80486DX2-66/86Box-fork
1
0
Fork 0
Code Issues Pull Requests Activity

Remove the previous kludge and #define BUS_ACK to 0x100 so it's different from BUS_ATN.

Browse Source
This commit is contained in:
OBattler
2024-05-06 13:24:04 +02:00
parent 4c84cccae7
commit 71c16a4481
2 changed files with 3 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
src/include/86box/scsi_device.h
View File

@@ -308,8 +308,7 @@
#define BUS_REQ 0x20
#define BUS_BSY 0x40
#define BUS_RST 0x80
#define BUS_ACK 0x200
/* TODO: Why is this defined to the same value as BUS_ACK?! */
#define BUS_ACK 0x100
#define BUS_ATN 0x200
#define BUS_ARB 0x8000
#define BUS_SETDATA(val) ((uint32_t) val << 16)

24
src/scsi/scsi_ncr5380.c
View File

@@ -161,28 +161,8 @@ ncr5380_get_bus_host(ncr_t *ncr)
if (ncr->icr & ICR_BSY)
bus_host |= BUS_BSY;
/*
TODO: See which method of fixing this is the most correct.
The #define's come from PCem and, for some reason, define
BUS_ATN to the same value as BUS_ACK (0x200). This breaks
the Corel driver for the Pro Audio Spectrum Trantor SCSI
controller, as it first asserts ATN without ACK, then ACK
without ATN, which should by definition result in ACK going
ON and OFF but because of these ambiguous #define's, it
instead manifests as ACK stuck on, therefore never clearing
BUS_REQ and progressing to the next phase.
Since I have no idea why BUS_ATN was #define's to the same
value as BUS_ACK, and the problem appears to only occur in
the Message Out phase, where ATN is not used, I have decided
to solve this by never asserting ATN in the Message Out phase
for time being.
*/
if (ncr->state != STATE_MESSAGEOUT) {
if (ncr->icr & ICR_ATN)
bus_host |= BUS_ATN;
}
if (ncr->icr & ICR_ATN)
bus_host |= BUS_ATN;
if (ncr->icr & ICR_ACK)
bus_host |= BUS_ACK;