1 /*******************************************************************************
3 * Intel Ethernet Controller XL710 Family Linux Virtual Function Driver
4 * Copyright(c) 2013 - 2016 Intel Corporation.
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms and conditions of the GNU General Public License,
8 * version 2, as published by the Free Software Foundation.
10 * This program is distributed in the hope it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
15 * You should have received a copy of the GNU General Public License along
16 * with this program. If not, see <http://www.gnu.org/licenses/>.
18 * The full GNU General Public License is included in this distribution in
19 * the file called "COPYING".
21 * Contact Information:
22 * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
23 * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
25 ******************************************************************************/
30 /* Interrupt Throttling and Rate Limiting Goodies */
32 #define I40E_MAX_ITR 0x0FF0 /* reg uses 2 usec resolution */
33 #define I40E_MIN_ITR 0x0001 /* reg uses 2 usec resolution */
34 #define I40E_ITR_100K 0x0005
35 #define I40E_ITR_50K 0x000A
36 #define I40E_ITR_20K 0x0019
37 #define I40E_ITR_18K 0x001B
38 #define I40E_ITR_8K 0x003E
39 #define I40E_ITR_4K 0x007A
40 #define I40E_MAX_INTRL 0x3B /* reg uses 4 usec resolution */
41 #define I40E_ITR_RX_DEF I40E_ITR_20K
42 #define I40E_ITR_TX_DEF I40E_ITR_20K
43 #define I40E_ITR_DYNAMIC 0x8000 /* use top bit as a flag */
44 #define I40E_MIN_INT_RATE 250 /* ~= 1000000 / (I40E_MAX_ITR * 2) */
45 #define I40E_MAX_INT_RATE 500000 /* == 1000000 / (I40E_MIN_ITR * 2) */
46 #define I40E_DEFAULT_IRQ_WORK 256
47 #define ITR_TO_REG(setting) ((setting & ~I40E_ITR_DYNAMIC) >> 1)
48 #define ITR_IS_DYNAMIC(setting) (!!(setting & I40E_ITR_DYNAMIC))
49 #define ITR_REG_TO_USEC(itr_reg) (itr_reg << 1)
50 /* 0x40 is the enable bit for interrupt rate limiting, and must be set if
51 * the value of the rate limit is non-zero
53 #define INTRL_ENA BIT(6)
54 #define INTRL_REG_TO_USEC(intrl) ((intrl & ~INTRL_ENA) << 2)
55 #define INTRL_USEC_TO_REG(set) ((set) ? ((set) >> 2) | INTRL_ENA : 0)
56 #define I40E_INTRL_8K 125 /* 8000 ints/sec */
57 #define I40E_INTRL_62K 16 /* 62500 ints/sec */
58 #define I40E_INTRL_83K 12 /* 83333 ints/sec */
60 #define I40E_QUEUE_END_OF_LIST 0x7FF
62 /* this enum matches hardware bits and is meant to be used by DYN_CTLN
63 * registers and QINT registers or more generally anywhere in the manual
64 * mentioning ITR_INDX, ITR_NONE cannot be used as an index 'n' into any
65 * register but instead is a special value meaning "don't update" ITR0/1/2.
71 I40E_ITR_NONE = 3 /* ITR_NONE must not be used as an index */
74 /* these are indexes into ITRN registers */
75 #define I40E_RX_ITR I40E_IDX_ITR0
76 #define I40E_TX_ITR I40E_IDX_ITR1
77 #define I40E_PE_ITR I40E_IDX_ITR2
79 /* Supported RSS offloads */
80 #define I40E_DEFAULT_RSS_HENA ( \
81 BIT_ULL(I40E_FILTER_PCTYPE_NONF_IPV4_UDP) | \
82 BIT_ULL(I40E_FILTER_PCTYPE_NONF_IPV4_SCTP) | \
83 BIT_ULL(I40E_FILTER_PCTYPE_NONF_IPV4_TCP) | \
84 BIT_ULL(I40E_FILTER_PCTYPE_NONF_IPV4_OTHER) | \
85 BIT_ULL(I40E_FILTER_PCTYPE_FRAG_IPV4) | \
86 BIT_ULL(I40E_FILTER_PCTYPE_NONF_IPV6_UDP) | \
87 BIT_ULL(I40E_FILTER_PCTYPE_NONF_IPV6_TCP) | \
88 BIT_ULL(I40E_FILTER_PCTYPE_NONF_IPV6_SCTP) | \
89 BIT_ULL(I40E_FILTER_PCTYPE_NONF_IPV6_OTHER) | \
90 BIT_ULL(I40E_FILTER_PCTYPE_FRAG_IPV6) | \
91 BIT_ULL(I40E_FILTER_PCTYPE_L2_PAYLOAD))
93 #define I40E_DEFAULT_RSS_HENA_EXPANDED (I40E_DEFAULT_RSS_HENA | \
94 BIT_ULL(I40E_FILTER_PCTYPE_NONF_IPV4_TCP_SYN_NO_ACK) | \
95 BIT_ULL(I40E_FILTER_PCTYPE_NONF_UNICAST_IPV4_UDP) | \
96 BIT_ULL(I40E_FILTER_PCTYPE_NONF_MULTICAST_IPV4_UDP) | \
97 BIT_ULL(I40E_FILTER_PCTYPE_NONF_IPV6_TCP_SYN_NO_ACK) | \
98 BIT_ULL(I40E_FILTER_PCTYPE_NONF_UNICAST_IPV6_UDP) | \
99 BIT_ULL(I40E_FILTER_PCTYPE_NONF_MULTICAST_IPV6_UDP))
101 #define i40e_pf_get_default_rss_hena(pf) \
102 (((pf)->flags & I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE) ? \
103 I40E_DEFAULT_RSS_HENA_EXPANDED : I40E_DEFAULT_RSS_HENA)
105 /* Supported Rx Buffer Sizes (a multiple of 128) */
106 #define I40E_RXBUFFER_256 256
107 #define I40E_RXBUFFER_2048 2048
108 #define I40E_RXBUFFER_3072 3072 /* For FCoE MTU of 2158 */
109 #define I40E_RXBUFFER_4096 4096
110 #define I40E_RXBUFFER_8192 8192
111 #define I40E_MAX_RXBUFFER 9728 /* largest size for single descriptor */
113 /* NOTE: netdev_alloc_skb reserves up to 64 bytes, NET_IP_ALIGN means we
114 * reserve 2 more, and skb_shared_info adds an additional 384 bytes more,
115 * this adds up to 512 bytes of extra data meaning the smallest allocation
116 * we could have is 1K.
117 * i.e. RXBUFFER_256 --> 960 byte skb (size-1024 slab)
118 * i.e. RXBUFFER_512 --> 1216 byte skb (size-2048 slab)
120 #define I40E_RX_HDR_SIZE I40E_RXBUFFER_256
121 #define i40e_rx_desc i40e_32byte_rx_desc
124 * i40e_test_staterr - tests bits in Rx descriptor status and error fields
125 * @rx_desc: pointer to receive descriptor (in le64 format)
126 * @stat_err_bits: value to mask
128 * This function does some fast chicanery in order to return the
129 * value of the mask which is really only used for boolean tests.
130 * The status_error_len doesn't need to be shifted because it begins
133 static inline bool i40e_test_staterr(union i40e_rx_desc *rx_desc,
134 const u64 stat_err_bits)
136 return !!(rx_desc->wb.qword1.status_error_len &
137 cpu_to_le64(stat_err_bits));
140 /* How many Rx Buffers do we bundle into one write to the hardware ? */
141 #define I40E_RX_BUFFER_WRITE 16 /* Must be power of 2 */
142 #define I40E_RX_INCREMENT(r, i) \
145 if ((i) == (r)->count) \
147 r->next_to_clean = i; \
150 #define I40E_RX_NEXT_DESC(r, i, n) \
153 if ((i) == (r)->count) \
155 (n) = I40E_RX_DESC((r), (i)); \
158 #define I40E_RX_NEXT_DESC_PREFETCH(r, i, n) \
160 I40E_RX_NEXT_DESC((r), (i), (n)); \
164 #define I40E_MAX_BUFFER_TXD 8
165 #define I40E_MIN_TX_LEN 17
167 /* The size limit for a transmit buffer in a descriptor is (16K - 1).
168 * In order to align with the read requests we will align the value to
169 * the nearest 4K which represents our maximum read request size.
171 #define I40E_MAX_READ_REQ_SIZE 4096
172 #define I40E_MAX_DATA_PER_TXD (16 * 1024 - 1)
173 #define I40E_MAX_DATA_PER_TXD_ALIGNED \
174 (I40E_MAX_DATA_PER_TXD & ~(I40E_MAX_READ_REQ_SIZE - 1))
176 /* This ugly bit of math is equivalent to DIV_ROUNDUP(size, X) where X is
177 * the value I40E_MAX_DATA_PER_TXD_ALIGNED. It is needed due to the fact
178 * that 12K is not a power of 2 and division is expensive. It is used to
179 * approximate the number of descriptors used per linear buffer. Note
180 * that this will overestimate in some cases as it doesn't account for the
181 * fact that we will add up to 4K - 1 in aligning the 12K buffer, however
182 * the error should not impact things much as large buffers usually mean
183 * we will use fewer descriptors then there are frags in an skb.
185 static inline unsigned int i40e_txd_use_count(unsigned int size)
187 const unsigned int max = I40E_MAX_DATA_PER_TXD_ALIGNED;
188 const unsigned int reciprocal = ((1ull << 32) - 1 + (max / 2)) / max;
189 unsigned int adjust = ~(u32)0;
191 /* if we rounded up on the reciprocal pull down the adjustment */
192 if ((max * reciprocal) > adjust)
193 adjust = ~(u32)(reciprocal - 1);
195 return (u32)((((u64)size * reciprocal) + adjust) >> 32);
198 /* Tx Descriptors needed, worst case */
199 #define DESC_NEEDED (MAX_SKB_FRAGS + 4)
200 #define I40E_MIN_DESC_PENDING 4
202 #define I40E_TX_FLAGS_HW_VLAN BIT(1)
203 #define I40E_TX_FLAGS_SW_VLAN BIT(2)
204 #define I40E_TX_FLAGS_TSO BIT(3)
205 #define I40E_TX_FLAGS_IPV4 BIT(4)
206 #define I40E_TX_FLAGS_IPV6 BIT(5)
207 #define I40E_TX_FLAGS_FCCRC BIT(6)
208 #define I40E_TX_FLAGS_FSO BIT(7)
209 #define I40E_TX_FLAGS_FD_SB BIT(9)
210 #define I40E_TX_FLAGS_VXLAN_TUNNEL BIT(10)
211 #define I40E_TX_FLAGS_VLAN_MASK 0xffff0000
212 #define I40E_TX_FLAGS_VLAN_PRIO_MASK 0xe0000000
213 #define I40E_TX_FLAGS_VLAN_PRIO_SHIFT 29
214 #define I40E_TX_FLAGS_VLAN_SHIFT 16
216 struct i40e_tx_buffer {
217 struct i40e_tx_desc *next_to_watch;
222 unsigned int bytecount;
223 unsigned short gso_segs;
225 DEFINE_DMA_UNMAP_ADDR(dma);
226 DEFINE_DMA_UNMAP_LEN(len);
230 struct i40e_rx_buffer {
234 unsigned int page_offset;
237 struct i40e_queue_stats {
242 struct i40e_tx_queue_stats {
248 u64 tx_lost_interrupt;
251 struct i40e_rx_queue_stats {
253 u64 alloc_page_failed;
254 u64 alloc_buff_failed;
255 u64 page_reuse_count;
259 enum i40e_ring_state_t {
260 __I40E_TX_FDIR_INIT_DONE,
261 __I40E_TX_XPS_INIT_DONE,
264 /* some useful defines for virtchannel interface, which
265 * is the only remaining user of header split
267 #define I40E_RX_DTYPE_NO_SPLIT 0
268 #define I40E_RX_DTYPE_HEADER_SPLIT 1
269 #define I40E_RX_DTYPE_SPLIT_ALWAYS 2
270 #define I40E_RX_SPLIT_L2 0x1
271 #define I40E_RX_SPLIT_IP 0x2
272 #define I40E_RX_SPLIT_TCP_UDP 0x4
273 #define I40E_RX_SPLIT_SCTP 0x8
275 /* struct that defines a descriptor ring, associated with a VSI */
277 struct i40e_ring *next; /* pointer to next ring in q_vector */
278 void *desc; /* Descriptor ring memory */
279 struct device *dev; /* Used for DMA mapping */
280 struct net_device *netdev; /* netdev ring maps to */
282 struct i40e_tx_buffer *tx_bi;
283 struct i40e_rx_buffer *rx_bi;
286 u16 queue_index; /* Queue number of ring */
287 u8 dcb_tc; /* Traffic class of ring */
290 /* high bit set means dynamic, use accessors routines to read/write.
291 * hardware only supports 2us resolution for the ITR registers.
292 * these values always store the USER setting, and must be converted
293 * before programming to a register.
298 u16 count; /* Number of descriptors */
299 u16 reg_idx; /* HW register index of the ring */
302 /* used in interrupt processing */
309 bool ring_active; /* is ring online or not */
310 bool arm_wb; /* do something to arm write back */
312 #define I40E_TXR_FLAGS_LAST_XMIT_MORE_SET BIT(2)
315 #define I40E_TXR_FLAGS_WB_ON_ITR BIT(0)
318 struct i40e_queue_stats stats;
319 struct u64_stats_sync syncp;
321 struct i40e_tx_queue_stats tx_stats;
322 struct i40e_rx_queue_stats rx_stats;
325 unsigned int size; /* length of descriptor ring in bytes */
326 dma_addr_t dma; /* physical address of ring */
328 struct i40e_vsi *vsi; /* Backreference to associated VSI */
329 struct i40e_q_vector *q_vector; /* Backreference to associated vector */
331 struct rcu_head rcu; /* to avoid race on free */
333 } ____cacheline_internodealigned_in_smp;
335 enum i40e_latency_range {
336 I40E_LOWEST_LATENCY = 0,
337 I40E_LOW_LATENCY = 1,
338 I40E_BULK_LATENCY = 2,
339 I40E_ULTRA_LATENCY = 3,
342 struct i40e_ring_container {
343 /* array of pointers to rings */
344 struct i40e_ring *ring;
345 unsigned int total_bytes; /* total bytes processed this int */
346 unsigned int total_packets; /* total packets processed this int */
348 enum i40e_latency_range latency_range;
352 /* iterator for handling rings in ring container */
353 #define i40e_for_each_ring(pos, head) \
354 for (pos = (head).ring; pos != NULL; pos = pos->next)
356 bool i40evf_alloc_rx_buffers(struct i40e_ring *rxr, u16 cleaned_count);
357 netdev_tx_t i40evf_xmit_frame(struct sk_buff *skb, struct net_device *netdev);
358 void i40evf_clean_tx_ring(struct i40e_ring *tx_ring);
359 void i40evf_clean_rx_ring(struct i40e_ring *rx_ring);
360 int i40evf_setup_tx_descriptors(struct i40e_ring *tx_ring);
361 int i40evf_setup_rx_descriptors(struct i40e_ring *rx_ring);
362 void i40evf_free_tx_resources(struct i40e_ring *tx_ring);
363 void i40evf_free_rx_resources(struct i40e_ring *rx_ring);
364 int i40evf_napi_poll(struct napi_struct *napi, int budget);
365 void i40evf_force_wb(struct i40e_vsi *vsi, struct i40e_q_vector *q_vector);
366 u32 i40evf_get_tx_pending(struct i40e_ring *ring, bool in_sw);
367 int __i40evf_maybe_stop_tx(struct i40e_ring *tx_ring, int size);
368 bool __i40evf_chk_linearize(struct sk_buff *skb);
371 * i40e_get_head - Retrieve head from head writeback
372 * @tx_ring: Tx ring to fetch head of
374 * Returns value of Tx ring head based on value stored
375 * in head write-back location
377 static inline u32 i40e_get_head(struct i40e_ring *tx_ring)
379 void *head = (struct i40e_tx_desc *)tx_ring->desc + tx_ring->count;
381 return le32_to_cpu(*(volatile __le32 *)head);
385 * i40e_xmit_descriptor_count - calculate number of Tx descriptors needed
387 * @tx_ring: ring to send buffer on
389 * Returns number of data descriptors needed for this skb. Returns 0 to indicate
390 * there is not enough descriptors available in this ring since we need at least
393 static inline int i40e_xmit_descriptor_count(struct sk_buff *skb)
395 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
396 unsigned int nr_frags = skb_shinfo(skb)->nr_frags;
397 int count = 0, size = skb_headlen(skb);
400 count += i40e_txd_use_count(size);
405 size = skb_frag_size(frag++);
412 * i40e_maybe_stop_tx - 1st level check for Tx stop conditions
413 * @tx_ring: the ring to be checked
414 * @size: the size buffer we want to assure is available
416 * Returns 0 if stop is not needed
418 static inline int i40e_maybe_stop_tx(struct i40e_ring *tx_ring, int size)
420 if (likely(I40E_DESC_UNUSED(tx_ring) >= size))
422 return __i40evf_maybe_stop_tx(tx_ring, size);
426 * i40e_chk_linearize - Check if there are more than 8 fragments per packet
428 * @count: number of buffers used
430 * Note: Our HW can't scatter-gather more than 8 fragments to build
431 * a packet on the wire and so we need to figure out the cases where we
432 * need to linearize the skb.
434 static inline bool i40e_chk_linearize(struct sk_buff *skb, int count)
436 /* Both TSO and single send will work if count is less than 8 */
437 if (likely(count < I40E_MAX_BUFFER_TXD))
441 return __i40evf_chk_linearize(skb);
443 /* we can support up to 8 data buffers for a single send */
444 return count != I40E_MAX_BUFFER_TXD;
448 * i40e_rx_is_fcoe - returns true if the Rx packet type is FCoE
449 * @ptype: the packet type field from Rx descriptor write-back
451 static inline bool i40e_rx_is_fcoe(u16 ptype)
453 return (ptype >= I40E_RX_PTYPE_L2_FCOE_PAY3) &&
454 (ptype <= I40E_RX_PTYPE_L2_FCOE_VFT_FCOTHER);
458 * txring_txq - Find the netdev Tx ring based on the i40e Tx ring
459 * @ring: Tx ring to find the netdev equivalent of
461 static inline struct netdev_queue *txring_txq(const struct i40e_ring *ring)
463 return netdev_get_tx_queue(ring->netdev, ring->queue_index);
465 #endif /* _I40E_TXRX_H_ */