/*-
* BSD LICENSE
- *
- * Copyright(c) 2010-2012 Intel Corporation. All rights reserved.
+ *
+ * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
* All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+/* BSD LICENSE
+ *
+ * Copyright(c) 2013 6WIND.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
* are met:
- *
- * * Redistributions of source code must retain the above copyright
+ *
+ * * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
- * * Redistributions in binary form must reproduce the above copyright
- * notice, this list of conditions and the following disclaimer in
- * the documentation and/or other materials provided with the
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
* distribution.
- * * Neither the name of Intel Corporation nor the names of its
- * contributors may be used to endorse or promote products derived
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- *
*/
+#define _FILE_OFFSET_BITS 64
#include <errno.h>
#include <stdarg.h>
#include <stdlib.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/queue.h>
-#include <fcntl.h>
+#include <sys/file.h>
#include <unistd.h>
#include <limits.h>
#include <errno.h>
#include <sys/ioctl.h>
+#include <sys/time.h>
#include <rte_log.h>
#include <rte_memory.h>
#include <rte_memzone.h>
#include <rte_launch.h>
-#include <rte_tailq.h>
#include <rte_eal.h>
+#include <rte_eal_memconfig.h>
#include <rte_per_lcore.h>
#include <rte_lcore.h>
#include <rte_common.h>
* zone as well as a physical contiguous zone.
*/
+static uint64_t baseaddr_offset;
#define RANDOMIZE_VA_SPACE_FILE "/proc/sys/kernel/randomize_va_space"
+/* Lock page in physical memory and prevent from swapping. */
+int
+rte_mem_lock_page(const void *virt)
+{
+ unsigned long virtual = (unsigned long)virt;
+ int page_size = getpagesize();
+ unsigned long aligned = (virtual & ~ (page_size - 1));
+ return mlock((void*)aligned, page_size);
+}
+
+/*
+ * Get physical address of any mapped virtual address in the current process.
+ */
+phys_addr_t
+rte_mem_virt2phy(const void *virtaddr)
+{
+ int fd;
+ uint64_t page, physaddr;
+ unsigned long virt_pfn;
+ int page_size;
+ off_t offset;
+
+ /* standard page size */
+ page_size = getpagesize();
+
+ fd = open("/proc/self/pagemap", O_RDONLY);
+ if (fd < 0) {
+ RTE_LOG(ERR, EAL, "%s(): cannot open /proc/self/pagemap: %s\n",
+ __func__, strerror(errno));
+ return RTE_BAD_PHYS_ADDR;
+ }
+
+ virt_pfn = (unsigned long)virtaddr / page_size;
+ offset = sizeof(uint64_t) * virt_pfn;
+ if (lseek(fd, offset, SEEK_SET) == (off_t) -1) {
+ RTE_LOG(ERR, EAL, "%s(): seek error in /proc/self/pagemap: %s\n",
+ __func__, strerror(errno));
+ close(fd);
+ return RTE_BAD_PHYS_ADDR;
+ }
+ if (read(fd, &page, sizeof(uint64_t)) < 0) {
+ RTE_LOG(ERR, EAL, "%s(): cannot read /proc/self/pagemap: %s\n",
+ __func__, strerror(errno));
+ close(fd);
+ return RTE_BAD_PHYS_ADDR;
+ }
+
+ /*
+ * the pfn (page frame number) are bits 0-54 (see
+ * pagemap.txt in linux Documentation)
+ */
+ physaddr = ((page & 0x7fffffffffffffULL) * page_size)
+ + ((unsigned long)virtaddr % page_size);
+ close(fd);
+ return physaddr;
+}
+
+/*
+ * For each hugepage in hugepg_tbl, fill the physaddr value. We find
+ * it by browsing the /proc/self/pagemap special file.
+ */
+static int
+find_physaddrs(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi)
+{
+ unsigned i;
+ phys_addr_t addr;
+
+ for (i = 0; i < hpi->num_pages[0]; i++) {
+ addr = rte_mem_virt2phy(hugepg_tbl[i].orig_va);
+ if (addr == RTE_BAD_PHYS_ADDR)
+ return -1;
+ hugepg_tbl[i].physaddr = addr;
+ }
+ return 0;
+}
+
/*
* Check whether address-space layout randomization is enabled in
* the kernel. This is important for multi-process as it can prevent
}
/*
- * Try to mmap *size bytes in /dev/zero. If it is succesful, return the
+ * Try to mmap *size bytes in /dev/zero. If it is successful, return the
* pointer to the mmap'd area and keep *size unmodified. Else, retry
* with a smaller zone: decrease *size by hugepage_sz until it reaches
* 0. In this case, return NULL. Note: this function returns an address
* which is a multiple of hugepage size.
*/
static void *
-get_virtual_area(uint64_t *size, uint64_t hugepage_sz)
+get_virtual_area(size_t *size, size_t hugepage_sz)
{
void *addr;
int fd;
long aligned_addr;
- RTE_LOG(INFO, EAL, "Ask a virtual area of 0x%"PRIx64" bytes\n", *size);
+ if (internal_config.base_virtaddr != 0) {
+ addr = (void*) (uintptr_t) (internal_config.base_virtaddr +
+ baseaddr_offset);
+ }
+ else addr = NULL;
+
+ RTE_LOG(INFO, EAL, "Ask a virtual area of 0x%zx bytes\n", *size);
fd = open("/dev/zero", O_RDONLY);
if (fd < 0){
return NULL;
}
do {
- addr = mmap(NULL, (*size) + hugepage_sz, PROT_READ, MAP_PRIVATE, fd, 0);
+ addr = mmap(addr,
+ (*size) + hugepage_sz, PROT_READ, MAP_PRIVATE, fd, 0);
if (addr == MAP_FAILED)
*size -= hugepage_sz;
} while (addr == MAP_FAILED && *size > 0);
aligned_addr &= (~(hugepage_sz - 1));
addr = (void *)(aligned_addr);
- RTE_LOG(INFO, EAL, "Virtual area found at %p (size = 0x%"PRIx64")\n",
+ RTE_LOG(INFO, EAL, "Virtual area found at %p (size = 0x%zx)\n",
addr, *size);
+ /* increment offset */
+ baseaddr_offset += *size;
+
return addr;
}
* map continguous physical blocks in contiguous virtual blocks.
*/
static int
-map_all_hugepages(struct hugepage *hugepg_tbl,
+map_all_hugepages(struct hugepage_file *hugepg_tbl,
struct hugepage_info *hpi, int orig)
{
int fd;
unsigned i;
void *virtaddr;
void *vma_addr = NULL;
- uint64_t vma_len = 0;
+ size_t vma_len = 0;
+
+#ifdef RTE_EAL_SINGLE_FILE_SEGMENTS
+ RTE_SET_USED(vma_len);
+#endif
for (i = 0; i < hpi->num_pages[0]; i++) {
uint64_t hugepage_sz = hpi->hugepage_sz;
if (orig) {
hugepg_tbl[i].file_id = i;
hugepg_tbl[i].size = hugepage_sz;
+#ifdef RTE_EAL_SINGLE_FILE_SEGMENTS
+ eal_get_hugefile_temp_path(hugepg_tbl[i].filepath,
+ sizeof(hugepg_tbl[i].filepath), hpi->hugedir,
+ hugepg_tbl[i].file_id);
+#else
eal_get_hugefile_path(hugepg_tbl[i].filepath,
sizeof(hugepg_tbl[i].filepath), hpi->hugedir,
hugepg_tbl[i].file_id);
+#endif
hugepg_tbl[i].filepath[sizeof(hugepg_tbl[i].filepath) - 1] = '\0';
}
-#ifndef RTE_ARCH_X86_64
- /* for 32-bit systems, don't remap 1G pages, just reuse original
- * map address as final map address.
+#ifndef RTE_ARCH_64
+ /* for 32-bit systems, don't remap 1G and 16G pages, just reuse
+ * original map address as final map address.
*/
- else if (hugepage_sz == RTE_PGSIZE_1G){
+ else if ((hugepage_sz == RTE_PGSIZE_1G)
+ || (hugepage_sz == RTE_PGSIZE_16G)) {
hugepg_tbl[i].final_va = hugepg_tbl[i].orig_va;
hugepg_tbl[i].orig_va = NULL;
continue;
}
#endif
+
+#ifndef RTE_EAL_SINGLE_FILE_SEGMENTS
else if (vma_len == 0) {
unsigned j, num_pages;
* physical block: count the number of
* contiguous physical pages. */
for (j = i+1; j < hpi->num_pages[0] ; j++) {
+#ifdef RTE_ARCH_PPC_64
+ /* The physical addresses are sorted in
+ * descending order on PPC64 */
+ if (hugepg_tbl[j].physaddr !=
+ hugepg_tbl[j-1].physaddr - hugepage_sz)
+ break;
+#else
if (hugepg_tbl[j].physaddr !=
hugepg_tbl[j-1].physaddr + hugepage_sz)
break;
+#endif
}
num_pages = j - i;
vma_len = num_pages * hugepage_sz;
if (vma_addr == NULL)
vma_len = hugepage_sz;
}
+#endif
+ /* try to create hugepage file */
fd = open(hugepg_tbl[i].filepath, O_CREAT | O_RDWR, 0755);
if (fd < 0) {
RTE_LOG(ERR, EAL, "%s(): open failed: %s\n", __func__,
hugepg_tbl[i].final_va = virtaddr;
}
+ /* set shared flock on the file. */
+ if (flock(fd, LOCK_SH | LOCK_NB) == -1) {
+ RTE_LOG(ERR, EAL, "%s(): Locking file failed:%s \n",
+ __func__, strerror(errno));
+ close(fd);
+ return -1;
+ }
+
+ close(fd);
+
vma_addr = (char *)vma_addr + hugepage_sz;
vma_len -= hugepage_sz;
- close(fd);
}
return 0;
}
-/* Unmap all hugepages from original mapping. */
-static int
-unmap_all_hugepages_orig(struct hugepage *hugepg_tbl, struct hugepage_info *hpi)
-{
- unsigned i;
- for (i = 0; i < hpi->num_pages[0]; i++) {
- if (hugepg_tbl[i].orig_va) {
- munmap(hugepg_tbl[i].orig_va, hpi->hugepage_sz);
- hugepg_tbl[i].orig_va = NULL;
- }
- }
- return 0;
-}
+#ifdef RTE_EAL_SINGLE_FILE_SEGMENTS
/*
- * For each hugepage in hugepg_tbl, fill the physaddr value. We find
- * it by browsing the /proc/self/pagemap special file.
+ * Remaps all hugepages into single file segments
*/
static int
-find_physaddr(struct hugepage *hugepg_tbl, struct hugepage_info *hpi)
+remap_all_hugepages(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi)
{
int fd;
- unsigned i;
- uint64_t page;
- unsigned long virt_pfn;
- int page_size;
+ unsigned i = 0, j, num_pages, page_idx = 0;
+ void *vma_addr = NULL, *old_addr = NULL, *page_addr = NULL;
+ size_t vma_len = 0;
+ size_t hugepage_sz = hpi->hugepage_sz;
+ size_t total_size, offset;
+ char filepath[MAX_HUGEPAGE_PATH];
+ phys_addr_t physaddr;
+ int socket;
+
+ while (i < hpi->num_pages[0]) {
+
+#ifndef RTE_ARCH_64
+ /* for 32-bit systems, don't remap 1G pages and 16G pages,
+ * just reuse original map address as final map address.
+ */
+ if ((hugepage_sz == RTE_PGSIZE_1G)
+ || (hugepage_sz == RTE_PGSIZE_16G)) {
+ hugepg_tbl[i].final_va = hugepg_tbl[i].orig_va;
+ hugepg_tbl[i].orig_va = NULL;
+ i++;
+ continue;
+ }
+#endif
- /* standard page size */
- page_size = getpagesize();
+ /* reserve a virtual area for next contiguous
+ * physical block: count the number of
+ * contiguous physical pages. */
+ for (j = i+1; j < hpi->num_pages[0] ; j++) {
+#ifdef RTE_ARCH_PPC_64
+ /* The physical addresses are sorted in descending
+ * order on PPC64 */
+ if (hugepg_tbl[j].physaddr !=
+ hugepg_tbl[j-1].physaddr - hugepage_sz)
+ break;
+#else
+ if (hugepg_tbl[j].physaddr !=
+ hugepg_tbl[j-1].physaddr + hugepage_sz)
+ break;
+#endif
+ }
+ num_pages = j - i;
+ vma_len = num_pages * hugepage_sz;
- fd = open("/proc/self/pagemap", O_RDONLY);
- if (fd < 0) {
- RTE_LOG(ERR, EAL, "%s(): cannot open /proc/self/pagemap: %s\n",
- __func__, strerror(errno));
- return -1;
- }
+ socket = hugepg_tbl[i].socket_id;
- for (i = 0; i < hpi->num_pages[0]; i++) {
- off_t offset;
- virt_pfn = (unsigned long)hugepg_tbl[i].orig_va /
- page_size;
- offset = sizeof(uint64_t) * virt_pfn;
- if (lseek(fd, offset, SEEK_SET) != offset){
- RTE_LOG(ERR, EAL, "%s(): seek error in /proc/self/pagemap: %s\n",
- __func__, strerror(errno));
- close(fd);
+ /* get the biggest virtual memory area up to
+ * vma_len. If it fails, vma_addr is NULL, so
+ * let the kernel provide the address. */
+ vma_addr = get_virtual_area(&vma_len, hpi->hugepage_sz);
+
+ /* If we can't find a big enough virtual area, work out how many pages
+ * we are going to get */
+ if (vma_addr == NULL)
+ j = i + 1;
+ else if (vma_len != num_pages * hugepage_sz) {
+ num_pages = vma_len / hugepage_sz;
+ j = i + num_pages;
+
+ }
+
+ hugepg_tbl[page_idx].file_id = page_idx;
+ eal_get_hugefile_path(filepath,
+ sizeof(filepath),
+ hpi->hugedir,
+ hugepg_tbl[page_idx].file_id);
+
+ /* try to create hugepage file */
+ fd = open(filepath, O_CREAT | O_RDWR, 0755);
+ if (fd < 0) {
+ RTE_LOG(ERR, EAL, "%s(): open failed: %s\n", __func__, strerror(errno));
return -1;
}
- if (read(fd, &page, sizeof(uint64_t)) < 0) {
- RTE_LOG(ERR, EAL, "%s(): cannot read /proc/self/pagemap: %s\n",
- __func__, strerror(errno));
+
+ total_size = 0;
+ for (;i < j; i++) {
+
+ /* unmap current segment */
+ if (total_size > 0)
+ munmap(vma_addr, total_size);
+
+ /* unmap original page */
+ munmap(hugepg_tbl[i].orig_va, hugepage_sz);
+ unlink(hugepg_tbl[i].filepath);
+
+ total_size += hugepage_sz;
+
+ old_addr = vma_addr;
+
+ /* map new, bigger segment */
+ vma_addr = mmap(vma_addr, total_size,
+ PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
+
+ if (vma_addr == MAP_FAILED || vma_addr != old_addr) {
+ RTE_LOG(ERR, EAL, "%s(): mmap failed: %s\n", __func__, strerror(errno));
+ close(fd);
+ return -1;
+ }
+
+ /* touch the page. this is needed because kernel postpones mapping
+ * creation until the first page fault. with this, we pin down
+ * the page and it is marked as used and gets into process' pagemap.
+ */
+ for (offset = 0; offset < total_size; offset += hugepage_sz)
+ *((volatile uint8_t*) RTE_PTR_ADD(vma_addr, offset));
+ }
+
+ /* set shared flock on the file. */
+ if (flock(fd, LOCK_SH | LOCK_NB) == -1) {
+ RTE_LOG(ERR, EAL, "%s(): Locking file failed:%s \n",
+ __func__, strerror(errno));
close(fd);
return -1;
}
- /*
- * the pfn (page frame number) are bits 0-54 (see
- * pagemap.txt in linux Documentation)
+ snprintf(hugepg_tbl[page_idx].filepath, MAX_HUGEPAGE_PATH, "%s",
+ filepath);
+
+ physaddr = rte_mem_virt2phy(vma_addr);
+
+ if (physaddr == RTE_BAD_PHYS_ADDR)
+ return -1;
+
+ hugepg_tbl[page_idx].final_va = vma_addr;
+
+ hugepg_tbl[page_idx].physaddr = physaddr;
+
+ hugepg_tbl[page_idx].repeated = num_pages;
+
+ hugepg_tbl[page_idx].socket_id = socket;
+
+ close(fd);
+
+ /* verify the memory segment - that is, check that every VA corresponds
+ * to the physical address we expect to see
*/
- hugepg_tbl[i].physaddr = ((page & 0x7fffffffffffffULL) * page_size);
+ for (offset = 0; offset < vma_len; offset += hugepage_sz) {
+ uint64_t expected_physaddr;
+
+ expected_physaddr = hugepg_tbl[page_idx].physaddr + offset;
+ page_addr = RTE_PTR_ADD(vma_addr, offset);
+ physaddr = rte_mem_virt2phy(page_addr);
+
+ if (physaddr != expected_physaddr) {
+ RTE_LOG(ERR, EAL, "Segment sanity check failed: wrong physaddr "
+ "at %p (offset 0x%" PRIx64 ": 0x%" PRIx64
+ " (expected 0x%" PRIx64 ")\n",
+ page_addr, offset, physaddr, expected_physaddr);
+ return -1;
+ }
+ }
+
+ /* zero out the whole segment */
+ memset(hugepg_tbl[page_idx].final_va, 0, total_size);
+
+ page_idx++;
}
- close(fd);
- return 0;
+
+ /* zero out the rest */
+ memset(&hugepg_tbl[page_idx], 0, (hpi->num_pages[0] - page_idx) * sizeof(struct hugepage_file));
+ return page_idx;
+}
+#else/* RTE_EAL_SINGLE_FILE_SEGMENTS=n */
+
+/* Unmap all hugepages from original mapping */
+static int
+unmap_all_hugepages_orig(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi)
+{
+ unsigned i;
+ for (i = 0; i < hpi->num_pages[0]; i++) {
+ if (hugepg_tbl[i].orig_va) {
+ munmap(hugepg_tbl[i].orig_va, hpi->hugepage_sz);
+ hugepg_tbl[i].orig_va = NULL;
+ }
+ }
+ return 0;
}
+#endif /* RTE_EAL_SINGLE_FILE_SEGMENTS */
/*
* Parse /proc/self/numa_maps to get the NUMA socket ID for each huge
* page.
*/
static int
-find_numasocket(struct hugepage *hugepg_tbl, struct hugepage_info *hpi)
+find_numasocket(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi)
{
int socket_id;
char *end, *nodestr;
return 0;
}
- rte_snprintf(hugedir_str, sizeof(hugedir_str),
- "%s/", hpi->hugedir);
+ snprintf(hugedir_str, sizeof(hugedir_str),
+ "%s/%s", hpi->hugedir, internal_config.hugefile_prefix);
/* parse numa map */
while (fgets(buf, sizeof(buf), f) != NULL) {
}
/*
- * Sort the hugepg_tbl by physical address (lower addresses first). We
- * use a slow algorithm, but we won't have millions of pages, and this
- * is only done at init time.
+ * Sort the hugepg_tbl by physical address (lower addresses first on x86,
+ * higher address first on powerpc). We use a slow algorithm, but we won't
+ * have millions of pages, and this is only done at init time.
*/
static int
-sort_by_physaddr(struct hugepage *hugepg_tbl, struct hugepage_info *hpi)
+sort_by_physaddr(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi)
{
unsigned i, j;
- int smallest_idx;
- uint64_t smallest_addr;
- struct hugepage tmp;
+ int compare_idx;
+ uint64_t compare_addr;
+ struct hugepage_file tmp;
for (i = 0; i < hpi->num_pages[0]; i++) {
- smallest_addr = 0;
- smallest_idx = -1;
+ compare_addr = 0;
+ compare_idx = -1;
/*
* browse all entries starting at 'i', and find the
*/
for (j=i; j< hpi->num_pages[0]; j++) {
- if (smallest_addr == 0 ||
- hugepg_tbl[j].physaddr < smallest_addr) {
- smallest_addr = hugepg_tbl[j].physaddr;
- smallest_idx = j;
+ if (compare_addr == 0 ||
+#ifdef RTE_ARCH_PPC_64
+ hugepg_tbl[j].physaddr > compare_addr) {
+#else
+ hugepg_tbl[j].physaddr < compare_addr) {
+#endif
+ compare_addr = hugepg_tbl[j].physaddr;
+ compare_idx = j;
}
}
/* should not happen */
- if (smallest_idx == -1) {
+ if (compare_idx == -1) {
RTE_LOG(ERR, EAL, "%s(): error in physaddr sorting\n", __func__);
return -1;
}
/* swap the 2 entries in the table */
- memcpy(&tmp, &hugepg_tbl[smallest_idx], sizeof(struct hugepage));
- memcpy(&hugepg_tbl[smallest_idx], &hugepg_tbl[i],
- sizeof(struct hugepage));
- memcpy(&hugepg_tbl[i], &tmp, sizeof(struct hugepage));
+ memcpy(&tmp, &hugepg_tbl[compare_idx],
+ sizeof(struct hugepage_file));
+ memcpy(&hugepg_tbl[compare_idx], &hugepg_tbl[i],
+ sizeof(struct hugepage_file));
+ memcpy(&hugepg_tbl[i], &tmp, sizeof(struct hugepage_file));
}
return 0;
}
* destination is typically the shared memory.
*/
static int
-copy_hugepages_to_shared_mem(struct hugepage * dst, int dest_size,
- const struct hugepage * src, int src_size)
+copy_hugepages_to_shared_mem(struct hugepage_file * dst, int dest_size,
+ const struct hugepage_file * src, int src_size)
{
int src_pos, dst_pos = 0;
/* error on overflow attempt */
if (dst_pos == dest_size)
return -1;
- memcpy(&dst[dst_pos], &src[src_pos], sizeof(struct hugepage));
+ memcpy(&dst[dst_pos], &src[src_pos], sizeof(struct hugepage_file));
dst_pos++;
}
}
* ALL hugepages (not just those we need), additional unmapping needs to be done.
*/
static int
-unmap_unneeded_hugepages(struct hugepage *hugepg_tbl,
+unmap_unneeded_hugepages(struct hugepage_file *hugepg_tbl,
struct hugepage_info *hpi,
unsigned num_hp_info)
{
unsigned socket, size;
int page, nrpages = 0;
- int fd;
/* get total number of hugepages */
for (size = 0; size < num_hp_info; size++)
for (size = 0; size < num_hp_info; size++) {
for (socket = 0; socket < RTE_MAX_NUMA_NODES; socket++) {
unsigned pages_found = 0;
+
/* traverse until we have unmapped all the unused pages */
for (page = 0; page < nrpages; page++) {
- struct hugepage *hp = &hugepg_tbl[page];
+ struct hugepage_file *hp = &hugepg_tbl[page];
+
+#ifdef RTE_EAL_SINGLE_FILE_SEGMENTS
+ /* if this page was already cleared */
+ if (hp->final_va == NULL)
+ continue;
+#endif
/* find a page that matches the criteria */
if ((hp->size == hpi[size].hugepage_sz) &&
/* if we skipped enough pages, unmap the rest */
if (pages_found == hpi[size].num_pages[socket]) {
- munmap(hp->final_va, hp->size);
+ uint64_t unmap_len;
+
+#ifdef RTE_EAL_SINGLE_FILE_SEGMENTS
+ unmap_len = hp->size * hp->repeated;
+#else
+ unmap_len = hp->size;
+#endif
+
+ /* get start addr and len of the remaining segment */
+ munmap(hp->final_va, (size_t) unmap_len);
+
hp->final_va = NULL;
+ if (unlink(hp->filepath) == -1) {
+ RTE_LOG(ERR, EAL, "%s(): Removing %s failed: %s\n",
+ __func__, hp->filepath, strerror(errno));
+ return -1;
+ }
}
+#ifdef RTE_EAL_SINGLE_FILE_SEGMENTS
+ /* else, check how much do we need to map */
else {
- pages_found++;
+ int nr_pg_left =
+ hpi[size].num_pages[socket] - pages_found;
+
+ /* if we need enough memory to fit into the segment */
+ if (hp->repeated <= nr_pg_left) {
+ pages_found += hp->repeated;
+ }
+ /* truncate the segment */
+ else {
+ uint64_t final_size = nr_pg_left * hp->size;
+ uint64_t seg_size = hp->repeated * hp->size;
+
+ void * unmap_va = RTE_PTR_ADD(hp->final_va,
+ final_size);
+ int fd;
+
+ munmap(unmap_va, seg_size - final_size);
+
+ fd = open(hp->filepath, O_RDWR);
+ if (fd < 0) {
+ RTE_LOG(ERR, EAL, "Cannot open %s: %s\n",
+ hp->filepath, strerror(errno));
+ return -1;
+ }
+ if (ftruncate(fd, final_size) < 0) {
+ RTE_LOG(ERR, EAL, "Cannot truncate %s: %s\n",
+ hp->filepath, strerror(errno));
+ return -1;
+ }
+ close(fd);
+
+ pages_found += nr_pg_left;
+ hp->repeated = nr_pg_left;
+ }
}
+#else
+ /* else, lock the page and skip */
+ else
+ pages_found++;
+#endif
+
} /* match page */
} /* foreach page */
} /* foreach socket */
if (num_hp_info == 0)
return -1;
- for (socket = 0; socket < RTE_MAX_NUMA_NODES && total_mem != 0; socket++) {
- /* if specific memory amounts per socket weren't requested */
- if (internal_config.force_sockets == 0) {
+ /* if specific memory amounts per socket weren't requested */
+ if (internal_config.force_sockets == 0) {
+ int cpu_per_socket[RTE_MAX_NUMA_NODES];
+ size_t default_size, total_size;
+ unsigned lcore_id;
+
+ /* Compute number of cores per socket */
+ memset(cpu_per_socket, 0, sizeof(cpu_per_socket));
+ RTE_LCORE_FOREACH(lcore_id) {
+ cpu_per_socket[rte_lcore_to_socket_id(lcore_id)]++;
+ }
+
+ /*
+ * Automatically spread requested memory amongst detected sockets according
+ * to number of cores from cpu mask present on each socket
+ */
+ total_size = internal_config.memory;
+ for (socket = 0; socket < RTE_MAX_NUMA_NODES && total_size != 0; socket++) {
+
+ /* Set memory amount per socket */
+ default_size = (internal_config.memory * cpu_per_socket[socket])
+ / rte_lcore_count();
+
+ /* Limit to maximum available memory on socket */
+ default_size = RTE_MIN(default_size, get_socket_mem_size(socket));
+
+ /* Update sizes */
+ memory[socket] = default_size;
+ total_size -= default_size;
+ }
+
+ /*
+ * If some memory is remaining, try to allocate it by getting all
+ * available memory from sockets, one after the other
+ */
+ for (socket = 0; socket < RTE_MAX_NUMA_NODES && total_size != 0; socket++) {
/* take whatever is available */
- memory[socket] = RTE_MIN(get_socket_mem_size(socket),
- total_mem);
+ default_size = RTE_MIN(get_socket_mem_size(socket) - memory[socket],
+ total_size);
+
+ /* Update sizes */
+ memory[socket] += default_size;
+ total_size -= default_size;
}
+ }
+
+ for (socket = 0; socket < RTE_MAX_NUMA_NODES && total_mem != 0; socket++) {
/* skips if the memory on specific socket wasn't requested */
for (i = 0; i < num_hp_info && memory[socket] != 0; i++){
hp_used[i].hugedir = hp_info[i].hugedir;
rte_eal_hugepage_init(void)
{
struct rte_mem_config *mcfg;
- struct hugepage *hugepage, *tmp_hp = NULL;
+ struct hugepage_file *hugepage, *tmp_hp = NULL;
struct hugepage_info used_hp[MAX_HUGEPAGE_SIZES];
uint64_t memory[RTE_MAX_NUMA_NODES];
unsigned hp_offset;
int i, j, new_memseg;
- int nrpages, total_pages = 0;
+ int nr_hugefiles, nr_hugepages = 0;
void *addr;
+#ifdef RTE_EAL_SINGLE_FILE_SEGMENTS
+ int new_pages_count[MAX_HUGEPAGE_SIZES];
+#endif
memset(used_hp, 0, sizeof(used_hp));
/* get pointer to global configuration */
mcfg = rte_eal_get_configuration()->mem_config;
- /* for debug purposes, hugetlbfs can be disabled */
+ /* hugetlbfs can be disabled */
if (internal_config.no_hugetlbfs) {
- addr = malloc(internal_config.memory);
+ addr = mmap(NULL, internal_config.memory, PROT_READ | PROT_WRITE,
+ MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
+ if (addr == MAP_FAILED) {
+ RTE_LOG(ERR, EAL, "%s: mmap() failed: %s\n", __func__,
+ strerror(errno));
+ return -1;
+ }
mcfg->memseg[0].phys_addr = (phys_addr_t)(uintptr_t)addr;
mcfg->memseg[0].addr = addr;
mcfg->memseg[0].len = internal_config.memory;
- mcfg->memseg[0].socket_id = 0;
+ mcfg->memseg[0].socket_id = SOCKET_ID_ANY;
return 0;
}
+/* check if app runs on Xen Dom0 */
+ if (internal_config.xen_dom0_support) {
+#ifdef RTE_LIBRTE_XEN_DOM0
+ /* use dom0_mm kernel driver to init memory */
+ if (rte_xen_dom0_memory_init() < 0)
+ return -1;
+ else
+ return 0;
+#endif
+ }
+
/* calculate total number of hugepages available. at this point we haven't
* yet started sorting them so they all are on socket 0 */
/* meanwhile, also initialize used_hp hugepage sizes in used_hp */
used_hp[i].hugepage_sz = internal_config.hugepage_info[i].hugepage_sz;
- total_pages += internal_config.hugepage_info[i].num_pages[0];
+ nr_hugepages += internal_config.hugepage_info[i].num_pages[0];
}
/*
* processing done on these pages, shared memory will be created
* at a later stage.
*/
- tmp_hp = malloc(total_pages * sizeof(struct hugepage));
+ tmp_hp = malloc(nr_hugepages * sizeof(struct hugepage_file));
if (tmp_hp == NULL)
goto fail;
- memset(tmp_hp, 0, total_pages * sizeof(struct hugepage));
+ memset(tmp_hp, 0, nr_hugepages * sizeof(struct hugepage_file));
hp_offset = 0; /* where we start the current page size entries */
*/
hpi = &internal_config.hugepage_info[i];
- if (hpi->num_pages == 0)
+ if (hpi->num_pages[0] == 0)
continue;
/* map all hugepages available */
}
/* find physical addresses and sockets for each hugepage */
- if (find_physaddr(&tmp_hp[hp_offset], hpi) < 0){
+ if (find_physaddrs(&tmp_hp[hp_offset], hpi) < 0){
RTE_LOG(DEBUG, EAL, "Failed to find phys addr for %u MB pages\n",
(unsigned)(hpi->hugepage_sz / 0x100000));
goto fail;
if (sort_by_physaddr(&tmp_hp[hp_offset], hpi) < 0)
goto fail;
+#ifdef RTE_EAL_SINGLE_FILE_SEGMENTS
+ /* remap all hugepages into single file segments */
+ new_pages_count[i] = remap_all_hugepages(&tmp_hp[hp_offset], hpi);
+ if (new_pages_count[i] < 0){
+ RTE_LOG(DEBUG, EAL, "Failed to remap %u MB pages\n",
+ (unsigned)(hpi->hugepage_sz / 0x100000));
+ goto fail;
+ }
+
+ /* we have processed a num of hugepages of this size, so inc offset */
+ hp_offset += new_pages_count[i];
+#else
/* remap all hugepages */
if (map_all_hugepages(&tmp_hp[hp_offset], hpi, 0) < 0){
RTE_LOG(DEBUG, EAL, "Failed to remap %u MB pages\n",
/* we have processed a num of hugepages of this size, so inc offset */
hp_offset += hpi->num_pages[0];
+#endif
+ }
+
+#ifdef RTE_EAL_SINGLE_FILE_SEGMENTS
+ nr_hugefiles = 0;
+ for (i = 0; i < (int) internal_config.num_hugepage_sizes; i++) {
+ nr_hugefiles += new_pages_count[i];
}
+#else
+ nr_hugefiles = nr_hugepages;
+#endif
+
/* clean out the numbers of pages */
for (i = 0; i < (int) internal_config.num_hugepage_sizes; i++)
internal_config.hugepage_info[i].num_pages[j] = 0;
/* get hugepages for each socket */
- for (i = 0; i < total_pages; i++) {
+ for (i = 0; i < nr_hugefiles; i++) {
int socket = tmp_hp[i].socket_id;
/* find a hugepage info with right size and increment num_pages */
for (j = 0; j < (int) internal_config.num_hugepage_sizes; j++) {
if (tmp_hp[i].size ==
internal_config.hugepage_info[j].hugepage_sz) {
+#ifdef RTE_EAL_SINGLE_FILE_SEGMENTS
+ internal_config.hugepage_info[j].num_pages[socket] +=
+ tmp_hp[i].repeated;
+#else
internal_config.hugepage_info[j].num_pages[socket]++;
+#endif
}
}
}
memory[i] = internal_config.socket_mem[i];
/* calculate final number of pages */
- nrpages = calc_num_pages_per_socket(memory,
+ nr_hugepages = calc_num_pages_per_socket(memory,
internal_config.hugepage_info, used_hp,
internal_config.num_hugepage_sizes);
/* error if not enough memory available */
- if (nrpages < 0)
+ if (nr_hugepages < 0)
goto fail;
/* reporting in! */
/* create shared memory */
hugepage = create_shared_memory(eal_hugepage_info_path(),
- nrpages * sizeof(struct hugepage));
+ nr_hugefiles * sizeof(struct hugepage_file));
if (hugepage == NULL) {
RTE_LOG(ERR, EAL, "Failed to create shared memory!\n");
goto fail;
}
+ memset(hugepage, 0, nr_hugefiles * sizeof(struct hugepage_file));
/*
* unmap pages that we won't need (looks at used_hp).
* this procedure only copies those hugepages that have final_va
* not NULL. has overflow protection.
*/
- if (copy_hugepages_to_shared_mem(hugepage, nrpages,
- tmp_hp, total_pages) < 0) {
+ if (copy_hugepages_to_shared_mem(hugepage, nr_hugefiles,
+ tmp_hp, nr_hugefiles) < 0) {
RTE_LOG(ERR, EAL, "Copying tables to shared memory failed!\n");
goto fail;
}
free(tmp_hp);
tmp_hp = NULL;
- memset(mcfg->memseg, 0, sizeof(mcfg->memseg));
- j = -1;
- for (i = 0; i < nrpages; i++) {
+ /* find earliest free memseg - this is needed because in case of IVSHMEM,
+ * segments might have already been initialized */
+ for (j = 0; j < RTE_MAX_MEMSEG; j++)
+ if (mcfg->memseg[j].addr == NULL) {
+ /* move to previous segment and exit loop */
+ j--;
+ break;
+ }
+
+ for (i = 0; i < nr_hugefiles; i++) {
new_memseg = 0;
/* if this is a new section, create a new memseg */
new_memseg = 1;
else if (hugepage[i].size != hugepage[i-1].size)
new_memseg = 1;
+
+#ifdef RTE_ARCH_PPC_64
+ /* On PPC64 architecture, the mmap always start from higher
+ * virtual address to lower address. Here, both the physical
+ * address and virtual address are in descending order */
+ else if ((hugepage[i-1].physaddr - hugepage[i].physaddr) !=
+ hugepage[i].size)
+ new_memseg = 1;
+ else if (((unsigned long)hugepage[i-1].final_va -
+ (unsigned long)hugepage[i].final_va) != hugepage[i].size)
+ new_memseg = 1;
+#else
else if ((hugepage[i].physaddr - hugepage[i-1].physaddr) !=
hugepage[i].size)
new_memseg = 1;
else if (((unsigned long)hugepage[i].final_va -
(unsigned long)hugepage[i-1].final_va) != hugepage[i].size)
new_memseg = 1;
+#endif
if (new_memseg) {
j += 1;
mcfg->memseg[j].phys_addr = hugepage[i].physaddr;
mcfg->memseg[j].addr = hugepage[i].final_va;
+#ifdef RTE_EAL_SINGLE_FILE_SEGMENTS
+ mcfg->memseg[j].len = hugepage[i].size * hugepage[i].repeated;
+#else
mcfg->memseg[j].len = hugepage[i].size;
+#endif
mcfg->memseg[j].socket_id = hugepage[i].socket_id;
mcfg->memseg[j].hugepage_sz = hugepage[i].size;
}
/* continuation of previous memseg */
else {
+#ifdef RTE_ARCH_PPC_64
+ /* Use the phy and virt address of the last page as segment
+ * address for IBM Power architecture */
+ mcfg->memseg[j].phys_addr = hugepage[i].physaddr;
+ mcfg->memseg[j].addr = hugepage[i].final_va;
+#endif
mcfg->memseg[j].len += mcfg->memseg[j].hugepage_sz;
}
hugepage[i].memseg_id = j;
}
- return 0;
+ if (i < nr_hugefiles) {
+ RTE_LOG(ERR, EAL, "Can only reserve %d pages "
+ "from %d requested\n"
+ "Current %s=%d is not enough\n"
+ "Please either increase it or request less amount "
+ "of memory.\n",
+ i, nr_hugefiles, RTE_STR(CONFIG_RTE_MAX_MEMSEG),
+ RTE_MAX_MEMSEG);
+ return (-ENOMEM);
+ }
+ return 0;
fail:
if (tmp_hp)
rte_eal_hugepage_attach(void)
{
const struct rte_mem_config *mcfg = rte_eal_get_configuration()->mem_config;
- const struct hugepage *hp = NULL;
+ const struct hugepage_file *hp = NULL;
unsigned num_hp = 0;
unsigned i, s = 0; /* s used to track the segment number */
off_t size;
"into secondary processes\n");
}
+ if (internal_config.xen_dom0_support) {
+#ifdef RTE_LIBRTE_XEN_DOM0
+ if (rte_xen_dom0_memory_attach() < 0) {
+ RTE_LOG(ERR, EAL,"Failed to attach memory setments of primay "
+ "process\n");
+ return -1;
+ }
+ return 0;
+#endif
+ }
+
fd_zero = open("/dev/zero", O_RDONLY);
if (fd_zero < 0) {
RTE_LOG(ERR, EAL, "Could not open /dev/zero\n");
goto error;
}
+ /* map all segments into memory to make sure we get the addrs */
+ for (s = 0; s < RTE_MAX_MEMSEG; ++s) {
+ void *base_addr;
+
+ /*
+ * the first memory segment with len==0 is the one that
+ * follows the last valid segment.
+ */
+ if (mcfg->memseg[s].len == 0)
+ break;
+
+#ifdef RTE_LIBRTE_IVSHMEM
+ /*
+ * if segment has ioremap address set, it's an IVSHMEM segment and
+ * doesn't need mapping as it was already mapped earlier
+ */
+ if (mcfg->memseg[s].ioremap_addr != 0)
+ continue;
+#endif
+
+ /*
+ * fdzero is mmapped to get a contiguous block of virtual
+ * addresses of the appropriate memseg size.
+ * use mmap to get identical addresses as the primary process.
+ */
+ base_addr = mmap(mcfg->memseg[s].addr, mcfg->memseg[s].len,
+ PROT_READ, MAP_PRIVATE, fd_zero, 0);
+ if (base_addr == MAP_FAILED ||
+ base_addr != mcfg->memseg[s].addr) {
+ RTE_LOG(ERR, EAL, "Could not mmap %llu bytes "
+ "in /dev/zero to requested address [%p]: '%s'\n",
+ (unsigned long long)mcfg->memseg[s].len,
+ mcfg->memseg[s].addr, strerror(errno));
+ if (aslr_enabled() > 0) {
+ RTE_LOG(ERR, EAL, "It is recommended to "
+ "disable ASLR in the kernel "
+ "and retry running both primary "
+ "and secondary processes\n");
+ }
+ goto error;
+ }
+ }
+
size = getFileSize(fd_hugepage);
hp = mmap(NULL, size, PROT_READ, MAP_PRIVATE, fd_hugepage, 0);
if (hp == NULL) {
goto error;
}
- num_hp = size / sizeof(struct hugepage);
- RTE_LOG(DEBUG, EAL, "Analysing %u hugepages\n", num_hp);
+ num_hp = size / sizeof(struct hugepage_file);
+ RTE_LOG(DEBUG, EAL, "Analysing %u files\n", num_hp);
+ s = 0;
while (s < RTE_MAX_MEMSEG && mcfg->memseg[s].len > 0){
void *addr, *base_addr;
uintptr_t offset = 0;
-
- /* fdzero is mmapped to get a contiguous block of virtual addresses
- * get a block of free memory of the appropriate size -
- * use mmap to attempt to get an identical address as server.
+ size_t mapping_size;
+#ifdef RTE_LIBRTE_IVSHMEM
+ /*
+ * if segment has ioremap address set, it's an IVSHMEM segment and
+ * doesn't need mapping as it was already mapped earlier
*/
- base_addr = mmap(mcfg->memseg[s].addr, mcfg->memseg[s].len,
- PROT_READ, MAP_PRIVATE, fd_zero, 0);
- if (base_addr == MAP_FAILED || base_addr != mcfg->memseg[s].addr) {
- RTE_LOG(ERR, EAL, "Could not mmap %llu bytes "
- "in /dev/zero to requested address [%p]\n",
- (unsigned long long)mcfg->memseg[s].len,
- mcfg->memseg[s].addr);
- if (aslr_enabled() > 0)
- RTE_LOG(ERR, EAL, "It is recommended to disable ASLR in the kernel "
- "and retry running both primary and secondary processes\n");
- goto error;
+ if (mcfg->memseg[s].ioremap_addr != 0) {
+ s++;
+ continue;
}
- /* free memory so we can map the hugepages into the space */
+#endif
+ /*
+ * free previously mapped memory so we can map the
+ * hugepages into the space
+ */
+ base_addr = mcfg->memseg[s].addr;
munmap(base_addr, mcfg->memseg[s].len);
/* find the hugepages for this segment and map them
hp[i].filepath);
goto error;
}
+#ifdef RTE_EAL_SINGLE_FILE_SEGMENTS
+ mapping_size = hp[i].size * hp[i].repeated;
+#else
+ mapping_size = hp[i].size;
+#endif
addr = mmap(RTE_PTR_ADD(base_addr, offset),
- hp[i].size, PROT_READ | PROT_WRITE,
- MAP_SHARED | MAP_FIXED, fd, 0);
+ mapping_size, PROT_READ | PROT_WRITE,
+ MAP_SHARED, fd, 0);
close(fd); /* close file both on success and on failure */
- if (addr == MAP_FAILED) {
+ if (addr == MAP_FAILED ||
+ addr != RTE_PTR_ADD(base_addr, offset)) {
RTE_LOG(ERR, EAL, "Could not mmap %s\n",
hp[i].filepath);
goto error;
}
- offset+=hp[i].size;
+ offset+=mapping_size;
}
}
RTE_LOG(DEBUG, EAL, "Mapped segment %u of size 0x%llx\n", s,
(unsigned long long)mcfg->memseg[s].len);
s++;
}
+ /* unmap the hugepage config file, since we are done using it */
+ munmap((void *)(uintptr_t)hp, size);
close(fd_zero);
close(fd_hugepage);
return 0;
int
rte_eal_memory_init(void)
{
- RTE_LOG(INFO, EAL, "Setting up hugepage memory...\n");
+ RTE_LOG(INFO, EAL, "Setting up memory...\n");
const int retval = rte_eal_process_type() == RTE_PROC_PRIMARY ?
rte_eal_hugepage_init() :
rte_eal_hugepage_attach();
git.droids-corp.org / dpdk.git / blobdiff