Messages in this thread | | | From | Arnd Bergmann <> | Subject | Re: [PATCH 4/5] tty/hw: move hardware drivers to drivers/tty/hw | Date | Mon, 30 Aug 2010 17:25:01 +0200 |
| |
On Monday 30 August 2010, Mike Frysinger wrote: > > On Mon, Aug 30, 2010 at 08:28, Arnd Bergmann wrote: > > On Monday 30 August 2010, Alan Cox wrote: > >> > My impression was that it should actually be converted into an hvc > >> > driver as a cleanup and moved to drivers/tty/hvc. It can probably > >> > get reduced to one third of its size if you do that. > >> > >> Not sure it makes sense - hvc seems the wrong abstraction. > > > > It seems to be very similar to hvc_tile.c in purpose, which is > > the most trivial driver we have. Why would that not work? > > > > Most of bfin_jtag_comm is about interfacing with the tty layer > > and providing a kernel thread to do the polling, which is exactly > > what hvc_console.c does. > > i have no idea what "hvc" is, but my limited knowledge was that it was > for large powerpc servers, so i didnt bother looking at it too much. > i simply wanted a way to create a tty on top of the hardware channel. > if there is a framework that doesnt break this usage and lets me > simplify my driver, then either works for me.
HVC started out as the console for large powerpc, but has turned into a generic framework for TTY drivers that are based on simple read/write operations with or without interrupts.
Please take a look at drivers/char/hvc_tile.c, which was added recently for a scenario quite similar to yours (debug port, not irq).
Arnd
| |