Re: [dpdk-dev] [PATCH v2] net/pcap: support snaplen option to truncate packet

[Ferruh Yigit <ferruh.yigit@intel.com>]

On 7/3/2020 7:47 AM, Zhike Wang wrote:
> Introduced "snaplen=<length>" option. It is convenient to truncate
> large packets to only capture necessary headers.
> 
> Signed-off-by: Zhike Wang <wangzhike@jd.com>
> ---
>  app/pdump/main.c                | 32 +++++++++++++++++++++++++++++++-
>  drivers/net/pcap/rte_eth_pcap.c | 35 ++++++++++++++++++++++++++++++++++-
>  2 files changed, 65 insertions(+), 2 deletions(-)
> 
> diff --git a/app/pdump/main.c b/app/pdump/main.c
> index c38c537..241346b 100644
> --- a/app/pdump/main.c
> +++ b/app/pdump/main.c
> @@ -41,10 +41,12 @@
>  #define PDUMP_RING_SIZE_ARG "ring-size"
>  #define PDUMP_MSIZE_ARG "mbuf-size"
>  #define PDUMP_NUM_MBUFS_ARG "total-num-mbufs"
> +#define PDUMP_SNAPLEN_ARG "snaplen"
>  
>  #define VDEV_NAME_FMT "net_pcap_%s_%d"
>  #define VDEV_PCAP_ARGS_FMT "tx_pcap=%s"
>  #define VDEV_IFACE_ARGS_FMT "tx_iface=%s"
> +#define VDEV_SNAPLEN_ARGS_FMT "snaplen=%d"
>  #define TX_STREAM_SIZE 64
>  
>  #define MP_NAME "pdump_pool_%d"
> @@ -97,6 +99,7 @@ enum pdump_by {
>  	PDUMP_RING_SIZE_ARG,
>  	PDUMP_MSIZE_ARG,
>  	PDUMP_NUM_MBUFS_ARG,
> +	PDUMP_SNAPLEN_ARG,
>  	NULL
>  };
>  
> @@ -116,6 +119,7 @@ struct pdump_tuples {
>  	uint32_t ring_size;
>  	uint16_t mbuf_data_size;
>  	uint32_t total_num_mbufs;
> +	uint16_t snaplen;
>  
>  	/* params for library API call */
>  	uint32_t dir;
> @@ -160,7 +164,8 @@ struct parse_val {
>  			" tx-dev=<iface or pcap file>,"
>  			"[ring-size=<ring size>default:16384],"
>  			"[mbuf-size=<mbuf data size>default:2176],"
> -			"[total-num-mbufs=<number of mbufs>default:65535]'\n",
> +			"[total-num-mbufs=<number of mbufs>default:65535],",
> +			"[snaplen=<snap length>default:0, meaning no truncation]'\n",
>  			prgname);
>  }
>  
> @@ -370,6 +375,19 @@ struct parse_val {
>  	} else
>  		pt->total_num_mbufs = MBUFS_PER_POOL;
>  
> +	/* snaplen parsing and validation */
> +	cnt1 = rte_kvargs_count(kvlist, PDUMP_SNAPLEN_ARG);
> +	if (cnt1 == 1) {
> +		v.min = 1;
> +		v.max = UINT16_MAX;
> +		ret = rte_kvargs_process(kvlist, PDUMP_SNAPLEN_ARG,
> +						&parse_uint_value, &v);
> +		if (ret < 0)
> +			goto free_kvlist;
> +		pt->snaplen = (uint16_t) v.val;
> +	} else
> +		pt->snaplen = 0;
> +
>  	num_tuples++;
>  
>  free_kvlist:
> @@ -692,6 +710,9 @@ struct parse_val {
>  				 VDEV_IFACE_ARGS_FMT, pt->rx_dev) :
>  			snprintf(vdev_args, sizeof(vdev_args),
>  				 VDEV_PCAP_ARGS_FMT, pt->rx_dev);
> +			snprintf(vdev_args + strlen(vdev_args),
> +				 sizeof(vdev_args) - strlen(vdev_args),
> +				 ","VDEV_SNAPLEN_ARGS_FMT, pt->snaplen);
>  			if (rte_eal_hotplug_add("vdev", vdev_name,
>  						vdev_args) < 0) {
>  				cleanup_rings();
> @@ -722,6 +743,9 @@ struct parse_val {
>  					 VDEV_IFACE_ARGS_FMT, pt->tx_dev) :
>  				snprintf(vdev_args, sizeof(vdev_args),
>  					 VDEV_PCAP_ARGS_FMT, pt->tx_dev);
> +				snprintf(vdev_args + strlen(vdev_args),
> +					 sizeof(vdev_args) - strlen(vdev_args),
> +					 ","VDEV_SNAPLEN_ARGS_FMT, pt->snaplen);
>  				if (rte_eal_hotplug_add("vdev", vdev_name,
>  							vdev_args) < 0) {
>  					cleanup_rings();
> @@ -762,6 +786,9 @@ struct parse_val {
>  				 VDEV_IFACE_ARGS_FMT, pt->rx_dev) :
>  			snprintf(vdev_args, sizeof(vdev_args),
>  				 VDEV_PCAP_ARGS_FMT, pt->rx_dev);
> +			snprintf(vdev_args + strlen(vdev_args),
> +				 sizeof(vdev_args) - strlen(vdev_args),
> +				 ","VDEV_SNAPLEN_ARGS_FMT, pt->snaplen);
>  			if (rte_eal_hotplug_add("vdev", vdev_name,
>  						vdev_args) < 0) {
>  				cleanup_rings();
> @@ -799,6 +826,9 @@ struct parse_val {
>  				 VDEV_IFACE_ARGS_FMT, pt->tx_dev) :
>  			snprintf(vdev_args, sizeof(vdev_args),
>  				 VDEV_PCAP_ARGS_FMT, pt->tx_dev);
> +			snprintf(vdev_args + strlen(vdev_args),
> +				 sizeof(vdev_args) - strlen(vdev_args),
> +				 ","VDEV_SNAPLEN_ARGS_FMT, pt->snaplen);
>  			if (rte_eal_hotplug_add("vdev", vdev_name,
>  						vdev_args) < 0) {
>  				cleanup_rings();
> diff --git a/drivers/net/pcap/rte_eth_pcap.c b/drivers/net/pcap/rte_eth_pcap.c
> index 13a3d0a..d51ea48 100644
> --- a/drivers/net/pcap/rte_eth_pcap.c
> +++ b/drivers/net/pcap/rte_eth_pcap.c
> @@ -40,6 +40,7 @@
>  #define ETH_PCAP_IFACE_ARG    "iface"
>  #define ETH_PCAP_PHY_MAC_ARG  "phy_mac"
>  #define ETH_PCAP_INFINITE_RX_ARG  "infinite_rx"
> +#define ETH_PCAP_SNAPLEN_ARG  "snaplen"
>  
>  #define ETH_PCAP_ARG_MAXLEN	64
>  
> @@ -92,6 +93,7 @@ struct pmd_process_private {
>  	pcap_t *rx_pcap[RTE_PMD_PCAP_MAX_QUEUES];
>  	pcap_t *tx_pcap[RTE_PMD_PCAP_MAX_QUEUES];
>  	pcap_dumper_t *tx_dumper[RTE_PMD_PCAP_MAX_QUEUES];
> +	int snaplen;

As it shouldn't be negative, choosing "unsigned int" or "size_t" type can be better.

Also is the variable added to 'process_private' memmory intentionally? Why not
in the shared device data ('pmd_internals') memory?

>  };
>  
>  struct pmd_devargs {
> @@ -114,6 +116,7 @@ struct pmd_devargs_all {
>  	unsigned int is_rx_pcap;
>  	unsigned int is_rx_iface;
>  	unsigned int infinite_rx;
> +	int snaplen;
>  };
>  
>  static const char *valid_arguments[] = {
> @@ -125,6 +128,7 @@ struct pmd_devargs_all {
>  	ETH_PCAP_IFACE_ARG,
>  	ETH_PCAP_PHY_MAC_ARG,
>  	ETH_PCAP_INFINITE_RX_ARG,
> +	ETH_PCAP_SNAPLEN_ARG,
>  	NULL
>  };
>  
> @@ -339,6 +343,9 @@ struct pmd_devargs_all {
>  			caplen = sizeof(temp_data);
>  		}
>  
> +		if ((pp->snaplen > 0) && (caplen > pp->snaplen))
> +			caplen = pp->snaplen;

This "is 'snaplen' set" check done per packet, and it will hit both 'snaplen'
set or not, I wonder if there is a way to get rid of per packet check?

>  		calculate_timestamp(&header.ts);
>  		header.len = len;
>  		header.caplen = caplen;
> @@ -949,6 +956,7 @@ struct pmd_devargs_all {
>  {
>  	const char *pcap_filename = value;
>  	struct pmd_devargs *dumpers = extra_args;
> +	int snaplen;

This is not used at all, I guess added by mistake.

>  	pcap_dumper_t *dumper;
>  
>  	if (open_single_tx_pcap(pcap_filename, &dumper) < 0)
> @@ -1083,6 +1091,19 @@ struct pmd_devargs_all {
>  }
>  
>  static int
> +get_snaplen_arg(const char *key __rte_unused,
> +		const char *value, void *extra_args)
> +{
> +	if (extra_args) {
> +		const int snaplen = atoi(value);

'const' qualifier is not needed.

And some error checks can be better.

> +		int *snaplen_p = extra_args;
> +
> +		*snaplen_p = snaplen;
> +	}
> +	return 0;
> +}
> +
> +static int
>  pmd_init_internals(struct rte_vdev_device *vdev,
>  		const unsigned int nb_rx_queues,
>  		const unsigned int nb_tx_queues,
> @@ -1335,6 +1356,7 @@ struct pmd_devargs_all {
>  	struct rte_eth_dev *eth_dev =  NULL;
>  	struct pmd_internals *internal;
>  	int ret = 0;
> +	unsigned int snaplen_cnt;
>  
>  	struct pmd_devargs_all devargs_all = {
>  		.single_iface = 0,
> @@ -1412,6 +1434,15 @@ struct pmd_devargs_all {
>  		rte_kvargs_count(kvlist, ETH_PCAP_TX_IFACE_ARG) ? 1 : 0;
>  	dumpers.num_of_queue = 0;
>  
> +	snaplen_cnt = rte_kvargs_count(kvlist,
> +				ETH_PCAP_SNAPLEN_ARG);
> +	if (snaplen_cnt == 1) {

No need to get the count of the devargs. Just calling 'rte_kvargs_process()'
should be enough.

> +		ret = rte_kvargs_process(kvlist, ETH_PCAP_SNAPLEN_ARG,
> +				&get_snaplen_arg, &devargs_all.snaplen);

'snaplen' make sense only for 'is_tx_pcap' case, so you can process the
'ETH_PCAP_SNAPLEN_ARG' arg only for 'is_tx_pcap' case.

> +		if (ret < 0)
> +			goto free_kvlist;
> +	}
> +
>  	if (devargs_all.is_rx_pcap) {
>  		/*
>  		 * We check whether we want to infinitely rx the pcap file.
> @@ -1518,6 +1549,7 @@ struct pmd_devargs_all {
>  			pp->tx_pcap[i] = dumpers.queue[i].pcap;
>  		}
>  
> +		pp->snaplen = devargs_all.snaplen;

This is inside the "rte_eal_process_type() == RTE_PROC_SECONDARY" block, why
'snaplen' only used for secondary process?

>  		eth_dev->process_private = pp;
>  		eth_dev->rx_pkt_burst = eth_pcap_rx;
>  		if (devargs_all.is_tx_pcap)
> @@ -1587,7 +1619,8 @@ struct pmd_devargs_all {
>  	ETH_PCAP_TX_IFACE_ARG "=<ifc> "
>  	ETH_PCAP_IFACE_ARG "=<ifc> "
>  	ETH_PCAP_PHY_MAC_ARG "=<int>"
> -	ETH_PCAP_INFINITE_RX_ARG "=<0|1>");
> +	ETH_PCAP_INFINITE_RX_ARG "=<0|1>"
> +	ETH_PCAP_SNAPLEN_ARG "=<int>");
>  
>  RTE_INIT(eth_pcap_init_log)
>  {
>