On Mon, Feb 24, 2003 at 03:51:22PM +0100, fcorneli@elis.rug.ac.be wrote:
> Hi,
>
> > FYI Frank, three things. First of all, I really don't like the
> > interface of adding a second address to ptrace; I believe it interferes
> > with PIC on x86, since IIRC the extra argument would go in %ebx.
> > The BSDs have a nice interface involving passing a request structure.
>
> I don't see the problem since we can pass up to 6 parameters on the i386
> architecture. The extra argument will be passed on using the stack as the
> other arguments do because of the asmlinkage directive. Using a structure
> slows everything down too much; if you can use the stack I think it's
> better to do so. What about that PIC?
I seem to remember this (five-arg syscalls) causing problems before.
Maybe it was on a different platform.
> > Secondly, the implementation should be in kernel/ptrace.c not under
> > i386, we're trying to stop doing that.
>
> The implementation is already in kernel/ptrace.c, only the usage lives
> under the arch-dependent directories since there the sys_ptrace entries
> are located.
Not any more; it should be in ptrace_request for anything new. Yes, if
you're adding an argument, that makes this more work.
> > Thirdly, I was going to do this, but I ended up making GDB use pread64
> > on /dev/mem instead. It works with no kernel modifications, and is
> > just as fast.
>
> mmm... I thought it would be convenient to use ptrace for all the trace
> work.
I've found it really doesn't make a difference.
-- Daniel Jacobowitz MontaVista Software Debian GNU/Linux Developer - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Feb 28 2003 - 22:00:19 EST