[LTP] [RFC PATCH v7 02/11] network/stress: Add library test_net_stress.sh

[Alexey Kodanev]

On 07/25/2017 05:52 PM, Petr Vorel wrote:
>> On 07/25/2017 12:57 PM, Petr Vorel wrote:
>>> Hi Alexey,
>>>>> +export OCTET_3_IPV4_UNUSED=${OCTET_3_IPV4_UNUSED:-23} # x.x.N.x
>>>>> +export OCTET_4_LHOST_IPV4_HOST_UNUSED=${OCTET_4_LHOST_IPV4_HOST_UNUSED:-2} # x.x.x.N
>>>>> +export OCTET_4_RHOST_IPV4_HOST_UNUSED=${OCTET_4_RHOST_IPV4_HOST_UNUSED:-1} # x.x.x.N
>>>>> +export LHOST_IPV4_HOST_UNUSED="${OCTET_3_IPV4_UNUSED}.${OCTET_4_LHOST_IPV4_HOST_UNUSED}"
>>>>> +export RHOST_IPV4_HOST_UNUSED="${OCTET_3_IPV4_UNUSED}.${OCTET_4_RHOST_IPV4_HOST_UNUSED}"
>>>>> +export LHOST_IPV4_UNUSED="${IPV4_NET16_UNUSED}.$LHOST_IPV4_HOST_UNUSED"
>>>>> +export RHOST_IPV4_UNUSED="${IPV4_NET16_UNUSED}.$RHOST_IPV4_HOST_UNUSED"
>>>>> +
>>>>> +export OCTET_3_IPV6_UNUSED="${OCTET_3_IPV6_UNUSED:-:0}"
>>>>> +export OCTET_4_LHOST_IPV6_HOST_UNUSED=${OCTET_4_LHOST_IPV6_HOST_UNUSED:-2}
>>>>> +export OCTET_4_RHOST_IPV6_HOST_UNUSED=${OCTET_4_RHOST_IPV6_HOST_UNUSED:-1}
>>>>> +export LHOST_IPV6_HOST_UNUSED="${OCTET_3_IPV6_UNUSED}:${OCTET_4_LHOST_IPV6_HOST_UNUSED}"
>>>>> +export RHOST_IPV6_HOST_UNUSED="${OCTET_3_IPV6_UNUSED}:${OCTET_4_RHOST_IPV6_HOST_UNUSED}"
>>>>> +export LHOST_IPV6_UNUSED="${IPV6_NET32_UNUSED}:$LHOST_IPV6_HOST_UNUSED"
>>>>> +export RHOST_IPV6_UNUSED="${IPV6_NET32_UNUSED}:$RHOST_IPV6_HOST_UNUSED"
>>>> Hmm, what is the purpose of adding these variables to the library?
>>>> Why someone would need to redefine any of them?
>>> I thought someone might appreciate ability to full control of IP
>>> addresses, but you're right it's probably useless.
>>> If you don't mind, I'll keep {L,R}HOST_IPV{4,6}_UNUSED variables as they are used in many scripts
>>> (DRY approach) and delete the rest of variables.
>
>> What about adding a new function similar to tst_ipaddr()?
>> tst_ipaddr_un(), tst_ipaddr_new(), etc.? Probably we would need to add
>> some optional counter parameter along with lhost/rhost...
> How about this?
>
> tst_ipaddr_un()
> {
> 	local type="${1:-lhost}"
> 	local counter="${2:-1}"
> 	local host net max_counter max_net max_host tmp
>
> 	[ "$TST_IPV6" ] && max_net=65536 || max_net=256
> 	max_host=$((max_net - 2))
> 	max_counter=$((max_host * max_net / 2))
>
> 	if [ $counter -gt $max_counter ]; then
> 		tst_resm TCONF "'$counter' higher than max counter limit '$max_counter'"
> 	fi

I guess we can safely omit this check.

>
> 	tmp=$((counter * 2))
> 	[ "$type" = "rhost" ] && tmp=$((tmp - 1))
>
> 	host=$((tmp % max_host))
> 	net=$((tmp / max_host))
> 	[ $host -eq 0 ] && { host=$max_host; net=$((net - 1)); }
>
> 	if [ "$TST_IPV6" ]; then
> 		echo "${IPV6_NET32_UNUSED}::$(printf %x:%x ${net} ${host})"
> 	else
> 		echo "${IPV4_NET16_UNUSED}.${net}.${host}"
> 	fi
> }

Looks good to me.

>
> Given the fact that most of the tests needs just one IP address (the
> exception are route4-change-{if,gw}, if-{addr,route}-{adddel,addlarge}) it's probably too
> overhead and error prone. I mean working with last octet/hextet might be enough.
>
> And if we want to work with more than last octet, we might appreciate to be able to
> specify (optionally) third octet as well ($net variable).
> Scripts which could benefit from it are at least:
> if-{addr,route}-addlarge, route4-change-if.
>
>
> Kind regards,
> Petr