4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
5 * Copyright 2013-2014 6WIND S.A.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * * Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * * Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in
16 * the documentation and/or other materials provided with the
18 * * Neither the name of Intel Corporation nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
25 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
26 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
27 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
28 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
32 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
40 #include <sys/queue.h>
43 #include <rte_errno.h>
44 #include <rte_interrupts.h>
48 #include <rte_per_lcore.h>
49 #include <rte_memory.h>
50 #include <rte_memzone.h>
52 #include <rte_string_fns.h>
53 #include <rte_common.h>
54 #include <rte_devargs.h>
56 #include "eal_private.h"
58 extern struct rte_pci_bus rte_pci_bus;
60 #define SYSFS_PCI_DEVICES "/sys/bus/pci/devices"
62 const char *pci_get_sysfs_path(void)
64 const char *path = NULL;
66 path = getenv("SYSFS_PCI_DEVICES");
68 return SYSFS_PCI_DEVICES;
73 static struct rte_devargs *pci_devargs_lookup(struct rte_pci_device *dev)
75 struct rte_devargs *devargs;
76 struct rte_pci_addr addr;
79 pbus = rte_bus_find_by_name("pci");
80 TAILQ_FOREACH(devargs, &devargs_list, next) {
81 if (devargs->bus != pbus)
83 devargs->bus->parse(devargs->name, &addr);
84 if (!rte_eal_compare_pci_addr(&dev->addr, &addr))
90 static inline const char *
91 get_u8_pciaddr_field(const char *in, void *_u8, char dlm)
98 val = strtoul(in, &end, 16);
99 if (errno != 0 || end[0] != dlm || val > UINT8_MAX) {
100 errno = errno ? errno : EINVAL;
109 eal_parse_pci_BDF(const char *input, struct rte_pci_addr *dev_addr)
111 const char *in = input;
113 dev_addr->domain = 0;
114 in = get_u8_pciaddr_field(in, &dev_addr->bus, ':');
117 in = get_u8_pciaddr_field(in, &dev_addr->devid, '.');
120 in = get_u8_pciaddr_field(in, &dev_addr->function, '\0');
127 eal_parse_pci_DomBDF(const char *input, struct rte_pci_addr *dev_addr)
129 const char *in = input;
134 val = strtoul(in, &end, 16);
135 if (errno != 0 || end[0] != ':' || val > UINT16_MAX)
137 dev_addr->domain = (uint16_t)val;
139 in = get_u8_pciaddr_field(in, &dev_addr->bus, ':');
142 in = get_u8_pciaddr_field(in, &dev_addr->devid, '.');
145 in = get_u8_pciaddr_field(in, &dev_addr->function, '\0');
152 rte_pci_device_name(const struct rte_pci_addr *addr,
153 char *output, size_t size)
155 RTE_VERIFY(size >= PCI_PRI_STR_SIZE);
156 RTE_VERIFY(snprintf(output, size, PCI_PRI_FMT,
157 addr->domain, addr->bus,
158 addr->devid, addr->function) >= 0);
162 rte_eal_compare_pci_addr(const struct rte_pci_addr *addr,
163 const struct rte_pci_addr *addr2)
165 uint64_t dev_addr, dev_addr2;
167 if ((addr == NULL) || (addr2 == NULL))
170 dev_addr = ((uint64_t)addr->domain << 24) |
171 (addr->bus << 16) | (addr->devid << 8) | addr->function;
172 dev_addr2 = ((uint64_t)addr2->domain << 24) |
173 (addr2->bus << 16) | (addr2->devid << 8) | addr2->function;
175 if (dev_addr > dev_addr2)
177 else if (dev_addr < dev_addr2)
184 pci_name_set(struct rte_pci_device *dev)
186 struct rte_devargs *devargs;
188 /* Each device has its internal, canonical name set. */
189 rte_pci_device_name(&dev->addr,
190 dev->name, sizeof(dev->name));
191 devargs = pci_devargs_lookup(dev);
192 dev->device.devargs = devargs;
193 /* In blacklist mode, if the device is not blacklisted, no
194 * rte_devargs exists for it.
197 /* If an rte_devargs exists, the generic rte_device uses the
198 * given name as its namea
200 dev->device.name = dev->device.devargs->name;
202 /* Otherwise, it uses the internal, canonical form. */
203 dev->device.name = dev->name;
206 /* map a particular resource from a file */
208 pci_map_resource(void *requested_addr, int fd, off_t offset, size_t size,
209 int additional_flags)
213 /* Map the PCI memory resource of device */
214 mapaddr = mmap(requested_addr, size, PROT_READ | PROT_WRITE,
215 MAP_SHARED | additional_flags, fd, offset);
216 if (mapaddr == MAP_FAILED) {
217 RTE_LOG(ERR, EAL, "%s(): cannot mmap(%d, %p, 0x%lx, 0x%lx): %s (%p)\n",
218 __func__, fd, requested_addr,
219 (unsigned long)size, (unsigned long)offset,
220 strerror(errno), mapaddr);
222 RTE_LOG(DEBUG, EAL, " PCI memory mapped at %p\n", mapaddr);
227 /* unmap a particular resource */
229 pci_unmap_resource(void *requested_addr, size_t size)
231 if (requested_addr == NULL)
234 /* Unmap the PCI memory resource of device */
235 if (munmap(requested_addr, size)) {
236 RTE_LOG(ERR, EAL, "%s(): cannot munmap(%p, 0x%lx): %s\n",
237 __func__, requested_addr, (unsigned long)size,
240 RTE_LOG(DEBUG, EAL, " PCI memory unmapped at %p\n",
245 * Match the PCI Driver and Device using the ID Table
248 rte_pci_match(const struct rte_pci_driver *pci_drv,
249 const struct rte_pci_device *pci_dev)
251 const struct rte_pci_id *id_table;
253 for (id_table = pci_drv->id_table; id_table->vendor_id != 0;
255 /* check if device's identifiers match the driver's ones */
256 if (id_table->vendor_id != pci_dev->id.vendor_id &&
257 id_table->vendor_id != PCI_ANY_ID)
259 if (id_table->device_id != pci_dev->id.device_id &&
260 id_table->device_id != PCI_ANY_ID)
262 if (id_table->subsystem_vendor_id !=
263 pci_dev->id.subsystem_vendor_id &&
264 id_table->subsystem_vendor_id != PCI_ANY_ID)
266 if (id_table->subsystem_device_id !=
267 pci_dev->id.subsystem_device_id &&
268 id_table->subsystem_device_id != PCI_ANY_ID)
270 if (id_table->class_id != pci_dev->id.class_id &&
271 id_table->class_id != RTE_CLASS_ANY_ID)
281 * If vendor/device ID match, call the probe() function of the
285 rte_pci_probe_one_driver(struct rte_pci_driver *dr,
286 struct rte_pci_device *dev)
289 struct rte_pci_addr *loc;
291 if ((dr == NULL) || (dev == NULL))
296 /* The device is not blacklisted; Check if driver supports it */
297 if (!rte_pci_match(dr, dev))
298 /* Match of device and driver failed */
301 RTE_LOG(INFO, EAL, "PCI device "PCI_PRI_FMT" on NUMA socket %i\n",
302 loc->domain, loc->bus, loc->devid, loc->function,
303 dev->device.numa_node);
305 /* no initialization when blacklisted, return without error */
306 if (dev->device.devargs != NULL &&
307 dev->device.devargs->policy ==
308 RTE_DEV_BLACKLISTED) {
309 RTE_LOG(INFO, EAL, " Device is blacklisted, not"
314 if (dev->device.numa_node < 0) {
315 RTE_LOG(WARNING, EAL, " Invalid NUMA socket, default to 0\n");
316 dev->device.numa_node = 0;
319 RTE_LOG(INFO, EAL, " probe driver: %x:%x %s\n", dev->id.vendor_id,
320 dev->id.device_id, dr->driver.name);
322 if (dr->drv_flags & RTE_PCI_DRV_NEED_MAPPING) {
323 /* map resources for devices that use igb_uio */
324 ret = rte_pci_map_device(dev);
329 /* reference driver structure */
331 dev->device.driver = &dr->driver;
333 /* call the driver probe() function */
334 ret = dr->probe(dr, dev);
337 dev->device.driver = NULL;
338 if ((dr->drv_flags & RTE_PCI_DRV_NEED_MAPPING) &&
339 /* Don't unmap if device is unsupported and
340 * driver needs mapped resources.
343 (dr->drv_flags & RTE_PCI_DRV_KEEP_MAPPED_RES)))
344 rte_pci_unmap_device(dev);
351 * If vendor/device ID match, call the remove() function of the
355 rte_pci_detach_dev(struct rte_pci_device *dev)
357 struct rte_pci_addr *loc;
358 struct rte_pci_driver *dr;
367 RTE_LOG(DEBUG, EAL, "PCI device "PCI_PRI_FMT" on NUMA socket %i\n",
368 loc->domain, loc->bus, loc->devid,
369 loc->function, dev->device.numa_node);
371 RTE_LOG(DEBUG, EAL, " remove driver: %x:%x %s\n", dev->id.vendor_id,
372 dev->id.device_id, dr->driver.name);
375 ret = dr->remove(dev);
380 /* clear driver structure */
383 if (dr->drv_flags & RTE_PCI_DRV_NEED_MAPPING)
384 /* unmap resources for devices that use igb_uio */
385 rte_pci_unmap_device(dev);
391 * If vendor/device ID match, call the probe() function of all
392 * registered driver for the given device. Return -1 if initialization
393 * failed, return 1 if no driver is found for this device.
396 pci_probe_all_drivers(struct rte_pci_device *dev)
398 struct rte_pci_driver *dr = NULL;
404 /* Check if a driver is already loaded */
405 if (dev->driver != NULL)
408 FOREACH_DRIVER_ON_PCIBUS(dr) {
409 rc = rte_pci_probe_one_driver(dr, dev);
411 /* negative value is an error */
414 /* positive value means driver doesn't support it */
422 * Find the pci device specified by pci address, then invoke probe function of
423 * the driver of the device.
426 rte_pci_probe_one(const struct rte_pci_addr *addr)
428 struct rte_pci_device *dev = NULL;
435 /* update current pci device in global list, kernel bindings might have
436 * changed since last time we looked at it.
438 if (pci_update_device(addr) < 0)
441 FOREACH_DEVICE_ON_PCIBUS(dev) {
442 if (rte_eal_compare_pci_addr(&dev->addr, addr))
445 ret = pci_probe_all_drivers(dev);
453 RTE_LOG(WARNING, EAL,
454 "Requested device " PCI_PRI_FMT " cannot be used\n",
455 addr->domain, addr->bus, addr->devid, addr->function);
460 * Detach device specified by its pci address.
463 rte_pci_detach(const struct rte_pci_addr *addr)
465 struct rte_pci_device *dev = NULL;
471 FOREACH_DEVICE_ON_PCIBUS(dev) {
472 if (rte_eal_compare_pci_addr(&dev->addr, addr))
475 ret = rte_pci_detach_dev(dev);
477 /* negative value is an error */
480 /* positive value means driver doesn't support it */
483 rte_pci_remove_device(dev);
490 RTE_LOG(WARNING, EAL, "Requested device " PCI_PRI_FMT
491 " cannot be used\n", dev->addr.domain, dev->addr.bus,
492 dev->addr.devid, dev->addr.function);
497 * Scan the content of the PCI bus, and call the probe() function for
498 * all registered drivers that have a matching entry in its id_table
499 * for discovered devices.
504 struct rte_pci_device *dev = NULL;
505 size_t probed = 0, failed = 0;
506 struct rte_devargs *devargs;
510 if (rte_pci_bus.bus.conf.scan_mode != RTE_BUS_SCAN_WHITELIST)
513 FOREACH_DEVICE_ON_PCIBUS(dev) {
516 devargs = dev->device.devargs;
517 /* probe all or only whitelisted devices */
519 ret = pci_probe_all_drivers(dev);
520 else if (devargs != NULL &&
521 devargs->policy == RTE_DEV_WHITELISTED)
522 ret = pci_probe_all_drivers(dev);
524 RTE_LOG(ERR, EAL, "Requested device " PCI_PRI_FMT
525 " cannot be used\n", dev->addr.domain, dev->addr.bus,
526 dev->addr.devid, dev->addr.function);
533 return (probed && probed == failed) ? -1 : 0;
536 /* dump one device */
538 pci_dump_one_device(FILE *f, struct rte_pci_device *dev)
542 fprintf(f, PCI_PRI_FMT, dev->addr.domain, dev->addr.bus,
543 dev->addr.devid, dev->addr.function);
544 fprintf(f, " - vendor:%x device:%x\n", dev->id.vendor_id,
547 for (i = 0; i != sizeof(dev->mem_resource) /
548 sizeof(dev->mem_resource[0]); i++) {
549 fprintf(f, " %16.16"PRIx64" %16.16"PRIx64"\n",
550 dev->mem_resource[i].phys_addr,
551 dev->mem_resource[i].len);
556 /* dump devices on the bus */
558 rte_pci_dump(FILE *f)
560 struct rte_pci_device *dev = NULL;
562 FOREACH_DEVICE_ON_PCIBUS(dev) {
563 pci_dump_one_device(f, dev);
568 pci_parse(const char *name, void *addr)
570 struct rte_pci_addr *out = addr;
571 struct rte_pci_addr pci_addr;
574 parse = (eal_parse_pci_BDF(name, &pci_addr) == 0 ||
575 eal_parse_pci_DomBDF(name, &pci_addr) == 0);
576 if (parse && addr != NULL)
578 return parse == false;
581 /* register a driver */
583 rte_pci_register(struct rte_pci_driver *driver)
585 TAILQ_INSERT_TAIL(&rte_pci_bus.driver_list, driver, next);
586 driver->bus = &rte_pci_bus;
589 /* unregister a driver */
591 rte_pci_unregister(struct rte_pci_driver *driver)
593 TAILQ_REMOVE(&rte_pci_bus.driver_list, driver, next);
597 /* Add a device to PCI bus */
599 rte_pci_add_device(struct rte_pci_device *pci_dev)
601 TAILQ_INSERT_TAIL(&rte_pci_bus.device_list, pci_dev, next);
604 /* Insert a device into a predefined position in PCI bus */
606 rte_pci_insert_device(struct rte_pci_device *exist_pci_dev,
607 struct rte_pci_device *new_pci_dev)
609 TAILQ_INSERT_BEFORE(exist_pci_dev, new_pci_dev, next);
612 /* Remove a device from PCI bus */
614 rte_pci_remove_device(struct rte_pci_device *pci_dev)
616 TAILQ_REMOVE(&rte_pci_bus.device_list, pci_dev, next);
619 static struct rte_device *
620 pci_find_device(const struct rte_device *start, rte_dev_cmp_t cmp,
623 struct rte_pci_device *dev;
625 FOREACH_DEVICE_ON_PCIBUS(dev) {
626 if (start && &dev->device == start) {
627 start = NULL; /* starting point found */
630 if (cmp(&dev->device, data) == 0)
638 pci_plug(struct rte_device *dev)
640 return pci_probe_all_drivers(RTE_DEV_TO_PCI(dev));
644 pci_unplug(struct rte_device *dev)
646 struct rte_pci_device *pdev;
649 pdev = RTE_DEV_TO_PCI(dev);
650 ret = rte_pci_detach_dev(pdev);
652 rte_pci_remove_device(pdev);
658 struct rte_pci_bus rte_pci_bus = {
660 .scan = rte_pci_scan,
661 .probe = rte_pci_probe,
662 .find_device = pci_find_device,
664 .unplug = pci_unplug,
666 .get_iommu_class = rte_pci_get_iommu_class,
668 .device_list = TAILQ_HEAD_INITIALIZER(rte_pci_bus.device_list),
669 .driver_list = TAILQ_HEAD_INITIALIZER(rte_pci_bus.driver_list),
672 RTE_REGISTER_BUS(pci, rte_pci_bus.bus);