Hide Forgot
Description of problem: when shared connection with specified address range via ipv4.addresses field in nmcli exists and NM service restarts that address field is removed and thus dhcp operates with different addresses (10.42.0.1/24) Version-Release number of selected component (if applicable): NetworkManager-0.995.0.0-1.el7 How reproducible: deterministic Steps to Reproduce: 1.nmcli con add type ethernet ifname eth0 con-name dhcp 2.nmcli con mod dhcp ipv4.addresses 192.168.100.1/24 ipv4.method shared 3.nmcli con up id dhcp 4.systemctl restart NetworkManager 5.nmcli c s dhcp Actual results: 10.42.0.1/24 Expected results: 192.168.100.1/24 Additional info:
A fix is available in NetworkManager upstream git branch jk/shared-ip-rh1174632.
>+is_any_ip4_address_defined (shvarFile *ifcfg, int *idx) why do you need this rather than just assuming/requiring that it's index 0? >+ if (!read_ip4_address (ifcfg, "GATEWAY", &gateway, error)) >+ goto done; hm... NMSettingIP4Config::verify() doesn't forbid it, but gateway doesn't really make sense for SHARED does it? (Or do we allow having a SHARED interface also have the default route?)
(In reply to Dan Winship from comment #2) > >+is_any_ip4_address_defined (shvarFile *ifcfg, int *idx) > > why do you need this rather than just assuming/requiring that it's index 0? > > >+ if (!read_ip4_address (ifcfg, "GATEWAY", &gateway, error)) > >+ goto done; > > hm... NMSettingIP4Config::verify() doesn't forbid it, but gateway doesn't > really make sense for SHARED does it? (Or do we allow having a SHARED > interface also have the default route?) I don't know. But I agree, it it makes no sense, we should tighten verify() to reject it. - g_assert (connection); - success = nm_connection_verify (connection, &error); - g_assert_no_error (error); - g_assert (success); + nmtst_assert_connection_verifies_without_normalization (connection); Rest LGTM
LGTM too, though I think the gateway should probably optional for the shared method since it doesn't really matter for sharing, since NAT handles routing packets instead of any default gateway.
(In reply to Dan Winship from comment #2) > >+is_any_ip4_address_defined (shvarFile *ifcfg, int *idx) > > why do you need this rather than just assuming/requiring that it's index 0? The first address can also start at index 1 or 2. > > >+ if (!read_ip4_address (ifcfg, "GATEWAY", &gateway, error)) > >+ goto done; > The gateway *is* optional, because read_ip4_address() returns success if GATEWAY is missing. (In reply to Thomas Haller from comment #3) > (In reply to Dan Winship from comment #2) > > - g_assert (connection); > - success = nm_connection_verify (connection, &error); > - g_assert_no_error (error); > - g_assert (success); > + nmtst_assert_connection_verifies_without_normalization (connection); > Fixed.
Commits in upstream master: 233498e ifcfg-rh: update test files not to use obsolete GATEWAY<n> variables a3c52d7 man: add description of BOOTPROTO to nm-settings-ifcfg-rh manpage c8fe3bb ifcfg-rh: read custom IP address for shared connections (rh #1174632)
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://rhn.redhat.com/errata/RHBA-2015-0311.html