Jan Kiszka wrote: > [EMAIL PROTECTED] wrote: >> Jan Kiszka wrote: >>> Tom Panis wrote: >>>> Hi, >>>> >>>> I'm trying to send raw ethernet packages on Xenomai, I'm also working >>>> with the trunk version of RTnet. I looked at the example >>>> \"raw-ethernet.c\" and I'm wondering why the functions like socket_rt, >>>> send_rt ,.. aren't used. Is it still real-time? I tried to use the >>> It is, but only when you link against Xenomai's POSIX skin (xeno-config >>> --posix-ldflags). This step will wrap all POSIX calls like socket, send, >>> bind to RTDM services, finally invoking RTnet if it is installed >>> (otherwise you will end up in Linux again). >>> >>>> real-time functions like socket_rt,... but then I can't send or receive >>>> raw ethernet packages anymore: the function send_rt returns -1. >>> All _rt services are deprecated now (we should mark them...). They were >>> replaced by RTDM's rt_dev_XXX. But only the name changed here. >>> >>> The difference between rt_dev_XXX (or old XXX_rt) services and the true >>> POSIX calls are the way error codes are returned. For POSIX services you >>> have to check errno, for the RTDM API you get the error directly via the >>> return code. Thus, one time -1 only means some error occurred, the other >>> it means -EPERM (here: RT context required). >>> >>>> I attached the program code. Running the program (I use rtlo as >>>> interface) results in the following output: >>>> >>>> Socket created: socket id: 0 >>>> Got interface: index: 0 >>>> Sending frames... >>>> Cannot send: Success >>>> Length: -1 >>>> shutting down >>> How did you compile that program? It doesn't map itself to RT context >>> unless you compiled it against the POSIX skin (which does so >>> automatically for main). >> I didn't compiled it against the POSIX skin. Now I just edited >> raw-ethertnet.c and kept it in the same directory. I compile it with >> makefile that's already there. When I do this I can send messages, but I >> can't receive any. When rt_dev_recv() is called, the program hangs, he >> keeps waiting for a package to arrive. When I send messages to my computer >> without this program running dmesg results : >> [4299353.146000] RTnet: no one cared for packet with layer 3 protocol >> type 0x1234 >> [4299354.150000] RTnet: no one cared for packet with layer 3 protocol >> type 0x1234 >> .... >> When my program runs, these messages aren't given anymore, so the packages >> are taking care of I guess... >> I added the source. The output of the program is: >> >> Socket created: socket id: 1 >> Got interface: index: 0 >> Receiving >> >> After stopping the program with ctrl+c the following is added: >> >> Length: -9 >> shutting down > > Your test case looks sane on first sight. You ran it against the > original raw-ethernet example as sender? Over a cable or also via local > loopback? Does loopback also fail? > > I cannot promise when I'll have time to look at this. Meanwhile you > could try to track a bit down what happens in the core, from the stack > manager to the af_packet code, by adding printks. >
Ok, effect confirmed over loopback here in qemu, and also with the original raw-packets example of RTnet. Obviously a regression of recent changes to the packet socket module. Investigating... Jan
signature.asc
Description: OpenPGP digital signature
------------------------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________ RTnet-users mailing list RTnet-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rtnet-users
