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 "vhost-net-cdev.h"
49 #define FUSE_OPT_DUMMY "\0\0"
50 #define FUSE_OPT_FORE "-f\0\0"
51 #define FUSE_OPT_NOMULTI "-s\0\0"
53 static const uint32_t default_major = 231;
54 static const uint32_t default_minor = 1;
55 static const char cuse_device_name[] = "/dev/cuse";
56 static const char default_cdev[] = "vhost-net";
58 static struct fuse_session *session;
59 static struct vhost_net_device_ops const *ops;
62 * Returns vhost_device_ctx from given fuse_req_t. The index is populated later
63 * when the device is added to the device linked list.
65 static struct vhost_device_ctx
66 fuse_req_to_vhost_ctx(fuse_req_t req, struct fuse_file_info *fi)
68 struct vhost_device_ctx ctx;
69 struct fuse_ctx const *const req_ctx = fuse_req_ctx(req);
71 ctx.pid = req_ctx->pid;
78 * When the device is created in QEMU it gets initialised here and
79 * added to the device linked list.
82 vhost_net_open(fuse_req_t req, struct fuse_file_info *fi)
84 struct vhost_device_ctx ctx = fuse_req_to_vhost_ctx(req, fi);
87 err = ops->new_device(ctx);
89 fuse_reply_err(req, EPERM);
95 RTE_LOG(INFO, VHOST_CONFIG,
96 "(%"PRIu64") Device configuration started\n", fi->fh);
97 fuse_reply_open(req, fi);
101 * When QEMU is shutdown or killed the device gets released.
104 vhost_net_release(fuse_req_t req, struct fuse_file_info *fi)
107 struct vhost_device_ctx ctx = fuse_req_to_vhost_ctx(req, fi);
109 ops->destroy_device(ctx);
110 RTE_LOG(INFO, VHOST_CONFIG, "(%"PRIu64") Device released\n", ctx.fh);
111 fuse_reply_err(req, err);
115 * Boilerplate code for CUSE IOCTL
116 * Implicit arguments: ctx, req, result.
118 #define VHOST_IOCTL(func) do { \
119 result = (func)(ctx); \
120 fuse_reply_ioctl(req, result, NULL, 0); \
124 * Boilerplate IOCTL RETRY
125 * Implicit arguments: req.
127 #define VHOST_IOCTL_RETRY(size_r, size_w) do { \
128 struct iovec iov_r = { arg, (size_r) }; \
129 struct iovec iov_w = { arg, (size_w) }; \
130 fuse_reply_ioctl_retry(req, &iov_r, \
131 (size_r) ? 1 : 0, &iov_w, (size_w) ? 1 : 0);\
135 * Boilerplate code for CUSE Read IOCTL
136 * Implicit arguments: ctx, req, result, in_bufsz, in_buf.
138 #define VHOST_IOCTL_R(type, var, func) do { \
140 VHOST_IOCTL_RETRY(sizeof(type), 0);\
142 (var) = *(const type*)in_buf; \
143 result = func(ctx, &(var)); \
144 fuse_reply_ioctl(req, result, NULL, 0);\
149 * Boilerplate code for CUSE Write IOCTL
150 * Implicit arguments: ctx, req, result, out_bufsz.
152 #define VHOST_IOCTL_W(type, var, func) do { \
154 VHOST_IOCTL_RETRY(0, sizeof(type));\
156 result = (func)(ctx, &(var));\
157 fuse_reply_ioctl(req, result, &(var), sizeof(type));\
162 * Boilerplate code for CUSE Read/Write IOCTL
163 * Implicit arguments: ctx, req, result, in_bufsz, in_buf.
165 #define VHOST_IOCTL_RW(type1, var1, type2, var2, func) do { \
167 VHOST_IOCTL_RETRY(sizeof(type1), sizeof(type2));\
169 (var1) = *(const type1*) (in_buf); \
170 result = (func)(ctx, (var1), &(var2)); \
171 fuse_reply_ioctl(req, result, &(var2), sizeof(type2));\
176 * The IOCTLs are handled using CUSE/FUSE in userspace. Depending on the type
177 * of IOCTL a buffer is requested to read or to write. This request is handled
178 * by FUSE and the buffer is then given to CUSE.
181 vhost_net_ioctl(fuse_req_t req, int cmd, void *arg,
182 struct fuse_file_info *fi, __rte_unused unsigned flags,
183 const void *in_buf, size_t in_bufsz, size_t out_bufsz)
185 struct vhost_device_ctx ctx = fuse_req_to_vhost_ctx(req, fi);
186 struct vhost_vring_file file;
187 struct vhost_vring_state state;
188 struct vhost_vring_addr addr;
194 case VHOST_NET_SET_BACKEND:
195 LOG_DEBUG(VHOST_CONFIG,
196 "(%"PRIu64") IOCTL: VHOST_NET_SET_BACKEND\n", ctx.fh);
197 VHOST_IOCTL_R(struct vhost_vring_file, file, ops->set_backend);
200 case VHOST_GET_FEATURES:
201 LOG_DEBUG(VHOST_CONFIG,
202 "(%"PRIu64") IOCTL: VHOST_GET_FEATURES\n", ctx.fh);
203 VHOST_IOCTL_W(uint64_t, features, ops->get_features);
206 case VHOST_SET_FEATURES:
207 LOG_DEBUG(VHOST_CONFIG,
208 "(%"PRIu64") IOCTL: VHOST_SET_FEATURES\n", ctx.fh);
209 VHOST_IOCTL_R(uint64_t, features, ops->set_features);
212 case VHOST_RESET_OWNER:
213 LOG_DEBUG(VHOST_CONFIG,
214 "(%"PRIu64") IOCTL: VHOST_RESET_OWNER\n", ctx.fh);
215 VHOST_IOCTL(ops->reset_owner);
218 case VHOST_SET_OWNER:
219 LOG_DEBUG(VHOST_CONFIG,
220 "(%"PRIu64") IOCTL: VHOST_SET_OWNER\n", ctx.fh);
221 VHOST_IOCTL(ops->set_owner);
224 case VHOST_SET_MEM_TABLE:
225 /*TODO fix race condition.*/
226 LOG_DEBUG(VHOST_CONFIG,
227 "(%"PRIu64") IOCTL: VHOST_SET_MEM_TABLE\n", ctx.fh);
228 static struct vhost_memory mem_temp;
232 VHOST_IOCTL_RETRY(sizeof(struct vhost_memory), 0);
235 case sizeof(struct vhost_memory):
236 mem_temp = *(const struct vhost_memory *) in_buf;
238 if (mem_temp.nregions > 0) {
239 VHOST_IOCTL_RETRY(sizeof(struct vhost_memory) +
240 (sizeof(struct vhost_memory_region) *
241 mem_temp.nregions), 0);
244 fuse_reply_ioctl(req, result, NULL, 0);
249 result = ops->set_mem_table(ctx,
250 in_buf, mem_temp.nregions);
252 fuse_reply_err(req, EINVAL);
254 fuse_reply_ioctl(req, result, NULL, 0);
258 case VHOST_SET_VRING_NUM:
259 LOG_DEBUG(VHOST_CONFIG,
260 "(%"PRIu64") IOCTL: VHOST_SET_VRING_NUM\n", ctx.fh);
261 VHOST_IOCTL_R(struct vhost_vring_state, state,
265 case VHOST_SET_VRING_BASE:
266 LOG_DEBUG(VHOST_CONFIG,
267 "(%"PRIu64") IOCTL: VHOST_SET_VRING_BASE\n", ctx.fh);
268 VHOST_IOCTL_R(struct vhost_vring_state, state,
269 ops->set_vring_base);
272 case VHOST_GET_VRING_BASE:
273 LOG_DEBUG(VHOST_CONFIG,
274 "(%"PRIu64") IOCTL: VHOST_GET_VRING_BASE\n", ctx.fh);
275 VHOST_IOCTL_RW(uint32_t, index,
276 struct vhost_vring_state, state, ops->get_vring_base);
279 case VHOST_SET_VRING_ADDR:
280 LOG_DEBUG(VHOST_CONFIG,
281 "(%"PRIu64") IOCTL: VHOST_SET_VRING_ADDR\n", ctx.fh);
282 VHOST_IOCTL_R(struct vhost_vring_addr, addr,
283 ops->set_vring_addr);
286 case VHOST_SET_VRING_KICK:
287 LOG_DEBUG(VHOST_CONFIG,
288 "(%"PRIu64") IOCTL: VHOST_SET_VRING_KICK\n", ctx.fh);
289 VHOST_IOCTL_R(struct vhost_vring_file, file,
290 ops->set_vring_kick);
293 case VHOST_SET_VRING_CALL:
294 LOG_DEBUG(VHOST_CONFIG,
295 "(%"PRIu64") IOCTL: VHOST_SET_VRING_CALL\n", ctx.fh);
296 VHOST_IOCTL_R(struct vhost_vring_file, file,
297 ops->set_vring_call);
301 RTE_LOG(ERR, VHOST_CONFIG,
302 "(%"PRIu64") IOCTL: DOESN NOT EXIST\n", ctx.fh);
304 fuse_reply_ioctl(req, result, NULL, 0);
308 LOG_DEBUG(VHOST_CONFIG,
309 "(%"PRIu64") IOCTL: FAIL\n", ctx.fh);
311 LOG_DEBUG(VHOST_CONFIG,
312 "(%"PRIu64") IOCTL: SUCCESS\n", ctx.fh);
316 * Structure handling open, release and ioctl function pointers is populated.
318 static const struct cuse_lowlevel_ops vhost_net_ops = {
319 .open = vhost_net_open,
320 .release = vhost_net_release,
321 .ioctl = vhost_net_ioctl,
325 * cuse_info is populated and used to register the cuse device.
326 * vhost_net_device_ops are also passed when the device is registered in app.
329 rte_vhost_driver_register(const char *dev_name)
331 struct cuse_info cuse_info;
332 char device_name[PATH_MAX] = "";
333 char char_device_name[PATH_MAX] = "";
334 const char *device_argv[] = { device_name };
336 char fuse_opt_dummy[] = FUSE_OPT_DUMMY;
337 char fuse_opt_fore[] = FUSE_OPT_FORE;
338 char fuse_opt_nomulti[] = FUSE_OPT_NOMULTI;
339 char *fuse_argv[] = {fuse_opt_dummy, fuse_opt_fore, fuse_opt_nomulti};
341 if (access(cuse_device_name, R_OK | W_OK) < 0) {
342 RTE_LOG(ERR, VHOST_CONFIG,
343 "char device %s can't be accessed, maybe not exist\n",
349 * The device name is created. This is passed to QEMU so that it can
350 * register the device with our application.
352 snprintf(device_name, PATH_MAX, "DEVNAME=%s", dev_name);
353 snprintf(char_device_name, PATH_MAX, "/dev/%s", dev_name);
355 /* Check if device already exists. */
356 if (access(char_device_name, F_OK) != -1) {
357 RTE_LOG(ERR, VHOST_CONFIG,
358 "char device %s already exists\n", char_device_name);
362 memset(&cuse_info, 0, sizeof(cuse_info));
363 cuse_info.dev_major = default_major;
364 cuse_info.dev_minor = default_minor;
365 cuse_info.dev_info_argc = 1;
366 cuse_info.dev_info_argv = device_argv;
367 cuse_info.flags = CUSE_UNRESTRICTED_IOCTL;
369 ops = get_virtio_net_callbacks();
371 session = cuse_lowlevel_setup(3, fuse_argv,
372 &cuse_info, &vhost_net_ops, 0, NULL);
380 * The CUSE session is launched allowing the application to receive open,
381 * release and ioctl calls.
384 rte_vhost_driver_session_start(void)
386 fuse_session_loop(session);