Messages in this thread | | | Date | Tue, 22 Sep 1998 23:54:29 +1000 | From | Richard Gooch <> | Subject | Re: Today Linus redesigns the networking driver interface (was Re: tulip driver in ...) |
| |
Alan Cox writes: > > But if your interrupt handler doesn't register/mark a BH, then the > > costs of "BH processing" (i.e. (bh_active & bh_mask)) are > > trivial. Where is the gain in that extra flag? > > Probably small. However > > 1. It also does signal processing and rescheduling on the return path > (the latter is now a lot cheaper with current->need_resched on smp)
That's a separate issue from BH processing, though.
> 2. You often have woken a task but dont want it to leap into life > instantly
I've recently looked at this, and if you wake an RT process then current->need_resched is set to 1 anyway. Same story if the woken process has a higher dynamic priority than current. So in fact you can't avoid the context switch anyway.
> > >From looking at the current Linux capabilities, it looks to me like we > > can indeed give hard-RT performance. Sure, it may mean not using > > broken 8390 drivers which globally disable interrupts while spending > > 1.6 ms reading a packet, but hey, we can live with that ;-) > > I've been working on two things where hard real time is an important > matter. That is localtalk cards and dumb synchronous cards. The former > requires I hit 90uS or so worst case latency on receive, the latter to > do 2Mbits is even tighter. > > > Being able to do all this with normal Linux without having to resort > > to RTLinux is a goal worthy of pursuit. > > No argument ;)
:-)
Regards,
Richard....
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |