LVS
lvs-devel
Google
 
Web LinuxVirtualServer.org

Re: [bug] FWMARKs and persistence in IPVS: The Use of Unions

To: Simon Horman <horms@xxxxxxxxxxxx>
Subject: Re: [bug] FWMARKs and persistence in IPVS: The Use of Unions
Cc: Jan Engelhardt <jengelh@xxxxxxxxxx>, netfilter-devel <netfilter-devel@xxxxxxxxxxxxxxx>, lvs-devel@xxxxxxxxxxxxxxx, Fabien Duchêne <fabien.duchene@xxxxxxxxxxxxxxxxxxxx>, Joseph Mack NA3T <jmack@xxxxxxxx>
From: Julius Volz <julius.volz@xxxxxxxxx>
Date: Tue, 28 Apr 2009 13:59:22 +0200
On Tue, Apr 28, 2009 at 12:59 PM, Simon Horman <horms@xxxxxxxxxxxx> wrote:
> On Tue, Apr 28, 2009 at 07:23:55PM +1000, Simon Horman wrote:
>> On Tue, Apr 28, 2009 at 11:07:40AM +0200, Jan Engelhardt wrote:
>> >
>> > On Tuesday 2009-04-28 10:15, Simon Horman wrote:
>> > >
>> > >It seems to me that it should be easy enough to fix by changing
>> > >fwmark in ip_vs_sched_persist() from:
>> > >
>> > >union nf_inet_addr fwmark = {
>> > >   .all = { 0, 0, 0, htonl(svc->fwmark) }
>> > >};
>> > >
>> > >to:
>> > >
>> > >union nf_inet_addr fwmark = {
>> > >   .all = { htonl(svc->fwmark), 0, 0, 0 }
>> > >};
>> > >
>> > >Assuming that this would result in fwmark->ip being set to
>> > >htonl(svc->fwmark), which is relevant if svc->af is AF_INET - that is,
>> > >for IPv4.[...]
>> > >An alternate idea would be to change the af value used for fwmarks,
>> > >but this seems to be even less clean than the current (slightly broken)
>> > >technique of using nf_inet_addr for IPv4 or IPv6 addresses, or fwmarks.
>> >
>> > If you use ->all, then using NFPROTO_UNSPEC as af
>> > seems to me like a good match.
>
> I am guessing that AF_UNSPEC is more appropriate than NFPROTO_UNSPEC.
> Please correct me if I am wrong.
>
>> That seems reasonable, though ip_vs_ct_in_get() would still
>> need to use the real af for the cp->af == af and
>> ip_vs_addr_equal(af, s_addr, &cp->caddr) portinos of the check.
>
> It looks like checking for proto == IPPROTO_IP can tell us if
> the destination is a fwmark. This is based on the assumption that
> iph.protocol can never be IPPROTO_IP in ip_vs_sched_persist().
>
> The following patch expresses these ideas as they crrently stand.
> Fabien, is it possible for you to test this?
>
> Index: net-next-2.6/net/netfilter/ipvs/ip_vs_conn.c
> ===================================================================
> --- net-next-2.6.orig/net/netfilter/ipvs/ip_vs_conn.c   2009-04-28 
> 20:37:48.000000000 +1000
> +++ net-next-2.6/net/netfilter/ipvs/ip_vs_conn.c        2009-04-28 
> 20:37:51.000000000 +1000
> @@ -260,7 +260,10 @@ struct ip_vs_conn *ip_vs_ct_in_get
>        list_for_each_entry(cp, &ip_vs_conn_tab[hash], c_list) {
>                if (cp->af == af &&
>                    ip_vs_addr_equal(af, s_addr, &cp->caddr) &&
> -                   ip_vs_addr_equal(af, d_addr, &cp->vaddr) &&
> +                   /* protocol should only be IPPROTO_IP if
> +                    * d_addr is a fwmark */
> +                   ip_vs_addr_equal(protocol == IPPROTO_IP ? AF_UNSPEC : af,
> +                                    d_addr, &cp->vaddr) &&
>                    s_port == cp->cport && d_port == cp->vport &&
>                    cp->flags & IP_VS_CONN_F_TEMPLATE &&
>                    protocol == cp->protocol) {
> @@ -698,7 +701,9 @@ ip_vs_conn_new(int af, int proto, const
>        cp->cport          = cport;
>        ip_vs_addr_copy(af, &cp->vaddr, vaddr);
>        cp->vport          = vport;
> -       ip_vs_addr_copy(af, &cp->daddr, daddr);
> +       /* proto should only be IPPROTO_IP if d_addr is a fwmark */
> +       ip_vs_addr_copy(proto == IPPROTO_IP ? AF_UNSPEC : af,
> +                       &cp->daddr, daddr);
>        cp->dport          = dport;
>        cp->flags          = flags;
>        spin_lock_init(&cp->lock);
> Index: net-next-2.6/net/netfilter/ipvs/ip_vs_core.c
> ===================================================================
> --- net-next-2.6.orig/net/netfilter/ipvs/ip_vs_core.c   2009-04-28 
> 20:37:48.000000000 +1000
> +++ net-next-2.6/net/netfilter/ipvs/ip_vs_core.c        2009-04-28 
> 20:37:51.000000000 +1000
> @@ -278,7 +278,7 @@ ip_vs_sched_persist(struct ip_vs_service
>                 */
>                if (svc->fwmark) {
>                        union nf_inet_addr fwmark = {
> -                               .all = { 0, 0, 0, htonl(svc->fwmark) }
> +                               .ip = htonl(svc->fwmark)
>                        };
>
>                        ct = ip_vs_ct_in_get(svc->af, IPPROTO_IP, &snet, 0,
> @@ -306,7 +306,7 @@ ip_vs_sched_persist(struct ip_vs_service
>                         */
>                        if (svc->fwmark) {
>                                union nf_inet_addr fwmark = {
> -                                       .all = { 0, 0, 0, htonl(svc->fwmark) }
> +                                       .ip = htonl(svc->fwmark)
>                                };
>
>                                ct = ip_vs_conn_new(svc->af, IPPROTO_IP,

Looks good to me, without being able to test it now.

I earlier mentioned also changing __ip_vs_conn_in_get(), but now
realized that the problem exists only during connection template
lookup, not for regular connections.

Julius
--
To unsubscribe from this list: send the line "unsubscribe lvs-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html

<Prev in Thread] Current Thread [Next in Thread>