path: root/net9p_defs.gen

#!/usr/bin/env python

import enum
import re


# Parse net9p_defs.txt #########################################################

class Atom(enum.Enum):
    u8 = 1
    u16 = 2
    u32 = 3
    u64 = 4


class Struct:
    name: str
    members: list["Member"]


class List:
    cnt: str
    typ: Atom | Struct

    def __init__(self, /, *, cnt: str, typ: Atom | Struct) -> None:
        self.cnt = cnt
        self.typ = typ


class Member:
    name: str
    typ: Atom | Struct | List

    def __init__(self, /, *, name: str, typ: Atom | Struct | List) -> None:
        self.name = name
        self.typ = typ


def parse_members(
    env: dict[str, Atom | Struct], existing: list[Member], specs: str
) -> list[Member]:
    ret = existing
    for spec in specs.split():
        m = re.fullmatch(r"(.+)\[([^*]+)(?:\*([^*]+))?\]", spec)
        if not m:
            raise SyntaxError(f"invalid member spec {repr(spec)}")
        if m.group(2) not in env:
            raise NameError(f"Unknown type {repr(m.group(2))}")
        name = m.group(1)
        typ = env[m.group(2)]
        if any(x.name == name for x in ret):
            raise ValueError(f"duplicate member name {repr(name)}")
        if cnt := m.group(3):
            if len(ret) == 0 or ret[-1].name != cnt:
                raise ValueError(f"list count must be previous item: {repr(cnt)}")
            if not isinstance(ret[-1].typ, Atom):
                raise ValueError(f"list count must be an integer type: {repr(cnt)}")
            ret += [Member(name=name, typ=List(cnt=cnt, typ=typ))]
        else:
            ret += [Member(name=name, typ=typ)]
    return ret


class Message:
    id: int
    name: str
    members: list[Member]


def parse_file(filename: str) -> tuple[list[Struct], list[Message]]:
    msgs: list[Message] = []
    env: dict[str, Atom | Struct] = {
        "1": Atom.u8,
        "2": Atom.u16,
        "4": Atom.u32,
        "8": Atom.u64,
    }
    with open(filename, "r") as fh:
        prev: Struct | Message | None = None
        for line in fh:
            line = line.split("#", 1)[0].strip()
            if not line:
                continue
            if m := re.fullmatch(r'([0-9]+)\s*=\s*(\S+)\s*"([^"]*)"', line):
                msg = Message()
                msg.id = int(m.group(1))
                msg.name = m.group(2)
                msg.members = parse_members(env, [], m.group(3))
                msgs += [msg]
                prev = msg
            elif m := re.fullmatch(r'(\S+)\s*=\s*"([^"]*)"', line):
                struct = Struct()
                struct.name = m.group(1)
                struct.members = parse_members(env, [], m.group(2))
                env[struct.name] = struct
                prev = struct
            elif m := re.fullmatch(r'"([^"]*)"', line):
                if not prev:
                    raise SyntaxError(
                        "a continuation line must come after a struct line"
                    )
                prev.members = parse_members(env, prev.members, line.strip('"'))
            else:
                raise SyntaxError(f"invalid line {repr(line)}")
    structs = [x for x in env.values() if isinstance(x, Struct)]
    return structs, msgs

# Generate C ###################################################################

def shortname(typ: Atom | Struct | Message) -> str:
    match typ:
        case Atom.u8:
            return "1"
        case Atom.u16:
            return "2"
        case Atom.u32:
            return "4"
        case Atom.u64:
            return "8"
        case Struct():
            return typ.name
        case Message():
            return 'msg_'+typ.name
        case _:
            raise ValueError(f"not a type: {typ.__class__.__name__}")

def c_typename(typ: Atom | Struct | List | Message) -> str:
    match typ:
        case Atom.u8:
            return "uint8_t"
        case Atom.u16:
            return "uint16_t"
        case Atom.u32:
            return "uint32_t"
        case Atom.u64:
            return "uint64_t"
        case Struct():
            return "struct v9fs_" + typ.name
        case Message():
            return "struct v9fs_msg_" + typ.name
        case List():
            return c_typename(typ.typ) + "*"
        case _:
            raise ValueError(f"not a type: {typ.__class__.__name__}")


def static_size(typ: Atom | Struct | List | Message) -> int | None:
    match typ:
        case Atom.u8:
            return 1
        case Atom.u16:
            return 2
        case Atom.u32:
            return 4
        case Atom.u64:
            return 8
        case Struct() | Message():
            size = 0
            for member in typ.members:
                msize = static_size(member.typ)
                if msize is None:
                    return None
                size += msize
            return size
        case List():
            return None
        case _:
            raise ValueError(f"not a type: {typ.__class__.__name__}")

        
def gen_h(structs: list[Struct], msgs: list[Message]) -> str:
    ret = ""
    ret += "/* Generated by ./net9p_defs.gen.  DO NOT EDIT!  */\n"
    ret += "\n"
    ret += "#ifndef _NET9P_DEFS_H_\n"
    ret += "#define _NET9P_DEFS_H_\n"
    ret += "\n"

    for struct in structs:
        ret += c_typename(struct) + " {\n"
        typewidth = max(len(c_typename(member.typ)) for member in struct.members)
        for member in struct.members:
            ret += f"\t{c_typename(member.typ).ljust(typewidth)} {member.name};\n"
        ret += "};\n"
        ret += "\n"

    ret += "enum v9fs_msg_type {\n"
    namewidth = max(len(msg.name) for msg in msgs)
    for msg in msgs:
        ret += f"\tV9FS_TYP_{msg.name.ljust(namewidth)} = {msg.id},\n"
    ret += "};\n"

    for msg in msgs:
        if not msg.members:
            ret += c_typename(msg) + " {};\n"
        else:
            ret += c_typename(msg) + " {\n"
            typewidth = max(len(c_typename(member.typ)) for member in msg.members)
            for member in msg.members:
                ret += f"\t{c_typename(member.typ).ljust(typewidth)} {member.name};\n"
            ret += "};\n"
        ret += "\n"

    ret += "#endif /* _NET9P_DEFS_H_ */\n"
    return ret


def gen_c(structs: list[Struct], msgs: list[Message]) -> str:
    ret = """
/* Generated by ./net9p_defs.gen.  DO NOT EDIT!  */

#include <stdint.h> /* for size_t, uint{n}_t */
#include <stdlib.h> /* for malloc() */

#include "net9p_defs.h"
"""

    # basic utilities ##########################################################
    ret += """
/* basic utilities ************************************************************/

#define UNUSED __attribute__ ((unused))

static inline uint16_t decode_u16le(uint8_t *bytes) {
        return (((uint16_t)(bytes[0])) <<  0)
             | (((uint16_t)(bytes[1])) <<  8)
             ;
}
static inline uint32_t decode_u32le(uint8_t *bytes) {
        return (((uint16_t)(bytes[0])) <<  0)
             | (((uint16_t)(bytes[1])) <<  8)
             | (((uint16_t)(bytes[2])) << 16)
             | (((uint16_t)(bytes[3])) << 24)
             ;
}
static inline uint64_t decode_u64le(uint8_t *bytes) {
        return (((uint16_t)(bytes[0])) <<  0)
             | (((uint16_t)(bytes[1])) <<  8)
             | (((uint16_t)(bytes[2])) << 16)
             | (((uint16_t)(bytes[3])) << 24)
             | (((uint16_t)(bytes[4])) << 32)
             | (((uint16_t)(bytes[5])) << 40)
             | (((uint16_t)(bytes[6])) << 48)
             | (((uint16_t)(bytes[7])) << 56)
             ;
}

static inline void decode_u16le(val uint16_t, uint8_t *bytes) {
	bytes[0] = (uint8_t)((val >>  0) & 0xFF);
	bytes[1] = (uint8_t)((val >>  8) & 0xFF);
}
static inline void decode_u32le(val uint32_t, uint8_t *bytes) {
	bytes[0] = (uint8_t)((val >>  0) & 0xFF);
	bytes[1] = (uint8_t)((val >>  8) & 0xFF);
	bytes[2] = (uint8_t)((val >> 16) & 0xFF);
	bytes[3] = (uint8_t)((val >> 24) & 0xFF);
}
static inline void decode_u64le(val uint64_t, uint8_t *bytes) {
	bytes[0] = (uint8_t)((val >>  0) & 0xFF);
	bytes[1] = (uint8_t)((val >>  8) & 0xFF);
	bytes[2] = (uint8_t)((val >> 16) & 0xFF);
	bytes[3] = (uint8_t)((val >> 24) & 0xFF);
	bytes[4] = (uint8_t)((val >> 32) & 0xFF);
	bytes[5] = (uint8_t)((val >> 40) & 0xFF);
	bytes[6] = (uint8_t)((val >> 48) & 0xFF);
	bytes[7] = (uint8_t)((val >> 56) & 0xFF);
}
"""

    # checksize ################################################################
    ret += """
/* checksize ******************************************************************/

typedef bool (*_checksize_fn_t)(uint32_t net_len, uint8_t *net_bytes, uint32_t *mut_net_offset, size_t *mut_host_extra);
static inline bool _checksize_list(size_t cnt, checksize_fn_t fn, size_t host_size,
                                   uint32_t net_len, uint8_t *net_bytes, uint32_t *mut_net_offset, size_t *mut_host_extra) {
	for (size_t i = 0; i < cnt; i++)
		if (__builtin_add_overflow(*mut_host_extra, host_size, mut_host_extra)
		 || fn(net_len, net_bytes, mut_net_offset, mut_host_extra))
			return true;
	return false;
}
static inline bool checksize_1(uint32_t net_len, uint8_t *net_bytes UNUSED, uint32_t *mut_net_offset, size_t *mut_host_extra UNUSED) {
	return __builtin_add_overflow(*mut_net_offset, 1, mut_net_offset) || net_len < *mut_net_offset;
}
static inline bool checksize_2(uint32_t net_len, uint8_t *net_bytes UNUSED, uint32_t *mut_net_offset, size_t *mut_host_extra UNUSED) {
	return __builtin_add_overflow(*mut_net_offset, 2, mut_net_offset) || net_len < *mut_net_offset;
}
static inline bool checksize_4(uint32_t net_len, uint8_t *net_bytes UNUSED, uint32_t *mut_net_offset, size_t *mut_host_extra UNUSED) {
	return __builtin_add_overflow(*mut_net_offset, 4, mut_net_offset) || net_len < *mut_net_offset;
}
static inline bool checksize_8(uint32_t net_len, uint8_t *net_bytes UNUSED, uint32_t *mut_net_offset, size_t *mut_host_extra UNUSED) {
	return __builtin_add_overflow(*mut_net_offset, 8, mut_net_offset) || net_len < *mut_net_offset;
}
"""
    for struct in structs + msgs:
        argattr = ' UNUSED' if len(struct.members) == 0 else ''
        ret += f"static inline bool checksize_{shortname(struct)}(uint32_t net_len{argattr}, uint8_t *net_bytes{argattr}, uint32_t *mut_net_offset{argattr}, size_t mut_host_extra{argattr}) {{"
        if len(struct.members) == 0:
            ret += "}\n"
            continue
        prefix0 = "\treturn "
        prefix1 = "\t    || "
        prefix2 = "\t       "
        prefix = prefix0
        prev_size = 0
        for member in struct.members:
            if isinstance(member.typ, List):
                ret += f"\n{prefix }_checksize_list(decode_u{prev_size*8}le(&net_bytes[(*mut_net_offset)-{prev_size}]), checksize_{shortname(member.typ.typ)}, sizeof({c_typename(member.typ.typ)}),"
                ret += f"\n{prefix2}                net_len, net_bytes, mut_net_offset, mut_host_extra)"
            else:
                ret += f"\n{prefix}checksize_{shortname(member.typ)}(net_len, net_bytes, mut_net_offset, mut_host_extra)"
            prefix = prefix1
        if struct.name == 's':
            ret += f"\n{prefix}__builtin_add_overflow(*mut_host_extra, 1, mut_host_extra)"
        ret += ";\n}\n"

    # unmarshal ################################################################
    ret += """
/* unmarshal ******************************************************************/
/* checksize_XXX() should be called before unmarshal_XXX().  */

static inline void unmarshal_1(uint32_t net_len, uint8_t *net_bytes UNUSED, uint32_t *mut_net_offset, void *mut_host_extra UNUSED, uint8_t *out) {
	*out = decode_u8le(&net_bytes[*mut_net_offset]);
	*mut_net_offset += 1;
}
static inline void unmarshal_2(uint32_t net_len, uint8_t *net_bytes UNUSED, uint32_t *mut_net_offset, void *mut_host_extra UNUSED, uint16_t *out) {
	*out = decode_u16le(&net_bytes[*mut_net_offset]);
	*mut_net_offset += 2;
}
static inline void unmarshal_4(uint32_t net_len, uint8_t *net_bytes UNUSED, uint32_t *mut_net_offset, void *mut_host_extra UNUSED, uint32_t *out) {
	*out = decode_u32le(&net_bytes[*mut_net_offset]);
	*mut_net_offset += 4;
}
static inline void unmarshal_8(uint32_t net_len, uint8_t *net_bytes UNUSED, uint32_t *mut_net_offset, void *mut_host_extra UNUSED, uint64_t *out) {
	*out = decode_u64le(&net_bytes[*mut_net_offset]);
	*mut_net_offset += 8;
}
"""
    for struct in structs + msgs:
        ret += f"static inline void unmarshal_{shortname(struct)}(uint32_t net_len, uint8_t *net_bytes, uint32_t *mut_net_offset, void *mut_host_extra, {c_typename(struct)} *out) {{"
        if len(struct.members) == 0:
            ret += "}\n"
            continue
        ret += "\n"
        for member in struct.members:
            if isinstance(member.typ, List):
                ret += f"\tout->{member.name} = mut_host_extra;\n"
                ret += f"\t*mut_host_extra += sizeof(out->{member.name}) * out->{member.typ.cnt};\n"
                ret += f"\tfor (typeof(out->{member.typ.cnt}) i = 0; i < out->{member.typ.cnt}; i++)\n"
                ret += f"\t\tunmarshal_{shortname(member.typ.typ)}(net_len, net_bytes, mut_net_offset, mut_host_extra, &(out->{member.name}[i]));\n"
            else:
                ret += f"\tunmarshal_{shortname(member.typ)}(net_len, net_bytes, mut_net_offset, mut_host_extra, &(out->{member.name}));\n"
        ret += "}\n"

    ############################################################################
    return ret


################################################################################

if __name__ == "__main__":
    structs, msgs = parse_file("net9p_defs.txt")
    with open('net9p_defs.h', 'w') as fh:
        fh.write(gen_h(structs, msgs))
    with open('net9p_defs.c', 'w') as fh:
        fh.write(gen_c(structs, msgs))