X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=lib%2Flibrte_eal%2Flinuxapp%2Feal%2Feal_memory.c;h=38853b753aabbec6069dac3b43342d76d13d0ce3;hb=5c7472135ba8d91d608d89c9311a7d80bda7bee7;hp=fcc7db33cb0de50d6785638b037d66ac46ea8f1e;hpb=b6df9fc8715f9a925136006b18fdd65f9c621757;p=dpdk.git diff --git a/lib/librte_eal/linuxapp/eal/eal_memory.c b/lib/librte_eal/linuxapp/eal/eal_memory.c index fcc7db33cb..38853b753a 100644 --- a/lib/librte_eal/linuxapp/eal/eal_memory.c +++ b/lib/librte_eal/linuxapp/eal/eal_memory.c @@ -1,75 +1,17 @@ -/*- - * BSD LICENSE - * - * Copyright(c) 2010-2013 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 - * 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 - * 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 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 - * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2010-2014 Intel Corporation. + * Copyright(c) 2013 6WIND S.A. */ +#define _FILE_OFFSET_BITS 64 #include #include +#include #include #include #include #include #include -#include #include #include #include @@ -77,16 +19,18 @@ #include #include #include -#include #include #include -#include +#include +#include +#ifdef RTE_EAL_NUMA_AWARE_HUGEPAGES +#include +#include +#endif #include #include -#include #include -#include #include #include #include @@ -99,6 +43,8 @@ #include "eal_filesystem.h" #include "eal_hugepages.h" +#define PFN_MASK_SIZE 8 + /** * @file * Huge page mapping under linux @@ -111,9 +57,140 @@ * zone as well as a physical contiguous zone. */ +static uint64_t baseaddr_offset; + +static bool phys_addrs_available = true; #define RANDOMIZE_VA_SPACE_FILE "/proc/sys/kernel/randomize_va_space" +static void +test_phys_addrs_available(void) +{ + uint64_t tmp; + phys_addr_t physaddr; + + if (!rte_eal_has_hugepages()) { + RTE_LOG(ERR, EAL, + "Started without hugepages support, physical addresses not available\n"); + phys_addrs_available = false; + return; + } + + physaddr = rte_mem_virt2phy(&tmp); + if (physaddr == RTE_BAD_PHYS_ADDR) { + if (rte_eal_iova_mode() == RTE_IOVA_PA) + RTE_LOG(ERR, EAL, + "Cannot obtain physical addresses: %s. " + "Only vfio will function.\n", + strerror(errno)); + phys_addrs_available = false; + } +} + +/* + * Get physical address of any mapped virtual address in the current process. + */ +phys_addr_t +rte_mem_virt2phy(const void *virtaddr) +{ + int fd, retval; + uint64_t page, physaddr; + unsigned long virt_pfn; + int page_size; + off_t offset; + + /* Cannot parse /proc/self/pagemap, no need to log errors everywhere */ + if (!phys_addrs_available) + return RTE_BAD_IOVA; + + /* 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_IOVA; + } + + 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_IOVA; + } + + retval = read(fd, &page, PFN_MASK_SIZE); + close(fd); + if (retval < 0) { + RTE_LOG(ERR, EAL, "%s(): cannot read /proc/self/pagemap: %s\n", + __func__, strerror(errno)); + return RTE_BAD_IOVA; + } else if (retval != PFN_MASK_SIZE) { + RTE_LOG(ERR, EAL, "%s(): read %d bytes from /proc/self/pagemap " + "but expected %d:\n", + __func__, retval, PFN_MASK_SIZE); + return RTE_BAD_IOVA; + } + + /* + * the pfn (page frame number) are bits 0-54 (see + * pagemap.txt in linux Documentation) + */ + if ((page & 0x7fffffffffffffULL) == 0) + return RTE_BAD_IOVA; + + physaddr = ((page & 0x7fffffffffffffULL) * page_size) + + ((unsigned long)virtaddr % page_size); + + return physaddr; +} + +rte_iova_t +rte_mem_virt2iova(const void *virtaddr) +{ + if (rte_eal_iova_mode() == RTE_IOVA_VA) + return (uintptr_t)virtaddr; + return rte_mem_virt2phy(virtaddr); +} + +/* + * 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 int 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; +} + +/* + * For each hugepage in hugepg_tbl, fill the physaddr value sequentially. + */ +static int +set_physaddrs(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi) +{ + unsigned int i; + static phys_addr_t addr; + + for (i = 0; i < hpi->num_pages[0]; i++) { + hugepg_tbl[i].physaddr = addr; + addr += hugepg_tbl[i].size; + } + return 0; +} + /* * Check whether address-space layout randomization is enabled in * the kernel. This is important for multi-process as it can prevent @@ -145,58 +222,31 @@ aslr_enabled(void) } /* - * Increase limit for open files for current process - */ -static int -increase_open_file_limit(void) -{ - struct rlimit limit; - - /* read current limits */ - if (getrlimit(RLIMIT_NOFILE, &limit) != 0) { - RTE_LOG(ERR, EAL, "Error reading resource limit: %s\n", - strerror(errno)); - return -1; - } - - /* check if current soft limit matches the hard limit */ - if (limit.rlim_cur < limit.rlim_max) { - /* set soft limit to match hard limit */ - limit.rlim_cur = limit.rlim_max; - } - else { - /* we can't increase the soft limit so now we try to increase - * soft and hard limit. this might fail when run as non-root. - */ - limit.rlim_cur *= 2; - limit.rlim_max *= 2; - } - - /* set current resource limit */ - if (setrlimit(RLIMIT_NOFILE, &limit) != 0) { - RTE_LOG(ERR, EAL, "Error increasing open files limit: %s\n", - strerror(errno)); - return -1; - } - - return 0; -} - -/* - * 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; + void *addr_hint; 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) { + int page_size = sysconf(_SC_PAGE_SIZE); + addr_hint = (void *) (uintptr_t) + (internal_config.base_virtaddr + baseaddr_offset); + addr_hint = RTE_PTR_ALIGN_FLOOR(addr_hint, page_size); + } else { + addr_hint = NULL; + } + + RTE_LOG(DEBUG, EAL, "Ask a virtual area of 0x%zx bytes\n", *size); + fd = open("/dev/zero", O_RDONLY); if (fd < 0){ @@ -204,14 +254,28 @@ get_virtual_area(uint64_t *size, uint64_t hugepage_sz) return NULL; } do { - addr = mmap(NULL, (*size) + hugepage_sz, PROT_READ, MAP_PRIVATE, fd, 0); - if (addr == MAP_FAILED) + addr = mmap(addr_hint, (*size) + hugepage_sz, PROT_READ, +#ifdef RTE_ARCH_PPC_64 + MAP_PRIVATE | MAP_ANONYMOUS | MAP_HUGETLB, +#else + MAP_PRIVATE, +#endif + fd, 0); + if (addr == MAP_FAILED) { *size -= hugepage_sz; + } else if (addr_hint != NULL && addr != addr_hint) { + RTE_LOG(WARNING, EAL, "WARNING! Base virtual address " + "hint (%p != %p) not respected!\n", + addr_hint, addr); + RTE_LOG(WARNING, EAL, " This may cause issues with " + "mapping memory into secondary processes\n"); + } } while (addr == MAP_FAILED && *size > 0); if (addr == MAP_FAILED) { close(fd); - RTE_LOG(INFO, EAL, "Cannot get a virtual area\n"); + RTE_LOG(ERR, EAL, "Cannot get a virtual area: %s\n", + strerror(errno)); return NULL; } @@ -224,32 +288,119 @@ get_virtual_area(uint64_t *size, uint64_t hugepage_sz) aligned_addr &= (~(hugepage_sz - 1)); addr = (void *)(aligned_addr); - RTE_LOG(INFO, EAL, "Virtual area found at %p (size = 0x%"PRIx64")\n", + RTE_LOG(DEBUG, EAL, "Virtual area found at %p (size = 0x%zx)\n", addr, *size); + /* increment offset */ + baseaddr_offset += *size; + return addr; } +static sigjmp_buf huge_jmpenv; + +static void huge_sigbus_handler(int signo __rte_unused) +{ + siglongjmp(huge_jmpenv, 1); +} + +/* Put setjmp into a wrap method to avoid compiling error. Any non-volatile, + * non-static local variable in the stack frame calling sigsetjmp might be + * clobbered by a call to longjmp. + */ +static int huge_wrap_sigsetjmp(void) +{ + return sigsetjmp(huge_jmpenv, 1); +} + +#ifdef RTE_EAL_NUMA_AWARE_HUGEPAGES +/* Callback for numa library. */ +void numa_error(char *where) +{ + RTE_LOG(ERR, EAL, "%s failed: %s\n", where, strerror(errno)); +} +#endif + /* * Mmap all hugepages of hugepage table: it first open a file in * hugetlbfs, then mmap() hugepage_sz data in it. If orig is set, the * virtual address is stored in hugepg_tbl[i].orig_va, else it is stored * in hugepg_tbl[i].final_va. The second mapping (when orig is 0) tries to - * map continguous physical blocks in contiguous virtual blocks. + * map contiguous physical blocks in contiguous virtual blocks. */ -static int -map_all_hugepages(struct hugepage *hugepg_tbl, - struct hugepage_info *hpi, int orig) +static unsigned +map_all_hugepages(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi, + uint64_t *essential_memory __rte_unused, 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_NUMA_AWARE_HUGEPAGES + int node_id = -1; + int essential_prev = 0; + int oldpolicy; + struct bitmask *oldmask = numa_allocate_nodemask(); + bool have_numa = true; + unsigned long maxnode = 0; + + /* Check if kernel supports NUMA. */ + if (numa_available() != 0) { + RTE_LOG(DEBUG, EAL, "NUMA is not supported.\n"); + have_numa = false; + } + + if (orig && have_numa) { + RTE_LOG(DEBUG, EAL, "Trying to obtain current memory policy.\n"); + if (get_mempolicy(&oldpolicy, oldmask->maskp, + oldmask->size + 1, 0, 0) < 0) { + RTE_LOG(ERR, EAL, + "Failed to get current mempolicy: %s. " + "Assuming MPOL_DEFAULT.\n", strerror(errno)); + oldpolicy = MPOL_DEFAULT; + } + for (i = 0; i < RTE_MAX_NUMA_NODES; i++) + if (internal_config.socket_mem[i]) + maxnode = i + 1; + } +#endif for (i = 0; i < hpi->num_pages[0]; i++) { uint64_t hugepage_sz = hpi->hugepage_sz; +#ifdef RTE_EAL_NUMA_AWARE_HUGEPAGES + if (maxnode) { + unsigned int j; + + for (j = 0; j < maxnode; j++) + if (essential_memory[j]) + break; + + if (j == maxnode) { + node_id = (node_id + 1) % maxnode; + while (!internal_config.socket_mem[node_id]) { + node_id++; + node_id %= maxnode; + } + essential_prev = 0; + } else { + node_id = j; + essential_prev = essential_memory[j]; + + if (essential_memory[j] < hugepage_sz) + essential_memory[j] = 0; + else + essential_memory[j] -= hugepage_sz; + } + + RTE_LOG(DEBUG, EAL, + "Setting policy MPOL_PREFERRED for socket %d\n", + node_id); + numa_set_preferred(node_id); + } +#endif + if (orig) { hugepg_tbl[i].file_id = i; hugepg_tbl[i].size = hugepage_sz; @@ -258,11 +409,12 @@ map_all_hugepages(struct hugepage *hugepg_tbl, hugepg_tbl[i].file_id); 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; @@ -275,9 +427,17 @@ map_all_hugepages(struct hugepage *hugepg_tbl, * 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; @@ -291,102 +451,102 @@ map_all_hugepages(struct hugepage *hugepg_tbl, } /* try to create hugepage file */ - fd = open(hugepg_tbl[i].filepath, O_CREAT | O_RDWR, 0755); + fd = open(hugepg_tbl[i].filepath, O_CREAT | O_RDWR, 0600); if (fd < 0) { - RTE_LOG(ERR, EAL, "%s(): open failed: %s\n", __func__, + RTE_LOG(DEBUG, EAL, "%s(): open failed: %s\n", __func__, strerror(errno)); - return -1; + goto out; } + /* map the segment, and populate page tables, + * the kernel fills this segment with zeros */ virtaddr = mmap(vma_addr, hugepage_sz, PROT_READ | PROT_WRITE, - MAP_SHARED, fd, 0); + MAP_SHARED | MAP_POPULATE, fd, 0); if (virtaddr == MAP_FAILED) { - RTE_LOG(ERR, EAL, "%s(): mmap failed: %s\n", __func__, + RTE_LOG(DEBUG, EAL, "%s(): mmap failed: %s\n", __func__, strerror(errno)); close(fd); - return -1; + goto out; } if (orig) { hugepg_tbl[i].orig_va = virtaddr; - memset(virtaddr, 0, hugepage_sz); } else { hugepg_tbl[i].final_va = virtaddr; } - /* close the file descriptor, files will be locked later */ + if (orig) { + /* In linux, hugetlb limitations, like cgroup, are + * enforced at fault time instead of mmap(), even + * with the option of MAP_POPULATE. Kernel will send + * a SIGBUS signal. To avoid to be killed, save stack + * environment here, if SIGBUS happens, we can jump + * back here. + */ + if (huge_wrap_sigsetjmp()) { + RTE_LOG(DEBUG, EAL, "SIGBUS: Cannot mmap more " + "hugepages of size %u MB\n", + (unsigned)(hugepage_sz / 0x100000)); + munmap(virtaddr, hugepage_sz); + close(fd); + unlink(hugepg_tbl[i].filepath); +#ifdef RTE_EAL_NUMA_AWARE_HUGEPAGES + if (maxnode) + essential_memory[node_id] = + essential_prev; +#endif + goto out; + } + *(int *)virtaddr = 0; + } + + + /* set shared flock on the file. */ + if (flock(fd, LOCK_SH | LOCK_NB) == -1) { + RTE_LOG(DEBUG, EAL, "%s(): Locking file failed:%s \n", + __func__, strerror(errno)); + close(fd); + goto out; + } + close(fd); vma_addr = (char *)vma_addr + hugepage_sz; vma_len -= hugepage_sz; } - 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; +out: +#ifdef RTE_EAL_NUMA_AWARE_HUGEPAGES + if (maxnode) { + RTE_LOG(DEBUG, EAL, + "Restoring previous memory policy: %d\n", oldpolicy); + if (oldpolicy == MPOL_DEFAULT) { + numa_set_localalloc(); + } else if (set_mempolicy(oldpolicy, oldmask->maskp, + oldmask->size + 1) < 0) { + RTE_LOG(ERR, EAL, "Failed to restore mempolicy: %s\n", + strerror(errno)); + numa_set_localalloc(); } } - return 0; + numa_free_cpumask(oldmask); +#endif + return i; } -/* - * For each hugepage in hugepg_tbl, fill the physaddr value. We find - * it by browsing the /proc/self/pagemap special file. - */ +/* Unmap all hugepages from original mapping */ static int -find_physaddr(struct hugepage *hugepg_tbl, struct hugepage_info *hpi) +unmap_all_hugepages_orig(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi) { - int fd; - unsigned i; - uint64_t page; - unsigned long virt_pfn; - int page_size; - - /* 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 -1; - } - - 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) == (off_t) -1) { - RTE_LOG(ERR, EAL, "%s(): seek error in /proc/self/pagemap: %s\n", - __func__, strerror(errno)); - close(fd); - 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)); - close(fd); - return -1; - } - - /* - * the pfn (page frame number) are bits 0-54 (see - * pagemap.txt in linux Documentation) - */ - hugepg_tbl[i].physaddr = ((page & 0x7fffffffffffffULL) * page_size); - } - close(fd); - return 0; + 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; } /* @@ -394,7 +554,7 @@ find_physaddr(struct hugepage *hugepg_tbl, struct hugepage_info *hpi) * 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; @@ -406,13 +566,13 @@ find_numasocket(struct hugepage *hugepg_tbl, struct hugepage_info *hpi) f = fopen("/proc/self/numa_maps", "r"); if (f == NULL) { - RTE_LOG(INFO, EAL, "cannot open /proc/self/numa_maps," - " consider that all memory is in socket_id 0\n"); + RTE_LOG(NOTICE, EAL, "NUMA support not available" + " consider that all memory is in socket_id 0\n"); 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) { @@ -456,6 +616,11 @@ find_numasocket(struct hugepage *hugepg_tbl, struct hugepage_info *hpi) if (hugepg_tbl[i].orig_va == va) { hugepg_tbl[i].socket_id = socket_id; hp_count++; +#ifdef RTE_EAL_NUMA_AWARE_HUGEPAGES + RTE_LOG(DEBUG, EAL, + "Hugepage %s is on socket %d\n", + hugepg_tbl[i].filepath, socket_id); +#endif } } } @@ -471,49 +636,23 @@ error: return -1; } -/* - * 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. - */ static int -sort_by_physaddr(struct hugepage *hugepg_tbl, struct hugepage_info *hpi) +cmp_physaddr(const void *a, const void *b) { - unsigned i, j; - int smallest_idx; - uint64_t smallest_addr; - struct hugepage tmp; - - for (i = 0; i < hpi->num_pages[0]; i++) { - smallest_addr = 0; - smallest_idx = -1; - - /* - * browse all entries starting at 'i', and find the - * entry with the smallest addr - */ - 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; - } - } - - /* should not happen */ - if (smallest_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)); - } - return 0; +#ifndef RTE_ARCH_PPC_64 + const struct hugepage_file *p1 = a; + const struct hugepage_file *p2 = b; +#else + /* PowerPC needs memory sorted in reverse order from x86 */ + const struct hugepage_file *p1 = b; + const struct hugepage_file *p2 = a; +#endif + if (p1->physaddr < p2->physaddr) + return -1; + else if (p1->physaddr > p2->physaddr) + return 1; + else + return 0; } /* @@ -533,6 +672,8 @@ create_shared_memory(const char *filename, const size_t mem_size) } retval = mmap(NULL, mem_size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0); close(fd); + if (retval == MAP_FAILED) + return NULL; return retval; } @@ -541,8 +682,8 @@ create_shared_memory(const char *filename, const size_t mem_size) * 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; @@ -551,25 +692,48 @@ copy_hugepages_to_shared_mem(struct hugepage * dst, int dest_size, /* 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++; } } return 0; } +static int +unlink_hugepage_files(struct hugepage_file *hugepg_tbl, + unsigned num_hp_info) +{ + unsigned socket, size; + int page, nrpages = 0; + + /* get total number of hugepages */ + for (size = 0; size < num_hp_info; size++) + for (socket = 0; socket < RTE_MAX_NUMA_NODES; socket++) + nrpages += + internal_config.hugepage_info[size].num_pages[socket]; + + for (page = 0; page < nrpages; page++) { + struct hugepage_file *hp = &hugepg_tbl[page]; + + if (hp->final_va != NULL && unlink(hp->filepath)) { + RTE_LOG(WARNING, EAL, "%s(): Removing %s failed: %s\n", + __func__, hp->filepath, strerror(errno)); + } + } + return 0; +} + /* * unmaps hugepages that are not going to be used. since we originally allocate * 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++) @@ -579,9 +743,10 @@ unmap_unneeded_hugepages(struct hugepage *hugepg_tbl, 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]; /* find a page that matches the criteria */ if ((hp->size == hpi[size].hugepage_sz) && @@ -589,35 +754,24 @@ unmap_unneeded_hugepages(struct hugepage *hugepg_tbl, /* 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; + + unmap_len = hp->size; + + /* get start addr and len of the remaining segment */ + munmap(hp->final_va, (size_t) unmap_len); + hp->final_va = NULL; - } - /* lock the page and skip */ - else { - /* try and open the hugepage file */ - while ((fd = open(hp->filepath, O_CREAT | O_RDWR, 0755)) < 0) { - /* if we can't open due to resource limits */ - if (errno == EMFILE) { - RTE_LOG(INFO, EAL, "Increasing open file limit\n"); - - /* if we manage to increase resource limit, try again */ - if (increase_open_file_limit() == 0) - continue; - } - else - RTE_LOG(ERR, EAL, "%s(): open failed: %s\n", __func__, - strerror(errno)); - return -1; - } - /* try and lock the hugepage */ - if (flock(fd, LOCK_SH | LOCK_NB) == -1) { - RTE_LOG(ERR, EAL, "Locking hugepage file failed!\n"); - close(fd); + if (unlink(hp->filepath) == -1) { + RTE_LOG(ERR, EAL, "%s(): Removing %s failed: %s\n", + __func__, hp->filepath, strerror(errno)); return -1; } - hp->page_lock = fd; + } else { + /* lock the page and skip */ pages_found++; } + } /* match page */ } /* foreach page */ } /* foreach socket */ @@ -638,7 +792,7 @@ get_socket_mem_size(int socket) size += hpi->hugepage_sz * hpi->num_pages[socket]; } - return (size); + return size; } /* @@ -662,13 +816,53 @@ calc_num_pages_per_socket(uint64_t * memory, 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; @@ -719,7 +913,7 @@ calc_num_pages_per_socket(uint64_t * memory, 0x100000); available = requested - ((unsigned) (memory[socket] / 0x100000)); - RTE_LOG(INFO, EAL, "Not enough memory available on socket %u! " + RTE_LOG(ERR, EAL, "Not enough memory available on socket %u! " "Requested: %uMB, available: %uMB\n", socket, requested, available); return -1; @@ -730,13 +924,58 @@ calc_num_pages_per_socket(uint64_t * memory, if (total_mem > 0) { requested = (unsigned) (internal_config.memory / 0x100000); available = requested - (unsigned) (total_mem / 0x100000); - RTE_LOG(INFO, EAL, "Not enough memory available! Requested: %uMB," + RTE_LOG(ERR, EAL, "Not enough memory available! Requested: %uMB," " available: %uMB\n", requested, available); return -1; } return total_num_pages; } +static inline size_t +eal_get_hugepage_mem_size(void) +{ + uint64_t size = 0; + unsigned i, j; + + for (i = 0; i < internal_config.num_hugepage_sizes; i++) { + struct hugepage_info *hpi = &internal_config.hugepage_info[i]; + if (hpi->hugedir != NULL) { + for (j = 0; j < RTE_MAX_NUMA_NODES; j++) { + size += hpi->hugepage_sz * hpi->num_pages[j]; + } + } + } + + return (size < SIZE_MAX) ? (size_t)(size) : SIZE_MAX; +} + +static struct sigaction huge_action_old; +static int huge_need_recover; + +static void +huge_register_sigbus(void) +{ + sigset_t mask; + struct sigaction action; + + sigemptyset(&mask); + sigaddset(&mask, SIGBUS); + action.sa_flags = 0; + action.sa_mask = mask; + action.sa_handler = huge_sigbus_handler; + + huge_need_recover = !sigaction(SIGBUS, &action, &huge_action_old); +} + +static void +huge_recover_sigbus(void) +{ + if (huge_need_recover) { + sigaction(SIGBUS, &huge_action_old, NULL); + huge_need_recover = 0; + } +} + /* * Prepare physical memory mapping: fill configuration structure with * these infos, return 0 on success. @@ -748,43 +987,54 @@ calc_num_pages_per_socket(uint64_t * memory, * 6. unmap the first mapping * 7. fill memsegs in configuration with contiguous zones */ -static int +int rte_eal_hugepage_init(void) { struct rte_mem_config *mcfg; - struct hugepage *hugepage, *tmp_hp = NULL; + struct hugepage_file *hugepage = NULL, *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; + test_phys_addrs_available(); + 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); - mcfg->memseg[0].phys_addr = (phys_addr_t)(uintptr_t)addr; + 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; + } + if (rte_eal_iova_mode() == RTE_IOVA_VA) + mcfg->memseg[0].iova = (uintptr_t)addr; + else + mcfg->memseg[0].iova = RTE_BAD_IOVA; mcfg->memseg[0].addr = addr; + mcfg->memseg[0].hugepage_sz = RTE_PGSIZE_4K; mcfg->memseg[0].len = internal_config.memory; mcfg->memseg[0].socket_id = 0; return 0; } - /* calculate total number of hugepages available. at this point we haven't * yet started sorting them so they all are on socket 0 */ for (i = 0; i < (int) internal_config.num_hugepage_sizes; i++) { /* 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]; } /* @@ -793,16 +1043,24 @@ rte_eal_hugepage_init(void) * 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 */ + huge_register_sigbus(); + + /* make a copy of socket_mem, needed for balanced allocation. */ + for (i = 0; i < RTE_MAX_NUMA_NODES; i++) + memory[i] = internal_config.socket_mem[i]; + + /* map all hugepages and sort them */ for (i = 0; i < (int)internal_config.num_hugepage_sizes; i ++){ + unsigned pages_old, pages_new; struct hugepage_info *hpi; /* @@ -812,21 +1070,43 @@ rte_eal_hugepage_init(void) */ hpi = &internal_config.hugepage_info[i]; - if (hpi->num_pages == 0) + if (hpi->num_pages[0] == 0) continue; /* map all hugepages available */ - if (map_all_hugepages(&tmp_hp[hp_offset], hpi, 1) < 0){ - RTE_LOG(DEBUG, EAL, "Failed to mmap %u MB hugepages\n", - (unsigned)(hpi->hugepage_sz / 0x100000)); - goto fail; + pages_old = hpi->num_pages[0]; + pages_new = map_all_hugepages(&tmp_hp[hp_offset], hpi, + memory, 1); + if (pages_new < pages_old) { + RTE_LOG(DEBUG, EAL, + "%d not %d hugepages of size %u MB allocated\n", + pages_new, pages_old, + (unsigned)(hpi->hugepage_sz / 0x100000)); + + int pages = pages_old - pages_new; + + nr_hugepages -= pages; + hpi->num_pages[0] = pages_new; + if (pages_new == 0) + continue; } - /* find physical addresses and sockets for each hugepage */ - if (find_physaddr(&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 (phys_addrs_available) { + /* find physical addresses for each hugepage */ + if (find_physaddrs(&tmp_hp[hp_offset], hpi) < 0) { + RTE_LOG(DEBUG, EAL, "Failed to find phys addr " + "for %u MB pages\n", + (unsigned int)(hpi->hugepage_sz / 0x100000)); + goto fail; + } + } else { + /* set physical addresses for each hugepage */ + if (set_physaddrs(&tmp_hp[hp_offset], hpi) < 0) { + RTE_LOG(DEBUG, EAL, "Failed to set phys addr " + "for %u MB pages\n", + (unsigned int)(hpi->hugepage_sz / 0x100000)); + goto fail; + } } if (find_numasocket(&tmp_hp[hp_offset], hpi) < 0){ @@ -835,12 +1115,13 @@ rte_eal_hugepage_init(void) goto fail; } - if (sort_by_physaddr(&tmp_hp[hp_offset], hpi) < 0) - goto fail; + qsort(&tmp_hp[hp_offset], hpi->num_pages[0], + sizeof(struct hugepage_file), cmp_physaddr); /* 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", + if (map_all_hugepages(&tmp_hp[hp_offset], hpi, NULL, 0) != + hpi->num_pages[0]) { + RTE_LOG(ERR, EAL, "Failed to remap %u MB pages\n", (unsigned)(hpi->hugepage_sz / 0x100000)); goto fail; } @@ -853,17 +1134,27 @@ rte_eal_hugepage_init(void) hp_offset += hpi->num_pages[0]; } + huge_recover_sigbus(); + + if (internal_config.memory == 0 && internal_config.force_sockets == 0) + internal_config.memory = eal_get_hugepage_mem_size(); + + nr_hugefiles = nr_hugepages; + + /* clean out the numbers of pages */ for (i = 0; i < (int) internal_config.num_hugepage_sizes; i++) for (j = 0; j < RTE_MAX_NUMA_NODES; j++) 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++) { + const int nb_hpsizes = RTE_MIN(MAX_HUGEPAGE_SIZES, + (int)internal_config.num_hugepage_sizes); + for (j = 0; j < nb_hpsizes; j++) { if (tmp_hp[i].size == internal_config.hugepage_info[j].hugepage_sz) { internal_config.hugepage_info[j].num_pages[socket]++; @@ -876,37 +1167,38 @@ rte_eal_hugepage_init(void) 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! */ for (i = 0; i < (int) internal_config.num_hugepage_sizes; i++) { for (j = 0; j < RTE_MAX_NUMA_NODES; j++) { if (used_hp[i].num_pages[j] > 0) { - RTE_LOG(INFO, EAL, - "Requesting %u pages of size %uMB" - " from socket %i\n", - used_hp[i].num_pages[j], - (unsigned) - (used_hp[i].hugepage_sz / 0x100000), - j); + RTE_LOG(DEBUG, EAL, + "Requesting %u pages of size %uMB" + " from socket %i\n", + used_hp[i].num_pages[j], + (unsigned) + (used_hp[i].hugepage_sz / 0x100000), + j); } } } /* 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). @@ -923,19 +1215,26 @@ rte_eal_hugepage_init(void) * 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 the hugepage backing files */ + if (internal_config.hugepage_unlink && + unlink_hugepage_files(tmp_hp, internal_config.num_hugepage_sizes) < 0) { + RTE_LOG(ERR, EAL, "Unlinking hugepage files failed!\n"); + goto fail; + } + /* free the temporary hugepage table */ free(tmp_hp); tmp_hp = NULL; - memset(mcfg->memseg, 0, sizeof(mcfg->memseg)); + /* first memseg index shall be 0 after incrementing it below */ j = -1; - for (i = 0; i < nrpages; i++) { + for (i = 0; i < nr_hugefiles; i++) { new_memseg = 0; /* if this is a new section, create a new memseg */ @@ -945,19 +1244,32 @@ rte_eal_hugepage_init(void) 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; if (j == RTE_MAX_MEMSEG) break; - mcfg->memseg[j].phys_addr = hugepage[i].physaddr; + mcfg->memseg[j].iova = hugepage[i].physaddr; mcfg->memseg[j].addr = hugepage[i].final_va; mcfg->memseg[j].len = hugepage[i].size; mcfg->memseg[j].socket_id = hugepage[i].socket_id; @@ -965,29 +1277,38 @@ rte_eal_hugepage_init(void) } /* 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].iova = 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; } - if (i < nrpages) { + 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, nrpages, RTE_STR(CONFIG_RTE_MAX_MEMSEG), + i, nr_hugefiles, RTE_STR(CONFIG_RTE_MAX_MEMSEG), RTE_MAX_MEMSEG); - return (-ENOMEM); + goto fail; } - - return 0; + munmap(hugepage, nr_hugefiles * sizeof(struct hugepage_file)); + return 0; fail: - if (tmp_hp) - free(tmp_hp); + huge_recover_sigbus(); + free(tmp_hp); + if (hugepage != NULL) + munmap(hugepage, nr_hugefiles * sizeof(struct hugepage_file)); + return -1; } @@ -1009,14 +1330,15 @@ getFileSize(int fd) * configuration and finds the hugepages which form that segment, mapping them * in order to form a contiguous block in the virtual memory space */ -static int +int rte_eal_hugepage_attach(void) { const struct rte_mem_config *mcfg = rte_eal_get_configuration()->mem_config; - const struct hugepage *hp = NULL; + struct hugepage_file *hp = NULL; unsigned num_hp = 0; unsigned i, s = 0; /* s used to track the segment number */ - off_t size; + unsigned max_seg = RTE_MAX_MEMSEG; + off_t size = 0; int fd, fd_zero = -1, fd_hugepage = -1; if (aslr_enabled() > 0) { @@ -1026,6 +1348,8 @@ rte_eal_hugepage_attach(void) "into secondary processes\n"); } + test_phys_addrs_available(); + fd_zero = open("/dev/zero", O_RDONLY); if (fd_zero < 0) { RTE_LOG(ERR, EAL, "Could not open /dev/zero\n"); @@ -1037,37 +1361,77 @@ rte_eal_hugepage_attach(void) 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; + + /* + * 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, +#ifdef RTE_ARCH_PPC_64 + MAP_PRIVATE | MAP_ANONYMOUS | MAP_HUGETLB, +#else + MAP_PRIVATE, +#endif + fd_zero, 0); + if (base_addr == MAP_FAILED || + base_addr != mcfg->memseg[s].addr) { + max_seg = s; + if (base_addr != MAP_FAILED) { + /* errno is stale, don't use */ + RTE_LOG(ERR, EAL, "Could not mmap %llu bytes " + "in /dev/zero at [%p], got [%p] - " + "please use '--base-virtaddr' option\n", + (unsigned long long)mcfg->memseg[s].len, + mcfg->memseg[s].addr, base_addr); + munmap(base_addr, mcfg->memseg[s].len); + } else { + RTE_LOG(ERR, EAL, "Could not mmap %llu bytes " + "in /dev/zero at [%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) { + if (hp == MAP_FAILED) { RTE_LOG(ERR, EAL, "Could not mmap %s\n", eal_hugepage_info_path()); 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; + /* + * free previously mapped memory so we can map the + * hugepages into the space */ - 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; - } - /* free 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 @@ -1081,27 +1445,35 @@ rte_eal_hugepage_attach(void) hp[i].filepath); goto error; } + mapping_size = hp[i].size; 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(hp, size); close(fd_zero); close(fd_hugepage); return 0; error: + for (i = 0; i < max_seg && mcfg->memseg[i].len > 0; i++) + munmap(mcfg->memseg[i].addr, mcfg->memseg[i].len); + if (hp != NULL && hp != MAP_FAILED) + munmap(hp, size); if (fd_zero >= 0) close(fd_zero); if (fd_hugepage >= 0) @@ -1109,35 +1481,8 @@ error: return -1; } -static int -rte_eal_memdevice_init(void) -{ - struct rte_config *config; - - if (rte_eal_process_type() == RTE_PROC_SECONDARY) - return 0; - - config = rte_eal_get_configuration(); - config->mem_config->nchannel = internal_config.force_nchannel; - config->mem_config->nrank = internal_config.force_nrank; - - return 0; -} - - -/* init memory subsystem */ int -rte_eal_memory_init(void) +rte_eal_using_phys_addrs(void) { - RTE_LOG(INFO, EAL, "Setting up hugepage memory...\n"); - const int retval = rte_eal_process_type() == RTE_PROC_PRIMARY ? - rte_eal_hugepage_init() : - rte_eal_hugepage_attach(); - if (retval < 0) - return -1; - - if (internal_config.no_shconf == 0 && rte_eal_memdevice_init() < 0) - return -1; - - return 0; + return phys_addrs_available; }