Changeset 478 in xtideuniversalbios for trunk/XTIDE_Universal_BIOS/Src

Timestamp:

Oct 20, 2012, 7:37:28 PM (12 years ago)

Author:

aitotat@…

google:author:

aitotat@gmail.com

Message:

Changes to XTIDE Universal BIOS:

• Fixes to DMA transfers but it still does not work.

File:

• trunk/XTIDE_Universal_BIOS/Src/Device/IDE/IdeDmaTransfer.asm (modified) (8 diffs)

trunk/XTIDE_Universal_BIOS/Src/Device/IDE/IdeDmaTransfer.asm

@@ -104 +104 @@
     mov     ah, CHANNEL_3 | WRITE | AUTOINIT_DISABLE | ADDRESS_INCREMENT | DEMAND_MODE
     call    StartDMAtransferForXTCFwithDmaModeInAH
-    jc      SHORT ReturnNumberOfSectorsXferred
+    call    UpdateVariablesForSecondPageIfRequired
+    jc      SHORT ReturnNumberOfSectorsXferred      ; Second page not needed
 
     ; Transfer second DMA page if necessary (always less than 64k)
-    call    UpdateVariablesForSecondPageIfRequired
-    jc      SHORT SecondDmaPageIsNotRequired
     mov     ah, CHANNEL_3 | WRITE | AUTOINIT_DISABLE | ADDRESS_INCREMENT | DEMAND_MODE
     call    StartDMAtransferForXTCFwithDmaModeInAH
-    jc      SHORT ReturnNumberOfSectorsXferred
-    ; Fall to BothDmaPagesTransferredSuccessfully
-
-BothDmaPagesTransferredSuccessfully:
+    ; Fall to BothDmaPagesTransferred
+
+BothDmaPagesTransferred:
     inc     cx          ; Never overflows since second page always less than 64k
     shr     cx, 1       ; BYTEs to WORDs
     add     [bp+DMAVARS.wTotalWordsXferred], cx
-SecondDmaPageIsNotRequired:
-    xor     ah, ah
 ReturnNumberOfSectorsXferred:
+    mov     bx, TIMEOUT_AND_STATUS_TO_WAIT(TIMEOUT_DRQ, FLG_STATUS_BSY)
+    call    IdeWait_PollStatusFlagInBLwithTimeoutInBH
+    jc      SHORT .ErrorInTransfer
     mov     cx, [bp+DMAVARS.wTotalWordsXferred]
     xchg    cl, ch      ; WORDs to sectors
+    ret
+
+.ErrorInTransfer:
+    mov     cx, 0       ; No way to know how many bytes got xferred
     ret
 
@@ -141 +144 @@
 ;       AL, BX, DX
 ;--------------------------------------------------------------------
+ALIGN JUMP_ALIGN
 WriteBlockToXTCF:
     ; Transfer first DMA page
     mov     ah, CHANNEL_3 | READ | AUTOINIT_DISABLE | ADDRESS_INCREMENT | DEMAND_MODE
     call    StartDMAtransferForXTCFwithDmaModeInAH
-    jc      SHORT ReturnNumberOfSectorsXferred
+    call    UpdateVariablesForSecondPageIfRequired
+    jc      SHORT ReturnNumberOfSectorsXferred      ; Second page not needed
 
     ; Transfer second DMA page if necessary (always less than 64k)
-    call    UpdateVariablesForSecondPageIfRequired
-    jc      SHORT SecondDmaPageIsNotRequired
     mov     ah, CHANNEL_3 | READ | AUTOINIT_DISABLE | ADDRESS_INCREMENT | DEMAND_MODE
     call    StartDMAtransferForXTCFwithDmaModeInAH
-    jc      SHORT ReturnNumberOfSectorsXferred
-    jmp     SHORT BothDmaPagesTransferredSuccessfully
+    jmp     SHORT BothDmaPagesTransferred
 
 
@@ -163 +165 @@
 ;       DS:DI:  Ptr to DPT (in RAMVARS segment)
 ;   Returns:
-;       AH:     BIOS Error code
-;       CF:     0 if transfer successful
-;               1 if any error
+;       Nothing
 ;   Corrupts registers:
-;       AL, BX, DX
-;--------------------------------------------------------------------
+;       AL, DX
+;--------------------------------------------------------------------
+ALIGN JUMP_ALIGN
 StartDMAtransferForXTCFwithDmaModeInAH:
     push    cx
@@ -210 +211 @@
     ; total block requests.  The 8237 is programmed with the actual byte
     ; count and will end the transfer by asserting TC when done.
-    xor     ax, ax
-    add     cx, BYTE 1 + 15     ; Number of BYTEs to xfer + 15
-    adc     al, ah
-    shr     ax, 1
+    add     cx, BYTE 1 + 15     ; Number of BYTEs to xfer + 15 (bit 16 in CF)
     rcr     cx, 1
     eSHR_IM cx, 3               ; CX = Number of 16 byte blocks
@@ -220 +218 @@
 
 .MoreToDo:                      ; at this point, cx must be >0
-    mov     al, 0x40            ; 0x40 = Raise DRQ and clear XT-CFv2 transfer counter
+    mov     al, 40h             ; 0x40 = Raise DRQ and clear XT-CFv2 transfer counter
 .NextDemandBlock:
     out     dx, al              ; get up to 16 bytes from XT-CF card
@@ -234 +232 @@
 
 .EndDMA:
-    mov     al, 0x10            ; 
+    mov     al, 10h             ; 
     out     dx, al              ; set back to DMA enabled status
-
-    ; Check IDE Status Register for errors
-    mov     bx, TIMEOUT_AND_STATUS_TO_WAIT(TIMEOUT_DRQ, FLG_STATUS_BSY)
-    call    IdeWait_PollStatusFlagInBLwithTimeoutInBH
     pop     cx
     ret
@@ -256 +250 @@
 ;       AX, (CX)
 ;--------------------------------------------------------------------
+ALIGN JUMP_ALIGN
 UpdateVariablesForSecondPageIfRequired:
     inc     cx                          ; BYTEs in first page
@@ -281 +276 @@
 .OnePageWasEnough:
     ret
+
+
+
+
+%if 0
+;--------------------------------------------------------------------
+ALIGN JUMP_ALIGN
+ReadBlockFromXTCF:
+;   Parameters:
+;       CX:     Block size in 512 byte sectors (1..64)
+;       DX:     IDE Data port address
+;       ES:DI:      Normalized ptr to buffer to receive data
+;   Returns:
+;       Nothing
+;   Corrupts registers:
+;       AX, BX, CX
+
+    ; use config register to see what we're doing:
+    ; < A0 = DMA via ch.3
+    ; > A0 = mem-mapped IO (word-length)
+    ; 
+    
+    or      dl, 0x1f    ; XT-CF board control register address (base + 1fh)
+    in      al, dx      ; get control register
+    cmp     al, 0xA0    ; test al against 0xA0:
+    jae .MemMapIO       ; - use memory-mapped IO if >=A0h
+    or      al, al      ; test al against 0 (quicker than cmp):
+    jz  .PortIO         ; - use port-based IO (the default) if it was zero
+                    ; otherwise, 0 < al < A0h, so fall to DMA mode
+
+.DMAIO:
+    ; work out how much we're transferring.  We can't cross a physical 64KB boundary
+    ; but since the max transfer size is 64KB, we only ever need to do one or two DMA operations
+    
+    ; first, normalise the pointer - we need ES to be a physical page address 
+    
+    mov     ax, es
+    mov     ch, cl          ; max sectors is 128; also sectors << 8 = words, and we need cl...
+    mov     cl, 4           ; ... for the rotate parameter
+    rol     ax, cl          ; ax = es rol 4
+    mov     bx, ax          ; bx = ax = es rol 4
+    and     al, 0xf0        ; ax (15..4) now has es (11..0)
+    and     bx, 0x000f      ; bx (3..0) now has es (15..12)
+    add     di, ax          ; add offset portion of es (in ax) to di...
+    adc     bl, 0           ; ... and if it overflowed, increment bl
+    mov     es, bx          ; es now has physical segment address >> 12
+
+    ; now check how much we can transfer without crossing a physical page boundary
+    mov     bx, di          ; 
+    not     bx          ; 65535 - di = number of bytes we could transfer -1 now in bx
+    xor     cl, cl          ; zero cl; cx now has transfer size in words
+    shl     cx, 1           ; words to bytes; CX has total byte count
+    dec     cx          ; calling DMA with 0 implies 1 byte transferred (so max is 64k exactly)
+    cmp     bx, cx          ; can we do it in one hit?
+    jae .LastDMA
+
+    ; at this point, the (bytes-1) for this transfer are in bx, total byte count -1 in cx
+    ; and we need to do 2 DMA operations as the buffer straddles a physical 64k boundary
+    
+    sub     cx, bx          ; cx has bytes for 2nd transfer (as (x-1)-(y-1) = x-y)
+    dec     cx          ; cx has bytes-1 for 2nd transfer
+    xchg        bx, cx          ; bx = bytes-1 for 2nd transfer; cx = bytes-1 for this transfer
+    mov     ah, 0x07        ; request a read DMA transfer
+    call    DemandBasedTransferWithBytesInCX
+
+    ; DMA 1 of 2 done - set up registers for second transfer
+    mov     cx, bx          ; bytes-1 for the 2nd transfer back in cx
+    ; 1st transfer is done, now for the second
+ALIGN JUMP_ALIGN
+.LastDMA:
+    ; at this point, (bytes-1) for this transfer are in CX
+    mov     ah, 0x07        ; request a read DMA transfer
+    call    DemandBasedTransferWithBytesInCX
+
+    ; transfer is done, set ES back to a physical segment address
+    mov     ax, es          ; 
+    mov     cl, 4           ; 
+    ror     ax, cl          ; ax = es ror 4.  Since ES was >> 12 (for DMA controller) so
+    mov     es, ax          ; now it's back to normal format
+    
+    ; pointer format restored - we're done
+    and     dl, 0xE0        ; restore register values
+    ret
+
+
+;--------------------------------------------------------------------
+ALIGN JUMP_ALIGN
+WriteBlockToXTCF:
+    ; use config register to see what we're doing:
+    ; 0 = not set; use byte-length port-IO
+    ; < A0 = DMA via ch.3
+    ; > A0 = mem-mapped IO (word-length)
+    ; 
+    
+    or      dl, 0x1f    ; XT-CF board control register address (base + 1fh)
+    in      al, dx      ; get control register
+    cmp     al, 0xA0    ; test al against 0xA0:
+    jae .MemMapIO       ; - use memory-mapped IO if >=A0h
+    or      al, al      ; test al against 0 (quicker than cmp):
+    jz  .PortIO         ; - use port-based IO (the default) if it was zero
+                    ; otherwise, 0 < al < A0h, so fall to DMA mode
+
+.DMAIO:
+    ; work out how much we're transferring.  We can't cross a physical 64KB boundary
+    ; but since the max transfer size is 64KB, we only ever need to do one or two DMA operations
+
+    push        di          ; save di...
+    mov     di, si          ; ...and move si to di (for DMA routine)
+    
+    ; first, normalise the pointer - we need ES to be a physical page address 
+    
+    mov     ax, es
+    mov     ch, cl          ; max sectors is 64; also sectors << 8 = words, and we need cl...
+    mov     cl, 4           ; ... for the rotate parameter
+    rol     ax, cl          ; ax = es rol 4
+    mov     bx, ax          ; bx = ax = es rol 4
+    and     al, 0xf0        ; ax (15..4) now has es (11..0)
+    and     bx, 0x000f      ; bx (3..0) now has es (15..12)
+    add     di, ax          ; add offset portion of es (in ax) to si...
+    adc     bl, 0           ; ... and if it overflowed, increment bl
+    mov     es, bx          ; es now has physical segment address >> 12
+
+    ; now check how much we can transfer without crossing a physical page boundary
+    mov     bx, di          ; 
+    not     bx          ; 65535 - di = number of bytes we could transfer -1 now in bx
+    xor     cl, cl          ; zero cl; cx now has transfer size in words
+    shl     cx, 1           ; words to bytes; CX has total byte count
+    dec     cx          ; calling DMA with 0 implies 1 byte transferred (so max is 64k exactly)
+    cmp     bx, cx          ; can we do it in one hit?
+    jae .LastDMA
+
+    ; at this point, the (bytes-1) for this transfer are in bx, total byte count -1 in cx
+    ; and we need to do 2 DMA operations as the buffer straddles a physical 64k boundary
+    
+    sub     cx, bx          ; cx has bytes for 2nd transfer (as (x-1)-(y-1) = x-y)
+    dec     cx          ; cx has bytes-1 for 2nd transfer
+    xchg        bx, cx          ; bx = bytes-1 for 2nd transfer; cx = bytes-1 for this transfer
+    mov     ah, 0x0b        ; request a write DMA transfer
+    call    DemandBasedTransferWithBytesInCX
+
+    ; DMA 1 of 2 done - set up registers for second transfer
+    mov     cx, bx          ; bytes-1 for the 2nd transfer back in cx
+    ; 1st transfer is done, now for the second
+ALIGN JUMP_ALIGN
+.LastDMA:
+    ; at this point, (bytes-1) for this transfer are in CX
+    mov     ah, 0x0b        ; request a write DMA transfer
+    call    DemandBasedTransferWithBytesInCX
+
+    ; transfer is done, set ES back to a physical segment address
+    mov     ax, es          ; 
+    mov     cl, 4           ; 
+    ror     ax, cl          ; ax = es ror 4.  Since ES was >> 12 (for DMA controller) so
+    mov     es, ax          ; now it's back to normal format
+
+    mov     si, di          ; move di (used by DMA routine) back to si
+    
+    ; pointers updated - we're done
+    pop     di          ; 
+    and     dl, 0xE0        ; restore register values
+    ret
+
+
+; -------------------------------------------------------------------------------------------------------------
+; 
+; DemandBasedTransferWithBytesInCX
+; ================================
+; 
+; DMA Transfer function:
+; - AH = 0x07 for read from media - demand/inc/non-auto-init/write(to memory)/ch3
+; - AH = 0x0b for write to media  - demand/inc/non-auto-init/read(from memory)/ch3
+; - byte count -1 in CX
+; - XT-CF control register in DX
+; - physical segment address in ES *but* high four bits in low four bits (i.e. shr 12)
+; - buffer offset (from physical segment) in DI
+;
+; Note - cannot cross a physical segment boundary, but ES will be updated (maintaining the >>12
+;        format) if after the last byte has been transferred the pointer needs to be updated to
+;        the next physical segment.
+; 
+; Preserves:    BX, DX
+; Corrupts:     AX, CX
+; Updates:  ES, DI
+; 
+; -------------------------------------------------------------------------------------------------------------
+
+ALIGN JUMP_ALIGN
+DemandBasedTransferWithBytesInCX:
+    cli                 ; clear interrupts while we set up the actual DMA transfer
+    mov     al, 0x07        ; mask (4) + channel (3)
+    out     0x0a, al        ; send to DMA mask register
+    mov     al, ah          ; retrieve the transfer mode passed in...
+    out     0x0b, al        ; and send mode to DMA mode register
+    out     0x0c, al        ; clear DMA byte-order flip-flop (write any value)
+    mov     ax, di          ; di has buffer offset
+    out     0x06, al        ; 
+    mov     al, ah          ; 
+    out     0x06, al        ; send offset to DMA controller address port for ch.3
+    mov     ax, es          ; es has physical segment address >> 12
+    out     0x82, al        ; send the relavent 4-bits to DMA controller page for ch.3
+    out     0x0c, al        ; clear DMA byte-order flip-flop (write any value)
+    mov     ax, cx          ; byte count to AX
+    out     0x07, al        ; send low-byte of transfer size in bytes...
+    mov     al, ah          ; mov high byte to low byte...
+    out     0x07, al        ; ...and high byte to DMA controller count port for ch.3
+    mov     al, 0x03        ; clear bit mask (0) + channel (3)
+    out     0x0a, al        ; send to DMA mask register - enable the DMA!
+    sti                 ; enable interrutps; let the CPU see to anything outstanding
+
+    mov     ax, es          ; update ES:DI (values have been loaded into DMA controller)
+    inc     cx          ; CX back to actual bytes
+    add     di, cx          ; add bytes to DI...
+    adc     ax, 0           ; ...and if it overflowed, increment...
+    mov     es, ax          ; ...ES
+
+    add     cx, 15          ; XT-CFv2 will present data in 16-byte blocks, but byte count may not
+    shr     cx, 1           ; be divisable by 16 due to boundary crossing.  So catch any < 16-byte
+    shr     cx, 1           ; block by adding 15, then dividing bytes (in CX) by 16 to get the
+    shr     cx, 1           ; total block requests.  The 8237 is programmed with the actual byte
+    shr     cx, 1           ; count and will end the transfer by asserting TC when done.
+
+.MoreToDo:                  ; at this point, cx must be >0
+    mov     al, 0x40        ; 0x40 = Raise DRQ and clear XT-CFv2 transfer counter
+.NextDemandBlock:
+    out     dx, al          ; get up to 16 bytes from XT-CF card
+    loop    .NextDemandBlock        ; decrement CX and loop if <> 0
+                        ; (Loop provides a wait-state between 16-byte blocks; do not unroll)
+.CleanUp:
+    ; check the transfer is actually done - in case another DMA operation messed things up
+    inc     cx          ; set up CX, in case we need to do an extra iteration
+    in      al, 0x08        ; get DMA status register
+    and     al, 0x08        ; test DMA ch.3 TC bit
+    jz  .MoreToDo           ; it wasn't set so get more bytes
+.EndDMA:
+    mov     al, 0x10        ; 
+    out     dx, al          ; set back to DMA enabled status
+    ret
+    
+%endif ; 0
+
+
+