From mboxrd@z Thu Jan 1 00:00:00 1970 From: Petr Vorel Date: Thu, 23 May 2019 15:05:47 +0200 Subject: [LTP] [RFC PATCH v2 1/6] net/route: Remove route{4, 6}-change-if In-Reply-To: <25af53c5-dcfc-93b0-b9cb-4d179d485548@oracle.com> References: <20190510183132.31039-1-pvorel@suse.cz> <20190510183132.31039-2-pvorel@suse.cz> <25af53c5-dcfc-93b0-b9cb-4d179d485548@oracle.com> Message-ID: <20190523130547.GA18816@dell5510> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: ltp@lists.linux.it Hi Alexey, > Hi Petr, > On 5/10/19 9:31 PM, Petr Vorel wrote: > > These testsuites requires 3 NIC setup, which is not supported by network > > namespace setup (single host configuration) and IMHO not common for two > > host configuration either. > I guess, the third interface can be easily created during the setup, on the > lhost... this is what's required, right? > For example, by adding macvlan to the test interface: > ip link add ltp_mv0 link $(tst_iface) type macvlan mode bridge > Then, for the test route we can set either ltp_mv0 or $(tst_iface) interface. Thanks for review. I'd prefer to have this support in tst_net.sh as it can be needed for some tests in the future, so why reinvent a wheel? Probably via some variable (e.g.: TST_NET_EXTRA_NETNS_IFACES=2 => would add 2 more interfaces). But maybe not every single thing must be in the library... Much more important thing is how to solve validation when route-change-dst on IPv6 requires using rhost and whether using C brings a benefit (next patches in this patch-set). First I wanted to use it because speed: faster changes stress system more efficiently, but validation might slows it down anyway. Validation: I give up because server app using threads (each binded to it's own IP) was too slow and with client waiting for server to be up looked to me that validation complicates the test a lot. But maybe instead of threading running just several single thread instances would be faster. Kind regards, Petr