net/rds/message.c

   1 /*
   2  * Copyright (c) 2006, 2020 Oracle and/or its affiliates.
   3  *
   4  * This software is available to you under a choice of one of two
   5  * licenses.  You may choose to be licensed under the terms of the GNU
   6  * General Public License (GPL) Version 2, available from the file
   7  * COPYING in the main directory of this source tree, or the
   8  * OpenIB.org BSD license below:
   9  *
  10  *     Redistribution and use in source and binary forms, with or
  11  *     without modification, are permitted provided that the following
  12  *     conditions are met:
  13  *
  14  *      - Redistributions of source code must retain the above
  15  *        copyright notice, this list of conditions and the following
  16  *        disclaimer.
  17  *
  18  *      - Redistributions in binary form must reproduce the above
  19  *        copyright notice, this list of conditions and the following
  20  *        disclaimer in the documentation and/or other materials
  21  *        provided with the distribution.
  22  *
  23  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  24  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  25  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  26  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
  27  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
  28  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  29  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  30  * SOFTWARE.
  31  *
  32  */
  33 #include <linux/kernel.h>
  34 #include <linux/slab.h>
  35 #include <linux/export.h>
  36 #include <linux/skbuff.h>
  37 #include <linux/list.h>
  38 #include <linux/errqueue.h>
  39
  40 #include "rds.h"
  41
  42 static unsigned int     rds_exthdr_size[__RDS_EXTHDR_MAX] = {
  43 [RDS_EXTHDR_NONE]       = 0,
  44 [RDS_EXTHDR_VERSION]    = sizeof(struct rds_ext_header_version),
  45 [RDS_EXTHDR_RDMA]       = sizeof(struct rds_ext_header_rdma),
  46 [RDS_EXTHDR_RDMA_DEST]  = sizeof(struct rds_ext_header_rdma_dest),
  47 [RDS_EXTHDR_NPATHS]     = sizeof(u16),
  48 [RDS_EXTHDR_GEN_NUM]    = sizeof(u32),
  49 };
  50
  51 void rds_message_addref(struct rds_message *rm)
  52 {
  53         rdsdebug("addref rm %p ref %d\n", rm, refcount_read(&rm->m_refcount));
  54         refcount_inc(&rm->m_refcount);
  55 }
  56 EXPORT_SYMBOL_GPL(rds_message_addref);
  57
  58 static inline bool rds_zcookie_add(struct rds_msg_zcopy_info *info, u32 cookie)
  59 {
  60         struct rds_zcopy_cookies *ck = &info->zcookies;
  61         int ncookies = ck->num;
  62
  63         if (ncookies == RDS_MAX_ZCOOKIES)
  64                 return false;
  65         ck->cookies[ncookies] = cookie;
  66         ck->num =  ++ncookies;
  67         return true;
  68 }
  69
  70 static struct rds_msg_zcopy_info *rds_info_from_znotifier(struct rds_znotifier *znotif)
  71 {
  72         return container_of(znotif, struct rds_msg_zcopy_info, znotif);
  73 }
  74
  75 void rds_notify_msg_zcopy_purge(struct rds_msg_zcopy_queue *q)
  76 {
  77         unsigned long flags;
  78         LIST_HEAD(copy);
  79         struct rds_msg_zcopy_info *info, *tmp;
  80
  81         spin_lock_irqsave(&q->lock, flags);
  82         list_splice(&q->zcookie_head, &copy);
  83         INIT_LIST_HEAD(&q->zcookie_head);
  84         spin_unlock_irqrestore(&q->lock, flags);
  85
  86         list_for_each_entry_safe(info, tmp, &copy, rs_zcookie_next) {
  87                 list_del(&info->rs_zcookie_next);
  88                 kfree(info);
  89         }
  90 }
  91
  92 static void rds_rm_zerocopy_callback(struct rds_sock *rs,
  93                                      struct rds_znotifier *znotif)
  94 {
  95         struct rds_msg_zcopy_info *info;
  96         struct rds_msg_zcopy_queue *q;
  97         u32 cookie = znotif->z_cookie;
  98         struct rds_zcopy_cookies *ck;
  99         struct list_head *head;
 100         unsigned long flags;
 101
 102         mm_unaccount_pinned_pages(&znotif->z_mmp);
 103         q = &rs->rs_zcookie_queue;
 104         spin_lock_irqsave(&q->lock, flags);
 105         head = &q->zcookie_head;
 106         if (!list_empty(head)) {
 107                 info = list_entry(head, struct rds_msg_zcopy_info,
 108                                   rs_zcookie_next);
 109                 if (info && rds_zcookie_add(info, cookie)) {
 110                         spin_unlock_irqrestore(&q->lock, flags);
 111                         kfree(rds_info_from_znotifier(znotif));
 112                         /* caller invokes rds_wake_sk_sleep() */
 113                         return;
 114                 }
 115         }
 116
 117         info = rds_info_from_znotifier(znotif);
 118         ck = &info->zcookies;
 119         memset(ck, 0, sizeof(*ck));
 120         WARN_ON(!rds_zcookie_add(info, cookie));
 121         list_add_tail(&q->zcookie_head, &info->rs_zcookie_next);
 122
 123         spin_unlock_irqrestore(&q->lock, flags);
 124         /* caller invokes rds_wake_sk_sleep() */
 125 }
 126
 127 /*
 128  * This relies on dma_map_sg() not touching sg[].page during merging.
 129  */
 130 static void rds_message_purge(struct rds_message *rm)
 131 {
 132         unsigned long i, flags;
 133         bool zcopy = false;
 134
 135         if (unlikely(test_bit(RDS_MSG_PAGEVEC, &rm->m_flags)))
 136                 return;
 137
 138         spin_lock_irqsave(&rm->m_rs_lock, flags);
 139         if (rm->m_rs) {
 140                 struct rds_sock *rs = rm->m_rs;
 141
 142                 if (rm->data.op_mmp_znotifier) {
 143                         zcopy = true;
 144                         rds_rm_zerocopy_callback(rs, rm->data.op_mmp_znotifier);
 145                         rds_wake_sk_sleep(rs);
 146                         rm->data.op_mmp_znotifier = NULL;
 147                 }
 148                 sock_put(rds_rs_to_sk(rs));
 149                 rm->m_rs = NULL;
 150         }
 151         spin_unlock_irqrestore(&rm->m_rs_lock, flags);
 152
 153         for (i = 0; i < rm->data.op_nents; i++) {
 154                 /* XXX will have to put_page for page refs */
 155                 if (!zcopy)
 156                         __free_page(sg_page(&rm->data.op_sg[i]));
 157                 else
 158                         put_page(sg_page(&rm->data.op_sg[i]));
 159         }
 160         rm->data.op_nents = 0;
 161
 162         if (rm->rdma.op_active)
 163                 rds_rdma_free_op(&rm->rdma);
 164         if (rm->rdma.op_rdma_mr)
 165                 kref_put(&rm->rdma.op_rdma_mr->r_kref, __rds_put_mr_final);
 166
 167         if (rm->atomic.op_active)
 168                 rds_atomic_free_op(&rm->atomic);
 169         if (rm->atomic.op_rdma_mr)
 170                 kref_put(&rm->atomic.op_rdma_mr->r_kref, __rds_put_mr_final);
 171 }
 172
 173 void rds_message_put(struct rds_message *rm)
 174 {
 175         rdsdebug("put rm %p ref %d\n", rm, refcount_read(&rm->m_refcount));
 176         WARN(!refcount_read(&rm->m_refcount), "danger refcount zero on %p\n", rm);
 177         if (refcount_dec_and_test(&rm->m_refcount)) {
 178                 BUG_ON(!list_empty(&rm->m_sock_item));
 179                 BUG_ON(!list_empty(&rm->m_conn_item));
 180                 rds_message_purge(rm);
 181
 182                 kfree(rm);
 183         }
 184 }
 185 EXPORT_SYMBOL_GPL(rds_message_put);
 186
 187 void rds_message_populate_header(struct rds_header *hdr, __be16 sport,
 188                                  __be16 dport, u64 seq)
 189 {
 190         hdr->h_flags = 0;
 191         hdr->h_sport = sport;
 192         hdr->h_dport = dport;
 193         hdr->h_sequence = cpu_to_be64(seq);
 194         hdr->h_exthdr[0] = RDS_EXTHDR_NONE;
 195 }
 196 EXPORT_SYMBOL_GPL(rds_message_populate_header);
 197
 198 int rds_message_add_extension(struct rds_header *hdr, unsigned int type,
 199                               const void *data, unsigned int len)
 200 {
 201         unsigned int ext_len = sizeof(u8) + len;
 202         unsigned char *dst;
 203
 204         /* For now, refuse to add more than one extension header */
 205         if (hdr->h_exthdr[0] != RDS_EXTHDR_NONE)
 206                 return 0;
 207
 208         if (type >= __RDS_EXTHDR_MAX || len != rds_exthdr_size[type])
 209                 return 0;
 210
 211         if (ext_len >= RDS_HEADER_EXT_SPACE)
 212                 return 0;
 213         dst = hdr->h_exthdr;
 214
 215         *dst++ = type;
 216         memcpy(dst, data, len);
 217
 218         dst[len] = RDS_EXTHDR_NONE;
 219         return 1;
 220 }
 221 EXPORT_SYMBOL_GPL(rds_message_add_extension);
 222
 223 /*
 224  * If a message has extension headers, retrieve them here.
 225  * Call like this:
 226  *
 227  * unsigned int pos = 0;
 228  *
 229  * while (1) {
 230  *      buflen = sizeof(buffer);
 231  *      type = rds_message_next_extension(hdr, &pos, buffer, &buflen);
 232  *      if (type == RDS_EXTHDR_NONE)
 233  *              break;
 234  *      ...
 235  * }
 236  */
 237 int rds_message_next_extension(struct rds_header *hdr,
 238                 unsigned int *pos, void *buf, unsigned int *buflen)
 239 {
 240         unsigned int offset, ext_type, ext_len;
 241         u8 *src = hdr->h_exthdr;
 242
 243         offset = *pos;
 244         if (offset >= RDS_HEADER_EXT_SPACE)
 245                 goto none;
 246
 247         /* Get the extension type and length. For now, the
 248          * length is implied by the extension type. */
 249         ext_type = src[offset++];
 250
 251         if (ext_type == RDS_EXTHDR_NONE || ext_type >= __RDS_EXTHDR_MAX)
 252                 goto none;
 253         ext_len = rds_exthdr_size[ext_type];
 254         if (offset + ext_len > RDS_HEADER_EXT_SPACE)
 255                 goto none;
 256
 257         *pos = offset + ext_len;
 258         if (ext_len < *buflen)
 259                 *buflen = ext_len;
 260         memcpy(buf, src + offset, *buflen);
 261         return ext_type;
 262
 263 none:
 264         *pos = RDS_HEADER_EXT_SPACE;
 265         *buflen = 0;
 266         return RDS_EXTHDR_NONE;
 267 }
 268
 269 int rds_message_add_rdma_dest_extension(struct rds_header *hdr, u32 r_key, u32 offset)
 270 {
 271         struct rds_ext_header_rdma_dest ext_hdr;
 272
 273         ext_hdr.h_rdma_rkey = cpu_to_be32(r_key);
 274         ext_hdr.h_rdma_offset = cpu_to_be32(offset);
 275         return rds_message_add_extension(hdr, RDS_EXTHDR_RDMA_DEST, &ext_hdr, sizeof(ext_hdr));
 276 }
 277 EXPORT_SYMBOL_GPL(rds_message_add_rdma_dest_extension);
 278
 279 /*
 280  * Each rds_message is allocated with extra space for the scatterlist entries
 281  * rds ops will need. This is to minimize memory allocation count. Then, each rds op
 282  * can grab SGs when initializing its part of the rds_message.
 283  */
 284 struct rds_message *rds_message_alloc(unsigned int extra_len, gfp_t gfp)
 285 {
 286         struct rds_message *rm;
 287
 288         if (extra_len > KMALLOC_MAX_SIZE - sizeof(struct rds_message))
 289                 return NULL;
 290
 291         rm = kzalloc(sizeof(struct rds_message) + extra_len, gfp);
 292         if (!rm)
 293                 goto out;
 294
 295         rm->m_used_sgs = 0;
 296         rm->m_total_sgs = extra_len / sizeof(struct scatterlist);
 297
 298         refcount_set(&rm->m_refcount, 1);
 299         INIT_LIST_HEAD(&rm->m_sock_item);
 300         INIT_LIST_HEAD(&rm->m_conn_item);
 301         spin_lock_init(&rm->m_rs_lock);
 302         init_waitqueue_head(&rm->m_flush_wait);
 303
 304 out:
 305         return rm;
 306 }
 307
 308 /*
 309  * RDS ops use this to grab SG entries from the rm's sg pool.
 310  */
 311 struct scatterlist *rds_message_alloc_sgs(struct rds_message *rm, int nents)
 312 {
 313         struct scatterlist *sg_first = (struct scatterlist *) &rm[1];
 314         struct scatterlist *sg_ret;
 315
 316         if (nents <= 0) {
 317                 pr_warn("rds: alloc sgs failed! nents <= 0\n");
 318                 return ERR_PTR(-EINVAL);
 319         }
 320
 321         if (rm->m_used_sgs + nents > rm->m_total_sgs) {
 322                 pr_warn("rds: alloc sgs failed! total %d used %d nents %d\n",
 323                         rm->m_total_sgs, rm->m_used_sgs, nents);
 324                 return ERR_PTR(-ENOMEM);
 325         }
 326
 327         sg_ret = &sg_first[rm->m_used_sgs];
 328         sg_init_table(sg_ret, nents);
 329         rm->m_used_sgs += nents;
 330
 331         return sg_ret;
 332 }
 333
 334 struct rds_message *rds_message_map_pages(unsigned long *page_addrs, unsigned int total_len)
 335 {
 336         struct rds_message *rm;
 337         unsigned int i;
 338         int num_sgs = DIV_ROUND_UP(total_len, PAGE_SIZE);
 339         int extra_bytes = num_sgs * sizeof(struct scatterlist);
 340
 341         rm = rds_message_alloc(extra_bytes, GFP_NOWAIT);
 342         if (!rm)
 343                 return ERR_PTR(-ENOMEM);
 344
 345         set_bit(RDS_MSG_PAGEVEC, &rm->m_flags);
 346         rm->m_inc.i_hdr.h_len = cpu_to_be32(total_len);
 347         rm->data.op_nents = DIV_ROUND_UP(total_len, PAGE_SIZE);
 348         rm->data.op_sg = rds_message_alloc_sgs(rm, num_sgs);
 349         if (IS_ERR(rm->data.op_sg)) {
 350                 rds_message_put(rm);
 351                 return ERR_CAST(rm->data.op_sg);
 352         }
 353
 354         for (i = 0; i < rm->data.op_nents; ++i) {
 355                 sg_set_page(&rm->data.op_sg[i],
 356                                 virt_to_page(page_addrs[i]),
 357                                 PAGE_SIZE, 0);
 358         }
 359
 360         return rm;
 361 }
 362
 363 static int rds_message_zcopy_from_user(struct rds_message *rm, struct iov_iter *from)
 364 {
 365         struct scatterlist *sg;
 366         int ret = 0;
 367         int length = iov_iter_count(from);
 368         int total_copied = 0;
 369         struct rds_msg_zcopy_info *info;
 370
 371         rm->m_inc.i_hdr.h_len = cpu_to_be32(iov_iter_count(from));
 372
 373         /*
 374          * now allocate and copy in the data payload.
 375          */
 376         sg = rm->data.op_sg;
 377
 378         info = kzalloc(sizeof(*info), GFP_KERNEL);
 379         if (!info)
 380                 return -ENOMEM;
 381         INIT_LIST_HEAD(&info->rs_zcookie_next);
 382         rm->data.op_mmp_znotifier = &info->znotif;
 383         if (mm_account_pinned_pages(&rm->data.op_mmp_znotifier->z_mmp,
 384                                     length)) {
 385                 ret = -ENOMEM;
 386                 goto err;
 387         }
 388         while (iov_iter_count(from)) {
 389                 struct page *pages;
 390                 size_t start;
 391                 ssize_t copied;
 392
 393                 copied = iov_iter_get_pages(from, &pages, PAGE_SIZE,
 394                                             1, &start);
 395                 if (copied < 0) {
 396                         struct mmpin *mmp;
 397                         int i;
 398
 399                         for (i = 0; i < rm->data.op_nents; i++)
 400                                 put_page(sg_page(&rm->data.op_sg[i]));
 401                         mmp = &rm->data.op_mmp_znotifier->z_mmp;
 402                         mm_unaccount_pinned_pages(mmp);
 403                         ret = -EFAULT;
 404                         goto err;
 405                 }
 406                 total_copied += copied;
 407                 iov_iter_advance(from, copied);
 408                 length -= copied;
 409                 sg_set_page(sg, pages, copied, start);
 410                 rm->data.op_nents++;
 411                 sg++;
 412         }
 413         WARN_ON_ONCE(length != 0);
 414         return ret;
 415 err:
 416         kfree(info);
 417         rm->data.op_mmp_znotifier = NULL;
 418         return ret;
 419 }
 420
 421 int rds_message_copy_from_user(struct rds_message *rm, struct iov_iter *from,
 422                                bool zcopy)
 423 {
 424         unsigned long to_copy, nbytes;
 425         unsigned long sg_off;
 426         struct scatterlist *sg;
 427         int ret = 0;
 428
 429         rm->m_inc.i_hdr.h_len = cpu_to_be32(iov_iter_count(from));
 430
 431         /* now allocate and copy in the data payload.  */
 432         sg = rm->data.op_sg;
 433         sg_off = 0; /* Dear gcc, sg->page will be null from kzalloc. */
 434
 435         if (zcopy)
 436                 return rds_message_zcopy_from_user(rm, from);
 437
 438         while (iov_iter_count(from)) {
 439                 if (!sg_page(sg)) {
 440                         ret = rds_page_remainder_alloc(sg, iov_iter_count(from),
 441                                                        GFP_HIGHUSER);
 442                         if (ret)
 443                                 return ret;
 444                         rm->data.op_nents++;
 445                         sg_off = 0;
 446                 }
 447
 448                 to_copy = min_t(unsigned long, iov_iter_count(from),
 449                                 sg->length - sg_off);
 450
 451                 rds_stats_add(s_copy_from_user, to_copy);
 452                 nbytes = copy_page_from_iter(sg_page(sg), sg->offset + sg_off,
 453                                              to_copy, from);
 454                 if (nbytes != to_copy)
 455                         return -EFAULT;
 456
 457                 sg_off += to_copy;
 458
 459                 if (sg_off == sg->length)
 460                         sg++;
 461         }
 462
 463         return ret;
 464 }
 465
 466 int rds_message_inc_copy_to_user(struct rds_incoming *inc, struct iov_iter *to)
 467 {
 468         struct rds_message *rm;
 469         struct scatterlist *sg;
 470         unsigned long to_copy;
 471         unsigned long vec_off;
 472         int copied;
 473         int ret;
 474         u32 len;
 475
 476         rm = container_of(inc, struct rds_message, m_inc);
 477         len = be32_to_cpu(rm->m_inc.i_hdr.h_len);
 478
 479         sg = rm->data.op_sg;
 480         vec_off = 0;
 481         copied = 0;
 482
 483         while (iov_iter_count(to) && copied < len) {
 484                 to_copy = min_t(unsigned long, iov_iter_count(to),
 485                                 sg->length - vec_off);
 486                 to_copy = min_t(unsigned long, to_copy, len - copied);
 487
 488                 rds_stats_add(s_copy_to_user, to_copy);
 489                 ret = copy_page_to_iter(sg_page(sg), sg->offset + vec_off,
 490                                         to_copy, to);
 491                 if (ret != to_copy)
 492                         return -EFAULT;
 493
 494                 vec_off += to_copy;
 495                 copied += to_copy;
 496
 497                 if (vec_off == sg->length) {
 498                         vec_off = 0;
 499                         sg++;
 500                 }
 501         }
 502
 503         return copied;
 504 }
 505
 506 /*
 507  * If the message is still on the send queue, wait until the transport
 508  * is done with it. This is particularly important for RDMA operations.
 509  */
 510 void rds_message_wait(struct rds_message *rm)
 511 {
 512         wait_event_interruptible(rm->m_flush_wait,
 513                         !test_bit(RDS_MSG_MAPPED, &rm->m_flags));
 514 }
 515
 516 void rds_message_unmapped(struct rds_message *rm)
 517 {
 518         clear_bit(RDS_MSG_MAPPED, &rm->m_flags);
 519         wake_up_interruptible(&rm->m_flush_wait);
 520 }
 521 EXPORT_SYMBOL_GPL(rds_message_unmapped);