diff options
Diffstat (limited to 'drivers/net/mlx4/mlx4_mr.c')
| -rw-r--r-- | drivers/net/mlx4/mlx4_mr.c | 1247 |
1 files changed, 1081 insertions, 166 deletions
diff --git a/drivers/net/mlx4/mlx4_mr.c b/drivers/net/mlx4/mlx4_mr.c index 9a1e4de3..d23d3c61 100644 --- a/drivers/net/mlx4/mlx4_mr.c +++ b/drivers/net/mlx4/mlx4_mr.c @@ -1,6 +1,6 @@ /* SPDX-License-Identifier: BSD-3-Clause * Copyright 2017 6WIND S.A. - * Copyright 2017 Mellanox + * Copyright 2017 Mellanox Technologies, Ltd */ /** @@ -30,237 +30,1152 @@ #include <rte_malloc.h> #include <rte_memory.h> #include <rte_mempool.h> -#include <rte_spinlock.h> +#include <rte_rwlock.h> #include "mlx4_glue.h" +#include "mlx4_mr.h" #include "mlx4_rxtx.h" #include "mlx4_utils.h" -struct mlx4_check_mempool_data { +struct mr_find_contig_memsegs_data { + uintptr_t addr; + uintptr_t start; + uintptr_t end; + const struct rte_memseg_list *msl; +}; + +struct mr_update_mp_data { + struct rte_eth_dev *dev; + struct mlx4_mr_ctrl *mr_ctrl; int ret; - char *start; - char *end; }; /** - * Called by mlx4_check_mempool() when iterating the memory chunks. - * - * @param[in] mp - * Pointer to memory pool (unused). - * @param[in, out] data - * Pointer to shared buffer with mlx4_check_mempool(). - * @param[in] memhdr - * Pointer to mempool chunk header. - * @param mem_idx - * Mempool element index (unused). + * Expand B-tree table to a given size. Can't be called with holding + * memory_hotplug_lock or priv->mr.rwlock due to rte_realloc(). + * + * @param bt + * Pointer to B-tree structure. + * @param n + * Number of entries for expansion. + * + * @return + * 0 on success, -1 on failure. */ -static void -mlx4_check_mempool_cb(struct rte_mempool *mp, void *opaque, - struct rte_mempool_memhdr *memhdr, - unsigned int mem_idx) +static int +mr_btree_expand(struct mlx4_mr_btree *bt, int n) +{ + void *mem; + int ret = 0; + + if (n <= bt->size) + return ret; + /* + * Downside of directly using rte_realloc() is that SOCKET_ID_ANY is + * used inside if there's no room to expand. Because this is a quite + * rare case and a part of very slow path, it is very acceptable. + * Initially cache_bh[] will be given practically enough space and once + * it is expanded, expansion wouldn't be needed again ever. + */ + mem = rte_realloc(bt->table, n * sizeof(struct mlx4_mr_cache), 0); + if (mem == NULL) { + /* Not an error, B-tree search will be skipped. */ + WARN("failed to expand MR B-tree (%p) table", (void *)bt); + ret = -1; + } else { + DEBUG("expanded MR B-tree table (size=%u)", n); + bt->table = mem; + bt->size = n; + } + return ret; +} + +/** + * Look up LKey from given B-tree lookup table, store the last index and return + * searched LKey. + * + * @param bt + * Pointer to B-tree structure. + * @param[out] idx + * Pointer to index. Even on search failure, returns index where it stops + * searching so that index can be used when inserting a new entry. + * @param addr + * Search key. + * + * @return + * Searched LKey on success, UINT32_MAX on no match. + */ +static uint32_t +mr_btree_lookup(struct mlx4_mr_btree *bt, uint16_t *idx, uintptr_t addr) +{ + struct mlx4_mr_cache *lkp_tbl; + uint16_t n; + uint16_t base = 0; + + assert(bt != NULL); + lkp_tbl = *bt->table; + n = bt->len; + /* First entry must be NULL for comparison. */ + assert(bt->len > 0 || (lkp_tbl[0].start == 0 && + lkp_tbl[0].lkey == UINT32_MAX)); + /* Binary search. */ + do { + register uint16_t delta = n >> 1; + + if (addr < lkp_tbl[base + delta].start) { + n = delta; + } else { + base += delta; + n -= delta; + } + } while (n > 1); + assert(addr >= lkp_tbl[base].start); + *idx = base; + if (addr < lkp_tbl[base].end) + return lkp_tbl[base].lkey; + /* Not found. */ + return UINT32_MAX; +} + +/** + * Insert an entry to B-tree lookup table. + * + * @param bt + * Pointer to B-tree structure. + * @param entry + * Pointer to new entry to insert. + * + * @return + * 0 on success, -1 on failure. + */ +static int +mr_btree_insert(struct mlx4_mr_btree *bt, struct mlx4_mr_cache *entry) { - struct mlx4_check_mempool_data *data = opaque; + struct mlx4_mr_cache *lkp_tbl; + uint16_t idx = 0; + size_t shift; - (void)mp; - (void)mem_idx; - /* It already failed, skip the next chunks. */ - if (data->ret != 0) + assert(bt != NULL); + assert(bt->len <= bt->size); + assert(bt->len > 0); + lkp_tbl = *bt->table; + /* Find out the slot for insertion. */ + if (mr_btree_lookup(bt, &idx, entry->start) != UINT32_MAX) { + DEBUG("abort insertion to B-tree(%p): already exist at" + " idx=%u [0x%" PRIxPTR ", 0x%" PRIxPTR ") lkey=0x%x", + (void *)bt, idx, entry->start, entry->end, entry->lkey); + /* Already exist, return. */ + return 0; + } + /* If table is full, return error. */ + if (unlikely(bt->len == bt->size)) { + bt->overflow = 1; + return -1; + } + /* Insert entry. */ + ++idx; + shift = (bt->len - idx) * sizeof(struct mlx4_mr_cache); + if (shift) + memmove(&lkp_tbl[idx + 1], &lkp_tbl[idx], shift); + lkp_tbl[idx] = *entry; + bt->len++; + DEBUG("inserted B-tree(%p)[%u]," + " [0x%" PRIxPTR ", 0x%" PRIxPTR ") lkey=0x%x", + (void *)bt, idx, entry->start, entry->end, entry->lkey); + return 0; +} + +/** + * Initialize B-tree and allocate memory for lookup table. + * + * @param bt + * Pointer to B-tree structure. + * @param n + * Number of entries to allocate. + * @param socket + * NUMA socket on which memory must be allocated. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx4_mr_btree_init(struct mlx4_mr_btree *bt, int n, int socket) +{ + if (bt == NULL) { + rte_errno = EINVAL; + return -rte_errno; + } + memset(bt, 0, sizeof(*bt)); + bt->table = rte_calloc_socket("B-tree table", + n, sizeof(struct mlx4_mr_cache), + 0, socket); + if (bt->table == NULL) { + rte_errno = ENOMEM; + ERROR("failed to allocate memory for btree cache on socket %d", + socket); + return -rte_errno; + } + bt->size = n; + /* First entry must be NULL for binary search. */ + (*bt->table)[bt->len++] = (struct mlx4_mr_cache) { + .lkey = UINT32_MAX, + }; + DEBUG("initialized B-tree %p with table %p", + (void *)bt, (void *)bt->table); + return 0; +} + +/** + * Free B-tree resources. + * + * @param bt + * Pointer to B-tree structure. + */ +void +mlx4_mr_btree_free(struct mlx4_mr_btree *bt) +{ + if (bt == NULL) return; - /* It is the first chunk. */ - if (data->start == NULL && data->end == NULL) { - data->start = memhdr->addr; - data->end = data->start + memhdr->len; + DEBUG("freeing B-tree %p with table %p", (void *)bt, (void *)bt->table); + rte_free(bt->table); + memset(bt, 0, sizeof(*bt)); +} + +#ifndef NDEBUG +/** + * Dump all the entries in a B-tree + * + * @param bt + * Pointer to B-tree structure. + */ +void +mlx4_mr_btree_dump(struct mlx4_mr_btree *bt) +{ + int idx; + struct mlx4_mr_cache *lkp_tbl; + + if (bt == NULL) return; + lkp_tbl = *bt->table; + for (idx = 0; idx < bt->len; ++idx) { + struct mlx4_mr_cache *entry = &lkp_tbl[idx]; + + DEBUG("B-tree(%p)[%u]," + " [0x%" PRIxPTR ", 0x%" PRIxPTR ") lkey=0x%x", + (void *)bt, idx, entry->start, entry->end, entry->lkey); } - if (data->end == memhdr->addr) { - data->end += memhdr->len; - return; +} +#endif + +/** + * Find virtually contiguous memory chunk in a given MR. + * + * @param dev + * Pointer to MR structure. + * @param[out] entry + * Pointer to returning MR cache entry. If not found, this will not be + * updated. + * @param start_idx + * Start index of the memseg bitmap. + * + * @return + * Next index to go on lookup. + */ +static int +mr_find_next_chunk(struct mlx4_mr *mr, struct mlx4_mr_cache *entry, + int base_idx) +{ + uintptr_t start = 0; + uintptr_t end = 0; + uint32_t idx = 0; + + for (idx = base_idx; idx < mr->ms_bmp_n; ++idx) { + if (rte_bitmap_get(mr->ms_bmp, idx)) { + const struct rte_memseg_list *msl; + const struct rte_memseg *ms; + + msl = mr->msl; + ms = rte_fbarray_get(&msl->memseg_arr, + mr->ms_base_idx + idx); + assert(msl->page_sz == ms->hugepage_sz); + if (!start) + start = ms->addr_64; + end = ms->addr_64 + ms->hugepage_sz; + } else if (start) { + /* Passed the end of a fragment. */ + break; + } } - if (data->start == (char *)memhdr->addr + memhdr->len) { - data->start -= memhdr->len; - return; + if (start) { + /* Found one chunk. */ + entry->start = start; + entry->end = end; + entry->lkey = rte_cpu_to_be_32(mr->ibv_mr->lkey); } - /* Error, mempool is not virtually contiguous. */ - data->ret = -1; + return idx; } /** - * Check if a mempool can be used: it must be virtually contiguous. + * Insert a MR to the global B-tree cache. It may fail due to low-on-memory. + * Then, this entry will have to be searched by mr_lookup_dev_list() in + * mlx4_mr_create() on miss. * - * @param[in] mp - * Pointer to memory pool. - * @param[out] start - * Pointer to the start address of the mempool virtual memory area. - * @param[out] end - * Pointer to the end address of the mempool virtual memory area. + * @param dev + * Pointer to Ethernet device. + * @param mr + * Pointer to MR to insert. * * @return - * 0 on success (mempool is virtually contiguous), -1 on error. + * 0 on success, -1 on failure. */ static int -mlx4_check_mempool(struct rte_mempool *mp, uintptr_t *start, uintptr_t *end) +mr_insert_dev_cache(struct rte_eth_dev *dev, struct mlx4_mr *mr) { - struct mlx4_check_mempool_data data; + struct priv *priv = dev->data->dev_private; + unsigned int n; - memset(&data, 0, sizeof(data)); - rte_mempool_mem_iter(mp, mlx4_check_mempool_cb, &data); - *start = (uintptr_t)data.start; - *end = (uintptr_t)data.end; - return data.ret; + DEBUG("port %u inserting MR(%p) to global cache", + dev->data->port_id, (void *)mr); + for (n = 0; n < mr->ms_bmp_n; ) { + struct mlx4_mr_cache entry = { 0, }; + + /* Find a contiguous chunk and advance the index. */ + n = mr_find_next_chunk(mr, &entry, n); + if (!entry.end) + break; + if (mr_btree_insert(&priv->mr.cache, &entry) < 0) { + /* + * Overflowed, but the global table cannot be expanded + * because of deadlock. + */ + return -1; + } + } + return 0; } /** - * Obtain a memory region from a memory pool. + * Look up address in the original global MR list. * - * If a matching memory region already exists, it is returned with its - * reference count incremented, otherwise a new one is registered. + * @param dev + * Pointer to Ethernet device. + * @param[out] entry + * Pointer to returning MR cache entry. If no match, this will not be updated. + * @param addr + * Search key. * - * @param priv - * Pointer to private structure. - * @param mp - * Pointer to memory pool. + * @return + * Found MR on match, NULL otherwise. + */ +static struct mlx4_mr * +mr_lookup_dev_list(struct rte_eth_dev *dev, struct mlx4_mr_cache *entry, + uintptr_t addr) +{ + struct priv *priv = dev->data->dev_private; + struct mlx4_mr *mr; + + /* Iterate all the existing MRs. */ + LIST_FOREACH(mr, &priv->mr.mr_list, mr) { + unsigned int n; + + if (mr->ms_n == 0) + continue; + for (n = 0; n < mr->ms_bmp_n; ) { + struct mlx4_mr_cache ret = { 0, }; + + n = mr_find_next_chunk(mr, &ret, n); + if (addr >= ret.start && addr < ret.end) { + /* Found. */ + *entry = ret; + return mr; + } + } + } + return NULL; +} + +/** + * Look up address on device. + * + * @param dev + * Pointer to Ethernet device. + * @param[out] entry + * Pointer to returning MR cache entry. If no match, this will not be updated. + * @param addr + * Search key. * * @return - * Memory region pointer, NULL in case of error and rte_errno is set. + * Searched LKey on success, UINT32_MAX on failure and rte_errno is set. */ -struct mlx4_mr * -mlx4_mr_get(struct priv *priv, struct rte_mempool *mp) +static uint32_t +mr_lookup_dev(struct rte_eth_dev *dev, struct mlx4_mr_cache *entry, + uintptr_t addr) { - const struct rte_memseg *ms = rte_eal_get_physmem_layout(); - uintptr_t start; - uintptr_t end; - unsigned int i; + struct priv *priv = dev->data->dev_private; + uint16_t idx; + uint32_t lkey = UINT32_MAX; struct mlx4_mr *mr; - if (mlx4_check_mempool(mp, &start, &end) != 0) { - rte_errno = EINVAL; - ERROR("mempool %p: not virtually contiguous", - (void *)mp); - return NULL; + /* + * If the global cache has overflowed since it failed to expand the + * B-tree table, it can't have all the existing MRs. Then, the address + * has to be searched by traversing the original MR list instead, which + * is very slow path. Otherwise, the global cache is all inclusive. + */ + if (!unlikely(priv->mr.cache.overflow)) { + lkey = mr_btree_lookup(&priv->mr.cache, &idx, addr); + if (lkey != UINT32_MAX) + *entry = (*priv->mr.cache.table)[idx]; + } else { + /* Falling back to the slowest path. */ + mr = mr_lookup_dev_list(dev, entry, addr); + if (mr != NULL) + lkey = entry->lkey; } - DEBUG("mempool %p area start=%p end=%p size=%zu", - (void *)mp, (void *)start, (void *)end, - (size_t)(end - start)); - /* Round start and end to page boundary if found in memory segments. */ - for (i = 0; (i < RTE_MAX_MEMSEG) && (ms[i].addr != NULL); ++i) { - uintptr_t addr = (uintptr_t)ms[i].addr; - size_t len = ms[i].len; - unsigned int align = ms[i].hugepage_sz; - - if ((start > addr) && (start < addr + len)) - start = RTE_ALIGN_FLOOR(start, align); - if ((end > addr) && (end < addr + len)) - end = RTE_ALIGN_CEIL(end, align); + assert(lkey == UINT32_MAX || (addr >= entry->start && + addr < entry->end)); + return lkey; +} + +/** + * Free MR resources. MR lock must not be held to avoid a deadlock. rte_free() + * can raise memory free event and the callback function will spin on the lock. + * + * @param mr + * Pointer to MR to free. + */ +static void +mr_free(struct mlx4_mr *mr) +{ + if (mr == NULL) + return; + DEBUG("freeing MR(%p):", (void *)mr); + if (mr->ibv_mr != NULL) + claim_zero(mlx4_glue->dereg_mr(mr->ibv_mr)); + if (mr->ms_bmp != NULL) + rte_bitmap_free(mr->ms_bmp); + rte_free(mr); +} + +/** + * Releass resources of detached MR having no online entry. + * + * @param dev + * Pointer to Ethernet device. + */ +static void +mlx4_mr_garbage_collect(struct rte_eth_dev *dev) +{ + struct priv *priv = dev->data->dev_private; + struct mlx4_mr *mr_next; + struct mlx4_mr_list free_list = LIST_HEAD_INITIALIZER(free_list); + + /* + * MR can't be freed with holding the lock because rte_free() could call + * memory free callback function. This will be a deadlock situation. + */ + rte_rwlock_write_lock(&priv->mr.rwlock); + /* Detach the whole free list and release it after unlocking. */ + free_list = priv->mr.mr_free_list; + LIST_INIT(&priv->mr.mr_free_list); + rte_rwlock_write_unlock(&priv->mr.rwlock); + /* Release resources. */ + mr_next = LIST_FIRST(&free_list); + while (mr_next != NULL) { + struct mlx4_mr *mr = mr_next; + + mr_next = LIST_NEXT(mr, mr); + mr_free(mr); } - DEBUG("mempool %p using start=%p end=%p size=%zu for MR", - (void *)mp, (void *)start, (void *)end, - (size_t)(end - start)); - rte_spinlock_lock(&priv->mr_lock); - LIST_FOREACH(mr, &priv->mr, next) - if (mp == mr->mp && start >= mr->start && end <= mr->end) - break; - if (mr) { - ++mr->refcnt; - goto release; +} + +/* Called during rte_memseg_contig_walk() by mlx4_mr_create(). */ +static int +mr_find_contig_memsegs_cb(const struct rte_memseg_list *msl, + const struct rte_memseg *ms, size_t len, void *arg) +{ + struct mr_find_contig_memsegs_data *data = arg; + + if (data->addr < ms->addr_64 || data->addr >= ms->addr_64 + len) + return 0; + /* Found, save it and stop walking. */ + data->start = ms->addr_64; + data->end = ms->addr_64 + len; + data->msl = msl; + return 1; +} + +/** + * Create a new global Memroy Region (MR) for a missing virtual address. + * Register entire virtually contiguous memory chunk around the address. + * + * @param dev + * Pointer to Ethernet device. + * @param[out] entry + * Pointer to returning MR cache entry, found in the global cache or newly + * created. If failed to create one, this will not be updated. + * @param addr + * Target virtual address to register. + * + * @return + * Searched LKey on success, UINT32_MAX on failure and rte_errno is set. + */ +static uint32_t +mlx4_mr_create(struct rte_eth_dev *dev, struct mlx4_mr_cache *entry, + uintptr_t addr) +{ + struct priv *priv = dev->data->dev_private; + struct rte_mem_config *mcfg = rte_eal_get_configuration()->mem_config; + const struct rte_memseg_list *msl; + const struct rte_memseg *ms; + struct mlx4_mr *mr = NULL; + size_t len; + uint32_t ms_n; + uint32_t bmp_size; + void *bmp_mem; + int ms_idx_shift = -1; + unsigned int n; + struct mr_find_contig_memsegs_data data = { + .addr = addr, + }; + struct mr_find_contig_memsegs_data data_re; + + DEBUG("port %u creating a MR using address (%p)", + dev->data->port_id, (void *)addr); + /* + * Release detached MRs if any. This can't be called with holding either + * memory_hotplug_lock or priv->mr.rwlock. MRs on the free list have + * been detached by the memory free event but it couldn't be released + * inside the callback due to deadlock. As a result, releasing resources + * is quite opportunistic. + */ + mlx4_mr_garbage_collect(dev); + /* + * Find out a contiguous virtual address chunk in use, to which the + * given address belongs, in order to register maximum range. In the + * best case where mempools are not dynamically recreated and + * '--socket-mem' is speicified as an EAL option, it is very likely to + * have only one MR(LKey) per a socket and per a hugepage-size even + * though the system memory is highly fragmented. + */ + if (!rte_memseg_contig_walk(mr_find_contig_memsegs_cb, &data)) { + WARN("port %u unable to find virtually contiguous" + " chunk for address (%p)." + " rte_memseg_contig_walk() failed.", + dev->data->port_id, (void *)addr); + rte_errno = ENXIO; + goto err_nolock; } - mr = rte_malloc(__func__, sizeof(*mr), 0); - if (!mr) { +alloc_resources: + /* Addresses must be page-aligned. */ + assert(rte_is_aligned((void *)data.start, data.msl->page_sz)); + assert(rte_is_aligned((void *)data.end, data.msl->page_sz)); + msl = data.msl; + ms = rte_mem_virt2memseg((void *)data.start, msl); + len = data.end - data.start; + assert(msl->page_sz == ms->hugepage_sz); + /* Number of memsegs in the range. */ + ms_n = len / msl->page_sz; + DEBUG("port %u extending %p to [0x%" PRIxPTR ", 0x%" PRIxPTR ")," + " page_sz=0x%" PRIx64 ", ms_n=%u", + dev->data->port_id, (void *)addr, + data.start, data.end, msl->page_sz, ms_n); + /* Size of memory for bitmap. */ + bmp_size = rte_bitmap_get_memory_footprint(ms_n); + mr = rte_zmalloc_socket(NULL, + RTE_ALIGN_CEIL(sizeof(*mr), + RTE_CACHE_LINE_SIZE) + + bmp_size, + RTE_CACHE_LINE_SIZE, msl->socket_id); + if (mr == NULL) { + WARN("port %u unable to allocate memory for a new MR of" + " address (%p).", + dev->data->port_id, (void *)addr); rte_errno = ENOMEM; - goto release; + goto err_nolock; } - *mr = (struct mlx4_mr){ - .start = start, - .end = end, - .refcnt = 1, - .priv = priv, - .mr = mlx4_glue->reg_mr(priv->pd, (void *)start, end - start, - IBV_ACCESS_LOCAL_WRITE), - .mp = mp, - }; - if (mr->mr) { - mr->lkey = mr->mr->lkey; - LIST_INSERT_HEAD(&priv->mr, mr, next); - } else { - rte_free(mr); - mr = NULL; - rte_errno = errno ? errno : EINVAL; + mr->msl = msl; + /* + * Save the index of the first memseg and initialize memseg bitmap. To + * see if a memseg of ms_idx in the memseg-list is still valid, check: + * rte_bitmap_get(mr->bmp, ms_idx - mr->ms_base_idx) + */ + mr->ms_base_idx = rte_fbarray_find_idx(&msl->memseg_arr, ms); + bmp_mem = RTE_PTR_ALIGN_CEIL(mr + 1, RTE_CACHE_LINE_SIZE); + mr->ms_bmp = rte_bitmap_init(ms_n, bmp_mem, bmp_size); + if (mr->ms_bmp == NULL) { + WARN("port %u unable to initialize bitamp for a new MR of" + " address (%p).", + dev->data->port_id, (void *)addr); + rte_errno = EINVAL; + goto err_nolock; + } + /* + * Should recheck whether the extended contiguous chunk is still valid. + * Because memory_hotplug_lock can't be held if there's any memory + * related calls in a critical path, resource allocation above can't be + * locked. If the memory has been changed at this point, try again with + * just single page. If not, go on with the big chunk atomically from + * here. + */ + rte_rwlock_read_lock(&mcfg->memory_hotplug_lock); + data_re = data; + if (len > msl->page_sz && + !rte_memseg_contig_walk(mr_find_contig_memsegs_cb, &data_re)) { + WARN("port %u unable to find virtually contiguous" + " chunk for address (%p)." + " rte_memseg_contig_walk() failed.", + dev->data->port_id, (void *)addr); + rte_errno = ENXIO; + goto err_memlock; + } + if (data.start != data_re.start || data.end != data_re.end) { + /* + * The extended contiguous chunk has been changed. Try again + * with single memseg instead. + */ + data.start = RTE_ALIGN_FLOOR(addr, msl->page_sz); + data.end = data.start + msl->page_sz; + rte_rwlock_read_unlock(&mcfg->memory_hotplug_lock); + mr_free(mr); + goto alloc_resources; } -release: - rte_spinlock_unlock(&priv->mr_lock); - return mr; + assert(data.msl == data_re.msl); + rte_rwlock_write_lock(&priv->mr.rwlock); + /* + * Check the address is really missing. If other thread already created + * one or it is not found due to overflow, abort and return. + */ + if (mr_lookup_dev(dev, entry, addr) != UINT32_MAX) { + /* + * Insert to the global cache table. It may fail due to + * low-on-memory. Then, this entry will have to be searched + * here again. + */ + mr_btree_insert(&priv->mr.cache, entry); + DEBUG("port %u found MR for %p on final lookup, abort", + dev->data->port_id, (void *)addr); + rte_rwlock_write_unlock(&priv->mr.rwlock); + rte_rwlock_read_unlock(&mcfg->memory_hotplug_lock); + /* + * Must be unlocked before calling rte_free() because + * mlx4_mr_mem_event_free_cb() can be called inside. + */ + mr_free(mr); + return entry->lkey; + } + /* + * Trim start and end addresses for verbs MR. Set bits for registering + * memsegs but exclude already registered ones. Bitmap can be + * fragmented. + */ + for (n = 0; n < ms_n; ++n) { + uintptr_t start; + struct mlx4_mr_cache ret = { 0, }; + + start = data_re.start + n * msl->page_sz; + /* Exclude memsegs already registered by other MRs. */ + if (mr_lookup_dev(dev, &ret, start) == UINT32_MAX) { + /* + * Start from the first unregistered memseg in the + * extended range. + */ + if (ms_idx_shift == -1) { + mr->ms_base_idx += n; + data.start = start; + ms_idx_shift = n; + } + data.end = start + msl->page_sz; + rte_bitmap_set(mr->ms_bmp, n - ms_idx_shift); + ++mr->ms_n; + } + } + len = data.end - data.start; + mr->ms_bmp_n = len / msl->page_sz; + assert(ms_idx_shift + mr->ms_bmp_n <= ms_n); + /* + * Finally create a verbs MR for the memory chunk. ibv_reg_mr() can be + * called with holding the memory lock because it doesn't use + * mlx4_alloc_buf_extern() which eventually calls rte_malloc_socket() + * through mlx4_alloc_verbs_buf(). + */ + mr->ibv_mr = mlx4_glue->reg_mr(priv->pd, (void *)data.start, len, + IBV_ACCESS_LOCAL_WRITE); + if (mr->ibv_mr == NULL) { + WARN("port %u fail to create a verbs MR for address (%p)", + dev->data->port_id, (void *)addr); + rte_errno = EINVAL; + goto err_mrlock; + } + assert((uintptr_t)mr->ibv_mr->addr == data.start); + assert(mr->ibv_mr->length == len); + LIST_INSERT_HEAD(&priv->mr.mr_list, mr, mr); + DEBUG("port %u MR CREATED (%p) for %p:\n" + " [0x%" PRIxPTR ", 0x%" PRIxPTR ")," + " lkey=0x%x base_idx=%u ms_n=%u, ms_bmp_n=%u", + dev->data->port_id, (void *)mr, (void *)addr, + data.start, data.end, rte_cpu_to_be_32(mr->ibv_mr->lkey), + mr->ms_base_idx, mr->ms_n, mr->ms_bmp_n); + /* Insert to the global cache table. */ + mr_insert_dev_cache(dev, mr); + /* Fill in output data. */ + mr_lookup_dev(dev, entry, addr); + /* Lookup can't fail. */ + assert(entry->lkey != UINT32_MAX); + rte_rwlock_write_unlock(&priv->mr.rwlock); + rte_rwlock_read_unlock(&mcfg->memory_hotplug_lock); + return entry->lkey; +err_mrlock: + rte_rwlock_write_unlock(&priv->mr.rwlock); +err_memlock: + rte_rwlock_read_unlock(&mcfg->memory_hotplug_lock); +err_nolock: + /* + * In case of error, as this can be called in a datapath, a warning + * message per an error is preferable instead. Must be unlocked before + * calling rte_free() because mlx4_mr_mem_event_free_cb() can be called + * inside. + */ + mr_free(mr); + return UINT32_MAX; } /** - * Release a memory region. + * Rebuild the global B-tree cache of device from the original MR list. * - * This function decrements its reference count and destroys it after - * reaching 0. + * @param dev + * Pointer to Ethernet device. + */ +static void +mr_rebuild_dev_cache(struct rte_eth_dev *dev) +{ + struct priv *priv = dev->data->dev_private; + struct mlx4_mr *mr; + + DEBUG("port %u rebuild dev cache[]", dev->data->port_id); + /* Flush cache to rebuild. */ + priv->mr.cache.len = 1; + priv->mr.cache.overflow = 0; + /* Iterate all the existing MRs. */ + LIST_FOREACH(mr, &priv->mr.mr_list, mr) + if (mr_insert_dev_cache(dev, mr) < 0) + return; +} + +/** + * Callback for memory free event. Iterate freed memsegs and check whether it + * belongs to an existing MR. If found, clear the bit from bitmap of MR. As a + * result, the MR would be fragmented. If it becomes empty, the MR will be freed + * later by mlx4_mr_garbage_collect(). * - * Note to avoid race conditions given this function may be used from the - * data plane, it's extremely important that each user holds its own - * reference. + * The global cache must be rebuilt if there's any change and this event has to + * be propagated to dataplane threads to flush the local caches. * - * @param mr - * Memory region to release. + * @param dev + * Pointer to Ethernet device. + * @param addr + * Address of freed memory. + * @param len + * Size of freed memory. + */ +static void +mlx4_mr_mem_event_free_cb(struct rte_eth_dev *dev, const void *addr, size_t len) +{ + struct priv *priv = dev->data->dev_private; + const struct rte_memseg_list *msl; + struct mlx4_mr *mr; + int ms_n; + int i; + int rebuild = 0; + + DEBUG("port %u free callback: addr=%p, len=%zu", + dev->data->port_id, addr, len); + msl = rte_mem_virt2memseg_list(addr); + /* addr and len must be page-aligned. */ + assert((uintptr_t)addr == RTE_ALIGN((uintptr_t)addr, msl->page_sz)); + assert(len == RTE_ALIGN(len, msl->page_sz)); + ms_n = len / msl->page_sz; + rte_rwlock_write_lock(&priv->mr.rwlock); + /* Clear bits of freed memsegs from MR. */ + for (i = 0; i < ms_n; ++i) { + const struct rte_memseg *ms; + struct mlx4_mr_cache entry; + uintptr_t start; + int ms_idx; + uint32_t pos; + + /* Find MR having this memseg. */ + start = (uintptr_t)addr + i * msl->page_sz; + mr = mr_lookup_dev_list(dev, &entry, start); + if (mr == NULL) + continue; + ms = rte_mem_virt2memseg((void *)start, msl); + assert(ms != NULL); + assert(msl->page_sz == ms->hugepage_sz); + ms_idx = rte_fbarray_find_idx(&msl->memseg_arr, ms); + pos = ms_idx - mr->ms_base_idx; + assert(rte_bitmap_get(mr->ms_bmp, pos)); + assert(pos < mr->ms_bmp_n); + DEBUG("port %u MR(%p): clear bitmap[%u] for addr %p", + dev->data->port_id, (void *)mr, pos, (void *)start); + rte_bitmap_clear(mr->ms_bmp, pos); + if (--mr->ms_n == 0) { + LIST_REMOVE(mr, mr); + LIST_INSERT_HEAD(&priv->mr.mr_free_list, mr, mr); + DEBUG("port %u remove MR(%p) from list", + dev->data->port_id, (void *)mr); + } + /* + * MR is fragmented or will be freed. the global cache must be + * rebuilt. + */ + rebuild = 1; + } + if (rebuild) { + mr_rebuild_dev_cache(dev); + /* + * Flush local caches by propagating invalidation across cores. + * rte_smp_wmb() is enough to synchronize this event. If one of + * freed memsegs is seen by other core, that means the memseg + * has been allocated by allocator, which will come after this + * free call. Therefore, this store instruction (incrementing + * generation below) will be guaranteed to be seen by other core + * before the core sees the newly allocated memory. + */ + ++priv->mr.dev_gen; + DEBUG("broadcasting local cache flush, gen=%d", + priv->mr.dev_gen); + rte_smp_wmb(); + } + rte_rwlock_write_unlock(&priv->mr.rwlock); +#ifndef NDEBUG + if (rebuild) + mlx4_mr_dump_dev(dev); +#endif +} + +/** + * Callback for memory event. + * + * @param event_type + * Memory event type. + * @param addr + * Address of memory. + * @param len + * Size of memory. */ void -mlx4_mr_put(struct mlx4_mr *mr) +mlx4_mr_mem_event_cb(enum rte_mem_event event_type, const void *addr, + size_t len, void *arg __rte_unused) { - struct priv *priv = mr->priv; - - rte_spinlock_lock(&priv->mr_lock); - assert(mr->refcnt); - if (--mr->refcnt) - goto release; - LIST_REMOVE(mr, next); - claim_zero(mlx4_glue->dereg_mr(mr->mr)); - rte_free(mr); -release: - rte_spinlock_unlock(&priv->mr_lock); + struct priv *priv; + + switch (event_type) { + case RTE_MEM_EVENT_FREE: + rte_rwlock_read_lock(&mlx4_mem_event_rwlock); + /* Iterate all the existing mlx4 devices. */ + LIST_FOREACH(priv, &mlx4_mem_event_cb_list, mem_event_cb) + mlx4_mr_mem_event_free_cb(priv->dev, addr, len); + rte_rwlock_read_unlock(&mlx4_mem_event_rwlock); + break; + case RTE_MEM_EVENT_ALLOC: + default: + break; + } } /** - * Add memory region (MR) <-> memory pool (MP) association to txq->mp2mr[]. - * If mp2mr[] is full, remove an entry first. + * Look up address in the global MR cache table. If not found, create a new MR. + * Insert the found/created entry to local bottom-half cache table. + * + * @param dev + * Pointer to Ethernet device. + * @param mr_ctrl + * Pointer to per-queue MR control structure. + * @param[out] entry + * Pointer to returning MR cache entry, found in the global cache or newly + * created. If failed to create one, this is not written. + * @param addr + * Search key. + * + * @return + * Searched LKey on success, UINT32_MAX on no match. + */ +static uint32_t +mlx4_mr_lookup_dev(struct rte_eth_dev *dev, struct mlx4_mr_ctrl *mr_ctrl, + struct mlx4_mr_cache *entry, uintptr_t addr) +{ + struct priv *priv = dev->data->dev_private; + struct mlx4_mr_btree *bt = &mr_ctrl->cache_bh; + uint16_t idx; + uint32_t lkey; + + /* If local cache table is full, try to double it. */ + if (unlikely(bt->len == bt->size)) + mr_btree_expand(bt, bt->size << 1); + /* Look up in the global cache. */ + rte_rwlock_read_lock(&priv->mr.rwlock); + lkey = mr_btree_lookup(&priv->mr.cache, &idx, addr); + if (lkey != UINT32_MAX) { + /* Found. */ + *entry = (*priv->mr.cache.table)[idx]; + rte_rwlock_read_unlock(&priv->mr.rwlock); + /* + * Update local cache. Even if it fails, return the found entry + * to update top-half cache. Next time, this entry will be found + * in the global cache. + */ + mr_btree_insert(bt, entry); + return lkey; + } + rte_rwlock_read_unlock(&priv->mr.rwlock); + /* First time to see the address? Create a new MR. */ + lkey = mlx4_mr_create(dev, entry, addr); + /* + * Update the local cache if successfully created a new global MR. Even + * if failed to create one, there's no action to take in this datapath + * code. As returning LKey is invalid, this will eventually make HW + * fail. + */ + if (lkey != UINT32_MAX) + mr_btree_insert(bt, entry); + return lkey; +} + +/** + * Bottom-half of LKey search on datapath. Firstly search in cache_bh[] and if + * misses, search in the global MR cache table and update the new entry to + * per-queue local caches. + * + * @param dev + * Pointer to Ethernet device. + * @param mr_ctrl + * Pointer to per-queue MR control structure. + * @param addr + * Search key. + * + * @return + * Searched LKey on success, UINT32_MAX on no match. + */ +static uint32_t +mlx4_mr_addr2mr_bh(struct rte_eth_dev *dev, struct mlx4_mr_ctrl *mr_ctrl, + uintptr_t addr) +{ + uint32_t lkey; + uint16_t bh_idx = 0; + /* Victim in top-half cache to replace with new entry. */ + struct mlx4_mr_cache *repl = &mr_ctrl->cache[mr_ctrl->head]; + + /* Binary-search MR translation table. */ + lkey = mr_btree_lookup(&mr_ctrl->cache_bh, &bh_idx, addr); + /* Update top-half cache. */ + if (likely(lkey != UINT32_MAX)) { + *repl = (*mr_ctrl->cache_bh.table)[bh_idx]; + } else { + /* + * If missed in local lookup table, search in the global cache + * and local cache_bh[] will be updated inside if possible. + * Top-half cache entry will also be updated. + */ + lkey = mlx4_mr_lookup_dev(dev, mr_ctrl, repl, addr); + if (unlikely(lkey == UINT32_MAX)) + return UINT32_MAX; + } + /* Update the most recently used entry. */ + mr_ctrl->mru = mr_ctrl->head; + /* Point to the next victim, the oldest. */ + mr_ctrl->head = (mr_ctrl->head + 1) % MLX4_MR_CACHE_N; + return lkey; +} + +/** + * Bottom-half of LKey search on Rx. + * + * @param rxq + * Pointer to Rx queue structure. + * @param addr + * Search key. + * + * @return + * Searched LKey on success, UINT32_MAX on no match. + */ +uint32_t +mlx4_rx_addr2mr_bh(struct rxq *rxq, uintptr_t addr) +{ + struct mlx4_mr_ctrl *mr_ctrl = &rxq->mr_ctrl; + struct priv *priv = rxq->priv; + + DEBUG("Rx queue %u: miss on top-half, mru=%u, head=%u, addr=%p", + rxq->stats.idx, mr_ctrl->mru, mr_ctrl->head, (void *)addr); + return mlx4_mr_addr2mr_bh(priv->dev, mr_ctrl, addr); +} + +/** + * Bottom-half of LKey search on Tx. * * @param txq * Pointer to Tx queue structure. - * @param[in] mp - * Memory pool for which a memory region lkey must be added. - * @param[in] i - * Index in memory pool (MP) where to add memory region (MR). + * @param addr + * Search key. * * @return - * Added mr->lkey on success, (uint32_t)-1 on failure. + * Searched LKey on success, UINT32_MAX on no match. */ uint32_t -mlx4_txq_add_mr(struct txq *txq, struct rte_mempool *mp, uint32_t i) +mlx4_tx_addr2mr_bh(struct txq *txq, uintptr_t addr) { + struct mlx4_mr_ctrl *mr_ctrl = &txq->mr_ctrl; + struct priv *priv = txq->priv; + + DEBUG("Tx queue %u: miss on top-half, mru=%u, head=%u, addr=%p", + txq->stats.idx, mr_ctrl->mru, mr_ctrl->head, (void *)addr); + return mlx4_mr_addr2mr_bh(priv->dev, mr_ctrl, addr); +} + +/** + * Flush all of the local cache entries. + * + * @param mr_ctrl + * Pointer to per-queue MR control structure. + */ +void +mlx4_mr_flush_local_cache(struct mlx4_mr_ctrl *mr_ctrl) +{ + /* Reset the most-recently-used index. */ + mr_ctrl->mru = 0; + /* Reset the linear search array. */ + mr_ctrl->head = 0; + memset(mr_ctrl->cache, 0, sizeof(mr_ctrl->cache)); + /* Reset the B-tree table. */ + mr_ctrl->cache_bh.len = 1; + mr_ctrl->cache_bh.overflow = 0; + /* Update the generation number. */ + mr_ctrl->cur_gen = *mr_ctrl->dev_gen_ptr; + DEBUG("mr_ctrl(%p): flushed, cur_gen=%d", + (void *)mr_ctrl, mr_ctrl->cur_gen); +} + +/* Called during rte_mempool_mem_iter() by mlx4_mr_update_mp(). */ +static void +mlx4_mr_update_mp_cb(struct rte_mempool *mp __rte_unused, void *opaque, + struct rte_mempool_memhdr *memhdr, + unsigned mem_idx __rte_unused) +{ + struct mr_update_mp_data *data = opaque; + uint32_t lkey; + + /* Stop iteration if failed in the previous walk. */ + if (data->ret < 0) + return; + /* Register address of the chunk and update local caches. */ + lkey = mlx4_mr_addr2mr_bh(data->dev, data->mr_ctrl, + (uintptr_t)memhdr->addr); + if (lkey == UINT32_MAX) + data->ret = -1; +} + +/** + * Register entire memory chunks in a Mempool. + * + * @param dev + * Pointer to Ethernet device. + * @param mr_ctrl + * Pointer to per-queue MR control structure. + * @param mp + * Pointer to registering Mempool. + * + * @return + * 0 on success, -1 on failure. + */ +int +mlx4_mr_update_mp(struct rte_eth_dev *dev, struct mlx4_mr_ctrl *mr_ctrl, + struct rte_mempool *mp) +{ + struct mr_update_mp_data data = { + .dev = dev, + .mr_ctrl = mr_ctrl, + .ret = 0, + }; + + rte_mempool_mem_iter(mp, mlx4_mr_update_mp_cb, &data); + return data.ret; +} + +#ifndef NDEBUG +/** + * Dump all the created MRs and the global cache entries. + * + * @param dev + * Pointer to Ethernet device. + */ +void +mlx4_mr_dump_dev(struct rte_eth_dev *dev) +{ + struct priv *priv = dev->data->dev_private; struct mlx4_mr *mr; + int mr_n = 0; + int chunk_n = 0; + + rte_rwlock_read_lock(&priv->mr.rwlock); + /* Iterate all the existing MRs. */ + LIST_FOREACH(mr, &priv->mr.mr_list, mr) { + unsigned int n; + + DEBUG("port %u MR[%u], LKey = 0x%x, ms_n = %u, ms_bmp_n = %u", + dev->data->port_id, mr_n++, + rte_cpu_to_be_32(mr->ibv_mr->lkey), + mr->ms_n, mr->ms_bmp_n); + if (mr->ms_n == 0) + continue; + for (n = 0; n < mr->ms_bmp_n; ) { + struct mlx4_mr_cache ret = { 0, }; - /* Add a new entry, register MR first. */ - DEBUG("%p: discovered new memory pool \"%s\" (%p)", - (void *)txq, mp->name, (void *)mp); - mr = mlx4_mr_get(txq->priv, mp); - if (unlikely(mr == NULL)) { - DEBUG("%p: unable to configure MR, mlx4_mr_get() failed", - (void *)txq); - return (uint32_t)-1; + n = mr_find_next_chunk(mr, &ret, n); + if (!ret.end) + break; + DEBUG(" chunk[%u], [0x%" PRIxPTR ", 0x%" PRIxPTR ")", + chunk_n++, ret.start, ret.end); + } } - if (unlikely(i == RTE_DIM(txq->mp2mr))) { - /* Table is full, remove oldest entry. */ - DEBUG("%p: MR <-> MP table full, dropping oldest entry.", - (void *)txq); - --i; - mlx4_mr_put(txq->mp2mr[0].mr); - memmove(&txq->mp2mr[0], &txq->mp2mr[1], - (sizeof(txq->mp2mr) - sizeof(txq->mp2mr[0]))); + DEBUG("port %u dumping global cache", dev->data->port_id); + mlx4_mr_btree_dump(&priv->mr.cache); + rte_rwlock_read_unlock(&priv->mr.rwlock); +} +#endif + +/** + * Release all the created MRs and resources. Remove device from memory callback + * list. + * + * @param dev + * Pointer to Ethernet device. + */ +void +mlx4_mr_release(struct rte_eth_dev *dev) +{ + struct priv *priv = dev->data->dev_private; + struct mlx4_mr *mr_next = LIST_FIRST(&priv->mr.mr_list); + + /* Remove from memory callback device list. */ + rte_rwlock_write_lock(&mlx4_mem_event_rwlock); + LIST_REMOVE(priv, mem_event_cb); + rte_rwlock_write_unlock(&mlx4_mem_event_rwlock); +#ifndef NDEBUG + mlx4_mr_dump_dev(dev); +#endif + rte_rwlock_write_lock(&priv->mr.rwlock); + /* Detach from MR list and move to free list. */ + while (mr_next != NULL) { + struct mlx4_mr *mr = mr_next; + + mr_next = LIST_NEXT(mr, mr); + LIST_REMOVE(mr, mr); + LIST_INSERT_HEAD(&priv->mr.mr_free_list, mr, mr); } - /* Store the new entry. */ - txq->mp2mr[i].mp = mp; - txq->mp2mr[i].mr = mr; - txq->mp2mr[i].lkey = mr->lkey; - DEBUG("%p: new MR lkey for MP \"%s\" (%p): 0x%08" PRIu32, - (void *)txq, mp->name, (void *)mp, txq->mp2mr[i].lkey); - return txq->mp2mr[i].lkey; + LIST_INIT(&priv->mr.mr_list); + /* Free global cache. */ + mlx4_mr_btree_free(&priv->mr.cache); + rte_rwlock_write_unlock(&priv->mr.rwlock); + /* Free all remaining MRs. */ + mlx4_mr_garbage_collect(dev); } |