I was going through openvswitch docs (OpenStack-131107.pdf and page number 13; Attached snapshot of page with this mail). I happen to see that first packet (of flow) goes to ovs-vswitchd (user space) and this makes decision of how to process THIS FLOW. Later, subsequent packets hit cached entry in ovs kernel module. Is this design still valid with latest code ? If yes, why don’t first packet of every new TCP connection does not go up to ovs-vswitchd (user space) ?
Please don’t mind even though I am asking basic questions. Regards, Vishwa. -----Original Message----- From: Badiger, Vishwanath (Vishwa) Sent: Monday, January 19, 2015 5:46 PM To: 'Jesse Gross' Cc: Uri Simchoni; discuss@openvswitch.org Subject: RE: [ovs-discuss] MSS Clamping I was going through openvswitch docs (OpenStack-131107.pdf and page number 13). I happen to see that first goes to ovs-vswitchd (user space) and this makes decision of how to process THIS FLOW. Later, subsequent packets hit cached entry in ovs kernel module. Is this design still valid with latest code ? If yes, why don’t first packet of every new TCP connection does not go up to ovs-vswitchd (user space) ? Please don’t mind even though I am asking basic questions. Regards, Vishwa. -----Original Message----- From: Jesse Gross [mailto:je...@nicira.com] Sent: Friday, January 16, 2015 2:08 AM To: Badiger, Vishwanath (Vishwa) Cc: Uri Simchoni; discuss@openvswitch.org Subject: Re: [ovs-discuss] MSS Clamping Thanks for the reminder. However, that patch will basically not work at all any more due to the fact that we try to avoid sending new TCP connections up to usespace. Therefore, I stand by my previous comments that you should look at the more recent thread. On Wed, Jan 14, 2015 at 10:29 PM, Badiger, Vishwanath (Vishwa) <vbadi...@avaya.com> wrote: > I am referring to below patch, > https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_pi > permail_dev_2012-2DJune_017640.html&d=AwIFaQ&c=BFpWQw8bsuKpl1SgiZH64Q& > r=QlZE3gUHm8jGRx97BmCtXqcaB_-3nXYBGm5kpCxsMks&m=96CZVpeBBqwkcX1tMaSfP8 > w_SAXSDy4xwyAna_dWMWo&s=J-6TI9tB4hAMuIGSNkKCUioTRyr-fcLk6Nx-uw-GI1g&e= > > MSS clamping even avoids packet getting fragmented by modifying MSS value in > TCP SYN packets. > > > Regards, > Vishwa. > > > -----Original Message----- > From: Jesse Gross [mailto:je...@nicira.com] > Sent: Thursday, January 15, 2015 11:04 AM > To: Badiger, Vishwanath (Vishwa) > Cc: Uri Simchoni; discuss@openvswitch.org > Subject: Re: [ovs-discuss] MSS Clamping > > On Wed, Jan 14, 2015 at 9:26 PM, Badiger, Vishwanath (Vishwa) > <vbadi...@avaya.com> wrote: >> Hi, >> I am in the process of migrating MSS clamping patch into recent OVS >> code base 2.3.0. But not finding functions like, 1. >> subfacet_make_actions() 2. iface_parse_tunnel() >> >> Looks like OVS has evolved lot compared to the code base to which MSS >> clamping patch created. >> I am planning to continue this effort and appreciate any help in >> understanding latest OVS implementation or migrating MSS clamping patch. >> >> Also, Is there any other way of implementing MSS clamping than how patch >> implements it ? >> Thanks. > > I don't actually know what patch you are referring to. However, I doubt that > a patch that is just updated to the latest code base will be accepted if the > original wasn't. > > Please see this (long) thread on more recent discussion of how to handle this > problem: > https://urldefense.proofpoint.com/v2/url?u=http-3A__www.spinics.net_li > sts_netdev_msg306502.html&d=AwIBaQ&c=BFpWQw8bsuKpl1SgiZH64Q&r=QlZE3gUH > m8jGRx97BmCtXqcaB_-3nXYBGm5kpCxsMks&m=7iNT4HO2rZcKPACWsmsxGFg6zXdLUpic > i8mppaeu1fQ&s=bj8J0oS15-pq-ZZK9elQoYnXSOpnzmICZu52c2-opYQ&e=
_______________________________________________ discuss mailing list discuss@openvswitch.org http://openvswitch.org/mailman/listinfo/discuss
