Changeset 598 in xtideuniversalbios for trunk/XTIDE_Universal_BIOS/Src/Initialization

Timestamp:

Jul 13, 2018, 10:23:54 AM (6 years ago)

Author:

Tomi Tilli

Message:

Windows 95 CMOS hack broke on last commit. Windows 95 DID require that CMOS byte 12h had drive 0 set. Drive 1 was not enough. Now there is detection for Award BIOSes and the CMOS hack won't be applied on those.

File:

• trunk/XTIDE_Universal_BIOS/Src/Initialization/DetectDrives.asm (modified) (1 diff)

trunk/XTIDE_Universal_BIOS/Src/Initialization/DetectDrives.asm

@@ -113 +113 @@
         mov     dl, HARD_DISK_TYPES
         call    CMOS_ReadFromIndexInDLtoAL
-        test    al, al
-        jnz     SHORT .ContinueInitialization   ; CMOS byte 12h is ready for Windows 95
+        test    al, 0F0h
+        jnz     SHORT .ClearBdaDriveCount       ; CMOS byte 12h is ready for Windows 95
         call    CMOS_Verify10hTo2Dh             ; Can we modify CMOS?
         jnz     SHORT .ClearBdaDriveCount       ; Unsupported BIOS, use plan B
 
-        ; Now we can alter CMOS location 12h. Important! We set type for drive 1
-        ; (primary slave) and not for drive 0! Award BIOS locks if we set drive 0 type to Fh.
-        ; We cannot set it to less either since that will fully set predefined hard drive type to the BIOS.
-        ; Windows 95 only cares that the CMOS location 12h is non-zero.
+        ; Now we can alter CMOS location 12h. Award BIOS locks if we set drive 0 type to Fh
+        ; (but accept changes to drive type 1). Windows 95 requires that the drive 0 type is
+        ; non zero and ignores drive 1 type. So if we only set drive 1, then Award BIOS
+        ; won't give problems but Windows 95 stays in MS-DOS compatibility mode.
+        ;
+        ; For Award BIOSes we could set the Drive 0 type to 1 and then clear the BDA drive count.
+        ; So essentially we could automatically do what user needs to do manually to get Windows 95
+        ; working on Award BIOSes. However, I think that should be left to do manually since
+        ; there may be SCSI drives on the system or FLG_ROMVARS_IGNORE_MOTHERBOARD_DRIVES could
+        ; be intentionally cleared and forcing the dummy drive might cause only trouble.
+
+        ; Try to detect Award BIOS (Seems to work on a tested 128k BIOS so hopefully
+        ; there will be no need to scan E000h segment)
+        mov     cx, 65536 - 4
+        mov     eax, 'Awar'     ; Four characters should be enough
+        mov     di, 0F000h      ; Scan 64k starting from segment F000h
+        mov     fs, di          ; No need to preserve FS since we set it to zero soon when we boot
+        xor     di, di
+    .ScanNextCharacters:
+        cmp     [fs:di], eax
+        je      SHORT .ClearBdaDriveCount   ; Award detected, cannot modify CMOS
+        inc     di              ; Increment offset by one character (not four)
+        loop    .ScanNextCharacters
+
+        ; Now it should be safe to write
         mov     dl, HARD_DISK_TYPES
-        mov     al, 0Fh ; Drive 1 type 16...47 (defined elsewhere in the CMOS)
+        mov     al, 0F0h    ; Drive 0 type 16...47 (supposed to be defined elsewhere in the CMOS)
         call    CMOS_WriteALtoIndexInDL
         call    CMOS_StoreNewChecksumFor10hto2Dh