X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=lib%2Flibrte_eal%2Fcommon%2Feal_common_memzone.c;h=ea072a25b5195be8e991a0990ca0b7b168d087a1;hb=224374cc0e3ca44af5141fb7035a97f338d00c18;hp=dae4ea074eb7505fa099810feb26a0ea7530ec70;hpb=af75078fece3615088e561357c1e97603e43a5fe;p=dpdk.git diff --git a/lib/librte_eal/common/eal_common_memzone.c b/lib/librte_eal/common/eal_common_memzone.c index dae4ea074e..ea072a25b5 100644 --- a/lib/librte_eal/common/eal_common_memzone.c +++ b/lib/librte_eal/common/eal_common_memzone.c @@ -1,36 +1,34 @@ /*- * 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 + * + * * 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 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 + * + * 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. - * - * version: DPDK.L.1.2.3-3 */ #include @@ -45,291 +43,364 @@ #include #include #include -#include #include +#include #include #include #include #include +#include "malloc_heap.h" +#include "malloc_elem.h" #include "eal_private.h" -/* internal copy of free memory segments */ -static struct rte_memseg free_memseg[RTE_MAX_MEMSEG]; +static inline const struct rte_memzone * +memzone_lookup_thread_unsafe(const char *name) +{ + const struct rte_mem_config *mcfg; + const struct rte_memzone *mz; + unsigned i = 0; -/* pointer to last reserved memzone */ -static unsigned memzone_idx; + /* get pointer to global configuration */ + mcfg = rte_eal_get_configuration()->mem_config; -/* - * Return a pointer to a correctly filled memzone descriptor. If the - * allocation cannot be done, return NULL. - */ -const struct rte_memzone * -rte_memzone_reserve(const char *name, uint64_t len, int socket_id, - unsigned flags) -{ - return rte_memzone_reserve_aligned(name, - len, socket_id, flags, CACHE_LINE_SIZE); + /* + * the algorithm is not optimal (linear), but there are few + * zones and this function should be called at init only + */ + for (i = 0; i < RTE_MAX_MEMZONE; i++) { + mz = &mcfg->memzone[i]; + if (mz->addr != NULL && !strncmp(name, mz->name, RTE_MEMZONE_NAMESIZE)) + return &mcfg->memzone[i]; + } + + return NULL; } -/* - * Return a pointer to a correctly filled memzone descriptor (with a - * specified alignment). If the allocation cannot be done, return NULL. - */ -const struct rte_memzone * -rte_memzone_reserve_aligned(const char *name, uint64_t len, - int socket_id, unsigned flags, unsigned align) +static inline struct rte_memzone * +get_next_free_memzone(void) { - struct rte_config *config; + struct rte_mem_config *mcfg; unsigned i = 0; - int memseg_idx = -1; - uint64_t requested_len; - uint64_t memseg_len = 0; - phys_addr_t memseg_physaddr; - void *memseg_addr; - uintptr_t addr_offset; - - /* if secondary processes return error */ - if (rte_eal_process_type() == RTE_PROC_SECONDARY){ - RTE_LOG(ERR, EAL, "%s(): Not allowed in secondary process\n", __func__); - rte_errno = E_RTE_SECONDARY; - return NULL; + + /* get pointer to global configuration */ + mcfg = rte_eal_get_configuration()->mem_config; + + for (i = 0; i < RTE_MAX_MEMZONE; i++) { + if (mcfg->memzone[i].addr == NULL) + return &mcfg->memzone[i]; } - /* if alignment is not a power of two */ - if (!rte_is_power_of_2(align)) { - RTE_LOG(ERR, EAL, "%s(): Invalid alignment: %u\n", __func__, - align); - rte_errno = EINVAL; - return NULL; + return NULL; +} + +/* This function will return the greatest free block if a heap has been + * specified. If no heap has been specified, it will return the heap and + * length of the greatest free block available in all heaps */ +static size_t +find_heap_max_free_elem(int *s, unsigned align) +{ + struct rte_mem_config *mcfg; + struct rte_malloc_socket_stats stats; + int i, socket = *s; + size_t len = 0; + + /* get pointer to global configuration */ + mcfg = rte_eal_get_configuration()->mem_config; + + for (i = 0; i < RTE_MAX_NUMA_NODES; i++) { + if ((socket != SOCKET_ID_ANY) && (socket != i)) + continue; + + malloc_heap_get_stats(&mcfg->malloc_heaps[i], &stats); + if (stats.greatest_free_size > len) { + len = stats.greatest_free_size; + *s = i; + } } - /* alignment less than cache size is not allowed */ - if (align < CACHE_LINE_SIZE) - align = CACHE_LINE_SIZE; + if (len < MALLOC_ELEM_OVERHEAD + align) + return 0; + + return len - MALLOC_ELEM_OVERHEAD - align; +} + +static const struct rte_memzone * +memzone_reserve_aligned_thread_unsafe(const char *name, size_t len, + int socket_id, unsigned flags, unsigned align, unsigned bound) +{ + struct rte_memzone *mz; + struct rte_mem_config *mcfg; + size_t requested_len; + int socket, i; /* get pointer to global configuration */ - config = rte_eal_get_configuration(); + mcfg = rte_eal_get_configuration()->mem_config; /* no more room in config */ - if (memzone_idx >= RTE_MAX_MEMZONE) { + if (mcfg->memzone_cnt >= RTE_MAX_MEMZONE) { RTE_LOG(ERR, EAL, "%s(): No more room in config\n", __func__); rte_errno = ENOSPC; return NULL; } - /* both sizes cannot be explicitly called for */ - if ((flags & RTE_MEMZONE_1GB) && (flags & RTE_MEMZONE_2MB)) { - rte_errno = EINVAL; + if (strlen(name) > sizeof(mz->name) - 1) { + RTE_LOG(DEBUG, EAL, "%s(): memzone <%s>: name too long\n", + __func__, name); + rte_errno = ENAMETOOLONG; return NULL; } /* zone already exist */ - if (rte_memzone_lookup(name) != NULL) { + if ((memzone_lookup_thread_unsafe(name)) != NULL) { RTE_LOG(DEBUG, EAL, "%s(): memzone <%s> already exists\n", __func__, name); rte_errno = EEXIST; return NULL; } - /* align length on cache boundary */ - len += CACHE_LINE_MASK; - len &= ~((uint64_t)CACHE_LINE_MASK); - - - - /* save requested length */ - requested_len = len; - - /* reserve extra space for future alignment */ - if (len) - len += align; + /* if alignment is not a power of two */ + if (align && !rte_is_power_of_2(align)) { + RTE_LOG(ERR, EAL, "%s(): Invalid alignment: %u\n", __func__, + align); + rte_errno = EINVAL; + return NULL; + } - /* find the smallest segment matching requirements */ - for (i = 0; i < RTE_MAX_MEMSEG; i++) { + /* alignment less than cache size is not allowed */ + if (align < RTE_CACHE_LINE_SIZE) + align = RTE_CACHE_LINE_SIZE; - /* last segment */ - if (free_memseg[i].addr == NULL) - break; + /* align length on cache boundary. Check for overflow before doing so */ + if (len > SIZE_MAX - RTE_CACHE_LINE_MASK) { + rte_errno = EINVAL; /* requested size too big */ + return NULL; + } - /* empty segment, skip it */ - if (free_memseg[i].len == 0) - continue; + len += RTE_CACHE_LINE_MASK; + len &= ~((size_t) RTE_CACHE_LINE_MASK); - /* bad socket ID */ - if (socket_id != SOCKET_ID_ANY && - socket_id != free_memseg[i].socket_id) - continue; + /* save minimal requested length */ + requested_len = RTE_MAX((size_t)RTE_CACHE_LINE_SIZE, len); - /* check len */ - if (len != 0 && len > free_memseg[i].len) - continue; + /* check that boundary condition is valid */ + if (bound != 0 && (requested_len > bound || !rte_is_power_of_2(bound))) { + rte_errno = EINVAL; + return NULL; + } - /* check flags for hugepage sizes */ - if ((flags & RTE_MEMZONE_2MB) && - free_memseg[i].hugepage_sz == RTE_PGSIZE_1G ) - continue; - if ((flags & RTE_MEMZONE_1GB) && - free_memseg[i].hugepage_sz == RTE_PGSIZE_2M ) - continue; + if ((socket_id != SOCKET_ID_ANY) && + (socket_id >= RTE_MAX_NUMA_NODES || socket_id < 0)) { + rte_errno = EINVAL; + return NULL; + } - /* this segment is the best until now */ - if (memseg_idx == -1) { - memseg_idx = i; - memseg_len = free_memseg[i].len; - } - /* find the biggest contiguous zone */ - else if (len == 0) { - if (free_memseg[i].len > memseg_len) { - memseg_idx = i; - memseg_len = free_memseg[i].len; + if (!rte_eal_has_hugepages()) + socket_id = SOCKET_ID_ANY; + + if (len == 0) { + if (bound != 0) + requested_len = bound; + else { + requested_len = find_heap_max_free_elem(&socket_id, align); + if (requested_len == 0) { + rte_errno = ENOMEM; + return NULL; } } - /* - * find the smallest (we already checked that current - * zone length is > len - */ - else if (free_memseg[i].len < memseg_len) { - memseg_idx = i; - memseg_len = free_memseg[i].len; + } + + if (socket_id == SOCKET_ID_ANY) + socket = malloc_get_numa_socket(); + else + socket = socket_id; + + /* allocate memory on heap */ + void *mz_addr = malloc_heap_alloc(&mcfg->malloc_heaps[socket], NULL, + requested_len, flags, align, bound); + + if ((mz_addr == NULL) && (socket_id == SOCKET_ID_ANY)) { + /* try other heaps */ + for (i = 0; i < RTE_MAX_NUMA_NODES; i++) { + if (socket == i) + continue; + + mz_addr = malloc_heap_alloc(&mcfg->malloc_heaps[i], + NULL, requested_len, flags, align, bound); + if (mz_addr != NULL) + break; } } - /* no segment found */ - if (memseg_idx == -1) { - /* - * If RTE_MEMZONE_SIZE_HINT_ONLY flag is specified, - * try allocating again without the size parameter otherwise -fail. - */ - if ((flags & RTE_MEMZONE_SIZE_HINT_ONLY) && - ((flags & RTE_MEMZONE_1GB) || (flags & RTE_MEMZONE_2MB))) - return rte_memzone_reserve_aligned(name, len - align, - socket_id, 0, align); - - RTE_LOG(ERR, EAL, "%s(): No appropriate segment found\n", __func__); + if (mz_addr == NULL) { rte_errno = ENOMEM; return NULL; } - /* get offset needed to adjust alignment */ - addr_offset = (uintptr_t) RTE_PTR_SUB( - RTE_ALIGN_CEIL(free_memseg[memseg_idx].addr, (uintptr_t) align), - (uintptr_t) free_memseg[memseg_idx].addr); + const struct malloc_elem *elem = malloc_elem_from_data(mz_addr); - /* save aligned physical and virtual addresses */ - memseg_physaddr = free_memseg[memseg_idx].phys_addr + addr_offset; - memseg_addr = RTE_PTR_ADD(free_memseg[memseg_idx].addr, addr_offset); + /* fill the zone in config */ + mz = get_next_free_memzone(); - /* if we are looking for a biggest memzone */ - if (requested_len == 0) - requested_len = memseg_len - addr_offset; + if (mz == NULL) { + RTE_LOG(ERR, EAL, "%s(): Cannot find free memzone but there is room " + "in config!\n", __func__); + rte_errno = ENOSPC; + return NULL; + } - /* set length to correct value */ - len = addr_offset + requested_len; + mcfg->memzone_cnt++; + snprintf(mz->name, sizeof(mz->name), "%s", name); + mz->iova = rte_malloc_virt2iova(mz_addr); + mz->addr = mz_addr; + mz->len = (requested_len == 0 ? elem->size : requested_len); + mz->hugepage_sz = elem->ms->hugepage_sz; + mz->socket_id = elem->ms->socket_id; + mz->flags = 0; + mz->memseg_id = elem->ms - rte_eal_get_configuration()->mem_config->memseg; - /* update our internal state */ - free_memseg[memseg_idx].len -= len; - free_memseg[memseg_idx].phys_addr += len; - free_memseg[memseg_idx].addr = - (char *)free_memseg[memseg_idx].addr + len; + return mz; +} - /* fill the zone in config */ - struct rte_memzone *mz = &config->mem_config->memzone[memzone_idx++]; - rte_snprintf(mz->name, sizeof(mz->name), "%s", name); - mz->phys_addr = memseg_physaddr; - mz->addr = memseg_addr; - mz->len = requested_len; - mz->hugepage_sz = free_memseg[memseg_idx].hugepage_sz; - mz->socket_id = free_memseg[memseg_idx].socket_id; - mz->flags = 0; +static const struct rte_memzone * +rte_memzone_reserve_thread_safe(const char *name, size_t len, + int socket_id, unsigned flags, unsigned align, + unsigned bound) +{ + struct rte_mem_config *mcfg; + const struct rte_memzone *mz = NULL; + + /* get pointer to global configuration */ + mcfg = rte_eal_get_configuration()->mem_config; + + rte_rwlock_write_lock(&mcfg->mlock); + + mz = memzone_reserve_aligned_thread_unsafe( + name, len, socket_id, flags, align, bound); + + rte_rwlock_write_unlock(&mcfg->mlock); return mz; } +/* + * Return a pointer to a correctly filled memzone descriptor (with a + * specified alignment and boundary). If the allocation cannot be done, + * return NULL. + */ +const struct rte_memzone * +rte_memzone_reserve_bounded(const char *name, size_t len, int socket_id, + unsigned flags, unsigned align, unsigned bound) +{ + return rte_memzone_reserve_thread_safe(name, len, socket_id, flags, + align, bound); +} + +/* + * Return a pointer to a correctly filled memzone descriptor (with a + * specified alignment). If the allocation cannot be done, return NULL. + */ +const struct rte_memzone * +rte_memzone_reserve_aligned(const char *name, size_t len, int socket_id, + unsigned flags, unsigned align) +{ + return rte_memzone_reserve_thread_safe(name, len, socket_id, flags, + align, 0); +} + +/* + * Return a pointer to a correctly filled memzone descriptor. If the + * allocation cannot be done, return NULL. + */ +const struct rte_memzone * +rte_memzone_reserve(const char *name, size_t len, int socket_id, + unsigned flags) +{ + return rte_memzone_reserve_thread_safe(name, len, socket_id, + flags, RTE_CACHE_LINE_SIZE, 0); +} + +int +rte_memzone_free(const struct rte_memzone *mz) +{ + struct rte_mem_config *mcfg; + int ret = 0; + void *addr; + unsigned idx; + + if (mz == NULL) + return -EINVAL; + + mcfg = rte_eal_get_configuration()->mem_config; + + rte_rwlock_write_lock(&mcfg->mlock); + + idx = ((uintptr_t)mz - (uintptr_t)mcfg->memzone); + idx = idx / sizeof(struct rte_memzone); + + addr = mcfg->memzone[idx].addr; + if (addr == NULL) + ret = -EINVAL; + else if (mcfg->memzone_cnt == 0) { + rte_panic("%s(): memzone address not NULL but memzone_cnt is 0!\n", + __func__); + } else { + memset(&mcfg->memzone[idx], 0, sizeof(mcfg->memzone[idx])); + mcfg->memzone_cnt--; + } + + rte_rwlock_write_unlock(&mcfg->mlock); + + rte_free(addr); + + return ret; +} + /* * Lookup for the memzone identified by the given name */ const struct rte_memzone * rte_memzone_lookup(const char *name) { - const struct rte_mem_config *mcfg; - unsigned i = 0; + struct rte_mem_config *mcfg; + const struct rte_memzone *memzone = NULL; - /* get pointer to global configuration */ mcfg = rte_eal_get_configuration()->mem_config; - /* - * the algorithm is not optimal (linear), but there are few - * zones and this function should be called at init only - */ - for (i = 0; i < RTE_MAX_MEMZONE && mcfg->memzone[i].addr != NULL; i++) { - if (!strncmp(name, mcfg->memzone[i].name, RTE_MEMZONE_NAMESIZE)) - return &mcfg->memzone[i]; - } - return NULL; + rte_rwlock_read_lock(&mcfg->mlock); + + memzone = memzone_lookup_thread_unsafe(name); + + rte_rwlock_read_unlock(&mcfg->mlock); + + return memzone; } /* Dump all reserved memory zones on console */ void -rte_memzone_dump(void) +rte_memzone_dump(FILE *f) { - const struct rte_mem_config *mcfg; + struct rte_mem_config *mcfg; unsigned i = 0; /* get pointer to global configuration */ mcfg = rte_eal_get_configuration()->mem_config; + rte_rwlock_read_lock(&mcfg->mlock); /* dump all zones */ for (i=0; imemzone[i].addr == NULL) break; - printf("name:<%s>, phys:0x%"PRIx64", len:0x%"PRIx64"" - ", virt:%p, socket_id:%"PRId32"\n", + fprintf(f, "Zone %u: name:<%s>, IO:0x%"PRIx64", len:0x%zx" + ", virt:%p, socket_id:%"PRId32", flags:%"PRIx32"\n", i, mcfg->memzone[i].name, - mcfg->memzone[i].phys_addr, + mcfg->memzone[i].iova, mcfg->memzone[i].len, mcfg->memzone[i].addr, - mcfg->memzone[i].socket_id); - } -} - -/* - * called by init: modify the free memseg list to have cache-aligned - * addresses and cache-aligned lengths - */ -static int -memseg_sanitize(struct rte_memseg *memseg) -{ - unsigned phys_align; - unsigned virt_align; - unsigned off; - - phys_align = memseg->phys_addr & CACHE_LINE_MASK; - virt_align = (unsigned long)memseg->addr & CACHE_LINE_MASK; - - /* - * sanity check: phys_addr and addr must have the same - * alignment - */ - if (phys_align != virt_align) - return -1; - - /* memseg is really too small, don't bother with it */ - if (memseg->len < (2 * CACHE_LINE_SIZE)) { - memseg->len = 0; - return 0; + mcfg->memzone[i].socket_id, + mcfg->memzone[i].flags); } - - /* align start address */ - off = (CACHE_LINE_SIZE - phys_align) & CACHE_LINE_MASK; - memseg->phys_addr += off; - memseg->addr = (char *)memseg->addr + off; - memseg->len -= off; - - /* align end address */ - memseg->len &= ~((uint64_t)CACHE_LINE_MASK); - - return 0; + rte_rwlock_read_unlock(&mcfg->mlock); } /* @@ -338,39 +409,46 @@ memseg_sanitize(struct rte_memseg *memseg) int rte_eal_memzone_init(void) { - struct rte_config *config; + struct rte_mem_config *mcfg; const struct rte_memseg *memseg; - unsigned i = 0; + + /* get pointer to global configuration */ + mcfg = rte_eal_get_configuration()->mem_config; /* secondary processes don't need to initialise anything */ if (rte_eal_process_type() == RTE_PROC_SECONDARY) return 0; - /* get pointer to global configuration */ - config = rte_eal_get_configuration(); - memseg = rte_eal_get_physmem_layout(); if (memseg == NULL) { RTE_LOG(ERR, EAL, "%s(): Cannot get physical layout\n", __func__); return -1; } - /* duplicate the memsegs from config */ - memcpy(free_memseg, memseg, sizeof(free_memseg)); - - /* make all zones cache-aligned */ - for (i=0; imlock); /* delete all zones */ - memzone_idx = 0; - memset(config->mem_config->memzone, 0, sizeof(config->mem_config->memzone)); + mcfg->memzone_cnt = 0; + memset(mcfg->memzone, 0, sizeof(mcfg->memzone)); + + rte_rwlock_write_unlock(&mcfg->mlock); - return 0; + return rte_eal_malloc_heap_init(); +} + +/* Walk all reserved memory zones */ +void rte_memzone_walk(void (*func)(const struct rte_memzone *, void *), + void *arg) +{ + struct rte_mem_config *mcfg; + unsigned i; + + mcfg = rte_eal_get_configuration()->mem_config; + + rte_rwlock_read_lock(&mcfg->mlock); + for (i=0; imemzone[i].addr != NULL) + (*func)(&mcfg->memzone[i], arg); + } + rte_rwlock_read_unlock(&mcfg->mlock); }