If the SCI_EVT flag is no longer set after processing an event, we
should clear sc->sc_gotsci, otherwise we'll do a pointless query of
the embedded controller. Such a query is actually hazardous on some
embedded controllers. At least I have an HP laptop where this makes
the embedded controller block, hanging the acpi thread. This makes
suspend/resume work on that laptop. Without this diff it doesn't even
attempt to suspend; I just get the prompt back after typing zzz.
As with all acpi diffs, this needs testing. So if your machine has
acpiec0 in its dmesg, please give this a shot, even if your machines
works perfectly already.
Index: acpiec.c
===================================================================
RCS file: /home/cvs/src/sys/dev/acpi/acpiec.c,v
retrieving revision 1.47
diff -u -p -r1.47 acpiec.c
--- acpiec.c 18 Jan 2013 06:02:51 -0000 1.47
+++ acpiec.c 30 Jun 2013 10:03:58 -0000
@@ -343,6 +343,8 @@ acpiec_gpehandler(struct acpi_softc *acp
if (stat & EC_STAT_SCI_EVT)
sc->sc_gotsci = 1;
+ else
+ sc->sc_gotsci = 0;
} while (sc->sc_gotsci);
/* Unmask the GPE which was blocked at interrupt time */
