I think this is a good idea.
>
> Now, before someone kills me for suggesting that I want to deploy a debugging
> patch in critical operations ;), that also means that we can track down bugs
> which occur as efficient as possible.
And in fact, there is no interaction with kdb and the rest of the
system under normal operations. kdb hooks are limited to:
a) the serial port (or graphics console) via ^A or <pause>
b) The panic routines
c) The debug exception.
So long as these aren't invoked, kdb is not involved in system operations.
>
> If we find a bug and then switch to a debugging kernel on that box, we all
> know what happens: first, we will not be able to reproduce the crash with the
> debugging kernel, second, it will take ages for this to occur and occur on
> some other box which is not running the debugging kernel yet, and third, this
> also means an additional down time for switching to the debugging kernel (and
> also the 2nd crash which we basically wait for even).
>
> Now, in general having an extended debug utility on the serial console appears
> like a real good idea. But I am not sure whether having kdb running in
> "standby" affects performance much if it is not used or if there are any
> additional caveats we may run into.
KDB doesn't affect performance at all during normal system operations.
>
> Does it interfere with a getty running on the serial console?
The serial console is still usable as a serial console, and runs
mingetty and you can log in as normal. Control-A will drop you
into the debugger at that point.
>
> The default command sequence ^A appears to be a real bad choice, since it
It was a legacy choice. It can be changed as necessary in
drivers/char/serial.c
> means "move to beginning of line" in quite a few programs, and is already used
> as a command sequence in minicom and screen, both very useful programs if
> running on a serial console. I assume this can be changed easily?
>
> Sincerely,
> Lars Marowsky-Brée
scott
(new kdb home page:
http://oss.sgi.com/projects/kdb
)
-
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/