Attachment 1446697 Details for Bug 1585078 – [NM-CI PATCH] ipv6: drop 1.12+ variant of @ipv6_add_static_address_manually_not_active

[patch] [NM-CI PATCH] ipv6: drop 1.12+ variant of @ipv6_add_static_address_manually_not_active

0001-ipv6-drop-1.12-variant-of-ipv6_add_static_address_ma.patch (text/plain), 5.28 KB, created by Beniamino Galvani on 2018-06-01 13:37:53 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Beniamino Galvani

Created: 2018-06-01 13:37:53 UTC

Size: 5.28 KB

patch

obsolete

>From 3f9a14408593b83d014c651b48786244d051412d Mon Sep 17 00:00:00 2001
>From: Beniamino Galvani <bgalvani@redhat.com>
>Date: Fri, 1 Jun 2018 15:18:03 +0200
>Subject: [PATCH] ipv6: drop 1.12+ variant of
> @ipv6_add_static_address_manually_not_active
>
>NM 1.12 still adds the IPv6LL address when an IPv6 address is added
>externally. The reason why this test passed is that it triggered an
>assertion:
>
> src/devices/nm-device.c: line 8080
> (nm_device_get_configured_mtu_for_wired): should not be reached
>
>which caused a restart of the NM process and therefore the LL address
>wasn't added.
>
>Once the assertion is fixed, the LL gets added because the following
>happens:
>
> - NM detects that the address is added externally
>
> - in queued_ip_config_change() we call check_and_add_ipv6ll_addr()
>   for the reason explained by this code comment:
>
>  "If no IPv6 link-local address exists but other addresses do then we
>   must add the LL address to remain conformant with RFC 3513 chapter 2.1
>   ("Addressing Model"): "All interfaces are required to have at least
>   one link-local unicast address".
>
> - check_and_add_ipv6ll_addr() generates and adds a LL address
>
> - then, an in-memory connection is created and device becomes
>   "external". From this point NM no longer touches the interface.
>
>I think changing this behavior and requiring the user to add a LL
>address (or change the addrgenmode and cycle the interface down+up)
>would be very user-unfriendly and so I prefer we keep adding address
>even if the connection is going to become "external".
>---
> nmcli/features/ipv6.feature | 30 +-----------------------------
> 1 file changed, 1 insertion(+), 29 deletions(-)
>
>diff --git a/nmcli/features/ipv6.feature b/nmcli/features/ipv6.feature
>index c639eb9..8e3884f 100644
>--- a/nmcli/features/ipv6.feature
>+++ b/nmcli/features/ipv6.feature
>@@ -1062,59 +1062,31 @@
>     * Add a new connection of type "ethernet" and options "con-name con_ipv6 ifname eth10 autoconnect no"
>     * Bring up connection "con_ipv6"
>     * Wait for at least "20" seconds
>     Then Check ipv6 connectivity is stable on assuming connection profile "con_ipv6" for device "eth10"
> 
> 
>     @rhbz1083133 @rhbz1098319 @rhbz1127718
>     @veth @eth3_disconnect
>-    @ver-=1.11.1
>     @ipv6_add_static_address_manually_not_active
>-    Scenario: NM - ipv6 - add a static address manually to non-active interface (legacy 1.10 behavior and older)
>+    Scenario: NM - ipv6 - add a static address manually to non-active interface
>     Given "testeth3" is visible with command "nmcli connection"
>     Given "eth3\s+ethernet\s+connected" is not visible with command "nmcli device"
>     Given "state UP" is visible with command "ip a s eth3"
>     * "0" is visible with command "cat /proc/sys/net/ipv6/conf/eth3/disable_ipv6"
>     * Execute "ip -6 addr add 2001::dead:beef:01/64 dev eth3"
>     Then "0" is visible with command "cat /proc/sys/net/ipv6/conf/eth3/disable_ipv6"
>     Then "inet6 2001::dead:beef:1/64 scope global" is visible with command "ip a s eth3"
>-    # Newer versions of NM no longer create IPv6 LL addresses for externally assumed devices.
>-    # This test is obsoleted by @ipv6_add_static_address_manually_not_active (1.12+), but this
>-    # behavior won't be backported to older versions.
>     Then "addrgenmode none " is visible with command "ip -d l show eth3"
>     Then "inet6 fe80" is visible with command "ip a s eth3" in "5" seconds
>     # the assumed connection is created, give just some time for DAD to complete
>     Then "eth3\s+ethernet\s+connected\s+eth3" is visible with command "nmcli device" in "5" seconds
> 
> 
>-    @rhbz1083133 @rhbz1098319 @rhbz1127718
>-    @veth @eth3_disconnect
>-    @ver+=1.11.2
>-    @ipv6_add_static_address_manually_not_active
>-    Scenario: NM - ipv6 - add a static address manually to non-active interface
>-    Given "testeth3" is visible with command "nmcli connection"
>-    Given "eth3\s+ethernet\s+connected" is not visible with command "nmcli device"
>-    Given "state UP" is visible with command "ip a s eth3"
>-    * "0" is visible with command "cat /proc/sys/net/ipv6/conf/eth3/disable_ipv6"
>-    * Execute "ip -6 addr add 2001::dead:beef:01/64 dev eth3"
>-    Then "0" is visible with command "cat /proc/sys/net/ipv6/conf/eth3/disable_ipv6"
>-    Then "inet6 2001::dead:beef:1/64 scope global" is visible with command "ip a s eth3"
>-    #
>-    # the connection is assumed externally, meaning it has "addrgenmode none". NM is not
>-    # interferring with the device, hence there is no IPv6 LL address. Which is a problem,
>-    # but a problem of the user who takes over the device without setting the addrgenmode
>-    # to its liking.
>-    Then "addrgenmode none " is visible with command "ip -d l show eth3"
>-    Then "inet6 fe80" is not visible with command "ip a s eth3" for full "5" seconds
>-    #
>-    # the assumed connection is created, give just some time for DAD to complete
>-    Then "eth3\s+ethernet\s+connected\s+eth3" is visible with command "nmcli device"
>-
>-
>     @rhbz1138426
>     @restart @add_testeth10
>     @ipv6_no_assumed_connection_for_ipv6ll_only
>     Scenario: NM - ipv6 - no assumed connection on IPv6LL only device
>     * Delete connection "testeth10"
>     * Execute "systemctl stop NetworkManager.service"
>     * Execute "ip a flush dev eth10; ip l set eth10 down; ip l set eth10 up"
>     When "fe80" is visible with command "ip a s eth10" in "5" seconds
>-- 
>2.17.0
>

Actions: View | Diff

Attachments on bug 1585078: 1446696 | 1446697