1 /*
2  * Copyright 2015 Amazon.com, Inc. or its affiliates.
3  *
4  * This software is available to you under a choice of one of two
5  * licenses.  You may choose to be licensed under the terms of the GNU
6  * General Public License (GPL) Version 2, available from the file
7  * COPYING in the main directory of this source tree, or the
8  * BSD license below:
9  *
10  *     Redistribution and use in source and binary forms, with or
11  *     without modification, are permitted provided that the following
12  *     conditions are met:
13  *
14  *      - Redistributions of source code must retain the above
15  *        copyright notice, this list of conditions and the following
16  *        disclaimer.
17  *
18  *      - Redistributions in binary form must reproduce the above
19  *        copyright notice, this list of conditions and the following
20  *        disclaimer in the documentation and/or other materials
21  *        provided with the distribution.
22  *
23  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
30  * SOFTWARE.
31  */
32 
33 #ifndef ENA_H
34 #define ENA_H
35 
36 #include <linux/bitops.h>
37 #include <linux/dim.h>
38 #include <linux/etherdevice.h>
39 #include <linux/if_vlan.h>
40 #include <linux/inetdevice.h>
41 #include <linux/interrupt.h>
42 #include <linux/netdevice.h>
43 #include <linux/skbuff.h>
44 
45 #include "ena_com.h"
46 #include "ena_eth_com.h"
47 
48 #define DRV_MODULE_VER_MAJOR	2
49 #define DRV_MODULE_VER_MINOR	1
50 #define DRV_MODULE_VER_SUBMINOR 0
51 
52 #define DRV_MODULE_NAME		"ena"
53 #ifndef DRV_MODULE_VERSION
54 #define DRV_MODULE_VERSION \
55 	__stringify(DRV_MODULE_VER_MAJOR) "."	\
56 	__stringify(DRV_MODULE_VER_MINOR) "."	\
57 	__stringify(DRV_MODULE_VER_SUBMINOR) "K"
58 #endif
59 
60 #define DEVICE_NAME	"Elastic Network Adapter (ENA)"
61 
62 /* 1 for AENQ + ADMIN */
63 #define ENA_ADMIN_MSIX_VEC		1
64 #define ENA_MAX_MSIX_VEC(io_queues)	(ENA_ADMIN_MSIX_VEC + (io_queues))
65 
66 /* The ENA buffer length fields is 16 bit long. So when PAGE_SIZE == 64kB the
67  * driver passes 0.
68  * Since the max packet size the ENA handles is ~9kB limit the buffer length to
69  * 16kB.
70  */
71 #if PAGE_SIZE > SZ_16K
72 #define ENA_PAGE_SIZE SZ_16K
73 #else
74 #define ENA_PAGE_SIZE PAGE_SIZE
75 #endif
76 
77 #define ENA_MIN_MSIX_VEC		2
78 
79 #define ENA_REG_BAR			0
80 #define ENA_MEM_BAR			2
81 #define ENA_BAR_MASK (BIT(ENA_REG_BAR) | BIT(ENA_MEM_BAR))
82 
83 #define ENA_DEFAULT_RING_SIZE	(1024)
84 #define ENA_MIN_RING_SIZE	(256)
85 
86 #define ENA_MIN_NUM_IO_QUEUES	(1)
87 
88 #define ENA_TX_WAKEUP_THRESH		(MAX_SKB_FRAGS + 2)
89 #define ENA_DEFAULT_RX_COPYBREAK	(256 - NET_IP_ALIGN)
90 
91 /* limit the buffer size to 600 bytes to handle MTU changes from very
92  * small to very large, in which case the number of buffers per packet
93  * could exceed ENA_PKT_MAX_BUFS
94  */
95 #define ENA_DEFAULT_MIN_RX_BUFF_ALLOC_SIZE 600
96 
97 #define ENA_MIN_MTU		128
98 
99 #define ENA_NAME_MAX_LEN	20
100 #define ENA_IRQNAME_SIZE	40
101 
102 #define ENA_PKT_MAX_BUFS	19
103 
104 #define ENA_RX_RSS_TABLE_LOG_SIZE  7
105 #define ENA_RX_RSS_TABLE_SIZE	(1 << ENA_RX_RSS_TABLE_LOG_SIZE)
106 
107 #define ENA_HASH_KEY_SIZE	40
108 
109 /* The number of tx packet completions that will be handled each NAPI poll
110  * cycle is ring_size / ENA_TX_POLL_BUDGET_DIVIDER.
111  */
112 #define ENA_TX_POLL_BUDGET_DIVIDER	4
113 
114 /* Refill Rx queue when number of required descriptors is above
115  * QUEUE_SIZE / ENA_RX_REFILL_THRESH_DIVIDER or ENA_RX_REFILL_THRESH_PACKET
116  */
117 #define ENA_RX_REFILL_THRESH_DIVIDER	8
118 #define ENA_RX_REFILL_THRESH_PACKET	256
119 
120 /* Number of queues to check for missing queues per timer service */
121 #define ENA_MONITORED_TX_QUEUES	4
122 /* Max timeout packets before device reset */
123 #define MAX_NUM_OF_TIMEOUTED_PACKETS 128
124 
125 #define ENA_TX_RING_IDX_NEXT(idx, ring_size) (((idx) + 1) & ((ring_size) - 1))
126 
127 #define ENA_RX_RING_IDX_NEXT(idx, ring_size) (((idx) + 1) & ((ring_size) - 1))
128 #define ENA_RX_RING_IDX_ADD(idx, n, ring_size) \
129 	(((idx) + (n)) & ((ring_size) - 1))
130 
131 #define ENA_IO_TXQ_IDX(q)	(2 * (q))
132 #define ENA_IO_RXQ_IDX(q)	(2 * (q) + 1)
133 
134 #define ENA_MGMNT_IRQ_IDX		0
135 #define ENA_IO_IRQ_FIRST_IDX		1
136 #define ENA_IO_IRQ_IDX(q)		(ENA_IO_IRQ_FIRST_IDX + (q))
137 
138 /* ENA device should send keep alive msg every 1 sec.
139  * We wait for 6 sec just to be on the safe side.
140  */
141 #define ENA_DEVICE_KALIVE_TIMEOUT	(6 * HZ)
142 #define ENA_MAX_NO_INTERRUPT_ITERATIONS 3
143 
144 #define ENA_MMIO_DISABLE_REG_READ	BIT(0)
145 
146 /* The max MTU size is configured to be the ethernet frame size without
147  * the overhead of the ethernet header, which can have a VLAN header, and
148  * a frame check sequence (FCS).
149  * The buffer size we share with the device is defined to be ENA_PAGE_SIZE
150  */
151 
152 #define ENA_XDP_MAX_MTU (ENA_PAGE_SIZE - ETH_HLEN - ETH_FCS_LEN - \
153 				VLAN_HLEN - XDP_PACKET_HEADROOM)
154 
155 #define ENA_IS_XDP_INDEX(adapter, index) (((index) >= (adapter)->xdp_first_ring) && \
156 	((index) < (adapter)->xdp_first_ring + (adapter)->xdp_num_queues))
157 
158 struct ena_irq {
159 	irq_handler_t handler;
160 	void *data;
161 	int cpu;
162 	u32 vector;
163 	cpumask_t affinity_hint_mask;
164 	char name[ENA_IRQNAME_SIZE];
165 };
166 
167 struct ena_napi {
168 	struct napi_struct napi ____cacheline_aligned;
169 	struct ena_ring *tx_ring;
170 	struct ena_ring *rx_ring;
171 	struct ena_ring *xdp_ring;
172 	bool first_interrupt;
173 	u32 qid;
174 	struct dim dim;
175 };
176 
177 struct ena_calc_queue_size_ctx {
178 	struct ena_com_dev_get_features_ctx *get_feat_ctx;
179 	struct ena_com_dev *ena_dev;
180 	struct pci_dev *pdev;
181 	u32 tx_queue_size;
182 	u32 rx_queue_size;
183 	u32 max_tx_queue_size;
184 	u32 max_rx_queue_size;
185 	u16 max_tx_sgl_size;
186 	u16 max_rx_sgl_size;
187 };
188 
189 struct ena_tx_buffer {
190 	struct sk_buff *skb;
191 	/* num of ena desc for this specific skb
192 	 * (includes data desc and metadata desc)
193 	 */
194 	u32 tx_descs;
195 	/* num of buffers used by this skb */
196 	u32 num_of_bufs;
197 
198 	/* XDP buffer structure which is used for sending packets in
199 	 * the xdp queues
200 	 */
201 	struct xdp_frame *xdpf;
202 	/* The rx page for the rx buffer that was received in rx and
203 	 * re transmitted on xdp tx queues as a result of XDP_TX action.
204 	 * We need to free the page once we finished cleaning the buffer in
205 	 * clean_xdp_irq()
206 	 */
207 	struct page *xdp_rx_page;
208 
209 	/* Indicate if bufs[0] map the linear data of the skb. */
210 	u8 map_linear_data;
211 
212 	/* Used for detect missing tx packets to limit the number of prints */
213 	u32 print_once;
214 	/* Save the last jiffies to detect missing tx packets
215 	 *
216 	 * sets to non zero value on ena_start_xmit and set to zero on
217 	 * napi and timer_Service_routine.
218 	 *
219 	 * while this value is not protected by lock,
220 	 * a given packet is not expected to be handled by ena_start_xmit
221 	 * and by napi/timer_service at the same time.
222 	 */
223 	unsigned long last_jiffies;
224 	struct ena_com_buf bufs[ENA_PKT_MAX_BUFS];
225 } ____cacheline_aligned;
226 
227 struct ena_rx_buffer {
228 	struct sk_buff *skb;
229 	struct page *page;
230 	u32 page_offset;
231 	struct ena_com_buf ena_buf;
232 } ____cacheline_aligned;
233 
234 struct ena_stats_tx {
235 	u64 cnt;
236 	u64 bytes;
237 	u64 queue_stop;
238 	u64 prepare_ctx_err;
239 	u64 queue_wakeup;
240 	u64 dma_mapping_err;
241 	u64 linearize;
242 	u64 linearize_failed;
243 	u64 napi_comp;
244 	u64 tx_poll;
245 	u64 doorbells;
246 	u64 bad_req_id;
247 	u64 llq_buffer_copy;
248 	u64 missed_tx;
249 };
250 
251 struct ena_stats_rx {
252 	u64 cnt;
253 	u64 bytes;
254 	u64 rx_copybreak_pkt;
255 	u64 csum_good;
256 	u64 refil_partial;
257 	u64 bad_csum;
258 	u64 page_alloc_fail;
259 	u64 skb_alloc_fail;
260 	u64 dma_mapping_err;
261 	u64 bad_desc_num;
262 	u64 bad_req_id;
263 	u64 empty_rx_ring;
264 	u64 csum_unchecked;
265 };
266 
267 struct ena_ring {
268 	/* Holds the empty requests for TX/RX
269 	 * out of order completions
270 	 */
271 	u16 *free_ids;
272 
273 	union {
274 		struct ena_tx_buffer *tx_buffer_info;
275 		struct ena_rx_buffer *rx_buffer_info;
276 	};
277 
278 	/* cache ptr to avoid using the adapter */
279 	struct device *dev;
280 	struct pci_dev *pdev;
281 	struct napi_struct *napi;
282 	struct net_device *netdev;
283 	struct ena_com_dev *ena_dev;
284 	struct ena_adapter *adapter;
285 	struct ena_com_io_cq *ena_com_io_cq;
286 	struct ena_com_io_sq *ena_com_io_sq;
287 	struct bpf_prog *xdp_bpf_prog;
288 	struct xdp_rxq_info xdp_rxq;
289 
290 	u16 next_to_use;
291 	u16 next_to_clean;
292 	u16 rx_copybreak;
293 	u16 rx_headroom;
294 	u16 qid;
295 	u16 mtu;
296 	u16 sgl_size;
297 
298 	/* The maximum header length the device can handle */
299 	u8 tx_max_header_size;
300 
301 	bool first_interrupt;
302 	u16 no_interrupt_event_cnt;
303 
304 	/* cpu for TPH */
305 	int cpu;
306 	 /* number of tx/rx_buffer_info's entries */
307 	int ring_size;
308 
309 	enum ena_admin_placement_policy_type tx_mem_queue_type;
310 
311 	struct ena_com_rx_buf_info ena_bufs[ENA_PKT_MAX_BUFS];
312 	u32  smoothed_interval;
313 	u32  per_napi_packets;
314 	u16 non_empty_napi_events;
315 	struct u64_stats_sync syncp;
316 	union {
317 		struct ena_stats_tx tx_stats;
318 		struct ena_stats_rx rx_stats;
319 	};
320 
321 	u8 *push_buf_intermediate_buf;
322 	int empty_rx_queue;
323 } ____cacheline_aligned;
324 
325 struct ena_stats_dev {
326 	u64 tx_timeout;
327 	u64 suspend;
328 	u64 resume;
329 	u64 wd_expired;
330 	u64 interface_up;
331 	u64 interface_down;
332 	u64 admin_q_pause;
333 	u64 rx_drops;
334 };
335 
336 enum ena_flags_t {
337 	ENA_FLAG_DEVICE_RUNNING,
338 	ENA_FLAG_DEV_UP,
339 	ENA_FLAG_LINK_UP,
340 	ENA_FLAG_MSIX_ENABLED,
341 	ENA_FLAG_TRIGGER_RESET,
342 	ENA_FLAG_ONGOING_RESET
343 };
344 
345 /* adapter specific private data structure */
346 struct ena_adapter {
347 	struct ena_com_dev *ena_dev;
348 	/* OS defined structs */
349 	struct net_device *netdev;
350 	struct pci_dev *pdev;
351 
352 	/* rx packets that shorter that this len will be copied to the skb
353 	 * header
354 	 */
355 	u32 rx_copybreak;
356 	u32 max_mtu;
357 
358 	u32 num_io_queues;
359 	u32 max_num_io_queues;
360 
361 	int msix_vecs;
362 
363 	u32 missing_tx_completion_threshold;
364 
365 	u32 requested_tx_ring_size;
366 	u32 requested_rx_ring_size;
367 
368 	u32 max_tx_ring_size;
369 	u32 max_rx_ring_size;
370 
371 	u32 msg_enable;
372 
373 	u16 max_tx_sgl_size;
374 	u16 max_rx_sgl_size;
375 
376 	u8 mac_addr[ETH_ALEN];
377 
378 	unsigned long keep_alive_timeout;
379 	unsigned long missing_tx_completion_to;
380 
381 	char name[ENA_NAME_MAX_LEN];
382 
383 	unsigned long flags;
384 	/* TX */
385 	struct ena_ring tx_ring[ENA_MAX_NUM_IO_QUEUES]
386 		____cacheline_aligned_in_smp;
387 
388 	/* RX */
389 	struct ena_ring rx_ring[ENA_MAX_NUM_IO_QUEUES]
390 		____cacheline_aligned_in_smp;
391 
392 	struct ena_napi ena_napi[ENA_MAX_NUM_IO_QUEUES];
393 
394 	struct ena_irq irq_tbl[ENA_MAX_MSIX_VEC(ENA_MAX_NUM_IO_QUEUES)];
395 
396 	/* timer service */
397 	struct work_struct reset_task;
398 	struct timer_list timer_service;
399 
400 	bool wd_state;
401 	bool dev_up_before_reset;
402 	unsigned long last_keep_alive_jiffies;
403 
404 	struct u64_stats_sync syncp;
405 	struct ena_stats_dev dev_stats;
406 
407 	/* last queue index that was checked for uncompleted tx packets */
408 	u32 last_monitored_tx_qid;
409 
410 	enum ena_regs_reset_reason_types reset_reason;
411 
412 	struct bpf_prog *xdp_bpf_prog;
413 	u32 xdp_first_ring;
414 	u32 xdp_num_queues;
415 };
416 
417 void ena_set_ethtool_ops(struct net_device *netdev);
418 
419 void ena_dump_stats_to_dmesg(struct ena_adapter *adapter);
420 
421 void ena_dump_stats_to_buf(struct ena_adapter *adapter, u8 *buf);
422 
423 int ena_update_queue_sizes(struct ena_adapter *adapter,
424 			   u32 new_tx_size,
425 			   u32 new_rx_size);
426 
427 int ena_update_queue_count(struct ena_adapter *adapter, u32 new_channel_count);
428 
429 int ena_get_sset_count(struct net_device *netdev, int sset);
430 
431 enum ena_xdp_errors_t {
432 	ENA_XDP_ALLOWED = 0,
433 	ENA_XDP_CURRENT_MTU_TOO_LARGE,
434 	ENA_XDP_NO_ENOUGH_QUEUES,
435 };
436 
437 static inline bool ena_xdp_queues_present(struct ena_adapter *adapter)
438 {
439 	return adapter->xdp_first_ring != 0;
440 }
441 
442 static inline bool ena_xdp_present(struct ena_adapter *adapter)
443 {
444 	return !!adapter->xdp_bpf_prog;
445 }
446 
447 static inline bool ena_xdp_present_ring(struct ena_ring *ring)
448 {
449 	return !!ring->xdp_bpf_prog;
450 }
451 
452 static inline int ena_xdp_legal_queue_count(struct ena_adapter *adapter,
453 					    u32 queues)
454 {
455 	return 2 * queues <= adapter->max_num_io_queues;
456 }
457 
458 static inline enum ena_xdp_errors_t ena_xdp_allowed(struct ena_adapter *adapter)
459 {
460 	enum ena_xdp_errors_t rc = ENA_XDP_ALLOWED;
461 
462 	if (adapter->netdev->mtu > ENA_XDP_MAX_MTU)
463 		rc = ENA_XDP_CURRENT_MTU_TOO_LARGE;
464 	else if (!ena_xdp_legal_queue_count(adapter, adapter->num_io_queues))
465 		rc = ENA_XDP_NO_ENOUGH_QUEUES;
466 
467 	return rc;
468 }
469 
470 #endif /* !(ENA_H) */
471