[dpdk-dev] [PATCH] examples/ip_fragmentation: add fragmentation size support

[Jain Ashish-B46179]

Hi Konstantin,

On 6/4/2017 9:14 PM, Ananyev, Konstantin wrote:
> Hi,
> 
>> -----Original Message-----
>> From: Ashish Jain [mailto:ashish.jain at nxp.com]
>> Sent: Thursday, April 20, 2017 8:18 AM
>> To: dev at dpdk.org; Ananyev, Konstantin <konstantin.ananyev at intel.com>
>> Subject: [PATCH] examples/ip_fragmentation: add fragmentation size support
>>
>> Adding support for determining fragmentation size for both
>> ipv4 and ipv6 traffic dynamically through command line.
>> It is helpful in testing to configure different fragmentation
>> sizes and validate the packets.
>>
>> Signed-off-by: Ashish Jain <ashish.jain at nxp.com>
>> Signed-off-by: Hemant Agrawal <hemant.agrawal at nxp.com>
>> ---
>>   examples/ip_fragmentation/main.c | 89 ++++++++++++++++++++++++++++++++++++----
>>   1 file changed, 81 insertions(+), 8 deletions(-)
>>
>> diff --git a/examples/ip_fragmentation/main.c b/examples/ip_fragmentation/main.c
>> index 815b225..436755b 100644
>> --- a/examples/ip_fragmentation/main.c
>> +++ b/examples/ip_fragmentation/main.c
>> @@ -94,6 +94,16 @@
>>   #define	IPV6_DEFAULT_PAYLOAD	(IPV6_MTU_DEFAULT - sizeof(struct ipv6_hdr))
>>
>>   /*
>> + * Configure fragmentation size for IPv4 and IPv6 packets
>> + */
>> +static uint32_t frag_size_v4 = IPV4_MTU_DEFAULT;
>> +static uint32_t frag_size_v6 = IPV6_MTU_DEFAULT;
>> +#define MIN_IPV4_FRAG_SIZE 64
>> +#define MAX_IPV4_FRAG_SIZE 9600
>> +#define MIN_IPV6_FRAG_SIZE 1280
>> +#define MAX_IPV6_FRAG_SIZE 9600
>> +
>> +/*
>>    * Max number of fragments per packet expected - defined by config file.
>>    */
>>   #define	MAX_PACKET_FRAG RTE_LIBRTE_IP_FRAG_MAX_FRAG
>> @@ -299,14 +309,14 @@ struct rte_lpm6_config lpm6_config = {
>>   		}
>>
>>   		/* if we don't need to do any fragmentation */
>> -		if (likely (IPV4_MTU_DEFAULT >= m->pkt_len)) {
>> +		if (likely (frag_size_v4 >= m->pkt_len)) {
>>   			qconf->tx_mbufs[port_out].m_table[len] = m;
>>   			len2 = 1;
>>   		} else {
>>   			len2 = rte_ipv4_fragment_packet(m,
>>   				&qconf->tx_mbufs[port_out].m_table[len],
>>   				(uint16_t)(MBUF_TABLE_SIZE - len),
>> -				IPV4_MTU_DEFAULT,
>> +				frag_size_v4,
>>   				rxq->direct_pool, rxq->indirect_pool);
>>
>>   			/* Free input packet */
>> @@ -336,14 +346,14 @@ struct rte_lpm6_config lpm6_config = {
>>   		}
>>
>>   		/* if we don't need to do any fragmentation */
>> -		if (likely (IPV6_MTU_DEFAULT >= m->pkt_len)) {
>> +		if (likely (frag_size_v6 >= m->pkt_len)) {
>>   			qconf->tx_mbufs[port_out].m_table[len] = m;
>>   			len2 = 1;
>>   		} else {
>>   			len2 = rte_ipv6_fragment_packet(m,
>>   				&qconf->tx_mbufs[port_out].m_table[len],
>>   				(uint16_t)(MBUF_TABLE_SIZE - len),
>> -				IPV6_MTU_DEFAULT,
>> +				frag_size_v6,
>>   				rxq->direct_pool, rxq->indirect_pool);
>>
>>   			/* Free input packet */
>> @@ -489,8 +499,14 @@ struct rte_lpm6_config lpm6_config = {
>>   {
>>   	printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
>>   	       "  -p PORTMASK: hexadecimal bitmask of ports to configure\n"
>> -	       "  -q NQ: number of queue (=ports) per lcore (default is 1)\n",
>> -	       prgname);
>> +	       "  -q NQ: number of queue (=ports) per lcore (default is 1)\n"
>> +	       "  --frag_size_v4=<num>:optional,IPv4 fragment size in decimal"
>> +	       ",Condition:(frag_size_v4 - 20) should be a multiple of 8,"
>> +	       " default is %d \n"
>> +	       "  --frag_size_v6=<num>:optional,IPv6 fragment size in decimal"
>> +	       ",Condition:(frag_size_v6 - 40) should be a multiple of 8,"
>> +	       " default is %d\n",
>> +	       prgname, frag_size_v4, frag_size_v6);
>>   }
>>
>>   static int
>> @@ -528,6 +544,29 @@ struct rte_lpm6_config lpm6_config = {
>>   	return n;
>>   }
>>
>> +static int
>> +parse_frag_size(const char *str, uint32_t min, uint32_t max,
>> +		uint8_t hdr_size, uint32_t *val)
>> +{
>> +	char *end;
>> +	uint64_t v;
>> +
>> +	/* parse decimal string */
>> +	errno = 0;
>> +	v = strtoul(str, &end, 10);
>> +	if (errno != 0 || *end != '\0')
>> +		return -EINVAL;
>> +
>> +	if (v < min || v > max)
>> +		return -EINVAL;
>> +
>> +	if ((v - hdr_size) % 8)
>> +		return -EINVAL;
>> +
>> +	*val = (uint32_t)v;
>> +	return 0;
>> +}
>> +
>>   /* Parse the argument given in the command line of the application */
>>   static int
>>   parse_args(int argc, char **argv)
>> @@ -537,6 +576,8 @@ struct rte_lpm6_config lpm6_config = {
>>   	int option_index;
>>   	char *prgname = argv[0];
>>   	static struct option lgopts[] = {
>> +		{"frag_size_v4", 1, 0, 0},
>> +		{"frag_size_v6", 1, 0, 0},
>>   		{NULL, 0, 0, 0}
>>   	};
>>
>> @@ -568,8 +609,40 @@ struct rte_lpm6_config lpm6_config = {
>>
>>   		/* long options */
>>   		case 0:
>> -			print_usage(prgname);
>> -			return -1;
>> +			if (!strncmp(lgopts[option_index].name,
>> +				"frag_size_v4", 12)) {
> 
> 
> Why just not strcmp() here?
strncmp() here is adhering to its similar use in other applications like 
ip_reassemble, l3fwd etc.
> 
>> +				ret = parse_frag_size(optarg,
>> +						MIN_IPV4_FRAG_SIZE,
>> +						MAX_IPV4_FRAG_SIZE,
>> +						sizeof(struct ipv4_hdr),
>> +						&frag_size_v4);
>> +				if (ret) {
>> +					printf("invalid value: \"%s\" for "
>> +						"parameter %s\n",
>> +						optarg,
>> +						lgopts[option_index].name);
>> +					print_usage(prgname);
>> +					return ret;
>> +				}
>> +			}
>> +			if (!strncmp(lgopts[option_index].name,
>> +				"frag_size_v6", 12)) {
>> +				ret = parse_frag_size(optarg,
>> +						MIN_IPV6_FRAG_SIZE,
>> +						MAX_IPV6_FRAG_SIZE,
>> +						sizeof(struct ipv6_hdr),
>> +						&frag_size_v6);
>> +				if (ret) {
>> +					printf("invalid value: \"%s\" for "
>> +						"parameter %s\n",
>> +						optarg,
>> +						lgopts[option_index].name);
>> +					print_usage(prgname);
>> +					return ret;
>> +				}
>> +			}
>> +
>> +			break;
> 
> Hmm, why do we need that break here?
> Konstantin
'break' is needed here to exit from the switch case

Ashish
> 
> 
>>
>>   		default:
>>   			print_usage(prgname);
>> --
>> 1.9.1
> 
>