4 * Copyright(c) 2010-2013 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.
36 #include <sys/types.h>
47 #include <sys/queue.h>
49 #include <rte_memory.h>
50 #include <rte_memzone.h>
51 #include <rte_tailq.h>
53 #include <rte_launch.h>
54 #include <rte_per_lcore.h>
55 #include <rte_lcore.h>
56 #include <rte_debug.h>
58 #include <rte_common.h>
59 #include "rte_string_fns.h"
60 #include "eal_internal_cfg.h"
61 #include "eal_hugepages.h"
62 #include "eal_filesystem.h"
64 static const char sys_dir_path[] = "/sys/kernel/mm/hugepages";
67 get_num_hugepages(const char *subdir)
70 long unsigned num_pages = 0;
71 const char *nr_hp_file;
73 /* if secondary process, just look at the number of hugepages,
74 * otherwise look at number of free hugepages */
75 if (internal_config.process_type == RTE_PROC_SECONDARY)
76 nr_hp_file = "nr_hugepages";
78 nr_hp_file = "free_hugepages";
80 rte_snprintf(path, sizeof(path), "%s/%s/%s",
81 sys_dir_path, subdir, nr_hp_file);
83 if (eal_parse_sysfs_value(path, &num_pages) < 0)
87 RTE_LOG(WARNING, EAL, "No free hugepages reported in %s\n",
90 return (int32_t)num_pages;
94 get_default_hp_size(void)
96 const char proc_meminfo[] = "/proc/meminfo";
97 const char str_hugepagesz[] = "Hugepagesize:";
98 unsigned hugepagesz_len = sizeof(str_hugepagesz) - 1;
100 unsigned long long size = 0;
102 FILE *fd = fopen(proc_meminfo, "r");
104 rte_panic("Cannot open %s\n", proc_meminfo);
105 while(fgets(buffer, sizeof(buffer), fd)){
106 if (strncmp(buffer, str_hugepagesz, hugepagesz_len) == 0){
107 size = rte_str_to_size(&buffer[hugepagesz_len]);
113 rte_panic("Cannot get default hugepage size from %s\n", proc_meminfo);
118 get_hugepage_dir(uint64_t hugepage_sz)
120 enum proc_mount_fieldnames {
127 static uint64_t default_size = 0;
128 const char proc_mounts[] = "/proc/mounts";
129 const char hugetlbfs_str[] = "hugetlbfs";
130 const size_t htlbfs_str_len = sizeof(hugetlbfs_str) - 1;
131 const char pagesize_opt[] = "pagesize=";
132 const size_t pagesize_opt_len = sizeof(pagesize_opt) - 1;
133 const char split_tok = ' ';
134 char *splitstr[_FIELDNAME_MAX];
138 FILE *fd = fopen(proc_mounts, "r");
140 rte_panic("Cannot open %s\n", proc_mounts);
142 if (default_size == 0)
143 default_size = get_default_hp_size();
145 while (fgets(buf, sizeof(buf), fd)){
146 if (rte_strsplit(buf, sizeof(buf), splitstr, _FIELDNAME_MAX,
147 split_tok) != _FIELDNAME_MAX) {
148 RTE_LOG(ERR, EAL, "Error parsing %s\n", proc_mounts);
149 break; /* return NULL */
152 /* we have a specified --huge-dir option, only examine that dir */
153 if (internal_config.hugepage_dir != NULL &&
154 strcmp(splitstr[MOUNTPT], internal_config.hugepage_dir) != 0)
157 if (strncmp(splitstr[FSTYPE], hugetlbfs_str, htlbfs_str_len) == 0){
158 const char *pagesz_str = strstr(splitstr[OPTIONS], pagesize_opt);
160 /* if no explicit page size, the default page size is compared */
161 if (pagesz_str == NULL){
162 if (hugepage_sz == default_size){
163 retval = strdup(splitstr[MOUNTPT]);
167 /* there is an explicit page size, so check it */
169 uint64_t pagesz = rte_str_to_size(&pagesz_str[pagesize_opt_len]);
170 if (pagesz == hugepage_sz) {
171 retval = strdup(splitstr[MOUNTPT]);
175 } /* end if strncmp hugetlbfs */
176 } /* end while fgets */
183 swap_hpi(struct hugepage_info *a, struct hugepage_info *b)
185 char buf[sizeof(*a)];
186 memcpy(buf, a, sizeof(buf));
187 memcpy(a, b, sizeof(buf));
188 memcpy(b, buf, sizeof(buf));
192 * Clear the hugepage directory of whatever hugepage files
193 * there are. Checks if the file is locked (i.e.
194 * if it's in use by another DPDK process).
197 clear_hugedir(const char * hugedir)
200 struct dirent *dirent;
201 int dir_fd, fd, lck_result;
202 const char filter[] = "*map_*"; /* matches hugepage files */
205 dir = opendir(hugedir);
207 RTE_LOG(INFO, EAL, "Unable to open hugepage directory %s\n",
213 dirent = readdir(dir);
215 RTE_LOG(INFO, EAL, "Unable to read hugepage directory %s\n",
220 while(dirent != NULL){
221 /* skip files that don't match the hugepage pattern */
222 if (fnmatch(filter, dirent->d_name, 0) > 0) {
223 dirent = readdir(dir);
227 /* try and lock the file */
228 fd = openat(dir_fd, dirent->d_name, O_RDONLY);
230 /* skip to next file */
232 dirent = readdir(dir);
236 /* non-blocking lock */
237 lck_result = flock(fd, LOCK_EX | LOCK_NB);
239 /* if lock succeeds, unlock and remove the file */
240 if (lck_result != -1) {
242 unlinkat(dir_fd, dirent->d_name, 0);
245 dirent = readdir(dir);
255 RTE_LOG(INFO, EAL, "Error while clearing hugepage dir: %s\n",
262 * when we initialize the hugepage info, everything goes
263 * to socket 0 by default. it will later get sorted by memory
264 * initialization procedure.
267 eal_hugepage_info_init(void)
269 const char dirent_start_text[] = "hugepages-";
270 const size_t dirent_start_len = sizeof(dirent_start_text) - 1;
271 unsigned i, num_sizes = 0;
273 DIR *dir = opendir(sys_dir_path);
275 rte_panic("Cannot open directory %s to read system hugepage info\n",
278 struct dirent *dirent = readdir(dir);
279 while(dirent != NULL){
280 if (strncmp(dirent->d_name, dirent_start_text, dirent_start_len) == 0){
281 struct hugepage_info *hpi = \
282 &internal_config.hugepage_info[num_sizes];
283 hpi->hugepage_sz = rte_str_to_size(&dirent->d_name[dirent_start_len]);
284 hpi->hugedir = get_hugepage_dir(hpi->hugepage_sz);
286 /* first, check if we have a mountpoint */
287 if (hpi->hugedir == NULL){
289 if ((num_pages = get_num_hugepages(dirent->d_name)) > 0)
290 RTE_LOG(INFO, EAL, "%u hugepages of size %llu reserved, "\
291 "but no mounted hugetlbfs found for that size\n",
293 (unsigned long long)hpi->hugepage_sz);
295 /* try to obtain a writelock */
296 hpi->lock_descriptor = open(hpi->hugedir, O_RDONLY);
298 /* if blocking lock failed */
299 if (flock(hpi->lock_descriptor, LOCK_EX) == -1) {
300 RTE_LOG(CRIT, EAL, "Failed to lock hugepage directory!\n");
303 /* clear out the hugepages dir from unused pages */
304 if (clear_hugedir(hpi->hugedir) == -1)
307 /* for now, put all pages into socket 0,
308 * later they will be sorted */
309 hpi->num_pages[0] = get_num_hugepages(dirent->d_name);
311 #ifndef RTE_ARCH_X86_64
312 /* for 32-bit systems, limit number of hugepages to 1GB per page size */
313 hpi->num_pages[0] = RTE_MIN(hpi->num_pages[0],
314 RTE_PGSIZE_1G / hpi->hugepage_sz);
320 dirent = readdir(dir);
323 internal_config.num_hugepage_sizes = num_sizes;
325 /* sort the page directory entries by size, largest to smallest */
326 for (i = 0; i < num_sizes; i++){
328 for (j = i+1; j < num_sizes; j++)
329 if (internal_config.hugepage_info[j-1].hugepage_sz < \
330 internal_config.hugepage_info[j].hugepage_sz)
331 swap_hpi(&internal_config.hugepage_info[j-1],
332 &internal_config.hugepage_info[j]);
335 /* now we have all info, check we have at least one valid size */
336 for (i = 0; i < num_sizes; i++)
337 if (internal_config.hugepage_info[i].hugedir != NULL &&
338 internal_config.hugepage_info[i].num_pages[0] > 0)
341 /* no valid hugepage mounts available, return error */