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.
43 #include <sys/types.h>
45 #include <sys/queue.h>
50 #include <sys/ioctl.h>
54 #include <rte_memory.h>
55 #include <rte_memzone.h>
56 #include <rte_launch.h>
57 #include <rte_tailq.h>
59 #include <rte_eal_memconfig.h>
60 #include <rte_per_lcore.h>
61 #include <rte_lcore.h>
62 #include <rte_common.h>
63 #include <rte_string_fns.h>
65 #include "eal_private.h"
66 #include "eal_internal_cfg.h"
67 #include "eal_filesystem.h"
68 #include <exec-env/rte_dom0_common.h>
70 #define PAGE_SIZE RTE_PGSIZE_4K
71 #define DEFAUL_DOM0_NAME "dom0-mem"
73 static int xen_fd = -1;
74 static const char sys_dir_path[] = "/sys/kernel/mm/dom0-mm/memsize-mB";
77 * Try to mmap *size bytes in /dev/zero. If it is succesful, return the
78 * pointer to the mmap'd area and keep *size unmodified. Else, retry
79 * with a smaller zone: decrease *size by mem_size until it reaches
80 * 0. In this case, return NULL. Note: this function returns an address
81 * which is a multiple of mem_size size.
84 xen_get_virtual_area(size_t *size, size_t mem_size)
90 RTE_LOG(INFO, EAL, "Ask a virtual area of 0x%zu bytes\n", *size);
92 fd = open("/dev/zero", O_RDONLY);
94 RTE_LOG(ERR, EAL, "Cannot open /dev/zero\n");
98 addr = mmap(NULL, (*size) + mem_size, PROT_READ,
100 if (addr == MAP_FAILED)
102 } while (addr == MAP_FAILED && *size > 0);
104 if (addr == MAP_FAILED) {
106 RTE_LOG(INFO, EAL, "Cannot get a virtual area\n");
110 munmap(addr, (*size) + mem_size);
113 /* align addr to a mem_size boundary */
114 aligned_addr = (uintptr_t)addr;
115 aligned_addr = RTE_ALIGN_CEIL(aligned_addr, mem_size);
116 addr = (void *)(aligned_addr);
118 RTE_LOG(INFO, EAL, "Virtual area found at %p (size = 0x%zx)\n",
125 * Get memory size configuration from /sys/devices/virtual/misc/dom0_mm
126 * /memsize-mB/memsize file, and the size unit is mB.
129 get_xen_memory_size(void)
132 unsigned long mem_size = 0;
133 static const char *file_name;
135 file_name = "memsize";
136 rte_snprintf(path, sizeof(path), "%s/%s",
137 sys_dir_path, file_name);
139 if (eal_parse_sysfs_value(path, &mem_size) < 0)
143 rte_exit(EXIT_FAILURE,"XEN-DOM0:the %s/%s was not"
144 " configured.\n",sys_dir_path, file_name);
146 rte_exit(EXIT_FAILURE,"XEN-DOM0:the %s/%s must be"
147 " even number.\n",sys_dir_path, file_name);
149 if (mem_size > DOM0_CONFIG_MEMSIZE)
150 rte_exit(EXIT_FAILURE,"XEN-DOM0:the %s/%s should not be larger"
151 " than %d mB\n",sys_dir_path, file_name, DOM0_CONFIG_MEMSIZE);
157 * Based on physical address to caculate MFN in Xen Dom0.
160 rte_mem_phy2mch(uint32_t memseg_id, const phys_addr_t phy_addr)
163 uint64_t mfn, mfn_offset;
164 struct rte_mem_config *mcfg = rte_eal_get_configuration()->mem_config;
165 struct rte_memseg *memseg = mcfg->memseg;
167 mfn_id = (phy_addr - memseg[memseg_id].phys_addr) / RTE_PGSIZE_2M;
169 /*the MFN is contiguous in 2M */
170 mfn_offset = (phy_addr - memseg[memseg_id].phys_addr) %
171 RTE_PGSIZE_2M / PAGE_SIZE;
172 mfn = mfn_offset + memseg[memseg_id].mfn[mfn_id];
174 /** return mechine address */
175 return (mfn * PAGE_SIZE + phy_addr % PAGE_SIZE);
179 rte_xen_dom0_memory_init(void)
181 void *vir_addr, *vma_addr = NULL;
183 uint32_t i, requested, mem_size, memseg_idx, num_memseg = 0;
185 struct memory_info meminfo;
186 struct memseg_info seginfo[RTE_MAX_MEMSEG];
187 int flags, page_size = getpagesize();
188 struct rte_mem_config *mcfg = rte_eal_get_configuration()->mem_config;
189 struct rte_memseg *memseg = mcfg->memseg;
190 uint64_t total_mem = internal_config.memory;
192 memset(seginfo, 0, sizeof(seginfo));
193 memset(&meminfo, 0, sizeof(struct memory_info));
195 mem_size = get_xen_memory_size();
196 requested = (unsigned) (total_mem / 0x100000);
197 if (requested > mem_size)
198 /* if we didn't satisfy total memory requirements */
199 rte_exit(EXIT_FAILURE,"Not enough memory available! Requested: %uMB,"
200 " available: %uMB\n", requested, mem_size);
201 else if (total_mem != 0)
202 mem_size = requested;
204 /* Check FD and open once */
206 xen_fd = open(DOM0_MM_DEV, O_RDWR);
208 RTE_LOG(ERR, EAL, "Can not open %s\n",DOM0_MM_DEV);
213 meminfo.size = mem_size;
215 /* construct memory mangement name for Dom0 */
216 rte_snprintf(meminfo.name, DOM0_NAME_MAX, "%s-%s",
217 internal_config.hugefile_prefix, DEFAUL_DOM0_NAME);
219 /* Notify kernel driver to allocate memory */
220 ret = ioctl(xen_fd, RTE_DOM0_IOCTL_PREPARE_MEMSEG, &meminfo);
222 RTE_LOG(ERR, EAL, "XEN DOM0:failed to get memory\n");
227 /* Get number of memory segment from driver */
228 ret = ioctl(xen_fd, RTE_DOM0_IOCTL_GET_NUM_MEMSEG, &num_memseg);
230 RTE_LOG(ERR, EAL, "XEN DOM0:failed to get memseg count.\n");
235 if(num_memseg > RTE_MAX_MEMSEG){
236 RTE_LOG(ERR, EAL, "XEN DOM0: the memseg count %d is greater"
237 " than max memseg %d.\n",num_memseg, RTE_MAX_MEMSEG);
242 /* get all memory segements information */
243 ret = ioctl(xen_fd, RTE_DOM0_IOCTL_GET_MEMSEG_INFO, seginfo);
245 RTE_LOG(ERR, EAL, "XEN DOM0:failed to get memseg info.\n");
250 /* map all memory segments to contiguous user space */
251 for (memseg_idx = 0; memseg_idx < num_memseg; memseg_idx++)
253 vma_len = seginfo[memseg_idx].size;
256 * get the biggest virtual memory area up to vma_len. If it fails,
257 * vma_addr is NULL, so let the kernel provide the address.
259 vma_addr = xen_get_virtual_area(&vma_len, RTE_PGSIZE_2M);
260 if (vma_addr == NULL) {
262 vma_len = RTE_PGSIZE_2M;
264 flags = MAP_SHARED | MAP_FIXED;
266 seginfo[memseg_idx].size = vma_len;
267 vir_addr = mmap(vma_addr, seginfo[memseg_idx].size,
268 PROT_READ|PROT_WRITE, flags, xen_fd,
269 memseg_idx * page_size);
270 if (vir_addr == MAP_FAILED) {
271 RTE_LOG(ERR, EAL, "XEN DOM0:Could not mmap %s\n",
277 memseg[memseg_idx].addr = vir_addr;
278 memseg[memseg_idx].phys_addr = page_size *
279 seginfo[memseg_idx].pfn ;
280 memseg[memseg_idx].len = seginfo[memseg_idx].size;
281 for ( i = 0; i < seginfo[memseg_idx].size / RTE_PGSIZE_2M; i++)
282 memseg[memseg_idx].mfn[i] = seginfo[memseg_idx].mfn[i];
284 /* MFNs are continuous in 2M, so assume that page size is 2M */
285 memseg[memseg_idx].hugepage_sz = RTE_PGSIZE_2M;
287 memseg[memseg_idx].nchannel = mcfg->nchannel;
288 memseg[memseg_idx].nrank = mcfg->nrank;
290 /* NUMA is not suppoted in Xen Dom0, so only set socket 0*/
291 memseg[memseg_idx].socket_id = 0;
304 * This creates the memory mappings in the secondary process to match that of
305 * the server process. It goes through each memory segment in the DPDK runtime
306 * configuration, mapping them in order to form a contiguous block in the
307 * virtual memory space
310 rte_xen_dom0_memory_attach(void)
312 const struct rte_mem_config *mcfg;
313 unsigned s = 0; /* s used to track the segment number */
317 char name[DOM0_NAME_MAX] = {0};
318 int page_size = getpagesize();
320 mcfg = rte_eal_get_configuration()->mem_config;
322 /* Check FD and open once */
324 xen_fd = open(DOM0_MM_DEV, O_RDWR);
326 RTE_LOG(ERR, EAL, "Can not open %s\n",DOM0_MM_DEV);
331 /* construct memory mangement name for Dom0 */
332 rte_snprintf(name, DOM0_NAME_MAX, "%s-%s",
333 internal_config.hugefile_prefix, DEFAUL_DOM0_NAME);
334 /* attach to memory segments of primary process */
335 ret = ioctl(xen_fd, RTE_DOM0_IOCTL_ATTACH_TO_MEMSEG, name);
337 RTE_LOG(ERR, EAL,"attach memory segments fail.\n");
341 /* map all segments into memory to make sure we get the addrs */
342 for (s = 0; s < RTE_MAX_MEMSEG; ++s) {
345 * the first memory segment with len==0 is the one that
346 * follows the last valid segment.
348 if (mcfg->memseg[s].len == 0)
351 vir_addr = mmap(mcfg->memseg[s].addr, mcfg->memseg[s].len,
352 PROT_READ|PROT_WRITE, MAP_SHARED|MAP_FIXED, xen_fd,
354 if (vir_addr == MAP_FAILED) {
355 RTE_LOG(ERR, EAL, "Could not mmap %llu bytes "
356 "in %s to requested address [%p]\n",
357 (unsigned long long)mcfg->memseg[s].len, DOM0_MM_DEV,
358 mcfg->memseg[s].addr);