- From: <vgarrin_at_free.fr>
- Date: Wed, 14 Feb 2007 13:51:09 +0100
Selon Tabish Mustufa <tmustufa_at_jhu.edu>: > > I have read docs, but can't find really up 2 date explanations. and also > > have a quick look at some code in rt.c since comedi_request_irq looks to > > be defined there, but this seems to use request_irq, not rt_request_irq > > ???? > > There are a couple versions of comedi_request_irq in rt.c. Which one ends > up being used depends on which realtime package comedi gets linked > against. > > To my recollection, comedi tries to get an unshared IRQ using request_irq > before calling rt_request_irq. This will fail if any other device is > using the same IRQ line. If it succeeds, you'll get an entry in > /proc/interrupts for your device that always shows zero interrupts (since > the IRQ's never get to the Linux domain). Thanks for your reply, Well, this is what i get!! my understanding is that comedi_request_irq just try to install a linux interrupt handler but do not call rt_request_irq. What I can see is that if the TRIG_RT flag is set, do_cmd_ioctl will call comedi_switch_to_rt which will then call rt_request_irq. The good news here is that the way interrupts from my card are handled are "hard real-time" interrupts. The bad news is how can i get this data if i am in a real time userland task ? My understanding is that a syscall to read will bring me back to linux so no longer real time ? > I don't know if that helps, but for what it's worth, I've been having > trouble with RTAI interrupts in a driver for some custom in-house hardware > at my employer. Everything works fine under RTAI 3.3, but I tried > upgrading one of our computers to 3.4 and the computer hard-locks on > interrupts. I haven't had a chance to figure out what's wrong yet... What still looks surprising is that "read" hangs if i do not specify TRIG_RT flag, in the cmd flags. What does really happens ?? the "linux" interrupt handler is set, so the interrupt should trigger this handler ??!! Thanks, > > -Tabish > > _______________________________________________ > comedi mailing list > comedi_at_comedi.org > https://cvs.comedi.org/cgi-bin/mailman/listinfo/comedi >
Received on 2007-02-14Z12:51:09