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 /* It's possible for an skb to have a maximal number of frags
95 * but still be less than MAX_BUFFER_OFFSET in size. Thus the
96 * worst-case number of copy operations is MAX_XEN_SKB_FRAGS per
99 #define MAX_GRANT_COPY_OPS (MAX_XEN_SKB_FRAGS * XEN_NETIF_RX_RING_SIZE)
101 #define NETBACK_INVALID_HANDLE -1
103 /* To avoid confusion, we define XEN_NETBK_LEGACY_SLOTS_MAX indicating
104 * the maximum slots a valid packet can use. Now this value is defined
105 * to be XEN_NETIF_NR_SLOTS_MIN, which is supposed to be supported by
108 #define XEN_NETBK_LEGACY_SLOTS_MAX XEN_NETIF_NR_SLOTS_MIN
110 /* Queue name is interface name with "-qNNN" appended */
111 #define QUEUE_NAME_SIZE (IFNAMSIZ + 5)
113 /* IRQ name is queue name with "-tx" or "-rx" appended */
114 #define IRQ_NAME_SIZE (QUEUE_NAME_SIZE + 3)
118 struct xenvif_stats {
119 /* Stats fields to be updated per-queue.
120 * A subset of struct net_device_stats that contains only the
121 * fields that are updated in netback.c for each queue.
123 unsigned int rx_bytes;
124 unsigned int rx_packets;
125 unsigned int tx_bytes;
126 unsigned int tx_packets;
128 /* Additional stats used by xenvif */
129 unsigned long rx_gso_checksum_fixup;
130 unsigned long tx_zerocopy_sent;
131 unsigned long tx_zerocopy_success;
132 unsigned long tx_zerocopy_fail;
133 unsigned long tx_frag_overflow;
136 struct xenvif_queue { /* Per-queue data for xenvif */
137 unsigned int id; /* Queue ID, 0-based */
138 char name[QUEUE_NAME_SIZE]; /* DEVNAME-qN */
139 struct xenvif *vif; /* Parent VIF */
141 /* Use NAPI for guest TX */
142 struct napi_struct napi;
143 /* When feature-split-event-channels = 0, tx_irq = rx_irq. */
145 /* Only used when feature-split-event-channels = 1 */
146 char tx_irq_name[IRQ_NAME_SIZE]; /* DEVNAME-qN-tx */
147 struct xen_netif_tx_back_ring tx;
148 struct sk_buff_head tx_queue;
149 struct page *mmap_pages[MAX_PENDING_REQS];
150 pending_ring_idx_t pending_prod;
151 pending_ring_idx_t pending_cons;
152 u16 pending_ring[MAX_PENDING_REQS];
153 struct pending_tx_info pending_tx_info[MAX_PENDING_REQS];
154 grant_handle_t grant_tx_handle[MAX_PENDING_REQS];
156 struct gnttab_copy tx_copy_ops[MAX_PENDING_REQS];
157 struct gnttab_map_grant_ref tx_map_ops[MAX_PENDING_REQS];
158 struct gnttab_unmap_grant_ref tx_unmap_ops[MAX_PENDING_REQS];
159 /* passed to gnttab_[un]map_refs with pages under (un)mapping */
160 struct page *pages_to_map[MAX_PENDING_REQS];
161 struct page *pages_to_unmap[MAX_PENDING_REQS];
163 /* This prevents zerocopy callbacks to race over dealloc_ring */
164 spinlock_t callback_lock;
165 /* This prevents dealloc thread and NAPI instance to race over response
166 * creation and pending_ring in xenvif_idx_release. In xenvif_tx_err
167 * it only protect response creation
169 spinlock_t response_lock;
170 pending_ring_idx_t dealloc_prod;
171 pending_ring_idx_t dealloc_cons;
172 u16 dealloc_ring[MAX_PENDING_REQS];
173 struct task_struct *dealloc_task;
174 wait_queue_head_t dealloc_wq;
175 atomic_t inflight_packets;
177 /* Use kthread for guest RX */
178 struct task_struct *task;
179 wait_queue_head_t wq;
180 /* When feature-split-event-channels = 0, tx_irq = rx_irq. */
182 /* Only used when feature-split-event-channels = 1 */
183 char rx_irq_name[IRQ_NAME_SIZE]; /* DEVNAME-qN-rx */
184 struct xen_netif_rx_back_ring rx;
185 struct sk_buff_head rx_queue;
187 unsigned int rx_queue_max;
188 unsigned int rx_queue_len;
189 unsigned long last_rx_time;
192 struct gnttab_copy grant_copy_op[MAX_GRANT_COPY_OPS];
194 /* We create one meta structure per ring request we consume, so
195 * the maximum number is the same as the ring size.
197 struct xenvif_rx_meta meta[XEN_NETIF_RX_RING_SIZE];
199 /* Transmit shaping: allow 'credit_bytes' every 'credit_usec'. */
200 unsigned long credit_bytes;
201 unsigned long credit_usec;
202 unsigned long remaining_credit;
203 struct timer_list credit_timeout;
204 u64 credit_window_start;
207 struct xenvif_stats stats;
210 enum state_bit_shift {
211 /* This bit marks that the vif is connected */
212 VIF_STATUS_CONNECTED,
215 struct xenvif_mcast_addr {
216 struct list_head entry;
221 #define XEN_NETBK_MCAST_MAX 64
223 #define XEN_NETBK_MAX_HASH_KEY_SIZE 40
224 #define XEN_NETBK_MAX_HASH_MAPPING_SIZE 128
225 #define XEN_NETBK_HASH_TAG_SIZE 40
227 struct xenvif_hash_cache_entry {
228 struct list_head link;
230 u8 tag[XEN_NETBK_HASH_TAG_SIZE];
236 struct xenvif_hash_cache {
238 struct list_head list;
246 u8 key[XEN_NETBK_MAX_HASH_KEY_SIZE];
247 u32 mapping[XEN_NETBK_MAX_HASH_MAPPING_SIZE];
249 struct xenvif_hash_cache cache;
253 /* Unique identifier for this interface. */
258 struct list_head fe_mcast_addr;
259 unsigned int fe_mcast_count;
261 /* Frontend feature information. */
268 u8 multicast_control:1;
270 /* Is this interface disabled? True when backend discovers
274 unsigned long status;
275 unsigned long drain_timeout;
276 unsigned long stall_timeout;
279 struct xenvif_queue *queues;
280 unsigned int num_queues; /* active queues, resource allocated */
281 unsigned int stalled_queues;
283 struct xenvif_hash hash;
285 struct xenbus_watch credit_watch;
286 struct xenbus_watch mcast_ctrl_watch;
290 #ifdef CONFIG_DEBUG_FS
291 struct dentry *xenvif_dbg_root;
294 struct xen_netif_ctrl_back_ring ctrl;
295 unsigned int ctrl_irq;
297 /* Miscellaneous private stuff. */
298 struct net_device *dev;
301 struct xenvif_rx_cb {
302 unsigned long expires;
306 #define XENVIF_RX_CB(skb) ((struct xenvif_rx_cb *)(skb)->cb)
308 static inline struct xenbus_device *xenvif_to_xenbus_device(struct xenvif *vif)
310 return to_xenbus_device(vif->dev->dev.parent);
313 void xenvif_tx_credit_callback(unsigned long data);
315 struct xenvif *xenvif_alloc(struct device *parent,
317 unsigned int handle);
319 int xenvif_init_queue(struct xenvif_queue *queue);
320 void xenvif_deinit_queue(struct xenvif_queue *queue);
322 int xenvif_connect_data(struct xenvif_queue *queue,
323 unsigned long tx_ring_ref,
324 unsigned long rx_ring_ref,
325 unsigned int tx_evtchn,
326 unsigned int rx_evtchn);
327 void xenvif_disconnect_data(struct xenvif *vif);
328 int xenvif_connect_ctrl(struct xenvif *vif, grant_ref_t ring_ref,
329 unsigned int evtchn);
330 void xenvif_disconnect_ctrl(struct xenvif *vif);
331 void xenvif_free(struct xenvif *vif);
333 int xenvif_xenbus_init(void);
334 void xenvif_xenbus_fini(void);
336 int xenvif_schedulable(struct xenvif *vif);
338 int xenvif_queue_stopped(struct xenvif_queue *queue);
339 void xenvif_wake_queue(struct xenvif_queue *queue);
341 /* (Un)Map communication rings. */
342 void xenvif_unmap_frontend_data_rings(struct xenvif_queue *queue);
343 int xenvif_map_frontend_data_rings(struct xenvif_queue *queue,
344 grant_ref_t tx_ring_ref,
345 grant_ref_t rx_ring_ref);
347 /* Check for SKBs from frontend and schedule backend processing */
348 void xenvif_napi_schedule_or_enable_events(struct xenvif_queue *queue);
350 /* Prevent the device from generating any further traffic. */
351 void xenvif_carrier_off(struct xenvif *vif);
353 int xenvif_tx_action(struct xenvif_queue *queue, int budget);
355 int xenvif_kthread_guest_rx(void *data);
356 void xenvif_kick_thread(struct xenvif_queue *queue);
358 int xenvif_dealloc_kthread(void *data);
360 irqreturn_t xenvif_ctrl_irq_fn(int irq, void *data);
362 void xenvif_rx_queue_tail(struct xenvif_queue *queue, struct sk_buff *skb);
364 void xenvif_carrier_on(struct xenvif *vif);
366 /* Callback from stack when TX packet can be released */
367 void xenvif_zerocopy_callback(struct ubuf_info *ubuf, bool zerocopy_success);
369 /* Unmap a pending page and release it back to the guest */
370 void xenvif_idx_unmap(struct xenvif_queue *queue, u16 pending_idx);
372 static inline pending_ring_idx_t nr_pending_reqs(struct xenvif_queue *queue)
374 return MAX_PENDING_REQS -
375 queue->pending_prod + queue->pending_cons;
378 irqreturn_t xenvif_interrupt(int irq, void *dev_id);
380 extern bool separate_tx_rx_irq;
382 extern unsigned int rx_drain_timeout_msecs;
383 extern unsigned int rx_stall_timeout_msecs;
384 extern unsigned int xenvif_max_queues;
385 extern unsigned int xenvif_hash_cache_size;
387 #ifdef CONFIG_DEBUG_FS
388 extern struct dentry *xen_netback_dbg_root;
391 void xenvif_skb_zerocopy_prepare(struct xenvif_queue *queue,
392 struct sk_buff *skb);
393 void xenvif_skb_zerocopy_complete(struct xenvif_queue *queue);
395 /* Multicast control */
396 bool xenvif_mcast_match(struct xenvif *vif, const u8 *addr);
397 void xenvif_mcast_addr_list_free(struct xenvif *vif);
400 void xenvif_init_hash(struct xenvif *vif);
401 void xenvif_deinit_hash(struct xenvif *vif);
403 u32 xenvif_set_hash_alg(struct xenvif *vif, u32 alg);
404 u32 xenvif_get_hash_flags(struct xenvif *vif, u32 *flags);
405 u32 xenvif_set_hash_flags(struct xenvif *vif, u32 flags);
406 u32 xenvif_set_hash_key(struct xenvif *vif, u32 gref, u32 len);
407 u32 xenvif_set_hash_mapping_size(struct xenvif *vif, u32 size);
408 u32 xenvif_set_hash_mapping(struct xenvif *vif, u32 gref, u32 len,
411 void xenvif_set_skb_hash(struct xenvif *vif, struct sk_buff *skb);
413 #endif /* __XEN_NETBACK__COMMON_H__ */