4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
5 * Copyright 2014 6WIND S.A.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * * Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * * Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in
16 * the documentation and/or other materials provided with the
18 * * Neither the name of Intel Corporation nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
25 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
26 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
27 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
28 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
32 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
42 * The mbuf library provides the ability to create and destroy buffers
43 * that may be used by the RTE application to store message
44 * buffers. The message buffers are stored in a mempool, using the
45 * RTE mempool library.
47 * This library provides an API to allocate/free packet mbufs, which are
48 * used to carry network packets.
50 * To understand the concepts of packet buffers or mbufs, you
51 * should read "TCP/IP Illustrated, Volume 2: The Implementation,
52 * Addison-Wesley, 1995, ISBN 0-201-63354-X from Richard Stevens"
53 * http://www.kohala.com/start/tcpipiv2.html
57 #include <rte_common.h>
58 #include <rte_mempool.h>
59 #include <rte_memory.h>
60 #include <rte_atomic.h>
61 #include <rte_prefetch.h>
62 #include <rte_branch_prediction.h>
63 #include <rte_mbuf_ptype.h>
70 * Packet Offload Features Flags. It also carry packet type information.
71 * Critical resources. Both rx/tx shared these bits. Be cautious on any change
73 * - RX flags start at bit position zero, and get added to the left of previous
75 * - The most-significant 3 bits are reserved for generic mbuf flags
76 * - TX flags therefore start at bit position 60 (i.e. 63-3), and new flags get
77 * added to the right of the previously defined flags i.e. they should count
78 * downwards, not upwards.
80 * Keep these flags synchronized with rte_get_rx_ol_flag_name() and
81 * rte_get_tx_ol_flag_name().
85 * RX packet is a 802.1q VLAN packet. This flag was set by PMDs when
86 * the packet is recognized as a VLAN, but the behavior between PMDs
87 * was not the same. This flag is kept for some time to avoid breaking
88 * applications and should be replaced by PKT_RX_VLAN_STRIPPED.
90 #define PKT_RX_VLAN_PKT (1ULL << 0)
92 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
93 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
97 * Checking this flag alone is deprecated: check the 2 bits of
98 * PKT_RX_L4_CKSUM_MASK.
99 * This flag was set when the L4 checksum of a packet was detected as
100 * wrong by the hardware.
102 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3)
106 * Checking this flag alone is deprecated: check the 2 bits of
107 * PKT_RX_IP_CKSUM_MASK.
108 * This flag was set when the IP checksum of a packet was detected as
109 * wrong by the hardware.
111 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4)
113 #define PKT_RX_EIP_CKSUM_BAD (1ULL << 5) /**< External IP header checksum error. */
116 * A vlan has been stripped by the hardware and its tci is saved in
117 * mbuf->vlan_tci. This can only happen if vlan stripping is enabled
118 * in the RX configuration of the PMD.
120 #define PKT_RX_VLAN_STRIPPED (1ULL << 6)
123 * Mask of bits used to determine the status of RX IP checksum.
124 * - PKT_RX_IP_CKSUM_UNKNOWN: no information about the RX IP checksum
125 * - PKT_RX_IP_CKSUM_BAD: the IP checksum in the packet is wrong
126 * - PKT_RX_IP_CKSUM_GOOD: the IP checksum in the packet is valid
127 * - PKT_RX_IP_CKSUM_NONE: the IP checksum is not correct in the packet
128 * data, but the integrity of the IP header is verified.
130 #define PKT_RX_IP_CKSUM_MASK ((1ULL << 4) | (1ULL << 7))
132 #define PKT_RX_IP_CKSUM_UNKNOWN 0
133 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4)
134 #define PKT_RX_IP_CKSUM_GOOD (1ULL << 7)
135 #define PKT_RX_IP_CKSUM_NONE ((1ULL << 4) | (1ULL << 7))
138 * Mask of bits used to determine the status of RX L4 checksum.
139 * - PKT_RX_L4_CKSUM_UNKNOWN: no information about the RX L4 checksum
140 * - PKT_RX_L4_CKSUM_BAD: the L4 checksum in the packet is wrong
141 * - PKT_RX_L4_CKSUM_GOOD: the L4 checksum in the packet is valid
142 * - PKT_RX_L4_CKSUM_NONE: the L4 checksum is not correct in the packet
143 * data, but the integrity of the L4 data is verified.
145 #define PKT_RX_L4_CKSUM_MASK ((1ULL << 3) | (1ULL << 8))
147 #define PKT_RX_L4_CKSUM_UNKNOWN 0
148 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3)
149 #define PKT_RX_L4_CKSUM_GOOD (1ULL << 8)
150 #define PKT_RX_L4_CKSUM_NONE ((1ULL << 3) | (1ULL << 8))
152 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
153 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
154 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
155 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
158 * The 2 vlans have been stripped by the hardware and their tci are
159 * saved in mbuf->vlan_tci (inner) and mbuf->vlan_tci_outer (outer).
160 * This can only happen if vlan stripping is enabled in the RX
161 * configuration of the PMD. If this flag is set, PKT_RX_VLAN_STRIPPED
164 #define PKT_RX_QINQ_STRIPPED (1ULL << 15)
168 * RX packet with double VLAN stripped.
169 * This flag is replaced by PKT_RX_QINQ_STRIPPED.
171 #define PKT_RX_QINQ_PKT PKT_RX_QINQ_STRIPPED
174 * When packets are coalesced by a hardware or virtual driver, this flag
175 * can be set in the RX mbuf, meaning that the m->tso_segsz field is
176 * valid and is set to the segment size of original packets.
178 #define PKT_RX_LRO (1ULL << 16)
180 /* add new RX flags here */
182 /* add new TX flags here */
185 * Offload the MACsec. This flag must be set by the application to enable
186 * this offload feature for a packet to be transmitted.
188 #define PKT_TX_MACSEC (1ULL << 44)
191 * Bits 45:48 used for the tunnel type.
192 * When doing Tx offload like TSO or checksum, the HW needs to configure the
193 * tunnel type into the HW descriptors.
195 #define PKT_TX_TUNNEL_VXLAN (0x1ULL << 45)
196 #define PKT_TX_TUNNEL_GRE (0x2ULL << 45)
197 #define PKT_TX_TUNNEL_IPIP (0x3ULL << 45)
198 #define PKT_TX_TUNNEL_GENEVE (0x4ULL << 45)
199 /* add new TX TUNNEL type here */
200 #define PKT_TX_TUNNEL_MASK (0xFULL << 45)
203 * Second VLAN insertion (QinQ) flag.
205 #define PKT_TX_QINQ_PKT (1ULL << 49) /**< TX packet with double VLAN inserted. */
208 * TCP segmentation offload. To enable this offload feature for a
209 * packet to be transmitted on hardware supporting TSO:
210 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
212 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
213 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
215 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
216 * - calculate the pseudo header checksum without taking ip_len in account,
217 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
218 * rte_ipv6_phdr_cksum() that can be used as helpers.
220 #define PKT_TX_TCP_SEG (1ULL << 50)
222 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
225 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
226 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
227 * L4 checksum offload, the user needs to:
228 * - fill l2_len and l3_len in mbuf
229 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
230 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
231 * - calculate the pseudo header checksum and set it in the L4 header (only
232 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
233 * For SCTP, set the crc field to 0.
235 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
236 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
237 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
238 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
239 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
242 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
243 * also be set by the application, although a PMD will only check
245 * - set the IP checksum field in the packet to 0
246 * - fill the mbuf offload information: l2_len, l3_len
248 #define PKT_TX_IP_CKSUM (1ULL << 54)
251 * Packet is IPv4. This flag must be set when using any offload feature
252 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
253 * packet. If the packet is a tunneled packet, this flag is related to
256 #define PKT_TX_IPV4 (1ULL << 55)
259 * Packet is IPv6. This flag must be set when using an offload feature
260 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
261 * packet. If the packet is a tunneled packet, this flag is related to
264 #define PKT_TX_IPV6 (1ULL << 56)
266 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
269 * Offload the IP checksum of an external header in the hardware. The
270 * flag PKT_TX_OUTER_IPV4 should also be set by the application, alto ugh
271 * a PMD will only check PKT_TX_IP_CKSUM. The IP checksum field in the
272 * packet must be set to 0.
273 * - set the outer IP checksum field in the packet to 0
274 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
276 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
279 * Packet outer header is IPv4. This flag must be set when using any
280 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
281 * outer header of the tunneled packet is an IPv4 packet.
283 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
286 * Packet outer header is IPv6. This flag must be set when using any
287 * outer offload feature (L4 checksum) to tell the NIC that the outer
288 * header of the tunneled packet is an IPv6 packet.
290 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
293 * Bitmask of all supported packet Tx offload features flags,
294 * which can be set for packet.
296 #define PKT_TX_OFFLOAD_MASK ( \
299 PKT_TX_OUTER_IP_CKSUM | \
305 #define __RESERVED (1ULL << 61) /**< reserved for future mbuf use */
307 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
309 /* Use final bit of flags to indicate a control mbuf */
310 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
312 /** Alignment constraint of mbuf private area. */
313 #define RTE_MBUF_PRIV_ALIGN 8
316 * Get the name of a RX offload flag
319 * The mask describing the flag.
321 * The name of this flag, or NULL if it's not a valid RX flag.
323 const char *rte_get_rx_ol_flag_name(uint64_t mask);
326 * Dump the list of RX offload flags in a buffer
329 * The mask describing the RX flags.
333 * The length of the buffer.
335 * 0 on success, (-1) on error.
337 int rte_get_rx_ol_flag_list(uint64_t mask, char *buf, size_t buflen);
340 * Get the name of a TX offload flag
343 * The mask describing the flag. Usually only one bit must be set.
344 * Several bits can be given if they belong to the same mask.
345 * Ex: PKT_TX_L4_MASK.
347 * The name of this flag, or NULL if it's not a valid TX flag.
349 const char *rte_get_tx_ol_flag_name(uint64_t mask);
352 * Dump the list of TX offload flags in a buffer
355 * The mask describing the TX flags.
359 * The length of the buffer.
361 * 0 on success, (-1) on error.
363 int rte_get_tx_ol_flag_list(uint64_t mask, char *buf, size_t buflen);
366 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
367 * splitting it into multiple segments.
368 * So, for mbufs that planned to be involved into RX/TX, the recommended
369 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
371 #define RTE_MBUF_DEFAULT_DATAROOM 2048
372 #define RTE_MBUF_DEFAULT_BUF_SIZE \
373 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
375 /* define a set of marker types that can be used to refer to set points in the
378 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
380 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
382 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
383 * with a single assignment */
386 * The generic rte_mbuf, containing a packet mbuf.
391 void *buf_addr; /**< Virtual address of segment buffer. */
392 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
394 uint16_t buf_len; /**< Length of segment buffer. */
396 /* next 6 bytes are initialised on RX descriptor rearm */
401 * 16-bit Reference counter.
402 * It should only be accessed using the following functions:
403 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
404 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
405 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
410 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
411 uint16_t refcnt; /**< Non-atomically accessed refcnt */
413 uint8_t nb_segs; /**< Number of segments. */
414 uint8_t port; /**< Input port. */
416 uint64_t ol_flags; /**< Offload features. */
418 /* remaining bytes are set on RX when pulling packet from descriptor */
419 MARKER rx_descriptor_fields1;
422 * The packet type, which is the combination of outer/inner L2, L3, L4
423 * and tunnel types. The packet_type is about data really present in the
424 * mbuf. Example: if vlan stripping is enabled, a received vlan packet
425 * would have RTE_PTYPE_L2_ETHER and not RTE_PTYPE_L2_VLAN because the
426 * vlan is stripped from the data.
430 uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
432 uint32_t l2_type:4; /**< (Outer) L2 type. */
433 uint32_t l3_type:4; /**< (Outer) L3 type. */
434 uint32_t l4_type:4; /**< (Outer) L4 type. */
435 uint32_t tun_type:4; /**< Tunnel type. */
436 uint32_t inner_l2_type:4; /**< Inner L2 type. */
437 uint32_t inner_l3_type:4; /**< Inner L3 type. */
438 uint32_t inner_l4_type:4; /**< Inner L4 type. */
442 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
443 uint16_t data_len; /**< Amount of data in segment buffer. */
444 /** VLAN TCI (CPU order), valid if PKT_RX_VLAN_STRIPPED is set. */
448 uint32_t rss; /**< RSS hash result if RSS enabled */
457 /**< Second 4 flexible bytes */
460 /**< First 4 flexible bytes or FD ID, dependent on
461 PKT_RX_FDIR_* flag in ol_flags. */
462 } fdir; /**< Filter identifier if FDIR enabled */
466 } sched; /**< Hierarchical scheduler */
467 uint32_t usr; /**< User defined tags. See rte_distributor_process() */
468 } hash; /**< hash information */
470 uint32_t seqn; /**< Sequence number. See also rte_reorder_insert() */
472 /** Outer VLAN TCI (CPU order), valid if PKT_RX_QINQ_STRIPPED is set. */
473 uint16_t vlan_tci_outer;
475 /* second cache line - fields only used in slow path or on TX */
476 MARKER cacheline1 __rte_cache_min_aligned;
480 void *userdata; /**< Can be used for external metadata */
481 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
484 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
485 struct rte_mbuf *next; /**< Next segment of scattered packet. */
487 /* fields to support TX offloads */
490 uint64_t tx_offload; /**< combined for easy fetch */
494 /**< L2 (MAC) Header Length for non-tunneling pkt.
495 * Outer_L4_len + ... + Inner_L2_len for tunneling pkt.
497 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
498 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
499 uint64_t tso_segsz:16; /**< TCP TSO segment size */
501 /* fields for TX offloading of tunnels */
502 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
503 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
505 /* uint64_t unused:8; */
509 /** Size of the application private data. In case of an indirect
510 * mbuf, it stores the direct mbuf private data size. */
513 /** Timesync flags for use with IEEE1588. */
515 } __rte_cache_aligned;
518 * Prefetch the first part of the mbuf
520 * The first 64 bytes of the mbuf corresponds to fields that are used early
521 * in the receive path. If the cache line of the architecture is higher than
522 * 64B, the second part will also be prefetched.
525 * The pointer to the mbuf.
528 rte_mbuf_prefetch_part1(struct rte_mbuf *m)
530 rte_prefetch0(&m->cacheline0);
534 * Prefetch the second part of the mbuf
536 * The next 64 bytes of the mbuf corresponds to fields that are used in the
537 * transmit path. If the cache line of the architecture is higher than 64B,
538 * this function does nothing as it is expected that the full mbuf is
542 * The pointer to the mbuf.
545 rte_mbuf_prefetch_part2(struct rte_mbuf *m)
547 #if RTE_CACHE_LINE_SIZE == 64
548 rte_prefetch0(&m->cacheline1);
555 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
558 * Return the DMA address of the beginning of the mbuf data
561 * The pointer to the mbuf.
563 * The physical address of the beginning of the mbuf data
565 static inline phys_addr_t
566 rte_mbuf_data_dma_addr(const struct rte_mbuf *mb)
568 return mb->buf_physaddr + mb->data_off;
572 * Return the default DMA address of the beginning of the mbuf data
574 * This function is used by drivers in their receive function, as it
575 * returns the location where data should be written by the NIC, taking
576 * the default headroom in account.
579 * The pointer to the mbuf.
581 * The physical address of the beginning of the mbuf data
583 static inline phys_addr_t
584 rte_mbuf_data_dma_addr_default(const struct rte_mbuf *mb)
586 return mb->buf_physaddr + RTE_PKTMBUF_HEADROOM;
590 * Return the mbuf owning the data buffer address of an indirect mbuf.
593 * The pointer to the indirect mbuf.
595 * The address of the direct mbuf corresponding to buffer_addr.
597 static inline struct rte_mbuf *
598 rte_mbuf_from_indirect(struct rte_mbuf *mi)
600 return (struct rte_mbuf *)RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
604 * Return the buffer address embedded in the given mbuf.
607 * The pointer to the mbuf.
609 * The address of the data buffer owned by the mbuf.
612 rte_mbuf_to_baddr(struct rte_mbuf *md)
615 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
620 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
622 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
625 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
627 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
630 * Private data in case of pktmbuf pool.
632 * A structure that contains some pktmbuf_pool-specific data that are
633 * appended after the mempool structure (in private data).
635 struct rte_pktmbuf_pool_private {
636 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
637 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
640 #ifdef RTE_LIBRTE_MBUF_DEBUG
642 /** check mbuf type in debug mode */
643 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
645 #else /* RTE_LIBRTE_MBUF_DEBUG */
647 /** check mbuf type in debug mode */
648 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
650 #endif /* RTE_LIBRTE_MBUF_DEBUG */
652 #ifdef RTE_MBUF_REFCNT_ATOMIC
655 * Reads the value of an mbuf's refcnt.
659 * Reference count number.
661 static inline uint16_t
662 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
664 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
668 * Sets an mbuf's refcnt to a defined value.
675 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
677 rte_atomic16_set(&m->refcnt_atomic, new_value);
681 * Adds given value to an mbuf's refcnt and returns its new value.
685 * Value to add/subtract
689 static inline uint16_t
690 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
693 * The atomic_add is an expensive operation, so we don't want to
694 * call it in the case where we know we are the uniq holder of
695 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
696 * operation has to be used because concurrent accesses on the
697 * reference counter can occur.
699 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
700 rte_mbuf_refcnt_set(m, 1 + value);
704 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
707 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
710 * Adds given value to an mbuf's refcnt and returns its new value.
712 static inline uint16_t
713 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
715 m->refcnt = (uint16_t)(m->refcnt + value);
720 * Reads the value of an mbuf's refcnt.
722 static inline uint16_t
723 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
729 * Sets an mbuf's refcnt to the defined value.
732 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
734 m->refcnt = new_value;
737 #endif /* RTE_MBUF_REFCNT_ATOMIC */
740 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
747 * Sanity checks on an mbuf.
749 * Check the consistency of the given mbuf. The function will cause a
750 * panic if corruption is detected.
753 * The mbuf to be checked.
755 * True if the mbuf is a packet header, false if it is a sub-segment
756 * of a packet (in this case, some fields like nb_segs are not checked)
759 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
762 * Allocate an unitialized mbuf from mempool *mp*.
764 * This function can be used by PMDs (especially in RX functions) to
765 * allocate an unitialized mbuf. The driver is responsible of
766 * initializing all the required fields. See rte_pktmbuf_reset().
767 * For standard needs, prefer rte_pktmbuf_alloc().
770 * The mempool from which mbuf is allocated.
772 * - The pointer to the new mbuf on success.
773 * - NULL if allocation failed.
775 static inline struct rte_mbuf *rte_mbuf_raw_alloc(struct rte_mempool *mp)
780 if (rte_mempool_get(mp, &mb) < 0)
782 m = (struct rte_mbuf *)mb;
783 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 0);
784 rte_mbuf_refcnt_set(m, 1);
785 __rte_mbuf_sanity_check(m, 0);
791 * @internal Put mbuf back into its original mempool.
792 * The use of that function is reserved for RTE internal needs.
793 * Please use rte_pktmbuf_free().
796 * The mbuf to be freed.
798 static inline void __attribute__((always_inline))
799 __rte_mbuf_raw_free(struct rte_mbuf *m)
801 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 0);
802 rte_mempool_put(m->pool, m);
805 /* Operations on ctrl mbuf */
808 * The control mbuf constructor.
810 * This function initializes some fields in an mbuf structure that are
811 * not modified by the user once created (mbuf type, origin pool, buffer
812 * start address, and so on). This function is given as a callback function
813 * to rte_mempool_create() at pool creation time.
816 * The mempool from which the mbuf is allocated.
818 * A pointer that can be used by the user to retrieve useful information
819 * for mbuf initialization. This pointer comes from the ``init_arg``
820 * parameter of rte_mempool_create().
822 * The mbuf to initialize.
824 * The index of the mbuf in the pool table.
826 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
827 void *m, unsigned i);
830 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
832 * This new mbuf is initialized with data pointing to the beginning of
833 * buffer, and with a length of zero.
836 * The mempool from which the mbuf is allocated.
838 * - The pointer to the new mbuf on success.
839 * - NULL if allocation failed.
841 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
844 * Free a control mbuf back into its original mempool.
847 * The control mbuf to be freed.
849 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
852 * A macro that returns the pointer to the carried data.
854 * The value that can be read or assigned.
859 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
862 * A macro that returns the length of the carried data.
864 * The value that can be read or assigned.
869 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
872 * Tests if an mbuf is a control mbuf
875 * The mbuf to be tested
877 * - True (1) if the mbuf is a control mbuf
878 * - False(0) otherwise
881 rte_is_ctrlmbuf(struct rte_mbuf *m)
883 return !!(m->ol_flags & CTRL_MBUF_FLAG);
886 /* Operations on pkt mbuf */
889 * The packet mbuf constructor.
891 * This function initializes some fields in the mbuf structure that are
892 * not modified by the user once created (origin pool, buffer start
893 * address, and so on). This function is given as a callback function to
894 * rte_mempool_create() at pool creation time.
897 * The mempool from which mbufs originate.
899 * A pointer that can be used by the user to retrieve useful information
900 * for mbuf initialization. This pointer comes from the ``init_arg``
901 * parameter of rte_mempool_create().
903 * The mbuf to initialize.
905 * The index of the mbuf in the pool table.
907 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
908 void *m, unsigned i);
912 * A packet mbuf pool constructor.
914 * This function initializes the mempool private data in the case of a
915 * pktmbuf pool. This private data is needed by the driver. The
916 * function is given as a callback function to rte_mempool_create() at
917 * pool creation. It can be extended by the user, for example, to
918 * provide another packet size.
921 * The mempool from which mbufs originate.
923 * A pointer that can be used by the user to retrieve useful information
924 * for mbuf initialization. This pointer comes from the ``init_arg``
925 * parameter of rte_mempool_create().
927 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
930 * Create a mbuf pool.
932 * This function creates and initializes a packet mbuf pool. It is
933 * a wrapper to rte_mempool_create() with the proper packet constructor
934 * and mempool constructor.
937 * The name of the mbuf pool.
939 * The number of elements in the mbuf pool. The optimum size (in terms
940 * of memory usage) for a mempool is when n is a power of two minus one:
943 * Size of the per-core object cache. See rte_mempool_create() for
946 * Size of application private are between the rte_mbuf structure
947 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
948 * @param data_room_size
949 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
951 * The socket identifier where the memory should be allocated. The
952 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
955 * The pointer to the new allocated mempool, on success. NULL on error
956 * with rte_errno set appropriately. Possible rte_errno values include:
957 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
958 * - E_RTE_SECONDARY - function was called from a secondary process instance
959 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
960 * - ENOSPC - the maximum number of memzones has already been allocated
961 * - EEXIST - a memzone with the same name already exists
962 * - ENOMEM - no appropriate memory area found in which to create memzone
965 rte_pktmbuf_pool_create(const char *name, unsigned n,
966 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
970 * Get the data room size of mbufs stored in a pktmbuf_pool
972 * The data room size is the amount of data that can be stored in a
973 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
976 * The packet mbuf pool.
978 * The data room size of mbufs stored in this mempool.
980 static inline uint16_t
981 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
983 struct rte_pktmbuf_pool_private *mbp_priv;
985 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
986 return mbp_priv->mbuf_data_room_size;
990 * Get the application private size of mbufs stored in a pktmbuf_pool
992 * The private size of mbuf is a zone located between the rte_mbuf
993 * structure and the data buffer where an application can store data
994 * associated to a packet.
997 * The packet mbuf pool.
999 * The private size of mbufs stored in this mempool.
1001 static inline uint16_t
1002 rte_pktmbuf_priv_size(struct rte_mempool *mp)
1004 struct rte_pktmbuf_pool_private *mbp_priv;
1006 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1007 return mbp_priv->mbuf_priv_size;
1011 * Reset the data_off field of a packet mbuf to its default value.
1013 * The given mbuf must have only one segment, which should be empty.
1016 * The packet mbuf's data_off field has to be reset.
1018 static inline void rte_pktmbuf_reset_headroom(struct rte_mbuf *m)
1020 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
1024 * Reset the fields of a packet mbuf to their default values.
1026 * The given mbuf must have only one segment.
1029 * The packet mbuf to be resetted.
1031 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
1037 m->vlan_tci_outer = 0;
1043 rte_pktmbuf_reset_headroom(m);
1046 __rte_mbuf_sanity_check(m, 1);
1050 * Allocate a new mbuf from a mempool.
1052 * This new mbuf contains one segment, which has a length of 0. The pointer
1053 * to data is initialized to have some bytes of headroom in the buffer
1054 * (if buffer size allows).
1057 * The mempool from which the mbuf is allocated.
1059 * - The pointer to the new mbuf on success.
1060 * - NULL if allocation failed.
1062 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
1065 if ((m = rte_mbuf_raw_alloc(mp)) != NULL)
1066 rte_pktmbuf_reset(m);
1071 * Allocate a bulk of mbufs, initialize refcnt and reset the fields to default
1075 * The mempool from which mbufs are allocated.
1077 * Array of pointers to mbufs
1083 static inline int rte_pktmbuf_alloc_bulk(struct rte_mempool *pool,
1084 struct rte_mbuf **mbufs, unsigned count)
1089 rc = rte_mempool_get_bulk(pool, (void **)mbufs, count);
1093 /* To understand duff's device on loop unwinding optimization, see
1094 * https://en.wikipedia.org/wiki/Duff's_device.
1095 * Here while() loop is used rather than do() while{} to avoid extra
1096 * check if count is zero.
1098 switch (count % 4) {
1100 while (idx != count) {
1101 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1102 rte_mbuf_refcnt_set(mbufs[idx], 1);
1103 rte_pktmbuf_reset(mbufs[idx]);
1106 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1107 rte_mbuf_refcnt_set(mbufs[idx], 1);
1108 rte_pktmbuf_reset(mbufs[idx]);
1111 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1112 rte_mbuf_refcnt_set(mbufs[idx], 1);
1113 rte_pktmbuf_reset(mbufs[idx]);
1116 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1117 rte_mbuf_refcnt_set(mbufs[idx], 1);
1118 rte_pktmbuf_reset(mbufs[idx]);
1126 * Attach packet mbuf to another packet mbuf.
1128 * After attachment we refer the mbuf we attached as 'indirect',
1129 * while mbuf we attached to as 'direct'.
1130 * The direct mbuf's reference counter is incremented.
1132 * Right now, not supported:
1133 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
1134 * - mbuf we trying to attach (mi) is used by someone else
1135 * e.g. it's reference counter is greater then 1.
1138 * The indirect packet mbuf.
1140 * The packet mbuf we're attaching to.
1142 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
1144 struct rte_mbuf *md;
1146 RTE_ASSERT(RTE_MBUF_DIRECT(mi) &&
1147 rte_mbuf_refcnt_read(mi) == 1);
1149 /* if m is not direct, get the mbuf that embeds the data */
1150 if (RTE_MBUF_DIRECT(m))
1153 md = rte_mbuf_from_indirect(m);
1155 rte_mbuf_refcnt_update(md, 1);
1156 mi->priv_size = m->priv_size;
1157 mi->buf_physaddr = m->buf_physaddr;
1158 mi->buf_addr = m->buf_addr;
1159 mi->buf_len = m->buf_len;
1162 mi->data_off = m->data_off;
1163 mi->data_len = m->data_len;
1165 mi->vlan_tci = m->vlan_tci;
1166 mi->vlan_tci_outer = m->vlan_tci_outer;
1167 mi->tx_offload = m->tx_offload;
1171 mi->pkt_len = mi->data_len;
1173 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
1174 mi->packet_type = m->packet_type;
1176 __rte_mbuf_sanity_check(mi, 1);
1177 __rte_mbuf_sanity_check(m, 0);
1181 * Detach an indirect packet mbuf.
1183 * - restore original mbuf address and length values.
1184 * - reset pktmbuf data and data_len to their default values.
1185 * - decrement the direct mbuf's reference counter. When the
1186 * reference counter becomes 0, the direct mbuf is freed.
1188 * All other fields of the given packet mbuf will be left intact.
1191 * The indirect attached packet mbuf.
1193 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
1195 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
1196 struct rte_mempool *mp = m->pool;
1197 uint32_t mbuf_size, buf_len, priv_size;
1199 priv_size = rte_pktmbuf_priv_size(mp);
1200 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
1201 buf_len = rte_pktmbuf_data_room_size(mp);
1203 m->priv_size = priv_size;
1204 m->buf_addr = (char *)m + mbuf_size;
1205 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + mbuf_size;
1206 m->buf_len = (uint16_t)buf_len;
1207 rte_pktmbuf_reset_headroom(m);
1211 if (rte_mbuf_refcnt_update(md, -1) == 0)
1212 __rte_mbuf_raw_free(md);
1215 static inline struct rte_mbuf* __attribute__((always_inline))
1216 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1218 __rte_mbuf_sanity_check(m, 0);
1220 if (likely(rte_mbuf_refcnt_update(m, -1) == 0)) {
1221 /* if this is an indirect mbuf, it is detached. */
1222 if (RTE_MBUF_INDIRECT(m))
1223 rte_pktmbuf_detach(m);
1230 * Free a segment of a packet mbuf into its original mempool.
1232 * Free an mbuf, without parsing other segments in case of chained
1236 * The packet mbuf segment to be freed.
1238 static inline void __attribute__((always_inline))
1239 rte_pktmbuf_free_seg(struct rte_mbuf *m)
1241 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
1243 __rte_mbuf_raw_free(m);
1248 * Free a packet mbuf back into its original mempool.
1250 * Free an mbuf, and all its segments in case of chained buffers. Each
1251 * segment is added back into its original mempool.
1254 * The packet mbuf to be freed.
1256 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
1258 struct rte_mbuf *m_next;
1260 __rte_mbuf_sanity_check(m, 1);
1264 rte_pktmbuf_free_seg(m);
1270 * Creates a "clone" of the given packet mbuf.
1272 * Walks through all segments of the given packet mbuf, and for each of them:
1273 * - Creates a new packet mbuf from the given pool.
1274 * - Attaches newly created mbuf to the segment.
1275 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
1276 * from the original packet mbuf.
1279 * The packet mbuf to be cloned.
1281 * The mempool from which the "clone" mbufs are allocated.
1283 * - The pointer to the new "clone" mbuf on success.
1284 * - NULL if allocation fails.
1286 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
1287 struct rte_mempool *mp)
1289 struct rte_mbuf *mc, *mi, **prev;
1293 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
1298 pktlen = md->pkt_len;
1303 rte_pktmbuf_attach(mi, md);
1306 } while ((md = md->next) != NULL &&
1307 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1311 mc->pkt_len = pktlen;
1313 /* Allocation of new indirect segment failed */
1314 if (unlikely (mi == NULL)) {
1315 rte_pktmbuf_free(mc);
1319 __rte_mbuf_sanity_check(mc, 1);
1324 * Adds given value to the refcnt of all packet mbuf segments.
1326 * Walks through all segments of given packet mbuf and for each of them
1327 * invokes rte_mbuf_refcnt_update().
1330 * The packet mbuf whose refcnt to be updated.
1332 * The value to add to the mbuf's segments refcnt.
1334 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1336 __rte_mbuf_sanity_check(m, 1);
1339 rte_mbuf_refcnt_update(m, v);
1340 } while ((m = m->next) != NULL);
1344 * Get the headroom in a packet mbuf.
1349 * The length of the headroom.
1351 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1353 __rte_mbuf_sanity_check(m, 1);
1358 * Get the tailroom of a packet mbuf.
1363 * The length of the tailroom.
1365 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1367 __rte_mbuf_sanity_check(m, 1);
1368 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1373 * Get the last segment of the packet.
1378 * The last segment of the given mbuf.
1380 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1382 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
1384 __rte_mbuf_sanity_check(m, 1);
1385 while (m2->next != NULL)
1391 * A macro that points to an offset into the data in the mbuf.
1393 * The returned pointer is cast to type t. Before using this
1394 * function, the user must ensure that the first segment is large
1395 * enough to accommodate its data.
1400 * The offset into the mbuf data.
1402 * The type to cast the result into.
1404 #define rte_pktmbuf_mtod_offset(m, t, o) \
1405 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
1408 * A macro that points to the start of the data in the mbuf.
1410 * The returned pointer is cast to type t. Before using this
1411 * function, the user must ensure that the first segment is large
1412 * enough to accommodate its data.
1417 * The type to cast the result into.
1419 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
1422 * A macro that returns the physical address that points to an offset of the
1423 * start of the data in the mbuf
1428 * The offset into the data to calculate address from.
1430 #define rte_pktmbuf_mtophys_offset(m, o) \
1431 (phys_addr_t)((m)->buf_physaddr + (m)->data_off + (o))
1434 * A macro that returns the physical address that points to the start of the
1440 #define rte_pktmbuf_mtophys(m) rte_pktmbuf_mtophys_offset(m, 0)
1443 * A macro that returns the length of the packet.
1445 * The value can be read or assigned.
1450 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1453 * A macro that returns the length of the segment.
1455 * The value can be read or assigned.
1460 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1463 * Prepend len bytes to an mbuf data area.
1465 * Returns a pointer to the new
1466 * data start address. If there is not enough headroom in the first
1467 * segment, the function will return NULL, without modifying the mbuf.
1472 * The amount of data to prepend (in bytes).
1474 * A pointer to the start of the newly prepended data, or
1475 * NULL if there is not enough headroom space in the first segment
1477 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1480 __rte_mbuf_sanity_check(m, 1);
1482 if (unlikely(len > rte_pktmbuf_headroom(m)))
1486 m->data_len = (uint16_t)(m->data_len + len);
1487 m->pkt_len = (m->pkt_len + len);
1489 return (char *)m->buf_addr + m->data_off;
1493 * Append len bytes to an mbuf.
1495 * Append len bytes to an mbuf and return a pointer to the start address
1496 * of the added data. If there is not enough tailroom in the last
1497 * segment, the function will return NULL, without modifying the mbuf.
1502 * The amount of data to append (in bytes).
1504 * A pointer to the start of the newly appended data, or
1505 * NULL if there is not enough tailroom space in the last segment
1507 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1510 struct rte_mbuf *m_last;
1512 __rte_mbuf_sanity_check(m, 1);
1514 m_last = rte_pktmbuf_lastseg(m);
1515 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1518 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1519 m_last->data_len = (uint16_t)(m_last->data_len + len);
1520 m->pkt_len = (m->pkt_len + len);
1521 return (char*) tail;
1525 * Remove len bytes at the beginning of an mbuf.
1527 * Returns a pointer to the start address of the new data area. If the
1528 * length is greater than the length of the first segment, then the
1529 * function will fail and return NULL, without modifying the mbuf.
1534 * The amount of data to remove (in bytes).
1536 * A pointer to the new start of the data.
1538 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1540 __rte_mbuf_sanity_check(m, 1);
1542 if (unlikely(len > m->data_len))
1545 m->data_len = (uint16_t)(m->data_len - len);
1547 m->pkt_len = (m->pkt_len - len);
1548 return (char *)m->buf_addr + m->data_off;
1552 * Remove len bytes of data at the end of the mbuf.
1554 * If the length is greater than the length of the last segment, the
1555 * function will fail and return -1 without modifying the mbuf.
1560 * The amount of data to remove (in bytes).
1565 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1567 struct rte_mbuf *m_last;
1569 __rte_mbuf_sanity_check(m, 1);
1571 m_last = rte_pktmbuf_lastseg(m);
1572 if (unlikely(len > m_last->data_len))
1575 m_last->data_len = (uint16_t)(m_last->data_len - len);
1576 m->pkt_len = (m->pkt_len - len);
1581 * Test if mbuf data is contiguous.
1586 * - 1, if all data is contiguous (one segment).
1587 * - 0, if there is several segments.
1589 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1591 __rte_mbuf_sanity_check(m, 1);
1592 return !!(m->nb_segs == 1);
1596 * @internal used by rte_pktmbuf_read().
1598 const void *__rte_pktmbuf_read(const struct rte_mbuf *m, uint32_t off,
1599 uint32_t len, void *buf);
1602 * Read len data bytes in a mbuf at specified offset.
1604 * If the data is contiguous, return the pointer in the mbuf data, else
1605 * copy the data in the buffer provided by the user and return its
1609 * The pointer to the mbuf.
1611 * The offset of the data in the mbuf.
1613 * The amount of bytes to read.
1615 * The buffer where data is copied if it is not contigous in mbuf
1616 * data. Its length should be at least equal to the len parameter.
1618 * The pointer to the data, either in the mbuf if it is contiguous,
1619 * or in the user buffer. If mbuf is too small, NULL is returned.
1621 static inline const void *rte_pktmbuf_read(const struct rte_mbuf *m,
1622 uint32_t off, uint32_t len, void *buf)
1624 if (likely(off + len <= rte_pktmbuf_data_len(m)))
1625 return rte_pktmbuf_mtod_offset(m, char *, off);
1627 return __rte_pktmbuf_read(m, off, len, buf);
1631 * Chain an mbuf to another, thereby creating a segmented packet.
1633 * Note: The implementation will do a linear walk over the segments to find
1634 * the tail entry. For cases when there are many segments, it's better to
1635 * chain the entries manually.
1638 * The head of the mbuf chain (the first packet)
1640 * The mbuf to put last in the chain
1644 * - -EOVERFLOW, if the chain is full (256 entries)
1646 static inline int rte_pktmbuf_chain(struct rte_mbuf *head, struct rte_mbuf *tail)
1648 struct rte_mbuf *cur_tail;
1650 /* Check for number-of-segments-overflow */
1651 if (head->nb_segs + tail->nb_segs >= 1 << (sizeof(head->nb_segs) * 8))
1654 /* Chain 'tail' onto the old tail */
1655 cur_tail = rte_pktmbuf_lastseg(head);
1656 cur_tail->next = tail;
1658 /* accumulate number of segments and total length. */
1659 head->nb_segs = (uint8_t)(head->nb_segs + tail->nb_segs);
1660 head->pkt_len += tail->pkt_len;
1662 /* pkt_len is only set in the head */
1663 tail->pkt_len = tail->data_len;
1669 * Validate general requirements for Tx offload in mbuf.
1671 * This function checks correctness and completeness of Tx offload settings.
1674 * The packet mbuf to be validated.
1676 * 0 if packet is valid
1679 rte_validate_tx_offload(const struct rte_mbuf *m)
1681 uint64_t ol_flags = m->ol_flags;
1682 uint64_t inner_l3_offset = m->l2_len;
1684 /* Does packet set any of available offloads? */
1685 if (!(ol_flags & PKT_TX_OFFLOAD_MASK))
1688 if (ol_flags & PKT_TX_OUTER_IP_CKSUM)
1689 inner_l3_offset += m->outer_l2_len + m->outer_l3_len;
1691 /* Headers are fragmented */
1692 if (rte_pktmbuf_data_len(m) < inner_l3_offset + m->l3_len + m->l4_len)
1695 /* IP checksum can be counted only for IPv4 packet */
1696 if ((ol_flags & PKT_TX_IP_CKSUM) && (ol_flags & PKT_TX_IPV6))
1699 /* IP type not set when required */
1700 if (ol_flags & (PKT_TX_L4_MASK | PKT_TX_TCP_SEG))
1701 if (!(ol_flags & (PKT_TX_IPV4 | PKT_TX_IPV6)))
1704 /* Check requirements for TSO packet */
1705 if (ol_flags & PKT_TX_TCP_SEG)
1706 if ((m->tso_segsz == 0) ||
1707 ((ol_flags & PKT_TX_IPV4) &&
1708 !(ol_flags & PKT_TX_IP_CKSUM)))
1711 /* PKT_TX_OUTER_IP_CKSUM set for non outer IPv4 packet. */
1712 if ((ol_flags & PKT_TX_OUTER_IP_CKSUM) &&
1713 !(ol_flags & PKT_TX_OUTER_IPV4))
1720 * Linearize data in mbuf.
1722 * This function moves the mbuf data in the first segment if there is enough
1723 * tailroom. The subsequent segments are unchained and freed.
1732 rte_pktmbuf_linearize(struct rte_mbuf *mbuf)
1734 int seg_len, copy_len;
1736 struct rte_mbuf *m_next;
1739 if (rte_pktmbuf_is_contiguous(mbuf))
1742 /* Extend first segment to the total packet length */
1743 copy_len = rte_pktmbuf_pkt_len(mbuf) - rte_pktmbuf_data_len(mbuf);
1745 if (unlikely(copy_len > rte_pktmbuf_tailroom(mbuf)))
1748 buffer = rte_pktmbuf_mtod_offset(mbuf, char *, mbuf->data_len);
1749 mbuf->data_len = (uint16_t)(mbuf->pkt_len);
1751 /* Append data from next segments to the first one */
1756 seg_len = rte_pktmbuf_data_len(m);
1757 rte_memcpy(buffer, rte_pktmbuf_mtod(m, char *), seg_len);
1760 rte_pktmbuf_free_seg(m);
1771 * Dump an mbuf structure to a file.
1773 * Dump all fields for the given packet mbuf and all its associated
1774 * segments (in the case of a chained buffer).
1777 * A pointer to a file for output
1781 * If dump_len != 0, also dump the "dump_len" first data bytes of
1784 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1790 #endif /* _RTE_MBUF_H_ */