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)
254 #define PKT_TX_TUNNEL_GTP (0x7ULL << 45)
257 * Generic IP encapsulated tunnel type, used for TSO and checksum offload.
258 * It can be used for tunnels which are not standards or listed above.
259 * It is preferred to use specific tunnel flags like PKT_TX_TUNNEL_GRE
260 * or PKT_TX_TUNNEL_IPIP if possible.
261 * The ethdev must be configured with DEV_TX_OFFLOAD_IP_TNL_TSO.
262 * Outer and inner checksums are done according to the existing flags like
264 * Specific tunnel headers that contain payload length, sequence id
265 * or checksum are not expected to be updated.
267 #define PKT_TX_TUNNEL_IP (0xDULL << 45)
269 * Generic UDP encapsulated tunnel type, used for TSO and checksum offload.
270 * UDP tunnel type implies outer IP layer.
271 * It can be used for tunnels which are not standards or listed above.
272 * It is preferred to use specific tunnel flags like PKT_TX_TUNNEL_VXLAN
274 * The ethdev must be configured with DEV_TX_OFFLOAD_UDP_TNL_TSO.
275 * Outer and inner checksums are done according to the existing flags like
277 * Specific tunnel headers that contain payload length, sequence id
278 * or checksum are not expected to be updated.
280 #define PKT_TX_TUNNEL_UDP (0xEULL << 45)
281 /* add new TX TUNNEL type here */
282 #define PKT_TX_TUNNEL_MASK (0xFULL << 45)
285 * Double VLAN insertion (QinQ) request to driver, driver may offload the
286 * insertion based on device capability.
287 * mbuf 'vlan_tci' & 'vlan_tci_outer' must be valid when this flag is set.
289 #define PKT_TX_QINQ (1ULL << 49)
290 /* this old name is deprecated */
291 #define PKT_TX_QINQ_PKT PKT_TX_QINQ
294 * TCP segmentation offload. To enable this offload feature for a
295 * packet to be transmitted on hardware supporting TSO:
296 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
298 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
299 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag
300 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
302 #define PKT_TX_TCP_SEG (1ULL << 50)
304 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
307 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
308 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
309 * L4 checksum offload, the user needs to:
310 * - fill l2_len and l3_len in mbuf
311 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
312 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
314 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
315 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
316 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
317 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
318 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
321 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
322 * also be set by the application, although a PMD will only check
324 * - fill the mbuf offload information: l2_len, l3_len
326 #define PKT_TX_IP_CKSUM (1ULL << 54)
329 * Packet is IPv4. This flag must be set when using any offload feature
330 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
331 * packet. If the packet is a tunneled packet, this flag is related to
334 #define PKT_TX_IPV4 (1ULL << 55)
337 * Packet is IPv6. This flag must be set when using an offload feature
338 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
339 * packet. If the packet is a tunneled packet, this flag is related to
342 #define PKT_TX_IPV6 (1ULL << 56)
345 * VLAN tag insertion request to driver, driver may offload the insertion
346 * based on the device capability.
347 * mbuf 'vlan_tci' field must be valid when this flag is set.
349 #define PKT_TX_VLAN (1ULL << 57)
350 /* this old name is deprecated */
351 #define PKT_TX_VLAN_PKT PKT_TX_VLAN
354 * Offload the IP checksum of an external header in the hardware. The
355 * flag PKT_TX_OUTER_IPV4 should also be set by the application, although
356 * a PMD will only check PKT_TX_OUTER_IP_CKSUM.
357 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
359 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
362 * Packet outer header is IPv4. This flag must be set when using any
363 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
364 * outer header of the tunneled packet is an IPv4 packet.
366 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
369 * Packet outer header is IPv6. This flag must be set when using any
370 * outer offload feature (L4 checksum) to tell the NIC that the outer
371 * header of the tunneled packet is an IPv6 packet.
373 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
376 * Bitmask of all supported packet Tx offload features flags,
377 * which can be set for packet.
379 #define PKT_TX_OFFLOAD_MASK ( \
380 PKT_TX_OUTER_IPV6 | \
381 PKT_TX_OUTER_IPV4 | \
382 PKT_TX_OUTER_IP_CKSUM | \
388 PKT_TX_IEEE1588_TMST | \
391 PKT_TX_TUNNEL_MASK | \
393 PKT_TX_SEC_OFFLOAD | \
395 PKT_TX_OUTER_UDP_CKSUM | \
399 * Mbuf having an external buffer attached. shinfo in mbuf must be filled.
401 #define EXT_ATTACHED_MBUF (1ULL << 61)
403 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
405 /** Alignment constraint of mbuf private area. */
406 #define RTE_MBUF_PRIV_ALIGN 8
409 * Get the name of a RX offload flag
412 * The mask describing the flag.
414 * The name of this flag, or NULL if it's not a valid RX flag.
416 const char *rte_get_rx_ol_flag_name(uint64_t mask);
419 * Dump the list of RX offload flags in a buffer
422 * The mask describing the RX flags.
426 * The length of the buffer.
428 * 0 on success, (-1) on error.
430 int rte_get_rx_ol_flag_list(uint64_t mask, char *buf, size_t buflen);
433 * Get the name of a TX offload flag
436 * The mask describing the flag. Usually only one bit must be set.
437 * Several bits can be given if they belong to the same mask.
438 * Ex: PKT_TX_L4_MASK.
440 * The name of this flag, or NULL if it's not a valid TX flag.
442 const char *rte_get_tx_ol_flag_name(uint64_t mask);
445 * Dump the list of TX offload flags in a buffer
448 * The mask describing the TX flags.
452 * The length of the buffer.
454 * 0 on success, (-1) on error.
456 int rte_get_tx_ol_flag_list(uint64_t mask, char *buf, size_t buflen);
459 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
460 * splitting it into multiple segments.
461 * So, for mbufs that planned to be involved into RX/TX, the recommended
462 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
464 #define RTE_MBUF_DEFAULT_DATAROOM 2048
465 #define RTE_MBUF_DEFAULT_BUF_SIZE \
466 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
468 /* define a set of marker types that can be used to refer to set points in the
471 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
473 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
475 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
476 * with a single assignment */
478 struct rte_mbuf_sched {
479 uint32_t queue_id; /**< Queue ID. */
480 uint8_t traffic_class;
481 /**< Traffic class ID. Traffic class 0
482 * is the highest priority traffic class.
485 /**< Color. @see enum rte_color.*/
486 uint16_t reserved; /**< Reserved. */
487 }; /**< Hierarchical scheduler */
490 * enum for the tx_offload bit-fields lengths and offsets.
491 * defines the layout of rte_mbuf tx_offload field.
494 RTE_MBUF_L2_LEN_BITS = 7,
495 RTE_MBUF_L3_LEN_BITS = 9,
496 RTE_MBUF_L4_LEN_BITS = 8,
497 RTE_MBUF_TSO_SEGSZ_BITS = 16,
498 RTE_MBUF_OUTL3_LEN_BITS = 9,
499 RTE_MBUF_OUTL2_LEN_BITS = 7,
500 RTE_MBUF_TXOFLD_UNUSED_BITS = sizeof(uint64_t) * CHAR_BIT -
501 RTE_MBUF_L2_LEN_BITS -
502 RTE_MBUF_L3_LEN_BITS -
503 RTE_MBUF_L4_LEN_BITS -
504 RTE_MBUF_TSO_SEGSZ_BITS -
505 RTE_MBUF_OUTL3_LEN_BITS -
506 RTE_MBUF_OUTL2_LEN_BITS,
507 #if RTE_BYTE_ORDER == RTE_BIG_ENDIAN
508 RTE_MBUF_L2_LEN_OFS =
509 sizeof(uint64_t) * CHAR_BIT - RTE_MBUF_L2_LEN_BITS,
510 RTE_MBUF_L3_LEN_OFS = RTE_MBUF_L2_LEN_OFS - RTE_MBUF_L3_LEN_BITS,
511 RTE_MBUF_L4_LEN_OFS = RTE_MBUF_L3_LEN_OFS - RTE_MBUF_L4_LEN_BITS,
512 RTE_MBUF_TSO_SEGSZ_OFS = RTE_MBUF_L4_LEN_OFS - RTE_MBUF_TSO_SEGSZ_BITS,
513 RTE_MBUF_OUTL3_LEN_OFS =
514 RTE_MBUF_TSO_SEGSZ_OFS - RTE_MBUF_OUTL3_LEN_BITS,
515 RTE_MBUF_OUTL2_LEN_OFS =
516 RTE_MBUF_OUTL3_LEN_OFS - RTE_MBUF_OUTL2_LEN_BITS,
517 RTE_MBUF_TXOFLD_UNUSED_OFS =
518 RTE_MBUF_OUTL2_LEN_OFS - RTE_MBUF_TXOFLD_UNUSED_BITS,
520 RTE_MBUF_L2_LEN_OFS = 0,
521 RTE_MBUF_L3_LEN_OFS = RTE_MBUF_L2_LEN_OFS + RTE_MBUF_L2_LEN_BITS,
522 RTE_MBUF_L4_LEN_OFS = RTE_MBUF_L3_LEN_OFS + RTE_MBUF_L3_LEN_BITS,
523 RTE_MBUF_TSO_SEGSZ_OFS = RTE_MBUF_L4_LEN_OFS + RTE_MBUF_L4_LEN_BITS,
524 RTE_MBUF_OUTL3_LEN_OFS =
525 RTE_MBUF_TSO_SEGSZ_OFS + RTE_MBUF_TSO_SEGSZ_BITS,
526 RTE_MBUF_OUTL2_LEN_OFS =
527 RTE_MBUF_OUTL3_LEN_OFS + RTE_MBUF_OUTL3_LEN_BITS,
528 RTE_MBUF_TXOFLD_UNUSED_OFS =
529 RTE_MBUF_OUTL2_LEN_OFS + RTE_MBUF_OUTL2_LEN_BITS,
534 * The generic rte_mbuf, containing a packet mbuf.
539 void *buf_addr; /**< Virtual address of segment buffer. */
541 * Physical address of segment buffer.
542 * Force alignment to 8-bytes, so as to ensure we have the exact
543 * same mbuf cacheline0 layout for 32-bit and 64-bit. This makes
544 * working on vector drivers easier.
549 rte_iova_t buf_physaddr; /**< deprecated */
550 } __rte_aligned(sizeof(rte_iova_t));
552 /* next 8 bytes are initialised on RX descriptor rearm */
557 * Reference counter. Its size should at least equal to the size
558 * of port field (16 bits), to support zero-copy broadcast.
559 * It should only be accessed using the following functions:
560 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
561 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
562 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
567 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
568 uint16_t refcnt; /**< Non-atomically accessed refcnt */
570 uint16_t nb_segs; /**< Number of segments. */
572 /** Input port (16 bits to support more than 256 virtual ports).
573 * The event eth Tx adapter uses this field to specify the output port.
577 uint64_t ol_flags; /**< Offload features. */
579 /* remaining bytes are set on RX when pulling packet from descriptor */
580 MARKER rx_descriptor_fields1;
583 * The packet type, which is the combination of outer/inner L2, L3, L4
584 * and tunnel types. The packet_type is about data really present in the
585 * mbuf. Example: if vlan stripping is enabled, a received vlan packet
586 * would have RTE_PTYPE_L2_ETHER and not RTE_PTYPE_L2_VLAN because the
587 * vlan is stripped from the data.
591 uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
593 uint32_t l2_type:4; /**< (Outer) L2 type. */
594 uint32_t l3_type:4; /**< (Outer) L3 type. */
595 uint32_t l4_type:4; /**< (Outer) L4 type. */
596 uint32_t tun_type:4; /**< Tunnel type. */
599 uint8_t inner_esp_next_proto;
600 /**< ESP next protocol type, valid if
601 * RTE_PTYPE_TUNNEL_ESP tunnel type is set
606 uint8_t inner_l2_type:4;
607 /**< Inner L2 type. */
608 uint8_t inner_l3_type:4;
609 /**< Inner L3 type. */
612 uint32_t inner_l4_type:4; /**< Inner L4 type. */
616 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
617 uint16_t data_len; /**< Amount of data in segment buffer. */
618 /** VLAN TCI (CPU order), valid if PKT_RX_VLAN is set. */
624 uint32_t rss; /**< RSS hash result if RSS enabled */
632 /**< Second 4 flexible bytes */
635 /**< First 4 flexible bytes or FD ID, dependent
636 * on PKT_RX_FDIR_* flag in ol_flags.
638 } fdir; /**< Filter identifier if FDIR enabled */
639 struct rte_mbuf_sched sched;
640 /**< Hierarchical scheduler : 8 bytes */
645 /**< The event eth Tx adapter uses this field
646 * to store Tx queue id.
647 * @see rte_event_eth_tx_adapter_txq_set()
649 } txadapter; /**< Eventdev ethdev Tx adapter */
650 /**< User defined tags. See rte_distributor_process() */
652 } hash; /**< hash information */
655 * Application specific metadata value
656 * for egress flow rule match.
657 * Valid if PKT_TX_METADATA is set.
658 * Located here to allow conjunct use
659 * with hash.sched.hi.
661 uint32_t tx_metadata;
666 /** Outer VLAN TCI (CPU order), valid if PKT_RX_QINQ is set. */
667 uint16_t vlan_tci_outer;
669 uint16_t buf_len; /**< Length of segment buffer. */
671 /** Valid if PKT_RX_TIMESTAMP is set. The unit and time reference
672 * are not normalized but are always the same for a given port.
673 * Some devices allow to query rte_eth_read_clock that will return the
674 * current device timestamp.
678 /* second cache line - fields only used in slow path or on TX */
679 MARKER cacheline1 __rte_cache_min_aligned;
683 void *userdata; /**< Can be used for external metadata */
684 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
687 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
688 struct rte_mbuf *next; /**< Next segment of scattered packet. */
690 /* fields to support TX offloads */
693 uint64_t tx_offload; /**< combined for easy fetch */
696 uint64_t l2_len:RTE_MBUF_L2_LEN_BITS;
697 /**< L2 (MAC) Header Length for non-tunneling pkt.
698 * Outer_L4_len + ... + Inner_L2_len for tunneling pkt.
700 uint64_t l3_len:RTE_MBUF_L3_LEN_BITS;
701 /**< L3 (IP) Header Length. */
702 uint64_t l4_len:RTE_MBUF_L4_LEN_BITS;
703 /**< L4 (TCP/UDP) Header Length. */
704 uint64_t tso_segsz:RTE_MBUF_TSO_SEGSZ_BITS;
705 /**< TCP TSO segment size */
708 * Fields for Tx offloading of tunnels.
709 * These are undefined for packets which don't request
710 * any tunnel offloads (outer IP or UDP checksum,
713 * PMDs should not use these fields unconditionally
714 * when calculating offsets.
716 * Applications are expected to set appropriate tunnel
717 * offload flags when they fill in these fields.
719 uint64_t outer_l3_len:RTE_MBUF_OUTL3_LEN_BITS;
720 /**< Outer L3 (IP) Hdr Length. */
721 uint64_t outer_l2_len:RTE_MBUF_OUTL2_LEN_BITS;
722 /**< Outer L2 (MAC) Hdr Length. */
724 /* uint64_t unused:RTE_MBUF_TXOFLD_UNUSED_BITS; */
728 /** Size of the application private data. In case of an indirect
729 * mbuf, it stores the direct mbuf private data size. */
732 /** Timesync flags for use with IEEE1588. */
735 /** Sequence number. See also rte_reorder_insert(). */
738 /** Shared data for external buffer attached to mbuf. See
739 * rte_pktmbuf_attach_extbuf().
741 struct rte_mbuf_ext_shared_info *shinfo;
743 } __rte_cache_aligned;
746 * Function typedef of callback to free externally attached buffer.
748 typedef void (*rte_mbuf_extbuf_free_callback_t)(void *addr, void *opaque);
751 * Shared data at the end of an external buffer.
753 struct rte_mbuf_ext_shared_info {
754 rte_mbuf_extbuf_free_callback_t free_cb; /**< Free callback function */
755 void *fcb_opaque; /**< Free callback argument */
756 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
759 /**< Maximum number of nb_segs allowed. */
760 #define RTE_MBUF_MAX_NB_SEGS UINT16_MAX
763 * Prefetch the first part of the mbuf
765 * The first 64 bytes of the mbuf corresponds to fields that are used early
766 * in the receive path. If the cache line of the architecture is higher than
767 * 64B, the second part will also be prefetched.
770 * The pointer to the mbuf.
773 rte_mbuf_prefetch_part1(struct rte_mbuf *m)
775 rte_prefetch0(&m->cacheline0);
779 * Prefetch the second part of the mbuf
781 * The next 64 bytes of the mbuf corresponds to fields that are used in the
782 * transmit path. If the cache line of the architecture is higher than 64B,
783 * this function does nothing as it is expected that the full mbuf is
787 * The pointer to the mbuf.
790 rte_mbuf_prefetch_part2(struct rte_mbuf *m)
792 #if RTE_CACHE_LINE_SIZE == 64
793 rte_prefetch0(&m->cacheline1);
800 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
803 * Return the IO address of the beginning of the mbuf data
806 * The pointer to the mbuf.
808 * The IO address of the beginning of the mbuf data
810 static inline rte_iova_t
811 rte_mbuf_data_iova(const struct rte_mbuf *mb)
813 return mb->buf_iova + mb->data_off;
817 static inline phys_addr_t
818 rte_mbuf_data_dma_addr(const struct rte_mbuf *mb)
820 return rte_mbuf_data_iova(mb);
824 * Return the default IO address of the beginning of the mbuf data
826 * This function is used by drivers in their receive function, as it
827 * returns the location where data should be written by the NIC, taking
828 * the default headroom in account.
831 * The pointer to the mbuf.
833 * The IO address of the beginning of the mbuf data
835 static inline rte_iova_t
836 rte_mbuf_data_iova_default(const struct rte_mbuf *mb)
838 return mb->buf_iova + RTE_PKTMBUF_HEADROOM;
842 static inline phys_addr_t
843 rte_mbuf_data_dma_addr_default(const struct rte_mbuf *mb)
845 return rte_mbuf_data_iova_default(mb);
849 * Return the mbuf owning the data buffer address of an indirect mbuf.
852 * The pointer to the indirect mbuf.
854 * The address of the direct mbuf corresponding to buffer_addr.
856 static inline struct rte_mbuf *
857 rte_mbuf_from_indirect(struct rte_mbuf *mi)
859 return (struct rte_mbuf *)RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
863 * Return address of buffer embedded in the given mbuf.
865 * The return value shall be same as mb->buf_addr if the mbuf is already
866 * initialized and direct. However, this API is useful if mempool of the
867 * mbuf is already known because it doesn't need to access mbuf contents in
868 * order to get the mempool pointer.
871 * @b EXPERIMENTAL: This API may change without prior notice.
872 * This will be used by rte_mbuf_to_baddr() which has redundant code once
873 * experimental tag is removed.
876 * The pointer to the mbuf.
878 * The pointer to the mempool of the mbuf.
880 * The pointer of the mbuf buffer.
884 rte_mbuf_buf_addr(struct rte_mbuf *mb, struct rte_mempool *mp)
886 return (char *)mb + sizeof(*mb) + rte_pktmbuf_priv_size(mp);
890 * Return the default address of the beginning of the mbuf data.
893 * @b EXPERIMENTAL: This API may change without prior notice.
896 * The pointer to the mbuf.
898 * The pointer of the beginning of the mbuf data.
902 rte_mbuf_data_addr_default(__rte_unused struct rte_mbuf *mb)
904 /* gcc complains about calling this experimental function even
905 * when not using it. Hide it with ALLOW_EXPERIMENTAL_API.
907 #ifdef ALLOW_EXPERIMENTAL_API
908 return rte_mbuf_buf_addr(mb, mb->pool) + RTE_PKTMBUF_HEADROOM;
915 * Return address of buffer embedded in the given mbuf.
917 * @note: Accessing mempool pointer of a mbuf is expensive because the
918 * pointer is stored in the 2nd cache line of mbuf. If mempool is known, it
919 * is better not to reference the mempool pointer in mbuf but calling
920 * rte_mbuf_buf_addr() would be more efficient.
923 * The pointer to the mbuf.
925 * The address of the data buffer owned by the mbuf.
928 rte_mbuf_to_baddr(struct rte_mbuf *md)
930 #ifdef ALLOW_EXPERIMENTAL_API
931 return rte_mbuf_buf_addr(md, md->pool);
934 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
940 * Return the starting address of the private data area embedded in
943 * Note that no check is made to ensure that a private data area
944 * actually exists in the supplied mbuf.
947 * The pointer to the mbuf.
949 * The starting address of the private data area of the given mbuf.
953 rte_mbuf_to_priv(struct rte_mbuf *m)
955 return RTE_PTR_ADD(m, sizeof(struct rte_mbuf));
959 * Returns TRUE if given mbuf is cloned by mbuf indirection, or FALSE
962 * If a mbuf has its data in another mbuf and references it by mbuf
963 * indirection, this mbuf can be defined as a cloned mbuf.
965 #define RTE_MBUF_CLONED(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
968 * Returns TRUE if given mbuf has an external buffer, or FALSE otherwise.
970 * External buffer is a user-provided anonymous buffer.
972 #define RTE_MBUF_HAS_EXTBUF(mb) ((mb)->ol_flags & EXT_ATTACHED_MBUF)
975 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
977 * If a mbuf embeds its own data after the rte_mbuf structure, this mbuf
978 * can be defined as a direct mbuf.
980 #define RTE_MBUF_DIRECT(mb) \
981 (!((mb)->ol_flags & (IND_ATTACHED_MBUF | EXT_ATTACHED_MBUF)))
984 * Private data in case of pktmbuf pool.
986 * A structure that contains some pktmbuf_pool-specific data that are
987 * appended after the mempool structure (in private data).
989 struct rte_pktmbuf_pool_private {
990 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
991 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
994 #ifdef RTE_LIBRTE_MBUF_DEBUG
996 /** check mbuf type in debug mode */
997 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
999 #else /* RTE_LIBRTE_MBUF_DEBUG */
1001 /** check mbuf type in debug mode */
1002 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
1004 #endif /* RTE_LIBRTE_MBUF_DEBUG */
1006 #ifdef RTE_MBUF_REFCNT_ATOMIC
1009 * Reads the value of an mbuf's refcnt.
1013 * Reference count number.
1015 static inline uint16_t
1016 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
1018 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
1022 * Sets an mbuf's refcnt to a defined value.
1029 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
1031 rte_atomic16_set(&m->refcnt_atomic, (int16_t)new_value);
1035 static inline uint16_t
1036 __rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1038 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
1042 * Adds given value to an mbuf's refcnt and returns its new value.
1046 * Value to add/subtract
1050 static inline uint16_t
1051 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1054 * The atomic_add is an expensive operation, so we don't want to
1055 * call it in the case where we know we are the unique holder of
1056 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
1057 * operation has to be used because concurrent accesses on the
1058 * reference counter can occur.
1060 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
1062 rte_mbuf_refcnt_set(m, (uint16_t)value);
1063 return (uint16_t)value;
1066 return __rte_mbuf_refcnt_update(m, value);
1069 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
1072 static inline uint16_t
1073 __rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1075 m->refcnt = (uint16_t)(m->refcnt + value);
1080 * Adds given value to an mbuf's refcnt and returns its new value.
1082 static inline uint16_t
1083 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1085 return __rte_mbuf_refcnt_update(m, value);
1089 * Reads the value of an mbuf's refcnt.
1091 static inline uint16_t
1092 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
1098 * Sets an mbuf's refcnt to the defined value.
1101 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
1103 m->refcnt = new_value;
1106 #endif /* RTE_MBUF_REFCNT_ATOMIC */
1109 * Reads the refcnt of an external buffer.
1112 * Shared data of the external buffer.
1114 * Reference count number.
1116 static inline uint16_t
1117 rte_mbuf_ext_refcnt_read(const struct rte_mbuf_ext_shared_info *shinfo)
1119 return (uint16_t)(rte_atomic16_read(&shinfo->refcnt_atomic));
1123 * Set refcnt of an external buffer.
1126 * Shared data of the external buffer.
1131 rte_mbuf_ext_refcnt_set(struct rte_mbuf_ext_shared_info *shinfo,
1134 rte_atomic16_set(&shinfo->refcnt_atomic, (int16_t)new_value);
1138 * Add given value to refcnt of an external buffer and return its new
1142 * Shared data of the external buffer.
1144 * Value to add/subtract
1148 static inline uint16_t
1149 rte_mbuf_ext_refcnt_update(struct rte_mbuf_ext_shared_info *shinfo,
1152 if (likely(rte_mbuf_ext_refcnt_read(shinfo) == 1)) {
1154 rte_mbuf_ext_refcnt_set(shinfo, (uint16_t)value);
1155 return (uint16_t)value;
1158 return (uint16_t)rte_atomic16_add_return(&shinfo->refcnt_atomic, value);
1161 /** Mbuf prefetch */
1162 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
1169 * Sanity checks on an mbuf.
1171 * Check the consistency of the given mbuf. The function will cause a
1172 * panic if corruption is detected.
1175 * The mbuf to be checked.
1177 * True if the mbuf is a packet header, false if it is a sub-segment
1178 * of a packet (in this case, some fields like nb_segs are not checked)
1181 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
1184 * Sanity checks on a mbuf.
1186 * Almost like rte_mbuf_sanity_check(), but this function gives the reason
1187 * if corruption is detected rather than panic.
1190 * The mbuf to be checked.
1192 * True if the mbuf is a packet header, false if it is a sub-segment
1193 * of a packet (in this case, some fields like nb_segs are not checked)
1195 * A reference to a string pointer where to store the reason why a mbuf is
1196 * considered invalid.
1198 * - 0 if no issue has been found, reason is left untouched.
1199 * - -1 if a problem is detected, reason then points to a string describing
1200 * the reason why the mbuf is deemed invalid.
1203 int rte_mbuf_check(const struct rte_mbuf *m, int is_header,
1204 const char **reason);
1206 #define MBUF_RAW_ALLOC_CHECK(m) do { \
1207 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1); \
1208 RTE_ASSERT((m)->next == NULL); \
1209 RTE_ASSERT((m)->nb_segs == 1); \
1210 __rte_mbuf_sanity_check(m, 0); \
1214 * Allocate an uninitialized mbuf from mempool *mp*.
1216 * This function can be used by PMDs (especially in RX functions) to
1217 * allocate an uninitialized mbuf. The driver is responsible of
1218 * initializing all the required fields. See rte_pktmbuf_reset().
1219 * For standard needs, prefer rte_pktmbuf_alloc().
1221 * The caller can expect that the following fields of the mbuf structure
1222 * are initialized: buf_addr, buf_iova, buf_len, refcnt=1, nb_segs=1,
1223 * next=NULL, pool, priv_size. The other fields must be initialized
1227 * The mempool from which mbuf is allocated.
1229 * - The pointer to the new mbuf on success.
1230 * - NULL if allocation failed.
1232 static inline struct rte_mbuf *rte_mbuf_raw_alloc(struct rte_mempool *mp)
1236 if (rte_mempool_get(mp, (void **)&m) < 0)
1238 MBUF_RAW_ALLOC_CHECK(m);
1243 * Put mbuf back into its original mempool.
1245 * The caller must ensure that the mbuf is direct and properly
1246 * reinitialized (refcnt=1, next=NULL, nb_segs=1), as done by
1247 * rte_pktmbuf_prefree_seg().
1249 * This function should be used with care, when optimization is
1250 * required. For standard needs, prefer rte_pktmbuf_free() or
1251 * rte_pktmbuf_free_seg().
1254 * The mbuf to be freed.
1256 static __rte_always_inline void
1257 rte_mbuf_raw_free(struct rte_mbuf *m)
1259 RTE_ASSERT(RTE_MBUF_DIRECT(m));
1260 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1);
1261 RTE_ASSERT(m->next == NULL);
1262 RTE_ASSERT(m->nb_segs == 1);
1263 __rte_mbuf_sanity_check(m, 0);
1264 rte_mempool_put(m->pool, m);
1268 * The packet mbuf constructor.
1270 * This function initializes some fields in the mbuf structure that are
1271 * not modified by the user once created (origin pool, buffer start
1272 * address, and so on). This function is given as a callback function to
1273 * rte_mempool_obj_iter() or rte_mempool_create() at pool creation time.
1276 * The mempool from which mbufs originate.
1278 * A pointer that can be used by the user to retrieve useful information
1279 * for mbuf initialization. This pointer is the opaque argument passed to
1280 * rte_mempool_obj_iter() or rte_mempool_create().
1282 * The mbuf to initialize.
1284 * The index of the mbuf in the pool table.
1286 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1287 void *m, unsigned i);
1291 * A packet mbuf pool constructor.
1293 * This function initializes the mempool private data in the case of a
1294 * pktmbuf pool. This private data is needed by the driver. The
1295 * function must be called on the mempool before it is used, or it
1296 * can be given as a callback function to rte_mempool_create() at
1297 * pool creation. It can be extended by the user, for example, to
1298 * provide another packet size.
1301 * The mempool from which mbufs originate.
1303 * A pointer that can be used by the user to retrieve useful information
1304 * for mbuf initialization. This pointer is the opaque argument passed to
1305 * rte_mempool_create().
1307 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
1310 * Create a mbuf pool.
1312 * This function creates and initializes a packet mbuf pool. It is
1313 * a wrapper to rte_mempool functions.
1316 * The name of the mbuf pool.
1318 * The number of elements in the mbuf pool. The optimum size (in terms
1319 * of memory usage) for a mempool is when n is a power of two minus one:
1322 * Size of the per-core object cache. See rte_mempool_create() for
1325 * Size of application private are between the rte_mbuf structure
1326 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
1327 * @param data_room_size
1328 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
1330 * The socket identifier where the memory should be allocated. The
1331 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
1334 * The pointer to the new allocated mempool, on success. NULL on error
1335 * with rte_errno set appropriately. Possible rte_errno values include:
1336 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
1337 * - E_RTE_SECONDARY - function was called from a secondary process instance
1338 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
1339 * - ENOSPC - the maximum number of memzones has already been allocated
1340 * - EEXIST - a memzone with the same name already exists
1341 * - ENOMEM - no appropriate memory area found in which to create memzone
1343 struct rte_mempool *
1344 rte_pktmbuf_pool_create(const char *name, unsigned n,
1345 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
1349 * Create a mbuf pool with a given mempool ops name
1351 * This function creates and initializes a packet mbuf pool. It is
1352 * a wrapper to rte_mempool functions.
1355 * The name of the mbuf pool.
1357 * The number of elements in the mbuf pool. The optimum size (in terms
1358 * of memory usage) for a mempool is when n is a power of two minus one:
1361 * Size of the per-core object cache. See rte_mempool_create() for
1364 * Size of application private are between the rte_mbuf structure
1365 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
1366 * @param data_room_size
1367 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
1369 * The socket identifier where the memory should be allocated. The
1370 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
1373 * The mempool ops name to be used for this mempool instead of
1374 * default mempool. The value can be *NULL* to use default mempool.
1376 * The pointer to the new allocated mempool, on success. NULL on error
1377 * with rte_errno set appropriately. Possible rte_errno values include:
1378 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
1379 * - E_RTE_SECONDARY - function was called from a secondary process instance
1380 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
1381 * - ENOSPC - the maximum number of memzones has already been allocated
1382 * - EEXIST - a memzone with the same name already exists
1383 * - ENOMEM - no appropriate memory area found in which to create memzone
1385 struct rte_mempool *
1386 rte_pktmbuf_pool_create_by_ops(const char *name, unsigned int n,
1387 unsigned int cache_size, uint16_t priv_size, uint16_t data_room_size,
1388 int socket_id, const char *ops_name);
1391 * Get the data room size of mbufs stored in a pktmbuf_pool
1393 * The data room size is the amount of data that can be stored in a
1394 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
1397 * The packet mbuf pool.
1399 * The data room size of mbufs stored in this mempool.
1401 static inline uint16_t
1402 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
1404 struct rte_pktmbuf_pool_private *mbp_priv;
1406 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1407 return mbp_priv->mbuf_data_room_size;
1411 * Get the application private size of mbufs stored in a pktmbuf_pool
1413 * The private size of mbuf is a zone located between the rte_mbuf
1414 * structure and the data buffer where an application can store data
1415 * associated to a packet.
1418 * The packet mbuf pool.
1420 * The private size of mbufs stored in this mempool.
1422 static inline uint16_t
1423 rte_pktmbuf_priv_size(struct rte_mempool *mp)
1425 struct rte_pktmbuf_pool_private *mbp_priv;
1427 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1428 return mbp_priv->mbuf_priv_size;
1432 * Reset the data_off field of a packet mbuf to its default value.
1434 * The given mbuf must have only one segment, which should be empty.
1437 * The packet mbuf's data_off field has to be reset.
1439 static inline void rte_pktmbuf_reset_headroom(struct rte_mbuf *m)
1441 m->data_off = (uint16_t)RTE_MIN((uint16_t)RTE_PKTMBUF_HEADROOM,
1442 (uint16_t)m->buf_len);
1446 * Reset the fields of a packet mbuf to their default values.
1448 * The given mbuf must have only one segment.
1451 * The packet mbuf to be reset.
1453 #define MBUF_INVALID_PORT UINT16_MAX
1455 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
1461 m->vlan_tci_outer = 0;
1463 m->port = MBUF_INVALID_PORT;
1467 rte_pktmbuf_reset_headroom(m);
1470 __rte_mbuf_sanity_check(m, 1);
1474 * Allocate a new mbuf from a mempool.
1476 * This new mbuf contains one segment, which has a length of 0. The pointer
1477 * to data is initialized to have some bytes of headroom in the buffer
1478 * (if buffer size allows).
1481 * The mempool from which the mbuf is allocated.
1483 * - The pointer to the new mbuf on success.
1484 * - NULL if allocation failed.
1486 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
1489 if ((m = rte_mbuf_raw_alloc(mp)) != NULL)
1490 rte_pktmbuf_reset(m);
1495 * Allocate a bulk of mbufs, initialize refcnt and reset the fields to default
1499 * The mempool from which mbufs are allocated.
1501 * Array of pointers to mbufs
1506 * - -ENOENT: Not enough entries in the mempool; no mbufs are retrieved.
1508 static inline int rte_pktmbuf_alloc_bulk(struct rte_mempool *pool,
1509 struct rte_mbuf **mbufs, unsigned count)
1514 rc = rte_mempool_get_bulk(pool, (void **)mbufs, count);
1518 /* To understand duff's device on loop unwinding optimization, see
1519 * https://en.wikipedia.org/wiki/Duff's_device.
1520 * Here while() loop is used rather than do() while{} to avoid extra
1521 * check if count is zero.
1523 switch (count % 4) {
1525 while (idx != count) {
1526 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1527 rte_pktmbuf_reset(mbufs[idx]);
1531 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1532 rte_pktmbuf_reset(mbufs[idx]);
1536 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1537 rte_pktmbuf_reset(mbufs[idx]);
1541 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1542 rte_pktmbuf_reset(mbufs[idx]);
1551 * Initialize shared data at the end of an external buffer before attaching
1552 * to a mbuf by ``rte_pktmbuf_attach_extbuf()``. This is not a mandatory
1553 * initialization but a helper function to simply spare a few bytes at the
1554 * end of the buffer for shared data. If shared data is allocated
1555 * separately, this should not be called but application has to properly
1556 * initialize the shared data according to its need.
1558 * Free callback and its argument is saved and the refcnt is set to 1.
1561 * The value of buf_len will be reduced to RTE_PTR_DIFF(shinfo, buf_addr)
1562 * after this initialization. This shall be used for
1563 * ``rte_pktmbuf_attach_extbuf()``
1566 * The pointer to the external buffer.
1567 * @param [in,out] buf_len
1568 * The pointer to length of the external buffer. Input value must be
1569 * larger than the size of ``struct rte_mbuf_ext_shared_info`` and
1570 * padding for alignment. If not enough, this function will return NULL.
1571 * Adjusted buffer length will be returned through this pointer.
1573 * Free callback function to call when the external buffer needs to be
1576 * Argument for the free callback function.
1579 * A pointer to the initialized shared data on success, return NULL
1582 static inline struct rte_mbuf_ext_shared_info *
1583 rte_pktmbuf_ext_shinfo_init_helper(void *buf_addr, uint16_t *buf_len,
1584 rte_mbuf_extbuf_free_callback_t free_cb, void *fcb_opaque)
1586 struct rte_mbuf_ext_shared_info *shinfo;
1587 void *buf_end = RTE_PTR_ADD(buf_addr, *buf_len);
1590 addr = RTE_PTR_ALIGN_FLOOR(RTE_PTR_SUB(buf_end, sizeof(*shinfo)),
1592 if (addr <= buf_addr)
1595 shinfo = (struct rte_mbuf_ext_shared_info *)addr;
1596 shinfo->free_cb = free_cb;
1597 shinfo->fcb_opaque = fcb_opaque;
1598 rte_mbuf_ext_refcnt_set(shinfo, 1);
1600 *buf_len = (uint16_t)RTE_PTR_DIFF(shinfo, buf_addr);
1605 * Attach an external buffer to a mbuf.
1607 * User-managed anonymous buffer can be attached to an mbuf. When attaching
1608 * it, corresponding free callback function and its argument should be
1609 * provided via shinfo. This callback function will be called once all the
1610 * mbufs are detached from the buffer (refcnt becomes zero).
1612 * The headroom for the attaching mbuf will be set to zero and this can be
1613 * properly adjusted after attachment. For example, ``rte_pktmbuf_adj()``
1614 * or ``rte_pktmbuf_reset_headroom()`` might be used.
1616 * More mbufs can be attached to the same external buffer by
1617 * ``rte_pktmbuf_attach()`` once the external buffer has been attached by
1620 * Detachment can be done by either ``rte_pktmbuf_detach_extbuf()`` or
1621 * ``rte_pktmbuf_detach()``.
1623 * Memory for shared data must be provided and user must initialize all of
1624 * the content properly, especially free callback and refcnt. The pointer
1625 * of shared data will be stored in m->shinfo.
1626 * ``rte_pktmbuf_ext_shinfo_init_helper`` can help to simply spare a few
1627 * bytes at the end of buffer for the shared data, store free callback and
1628 * its argument and set the refcnt to 1. The following is an example:
1630 * struct rte_mbuf_ext_shared_info *shinfo =
1631 * rte_pktmbuf_ext_shinfo_init_helper(buf_addr, &buf_len,
1632 * free_cb, fcb_arg);
1633 * rte_pktmbuf_attach_extbuf(m, buf_addr, buf_iova, buf_len, shinfo);
1634 * rte_pktmbuf_reset_headroom(m);
1635 * rte_pktmbuf_adj(m, data_len);
1637 * Attaching an external buffer is quite similar to mbuf indirection in
1638 * replacing buffer addresses and length of a mbuf, but a few differences:
1639 * - When an indirect mbuf is attached, refcnt of the direct mbuf would be
1640 * 2 as long as the direct mbuf itself isn't freed after the attachment.
1641 * In such cases, the buffer area of a direct mbuf must be read-only. But
1642 * external buffer has its own refcnt and it starts from 1. Unless
1643 * multiple mbufs are attached to a mbuf having an external buffer, the
1644 * external buffer is writable.
1645 * - There's no need to allocate buffer from a mempool. Any buffer can be
1646 * attached with appropriate free callback and its IO address.
1647 * - Smaller metadata is required to maintain shared data such as refcnt.
1650 * The pointer to the mbuf.
1652 * The pointer to the external buffer.
1654 * IO address of the external buffer.
1656 * The size of the external buffer.
1658 * User-provided memory for shared data of the external buffer.
1661 rte_pktmbuf_attach_extbuf(struct rte_mbuf *m, void *buf_addr,
1662 rte_iova_t buf_iova, uint16_t buf_len,
1663 struct rte_mbuf_ext_shared_info *shinfo)
1665 /* mbuf should not be read-only */
1666 RTE_ASSERT(RTE_MBUF_DIRECT(m) && rte_mbuf_refcnt_read(m) == 1);
1667 RTE_ASSERT(shinfo->free_cb != NULL);
1669 m->buf_addr = buf_addr;
1670 m->buf_iova = buf_iova;
1671 m->buf_len = buf_len;
1676 m->ol_flags |= EXT_ATTACHED_MBUF;
1681 * Detach the external buffer attached to a mbuf, same as
1682 * ``rte_pktmbuf_detach()``
1685 * The mbuf having external buffer.
1687 #define rte_pktmbuf_detach_extbuf(m) rte_pktmbuf_detach(m)
1691 __rte_pktmbuf_copy_hdr(struct rte_mbuf *mdst, const struct rte_mbuf *msrc)
1693 mdst->port = msrc->port;
1694 mdst->vlan_tci = msrc->vlan_tci;
1695 mdst->vlan_tci_outer = msrc->vlan_tci_outer;
1696 mdst->tx_offload = msrc->tx_offload;
1697 mdst->hash = msrc->hash;
1698 mdst->packet_type = msrc->packet_type;
1699 mdst->timestamp = msrc->timestamp;
1703 * Attach packet mbuf to another packet mbuf.
1705 * If the mbuf we are attaching to isn't a direct buffer and is attached to
1706 * an external buffer, the mbuf being attached will be attached to the
1707 * external buffer instead of mbuf indirection.
1709 * Otherwise, the mbuf will be indirectly attached. After attachment we
1710 * refer the mbuf we attached as 'indirect', while mbuf we attached to as
1711 * 'direct'. The direct mbuf's reference counter is incremented.
1713 * Right now, not supported:
1714 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
1715 * - mbuf we trying to attach (mi) is used by someone else
1716 * e.g. it's reference counter is greater then 1.
1719 * The indirect packet mbuf.
1721 * The packet mbuf we're attaching to.
1723 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
1725 RTE_ASSERT(RTE_MBUF_DIRECT(mi) &&
1726 rte_mbuf_refcnt_read(mi) == 1);
1728 if (RTE_MBUF_HAS_EXTBUF(m)) {
1729 rte_mbuf_ext_refcnt_update(m->shinfo, 1);
1730 mi->ol_flags = m->ol_flags;
1731 mi->shinfo = m->shinfo;
1733 /* if m is not direct, get the mbuf that embeds the data */
1734 rte_mbuf_refcnt_update(rte_mbuf_from_indirect(m), 1);
1735 mi->priv_size = m->priv_size;
1736 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
1739 __rte_pktmbuf_copy_hdr(mi, m);
1741 mi->data_off = m->data_off;
1742 mi->data_len = m->data_len;
1743 mi->buf_iova = m->buf_iova;
1744 mi->buf_addr = m->buf_addr;
1745 mi->buf_len = m->buf_len;
1748 mi->pkt_len = mi->data_len;
1751 __rte_mbuf_sanity_check(mi, 1);
1752 __rte_mbuf_sanity_check(m, 0);
1756 * @internal used by rte_pktmbuf_detach().
1758 * Decrement the reference counter of the external buffer. When the
1759 * reference counter becomes 0, the buffer is freed by pre-registered
1763 __rte_pktmbuf_free_extbuf(struct rte_mbuf *m)
1765 RTE_ASSERT(RTE_MBUF_HAS_EXTBUF(m));
1766 RTE_ASSERT(m->shinfo != NULL);
1768 if (rte_mbuf_ext_refcnt_update(m->shinfo, -1) == 0)
1769 m->shinfo->free_cb(m->buf_addr, m->shinfo->fcb_opaque);
1773 * @internal used by rte_pktmbuf_detach().
1775 * Decrement the direct mbuf's reference counter. When the reference
1776 * counter becomes 0, the direct mbuf is freed.
1779 __rte_pktmbuf_free_direct(struct rte_mbuf *m)
1781 struct rte_mbuf *md;
1783 RTE_ASSERT(RTE_MBUF_CLONED(m));
1785 md = rte_mbuf_from_indirect(m);
1787 if (rte_mbuf_refcnt_update(md, -1) == 0) {
1790 rte_mbuf_refcnt_set(md, 1);
1791 rte_mbuf_raw_free(md);
1796 * Detach a packet mbuf from external buffer or direct buffer.
1798 * - decrement refcnt and free the external/direct buffer if refcnt
1800 * - restore original mbuf address and length values.
1801 * - reset pktmbuf data and data_len to their default values.
1803 * All other fields of the given packet mbuf will be left intact.
1806 * The indirect attached packet mbuf.
1808 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
1810 struct rte_mempool *mp = m->pool;
1811 uint32_t mbuf_size, buf_len;
1814 if (RTE_MBUF_HAS_EXTBUF(m))
1815 __rte_pktmbuf_free_extbuf(m);
1817 __rte_pktmbuf_free_direct(m);
1819 priv_size = rte_pktmbuf_priv_size(mp);
1820 mbuf_size = (uint32_t)(sizeof(struct rte_mbuf) + priv_size);
1821 buf_len = rte_pktmbuf_data_room_size(mp);
1823 m->priv_size = priv_size;
1824 m->buf_addr = (char *)m + mbuf_size;
1825 m->buf_iova = rte_mempool_virt2iova(m) + mbuf_size;
1826 m->buf_len = (uint16_t)buf_len;
1827 rte_pktmbuf_reset_headroom(m);
1833 * Decrease reference counter and unlink a mbuf segment
1835 * This function does the same than a free, except that it does not
1836 * return the segment to its pool.
1837 * It decreases the reference counter, and if it reaches 0, it is
1838 * detached from its parent for an indirect mbuf.
1841 * The mbuf to be unlinked
1843 * - (m) if it is the last reference. It can be recycled or freed.
1844 * - (NULL) if the mbuf still has remaining references on it.
1846 static __rte_always_inline struct rte_mbuf *
1847 rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1849 __rte_mbuf_sanity_check(m, 0);
1851 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
1853 if (!RTE_MBUF_DIRECT(m))
1854 rte_pktmbuf_detach(m);
1856 if (m->next != NULL) {
1863 } else if (__rte_mbuf_refcnt_update(m, -1) == 0) {
1865 if (!RTE_MBUF_DIRECT(m))
1866 rte_pktmbuf_detach(m);
1868 if (m->next != NULL) {
1872 rte_mbuf_refcnt_set(m, 1);
1880 * Free a segment of a packet mbuf into its original mempool.
1882 * Free an mbuf, without parsing other segments in case of chained
1886 * The packet mbuf segment to be freed.
1888 static __rte_always_inline void
1889 rte_pktmbuf_free_seg(struct rte_mbuf *m)
1891 m = rte_pktmbuf_prefree_seg(m);
1892 if (likely(m != NULL))
1893 rte_mbuf_raw_free(m);
1897 * Free a packet mbuf back into its original mempool.
1899 * Free an mbuf, and all its segments in case of chained buffers. Each
1900 * segment is added back into its original mempool.
1903 * The packet mbuf to be freed. If NULL, the function does nothing.
1905 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
1907 struct rte_mbuf *m_next;
1910 __rte_mbuf_sanity_check(m, 1);
1914 rte_pktmbuf_free_seg(m);
1920 * Free a bulk of packet mbufs back into their original mempools.
1922 * Free a bulk of mbufs, and all their segments in case of chained buffers.
1923 * Each segment is added back into its original mempool.
1926 * Array of pointers to packet mbufs.
1927 * The array may contain NULL pointers.
1932 void rte_pktmbuf_free_bulk(struct rte_mbuf **mbufs, unsigned int count);
1935 * Create a "clone" of the given packet mbuf.
1937 * Walks through all segments of the given packet mbuf, and for each of them:
1938 * - Creates a new packet mbuf from the given pool.
1939 * - Attaches newly created mbuf to the segment.
1940 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
1941 * from the original packet mbuf.
1944 * The packet mbuf to be cloned.
1946 * The mempool from which the "clone" mbufs are allocated.
1948 * - The pointer to the new "clone" mbuf on success.
1949 * - NULL if allocation fails.
1952 rte_pktmbuf_clone(struct rte_mbuf *md, struct rte_mempool *mp);
1955 * Create a full copy of a given packet mbuf.
1957 * Copies all the data from a given packet mbuf to a newly allocated
1958 * set of mbufs. The private data are is not copied.
1961 * The packet mbuf to be copiedd.
1963 * The mempool from which the "clone" mbufs are allocated.
1965 * The number of bytes to skip before copying.
1966 * If the mbuf does not have that many bytes, it is an error
1967 * and NULL is returned.
1969 * The upper limit on bytes to copy. Passing UINT32_MAX
1970 * means all data (after offset).
1972 * - The pointer to the new "clone" mbuf on success.
1973 * - NULL if allocation fails.
1977 rte_pktmbuf_copy(const struct rte_mbuf *m, struct rte_mempool *mp,
1978 uint32_t offset, uint32_t length);
1981 * Adds given value to the refcnt of all packet mbuf segments.
1983 * Walks through all segments of given packet mbuf and for each of them
1984 * invokes rte_mbuf_refcnt_update().
1987 * The packet mbuf whose refcnt to be updated.
1989 * The value to add to the mbuf's segments refcnt.
1991 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1993 __rte_mbuf_sanity_check(m, 1);
1996 rte_mbuf_refcnt_update(m, v);
1997 } while ((m = m->next) != NULL);
2001 * Get the headroom in a packet mbuf.
2006 * The length of the headroom.
2008 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
2010 __rte_mbuf_sanity_check(m, 0);
2015 * Get the tailroom of a packet mbuf.
2020 * The length of the tailroom.
2022 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
2024 __rte_mbuf_sanity_check(m, 0);
2025 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
2030 * Get the last segment of the packet.
2035 * The last segment of the given mbuf.
2037 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
2039 __rte_mbuf_sanity_check(m, 1);
2040 while (m->next != NULL)
2046 * A macro that points to an offset into the data in the mbuf.
2048 * The returned pointer is cast to type t. Before using this
2049 * function, the user must ensure that the first segment is large
2050 * enough to accommodate its data.
2055 * The offset into the mbuf data.
2057 * The type to cast the result into.
2059 #define rte_pktmbuf_mtod_offset(m, t, o) \
2060 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
2063 * A macro that points to the start of the data in the mbuf.
2065 * The returned pointer is cast to type t. Before using this
2066 * function, the user must ensure that the first segment is large
2067 * enough to accommodate its data.
2072 * The type to cast the result into.
2074 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
2077 * A macro that returns the IO address that points to an offset of the
2078 * start of the data in the mbuf
2083 * The offset into the data to calculate address from.
2085 #define rte_pktmbuf_iova_offset(m, o) \
2086 (rte_iova_t)((m)->buf_iova + (m)->data_off + (o))
2089 #define rte_pktmbuf_mtophys_offset(m, o) \
2090 rte_pktmbuf_iova_offset(m, o)
2093 * A macro that returns the IO address that points to the start of the
2099 #define rte_pktmbuf_iova(m) rte_pktmbuf_iova_offset(m, 0)
2102 #define rte_pktmbuf_mtophys(m) rte_pktmbuf_iova(m)
2105 * A macro that returns the length of the packet.
2107 * The value can be read or assigned.
2112 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
2115 * A macro that returns the length of the segment.
2117 * The value can be read or assigned.
2122 #define rte_pktmbuf_data_len(m) ((m)->data_len)
2125 * Prepend len bytes to an mbuf data area.
2127 * Returns a pointer to the new
2128 * data start address. If there is not enough headroom in the first
2129 * segment, the function will return NULL, without modifying the mbuf.
2134 * The amount of data to prepend (in bytes).
2136 * A pointer to the start of the newly prepended data, or
2137 * NULL if there is not enough headroom space in the first segment
2139 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
2142 __rte_mbuf_sanity_check(m, 1);
2144 if (unlikely(len > rte_pktmbuf_headroom(m)))
2147 /* NB: elaborating the subtraction like this instead of using
2148 * -= allows us to ensure the result type is uint16_t
2149 * avoiding compiler warnings on gcc 8.1 at least */
2150 m->data_off = (uint16_t)(m->data_off - len);
2151 m->data_len = (uint16_t)(m->data_len + len);
2152 m->pkt_len = (m->pkt_len + len);
2154 return (char *)m->buf_addr + m->data_off;
2158 * Append len bytes to an mbuf.
2160 * Append len bytes to an mbuf and return a pointer to the start address
2161 * of the added data. If there is not enough tailroom in the last
2162 * segment, the function will return NULL, without modifying the mbuf.
2167 * The amount of data to append (in bytes).
2169 * A pointer to the start of the newly appended data, or
2170 * NULL if there is not enough tailroom space in the last segment
2172 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
2175 struct rte_mbuf *m_last;
2177 __rte_mbuf_sanity_check(m, 1);
2179 m_last = rte_pktmbuf_lastseg(m);
2180 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
2183 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
2184 m_last->data_len = (uint16_t)(m_last->data_len + len);
2185 m->pkt_len = (m->pkt_len + len);
2186 return (char*) tail;
2190 * Remove len bytes at the beginning of an mbuf.
2192 * Returns a pointer to the start address of the new data area. If the
2193 * length is greater than the length of the first segment, then the
2194 * function will fail and return NULL, without modifying the mbuf.
2199 * The amount of data to remove (in bytes).
2201 * A pointer to the new start of the data.
2203 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
2205 __rte_mbuf_sanity_check(m, 1);
2207 if (unlikely(len > m->data_len))
2210 /* NB: elaborating the addition like this instead of using
2211 * += allows us to ensure the result type is uint16_t
2212 * avoiding compiler warnings on gcc 8.1 at least */
2213 m->data_len = (uint16_t)(m->data_len - len);
2214 m->data_off = (uint16_t)(m->data_off + len);
2215 m->pkt_len = (m->pkt_len - len);
2216 return (char *)m->buf_addr + m->data_off;
2220 * Remove len bytes of data at the end of the mbuf.
2222 * If the length is greater than the length of the last segment, the
2223 * function will fail and return -1 without modifying the mbuf.
2228 * The amount of data to remove (in bytes).
2233 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
2235 struct rte_mbuf *m_last;
2237 __rte_mbuf_sanity_check(m, 1);
2239 m_last = rte_pktmbuf_lastseg(m);
2240 if (unlikely(len > m_last->data_len))
2243 m_last->data_len = (uint16_t)(m_last->data_len - len);
2244 m->pkt_len = (m->pkt_len - len);
2249 * Test if mbuf data is contiguous.
2254 * - 1, if all data is contiguous (one segment).
2255 * - 0, if there is several segments.
2257 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
2259 __rte_mbuf_sanity_check(m, 1);
2260 return !!(m->nb_segs == 1);
2264 * @internal used by rte_pktmbuf_read().
2266 const void *__rte_pktmbuf_read(const struct rte_mbuf *m, uint32_t off,
2267 uint32_t len, void *buf);
2270 * Read len data bytes in a mbuf at specified offset.
2272 * If the data is contiguous, return the pointer in the mbuf data, else
2273 * copy the data in the buffer provided by the user and return its
2277 * The pointer to the mbuf.
2279 * The offset of the data in the mbuf.
2281 * The amount of bytes to read.
2283 * The buffer where data is copied if it is not contiguous in mbuf
2284 * data. Its length should be at least equal to the len parameter.
2286 * The pointer to the data, either in the mbuf if it is contiguous,
2287 * or in the user buffer. If mbuf is too small, NULL is returned.
2289 static inline const void *rte_pktmbuf_read(const struct rte_mbuf *m,
2290 uint32_t off, uint32_t len, void *buf)
2292 if (likely(off + len <= rte_pktmbuf_data_len(m)))
2293 return rte_pktmbuf_mtod_offset(m, char *, off);
2295 return __rte_pktmbuf_read(m, off, len, buf);
2299 * Chain an mbuf to another, thereby creating a segmented packet.
2301 * Note: The implementation will do a linear walk over the segments to find
2302 * the tail entry. For cases when there are many segments, it's better to
2303 * chain the entries manually.
2306 * The head of the mbuf chain (the first packet)
2308 * The mbuf to put last in the chain
2312 * - -EOVERFLOW, if the chain segment limit exceeded
2314 static inline int rte_pktmbuf_chain(struct rte_mbuf *head, struct rte_mbuf *tail)
2316 struct rte_mbuf *cur_tail;
2318 /* Check for number-of-segments-overflow */
2319 if (head->nb_segs + tail->nb_segs > RTE_MBUF_MAX_NB_SEGS)
2322 /* Chain 'tail' onto the old tail */
2323 cur_tail = rte_pktmbuf_lastseg(head);
2324 cur_tail->next = tail;
2326 /* accumulate number of segments and total length.
2327 * NB: elaborating the addition like this instead of using
2328 * -= allows us to ensure the result type is uint16_t
2329 * avoiding compiler warnings on gcc 8.1 at least */
2330 head->nb_segs = (uint16_t)(head->nb_segs + tail->nb_segs);
2331 head->pkt_len += tail->pkt_len;
2333 /* pkt_len is only set in the head */
2334 tail->pkt_len = tail->data_len;
2341 * @b EXPERIMENTAL: This API may change without prior notice.
2343 * For given input values generate raw tx_offload value.
2344 * Note that it is caller responsibility to make sure that input parameters
2345 * don't exceed maximum bit-field values.
2355 * outer_l3_len value.
2357 * outer_l2_len value.
2361 * raw tx_offload value.
2363 static __rte_always_inline uint64_t
2364 rte_mbuf_tx_offload(uint64_t il2, uint64_t il3, uint64_t il4, uint64_t tso,
2365 uint64_t ol3, uint64_t ol2, uint64_t unused)
2367 return il2 << RTE_MBUF_L2_LEN_OFS |
2368 il3 << RTE_MBUF_L3_LEN_OFS |
2369 il4 << RTE_MBUF_L4_LEN_OFS |
2370 tso << RTE_MBUF_TSO_SEGSZ_OFS |
2371 ol3 << RTE_MBUF_OUTL3_LEN_OFS |
2372 ol2 << RTE_MBUF_OUTL2_LEN_OFS |
2373 unused << RTE_MBUF_TXOFLD_UNUSED_OFS;
2377 * Validate general requirements for Tx offload in mbuf.
2379 * This function checks correctness and completeness of Tx offload settings.
2382 * The packet mbuf to be validated.
2384 * 0 if packet is valid
2387 rte_validate_tx_offload(const struct rte_mbuf *m)
2389 uint64_t ol_flags = m->ol_flags;
2391 /* Does packet set any of available offloads? */
2392 if (!(ol_flags & PKT_TX_OFFLOAD_MASK))
2395 /* IP checksum can be counted only for IPv4 packet */
2396 if ((ol_flags & PKT_TX_IP_CKSUM) && (ol_flags & PKT_TX_IPV6))
2399 /* IP type not set when required */
2400 if (ol_flags & (PKT_TX_L4_MASK | PKT_TX_TCP_SEG))
2401 if (!(ol_flags & (PKT_TX_IPV4 | PKT_TX_IPV6)))
2404 /* Check requirements for TSO packet */
2405 if (ol_flags & PKT_TX_TCP_SEG)
2406 if ((m->tso_segsz == 0) ||
2407 ((ol_flags & PKT_TX_IPV4) &&
2408 !(ol_flags & PKT_TX_IP_CKSUM)))
2411 /* PKT_TX_OUTER_IP_CKSUM set for non outer IPv4 packet. */
2412 if ((ol_flags & PKT_TX_OUTER_IP_CKSUM) &&
2413 !(ol_flags & PKT_TX_OUTER_IPV4))
2420 * @internal used by rte_pktmbuf_linearize().
2422 int __rte_pktmbuf_linearize(struct rte_mbuf *mbuf);
2425 * Linearize data in mbuf.
2427 * This function moves the mbuf data in the first segment if there is enough
2428 * tailroom. The subsequent segments are unchained and freed.
2437 rte_pktmbuf_linearize(struct rte_mbuf *mbuf)
2439 if (rte_pktmbuf_is_contiguous(mbuf))
2441 return __rte_pktmbuf_linearize(mbuf);
2445 * Dump an mbuf structure to a file.
2447 * Dump all fields for the given packet mbuf and all its associated
2448 * segments (in the case of a chained buffer).
2451 * A pointer to a file for output
2455 * If dump_len != 0, also dump the "dump_len" first data bytes of
2458 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
2461 * Get the value of mbuf sched queue_id field.
2463 static inline uint32_t
2464 rte_mbuf_sched_queue_get(const struct rte_mbuf *m)
2466 return m->hash.sched.queue_id;
2470 * Get the value of mbuf sched traffic_class field.
2472 static inline uint8_t
2473 rte_mbuf_sched_traffic_class_get(const struct rte_mbuf *m)
2475 return m->hash.sched.traffic_class;
2479 * Get the value of mbuf sched color field.
2481 static inline uint8_t
2482 rte_mbuf_sched_color_get(const struct rte_mbuf *m)
2484 return m->hash.sched.color;
2488 * Get the values of mbuf sched queue_id, traffic_class and color.
2493 * Returns the queue id
2494 * @param traffic_class
2495 * Returns the traffic class id
2497 * Returns the colour id
2500 rte_mbuf_sched_get(const struct rte_mbuf *m, uint32_t *queue_id,
2501 uint8_t *traffic_class,
2504 struct rte_mbuf_sched sched = m->hash.sched;
2506 *queue_id = sched.queue_id;
2507 *traffic_class = sched.traffic_class;
2508 *color = sched.color;
2512 * Set the mbuf sched queue_id to the defined value.
2515 rte_mbuf_sched_queue_set(struct rte_mbuf *m, uint32_t queue_id)
2517 m->hash.sched.queue_id = queue_id;
2521 * Set the mbuf sched traffic_class id to the defined value.
2524 rte_mbuf_sched_traffic_class_set(struct rte_mbuf *m, uint8_t traffic_class)
2526 m->hash.sched.traffic_class = traffic_class;
2530 * Set the mbuf sched color id to the defined value.
2533 rte_mbuf_sched_color_set(struct rte_mbuf *m, uint8_t color)
2535 m->hash.sched.color = color;
2539 * Set the mbuf sched queue_id, traffic_class and color.
2544 * Queue id value to be set
2545 * @param traffic_class
2546 * Traffic class id value to be set
2548 * Color id to be set
2551 rte_mbuf_sched_set(struct rte_mbuf *m, uint32_t queue_id,
2552 uint8_t traffic_class,
2555 m->hash.sched = (struct rte_mbuf_sched){
2556 .queue_id = queue_id,
2557 .traffic_class = traffic_class,
2567 #endif /* _RTE_MBUF_H_ */