[RFC] graph: add support for pcap trace for graph

Message ID 20221223120235.3171516-1-amitprakashs@marvell.com (mailing list archive)
State Superseded, archived
Delegated to: Thomas Monjalon
Headers
Series [RFC] graph: add support for pcap trace for graph |

Checks

Context Check Description
ci/checkpatch warning coding style issues
ci/loongarch-compilation success Compilation OK
ci/loongarch-unit-testing success Unit Testing PASS
ci/Intel-compilation fail Compilation issues
ci/intel-Testing success Testing PASS

Commit Message

Amit Prakash Shukla Dec. 23, 2022, 12:02 p.m. UTC
  Packets are captured at each graph node with the node and mbuf metadata
as part of the pcap.

This is inspired from VPP.

Signed-off-by: Amit Prakash Shukla <amitprakashs@marvell.com>
---
 doc/guides/sample_app_ug/l3_forward_graph.rst |   3 +
 examples/l3fwd-graph/main.c                   |  12 +
 lib/graph/graph_pcap_trace.c                  | 337 ++++++++++++++++++
 lib/graph/graph_populate.c                    |   6 +-
 lib/graph/meson.build                         |   5 +-
 lib/graph/rte_graph_pcap_trace.h              | 149 ++++++++
 lib/graph/rte_graph_worker.h                  |   6 +
 7 files changed, 515 insertions(+), 3 deletions(-)
 create mode 100644 lib/graph/graph_pcap_trace.c
 create mode 100644 lib/graph/rte_graph_pcap_trace.h
  

Comments

Stephen Hemminger Dec. 23, 2022, 4:47 p.m. UTC | #1
On Fri, 23 Dec 2022 17:32:35 +0530
Amit Prakash Shukla <amitprakashs@marvell.com> wrote:

> +
> +	pcap_trace.file_descriptor = open(pcap_trace.file_name,
> +					  O_CREAT | O_TRUNC | O_WRONLY, 0664);
> +	if (pcap_trace.file_descriptor < 0) {
> +		ret = 1;
> +		goto done;
> +	}
> +	pcap_trace.n_pcap_data_written = 0;
> +
> +	/* Write file header. */
> +	memset(&file_hdr, 0, sizeof(file_hdr));
> +	file_hdr.magic = 0xa1b2c3d4;
> +	file_hdr.major_version = 2;
> +	file_hdr.minor_version = 4;
> +	file_hdr.time_zone = 0;
> +	file_hdr.max_packet_size_in_bytes = ((1 << 16) - 1);
> +	file_hdr.packet_type = pcap_trace.packet_type;
> +	n = write(pcap_trace.file_descriptor, &file_hdr, sizeof(file_hdr));
> +	if (n != sizeof(file_hdr)) {
> +		ret = 1;
> +		goto done;
> +	}
> +
> +	while (pcap_trace.n_bytes > pcap_trace.n_pcap_data_written) {
> +		int n = pcap_trace.n_bytes - pcap_trace.n_pcap_data_written;
> +
> +		n = write(pcap_trace.file_descriptor,

NAK please use lib/pcapng rather than rolling your own pcap format code
  
Amit Prakash Shukla Jan. 6, 2023, 10:40 a.m. UTC | #2
Thanks Stephen for the review. Sure, will use lib/pcapng.

I see dpdk libpcapng adds most of the debugging data, however I would like to add a node name to the packets which I am thinking of adding using 'comment' option under 'Enhanced Packet Block' .

Please let me know if that's fine.

Thanks,
Amit Shukla

> -----Original Message-----
> From: Stephen Hemminger <stephen@networkplumber.org>
> Sent: Friday, December 23, 2022 10:17 PM
> To: Amit Prakash Shukla <amitprakashs@marvell.com>
> Cc: Jerin Jacob Kollanukkaran <jerinj@marvell.com>; Kiran Kumar
> Kokkilagadda <kirankumark@marvell.com>; Nithin Kumar Dabilpuram
> <ndabilpuram@marvell.com>; dev@dpdk.org
> Subject: [EXT] Re: [RFC PATCH] graph: add support for pcap trace for graph
> 
> External Email
> 
> ----------------------------------------------------------------------
> On Fri, 23 Dec 2022 17:32:35 +0530
> Amit Prakash Shukla <amitprakashs@marvell.com> wrote:
> 
> > +
> > +	pcap_trace.file_descriptor = open(pcap_trace.file_name,
> > +					  O_CREAT | O_TRUNC | O_WRONLY,
> 0664);
> > +	if (pcap_trace.file_descriptor < 0) {
> > +		ret = 1;
> > +		goto done;
> > +	}
> > +	pcap_trace.n_pcap_data_written = 0;
> > +
> > +	/* Write file header. */
> > +	memset(&file_hdr, 0, sizeof(file_hdr));
> > +	file_hdr.magic = 0xa1b2c3d4;
> > +	file_hdr.major_version = 2;
> > +	file_hdr.minor_version = 4;
> > +	file_hdr.time_zone = 0;
> > +	file_hdr.max_packet_size_in_bytes = ((1 << 16) - 1);
> > +	file_hdr.packet_type = pcap_trace.packet_type;
> > +	n = write(pcap_trace.file_descriptor, &file_hdr, sizeof(file_hdr));
> > +	if (n != sizeof(file_hdr)) {
> > +		ret = 1;
> > +		goto done;
> > +	}
> > +
> > +	while (pcap_trace.n_bytes > pcap_trace.n_pcap_data_written) {
> > +		int n = pcap_trace.n_bytes -
> pcap_trace.n_pcap_data_written;
> > +
> > +		n = write(pcap_trace.file_descriptor,
> 
> NAK please use lib/pcapng rather than rolling your own pcap format code
  
Stephen Hemminger Jan. 6, 2023, 4:41 p.m. UTC | #3
On Fri, 6 Jan 2023 10:40:30 +0000
Amit Prakash Shukla <amitprakashs@marvell.com> wrote:

> Thanks Stephen for the review. Sure, will use lib/pcapng.
> 
> I see dpdk libpcapng adds most of the debugging data, however I would like to add a node name to the packets which I am thinking of adding using 'comment' option under 'Enhanced Packet Block' .
> 
> Please let me know if that's fine.
> 
> Thanks,
> Amit Shukla

The pcapng supports comment on file, and the revised version has more interface info.
If you want comment on a per-packet basis that should not be a major change.

Rather than node name, maybe packet id would be better (epb_packetid)
  
Stephen Hemminger Jan. 6, 2023, 6:56 p.m. UTC | #4
On Fri, 6 Jan 2023 10:40:30 +0000
Amit Prakash Shukla <amitprakashs@marvell.com> wrote:

> Thanks Stephen for the review. Sure, will use lib/pcapng.
> 
> I see dpdk libpcapng adds most of the debugging data, however I would like to add a node name to the packets which I am thinking of adding using 'comment' option under 'Enhanced Packet Block' .
> 
> Please let me know if that's fine.
> 
> Thanks,
> Amit Shukla
> 
> > -----Original Message-----
> > From: Stephen Hemminger <stephen@networkplumber.org>
> > Sent: Friday, December 23, 2022 10:17 PM
> > To: Amit Prakash Shukla <amitprakashs@marvell.com>
> > Cc: Jerin Jacob Kollanukkaran <jerinj@marvell.com>; Kiran Kumar
> > Kokkilagadda <kirankumark@marvell.com>; Nithin Kumar Dabilpuram
> > <ndabilpuram@marvell.com>; dev@dpdk.org
> > Subject: [EXT] Re: [RFC PATCH] graph: add support for pcap trace for graph
> > 
> > External Email
> > 
> > ----------------------------------------------------------------------
> > On Fri, 23 Dec 2022 17:32:35 +0530
> > Amit Prakash Shukla <amitprakashs@marvell.com> wrote:
> >   
> > > +
> > > +	pcap_trace.file_descriptor = open(pcap_trace.file_name,
> > > +					  O_CREAT | O_TRUNC | O_WRONLY,  
> > 0664);  
> > > +	if (pcap_trace.file_descriptor < 0) {
> > > +		ret = 1;
> > > +		goto done;
> > > +	}
> > > +	pcap_trace.n_pcap_data_written = 0;
> > > +
> > > +	/* Write file header. */
> > > +	memset(&file_hdr, 0, sizeof(file_hdr));
> > > +	file_hdr.magic = 0xa1b2c3d4;
> > > +	file_hdr.major_version = 2;
> > > +	file_hdr.minor_version = 4;
> > > +	file_hdr.time_zone = 0;
> > > +	file_hdr.max_packet_size_in_bytes = ((1 << 16) - 1);
> > > +	file_hdr.packet_type = pcap_trace.packet_type;
> > > +	n = write(pcap_trace.file_descriptor, &file_hdr, sizeof(file_hdr));
> > > +	if (n != sizeof(file_hdr)) {
> > > +		ret = 1;
> > > +		goto done;
> > > +	}
> > > +
> > > +	while (pcap_trace.n_bytes > pcap_trace.n_pcap_data_written) {
> > > +		int n = pcap_trace.n_bytes -  
> > pcap_trace.n_pcap_data_written;  
> > > +
> > > +		n = write(pcap_trace.file_descriptor,  
> > 
> > NAK please use lib/pcapng rather than rolling your own pcap format code  

Something like this:

diff --git a/lib/pcapng/rte_pcapng.c b/lib/pcapng/rte_pcapng.c
index cb590ea0096c..216ad80f7ed6 100644
--- a/lib/pcapng/rte_pcapng.c
+++ b/lib/pcapng/rte_pcapng.c
@@ -492,7 +492,8 @@ rte_pcapng_copy(uint16_t port_id, uint32_t queue,
 		const struct rte_mbuf *md,
 		struct rte_mempool *mp,
 		uint32_t length, uint64_t cycles,
-		enum rte_pcapng_direction direction)
+		enum rte_pcapng_direction direction,
+		const char *comment)
 {
 	struct pcapng_enhance_packet_block *epb;
 	uint32_t orig_len, data_len, padding, flags;
@@ -552,6 +553,8 @@ rte_pcapng_copy(uint16_t port_id, uint32_t queue,
 	optlen += pcapng_optlen(sizeof(queue));
 	if (rss_hash)
 		optlen += pcapng_optlen(sizeof(uint8_t) + sizeof(uint32_t));
+	if (comment)
+		optlen += pcapng_optlen(strlen(comment));

 	/* reserve trailing options and block length */
 	opt = (struct pcapng_option *)
@@ -590,6 +593,10 @@ rte_pcapng_copy(uint16_t port_id, uint32_t queue,
 					&hash_opt, sizeof(hash_opt));
 	}

+	if (comment)
+		opt = pcapng_add_option(opt, PCAPNG_OPT_COMMENT,
+					comment, strlen(comment));
+
 	/* Note: END_OPT necessary here. Wireshark doesn't do it. */

 	/* Add PCAPNG packet header */
diff --git a/lib/pcapng/rte_pcapng.h b/lib/pcapng/rte_pcapng.h
index 6b8aaffc6e0f..98f35e8ea177 100644
--- a/lib/pcapng/rte_pcapng.h
+++ b/lib/pcapng/rte_pcapng.h
@@ -126,6 +126,8 @@ enum rte_pcapng_direction {
  *   The timestamp in TSC cycles.
  * @param direction
  *   The direction of the packer: receive, transmit or unknown.
+ * @param comment
+ *  Optional: comment on packet
  *
  * @return
  *   - The pointer to the new mbuf formatted for pcapng_write
@@ -137,7 +139,8 @@ struct rte_mbuf *
 rte_pcapng_copy(uint16_t port_id, uint32_t queue,
 		const struct rte_mbuf *m, struct rte_mempool *mp,
 		uint32_t length, uint64_t timestamp,
-		enum rte_pcapng_direction direction);
+		enum rte_pcapng_direction direction,
+		const char *comment);


 /**
  
Amit Prakash Shukla Jan. 10, 2023, 11:26 a.m. UTC | #5
Thanks Stephen for the pointers. I will post next version of the patch with the change.


> -----Original Message-----
> From: Stephen Hemminger <stephen@networkplumber.org>
> Sent: Saturday, January 7, 2023 12:27 AM
> To: Amit Prakash Shukla <amitprakashs@marvell.com>
> Cc: Jerin Jacob Kollanukkaran <jerinj@marvell.com>; Kiran Kumar
> Kokkilagadda <kirankumark@marvell.com>; Nithin Kumar Dabilpuram
> <ndabilpuram@marvell.com>; dev@dpdk.org
> Subject: Re: [EXT] Re: [RFC PATCH] graph: add support for pcap trace for
> graph
> 
> On Fri, 6 Jan 2023 10:40:30 +0000
> Amit Prakash Shukla <amitprakashs@marvell.com> wrote:
> 
> > Thanks Stephen for the review. Sure, will use lib/pcapng.
> >
> > I see dpdk libpcapng adds most of the debugging data, however I would
> like to add a node name to the packets which I am thinking of adding using
> 'comment' option under 'Enhanced Packet Block' .
> >
> > Please let me know if that's fine.
> >
> > Thanks,
> > Amit Shukla
> >
> > > -----Original Message-----
> > > From: Stephen Hemminger <stephen@networkplumber.org>
> > > Sent: Friday, December 23, 2022 10:17 PM
> > > To: Amit Prakash Shukla <amitprakashs@marvell.com>
> > > Cc: Jerin Jacob Kollanukkaran <jerinj@marvell.com>; Kiran Kumar
> > > Kokkilagadda <kirankumark@marvell.com>; Nithin Kumar Dabilpuram
> > > <ndabilpuram@marvell.com>; dev@dpdk.org
> > > Subject: [EXT] Re: [RFC PATCH] graph: add support for pcap trace for
> > > graph
> > >
> > > External Email
> > >
> > > --------------------------------------------------------------------
> > > --
> > > On Fri, 23 Dec 2022 17:32:35 +0530
> > > Amit Prakash Shukla <amitprakashs@marvell.com> wrote:
> > >
> > > > +
> > > > +	pcap_trace.file_descriptor = open(pcap_trace.file_name,
> > > > +					  O_CREAT | O_TRUNC | O_WRONLY,
> > > 0664);
> > > > +	if (pcap_trace.file_descriptor < 0) {
> > > > +		ret = 1;
> > > > +		goto done;
> > > > +	}
> > > > +	pcap_trace.n_pcap_data_written = 0;
> > > > +
> > > > +	/* Write file header. */
> > > > +	memset(&file_hdr, 0, sizeof(file_hdr));
> > > > +	file_hdr.magic = 0xa1b2c3d4;
> > > > +	file_hdr.major_version = 2;
> > > > +	file_hdr.minor_version = 4;
> > > > +	file_hdr.time_zone = 0;
> > > > +	file_hdr.max_packet_size_in_bytes = ((1 << 16) - 1);
> > > > +	file_hdr.packet_type = pcap_trace.packet_type;
> > > > +	n = write(pcap_trace.file_descriptor, &file_hdr, sizeof(file_hdr));
> > > > +	if (n != sizeof(file_hdr)) {
> > > > +		ret = 1;
> > > > +		goto done;
> > > > +	}
> > > > +
> > > > +	while (pcap_trace.n_bytes > pcap_trace.n_pcap_data_written) {
> > > > +		int n = pcap_trace.n_bytes -
> > > pcap_trace.n_pcap_data_written;
> > > > +
> > > > +		n = write(pcap_trace.file_descriptor,
> > >
> > > NAK please use lib/pcapng rather than rolling your own pcap format
> > > code
> 
> Something like this:
> 
> diff --git a/lib/pcapng/rte_pcapng.c b/lib/pcapng/rte_pcapng.c index
> cb590ea0096c..216ad80f7ed6 100644
> --- a/lib/pcapng/rte_pcapng.c
> +++ b/lib/pcapng/rte_pcapng.c
> @@ -492,7 +492,8 @@ rte_pcapng_copy(uint16_t port_id, uint32_t queue,
>  		const struct rte_mbuf *md,
>  		struct rte_mempool *mp,
>  		uint32_t length, uint64_t cycles,
> -		enum rte_pcapng_direction direction)
> +		enum rte_pcapng_direction direction,
> +		const char *comment)
>  {
>  	struct pcapng_enhance_packet_block *epb;
>  	uint32_t orig_len, data_len, padding, flags; @@ -552,6 +553,8 @@
> rte_pcapng_copy(uint16_t port_id, uint32_t queue,
>  	optlen += pcapng_optlen(sizeof(queue));
>  	if (rss_hash)
>  		optlen += pcapng_optlen(sizeof(uint8_t) + sizeof(uint32_t));
> +	if (comment)
> +		optlen += pcapng_optlen(strlen(comment));
> 
>  	/* reserve trailing options and block length */
>  	opt = (struct pcapng_option *)
> @@ -590,6 +593,10 @@ rte_pcapng_copy(uint16_t port_id, uint32_t queue,
>  					&hash_opt, sizeof(hash_opt));
>  	}
> 
> +	if (comment)
> +		opt = pcapng_add_option(opt, PCAPNG_OPT_COMMENT,
> +					comment, strlen(comment));
> +
>  	/* Note: END_OPT necessary here. Wireshark doesn't do it. */
> 
>  	/* Add PCAPNG packet header */
> diff --git a/lib/pcapng/rte_pcapng.h b/lib/pcapng/rte_pcapng.h index
> 6b8aaffc6e0f..98f35e8ea177 100644
> --- a/lib/pcapng/rte_pcapng.h
> +++ b/lib/pcapng/rte_pcapng.h
> @@ -126,6 +126,8 @@ enum rte_pcapng_direction {
>   *   The timestamp in TSC cycles.
>   * @param direction
>   *   The direction of the packer: receive, transmit or unknown.
> + * @param comment
> + *  Optional: comment on packet
>   *
>   * @return
>   *   - The pointer to the new mbuf formatted for pcapng_write
> @@ -137,7 +139,8 @@ struct rte_mbuf *
>  rte_pcapng_copy(uint16_t port_id, uint32_t queue,
>  		const struct rte_mbuf *m, struct rte_mempool *mp,
>  		uint32_t length, uint64_t timestamp,
> -		enum rte_pcapng_direction direction);
> +		enum rte_pcapng_direction direction,
> +		const char *comment);
> 
> 
>  /**
  

Patch

diff --git a/doc/guides/sample_app_ug/l3_forward_graph.rst b/doc/guides/sample_app_ug/l3_forward_graph.rst
index 0a3e0d44ec..cf199bcf81 100644
--- a/doc/guides/sample_app_ug/l3_forward_graph.rst
+++ b/doc/guides/sample_app_ug/l3_forward_graph.rst
@@ -51,6 +51,7 @@  The application has a number of command line options similar to l3fwd::
                                    [--max-pkt-len PKTLEN]
                                    [--no-numa]
                                    [--per-port-pool]
+                                   [--pcap-enable]
 
 Where,
 
@@ -69,6 +70,8 @@  Where,
 
 * ``--per-port-pool:`` Optional, set to use independent buffer pools per port. Without this option, single buffer pool is used for all ports.
 
+* ``--pcap-enable:`` Optional, Enables packet capture in pcap format on each node with mbuf and node metadata.
+
 For example, consider a dual processor socket platform with 8 physical cores, where cores 0-7 and 16-23 appear on socket 0,
 while cores 8-15 and 24-31 appear on socket 1.
 
diff --git a/examples/l3fwd-graph/main.c b/examples/l3fwd-graph/main.c
index 6dcb6ee92b..b6408310aa 100644
--- a/examples/l3fwd-graph/main.c
+++ b/examples/l3fwd-graph/main.c
@@ -404,6 +404,7 @@  static const char short_options[] = "p:" /* portmask */
 #define CMD_LINE_OPT_NO_NUMA	   "no-numa"
 #define CMD_LINE_OPT_MAX_PKT_LEN   "max-pkt-len"
 #define CMD_LINE_OPT_PER_PORT_POOL "per-port-pool"
+#define CMD_LINE_OPT_PCAP_ENABLE   "pcap-enable"
 enum {
 	/* Long options mapped to a short option */
 
@@ -416,6 +417,7 @@  enum {
 	CMD_LINE_OPT_NO_NUMA_NUM,
 	CMD_LINE_OPT_MAX_PKT_LEN_NUM,
 	CMD_LINE_OPT_PARSE_PER_PORT_POOL,
+	CMD_LINE_OPT_PARSE_PCAP_ENABLE,
 };
 
 static const struct option lgopts[] = {
@@ -424,6 +426,7 @@  static const struct option lgopts[] = {
 	{CMD_LINE_OPT_NO_NUMA, 0, 0, CMD_LINE_OPT_NO_NUMA_NUM},
 	{CMD_LINE_OPT_MAX_PKT_LEN, 1, 0, CMD_LINE_OPT_MAX_PKT_LEN_NUM},
 	{CMD_LINE_OPT_PER_PORT_POOL, 0, 0, CMD_LINE_OPT_PARSE_PER_PORT_POOL},
+	{CMD_LINE_OPT_PCAP_ENABLE, 0, 0, CMD_LINE_OPT_PARSE_PCAP_ENABLE},
 	{NULL, 0, 0, 0},
 };
 
@@ -498,6 +501,11 @@  parse_args(int argc, char **argv)
 			per_port_pool = 1;
 			break;
 
+		case CMD_LINE_OPT_PARSE_PCAP_ENABLE:
+			printf("Packet capture enabled\n");
+			set_pcap_trace(1);
+			break;
+
 		default:
 			print_usage(prgname);
 			return -1;
@@ -831,6 +839,7 @@  main(int argc, char **argv)
 			local_port_conf.txmode.offloads |=
 				RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
 
+		local_port_conf.rxmode.offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
 		local_port_conf.rx_adv_conf.rss_conf.rss_hf &=
 			dev_info.flow_type_rss_offloads;
 		if (local_port_conf.rx_adv_conf.rss_conf.rss_hf !=
@@ -1116,6 +1125,9 @@  main(int argc, char **argv)
 	}
 	/* >8 End of adding route to ip4 graph infa. */
 
+	if (is_pcap_trace_enable())
+		rte_graph_pcap_trace_init();
+
 	/* Launch per-lcore init on every worker lcore */
 	rte_eal_mp_remote_launch(graph_main_loop, NULL, SKIP_MAIN);
 
diff --git a/lib/graph/graph_pcap_trace.c b/lib/graph/graph_pcap_trace.c
new file mode 100644
index 0000000000..c5be07de6d
--- /dev/null
+++ b/lib/graph/graph_pcap_trace.c
@@ -0,0 +1,337 @@ 
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(C) 2022 Marvell International Ltd.
+ */
+
+#include <unistd.h>
+#include <sys/time.h>
+#include <fcntl.h>
+
+#include <rte_malloc.h>
+#include <rte_mbuf.h>
+#include <rte_net.h>
+#include <rte_flow.h>
+
+#include "rte_graph_worker.h"
+
+#define RTE_PCAP_MAJOR_VERSION 1
+#define RTE_PCAP_MINOR_VERSION 0
+#define MAX_PKT_TO_CAPTURE 200
+#define MAX_PCAP_BUF_SZ 2048
+
+#define PCAP_DUMP_STR(buf, buf_size, cur_len, ...)                           \
+do {                                                                         \
+	if (cur_len >= buf_size)                                             \
+		break;                                                       \
+	cur_len += snprintf(buf + cur_len, buf_size - cur_len, __VA_ARGS__); \
+} while (0)
+
+#define PCAP_DUMP_DATA(dbuf, buf_size, cur_len, sbuf, len)                   \
+do {                                                                         \
+	if ((cur_len + len) >= buf_size)                                     \
+		break;                                                       \
+	rte_memcpy(dbuf + cur_len, sbuf, len);                               \
+	cur_len += len;                                                      \
+} while (0)
+
+static int pcap_trace_enable;
+static pcap_trace_t pcap_trace;
+
+static uint16_t pkt_metadata_dump(struct rte_mbuf *mbuf, char *buffer,
+				  size_t buf_size, uint16_t cur_len);
+static int pcap_trace_close(void);
+static int pcap_trace_write(void);
+
+void
+rte_graph_pcap_trace_init(void)
+{
+	memset(&pcap_trace, 0, sizeof(pcap_trace_t));
+	rte_spinlock_init(&pcap_trace.lock);
+	pcap_trace.packet_type = PCAP_PACKET_TYPE_USER13;
+	pcap_trace.n_packets_to_capture = MAX_PKT_TO_CAPTURE;
+	pcap_trace.file_name = "/tmp/dpdk.pcap";
+	pcap_trace.pcap_data = NULL;
+}
+
+void
+set_pcap_trace(int val)
+{
+	pcap_trace_enable = val;
+}
+
+int
+is_pcap_trace_enable(void)
+{
+	return pcap_trace_enable;
+}
+
+static int
+pcap_trace_close(void)
+{
+	close(pcap_trace.file_descriptor);
+	pcap_trace.file_descriptor = -1;
+	return 0;
+}
+
+static int
+pcap_trace_write(void)
+{
+	pcap_file_header_t file_hdr;
+	int ret = 0;
+	int n;
+
+	if (!pcap_trace.file_name)
+		pcap_trace.file_name = "/tmp/dpdk.pcap";
+
+	pcap_trace.file_descriptor = open(pcap_trace.file_name,
+					  O_CREAT | O_TRUNC | O_WRONLY, 0664);
+	if (pcap_trace.file_descriptor < 0) {
+		ret = 1;
+		goto done;
+	}
+	pcap_trace.n_pcap_data_written = 0;
+
+	/* Write file header. */
+	memset(&file_hdr, 0, sizeof(file_hdr));
+	file_hdr.magic = 0xa1b2c3d4;
+	file_hdr.major_version = 2;
+	file_hdr.minor_version = 4;
+	file_hdr.time_zone = 0;
+	file_hdr.max_packet_size_in_bytes = ((1 << 16) - 1);
+	file_hdr.packet_type = pcap_trace.packet_type;
+	n = write(pcap_trace.file_descriptor, &file_hdr, sizeof(file_hdr));
+	if (n != sizeof(file_hdr)) {
+		ret = 1;
+		goto done;
+	}
+
+	while (pcap_trace.n_bytes > pcap_trace.n_pcap_data_written) {
+		int n = pcap_trace.n_bytes - pcap_trace.n_pcap_data_written;
+
+		n = write(pcap_trace.file_descriptor,
+			  (pcap_trace.pcap_data +
+			   pcap_trace.n_pcap_data_written), n);
+
+		if (n < 0 && errno != 0) {
+			ret = 1;
+			goto done;
+		}
+
+		pcap_trace.n_pcap_data_written += n;
+	}
+
+	if (pcap_trace.n_pcap_data_written >= pcap_trace.n_bytes)	{
+		rte_free(pcap_trace.pcap_data);
+		pcap_trace.pcap_data = NULL;
+		pcap_trace.n_pcap_data_written = 0;
+	}
+
+	if (pcap_trace.n_packets_captured >=
+	    pcap_trace.n_packets_to_capture)
+		pcap_trace_close();
+
+done:
+	if (ret) {
+		if (pcap_trace.file_descriptor >= 0)
+			pcap_trace_close();
+	}
+
+	return ret;
+}
+
+static uint16_t
+pkt_metadata_dump(struct rte_mbuf *mbuf, char *buffer, size_t buf_size,
+		  uint16_t cur_len)
+{
+	struct rte_flow_restore_info info = { 0, };
+	struct rte_net_hdr_lens hdr_lens;
+	struct rte_flow_error error;
+	uint32_t sw_packet_type;
+	uint64_t ol_flags;
+	char buf[256];
+	int ret;
+
+	ret = rte_flow_get_restore_info(mbuf->port, mbuf, &info, &error);
+	if (!ret) {
+		PCAP_DUMP_STR(buffer, buf_size, cur_len, "restore info:");
+
+		if (info.flags & RTE_FLOW_RESTORE_INFO_ENCAPSULATED)
+			PCAP_DUMP_STR(buffer, buf_size, cur_len,
+				      "outer header present. ");
+		else
+			PCAP_DUMP_STR(buffer, buf_size, cur_len,
+				      "no outer header. ");
+		if (info.flags & RTE_FLOW_RESTORE_INFO_GROUP_ID)
+			PCAP_DUMP_STR(buffer, buf_size, cur_len,
+				      "miss group %u", info.group_id);
+		else
+			PCAP_DUMP_STR(buffer, buf_size, cur_len,
+				      "no miss group");
+		PCAP_DUMP_STR(buffer, buf_size, cur_len, "\n");
+	}
+
+	PCAP_DUMP_STR(buffer, buf_size, cur_len,
+		      "pool=%s  length=%u  nb_segs=%d\n", mbuf->pool->name,
+		      (unsigned int) mbuf->pkt_len, (int)mbuf->nb_segs);
+
+	ol_flags = mbuf->ol_flags;
+	if (ol_flags & RTE_MBUF_F_RX_RSS_HASH) {
+		PCAP_DUMP_STR(buffer, buf_size, cur_len,
+			      "RSS hash=0x%x\n",
+			      (unsigned int) mbuf->hash.rss);
+	}
+
+	if (ol_flags & RTE_MBUF_F_RX_FDIR) {
+		PCAP_DUMP_STR(buffer, buf_size, cur_len, "FDIR matched ");
+		if (ol_flags & RTE_MBUF_F_RX_FDIR_ID)
+			PCAP_DUMP_STR(buffer, buf_size, cur_len,
+				      "ID=0x%x\n", mbuf->hash.fdir.hi);
+		else if (ol_flags & RTE_MBUF_F_RX_FDIR_FLX)
+			PCAP_DUMP_STR(buffer, buf_size, cur_len,
+				      "flex bytes=0x%08x %08x\n",
+				      mbuf->hash.fdir.hi, mbuf->hash.fdir.lo);
+		else
+			PCAP_DUMP_STR(buffer, buf_size, cur_len,
+				      "hash=0x%x ID=0x%x\n",
+				      mbuf->hash.fdir.hash, mbuf->hash.fdir.id);
+	}
+
+	if (ol_flags & RTE_MBUF_F_RX_QINQ)
+		PCAP_DUMP_STR(buffer, buf_size, cur_len,
+			      "QinQ VLAN tci=0x%x, VLAN tci outer=0x%x\n",
+			      mbuf->vlan_tci, mbuf->vlan_tci_outer);
+	else if (ol_flags & RTE_MBUF_F_RX_VLAN)
+		PCAP_DUMP_STR(buffer, buf_size, cur_len, "VLAN tci=0x%x\n",
+			      mbuf->vlan_tci);
+
+	if (mbuf->packet_type) {
+		rte_get_ptype_name(mbuf->packet_type, buf, sizeof(buf));
+		PCAP_DUMP_STR(buffer, buf_size, cur_len, "hw ptype: %s\n",
+			      buf);
+	}
+
+	sw_packet_type = rte_net_get_ptype(mbuf, &hdr_lens, RTE_PTYPE_ALL_MASK);
+	rte_get_ptype_name(sw_packet_type, buf, sizeof(buf));
+	PCAP_DUMP_STR(buffer, buf_size, cur_len, "sw ptype: %s\n", buf);
+
+	if (sw_packet_type & RTE_PTYPE_L2_MASK)
+		PCAP_DUMP_STR(buffer, buf_size, cur_len, "l2_len=%d  ",
+			      hdr_lens.l2_len);
+	if (sw_packet_type & RTE_PTYPE_L3_MASK)
+		PCAP_DUMP_STR(buffer, buf_size, cur_len, "l3_len=%d  ",
+			      hdr_lens.l3_len);
+	if (sw_packet_type & RTE_PTYPE_L4_MASK)
+		PCAP_DUMP_STR(buffer, buf_size, cur_len, "l4_len=%d  ",
+			      hdr_lens.l4_len);
+	if (sw_packet_type & RTE_PTYPE_TUNNEL_MASK)
+		PCAP_DUMP_STR(buffer, buf_size, cur_len,
+			      "tunnel_len=%d  ", hdr_lens.tunnel_len);
+	if (sw_packet_type & RTE_PTYPE_INNER_L2_MASK)
+		PCAP_DUMP_STR(buffer, buf_size, cur_len,
+			      "inner_l2_len=%d  ", hdr_lens.inner_l2_len);
+	if (sw_packet_type & RTE_PTYPE_INNER_L3_MASK)
+		PCAP_DUMP_STR(buffer, buf_size, cur_len,
+			      "inner_l3_len=%d  ", hdr_lens.inner_l3_len);
+	if (sw_packet_type & RTE_PTYPE_INNER_L4_MASK)
+		PCAP_DUMP_STR(buffer, buf_size, cur_len,
+			      "inner_l4_len=%d  ", hdr_lens.inner_l4_len);
+
+	PCAP_DUMP_STR(buffer, buf_size, cur_len, "\n");
+
+	rte_get_rx_ol_flag_list(mbuf->ol_flags, buf, sizeof(buf));
+	PCAP_DUMP_STR(buffer, buf_size, cur_len, "Rx ol_flags: %s\n", buf);
+
+	rte_get_tx_ol_flag_list(mbuf->ol_flags, buf, sizeof(buf));
+	PCAP_DUMP_STR(buffer, buf_size, cur_len, "Tx ol_flags: %s\n", buf);
+
+	/* Keep it last. */
+	buffer[cur_len++] = '\0';
+	return cur_len;
+}
+
+uint16_t
+rte_graph_pcap_trace_dispatch(struct rte_graph *graph __rte_unused,
+			      struct rte_node *node, void **objs,
+			      uint16_t nb_objs)
+{
+	struct rte_mbuf *mbuf;
+	pcap_packet_header_t *phdr;
+	pcap_version_hdr_t vhdr = {0};
+	struct timeval ts = {0};
+	char buffer[MAX_PCAP_BUF_SZ] = {0};
+	uint16_t total_len, n_bytes, cb_len;
+	uint8_t *pkt_data;
+	uint8_t *data;
+	int i;
+
+	gettimeofday(&ts, NULL);
+
+	for (i = 0; i < nb_objs; i++) {
+		if (pcap_trace.n_packets_captured >=
+		    pcap_trace.n_packets_to_capture)
+			break;
+
+		mbuf = (struct rte_mbuf *)objs[i];
+
+		memset(buffer, 0, sizeof(buffer));
+		cb_len = 0;
+		total_len = 0;
+		n_bytes = 0;
+
+		vhdr.pcap_major_version = (uint8_t)RTE_PCAP_MAJOR_VERSION;
+		vhdr.pcap_minor_version = (uint8_t)RTE_PCAP_MINOR_VERSION;
+		PCAP_DUMP_DATA(buffer, MAX_PCAP_BUF_SZ, cb_len, &vhdr,
+			       sizeof(pcap_version_hdr_t));
+
+		PCAP_DUMP_DATA(buffer, MAX_PCAP_BUF_SZ, cb_len, node->name,
+			       (strlen(node->name) + 1));
+
+		PCAP_DUMP_DATA(buffer, MAX_PCAP_BUF_SZ, cb_len, node->ctx,
+			       RTE_NODE_CTX_SZ);
+
+		cb_len = pkt_metadata_dump(mbuf, buffer, sizeof(buffer),
+					   cb_len);
+
+		total_len = cb_len + mbuf->pkt_len +
+				sizeof(pcap_packet_header_t);
+		n_bytes = RTE_MIN((int)total_len, 16384);
+
+		rte_spinlock_lock(&pcap_trace.lock);
+
+		if (pcap_trace.pcap_data == NULL) {
+			pcap_trace.pcap_data = rte_malloc(NULL, n_bytes, 0);
+			phdr = (pcap_packet_header_t *)pcap_trace.pcap_data;
+		} else {
+			pcap_trace.pcap_data = rte_realloc(
+					pcap_trace.pcap_data,
+					(pcap_trace.n_bytes + n_bytes), 0);
+
+			phdr = (pcap_packet_header_t *)(pcap_trace.pcap_data +
+					pcap_trace.n_bytes);
+		}
+
+		phdr->time_in_sec = (uint32_t)ts.tv_sec;
+		phdr->time_in_usec = (uint32_t)ts.tv_usec;
+		phdr->n_packet_bytes_stored_in_file =
+				(n_bytes - sizeof(pcap_packet_header_t));
+		phdr->n_bytes_in_packet =
+				(total_len - sizeof(pcap_packet_header_t));
+		data = phdr->data;
+
+		rte_memcpy(data, buffer, cb_len);
+		data += cb_len;
+
+		pkt_data = rte_pktmbuf_mtod(mbuf, uint8_t *);
+		rte_memcpy(data, pkt_data, (RTE_MIN((n_bytes - cb_len),
+			   mbuf->data_len)));
+
+		pcap_trace.n_bytes += n_bytes;
+		pcap_trace.n_packets_captured++;
+
+		if (pcap_trace.n_packets_captured == MAX_PKT_TO_CAPTURE)
+			pcap_trace_write();
+
+		rte_spinlock_unlock(&pcap_trace.lock);
+	}
+
+	return node->p_process(graph, node, objs, nb_objs);
+}
diff --git a/lib/graph/graph_populate.c b/lib/graph/graph_populate.c
index 102fd6c29b..36f81505df 100644
--- a/lib/graph/graph_populate.c
+++ b/lib/graph/graph_populate.c
@@ -75,7 +75,11 @@  graph_nodes_populate(struct graph *_graph)
 		memset(node, 0, sizeof(*node));
 		node->fence = RTE_GRAPH_FENCE;
 		node->off = off;
-		node->process = graph_node->node->process;
+		if (is_pcap_trace_enable()) {
+			node->process = rte_graph_pcap_trace_dispatch;
+			node->p_process = graph_node->node->process;
+		} else
+			node->process = graph_node->node->process;
 		memcpy(node->name, graph_node->node->name, RTE_GRAPH_NAMESIZE);
 		pid = graph_node->node->parent_id;
 		if (pid != RTE_NODE_ID_INVALID) { /* Cloned node */
diff --git a/lib/graph/meson.build b/lib/graph/meson.build
index c7327549e8..6cd9d836d9 100644
--- a/lib/graph/meson.build
+++ b/lib/graph/meson.build
@@ -14,7 +14,8 @@  sources = files(
         'graph_debug.c',
         'graph_stats.c',
         'graph_populate.c',
+        'graph_pcap_trace.c',
 )
-headers = files('rte_graph.h', 'rte_graph_worker.h')
+headers = files('rte_graph.h', 'rte_graph_worker.h', 'rte_graph_pcap_trace.h')
 
-deps += ['eal']
+deps += ['eal', 'mbuf', 'mempool', 'net', 'ethdev']
diff --git a/lib/graph/rte_graph_pcap_trace.h b/lib/graph/rte_graph_pcap_trace.h
new file mode 100644
index 0000000000..e2faf6205f
--- /dev/null
+++ b/lib/graph/rte_graph_pcap_trace.h
@@ -0,0 +1,149 @@ 
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(C) 2022 Marvell International Ltd.
+ */
+
+#ifndef _DPDK_GRAPH_PCAP_TRACE_H_
+#define _DPDK_GRAPH_PCAP_TRACE_H_
+
+/**
+ * @file rte_graph_pcap_trace.h
+ *
+ * @warning
+ * @b EXPERIMENTAL:
+ * All functions in this file may be changed or removed without prior notice.
+ *
+ * This API enables to capture packet at each node with mbuf and node metadata.
+ *
+ */
+
+#include <rte_spinlock.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * User packet type used in pcap file is used by wireshark to lookup into the
+ * DLT table to know which protocol(s) to use for each DLT.
+ */
+typedef enum {
+	PCAP_PACKET_TYPE_USER0 = 147,
+	PCAP_PACKET_TYPE_USER1,
+	PCAP_PACKET_TYPE_USER2,
+	PCAP_PACKET_TYPE_USER3,
+	PCAP_PACKET_TYPE_USER4,
+	PCAP_PACKET_TYPE_USER5,
+	PCAP_PACKET_TYPE_USER6,
+	PCAP_PACKET_TYPE_USER7,
+	PCAP_PACKET_TYPE_USER8,
+	PCAP_PACKET_TYPE_USER9,
+	PCAP_PACKET_TYPE_USER10,
+	PCAP_PACKET_TYPE_USER11,
+	PCAP_PACKET_TYPE_USER12,
+	PCAP_PACKET_TYPE_USER13,
+	PCAP_PACKET_TYPE_USER14,
+	PCAP_PACKET_TYPE_USER15,
+} pcap_packet_type_t;
+
+/**
+ * Pcap file shall contain a header at the start of the file. Parameters are
+ * dissected by the wireshark to display the data.
+ */
+typedef struct pcap_file_header {
+	uint32_t magic;
+	uint16_t major_version;
+	uint16_t minor_version;
+	uint32_t time_zone;
+	uint32_t sigfigs;
+	uint32_t max_packet_size_in_bytes;
+	uint32_t packet_type;
+} pcap_file_header_t;
+
+/**
+ * Each packet shall be prepended by the packet header.
+ */
+typedef struct pcap_packet_header_t {
+	uint32_t time_in_sec;
+	uint32_t time_in_usec;
+	uint32_t n_packet_bytes_stored_in_file;
+	uint32_t n_bytes_in_packet;
+	/** Packet data follows. */
+	uint8_t data[0];
+} pcap_packet_header_t;
+
+/**
+ * Pcap version header.
+ */
+typedef struct pcap_version_hdr {
+	uint8_t pcap_major_version;
+	uint8_t pcap_minor_version;
+} pcap_version_hdr_t;
+
+/**
+ * Book-keeping of the packets captured.
+ */
+typedef struct pcap_trace {
+	rte_spinlock_t lock;
+	const char *file_name;
+	uint32_t n_packets_to_capture;
+	uint32_t n_bytes;
+	pcap_packet_type_t packet_type;
+	uint32_t n_packets_captured;
+	int file_descriptor;
+	uint32_t n_pcap_data_written;
+	uint8_t *pcap_data;
+} pcap_trace_t;
+
+/**
+ * Pcap trace enable/disable function.
+ *
+ * The function is called to enable/disable graph pcap trace functionality.
+ *
+ * @param val
+ *   Value to be set to enable/disable graph pcap trace.
+ */
+void set_pcap_trace(int val);
+
+/**
+ * Check graph pcap trace is enable/disable.
+ *
+ * The function is called to check if the graph pcap trace is enabled/disabled.
+ *
+ * @return
+ *   - 1: Enable
+ *   - 0: Disable
+ */
+int is_pcap_trace_enable(void);
+
+/**
+ * Initialise graph pcap trace functionality.
+ *
+ * The function invoked when the graph pcap trace is enabled from the
+ * application.
+ *
+ */
+void rte_graph_pcap_trace_init(void);
+
+/**
+ * Capture mbuf metadata and node metadata to a pcap file.
+ *
+ * When graph pcap trace enabled, this function is invoked prior to each node
+ * and mbuf, node metadata is parsed and captured in a pcap file.
+ *
+ * @param graph
+ *   Pointer to the graph object.
+ * @param node
+ *   Pointer to the node object.
+ * @param objs
+ *   Pointer to an array of objects to be processed.
+ * @param nb_objs
+ *   Number of objects in the array.
+ */
+uint16_t rte_graph_pcap_trace_dispatch(struct rte_graph *graph __rte_unused,
+				       struct rte_node *node, void **objs,
+				       uint16_t nb_objs);
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _DPDK_GRAPH_PCAP_TRACE_H_ */
diff --git a/lib/graph/rte_graph_worker.h b/lib/graph/rte_graph_worker.h
index fc6fee48c8..64c18421d5 100644
--- a/lib/graph/rte_graph_worker.h
+++ b/lib/graph/rte_graph_worker.h
@@ -24,6 +24,7 @@ 
 #include <rte_memory.h>
 
 #include "rte_graph.h"
+#include "rte_graph_pcap_trace.h"
 
 #ifdef __cplusplus
 extern "C" {
@@ -64,6 +65,11 @@  struct rte_node {
 	char parent[RTE_NODE_NAMESIZE];	/**< Parent node name. */
 	char name[RTE_NODE_NAMESIZE];	/**< Name of the node. */
 
+	union {
+		rte_node_process_t p_process; /**< Process function. */
+		uint64_t p_process_u64;
+	};
+
 	/* Fast path area  */
 #define RTE_NODE_CTX_SZ 16
 	uint8_t ctx[RTE_NODE_CTX_SZ] __rte_cache_aligned; /**< Node Context. */