* Create dirs from userspace * Add delete syscall * Add name to FileStat * Add offset to Dir * Add dynamic size to dir * Add FileStat (de)serializer * Add ReadDir#block_index * Add FileStat::root * Implement FileIO#read for Dir * Add api::fs::read_dir * Use read_dir for list command * Allow read /path/to/dir/ commands * Save dir size to dir entry * Fix return value of api::fs::create_device * Read dir from userspace * Add proc command * Close file handle when creating dir * Add /dev/null * Remove path restriction * Allow listing of singular file * Update doc * Re-export fs struc in API * Use read_dir in shell autocomplete * Use fs::delete in httpd * Rename stat to info * Update doc * Rename attributes
5.4 KiB
MOROS Filesystem
Hard drive
A hard drive is separated in blocks of 512 bytes, grouped into 4 areas:
+------------+
| Boot | (4096 blocks)
+------------+
| Superblock | (2 blocks)
+------------+
| Bitmap | (n / (8 * 512) blocks)
+------------+
| Data | (n blocks)
+------------+
The first area contains the bootloader and the kernel, the second is a superblock with a magic string to identify the file system, the third is a bitmap mapping the allocated data blocks of the last area.
A location on the tree of dirs and files is named a path:
- The root dir is represented by a slash:
/
- A dir inside the root will have its name appended to the slash:
/usr
- Subsequent dirs will append a slash and their names:
/usr/admin
Creation with QEMU
$ qemu-img create disk.img 128M
Formatting 'disk.img', fmt=raw size=134217728
Setup in diskless console
During boot MOROS will detect any hard drives present on the ATA buses, then
look for a filesystem on those hard drives. If no filesystem is found, MOROS
will open a console in diskless mode to allow the user to create one with
the disk format
command:
> disk format /dev/ata/0/0
This command will format the first disk on the first ATA bus by writing a magic string in a superblock, mounting the filesystem, and allocating the root directory.
The next step during setup is to create the directory structure:
> write /bin/ # Binaries
> write /dev/ # Devices
> write /ini/ # Initialisation files
> write /lib/ # Libraries
> write /net/ # Network
> write /src/ # Sources
> write /tmp/ # Temporary files
> write /usr/ # User directories
> write /var/ # Variable files
Then the following should be added to the boot script with the
command edit /ini/boot.sh
to allow MOROS to finish booting:
user login
shell
Finally a user can be created with the following command:
> user create
All of this can be made more easily by running the install
command instead.
This installer will also add additional files contained in the dsk
repository of the source code, like a nice login banner :)
Data Structures
BlockBitmap
Bitmap of allocated blocks in the data area.
Block
A block is small area of 512 bytes on a hard drive, and it is also part of linked list representing a file or a directory.
The first 4 bytes of a block is the address of the next block on the list and the rest of block is the data stored in the block.
Structure:
0
0 1 2 3 4 5 6 n
+-+-+-+-+-+-+-+ // +-+
| addr | data |
+-+-+-+-+-+-+-+ // +-+
n = 512
Superblock
0 1 2
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 n
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ // +-+
| signature |v|b| count | alloc | reserved |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ // +-+
signature = "MOROS FS"
v = version number of the FS
b = size of a block in 2 ^ (9 + b) bytes
count = number of blocks
alloc = number of allocated blocks
File
The first block of a contains the address of the next block where its contents is stored and the beginning of its contents in the rest of the block.
If all contents can fit into one block the address of the next block will be empty.
Structure:
0
0 1 2 3 4 5 6 7 8 n
+-+-+-+-+-+-+-+-+-+ // +-+
| addr | contents |
+-+-+-+-+-+-+-+-+-+ // +-+
n = 512
Dir
The first block of a directory contains the address of the next block where its directory entries are stored and the first entries in the rest of the block.
If all entries can fit into one block the address of the next block will be empty.
Structure:
0 1
0 1 2 3 4 5 6 7 8 9 0 n
+-+-+-+-+-+-+-+-+-+-+-+ // +-+-+-+-+-+-+-+-+ // +-+
| addr | dir entry 1 | dir entry 2 |
+-+-+-+-+-+-+-+-+-+-+-+ // +-+-+-+-+-+-+-+-+ // +-+
n = 512
DirEntry
A directory entry represents a file or a directory contained inside a directory. Each entry use a variable number of bytes that must fit inside the data of one block. Those bytes represent the kind of entry (file or dir), the address of the first block, the filesize (max 4GB), the last modified time in seconds since Unix Epoch, the length of the filename, and the filename (max 255 chars) of the entry.
Structure:
0 1 2
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 m
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ // +-+
|k| addr | size | time |n| name buffer |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ // +-+
k = kind of entry
n = length of name buffer
m = 17 + n
FileInfo
The info
syscall on a file or directory and the read
syscall on a directory
return a subset of a directory entry for userspace programs.
Structure:
0 1 2
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 m
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ // +-+
|k| size | time |n| name buffer |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ // +-+
k = kind of entry
n = length of name buffer
m = 13 + n