Messages in this thread | | | Date | Sun, 25 Oct 1998 22:14:57 -0500 (EST) | From | Kenneth Albanowski <> | Subject | Re: PCI Modem Support |
| |
On Sun, 25 Oct 1998, Benjamin Herrenschmidt wrote:
> On Sun, Oct 25, 1998, Vojtech Pavlik <vojtech-lists@twilight.ucw.cz> wrote: > > >Thus writing a winmodem driver should be possible even on normal Linux, > >while softmodem might be much harder to implement, and might need a > >fair amount of either RT or kludges in the kernel. > > There is at least two softmodems on MacOS (Apple's Geoport and SAT ISDN > adapter) and they work.
Poor example. Both Windows and MacOS are highly amenable to kernel hacks. Now if you wanted to say a software modem was available for NeXT, this would be more interesting. (In fact, I believe NeXT included a software modem from the beginning -- but it only ran at 300 baud. Not so interesting after all.)
> The trick is to have some hardware support for DMA transfer of the > synchronous data flow. This way, you can have the HW transfert quite > large frames of samples and separately (asynchronously) process them > (from a kernel thread or an interrupt handler). This will slow down the > kernel but it looks possible. I think the SAT ISDN adapter also works > without DMA support (polling or interrupt every 3 byte) but the machines > slows down horribly.
Yes, that's part of it, but another part is sufficient RT support to explain to the kernel that this process has priority, and, conversely, explain when a slightly lossy modem is OK. Remember, you aren't just getting the data here and back again, but doing a bit of math on it in the meantime.
-- Kenneth Albanowski (kjahds@kjahds.com, CIS: 70705,126)
- 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/
| |