Re: tp_smapi conflict with IDE, hdaps

[Shem Multinymous <multinymous@gmail.com>]

On 12/13/05, Robert Love <rlove@rlove.org> wrote:
> On Tue, 2005-12-13 at 16:35 +0200, Shem Multinymous wrote:
> Alan's response is the correct course of action here,

Ideally, but I'm not sure we understand the interface sufficiently
well to abstract it beyond a mere mutex. Compare your hdaps.c code to
the following relevant code from tp_smapi.c (stripped down a bit for
simplicty):

----------------------------------------
#define APS_ROW_LEN 16
static int read_aps_row(u8 arg1610, u8 arg161F, u8* buf) {
	int retries, i;
	int ret = -EIO;

	for (retries=APS_MAX_RETRIES; retries>0; --retries) {
		if (inb(0x1604)&0x40) { /* readout pending? */
			inb(0x161F); /* discard it */
			udelay(10);
		} else {
			outb(arg1610, 0x1610);
			if (inb(0x1604)&0x20)
				goto wrote1610;
		}
	}
	goto out;
wrote1610:
	outb(arg161F, 0x161F);
	for (retries=APS_MAX_RETRIES; retries>0; --retries) {
		if (inb(0x1604)&0x40) /* readout pending? */
			goto gotdata;
			udelay(10);
	}
	goto out;
gotdata:
	for (i=0; i<APS_ROW_LEN; ++i) {
		buf[i] = inb(0x1610+i);
	}
	ret = 0;
out:
	return ret;
}
----------------------------------------
(Yes, the loop/goto structure should be cleanup up a bit.)

Not to mention the HDAPS init code, which is voodoo.


> question: What other data in 0x1604-0x161F is there?

I looked only at the readout "rows" given by arg1610=1,..,18 and
arg161F=0,1. In that range, I didn't see anything obviously
interesting in the "rows" not already used by the battery readout and
HDAPS. I don't think any of the Windows driver reads other "rows". But
for all we know, this could be a window into the embedded controller's
memory or something of the sorts. Also, some of these "rows" are
actually commands given during HDAPS init, so the "read_row"
abstraction is obviously not accurate.

  Shem