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 "eal_filesystem.h"
71 #include "eal_private.h"
75 * PCI probing under linux
77 * This code is used to simulate a PCI probe by parsing information in
78 * sysfs. Moreover, when a registered driver matches a device, the
79 * kernel driver currently using it is unloaded and replaced by
80 * igb_uio module, which is a very minimal userland driver for Intel
81 * network card, only providing access to PCI BAR to applications, and
82 * enabling bus master.
85 /* unbind kernel driver for this device */
87 pci_unbind_kernel_driver(struct rte_pci_device *dev __rte_unused)
89 RTE_LOG(ERR, EAL, "RTE_PCI_DRV_FORCE_UNBIND flag is not implemented "
96 pci_map_device(struct rte_pci_device *dev)
100 /* try mapping the NIC resources */
102 case RTE_KDRV_NIC_UIO:
103 /* map resources for devices that use uio */
104 ret = pci_uio_map_resource(dev);
108 " Not managed by a supported kernel driver, skipped\n");
116 /* Unmap pci device */
118 pci_unmap_device(struct rte_pci_device *dev)
120 /* try unmapping the NIC resources */
122 case RTE_KDRV_NIC_UIO:
123 /* unmap resources for devices that use uio */
124 pci_uio_unmap_resource(dev);
128 " Not managed by a supported kernel driver, skipped\n");
134 pci_uio_free_resource(struct rte_pci_device *dev,
135 struct mapped_pci_resource *uio_res)
139 if (dev->intr_handle.fd) {
140 close(dev->intr_handle.fd);
141 dev->intr_handle.fd = -1;
142 dev->intr_handle.type = RTE_INTR_HANDLE_UNKNOWN;
147 pci_uio_alloc_resource(struct rte_pci_device *dev,
148 struct mapped_pci_resource **uio_res)
150 char devname[PATH_MAX]; /* contains the /dev/uioX */
151 struct rte_pci_addr *loc;
155 snprintf(devname, sizeof(devname), "/dev/uio@pci:%u:%u:%u",
156 dev->addr.bus, dev->addr.devid, dev->addr.function);
158 if (access(devname, O_RDWR) < 0) {
159 RTE_LOG(WARNING, EAL, " "PCI_PRI_FMT" not managed by UIO driver, "
160 "skipping\n", loc->domain, loc->bus, loc->devid, loc->function);
164 /* save fd if in primary process */
165 dev->intr_handle.fd = open(devname, O_RDWR);
166 if (dev->intr_handle.fd < 0) {
167 RTE_LOG(ERR, EAL, "Cannot open %s: %s\n",
168 devname, strerror(errno));
171 dev->intr_handle.type = RTE_INTR_HANDLE_UIO;
173 /* allocate the mapping details for secondary processes*/
174 *uio_res = rte_zmalloc("UIO_RES", sizeof(**uio_res), 0);
175 if (*uio_res == NULL) {
177 "%s(): cannot store uio mmap details\n", __func__);
181 snprintf((*uio_res)->path, sizeof((*uio_res)->path), "%s", devname);
182 memcpy(&(*uio_res)->pci_addr, &dev->addr, sizeof((*uio_res)->pci_addr));
187 pci_uio_free_resource(dev, *uio_res);
192 pci_uio_map_resource_by_index(struct rte_pci_device *dev, int res_idx,
193 struct mapped_pci_resource *uio_res, int map_idx)
200 struct pci_map *maps;
202 maps = uio_res->maps;
203 devname = uio_res->path;
204 pagesz = sysconf(_SC_PAGESIZE);
206 /* allocate memory to keep path */
207 maps[map_idx].path = rte_malloc(NULL, strlen(devname) + 1, 0);
208 if (maps[map_idx].path == NULL) {
209 RTE_LOG(ERR, EAL, "Cannot allocate memory for path: %s\n",
215 * open resource file, to mmap it
217 fd = open(devname, O_RDWR);
219 RTE_LOG(ERR, EAL, "Cannot open %s: %s\n",
220 devname, strerror(errno));
224 /* if matching map is found, then use it */
225 offset = res_idx * pagesz;
226 mapaddr = pci_map_resource(NULL, fd, (off_t)offset,
227 (size_t)dev->mem_resource[res_idx].len, 0);
229 if (mapaddr == MAP_FAILED)
232 maps[map_idx].phaddr = dev->mem_resource[res_idx].phys_addr;
233 maps[map_idx].size = dev->mem_resource[res_idx].len;
234 maps[map_idx].addr = mapaddr;
235 maps[map_idx].offset = offset;
236 strcpy(maps[map_idx].path, devname);
237 dev->mem_resource[res_idx].addr = mapaddr;
242 rte_free(maps[map_idx].path);
247 pci_scan_one(int dev_pci_fd, struct pci_conf *conf)
249 struct rte_pci_device *dev;
250 struct pci_bar_io bar;
253 dev = malloc(sizeof(*dev));
258 memset(dev, 0, sizeof(*dev));
259 dev->addr.domain = conf->pc_sel.pc_domain;
260 dev->addr.bus = conf->pc_sel.pc_bus;
261 dev->addr.devid = conf->pc_sel.pc_dev;
262 dev->addr.function = conf->pc_sel.pc_func;
265 dev->id.vendor_id = conf->pc_vendor;
268 dev->id.device_id = conf->pc_device;
270 /* get subsystem_vendor id */
271 dev->id.subsystem_vendor_id = conf->pc_subvendor;
273 /* get subsystem_device id */
274 dev->id.subsystem_device_id = conf->pc_subdevice;
276 /* TODO: get max_vfs */
279 /* FreeBSD has no NUMA support (yet) */
282 /* FreeBSD has only one pass through driver */
283 dev->kdrv = RTE_KDRV_NIC_UIO;
285 /* parse resources */
286 switch (conf->pc_hdr & PCIM_HDRTYPE) {
287 case PCIM_HDRTYPE_NORMAL:
288 max = PCIR_MAX_BAR_0;
290 case PCIM_HDRTYPE_BRIDGE:
291 max = PCIR_MAX_BAR_1;
293 case PCIM_HDRTYPE_CARDBUS:
294 max = PCIR_MAX_BAR_2;
300 for (i = 0; i <= max; i++) {
301 bar.pbi_sel = conf->pc_sel;
302 bar.pbi_reg = PCIR_BAR(i);
303 if (ioctl(dev_pci_fd, PCIOCGETBAR, &bar) < 0)
306 dev->mem_resource[i].len = bar.pbi_length;
307 if (PCI_BAR_IO(bar.pbi_base)) {
308 dev->mem_resource[i].addr = (void *)(bar.pbi_base & ~((uint64_t)0xf));
311 dev->mem_resource[i].phys_addr = bar.pbi_base & ~((uint64_t)0xf);
314 /* device is valid, add in list (sorted) */
315 if (TAILQ_EMPTY(&pci_device_list)) {
316 TAILQ_INSERT_TAIL(&pci_device_list, dev, next);
319 struct rte_pci_device *dev2 = NULL;
322 TAILQ_FOREACH(dev2, &pci_device_list, next) {
323 ret = rte_eal_compare_pci_addr(&dev->addr, &dev2->addr);
327 TAILQ_INSERT_BEFORE(dev2, dev, next);
329 } else { /* already registered */
330 dev2->kdrv = dev->kdrv;
331 dev2->max_vfs = dev->max_vfs;
332 memmove(dev2->mem_resource,
334 sizeof(dev->mem_resource));
339 TAILQ_INSERT_TAIL(&pci_device_list, dev, next);
350 * Scan the content of the PCI bus, and add the devices in the devices
351 * list. Call pci_scan_one() for each pci entry found.
354 rte_eal_pci_scan(void)
357 unsigned dev_count = 0;
358 struct pci_conf matches[16];
359 struct pci_conf_io conf_io = {
363 .match_buf_len = sizeof(matches),
364 .matches = &matches[0],
367 fd = open("/dev/pci", O_RDONLY);
369 RTE_LOG(ERR, EAL, "%s(): error opening /dev/pci\n", __func__);
375 if (ioctl(fd, PCIOCGETCONF, &conf_io) < 0) {
376 RTE_LOG(ERR, EAL, "%s(): error with ioctl on /dev/pci: %s\n",
377 __func__, strerror(errno));
381 for (i = 0; i < conf_io.num_matches; i++)
382 if (pci_scan_one(fd, &matches[i]) < 0)
385 dev_count += conf_io.num_matches;
386 } while(conf_io.status == PCI_GETCONF_MORE_DEVS);
390 RTE_LOG(ERR, EAL, "PCI scan found %u devices\n", dev_count);
399 /* Read PCI config space. */
400 int rte_eal_pci_read_config(const struct rte_pci_device *dev,
401 void *buf, size_t len, off_t offset)
406 .pc_domain = dev->addr.domain,
407 .pc_bus = dev->addr.bus,
408 .pc_dev = dev->addr.devid,
409 .pc_func = dev->addr.function,
415 if (len == 3 || len > sizeof(pi.pi_data)) {
416 RTE_LOG(ERR, EAL, "%s(): invalid pci read length\n", __func__);
420 fd = open("/dev/pci", O_RDONLY);
422 RTE_LOG(ERR, EAL, "%s(): error opening /dev/pci\n", __func__);
426 if (ioctl(fd, PCIOCREAD, &pi) < 0)
430 memcpy(buf, &pi.pi_data, len);
439 /* Write PCI config space. */
440 int rte_eal_pci_write_config(const struct rte_pci_device *dev,
441 const void *buf, size_t len, off_t offset)
447 .pc_domain = dev->addr.domain,
448 .pc_bus = dev->addr.bus,
449 .pc_dev = dev->addr.devid,
450 .pc_func = dev->addr.function,
453 .pi_data = *(const uint32_t *)buf,
457 if (len == 3 || len > sizeof(pi.pi_data)) {
458 RTE_LOG(ERR, EAL, "%s(): invalid pci read length\n", __func__);
462 memcpy(&pi.pi_data, buf, len);
464 fd = open("/dev/pci", O_RDONLY);
466 RTE_LOG(ERR, EAL, "%s(): error opening /dev/pci\n", __func__);
470 if (ioctl(fd, PCIOCWRITE, &pi) < 0)
482 /* Init the PCI EAL subsystem */
484 rte_eal_pci_init(void)
486 TAILQ_INIT(&pci_driver_list);
487 TAILQ_INIT(&pci_device_list);
489 /* for debug purposes, PCI can be disabled */
490 if (internal_config.no_pci)
493 if (rte_eal_pci_scan() < 0) {
494 RTE_LOG(ERR, EAL, "%s(): Cannot scan PCI bus\n", __func__);