cvs commit: src/sys/dev/dcons dcons.c
simokawa at sat.t.u-tokyo.ac.jp
Sun Jul 11 01:06:07 PDT 2004
Thank you for your explanation.
I found that the interface is very similar to the low level console
interface. dcons(4) is actually character oriented and the current
interface seems sufficient.
/\ Hidetoshi Shimokawa
\/ simokawa at sat.t.u-tokyo.ac.jp
PGP public key: http://www.sat.t.u-tokyo.ac.jp/~simokawa/pgp.html
At Sun, 11 Jul 2004 00:49:01 -0700,
Marcel Moolenaar wrote:
> On Sun, Jul 11, 2004 at 04:11:16PM +0900, Hidetoshi Shimokawa wrote:
> > I'd like to re-support remote debugging by dcons(4).
> > Do you have any pointer for "new GDB debug port interface"?
> The interface is defined in src/sys/gdb/gdb.h. A good example can
> be found in src/sys/dev/uart/uart_dbg.c.
> First GDB calls the probe function. The probe function is supposed
> to return a priority (>=0). A negative priority means that the
> debug port is dead or unusable. The uart(4) driver returns 0. So
> does the sio(4) driver. If you want to give dcons priority over a
> serial port, have the probe function return 1 or more.
> After all the probe functions have been called, one debug port is
> seleced as the current port. The init function for that port is
> called so that it can initialize itself.
> After that, the GDB backend will call the getc and putc functions
> to send and receive packets. The interface is mostly character
> oriented now, even though the GDB protocol is based on packets. If
> this doesn't work as efficiently for firewire (or ethernet for that
> matter) and have ideas to make it more friendly for you, let me
> know. I welcome improvements and enhancements.
> Marcel Moolenaar USPA: A-39004 marcel at xcllnt.net
More information about the cvs-all