xref: /dpdk/lib/mbuf/rte_mbuf_core.h (revision 9e152e674c77dd14c4f324a832ece4498b1c6470)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2010-2014 Intel Corporation.
3  * Copyright 2014 6WIND S.A.
4  */
5 
6 #ifndef _RTE_MBUF_CORE_H_
7 #define _RTE_MBUF_CORE_H_
8 
9 /**
10  * @file
11  * This file contains definition of RTE mbuf structure itself,
12  * packet offload flags and some related macros.
13  * For majority of DPDK entities, it is not recommended to include
14  * this file directly, use include <rte_mbuf.h> instead.
15  *
16  * New fields and flags should fit in the "dynamic space".
17  */
18 
19 #include <stdalign.h>
20 #include <stdint.h>
21 
22 #include <rte_byteorder.h>
23 #include <rte_stdatomic.h>
24 
25 #ifdef __cplusplus
26 extern "C" {
27 #endif
28 
29 /*
30  * Packet Offload Features Flags. It also carry packet type information.
31  * Critical resources. Both rx/tx shared these bits. Be cautious on any change
32  *
33  * - RX flags start at bit position zero, and get added to the left of previous
34  *   flags.
35  * - The most-significant 3 bits are reserved for generic mbuf flags
36  * - TX flags therefore start at bit position 60 (i.e. 63-3), and new flags get
37  *   added to the right of the previously defined flags i.e. they should count
38  *   downwards, not upwards.
39  *
40  * Keep these flags synchronized with rte_get_rx_ol_flag_name() and
41  * rte_get_tx_ol_flag_name().
42  */
43 
44 /**
45  * The RX packet is a 802.1q VLAN packet, and the tci has been
46  * saved in mbuf->vlan_tci.
47  * If the flag RTE_MBUF_F_RX_VLAN_STRIPPED is also present, the VLAN
48  * header has been stripped from mbuf data, else it is still
49  * present.
50  */
51 #define RTE_MBUF_F_RX_VLAN          (1ULL << 0)
52 
53 /** RX packet with RSS hash result. */
54 #define RTE_MBUF_F_RX_RSS_HASH      (1ULL << 1)
55 
56  /** RX packet with FDIR match indicate. */
57 #define RTE_MBUF_F_RX_FDIR          (1ULL << 2)
58 
59 /**
60  * This flag is set when the outermost IP header checksum is detected as
61  * wrong by the hardware.
62  */
63 #define RTE_MBUF_F_RX_OUTER_IP_CKSUM_BAD (1ULL << 5)
64 
65 /**
66  * A vlan has been stripped by the hardware and its tci is saved in
67  * mbuf->vlan_tci. This can only happen if vlan stripping is enabled
68  * in the RX configuration of the PMD.
69  * When RTE_MBUF_F_RX_VLAN_STRIPPED is set, RTE_MBUF_F_RX_VLAN must also be set.
70  */
71 #define RTE_MBUF_F_RX_VLAN_STRIPPED (1ULL << 6)
72 
73 /**
74  * Mask of bits used to determine the status of RX IP checksum.
75  * - RTE_MBUF_F_RX_IP_CKSUM_UNKNOWN: no information about the RX IP checksum
76  * - RTE_MBUF_F_RX_IP_CKSUM_BAD: the IP checksum in the packet is wrong
77  * - RTE_MBUF_F_RX_IP_CKSUM_GOOD: the IP checksum in the packet is valid
78  * - RTE_MBUF_F_RX_IP_CKSUM_NONE: the IP checksum is not correct in the packet
79  *   data, but the integrity of the IP header is verified.
80  */
81 #define RTE_MBUF_F_RX_IP_CKSUM_MASK ((1ULL << 4) | (1ULL << 7))
82 
83 #define RTE_MBUF_F_RX_IP_CKSUM_UNKNOWN 0
84 #define RTE_MBUF_F_RX_IP_CKSUM_BAD     (1ULL << 4)
85 #define RTE_MBUF_F_RX_IP_CKSUM_GOOD    (1ULL << 7)
86 #define RTE_MBUF_F_RX_IP_CKSUM_NONE    ((1ULL << 4) | (1ULL << 7))
87 
88 /**
89  * Mask of bits used to determine the status of RX L4 checksum.
90  * - RTE_MBUF_F_RX_L4_CKSUM_UNKNOWN: no information about the RX L4 checksum
91  * - RTE_MBUF_F_RX_L4_CKSUM_BAD: the L4 checksum in the packet is wrong
92  * - RTE_MBUF_F_RX_L4_CKSUM_GOOD: the L4 checksum in the packet is valid
93  * - RTE_MBUF_F_RX_L4_CKSUM_NONE: the L4 checksum is not correct in the packet
94  *   data, but the integrity of the L4 data is verified.
95  */
96 #define RTE_MBUF_F_RX_L4_CKSUM_MASK ((1ULL << 3) | (1ULL << 8))
97 
98 #define RTE_MBUF_F_RX_L4_CKSUM_UNKNOWN 0
99 #define RTE_MBUF_F_RX_L4_CKSUM_BAD     (1ULL << 3)
100 #define RTE_MBUF_F_RX_L4_CKSUM_GOOD    (1ULL << 8)
101 #define RTE_MBUF_F_RX_L4_CKSUM_NONE    ((1ULL << 3) | (1ULL << 8))
102 
103 /** RX IEEE1588 L2 Ethernet PT Packet. */
104 #define RTE_MBUF_F_RX_IEEE1588_PTP  (1ULL << 9)
105 
106 /** RX IEEE1588 L2/L4 timestamped packet.*/
107 #define RTE_MBUF_F_RX_IEEE1588_TMST (1ULL << 10)
108 
109 /** FD id reported if FDIR match. */
110 #define RTE_MBUF_F_RX_FDIR_ID       (1ULL << 13)
111 
112 /** Flexible bytes reported if FDIR match. */
113 #define RTE_MBUF_F_RX_FDIR_FLX      (1ULL << 14)
114 
115 /**
116  * The outer VLAN has been stripped by the hardware and its TCI is
117  * saved in mbuf->vlan_tci_outer.
118  * This can only happen if VLAN stripping is enabled in the Rx
119  * configuration of the PMD.
120  * When RTE_MBUF_F_RX_QINQ_STRIPPED is set, the flags RTE_MBUF_F_RX_VLAN
121  * and RTE_MBUF_F_RX_QINQ must also be set.
122  *
123  * - If both RTE_MBUF_F_RX_QINQ_STRIPPED and RTE_MBUF_F_RX_VLAN_STRIPPED are
124  *   set, the 2 VLANs have been stripped by the hardware and their TCIs are
125  *   saved in mbuf->vlan_tci (inner) and mbuf->vlan_tci_outer (outer).
126  * - If RTE_MBUF_F_RX_QINQ_STRIPPED is set and RTE_MBUF_F_RX_VLAN_STRIPPED
127  *   is unset, only the outer VLAN is removed from packet data, but both tci
128  *   are saved in mbuf->vlan_tci (inner) and mbuf->vlan_tci_outer (outer).
129  */
130 #define RTE_MBUF_F_RX_QINQ_STRIPPED (1ULL << 15)
131 
132 /**
133  * When packets are coalesced by a hardware or virtual driver, this flag
134  * can be set in the RX mbuf, meaning that the m->tso_segsz field is
135  * valid and is set to the segment size of original packets.
136  */
137 #define RTE_MBUF_F_RX_LRO           (1ULL << 16)
138 
139 /* There is no flag defined at offset 17. It is free for any future use. */
140 
141 /**
142  * Indicate that security offload processing was applied on the RX packet.
143  */
144 #define RTE_MBUF_F_RX_SEC_OFFLOAD	(1ULL << 18)
145 
146 /**
147  * Indicate that security offload processing failed on the RX packet.
148  */
149 #define RTE_MBUF_F_RX_SEC_OFFLOAD_FAILED	(1ULL << 19)
150 
151 /**
152  * The RX packet is a double VLAN, and the outer tci has been
153  * saved in mbuf->vlan_tci_outer. If this flag is set, RTE_MBUF_F_RX_VLAN
154  * must also be set and the inner tci is saved in mbuf->vlan_tci.
155  * If the flag RTE_MBUF_F_RX_QINQ_STRIPPED is also present, both VLANs
156  * headers have been stripped from mbuf data, else they are still
157  * present.
158  */
159 #define RTE_MBUF_F_RX_QINQ          (1ULL << 20)
160 
161 /**
162  * Mask of bits used to determine the status of outer RX L4 checksum.
163  * - RTE_MBUF_F_RX_OUTER_L4_CKSUM_UNKNOWN: no info about the outer RX L4
164  *   checksum
165  * - RTE_MBUF_F_RX_OUTER_L4_CKSUM_BAD: the outer L4 checksum in the packet
166  *   is wrong
167  * - RTE_MBUF_F_RX_OUTER_L4_CKSUM_GOOD: the outer L4 checksum in the packet
168  *   is valid
169  * - RTE_MBUF_F_RX_OUTER_L4_CKSUM_INVALID: invalid outer L4 checksum state.
170  *
171  * The detection of RTE_MBUF_F_RX_OUTER_L4_CKSUM_GOOD shall be based on the
172  * given HW capability, At minimum, the PMD should support
173  * RTE_MBUF_F_RX_OUTER_L4_CKSUM_UNKNOWN and RTE_MBUF_F_RX_OUTER_L4_CKSUM_BAD
174  * states if the RTE_ETH_RX_OFFLOAD_OUTER_UDP_CKSUM offload is available.
175  */
176 #define RTE_MBUF_F_RX_OUTER_L4_CKSUM_MASK	((1ULL << 21) | (1ULL << 22))
177 
178 #define RTE_MBUF_F_RX_OUTER_L4_CKSUM_UNKNOWN	0
179 #define RTE_MBUF_F_RX_OUTER_L4_CKSUM_BAD	(1ULL << 21)
180 #define RTE_MBUF_F_RX_OUTER_L4_CKSUM_GOOD	(1ULL << 22)
181 #define RTE_MBUF_F_RX_OUTER_L4_CKSUM_INVALID	((1ULL << 21) | (1ULL << 22))
182 
183 /* add new RX flags here, don't forget to update RTE_MBUF_F_FIRST_FREE */
184 
185 #define RTE_MBUF_F_FIRST_FREE (1ULL << 23)
186 #define RTE_MBUF_F_LAST_FREE (1ULL << 40)
187 
188 /* add new TX flags here, don't forget to update RTE_MBUF_F_LAST_FREE  */
189 
190 /**
191  * Outer UDP checksum offload flag. This flag is used for enabling
192  * outer UDP checksum in PMD. To use outer UDP checksum, the user needs to
193  * 1) Enable the following in mbuf,
194  * a) Fill outer_l2_len and outer_l3_len in mbuf.
195  * b) Set the RTE_MBUF_F_TX_OUTER_UDP_CKSUM flag.
196  * c) Set the RTE_MBUF_F_TX_OUTER_IPV4 or RTE_MBUF_F_TX_OUTER_IPV6 flag.
197  * 2) Configure RTE_ETH_TX_OFFLOAD_OUTER_UDP_CKSUM offload flag.
198  */
199 #define RTE_MBUF_F_TX_OUTER_UDP_CKSUM     (1ULL << 41)
200 
201 /**
202  * UDP Fragmentation Offload flag. This flag is used for enabling UDP
203  * fragmentation in SW or in HW. When use UFO, mbuf->tso_segsz is used
204  * to store the MSS of UDP fragments.
205  */
206 #define RTE_MBUF_F_TX_UDP_SEG	(1ULL << 42)
207 
208 /**
209  * Request security offload processing on the TX packet.
210  * To use Tx security offload, the user needs to fill l2_len in mbuf
211  * indicating L2 header size and where L3 header starts. Similarly,
212  * l3_len should also be filled along with ol_flags reflecting current L3 type.
213  */
214 #define RTE_MBUF_F_TX_SEC_OFFLOAD	(1ULL << 43)
215 
216 /**
217  * Offload the MACsec. This flag must be set by the application to enable
218  * this offload feature for a packet to be transmitted.
219  */
220 #define RTE_MBUF_F_TX_MACSEC        (1ULL << 44)
221 
222 /**
223  * Bits 45:48 used for the tunnel type.
224  * The tunnel type must be specified for TSO or checksum on the inner part
225  * of tunnel packets.
226  * These flags can be used with RTE_MBUF_F_TX_TCP_SEG for TSO, or
227  * RTE_MBUF_F_TX_xxx_CKSUM.
228  * The mbuf fields for inner and outer header lengths are required:
229  * outer_l2_len, outer_l3_len, l2_len, l3_len, l4_len and tso_segsz for TSO.
230  */
231 #define RTE_MBUF_F_TX_TUNNEL_VXLAN   (0x1ULL << 45)
232 #define RTE_MBUF_F_TX_TUNNEL_GRE     (0x2ULL << 45)
233 #define RTE_MBUF_F_TX_TUNNEL_IPIP    (0x3ULL << 45)
234 #define RTE_MBUF_F_TX_TUNNEL_GENEVE  (0x4ULL << 45)
235 /** TX packet with MPLS-in-UDP RFC 7510 header. */
236 #define RTE_MBUF_F_TX_TUNNEL_MPLSINUDP (0x5ULL << 45)
237 #define RTE_MBUF_F_TX_TUNNEL_VXLAN_GPE (0x6ULL << 45)
238 #define RTE_MBUF_F_TX_TUNNEL_GTP       (0x7ULL << 45)
239 #define RTE_MBUF_F_TX_TUNNEL_ESP       (0x8ULL << 45)
240 /**
241  * Generic IP encapsulated tunnel type, used for TSO and checksum offload.
242  * It can be used for tunnels which are not standards or listed above.
243  * It is preferred to use specific tunnel flags like RTE_MBUF_F_TX_TUNNEL_GRE
244  * or RTE_MBUF_F_TX_TUNNEL_IPIP if possible.
245  * The ethdev must be configured with RTE_ETH_TX_OFFLOAD_IP_TNL_TSO.
246  * Outer and inner checksums are done according to the existing flags like
247  * RTE_MBUF_F_TX_xxx_CKSUM.
248  * Specific tunnel headers that contain payload length, sequence id
249  * or checksum are not expected to be updated.
250  */
251 #define RTE_MBUF_F_TX_TUNNEL_IP (0xDULL << 45)
252 /**
253  * Generic UDP encapsulated tunnel type, used for TSO and checksum offload.
254  * UDP tunnel type implies outer IP layer.
255  * It can be used for tunnels which are not standards or listed above.
256  * It is preferred to use specific tunnel flags like RTE_MBUF_F_TX_TUNNEL_VXLAN
257  * if possible.
258  * The ethdev must be configured with RTE_ETH_TX_OFFLOAD_UDP_TNL_TSO.
259  * Outer and inner checksums are done according to the existing flags like
260  * RTE_MBUF_F_TX_xxx_CKSUM.
261  * Specific tunnel headers that contain payload length, sequence id
262  * or checksum are not expected to be updated.
263  */
264 #define RTE_MBUF_F_TX_TUNNEL_UDP (0xEULL << 45)
265 /* add new TX TUNNEL type here */
266 #define RTE_MBUF_F_TX_TUNNEL_MASK    (0xFULL << 45)
267 
268 /**
269  * Double VLAN insertion (QinQ) request to driver, driver may offload the
270  * insertion based on device capability.
271  * mbuf 'vlan_tci' & 'vlan_tci_outer' must be valid when this flag is set.
272  */
273 #define RTE_MBUF_F_TX_QINQ        (1ULL << 49)
274 
275 /**
276  * TCP segmentation offload. To enable this offload feature for a
277  * packet to be transmitted on hardware supporting TSO:
278  *  - set the RTE_MBUF_F_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
279  *    RTE_MBUF_F_TX_TCP_CKSUM)
280  *  - set the flag RTE_MBUF_F_TX_IPV4 or RTE_MBUF_F_TX_IPV6
281  *  - if it's IPv4, set the RTE_MBUF_F_TX_IP_CKSUM flag
282  *  - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
283  */
284 #define RTE_MBUF_F_TX_TCP_SEG       (1ULL << 50)
285 
286 /** TX IEEE1588 packet to timestamp. */
287 #define RTE_MBUF_F_TX_IEEE1588_TMST (1ULL << 51)
288 
289 /*
290  * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
291  * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
292  * L4 checksum offload, the user needs to:
293  *  - fill l2_len and l3_len in mbuf
294  *  - set the flags RTE_MBUF_F_TX_TCP_CKSUM, RTE_MBUF_F_TX_SCTP_CKSUM or
295  *    RTE_MBUF_F_TX_UDP_CKSUM
296  *  - set the flag RTE_MBUF_F_TX_IPV4 or RTE_MBUF_F_TX_IPV6
297  */
298 
299 /** Disable L4 cksum of TX pkt. */
300 #define RTE_MBUF_F_TX_L4_NO_CKSUM   (0ULL << 52)
301 
302 /** TCP cksum of TX pkt. computed by NIC. */
303 #define RTE_MBUF_F_TX_TCP_CKSUM     (1ULL << 52)
304 
305 /** SCTP cksum of TX pkt. computed by NIC. */
306 #define RTE_MBUF_F_TX_SCTP_CKSUM    (2ULL << 52)
307 
308 /** UDP cksum of TX pkt. computed by NIC. */
309 #define RTE_MBUF_F_TX_UDP_CKSUM     (3ULL << 52)
310 
311 /** Mask for L4 cksum offload request. */
312 #define RTE_MBUF_F_TX_L4_MASK       (3ULL << 52)
313 
314 /**
315  * Offload the IP checksum in the hardware. The flag RTE_MBUF_F_TX_IPV4 should
316  * also be set by the application, although a PMD will only check
317  * RTE_MBUF_F_TX_IP_CKSUM.
318  *  - fill the mbuf offload information: l2_len, l3_len
319  */
320 #define RTE_MBUF_F_TX_IP_CKSUM      (1ULL << 54)
321 
322 /**
323  * Packet is IPv4. This flag must be set when using any offload feature
324  * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
325  * packet. If the packet is a tunneled packet, this flag is related to
326  * the inner headers.
327  */
328 #define RTE_MBUF_F_TX_IPV4          (1ULL << 55)
329 
330 /**
331  * Packet is IPv6. This flag must be set when using an offload feature
332  * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
333  * packet. If the packet is a tunneled packet, this flag is related to
334  * the inner headers.
335  */
336 #define RTE_MBUF_F_TX_IPV6          (1ULL << 56)
337 
338 /**
339  * VLAN tag insertion request to driver, driver may offload the insertion
340  * based on the device capability.
341  * mbuf 'vlan_tci' field must be valid when this flag is set.
342  */
343 #define RTE_MBUF_F_TX_VLAN          (1ULL << 57)
344 
345 /**
346  * Offload the IP checksum of an external header in the hardware. The
347  * flag RTE_MBUF_F_TX_OUTER_IPV4 should also be set by the application, although
348  * a PMD will only check RTE_MBUF_F_TX_OUTER_IP_CKSUM.
349  *  - fill the mbuf offload information: outer_l2_len, outer_l3_len
350  */
351 #define RTE_MBUF_F_TX_OUTER_IP_CKSUM   (1ULL << 58)
352 
353 /**
354  * Packet outer header is IPv4. This flag must be set when using any
355  * outer offload feature (L3 or L4 checksum) to tell the NIC that the
356  * outer header of the tunneled packet is an IPv4 packet.
357  */
358 #define RTE_MBUF_F_TX_OUTER_IPV4   (1ULL << 59)
359 
360 /**
361  * Packet outer header is IPv6. This flag must be set when using any
362  * outer offload feature (L4 checksum) to tell the NIC that the outer
363  * header of the tunneled packet is an IPv6 packet.
364  */
365 #define RTE_MBUF_F_TX_OUTER_IPV6    (1ULL << 60)
366 
367 /**
368  * Bitmask of all supported packet Tx offload features flags,
369  * which can be set for packet.
370  */
371 #define RTE_MBUF_F_TX_OFFLOAD_MASK (    \
372 		RTE_MBUF_F_TX_OUTER_IPV6 |	 \
373 		RTE_MBUF_F_TX_OUTER_IPV4 |	 \
374 		RTE_MBUF_F_TX_OUTER_IP_CKSUM |  \
375 		RTE_MBUF_F_TX_VLAN |        \
376 		RTE_MBUF_F_TX_IPV6 |		 \
377 		RTE_MBUF_F_TX_IPV4 |		 \
378 		RTE_MBUF_F_TX_IP_CKSUM |        \
379 		RTE_MBUF_F_TX_L4_MASK |         \
380 		RTE_MBUF_F_TX_IEEE1588_TMST |	 \
381 		RTE_MBUF_F_TX_TCP_SEG |         \
382 		RTE_MBUF_F_TX_QINQ |        \
383 		RTE_MBUF_F_TX_TUNNEL_MASK |	 \
384 		RTE_MBUF_F_TX_MACSEC |		 \
385 		RTE_MBUF_F_TX_SEC_OFFLOAD |	 \
386 		RTE_MBUF_F_TX_UDP_SEG |	 \
387 		RTE_MBUF_F_TX_OUTER_UDP_CKSUM)
388 
389 /**
390  * Mbuf having an external buffer attached. shinfo in mbuf must be filled.
391  */
392 #define RTE_MBUF_F_EXTERNAL    (1ULL << 61)
393 
394 #define RTE_MBUF_F_INDIRECT    (1ULL << 62) /**< Indirect attached mbuf */
395 
396 /** Alignment constraint of mbuf private area. */
397 #define RTE_MBUF_PRIV_ALIGN 8
398 
399 /**
400  * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
401  * splitting it into multiple segments.
402  * So, for mbufs that planned to be involved into RX/TX, the recommended
403  * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
404  */
405 #define	RTE_MBUF_DEFAULT_DATAROOM	2048
406 #define	RTE_MBUF_DEFAULT_BUF_SIZE	\
407 	(RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
408 
409 struct rte_mbuf_sched {
410 	uint32_t queue_id;   /**< Queue ID. */
411 	uint8_t traffic_class;
412 	/**< Traffic class ID. Traffic class 0
413 	 * is the highest priority traffic class.
414 	 */
415 	uint8_t color;
416 	/**< Color. @see enum rte_color.*/
417 	uint16_t reserved;   /**< Reserved. */
418 }; /**< Hierarchical scheduler */
419 
420 /**
421  * enum for the tx_offload bit-fields lengths and offsets.
422  * defines the layout of rte_mbuf tx_offload field.
423  */
424 enum {
425 	RTE_MBUF_L2_LEN_BITS = 7,
426 	RTE_MBUF_L3_LEN_BITS = 9,
427 	RTE_MBUF_L4_LEN_BITS = 8,
428 	RTE_MBUF_TSO_SEGSZ_BITS = 16,
429 	RTE_MBUF_OUTL3_LEN_BITS = 9,
430 	RTE_MBUF_OUTL2_LEN_BITS = 7,
431 	RTE_MBUF_TXOFLD_UNUSED_BITS = sizeof(uint64_t) * CHAR_BIT -
432 		RTE_MBUF_L2_LEN_BITS -
433 		RTE_MBUF_L3_LEN_BITS -
434 		RTE_MBUF_L4_LEN_BITS -
435 		RTE_MBUF_TSO_SEGSZ_BITS -
436 		RTE_MBUF_OUTL3_LEN_BITS -
437 		RTE_MBUF_OUTL2_LEN_BITS,
438 #if RTE_BYTE_ORDER == RTE_BIG_ENDIAN
439 	RTE_MBUF_L2_LEN_OFS =
440 		sizeof(uint64_t) * CHAR_BIT - RTE_MBUF_L2_LEN_BITS,
441 	RTE_MBUF_L3_LEN_OFS = RTE_MBUF_L2_LEN_OFS - RTE_MBUF_L3_LEN_BITS,
442 	RTE_MBUF_L4_LEN_OFS = RTE_MBUF_L3_LEN_OFS - RTE_MBUF_L4_LEN_BITS,
443 	RTE_MBUF_TSO_SEGSZ_OFS = RTE_MBUF_L4_LEN_OFS - RTE_MBUF_TSO_SEGSZ_BITS,
444 	RTE_MBUF_OUTL3_LEN_OFS =
445 		RTE_MBUF_TSO_SEGSZ_OFS - RTE_MBUF_OUTL3_LEN_BITS,
446 	RTE_MBUF_OUTL2_LEN_OFS =
447 		RTE_MBUF_OUTL3_LEN_OFS - RTE_MBUF_OUTL2_LEN_BITS,
448 	RTE_MBUF_TXOFLD_UNUSED_OFS =
449 		RTE_MBUF_OUTL2_LEN_OFS - RTE_MBUF_TXOFLD_UNUSED_BITS,
450 #else
451 	RTE_MBUF_L2_LEN_OFS = 0,
452 	RTE_MBUF_L3_LEN_OFS = RTE_MBUF_L2_LEN_OFS + RTE_MBUF_L2_LEN_BITS,
453 	RTE_MBUF_L4_LEN_OFS = RTE_MBUF_L3_LEN_OFS + RTE_MBUF_L3_LEN_BITS,
454 	RTE_MBUF_TSO_SEGSZ_OFS = RTE_MBUF_L4_LEN_OFS + RTE_MBUF_L4_LEN_BITS,
455 	RTE_MBUF_OUTL3_LEN_OFS =
456 		RTE_MBUF_TSO_SEGSZ_OFS + RTE_MBUF_TSO_SEGSZ_BITS,
457 	RTE_MBUF_OUTL2_LEN_OFS =
458 		RTE_MBUF_OUTL3_LEN_OFS + RTE_MBUF_OUTL3_LEN_BITS,
459 	RTE_MBUF_TXOFLD_UNUSED_OFS =
460 		RTE_MBUF_OUTL2_LEN_OFS + RTE_MBUF_OUTL2_LEN_BITS,
461 #endif
462 };
463 
464 /**
465  * The generic rte_mbuf, containing a packet mbuf.
466  */
467 struct __rte_cache_aligned rte_mbuf {
468 	void *buf_addr;           /**< Virtual address of segment buffer. */
469 #if RTE_IOVA_IN_MBUF
470 	/**
471 	 * Physical address of segment buffer.
472 	 * This field is undefined if the build is configured to use only
473 	 * virtual address as IOVA (i.e. RTE_IOVA_IN_MBUF is 0).
474 	 * Force alignment to 8-bytes, so as to ensure we have the exact
475 	 * layout for the first cache line for 32-bit and 64-bit. This makes
476 	 * working on vector drivers easier.
477 	 */
478 	alignas(sizeof(rte_iova_t)) rte_iova_t buf_iova;
479 #else
480 	/**
481 	 * Next segment of scattered packet.
482 	 * This field is valid when physical address field is undefined.
483 	 * Otherwise next pointer in the second cache line will be used.
484 	 */
485 	struct rte_mbuf *next;
486 #endif
487 
488 	/* next 8 bytes are initialised on RX descriptor rearm */
489 	union {
490 		uint64_t rearm_data[1];
491 		__extension__
492 		struct {
493 			uint16_t data_off;
494 
495 			/**
496 			 * Reference counter. Its size should at least equal to the size
497 			 * of port field (16 bits), to support zero-copy broadcast.
498 			 * It should only be accessed using the following functions:
499 			 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
500 			 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
501 			 * or non-atomic) is controlled by the RTE_MBUF_REFCNT_ATOMIC flag.
502 			 */
503 			RTE_ATOMIC(uint16_t) refcnt;
504 
505 			/**
506 			 * Number of segments. Only valid for the first segment of an mbuf
507 			 * chain.
508 			 */
509 			uint16_t nb_segs;
510 
511 			/** Input port (16 bits to support more than 256 virtual ports).
512 			 * The event eth Tx adapter uses this field to specify the output port.
513 			 */
514 			uint16_t port;
515 		};
516 	};
517 
518 	uint64_t ol_flags;        /**< Offload features. */
519 
520 	/* remaining 24 bytes are set on RX when pulling packet from descriptor */
521 	union {
522 		/* void * type of the array elements is retained for driver compatibility. */
523 		void *rx_descriptor_fields1[24 / sizeof(void *)];
524 		__extension__
525 		struct {
526 			/*
527 			 * The packet type, which is the combination of outer/inner L2, L3, L4
528 			 * and tunnel types. The packet_type is about data really present in the
529 			 * mbuf. Example: if vlan stripping is enabled, a received vlan packet
530 			 * would have RTE_PTYPE_L2_ETHER and not RTE_PTYPE_L2_VLAN because the
531 			 * vlan is stripped from the data.
532 			 */
533 			union {
534 				uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
535 				__extension__
536 				struct {
537 					uint8_t l2_type:4;   /**< (Outer) L2 type. */
538 					uint8_t l3_type:4;   /**< (Outer) L3 type. */
539 					uint8_t l4_type:4;   /**< (Outer) L4 type. */
540 					uint8_t tun_type:4;  /**< Tunnel type. */
541 					union {
542 						uint8_t inner_esp_next_proto;
543 						/**< ESP next protocol type, valid if
544 						 * RTE_PTYPE_TUNNEL_ESP tunnel type is set
545 						 * on both Tx and Rx.
546 						 */
547 						__extension__
548 						struct {
549 							uint8_t inner_l2_type:4;
550 							/**< Inner L2 type. */
551 							uint8_t inner_l3_type:4;
552 							/**< Inner L3 type. */
553 						};
554 					};
555 					uint8_t inner_l4_type:4; /**< Inner L4 type. */
556 				};
557 			};
558 
559 			uint32_t pkt_len;         /**< Total pkt len: sum of all segments. */
560 			uint16_t data_len;        /**< Amount of data in segment buffer. */
561 			/** VLAN TCI (CPU order), valid if RTE_MBUF_F_RX_VLAN is set. */
562 			uint16_t vlan_tci;
563 
564 			union {
565 				union {
566 					uint32_t rss;     /**< RSS hash result if RSS enabled */
567 					struct {
568 						union {
569 							struct {
570 								uint16_t hash;
571 								uint16_t id;
572 							};
573 							uint32_t lo;
574 							/**< Second 4 flexible bytes */
575 						};
576 						uint32_t hi;
577 						/**< First 4 flexible bytes or FD ID, dependent
578 						 * on RTE_MBUF_F_RX_FDIR_* flag in ol_flags.
579 						 */
580 					} fdir;	/**< Filter identifier if FDIR enabled */
581 					struct rte_mbuf_sched sched;
582 					/**< Hierarchical scheduler : 8 bytes */
583 					struct {
584 						uint32_t reserved1;
585 						uint16_t reserved2;
586 						uint16_t txq;
587 						/**< The event eth Tx adapter uses this field
588 						 * to store Tx queue id.
589 						 * @see rte_event_eth_tx_adapter_txq_set()
590 						 */
591 					} txadapter; /**< Eventdev ethdev Tx adapter */
592 					uint32_t usr;
593 					/**< User defined tags. See rte_distributor_process() */
594 				} hash;                   /**< hash information */
595 			};
596 
597 			/** Outer VLAN TCI (CPU order), valid if RTE_MBUF_F_RX_QINQ is set. */
598 			uint16_t vlan_tci_outer;
599 
600 			uint16_t buf_len;         /**< Length of segment buffer. */
601 		};
602 	};
603 
604 	struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
605 
606 	/* second cache line - fields only used in slow path or on TX */
607 #if RTE_IOVA_IN_MBUF
608 	/**
609 	 * Next segment of scattered packet. Must be NULL in the last
610 	 * segment or in case of non-segmented packet.
611 	 */
612 	alignas(RTE_CACHE_LINE_MIN_SIZE)
613 	struct rte_mbuf *next;
614 #else
615 	/**
616 	 * Reserved for dynamic fields
617 	 * when the next pointer is in first cache line (i.e. RTE_IOVA_IN_MBUF is 0).
618 	 */
619 	alignas(RTE_CACHE_LINE_MIN_SIZE)
620 	uint64_t dynfield2;
621 #endif
622 
623 	/* fields to support TX offloads */
624 	union {
625 		uint64_t tx_offload;       /**< combined for easy fetch */
626 		__extension__
627 		struct {
628 			uint64_t l2_len:RTE_MBUF_L2_LEN_BITS;
629 			/**< L2 (MAC) Header Length for non-tunneling pkt.
630 			 * Outer_L4_len + ... + Inner_L2_len for tunneling pkt.
631 			 */
632 			uint64_t l3_len:RTE_MBUF_L3_LEN_BITS;
633 			/**< L3 (IP) Header Length. */
634 			uint64_t l4_len:RTE_MBUF_L4_LEN_BITS;
635 			/**< L4 (TCP/UDP) Header Length. */
636 			uint64_t tso_segsz:RTE_MBUF_TSO_SEGSZ_BITS;
637 			/**< TCP TSO segment size */
638 
639 			/*
640 			 * Fields for Tx offloading of tunnels.
641 			 * These are undefined for packets which don't request
642 			 * any tunnel offloads (outer IP or UDP checksum,
643 			 * tunnel TSO).
644 			 *
645 			 * PMDs should not use these fields unconditionally
646 			 * when calculating offsets.
647 			 *
648 			 * Applications are expected to set appropriate tunnel
649 			 * offload flags when they fill in these fields.
650 			 */
651 			uint64_t outer_l3_len:RTE_MBUF_OUTL3_LEN_BITS;
652 			/**< Outer L3 (IP) Hdr Length. */
653 			uint64_t outer_l2_len:RTE_MBUF_OUTL2_LEN_BITS;
654 			/**< Outer L2 (MAC) Hdr Length. */
655 
656 			/* uint64_t unused:RTE_MBUF_TXOFLD_UNUSED_BITS; */
657 		};
658 	};
659 
660 	/** Shared data for external buffer attached to mbuf. See
661 	 * rte_pktmbuf_attach_extbuf().
662 	 */
663 	struct rte_mbuf_ext_shared_info *shinfo;
664 
665 	/** Size of the application private data. In case of an indirect
666 	 * mbuf, it stores the direct mbuf private data size.
667 	 */
668 	uint16_t priv_size;
669 
670 	/** Timesync flags for use with IEEE1588. */
671 	uint16_t timesync;
672 
673 	uint32_t dynfield1[9]; /**< Reserved for dynamic fields. */
674 };
675 
676 /**
677  * Function typedef of callback to free externally attached buffer.
678  */
679 typedef void (*rte_mbuf_extbuf_free_callback_t)(void *addr, void *opaque);
680 
681 /**
682  * Shared data at the end of an external buffer.
683  */
684 struct rte_mbuf_ext_shared_info {
685 	rte_mbuf_extbuf_free_callback_t free_cb; /**< Free callback function */
686 	void *fcb_opaque;                        /**< Free callback argument */
687 	RTE_ATOMIC(uint16_t) refcnt;
688 };
689 
690 /** Maximum number of nb_segs allowed. */
691 #define RTE_MBUF_MAX_NB_SEGS	UINT16_MAX
692 
693 /**
694  * Returns TRUE if given mbuf is cloned by mbuf indirection, or FALSE
695  * otherwise.
696  *
697  * If a mbuf has its data in another mbuf and references it by mbuf
698  * indirection, this mbuf can be defined as a cloned mbuf.
699  */
700 #define RTE_MBUF_CLONED(mb)     ((mb)->ol_flags & RTE_MBUF_F_INDIRECT)
701 
702 /**
703  * Returns TRUE if given mbuf has an external buffer, or FALSE otherwise.
704  *
705  * External buffer is a user-provided anonymous buffer.
706  */
707 #define RTE_MBUF_HAS_EXTBUF(mb) ((mb)->ol_flags & RTE_MBUF_F_EXTERNAL)
708 
709 /**
710  * Returns TRUE if given mbuf is direct, or FALSE otherwise.
711  *
712  * If a mbuf embeds its own data after the rte_mbuf structure, this mbuf
713  * can be defined as a direct mbuf.
714  */
715 #define RTE_MBUF_DIRECT(mb) \
716 	(!((mb)->ol_flags & (RTE_MBUF_F_INDIRECT | RTE_MBUF_F_EXTERNAL)))
717 
718 /** Uninitialized or unspecified port. */
719 #define RTE_MBUF_PORT_INVALID UINT16_MAX
720 /** For backwards compatibility. */
721 #define MBUF_INVALID_PORT RTE_MBUF_PORT_INVALID
722 
723 /**
724  * A macro that points to an offset into the data in the mbuf.
725  *
726  * The returned pointer is cast to type t. Before using this
727  * function, the user must ensure that the first segment is large
728  * enough to accommodate its data.
729  *
730  * @param m
731  *   The packet mbuf.
732  * @param o
733  *   The offset into the mbuf data.
734  * @param t
735  *   The type to cast the result into.
736  */
737 #define rte_pktmbuf_mtod_offset(m, t, o)	\
738 	((t)(void *)((char *)(m)->buf_addr + (m)->data_off + (o)))
739 
740 /**
741  * A macro that points to the start of the data in the mbuf.
742  *
743  * The returned pointer is cast to type t. Before using this
744  * function, the user must ensure that the first segment is large
745  * enough to accommodate its data.
746  *
747  * @param m
748  *   The packet mbuf.
749  * @param t
750  *   The type to cast the result into.
751  */
752 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
753 
754 /**
755  * A macro that returns the IO address that points to an offset of the
756  * start of the data in the mbuf
757  *
758  * @param m
759  *   The packet mbuf.
760  * @param o
761  *   The offset into the data to calculate address from.
762  */
763 #define rte_pktmbuf_iova_offset(m, o) \
764 	(rte_iova_t)(rte_mbuf_iova_get(m) + (m)->data_off + (o))
765 
766 /**
767  * A macro that returns the IO address that points to the start of the
768  * data in the mbuf
769  *
770  * @param m
771  *   The packet mbuf.
772  */
773 #define rte_pktmbuf_iova(m) rte_pktmbuf_iova_offset(m, 0)
774 
775 #ifdef __cplusplus
776 }
777 #endif
778 
779 #endif /* _RTE_MBUF_CORE_H_ */
780