Dear Sean

I undersood what you mean. Anyway I still could not find any way to get data from drvier to application by using ioctl. 


Only ath_diag data is delivered but the data what I'd like to receive is out of this(ath_diag) data structure.  And I 
could not find any proper example code. 

Could you provide me some sample code or please let me know where I can get sample code. 

Best Regards,
Dooeui

Good chance here for a re-cap of the difference between our micro-kernel and
the typical monolithic kernel :> (and I should have been a bit more specific
when I said "just look at the source"!)

With a monolithic kernel, when you pass in a process address to the ioctl,
the ioctl results in a kernel call and the ioctl processing actually happens
in the stack which is embedded in the kernel.  Since the stack is part of
the kernel, it can do pretty much anything it wants... Including reaching
into the application's address space to read / write data.  

With the microkernel, the ioctl results in a message pass under the covers
between the application process and the stack process.  The kernel doesn't
know that this is an "ioctl function to be processed by the stack" and
simply copies the contents of the data buffers from one address space into
buffers in the other address space (note: they are different buffers).  This
means that the stack doesn't reach into the application's address space
directly and leads to one of the more difficult parts of implementing the
port of the stack from the monolithic to microkernel architecture.
Essentially, all data app-to-stack interactions have to be covered with
message passes instead of straight copies.

Most of the ioctl functional compatibility has been taking care of inside of
libsocket (take a look at the source in lib/socket/ioctlsocket.c if you want
all of the details), but for "specialized" (non-stack standard) ioctls, you
need to modify the way that the call is made in order to have the data
properly returned to you.   Unfortunately, you can't actually just adjust
the way that the data is formatted in the ioctl because the size of the
ioctl message is encoded in the ioctl macro, so you'd end up having to
modify the driver as well as the application.

I picked out the relevant portion of the ioctlsocket modified it for the
DIAG ioctl and I'm pasting it in here for you to try out.  Let me know how
it goes.  You'll also have to change it (of course) for any of the other
Atheros specific ioctls that you use (since they may have different ways of
determining address / size parameters for the second IOV).

You can then just replace the ioctl call in the code with ath_ioctl and
you're set.

		Robert.


#include <alloca.h>
#include <errno.h>
#include <fcntl.h>
#include <netinet/in.h>
#include <share.h>
#include <stdlib.h>
#include <stdarg.h>
#include <string.h>
#include <sys/types.h>
#include <sys/iomsg.h>
#include <sys/ioctl.h>
#include <sys/socket.h>
#include <sys/sockmsg.h>
#include <sys/ioctl.h>
#include <assert.h>


int ath_ioctl(int fd, int cmd, ...) {
	va_list		vl;
	void		*data;
	iov_t		siov[6], *siovp;
	int			sniov;
	iov_t		riov[6], *riovp;
	int			rniov;
	io_devctl_t	msg;
	int			ret, nbytes;

	/* Pull out the data */
	va_start(vl, cmd);
	data = va_arg(vl, void *);
	va_end(vl);

	nbytes = 0;

	switch (cmd) {
	case SIOCGATHDIAG:
		{
			struct ath_diag *diag = data;
			SETIOV(&siov[1], data, sizeof(*diag));
			SETIOV(&siov[2], diag->ad_out_data,
diag->ad_out_size);
		}
		nbytes += GETIOVLEN(&siov[1]);
		nbytes += GETIOVLEN(&siov[2]);
		
		siovp = riovp = siov;
		sniov = rniov = 2;
		break;
		
	}

	/* The following may set the same iov twice if siovp == riovp */
	SETIOV(siovp, &msg, sizeof(msg)); /* msg.o same size as msg.i */
	SETIOV(riovp, &msg, sizeof(msg)); /* msg.o same size as msg.i */
	sniov++;
	rniov++;

	msg.i.type = _IO_DEVCTL;
	msg.i.combine_len = sizeof msg.i;
	msg.i.dcmd = cmd;
	msg.i.nbytes = IOCPARM_LEN((unsigned)cmd);
	msg.i.zero = 0;

	if ((ret = MsgSendv_r(fd, siovp, sniov, riovp, rniov)) != EOK) {
		errno = -ret;
		return -1;
	}


	return 0;
}



-----Original Message-----
From: Sean Boudreau [mailto:seanb@qnx.com] 
Sent: Tuesday, January 29, 2008 9:20...