1 /* SPDX-License-Identifier: BSD-3-Clause
2 * Copyright(c) 2010-2014 Intel Corporation.
3 * Copyright 2014 6WIND S.A.
13 * The mbuf library provides the ability to create and destroy buffers
14 * that may be used by the RTE application to store message
15 * buffers. The message buffers are stored in a mempool, using the
16 * RTE mempool library.
18 * The preferred way to create a mbuf pool is to use
19 * rte_pktmbuf_pool_create(). However, in some situations, an
20 * application may want to have more control (ex: populate the pool with
21 * specific memory), in this case it is possible to use functions from
22 * rte_mempool. See how rte_pktmbuf_pool_create() is implemented for
25 * This library provides an API to allocate/free packet mbufs, which are
26 * used to carry network packets.
28 * To understand the concepts of packet buffers or mbufs, you
29 * should read "TCP/IP Illustrated, Volume 2: The Implementation,
30 * Addison-Wesley, 1995, ISBN 0-201-63354-X from Richard Stevens"
31 * http://www.kohala.com/start/tcpipiv2.html
35 #include <rte_compat.h>
36 #include <rte_common.h>
37 #include <rte_config.h>
38 #include <rte_mempool.h>
39 #include <rte_memory.h>
40 #include <rte_atomic.h>
41 #include <rte_prefetch.h>
42 #include <rte_branch_prediction.h>
43 #include <rte_byteorder.h>
44 #include <rte_mbuf_ptype.h>
51 * Packet Offload Features Flags. It also carry packet type information.
52 * Critical resources. Both rx/tx shared these bits. Be cautious on any change
54 * - RX flags start at bit position zero, and get added to the left of previous
56 * - The most-significant 3 bits are reserved for generic mbuf flags
57 * - TX flags therefore start at bit position 60 (i.e. 63-3), and new flags get
58 * added to the right of the previously defined flags i.e. they should count
59 * downwards, not upwards.
61 * Keep these flags synchronized with rte_get_rx_ol_flag_name() and
62 * rte_get_tx_ol_flag_name().
66 * The RX packet is a 802.1q VLAN packet, and the tci has been
67 * saved in in mbuf->vlan_tci.
68 * If the flag PKT_RX_VLAN_STRIPPED is also present, the VLAN
69 * header has been stripped from mbuf data, else it is still
72 #define PKT_RX_VLAN (1ULL << 0)
74 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
75 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
79 * Checking this flag alone is deprecated: check the 2 bits of
80 * PKT_RX_L4_CKSUM_MASK.
81 * This flag was set when the L4 checksum of a packet was detected as
82 * wrong by the hardware.
84 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3)
88 * Checking this flag alone is deprecated: check the 2 bits of
89 * PKT_RX_IP_CKSUM_MASK.
90 * This flag was set when the IP checksum of a packet was detected as
91 * wrong by the hardware.
93 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4)
95 #define PKT_RX_EIP_CKSUM_BAD (1ULL << 5) /**< External IP header checksum error. */
98 * A vlan has been stripped by the hardware and its tci is saved in
99 * mbuf->vlan_tci. This can only happen if vlan stripping is enabled
100 * in the RX configuration of the PMD.
101 * When PKT_RX_VLAN_STRIPPED is set, PKT_RX_VLAN must also be set.
103 #define PKT_RX_VLAN_STRIPPED (1ULL << 6)
106 * Mask of bits used to determine the status of RX IP checksum.
107 * - PKT_RX_IP_CKSUM_UNKNOWN: no information about the RX IP checksum
108 * - PKT_RX_IP_CKSUM_BAD: the IP checksum in the packet is wrong
109 * - PKT_RX_IP_CKSUM_GOOD: the IP checksum in the packet is valid
110 * - PKT_RX_IP_CKSUM_NONE: the IP checksum is not correct in the packet
111 * data, but the integrity of the IP header is verified.
113 #define PKT_RX_IP_CKSUM_MASK ((1ULL << 4) | (1ULL << 7))
115 #define PKT_RX_IP_CKSUM_UNKNOWN 0
116 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4)
117 #define PKT_RX_IP_CKSUM_GOOD (1ULL << 7)
118 #define PKT_RX_IP_CKSUM_NONE ((1ULL << 4) | (1ULL << 7))
121 * Mask of bits used to determine the status of RX L4 checksum.
122 * - PKT_RX_L4_CKSUM_UNKNOWN: no information about the RX L4 checksum
123 * - PKT_RX_L4_CKSUM_BAD: the L4 checksum in the packet is wrong
124 * - PKT_RX_L4_CKSUM_GOOD: the L4 checksum in the packet is valid
125 * - PKT_RX_L4_CKSUM_NONE: the L4 checksum is not correct in the packet
126 * data, but the integrity of the L4 data is verified.
128 #define PKT_RX_L4_CKSUM_MASK ((1ULL << 3) | (1ULL << 8))
130 #define PKT_RX_L4_CKSUM_UNKNOWN 0
131 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3)
132 #define PKT_RX_L4_CKSUM_GOOD (1ULL << 8)
133 #define PKT_RX_L4_CKSUM_NONE ((1ULL << 3) | (1ULL << 8))
135 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
136 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
137 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
138 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
141 * The 2 vlans have been stripped by the hardware and their tci are
142 * saved in mbuf->vlan_tci (inner) and mbuf->vlan_tci_outer (outer).
143 * This can only happen if vlan stripping is enabled in the RX
144 * configuration of the PMD.
145 * When PKT_RX_QINQ_STRIPPED is set, the flags (PKT_RX_VLAN |
146 * PKT_RX_VLAN_STRIPPED | PKT_RX_QINQ) must also be set.
148 #define PKT_RX_QINQ_STRIPPED (1ULL << 15)
151 * When packets are coalesced by a hardware or virtual driver, this flag
152 * can be set in the RX mbuf, meaning that the m->tso_segsz field is
153 * valid and is set to the segment size of original packets.
155 #define PKT_RX_LRO (1ULL << 16)
158 * Indicate that the timestamp field in the mbuf is valid.
160 #define PKT_RX_TIMESTAMP (1ULL << 17)
163 * Indicate that security offload processing was applied on the RX packet.
165 #define PKT_RX_SEC_OFFLOAD (1ULL << 18)
168 * Indicate that security offload processing failed on the RX packet.
170 #define PKT_RX_SEC_OFFLOAD_FAILED (1ULL << 19)
173 * The RX packet is a double VLAN, and the outer tci has been
174 * saved in in mbuf->vlan_tci_outer. If PKT_RX_QINQ set, PKT_RX_VLAN
175 * also should be set and inner tci should be saved to mbuf->vlan_tci.
176 * If the flag PKT_RX_QINQ_STRIPPED is also present, both VLANs
177 * headers have been stripped from mbuf data, else they are still
180 #define PKT_RX_QINQ (1ULL << 20)
183 * Mask of bits used to determine the status of outer RX L4 checksum.
184 * - PKT_RX_OUTER_L4_CKSUM_UNKNOWN: no info about the outer RX L4 checksum
185 * - PKT_RX_OUTER_L4_CKSUM_BAD: the outer L4 checksum in the packet is wrong
186 * - PKT_RX_OUTER_L4_CKSUM_GOOD: the outer L4 checksum in the packet is valid
187 * - PKT_RX_OUTER_L4_CKSUM_INVALID: invalid outer L4 checksum state.
189 * The detection of PKT_RX_OUTER_L4_CKSUM_GOOD shall be based on the given
190 * HW capability, At minimum, the PMD should support
191 * PKT_RX_OUTER_L4_CKSUM_UNKNOWN and PKT_RX_OUTER_L4_CKSUM_BAD states
192 * if the DEV_RX_OFFLOAD_OUTER_UDP_CKSUM offload is available.
194 #define PKT_RX_OUTER_L4_CKSUM_MASK ((1ULL << 21) | (1ULL << 22))
196 #define PKT_RX_OUTER_L4_CKSUM_UNKNOWN 0
197 #define PKT_RX_OUTER_L4_CKSUM_BAD (1ULL << 21)
198 #define PKT_RX_OUTER_L4_CKSUM_GOOD (1ULL << 22)
199 #define PKT_RX_OUTER_L4_CKSUM_INVALID ((1ULL << 21) | (1ULL << 22))
201 /* add new RX flags here */
203 /* add new TX flags here */
206 * Indicate that the metadata field in the mbuf is in use.
208 #define PKT_TX_METADATA (1ULL << 40)
211 * Outer UDP checksum offload flag. This flag is used for enabling
212 * outer UDP checksum in PMD. To use outer UDP checksum, the user needs to
213 * 1) Enable the following in mbuf,
214 * a) Fill outer_l2_len and outer_l3_len in mbuf.
215 * b) Set the PKT_TX_OUTER_UDP_CKSUM flag.
216 * c) Set the PKT_TX_OUTER_IPV4 or PKT_TX_OUTER_IPV6 flag.
217 * 2) Configure DEV_TX_OFFLOAD_OUTER_UDP_CKSUM offload flag.
219 #define PKT_TX_OUTER_UDP_CKSUM (1ULL << 41)
222 * UDP Fragmentation Offload flag. This flag is used for enabling UDP
223 * fragmentation in SW or in HW. When use UFO, mbuf->tso_segsz is used
224 * to store the MSS of UDP fragments.
226 #define PKT_TX_UDP_SEG (1ULL << 42)
229 * Request security offload processing on the TX packet.
231 #define PKT_TX_SEC_OFFLOAD (1ULL << 43)
234 * Offload the MACsec. This flag must be set by the application to enable
235 * this offload feature for a packet to be transmitted.
237 #define PKT_TX_MACSEC (1ULL << 44)
240 * Bits 45:48 used for the tunnel type.
241 * The tunnel type must be specified for TSO or checksum on the inner part
243 * These flags can be used with PKT_TX_TCP_SEG for TSO, or PKT_TX_xxx_CKSUM.
244 * The mbuf fields for inner and outer header lengths are required:
245 * outer_l2_len, outer_l3_len, l2_len, l3_len, l4_len and tso_segsz for TSO.
247 #define PKT_TX_TUNNEL_VXLAN (0x1ULL << 45)
248 #define PKT_TX_TUNNEL_GRE (0x2ULL << 45)
249 #define PKT_TX_TUNNEL_IPIP (0x3ULL << 45)
250 #define PKT_TX_TUNNEL_GENEVE (0x4ULL << 45)
251 /** TX packet with MPLS-in-UDP RFC 7510 header. */
252 #define PKT_TX_TUNNEL_MPLSINUDP (0x5ULL << 45)
253 #define PKT_TX_TUNNEL_VXLAN_GPE (0x6ULL << 45)
255 * Generic IP encapsulated tunnel type, used for TSO and checksum offload.
256 * It can be used for tunnels which are not standards or listed above.
257 * It is preferred to use specific tunnel flags like PKT_TX_TUNNEL_GRE
258 * or PKT_TX_TUNNEL_IPIP if possible.
259 * The ethdev must be configured with DEV_TX_OFFLOAD_IP_TNL_TSO.
260 * Outer and inner checksums are done according to the existing flags like
262 * Specific tunnel headers that contain payload length, sequence id
263 * or checksum are not expected to be updated.
265 #define PKT_TX_TUNNEL_IP (0xDULL << 45)
267 * Generic UDP encapsulated tunnel type, used for TSO and checksum offload.
268 * UDP tunnel type implies outer IP layer.
269 * It can be used for tunnels which are not standards or listed above.
270 * It is preferred to use specific tunnel flags like PKT_TX_TUNNEL_VXLAN
272 * The ethdev must be configured with DEV_TX_OFFLOAD_UDP_TNL_TSO.
273 * Outer and inner checksums are done according to the existing flags like
275 * Specific tunnel headers that contain payload length, sequence id
276 * or checksum are not expected to be updated.
278 #define PKT_TX_TUNNEL_UDP (0xEULL << 45)
279 /* add new TX TUNNEL type here */
280 #define PKT_TX_TUNNEL_MASK (0xFULL << 45)
283 * Double VLAN insertion (QinQ) request to driver, driver may offload the
284 * insertion based on device capability.
285 * mbuf 'vlan_tci' & 'vlan_tci_outer' must be valid when this flag is set.
287 #define PKT_TX_QINQ (1ULL << 49)
288 /* this old name is deprecated */
289 #define PKT_TX_QINQ_PKT PKT_TX_QINQ
292 * TCP segmentation offload. To enable this offload feature for a
293 * packet to be transmitted on hardware supporting TSO:
294 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
296 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
297 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag
298 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
300 #define PKT_TX_TCP_SEG (1ULL << 50)
302 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
305 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
306 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
307 * L4 checksum offload, the user needs to:
308 * - fill l2_len and l3_len in mbuf
309 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
310 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
312 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
313 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
314 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
315 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
316 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
319 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
320 * also be set by the application, although a PMD will only check
322 * - fill the mbuf offload information: l2_len, l3_len
324 #define PKT_TX_IP_CKSUM (1ULL << 54)
327 * Packet is IPv4. This flag must be set when using any offload feature
328 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
329 * packet. If the packet is a tunneled packet, this flag is related to
332 #define PKT_TX_IPV4 (1ULL << 55)
335 * Packet is IPv6. This flag must be set when using an offload feature
336 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
337 * packet. If the packet is a tunneled packet, this flag is related to
340 #define PKT_TX_IPV6 (1ULL << 56)
343 * VLAN tag insertion request to driver, driver may offload the insertion
344 * based on the device capability.
345 * mbuf 'vlan_tci' field must be valid when this flag is set.
347 #define PKT_TX_VLAN (1ULL << 57)
348 /* this old name is deprecated */
349 #define PKT_TX_VLAN_PKT PKT_TX_VLAN
352 * Offload the IP checksum of an external header in the hardware. The
353 * flag PKT_TX_OUTER_IPV4 should also be set by the application, although
354 * a PMD will only check PKT_TX_OUTER_IP_CKSUM.
355 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
357 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
360 * Packet outer header is IPv4. This flag must be set when using any
361 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
362 * outer header of the tunneled packet is an IPv4 packet.
364 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
367 * Packet outer header is IPv6. This flag must be set when using any
368 * outer offload feature (L4 checksum) to tell the NIC that the outer
369 * header of the tunneled packet is an IPv6 packet.
371 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
374 * Bitmask of all supported packet Tx offload features flags,
375 * which can be set for packet.
377 #define PKT_TX_OFFLOAD_MASK ( \
378 PKT_TX_OUTER_IPV6 | \
379 PKT_TX_OUTER_IPV4 | \
380 PKT_TX_OUTER_IP_CKSUM | \
386 PKT_TX_IEEE1588_TMST | \
389 PKT_TX_TUNNEL_MASK | \
391 PKT_TX_SEC_OFFLOAD | \
393 PKT_TX_OUTER_UDP_CKSUM | \
397 * Mbuf having an external buffer attached. shinfo in mbuf must be filled.
399 #define EXT_ATTACHED_MBUF (1ULL << 61)
401 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
403 /** Alignment constraint of mbuf private area. */
404 #define RTE_MBUF_PRIV_ALIGN 8
407 * Get the name of a RX offload flag
410 * The mask describing the flag.
412 * The name of this flag, or NULL if it's not a valid RX flag.
414 const char *rte_get_rx_ol_flag_name(uint64_t mask);
417 * Dump the list of RX offload flags in a buffer
420 * The mask describing the RX flags.
424 * The length of the buffer.
426 * 0 on success, (-1) on error.
428 int rte_get_rx_ol_flag_list(uint64_t mask, char *buf, size_t buflen);
431 * Get the name of a TX offload flag
434 * The mask describing the flag. Usually only one bit must be set.
435 * Several bits can be given if they belong to the same mask.
436 * Ex: PKT_TX_L4_MASK.
438 * The name of this flag, or NULL if it's not a valid TX flag.
440 const char *rte_get_tx_ol_flag_name(uint64_t mask);
443 * Dump the list of TX offload flags in a buffer
446 * The mask describing the TX flags.
450 * The length of the buffer.
452 * 0 on success, (-1) on error.
454 int rte_get_tx_ol_flag_list(uint64_t mask, char *buf, size_t buflen);
457 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
458 * splitting it into multiple segments.
459 * So, for mbufs that planned to be involved into RX/TX, the recommended
460 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
462 #define RTE_MBUF_DEFAULT_DATAROOM 2048
463 #define RTE_MBUF_DEFAULT_BUF_SIZE \
464 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
466 /* define a set of marker types that can be used to refer to set points in the
469 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
471 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
473 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
474 * with a single assignment */
476 struct rte_mbuf_sched {
477 uint32_t queue_id; /**< Queue ID. */
478 uint8_t traffic_class;
479 /**< Traffic class ID. Traffic class 0
480 * is the highest priority traffic class.
483 /**< Color. @see enum rte_color.*/
484 uint16_t reserved; /**< Reserved. */
485 }; /**< Hierarchical scheduler */
488 * enum for the tx_offload bit-fields lengths and offsets.
489 * defines the layout of rte_mbuf tx_offload field.
492 RTE_MBUF_L2_LEN_BITS = 7,
493 RTE_MBUF_L3_LEN_BITS = 9,
494 RTE_MBUF_L4_LEN_BITS = 8,
495 RTE_MBUF_TSO_SEGSZ_BITS = 16,
496 RTE_MBUF_OUTL3_LEN_BITS = 9,
497 RTE_MBUF_OUTL2_LEN_BITS = 7,
498 RTE_MBUF_TXOFLD_UNUSED_BITS = sizeof(uint64_t) * CHAR_BIT -
499 RTE_MBUF_L2_LEN_BITS -
500 RTE_MBUF_L3_LEN_BITS -
501 RTE_MBUF_L4_LEN_BITS -
502 RTE_MBUF_TSO_SEGSZ_BITS -
503 RTE_MBUF_OUTL3_LEN_BITS -
504 RTE_MBUF_OUTL2_LEN_BITS,
505 #if RTE_BYTE_ORDER == RTE_BIG_ENDIAN
506 RTE_MBUF_L2_LEN_OFS =
507 sizeof(uint64_t) * CHAR_BIT - RTE_MBUF_L2_LEN_BITS,
508 RTE_MBUF_L3_LEN_OFS = RTE_MBUF_L2_LEN_OFS - RTE_MBUF_L3_LEN_BITS,
509 RTE_MBUF_L4_LEN_OFS = RTE_MBUF_L3_LEN_OFS - RTE_MBUF_L4_LEN_BITS,
510 RTE_MBUF_TSO_SEGSZ_OFS = RTE_MBUF_L4_LEN_OFS - RTE_MBUF_TSO_SEGSZ_BITS,
511 RTE_MBUF_OUTL3_LEN_OFS =
512 RTE_MBUF_TSO_SEGSZ_OFS - RTE_MBUF_OUTL3_LEN_BITS,
513 RTE_MBUF_OUTL2_LEN_OFS =
514 RTE_MBUF_OUTL3_LEN_OFS - RTE_MBUF_OUTL2_LEN_BITS,
515 RTE_MBUF_TXOFLD_UNUSED_OFS =
516 RTE_MBUF_OUTL2_LEN_OFS - RTE_MBUF_TXOFLD_UNUSED_BITS,
518 RTE_MBUF_L2_LEN_OFS = 0,
519 RTE_MBUF_L3_LEN_OFS = RTE_MBUF_L2_LEN_OFS + RTE_MBUF_L2_LEN_BITS,
520 RTE_MBUF_L4_LEN_OFS = RTE_MBUF_L3_LEN_OFS + RTE_MBUF_L3_LEN_BITS,
521 RTE_MBUF_TSO_SEGSZ_OFS = RTE_MBUF_L4_LEN_OFS + RTE_MBUF_L4_LEN_BITS,
522 RTE_MBUF_OUTL3_LEN_OFS =
523 RTE_MBUF_TSO_SEGSZ_OFS + RTE_MBUF_TSO_SEGSZ_BITS,
524 RTE_MBUF_OUTL2_LEN_OFS =
525 RTE_MBUF_OUTL3_LEN_OFS + RTE_MBUF_OUTL3_LEN_BITS,
526 RTE_MBUF_TXOFLD_UNUSED_OFS =
527 RTE_MBUF_OUTL2_LEN_OFS + RTE_MBUF_OUTL2_LEN_BITS,
532 * The generic rte_mbuf, containing a packet mbuf.
537 void *buf_addr; /**< Virtual address of segment buffer. */
539 * Physical address of segment buffer.
540 * Force alignment to 8-bytes, so as to ensure we have the exact
541 * same mbuf cacheline0 layout for 32-bit and 64-bit. This makes
542 * working on vector drivers easier.
547 rte_iova_t buf_physaddr; /**< deprecated */
548 } __rte_aligned(sizeof(rte_iova_t));
550 /* next 8 bytes are initialised on RX descriptor rearm */
555 * Reference counter. Its size should at least equal to the size
556 * of port field (16 bits), to support zero-copy broadcast.
557 * It should only be accessed using the following functions:
558 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
559 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
560 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
565 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
566 uint16_t refcnt; /**< Non-atomically accessed refcnt */
568 uint16_t nb_segs; /**< Number of segments. */
570 /** Input port (16 bits to support more than 256 virtual ports).
571 * The event eth Tx adapter uses this field to specify the output port.
575 uint64_t ol_flags; /**< Offload features. */
577 /* remaining bytes are set on RX when pulling packet from descriptor */
578 MARKER rx_descriptor_fields1;
581 * The packet type, which is the combination of outer/inner L2, L3, L4
582 * and tunnel types. The packet_type is about data really present in the
583 * mbuf. Example: if vlan stripping is enabled, a received vlan packet
584 * would have RTE_PTYPE_L2_ETHER and not RTE_PTYPE_L2_VLAN because the
585 * vlan is stripped from the data.
589 uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
591 uint32_t l2_type:4; /**< (Outer) L2 type. */
592 uint32_t l3_type:4; /**< (Outer) L3 type. */
593 uint32_t l4_type:4; /**< (Outer) L4 type. */
594 uint32_t tun_type:4; /**< Tunnel type. */
597 uint8_t inner_esp_next_proto;
598 /**< ESP next protocol type, valid if
599 * RTE_PTYPE_TUNNEL_ESP tunnel type is set
604 uint8_t inner_l2_type:4;
605 /**< Inner L2 type. */
606 uint8_t inner_l3_type:4;
607 /**< Inner L3 type. */
610 uint32_t inner_l4_type:4; /**< Inner L4 type. */
614 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
615 uint16_t data_len; /**< Amount of data in segment buffer. */
616 /** VLAN TCI (CPU order), valid if PKT_RX_VLAN is set. */
622 uint32_t rss; /**< RSS hash result if RSS enabled */
630 /**< Second 4 flexible bytes */
633 /**< First 4 flexible bytes or FD ID, dependent
634 * on PKT_RX_FDIR_* flag in ol_flags.
636 } fdir; /**< Filter identifier if FDIR enabled */
637 struct rte_mbuf_sched sched;
638 /**< Hierarchical scheduler : 8 bytes */
643 /**< The event eth Tx adapter uses this field
644 * to store Tx queue id.
645 * @see rte_event_eth_tx_adapter_txq_set()
647 } txadapter; /**< Eventdev ethdev Tx adapter */
648 /**< User defined tags. See rte_distributor_process() */
650 } hash; /**< hash information */
653 * Application specific metadata value
654 * for egress flow rule match.
655 * Valid if PKT_TX_METADATA is set.
656 * Located here to allow conjunct use
657 * with hash.sched.hi.
659 uint32_t tx_metadata;
664 /** Outer VLAN TCI (CPU order), valid if PKT_RX_QINQ is set. */
665 uint16_t vlan_tci_outer;
667 uint16_t buf_len; /**< Length of segment buffer. */
669 /** Valid if PKT_RX_TIMESTAMP is set. The unit and time reference
670 * are not normalized but are always the same for a given port.
671 * Some devices allow to query rte_eth_read_clock that will return the
672 * current device timestamp.
676 /* second cache line - fields only used in slow path or on TX */
677 MARKER cacheline1 __rte_cache_min_aligned;
681 void *userdata; /**< Can be used for external metadata */
682 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
685 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
686 struct rte_mbuf *next; /**< Next segment of scattered packet. */
688 /* fields to support TX offloads */
691 uint64_t tx_offload; /**< combined for easy fetch */
694 uint64_t l2_len:RTE_MBUF_L2_LEN_BITS;
695 /**< L2 (MAC) Header Length for non-tunneling pkt.
696 * Outer_L4_len + ... + Inner_L2_len for tunneling pkt.
698 uint64_t l3_len:RTE_MBUF_L3_LEN_BITS;
699 /**< L3 (IP) Header Length. */
700 uint64_t l4_len:RTE_MBUF_L4_LEN_BITS;
701 /**< L4 (TCP/UDP) Header Length. */
702 uint64_t tso_segsz:RTE_MBUF_TSO_SEGSZ_BITS;
703 /**< TCP TSO segment size */
705 /* fields for TX offloading of tunnels */
706 uint64_t outer_l3_len:RTE_MBUF_OUTL3_LEN_BITS;
707 /**< Outer L3 (IP) Hdr Length. */
708 uint64_t outer_l2_len:RTE_MBUF_OUTL2_LEN_BITS;
709 /**< Outer L2 (MAC) Hdr Length. */
711 /* uint64_t unused:RTE_MBUF_TXOFLD_UNUSED_BITS; */
715 /** Size of the application private data. In case of an indirect
716 * mbuf, it stores the direct mbuf private data size. */
719 /** Timesync flags for use with IEEE1588. */
722 /** Sequence number. See also rte_reorder_insert(). */
725 /** Shared data for external buffer attached to mbuf. See
726 * rte_pktmbuf_attach_extbuf().
728 struct rte_mbuf_ext_shared_info *shinfo;
730 } __rte_cache_aligned;
733 * Function typedef of callback to free externally attached buffer.
735 typedef void (*rte_mbuf_extbuf_free_callback_t)(void *addr, void *opaque);
738 * Shared data at the end of an external buffer.
740 struct rte_mbuf_ext_shared_info {
741 rte_mbuf_extbuf_free_callback_t free_cb; /**< Free callback function */
742 void *fcb_opaque; /**< Free callback argument */
743 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
746 /**< Maximum number of nb_segs allowed. */
747 #define RTE_MBUF_MAX_NB_SEGS UINT16_MAX
750 * Prefetch the first part of the mbuf
752 * The first 64 bytes of the mbuf corresponds to fields that are used early
753 * in the receive path. If the cache line of the architecture is higher than
754 * 64B, the second part will also be prefetched.
757 * The pointer to the mbuf.
760 rte_mbuf_prefetch_part1(struct rte_mbuf *m)
762 rte_prefetch0(&m->cacheline0);
766 * Prefetch the second part of the mbuf
768 * The next 64 bytes of the mbuf corresponds to fields that are used in the
769 * transmit path. If the cache line of the architecture is higher than 64B,
770 * this function does nothing as it is expected that the full mbuf is
774 * The pointer to the mbuf.
777 rte_mbuf_prefetch_part2(struct rte_mbuf *m)
779 #if RTE_CACHE_LINE_SIZE == 64
780 rte_prefetch0(&m->cacheline1);
787 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
790 * Return the IO address of the beginning of the mbuf data
793 * The pointer to the mbuf.
795 * The IO address of the beginning of the mbuf data
797 static inline rte_iova_t
798 rte_mbuf_data_iova(const struct rte_mbuf *mb)
800 return mb->buf_iova + mb->data_off;
804 static inline phys_addr_t
805 rte_mbuf_data_dma_addr(const struct rte_mbuf *mb)
807 return rte_mbuf_data_iova(mb);
811 * Return the default IO address of the beginning of the mbuf data
813 * This function is used by drivers in their receive function, as it
814 * returns the location where data should be written by the NIC, taking
815 * the default headroom in account.
818 * The pointer to the mbuf.
820 * The IO address of the beginning of the mbuf data
822 static inline rte_iova_t
823 rte_mbuf_data_iova_default(const struct rte_mbuf *mb)
825 return mb->buf_iova + RTE_PKTMBUF_HEADROOM;
829 static inline phys_addr_t
830 rte_mbuf_data_dma_addr_default(const struct rte_mbuf *mb)
832 return rte_mbuf_data_iova_default(mb);
836 * Return the mbuf owning the data buffer address of an indirect mbuf.
839 * The pointer to the indirect mbuf.
841 * The address of the direct mbuf corresponding to buffer_addr.
843 static inline struct rte_mbuf *
844 rte_mbuf_from_indirect(struct rte_mbuf *mi)
846 return (struct rte_mbuf *)RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
850 * Return address of buffer embedded in the given mbuf.
852 * The return value shall be same as mb->buf_addr if the mbuf is already
853 * initialized and direct. However, this API is useful if mempool of the
854 * mbuf is already known because it doesn't need to access mbuf contents in
855 * order to get the mempool pointer.
858 * @b EXPERIMENTAL: This API may change without prior notice.
859 * This will be used by rte_mbuf_to_baddr() which has redundant code once
860 * experimental tag is removed.
863 * The pointer to the mbuf.
865 * The pointer to the mempool of the mbuf.
867 * The pointer of the mbuf buffer.
871 rte_mbuf_buf_addr(struct rte_mbuf *mb, struct rte_mempool *mp)
873 return (char *)mb + sizeof(*mb) + rte_pktmbuf_priv_size(mp);
877 * Return the default address of the beginning of the mbuf data.
880 * @b EXPERIMENTAL: This API may change without prior notice.
883 * The pointer to the mbuf.
885 * The pointer of the beginning of the mbuf data.
889 rte_mbuf_data_addr_default(__rte_unused struct rte_mbuf *mb)
891 /* gcc complains about calling this experimental function even
892 * when not using it. Hide it with ALLOW_EXPERIMENTAL_API.
894 #ifdef ALLOW_EXPERIMENTAL_API
895 return rte_mbuf_buf_addr(mb, mb->pool) + RTE_PKTMBUF_HEADROOM;
902 * Return address of buffer embedded in the given mbuf.
904 * @note: Accessing mempool pointer of a mbuf is expensive because the
905 * pointer is stored in the 2nd cache line of mbuf. If mempool is known, it
906 * is better not to reference the mempool pointer in mbuf but calling
907 * rte_mbuf_buf_addr() would be more efficient.
910 * The pointer to the mbuf.
912 * The address of the data buffer owned by the mbuf.
915 rte_mbuf_to_baddr(struct rte_mbuf *md)
917 #ifdef ALLOW_EXPERIMENTAL_API
918 return rte_mbuf_buf_addr(md, md->pool);
921 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
927 * Return the starting address of the private data area embedded in
930 * Note that no check is made to ensure that a private data area
931 * actually exists in the supplied mbuf.
934 * The pointer to the mbuf.
936 * The starting address of the private data area of the given mbuf.
940 rte_mbuf_to_priv(struct rte_mbuf *m)
942 return RTE_PTR_ADD(m, sizeof(struct rte_mbuf));
946 * Returns TRUE if given mbuf is cloned by mbuf indirection, or FALSE
949 * If a mbuf has its data in another mbuf and references it by mbuf
950 * indirection, this mbuf can be defined as a cloned mbuf.
952 #define RTE_MBUF_CLONED(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
955 * Returns TRUE if given mbuf has an external buffer, or FALSE otherwise.
957 * External buffer is a user-provided anonymous buffer.
959 #define RTE_MBUF_HAS_EXTBUF(mb) ((mb)->ol_flags & EXT_ATTACHED_MBUF)
962 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
964 * If a mbuf embeds its own data after the rte_mbuf structure, this mbuf
965 * can be defined as a direct mbuf.
967 #define RTE_MBUF_DIRECT(mb) \
968 (!((mb)->ol_flags & (IND_ATTACHED_MBUF | EXT_ATTACHED_MBUF)))
971 * Private data in case of pktmbuf pool.
973 * A structure that contains some pktmbuf_pool-specific data that are
974 * appended after the mempool structure (in private data).
976 struct rte_pktmbuf_pool_private {
977 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
978 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
981 #ifdef RTE_LIBRTE_MBUF_DEBUG
983 /** check mbuf type in debug mode */
984 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
986 #else /* RTE_LIBRTE_MBUF_DEBUG */
988 /** check mbuf type in debug mode */
989 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
991 #endif /* RTE_LIBRTE_MBUF_DEBUG */
993 #ifdef RTE_MBUF_REFCNT_ATOMIC
996 * Reads the value of an mbuf's refcnt.
1000 * Reference count number.
1002 static inline uint16_t
1003 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
1005 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
1009 * Sets an mbuf's refcnt to a defined value.
1016 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
1018 rte_atomic16_set(&m->refcnt_atomic, (int16_t)new_value);
1022 static inline uint16_t
1023 __rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1025 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
1029 * Adds given value to an mbuf's refcnt and returns its new value.
1033 * Value to add/subtract
1037 static inline uint16_t
1038 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1041 * The atomic_add is an expensive operation, so we don't want to
1042 * call it in the case where we know we are the unique holder of
1043 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
1044 * operation has to be used because concurrent accesses on the
1045 * reference counter can occur.
1047 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
1049 rte_mbuf_refcnt_set(m, (uint16_t)value);
1050 return (uint16_t)value;
1053 return __rte_mbuf_refcnt_update(m, value);
1056 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
1059 static inline uint16_t
1060 __rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1062 m->refcnt = (uint16_t)(m->refcnt + value);
1067 * Adds given value to an mbuf's refcnt and returns its new value.
1069 static inline uint16_t
1070 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1072 return __rte_mbuf_refcnt_update(m, value);
1076 * Reads the value of an mbuf's refcnt.
1078 static inline uint16_t
1079 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
1085 * Sets an mbuf's refcnt to the defined value.
1088 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
1090 m->refcnt = new_value;
1093 #endif /* RTE_MBUF_REFCNT_ATOMIC */
1096 * Reads the refcnt of an external buffer.
1099 * Shared data of the external buffer.
1101 * Reference count number.
1103 static inline uint16_t
1104 rte_mbuf_ext_refcnt_read(const struct rte_mbuf_ext_shared_info *shinfo)
1106 return (uint16_t)(rte_atomic16_read(&shinfo->refcnt_atomic));
1110 * Set refcnt of an external buffer.
1113 * Shared data of the external buffer.
1118 rte_mbuf_ext_refcnt_set(struct rte_mbuf_ext_shared_info *shinfo,
1121 rte_atomic16_set(&shinfo->refcnt_atomic, (int16_t)new_value);
1125 * Add given value to refcnt of an external buffer and return its new
1129 * Shared data of the external buffer.
1131 * Value to add/subtract
1135 static inline uint16_t
1136 rte_mbuf_ext_refcnt_update(struct rte_mbuf_ext_shared_info *shinfo,
1139 if (likely(rte_mbuf_ext_refcnt_read(shinfo) == 1)) {
1141 rte_mbuf_ext_refcnt_set(shinfo, (uint16_t)value);
1142 return (uint16_t)value;
1145 return (uint16_t)rte_atomic16_add_return(&shinfo->refcnt_atomic, value);
1148 /** Mbuf prefetch */
1149 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
1156 * Sanity checks on an mbuf.
1158 * Check the consistency of the given mbuf. The function will cause a
1159 * panic if corruption is detected.
1162 * The mbuf to be checked.
1164 * True if the mbuf is a packet header, false if it is a sub-segment
1165 * of a packet (in this case, some fields like nb_segs are not checked)
1168 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
1171 * Sanity checks on a mbuf.
1173 * Almost like rte_mbuf_sanity_check(), but this function gives the reason
1174 * if corruption is detected rather than panic.
1177 * The mbuf to be checked.
1179 * True if the mbuf is a packet header, false if it is a sub-segment
1180 * of a packet (in this case, some fields like nb_segs are not checked)
1182 * A reference to a string pointer where to store the reason why a mbuf is
1183 * considered invalid.
1185 * - 0 if no issue has been found, reason is left untouched.
1186 * - -1 if a problem is detected, reason then points to a string describing
1187 * the reason why the mbuf is deemed invalid.
1190 int rte_mbuf_check(const struct rte_mbuf *m, int is_header,
1191 const char **reason);
1193 #define MBUF_RAW_ALLOC_CHECK(m) do { \
1194 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1); \
1195 RTE_ASSERT((m)->next == NULL); \
1196 RTE_ASSERT((m)->nb_segs == 1); \
1197 __rte_mbuf_sanity_check(m, 0); \
1201 * Allocate an uninitialized mbuf from mempool *mp*.
1203 * This function can be used by PMDs (especially in RX functions) to
1204 * allocate an uninitialized mbuf. The driver is responsible of
1205 * initializing all the required fields. See rte_pktmbuf_reset().
1206 * For standard needs, prefer rte_pktmbuf_alloc().
1208 * The caller can expect that the following fields of the mbuf structure
1209 * are initialized: buf_addr, buf_iova, buf_len, refcnt=1, nb_segs=1,
1210 * next=NULL, pool, priv_size. The other fields must be initialized
1214 * The mempool from which mbuf is allocated.
1216 * - The pointer to the new mbuf on success.
1217 * - NULL if allocation failed.
1219 static inline struct rte_mbuf *rte_mbuf_raw_alloc(struct rte_mempool *mp)
1223 if (rte_mempool_get(mp, (void **)&m) < 0)
1225 MBUF_RAW_ALLOC_CHECK(m);
1230 * Put mbuf back into its original mempool.
1232 * The caller must ensure that the mbuf is direct and properly
1233 * reinitialized (refcnt=1, next=NULL, nb_segs=1), as done by
1234 * rte_pktmbuf_prefree_seg().
1236 * This function should be used with care, when optimization is
1237 * required. For standard needs, prefer rte_pktmbuf_free() or
1238 * rte_pktmbuf_free_seg().
1241 * The mbuf to be freed.
1243 static __rte_always_inline void
1244 rte_mbuf_raw_free(struct rte_mbuf *m)
1246 RTE_ASSERT(RTE_MBUF_DIRECT(m));
1247 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1);
1248 RTE_ASSERT(m->next == NULL);
1249 RTE_ASSERT(m->nb_segs == 1);
1250 __rte_mbuf_sanity_check(m, 0);
1251 rte_mempool_put(m->pool, m);
1255 * The packet mbuf constructor.
1257 * This function initializes some fields in the mbuf structure that are
1258 * not modified by the user once created (origin pool, buffer start
1259 * address, and so on). This function is given as a callback function to
1260 * rte_mempool_obj_iter() or rte_mempool_create() at pool creation time.
1263 * The mempool from which mbufs originate.
1265 * A pointer that can be used by the user to retrieve useful information
1266 * for mbuf initialization. This pointer is the opaque argument passed to
1267 * rte_mempool_obj_iter() or rte_mempool_create().
1269 * The mbuf to initialize.
1271 * The index of the mbuf in the pool table.
1273 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1274 void *m, unsigned i);
1278 * A packet mbuf pool constructor.
1280 * This function initializes the mempool private data in the case of a
1281 * pktmbuf pool. This private data is needed by the driver. The
1282 * function must be called on the mempool before it is used, or it
1283 * can be given as a callback function to rte_mempool_create() at
1284 * pool creation. It can be extended by the user, for example, to
1285 * provide another packet size.
1288 * The mempool from which mbufs originate.
1290 * A pointer that can be used by the user to retrieve useful information
1291 * for mbuf initialization. This pointer is the opaque argument passed to
1292 * rte_mempool_create().
1294 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
1297 * Create a mbuf pool.
1299 * This function creates and initializes a packet mbuf pool. It is
1300 * a wrapper to rte_mempool functions.
1303 * The name of the mbuf pool.
1305 * The number of elements in the mbuf pool. The optimum size (in terms
1306 * of memory usage) for a mempool is when n is a power of two minus one:
1309 * Size of the per-core object cache. See rte_mempool_create() for
1312 * Size of application private are between the rte_mbuf structure
1313 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
1314 * @param data_room_size
1315 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
1317 * The socket identifier where the memory should be allocated. The
1318 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
1321 * The pointer to the new allocated mempool, on success. NULL on error
1322 * with rte_errno set appropriately. Possible rte_errno values include:
1323 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
1324 * - E_RTE_SECONDARY - function was called from a secondary process instance
1325 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
1326 * - ENOSPC - the maximum number of memzones has already been allocated
1327 * - EEXIST - a memzone with the same name already exists
1328 * - ENOMEM - no appropriate memory area found in which to create memzone
1330 struct rte_mempool *
1331 rte_pktmbuf_pool_create(const char *name, unsigned n,
1332 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
1336 * Create a mbuf pool with a given mempool ops name
1338 * This function creates and initializes a packet mbuf pool. It is
1339 * a wrapper to rte_mempool functions.
1342 * The name of the mbuf pool.
1344 * The number of elements in the mbuf pool. The optimum size (in terms
1345 * of memory usage) for a mempool is when n is a power of two minus one:
1348 * Size of the per-core object cache. See rte_mempool_create() for
1351 * Size of application private are between the rte_mbuf structure
1352 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
1353 * @param data_room_size
1354 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
1356 * The socket identifier where the memory should be allocated. The
1357 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
1360 * The mempool ops name to be used for this mempool instead of
1361 * default mempool. The value can be *NULL* to use default mempool.
1363 * The pointer to the new allocated mempool, on success. NULL on error
1364 * with rte_errno set appropriately. Possible rte_errno values include:
1365 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
1366 * - E_RTE_SECONDARY - function was called from a secondary process instance
1367 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
1368 * - ENOSPC - the maximum number of memzones has already been allocated
1369 * - EEXIST - a memzone with the same name already exists
1370 * - ENOMEM - no appropriate memory area found in which to create memzone
1372 struct rte_mempool *
1373 rte_pktmbuf_pool_create_by_ops(const char *name, unsigned int n,
1374 unsigned int cache_size, uint16_t priv_size, uint16_t data_room_size,
1375 int socket_id, const char *ops_name);
1378 * Get the data room size of mbufs stored in a pktmbuf_pool
1380 * The data room size is the amount of data that can be stored in a
1381 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
1384 * The packet mbuf pool.
1386 * The data room size of mbufs stored in this mempool.
1388 static inline uint16_t
1389 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
1391 struct rte_pktmbuf_pool_private *mbp_priv;
1393 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1394 return mbp_priv->mbuf_data_room_size;
1398 * Get the application private size of mbufs stored in a pktmbuf_pool
1400 * The private size of mbuf is a zone located between the rte_mbuf
1401 * structure and the data buffer where an application can store data
1402 * associated to a packet.
1405 * The packet mbuf pool.
1407 * The private size of mbufs stored in this mempool.
1409 static inline uint16_t
1410 rte_pktmbuf_priv_size(struct rte_mempool *mp)
1412 struct rte_pktmbuf_pool_private *mbp_priv;
1414 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1415 return mbp_priv->mbuf_priv_size;
1419 * Reset the data_off field of a packet mbuf to its default value.
1421 * The given mbuf must have only one segment, which should be empty.
1424 * The packet mbuf's data_off field has to be reset.
1426 static inline void rte_pktmbuf_reset_headroom(struct rte_mbuf *m)
1428 m->data_off = (uint16_t)RTE_MIN((uint16_t)RTE_PKTMBUF_HEADROOM,
1429 (uint16_t)m->buf_len);
1433 * Reset the fields of a packet mbuf to their default values.
1435 * The given mbuf must have only one segment.
1438 * The packet mbuf to be reset.
1440 #define MBUF_INVALID_PORT UINT16_MAX
1442 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
1448 m->vlan_tci_outer = 0;
1450 m->port = MBUF_INVALID_PORT;
1454 rte_pktmbuf_reset_headroom(m);
1457 __rte_mbuf_sanity_check(m, 1);
1461 * Allocate a new mbuf from a mempool.
1463 * This new mbuf contains one segment, which has a length of 0. The pointer
1464 * to data is initialized to have some bytes of headroom in the buffer
1465 * (if buffer size allows).
1468 * The mempool from which the mbuf is allocated.
1470 * - The pointer to the new mbuf on success.
1471 * - NULL if allocation failed.
1473 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
1476 if ((m = rte_mbuf_raw_alloc(mp)) != NULL)
1477 rte_pktmbuf_reset(m);
1482 * Allocate a bulk of mbufs, initialize refcnt and reset the fields to default
1486 * The mempool from which mbufs are allocated.
1488 * Array of pointers to mbufs
1493 * - -ENOENT: Not enough entries in the mempool; no mbufs are retrieved.
1495 static inline int rte_pktmbuf_alloc_bulk(struct rte_mempool *pool,
1496 struct rte_mbuf **mbufs, unsigned count)
1501 rc = rte_mempool_get_bulk(pool, (void **)mbufs, count);
1505 /* To understand duff's device on loop unwinding optimization, see
1506 * https://en.wikipedia.org/wiki/Duff's_device.
1507 * Here while() loop is used rather than do() while{} to avoid extra
1508 * check if count is zero.
1510 switch (count % 4) {
1512 while (idx != count) {
1513 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1514 rte_pktmbuf_reset(mbufs[idx]);
1518 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1519 rte_pktmbuf_reset(mbufs[idx]);
1523 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1524 rte_pktmbuf_reset(mbufs[idx]);
1528 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1529 rte_pktmbuf_reset(mbufs[idx]);
1538 * Initialize shared data at the end of an external buffer before attaching
1539 * to a mbuf by ``rte_pktmbuf_attach_extbuf()``. This is not a mandatory
1540 * initialization but a helper function to simply spare a few bytes at the
1541 * end of the buffer for shared data. If shared data is allocated
1542 * separately, this should not be called but application has to properly
1543 * initialize the shared data according to its need.
1545 * Free callback and its argument is saved and the refcnt is set to 1.
1548 * The value of buf_len will be reduced to RTE_PTR_DIFF(shinfo, buf_addr)
1549 * after this initialization. This shall be used for
1550 * ``rte_pktmbuf_attach_extbuf()``
1553 * The pointer to the external buffer.
1554 * @param [in,out] buf_len
1555 * The pointer to length of the external buffer. Input value must be
1556 * larger than the size of ``struct rte_mbuf_ext_shared_info`` and
1557 * padding for alignment. If not enough, this function will return NULL.
1558 * Adjusted buffer length will be returned through this pointer.
1560 * Free callback function to call when the external buffer needs to be
1563 * Argument for the free callback function.
1566 * A pointer to the initialized shared data on success, return NULL
1569 static inline struct rte_mbuf_ext_shared_info *
1570 rte_pktmbuf_ext_shinfo_init_helper(void *buf_addr, uint16_t *buf_len,
1571 rte_mbuf_extbuf_free_callback_t free_cb, void *fcb_opaque)
1573 struct rte_mbuf_ext_shared_info *shinfo;
1574 void *buf_end = RTE_PTR_ADD(buf_addr, *buf_len);
1577 addr = RTE_PTR_ALIGN_FLOOR(RTE_PTR_SUB(buf_end, sizeof(*shinfo)),
1579 if (addr <= buf_addr)
1582 shinfo = (struct rte_mbuf_ext_shared_info *)addr;
1583 shinfo->free_cb = free_cb;
1584 shinfo->fcb_opaque = fcb_opaque;
1585 rte_mbuf_ext_refcnt_set(shinfo, 1);
1587 *buf_len = (uint16_t)RTE_PTR_DIFF(shinfo, buf_addr);
1592 * Attach an external buffer to a mbuf.
1594 * User-managed anonymous buffer can be attached to an mbuf. When attaching
1595 * it, corresponding free callback function and its argument should be
1596 * provided via shinfo. This callback function will be called once all the
1597 * mbufs are detached from the buffer (refcnt becomes zero).
1599 * The headroom for the attaching mbuf will be set to zero and this can be
1600 * properly adjusted after attachment. For example, ``rte_pktmbuf_adj()``
1601 * or ``rte_pktmbuf_reset_headroom()`` might be used.
1603 * More mbufs can be attached to the same external buffer by
1604 * ``rte_pktmbuf_attach()`` once the external buffer has been attached by
1607 * Detachment can be done by either ``rte_pktmbuf_detach_extbuf()`` or
1608 * ``rte_pktmbuf_detach()``.
1610 * Memory for shared data must be provided and user must initialize all of
1611 * the content properly, especially free callback and refcnt. The pointer
1612 * of shared data will be stored in m->shinfo.
1613 * ``rte_pktmbuf_ext_shinfo_init_helper`` can help to simply spare a few
1614 * bytes at the end of buffer for the shared data, store free callback and
1615 * its argument and set the refcnt to 1. The following is an example:
1617 * struct rte_mbuf_ext_shared_info *shinfo =
1618 * rte_pktmbuf_ext_shinfo_init_helper(buf_addr, &buf_len,
1619 * free_cb, fcb_arg);
1620 * rte_pktmbuf_attach_extbuf(m, buf_addr, buf_iova, buf_len, shinfo);
1621 * rte_pktmbuf_reset_headroom(m);
1622 * rte_pktmbuf_adj(m, data_len);
1624 * Attaching an external buffer is quite similar to mbuf indirection in
1625 * replacing buffer addresses and length of a mbuf, but a few differences:
1626 * - When an indirect mbuf is attached, refcnt of the direct mbuf would be
1627 * 2 as long as the direct mbuf itself isn't freed after the attachment.
1628 * In such cases, the buffer area of a direct mbuf must be read-only. But
1629 * external buffer has its own refcnt and it starts from 1. Unless
1630 * multiple mbufs are attached to a mbuf having an external buffer, the
1631 * external buffer is writable.
1632 * - There's no need to allocate buffer from a mempool. Any buffer can be
1633 * attached with appropriate free callback and its IO address.
1634 * - Smaller metadata is required to maintain shared data such as refcnt.
1637 * The pointer to the mbuf.
1639 * The pointer to the external buffer.
1641 * IO address of the external buffer.
1643 * The size of the external buffer.
1645 * User-provided memory for shared data of the external buffer.
1648 rte_pktmbuf_attach_extbuf(struct rte_mbuf *m, void *buf_addr,
1649 rte_iova_t buf_iova, uint16_t buf_len,
1650 struct rte_mbuf_ext_shared_info *shinfo)
1652 /* mbuf should not be read-only */
1653 RTE_ASSERT(RTE_MBUF_DIRECT(m) && rte_mbuf_refcnt_read(m) == 1);
1654 RTE_ASSERT(shinfo->free_cb != NULL);
1656 m->buf_addr = buf_addr;
1657 m->buf_iova = buf_iova;
1658 m->buf_len = buf_len;
1663 m->ol_flags |= EXT_ATTACHED_MBUF;
1668 * Detach the external buffer attached to a mbuf, same as
1669 * ``rte_pktmbuf_detach()``
1672 * The mbuf having external buffer.
1674 #define rte_pktmbuf_detach_extbuf(m) rte_pktmbuf_detach(m)
1677 * Attach packet mbuf to another packet mbuf.
1679 * If the mbuf we are attaching to isn't a direct buffer and is attached to
1680 * an external buffer, the mbuf being attached will be attached to the
1681 * external buffer instead of mbuf indirection.
1683 * Otherwise, the mbuf will be indirectly attached. After attachment we
1684 * refer the mbuf we attached as 'indirect', while mbuf we attached to as
1685 * 'direct'. The direct mbuf's reference counter is incremented.
1687 * Right now, not supported:
1688 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
1689 * - mbuf we trying to attach (mi) is used by someone else
1690 * e.g. it's reference counter is greater then 1.
1693 * The indirect packet mbuf.
1695 * The packet mbuf we're attaching to.
1697 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
1699 RTE_ASSERT(RTE_MBUF_DIRECT(mi) &&
1700 rte_mbuf_refcnt_read(mi) == 1);
1702 if (RTE_MBUF_HAS_EXTBUF(m)) {
1703 rte_mbuf_ext_refcnt_update(m->shinfo, 1);
1704 mi->ol_flags = m->ol_flags;
1705 mi->shinfo = m->shinfo;
1707 /* if m is not direct, get the mbuf that embeds the data */
1708 rte_mbuf_refcnt_update(rte_mbuf_from_indirect(m), 1);
1709 mi->priv_size = m->priv_size;
1710 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
1713 mi->buf_iova = m->buf_iova;
1714 mi->buf_addr = m->buf_addr;
1715 mi->buf_len = m->buf_len;
1717 mi->data_off = m->data_off;
1718 mi->data_len = m->data_len;
1720 mi->vlan_tci = m->vlan_tci;
1721 mi->vlan_tci_outer = m->vlan_tci_outer;
1722 mi->tx_offload = m->tx_offload;
1726 mi->pkt_len = mi->data_len;
1728 mi->packet_type = m->packet_type;
1729 mi->timestamp = m->timestamp;
1731 __rte_mbuf_sanity_check(mi, 1);
1732 __rte_mbuf_sanity_check(m, 0);
1736 * @internal used by rte_pktmbuf_detach().
1738 * Decrement the reference counter of the external buffer. When the
1739 * reference counter becomes 0, the buffer is freed by pre-registered
1743 __rte_pktmbuf_free_extbuf(struct rte_mbuf *m)
1745 RTE_ASSERT(RTE_MBUF_HAS_EXTBUF(m));
1746 RTE_ASSERT(m->shinfo != NULL);
1748 if (rte_mbuf_ext_refcnt_update(m->shinfo, -1) == 0)
1749 m->shinfo->free_cb(m->buf_addr, m->shinfo->fcb_opaque);
1753 * @internal used by rte_pktmbuf_detach().
1755 * Decrement the direct mbuf's reference counter. When the reference
1756 * counter becomes 0, the direct mbuf is freed.
1759 __rte_pktmbuf_free_direct(struct rte_mbuf *m)
1761 struct rte_mbuf *md;
1763 RTE_ASSERT(RTE_MBUF_CLONED(m));
1765 md = rte_mbuf_from_indirect(m);
1767 if (rte_mbuf_refcnt_update(md, -1) == 0) {
1770 rte_mbuf_refcnt_set(md, 1);
1771 rte_mbuf_raw_free(md);
1776 * Detach a packet mbuf from external buffer or direct buffer.
1778 * - decrement refcnt and free the external/direct buffer if refcnt
1780 * - restore original mbuf address and length values.
1781 * - reset pktmbuf data and data_len to their default values.
1783 * All other fields of the given packet mbuf will be left intact.
1786 * The indirect attached packet mbuf.
1788 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
1790 struct rte_mempool *mp = m->pool;
1791 uint32_t mbuf_size, buf_len;
1794 if (RTE_MBUF_HAS_EXTBUF(m))
1795 __rte_pktmbuf_free_extbuf(m);
1797 __rte_pktmbuf_free_direct(m);
1799 priv_size = rte_pktmbuf_priv_size(mp);
1800 mbuf_size = (uint32_t)(sizeof(struct rte_mbuf) + priv_size);
1801 buf_len = rte_pktmbuf_data_room_size(mp);
1803 m->priv_size = priv_size;
1804 m->buf_addr = (char *)m + mbuf_size;
1805 m->buf_iova = rte_mempool_virt2iova(m) + mbuf_size;
1806 m->buf_len = (uint16_t)buf_len;
1807 rte_pktmbuf_reset_headroom(m);
1813 * Decrease reference counter and unlink a mbuf segment
1815 * This function does the same than a free, except that it does not
1816 * return the segment to its pool.
1817 * It decreases the reference counter, and if it reaches 0, it is
1818 * detached from its parent for an indirect mbuf.
1821 * The mbuf to be unlinked
1823 * - (m) if it is the last reference. It can be recycled or freed.
1824 * - (NULL) if the mbuf still has remaining references on it.
1826 static __rte_always_inline struct rte_mbuf *
1827 rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1829 __rte_mbuf_sanity_check(m, 0);
1831 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
1833 if (!RTE_MBUF_DIRECT(m))
1834 rte_pktmbuf_detach(m);
1836 if (m->next != NULL) {
1843 } else if (__rte_mbuf_refcnt_update(m, -1) == 0) {
1845 if (!RTE_MBUF_DIRECT(m))
1846 rte_pktmbuf_detach(m);
1848 if (m->next != NULL) {
1852 rte_mbuf_refcnt_set(m, 1);
1860 * Free a segment of a packet mbuf into its original mempool.
1862 * Free an mbuf, without parsing other segments in case of chained
1866 * The packet mbuf segment to be freed.
1868 static __rte_always_inline void
1869 rte_pktmbuf_free_seg(struct rte_mbuf *m)
1871 m = rte_pktmbuf_prefree_seg(m);
1872 if (likely(m != NULL))
1873 rte_mbuf_raw_free(m);
1877 * Free a packet mbuf back into its original mempool.
1879 * Free an mbuf, and all its segments in case of chained buffers. Each
1880 * segment is added back into its original mempool.
1883 * The packet mbuf to be freed. If NULL, the function does nothing.
1885 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
1887 struct rte_mbuf *m_next;
1890 __rte_mbuf_sanity_check(m, 1);
1894 rte_pktmbuf_free_seg(m);
1900 * Creates a "clone" of the given packet mbuf.
1902 * Walks through all segments of the given packet mbuf, and for each of them:
1903 * - Creates a new packet mbuf from the given pool.
1904 * - Attaches newly created mbuf to the segment.
1905 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
1906 * from the original packet mbuf.
1909 * The packet mbuf to be cloned.
1911 * The mempool from which the "clone" mbufs are allocated.
1913 * - The pointer to the new "clone" mbuf on success.
1914 * - NULL if allocation fails.
1916 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
1917 struct rte_mempool *mp)
1919 struct rte_mbuf *mc, *mi, **prev;
1923 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
1928 pktlen = md->pkt_len;
1933 rte_pktmbuf_attach(mi, md);
1936 } while ((md = md->next) != NULL &&
1937 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1941 mc->pkt_len = pktlen;
1943 /* Allocation of new indirect segment failed */
1944 if (unlikely (mi == NULL)) {
1945 rte_pktmbuf_free(mc);
1949 __rte_mbuf_sanity_check(mc, 1);
1954 * Adds given value to the refcnt of all packet mbuf segments.
1956 * Walks through all segments of given packet mbuf and for each of them
1957 * invokes rte_mbuf_refcnt_update().
1960 * The packet mbuf whose refcnt to be updated.
1962 * The value to add to the mbuf's segments refcnt.
1964 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1966 __rte_mbuf_sanity_check(m, 1);
1969 rte_mbuf_refcnt_update(m, v);
1970 } while ((m = m->next) != NULL);
1974 * Get the headroom in a packet mbuf.
1979 * The length of the headroom.
1981 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1983 __rte_mbuf_sanity_check(m, 0);
1988 * Get the tailroom of a packet mbuf.
1993 * The length of the tailroom.
1995 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1997 __rte_mbuf_sanity_check(m, 0);
1998 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
2003 * Get the last segment of the packet.
2008 * The last segment of the given mbuf.
2010 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
2012 __rte_mbuf_sanity_check(m, 1);
2013 while (m->next != NULL)
2019 * A macro that points to an offset into the data in the mbuf.
2021 * The returned pointer is cast to type t. Before using this
2022 * function, the user must ensure that the first segment is large
2023 * enough to accommodate its data.
2028 * The offset into the mbuf data.
2030 * The type to cast the result into.
2032 #define rte_pktmbuf_mtod_offset(m, t, o) \
2033 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
2036 * A macro that points to the start of the data in the mbuf.
2038 * The returned pointer is cast to type t. Before using this
2039 * function, the user must ensure that the first segment is large
2040 * enough to accommodate its data.
2045 * The type to cast the result into.
2047 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
2050 * A macro that returns the IO address that points to an offset of the
2051 * start of the data in the mbuf
2056 * The offset into the data to calculate address from.
2058 #define rte_pktmbuf_iova_offset(m, o) \
2059 (rte_iova_t)((m)->buf_iova + (m)->data_off + (o))
2062 #define rte_pktmbuf_mtophys_offset(m, o) \
2063 rte_pktmbuf_iova_offset(m, o)
2066 * A macro that returns the IO address that points to the start of the
2072 #define rte_pktmbuf_iova(m) rte_pktmbuf_iova_offset(m, 0)
2075 #define rte_pktmbuf_mtophys(m) rte_pktmbuf_iova(m)
2078 * A macro that returns the length of the packet.
2080 * The value can be read or assigned.
2085 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
2088 * A macro that returns the length of the segment.
2090 * The value can be read or assigned.
2095 #define rte_pktmbuf_data_len(m) ((m)->data_len)
2098 * Prepend len bytes to an mbuf data area.
2100 * Returns a pointer to the new
2101 * data start address. If there is not enough headroom in the first
2102 * segment, the function will return NULL, without modifying the mbuf.
2107 * The amount of data to prepend (in bytes).
2109 * A pointer to the start of the newly prepended data, or
2110 * NULL if there is not enough headroom space in the first segment
2112 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
2115 __rte_mbuf_sanity_check(m, 1);
2117 if (unlikely(len > rte_pktmbuf_headroom(m)))
2120 /* NB: elaborating the subtraction like this instead of using
2121 * -= allows us to ensure the result type is uint16_t
2122 * avoiding compiler warnings on gcc 8.1 at least */
2123 m->data_off = (uint16_t)(m->data_off - len);
2124 m->data_len = (uint16_t)(m->data_len + len);
2125 m->pkt_len = (m->pkt_len + len);
2127 return (char *)m->buf_addr + m->data_off;
2131 * Append len bytes to an mbuf.
2133 * Append len bytes to an mbuf and return a pointer to the start address
2134 * of the added data. If there is not enough tailroom in the last
2135 * segment, the function will return NULL, without modifying the mbuf.
2140 * The amount of data to append (in bytes).
2142 * A pointer to the start of the newly appended data, or
2143 * NULL if there is not enough tailroom space in the last segment
2145 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
2148 struct rte_mbuf *m_last;
2150 __rte_mbuf_sanity_check(m, 1);
2152 m_last = rte_pktmbuf_lastseg(m);
2153 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
2156 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
2157 m_last->data_len = (uint16_t)(m_last->data_len + len);
2158 m->pkt_len = (m->pkt_len + len);
2159 return (char*) tail;
2163 * Remove len bytes at the beginning of an mbuf.
2165 * Returns a pointer to the start address of the new data area. If the
2166 * length is greater than the length of the first segment, then the
2167 * function will fail and return NULL, without modifying the mbuf.
2172 * The amount of data to remove (in bytes).
2174 * A pointer to the new start of the data.
2176 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
2178 __rte_mbuf_sanity_check(m, 1);
2180 if (unlikely(len > m->data_len))
2183 /* NB: elaborating the addition like this instead of using
2184 * += allows us to ensure the result type is uint16_t
2185 * avoiding compiler warnings on gcc 8.1 at least */
2186 m->data_len = (uint16_t)(m->data_len - len);
2187 m->data_off = (uint16_t)(m->data_off + len);
2188 m->pkt_len = (m->pkt_len - len);
2189 return (char *)m->buf_addr + m->data_off;
2193 * Remove len bytes of data at the end of the mbuf.
2195 * If the length is greater than the length of the last segment, the
2196 * function will fail and return -1 without modifying the mbuf.
2201 * The amount of data to remove (in bytes).
2206 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
2208 struct rte_mbuf *m_last;
2210 __rte_mbuf_sanity_check(m, 1);
2212 m_last = rte_pktmbuf_lastseg(m);
2213 if (unlikely(len > m_last->data_len))
2216 m_last->data_len = (uint16_t)(m_last->data_len - len);
2217 m->pkt_len = (m->pkt_len - len);
2222 * Test if mbuf data is contiguous.
2227 * - 1, if all data is contiguous (one segment).
2228 * - 0, if there is several segments.
2230 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
2232 __rte_mbuf_sanity_check(m, 1);
2233 return !!(m->nb_segs == 1);
2237 * @internal used by rte_pktmbuf_read().
2239 const void *__rte_pktmbuf_read(const struct rte_mbuf *m, uint32_t off,
2240 uint32_t len, void *buf);
2243 * Read len data bytes in a mbuf at specified offset.
2245 * If the data is contiguous, return the pointer in the mbuf data, else
2246 * copy the data in the buffer provided by the user and return its
2250 * The pointer to the mbuf.
2252 * The offset of the data in the mbuf.
2254 * The amount of bytes to read.
2256 * The buffer where data is copied if it is not contiguous in mbuf
2257 * data. Its length should be at least equal to the len parameter.
2259 * The pointer to the data, either in the mbuf if it is contiguous,
2260 * or in the user buffer. If mbuf is too small, NULL is returned.
2262 static inline const void *rte_pktmbuf_read(const struct rte_mbuf *m,
2263 uint32_t off, uint32_t len, void *buf)
2265 if (likely(off + len <= rte_pktmbuf_data_len(m)))
2266 return rte_pktmbuf_mtod_offset(m, char *, off);
2268 return __rte_pktmbuf_read(m, off, len, buf);
2272 * Chain an mbuf to another, thereby creating a segmented packet.
2274 * Note: The implementation will do a linear walk over the segments to find
2275 * the tail entry. For cases when there are many segments, it's better to
2276 * chain the entries manually.
2279 * The head of the mbuf chain (the first packet)
2281 * The mbuf to put last in the chain
2285 * - -EOVERFLOW, if the chain segment limit exceeded
2287 static inline int rte_pktmbuf_chain(struct rte_mbuf *head, struct rte_mbuf *tail)
2289 struct rte_mbuf *cur_tail;
2291 /* Check for number-of-segments-overflow */
2292 if (head->nb_segs + tail->nb_segs > RTE_MBUF_MAX_NB_SEGS)
2295 /* Chain 'tail' onto the old tail */
2296 cur_tail = rte_pktmbuf_lastseg(head);
2297 cur_tail->next = tail;
2299 /* accumulate number of segments and total length.
2300 * NB: elaborating the addition like this instead of using
2301 * -= allows us to ensure the result type is uint16_t
2302 * avoiding compiler warnings on gcc 8.1 at least */
2303 head->nb_segs = (uint16_t)(head->nb_segs + tail->nb_segs);
2304 head->pkt_len += tail->pkt_len;
2306 /* pkt_len is only set in the head */
2307 tail->pkt_len = tail->data_len;
2314 * @b EXPERIMENTAL: This API may change without prior notice.
2316 * For given input values generate raw tx_offload value.
2317 * Note that it is caller responsibility to make sure that input parameters
2318 * don't exceed maximum bit-field values.
2328 * outer_l3_len value.
2330 * outer_l2_len value.
2334 * raw tx_offload value.
2336 static __rte_always_inline uint64_t
2337 rte_mbuf_tx_offload(uint64_t il2, uint64_t il3, uint64_t il4, uint64_t tso,
2338 uint64_t ol3, uint64_t ol2, uint64_t unused)
2340 return il2 << RTE_MBUF_L2_LEN_OFS |
2341 il3 << RTE_MBUF_L3_LEN_OFS |
2342 il4 << RTE_MBUF_L4_LEN_OFS |
2343 tso << RTE_MBUF_TSO_SEGSZ_OFS |
2344 ol3 << RTE_MBUF_OUTL3_LEN_OFS |
2345 ol2 << RTE_MBUF_OUTL2_LEN_OFS |
2346 unused << RTE_MBUF_TXOFLD_UNUSED_OFS;
2350 * Validate general requirements for Tx offload in mbuf.
2352 * This function checks correctness and completeness of Tx offload settings.
2355 * The packet mbuf to be validated.
2357 * 0 if packet is valid
2360 rte_validate_tx_offload(const struct rte_mbuf *m)
2362 uint64_t ol_flags = m->ol_flags;
2364 /* Does packet set any of available offloads? */
2365 if (!(ol_flags & PKT_TX_OFFLOAD_MASK))
2368 /* IP checksum can be counted only for IPv4 packet */
2369 if ((ol_flags & PKT_TX_IP_CKSUM) && (ol_flags & PKT_TX_IPV6))
2372 /* IP type not set when required */
2373 if (ol_flags & (PKT_TX_L4_MASK | PKT_TX_TCP_SEG))
2374 if (!(ol_flags & (PKT_TX_IPV4 | PKT_TX_IPV6)))
2377 /* Check requirements for TSO packet */
2378 if (ol_flags & PKT_TX_TCP_SEG)
2379 if ((m->tso_segsz == 0) ||
2380 ((ol_flags & PKT_TX_IPV4) &&
2381 !(ol_flags & PKT_TX_IP_CKSUM)))
2384 /* PKT_TX_OUTER_IP_CKSUM set for non outer IPv4 packet. */
2385 if ((ol_flags & PKT_TX_OUTER_IP_CKSUM) &&
2386 !(ol_flags & PKT_TX_OUTER_IPV4))
2393 * Linearize data in mbuf.
2395 * This function moves the mbuf data in the first segment if there is enough
2396 * tailroom. The subsequent segments are unchained and freed.
2405 rte_pktmbuf_linearize(struct rte_mbuf *mbuf)
2407 size_t seg_len, copy_len;
2409 struct rte_mbuf *m_next;
2412 if (rte_pktmbuf_is_contiguous(mbuf))
2415 /* Extend first segment to the total packet length */
2416 copy_len = rte_pktmbuf_pkt_len(mbuf) - rte_pktmbuf_data_len(mbuf);
2418 if (unlikely(copy_len > rte_pktmbuf_tailroom(mbuf)))
2421 buffer = rte_pktmbuf_mtod_offset(mbuf, char *, mbuf->data_len);
2422 mbuf->data_len = (uint16_t)(mbuf->pkt_len);
2424 /* Append data from next segments to the first one */
2429 seg_len = rte_pktmbuf_data_len(m);
2430 rte_memcpy(buffer, rte_pktmbuf_mtod(m, char *), seg_len);
2433 rte_pktmbuf_free_seg(m);
2444 * Dump an mbuf structure to a file.
2446 * Dump all fields for the given packet mbuf and all its associated
2447 * segments (in the case of a chained buffer).
2450 * A pointer to a file for output
2454 * If dump_len != 0, also dump the "dump_len" first data bytes of
2457 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
2460 * Get the value of mbuf sched queue_id field.
2462 static inline uint32_t
2463 rte_mbuf_sched_queue_get(const struct rte_mbuf *m)
2465 return m->hash.sched.queue_id;
2469 * Get the value of mbuf sched traffic_class field.
2471 static inline uint8_t
2472 rte_mbuf_sched_traffic_class_get(const struct rte_mbuf *m)
2474 return m->hash.sched.traffic_class;
2478 * Get the value of mbuf sched color field.
2480 static inline uint8_t
2481 rte_mbuf_sched_color_get(const struct rte_mbuf *m)
2483 return m->hash.sched.color;
2487 * Get the values of mbuf sched queue_id, traffic_class and color.
2492 * Returns the queue id
2493 * @param traffic_class
2494 * Returns the traffic class id
2496 * Returns the colour id
2499 rte_mbuf_sched_get(const struct rte_mbuf *m, uint32_t *queue_id,
2500 uint8_t *traffic_class,
2503 struct rte_mbuf_sched sched = m->hash.sched;
2505 *queue_id = sched.queue_id;
2506 *traffic_class = sched.traffic_class;
2507 *color = sched.color;
2511 * Set the mbuf sched queue_id to the defined value.
2514 rte_mbuf_sched_queue_set(struct rte_mbuf *m, uint32_t queue_id)
2516 m->hash.sched.queue_id = queue_id;
2520 * Set the mbuf sched traffic_class id to the defined value.
2523 rte_mbuf_sched_traffic_class_set(struct rte_mbuf *m, uint8_t traffic_class)
2525 m->hash.sched.traffic_class = traffic_class;
2529 * Set the mbuf sched color id to the defined value.
2532 rte_mbuf_sched_color_set(struct rte_mbuf *m, uint8_t color)
2534 m->hash.sched.color = color;
2538 * Set the mbuf sched queue_id, traffic_class and color.
2543 * Queue id value to be set
2544 * @param traffic_class
2545 * Traffic class id value to be set
2547 * Color id to be set
2550 rte_mbuf_sched_set(struct rte_mbuf *m, uint32_t queue_id,
2551 uint8_t traffic_class,
2554 m->hash.sched = (struct rte_mbuf_sched){
2555 .queue_id = queue_id,
2556 .traffic_class = traffic_class,
2566 #endif /* _RTE_MBUF_H_ */