Nsis_Imp February 2006

nsis_imp@informatik.uni-goettingen.de

5 participants
8 discussions

Re: [Nsis_imp] RMD with QoS-NSLP over GIST
by Christian Dickmann 15 Feb '06

15 Feb '06

Hi, > Thank you for your answer, it has proven most helpful. > However, I have a new problem: how can I use GIST with > RMD-QoS-NSLP in a stateless domain. > I mean: > 1. between Ingress and Egress stateful edge nodes, normal > qos-nslp operation is sufficient with the following: > - signaling messages sent between edge nodes over TCP > (but intra-domain stateless nodes are not processing > these edge2edge messages) > 2. for the intra-domain nodes: > - signaling is sent over UDP (and edge nodes are also > processing these intra-domain messages) > I believe TCP/UDP choice can be solved with the reliable flag > of the SendMessage API call, but I only have ideas regarding > the local message processing or no-processing in the > stateless intra-domain nodes. I do not have a deep knowledge about QoS NSLP, so I am not sure what you mean. With stateless, to which state do you refer? GIST state? QoS state? Any other? The choice between TCP/UDP is done by the reliable flag of the SendMessage API call as you said. > I read in the drafts I shall use the RAO value/NSLP-ID to > determine what message shall be passed on without processing, > and what shall be passed up to the QoS-NSLP for processing. > ( Shall I solve this by checking the NSLP-ID value at every > node when receiving message from GIST to decide to > process/pass on? ) > Could you specify how can I achieve this? Is there an option > in either the SendMessage/RecvMessage API call, or the mri > to indicate this to GIST? Which part of NSIS do you try to implement? Our implementation uses only one RAO because linux (at least some version) seems to react to RAO value 0 only. So we fully rely on the NSLP ID. QoS NSLP has one NSLP ID and for example NATFW has another. When a message is intercepted by a GIST router because of the RAO, the NSLP ID is checked and the message is bypassed if their is no NSLP currently connected that handles that specific NSLP ID. So this behaviour is completely hidden from the NSLP ID. Did I get you right or did you mean something different? > Perhaps it is not a GIST-relevant question, but can I configure > a linux-box to only handle packets with a specified RAO value? > (The one I hope I can force GIST to use). You do not need to tell linux this one (or do you want to filter?). The application opening the RAW socket tells linux to which RAO it should react. Other RAOs are simply forwarded like a normal packet. I hope this mail helps a little bit. Otherwise please explain a bit more, what you are trying to do and maybe why ;) Christian Dickmann

1 0

RMD with QoS-NSLP over GIST
by Ferenc Pintér (IJ/ETH) 15 Feb '06

15 Feb '06

Hi all! Thank you for your answer, it has proven most helpful. However, I have a new problem: how can I use GIST with RMD-QoS-NSLP in a stateless domain. I mean: 1. between Ingress and Egress stateful edge nodes, normal qos-nslp operation is sufficient with the following: - signaling messages sent between edge nodes over TCP (but intra-domain stateless nodes are not processing these edge2edge messages) 2. for the intra-domain nodes: - signaling is sent over UDP (and edge nodes are also processing these intra-domain messages) I believe TCP/UDP choice can be solved with the reliable flag of the SendMessage API call, but I only have ideas regarding the local message processing or no-processing in the stateless intra-domain nodes. I read in the drafts I shall use the RAO value/NSLP-ID to determine what message shall be passed on without processing, and what shall be passed up to the QoS-NSLP for processing. ( Shall I solve this by checking the NSLP-ID value at every node when receiving message from GIST to decide to process/pass on? ) Could you specify how can I achieve this? Is there an option in either the SendMessage/RecvMessage API call, or the mri to indicate this to GIST? Perhaps it is not a GIST-relevant question, but can I configure a linux-box to only handle packets with a specified RAO value? (The one I hope I can force GIST to use). Thank you for your help, Ferenc Pintér

1 0

Access forbiden to the archives list
by Jean-Noel Chardron 14 Feb '06

14 Feb '06

Hello, when I try to read the archive list of https://user.informatik.uni-goettingen.de/mailman/listinfo/nsis_imp [ To see the collection of prior postings to the list, visit the Nsis_Imp Archives <https://user.informatik.uni-goettingen.de/pipermail/nsis_imp/>. -> https://user.informatik.uni-goettingen.de/pipermail/nsis_imp/ ] I get an HTTP 403 Error : Forbidden You don't have permission to access /pipermail/nsis_imp/ on this server. ------------------------------------------------------------------------ Apache/1.3.33 Ben-SSL/1.55 Server at mail.ifi.informatik.uni-goettingen.de Port 443 is there a security option that I can try to browse the archive list ? Thanks, -- Jean-noel Chardron chardron(a)labri.fr Jean-Noel.Chardron(a)dr15.cnrs.fr

1 0

Documentation and UML diagram of GIST
by Chardron 13 Feb '06

13 Feb '06

This message is written in english and then in french Le même message est rédigé en anglais puis en français Hello, I'm currently working in the course of my study on the negociation of service level (SLS/SLA). the aim of this study is to write the implementation of an application SLS-NSLP in the NSIS framework. I suppose that I can develop it upon the GIST's implementation made by the team working at the University of Gottingen in C++ or the University of Coimbra in Java or may be (why not) doing a new implementation of Gist in Ada and therefore a Qos-NSLP and SLS-NSLP upon this new NTLP implementation. In all case, access to the documentation and the diagramm of Class (if possible in UML) of actual implementation of Gist and more specially the NSLP-API will be a great help for me. Could I expect get it from you, please. I'm looking for your answer. If you need more precise information on my project, please, do not hesitate to contact me. Thank you very much, Bonjour, Je travaille actuellement dans le cadre de mon projet d'étude sur la négociation de niveau de service (SLS). Comme point d"échéance je pense à implémenter une application NSLP de SLS s'intégrant dans le plan de travail de NSIS. Il est possible que je développe cette application au dessus de l'implementation de la couche transport GIST développé par l'université de göttingen en C++ ou par l'université de Coimbra en Java ou bien même -pourquoi pas- faire une autre implémentation de GIST en Ada et par la suite écrire sur celle-ci l'application QoS-NSLP et SLS-NSLP. Dans tous les cas, avoir accès à votre documenation et au diagramme de Classe (si possible en UML) des différentes implémentations de GIST (et en particulier de l'API NSLP) serait d'une trés grande aide pour moi. Vous serait-il possible de me communiquer cette information. Pour plus de précision sur mon projet je reste à votre disposition Merci beaucoup, -- Jean-Noël Chardron chardron(a)labri.fr

4 3

Re: [Nsis_imp] NSLP-GIST interface question
by Christian Dickmann 13 Feb '06

13 Feb '06

Hi, the source of your problem is most likely your nsis.conf. Please note, that the network parameters need to be interpreted like in a routing table. So you specify an address for a specific destination network. The example provided in our release contains a default route, which might help. Another possibility is that your code stores ip addresses in host byte order instead of network byte order. This would result in a failure when doing a lookup in the configuration table. I recommend two/three steps: 1. Use the PING NSLP to check your configuration. If that works your configuration is OK and the problem is in your code. 2. Use our ethereal dissector to check your packets. You find an old version at: http://user.informatik.uni-goettingen.de/~nsis/release/ndiss/ If you need a more recent version I can provide one on request. Using the dissector you see the IP addresses used in the MRI. 3. If you can't or do not want to use the ethereal dissector, you should start nsis with # ./nsis -debug 3 or -debug 2. Doing so, you see a debug output of the GIST packets going over the wire. This way, so can check the IP addresses. Of couse, this method is not as reliable as the dissector and not so comfortable. I hope this helps. If not, please explain the results of the steps listed above and show your nsis.conf of all machines. > if using msg_routing_info_base_t* finally, why fill parts of > msg_routing_info_path_coupled_t ? > > As I saw, it only uses srcAddr, destAddr, ipVersion, and > msgRoutingMethod. > (anyway, what is the difference between these msg_routing methods?) You should never create a datastructure of msg_routing_info_base_t directly. Always create msg_routing_info_path_coupled_t or loose end. The base type is simply a type used in pointers that are not specific to a certain message routing method (MRM). Think of it as an "abstract base class", although it is a struct and not a class. Our API is generic and so needs a pointer to msg_routing_info_base_t. Using the MRM field in the msg_routing_info_base_t it determines the REAL type internally. The GIST draft explains the difference between the MRMs. In general the path-coupled MRM is used, but currently NATFW NSLP defines one case where you need to use the loose-end MRM. For QoS you should stick to the path-coupled one. Christian Dickmann

1 0

Re: [Nsis_imp] NSLP-GIST interface question
by Christian Dickmann 13 Feb '06

13 Feb '06

I will answer this message as soon as I am back from lunch. Christian Dickmann

1 0

NSLP-GIST interface question
by Ferenc Pintér (IJ/ETH) 13 Feb '06

13 Feb '06

Hi all! I'm nearing to make my QoS-NSLP code work with the one of GIST, but I've ran into an issue I do not seem to be able to resolve. I get my RecvMsg callback executed fine when sending a sample Resv from another node running the GIST qos_server and qos_client, like this: Node A <-- Node B running: running: ./nsis ./nsis ./nsis_qos_nslp ./qos_server (my code) ./qos_client RESERVE 10.10.2.2 2 0 RII REP Any time I wish to send something from my code, GIST apologies with this: (I issue commands within the code instead of passing cmd line args, listening to stdin with select() next to the unix sockets) Node A --> Node B running: running: ./nsis ./nsis ./nsis_qos_nslp ./qos_server (my code) ./qos_client RESERVE 10.10.2.2 2 0 RII REP or Node A --> Node B running: running: ./nsis ./nsis ./nsis_qos_nslp ./nsis_qos_nslp (my code) (my code) -> GistException: "No local IP address found for this outgoing message" /Can this message occur because I'm using stdin() as well? / I believe it is originating from: //in: GistFsmBase.cpp /** * Generate Network Layer Information object for local maschine * Warning: Object has to be deleted after use. */ GistNetworkLayerInfo * GistFsmBase::generateNetworkLayerInfo(char * ident, int identLength, IPaddr destIP, bool preferIPv6) { GistNetworkLayerInfo *netlayerInfo; netlayerInfo = new GistNetworkLayerInfo(); netlayerInfo->setPeerIdentity(ident, identLength); IPaddr addr; addr.version = 0; /* IPaddr addr = GistConfiguration::general_localIP; if (preferIPv6) { addr.version = 6; }; */ if (!preferIPv6 && destIP.version == 4 && GistConfiguration::general_localIPv4count > 0) { for (int i = 0; i < GistConfiguration::general_localIPv4count; i++) { unsigned int mask = GistConfiguration::general_localIPv4Nets[i].mask; unsigned int tmpAddr = GistConfiguration::general_localIPv4Nets[i].net; if ((destIP.v4 & mask) == (tmpAddr & mask)) { addr = GistConfiguration::general_localIPv4s[i]; } } } if (preferIPv6 && destIP.version == 6 && GistConfiguration::general_localIPv6count > 0) { for (int i = 0; i < GistConfiguration::general_localIPv6count; i++) { unsigned char * mask = GistConfiguration::general_localIPv6Nets[i].mask; unsigned char * tmpAddr = GistConfiguration::general_localIPv6Nets[i].net; bool equal = true; for (int j = 0; j < 16; j++) { if ((destIP.v6[j] & mask[j]) != (tmpAddr[j] & mask[j])) { equal = false; break; }; }; if (equal) { addr = GistConfiguration::general_localIPv6s[i]; } } } if (addr.version == 0) { ------->> throw new GistException("No local IP address found for this outgoing message"); }; netlayerInfo->setInterfaceAddr(addr); netlayerInfo->setIpTTL(64); netlayerInfo->setRouteStateValidityTime(CONFIG_GIST_MRS_VALIDITY_TIME); return netlayerInfo; } I've checked my SendMessage(...) code, it seems to be fine. The nsis.conf and ifconfig and route configurations are the same. I've also traced the contents of each structure, etc. I gave to this call. Cheking the GIST code, I seached where you call SendMessage, and found: //in: QosFsmQni.cpp void QosFsmQni::sendMsg(QoSMessage *msg, void *args) { QoSFsmData *data = (QoSFsmData *)args; msg_routing_info_path_coupled_t pc_mri; memset(&pc_mri, 0, sizeof(msg_routing_info_path_coupled_t)); pc_mri.msgRoutingMethod = GIST_DEFAULT_MRM; pc_mri.ipVersion = 4; // TODO: Use the _right_ outgoing address (christian.dickmann) pc_mri.srcAddr.v4.addr = data->server->localIPv4List[0].v4; pc_mri.destAddr.v4.addr = data->addr.v4; pc_mri.srcPrefix = 32; pc_mri.destPrefix = 32; pc_mri.d = DOWNSTREAM; pc_mri.p = 1; pc_mri.a = 1; pc_mri.b = 1; printf("srcAddr : %s\n", IPaddr::printIP(data->server->localIPv4List[0].v4)); printf("destAddr: %s\n", IPaddr::printIP(data->addr.v4)); printf ("msg length = %i\n",msg->getLength()); ----->> data->server->api->SendMessage((char *)msg->getBuffer(), msg->getLength(), NULL, QOS_NSLP_ID, (unsigned char *)data->sid, (msg_routing_info_base_t *)&pc_mri, NULL, NULL, true, false, false, 2000, 10); } and have few questions about this: -> if using msg_routing_info_base_t* finally, why fill parts of msg_routing_info_path_coupled_t ? As I saw, it only uses srcAddr, destAddr, ipVersion, and msgRoutingMethod. (anyway, what is the difference between these msg_routing methods?) Thank you for your help and explanation, Ferenc Pintér (please reply to this letter, as I am not on the list) ( I tried to check the lists' archives without subscription, but got a forbidden error)

1 0

Re: [Nsis_imp] Documentation and UML diagram of GIST
by Henning Peters 12 Feb '06

12 Feb '06

Hi Jean-Noël, you're welcome. please don't forget to post/reply to the list, so other team members can join discussion freely: nsis_imp(a)informatik.uni-goettingen.de Henning Chardron sagte So, 12.02.2006, 22:20: > Hi Henning, > > Thank you so much for your fast answer, > I will follow your indication. I have already had a look on the QoS-NSLP > implementation, I found it quite complexe. but now I will restart with > the ping daemon, it would be a good starting point. > I'll tell you more on my project soon. > bye for now have a nice evening > > Best regards, > > Jean-Noël > > > Henning Peters wrote: > >>Hi Jean-Noël, >> >>unfortunately, we can't provide you an UML diagram as we don't have one >>created, yet - and I think we don't plan to do so. But I don't see this >> as >>a problem. It's fairly easy to start off with an NSLP using already >>existing code and modify it for your needs. The API is simple (and >>documented in GIST I-D) and there are also several sample NSLP >> (partially) >>implemented, including NAT/Firewall NSLP, QoS NSLP and Ping/Diag NSLP. >>There is also some documentation included that is worth reading. >> >>Please take a dive into the code (maybe start with a simple NSLP like >>Ping), especially the /nslp directory might be interesting for you as it >>contains the code of various NSLP daemons. GIST is in a pretty stable >>state, whereas the NSLP are less robust. Ping and NAT/Firewall are the >>most robust ones. >> >>The folks on this list will be happy to help you with your issues. Thank >>you for considering our implementation as a base for your work. >>Personally, I recommend you to try existing implementations before >>thinking coding GIST again. You might save a lot of time and energy as >> the >>NSLP are usually much more leightweight than NTLP/GIST. >> >>Independent from this decision, please consider also using our Ethereal >>dissectors. These are included in the distribution and have proven to be >>irreplaceable during development. Also, creating your own NSLP dissector >>will be probably easy taking another NSLP dissector as code base. >> >>Maybe you can give a short outline of your ideas. That would be not only >>interesting for us, but we could assist you with your future plans. >> >>Cheers, >>Henning >> >> >>Chardron sagte So, 12.02.2006, 13:41: >> >> >>>This message is written in english and then in french >>>Le même message est rédigé en anglais puis en français >>> >>>Hello, >>> >>>I'm currently working in the course of my study on the negociation of >>>service level (SLS/SLA). the aim of this study is to write the >>>implementation of an application SLS-NSLP in the NSIS framework. >>>I suppose that I can develop it upon the GIST's implementation made by >>>the team working at the University of Gottingen in C++ or the >>>University of Coimbra in Java or may be (why not) doing a new >>>implementation of Gist in Ada and therefore a Qos-NSLP and SLS-NSLP upon >>>this new NTLP implementation. >>>In all case, access to the documentation and the diagramm of Class (if >>>possible in UML) of actual implementation of Gist and more specially the >>>NSLP-API will be a great help for me. >>>Could I expect get it from you, please. I'm looking for your answer. >>>If you need more precise information on my project, please, do not >>>hesitate to contact me. >>>Thank you very much, >>> >>> >>>Bonjour, >>> >>>Je travaille actuellement dans le cadre de mon projet d'étude sur la >>>négociation de niveau de service (SLS). Comme point d"échéance je pense >>>à implémenter une application NSLP de SLS s'intégrant dans le plan de >>>travail de NSIS. >>>Il est possible que je développe cette application au dessus de >>>l'implementation de la couche transport GIST développé par l'université >>>de göttingen en C++ ou par l'université de Coimbra en Java ou bien même >>>-pourquoi pas- faire une autre implémentation de GIST en Ada et par la >>>suite écrire sur celle-ci l'application QoS-NSLP et SLS-NSLP. >>>Dans tous les cas, avoir accès à votre documenation et au diagramme de >>>Classe (si possible en UML) des différentes implémentations de GIST (et >>>en particulier de l'API NSLP) serait d'une trés grande aide pour moi. >>> >>>Vous serait-il possible de me communiquer cette information. >>>Pour plus de précision sur mon projet je reste à votre disposition >>> >>>Merci beaucoup, >>> >>>-- >>>Jean-Noël Chardron >>>chardron(a)labri.fr >>> >>>_______________________________________________ >>>Nsis_Imp mailing list >>>Nsis_Imp(a)informatik.uni-goettingen.de >>>https://user.informatik.uni-goettingen.de/mailman/listinfo/nsis_imp >>> >>> >>> >>> >> >> >> >> >> > >

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Nsis_Imp February 2006