exp.hpev_notified = false;
408 
409     qbus_set_hotplug_handler(BUS(pci_bridge_get_sec_bus(PCI_BRIDGE(dev))),
410                              DEVICE(dev), NULL);
411 }
412 
413 void pcie_cap_slot_reset(PCIDevice *dev)
414 {
415     uint8_t *exp_cap = dev->config + dev->exp.exp_cap;
416     uint8_t port_type = pcie_cap_get_type(dev);
417 
418     assert(port_type == PCI_EXP_TYPE_DOWNSTREAM ||
419            port_type == PCI_EXP_TYPE_ROOT_PORT);
420 
421     PCIE_DEV_PRINTF(dev, "reset\n");
422 
423     pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTCTL,
424                                  PCI_EXP_SLTCTL_EIC |
425                                  PCI_EXP_SLTCTL_PIC |
426                                  PCI_EXP_SLTCTL_AIC |
427                                  PCI_EXP_SLTCTL_HPIE |
428                                  PCI_EXP_SLTCTL_CCIE |
429                                  PCI_EXP_SLTCTL_PDCE |
430                                  PCI_EXP_SLTCTL_ABPE);
431     pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTCTL,
432                                PCI_EXP_SLTCTL_AIC_OFF);
433 
434     if (dev->cap_present & QEMU_PCIE_SLTCAP_PCP) {
435         /* Downstream ports enforce device number 0. */
436         bool populated = pci_bridge_get_sec_bus(PCI_BRIDGE(dev))->devices[0];
437         uint16_t pic;
438 
439         if (populated) {
440             pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTCTL,
441                                          PCI_EXP_SLTCTL_PCC);
442         } else {
443             pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTCTL,
444                                        PCI_EXP_SLTCTL_PCC);
445         }
446 
447         pic = populated ? PCI_EXP_SLTCTL_PIC_ON : PCI_EXP_SLTCTL_PIC_OFF;
448         pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTCTL, pic);
449     }
450 
451     pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTSTA,
452                                  PCI_EXP_SLTSTA_EIS |/* on reset,
453                                                         the lock is released */
454                                  PCI_EXP_SLTSTA_CC |
455                                  PCI_EXP_SLTSTA_PDC |
456                                  PCI_EXP_SLTSTA_ABP);
457 
458     hotplug_event_update_event_status(dev);
459 }
460 
461 void pcie_cap_slot_write_config(PCIDevice *dev,
462                                 uint32_t addr, uint32_t val, int len)
463 {
464     uint32_t pos = dev->exp.exp_cap;
465     uint8_t *exp_cap = dev->config + pos;
466     uint16_t sltsta = pci_get_word(exp_cap + PCI_EXP_SLTSTA);
467 
468     if (ranges_overlap(addr, len, pos + PCI_EXP_SLTSTA, 2)) {
469         hotplug_event_clear(dev);
470     }
471 
472     if (!ranges_overlap(addr, len, pos + PCI_EXP_SLTCTL, 2)) {
473         return;
474     }
475 
476     if (pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTCTL,
477                                      PCI_EXP_SLTCTL_EIC)) {
478         sltsta ^= PCI_EXP_SLTSTA_EIS; /* toggle PCI_EXP_SLTSTA_EIS bit */
479         pci_set_word(exp_cap + PCI_EXP_SLTSTA, sltsta);
480         PCIE_DEV_PRINTF(dev, "PCI_EXP_SLTCTL_EIC: "
481                         "sltsta -> 0x%02"PRIx16"\n",
482                         sltsta);
483     }
484 
485     /*
486      * If the slot is polulated, power indicator is off and power
487      * controller is off, it is safe to detach the devices.
488      */
489     if ((sltsta & PCI_EXP_SLTSTA_PDS) && (val & PCI_EXP_SLTCTL_PCC) &&
490         ((val & PCI_EXP_SLTCTL_PIC_OFF) == PCI_EXP_SLTCTL_PIC_OFF)) {
491         PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(dev));
492         pci_for_each_device(sec_bus, pci_bus_num(sec_bus),
493                             pcie_unplug_device, NULL);
494 
495         pci_word_test_and_clear_mask(exp_cap + PCI_EXP_SLTSTA,
496                                      PCI_EXP_SLTSTA_PDS);
497         pci_word_test_and_set_mask(exp_cap + PCI_EXP_SLTSTA,
498                                        PCI_EXP_SLTSTA_PDC);
499     }
500 
501     hotplug_event_notify(dev);
502 
503     /*
504      * 6.7.3.2 Command Completed Events
505      *
506      * Software issues a command to a hot-plug capable Downstream Port by
507      * issuing a write transaction that targets any portion of the Port’s Slot
508      * Control register. A single write to the Slot Control register is
509      * considered to be a single command, even if the write affects more than
510      * one field in the Slot Control register. In response to this transaction,
511      * the Port must carry out the requested actions and then set the
512      * associated status field for the command completed event. */
513 
514     /* Real hardware might take a while to complete requested command because
515      * physical movement would be involved like locking the electromechanical
516      * lock.  However in our case, command is completed instantaneously above,
517      * so send a command completion event right now.
518      */
519     pcie_cap_slot_event(dev, PCI_EXP_HP_EV_CCI);
520 }
521 
522 int pcie_cap_slot_post_load(void *opaque, int version_id)
523 {
524     PCIDevice *dev = opaque;
525     hotplug_event_update_event_status(dev);
526     return 0;
527 }
528 
529 void pcie_cap_slot_push_attention_button(PCIDevice *dev)
530 {
531     pcie_cap_slot_event(dev, PCI_EXP_HP_EV_ABP);
532 }
533 
534 /* root control/capabilities/status. PME isn't emulated for now */
535 void pcie_cap_root_init(PCIDevice *dev)
536 {
537     pci_set_word(dev->wmask + dev->exp.exp_cap + PCI_EXP_RTCTL,
538                  PCI_EXP_RTCTL_SECEE | PCI_EXP_RTCTL_SENFEE |
539                  PCI_EXP_RTCTL_SEFEE);
540 }
541 
542 void pcie_cap_root_reset(PCIDevice *dev)
543 {
544     pci_set_word(dev->config + dev->exp.exp_cap + PCI_EXP_RTCTL, 0);
545 }
546 
547 /* function level reset(FLR) */
548 void pcie_cap_flr_init(PCIDevice *dev)
549 {
550     pci_long_test_and_set_mask(dev->config + dev->exp.exp_cap + PCI_EXP_DEVCAP,
551                                PCI_EXP_DEVCAP_FLR);
552 
553     /* Although reading BCR_FLR returns always 0,
554      * the bit is made writable here in order to detect the 1b is written
555      * pcie_cap_flr_write_config() test-and-clear the bit, so
556      * this bit always returns 0 to the guest.
557      */
558     pci_word_test_and_set_mask(dev->wmask + dev->exp.exp_cap + PCI_EXP_DEVCTL,
559                                PCI_EXP_DEVCTL_BCR_FLR);
560 }
561 
562 void pcie_cap_flr_write_config(PCIDevice *dev,
563                                uint32_t addr, uint32_t val, int len)
564 {
565     uint8_t *devctl = dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL;
566     if (pci_get_word(devctl) & PCI_EXP_DEVCTL_BCR_FLR) {
567         /* Clear PCI_EXP_DEVCTL_BCR_FLR after invoking the reset handler
568            so the handler can detect FLR by looking at this bit. */
569         pci_device_reset(dev);
570         pci_word_test_and_clear_mask(devctl, PCI_EXP_DEVCTL_BCR_FLR);
571     }
572 }
573 
574 /* Alternative Routing-ID Interpretation (ARI)
575  * forwarding support for root and downstream ports
576  */
577 void pcie_cap_arifwd_init(PCIDevice *dev)
578 {
579     uint32_t pos = dev->exp.exp_cap;
580     pci_long_test_and_set_mask(dev->config + pos + PCI_EXP_DEVCAP2,
581                                PCI_EXP_DEVCAP2_ARI);
582     pci_long_test_and_set_mask(dev->wmask + pos + PCI_EXP_DEVCTL2,
583                                PCI_EXP_DEVCTL2_ARI);
584 }
585 
586 void pcie_cap_arifwd_reset(PCIDevice *dev)
587 {
588     uint8_t *devctl2 = dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL2;
589     pci_long_test_and_clear_mask(devctl2, PCI_EXP_DEVCTL2_ARI);
590 }
591 
592 bool pcie_cap_is_arifwd_enabled(const PCIDevice *dev)
593 {
594     if (!pci_is_express(dev)) {
595         return false;
596     }
597     if (!dev->exp.exp_cap) {
598         return false;
599     }
600 
601     return pci_get_long(dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL2) &
602         PCI_EXP_DEVCTL2_ARI;
603 }
604 
605 /**************************************************************************
606  * pci express extended capability list management functions
607  * uint16_t ext_cap_id (16 bit)
608  * uint8_t cap_ver (4 bit)
609  * uint16_t cap_offset (12 bit)
610  * uint16_t ext_cap_size
611  */
612 
613 static uint16_t pcie_find_capability_list(PCIDevice *dev, uint16_t cap_id,
614                                           uint16_t *prev_p)
615 {
616     uint16_t prev = 0;
617     uint16_t next;
618     uint32_t header = pci_get_long(dev->config + PCI_CONFIG_SPACE_SIZE);
619 
620     if (!header) {
621         /* no extended capability */
622         next = 0;
623         goto out;
624     }
625     for (next = PCI_CONFIG_SPACE_SIZE; next;
626          prev = next, next = PCI_EXT_CAP_NEXT(header)) {
627 
628         assert(next >= PCI_CONFIG_SPACE_SIZE);
629         assert(next <= PCIE_CONFIG_SPACE_SIZE - 8);
630 
631         header = pci_get_long(dev->config + next);
632         if (PCI_EXT_CAP_ID(header) == cap_id) {
633             break;
634         }
635     }
636 
637 out:
638     if (prev_p) {
639         *prev_p = prev;
640     }
641     return next;
642 }
643 
644 uint16_t pcie_find_capability(PCIDevice *dev, uint16_t cap_id)
645 {
646     return pcie_find_capability_list(dev, cap_id, NULL);
647 }
648 
649 static void pcie_ext_cap_set_next(PCIDevice *dev, uint16_t pos, uint16_t next)
650 {
651     uint32_t header = pci_get_long(dev->config + pos);
652     assert(!(next & (PCI_EXT_CAP_ALIGN - 1)));
653     header = (header & ~PCI_EXT_CAP_NEXT_MASK) |
654         ((next << PCI_EXT_CAP_NEXT_SHIFT) & PCI_EXT_CAP_NEXT_MASK);
655     pci_set_long(dev->config + pos, header);
656 }
657 
658 /*
659  * caller must supply valid (offset, size) * such that the range shouldn't
660  * overlap with other capability or other registers.
661  * This function doesn't check it.
662  */
663 void pcie_add_capability(PCIDevice *dev,
664                          uint16_t cap_id, uint8_t cap_ver,
665                          uint16_t offset, uint16_t size)
666 {
667     uint32_t header;
668     uint16_t next;
669 
670     assert(offset >= PCI_CONFIG_SPACE_SIZE);
671     assert(offset < offset + size);
672     assert(offset + size <= PCIE_CONFIG_SPACE_SIZE);
673     assert(size >= 8);
674     assert(pci_is_express(dev));
675 
676     if (offset == PCI_CONFIG_SPACE_SIZE) {
677         header = pci_get_long(dev->config + offset);
678         next = PCI_EXT_CAP_NEXT(header);
679     } else {
680         uint16_t prev;
681 
682         /* 0 is reserved cap id. use internally to find the last capability
683            in the linked list */
684         next = pcie_find_capability_list(dev, 0, &prev);
685 
686         assert(prev >= PCI_CONFIG_SPACE_SIZE);
687         assert(next == 0);
688         pcie_ext_cap_set_next(dev, prev, offset);
689     }
690     pci_set_long(dev->config + offset, PCI_EXT_CAP(cap_id, cap_ver, next));
691 
692     /* Make capability read-only by default */
693     memset(dev->wmask + offset, 0, size);
694     memset(dev->w1cmask + offset, 0, size);
695     /* Check capability by default */
696     memset(dev->cmask + offset, 0xFF, size);
697 }
698 
699 /**************************************************************************
700  * pci express extended capability helper functions
701  */
702 
703 /* ARI */
704 void pcie_ari_init(PCIDevice *dev, uint16_t offset, uint16_t nextfn)
705 {
706     pcie_add_capability(dev, PCI_EXT_CAP_ID_ARI, PCI_ARI_VER,
707                         offset, PCI_ARI_SIZEOF);
708     pci_set_long(dev->config + offset + PCI_ARI_CAP, (nextfn & 0xff) << 8);
709 }
710 
711 void pcie_dev_ser_num_init(PCIDevice *dev, uint16_t offset, uint64_t ser_num)
712 {
713     static const int pci_dsn_ver = 1;
714     static const int pci_dsn_cap = 4;
715 
716     pcie_add_capability(dev, PCI_EXT_CAP_ID_DSN, pci_dsn_ver, offset,
717                         PCI_EXT_CAP_DSN_SIZEOF);
718     pci_set_quad(dev->config + offset + pci_dsn_cap, ser_num);
719 }
720 
721 void pcie_ats_init(PCIDevice *dev, uint16_t offset)
722 {
723     pcie_add_capability(dev, PCI_EXT_CAP_ID_ATS, 0x1,
724                         offset, PCI_EXT_CAP_ATS_SIZEOF);
725 
726     dev->exp.ats_cap = offset;
727 
728     /* Invalidate Queue Depth 0, Page Aligned Request 0 */
729     pci_set_word(dev->config + offset + PCI_ATS_CAP, 0);
730     /* STU 0, Disabled by default */
731     pci_set_word(dev->config + offset + PCI_ATS_CTRL, 0);
732 
733     pci_set_word(dev->wmask + dev->exp.ats_cap + PCI_ATS_CTRL, 0x800f);
734 }
735