On 23/06/15 06:12, Lidong Zhong wrote: >>>> On 6/18/2015 at 10:03 PM, in message <[email protected]>, >>>> Christine > Caulfield <[email protected]> wrote: >> On 18/06/15 03:20, Lidong Zhong wrote: >>> Hi Christine, >>> >>> Any news about this patch? >>> >> >> Hi Lidong, >> > Hi Christine, > > Thanks for your reply. > >> I'm afraid I'm going to have to NAK it. Although it technically matches >> the man page, it breaks existing behaviour (the ability to bind to a >> host IP address) and that behaviour is also valid in corosync 2. If you >> can come up with a patch that allows that behaviour AND fixes this >> problem then I'll he happy to review it. >> > I think the condition you described here has already been considered in the > original logic. > The code is here > 611 if (tb[IFA_LOCAL]) { > 612 memcpy(ipaddr.addr, RTA_DATA(tb[IFA_LOCAL]), > TOTEMIP_ADDRLEN); > 613 if (totemip_equal(&ipaddr, bindnet)) { > 614 found_if = 1; > 615 exact_match_found = 1; > 616 } > 617 } > Then it will check if the bindnetwork address with netmask. >
You are correct, I took the time to test it today :) It's been such a long time since that code was looked at, I'd forgotten how that bit worked. I'll check it into the flatiron branch, but it's unlikely to make RHEL unless there's a compelling bugzilla to go with it. Thanks for being persistent and patient. Chrissie > Regards, > Lidong > >> Sorry for delay in responding, I was away for a while. >> >> Chrissie >> >>> Regards, >>> Lidong >>> >>>>>> On 5/15/2015 at 09:52 AM, in >>>> message >>> <[email protected]>, "Lidong >>>> Zhong" >>> <[email protected]> wrote: >>> On 5/14/2015 at 08:23 PM, in message >>>> <[email protected]>, Christine >>>> Caulfield <[email protected]> wrote: >>>>> On 14/05/15 06:36, Lidong Zhong wrote: >>>>>> For instance, the IPs are: >>>>>> eth0: 10.150.10.81/8 >>>>>> eth1: 10.150.10.91/24 >>>>>> The bindnetaddr in corosync is: 10.150.10.0 and the addr of eth1 is >>>>>> selected in UDPU mode. Then the bindto address probably mistakenly >>>>>> choose >>>>>> the addr of eth0. >>>>>> >>>>> >>>>> Hiya, >>>>> >>>>> Thanks for the patch but I'm not convinced it's the right thing to do >>>>> for all situations. If the bindnet_addr is set to an actual IP address >>>>> and not a subnet mask then it will fail I think. What does your >>>>> corosync.conf look like? >>>>> >>>> Hi Chrissie, >>>> >>>> Thanks for your reply. The full corosync.conf is attached below. >>>> As the man page says, the bindnetaddr always is the local interface ip >>>> with netmask. It should be a full ip address only in IPv6 case. And I >>>> think >> >>>> there >>>> is no problem for the current code to handle IPv6 address. >>>> >>>> Regards, >>>> Lidong >>>> >>>>> Chrissie >>>>> >>>>>> Signed-off-by: Lidong Zhong <[email protected]> >>>>>> --- >>>>>> exec/totemip.c | 2 +- >>>>>> 1 file changed, 1 insertion(+), 1 deletion(-) >>>>>> >>>>>> diff --git a/exec/totemip.c b/exec/totemip.c >>>>>> index ab175a8..c1fa800 100644 >>>>>> --- a/exec/totemip.c >>>>>> +++ b/exec/totemip.c >>>>>> @@ -626,7 +626,7 @@ int totemip_iface_check(struct totem_ip_address >>>>> *bindnet, >>>>>> memcpy(&network, >>>>>> RTA_DATA(tb[IFA_LOCAL]), sizeof(uint32_t)); >>>>>> memcpy(&addr, bindnet->addr, >>>>>> sizeof(uint32_t)); >>>>>> >>>>>> - if ((addr & netmask) == >>>>>> (network & netmask)) { >>>>>> + if (addr == (network & >>>>>> netmask)) { >>>>>> memcpy(ipaddr.addr, >>>>>> RTA_DATA(tb[IFA_ADDRESS]), TOTEMIP_ADDRLEN); >>>>>> found_if = 1; >>>>>> } >>>>>> >>>>> >>>>> _______________________________________________ >>>>> discuss mailing list >>>>> [email protected] >>>>> http://lists.corosync.org/mailman/listinfo/discuss >>>>> >>>>> >>>> >>>> >>>> >>> >>> >> >> >> > _______________________________________________ discuss mailing list [email protected] http://lists.corosync.org/mailman/listinfo/discuss
