assembly - Why isn't my root directory being loaded? (FAT12)

Question

I am writing a stage 1 bootloader in assembly with which I am attempting to load the FAT12 filesystem into memory so that I can load my stage 2 bootloader. I have managed to load the FATs into memory, however I am struggling to load the root directory into memory.

I am currently using this for reference and have produced the following:

.load_root:
    ;es is 0x7c0
    xor dx, dx              ; blank dx for division
    mov si, fat_loaded      ; inform user that FAT is loaded
    call print
    mov al, [FATcount]      ; calculate how many sectors into the disk must be loaded
    mul word [SectorsPerFAT]
    add al, [ReservedSectors]
    div byte [SectorsPerTrack]
    mov ch, ah              ; Store quotient in ch for cylinder number
    mov cl, al              ; Store remainder in cl for sector number

    xor dx, dx
    xor ax, ax
    mov al, ch              ; get back to "absolute" sector number
    mul byte [SectorsPerTrack]
    add al, cl
    mul word [BytesPerSector]
    mov bx,ax               ; Memory offset to load to data into memory after BOTH FATs (should be 0x2600, physical address should be 0xA200)

    xor dx, dx              ; blank dx for division
    mov ax, 32
    mul word [MaxDirEntries]
    div word [BytesPerSector] ; number of sectors root directory takes up (should be 14)

    xor dh, dh              ; head 0
    mov dl, [boot_device]   ; boot device

    mov ah, 0x02            ; select read mode

    int 13h
    cmp ah, 0
    je .load_OS
    mov si, error_text
    call print
    jmp $

However, if I inspect the memory at 0xA200 with gdb, I just see 0s. My root directory does contain a file -- I have put a file called OS.BIN in the root directory to test with.

Using info registers in gdb after the read operation gives the following output:

eax            0xe      14
ecx            0x101    257
edx            0x0      0
ebx            0x2600   9728
esp            0x76d0   0x76d0
ebp            0x0      0x0
esi            0x16d    365
edi            0x0      0
eip            0x7cdd   0x7cdd
eflags         0x246    [ PF ZF IF ]
cs             0x0      0
ss             0x53     83
ds             0x7c0    1984
es             0x7c0    1984
fs             0x0      0
gs             0x0      0

The status of the operation is 0, the number of sectors read is 14, and es:bx points to 0xA200, but x/32b 0xa200 shows 32 0s, when I would expecting to see the data for OS.BIN.

EDIT I did info registers before the interrupt and the output is the following:

eax            0x20e    526
ecx            0x101    257
edx            0x0      0
ebx            0x2600   9728
esp            0x76d0   0x76d0
ebp            0x0      0x0
esi            0x161    353
edi            0x0      0
eip            0x7cc8   0x7cc8
eflags         0x246    [ PF ZF IF ]
cs             0x0      0
ss             0x53     83
ds             0x7c0    1984
es             0x7c0    1984
fs             0x0      0
gs             0x0      0

Which is the same as after, except the function request number has been replaced with the status code.

Where am I going wrong? Am I reading from the wrong CHS address? Or some other simple mistake? And how can I correct this?

I am using fat_imgen to make my disk image. Command for creating the disk image is fat_imgen -c -f floppy.flp -F -s bootloader.bin and command for adding OS.BIN to the image is fat_imgen -m -f floppy.flp -i OS.BIN

---

I have a BIOS Parameter Block (BPB) that represents a 1.44MB floppy using FAT12:

jmp short loader
times 9 db 0

BytesPerSector: dw 512
SectorsPerCluster: db 1
ReservedSectors: dw 1
FATcount: db 2
MaxDirEntries: dw 224
TotalSectors: dw 2880
db 0
SectorsPerFAT: dw 9
SectorsPerTrack: dw 18
NumberOfHeads: dw 2
dd 0
dd 0
dw 0
BootSignature: db 0x29
VolumeID: dd 77
VolumeLabel: db "Bum'dOS   ",0
FSType: db "FAT12   "

I have another function that appears to work that loads the FAT12 table to memory address 0x7c0:0x0200 (physical address 0x07e00):

;;;Start loading File Allocation Table (FAT)
.load_fat:
    mov ax, 0x07c0          ; address from start of programs
    mov es, ax
    mov ah, 0x02            ; set to read
    mov al, [SectorsPerFAT]   ; how many sectors to load
    xor ch, ch              ; cylinder 0
    mov cl, [ReservedSectors]  ; Load FAT1
    add cl, byte 1
    xor dh, dh              ; head 0
    mov bx, 0x0200          ; read data to 512B after start of code
    int 13h
    cmp ah, 0
    je .load_root
    mov si, error_text
    call print
    hlt

Answer 1

Analysis of Problem

The issue with your code is that you aren't reading from the point on the disk that you are expecting. Although your disk read is successful it has loaded the wrong sectors into memory.

If we look at Ralph Brown's Interrupt List for Int 13h/AH=2 we'll see that the inputs look like this:

DISK - READ SECTOR(S) INTO MEMORY

AH = 02h
AL = number of sectors to read (must be nonzero)
CH = low eight bits of cylinder number
CL = sector number 1-63 (bits 0-5)
high two bits of cylinder (bits 6-7, hard disk only)
DH = head number
DL = drive number (bit 7 set for hard disk)
ES:BX -> data buffer

If we review your registers before you do int 13h in .load_root we see these registers with the following contents:

eax            0x20e   
ecx            0x101 
edx            0x0
ebx            0x2600 
es             0x7c0

So ES:BX is 0x7c0:0x2600 which is physical address 0xA200. That is correct. AH (0x02) is disk read and the number of sectors to read in AL is 14 (0x0e). This seems reasonable. The issue arises in ECX and EDX. If we review your code it appears you are attempting to find the sector (Logical Block Address) on disk where the root directory starts:

mov al, [FATcount]      ; calculate how many sectors into the disk must be loaded
mul word [SectorsPerFAT]
add al, [ReservedSectors]

In your BIOS Parameter Block you have SectorsPerFat = 9, ReservedSectors = 1, and FATCount = 2. If we review a FAT12 design document that shows this configuration it would look like:

Your calculation is correct. 2*9+1 = 19. The first 19 Logical Blocks run from LBA 0 to LBA 18. LBA 19 is where your root directory starts. We need to convert this to Cylinders/Heads/Sectors (CHS). Logical Block Address to CHS calculation:

CHS tuples can be mapped to LBA address with the following formula:

LBA = (C × HPC + H) × SPT + (S - 1)

where C, H and S are the cylinder number, the head number, and the sector number

LBA is the logical block address
HPC is the maximum number of heads per cylinder (reported by 
    disk drive, typically 16 for 28-bit LBA)
SPT is the maximum number of sectors per track (reported by
    disk drive, typically 63 for 28-bit LBA)
LBA addresses can be mapped to CHS tuples with the following formula 
    ("mod" is the modulo operation, i.e. the remainder, and "÷" is 
    integer division, i.e. the quotient of the division where any 
    fractional part is discarded):

    C = LBA ÷ (HPC × SPT)
    H = (LBA ÷ SPT) mod HPC
    S = (LBA mod SPT) + 1

In your code SPT = 18, HPC = 2. If we use an LBA of 19 we compute a CHS of C=0, H=1, S=2. If we look at the values you passed into the registers (CL, CH, DH) above we'd discover you used a CHS of C=1, H=0, S=1. This happens to be LBA 36, not 19. The issue is that your calculations are wrong. In particular .load_root:

div byte [SectorsPerTrack]
mov ch, ah              ; Store quotient in ch for cylinder number
mov cl, al              ; Store remainder in cl for sector number
[snip]
xor dh, dh              ; head 0
mov dl, [boot_device]   ; boot device
mov ah, 0x02            ; select read mode
int 13h

Unfortunately this isn't a correct way of calculating CHS from an LBA. You have a similar issue with .load_fat but you get lucky that you compute the right value. You are reading from the wrong sectors on the disk and that is causing data to be loaded at 0xA200 that you aren't expecting.

---

Translation of LBA to CHS

What you need is a proper LBA to CHS conversion routine. Since you will need such a function for different aspect of navigating FAT12 file structures it is best to create a function. We'll call it lba_to_chs.

Before we write such code we should revisit the equation earlier:

C = LBA ÷ (HPC × SPT)
H = (LBA ÷ SPT) mod HPC
S = (LBA mod SPT) + 1

We could implement this as is, but if we rework the equation for cylinders we can reduce the amount of work we have to do. C = LBA ÷ (HPC × SPT) can be rewritten as:

C = LBA ÷ (HPC × SPT)
C = LBA ÷ (SPT × HPC)
C = (LBA ÷ SPT) × (1 ÷ HPC)
C = (LBA ÷ SPT) ÷ HPC

If we now look at the revised formula we have:

C = (LBA ÷ SPT) ÷ HPC
H = (LBA ÷ SPT) mod HPC
S = (LBA mod SPT) + 1

Now we should notice that (LBA ÷ SPT) is duplicated in two places. We only have to do that equation once. As well since x86 DIV instruction computes the remainder and quotient at the same time we also end up computing LBA mod SPT for free when we do (LBA ÷ SPT). The code would follow this structure:

1. Compute LBA DIV SPT . This yields:
   • (LBA ÷ SPT) in the quotient
   • (LBA mod SPT) in the remainder
2. Take the remainder from step (1) and put in temporary register
3. Add 1 to the temporary in step (2). That register now contains the sector as computed by S = (LBA mod SPT) + 1
4. Take quotient from step (1) and divide by HPC.
   • Cylinder number will be the quotient
   • Head will be the remainder.

We have reduced the equation down to a couple DIV instructions and an increment/add. We can simplify things more. If we assume we are using well known IBM Compatible Disk formats then we can also say that Sectors per Track (SPT), Heads(HPC), Cylinder, Head, and Sector will always be less than 256. When the maximum LBA on any well known floppy disk format is divided by SPT the result will always be less than 256. Knowing this allows us to avoid bit twiddling the top two bits of the cylinder and placing them in the top two bits of CL. We can also use DIV instructions that do 16-bit by 8-bit unsigned division.

---

Translation Code

If we take the pseudo code above we can create a rather small lba_to_chs function that takes an LBA and converts it to CHS and works for all well known IBM compatible floppy disk formats.

;    Function: lba_to_chs
; Description: Translate Logical block address to CHS (Cylinder, Head, Sector).
;              Works for all valid FAT12 compatible disk geometries.
;
;   Resources: http://www.ctyme.com/intr/rb-0607.htm
;              https://en.wikipedia.org/wiki/Logical_block_addressing#CHS_conversion
;              https://stackoverflow.com/q/45434899/3857942
;              Sector    = (LBA mod SPT) + 1
;              Head      = (LBA / SPT) mod HEADS
;              Cylinder  = (LBA / SPT) / HEADS
;
;      Inputs: SI = LBA
;     Outputs: DL = Boot Drive Number
;              DH = Head
;              CH = Cylinder (lower 8 bits of 10-bit cylinder)
;              CL = Sector/Cylinder
;                   Upper 2 bits of 10-bit Cylinders in upper 2 bits of CL
;                   Sector in lower 6 bits of CL
;
;       Notes: Output registers match expectation of Int 13h/AH=2 inputs
;
lba_to_chs:
    push ax                    ; Preserve AX
    mov ax, si                 ; Copy LBA to AX
    xor dx, dx                 ; Upper 16-bit of 32-bit value set to 0 for DIV
    div word [SectorsPerTrack] ; 32-bit by 16-bit DIV : LBA / SPT
    mov cl, dl                 ; CL = S = LBA mod SPT
    inc cl                     ; CL = S = (LBA mod SPT) + 1
    xor dx, dx                 ; Upper 16-bit of 32-bit value set to 0 for DIV
    div word [NumberOfHeads]   ; 32-bit by 16-bit DIV : (LBA / SPT) / HEADS
    mov dh, dl                 ; DH = H = (LBA / SPT) mod HEADS
    mov dl, [boot_device]      ; boot device, not necessary to set but convenient
    mov ch, al                 ; CH = C(lower 8 bits) = (LBA / SPT) / HEADS
    shl ah, 6                  ; Store upper 2 bits of 10-bit Cylinder into
    or  cl, ah                 ;     upper 2 bits of Sector (CL)
    pop ax                     ; Restore scratch registers
    ret

You can use this lba_to_chs functions and integrate it into your .load_fat and .load_root code. Your code could look like:

;;;Start loading File Allocation Table (FAT)
.load_fat:
    mov ax, 0x07c0             ; address from start of programs
    mov es, ax
    mov ah, 0x02               ; set to read
    mov al, [SectorsPerFAT]    ; how many sectors to load

    mov si, [ReservedSectors]  ; Load FAT1 into SI for input to lba_to_chs
    call lba_to_chs            ; Retrieve CHS parameters and boot drive for LBA

    mov bx, 0x0200             ; read data to 512B after start of code
    int 13h
    cmp ah, 0
    je .load_root
    mov si, error_text
    call print
    hlt

;;;Start loading root directory
.load_root:
    mov si, fat_loaded
    call print
    xor ax, ax
    mov al, [FATcount]
    mul word [SectorsPerFAT]
    add ax, [ReservedSectors]  ; Compute LBA of oot directory entries
    mov si, ax                 ; Copy LBA to SI for later call to lba_to_chs

    mul word [BytesPerSector]
    mov bx,ax                  ; Load to after BOTH FATs in memory

    mov ax, 32
    cwd                        ; Zero dx for division
                               ;     (works since AX(32) < 0x8000)
    mul word [MaxDirEntries]
    div word [BytesPerSector]  ; number of sectors to read

    call lba_to_chs            ; Retrieve CHS values and load boot drive
    mov ah, 0x02
    int 13h
    cmp ah, 0
    je .load_OS
    mov si, error_text
    call print
    jmp $