net/virtio: support vhost-user protocol features
[dpdk.git] / drivers / net / virtio / virtio_user / vhost_user.c
index 4ad7b21..e330675 100644 (file)
@@ -1,34 +1,5 @@
-/*-
- *   BSD LICENSE
- *
- *   Copyright(c) 2010-2016 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.
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2016 Intel Corporation
  */
 
 #include <sys/socket.h>
@@ -40,6 +11,9 @@
 #include <string.h>
 #include <errno.h>
 
+#include <rte_string_fns.h>
+#include <rte_fbarray.h>
+
 #include "vhost.h"
 #include "virtio_user_dev.h"
 
@@ -69,7 +43,7 @@ struct vhost_user_msg {
                struct vhost_memory memory;
        } payload;
        int fds[VHOST_MEMORY_MAX_NREGIONS];
-} __attribute((packed));
+} __rte_packed;
 
 #define VHOST_USER_HDR_SIZE offsetof(struct vhost_user_msg, payload.u64)
 #define VHOST_USER_PAYLOAD_SIZE \
@@ -130,6 +104,10 @@ vhost_user_read(int fd, struct vhost_user_msg *msg)
        }
 
        sz_payload = msg->size;
+
+       if ((size_t)sz_payload > sizeof(msg->payload))
+               goto fail;
+
        if (sz_payload) {
                ret = recv(fd, (void *)((char *)msg + sz_hdr), sz_payload, 0);
                if (ret < sz_payload) {
@@ -146,109 +124,103 @@ fail:
        return -1;
 }
 
-struct hugepage_file_info {
-       uint64_t addr;            /**< virtual addr */
-       size_t   size;            /**< the file size */
-       char     path[PATH_MAX];  /**< path to backing file */
+struct walk_arg {
+       struct vhost_memory *vm;
+       int *fds;
+       int region_nr;
 };
 
-/* Two possible options:
- * 1. Match HUGEPAGE_INFO_FMT to find the file storing struct hugepage_file
- * array. This is simple but cannot be used in secondary process because
- * secondary process will close and munmap that file.
- * 2. Match HUGEFILE_FMT to find hugepage files directly.
- *
- * We choose option 2.
- */
 static int
-get_hugepage_file_info(struct hugepage_file_info huges[], int max)
+update_memory_region(const struct rte_memseg_list *msl __rte_unused,
+               const struct rte_memseg *ms, void *arg)
 {
-       int idx;
-       FILE *f;
-       char buf[BUFSIZ], *tmp, *tail;
-       char *str_underline, *str_start;
-       int huge_index;
-       uint64_t v_start, v_end;
-
-       f = fopen("/proc/self/maps", "r");
-       if (!f) {
-               PMD_DRV_LOG(ERR, "cannot open /proc/self/maps");
+       struct walk_arg *wa = arg;
+       struct vhost_memory_region *mr;
+       uint64_t start_addr, end_addr;
+       size_t offset;
+       int i, fd;
+
+       fd = rte_memseg_get_fd_thread_unsafe(ms);
+       if (fd < 0) {
+               PMD_DRV_LOG(ERR, "Failed to get fd, ms=%p rte_errno=%d",
+                       ms, rte_errno);
                return -1;
        }
 
-       idx = 0;
-       while (fgets(buf, sizeof(buf), f) != NULL) {
-               if (sscanf(buf, "%" PRIx64 "-%" PRIx64, &v_start, &v_end) < 2) {
-                       PMD_DRV_LOG(ERR, "Failed to parse address");
-                       goto error;
-               }
+       if (rte_memseg_get_fd_offset_thread_unsafe(ms, &offset) < 0) {
+               PMD_DRV_LOG(ERR, "Failed to get offset, ms=%p rte_errno=%d",
+                       ms, rte_errno);
+               return -1;
+       }
 
-               tmp = strchr(buf, ' ') + 1; /** skip address */
-               tmp = strchr(tmp, ' ') + 1; /** skip perm */
-               tmp = strchr(tmp, ' ') + 1; /** skip offset */
-               tmp = strchr(tmp, ' ') + 1; /** skip dev */
-               tmp = strchr(tmp, ' ') + 1; /** skip inode */
-               while (*tmp == ' ')         /** skip spaces */
-                       tmp++;
-               tail = strrchr(tmp, '\n');  /** remove newline if exists */
-               if (tail)
-                       *tail = '\0';
-
-               /* Match HUGEFILE_FMT, aka "%s/%smap_%d",
-                * which is defined in eal_filesystem.h
-                */
-               str_underline = strrchr(tmp, '_');
-               if (!str_underline)
-                       continue;
+       start_addr = (uint64_t)(uintptr_t)ms->addr;
+       end_addr = start_addr + ms->len;
 
-               str_start = str_underline - strlen("map");
-               if (str_start < tmp)
+       for (i = 0; i < wa->region_nr; i++) {
+               if (wa->fds[i] != fd)
                        continue;
 
-               if (sscanf(str_start, "map_%d", &huge_index) != 1)
-                       continue;
+               mr = &wa->vm->regions[i];
+
+               if (mr->userspace_addr + mr->memory_size < end_addr)
+                       mr->memory_size = end_addr - mr->userspace_addr;
 
-               if (idx >= max) {
-                       PMD_DRV_LOG(ERR, "Exceed maximum of %d", max);
-                       goto error;
+               if (mr->userspace_addr > start_addr) {
+                       mr->userspace_addr = start_addr;
+                       mr->guest_phys_addr = start_addr;
                }
-               huges[idx].addr = v_start;
-               huges[idx].size = v_end - v_start;
-               snprintf(huges[idx].path, PATH_MAX, "%s", tmp);
-               idx++;
+
+               if (mr->mmap_offset > offset)
+                       mr->mmap_offset = offset;
+
+               PMD_DRV_LOG(DEBUG, "index=%d fd=%d offset=0x%" PRIx64
+                       " addr=0x%" PRIx64 " len=%" PRIu64, i, fd,
+                       mr->mmap_offset, mr->userspace_addr,
+                       mr->memory_size);
+
+               return 0;
        }
 
-       fclose(f);
-       return idx;
+       if (i >= VHOST_MEMORY_MAX_NREGIONS) {
+               PMD_DRV_LOG(ERR, "Too many memory regions");
+               return -1;
+       }
 
-error:
-       fclose(f);
-       return -1;
+       mr = &wa->vm->regions[i];
+       wa->fds[i] = fd;
+
+       mr->guest_phys_addr = start_addr;
+       mr->userspace_addr = start_addr;
+       mr->memory_size = ms->len;
+       mr->mmap_offset = offset;
+
+       PMD_DRV_LOG(DEBUG, "index=%d fd=%d offset=0x%" PRIx64
+               " addr=0x%" PRIx64 " len=%" PRIu64, i, fd,
+               mr->mmap_offset, mr->userspace_addr,
+               mr->memory_size);
+
+       wa->region_nr++;
+
+       return 0;
 }
 
 static int
 prepare_vhost_memory_user(struct vhost_user_msg *msg, int fds[])
 {
-       int i, num;
-       struct hugepage_file_info huges[VHOST_MEMORY_MAX_NREGIONS];
-       struct vhost_memory_region *mr;
+       struct walk_arg wa;
 
-       num = get_hugepage_file_info(huges, VHOST_MEMORY_MAX_NREGIONS);
-       if (num < 0) {
-               PMD_INIT_LOG(ERR, "Failed to prepare memory for vhost-user");
-               return -1;
-       }
+       wa.region_nr = 0;
+       wa.vm = &msg->payload.memory;
+       wa.fds = fds;
 
-       for (i = 0; i < num; ++i) {
-               mr = &msg->payload.memory.regions[i];
-               mr->guest_phys_addr = huges[i].addr; /* use vaddr! */
-               mr->userspace_addr = huges[i].addr;
-               mr->memory_size = huges[i].size;
-               mr->mmap_offset = 0;
-               fds[i] = open(huges[i].path, O_RDWR);
-       }
+       /*
+        * The memory lock has already been taken by memory subsystem
+        * or virtio_user_start_device().
+        */
+       if (rte_memseg_walk_thread_unsafe(update_memory_region, &wa) < 0)
+               return -1;
 
-       msg->payload.memory.nregions = num;
+       msg->payload.memory.nregions = wa.region_nr;
        msg->payload.memory.padding = 0;
 
        return 0;
@@ -269,6 +241,8 @@ const char * const vhost_msg_strings[] = {
        [VHOST_USER_SET_VRING_KICK] = "VHOST_SET_VRING_KICK",
        [VHOST_USER_SET_MEM_TABLE] = "VHOST_SET_MEM_TABLE",
        [VHOST_USER_SET_VRING_ENABLE] = "VHOST_SET_VRING_ENABLE",
+       [VHOST_USER_GET_PROTOCOL_FEATURES] = "VHOST_USER_GET_PROTOCOL_FEATURES",
+       [VHOST_USER_SET_PROTOCOL_FEATURES] = "VHOST_USER_SET_PROTOCOL_FEATURES",
 };
 
 static int
@@ -281,23 +255,28 @@ vhost_user_sock(struct virtio_user_dev *dev,
        int need_reply = 0;
        int fds[VHOST_MEMORY_MAX_NREGIONS];
        int fd_num = 0;
-       int i, len;
+       int len;
        int vhostfd = dev->vhostfd;
 
        RTE_SET_USED(m);
 
        PMD_DRV_LOG(INFO, "%s", vhost_msg_strings[req]);
 
+       if (dev->is_server && vhostfd < 0)
+               return -1;
+
        msg.request = req;
        msg.flags = VHOST_USER_VERSION;
        msg.size = 0;
 
        switch (req) {
        case VHOST_USER_GET_FEATURES:
+       case VHOST_USER_GET_PROTOCOL_FEATURES:
                need_reply = 1;
                break;
 
        case VHOST_USER_SET_FEATURES:
+       case VHOST_USER_SET_PROTOCOL_FEATURES:
        case VHOST_USER_SET_LOG_BASE:
                msg.payload.u64 = *((__u64 *)arg);
                msg.size = sizeof(m.payload.u64);
@@ -362,10 +341,6 @@ vhost_user_sock(struct virtio_user_dev *dev,
                return -1;
        }
 
-       if (req == VHOST_USER_SET_MEM_TABLE)
-               for (i = 0; i < fd_num; ++i)
-                       close(fds[i]);
-
        if (need_reply) {
                if (vhost_user_read(vhostfd, &msg) < 0) {
                        PMD_DRV_LOG(ERR, "Received msg failed: %s",
@@ -380,6 +355,7 @@ vhost_user_sock(struct virtio_user_dev *dev,
 
                switch (req) {
                case VHOST_USER_GET_FEATURES:
+               case VHOST_USER_GET_PROTOCOL_FEATURES:
                        if (msg.size != sizeof(m.payload.u64)) {
                                PMD_DRV_LOG(ERR, "Received bad msg size");
                                return -1;
@@ -403,6 +379,33 @@ vhost_user_sock(struct virtio_user_dev *dev,
        return 0;
 }
 
+#define MAX_VIRTIO_USER_BACKLOG 1
+static int
+virtio_user_start_server(struct virtio_user_dev *dev, struct sockaddr_un *un)
+{
+       int ret;
+       int flag;
+       int fd = dev->listenfd;
+
+       ret = bind(fd, (struct sockaddr *)un, sizeof(*un));
+       if (ret < 0) {
+               PMD_DRV_LOG(ERR, "failed to bind to %s: %s; remove it and try again\n",
+                           dev->path, strerror(errno));
+               return -1;
+       }
+       ret = listen(fd, MAX_VIRTIO_USER_BACKLOG);
+       if (ret < 0)
+               return -1;
+
+       flag = fcntl(fd, F_GETFL);
+       if (fcntl(fd, F_SETFL, flag | O_NONBLOCK) < 0) {
+               PMD_DRV_LOG(ERR, "fcntl failed, %s", strerror(errno));
+               return -1;
+       }
+
+       return 0;
+}
+
 /**
  * Set up environment to talk with a vhost user backend.
  *
@@ -429,14 +432,25 @@ vhost_user_setup(struct virtio_user_dev *dev)
 
        memset(&un, 0, sizeof(un));
        un.sun_family = AF_UNIX;
-       snprintf(un.sun_path, sizeof(un.sun_path), "%s", dev->path);
-       if (connect(fd, (struct sockaddr *)&un, sizeof(un)) < 0) {
-               PMD_DRV_LOG(ERR, "connect error, %s", strerror(errno));
-               close(fd);
-               return -1;
+       strlcpy(un.sun_path, dev->path, sizeof(un.sun_path));
+
+       if (dev->is_server) {
+               dev->listenfd = fd;
+               if (virtio_user_start_server(dev, &un) < 0) {
+                       PMD_DRV_LOG(ERR, "virtio-user startup fails in server mode");
+                       close(fd);
+                       return -1;
+               }
+               dev->vhostfd = -1;
+       } else {
+               if (connect(fd, (struct sockaddr *)&un, sizeof(un)) < 0) {
+                       PMD_DRV_LOG(ERR, "connect error, %s", strerror(errno));
+                       close(fd);
+                       return -1;
+               }
+               dev->vhostfd = fd;
        }
 
-       dev->vhostfd = fd;
        return 0;
 }
 
@@ -447,6 +461,9 @@ vhost_user_enable_queue_pair(struct virtio_user_dev *dev,
 {
        int i;
 
+       if (dev->qp_enabled[pair_idx] == enable)
+               return 0;
+
        for (i = 0; i < 2; ++i) {
                struct vhost_vring_state state = {
                        .index = pair_idx * 2 + i,
@@ -457,10 +474,11 @@ vhost_user_enable_queue_pair(struct virtio_user_dev *dev,
                        return -1;
        }
 
+       dev->qp_enabled[pair_idx] = enable;
        return 0;
 }
 
-struct virtio_user_backend_ops ops_user = {
+struct virtio_user_backend_ops virtio_ops_user = {
        .setup = vhost_user_setup,
        .send_request = vhost_user_sock,
        .enable_qp = vhost_user_enable_queue_pair