4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * * Neither the name of Intel Corporation nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
41 #include <sys/types.h>
48 #include <sys/queue.h>
50 #include <sys/ioctl.h>
51 #include <sys/pciio.h>
52 #include <dev/pci/pcireg.h>
54 #include <rte_interrupts.h>
57 #include <rte_common.h>
58 #include <rte_launch.h>
59 #include <rte_memory.h>
60 #include <rte_memzone.h>
62 #include <rte_eal_memconfig.h>
63 #include <rte_per_lcore.h>
64 #include <rte_lcore.h>
65 #include <rte_malloc.h>
66 #include <rte_string_fns.h>
67 #include <rte_debug.h>
68 #include <rte_devargs.h>
70 #include "rte_pci_dev_ids.h"
71 #include "eal_filesystem.h"
72 #include "eal_private.h"
76 * PCI probing under linux
78 * This code is used to simulate a PCI probe by parsing information in
79 * sysfs. Moreover, when a registered driver matches a device, the
80 * kernel driver currently using it is unloaded and replaced by
81 * igb_uio module, which is a very minimal userland driver for Intel
82 * network card, only providing access to PCI BAR to applications, and
83 * enabling bus master.
85 static struct rte_tailq_elem rte_uio_tailq = {
86 .name = "UIO_RESOURCE_LIST",
88 EAL_REGISTER_TAILQ(rte_uio_tailq)
90 /* unbind kernel driver for this device */
92 pci_unbind_kernel_driver(struct rte_pci_device *dev __rte_unused)
94 RTE_LOG(ERR, EAL, "RTE_PCI_DRV_FORCE_UNBIND flag is not implemented "
100 pci_uio_map_secondary(struct rte_pci_device *dev)
103 struct mapped_pci_resource *uio_res;
104 struct mapped_pci_res_list *uio_res_list =
105 RTE_TAILQ_CAST(rte_uio_tailq.head, mapped_pci_res_list);
107 TAILQ_FOREACH(uio_res, uio_res_list, next) {
109 /* skip this element if it doesn't match our PCI address */
110 if (rte_eal_compare_pci_addr(&uio_res->pci_addr, &dev->addr))
113 for (i = 0; i != uio_res->nb_maps; i++) {
115 * open devname, to mmap it
117 fd = open(uio_res->maps[i].path, O_RDWR);
119 RTE_LOG(ERR, EAL, "Cannot open %s: %s\n",
120 uio_res->maps[i].path, strerror(errno));
124 void *mapaddr = pci_map_resource(uio_res->maps[i].addr,
125 fd, (off_t)uio_res->maps[i].offset,
126 (size_t)uio_res->maps[i].size, 0);
127 /* fd is not needed in slave process, close it */
129 if (mapaddr != uio_res->maps[i].addr) {
131 "Cannot mmap device resource file %s to address: %p\n",
132 uio_res->maps[i].path,
133 uio_res->maps[i].addr);
140 RTE_LOG(ERR, EAL, "Cannot find resource for device\n");
145 pci_uio_free_resource(struct rte_pci_device *dev,
146 struct mapped_pci_resource *uio_res)
150 if (dev->intr_handle.fd) {
151 close(dev->intr_handle.fd);
152 dev->intr_handle.fd = -1;
153 dev->intr_handle.type = RTE_INTR_HANDLE_UNKNOWN;
158 pci_uio_alloc_resource(struct rte_pci_device *dev,
159 struct mapped_pci_resource **uio_res)
161 char devname[PATH_MAX]; /* contains the /dev/uioX */
162 struct rte_pci_addr *loc;
166 snprintf(devname, sizeof(devname), "/dev/uio@pci:%u:%u:%u",
167 dev->addr.bus, dev->addr.devid, dev->addr.function);
169 if (access(devname, O_RDWR) < 0) {
170 RTE_LOG(WARNING, EAL, " "PCI_PRI_FMT" not managed by UIO driver, "
171 "skipping\n", loc->domain, loc->bus, loc->devid, loc->function);
175 /* save fd if in primary process */
176 dev->intr_handle.fd = open(devname, O_RDWR);
177 if (dev->intr_handle.fd < 0) {
178 RTE_LOG(ERR, EAL, "Cannot open %s: %s\n",
179 devname, strerror(errno));
182 dev->intr_handle.type = RTE_INTR_HANDLE_UIO;
184 /* allocate the mapping details for secondary processes*/
185 *uio_res = rte_zmalloc("UIO_RES", sizeof(**uio_res), 0);
186 if (*uio_res == NULL) {
188 "%s(): cannot store uio mmap details\n", __func__);
192 snprintf((*uio_res)->path, sizeof((*uio_res)->path), "%s", devname);
193 memcpy(&(*uio_res)->pci_addr, &dev->addr, sizeof((*uio_res)->pci_addr));
198 pci_uio_free_resource(dev, *uio_res);
203 pci_uio_map_resource_by_index(struct rte_pci_device *dev, int res_idx,
204 struct mapped_pci_resource *uio_res, int map_idx)
211 struct pci_map *maps;
213 maps = uio_res->maps;
214 devname = uio_res->path;
215 pagesz = sysconf(_SC_PAGESIZE);
217 /* allocate memory to keep path */
218 maps[map_idx].path = rte_malloc(NULL, strlen(devname) + 1, 0);
219 if (maps[map_idx].path == NULL) {
220 RTE_LOG(ERR, EAL, "Cannot allocate memory for path: %s\n",
226 * open resource file, to mmap it
228 fd = open(devname, O_RDWR);
230 RTE_LOG(ERR, EAL, "Cannot open %s: %s\n",
231 devname, strerror(errno));
235 /* if matching map is found, then use it */
236 offset = res_idx * pagesz;
237 mapaddr = pci_map_resource(NULL, fd, (off_t)offset,
238 (size_t)dev->mem_resource[res_idx].len, 0);
240 if (mapaddr == MAP_FAILED)
243 maps[map_idx].phaddr = dev->mem_resource[res_idx].phys_addr;
244 maps[map_idx].size = dev->mem_resource[res_idx].len;
245 maps[map_idx].addr = mapaddr;
246 maps[map_idx].offset = offset;
247 strcpy(maps[map_idx].path, devname);
248 dev->mem_resource[res_idx].addr = mapaddr;
253 rte_free(maps[map_idx].path);
257 /* map the PCI resource of a PCI device in virtual memory */
259 pci_uio_map_resource(struct rte_pci_device *dev)
261 int i, map_idx = 0, ret;
263 struct mapped_pci_resource *uio_res = NULL;
264 struct mapped_pci_res_list *uio_res_list =
265 RTE_TAILQ_CAST(rte_uio_tailq.head, mapped_pci_res_list);
267 dev->intr_handle.fd = -1;
268 dev->intr_handle.type = RTE_INTR_HANDLE_UNKNOWN;
270 /* secondary processes - use already recorded details */
271 if (rte_eal_process_type() != RTE_PROC_PRIMARY)
272 return pci_uio_map_secondary(dev);
274 /* allocate uio resource */
275 ret = pci_uio_alloc_resource(dev, &uio_res);
280 for (i = 0; i != PCI_MAX_RESOURCE; i++) {
282 if ((phaddr = dev->mem_resource[i].phys_addr) == 0)
285 ret = pci_uio_map_resource_by_index(dev, i,
293 uio_res->nb_maps = map_idx;
295 TAILQ_INSERT_TAIL(uio_res_list, uio_res, next);
299 for (i = 0; i < map_idx; i++)
300 rte_free(uio_res->maps[i].path);
301 pci_uio_free_resource(dev, uio_res);
305 /* Scan one pci sysfs entry, and fill the devices list from it. */
307 pci_scan_one(int dev_pci_fd, struct pci_conf *conf)
309 struct rte_pci_device *dev;
310 struct pci_bar_io bar;
313 dev = malloc(sizeof(*dev));
318 memset(dev, 0, sizeof(*dev));
319 dev->addr.domain = conf->pc_sel.pc_domain;
320 dev->addr.bus = conf->pc_sel.pc_bus;
321 dev->addr.devid = conf->pc_sel.pc_dev;
322 dev->addr.function = conf->pc_sel.pc_func;
325 dev->id.vendor_id = conf->pc_vendor;
328 dev->id.device_id = conf->pc_device;
330 /* get subsystem_vendor id */
331 dev->id.subsystem_vendor_id = conf->pc_subvendor;
333 /* get subsystem_device id */
334 dev->id.subsystem_device_id = conf->pc_subdevice;
336 /* TODO: get max_vfs */
339 /* FreeBSD has no NUMA support (yet) */
342 /* parse resources */
343 switch (conf->pc_hdr & PCIM_HDRTYPE) {
344 case PCIM_HDRTYPE_NORMAL:
345 max = PCIR_MAX_BAR_0;
347 case PCIM_HDRTYPE_BRIDGE:
348 max = PCIR_MAX_BAR_1;
350 case PCIM_HDRTYPE_CARDBUS:
351 max = PCIR_MAX_BAR_2;
357 for (i = 0; i <= max; i++) {
358 bar.pbi_sel = conf->pc_sel;
359 bar.pbi_reg = PCIR_BAR(i);
360 if (ioctl(dev_pci_fd, PCIOCGETBAR, &bar) < 0)
363 dev->mem_resource[i].len = bar.pbi_length;
364 if (PCI_BAR_IO(bar.pbi_base)) {
365 dev->mem_resource[i].addr = (void *)(bar.pbi_base & ~((uint64_t)0xf));
368 dev->mem_resource[i].phys_addr = bar.pbi_base & ~((uint64_t)0xf);
371 /* device is valid, add in list (sorted) */
372 if (TAILQ_EMPTY(&pci_device_list)) {
373 TAILQ_INSERT_TAIL(&pci_device_list, dev, next);
376 struct rte_pci_device *dev2 = NULL;
379 TAILQ_FOREACH(dev2, &pci_device_list, next) {
380 ret = rte_eal_compare_pci_addr(&dev->addr, &dev2->addr);
384 TAILQ_INSERT_BEFORE(dev2, dev, next);
386 } else { /* already registered */
387 dev2->kdrv = dev->kdrv;
388 dev2->max_vfs = dev->max_vfs;
389 memmove(dev2->mem_resource,
391 sizeof(dev->mem_resource));
396 TAILQ_INSERT_TAIL(&pci_device_list, dev, next);
407 * Scan the content of the PCI bus, and add the devices in the devices
408 * list. Call pci_scan_one() for each pci entry found.
414 unsigned dev_count = 0;
415 struct pci_conf matches[16];
416 struct pci_conf_io conf_io = {
420 .match_buf_len = sizeof(matches),
421 .matches = &matches[0],
424 fd = open("/dev/pci", O_RDONLY);
426 RTE_LOG(ERR, EAL, "%s(): error opening /dev/pci\n", __func__);
432 if (ioctl(fd, PCIOCGETCONF, &conf_io) < 0) {
433 RTE_LOG(ERR, EAL, "%s(): error with ioctl on /dev/pci: %s\n",
434 __func__, strerror(errno));
438 for (i = 0; i < conf_io.num_matches; i++)
439 if (pci_scan_one(fd, &matches[i]) < 0)
442 dev_count += conf_io.num_matches;
443 } while(conf_io.status == PCI_GETCONF_MORE_DEVS);
447 RTE_LOG(ERR, EAL, "PCI scan found %u devices\n", dev_count);
457 * If vendor/device ID match, call the devinit() function of the
461 rte_eal_pci_probe_one_driver(struct rte_pci_driver *dr, struct rte_pci_device *dev)
463 const struct rte_pci_id *id_table;
466 for (id_table = dr->id_table ; id_table->vendor_id != 0; id_table++) {
468 /* check if device's identifiers match the driver's ones */
469 if (id_table->vendor_id != dev->id.vendor_id &&
470 id_table->vendor_id != PCI_ANY_ID)
472 if (id_table->device_id != dev->id.device_id &&
473 id_table->device_id != PCI_ANY_ID)
475 if (id_table->subsystem_vendor_id != dev->id.subsystem_vendor_id &&
476 id_table->subsystem_vendor_id != PCI_ANY_ID)
478 if (id_table->subsystem_device_id != dev->id.subsystem_device_id &&
479 id_table->subsystem_device_id != PCI_ANY_ID)
482 struct rte_pci_addr *loc = &dev->addr;
484 RTE_LOG(DEBUG, EAL, "PCI device "PCI_PRI_FMT" on NUMA socket %i\n",
485 loc->domain, loc->bus, loc->devid, loc->function,
488 RTE_LOG(DEBUG, EAL, " probe driver: %x:%x %s\n", dev->id.vendor_id,
489 dev->id.device_id, dr->name);
491 /* no initialization when blacklisted, return without error */
492 if (dev->devargs != NULL &&
493 dev->devargs->type == RTE_DEVTYPE_BLACKLISTED_PCI) {
495 RTE_LOG(DEBUG, EAL, " Device is blacklisted, not initializing\n");
499 if (dr->drv_flags & RTE_PCI_DRV_NEED_MAPPING) {
500 /* map resources for devices that use igb_uio */
501 ret = pci_uio_map_resource(dev);
504 } else if (dr->drv_flags & RTE_PCI_DRV_FORCE_UNBIND &&
505 rte_eal_process_type() == RTE_PROC_PRIMARY) {
506 /* unbind current driver */
507 if (pci_unbind_kernel_driver(dev) < 0)
511 /* reference driver structure */
514 /* call the driver devinit() function */
515 return dr->devinit(dr, dev);
517 /* return positive value if driver is not found */
521 /* Init the PCI EAL subsystem */
523 rte_eal_pci_init(void)
525 TAILQ_INIT(&pci_driver_list);
526 TAILQ_INIT(&pci_device_list);
528 /* for debug purposes, PCI can be disabled */
529 if (internal_config.no_pci)
532 if (pci_scan() < 0) {
533 RTE_LOG(ERR, EAL, "%s(): Cannot scan PCI bus\n", __func__);