mcp_fs

A filesystem.

#define MY_BLOCK_SIZE 2048
#define MY_BLOCK_COUNT 16
static uint8_t block_buf[MY_BLOCK_SIZE];
static int read_block_cb(void * cb_ctx, int block_index){
	my_read_block(block_buf, block_index);
	return 0;
}
static int write_block_cb(void * cb_ctx, int block_index) {
	my_write_block(block_buf, block_index);
	return 0;
}
static uint8_t bb[4][MFS_BIT_BUF_SIZE_BYTES(MY_BLOCK_COUNT)];
static const mfs_conf_t conf = {
	.block_buf = block_buf,
	.bit_bufs = {bb[0], bb[1], bb[2], bb[3]},
	.block_size = MY_BLOCK_SIZE,
	.block_count = MY_BLOCK_COUNT,
	.cb_ctx = NULL,
	.read_block = read_block_cb,
	.write_block = write_block_cb
};
static mfs_t mfs;
int err = mfs_mount(&mfs, &conf);
assert(err = 0);
err = mfs_open(&mfs, "foo.txt", MFS_MODE_WRITE);
assert(err = 0);
/* ... */

Properties:

• No directories
• Only one open file at a time.
• Mounting an existing volume requires that all the blocks are scanned. Only practical for small volumes.
• 1 file, 1 or more blocks. No metadata blocks.
• A file can be stored across blocks that are not contiguous. This is fairly standard.
• Re-writing an existing file is power-failure tollerant. The changes are committed atomically. The original file contents will always be seen after a power interruption. Only a successful file close commits the new file contents.