Hi,
In the recent change made to the above function to register for ARP events,
it looks like there is an access to the pool elements using pointers across
a pool_get call.
--
while (p && *p != ~0)
{
mc = pool_elt_at_index (am->mac_changes, *p);
if (mc->node_index == node_index
Hi,
VPP in the latest builds crashes with the probe-neighbor command for a /31
neighbor.
That is, if I have configured 10.2.2.4/31 on an interface and probe
10.2.2.5, VPP crashes at adj_get() because in the
ip4_add_del_interface_route function, the neighbor_adj_index is not set for
a /31 route (i
Sorry, I meant "ip6_neighbor_sw_interface_add_del"
Thanks,
-nagp
On Fri, May 26, 2017 at 4:05 PM, Nagaprabhanjan Bellaru <
nagp.li...@gmail.com> wrote:
> When radv_info is created, it is initialized to zero, but mcast_adj_index
> is assigned in the same flow - in ip6_sw_in
adj_index defaults to ~0 when the
> radv_info is first created.
>
>
>
> /neale
>
>
>
> *From: *Nagaprabhanjan Bellaru
> *Date: *Friday, 26 May 2017 at 10:12
> *To: *"Neale Ranns (nranns)"
> *Cc: *vpp-dev
> *Subject: *Re: [vpp-dev] Is adj_index "0&qu
n older code this was
> not the case.
>
>
>
> Regards,
>
> neale
>
>
>
> *From: * on behalf of Nagaprabhanjan Bellaru
>
> *Date: *Friday, 26 May 2017 at 08:43
> *To: *vpp-dev
> *Subject: *[vpp-dev] Is adj_index "0" not valid?
>
>
>
Hi,
In the latest VPP code, in icmp6_router_solicitation - the code checks for
radv->mcast_adj_index and if it is "0" - it flags an error
ICMP6_DST_LOOKUP_MISS.
However, in ip6_neighbor_sw_interface_add_del - when the radv_info_t
structure is getting initialized - when mcast_adj_index is allocate
ation interface already carry vrfs but we’re not
> yet using them in tcp and for session lookup. We do however plan to support
> this in the future.
>
> Could you elaborate on why you would need support for this?
>
> HTH,
> Florin
>
> > On May 24, 2017, at 10:28 PM, Nagaprabh
I am asking this because the session lookup functions in tcp_input.c does
not seem to be taking a fib index or a table id to look up the session -
just the usual 4-tuple.
If it is done in a different way, please help me see it.
Thanks,
-nagp
___
vpp-dev
erver()).
>
>
>
> Post lookup, mcast packets will be sent to ipX_local and then onto your
> handler registered through ipX_register_protocol().
>
>
>
> Hth,
>
> neale
>
>
>
>
>
> *From: * on behalf of Nagaprabhanjan Bellaru
>
> *Date: *Mond
I was wondering if any of the following options work:
* Have a dpo-receive assigned to a multicast address and use the ip
protocol type to punt packets to a custom node in graph
* Have a custom dpo type created and redirect packets directly from ip
lookup with a "fib_table_entry_special_dpo_add"
Thanks, I have found the commit which fixes the busy loop issue. Now CPU
utilization is at bay. Thanks a lot! :-)
-nagp
On Thu, May 11, 2017 at 5:10 PM, Nagaprabhanjan Bellaru <
nagp.li...@gmail.com> wrote:
> Thanks a lot Damjan, can you please point me to the commit so that I can
&g
Thanks a lot Damjan, can you please point me to the commit so that I can
pick them?
Thanks,
-nagp
On Tue, May 9, 2017 at 7:02 PM, Damjan Marion
wrote:
>
>
> On 9 May 2017, at 10:30, Nagaprabhanjan Bellaru
> wrote:
>
> Hi,
>
> I am running vpp lite version with no n
ich rely on the etype of the packet only.
> -John
>
>
>
> *From:* vpp-dev-boun...@lists.fd.io [mailto:vpp-dev-boun...@lists.fd.io] *On
> Behalf Of *Nagaprabhanjan Bellaru
> *Sent:* Tuesday, May 09, 2017 5:49 AM
> *To:* Damjan Marion (damarion)
> *Cc:* vpp-dev
> *Su
> On 9 May 2017, at 10:27, Nagaprabhanjan Bellaru
> wrote:
>
> Yes, it is set correctly. To validate it, I bypassed the dpdk optimization
> and fed every packet to ethernet-input (by making "dpdk_rx_next_from_etype"
> always return VNET_DEVICE_INPUT_NEXT_ETHERNET_INPUT)
Hi,
I am running vpp lite version with no nodes in the polling mode, but still
unix-epoll-input goes to polling mode and vpp_main consumes 100% CPU.
Please find the "show run" output here:
https://pastebin.com/0STv9JQV
When I stepped through the unix-epoll-input code - timeout_ms calculations
al
rt_device_input" does not seem to be
doing anything here..
Thanks,
-nagp
On Mon, May 8, 2017 at 7:32 PM, Damjan Marion (damarion) wrote:
> >
> > On 3 May 2017, at 13:28, Nagaprabhanjan Bellaru
> wrote:
> >
> > Hi,
> >
> > It looks like dpdk_device_i
Hi,
It looks like dpdk_device_input() - is not checking if there is a vlan
header in the packet or not and always sets the buffer->current_data to 14
(smac+dmac+ethtype). Because of that ip4_input is not able to recognize a
correct IP packet.
For example, I have a subinterface created with vlan10
I see that the hash_delete function in vppinfra/hash.h is empty. If I want
to destroy a hash how to do that?
Thanks,
-nagp
___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev
Thanks Neale! That seems to help..
-nagp
On Fri, Apr 28, 2017 at 12:18 PM, wrote:
> > I’d suggest a modification to ip6_interface_first_address to indicate
> whether a link-local is ok. Then you can choose to ignore the link-local
> when the ping’s destination is global scope.
>
> At some point
Even though the destination IP6 is multi hop away, ping seems to choose
link local as its source address - any idea what can be done to override
this behavior?
Thanks,
-nagp
___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinf
’s
> source address, but we don’t want a FIB entry to be generated from it. This
> patch adds those functions:
>
> https://gerrit.fd.io/r/#/c/6264/
>
>
>
> please let me know if it fixes your issue. If so I’ll merge.
>
>
>
> Thanks,
>
> neale
>
>
>
solicitation request with
(global source ip6, link local multicast group address) - and an
advertisement is coming back with (link local source ip6, global
destination ip6)
Thanks,
-nagp
On Thu, Apr 13, 2017 at 10:49 AM, Nagaprabhanjan Bellaru <
nagp.li...@gmail.com> wrote:
> Makes sens
Neale
>
>
>
> *there are other ways to do this if you have an object you can ‘dpo_stack’
> and re-stack when the adj changes state. There are various examples of
> features stacking on a fib_entry (e.g. LISP and VXALN), but not on an
> adjacency – but the principle is the s
a per-interface link-local address only ‘FIB’. But
> VPP has no such function. The adjacency data-base serves this purpose*.
>
>
>
> Regards,
>
> Neale
>
>
>
> *if you are using this in the data-plane though, we should consider
> alternatives.
>
>
>
&g
x_t ai = adj_nbr_add_or_lock(FIB_PROTOCOL_IP6, VNET_LINK_IP6,
> , );
>
>
>
> now you have the adjacency, the lookup is not required, so you can send
> the packet directly to ip6_rewrite/ip6_discover_neighbor. There is an
> example of this in the BFD code (see bfd_udp_input()).
&
Hi,
How can I send packets to a link local IPv6 address? I know the interface
on which the packet has to be sent, but when I enqueue the packet to
"ip6_lookup" node, how can I encode the interface such that the neighor
resolution happens on that interface and the packet gets transmitted?
Thanks,
gt;
> */
>
> static f64 fib_walk_sleep_duration[] = {
>
> [FIB_WALK_LONG_SLEEP] = 1e-3,
>
> [FIB_WALK_SHORT_SLEEP] = 1e-8,
>
> };
>
>
>
> I’d suggest ‘[sudo] perf top’ as a means to determine what is consuming
> the CPU time.
>
>
>
> Regards,
>
>
> ‘sh fib entry 19’
>
>
>
> if no cycle exists let me know. If one does exist, please debug from the
> from a routing perspective.
>
>
>
> Regards,
>
> /neale
>
>
>
> *From: * on behalf of Nagaprabhanjan Bellaru
>
> *Date: *Monday, 10 April
There are a few routes in FIB, but fib-walk takes up 100% CPU. "show run"
shows fib-wak process running perpetually.
Is there anything that we can do to get around this?
Thanks,
-nagp
___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mail
Yourtchenko
wrote:
> Hello,
>
> > On 27 Jan 2017, at 04:12, Nagaprabhanjan Bellaru
> wrote:
> >
> > Hi,
> >
> > I am not sure if the ping debug CLI is being actively used, but the
> function "init_icmp46_echo_request" goes ahead and wri
Hi,
I am not sure if the ping debug CLI is being actively used, but the
function "init_icmp46_echo_request" goes ahead and writes 2000 bytes into
the vlib_buffer corrupting the surrounding memory area. After 3-4 pings,
vpp always crashes.
Instead of copying sizeof(icmp_echo_request->data) which i
Hi,
I am trying to understand how the nodes defined in the dpo files, for
example "lookup-ip4-dst-itf" in lookup_dpo.c are linked to the graph in the
data plane.
For example, "show vlib graph" shows a number of next nodes for
"lookup-ip4-dst-itf" like "ip4-local", "ip4-arp", "ip4-glean" etc., but
xperiment is because 1-2 and 2-1 translates were applied on both
> end of xconnect interfaces respectively. -John
>
>
>
> *From:* Nagaprabhanjan Bellaru [mailto:nagp.li...@gmail.com]
> *Sent:* Friday, January 13, 2017 2:21 PM
>
> *To:* John Lo (loj)
> *Cc:* Jan Gelety -X
why two vtr operations are performed per packet.
>
>
>
> Regards,
>
> John
>
>
>
> *From:* Nagaprabhanjan Bellaru [mailto:nagp.li...@gmail.com]
> *Sent:* Friday, January 13, 2017 1:50 PM
> *To:* John Lo (loj)
> *Cc:* Jan Gelety -X (jgelety - PANTHEON TECHNOLOGIES
>
> Regards,
>
> John
>
>
>
> *From:* vpp-dev-boun...@lists.fd.io [mailto:vpp-dev-boun...@lists.fd.io] *On
> Behalf Of *Jan Gelety -X (jgelety - PANTHEON TECHNOLOGIES at Cisco)
> *Sent:* Friday, January 13, 2017 12:24 PM
> *To:* Nagaprabhanjan Bellaru ; vpp-dev <
>
Hi,
I have a L2 cross-connect configuration where packets arrive on a sub-if1
(with vlan id 500) and go out on sub-if2 (with a vlan stack 200, 150).
When I send packets on sub-if1, it is getting out on sub-if2, but the vlan
tags are not getting re-written. Is there anything else that should be do
Hi,
Is it possible to do L2 cross-connect over MPLS?
For example:
I want traffic coming in on a L2 sub interface to get punted on a MPLS
tunnel (i.e. with MPLS encapsulation) and on the other side, pop the labels
and switch the packet on the corresponding sub interface. Technically, a
psuedowire
m ip6 neighbour discovery "service"
>
> This is my guess, please correct me somebody with more insight on this.
>
> Regards,
> Dmitry
>
> On Wed, Oct 19, 2016 at 8:54 AM, Nagaprabhanjan Bellaru <
> nagp.li...@gmail.com> wrote:
>
>> Hi,
>>
HI,
I am writing a small plugin to process packets from a graph node. I see
that FIB_PROTOCOL_IP4 is being mapped to 0 and fib_table_get(0, 0) (index 0
and protocol 0) is asserting.
However, fib_table_get(0, 1) and fib_table_get(0, 2) are returning ip4 and
ip6 fib tables respectively, indicating
Hi,
The sup_and_sub_key which gets `clib_mem_alloc`d during create sub
interface does not appear to be getting freed in the "delete_sub_interface"
function. Is this expected behavior?
Thanks,
-nagp
___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://li
gt; what they logically mean.
>
>
>
> Chris.
>
>
>
> *From:* vpp-dev-boun...@lists.fd.io [mailto:vpp-dev-boun...@lists.fd.io] *On
> Behalf Of *Nagaprabhanjan Bellaru
> *Sent:* Monday, November 14, 2016 6:27 AM
> *To:* vpp-dev
> *Subject:* Re: [vpp-dev] Creating an unta
Is my understanding correct? Or is this a bug?
-nagp
On Wed, Oct 26, 2016 at 3:48 PM, Nagaprabhanjan Bellaru <
nagp.li...@gmail.com> wrote:
> Is this expected behavior?
>
> DBGvpp# show interface
> Name Idx State Counte
00:06:00.08086:1583 5.0 GT/s x8igb_uio XL710
40GbE Controller
:06:00.18086:1583 5.0 GT/s x8igb_uio XL710
40GbE Controller
--
-nagp
On Sat, Nov 12, 2016 at 3:54 PM, Nagaprabhanjan Bellaru <
nagp.li...@gmail.com> wrote:
> Hi,
>
&
Hi,
I see that in VPP CLI, "show pci" shows the interfaces, but dpdk_lib_init
reports no ports. Please find the output below. Can someone help me by
identifying what is the problem here? When I download DPDK separately and
run testpmd, it detects the interfaces.
I tried both manually loading the
with .so files coming from ubuntu packages.
>
>
> > On 9 Nov 2016, at 17:42, Nagaprabhanjan Bellaru
> wrote:
> >
> > Hi,
> >
> > I want to know if we can build dpdk as .so files? I tried enabling the
> CONFIG_RTE_SHARED_LIB=y in common-base configura
Hi,
I want to know if we can build dpdk as .so files? I tried enabling the
CONFIG_RTE_SHARED_LIB=y in common-base configuration of DPDK and build vpp,
but it fails while building librte_sched.so because it fails to link with
librte_timer which is switched off in VPP.
Is there any workaround for t
Is this expected behavior?
DBGvpp# show interface
Name Idx State Counter
Count
local00down
eth-1 1 up
eth-2 2 up
eth-3 3 up
DBGvpp# cre
Hi,
I created a sub interface and configured IP6 address on it. After I deleted
the sub interface, the timer event function kicked in and tried to access
the sw interface and got aborted.
Am I doing something wrong here? Or should there be a defensive check in
the timer event?
Thanks,
-nagp
48 matches
Mail list logo
