DPDK patches and discussions
 help / color / mirror / Atom feed
From: Ferruh Yigit <ferruh.yigit@xilinx.com>
To: Tianli Lai <laitianli@tom.com>, <dev@dpdk.org>
Cc: Bruce Richardson <bruce.richardson@intel.com>
Subject: Re: [PATCH] net/pcap: add snaplen argument
Date: Wed, 20 Apr 2022 18:14:18 +0100	[thread overview]
Message-ID: <ebb8b401-bdd1-92b1-f4b7-25a4da9c9648@xilinx.com> (raw)
In-Reply-To: <20220313112638.3945-1-laitianli@tom.com>

On 3/13/2022 11:26 AM, Tianli Lai wrote:
> snaplen argument would set the length of each packet
> that will save to pcap file.
> 
> Signed-off-by: Tianli Lai <laitianli@tom.com>
> ---
>   doc/guides/nics/pcap_ring.rst  | 16 +++++++++-
>   drivers/net/pcap/pcap_ethdev.c | 58 +++++++++++++++++++++++++---------
>   2 files changed, 58 insertions(+), 16 deletions(-)
> 
> diff --git a/doc/guides/nics/pcap_ring.rst b/doc/guides/nics/pcap_ring.rst
> index acb1f00e30..8fb309212f 100644
> --- a/doc/guides/nics/pcap_ring.rst
> +++ b/doc/guides/nics/pcap_ring.rst
> @@ -94,6 +94,13 @@ The different stream types are:
>   
>           iface=eth0
>   
> +*   snaplen: Defines the length of one packet that will save pcap file.
> +    the length of each packets can be snap by set this value.
> +    this value between 64 and 65535.
> +
> +        snaplen=<packet len>
> +
> +

Hi Tianli,

This section is "Device Streams", it documents main Rx/Tx streams,
'snaplen' can be a detail for this section, what do you think
to drop it form here, it is already documented below?

>   Runtime Config Options
>   ^^^^^^^^^^^^^^^^^^^^^^
>   
> @@ -132,6 +139,13 @@ Runtime Config Options
>   
>   In this case, one dummy rx queue is created for each tx queue argument passed
>   
> + - Receive packets on Tx and set the length of packet, for example::
> +

What about focusing on the argument description, something like:

- `snaplen`: Set the length of the saved packet size, for example::

> +    --vdev 'net_pcap0,tx_pcap=file_tx.pcap,snaplen=100'
> +
> +In this case, one dummy rx queue is created for each tx queue argument passed and the length of each packet would not over 100 byte.
> +

Don't focus on other parameters (like dummy Rx), please only explain 
what 'snaplen' does.

> +
>   Examples of Usage
>   ^^^^^^^^^^^^^^^^^
>   
> @@ -140,7 +154,7 @@ Read packets from one pcap file and write them to another:
>   .. code-block:: console
>   
>       ./<build_dir>/app/dpdk-testpmd -l 0-3 -n 4 \
> -        --vdev 'net_pcap0,rx_pcap=file_rx.pcap,tx_pcap=file_tx.pcap' \
> +        --vdev 'net_pcap0,rx_pcap=file_rx.pcap,tx_pcap=file_tx.pcap,snaplen=100' \
>           -- --port-topology=chained
>   

Can you add a new sample with 'snaplen', instead of attaching to 
existing one?

>   Read packets from a network interface and write them to a pcap file:
> diff --git a/drivers/net/pcap/pcap_ethdev.c b/drivers/net/pcap/pcap_ethdev.c
> index ec29fd6bc5..6e111518e7 100644
> --- a/drivers/net/pcap/pcap_ethdev.c
> +++ b/drivers/net/pcap/pcap_ethdev.c
> @@ -33,10 +33,12 @@
>   #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
>   
>   #define RTE_PMD_PCAP_MAX_QUEUES 16
> +#define RTE_PCAP_MIN_SNAPLEN    64
>   
>   static char errbuf[PCAP_ERRBUF_SIZE];
>   static struct timespec start_time;
> @@ -93,6 +95,7 @@ struct pmd_internals {
>   	int single_iface;
>   	int phy_mac;
>   	unsigned int infinite_rx;
> +	int snaplen;
>   };
>   
>   struct pmd_process_private {
> @@ -121,6 +124,14 @@ struct pmd_devargs_all {
>   	unsigned int is_rx_pcap;
>   	unsigned int is_rx_iface;
>   	unsigned int infinite_rx;
> +	int snaplen;
> +};
> +struct pmd_devargs_all devargs_all = {
> +	.single_iface = 0,
> +	.is_tx_pcap = 0,
> +	.is_tx_iface = 0,
> +	.infinite_rx = 0,
> +	.snaplen = RTE_ETH_PCAP_SNAPSHOT_LEN,
>   };
>

'devargs_all' is to store user provided devargs per port,
making it a global variable is wrong, it causes problems where there are 
multiple ports, please leave it where it is.


>   static const char *valid_arguments[] = {
> @@ -132,6 +143,7 @@ static const char *valid_arguments[] = {
>   	ETH_PCAP_IFACE_ARG,
>   	ETH_PCAP_PHY_MAC_ARG,
>   	ETH_PCAP_INFINITE_RX_ARG,
> +	ETH_PCAP_SNAPLEN_ARG,
>   	NULL
>   };
>   
> @@ -384,8 +396,10 @@ eth_pcap_tx_dumper(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
>   	pcap_dumper_t *dumper;
>   	unsigned char temp_data[RTE_ETH_PCAP_SNAPLEN];
>   	size_t len, caplen;
> -
> -	pp = rte_eth_devices[dumper_q->port_id].process_private;
> +	struct pmd_internals *internals;
> +	struct rte_eth_dev *dev = &rte_eth_devices[dumper_q->port_id];
> +	internals = dev->data->dev_private;
> +	pp = dev->process_private;
>   	dumper = pp->tx_dumper[dumper_q->queue_id];
>   
>   	if (dumper == NULL || nb_pkts == 0)
> @@ -402,6 +416,8 @@ eth_pcap_tx_dumper(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
>   		}
>   
>   		calculate_timestamp(&header.ts);
> +		if ((typeof(internals->snaplen))caplen > internals->snaplen)
> +			caplen = internals->snaplen;

When opening pcap interface 'snaplen' is already provided,
what happens if 'caplen' is not set correctly here, won't pcap
library limit the packets to 'snaplen' already?

>   		header.len = len;
>   		header.caplen = caplen;
>   		/* rte_pktmbuf_read() returns a pointer to the data directly
> @@ -536,7 +552,7 @@ open_single_iface(const char *iface, pcap_t **pcap)
>   }
>   
>   static int
> -open_single_tx_pcap(const char *pcap_filename, pcap_dumper_t **dumper)
> +open_single_tx_pcap(const char *pcap_filename, pcap_dumper_t **dumper, int snaplen)
>   {
>   	pcap_t *tx_pcap;
>   
> @@ -546,7 +562,7 @@ open_single_tx_pcap(const char *pcap_filename, pcap_dumper_t **dumper)
>   	 * pcap holder.
>   	 */
>   	tx_pcap = pcap_open_dead_with_tstamp_precision(DLT_EN10MB,
> -			RTE_ETH_PCAP_SNAPSHOT_LEN, PCAP_TSTAMP_PRECISION_NANO);
> +			snaplen, PCAP_TSTAMP_PRECISION_NANO);
>   	if (tx_pcap == NULL) {
>   		PMD_LOG(ERR, "Couldn't create dead pcap");
>   		return -1;
> @@ -627,7 +643,7 @@ eth_dev_start(struct rte_eth_dev *dev)
>   		if (!pp->tx_dumper[i] &&
>   				strcmp(tx->type, ETH_PCAP_TX_PCAP_ARG) == 0) {
>   			if (open_single_tx_pcap(tx->name,
> -				&pp->tx_dumper[i]) < 0)
> +				&pp->tx_dumper[i], internals->snaplen) < 0)
>   				return -1;
>   		} else if (!pp->tx_pcap[i] &&
>   				strcmp(tx->type, ETH_PCAP_TX_IFACE_ARG) == 0) {
> @@ -1055,7 +1071,7 @@ open_tx_pcap(const char *key, const char *value, void *extra_args)
>   	struct pmd_devargs *dumpers = extra_args;
>   	pcap_dumper_t *dumper;
>   
> -	if (open_single_tx_pcap(pcap_filename, &dumper) < 0)
> +	if (open_single_tx_pcap(pcap_filename, &dumper, devargs_all.snaplen) < 0)

Here 'devargs_all' is used directly since patch makes it global, but 
when it is not global anymore, perhaps you can consider to add 'snaplen'
to "struct pmd_devargs" and get it to this function via it.

>   		return -1;
>   
>   	if (add_queue(dumpers, pcap_filename, key, NULL, dumper) < 0) {
> @@ -1066,6 +1082,15 @@ open_tx_pcap(const char *key, const char *value, void *extra_args)
>   	return 0;
>   }
>   
> +static int
> +parse_uint_value(const char *key __rte_unused, const char *value, void *extra_args)
> +{
> +	char *end;
> +	int *val = extra_args;
> +	*val = strtoul(value, &end, 10);

Should check 'strtoul' return value and return error accordingly.

> +	return 0;
> +}
> +
>   /*
>    * Opens an interface for reading and writing
>    */
> @@ -1325,10 +1350,10 @@ eth_from_pcaps(struct rte_vdev_device *vdev,
>   	int ret;
>   
>   	ret = eth_from_pcaps_common(vdev, devargs_all, &internals, &eth_dev);
> -

unrelated change

>   	if (ret < 0)
>   		return ret;
>   
> +	internals->snaplen = devargs_all->snaplen; >   	/* store weather we are using a single interface for rx/tx or not */
>   	internals->single_iface = single_iface;
>   
> @@ -1404,13 +1429,6 @@ pmd_pcap_probe(struct rte_vdev_device *dev)
>   	struct pmd_internals *internal;
>   	int ret = 0;
>   
> -	struct pmd_devargs_all devargs_all = {
> -		.single_iface = 0,
> -		.is_tx_pcap = 0,
> -		.is_tx_iface = 0,
> -		.infinite_rx = 0,
> -	};
> -
>   	name = rte_vdev_device_name(dev);
>   	PMD_LOG(INFO, "Initializing pmd_pcap for %s", name);
>   
> @@ -1444,6 +1462,16 @@ pmd_pcap_probe(struct rte_vdev_device *dev)
>   			return -1;
>   	}
>   
> +	if (rte_kvargs_count(kvlist, ETH_PCAP_SNAPLEN_ARG) == 1) {
> +		ret = rte_kvargs_process(kvlist, ETH_PCAP_SNAPLEN_ARG,
> +				&parse_uint_value, &devargs_all.snaplen);
> +		if (ret < 0)
> +			goto free_kvlist;

Can you print an error log here?

> +		if (devargs_all.snaplen < RTE_PCAP_MIN_SNAPLEN ||
> +			devargs_all.snaplen >= RTE_ETH_PCAP_SNAPSHOT_LEN)
> +			devargs_all.snaplen = RTE_PCAP_MIN_SNAPLEN;

I think better to return an error here, with a log.

> +	}

Can you please move parsing 'ETH_PCAP_SNAPLEN_ARG' argument below in the 
function, where Tx related parameters parsed. After 
'devargs_all.is_tx_pcap' checks.

And can you please define (and document in above documentation), when 
'snaplen' argument is valid?
Like can I use it with only 'iface=' argument? Or does it requires 
'tx_pcap' to work? If so this should be checked in the code.

> +
>   	/*
>   	 * If iface argument is passed we open the NICs and use them for
>   	 * reading / writing
> @@ -1593,7 +1621,7 @@ pmd_pcap_probe(struct rte_vdev_device *dev)
>   			pp->tx_dumper[i] = dumpers.queue[i].dumper;
>   			pp->tx_pcap[i] = dumpers.queue[i].pcap;
>   		}
> -
> +		internal->snaplen = devargs_all.snaplen;

This is not required, 'internals' is shared between primary and 
secondary, if primary has this parameter, secondary can use it
already.

>   		eth_dev->process_private = pp;
>   		eth_dev->rx_pkt_burst = eth_pcap_rx;
>   		if (devargs_all.is_tx_pcap)


  reply	other threads:[~2022-04-20 17:14 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-03-13 11:26 Tianli Lai
2022-04-20 17:14 ` Ferruh Yigit [this message]
  -- strict thread matches above, loose matches on Subject: below --
2022-03-13  9:26 Tianli Lai

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=ebb8b401-bdd1-92b1-f4b7-25a4da9c9648@xilinx.com \
    --to=ferruh.yigit@xilinx.com \
    --cc=bruce.richardson@intel.com \
    --cc=dev@dpdk.org \
    --cc=laitianli@tom.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).