4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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
17 * * Neither the name of Intel Corporation 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.
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.
35 #include <fuse/cuse_lowlevel.h>
36 #include <linux/limits.h>
37 #include <linux/vhost.h>
42 #include <rte_ethdev.h>
44 #include <rte_string_fns.h>
45 #include <rte_virtio_net.h>
47 #include "virtio-net-cdev.h"
48 #include "vhost-net.h"
49 #include "eventfd_copy.h"
51 #define FUSE_OPT_DUMMY "\0\0"
52 #define FUSE_OPT_FORE "-f\0\0"
53 #define FUSE_OPT_NOMULTI "-s\0\0"
55 static const uint32_t default_major = 231;
56 static const uint32_t default_minor = 1;
57 static const char cuse_device_name[] = "/dev/cuse";
58 static const char default_cdev[] = "vhost-net";
60 static struct fuse_session *session;
61 struct vhost_net_device_ops const *ops;
64 * Returns vhost_device_ctx from given fuse_req_t. The index is populated later
65 * when the device is added to the device linked list.
67 static struct vhost_device_ctx
68 fuse_req_to_vhost_ctx(fuse_req_t req, struct fuse_file_info *fi)
70 struct vhost_device_ctx ctx;
71 struct fuse_ctx const *const req_ctx = fuse_req_ctx(req);
73 ctx.pid = req_ctx->pid;
80 * When the device is created in QEMU it gets initialised here and
81 * added to the device linked list.
84 vhost_net_open(fuse_req_t req, struct fuse_file_info *fi)
86 struct vhost_device_ctx ctx = fuse_req_to_vhost_ctx(req, fi);
89 err = ops->new_device(ctx);
91 fuse_reply_err(req, EPERM);
97 RTE_LOG(INFO, VHOST_CONFIG,
98 "(%"PRIu64") Device configuration started\n", fi->fh);
99 fuse_reply_open(req, fi);
103 * When QEMU is shutdown or killed the device gets released.
106 vhost_net_release(fuse_req_t req, struct fuse_file_info *fi)
109 struct vhost_device_ctx ctx = fuse_req_to_vhost_ctx(req, fi);
111 ops->destroy_device(ctx);
112 RTE_LOG(INFO, VHOST_CONFIG, "(%"PRIu64") Device released\n", ctx.fh);
113 fuse_reply_err(req, err);
117 * Boilerplate code for CUSE IOCTL
118 * Implicit arguments: ctx, req, result.
120 #define VHOST_IOCTL(func) do { \
121 result = (func)(ctx); \
122 fuse_reply_ioctl(req, result, NULL, 0); \
126 * Boilerplate IOCTL RETRY
127 * Implicit arguments: req.
129 #define VHOST_IOCTL_RETRY(size_r, size_w) do { \
130 struct iovec iov_r = { arg, (size_r) }; \
131 struct iovec iov_w = { arg, (size_w) }; \
132 fuse_reply_ioctl_retry(req, &iov_r, \
133 (size_r) ? 1 : 0, &iov_w, (size_w) ? 1 : 0);\
137 * Boilerplate code for CUSE Read IOCTL
138 * Implicit arguments: ctx, req, result, in_bufsz, in_buf.
140 #define VHOST_IOCTL_R(type, var, func) do { \
142 VHOST_IOCTL_RETRY(sizeof(type), 0);\
144 (var) = *(const type*)in_buf; \
145 result = func(ctx, &(var)); \
146 fuse_reply_ioctl(req, result, NULL, 0);\
151 * Boilerplate code for CUSE Write IOCTL
152 * Implicit arguments: ctx, req, result, out_bufsz.
154 #define VHOST_IOCTL_W(type, var, func) do { \
156 VHOST_IOCTL_RETRY(0, sizeof(type));\
158 result = (func)(ctx, &(var));\
159 fuse_reply_ioctl(req, result, &(var), sizeof(type));\
164 * Boilerplate code for CUSE Read/Write IOCTL
165 * Implicit arguments: ctx, req, result, in_bufsz, in_buf.
167 #define VHOST_IOCTL_RW(type1, var1, type2, var2, func) do { \
169 VHOST_IOCTL_RETRY(sizeof(type1), sizeof(type2));\
171 (var1) = *(const type1*) (in_buf); \
172 result = (func)(ctx, (var1), &(var2)); \
173 fuse_reply_ioctl(req, result, &(var2), sizeof(type2));\
178 * The IOCTLs are handled using CUSE/FUSE in userspace. Depending on the type
179 * of IOCTL a buffer is requested to read or to write. This request is handled
180 * by FUSE and the buffer is then given to CUSE.
183 vhost_net_ioctl(fuse_req_t req, int cmd, void *arg,
184 struct fuse_file_info *fi, __rte_unused unsigned flags,
185 const void *in_buf, size_t in_bufsz, size_t out_bufsz)
187 struct vhost_device_ctx ctx = fuse_req_to_vhost_ctx(req, fi);
188 struct vhost_vring_file file;
189 struct vhost_vring_state state;
190 struct vhost_vring_addr addr;
196 case VHOST_NET_SET_BACKEND:
197 LOG_DEBUG(VHOST_CONFIG,
198 "(%"PRIu64") IOCTL: VHOST_NET_SET_BACKEND\n", ctx.fh);
199 VHOST_IOCTL_R(struct vhost_vring_file, file, ops->set_backend);
202 case VHOST_GET_FEATURES:
203 LOG_DEBUG(VHOST_CONFIG,
204 "(%"PRIu64") IOCTL: VHOST_GET_FEATURES\n", ctx.fh);
205 VHOST_IOCTL_W(uint64_t, features, ops->get_features);
208 case VHOST_SET_FEATURES:
209 LOG_DEBUG(VHOST_CONFIG,
210 "(%"PRIu64") IOCTL: VHOST_SET_FEATURES\n", ctx.fh);
211 VHOST_IOCTL_R(uint64_t, features, ops->set_features);
214 case VHOST_RESET_OWNER:
215 LOG_DEBUG(VHOST_CONFIG,
216 "(%"PRIu64") IOCTL: VHOST_RESET_OWNER\n", ctx.fh);
217 VHOST_IOCTL(ops->reset_owner);
220 case VHOST_SET_OWNER:
221 LOG_DEBUG(VHOST_CONFIG,
222 "(%"PRIu64") IOCTL: VHOST_SET_OWNER\n", ctx.fh);
223 VHOST_IOCTL(ops->set_owner);
226 case VHOST_SET_MEM_TABLE:
227 /*TODO fix race condition.*/
228 LOG_DEBUG(VHOST_CONFIG,
229 "(%"PRIu64") IOCTL: VHOST_SET_MEM_TABLE\n", ctx.fh);
230 static struct vhost_memory mem_temp;
234 VHOST_IOCTL_RETRY(sizeof(struct vhost_memory), 0);
237 case sizeof(struct vhost_memory):
238 mem_temp = *(const struct vhost_memory *) in_buf;
240 if (mem_temp.nregions > 0) {
241 VHOST_IOCTL_RETRY(sizeof(struct vhost_memory) +
242 (sizeof(struct vhost_memory_region) *
243 mem_temp.nregions), 0);
246 fuse_reply_ioctl(req, result, NULL, 0);
251 result = cuse_set_mem_table(ctx, in_buf,
254 fuse_reply_err(req, EINVAL);
256 fuse_reply_ioctl(req, result, NULL, 0);
260 case VHOST_SET_VRING_NUM:
261 LOG_DEBUG(VHOST_CONFIG,
262 "(%"PRIu64") IOCTL: VHOST_SET_VRING_NUM\n", ctx.fh);
263 VHOST_IOCTL_R(struct vhost_vring_state, state,
267 case VHOST_SET_VRING_BASE:
268 LOG_DEBUG(VHOST_CONFIG,
269 "(%"PRIu64") IOCTL: VHOST_SET_VRING_BASE\n", ctx.fh);
270 VHOST_IOCTL_R(struct vhost_vring_state, state,
271 ops->set_vring_base);
274 case VHOST_GET_VRING_BASE:
275 LOG_DEBUG(VHOST_CONFIG,
276 "(%"PRIu64") IOCTL: VHOST_GET_VRING_BASE\n", ctx.fh);
277 VHOST_IOCTL_RW(uint32_t, index,
278 struct vhost_vring_state, state, ops->get_vring_base);
281 case VHOST_SET_VRING_ADDR:
282 LOG_DEBUG(VHOST_CONFIG,
283 "(%"PRIu64") IOCTL: VHOST_SET_VRING_ADDR\n", ctx.fh);
284 VHOST_IOCTL_R(struct vhost_vring_addr, addr,
285 ops->set_vring_addr);
288 case VHOST_SET_VRING_KICK:
289 case VHOST_SET_VRING_CALL:
290 if (cmd == VHOST_SET_VRING_KICK)
291 LOG_DEBUG(VHOST_CONFIG,
292 "(%"PRIu64") IOCTL: VHOST_SET_VRING_KICK\n",
295 LOG_DEBUG(VHOST_CONFIG,
296 "(%"PRIu64") IOCTL: VHOST_SET_VRING_CALL\n",
299 VHOST_IOCTL_RETRY(sizeof(struct vhost_vring_file), 0);
302 file = *(const struct vhost_vring_file *)in_buf;
303 LOG_DEBUG(VHOST_CONFIG,
304 "idx:%d fd:%d\n", file.index, file.fd);
305 fd = eventfd_copy(file.fd, ctx.pid);
307 fuse_reply_ioctl(req, -1, NULL, 0);
312 if (cmd == VHOST_SET_VRING_KICK) {
313 result = ops->set_vring_kick(ctx, &file);
314 fuse_reply_ioctl(req, result, NULL, 0);
316 result = ops->set_vring_call(ctx, &file);
317 fuse_reply_ioctl(req, result, NULL, 0);
323 RTE_LOG(ERR, VHOST_CONFIG,
324 "(%"PRIu64") IOCTL: DOESN NOT EXIST\n", ctx.fh);
326 fuse_reply_ioctl(req, result, NULL, 0);
330 LOG_DEBUG(VHOST_CONFIG,
331 "(%"PRIu64") IOCTL: FAIL\n", ctx.fh);
333 LOG_DEBUG(VHOST_CONFIG,
334 "(%"PRIu64") IOCTL: SUCCESS\n", ctx.fh);
338 * Structure handling open, release and ioctl function pointers is populated.
340 static const struct cuse_lowlevel_ops vhost_net_ops = {
341 .open = vhost_net_open,
342 .release = vhost_net_release,
343 .ioctl = vhost_net_ioctl,
347 * cuse_info is populated and used to register the cuse device.
348 * vhost_net_device_ops are also passed when the device is registered in app.
351 rte_vhost_driver_register(const char *dev_name)
353 struct cuse_info cuse_info;
354 char device_name[PATH_MAX] = "";
355 char char_device_name[PATH_MAX] = "";
356 const char *device_argv[] = { device_name };
358 char fuse_opt_dummy[] = FUSE_OPT_DUMMY;
359 char fuse_opt_fore[] = FUSE_OPT_FORE;
360 char fuse_opt_nomulti[] = FUSE_OPT_NOMULTI;
361 char *fuse_argv[] = {fuse_opt_dummy, fuse_opt_fore, fuse_opt_nomulti};
363 if (access(cuse_device_name, R_OK | W_OK) < 0) {
364 RTE_LOG(ERR, VHOST_CONFIG,
365 "char device %s can't be accessed, maybe not exist\n",
371 * The device name is created. This is passed to QEMU so that it can
372 * register the device with our application.
374 snprintf(device_name, PATH_MAX, "DEVNAME=%s", dev_name);
375 snprintf(char_device_name, PATH_MAX, "/dev/%s", dev_name);
377 /* Check if device already exists. */
378 if (access(char_device_name, F_OK) != -1) {
379 RTE_LOG(ERR, VHOST_CONFIG,
380 "char device %s already exists\n", char_device_name);
384 memset(&cuse_info, 0, sizeof(cuse_info));
385 cuse_info.dev_major = default_major;
386 cuse_info.dev_minor = default_minor;
387 cuse_info.dev_info_argc = 1;
388 cuse_info.dev_info_argv = device_argv;
389 cuse_info.flags = CUSE_UNRESTRICTED_IOCTL;
391 ops = get_virtio_net_callbacks();
393 session = cuse_lowlevel_setup(3, fuse_argv,
394 &cuse_info, &vhost_net_ops, 0, NULL);
402 * The CUSE session is launched allowing the application to receive open,
403 * release and ioctl calls.
406 rte_vhost_driver_session_start(void)
408 fuse_session_loop(session);