On Thu, 2007-03-08 at 03:59 +0000, Linux Kernel Mailing List wrote:
> Gitweb:
> http://git.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=c7ababbdc647e67e953d153ddf62cbdc9fe3297e
> Commit: c7ababbdc647e67e953d153ddf62cbdc9fe3297e
> Parent: 16bec31db751030171b31d7767fa3a5bdbe980ea
> Author: Herbert Xu <[EMAIL PROTECTED]>
> AuthorDate: Wed Mar 7 16:02:40 2007 -0800
> Committer: David S. Miller <[EMAIL PROTECTED]>
> CommitDate: Wed Mar 7 16:08:12 2007 -0800
>
> [IPV6]: Do not set IF_READY if device is down
>
> Now that we add the IPv6 device at registration time we don't need
> to set IF_READY in ipv6_add_dev anymore because we will always get
> a NETDEV_UP event later on should the device ever become ready.
>
> Signed-off-by: Herbert Xu <[EMAIL PROTECTED]>
> Signed-off-by: David S. Miller <[EMAIL PROTECTED]>
This commit seems to have broken NetworkManager, which removes the
link-local IPv6 address from each wireless interface while it's only
intending to use it for scanning, then adds it back when it actually
wants to _use_ the interface in question.
After this commit, any IPv6 address added to an interface which had none
will be marked as 'tentative', and will never become real.
A failing test looks like this....
+ ip -6 addr del fe80::203:ccff:fe3c:27d/64 dev eth0
+ ip -6 addr del 2001:8b0:10b:1:203:ccff:fe3c:27d/64 dev eth0
+ ip -6 addr list eth0
+ ip -6 addr add fe80::203:ccff:fe3c:27d/64 dev eth0
+ ip -6 addr list eth0
4: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qlen 1000
inet6 fe80::203:ccff:fe3c:27d/64 scope link tentative
valid_lft forever preferred_lft forever
+ sleep 3
+ ip -6 addr list eth0
4: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qlen 1000
inet6 fe80::203:ccff:fe3c:27d/64 scope link tentative
valid_lft forever preferred_lft forever
... while on a good kernel it looks like this...
+ ip -6 addr del fe80::203:ccff:fe3c:27d/64 dev eth0
+ ip -6 addr del 2001:8b0:10b:1:203:ccff:fe3c:27d/64 dev eth0
+ ip -6 addr list eth0
+ ip -6 addr add fe80::203:ccff:fe3c:27d/64 dev eth0
+ ip -6 addr list eth0
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qlen 1000
inet6 fe80::203:ccff:fe3c:27d/64 scope link tentative
valid_lft forever preferred_lft forever
+ sleep 3
+ ip -6 addr list eth0
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qlen 1000
inet6 2001:8b0:10b:1:203:ccff:fe3c:27d/64 scope global dynamic
valid_lft 298sec preferred_lft 118sec
inet6 fe80::203:ccff:fe3c:27d/64 scope link
valid_lft forever preferred_lft forever
If I add an extra IPv6 address before running the script on a 'bad'
kernel, it behaves correctly. The problem only happens when you remove
_all_ IPv6 addresses from an interface.
Answers of "don't do that then" should be accompanied with an
alternative plan which allows NetworkManager to use a wireless device
for scanning without actually having any other effect.
--
dwmw2
-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
