]>
Commit | Line | Data |
---|---|---|
f942dc25 IC |
1 | /* |
2 | * This program is free software; you can redistribute it and/or | |
3 | * modify it under the terms of the GNU General Public License version 2 | |
4 | * as published by the Free Software Foundation; or, when distributed | |
5 | * separately from the Linux kernel or incorporated into other | |
6 | * software packages, subject to the following license: | |
7 | * | |
8 | * Permission is hereby granted, free of charge, to any person obtaining a copy | |
9 | * of this source file (the "Software"), to deal in the Software without | |
10 | * restriction, including without limitation the rights to use, copy, modify, | |
11 | * merge, publish, distribute, sublicense, and/or sell copies of the Software, | |
12 | * and to permit persons to whom the Software is furnished to do so, subject to | |
13 | * the following conditions: | |
14 | * | |
15 | * The above copyright notice and this permission notice shall be included in | |
16 | * all copies or substantial portions of the Software. | |
17 | * | |
18 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
19 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
20 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE | |
21 | * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
22 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING | |
23 | * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS | |
24 | * IN THE SOFTWARE. | |
25 | */ | |
26 | ||
27 | #ifndef __XEN_NETBACK__COMMON_H__ | |
28 | #define __XEN_NETBACK__COMMON_H__ | |
29 | ||
30 | #define pr_fmt(fmt) KBUILD_MODNAME ":%s: " fmt, __func__ | |
31 | ||
32 | #include <linux/module.h> | |
33 | #include <linux/interrupt.h> | |
34 | #include <linux/slab.h> | |
35 | #include <linux/ip.h> | |
36 | #include <linux/in.h> | |
37 | #include <linux/io.h> | |
38 | #include <linux/netdevice.h> | |
39 | #include <linux/etherdevice.h> | |
40 | #include <linux/wait.h> | |
41 | #include <linux/sched.h> | |
42 | ||
43 | #include <xen/interface/io/netif.h> | |
44 | #include <xen/interface/grant_table.h> | |
45 | #include <xen/grant_table.h> | |
46 | #include <xen/xenbus.h> | |
47 | ||
b3f980bd WL |
48 | typedef unsigned int pending_ring_idx_t; |
49 | #define INVALID_PENDING_RING_IDX (~0U) | |
50 | ||
b3f980bd | 51 | struct pending_tx_info { |
62bad319 | 52 | struct xen_netif_tx_request req; /* tx request */ |
f53c3fe8 ZK |
53 | /* Callback data for released SKBs. The callback is always |
54 | * xenvif_zerocopy_callback, desc contains the pending_idx, which is | |
55 | * also an index in pending_tx_info array. It is initialized in | |
56 | * xenvif_alloc and it never changes. | |
57 | * skb_shinfo(skb)->destructor_arg points to the first mapped slot's | |
58 | * callback_struct in this array of struct pending_tx_info's, then ctx | |
59 | * to the next, or NULL if there is no more slot for this skb. | |
60 | * ubuf_to_vif is a helper which finds the struct xenvif from a pointer | |
61 | * to this field. | |
62 | */ | |
63 | struct ubuf_info callback_struct; | |
b3f980bd WL |
64 | }; |
65 | ||
66 | #define XEN_NETIF_TX_RING_SIZE __CONST_RING_SIZE(xen_netif_tx, PAGE_SIZE) | |
67 | #define XEN_NETIF_RX_RING_SIZE __CONST_RING_SIZE(xen_netif_rx, PAGE_SIZE) | |
68 | ||
69 | struct xenvif_rx_meta { | |
70 | int id; | |
71 | int size; | |
82cada22 | 72 | int gso_type; |
b3f980bd WL |
73 | int gso_size; |
74 | }; | |
75 | ||
82cada22 PD |
76 | #define GSO_BIT(type) \ |
77 | (1 << XEN_NETIF_GSO_TYPE_ ## type) | |
78 | ||
b3f980bd WL |
79 | /* Discriminate from any valid pending_idx value. */ |
80 | #define INVALID_PENDING_IDX 0xFFFF | |
81 | ||
82 | #define MAX_BUFFER_OFFSET PAGE_SIZE | |
83 | ||
869b9b19 | 84 | #define MAX_PENDING_REQS XEN_NETIF_TX_RING_SIZE |
f942dc25 | 85 | |
ac3d5ac2 PD |
86 | /* It's possible for an skb to have a maximal number of frags |
87 | * but still be less than MAX_BUFFER_OFFSET in size. Thus the | |
88 | * worst-case number of copy operations is MAX_SKB_FRAGS per | |
89 | * ring slot. | |
90 | */ | |
91 | #define MAX_GRANT_COPY_OPS (MAX_SKB_FRAGS * XEN_NETIF_RX_RING_SIZE) | |
92 | ||
121fa4b7 ZK |
93 | #define NETBACK_INVALID_HANDLE -1 |
94 | ||
95 | /* To avoid confusion, we define XEN_NETBK_LEGACY_SLOTS_MAX indicating | |
96 | * the maximum slots a valid packet can use. Now this value is defined | |
97 | * to be XEN_NETIF_NR_SLOTS_MIN, which is supposed to be supported by | |
98 | * all backend. | |
99 | */ | |
100 | #define XEN_NETBK_LEGACY_SLOTS_MAX XEN_NETIF_NR_SLOTS_MIN | |
101 | ||
f942dc25 IC |
102 | struct xenvif { |
103 | /* Unique identifier for this interface. */ | |
104 | domid_t domid; | |
105 | unsigned int handle; | |
106 | ||
e9d8b2c2 WL |
107 | /* Is this interface disabled? True when backend discovers |
108 | * frontend is rogue. | |
109 | */ | |
110 | bool disabled; | |
111 | ||
b3f980bd WL |
112 | /* Use NAPI for guest TX */ |
113 | struct napi_struct napi; | |
114 | /* When feature-split-event-channels = 0, tx_irq = rx_irq. */ | |
115 | unsigned int tx_irq; | |
116 | /* Only used when feature-split-event-channels = 1 */ | |
117 | char tx_irq_name[IFNAMSIZ+4]; /* DEVNAME-tx */ | |
118 | struct xen_netif_tx_back_ring tx; | |
119 | struct sk_buff_head tx_queue; | |
120 | struct page *mmap_pages[MAX_PENDING_REQS]; | |
121 | pending_ring_idx_t pending_prod; | |
122 | pending_ring_idx_t pending_cons; | |
123 | u16 pending_ring[MAX_PENDING_REQS]; | |
124 | struct pending_tx_info pending_tx_info[MAX_PENDING_REQS]; | |
f53c3fe8 | 125 | grant_handle_t grant_tx_handle[MAX_PENDING_REQS]; |
b3f980bd | 126 | |
f53c3fe8 ZK |
127 | struct gnttab_map_grant_ref tx_map_ops[MAX_PENDING_REQS]; |
128 | struct gnttab_unmap_grant_ref tx_unmap_ops[MAX_PENDING_REQS]; | |
129 | /* passed to gnttab_[un]map_refs with pages under (un)mapping */ | |
130 | struct page *pages_to_map[MAX_PENDING_REQS]; | |
131 | struct page *pages_to_unmap[MAX_PENDING_REQS]; | |
132 | ||
133 | /* This prevents zerocopy callbacks to race over dealloc_ring */ | |
134 | spinlock_t callback_lock; | |
135 | /* This prevents dealloc thread and NAPI instance to race over response | |
136 | * creation and pending_ring in xenvif_idx_release. In xenvif_tx_err | |
137 | * it only protect response creation | |
138 | */ | |
139 | spinlock_t response_lock; | |
140 | pending_ring_idx_t dealloc_prod; | |
141 | pending_ring_idx_t dealloc_cons; | |
142 | u16 dealloc_ring[MAX_PENDING_REQS]; | |
143 | struct task_struct *dealloc_task; | |
144 | wait_queue_head_t dealloc_wq; | |
f942dc25 | 145 | |
b3f980bd WL |
146 | /* Use kthread for guest RX */ |
147 | struct task_struct *task; | |
148 | wait_queue_head_t wq; | |
e1f00a69 | 149 | /* When feature-split-event-channels = 0, tx_irq = rx_irq. */ |
e1f00a69 WL |
150 | unsigned int rx_irq; |
151 | /* Only used when feature-split-event-channels = 1 */ | |
e1f00a69 | 152 | char rx_irq_name[IFNAMSIZ+4]; /* DEVNAME-rx */ |
b3f980bd WL |
153 | struct xen_netif_rx_back_ring rx; |
154 | struct sk_buff_head rx_queue; | |
9ab9831b | 155 | RING_IDX rx_last_skb_slots; |
09350788 ZK |
156 | bool rx_queue_purge; |
157 | ||
158 | struct timer_list wake_queue; | |
b3f980bd | 159 | |
ac3d5ac2 PD |
160 | /* This array is allocated seperately as it is large */ |
161 | struct gnttab_copy *grant_copy_op; | |
f942dc25 | 162 | |
ac3d5ac2 PD |
163 | /* We create one meta structure per ring request we consume, so |
164 | * the maximum number is the same as the ring size. | |
165 | */ | |
166 | struct xenvif_rx_meta meta[XEN_NETIF_RX_RING_SIZE]; | |
b3f980bd WL |
167 | |
168 | u8 fe_dev_addr[6]; | |
f942dc25 | 169 | |
f942dc25 | 170 | /* Frontend feature information. */ |
82cada22 PD |
171 | int gso_mask; |
172 | int gso_prefix_mask; | |
173 | ||
f942dc25 | 174 | u8 can_sg:1; |
146c8a77 PD |
175 | u8 ip_csum:1; |
176 | u8 ipv6_csum:1; | |
f942dc25 IC |
177 | |
178 | /* Internal feature information. */ | |
179 | u8 can_queue:1; /* can queue packets for receiver? */ | |
180 | ||
f942dc25 IC |
181 | /* Transmit shaping: allow 'credit_bytes' every 'credit_usec'. */ |
182 | unsigned long credit_bytes; | |
183 | unsigned long credit_usec; | |
184 | unsigned long remaining_credit; | |
185 | struct timer_list credit_timeout; | |
059dfa6a | 186 | u64 credit_window_start; |
f942dc25 IC |
187 | |
188 | /* Statistics */ | |
189 | unsigned long rx_gso_checksum_fixup; | |
1bb332af ZK |
190 | unsigned long tx_zerocopy_sent; |
191 | unsigned long tx_zerocopy_success; | |
192 | unsigned long tx_zerocopy_fail; | |
e3377f36 | 193 | unsigned long tx_frag_overflow; |
f942dc25 IC |
194 | |
195 | /* Miscellaneous private stuff. */ | |
f942dc25 | 196 | struct net_device *dev; |
f942dc25 IC |
197 | }; |
198 | ||
c9d63699 DV |
199 | static inline struct xenbus_device *xenvif_to_xenbus_device(struct xenvif *vif) |
200 | { | |
201 | return to_xenbus_device(vif->dev->dev.parent); | |
202 | } | |
203 | ||
f942dc25 IC |
204 | struct xenvif *xenvif_alloc(struct device *parent, |
205 | domid_t domid, | |
206 | unsigned int handle); | |
207 | ||
208 | int xenvif_connect(struct xenvif *vif, unsigned long tx_ring_ref, | |
e1f00a69 WL |
209 | unsigned long rx_ring_ref, unsigned int tx_evtchn, |
210 | unsigned int rx_evtchn); | |
f942dc25 | 211 | void xenvif_disconnect(struct xenvif *vif); |
279f438e | 212 | void xenvif_free(struct xenvif *vif); |
f942dc25 | 213 | |
f942dc25 | 214 | int xenvif_xenbus_init(void); |
b103f358 | 215 | void xenvif_xenbus_fini(void); |
f942dc25 IC |
216 | |
217 | int xenvif_schedulable(struct xenvif *vif); | |
218 | ||
7376419a | 219 | int xenvif_must_stop_queue(struct xenvif *vif); |
f942dc25 IC |
220 | |
221 | /* (Un)Map communication rings. */ | |
7376419a WL |
222 | void xenvif_unmap_frontend_rings(struct xenvif *vif); |
223 | int xenvif_map_frontend_rings(struct xenvif *vif, | |
224 | grant_ref_t tx_ring_ref, | |
225 | grant_ref_t rx_ring_ref); | |
f942dc25 | 226 | |
f942dc25 | 227 | /* Check for SKBs from frontend and schedule backend processing */ |
7376419a | 228 | void xenvif_check_rx_xenvif(struct xenvif *vif); |
f942dc25 | 229 | |
48856286 IC |
230 | /* Prevent the device from generating any further traffic. */ |
231 | void xenvif_carrier_off(struct xenvif *vif); | |
232 | ||
7376419a | 233 | int xenvif_tx_action(struct xenvif *vif, int budget); |
b3f980bd | 234 | |
121fa4b7 | 235 | int xenvif_kthread_guest_rx(void *data); |
ca2f09f2 PD |
236 | void xenvif_kick_thread(struct xenvif *vif); |
237 | ||
f53c3fe8 ZK |
238 | int xenvif_dealloc_kthread(void *data); |
239 | ||
ca2f09f2 PD |
240 | /* Determine whether the needed number of slots (req) are available, |
241 | * and set req_event if not. | |
242 | */ | |
243 | bool xenvif_rx_ring_slots_available(struct xenvif *vif, int needed); | |
244 | ||
245 | void xenvif_stop_queue(struct xenvif *vif); | |
b3f980bd | 246 | |
f53c3fe8 ZK |
247 | /* Callback from stack when TX packet can be released */ |
248 | void xenvif_zerocopy_callback(struct ubuf_info *ubuf, bool zerocopy_success); | |
249 | ||
250 | /* Unmap a pending page and release it back to the guest */ | |
251 | void xenvif_idx_unmap(struct xenvif *vif, u16 pending_idx); | |
252 | ||
121fa4b7 ZK |
253 | static inline pending_ring_idx_t nr_pending_reqs(struct xenvif *vif) |
254 | { | |
255 | return MAX_PENDING_REQS - | |
256 | vif->pending_prod + vif->pending_cons; | |
257 | } | |
258 | ||
3e2234b3 ZK |
259 | /* Callback from stack when TX packet can be released */ |
260 | void xenvif_zerocopy_callback(struct ubuf_info *ubuf, bool zerocopy_success); | |
261 | ||
e1f00a69 WL |
262 | extern bool separate_tx_rx_irq; |
263 | ||
09350788 ZK |
264 | extern unsigned int rx_drain_timeout_msecs; |
265 | extern unsigned int rx_drain_timeout_jiffies; | |
266 | ||
f942dc25 | 267 | #endif /* __XEN_NETBACK__COMMON_H__ */ |