Messages in this thread | | | Date | Wed, 22 Oct 2008 13:25:30 +0100 | From | Alan Cox <> | Subject | Re: kgdb 2.6.28 merge plans |
| |
> This is a discussion that needs to be hashed out. It is a question of > where would you like it discussed (IE: new thread in lmkl, > linux-serial@...)?
Either of those. I'd also love to drive the sysrq hook out of every driver in the process. It seems the sysrq key, and debugger activation are all the same thing.
> It is a discussion more about general polled vs interrupt driven > serial/tty APIs, not all of which are used for debuggers. I am open > to help fix/change/extend/refactor the existing kernel's CONSOLE_POLL > API or even deprecate it in favor of something else so long as it > leads to an interface which can allow the sharing of the hardware > among various contexts with early_printk, console, tty, debugger, > etc...
What I've done so far is begin to create a "tty_port" object, the idea being that every tty device will eventually contain one which exists for the lifetime of the hardware object and have enough methods for printk console (and now for polled input)
> If it was all obvious and easy, it would have already been done. Rome > was not built in a day after all. :-)
Agreed entirely - and in this case we have to convert a twisty maze of ad-hoc building into Rome without breaking it during the rennovations
I certainly was not expecting a quick follow up patch to fix all these problems, rather that it gets tackled in an orderly manner for 2.6.29/30 somewhere.
Alan
| |