dwiic: stop enabling polling by default

The interrupt problem had nothing to do with changes in the Intel
100 Series and was due to most of these machines using the same DSDT
template where we were walking _INI nodes in the wrong order.

This lead to the touchpad device's _INI method being called before
\_SB_.PCI0._INI (where \OSYS is initialized), making the touchpad
device's interrupts setup for an OSYS of 0, but when its _CRS method
was called later during ihidev attachment (after \_SB_.PCI0._INI was
executed, modifying OSYS), it returned information for interrupts as
if were initially setup for the proper OSYS.

1 files changed, 4 insertions, 10 deletions

diff --git a/sys/dev/pci/dwiic_pci.c b/sys/dev/pci/dwiic_pci.c
index 520c1aaea81..d685e5dcc72 100644
--- a/sys/dev/pci/dwiic_pci.c
+++ b/sys/dev/pci/dwiic_pci.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: dwiic_pci.c,v 1.8 2019/07/16 19:12:32 jcs Exp $ */
+/* $OpenBSD: dwiic_pci.c,v 1.9 2019/07/31 16:04:16 jcs Exp $ */
 /*
  * Synopsys DesignWare I2C controller
  * PCI attachment
@@ -206,14 +206,8 @@ dwiic_pci_bus_scan(struct device *iic, struct i2cbus_attach_args *iba,
 
 	sc->sc_iic = iic;
 
-	if (sc->sc_devnode != NULL) {
-		/*
-		 * XXX: until we can figure out why interrupts don't arrive for
-		 * i2c slave devices on intel 100 series and newer, force
-		 * polling for ihidev.
-		 */
-		sc->sc_poll_ihidev = 1;
-
+#if NACPI > 0
+	if (sc->sc_devnode != NULL)
 		aml_find_node(sc->sc_devnode, "_HID", dwiic_acpi_found_hid, sc);
-	}
+#endif
 }