Re: [vpp-dev] memory tracer does not see free's

Let me take a look... Thanks for the report... -Original Message- From: vpp-dev@lists.fd.io On Behalf Of Andreas Schultz Sent: Thursday, May 16, 2019 12:57 PM To: vpp-dev@lists.fd.io Subject: [vpp-dev] memory tracer does not see free's Hi, VPPs memory trace seems to be missing the rele

[vpp-dev] memory tracer (leakfinder) seems OK

The output from this unit-test seems correct. Moreover, "memory-trace" ... do some stuff ... "show memory verbose" also seems pretty sensible... HTH... Dave static clib_error_t * test_memory_trace_command_fn (vlib_main_t * vm, unformat_input_t * input,

Re: [vpp-dev] VPP Unit test framework questions

The tests in src/vppinfra are lightly used - primarily by me - but yes, we do use them from time to time. test-gcov has worked as expected - at least on x86_64 - since we merged https://gerrit.fd.io/r/#/c/19379/. Running gcov manually is possible, but frankly it's not recommended. If you insi

Re: [vpp-dev] Process improvement proposal 3/3 - commit criteria

I'll add an item to the agenda right away... Thanks... Dave From: vpp-dev@lists.fd.io On Behalf Of Ole Troan Sent: Tuesday, May 21, 2019 1:43 PM To: vpp-dev Subject: [vpp-dev] Process improvement proposal 3/3 - commit criteria This is the third of three proposals for process changes. This is a

[vpp-dev] Historical ip4-icmp OAM scheme - OK to remove?

We recently tripped over the “vpe-icmp4-oam” scheme. It’s a historical artifact which consumes for-us icmp4 echo_reply packets. Net of switching init function ordering to a constraint-based topological sort, it reappeared. The ping CLI input node has been consuming for-us ip4 icmp packets for qu

Re: [vpp-dev] Regarding vlib_time_now

Vlib_time_now(...) works reasonably hard to return the same time on all worker threads: /* * Note when we let go of the barrier. * Workers can use this to derive a reasonably accurate * time offset. See vlib_time_now(...) */ vm->time_last_barrier_release = vlib_time_now (vm); al

Re: f64 vs u64 time values [Re: [vpp-dev] Regarding vlib_time_now]

alues to calculate periodic timer stuff within my packet processing routine. I wonder how much more expensive using vlib and the inherent divisions for converting to f64s here would be? Thanks, Chris. Dave Barach via Lists.Fd.Io writes: > Vlib_time_now(...) works reasonably

[vpp-dev] Periodic processes should be started on demand when practicable

Folks, We've accumulated quite a number of periodic processes which start whether or not the related feature will [ever] be used. See below. It's quite easy to switch to on-demand process creation. See https://gerrit.fd.io/r/#/c/19878 for a complete example. Here's a sketch of how to start a p

[vpp-dev] Messages stuck in the vpp-dev list moderation queue

Folks, Please note that list postings by non-list-members end up in an email moderation queue. Because 90+% of the emails in the moderation queue are spam / phish garbage, I tend not to look at the moderation queue very often. To ensure timely posting, please join the list. You can disable mes

Re: [vpp-dev] cache hits/misses counters per plugin/graph node

As Damjan wrote: “perf top” or the perfmon plugin will give you the info you’re looking for. “show pmc events” will list all of the available perf events Then: “set pmc event [...eventN]” “show pmc” to show results; per-node, per thread. HTH.. Dave From: vpp-dev@lists.fd.io On Behalf Of

Re: [vpp-dev] Unexpected vpp failure

This email sat in the moderation queue for several days, because you're not subscribed to vpp-dev@lists.fd.io. Please visit lists.fd.io and subscribe. You can disable mail delivery if you like. Since 97% of the message which land in the moderation queue are spam, I do

Re: [vpp-dev] Unexpected vpp failure

I downloaded the Debian package + core file. The core file is truncated, and gdb isn't able to show anything which would allow us to start work on the problem: Reading symbols from /usr/bin/vpp...Reading symbols from /usr/lib/debug/.build-id/5a/739f0daa7f81cd6b0e5529b422ce00b34ef8fb.debug...don

Re: [vpp-dev] sw_if_index reuse in tunnel implementations

Let me add a specific point on this subject: "the API or CLI invocation stops all workers somewhere between graph nodes." API / CLI barrier syncs stop worker threads at a very specific point: the top of the dispatch loop. At that point, all extant vectors (packets) have been fully processed.

Re: [vpp-dev] "owner node" help understanding.

Dear Chris, Please take a look at https://fdio-vpp.readthedocs.io/en/latest/gettingstarted/developers/vlib.html#complications et seq. I tried to document this exact bit of code. Let me know if the document helps, or if it raises more questions than it answers. Feel free to push doc patches t

Re: [vpp-dev] Passive close case, if FIN coming in with data, it seems don't count the FIN sequence, which lead to active close side retransmit FIN. #vnet

Please, please, please upgrade to 19.04. The TCP host stack in 19.04 has had a tremendous amount of hardening work, to the point where it passes all 1.2 million Codenomicon TCP test [read: attack] vectors. The 19.04 version is deployed in a least one critical production network role. If you ch

Re: [vpp-dev] How to get every log info print out to stdout #vpp

# /usr/bin/vpp unix interactive Not recommended for production use. Clib_warnings show up in syslog. D. From: vpp-dev@lists.fd.io On Behalf Of guangwei Sent: Tuesday, June 4, 2019 1:09 PM To: vpp-dev@lists.fd.io Subject: [vpp-dev] How to get every log info print out to stdout #vpp VPP run un

Re: [vpp-dev] About the order of  VLIB_INIT_FUNCTION called between different plugins

You can use vlib_get_plugin_symbol (char *plugin_name, char *symbol_name) to find the symbol you need. Clib_error_t (*fp)(vlib_main_t *) = vlib_get_plugin_symbo (“otherplugin”, “init_function_name”); if (fp == 0) Oops(other_plugin_AWOL) if ((error = (*fp)(vm))) return error; Fr

Re: [vpp-dev] undefined symbol: ip4_reass_node

Are you trying to send packets to ip4-reassembly? If so, ask the framework for a static and/or dynamic graph arc from to “ip4-reassembly”. You don’t need (or want) to link against the node registration to do that. D. From: vpp-dev@lists.fd.io On Behalf Of prashan...@gmail.com Sent: Thursday,

Re: [vpp-dev] undefined symbol: ip4_reass_node

Yes, that’s one way to do it... See also Ole’s comment... D. From: vpp-dev@lists.fd.io On Behalf Of prashan...@gmail.com Sent: Thursday, June 6, 2019 12:22 PM To: vpp-dev@lists.fd.io Subject: Re: [vpp-dev] undefined symbol: ip4_reass_node Thanks a lot Dave and Ole for your reply. I am new to V

Re: [vpp-dev] Approximate effective CPU utilization of VPP worker

The instantaneous vector size - in the stats segment, per-thread - is the best measure of how hard vpp is working. It's not a linear function of the offered load, but for any given offered load / feature set you can model it with three segments: * Dead asleep: vector size < 3, offered l

Re: [vpp-dev] Approximate effective CPU utilization of VPP worker

We've added purpose-built callback hooks in the main dispatch loop [in master/latest] so that folks can hook up whatever sort of instrumentation they like. Please have a look at that scheme. There's no point in attempting to upstream your work to stable/1801. We long since shut down the CSIT pe

Re: [vpp-dev] IPv6 packet total length

vlib_buffer_length_in_chain(vm, b0) with b0->current_data pointing to the ip6 header. From: vpp-dev@lists.fd.io On Behalf Of Esin Erenoglu Sent: Monday, June 10, 2019 5:16 AM To: vpp-dev@lists.fd.io Subject: [vpp-dev] IPv6 packet total length Hi everyone, I need total length of ipv6 packet. In

Re: [vpp-dev] IPv4 packet is dropped

If I had to bet: the FIB is not programmed correctly for the traffic that you're sending. "show error" or "trace add dpdk-input 100" - start pcap replay - "show trace". Please take a look at https://fdio-vpp.readthedocs.io/en/latest/troubleshooting/reportingissues/reportingissues.html. ___

Re: [vpp-dev] timer wheel usage queries

You didn't mention calling tw_timer_expire_timers_xxx(tw, now), which is the only force of physics which will actually cause a timer to expire. See also .../src/vppinfra/test_tw_timer.c, src/vlib/main.c or src/vnet/tcp/tcp.c for usage examples. The timer wheel code has been tested to N=60 milli

Re: [vpp-dev] IPv4 packet is dropped

Identify the input interface – “show interface” – identify the corresponding input node – “show run” - and then s/dpdk-input/actual-input-node-sourcing-traffic/ in the packet tracer “trace add” command. Please take the time to study available tools. D. From: Esin Erenoğlu Sent: Wednesday, Ju

Re: [vpp-dev] Vlib Buffer Size

How could anyone guess what’s wrong? No stack trace, no configuration, no access to the code you wrote, no idea what the plugin is supposed to do. We’d like to help, but you need to make it possible. See https://fdio-vpp.readthedocs.io/en/latest/troubleshooting/reportingissues/reportingissues.ht

Re: [vpp-dev] PerfMon Plugin Issue

It looks like you sent one ore more worker threads down a rabbit hole, to the point where they never make it back to the worker thread dispatch loop. Otherwise, the data collection would have stopped properly. I don’t know of any recent or broken commits to the performance monitor plugin. Pleas

Re: [vpp-dev] timer wheel usage queries

The only requirement is to call tw_timer_expire_timers_xxx (...) often enough so that timers expire accurately enough for correct operation. You can spin up an input node (as you describe), you can arrange to schedule a frame for an interior graph node, and so on. I would suggest placing sessio

Re: [vpp-dev] PerfMon Plugin Issue

Sounds like this should work. Please try to debug it and report results. From: vpp-dev@lists.fd.io On Behalf Of Satya Murthy Sent: Monday, June 17, 2019 3:10 AM To: vpp-dev@lists.fd.io Subject: Re: [vpp-dev] PerfMon Plugin Issue Hi Dave, Thanks for the reply. I am still seeing the same issue.

Re: [vpp-dev] timer wheel usage queries

Add a vector of tw_timer_wheel_xxx_t *’s to e.g. my_main_t. Vec_validate (...) so you have one wheel per thread. Initialize all of the wheels. Use e.g. vm->thread_index or os_get_cpu_index(..) to pick the wheel to process and/or use to start/stop timers. From: vpp-dev@lists.fd.io On Behalf Of

Re: [vpp-dev] VPP tap interface issue on Arm servers

Dear Lijan, The aarch64 development resources in the LF data center are in need of cleanup. I tried to use fdio-cavium5 @ 10.30.51.66 (Ubuntu1804), but I find that the login credentials have been changed. I finally managed to gain access to fdio-mcbin3 @ 10.30.51.43 (Ubuntu1604). It is in fact

Re: [vpp-dev] VPP tap interface issue on Arm servers

around for years so I will not he surprised that it is simply disabled in custom kernel built for mcbin. — Damjan On Jun 18, 2019, at 4:41 PM, Dave Barach via Lists.Fd.Io mailto:dbarach=cisco@lists.fd.io>> wrote: Dear Lijan, The aarch64 development resources in the LF data center are i

Re: [vpp-dev] tw_timer_first_expires_in_ticks problem

In real life - aka if you set now = vlib_time_now(vm) - the variable "now" would be the number of f64 seconds since vpp started, not the number of f64 seconds since 1/1/1970. Although you don't show the entire code, it looks like you built it into vpp instead of as a standalone test. In terms

Re: [vpp-dev] Vpp 19.04 (process showing 200Gig of virtual memory on VM)

Ack. I’ve asked folks to clean up a number of features which create [large] data structures only when features are enabled for the first time. That’s the fundamental problem. Virtual space not backed by physical pages costs very little, so it’s not a drop-everything come-running emergency. Wha

Re: [vpp-dev] Vpp 19.04 (process showing 200Gig of virtual memory on VM)

You’re right in terms of “ls -lh”, but remember that Linux is smart about this specific problem. gzip one of these core files, and the compressed file will probably consume 50mb. You will need 200gb to gunzip one of them. HTH... Dave From: on behalf of chetan bhasin Date: Tuesday, June 25, 2

Re: [vpp-dev] tw_timer_first_expires_in_ticks problem

Fix looks good, modulo a checkstyle failure which I fixed. Will merge when verified. From: on behalf of Andreas Schultz Date: Tuesday, June 25, 2019 at 9:27 AM To: Recipient Suppressed Cc: "vpp-dev@lists.fd.io" , "hagb...@gmail.com" , "winkelmann, holger" , "gumz, mathias" Subject: Re: [vp

Re: [vpp-dev] test-debug in the CI

The BIER tests cause ASSERT failures. I’ll have a quick look, then turn the problem over to Neale if I can’t solve it. D. From: vpp-dev@lists.fd.io On Behalf Of Paul Vinciguerra Sent: Tuesday, June 25, 2019 2:55 PM To: vpp-dev@lists.fd.io Subject: [vpp-dev] test-debug in the CI I set up a job

Re: [vpp-dev] test-debug in the CI

ilto:vpp-dev@lists.fd.io>" mailto:vpp-dev@lists.fd.io>> Objet : Re: [vpp-dev] test-debug in the CI This [1] seems to solve sctp issues with debug images locally. Florin [1] https://gerrit.fd.io/r/c/20337/ On Jun 25, 2019, at 12:10 PM, Dave Barach via Lists.Fd.Io mailto:dbarach=cisco@li

Re: [vpp-dev] VPP on POWER9

It worked a few years ago, at least up to a certain point. Given how long it's been since anyone tested on a ppc of any kind, one would have to invest a bunch of energy to get back to first base. FWIW... Dave -Original Message- From: vpp-dev@lists.fd.io On Behalf Of Eyle Brinkhuis Sen

Re: [vpp-dev] VPP on POWER9

hours to get it fixed... — Damjan > On 27 Jun 2019, at 16:41, Dave Barach via Lists.Fd.Io > wrote: > > It worked a few years ago, at least up to a certain point. Given how long > it's been since anyone tested on a ppc of any kind, one would have to invest > a bunch of

Re: [vpp-dev] Checking for enabled feature

Seems like your code is OK, see also vnet_interface_features_show(...) in .../src/vnet/feature/feature.c. We could add per-interface, per-feature arc "feature X is enabled" bitmaps to speed up this sort of processing, but there has to be a compelling reason to do so. What problem are you try

Re: [E] Re: [vpp-dev] VPP worker threads getting stuck

Please remember that the community actively supports the current release (19.04), and the prior release (19.01) on a best-effort basis. Please try 19.04 or master/latest. Thanks... Dave From: vpp-dev@lists.fd.io On Behalf Of Damjan Marion via Lists.Fd.Io Sent: Monday, July 1, 2019 7:58 AM To:

Re: [vpp-dev] centos jobs failing

I spun up a centos7 VM so I could investigate. * “make install-dep” works fine * “make install-ext-deps” complains about at the end of the build, prior to installing the .rpm. * The .rpm is built OK. * “rpm -i” installs the .rpm OK. With the vpp-ext-deps RPM installed man

Re: [vpp-dev] centos jobs failing

See https://gerrit.fd.io/r/#/c/20446/ - .../build/external/Makefile needed a bit of “|| true” action to avoid falling all over itself if no vpp-ext-deps .rpm was installed: ifneq ($(INSTALLED_RPM_VER),$(RPM_VER)-$(PKG_SUFFIX)) @$(MAKE) $(DEV_RPM) sudo rpm -e vpp-ext-deps || true

Re: [vpp-dev] SIGSEGV after calling vlib_get_frame_to_node

tps://github.com/travelping/vpp/blob/feature/master/upf%2Btdf/src/plugins/upf/upf_pfcp_server.c#L121 @Hugo: have you found the root cause for your problem? Regards Andreas Am Mi., 28. Nov. 2018 um 12:53 Uhr schrieb Dave Barach via Lists.Fd.Io<http://Lists.Fd.Io> mailto:cisco@lists.fd.io>>

Re: [vpp-dev] SIGSEGV after calling vlib_get_frame_to_node

s very similar to the one from Hugo. I'm also getting a invalid point from vlib_get_frame_to_node. It is crashing here: https://github.com/travelping/vpp/blob/feature/master/upf%2Btdf/src/plugins/upf/upf_pfcp_server.c#L121 @Hugo: have you found the root cause for your problem? Regards Andreas Am M

Re: [vpp-dev] SIGSEGV after calling vlib_get_frame_to_node

019 10:24 AM To: Dave Barach (dbarach) Cc: vpp-dev@lists.fd.io Subject: Re: [vpp-dev] SIGSEGV after calling vlib_get_frame_to_node Am Mi., 3. Juli 2019 um 15:55 Uhr schrieb Dave Barach via Lists.Fd.Io<http://Lists.Fd.Io> mailto:cisco@lists.fd.io>>: >>> vm->heap_a

Re: [vpp-dev] unsave tw_timer_expire + tw_timer_stop + tw_timer_start pattern

Thanks for highlighting this issue. Typically, the expired timer callback sets the timer_id to invalid which prevents confusion. I'll pass along the SCTP report. Dave -Original Message- From: vpp-dev@lists.fd.io On Behalf Of Andreas Schultz Sent: Thursday, July 4, 2019 8:21 AM To: vpp

Re: [E] Re: [vpp-dev] VPP worker threads getting stuck

Yes, it is a known issue: we do not support 18.07 at this point. Please see if you can repro w/ 19.04 or master/latest. From: vpp-dev@lists.fd.io On Behalf Of via Lists.Fd.Io Sent: Monday, July 1, 2019 6:56 AM To: Damjan Marion Cc: vpp-dev@lists.fd.io Subject: Re: [E] Re: [vpp-dev] VPP worker t

Re: [vpp-dev] vppctl makes VPP hang-up occasionally

The root cause is obvious from the backtrace in VPP-1711: memory allocator takes a spin-lock, timer interrupt fires, back into the memory allocator, deadlock. The [sic] timer interrupt function in .../src/vppinfra/timers.c is a mistake. You managed to find the only place in all of vpp which use

Re: [vpp-dev] vppctl makes VPP hang-up occasionally

://gerrit.fd.io/r/#/c/20573/ was merged earlier to resolve this issue. We are hoping Tatsumi is able to confirm this, but since it is a better approach regardless. Chris. From: vpp-dev@lists.fd.io<mailto:vpp-dev@lists.fd.io> mailto:vpp-dev@lists.fd.io>> On Behalf Of Dave Barach via List

Re: [vpp-dev] vppctl makes VPP hang-up occasionally

s a better approach regardless. Chris. From: vpp-dev@lists.fd.io<mailto:vpp-dev@lists.fd.io> mailto:vpp-dev@lists.fd.io>> On Behalf Of Dave Barach via Lists.Fd.Io Sent: Tuesday, July 9, 2019 08:06 To: Yusuke Tatsumi mailto:ytats...@yahoo-corp.jp>>; vpp-dev@lists.fd.io

Re: [vpp-dev] Drop vs Free?

I would suggest using error-drop unless the number of clock cycles saved by freeing packets directly somehow makes a first-order difference in your use-case. The "error-drop" scheme results in more obvious packet tracer output, supports N different drop causes by construction, and yields better

Re: [vpp-dev] bihash memory fragmentation

Try configuring approximately nbuckets = (nitems / BIHASH_KVP_PER_PAGE). Make sure that the hash function performs reasonably well on the particular keys involved. The scheme tolerates a mis-sized bucket array - lookup performance won't change, to a first approximation - but space utilization b

Re: [vpp-dev] vpp process stuck

Please fix your test client. It would be helpful to know exactly how and where the test client crashes, since it’s fairly hard to break the svm region mutex. Client-side code doesn’t hold that mutex while processing a message, on purpose to avoid issues of the form seen in the backtrace: whi

[vpp-dev] fd.io vpp spurious patch validation failures this AM

Here's an example. 07:41:52 Triggered by Gerrit: https://gerrit.fd.io/r/20739 07:41:52 ERROR: Issue with creating launcher for agent jenkins-fdda26a06e89c. The agent is being disconnected 07:41:52 [EnvInject] - Loading node environment variables. 07:41:52 ERROR: SEVERE ERROR occurs 07:41:52 org.j

Re: [vpp-dev] How to save VPP configuration

See https://fdio-vpp.readthedocs.io/en/latest/usecases/homegateway.html. Put the “vpp configuration” block into /etc/setup.gate. There’s nothing magic about the filename, except that /etc/vpp/startup.conf needs to point to it. D. From: vpp-dev@lists.fd.io On Behalf Of rya...@yunify.com Sent: F

[vpp-dev] Improper sw_if_index validation

Folks, The following code snippet is not sufficient to check a sw_if_index in a binary API message handler: hw = vnet_get_sup_hw_interface (vnm, ntohl(mp->sw_if_index)); if (hw == NULL || _device_class.index != hw->dev_class_index) return VNET_API_ERROR_INVALID_SW_IF_INDEX; If sw_if_ind

Re: [vpp-dev] To trigger a callback function in a regular x interval, looking for a sample timer API usage test program in VPP platform.

Need more detail: what is the interval x? 10 seconds, 10 microseconds, etc.? Is the code running in process context on thread 0, or in the data plane? D. From: vpp-dev@lists.fd.io On Behalf Of Ashish Prajapati Sent: Tuesday, July 23, 2019 7:50 AM To: vpp-dev@lists.fd.io Subject: [vpp-dev] To tr

Re: [vpp-dev] To trigger a callback function in a regular x interval, looking for a sample timer API usage test program in VPP platform.

Please https://fdio-vpp.readthedocs.io/en/latest/gettingstarted/developers/vlib.html, scroll down to Process / thread model and read through “Process events.” This shows how to start a lightweight (data-plane) thread which can wake up every N seconds and do something. This may not be the exact

Re: [vpp-dev] To trigger a callback function in a regular x interval, looking for a sample timer API usage test program in VPP platform.

Again, based on minimal information, here’s one possibility: Passive f64 expiration timer in the session object. Maintain a [per thread?] LRU list. See src/vppinfra/llist.h. Each time traffic hits a session, remove it from the LRU list, re-add it to the end of the LRU list, and set the expiratio

Re: [vpp-dev] node_funcs.h:61 (vlib_get_node) assertion `(i) < vec_len (vm->node_main.nodes)' fails #vpp

Backtrace, please. If you feed vlib_get_node(...) a bad node index – such as ~0 - it will ASSERT. Dave From: vpp-dev@lists.fd.io On Behalf Of prashan...@gmail.com Sent: Wednesday, July 24, 2019 9:07 AM To: vpp-dev@lists.fd.io Subject: [vpp-dev] node_funcs.h:61 (vlib_get_node) assertion `(i) <

Re: [vpp-dev] Time-consuming with 'show hardware-interfaces' command

Please submit a patch. From: vpp-dev@lists.fd.io On Behalf Of Lijian Zhang Sent: Wednesday, July 31, 2019 10:37 PM To: vpp-dev@lists.fd.io Cc: Juraj Linkeš ; Honnappa Nagarahalli ; nd Subject: [vpp-dev] Time-consuming with 'show hardware-interfaces' command Hi, It takes a long time to execute

Re: [vpp-dev] process node's thread

There's nothing inherently magic about thread-0, but one would have to drink a certain amount of Kool-Aid in terms of affecting worker-thread performance. Topics would include spinning up and servicing timer wheels, placing processes on threads, and coding carefully to avoid barrier-sync deadloc

Re: [vpp-dev] process node's thread

Dear Chris, You might look into using the [worker thread] main loop callback mechanism to schedule a vanilla interior graph node. See src/vlib/main.c:vlib_main_or_worker(...): if (PREDICT_FALSE (vec_len (vm->worker_thread_main_loop_callbacks))) clib_call_callbacks (vm->worke

Re: [vpp-dev] tracing packets of a specific 5-tuple

In your node, call vlib_trace_buffer (...) + vlib_add_trace(...) only if pkts match the N-tuple of interest. VLIB_BUFFER_TRACE_TRAJECTORY_INIT(...) is historical, and is only effective if you edit src/vlib/buffer.d and rebuild. No harm in adding the line of code. Packet tracing is not free. Th

Re: [vpp-dev] Is VPP vnet policer thread safe

Either protect the data structures with a clib_spin_lock, or switch to per-thread policer pools. The config pool might as well stay on the main thread, with it and the related name-to-index hash table protected by a spinlock. Patches welcome, after the 19.08 RC1 throttle-pull is complete. Pleas

Re: [vpp-dev] Can VPP detect illegal packet size?

Are you referring to checks of the form “ip-length <= L2 length,” which happen in .../src/vnet/ip/ip[46]_input.c? From: vpp-dev@lists.fd.io On Behalf Of kmp...@soosan.co.kr Sent: Friday, August 9, 2019 7:24 AM To: vpp-dev@lists.fd.io Subject: [vpp-dev] Can VPP detect illegal packet size? Hi

Re: [vpp-dev] query regarding memif

Please try w/ master/latest or stable/1908. From: vpp-dev@lists.fd.io On Behalf Of via Lists.Fd.Io Sent: Wednesday, August 7, 2019 9:22 AM To: vpp-dev@lists.fd.io Cc: vpp-dev@lists.fd.io Subject: [vpp-dev] query regarding memif Hi, I am using vpp version 18.01.01 and i am getting a crash in mem

Re: [vpp-dev] #vapi #vpp #vppcapi

Start vpp. Run vpp_api_test. Type help. From: vpp-dev@lists.fd.io On Behalf Of prashan...@gmail.com Sent: Monday, August 12, 2019 12:37 AM To: vpp-dev@lists.fd.io Subject: [vpp-dev] #vapi #vpp #vppcapi Have written a test plugin on vpp and trying to write a program on Linux control plane to co

Re: [vpp-dev] lockless architecture

What you’ve written is a best an oversimplification. It’s not correct to say that vpp lacks lockless data structures, and that all table-programming activities require coarse-grained locking and/or barrier synchronization. Rather than starting a discussion from first principles, would it be poss

Re: [vpp-dev] Crash when using dns_name_server

Four bits of config required: nat44 add identity mapping external GigabitEthernet3/0/0 udp 53053 binary-api dns_name_server_add_del 8.8.8.8 binary-api dns_enable_disable Inside network DHCP server needs to set option 6 (DNS name server) to the vpp gateway address. D. -Original Message-

Re: [vpp-dev] Crash when using dns_name_server

Folks, I’ll look at these issues. It would be helpful if people would contribute patches, or at least write Jira tickets. If we don’t know it’s broken, it won’t get fixed... To level-set: the DNS name resolver has been lightly used. Nothing would surprise me at this point. D. From: vpp-dev@l

Re: [vpp-dev] Crash when using dns_name_server

vs. unused_mumble_registration.node_index. Passing 0 when signaling name resolution events couldn’t possibly work. D. From: vpp-dev@lists.fd.io On Behalf Of Dave Barach via Lists.Fd.Io Sent: Thursday, August 15, 2019 2:54 PM To: anoopnairh...@gmail.com; vpp-dev@lists.fd.io Cc: vpp-dev@lists.fd

Re: [vpp-dev] Build errors

That setup will cause no end of pain if your git chops aren't up to it. I guess you've already discovered that. If you do things in a more normal way, you won't have to add tags or argue with the build system. A more straightforward way to work - which I use myself - goes like so: Create a loca

Re: [vpp-dev] Pcap trace doesn't work while using a seperate worker thread instead of main

Please submit a patch. From: vpp-dev@lists.fd.io On Behalf Of ?? Sent: Friday, August 16, 2019 8:19 AM To: vpp-dev@lists.fd.io Subject: [vpp-dev] Pcap trace doesn't work while using a seperate worker thread instead of main Dear Developers: If I use just one thread for configure and packet proc

Re: [vpp-dev] Crash when using dns_name_server

I am trying to use DNS server and on "ping google.com" VPP is crashing Color me unsurprised. Please fix your configuration, to the point where first-order obvious errors have been eliminated. It looks to me like the "binary-api dns_name_server_add_del 8.8.8.8" command has been truncated: Aug 1

Re: [vpp-dev] Crash when using dns_name_server

bridge GigabitEthernet3/0/0 1 comment { dhcp server listens on this interface } create tap id 0 host-ip4-addr 10.155.6.2/24<http://10.155.6.2/24> host-if-name novlan set int l2 bridge tap0 1 set int state tap0 up nat44 add interface address TenGigabitEthernet8/0/0 set interface

Re: [vpp-dev] VPP 19.08 release is available!

+1 congrats to Andrew and to the entire community on a timely release. Note that 19.08 will be supported for a year, so please remember that [non-API affecting] cherry-picks will be required. Otherwise, Andrew will be very busy pinging people... 😉... Dave From: vpp-dev@lists.fd.io On Behalf O

Re: [vpp-dev] VPP 19.08 release is available!

+1... Dave From: vpp-dev@lists.fd.io On Behalf Of Florin Coras Sent: Thursday, August 22, 2019 12:05 PM To: Andrew Yourtchenko Cc: vpp-dev ; csit-dev Subject: Re: [vpp-dev] VPP 19.08 release is available! Congrats to the entire community and thanks Andrew! Cheers, Florin > On Aug 21, 2019, a

Re: [vpp-dev] Crash when using dns_name_server

set int l2 bridge tap0 1 set int state tap0 up nat44 add interface address TenGigabitEthernet8/0/0 set interface nat44 in loop0 out TenGigabitEthernet8/0/0 - dnsmasq.conf # Gateway dhcp-option=novlan,3,10.155.6.1 # DNS server dhcp-option=novlan,6,10.155.6.1 ---

Re: [vpp-dev] Crash when using dns_name_server

reate tap id 0 host-ip4-addr 10.155.6.2/24<http://10.155.6.2/24> host-if-name novlan set int l2 bridge tap0 1 set int state tap0 up nat44 add interface address TenGigabitEthernet8/0/0 set interface nat44 in loop0 out TenGigabitEthernet8/0/0 - dnsmasq.conf # Gateway

Re: [vpp-dev] Build fails when cloning from gitlab. #vpp

If you commit and push build artifacts to your local gitlab, clone the resulting repo and attempt to build, it’s quite likely to fail as described. I use a local mirror of gerrit.fd.io myself. It works perfectly. Google / read docs for instructions. Anyhow, this subject is off-topic for vpp-dev

Re: [vpp-dev] How to use valgrind to check for memory errors in vpp?

Dlmalloc [aka "Doug Lea Malloc"] is a lightly modified copy of the allocator described here: http://gee.cs.oswego.edu/dl/html/malloc.html. If you've managed to find an issue in it, please share the details. Until proven otherwise, I suspect the report rather than dlmalloc itself. Vpp does indee

Re: [vpp-dev] Support for shared subnet

This is standard router behavior. Multiple interfaces on the same subnet are not allowed. Please explain the use case in detail. Otherwise, we won’t be able to help. D. From: vpp-dev@lists.fd.io On Behalf Of krishnamurthy.m...@gmail.com Sent: Tuesday, September 10, 2019 10:56 AM To: vpp-dev@l

Re: [vpp-dev] Support for shared subnet

+1, see https://fdio-vpp.readthedocs.io/en/latest/usecases/homegateway.html for an IRB configuration which may be what's needed... Dave -Original Message- From: Andrew 👽 Yourtchenko Sent: Tuesday, September 10, 2019 1:41 PM To: Burt Silverman Cc: Christian Hopps ; Dave Barach (dbarach

Re: [vpp-dev] Crash in qsort

As far as I know, there are no issues with vec_sort_with_function. It’s been unchanged for many years. We use it in lots of places, and it does not give trouble. Unless someone feeds it a dangling vector reference, or writes a broken element comparison function. If after taking a good hard loo

Re: [vpp-dev] Assert verifying node graph construction around ip4-arp.

Sounds like one of the threads hasn't finished rebuilding its graph replica after you crank up the ipsec tunnel. That operation occurs N-way parallel, with each thread rebuilding its own replica. Take a look at [aka ASSERT] *vlib_worker_threads->node_reforks_required == 0 at the moment when you

Re: [vpp-dev] Can vlib_time_now be used in worker threads

Each thread has its own vlib_main_t. From: vpp-dev@lists.fd.io On Behalf Of Vamsi Krishna Sent: Thursday, September 19, 2019 7:51 AM To: vpp-dev@lists.fd.io Subject: Re: [vpp-dev] Can vlib_time_now be used in worker threads Hi Damjan, I was going through the following code which vlib_time_now c

Re: [vpp-dev] [csit-dev] API check failure

Yes. Since folks have been unable to debug the problem so far, we need to disable voting on the API CRC checker job. Ed Kern [or anyone with appropriate access rights!]: please do that immediately, and send email to vpp-dev@lists.fd.io so folks can "recheck" patches which failed for no good reaso

Re: [vpp-dev] About VPP trace，why I trace add ip4_lookup, but when i show ,it does't have any record?

One can call vlib_trace_buffer(...) anywhere, but doing so on a packet which is already traced - from one of the input nodes - has a peculiar "fork" effect which may prove confusing. As Chris wrote: [we hope that] all device input nodes are capable of calling vlib_trace_buffer. Few if any inter

Re: [vpp-dev] [csit-dev] API check failure

I rebased one of my failing patches, and the API crc check passed. Thanks for your help! Dave 08:54:51 +++ export PYTHONPATH=/w/workspace/vpp-csit-verify-api-crc-master/csit 08:54:51 +++ PYTHONPATH=/w/workspace/vpp-csit-verify-api-crc-master/csit 08:54:51 +++ make json-api-files 08:54:51 /w/wor

Re: [vpp-dev] About VPP trace，why I trace add ip4_lookup, but when i show ,it does't have any record?

No it isn’t useless. The trace property is inherited. From: vpp-dev@lists.fd.io On Behalf Of wei_sky2...@163.com Sent: Sunday, September 22, 2019 9:11 PM To: vpp-dev@lists.fd.io Subject: Re: [vpp-dev] About VPP trace，why I trace add ip4_lookup, but when i show ,it does't have any record? [Edit

Re: [vpp-dev] Implementation of a hash table that can be both written and read by all VPP workers #vpp

Start with bihash and see how much serialization you encounter in practice. If your use case will write the hash table anywhere near as often as it reads the hash table, I wouldn’t think that’s a good sign... D. From: vpp-dev@lists.fd.io On Behalf Of krishnamurthy.m...@gmail.com Sent: Friday,

Re: [vpp-dev] Need help to configure NAT44, reply packets not working

I'd suggest using the packet tracer to see what happens to the return-path traffic. "trace add dpdk-input 100", send a ping, then "show trace". HTH... Dave From: vpp-dev@lists.fd.io On Behalf Of ?? Sent: Monday, September 30, 2019 3:43 AM To: Vpp-dev@lists.fd.io Subject: [vpp-dev] Need

Re: [vpp-dev] test_srmpls.py and test_ip_mcast.py fail under "make test-debug"

Dear Ben, Looping in Neale. The failure is 100% reproducible, due to the highlighted ASSERT. According to gdb, fib_table->ft_total_route_counts is 1 not zero. HTH... Dave void ip6_fib_table_destroy (u32 fib_index) { /* * all link local first ... */ fib_prefix_t

[vpp-dev] Jenkins queue stuck / patch validation halted

Linux Foundation sysadmin alerted. -=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#14142): https://lists.fd.io/g/vpp-dev/message/14142 Mute This Topic: https://lists.fd.io/mt/34441744/21656 Group Owner: vpp-dev+ow...@lists.fd.io Unsubscribe: https://

Re: [vpp-dev] What is the best way to pass packets from an internal node to a process node?

It would help to describe the problem at a higher level. First question: why pass packets to a process node? From: vpp-dev@lists.fd.io On Behalf Of Jacques Samain via Lists.Fd.Io Sent: Monday, October 14, 2019 11:02 AM To: vpp-dev@lists.fd.io Cc: vpp-dev@lists.fd.io Subject: [vpp-dev] What is t

Re: [vpp-dev] What is the best way to pass packets from an internal node to a process node?

At that data rate, I’d consider using [a per-thread variation of?] ../src/vppinfra/mpcap.[ch]. Push/pop the buffer length to truncate packets after the L3 header across the log operation. If the spinlock turns out to be a significant serialization point, use one mpcap log per worker thread. Sin

Re: [vpp-dev] What is the best way to pass packets from an internal node to a process node?

Perfect, glad it worked out for you... Dave From: Jacques Samain (jsamain) Sent: Wednesday, October 16, 2019 11:08 AM To: Dave Barach (dbarach) ; vpp-dev@lists.fd.io Subject: Re: What is the best way to pass packets from an internal node to a process node? Hello Dave, I ended up with one mpcap

  1   2   3   4   5   6   >