Doug / smarc-fsl-linux-kernel

Download as
Browse Code ยป

Commit c53cff5e42a06b81495983bd01741b9a954f11f0

Authored by Basil Gor
Committed by David S. Miller
1 parent 13a8e0c8cd
Exists in smarc-l5.0.0_1.0.0-ga and in 5 other branches imx_3.10.17_1.0.1_ga, imx_3.10.53_1.1.0_ga, imx_3.14.28_1.0.0_ga, smarc-imx_3.10.53_1.1.0_ga, smarc-imx_3.14.28_1.0.0_ga

vhost-net: fix handle_rx buffer size 

Take vlan header length into account, when vlan id is stored as
vlan_tci. Otherwise tagged packets coming from macvtap will be
truncated.

Signed-off-by: Basil Gor <basil.gor@gmail.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Showing 1 changed file with 6 additions and 1 deletions Inline Diff

1 /* Copyright (C) 2009 Red Hat, Inc. 1 /* Copyright (C) 2009 Red Hat, Inc.
2 * Author: Michael S. Tsirkin <mst@redhat.com> 2 * Author: Michael S. Tsirkin <mst@redhat.com>
3 * 3 *
4 * This work is licensed under the terms of the GNU GPL, version 2. 4 * This work is licensed under the terms of the GNU GPL, version 2.
5 * 5 *
6 * virtio-net server in host kernel. 6 * virtio-net server in host kernel.
7 */ 7 */
8 8
9 #include <linux/compat.h> 9 #include <linux/compat.h>
10 #include <linux/eventfd.h> 10 #include <linux/eventfd.h>
11 #include <linux/vhost.h> 11 #include <linux/vhost.h>
12 #include <linux/virtio_net.h> 12 #include <linux/virtio_net.h>
13 #include <linux/miscdevice.h> 13 #include <linux/miscdevice.h>
14 #include <linux/module.h> 14 #include <linux/module.h>
15 #include <linux/moduleparam.h> 15 #include <linux/moduleparam.h>
16 #include <linux/mutex.h> 16 #include <linux/mutex.h>
17 #include <linux/workqueue.h> 17 #include <linux/workqueue.h>
18 #include <linux/rcupdate.h> 18 #include <linux/rcupdate.h>
19 #include <linux/file.h> 19 #include <linux/file.h>
20 #include <linux/slab.h> 20 #include <linux/slab.h>
21 21
22 #include <linux/net.h> 22 #include <linux/net.h>
23 #include <linux/if_packet.h> 23 #include <linux/if_packet.h>
24 #include <linux/if_arp.h> 24 #include <linux/if_arp.h>
25 #include <linux/if_tun.h> 25 #include <linux/if_tun.h>
26 #include <linux/if_macvlan.h> 26 #include <linux/if_macvlan.h>
27 #include <linux/if_vlan.h>
27 28
28 #include <net/sock.h> 29 #include <net/sock.h>
29 30
30 #include "vhost.h" 31 #include "vhost.h"
31 32
32 static int experimental_zcopytx; 33 static int experimental_zcopytx;
33 module_param(experimental_zcopytx, int, 0444); 34 module_param(experimental_zcopytx, int, 0444);
34 MODULE_PARM_DESC(experimental_zcopytx, "Enable Experimental Zero Copy TX"); 35 MODULE_PARM_DESC(experimental_zcopytx, "Enable Experimental Zero Copy TX");
35 36
36 /* Max number of bytes transferred before requeueing the job. 37 /* Max number of bytes transferred before requeueing the job.
37 * Using this limit prevents one virtqueue from starving others. */ 38 * Using this limit prevents one virtqueue from starving others. */
38 #define VHOST_NET_WEIGHT 0x80000 39 #define VHOST_NET_WEIGHT 0x80000
39 40
40 /* MAX number of TX used buffers for outstanding zerocopy */ 41 /* MAX number of TX used buffers for outstanding zerocopy */
41 #define VHOST_MAX_PEND 128 42 #define VHOST_MAX_PEND 128
42 #define VHOST_GOODCOPY_LEN 256 43 #define VHOST_GOODCOPY_LEN 256
43 44
44 enum { 45 enum {
45 VHOST_NET_VQ_RX = 0, 46 VHOST_NET_VQ_RX = 0,
46 VHOST_NET_VQ_TX = 1, 47 VHOST_NET_VQ_TX = 1,
47 VHOST_NET_VQ_MAX = 2, 48 VHOST_NET_VQ_MAX = 2,
48 }; 49 };
49 50
50 enum vhost_net_poll_state { 51 enum vhost_net_poll_state {
51 VHOST_NET_POLL_DISABLED = 0, 52 VHOST_NET_POLL_DISABLED = 0,
52 VHOST_NET_POLL_STARTED = 1, 53 VHOST_NET_POLL_STARTED = 1,
53 VHOST_NET_POLL_STOPPED = 2, 54 VHOST_NET_POLL_STOPPED = 2,
54 }; 55 };
55 56
56 struct vhost_net { 57 struct vhost_net {
57 struct vhost_dev dev; 58 struct vhost_dev dev;
58 struct vhost_virtqueue vqs[VHOST_NET_VQ_MAX]; 59 struct vhost_virtqueue vqs[VHOST_NET_VQ_MAX];
59 struct vhost_poll poll[VHOST_NET_VQ_MAX]; 60 struct vhost_poll poll[VHOST_NET_VQ_MAX];
60 /* Tells us whether we are polling a socket for TX. 61 /* Tells us whether we are polling a socket for TX.
61 * We only do this when socket buffer fills up. 62 * We only do this when socket buffer fills up.
62 * Protected by tx vq lock. */ 63 * Protected by tx vq lock. */
63 enum vhost_net_poll_state tx_poll_state; 64 enum vhost_net_poll_state tx_poll_state;
64 }; 65 };
65 66
66 static bool vhost_sock_zcopy(struct socket *sock) 67 static bool vhost_sock_zcopy(struct socket *sock)
67 { 68 {
68 return unlikely(experimental_zcopytx) && 69 return unlikely(experimental_zcopytx) &&
69 sock_flag(sock->sk, SOCK_ZEROCOPY); 70 sock_flag(sock->sk, SOCK_ZEROCOPY);
70 } 71 }
71 72
72 /* Pop first len bytes from iovec. Return number of segments used. */ 73 /* Pop first len bytes from iovec. Return number of segments used. */
73 static int move_iovec_hdr(struct iovec *from, struct iovec *to, 74 static int move_iovec_hdr(struct iovec *from, struct iovec *to,
74 size_t len, int iov_count) 75 size_t len, int iov_count)
75 { 76 {
76 int seg = 0; 77 int seg = 0;
77 size_t size; 78 size_t size;
78 79
79 while (len && seg < iov_count) { 80 while (len && seg < iov_count) {
80 size = min(from->iov_len, len); 81 size = min(from->iov_len, len);
81 to->iov_base = from->iov_base; 82 to->iov_base = from->iov_base;
82 to->iov_len = size; 83 to->iov_len = size;
83 from->iov_len -= size; 84 from->iov_len -= size;
84 from->iov_base += size; 85 from->iov_base += size;
85 len -= size; 86 len -= size;
86 ++from; 87 ++from;
87 ++to; 88 ++to;
88 ++seg; 89 ++seg;
89 } 90 }
90 return seg; 91 return seg;
91 } 92 }
92 /* Copy iovec entries for len bytes from iovec. */ 93 /* Copy iovec entries for len bytes from iovec. */
93 static void copy_iovec_hdr(const struct iovec *from, struct iovec *to, 94 static void copy_iovec_hdr(const struct iovec *from, struct iovec *to,
94 size_t len, int iovcount) 95 size_t len, int iovcount)
95 { 96 {
96 int seg = 0; 97 int seg = 0;
97 size_t size; 98 size_t size;
98 99
99 while (len && seg < iovcount) { 100 while (len && seg < iovcount) {
100 size = min(from->iov_len, len); 101 size = min(from->iov_len, len);
101 to->iov_base = from->iov_base; 102 to->iov_base = from->iov_base;
102 to->iov_len = size; 103 to->iov_len = size;
103 len -= size; 104 len -= size;
104 ++from; 105 ++from;
105 ++to; 106 ++to;
106 ++seg; 107 ++seg;
107 } 108 }
108 } 109 }
109 110
110 /* Caller must have TX VQ lock */ 111 /* Caller must have TX VQ lock */
111 static void tx_poll_stop(struct vhost_net *net) 112 static void tx_poll_stop(struct vhost_net *net)
112 { 113 {
113 if (likely(net->tx_poll_state != VHOST_NET_POLL_STARTED)) 114 if (likely(net->tx_poll_state != VHOST_NET_POLL_STARTED))
114 return; 115 return;
115 vhost_poll_stop(net->poll + VHOST_NET_VQ_TX); 116 vhost_poll_stop(net->poll + VHOST_NET_VQ_TX);
116 net->tx_poll_state = VHOST_NET_POLL_STOPPED; 117 net->tx_poll_state = VHOST_NET_POLL_STOPPED;
117 } 118 }
118 119
119 /* Caller must have TX VQ lock */ 120 /* Caller must have TX VQ lock */
120 static void tx_poll_start(struct vhost_net *net, struct socket *sock) 121 static void tx_poll_start(struct vhost_net *net, struct socket *sock)
121 { 122 {
122 if (unlikely(net->tx_poll_state != VHOST_NET_POLL_STOPPED)) 123 if (unlikely(net->tx_poll_state != VHOST_NET_POLL_STOPPED))
123 return; 124 return;
124 vhost_poll_start(net->poll + VHOST_NET_VQ_TX, sock->file); 125 vhost_poll_start(net->poll + VHOST_NET_VQ_TX, sock->file);
125 net->tx_poll_state = VHOST_NET_POLL_STARTED; 126 net->tx_poll_state = VHOST_NET_POLL_STARTED;
126 } 127 }
127 128
128 /* Expects to be always run from workqueue - which acts as 129 /* Expects to be always run from workqueue - which acts as
129 * read-size critical section for our kind of RCU. */ 130 * read-size critical section for our kind of RCU. */
130 static void handle_tx(struct vhost_net *net) 131 static void handle_tx(struct vhost_net *net)
131 { 132 {
132 struct vhost_virtqueue *vq = &net->dev.vqs[VHOST_NET_VQ_TX]; 133 struct vhost_virtqueue *vq = &net->dev.vqs[VHOST_NET_VQ_TX];
133 unsigned out, in, s; 134 unsigned out, in, s;
134 int head; 135 int head;
135 struct msghdr msg = { 136 struct msghdr msg = {
136 .msg_name = NULL, 137 .msg_name = NULL,
137 .msg_namelen = 0, 138 .msg_namelen = 0,
138 .msg_control = NULL, 139 .msg_control = NULL,
139 .msg_controllen = 0, 140 .msg_controllen = 0,
140 .msg_iov = vq->iov, 141 .msg_iov = vq->iov,
141 .msg_flags = MSG_DONTWAIT, 142 .msg_flags = MSG_DONTWAIT,
142 }; 143 };
143 size_t len, total_len = 0; 144 size_t len, total_len = 0;
144 int err, wmem; 145 int err, wmem;
145 size_t hdr_size; 146 size_t hdr_size;
146 struct socket *sock; 147 struct socket *sock;
147 struct vhost_ubuf_ref *uninitialized_var(ubufs); 148 struct vhost_ubuf_ref *uninitialized_var(ubufs);
148 bool zcopy; 149 bool zcopy;
149 150
150 /* TODO: check that we are running from vhost_worker? */ 151 /* TODO: check that we are running from vhost_worker? */
151 sock = rcu_dereference_check(vq->private_data, 1); 152 sock = rcu_dereference_check(vq->private_data, 1);
152 if (!sock) 153 if (!sock)
153 return; 154 return;
154 155
155 wmem = atomic_read(&sock->sk->sk_wmem_alloc); 156 wmem = atomic_read(&sock->sk->sk_wmem_alloc);
156 if (wmem >= sock->sk->sk_sndbuf) { 157 if (wmem >= sock->sk->sk_sndbuf) {
157 mutex_lock(&vq->mutex); 158 mutex_lock(&vq->mutex);
158 tx_poll_start(net, sock); 159 tx_poll_start(net, sock);
159 mutex_unlock(&vq->mutex); 160 mutex_unlock(&vq->mutex);
160 return; 161 return;
161 } 162 }
162 163
163 mutex_lock(&vq->mutex); 164 mutex_lock(&vq->mutex);
164 vhost_disable_notify(&net->dev, vq); 165 vhost_disable_notify(&net->dev, vq);
165 166
166 if (wmem < sock->sk->sk_sndbuf / 2) 167 if (wmem < sock->sk->sk_sndbuf / 2)
167 tx_poll_stop(net); 168 tx_poll_stop(net);
168 hdr_size = vq->vhost_hlen; 169 hdr_size = vq->vhost_hlen;
169 zcopy = vhost_sock_zcopy(sock); 170 zcopy = vhost_sock_zcopy(sock);
170 171
171 for (;;) { 172 for (;;) {
172 /* Release DMAs done buffers first */ 173 /* Release DMAs done buffers first */
173 if (zcopy) 174 if (zcopy)
174 vhost_zerocopy_signal_used(vq); 175 vhost_zerocopy_signal_used(vq);
175 176
176 head = vhost_get_vq_desc(&net->dev, vq, vq->iov, 177 head = vhost_get_vq_desc(&net->dev, vq, vq->iov,
177 ARRAY_SIZE(vq->iov), 178 ARRAY_SIZE(vq->iov),
178 &out, &in, 179 &out, &in,
179 NULL, NULL); 180 NULL, NULL);
180 /* On error, stop handling until the next kick. */ 181 /* On error, stop handling until the next kick. */
181 if (unlikely(head < 0)) 182 if (unlikely(head < 0))
182 break; 183 break;
183 /* Nothing new? Wait for eventfd to tell us they refilled. */ 184 /* Nothing new? Wait for eventfd to tell us they refilled. */
184 if (head == vq->num) { 185 if (head == vq->num) {
185 int num_pends; 186 int num_pends;
186 187
187 wmem = atomic_read(&sock->sk->sk_wmem_alloc); 188 wmem = atomic_read(&sock->sk->sk_wmem_alloc);
188 if (wmem >= sock->sk->sk_sndbuf * 3 / 4) { 189 if (wmem >= sock->sk->sk_sndbuf * 3 / 4) {
189 tx_poll_start(net, sock); 190 tx_poll_start(net, sock);
190 set_bit(SOCK_ASYNC_NOSPACE, &sock->flags); 191 set_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
191 break; 192 break;
192 } 193 }
193 /* If more outstanding DMAs, queue the work. 194 /* If more outstanding DMAs, queue the work.
194 * Handle upend_idx wrap around 195 * Handle upend_idx wrap around
195 */ 196 */
196 num_pends = likely(vq->upend_idx >= vq->done_idx) ? 197 num_pends = likely(vq->upend_idx >= vq->done_idx) ?
197 (vq->upend_idx - vq->done_idx) : 198 (vq->upend_idx - vq->done_idx) :
198 (vq->upend_idx + UIO_MAXIOV - vq->done_idx); 199 (vq->upend_idx + UIO_MAXIOV - vq->done_idx);
199 if (unlikely(num_pends > VHOST_MAX_PEND)) { 200 if (unlikely(num_pends > VHOST_MAX_PEND)) {
200 tx_poll_start(net, sock); 201 tx_poll_start(net, sock);
201 set_bit(SOCK_ASYNC_NOSPACE, &sock->flags); 202 set_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
202 break; 203 break;
203 } 204 }
204 if (unlikely(vhost_enable_notify(&net->dev, vq))) { 205 if (unlikely(vhost_enable_notify(&net->dev, vq))) {
205 vhost_disable_notify(&net->dev, vq); 206 vhost_disable_notify(&net->dev, vq);
206 continue; 207 continue;
207 } 208 }
208 break; 209 break;
209 } 210 }
210 if (in) { 211 if (in) {
211 vq_err(vq, "Unexpected descriptor format for TX: " 212 vq_err(vq, "Unexpected descriptor format for TX: "
212 "out %d, int %d\n", out, in); 213 "out %d, int %d\n", out, in);
213 break; 214 break;
214 } 215 }
215 /* Skip header. TODO: support TSO. */ 216 /* Skip header. TODO: support TSO. */
216 s = move_iovec_hdr(vq->iov, vq->hdr, hdr_size, out); 217 s = move_iovec_hdr(vq->iov, vq->hdr, hdr_size, out);
217 msg.msg_iovlen = out; 218 msg.msg_iovlen = out;
218 len = iov_length(vq->iov, out); 219 len = iov_length(vq->iov, out);
219 /* Sanity check */ 220 /* Sanity check */
220 if (!len) { 221 if (!len) {
221 vq_err(vq, "Unexpected header len for TX: " 222 vq_err(vq, "Unexpected header len for TX: "
222 "%zd expected %zd\n", 223 "%zd expected %zd\n",
223 iov_length(vq->hdr, s), hdr_size); 224 iov_length(vq->hdr, s), hdr_size);
224 break; 225 break;
225 } 226 }
226 /* use msg_control to pass vhost zerocopy ubuf info to skb */ 227 /* use msg_control to pass vhost zerocopy ubuf info to skb */
227 if (zcopy) { 228 if (zcopy) {
228 vq->heads[vq->upend_idx].id = head; 229 vq->heads[vq->upend_idx].id = head;
229 if (len < VHOST_GOODCOPY_LEN) { 230 if (len < VHOST_GOODCOPY_LEN) {
230 /* copy don't need to wait for DMA done */ 231 /* copy don't need to wait for DMA done */
231 vq->heads[vq->upend_idx].len = 232 vq->heads[vq->upend_idx].len =
232 VHOST_DMA_DONE_LEN; 233 VHOST_DMA_DONE_LEN;
233 msg.msg_control = NULL; 234 msg.msg_control = NULL;
234 msg.msg_controllen = 0; 235 msg.msg_controllen = 0;
235 ubufs = NULL; 236 ubufs = NULL;
236 } else { 237 } else {
237 struct ubuf_info *ubuf = &vq->ubuf_info[head]; 238 struct ubuf_info *ubuf = &vq->ubuf_info[head];
238 239
239 vq->heads[vq->upend_idx].len = len; 240 vq->heads[vq->upend_idx].len = len;
240 ubuf->callback = vhost_zerocopy_callback; 241 ubuf->callback = vhost_zerocopy_callback;
241 ubuf->ctx = vq->ubufs; 242 ubuf->ctx = vq->ubufs;
242 ubuf->desc = vq->upend_idx; 243 ubuf->desc = vq->upend_idx;
243 msg.msg_control = ubuf; 244 msg.msg_control = ubuf;
244 msg.msg_controllen = sizeof(ubuf); 245 msg.msg_controllen = sizeof(ubuf);
245 ubufs = vq->ubufs; 246 ubufs = vq->ubufs;
246 kref_get(&ubufs->kref); 247 kref_get(&ubufs->kref);
247 } 248 }
248 vq->upend_idx = (vq->upend_idx + 1) % UIO_MAXIOV; 249 vq->upend_idx = (vq->upend_idx + 1) % UIO_MAXIOV;
249 } 250 }
250 /* TODO: Check specific error and bomb out unless ENOBUFS? */ 251 /* TODO: Check specific error and bomb out unless ENOBUFS? */
251 err = sock->ops->sendmsg(NULL, sock, &msg, len); 252 err = sock->ops->sendmsg(NULL, sock, &msg, len);
252 if (unlikely(err < 0)) { 253 if (unlikely(err < 0)) {
253 if (zcopy) { 254 if (zcopy) {
254 if (ubufs) 255 if (ubufs)
255 vhost_ubuf_put(ubufs); 256 vhost_ubuf_put(ubufs);
256 vq->upend_idx = ((unsigned)vq->upend_idx - 1) % 257 vq->upend_idx = ((unsigned)vq->upend_idx - 1) %
257 UIO_MAXIOV; 258 UIO_MAXIOV;
258 } 259 }
259 vhost_discard_vq_desc(vq, 1); 260 vhost_discard_vq_desc(vq, 1);
260 tx_poll_start(net, sock); 261 tx_poll_start(net, sock);
261 break; 262 break;
262 } 263 }
263 if (err != len) 264 if (err != len)
264 pr_debug("Truncated TX packet: " 265 pr_debug("Truncated TX packet: "
265 " len %d != %zd\n", err, len); 266 " len %d != %zd\n", err, len);
266 if (!zcopy) 267 if (!zcopy)
267 vhost_add_used_and_signal(&net->dev, vq, head, 0); 268 vhost_add_used_and_signal(&net->dev, vq, head, 0);
268 total_len += len; 269 total_len += len;
269 if (unlikely(total_len >= VHOST_NET_WEIGHT)) { 270 if (unlikely(total_len >= VHOST_NET_WEIGHT)) {
270 vhost_poll_queue(&vq->poll); 271 vhost_poll_queue(&vq->poll);
271 break; 272 break;
272 } 273 }
273 } 274 }
274 275
275 mutex_unlock(&vq->mutex); 276 mutex_unlock(&vq->mutex);
276 } 277 }
277 278
278 static int peek_head_len(struct sock *sk) 279 static int peek_head_len(struct sock *sk)
279 { 280 {
280 struct sk_buff *head; 281 struct sk_buff *head;
281 int len = 0; 282 int len = 0;
282 unsigned long flags; 283 unsigned long flags;
283 284
284 spin_lock_irqsave(&sk->sk_receive_queue.lock, flags); 285 spin_lock_irqsave(&sk->sk_receive_queue.lock, flags);
285 head = skb_peek(&sk->sk_receive_queue); 286 head = skb_peek(&sk->sk_receive_queue);
286 if (likely(head)) 287 if (likely(head)) {
287 len = head->len; 288 len = head->len;
289 if (vlan_tx_tag_present(head))
290 len += VLAN_HLEN;
291 }
292
288 spin_unlock_irqrestore(&sk->sk_receive_queue.lock, flags); 293 spin_unlock_irqrestore(&sk->sk_receive_queue.lock, flags);
289 return len; 294 return len;
290 } 295 }
291 296
292 /* This is a multi-buffer version of vhost_get_desc, that works if 297 /* This is a multi-buffer version of vhost_get_desc, that works if
293 * vq has read descriptors only. 298 * vq has read descriptors only.
294 * @vq - the relevant virtqueue 299 * @vq - the relevant virtqueue
295 * @datalen - data length we'll be reading 300 * @datalen - data length we'll be reading
296 * @iovcount - returned count of io vectors we fill 301 * @iovcount - returned count of io vectors we fill
297 * @log - vhost log 302 * @log - vhost log
298 * @log_num - log offset 303 * @log_num - log offset
299 * @quota - headcount quota, 1 for big buffer 304 * @quota - headcount quota, 1 for big buffer
300 * returns number of buffer heads allocated, negative on error 305 * returns number of buffer heads allocated, negative on error
301 */ 306 */
302 static int get_rx_bufs(struct vhost_virtqueue *vq, 307 static int get_rx_bufs(struct vhost_virtqueue *vq,
303 struct vring_used_elem *heads, 308 struct vring_used_elem *heads,
304 int datalen, 309 int datalen,
305 unsigned *iovcount, 310 unsigned *iovcount,
306 struct vhost_log *log, 311 struct vhost_log *log,
307 unsigned *log_num, 312 unsigned *log_num,
308 unsigned int quota) 313 unsigned int quota)
309 { 314 {
310 unsigned int out, in; 315 unsigned int out, in;
311 int seg = 0; 316 int seg = 0;
312 int headcount = 0; 317 int headcount = 0;
313 unsigned d; 318 unsigned d;
314 int r, nlogs = 0; 319 int r, nlogs = 0;
315 320
316 while (datalen > 0 && headcount < quota) { 321 while (datalen > 0 && headcount < quota) {
317 if (unlikely(seg >= UIO_MAXIOV)) { 322 if (unlikely(seg >= UIO_MAXIOV)) {
318 r = -ENOBUFS; 323 r = -ENOBUFS;
319 goto err; 324 goto err;
320 } 325 }
321 d = vhost_get_vq_desc(vq->dev, vq, vq->iov + seg, 326 d = vhost_get_vq_desc(vq->dev, vq, vq->iov + seg,
322 ARRAY_SIZE(vq->iov) - seg, &out, 327 ARRAY_SIZE(vq->iov) - seg, &out,
323 &in, log, log_num); 328 &in, log, log_num);
324 if (d == vq->num) { 329 if (d == vq->num) {
325 r = 0; 330 r = 0;
326 goto err; 331 goto err;
327 } 332 }
328 if (unlikely(out || in <= 0)) { 333 if (unlikely(out || in <= 0)) {
329 vq_err(vq, "unexpected descriptor format for RX: " 334 vq_err(vq, "unexpected descriptor format for RX: "
330 "out %d, in %d\n", out, in); 335 "out %d, in %d\n", out, in);
331 r = -EINVAL; 336 r = -EINVAL;
332 goto err; 337 goto err;
333 } 338 }
334 if (unlikely(log)) { 339 if (unlikely(log)) {
335 nlogs += *log_num; 340 nlogs += *log_num;
336 log += *log_num; 341 log += *log_num;
337 } 342 }
338 heads[headcount].id = d; 343 heads[headcount].id = d;
339 heads[headcount].len = iov_length(vq->iov + seg, in); 344 heads[headcount].len = iov_length(vq->iov + seg, in);
340 datalen -= heads[headcount].len; 345 datalen -= heads[headcount].len;
341 ++headcount; 346 ++headcount;
342 seg += in; 347 seg += in;
343 } 348 }
344 heads[headcount - 1].len += datalen; 349 heads[headcount - 1].len += datalen;
345 *iovcount = seg; 350 *iovcount = seg;
346 if (unlikely(log)) 351 if (unlikely(log))
347 *log_num = nlogs; 352 *log_num = nlogs;
348 return headcount; 353 return headcount;
349 err: 354 err:
350 vhost_discard_vq_desc(vq, headcount); 355 vhost_discard_vq_desc(vq, headcount);
351 return r; 356 return r;
352 } 357 }
353 358
354 /* Expects to be always run from workqueue - which acts as 359 /* Expects to be always run from workqueue - which acts as
355 * read-size critical section for our kind of RCU. */ 360 * read-size critical section for our kind of RCU. */
356 static void handle_rx(struct vhost_net *net) 361 static void handle_rx(struct vhost_net *net)
357 { 362 {
358 struct vhost_virtqueue *vq = &net->dev.vqs[VHOST_NET_VQ_RX]; 363 struct vhost_virtqueue *vq = &net->dev.vqs[VHOST_NET_VQ_RX];
359 unsigned uninitialized_var(in), log; 364 unsigned uninitialized_var(in), log;
360 struct vhost_log *vq_log; 365 struct vhost_log *vq_log;
361 struct msghdr msg = { 366 struct msghdr msg = {
362 .msg_name = NULL, 367 .msg_name = NULL,
363 .msg_namelen = 0, 368 .msg_namelen = 0,
364 .msg_control = NULL, /* FIXME: get and handle RX aux data. */ 369 .msg_control = NULL, /* FIXME: get and handle RX aux data. */
365 .msg_controllen = 0, 370 .msg_controllen = 0,
366 .msg_iov = vq->iov, 371 .msg_iov = vq->iov,
367 .msg_flags = MSG_DONTWAIT, 372 .msg_flags = MSG_DONTWAIT,
368 }; 373 };
369 struct virtio_net_hdr_mrg_rxbuf hdr = { 374 struct virtio_net_hdr_mrg_rxbuf hdr = {
370 .hdr.flags = 0, 375 .hdr.flags = 0,
371 .hdr.gso_type = VIRTIO_NET_HDR_GSO_NONE 376 .hdr.gso_type = VIRTIO_NET_HDR_GSO_NONE
372 }; 377 };
373 size_t total_len = 0; 378 size_t total_len = 0;
374 int err, headcount, mergeable; 379 int err, headcount, mergeable;
375 size_t vhost_hlen, sock_hlen; 380 size_t vhost_hlen, sock_hlen;
376 size_t vhost_len, sock_len; 381 size_t vhost_len, sock_len;
377 /* TODO: check that we are running from vhost_worker? */ 382 /* TODO: check that we are running from vhost_worker? */
378 struct socket *sock = rcu_dereference_check(vq->private_data, 1); 383 struct socket *sock = rcu_dereference_check(vq->private_data, 1);
379 384
380 if (!sock) 385 if (!sock)
381 return; 386 return;
382 387
383 mutex_lock(&vq->mutex); 388 mutex_lock(&vq->mutex);
384 vhost_disable_notify(&net->dev, vq); 389 vhost_disable_notify(&net->dev, vq);
385 vhost_hlen = vq->vhost_hlen; 390 vhost_hlen = vq->vhost_hlen;
386 sock_hlen = vq->sock_hlen; 391 sock_hlen = vq->sock_hlen;
387 392
388 vq_log = unlikely(vhost_has_feature(&net->dev, VHOST_F_LOG_ALL)) ? 393 vq_log = unlikely(vhost_has_feature(&net->dev, VHOST_F_LOG_ALL)) ?
389 vq->log : NULL; 394 vq->log : NULL;
390 mergeable = vhost_has_feature(&net->dev, VIRTIO_NET_F_MRG_RXBUF); 395 mergeable = vhost_has_feature(&net->dev, VIRTIO_NET_F_MRG_RXBUF);
391 396
392 while ((sock_len = peek_head_len(sock->sk))) { 397 while ((sock_len = peek_head_len(sock->sk))) {
393 sock_len += sock_hlen; 398 sock_len += sock_hlen;
394 vhost_len = sock_len + vhost_hlen; 399 vhost_len = sock_len + vhost_hlen;
395 headcount = get_rx_bufs(vq, vq->heads, vhost_len, 400 headcount = get_rx_bufs(vq, vq->heads, vhost_len,
396 &in, vq_log, &log, 401 &in, vq_log, &log,
397 likely(mergeable) ? UIO_MAXIOV : 1); 402 likely(mergeable) ? UIO_MAXIOV : 1);
398 /* On error, stop handling until the next kick. */ 403 /* On error, stop handling until the next kick. */
399 if (unlikely(headcount < 0)) 404 if (unlikely(headcount < 0))
400 break; 405 break;
401 /* OK, now we need to know about added descriptors. */ 406 /* OK, now we need to know about added descriptors. */
402 if (!headcount) { 407 if (!headcount) {
403 if (unlikely(vhost_enable_notify(&net->dev, vq))) { 408 if (unlikely(vhost_enable_notify(&net->dev, vq))) {
404 /* They have slipped one in as we were 409 /* They have slipped one in as we were
405 * doing that: check again. */ 410 * doing that: check again. */
406 vhost_disable_notify(&net->dev, vq); 411 vhost_disable_notify(&net->dev, vq);
407 continue; 412 continue;
408 } 413 }
409 /* Nothing new? Wait for eventfd to tell us 414 /* Nothing new? Wait for eventfd to tell us
410 * they refilled. */ 415 * they refilled. */
411 break; 416 break;
412 } 417 }
413 /* We don't need to be notified again. */ 418 /* We don't need to be notified again. */
414 if (unlikely((vhost_hlen))) 419 if (unlikely((vhost_hlen)))
415 /* Skip header. TODO: support TSO. */ 420 /* Skip header. TODO: support TSO. */
416 move_iovec_hdr(vq->iov, vq->hdr, vhost_hlen, in); 421 move_iovec_hdr(vq->iov, vq->hdr, vhost_hlen, in);
417 else 422 else
418 /* Copy the header for use in VIRTIO_NET_F_MRG_RXBUF: 423 /* Copy the header for use in VIRTIO_NET_F_MRG_RXBUF:
419 * needed because recvmsg can modify msg_iov. */ 424 * needed because recvmsg can modify msg_iov. */
420 copy_iovec_hdr(vq->iov, vq->hdr, sock_hlen, in); 425 copy_iovec_hdr(vq->iov, vq->hdr, sock_hlen, in);
421 msg.msg_iovlen = in; 426 msg.msg_iovlen = in;
422 err = sock->ops->recvmsg(NULL, sock, &msg, 427 err = sock->ops->recvmsg(NULL, sock, &msg,
423 sock_len, MSG_DONTWAIT | MSG_TRUNC); 428 sock_len, MSG_DONTWAIT | MSG_TRUNC);
424 /* Userspace might have consumed the packet meanwhile: 429 /* Userspace might have consumed the packet meanwhile:
425 * it's not supposed to do this usually, but might be hard 430 * it's not supposed to do this usually, but might be hard
426 * to prevent. Discard data we got (if any) and keep going. */ 431 * to prevent. Discard data we got (if any) and keep going. */
427 if (unlikely(err != sock_len)) { 432 if (unlikely(err != sock_len)) {
428 pr_debug("Discarded rx packet: " 433 pr_debug("Discarded rx packet: "
429 " len %d, expected %zd\n", err, sock_len); 434 " len %d, expected %zd\n", err, sock_len);
430 vhost_discard_vq_desc(vq, headcount); 435 vhost_discard_vq_desc(vq, headcount);
431 continue; 436 continue;
432 } 437 }
433 if (unlikely(vhost_hlen) && 438 if (unlikely(vhost_hlen) &&
434 memcpy_toiovecend(vq->hdr, (unsigned char *)&hdr, 0, 439 memcpy_toiovecend(vq->hdr, (unsigned char *)&hdr, 0,
435 vhost_hlen)) { 440 vhost_hlen)) {
436 vq_err(vq, "Unable to write vnet_hdr at addr %p\n", 441 vq_err(vq, "Unable to write vnet_hdr at addr %p\n",
437 vq->iov->iov_base); 442 vq->iov->iov_base);
438 break; 443 break;
439 } 444 }
440 /* TODO: Should check and handle checksum. */ 445 /* TODO: Should check and handle checksum. */
441 if (likely(mergeable) && 446 if (likely(mergeable) &&
442 memcpy_toiovecend(vq->hdr, (unsigned char *)&headcount, 447 memcpy_toiovecend(vq->hdr, (unsigned char *)&headcount,
443 offsetof(typeof(hdr), num_buffers), 448 offsetof(typeof(hdr), num_buffers),
444 sizeof hdr.num_buffers)) { 449 sizeof hdr.num_buffers)) {
445 vq_err(vq, "Failed num_buffers write"); 450 vq_err(vq, "Failed num_buffers write");
446 vhost_discard_vq_desc(vq, headcount); 451 vhost_discard_vq_desc(vq, headcount);
447 break; 452 break;
448 } 453 }
449 vhost_add_used_and_signal_n(&net->dev, vq, vq->heads, 454 vhost_add_used_and_signal_n(&net->dev, vq, vq->heads,
450 headcount); 455 headcount);
451 if (unlikely(vq_log)) 456 if (unlikely(vq_log))
452 vhost_log_write(vq, vq_log, log, vhost_len); 457 vhost_log_write(vq, vq_log, log, vhost_len);
453 total_len += vhost_len; 458 total_len += vhost_len;
454 if (unlikely(total_len >= VHOST_NET_WEIGHT)) { 459 if (unlikely(total_len >= VHOST_NET_WEIGHT)) {
455 vhost_poll_queue(&vq->poll); 460 vhost_poll_queue(&vq->poll);
456 break; 461 break;
457 } 462 }
458 } 463 }
459 464
460 mutex_unlock(&vq->mutex); 465 mutex_unlock(&vq->mutex);
461 } 466 }
462 467
463 static void handle_tx_kick(struct vhost_work *work) 468 static void handle_tx_kick(struct vhost_work *work)
464 { 469 {
465 struct vhost_virtqueue *vq = container_of(work, struct vhost_virtqueue, 470 struct vhost_virtqueue *vq = container_of(work, struct vhost_virtqueue,
466 poll.work); 471 poll.work);
467 struct vhost_net *net = container_of(vq->dev, struct vhost_net, dev); 472 struct vhost_net *net = container_of(vq->dev, struct vhost_net, dev);
468 473
469 handle_tx(net); 474 handle_tx(net);
470 } 475 }
471 476
472 static void handle_rx_kick(struct vhost_work *work) 477 static void handle_rx_kick(struct vhost_work *work)
473 { 478 {
474 struct vhost_virtqueue *vq = container_of(work, struct vhost_virtqueue, 479 struct vhost_virtqueue *vq = container_of(work, struct vhost_virtqueue,
475 poll.work); 480 poll.work);
476 struct vhost_net *net = container_of(vq->dev, struct vhost_net, dev); 481 struct vhost_net *net = container_of(vq->dev, struct vhost_net, dev);
477 482
478 handle_rx(net); 483 handle_rx(net);
479 } 484 }
480 485
481 static void handle_tx_net(struct vhost_work *work) 486 static void handle_tx_net(struct vhost_work *work)
482 { 487 {
483 struct vhost_net *net = container_of(work, struct vhost_net, 488 struct vhost_net *net = container_of(work, struct vhost_net,
484 poll[VHOST_NET_VQ_TX].work); 489 poll[VHOST_NET_VQ_TX].work);
485 handle_tx(net); 490 handle_tx(net);
486 } 491 }
487 492
488 static void handle_rx_net(struct vhost_work *work) 493 static void handle_rx_net(struct vhost_work *work)
489 { 494 {
490 struct vhost_net *net = container_of(work, struct vhost_net, 495 struct vhost_net *net = container_of(work, struct vhost_net,
491 poll[VHOST_NET_VQ_RX].work); 496 poll[VHOST_NET_VQ_RX].work);
492 handle_rx(net); 497 handle_rx(net);
493 } 498 }
494 499
495 static int vhost_net_open(struct inode *inode, struct file *f) 500 static int vhost_net_open(struct inode *inode, struct file *f)
496 { 501 {
497 struct vhost_net *n = kmalloc(sizeof *n, GFP_KERNEL); 502 struct vhost_net *n = kmalloc(sizeof *n, GFP_KERNEL);
498 struct vhost_dev *dev; 503 struct vhost_dev *dev;
499 int r; 504 int r;
500 505
501 if (!n) 506 if (!n)
502 return -ENOMEM; 507 return -ENOMEM;
503 508
504 dev = &n->dev; 509 dev = &n->dev;
505 n->vqs[VHOST_NET_VQ_TX].handle_kick = handle_tx_kick; 510 n->vqs[VHOST_NET_VQ_TX].handle_kick = handle_tx_kick;
506 n->vqs[VHOST_NET_VQ_RX].handle_kick = handle_rx_kick; 511 n->vqs[VHOST_NET_VQ_RX].handle_kick = handle_rx_kick;
507 r = vhost_dev_init(dev, n->vqs, VHOST_NET_VQ_MAX); 512 r = vhost_dev_init(dev, n->vqs, VHOST_NET_VQ_MAX);
508 if (r < 0) { 513 if (r < 0) {
509 kfree(n); 514 kfree(n);
510 return r; 515 return r;
511 } 516 }
512 517
513 vhost_poll_init(n->poll + VHOST_NET_VQ_TX, handle_tx_net, POLLOUT, dev); 518 vhost_poll_init(n->poll + VHOST_NET_VQ_TX, handle_tx_net, POLLOUT, dev);
514 vhost_poll_init(n->poll + VHOST_NET_VQ_RX, handle_rx_net, POLLIN, dev); 519 vhost_poll_init(n->poll + VHOST_NET_VQ_RX, handle_rx_net, POLLIN, dev);
515 n->tx_poll_state = VHOST_NET_POLL_DISABLED; 520 n->tx_poll_state = VHOST_NET_POLL_DISABLED;
516 521
517 f->private_data = n; 522 f->private_data = n;
518 523
519 return 0; 524 return 0;
520 } 525 }
521 526
522 static void vhost_net_disable_vq(struct vhost_net *n, 527 static void vhost_net_disable_vq(struct vhost_net *n,
523 struct vhost_virtqueue *vq) 528 struct vhost_virtqueue *vq)
524 { 529 {
525 if (!vq->private_data) 530 if (!vq->private_data)
526 return; 531 return;
527 if (vq == n->vqs + VHOST_NET_VQ_TX) { 532 if (vq == n->vqs + VHOST_NET_VQ_TX) {
528 tx_poll_stop(n); 533 tx_poll_stop(n);
529 n->tx_poll_state = VHOST_NET_POLL_DISABLED; 534 n->tx_poll_state = VHOST_NET_POLL_DISABLED;
530 } else 535 } else
531 vhost_poll_stop(n->poll + VHOST_NET_VQ_RX); 536 vhost_poll_stop(n->poll + VHOST_NET_VQ_RX);
532 } 537 }
533 538
534 static void vhost_net_enable_vq(struct vhost_net *n, 539 static void vhost_net_enable_vq(struct vhost_net *n,
535 struct vhost_virtqueue *vq) 540 struct vhost_virtqueue *vq)
536 { 541 {
537 struct socket *sock; 542 struct socket *sock;
538 543
539 sock = rcu_dereference_protected(vq->private_data, 544 sock = rcu_dereference_protected(vq->private_data,
540 lockdep_is_held(&vq->mutex)); 545 lockdep_is_held(&vq->mutex));
541 if (!sock) 546 if (!sock)
542 return; 547 return;
543 if (vq == n->vqs + VHOST_NET_VQ_TX) { 548 if (vq == n->vqs + VHOST_NET_VQ_TX) {
544 n->tx_poll_state = VHOST_NET_POLL_STOPPED; 549 n->tx_poll_state = VHOST_NET_POLL_STOPPED;
545 tx_poll_start(n, sock); 550 tx_poll_start(n, sock);
546 } else 551 } else
547 vhost_poll_start(n->poll + VHOST_NET_VQ_RX, sock->file); 552 vhost_poll_start(n->poll + VHOST_NET_VQ_RX, sock->file);
548 } 553 }
549 554
550 static struct socket *vhost_net_stop_vq(struct vhost_net *n, 555 static struct socket *vhost_net_stop_vq(struct vhost_net *n,
551 struct vhost_virtqueue *vq) 556 struct vhost_virtqueue *vq)
552 { 557 {
553 struct socket *sock; 558 struct socket *sock;
554 559
555 mutex_lock(&vq->mutex); 560 mutex_lock(&vq->mutex);
556 sock = rcu_dereference_protected(vq->private_data, 561 sock = rcu_dereference_protected(vq->private_data,
557 lockdep_is_held(&vq->mutex)); 562 lockdep_is_held(&vq->mutex));
558 vhost_net_disable_vq(n, vq); 563 vhost_net_disable_vq(n, vq);
559 rcu_assign_pointer(vq->private_data, NULL); 564 rcu_assign_pointer(vq->private_data, NULL);
560 mutex_unlock(&vq->mutex); 565 mutex_unlock(&vq->mutex);
561 return sock; 566 return sock;
562 } 567 }
563 568
564 static void vhost_net_stop(struct vhost_net *n, struct socket **tx_sock, 569 static void vhost_net_stop(struct vhost_net *n, struct socket **tx_sock,
565 struct socket **rx_sock) 570 struct socket **rx_sock)
566 { 571 {
567 *tx_sock = vhost_net_stop_vq(n, n->vqs + VHOST_NET_VQ_TX); 572 *tx_sock = vhost_net_stop_vq(n, n->vqs + VHOST_NET_VQ_TX);
568 *rx_sock = vhost_net_stop_vq(n, n->vqs + VHOST_NET_VQ_RX); 573 *rx_sock = vhost_net_stop_vq(n, n->vqs + VHOST_NET_VQ_RX);
569 } 574 }
570 575
571 static void vhost_net_flush_vq(struct vhost_net *n, int index) 576 static void vhost_net_flush_vq(struct vhost_net *n, int index)
572 { 577 {
573 vhost_poll_flush(n->poll + index); 578 vhost_poll_flush(n->poll + index);
574 vhost_poll_flush(&n->dev.vqs[index].poll); 579 vhost_poll_flush(&n->dev.vqs[index].poll);
575 } 580 }
576 581
577 static void vhost_net_flush(struct vhost_net *n) 582 static void vhost_net_flush(struct vhost_net *n)
578 { 583 {
579 vhost_net_flush_vq(n, VHOST_NET_VQ_TX); 584 vhost_net_flush_vq(n, VHOST_NET_VQ_TX);
580 vhost_net_flush_vq(n, VHOST_NET_VQ_RX); 585 vhost_net_flush_vq(n, VHOST_NET_VQ_RX);
581 } 586 }
582 587
583 static int vhost_net_release(struct inode *inode, struct file *f) 588 static int vhost_net_release(struct inode *inode, struct file *f)
584 { 589 {
585 struct vhost_net *n = f->private_data; 590 struct vhost_net *n = f->private_data;
586 struct socket *tx_sock; 591 struct socket *tx_sock;
587 struct socket *rx_sock; 592 struct socket *rx_sock;
588 593
589 vhost_net_stop(n, &tx_sock, &rx_sock); 594 vhost_net_stop(n, &tx_sock, &rx_sock);
590 vhost_net_flush(n); 595 vhost_net_flush(n);
591 vhost_dev_cleanup(&n->dev, false); 596 vhost_dev_cleanup(&n->dev, false);
592 if (tx_sock) 597 if (tx_sock)
593 fput(tx_sock->file); 598 fput(tx_sock->file);
594 if (rx_sock) 599 if (rx_sock)
595 fput(rx_sock->file); 600 fput(rx_sock->file);
596 /* We do an extra flush before freeing memory, 601 /* We do an extra flush before freeing memory,
597 * since jobs can re-queue themselves. */ 602 * since jobs can re-queue themselves. */
598 vhost_net_flush(n); 603 vhost_net_flush(n);
599 kfree(n); 604 kfree(n);
600 return 0; 605 return 0;
601 } 606 }
602 607
603 static struct socket *get_raw_socket(int fd) 608 static struct socket *get_raw_socket(int fd)
604 { 609 {
605 struct { 610 struct {
606 struct sockaddr_ll sa; 611 struct sockaddr_ll sa;
607 char buf[MAX_ADDR_LEN]; 612 char buf[MAX_ADDR_LEN];
608 } uaddr; 613 } uaddr;
609 int uaddr_len = sizeof uaddr, r; 614 int uaddr_len = sizeof uaddr, r;
610 struct socket *sock = sockfd_lookup(fd, &r); 615 struct socket *sock = sockfd_lookup(fd, &r);
611 616
612 if (!sock) 617 if (!sock)
613 return ERR_PTR(-ENOTSOCK); 618 return ERR_PTR(-ENOTSOCK);
614 619
615 /* Parameter checking */ 620 /* Parameter checking */
616 if (sock->sk->sk_type != SOCK_RAW) { 621 if (sock->sk->sk_type != SOCK_RAW) {
617 r = -ESOCKTNOSUPPORT; 622 r = -ESOCKTNOSUPPORT;
618 goto err; 623 goto err;
619 } 624 }
620 625
621 r = sock->ops->getname(sock, (struct sockaddr *)&uaddr.sa, 626 r = sock->ops->getname(sock, (struct sockaddr *)&uaddr.sa,
622 &uaddr_len, 0); 627 &uaddr_len, 0);
623 if (r) 628 if (r)
624 goto err; 629 goto err;
625 630
626 if (uaddr.sa.sll_family != AF_PACKET) { 631 if (uaddr.sa.sll_family != AF_PACKET) {
627 r = -EPFNOSUPPORT; 632 r = -EPFNOSUPPORT;
628 goto err; 633 goto err;
629 } 634 }
630 return sock; 635 return sock;
631 err: 636 err:
632 fput(sock->file); 637 fput(sock->file);
633 return ERR_PTR(r); 638 return ERR_PTR(r);
634 } 639 }
635 640
636 static struct socket *get_tap_socket(int fd) 641 static struct socket *get_tap_socket(int fd)
637 { 642 {
638 struct file *file = fget(fd); 643 struct file *file = fget(fd);
639 struct socket *sock; 644 struct socket *sock;
640 645
641 if (!file) 646 if (!file)
642 return ERR_PTR(-EBADF); 647 return ERR_PTR(-EBADF);
643 sock = tun_get_socket(file); 648 sock = tun_get_socket(file);
644 if (!IS_ERR(sock)) 649 if (!IS_ERR(sock))
645 return sock; 650 return sock;
646 sock = macvtap_get_socket(file); 651 sock = macvtap_get_socket(file);
647 if (IS_ERR(sock)) 652 if (IS_ERR(sock))
648 fput(file); 653 fput(file);
649 return sock; 654 return sock;
650 } 655 }
651 656
652 static struct socket *get_socket(int fd) 657 static struct socket *get_socket(int fd)
653 { 658 {
654 struct socket *sock; 659 struct socket *sock;
655 660
656 /* special case to disable backend */ 661 /* special case to disable backend */
657 if (fd == -1) 662 if (fd == -1)
658 return NULL; 663 return NULL;
659 sock = get_raw_socket(fd); 664 sock = get_raw_socket(fd);
660 if (!IS_ERR(sock)) 665 if (!IS_ERR(sock))
661 return sock; 666 return sock;
662 sock = get_tap_socket(fd); 667 sock = get_tap_socket(fd);
663 if (!IS_ERR(sock)) 668 if (!IS_ERR(sock))
664 return sock; 669 return sock;
665 return ERR_PTR(-ENOTSOCK); 670 return ERR_PTR(-ENOTSOCK);
666 } 671 }
667 672
668 static long vhost_net_set_backend(struct vhost_net *n, unsigned index, int fd) 673 static long vhost_net_set_backend(struct vhost_net *n, unsigned index, int fd)
669 { 674 {
670 struct socket *sock, *oldsock; 675 struct socket *sock, *oldsock;
671 struct vhost_virtqueue *vq; 676 struct vhost_virtqueue *vq;
672 struct vhost_ubuf_ref *ubufs, *oldubufs = NULL; 677 struct vhost_ubuf_ref *ubufs, *oldubufs = NULL;
673 int r; 678 int r;
674 679
675 mutex_lock(&n->dev.mutex); 680 mutex_lock(&n->dev.mutex);
676 r = vhost_dev_check_owner(&n->dev); 681 r = vhost_dev_check_owner(&n->dev);
677 if (r) 682 if (r)
678 goto err; 683 goto err;
679 684
680 if (index >= VHOST_NET_VQ_MAX) { 685 if (index >= VHOST_NET_VQ_MAX) {
681 r = -ENOBUFS; 686 r = -ENOBUFS;
682 goto err; 687 goto err;
683 } 688 }
684 vq = n->vqs + index; 689 vq = n->vqs + index;
685 mutex_lock(&vq->mutex); 690 mutex_lock(&vq->mutex);
686 691
687 /* Verify that ring has been setup correctly. */ 692 /* Verify that ring has been setup correctly. */
688 if (!vhost_vq_access_ok(vq)) { 693 if (!vhost_vq_access_ok(vq)) {
689 r = -EFAULT; 694 r = -EFAULT;
690 goto err_vq; 695 goto err_vq;
691 } 696 }
692 sock = get_socket(fd); 697 sock = get_socket(fd);
693 if (IS_ERR(sock)) { 698 if (IS_ERR(sock)) {
694 r = PTR_ERR(sock); 699 r = PTR_ERR(sock);
695 goto err_vq; 700 goto err_vq;
696 } 701 }
697 702
698 /* start polling new socket */ 703 /* start polling new socket */
699 oldsock = rcu_dereference_protected(vq->private_data, 704 oldsock = rcu_dereference_protected(vq->private_data,
700 lockdep_is_held(&vq->mutex)); 705 lockdep_is_held(&vq->mutex));
701 if (sock != oldsock) { 706 if (sock != oldsock) {
702 ubufs = vhost_ubuf_alloc(vq, sock && vhost_sock_zcopy(sock)); 707 ubufs = vhost_ubuf_alloc(vq, sock && vhost_sock_zcopy(sock));
703 if (IS_ERR(ubufs)) { 708 if (IS_ERR(ubufs)) {
704 r = PTR_ERR(ubufs); 709 r = PTR_ERR(ubufs);
705 goto err_ubufs; 710 goto err_ubufs;
706 } 711 }
707 oldubufs = vq->ubufs; 712 oldubufs = vq->ubufs;
708 vq->ubufs = ubufs; 713 vq->ubufs = ubufs;
709 vhost_net_disable_vq(n, vq); 714 vhost_net_disable_vq(n, vq);
710 rcu_assign_pointer(vq->private_data, sock); 715 rcu_assign_pointer(vq->private_data, sock);
711 vhost_net_enable_vq(n, vq); 716 vhost_net_enable_vq(n, vq);
712 717
713 r = vhost_init_used(vq); 718 r = vhost_init_used(vq);
714 if (r) 719 if (r)
715 goto err_vq; 720 goto err_vq;
716 } 721 }
717 722
718 mutex_unlock(&vq->mutex); 723 mutex_unlock(&vq->mutex);
719 724
720 if (oldubufs) { 725 if (oldubufs) {
721 vhost_ubuf_put_and_wait(oldubufs); 726 vhost_ubuf_put_and_wait(oldubufs);
722 mutex_lock(&vq->mutex); 727 mutex_lock(&vq->mutex);
723 vhost_zerocopy_signal_used(vq); 728 vhost_zerocopy_signal_used(vq);
724 mutex_unlock(&vq->mutex); 729 mutex_unlock(&vq->mutex);
725 } 730 }
726 731
727 if (oldsock) { 732 if (oldsock) {
728 vhost_net_flush_vq(n, index); 733 vhost_net_flush_vq(n, index);
729 fput(oldsock->file); 734 fput(oldsock->file);
730 } 735 }
731 736
732 mutex_unlock(&n->dev.mutex); 737 mutex_unlock(&n->dev.mutex);
733 return 0; 738 return 0;
734 739
735 err_ubufs: 740 err_ubufs:
736 fput(sock->file); 741 fput(sock->file);
737 err_vq: 742 err_vq:
738 mutex_unlock(&vq->mutex); 743 mutex_unlock(&vq->mutex);
739 err: 744 err:
740 mutex_unlock(&n->dev.mutex); 745 mutex_unlock(&n->dev.mutex);
741 return r; 746 return r;
742 } 747 }
743 748
744 static long vhost_net_reset_owner(struct vhost_net *n) 749 static long vhost_net_reset_owner(struct vhost_net *n)
745 { 750 {
746 struct socket *tx_sock = NULL; 751 struct socket *tx_sock = NULL;
747 struct socket *rx_sock = NULL; 752 struct socket *rx_sock = NULL;
748 long err; 753 long err;
749 754
750 mutex_lock(&n->dev.mutex); 755 mutex_lock(&n->dev.mutex);
751 err = vhost_dev_check_owner(&n->dev); 756 err = vhost_dev_check_owner(&n->dev);
752 if (err) 757 if (err)
753 goto done; 758 goto done;
754 vhost_net_stop(n, &tx_sock, &rx_sock); 759 vhost_net_stop(n, &tx_sock, &rx_sock);
755 vhost_net_flush(n); 760 vhost_net_flush(n);
756 err = vhost_dev_reset_owner(&n->dev); 761 err = vhost_dev_reset_owner(&n->dev);
757 done: 762 done:
758 mutex_unlock(&n->dev.mutex); 763 mutex_unlock(&n->dev.mutex);
759 if (tx_sock) 764 if (tx_sock)
760 fput(tx_sock->file); 765 fput(tx_sock->file);
761 if (rx_sock) 766 if (rx_sock)
762 fput(rx_sock->file); 767 fput(rx_sock->file);
763 return err; 768 return err;
764 } 769 }
765 770
766 static int vhost_net_set_features(struct vhost_net *n, u64 features) 771 static int vhost_net_set_features(struct vhost_net *n, u64 features)
767 { 772 {
768 size_t vhost_hlen, sock_hlen, hdr_len; 773 size_t vhost_hlen, sock_hlen, hdr_len;
769 int i; 774 int i;
770 775
771 hdr_len = (features & (1 << VIRTIO_NET_F_MRG_RXBUF)) ? 776 hdr_len = (features & (1 << VIRTIO_NET_F_MRG_RXBUF)) ?
772 sizeof(struct virtio_net_hdr_mrg_rxbuf) : 777 sizeof(struct virtio_net_hdr_mrg_rxbuf) :
773 sizeof(struct virtio_net_hdr); 778 sizeof(struct virtio_net_hdr);
774 if (features & (1 << VHOST_NET_F_VIRTIO_NET_HDR)) { 779 if (features & (1 << VHOST_NET_F_VIRTIO_NET_HDR)) {
775 /* vhost provides vnet_hdr */ 780 /* vhost provides vnet_hdr */
776 vhost_hlen = hdr_len; 781 vhost_hlen = hdr_len;
777 sock_hlen = 0; 782 sock_hlen = 0;
778 } else { 783 } else {
779 /* socket provides vnet_hdr */ 784 /* socket provides vnet_hdr */
780 vhost_hlen = 0; 785 vhost_hlen = 0;
781 sock_hlen = hdr_len; 786 sock_hlen = hdr_len;
782 } 787 }
783 mutex_lock(&n->dev.mutex); 788 mutex_lock(&n->dev.mutex);
784 if ((features & (1 << VHOST_F_LOG_ALL)) && 789 if ((features & (1 << VHOST_F_LOG_ALL)) &&
785 !vhost_log_access_ok(&n->dev)) { 790 !vhost_log_access_ok(&n->dev)) {
786 mutex_unlock(&n->dev.mutex); 791 mutex_unlock(&n->dev.mutex);
787 return -EFAULT; 792 return -EFAULT;
788 } 793 }
789 n->dev.acked_features = features; 794 n->dev.acked_features = features;
790 smp_wmb(); 795 smp_wmb();
791 for (i = 0; i < VHOST_NET_VQ_MAX; ++i) { 796 for (i = 0; i < VHOST_NET_VQ_MAX; ++i) {
792 mutex_lock(&n->vqs[i].mutex); 797 mutex_lock(&n->vqs[i].mutex);
793 n->vqs[i].vhost_hlen = vhost_hlen; 798 n->vqs[i].vhost_hlen = vhost_hlen;
794 n->vqs[i].sock_hlen = sock_hlen; 799 n->vqs[i].sock_hlen = sock_hlen;
795 mutex_unlock(&n->vqs[i].mutex); 800 mutex_unlock(&n->vqs[i].mutex);
796 } 801 }
797 vhost_net_flush(n); 802 vhost_net_flush(n);
798 mutex_unlock(&n->dev.mutex); 803 mutex_unlock(&n->dev.mutex);
799 return 0; 804 return 0;
800 } 805 }
801 806
802 static long vhost_net_ioctl(struct file *f, unsigned int ioctl, 807 static long vhost_net_ioctl(struct file *f, unsigned int ioctl,
803 unsigned long arg) 808 unsigned long arg)
804 { 809 {
805 struct vhost_net *n = f->private_data; 810 struct vhost_net *n = f->private_data;
806 void __user *argp = (void __user *)arg; 811 void __user *argp = (void __user *)arg;
807 u64 __user *featurep = argp; 812 u64 __user *featurep = argp;
808 struct vhost_vring_file backend; 813 struct vhost_vring_file backend;
809 u64 features; 814 u64 features;
810 int r; 815 int r;
811 816
812 switch (ioctl) { 817 switch (ioctl) {
813 case VHOST_NET_SET_BACKEND: 818 case VHOST_NET_SET_BACKEND:
814 if (copy_from_user(&backend, argp, sizeof backend)) 819 if (copy_from_user(&backend, argp, sizeof backend))
815 return -EFAULT; 820 return -EFAULT;
816 return vhost_net_set_backend(n, backend.index, backend.fd); 821 return vhost_net_set_backend(n, backend.index, backend.fd);
817 case VHOST_GET_FEATURES: 822 case VHOST_GET_FEATURES:
818 features = VHOST_FEATURES; 823 features = VHOST_FEATURES;
819 if (copy_to_user(featurep, &features, sizeof features)) 824 if (copy_to_user(featurep, &features, sizeof features))
820 return -EFAULT; 825 return -EFAULT;
821 return 0; 826 return 0;
822 case VHOST_SET_FEATURES: 827 case VHOST_SET_FEATURES:
823 if (copy_from_user(&features, featurep, sizeof features)) 828 if (copy_from_user(&features, featurep, sizeof features))
824 return -EFAULT; 829 return -EFAULT;
825 if (features & ~VHOST_FEATURES) 830 if (features & ~VHOST_FEATURES)
826 return -EOPNOTSUPP; 831 return -EOPNOTSUPP;
827 return vhost_net_set_features(n, features); 832 return vhost_net_set_features(n, features);
828 case VHOST_RESET_OWNER: 833 case VHOST_RESET_OWNER:
829 return vhost_net_reset_owner(n); 834 return vhost_net_reset_owner(n);
830 default: 835 default:
831 mutex_lock(&n->dev.mutex); 836 mutex_lock(&n->dev.mutex);
832 r = vhost_dev_ioctl(&n->dev, ioctl, arg); 837 r = vhost_dev_ioctl(&n->dev, ioctl, arg);
833 vhost_net_flush(n); 838 vhost_net_flush(n);
834 mutex_unlock(&n->dev.mutex); 839 mutex_unlock(&n->dev.mutex);
835 return r; 840 return r;
836 } 841 }
837 } 842 }
838 843
839 #ifdef CONFIG_COMPAT 844 #ifdef CONFIG_COMPAT
840 static long vhost_net_compat_ioctl(struct file *f, unsigned int ioctl, 845 static long vhost_net_compat_ioctl(struct file *f, unsigned int ioctl,
841 unsigned long arg) 846 unsigned long arg)
842 { 847 {
843 return vhost_net_ioctl(f, ioctl, (unsigned long)compat_ptr(arg)); 848 return vhost_net_ioctl(f, ioctl, (unsigned long)compat_ptr(arg));
844 } 849 }
845 #endif 850 #endif
846 851
847 static const struct file_operations vhost_net_fops = { 852 static const struct file_operations vhost_net_fops = {
848 .owner = THIS_MODULE, 853 .owner = THIS_MODULE,
849 .release = vhost_net_release, 854 .release = vhost_net_release,
850 .unlocked_ioctl = vhost_net_ioctl, 855 .unlocked_ioctl = vhost_net_ioctl,
851 #ifdef CONFIG_COMPAT 856 #ifdef CONFIG_COMPAT
852 .compat_ioctl = vhost_net_compat_ioctl, 857 .compat_ioctl = vhost_net_compat_ioctl,
853 #endif 858 #endif
854 .open = vhost_net_open, 859 .open = vhost_net_open,
855 .llseek = noop_llseek, 860 .llseek = noop_llseek,
856 }; 861 };
857 862
858 static struct miscdevice vhost_net_misc = { 863 static struct miscdevice vhost_net_misc = {
859 .minor = VHOST_NET_MINOR, 864 .minor = VHOST_NET_MINOR,
860 .name = "vhost-net", 865 .name = "vhost-net",
861 .fops = &vhost_net_fops, 866 .fops = &vhost_net_fops,
862 }; 867 };
863 868
864 static int vhost_net_init(void) 869 static int vhost_net_init(void)
865 { 870 {
866 if (experimental_zcopytx) 871 if (experimental_zcopytx)
867 vhost_enable_zcopy(VHOST_NET_VQ_TX); 872 vhost_enable_zcopy(VHOST_NET_VQ_TX);
868 return misc_register(&vhost_net_misc); 873 return misc_register(&vhost_net_misc);
869 } 874 }
870 module_init(vhost_net_init); 875 module_init(vhost_net_init);
871 876
872 static void vhost_net_exit(void) 877 static void vhost_net_exit(void)
873 { 878 {
874 misc_deregister(&vhost_net_misc); 879 misc_deregister(&vhost_net_misc);
875 } 880 }
876 module_exit(vhost_net_exit); 881 module_exit(vhost_net_exit);
877 882
878 MODULE_VERSION("0.0.1"); 883 MODULE_VERSION("0.0.1");
879 MODULE_LICENSE("GPL v2"); 884 MODULE_LICENSE("GPL v2");
880 MODULE_AUTHOR("Michael S. Tsirkin"); 885 MODULE_AUTHOR("Michael S. Tsirkin");
881 MODULE_DESCRIPTION("Host kernel accelerator for virtio net"); 886 MODULE_DESCRIPTION("Host kernel accelerator for virtio net");
882 MODULE_ALIAS_MISCDEV(VHOST_NET_MINOR); 887 MODULE_ALIAS_MISCDEV(VHOST_NET_MINOR);
883 MODULE_ALIAS("devname:vhost-net"); 888 MODULE_ALIAS("devname:vhost-net");
884 889