On 08/22/2017 at 02:11 PM, Pingfan Liu wrote:
----- Original Message -----
> From: "Ziyue Yang" <ziyang(a)redhat.com>
> To: "Xunlei Pang" <xlpang(a)redhat.com>
> Cc: kexec(a)lists.fedoraproject.org, "Baoquan He" <bhe(a)redhat.com>
> Sent: Tuesday, August 22, 2017 11:34:19 AM
> Subject: Re: [PATCH v5 1/3] kdump-lib.sh: support link local ipv6 address with
interface name
>
> Hi,
>
> On Tue, Aug 22, 2017 at 9:30 AM, Xunlei Pang < xpang(a)redhat.com > wrote:
>
>
>
> On 08/22/2017 at 09:10 AM, Baoquan He wrote:
>> On 08/22/17 at 08:59am, Xunlei Pang wrote:
>>> On 08/21/2017 at 10:07 AM, Ziyue Yang wrote:
>>>> On Fri, Aug 18, 2017 at 11:33 AM, Xunlei Pang < xpang(a)redhat.com
<mailto:
>>>> xpang(a)redhat.com >> wrote:
>>>>
>>>> On 08/13/2017 at 05:41 PM, Ziyue Yang wrote:
>>>>> When dumping to ssh via local ipv6 address, the ssh parameter
>>>>> in kdump.conf is supposed to have the form like
>>>>>
>>>>> ssh user@fe80::cc1:8bff:fe90:b95f%eth0
>>>>>
>>>>> where "%eth0" is an existing network interface supporting
ipv6.
>>>>>
>>>>> The get_remote_host function in kdump-lib.sh currently
>>>>> doesn't remove the network interface in the link local ipv6
>>>>> addresses, causing the ip command in kdump_install_net
>>>>> function to fail, leading to a "Bad kdump location"
message.
>>>>>
>>>>> This commit
>>>>> 1) makes get_remote_host function remove network interface
>>>>> in ipv6 addresses if there are any to support host finding;
>>>>> 2) adds is_prefixed_ipv6_link_local function to detect link local
>>>>> ipv6 address with 'kdump-' prefixed interface name.
>>>>>
>>>>> Signed-off-by: Ziyue Yang < ziyang(a)redhat.com <mailto:
>>>>> ziyang(a)redhat.com >>
>>>>> ---
>>>>> kdump-lib.sh | 12 ++++++++++++
>>>>> 1 file changed, 12 insertions(+)
>>>>>
>>>>> diff --git a/kdump-lib.sh b/kdump-lib.sh
>>>>> index 3f0af91..e67b28f 100755
>>>>> --- a/kdump-lib.sh
>>>>> +++ b/kdump-lib.sh
>>>>> @@ -312,6 +312,14 @@ is_ipv6_address()
>>>>> echo $1 | grep -q ":"
>>>>> }
>>>>>
>>>>> +is_prefixed_ipv6_link_local()
>>>>> +{
>>>>> + is_ipv6_address $1 && {
>>>>> + local _host_postfix=${1##*\%}
>>>>> + echo $_host_postfix | grep -q "^kdump-eth"
>>>>> + }
>>>>> +}
>>>>> +
>>>>> # get ip address or hostname from nfs/ssh config value
>>>>> get_remote_host()
>>>>> {
>>>>> @@ -323,6 +331,10 @@ get_remote_host()
>>>>> _config_val=${_config_val%:/*}
>>>>> _config_val=${_config_val#[}
>>>>> _config_val=${_config_val%]}
>>>>> + # factor out network inteface segment in local ipv6 address
>>>>> + if is_ipv6_address $_config_val; then
>>>>> + _config_val=${_config_val%\%*}
>>>>> + fi
>>>> There is something unclear to me, why does ipv6 address need an
interface
>>>> name followed?
>>>> We can see in kdump_install_net(), it uses "ip -o route get to
$_server"
>>>> to acquire the
>>>> interface name, is there any chance that the name got is different from
>>>> that in "ssh"?
>>>>
>>>>
>>>> ssh related commands needs the interface name for routing, since for
link
>>>> local ipv6 addresses there might be multiple interfaces that could be
>>>> used for routing.
>>> That's what I concerned about, e.g. there are multiple available
routing,
>>> via eth0 or eth1.
>>> You specified eth1 for "ssh" in /etc/kdump.conf, unfortunately
>>> kdump_install_net() gets
>>> eth0 as kdump's netdev from $(get_ip_route_field "$_route"
"dev"). Is this
>>> possible?
>> An available route entry must include destination ip and next hop
>> information, either next hop ip address, or interface to next hop.
>> Only a destination is not a complete route.
> I just simplified.
>
> To be specific, my question is that whether the netdev got from the route
> entry can be ensured
> to be the same as the one we specified in "ssh" directive, because from
the
> code we clearly see
> there are two different sources.
> Yes I agree.
> Since $1 in kdump_setup_ifname is promised to be an interface with valid
> route, maybe we can use the $1 in kdump_setup_ifname directly as the
> interface in the ssh_target related code?
>
After testing, I found: "ip route get to link-local addr" just returns the
default route-entry. It does not do some arp testing.
We can observe the packet by tcpdump, in that case, there is no packet. Also given any
invalid or non-exist link-local addr, the cmd always return the default entry.
To get the link-path for link-local addr, we should use "ip neigh" cmd
Maybe it has nothing to do with ARP(NDISC), we just need to know the output interface for
the ssh target.
See
https://en.wikipedia.org/wiki/IPv6_address#Link-local_addresses_and_zone_...
So I think we can't rely on route to get the netdev, in such cases(i.e. ipv6 link
local address),
we can simply get netdev from the zone index of "ssh" directive instead of using
"ip route".
Regards,
Xunlei