1/*-
2 *   BSD LICENSE
3 *
4 *   Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
5 *   All rights reserved.
6 *
7 *   Redistribution and use in source and binary forms, with or without
8 *   modification, are permitted provided that the following conditions
9 *   are met:
10 *
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
16 *       distribution.
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.
20 *
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.
32 */
33/*   BSD LICENSE
34 *
35 *   Copyright 2013-2014 6WIND S.A.
36 *
37 *   Redistribution and use in source and binary forms, with or without
38 *   modification, are permitted provided that the following conditions
39 *   are met:
40 *
41 *     * Redistributions of source code must retain the above copyright
42 *       notice, this list of conditions and the following disclaimer.
43 *     * Redistributions in binary form must reproduce the above copyright
44 *       notice, this list of conditions and the following disclaimer in
45 *       the documentation and/or other materials provided with the
46 *       distribution.
47 *     * Neither the name of 6WIND S.A. nor the names of its
48 *       contributors may be used to endorse or promote products derived
49 *       from this software without specific prior written permission.
50 *
51 *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
52 *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
53 *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
54 *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
55 *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
56 *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
57 *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
58 *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
59 *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
60 *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
61 *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
62 */
63
64#include <string.h>
65#include <inttypes.h>
66#include <stdint.h>
67#include <stdlib.h>
68#include <stdio.h>
69#include <sys/queue.h>
70#include <sys/mman.h>
71
72#include <rte_interrupts.h>
73#include <rte_log.h>
74#include <rte_pci.h>
75#include <rte_per_lcore.h>
76#include <rte_memory.h>
77#include <rte_memzone.h>
78#include <rte_eal.h>
79#include <rte_string_fns.h>
80#include <rte_common.h>
81#include <rte_devargs.h>
82
83#include "eal_private.h"
84
85struct pci_driver_list pci_driver_list =
86	TAILQ_HEAD_INITIALIZER(pci_driver_list);
87struct pci_device_list pci_device_list =
88	TAILQ_HEAD_INITIALIZER(pci_device_list);
89
90#define SYSFS_PCI_DEVICES "/sys/bus/pci/devices"
91
92const char *pci_get_sysfs_path(void)
93{
94	const char *path = NULL;
95
96	path = getenv("SYSFS_PCI_DEVICES");
97	if (path == NULL)
98		return SYSFS_PCI_DEVICES;
99
100	return path;
101}
102
103static struct rte_devargs *pci_devargs_lookup(struct rte_pci_device *dev)
104{
105	struct rte_devargs *devargs;
106
107	TAILQ_FOREACH(devargs, &devargs_list, next) {
108		if (devargs->type != RTE_DEVTYPE_BLACKLISTED_PCI &&
109			devargs->type != RTE_DEVTYPE_WHITELISTED_PCI)
110			continue;
111		if (!rte_eal_compare_pci_addr(&dev->addr, &devargs->pci.addr))
112			return devargs;
113	}
114	return NULL;
115}
116
117/* map a particular resource from a file */
118void *
119pci_map_resource(void *requested_addr, int fd, off_t offset, size_t size,
120		 int additional_flags)
121{
122	void *mapaddr;
123
124	/* Map the PCI memory resource of device */
125	mapaddr = mmap(requested_addr, size, PROT_READ | PROT_WRITE,
126			MAP_SHARED | additional_flags, fd, offset);
127	if (mapaddr == MAP_FAILED) {
128		RTE_LOG(ERR, EAL, "%s(): cannot mmap(%d, %p, 0x%lx, 0x%lx): %s (%p)\n",
129			__func__, fd, requested_addr,
130			(unsigned long)size, (unsigned long)offset,
131			strerror(errno), mapaddr);
132	} else
133		RTE_LOG(DEBUG, EAL, "  PCI memory mapped at %p\n", mapaddr);
134
135	return mapaddr;
136}
137
138/* unmap a particular resource */
139void
140pci_unmap_resource(void *requested_addr, size_t size)
141{
142	if (requested_addr == NULL)
143		return;
144
145	/* Unmap the PCI memory resource of device */
146	if (munmap(requested_addr, size)) {
147		RTE_LOG(ERR, EAL, "%s(): cannot munmap(%p, 0x%lx): %s\n",
148			__func__, requested_addr, (unsigned long)size,
149			strerror(errno));
150	} else
151		RTE_LOG(DEBUG, EAL, "  PCI memory unmapped at %p\n",
152				requested_addr);
153}
154
155/*
156 * If vendor/device ID match, call the probe() function of the
157 * driver.
158 */
159static int
160rte_eal_pci_probe_one_driver(struct rte_pci_driver *dr, struct rte_pci_device *dev)
161{
162	int ret;
163	const struct rte_pci_id *id_table;
164
165	for (id_table = dr->id_table; id_table->vendor_id != 0; id_table++) {
166
167		/* check if device's identifiers match the driver's ones */
168		if (id_table->vendor_id != dev->id.vendor_id &&
169				id_table->vendor_id != PCI_ANY_ID)
170			continue;
171		if (id_table->device_id != dev->id.device_id &&
172				id_table->device_id != PCI_ANY_ID)
173			continue;
174		if (id_table->subsystem_vendor_id != dev->id.subsystem_vendor_id &&
175				id_table->subsystem_vendor_id != PCI_ANY_ID)
176			continue;
177		if (id_table->subsystem_device_id != dev->id.subsystem_device_id &&
178				id_table->subsystem_device_id != PCI_ANY_ID)
179			continue;
180		if (id_table->class_id != dev->id.class_id &&
181				id_table->class_id != RTE_CLASS_ANY_ID)
182			continue;
183
184		struct rte_pci_addr *loc = &dev->addr;
185
186		RTE_LOG(INFO, EAL, "PCI device "PCI_PRI_FMT" on NUMA socket %i\n",
187				loc->domain, loc->bus, loc->devid, loc->function,
188				dev->device.numa_node);
189
190		/* no initialization when blacklisted, return without error */
191		if (dev->device.devargs != NULL &&
192			dev->device.devargs->type ==
193				RTE_DEVTYPE_BLACKLISTED_PCI) {
194			RTE_LOG(INFO, EAL, "  Device is blacklisted, not initializing\n");
195			return 1;
196		}
197
198		RTE_LOG(INFO, EAL, "  probe driver: %x:%x %s\n", dev->id.vendor_id,
199				dev->id.device_id, dr->driver.name);
200
201		if (dr->drv_flags & RTE_PCI_DRV_NEED_MAPPING) {
202			/* map resources for devices that use igb_uio */
203			ret = rte_eal_pci_map_device(dev);
204			if (ret != 0)
205				return ret;
206		} else if (dr->drv_flags & RTE_PCI_DRV_FORCE_UNBIND &&
207				rte_eal_process_type() == RTE_PROC_PRIMARY) {
208			/* unbind current driver */
209			if (pci_unbind_kernel_driver(dev) < 0)
210				return -1;
211		}
212
213		/* reference driver structure */
214		dev->driver = dr;
215
216		/* call the driver probe() function */
217		ret = dr->probe(dr, dev);
218		if (ret)
219			dev->driver = NULL;
220
221		return ret;
222	}
223	/* return positive value if driver doesn't support this device */
224	return 1;
225}
226
227/*
228 * If vendor/device ID match, call the remove() function of the
229 * driver.
230 */
231static int
232rte_eal_pci_detach_dev(struct rte_pci_driver *dr,
233		struct rte_pci_device *dev)
234{
235	const struct rte_pci_id *id_table;
236
237	if ((dr == NULL) || (dev == NULL))
238		return -EINVAL;
239
240	for (id_table = dr->id_table; id_table->vendor_id != 0; id_table++) {
241
242		/* check if device's identifiers match the driver's ones */
243		if (id_table->vendor_id != dev->id.vendor_id &&
244				id_table->vendor_id != PCI_ANY_ID)
245			continue;
246		if (id_table->device_id != dev->id.device_id &&
247				id_table->device_id != PCI_ANY_ID)
248			continue;
249		if (id_table->subsystem_vendor_id != dev->id.subsystem_vendor_id &&
250				id_table->subsystem_vendor_id != PCI_ANY_ID)
251			continue;
252		if (id_table->subsystem_device_id != dev->id.subsystem_device_id &&
253				id_table->subsystem_device_id != PCI_ANY_ID)
254			continue;
255
256		struct rte_pci_addr *loc = &dev->addr;
257
258		RTE_LOG(DEBUG, EAL, "PCI device "PCI_PRI_FMT" on NUMA socket %i\n",
259				loc->domain, loc->bus, loc->devid,
260				loc->function, dev->device.numa_node);
261
262		RTE_LOG(DEBUG, EAL, "  remove driver: %x:%x %s\n", dev->id.vendor_id,
263				dev->id.device_id, dr->driver.name);
264
265		if (dr->remove && (dr->remove(dev) < 0))
266			return -1;	/* negative value is an error */
267
268		/* clear driver structure */
269		dev->driver = NULL;
270
271		if (dr->drv_flags & RTE_PCI_DRV_NEED_MAPPING)
272			/* unmap resources for devices that use igb_uio */
273			rte_eal_pci_unmap_device(dev);
274
275		return 0;
276	}
277
278	/* return positive value if driver doesn't support this device */
279	return 1;
280}
281
282/*
283 * If vendor/device ID match, call the probe() function of all
284 * registered driver for the given device. Return -1 if initialization
285 * failed, return 1 if no driver is found for this device.
286 */
287static int
288pci_probe_all_drivers(struct rte_pci_device *dev)
289{
290	struct rte_pci_driver *dr = NULL;
291	int rc = 0;
292
293	if (dev == NULL)
294		return -1;
295
296	/* Check if a driver is already loaded */
297	if (dev->driver != NULL)
298		return 0;
299
300	TAILQ_FOREACH(dr, &pci_driver_list, next) {
301		rc = rte_eal_pci_probe_one_driver(dr, dev);
302		if (rc < 0)
303			/* negative value is an error */
304			return -1;
305		if (rc > 0)
306			/* positive value means driver doesn't support it */
307			continue;
308		return 0;
309	}
310	return 1;
311}
312
313/*
314 * If vendor/device ID match, call the remove() function of all
315 * registered driver for the given device. Return -1 if initialization
316 * failed, return 1 if no driver is found for this device.
317 */
318static int
319pci_detach_all_drivers(struct rte_pci_device *dev)
320{
321	struct rte_pci_driver *dr = NULL;
322	int rc = 0;
323
324	if (dev == NULL)
325		return -1;
326
327	TAILQ_FOREACH(dr, &pci_driver_list, next) {
328		rc = rte_eal_pci_detach_dev(dr, dev);
329		if (rc < 0)
330			/* negative value is an error */
331			return -1;
332		if (rc > 0)
333			/* positive value means driver doesn't support it */
334			continue;
335		return 0;
336	}
337	return 1;
338}
339
340/*
341 * Find the pci device specified by pci address, then invoke probe function of
342 * the driver of the devive.
343 */
344int
345rte_eal_pci_probe_one(const struct rte_pci_addr *addr)
346{
347	struct rte_pci_device *dev = NULL;
348	int ret = 0;
349
350	if (addr == NULL)
351		return -1;
352
353	/* update current pci device in global list, kernel bindings might have
354	 * changed since last time we looked at it.
355	 */
356	if (pci_update_device(addr) < 0)
357		goto err_return;
358
359	TAILQ_FOREACH(dev, &pci_device_list, next) {
360		if (rte_eal_compare_pci_addr(&dev->addr, addr))
361			continue;
362
363		ret = pci_probe_all_drivers(dev);
364		if (ret)
365			goto err_return;
366		return 0;
367	}
368	return -1;
369
370err_return:
371	RTE_LOG(WARNING, EAL,
372		"Requested device " PCI_PRI_FMT " cannot be used\n",
373		addr->domain, addr->bus, addr->devid, addr->function);
374	return -1;
375}
376
377/*
378 * Detach device specified by its pci address.
379 */
380int
381rte_eal_pci_detach(const struct rte_pci_addr *addr)
382{
383	struct rte_pci_device *dev = NULL;
384	int ret = 0;
385
386	if (addr == NULL)
387		return -1;
388
389	TAILQ_FOREACH(dev, &pci_device_list, next) {
390		if (rte_eal_compare_pci_addr(&dev->addr, addr))
391			continue;
392
393		ret = pci_detach_all_drivers(dev);
394		if (ret < 0)
395			goto err_return;
396
397		TAILQ_REMOVE(&pci_device_list, dev, next);
398		free(dev);
399		return 0;
400	}
401	return -1;
402
403err_return:
404	RTE_LOG(WARNING, EAL, "Requested device " PCI_PRI_FMT
405			" cannot be used\n", dev->addr.domain, dev->addr.bus,
406			dev->addr.devid, dev->addr.function);
407	return -1;
408}
409
410/*
411 * Scan the content of the PCI bus, and call the probe() function for
412 * all registered drivers that have a matching entry in its id_table
413 * for discovered devices.
414 */
415int
416rte_eal_pci_probe(void)
417{
418	struct rte_pci_device *dev = NULL;
419	struct rte_devargs *devargs;
420	int probe_all = 0;
421	int ret = 0;
422
423	if (rte_eal_devargs_type_count(RTE_DEVTYPE_WHITELISTED_PCI) == 0)
424		probe_all = 1;
425
426	TAILQ_FOREACH(dev, &pci_device_list, next) {
427
428		/* set devargs in PCI structure */
429		devargs = pci_devargs_lookup(dev);
430		if (devargs != NULL)
431			dev->device.devargs = devargs;
432
433		/* probe all or only whitelisted devices */
434		if (probe_all)
435			ret = pci_probe_all_drivers(dev);
436		else if (devargs != NULL &&
437			devargs->type == RTE_DEVTYPE_WHITELISTED_PCI)
438			ret = pci_probe_all_drivers(dev);
439		if (ret < 0)
440			rte_exit(EXIT_FAILURE, "Requested device " PCI_PRI_FMT
441				 " cannot be used\n", dev->addr.domain, dev->addr.bus,
442				 dev->addr.devid, dev->addr.function);
443	}
444
445	return 0;
446}
447
448/* dump one device */
449static int
450pci_dump_one_device(FILE *f, struct rte_pci_device *dev)
451{
452	int i;
453
454	fprintf(f, PCI_PRI_FMT, dev->addr.domain, dev->addr.bus,
455	       dev->addr.devid, dev->addr.function);
456	fprintf(f, " - vendor:%x device:%x\n", dev->id.vendor_id,
457	       dev->id.device_id);
458
459	for (i = 0; i != sizeof(dev->mem_resource) /
460		sizeof(dev->mem_resource[0]); i++) {
461		fprintf(f, "   %16.16"PRIx64" %16.16"PRIx64"\n",
462			dev->mem_resource[i].phys_addr,
463			dev->mem_resource[i].len);
464	}
465	return 0;
466}
467
468/* dump devices on the bus */
469void
470rte_eal_pci_dump(FILE *f)
471{
472	struct rte_pci_device *dev = NULL;
473
474	TAILQ_FOREACH(dev, &pci_device_list, next) {
475		pci_dump_one_device(f, dev);
476	}
477}
478
479/* register a driver */
480void
481rte_eal_pci_register(struct rte_pci_driver *driver)
482{
483	TAILQ_INSERT_TAIL(&pci_driver_list, driver, next);
484	rte_eal_driver_register(&driver->driver);
485}
486
487/* unregister a driver */
488void
489rte_eal_pci_unregister(struct rte_pci_driver *driver)
490{
491	rte_eal_driver_unregister(&driver->driver);
492	TAILQ_REMOVE(&pci_driver_list, driver, next);
493}
494