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:
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:
15 * The above copyright notice and this permission notice shall be included in
16 * all copies or substantial portions of the Software.
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
27 #ifndef __XEN_NETBACK__COMMON_H__
28 #define __XEN_NETBACK__COMMON_H__
30 #define pr_fmt(fmt) KBUILD_MODNAME ":%s: " fmt, __func__
32 #include <linux/module.h>
33 #include <linux/interrupt.h>
34 #include <linux/slab.h>
38 #include <linux/netdevice.h>
39 #include <linux/etherdevice.h>
40 #include <linux/wait.h>
41 #include <linux/sched.h>
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>
48 #include <linux/debugfs.h>
50 typedef unsigned int pending_ring_idx_t;
51 #define INVALID_PENDING_RING_IDX (~0U)
53 struct pending_tx_info {
54 struct xen_netif_tx_request req; /* tx request */
55 unsigned int extra_count;
56 /* Callback data for released SKBs. The callback is always
57 * xenvif_zerocopy_callback, desc contains the pending_idx, which is
58 * also an index in pending_tx_info array. It is initialized in
59 * xenvif_alloc and it never changes.
60 * skb_shinfo(skb)->destructor_arg points to the first mapped slot's
61 * callback_struct in this array of struct pending_tx_info's, then ctx
62 * to the next, or NULL if there is no more slot for this skb.
63 * ubuf_to_vif is a helper which finds the struct xenvif from a pointer
66 struct ubuf_info callback_struct;
69 #define XEN_NETIF_TX_RING_SIZE __CONST_RING_SIZE(xen_netif_tx, XEN_PAGE_SIZE)
70 #define XEN_NETIF_RX_RING_SIZE __CONST_RING_SIZE(xen_netif_rx, XEN_PAGE_SIZE)
72 struct xenvif_rx_meta {
79 #define GSO_BIT(type) \
80 (1 << XEN_NETIF_GSO_TYPE_ ## type)
82 /* Discriminate from any valid pending_idx value. */
83 #define INVALID_PENDING_IDX 0xFFFF
85 #define MAX_BUFFER_OFFSET XEN_PAGE_SIZE
87 #define MAX_PENDING_REQS XEN_NETIF_TX_RING_SIZE
89 /* The maximum number of frags is derived from the size of a grant (same
90 * as a Xen page size for now).
92 #define MAX_XEN_SKB_FRAGS (65536 / XEN_PAGE_SIZE + 1)
94 #define NETBACK_INVALID_HANDLE -1
96 /* To avoid confusion, we define XEN_NETBK_LEGACY_SLOTS_MAX indicating
97 * the maximum slots a valid packet can use. Now this value is defined
98 * to be XEN_NETIF_NR_SLOTS_MIN, which is supposed to be supported by
101 #define XEN_NETBK_LEGACY_SLOTS_MAX XEN_NETIF_NR_SLOTS_MIN
103 /* Queue name is interface name with "-qNNN" appended */
104 #define QUEUE_NAME_SIZE (IFNAMSIZ + 5)
106 /* IRQ name is queue name with "-tx" or "-rx" appended */
107 #define IRQ_NAME_SIZE (QUEUE_NAME_SIZE + 3)
111 struct xenvif_stats {
112 /* Stats fields to be updated per-queue.
113 * A subset of struct net_device_stats that contains only the
114 * fields that are updated in netback.c for each queue.
116 unsigned int rx_bytes;
117 unsigned int rx_packets;
118 unsigned int tx_bytes;
119 unsigned int tx_packets;
121 /* Additional stats used by xenvif */
122 unsigned long rx_gso_checksum_fixup;
123 unsigned long tx_zerocopy_sent;
124 unsigned long tx_zerocopy_success;
125 unsigned long tx_zerocopy_fail;
126 unsigned long tx_frag_overflow;
129 #define COPY_BATCH_SIZE 64
131 struct xenvif_copy_state {
132 struct gnttab_copy op[COPY_BATCH_SIZE];
133 RING_IDX idx[COPY_BATCH_SIZE];
137 struct xenvif_queue { /* Per-queue data for xenvif */
138 unsigned int id; /* Queue ID, 0-based */
139 char name[QUEUE_NAME_SIZE]; /* DEVNAME-qN */
140 struct xenvif *vif; /* Parent VIF */
142 /* Use NAPI for guest TX */
143 struct napi_struct napi;
144 /* When feature-split-event-channels = 0, tx_irq = rx_irq. */
146 /* Only used when feature-split-event-channels = 1 */
147 char tx_irq_name[IRQ_NAME_SIZE]; /* DEVNAME-qN-tx */
148 struct xen_netif_tx_back_ring tx;
149 struct sk_buff_head tx_queue;
150 struct page *mmap_pages[MAX_PENDING_REQS];
151 pending_ring_idx_t pending_prod;
152 pending_ring_idx_t pending_cons;
153 u16 pending_ring[MAX_PENDING_REQS];
154 struct pending_tx_info pending_tx_info[MAX_PENDING_REQS];
155 grant_handle_t grant_tx_handle[MAX_PENDING_REQS];
157 struct gnttab_copy tx_copy_ops[MAX_PENDING_REQS];
158 struct gnttab_map_grant_ref tx_map_ops[MAX_PENDING_REQS];
159 struct gnttab_unmap_grant_ref tx_unmap_ops[MAX_PENDING_REQS];
160 /* passed to gnttab_[un]map_refs with pages under (un)mapping */
161 struct page *pages_to_map[MAX_PENDING_REQS];
162 struct page *pages_to_unmap[MAX_PENDING_REQS];
164 /* This prevents zerocopy callbacks to race over dealloc_ring */
165 spinlock_t callback_lock;
166 /* This prevents dealloc thread and NAPI instance to race over response
167 * creation and pending_ring in xenvif_idx_release. In xenvif_tx_err
168 * it only protect response creation
170 spinlock_t response_lock;
171 pending_ring_idx_t dealloc_prod;
172 pending_ring_idx_t dealloc_cons;
173 u16 dealloc_ring[MAX_PENDING_REQS];
174 struct task_struct *dealloc_task;
175 wait_queue_head_t dealloc_wq;
176 atomic_t inflight_packets;
178 /* Use kthread for guest RX */
179 struct task_struct *task;
180 wait_queue_head_t wq;
181 /* When feature-split-event-channels = 0, tx_irq = rx_irq. */
183 /* Only used when feature-split-event-channels = 1 */
184 char rx_irq_name[IRQ_NAME_SIZE]; /* DEVNAME-qN-rx */
185 struct xen_netif_rx_back_ring rx;
186 struct sk_buff_head rx_queue;
188 unsigned int rx_queue_max;
189 unsigned int rx_queue_len;
190 unsigned long last_rx_time;
193 struct xenvif_copy_state rx_copy;
195 /* Transmit shaping: allow 'credit_bytes' every 'credit_usec'. */
196 unsigned long credit_bytes;
197 unsigned long credit_usec;
198 unsigned long remaining_credit;
199 struct timer_list credit_timeout;
200 u64 credit_window_start;
203 struct xenvif_stats stats;
206 enum state_bit_shift {
207 /* This bit marks that the vif is connected */
208 VIF_STATUS_CONNECTED,
211 struct xenvif_mcast_addr {
212 struct list_head entry;
217 #define XEN_NETBK_MCAST_MAX 64
219 #define XEN_NETBK_MAX_HASH_KEY_SIZE 40
220 #define XEN_NETBK_MAX_HASH_MAPPING_SIZE 128
221 #define XEN_NETBK_HASH_TAG_SIZE 40
223 struct xenvif_hash_cache_entry {
224 struct list_head link;
226 u8 tag[XEN_NETBK_HASH_TAG_SIZE];
232 struct xenvif_hash_cache {
234 struct list_head list;
242 u8 key[XEN_NETBK_MAX_HASH_KEY_SIZE];
243 u32 mapping[XEN_NETBK_MAX_HASH_MAPPING_SIZE];
245 struct xenvif_hash_cache cache;
249 /* Unique identifier for this interface. */
254 struct list_head fe_mcast_addr;
255 unsigned int fe_mcast_count;
257 /* Frontend feature information. */
263 u8 multicast_control:1;
265 /* Is this interface disabled? True when backend discovers
269 unsigned long status;
270 unsigned long drain_timeout;
271 unsigned long stall_timeout;
274 struct xenvif_queue *queues;
275 unsigned int num_queues; /* active queues, resource allocated */
276 unsigned int stalled_queues;
278 struct xenvif_hash hash;
280 struct xenbus_watch credit_watch;
281 struct xenbus_watch mcast_ctrl_watch;
285 #ifdef CONFIG_DEBUG_FS
286 struct dentry *xenvif_dbg_root;
289 struct xen_netif_ctrl_back_ring ctrl;
290 unsigned int ctrl_irq;
292 /* Miscellaneous private stuff. */
293 struct net_device *dev;
296 struct xenvif_rx_cb {
297 unsigned long expires;
301 #define XENVIF_RX_CB(skb) ((struct xenvif_rx_cb *)(skb)->cb)
303 static inline struct xenbus_device *xenvif_to_xenbus_device(struct xenvif *vif)
305 return to_xenbus_device(vif->dev->dev.parent);
308 void xenvif_tx_credit_callback(unsigned long data);
310 struct xenvif *xenvif_alloc(struct device *parent,
312 unsigned int handle);
314 int xenvif_init_queue(struct xenvif_queue *queue);
315 void xenvif_deinit_queue(struct xenvif_queue *queue);
317 int xenvif_connect_data(struct xenvif_queue *queue,
318 unsigned long tx_ring_ref,
319 unsigned long rx_ring_ref,
320 unsigned int tx_evtchn,
321 unsigned int rx_evtchn);
322 void xenvif_disconnect_data(struct xenvif *vif);
323 int xenvif_connect_ctrl(struct xenvif *vif, grant_ref_t ring_ref,
324 unsigned int evtchn);
325 void xenvif_disconnect_ctrl(struct xenvif *vif);
326 void xenvif_free(struct xenvif *vif);
328 int xenvif_xenbus_init(void);
329 void xenvif_xenbus_fini(void);
331 int xenvif_schedulable(struct xenvif *vif);
333 int xenvif_queue_stopped(struct xenvif_queue *queue);
334 void xenvif_wake_queue(struct xenvif_queue *queue);
336 /* (Un)Map communication rings. */
337 void xenvif_unmap_frontend_data_rings(struct xenvif_queue *queue);
338 int xenvif_map_frontend_data_rings(struct xenvif_queue *queue,
339 grant_ref_t tx_ring_ref,
340 grant_ref_t rx_ring_ref);
342 /* Check for SKBs from frontend and schedule backend processing */
343 void xenvif_napi_schedule_or_enable_events(struct xenvif_queue *queue);
345 /* Prevent the device from generating any further traffic. */
346 void xenvif_carrier_off(struct xenvif *vif);
348 int xenvif_tx_action(struct xenvif_queue *queue, int budget);
350 int xenvif_kthread_guest_rx(void *data);
351 void xenvif_kick_thread(struct xenvif_queue *queue);
353 int xenvif_dealloc_kthread(void *data);
355 irqreturn_t xenvif_ctrl_irq_fn(int irq, void *data);
357 void xenvif_rx_action(struct xenvif_queue *queue);
358 void xenvif_rx_queue_tail(struct xenvif_queue *queue, struct sk_buff *skb);
360 void xenvif_carrier_on(struct xenvif *vif);
362 /* Callback from stack when TX packet can be released */
363 void xenvif_zerocopy_callback(struct ubuf_info *ubuf, bool zerocopy_success);
365 /* Unmap a pending page and release it back to the guest */
366 void xenvif_idx_unmap(struct xenvif_queue *queue, u16 pending_idx);
368 static inline pending_ring_idx_t nr_pending_reqs(struct xenvif_queue *queue)
370 return MAX_PENDING_REQS -
371 queue->pending_prod + queue->pending_cons;
374 irqreturn_t xenvif_interrupt(int irq, void *dev_id);
376 extern bool separate_tx_rx_irq;
378 extern unsigned int rx_drain_timeout_msecs;
379 extern unsigned int rx_stall_timeout_msecs;
380 extern unsigned int xenvif_max_queues;
381 extern unsigned int xenvif_hash_cache_size;
383 #ifdef CONFIG_DEBUG_FS
384 extern struct dentry *xen_netback_dbg_root;
387 void xenvif_skb_zerocopy_prepare(struct xenvif_queue *queue,
388 struct sk_buff *skb);
389 void xenvif_skb_zerocopy_complete(struct xenvif_queue *queue);
391 /* Multicast control */
392 bool xenvif_mcast_match(struct xenvif *vif, const u8 *addr);
393 void xenvif_mcast_addr_list_free(struct xenvif *vif);
396 void xenvif_init_hash(struct xenvif *vif);
397 void xenvif_deinit_hash(struct xenvif *vif);
399 u32 xenvif_set_hash_alg(struct xenvif *vif, u32 alg);
400 u32 xenvif_get_hash_flags(struct xenvif *vif, u32 *flags);
401 u32 xenvif_set_hash_flags(struct xenvif *vif, u32 flags);
402 u32 xenvif_set_hash_key(struct xenvif *vif, u32 gref, u32 len);
403 u32 xenvif_set_hash_mapping_size(struct xenvif *vif, u32 size);
404 u32 xenvif_set_hash_mapping(struct xenvif *vif, u32 gref, u32 len,
407 void xenvif_set_skb_hash(struct xenvif *vif, struct sk_buff *skb);
409 #endif /* __XEN_NETBACK__COMMON_H__ */