From: Duje Mihanović Date: Sat, 30 Apr 2022 14:35:18 +0000 (+0200) Subject: Fixes and improvements to FAT32 driver X-Git-Tag: 0.1.0~10 X-Git-Url: http://git.dujemihanovic.xyz/html/static/%7B%7B%20%24.Site.BaseURL%20%7D%7Dposts/%7B%7B%20.Permalink%20%7D%7D?a=commitdiff_plain;h=2a70d72730b91905543e5d5e850bc00cf43bc5e6;p=nameless-os.git Fixes and improvements to FAT32 driver * Fix issue where multiple clusters would be loaded 0x800 apart rather than contiguously. * Get rid of some global variables, saving space. * si points to the start LBA address rather than the partition table entry, saving a bit of space as we no longer need to use offsets. * In read_cluster_chain, drop push/popad as they in fact do not save the upper 16 bytes of the registers. Instead, just push and pops the registers that are used. --- diff --git a/boot/x86/fat32-structs.s b/boot/x86/fat32-structs.s index 23980ba..e72735d 100644 --- a/boot/x86/fat32-structs.s +++ b/boot/x86/fat32-structs.s @@ -1,15 +1,6 @@ ; FAT32 data structures ; Because they have to be defined before use -struc part_entry - .attrib: resb 1 - .chs_start: resb 3 - .type: resb 1 - .chs_end: resb 3 - .lba_start: resd 1 - .lba_end: resd 1 -endstruc - struc dir_entry .name: resb 11 .attr: resb 1 diff --git a/boot/x86/fat32.s b/boot/x86/fat32.s index 31288cb..caed036 100644 --- a/boot/x86/fat32.s +++ b/boot/x86/fat32.s @@ -1,12 +1,20 @@ ; FAT32 driver -; BOOT_DRIVE, part_table_entry and bp must be set up by the code using this driver. -; first_data_sec is to be set up by calling get_1st_data_sec. +; BOOT_DRIVE, si and bp must be set up by the code using this driver. +; (Specifically, si must be a pointer to the partition's partition +; table entry and bp must be a pointer to the BPB.) +; Before reading any cluster chains, callers must call get_1st_data_sec +; and must be careful not to trash ecx after calling it. ; eax - start cluster number ; es:di - where to load the cluster chain ; NOTE: Cluster chain might be the root directory or a file. read_cluster_chain: - pusha + push eax + push ebx + push ecx + push edx + push es + push di .loop: ; get the first sector of the cluster to read push eax @@ -14,13 +22,15 @@ read_cluster_chain: xor ebx, ebx mov bl, BPB_SecPerClus mul ebx - add eax, [first_data_sec] + add eax, ecx mov ebx, eax pop eax ; read the cluster + push cx movzx cx, BPB_SecPerClus call read_sectors + pop cx ; increment the load offset (and segment if needed) push eax @@ -36,22 +46,21 @@ read_cluster_chain: .get_next_cluster: pop eax ; get FAT sector number and offset containing the next cluster number - xor edx, edx - mov ebx, 4 - mul dword ebx + xor ebx, ebx + mov bl, 4 ; put 4 in EBX, doing this instead of "mov ebx, 4" saves us an entire byte + mul ebx movzx ebx, word BPB_BytsPerSec - div dword ebx + div ebx movzx ebx, word BPB_RsvdSecCnt add eax, ebx - push bp - mov bp, [part_table_entry] - add eax, [bp+part_entry.lba_start] - pop bp + add eax, [si] mov ebx, edx ; reminder for myself: EAX is FAT sector number, (E)BX is offset into FAT sector ; load the FAT sector we're looking for + push cx + push di push ebx ; offset push es ; we want to read at 0:1000, not STAGE3_SEGMENT:1000 push eax ; desired LBA @@ -65,15 +74,20 @@ read_cluster_chain: ; find the cluster we're looking for pop es ; restore STAGE3_SEGMENT pop ebx ; pop FAT offset back into EBX for cmp - cmp dword [di+bx], 0xffffff7 - jge .done ; if cluster number is greater than or equal to the defective cluster value, we're done - ; TODO: should this perhaps be changed so that equal makes it error out? - ; otherwise, load the next sector number in eax and read again mov eax, [di+bx] - jmp .loop + pop di + pop cx + cmp eax, 0xffffff7 + jl .loop ; if cluster number is lower than the defective cluster value, we're not done + ; TODO: perhaps check is it equal defective and error out in that case? .done: ; cleanup and return - popa + pop di + pop es + pop edx + pop ecx + pop ebx + pop eax ret ; es:di - where to load the sector(s) @@ -101,18 +115,17 @@ read_sectors: hlt jmp $-1 -; no arguments +; ds:si - pointer to partition table entry +; RETURN: ecx - first data sector get_1st_data_sec: push eax push ebx - xor eax, eax movzx ax, BPB_NumFats mul dword BPB_FatSz32 ; space occupied by all FATs movzx ebx, word BPB_RsvdSecCnt add eax, ebx ; space occupied by the reserved area - mov bx, [part_table_entry] - add eax, [bx+part_entry.lba_start] ; space before the partition - mov [first_data_sec], eax + add eax, [si] ; space before the partition + mov ecx, eax pop ebx pop eax ret @@ -120,5 +133,3 @@ get_1st_data_sec: read_error: db "Read error", 0 BOOT_DRIVE: db 0 -part_table_entry: dw 0 -first_data_sec: dd 0 diff --git a/boot/x86/mbr.s b/boot/x86/mbr.s index 8089b2f..613b248 100644 --- a/boot/x86/mbr.s +++ b/boot/x86/mbr.s @@ -5,7 +5,7 @@ cpu 686 org 0x600 _start: - cli + cli ; we really don't want to be interrupted during relocation ; set up segment registers xor ax, ax mov ds, ax @@ -27,55 +27,51 @@ real_start: call check_int13_ext mov byte [BOOT_DRIVE], dl ; look for active partition - mov bx, part_1 + mov si, part_1 mov cx, 4 .check_part_loop: - mov al, [bx] + mov al, [si] test al, 0x80 jnz .part_found - add bx, 16 + add si, 16 loop .check_part_loop .error: mov si, no_os call print - cli hlt jmp short $-1 .part_found: ; look for any other active partitions, if they exist the partition table is invalid cmp cx, 0 je .load_vbr - push bx + push si .look_other_loop: - add bx, 16 - mov al, [bx] + add si, 16 + mov al, [si] test al, 0x80 jnz .invalid_mbr loop .look_other_loop .load_vbr: ; load active partition's VBR - pop bx + pop si mov ax, 0x7c0 - mov edx, [bx+0x8] - push bx ; save pointer to partition table entry + add si, 8 + mov edx, [si] xor bx, bx call read_sector ; check is the VBR bootable (ends with 0x55 0xaa), if not halt cmp word [0x7dfe], 0xaa55 jne .not_bootable mov dl, [BOOT_DRIVE] - pop si ; hand off partition table entry to VBR - jmp 0x7c00 + call 0x7c00 .not_bootable: mov si, no_os call print - cli hlt jmp short $-1 .invalid_mbr: mov si, invalid_mbr call print - cli hlt jmp short $-1 check_int13_ext: @@ -92,7 +88,6 @@ check_int13_ext: .no_ext: mov si, no_int13_ext call print - cli hlt jmp short $-1 @@ -119,7 +114,6 @@ read_sector: .error: mov si, read_fail call print - cli hlt jmp short $-1 diff --git a/boot/x86/vbr-fat32.s b/boot/x86/vbr-fat32.s index 061ca52..b4c7148 100644 --- a/boot/x86/vbr-fat32.s +++ b/boot/x86/vbr-fat32.s @@ -24,7 +24,6 @@ real_start: ; we expect the boot drive to be in DL and our partition table entry in DS:SI mov [BOOT_DRIVE], dl - mov [part_table_entry], si ; calculate the 1st sector of the data area call get_1st_data_sec @@ -37,7 +36,9 @@ real_start: mov eax, BPB_RootClus call read_cluster_chain + push cx mov cx, 11 + push si .find_stage_3: mov si, STAGE3_NAME cmp byte [es:di], 0 ; we have no more entries to look at @@ -53,11 +54,14 @@ real_start: add bx, 0x1000 mov es, bx .stage3_found: - sub di, 0xb ; "repe cmpsb" incremented this + pop si + pop cx + add di, 9 ; knowing that cmpsb incremented this by 11, we can also increment it by 9 + ; right here so we can use 1 less offset below ; stage 3 has been found and ES:DI points to its directory entry - mov ax, [es:di+dir_entry.firstclushi] ; load high half of the 1st cluster + mov ax, [es:di] ; load high half of the 1st cluster shl eax, 16 - mov ax, [es:di+dir_entry.firstcluslo] ; load low half of the 1st cluster + mov ax, [es:di+(dir_entry.firstcluslo-dir_entry.firstclushi)] ; load low half of the 1st cluster mov bx, STAGE3_SEGMENT mov es, bx mov di, STAGE3_OFFSET @@ -70,7 +74,6 @@ real_start: mov si, stage3_missing call print .halt: - cli hlt jmp short $-1