ceph/src/seastar/dpdk/drivers/net/mlx5/mlx5_mr.c

   1 /*-
   2  *   BSD LICENSE
   3  *
   4  *   Copyright 2016 6WIND S.A.
   5  *   Copyright 2016 Mellanox.
   6  *
   7  *   Redistribution and use in source and binary forms, with or without
   8  *   modification, are permitted provided that the following conditions
   9  *   are met:
  10  *
  11  *     * Redistributions of source code must retain the above copyright
  12  *       notice, this list of conditions and the following disclaimer.
  13  *     * Redistributions in binary form must reproduce the above copyright
  14  *       notice, this list of conditions and the following disclaimer in
  15  *       the documentation and/or other materials provided with the
  16  *       distribution.
  17  *     * Neither the name of 6WIND S.A. nor the names of its
  18  *       contributors may be used to endorse or promote products derived
  19  *       from this software without specific prior written permission.
  20  *
  21  *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22  *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23  *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
  24  *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
  25  *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  26  *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
  27  *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  28  *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  29  *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  30  *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  31  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  32  */
  33
  34 /* Verbs header. */
  35 /* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
  36 #ifdef PEDANTIC
  37 #pragma GCC diagnostic ignored "-Wpedantic"
  38 #endif
  39 #include <infiniband/verbs.h>
  40 #ifdef PEDANTIC
  41 #pragma GCC diagnostic error "-Wpedantic"
  42 #endif
  43
  44 /* DPDK headers don't like -pedantic. */
  45 #ifdef PEDANTIC
  46 #pragma GCC diagnostic ignored "-Wpedantic"
  47 #endif
  48 #include <rte_mempool.h>
  49 #ifdef PEDANTIC
  50 #pragma GCC diagnostic error "-Wpedantic"
  51 #endif
  52
  53 #include "mlx5.h"
  54 #include "mlx5_rxtx.h"
  55
  56 struct mlx5_check_mempool_data {
  57         int ret;
  58         char *start;
  59         char *end;
  60 };
  61
  62 /* Called by mlx5_check_mempool() when iterating the memory chunks. */
  63 static void
  64 mlx5_check_mempool_cb(struct rte_mempool *mp,
  65                       void *opaque, struct rte_mempool_memhdr *memhdr,
  66                       unsigned int mem_idx)
  67 {
  68         struct mlx5_check_mempool_data *data = opaque;
  69
  70         (void)mp;
  71         (void)mem_idx;
  72
  73         /* It already failed, skip the next chunks. */
  74         if (data->ret != 0)
  75                 return;
  76         /* It is the first chunk. */
  77         if (data->start == NULL && data->end == NULL) {
  78                 data->start = memhdr->addr;
  79                 data->end = data->start + memhdr->len;
  80                 return;
  81         }
  82         if (data->end == memhdr->addr) {
  83                 data->end += memhdr->len;
  84                 return;
  85         }
  86         if (data->start == (char *)memhdr->addr + memhdr->len) {
  87                 data->start -= memhdr->len;
  88                 return;
  89         }
  90         /* Error, mempool is not virtually contiguous. */
  91         data->ret = -1;
  92 }
  93
  94 /**
  95  * Check if a mempool can be used: it must be virtually contiguous.
  96  *
  97  * @param[in] mp
  98  *   Pointer to memory pool.
  99  * @param[out] start
 100  *   Pointer to the start address of the mempool virtual memory area
 101  * @param[out] end
 102  *   Pointer to the end address of the mempool virtual memory area
 103  *
 104  * @return
 105  *   0 on success (mempool is virtually contiguous), -1 on error.
 106  */
 107 static int mlx5_check_mempool(struct rte_mempool *mp, uintptr_t *start,
 108         uintptr_t *end)
 109 {
 110         struct mlx5_check_mempool_data data;
 111
 112         memset(&data, 0, sizeof(data));
 113         rte_mempool_mem_iter(mp, mlx5_check_mempool_cb, &data);
 114         *start = (uintptr_t)data.start;
 115         *end = (uintptr_t)data.end;
 116
 117         return data.ret;
 118 }
 119
 120 /**
 121  * Register mempool as a memory region.
 122  *
 123  * @param pd
 124  *   Pointer to protection domain.
 125  * @param mp
 126  *   Pointer to memory pool.
 127  *
 128  * @return
 129  *   Memory region pointer, NULL in case of error.
 130  */
 131 struct ibv_mr *
 132 mlx5_mp2mr(struct ibv_pd *pd, struct rte_mempool *mp)
 133 {
 134         const struct rte_memseg *ms = rte_eal_get_physmem_layout();
 135         uintptr_t start;
 136         uintptr_t end;
 137         unsigned int i;
 138
 139         if (mlx5_check_mempool(mp, &start, &end) != 0) {
 140                 ERROR("mempool %p: not virtually contiguous",
 141                       (void *)mp);
 142                 return NULL;
 143         }
 144
 145         DEBUG("mempool %p area start=%p end=%p size=%zu",
 146               (void *)mp, (void *)start, (void *)end,
 147               (size_t)(end - start));
 148         /* Round start and end to page boundary if found in memory segments. */
 149         for (i = 0; (i < RTE_MAX_MEMSEG) && (ms[i].addr != NULL); ++i) {
 150                 uintptr_t addr = (uintptr_t)ms[i].addr;
 151                 size_t len = ms[i].len;
 152                 unsigned int align = ms[i].hugepage_sz;
 153
 154                 if ((start > addr) && (start < addr + len))
 155                         start = RTE_ALIGN_FLOOR(start, align);
 156                 if ((end > addr) && (end < addr + len))
 157                         end = RTE_ALIGN_CEIL(end, align);
 158         }
 159         DEBUG("mempool %p using start=%p end=%p size=%zu for MR",
 160               (void *)mp, (void *)start, (void *)end,
 161               (size_t)(end - start));
 162         return ibv_reg_mr(pd,
 163                           (void *)start,
 164                           end - start,
 165                           IBV_ACCESS_LOCAL_WRITE);
 166 }
 167
 168 /**
 169  * Register a Memory Region (MR) <-> Memory Pool (MP) association in
 170  * txq->mp2mr[]. If mp2mr[] is full, remove an entry first.
 171  *
 172  * This function should only be called by txq_mp2mr().
 173  *
 174  * @param txq
 175  *   Pointer to TX queue structure.
 176  * @param[in] mp
 177  *   Memory Pool for which a Memory Region lkey must be returned.
 178  * @param idx
 179  *   Index of the next available entry.
 180  *
 181  * @return
 182  *   mr->lkey on success, (uint32_t)-1 on failure.
 183  */
 184 uint32_t
 185 txq_mp2mr_reg(struct txq *txq, struct rte_mempool *mp, unsigned int idx)
 186 {
 187         struct txq_ctrl *txq_ctrl = container_of(txq, struct txq_ctrl, txq);
 188         struct ibv_mr *mr;
 189
 190         /* Add a new entry, register MR first. */
 191         DEBUG("%p: discovered new memory pool \"%s\" (%p)",
 192               (void *)txq_ctrl, mp->name, (void *)mp);
 193         mr = mlx5_mp2mr(txq_ctrl->priv->pd, mp);
 194         if (unlikely(mr == NULL)) {
 195                 DEBUG("%p: unable to configure MR, ibv_reg_mr() failed.",
 196                       (void *)txq_ctrl);
 197                 return (uint32_t)-1;
 198         }
 199         if (unlikely(idx == RTE_DIM(txq_ctrl->txq.mp2mr))) {
 200                 /* Table is full, remove oldest entry. */
 201                 DEBUG("%p: MR <-> MP table full, dropping oldest entry.",
 202                       (void *)txq_ctrl);
 203                 --idx;
 204                 claim_zero(ibv_dereg_mr(txq_ctrl->txq.mp2mr[0].mr));
 205                 memmove(&txq_ctrl->txq.mp2mr[0], &txq_ctrl->txq.mp2mr[1],
 206                         (sizeof(txq_ctrl->txq.mp2mr) -
 207                          sizeof(txq_ctrl->txq.mp2mr[0])));
 208         }
 209         /* Store the new entry. */
 210         txq_ctrl->txq.mp2mr[idx].mp = mp;
 211         txq_ctrl->txq.mp2mr[idx].mr = mr;
 212         txq_ctrl->txq.mp2mr[idx].lkey = htonl(mr->lkey);
 213         DEBUG("%p: new MR lkey for MP \"%s\" (%p): 0x%08" PRIu32,
 214               (void *)txq_ctrl, mp->name, (void *)mp,
 215               txq_ctrl->txq.mp2mr[idx].lkey);
 216         return txq_ctrl->txq.mp2mr[idx].lkey;
 217 }
 218
 219 struct txq_mp2mr_mbuf_check_data {
 220         int ret;
 221 };
 222
 223 /**
 224  * Callback function for rte_mempool_obj_iter() to check whether a given
 225  * mempool object looks like a mbuf.
 226  *
 227  * @param[in] mp
 228  *   The mempool pointer
 229  * @param[in] arg
 230  *   Context data (struct txq_mp2mr_mbuf_check_data). Contains the
 231  *   return value.
 232  * @param[in] obj
 233  *   Object address.
 234  * @param index
 235  *   Object index, unused.
 236  */
 237 static void
 238 txq_mp2mr_mbuf_check(struct rte_mempool *mp, void *arg, void *obj,
 239         uint32_t index __rte_unused)
 240 {
 241         struct txq_mp2mr_mbuf_check_data *data = arg;
 242         struct rte_mbuf *buf = obj;
 243
 244         /*
 245          * Check whether mbuf structure fits element size and whether mempool
 246          * pointer is valid.
 247          */
 248         if (sizeof(*buf) > mp->elt_size || buf->pool != mp)
 249                 data->ret = -1;
 250 }
 251
 252 /**
 253  * Iterator function for rte_mempool_walk() to register existing mempools and
 254  * fill the MP to MR cache of a TX queue.
 255  *
 256  * @param[in] mp
 257  *   Memory Pool to register.
 258  * @param *arg
 259  *   Pointer to TX queue structure.
 260  */
 261 void
 262 txq_mp2mr_iter(struct rte_mempool *mp, void *arg)
 263 {
 264         struct txq_ctrl *txq_ctrl = arg;
 265         struct txq_mp2mr_mbuf_check_data data = {
 266                 .ret = 0,
 267         };
 268         unsigned int i;
 269
 270         /* Register mempool only if the first element looks like a mbuf. */
 271         if (rte_mempool_obj_iter(mp, txq_mp2mr_mbuf_check, &data) == 0 ||
 272                         data.ret == -1)
 273                 return;
 274         for (i = 0; (i != RTE_DIM(txq_ctrl->txq.mp2mr)); ++i) {
 275                 if (unlikely(txq_ctrl->txq.mp2mr[i].mp == NULL)) {
 276                         /* Unknown MP, add a new MR for it. */
 277                         break;
 278                 }
 279                 if (txq_ctrl->txq.mp2mr[i].mp == mp)
 280                         return;
 281         }
 282         txq_mp2mr_reg(&txq_ctrl->txq, mp, i);
 283 }