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);
200 VHOST_IOCTL_RETRY(sizeof(file), 0);
203 file = *(const struct vhost_vring_file *)in_buf;
204 result = cuse_set_backend(ctx, &file);
205 fuse_reply_ioctl(req, result, NULL, 0);
208 case VHOST_GET_FEATURES:
209 LOG_DEBUG(VHOST_CONFIG,
210 "(%"PRIu64") IOCTL: VHOST_GET_FEATURES\n", ctx.fh);
211 VHOST_IOCTL_W(uint64_t, features, ops->get_features);
214 case VHOST_SET_FEATURES:
215 LOG_DEBUG(VHOST_CONFIG,
216 "(%"PRIu64") IOCTL: VHOST_SET_FEATURES\n", ctx.fh);
217 VHOST_IOCTL_R(uint64_t, features, ops->set_features);
220 case VHOST_RESET_OWNER:
221 LOG_DEBUG(VHOST_CONFIG,
222 "(%"PRIu64") IOCTL: VHOST_RESET_OWNER\n", ctx.fh);
223 VHOST_IOCTL(ops->reset_owner);
226 case VHOST_SET_OWNER:
227 LOG_DEBUG(VHOST_CONFIG,
228 "(%"PRIu64") IOCTL: VHOST_SET_OWNER\n", ctx.fh);
229 VHOST_IOCTL(ops->set_owner);
232 case VHOST_SET_MEM_TABLE:
233 /*TODO fix race condition.*/
234 LOG_DEBUG(VHOST_CONFIG,
235 "(%"PRIu64") IOCTL: VHOST_SET_MEM_TABLE\n", ctx.fh);
236 static struct vhost_memory mem_temp;
240 VHOST_IOCTL_RETRY(sizeof(struct vhost_memory), 0);
243 case sizeof(struct vhost_memory):
244 mem_temp = *(const struct vhost_memory *) in_buf;
246 if (mem_temp.nregions > 0) {
247 VHOST_IOCTL_RETRY(sizeof(struct vhost_memory) +
248 (sizeof(struct vhost_memory_region) *
249 mem_temp.nregions), 0);
252 fuse_reply_ioctl(req, result, NULL, 0);
257 result = cuse_set_mem_table(ctx, in_buf,
260 fuse_reply_err(req, EINVAL);
262 fuse_reply_ioctl(req, result, NULL, 0);
266 case VHOST_SET_VRING_NUM:
267 LOG_DEBUG(VHOST_CONFIG,
268 "(%"PRIu64") IOCTL: VHOST_SET_VRING_NUM\n", ctx.fh);
269 VHOST_IOCTL_R(struct vhost_vring_state, state,
273 case VHOST_SET_VRING_BASE:
274 LOG_DEBUG(VHOST_CONFIG,
275 "(%"PRIu64") IOCTL: VHOST_SET_VRING_BASE\n", ctx.fh);
276 VHOST_IOCTL_R(struct vhost_vring_state, state,
277 ops->set_vring_base);
280 case VHOST_GET_VRING_BASE:
281 LOG_DEBUG(VHOST_CONFIG,
282 "(%"PRIu64") IOCTL: VHOST_GET_VRING_BASE\n", ctx.fh);
283 VHOST_IOCTL_RW(uint32_t, index,
284 struct vhost_vring_state, state, ops->get_vring_base);
287 case VHOST_SET_VRING_ADDR:
288 LOG_DEBUG(VHOST_CONFIG,
289 "(%"PRIu64") IOCTL: VHOST_SET_VRING_ADDR\n", ctx.fh);
290 VHOST_IOCTL_R(struct vhost_vring_addr, addr,
291 ops->set_vring_addr);
294 case VHOST_SET_VRING_KICK:
295 case VHOST_SET_VRING_CALL:
296 if (cmd == VHOST_SET_VRING_KICK)
297 LOG_DEBUG(VHOST_CONFIG,
298 "(%"PRIu64") IOCTL: VHOST_SET_VRING_KICK\n",
301 LOG_DEBUG(VHOST_CONFIG,
302 "(%"PRIu64") IOCTL: VHOST_SET_VRING_CALL\n",
305 VHOST_IOCTL_RETRY(sizeof(struct vhost_vring_file), 0);
308 file = *(const struct vhost_vring_file *)in_buf;
309 LOG_DEBUG(VHOST_CONFIG,
310 "idx:%d fd:%d\n", file.index, file.fd);
311 fd = eventfd_copy(file.fd, ctx.pid);
313 fuse_reply_ioctl(req, -1, NULL, 0);
318 if (cmd == VHOST_SET_VRING_KICK) {
319 result = ops->set_vring_kick(ctx, &file);
320 fuse_reply_ioctl(req, result, NULL, 0);
322 result = ops->set_vring_call(ctx, &file);
323 fuse_reply_ioctl(req, result, NULL, 0);
329 RTE_LOG(ERR, VHOST_CONFIG,
330 "(%"PRIu64") IOCTL: DOESN NOT EXIST\n", ctx.fh);
332 fuse_reply_ioctl(req, result, NULL, 0);
336 LOG_DEBUG(VHOST_CONFIG,
337 "(%"PRIu64") IOCTL: FAIL\n", ctx.fh);
339 LOG_DEBUG(VHOST_CONFIG,
340 "(%"PRIu64") IOCTL: SUCCESS\n", ctx.fh);
344 * Structure handling open, release and ioctl function pointers is populated.
346 static const struct cuse_lowlevel_ops vhost_net_ops = {
347 .open = vhost_net_open,
348 .release = vhost_net_release,
349 .ioctl = vhost_net_ioctl,
353 * cuse_info is populated and used to register the cuse device.
354 * vhost_net_device_ops are also passed when the device is registered in app.
357 rte_vhost_driver_register(const char *dev_name)
359 struct cuse_info cuse_info;
360 char device_name[PATH_MAX] = "";
361 char char_device_name[PATH_MAX] = "";
362 const char *device_argv[] = { device_name };
364 char fuse_opt_dummy[] = FUSE_OPT_DUMMY;
365 char fuse_opt_fore[] = FUSE_OPT_FORE;
366 char fuse_opt_nomulti[] = FUSE_OPT_NOMULTI;
367 char *fuse_argv[] = {fuse_opt_dummy, fuse_opt_fore, fuse_opt_nomulti};
369 if (access(cuse_device_name, R_OK | W_OK) < 0) {
370 RTE_LOG(ERR, VHOST_CONFIG,
371 "char device %s can't be accessed, maybe not exist\n",
376 if (eventfd_init() < 0)
380 * The device name is created. This is passed to QEMU so that it can
381 * register the device with our application.
383 snprintf(device_name, PATH_MAX, "DEVNAME=%s", dev_name);
384 snprintf(char_device_name, PATH_MAX, "/dev/%s", dev_name);
386 /* Check if device already exists. */
387 if (access(char_device_name, F_OK) != -1) {
388 RTE_LOG(ERR, VHOST_CONFIG,
389 "char device %s already exists\n", char_device_name);
393 memset(&cuse_info, 0, sizeof(cuse_info));
394 cuse_info.dev_major = default_major;
395 cuse_info.dev_minor = default_minor;
396 cuse_info.dev_info_argc = 1;
397 cuse_info.dev_info_argv = device_argv;
398 cuse_info.flags = CUSE_UNRESTRICTED_IOCTL;
400 ops = get_virtio_net_callbacks();
402 session = cuse_lowlevel_setup(3, fuse_argv,
403 &cuse_info, &vhost_net_ops, 0, NULL);
411 * An empty function for unregister
414 rte_vhost_driver_unregister(const char *dev_name __rte_unused)
420 * The CUSE session is launched allowing the application to receive open,
421 * release and ioctl calls.
424 rte_vhost_driver_session_start(void)
426 fuse_session_loop(session);