Move the code to write the interrupt number info back before the first card

reset (same place as in the Linux driver). Hopefully this would fix
the initialization problem I've seen previously (cannot repeat it
even without this change no matter how I play with the card).
Remove the comment about the initialization problem now.
This commit is contained in:
jdolecek 2001-03-17 16:59:29 +00:00
parent e241005cd9
commit bc5acf5f90
1 changed files with 9 additions and 14 deletions

View File

@ -1,4 +1,4 @@
/* $NetBSD: if_elmc_mca.c,v 1.1 2001/03/16 23:03:15 jdolecek Exp $ */
/* $NetBSD: if_elmc_mca.c,v 1.2 2001/03/17 16:59:29 jdolecek Exp $ */
/*-
* Copyright (c) 2001 The NetBSD Foundation, Inc.
@ -43,11 +43,6 @@
* drivers/net/3c523.[ch]).
*
* This driver uses generic i82586 stuff. See also ai(4), ef(4), ix(4).
*
* Known issues:
* o seems like the initialization is not bullet proof; initially, the
* card I had responded with "incorrect termination" warning and started
* to work correctly after I booted the PS/2 with Slackware 3.3 bootdisk
*/
#include <sys/param.h>
@ -225,6 +220,13 @@ elmc_mca_attach(struct device *parent, struct device *self, void *aux)
sc->buf_area = sc->scb + IE_SCB_SZ;
sc->buf_area_sz = sc->sc_msize - IE_ISCP_SZ - IE_SCB_SZ - IE_SCP_SZ;
/*
* According to docs, we might need to read the interrupt number and
* write it back to the IRQ select register, since the POST might not
* configure the IRQ properly.
*/
(void) mca_conf_write(ma->ma_mc, ma->ma_slot, 3, pos3 & 0x1f);
/* reset the card first */
elmc_mca_hwreset(sc, CARD_RESET);
delay(1000000 / ( 1<< 5));
@ -281,13 +283,6 @@ elmc_mca_attach(struct device *parent, struct device *self, void *aux)
else
printf("%s: interrupting at irq %d\n", sc->sc_dev.dv_xname,irq);
/*
* According to docs, we might need to read the interrupt number and
* write it back to the IRQ select register, since the POST might not
* configure the IRQ properly.
*/
(void) mca_conf_write(ma->ma_mc, ma->ma_slot, 3, pos3 & 0x1f);
return;
}