net: proper socket API in the kernel (SYS_NET) - no more SLIP in userland

The network stack moves into the kernel. src/socket.asm owns SLIP framing,
IPv4, RFC1071 checksums, and ICMP; programs now speak a socket API through one
syscall (SYS_NET, DE=&netreq dispatched by op): net_socket/connect/send/recv/
close/poll (c/sock.h). No program touches SLIP, IP headers, or checksums.

- Socket table (4 sockets) + tx/rx buffers in WRAM0; our IP = 10.0.0.2.
- net_pump: drains the link, reassembles SLIP frames, demuxes IPv4. Inbound
  ICMP echo requests are auto-answered in-kernel, so the GB replies to pings
  whenever any process pumps RX.
- ICMP sockets: send() emits an echo request to the connected peer; recv()
  returns the matching reply (with a spin/yield timeout).

ping.c is now a ~15-line socket client; netd.c is just `for(;;){net_poll();
yield();}`. Verified over tunbridge:
    /# ping 1.1.1.1     -> replies from the real internet (kernel builds it all)
    host# ping 10.0.0.2 -> 4/4, 0% loss (kernel auto-answers)

Gotchas recorded: gbos.inc isn't a make dep (touch asm after editing); this
crt0 doesn't copy initializers (fill arrays at runtime); and the arg string at
0xA000 overlaps _DATA, so parse targets must sit past it (big buffer first).
UDP and TCP sockets build on this same core next.

3 files changed, 756 insertions, 0 deletions

diff --git a/src/boot.asm b/src/boot.asm
index 23511c4..c29267d 100644
--- a/src/boot.asm
+++ b/src/boot.asm
@@ -56,6 +56,7 @@ KernelInit:
     ld a, HIGH(wKernelPCB)
     ld [wCurProc+1], a
     call pipe_init              ; mark all pipes free
+    call net_init               ; clear the socket table, set our IP (10.0.0.2)
     call fs_init                ; mount the disk (format on first boot; persists)
     call term_init              ; set up the LCD text terminal
     call osk_init               ; build the on-screen keyboard (window layer)
diff --git a/src/socket.asm b/src/socket.asm
new file mode 100644
index 0000000..f0d8444
--- /dev/null
+++ b/src/socket.asm
@@ -0,0 +1,754 @@
+; =============================================================================
+; socket.asm - the kernel network stack + socket layer (SYS_NET).
+;
+; The kernel owns the link port: SLIP framing, IPv4, checksums, and the per-
+; protocol logic all live here. Userland just does socket()/connect()/send()/
+; recv()/close() via one syscall (DE = &netreq). No program touches SLIP or IP.
+;
+; Our address is 10.0.0.2. Incoming ICMP echo requests are auto-answered while
+; any process pumps RX (recv/poll), so the Game Boy responds to pings for real.
+;
+; Phase 1: ICMP (echo).  UDP and TCP build on the same frame/checksum core.
+; =============================================================================
+INCLUDE "include/gbos.inc"
+
+; ---- socket entry layout ----------------------------------------------------
+DEF SK_TYPE    EQU 0     ; 0 = free
+DEF SK_RIP     EQU 1     ; 4  remote IP
+DEF SK_RPORT   EQU 5     ; 2  remote port
+DEF SK_LPORT   EQU 7     ; 2  local port
+DEF SK_HASRX   EQU 9     ; 1  rxbuf holds a datagram
+DEF SK_RXLEN   EQU 10    ; 1  payload length
+DEF SK_SRCIP   EQU 11    ; 4  sender IP
+DEF SK_SRCPORT EQU 15    ; 2  sender port
+DEF SK_RXBUF   EQU 17    ; 96 payload
+DEF SK_SIZE    EQU 113
+
+SECTION "net_bss", WRAM0
+wSockTab::     DS MAX_SOCKS * SK_SIZE
+wNetOurIP::    DS 4
+wNetReq::      DS NR_SIZE
+wNetSockPtr::  DS 2
+wNetIhl::      DS 1
+wNetSeq::      DS 1
+wNetTx::       DS 256      ; tx build buffer
+wNetRxBuf::    DS 256      ; rx frame assembly / parse
+wNetRxLen::    DS 1
+wNetRxEsc::    DS 1
+
+SECTION "socket", ROM0
+
+; net_init - clear the socket table and set our IP (called at boot).
+net_init::
+    ld hl, wSockTab
+    ld bc, MAX_SOCKS * SK_SIZE
+.clr
+    xor a
+    ld [hl+], a
+    dec bc
+    ld a, b
+    or c
+    jr nz, .clr
+    ld a, 10
+    ld [wNetOurIP+0], a
+    xor a
+    ld [wNetOurIP+1], a
+    ld [wNetOurIP+2], a
+    ld a, 2
+    ld [wNetOurIP+3], a
+    xor a
+    ld [wNetRxLen], a
+    ld [wNetRxEsc], a
+    ld [wNetSeq], a
+    ret
+
+; -----------------------------------------------------------------------------
+; low-level link port (byte at a time). Preserve BC/DE/HL; use A only.
+net_putbyte:                    ; A = byte to transmit
+    ld [rSB], a
+    ld a, $81
+    ld [rSC], a
+.w
+    ld a, [rSC]
+    bit 7, a
+    jr nz, .w
+    ret
+
+net_getbyte_nb:                 ; -> A = byte, CF set if none
+    ld a, $80
+    ld [rSC], a
+    ld a, [rSC]
+    bit 7, a
+    jr z, .got
+    scf
+    ret
+.got
+    ld a, [rSB]
+    and a                       ; CF = 0
+    ret
+
+; -----------------------------------------------------------------------------
+; net_slip_send(HL = buf, B = len) - transmit a SLIP frame.
+net_slip_send:
+    ld a, SLIP_END
+    call net_putbyte
+.loop
+    ld a, b
+    or a
+    jr z, .end
+    ld a, [hl+]
+    cp SLIP_END
+    jr z, .esc_end
+    cp SLIP_ESC
+    jr z, .esc_esc
+    call net_putbyte
+.cont
+    dec b
+    jr .loop
+.esc_end
+    ld a, SLIP_ESC
+    call net_putbyte
+    ld a, SLIP_ESC_END
+    call net_putbyte
+    jr .cont
+.esc_esc
+    ld a, SLIP_ESC
+    call net_putbyte
+    ld a, SLIP_ESC_ESC
+    call net_putbyte
+    jr .cont
+.end
+    ld a, SLIP_END
+    call net_putbyte
+    ret
+
+; -----------------------------------------------------------------------------
+; net_cksum(HL = buf, B = len, DE = seed) -> HL = inverted 16-bit checksum.
+; One's-complement sum (RFC 1071), carry-folded each add. len <= 255.
+net_cksum:
+.loop
+    ld a, b
+    cp 2
+    jr c, .tail
+    ld a, [hl+]                 ; high byte
+    ld c, a
+    ld a, [hl+]                 ; low byte
+    add e
+    ld e, a
+    ld a, c
+    adc d
+    ld d, a
+    jr nc, .nc1
+    inc de
+.nc1
+    dec b
+    dec b
+    jr .loop
+.tail
+    ld a, b
+    or a
+    jr z, .fold
+    ld a, [hl]                  ; last odd byte -> high half
+    add d
+    ld d, a
+    jr nc, .fold
+    inc e
+    jr nz, .fold
+    inc d
+.fold
+    ld a, d
+    cpl
+    ld h, a
+    ld a, e
+    cpl
+    ld l, a
+    ret
+
+; -----------------------------------------------------------------------------
+; net_sock_ptr(A = sockid) -> HL = &wSockTab[A]  (clobbers B, DE)
+net_sock_ptr:
+    ld hl, wSockTab
+    or a
+    ret z
+    ld b, a
+    ld de, SK_SIZE
+.l
+    add hl, de
+    dec b
+    jr nz, .l
+    ret
+
+; =============================================================================
+; sys_net(DE = &netreq) -> A = result
+; =============================================================================
+sys_net::
+    ld hl, wNetReq              ; copy the request out of process RAM
+    ld b, NR_SIZE
+.cp
+    ld a, [de]
+    ld [hl+], a
+    inc de
+    dec b
+    jr nz, .cp
+    ld a, [wNetReq + NR_OP]
+    cp NET_SOCKET
+    jp z, net_op_socket
+    cp NET_CONNECT
+    jp z, net_op_connect
+    cp NET_SEND
+    jp z, net_op_send
+    cp NET_RECV
+    jp z, net_op_recv
+    cp NET_CLOSE
+    jp z, net_op_close
+    cp NET_POLL
+    jp z, net_op_poll
+    ld a, $FF
+    ret
+
+; ---- socket(type) -> A = sockid or $FF ---------------------------------------
+net_op_socket:
+    ld c, 0
+.scan
+    ld a, c
+    call net_sock_ptr           ; HL = &sock[c]
+    ld a, [hl]
+    or a                        ; SK_TYPE == 0 (free)?
+    jr z, .free
+    inc c
+    ld a, c
+    cp MAX_SOCKS
+    jr c, .scan
+    ld a, $FF                   ; table full
+    ret
+.free
+    ld a, [wNetReq + NR_SOCK]   ; requested type
+    ld [hl], a                  ; SK_TYPE
+    ld a, l                     ; clear SK_HASRX
+    add SK_HASRX
+    ld l, a
+    ld a, h
+    adc 0
+    ld h, a
+    xor a
+    ld [hl], a
+    ld a, c                     ; return sockid
+    ret
+
+; ---- connect(sock, ip, port) - set remote --------------------------------
+net_op_connect:
+    ld a, [wNetReq + NR_SOCK]
+    call net_sock_ptr           ; HL = &sock
+    ld a, l
+    add SK_RIP
+    ld l, a
+    ld a, h
+    adc 0
+    ld h, a                     ; HL = &sock.RIP
+    ld de, wNetReq + NR_IP
+    ld b, 6                     ; 4 IP + 2 port (contiguous: RIP then RPORT)
+.cp
+    ld a, [de]
+    ld [hl+], a
+    inc de
+    dec b
+    jr nz, .cp
+    xor a
+    ret
+
+; ---- close(sock) -------------------------------------------------------------
+net_op_close:
+    ld a, [wNetReq + NR_SOCK]
+    call net_sock_ptr
+    xor a
+    ld [hl], a                  ; SK_TYPE = free
+    ret
+
+; ---- poll - pump RX once (answers pings) -------------------------------------
+net_op_poll:
+    call net_pump
+    xor a
+    ret
+
+; ---- send(sock, buf, len) ----------------------------------------------------
+net_op_send:
+    ld a, [wNetReq + NR_SOCK]
+    call net_sock_ptr
+    ld a, l
+    ld [wNetSockPtr], a
+    ld a, h
+    ld [wNetSockPtr+1], a
+    ld a, [hl]                  ; SK_TYPE
+    cp SOCK_ICMP
+    jp z, icmp_send
+    ld a, $FF
+    ret
+
+; ---- recv(sock, buf, maxlen) -> A = len or $FF (timeout) ---------------------
+net_op_recv:
+    ld a, [wNetReq + NR_SOCK]
+    call net_sock_ptr
+    ld a, l
+    ld [wNetSockPtr], a
+    ld a, h
+    ld [wNetSockPtr+1], a
+    ld b, 200                   ; outer timeout
+.outer
+    ld de, 0
+.inner
+    call net_pump
+    call sock_has_rx            ; Z clear (nz) if HASRX set
+    jr nz, .got
+    dec de
+    ld a, d
+    or e
+    jr nz, .inner
+    call SchedYield
+    dec b
+    jr nz, .outer
+    ld a, $FF                   ; timeout
+    ret
+.got
+    ; copy sock.RXBUF -> process NR_BUF, len = sock.RXLEN, NR_IP = sock.SRCIP
+    call sock_ptr_hl            ; HL = &sock
+    push hl
+    ld a, l
+    add SK_RXLEN
+    ld l, a
+    ld a, h
+    adc 0
+    ld h, a
+    ld a, [hl]                  ; RXLEN
+    ld c, a                     ; C = length
+    pop hl
+    ; source IP -> wNetReq.NR_IP (returned to caller via buffer? no - copy to caller req is skipped; give src via NR_IP in kernel copy only)
+    push bc
+    push hl
+    ; dst = process buf pointer
+    ld a, [wNetReq + NR_BUF]
+    ld e, a
+    ld a, [wNetReq + NR_BUF + 1]
+    ld d, a                     ; DE = process buf
+    ld a, l
+    add SK_RXBUF
+    ld l, a
+    ld a, h
+    adc 0
+    ld h, a                     ; HL = &sock.RXBUF
+    ld a, c
+    or a
+    jr z, .nocopy
+.cpl
+    ld a, [hl+]
+    ld [de], a
+    inc de
+    dec c
+    jr nz, .cpl
+.nocopy
+    pop hl                      ; &sock
+    pop bc                      ; C = length
+    ; clear HASRX
+    ld a, l
+    add SK_HASRX
+    ld l, a
+    ld a, h
+    adc 0
+    ld h, a
+    xor a
+    ld [hl], a
+    ld a, c                     ; return length
+    ret
+
+; sock_ptr_hl -> HL = current socket pointer (from wNetSockPtr)
+sock_ptr_hl:
+    ld a, [wNetSockPtr]
+    ld l, a
+    ld a, [wNetSockPtr+1]
+    ld h, a
+    ret
+
+; sock_has_rx -> Z clear (nz) if the current socket has a buffered datagram
+sock_has_rx:
+    call sock_ptr_hl
+    ld a, l
+    add SK_HASRX
+    ld l, a
+    ld a, h
+    adc 0
+    ld h, a
+    ld a, [hl]
+    or a
+    ret
+
+; =============================================================================
+; icmp_send - build + transmit an ICMP echo request from the current socket.
+; =============================================================================
+icmp_send:
+    ; IP header (20 bytes)
+    ld a, $45
+    ld [wNetTx+0], a
+    xor a
+    ld [wNetTx+1], a
+    ld [wNetTx+2], a            ; total len hi
+    ld a, [wNetReq+NR_LEN]
+    add 28
+    ld [wNetTx+3], a            ; total len lo = 28 + payload
+    xor a
+    ld [wNetTx+4], a
+    ld [wNetTx+5], a            ; id
+    ld [wNetTx+6], a
+    ld [wNetTx+7], a            ; flags/frag
+    ld a, 64
+    ld [wNetTx+8], a           ; ttl
+    ld a, 1
+    ld [wNetTx+9], a           ; proto ICMP
+    xor a
+    ld [wNetTx+10], a
+    ld [wNetTx+11], a          ; ip cksum
+    ld a, [wNetOurIP+0]
+    ld [wNetTx+12], a
+    ld a, [wNetOurIP+1]
+    ld [wNetTx+13], a
+    ld a, [wNetOurIP+2]
+    ld [wNetTx+14], a
+    ld a, [wNetOurIP+3]
+    ld [wNetTx+15], a
+    ; dst IP = sock.RIP
+    call sock_ptr_hl
+    ld a, l
+    add SK_RIP
+    ld l, a
+    ld a, h
+    adc 0
+    ld h, a                     ; HL = &sock.RIP
+    ld de, wNetTx+16
+    ld b, 4
+.dip
+    ld a, [hl+]
+    ld [de], a
+    inc de
+    dec b
+    jr nz, .dip
+    ; ICMP header (8 bytes) at wNetTx+20
+    ld a, 8
+    ld [wNetTx+20], a          ; type = echo request
+    xor a
+    ld [wNetTx+21], a          ; code
+    ld [wNetTx+22], a
+    ld [wNetTx+23], a          ; cksum
+    ld a, $12
+    ld [wNetTx+24], a          ; id
+    ld a, $34
+    ld [wNetTx+25], a
+    xor a
+    ld [wNetTx+26], a          ; seq hi
+    ld a, [wNetSeq]
+    inc a
+    ld [wNetSeq], a
+    ld [wNetTx+27], a          ; seq lo
+    ; payload from process buf
+    ld a, [wNetReq+NR_BUF]
+    ld l, a
+    ld a, [wNetReq+NR_BUF+1]
+    ld h, a
+    ld de, wNetTx+28
+    ld a, [wNetReq+NR_LEN]
+    ld b, a
+    or a
+    jr z, .nopl
+.cppl
+    ld a, [hl+]
+    ld [de], a
+    inc de
+    dec b
+    jr nz, .cppl
+.nopl
+    ; ICMP checksum over 8 + payload
+    ld a, [wNetReq+NR_LEN]
+    add 8
+    ld b, a
+    ld hl, wNetTx+20
+    ld de, 0
+    call net_cksum
+    ld a, h
+    ld [wNetTx+22], a
+    ld a, l
+    ld [wNetTx+23], a
+    ; IP checksum over 20
+    ld hl, wNetTx
+    ld b, 20
+    ld de, 0
+    call net_cksum
+    ld a, h
+    ld [wNetTx+10], a
+    ld a, l
+    ld [wNetTx+11], a
+    ; transmit: 28 + payload
+    ld a, [wNetReq+NR_LEN]
+    add 28
+    ld b, a
+    ld hl, wNetTx
+    call net_slip_send
+    xor a
+    ret
+
+; =============================================================================
+; net_pump - read available serial bytes, assemble SLIP frames, process each.
+; Non-blocking: returns when the link has no more bytes ready.
+; =============================================================================
+net_pump:
+.next
+    call net_getbyte_nb
+    ret c                       ; no byte ready
+    ; SLIP decode using wNetRxLen / wNetRxEsc
+    cp SLIP_END
+    jr z, .frame
+    ld c, a                     ; save byte
+    ld a, [wNetRxEsc]
+    or a
+    jr nz, .escaped
+    ld a, c
+    cp SLIP_ESC
+    jr z, .setesc
+    jr .store
+.escaped
+    ld a, c
+    cp SLIP_ESC_END
+    jr nz, .notend
+    ld c, SLIP_END
+.notend
+    ld a, c
+    cp SLIP_ESC_ESC
+    jr nz, .noteesc
+    ld c, SLIP_ESC
+.noteesc
+    xor a
+    ld [wNetRxEsc], a
+    ld a, c
+    jr .store2
+.setesc
+    ld a, 1
+    ld [wNetRxEsc], a
+    jr .next
+.store
+    ld a, c
+.store2
+    ; store A at wNetRxBuf[wNetRxLen] if room
+    ld c, a
+    ld a, [wNetRxLen]
+    cp 255
+    jr nc, .next                ; overflow: drop
+    ld e, a
+    ld d, 0
+    ld hl, wNetRxBuf
+    add hl, de
+    ld a, c
+    ld [hl], a
+    ld a, [wNetRxLen]
+    inc a
+    ld [wNetRxLen], a
+    jr .next
+.frame
+    ; END: if we have a frame, process it
+    xor a
+    ld [wNetRxEsc], a
+    ld a, [wNetRxLen]
+    or a
+    jr z, .next                 ; empty frame
+    push bc
+    call process_frame
+    pop bc
+    xor a
+    ld [wNetRxLen], a
+    jr .next
+
+; -----------------------------------------------------------------------------
+; process_frame - IPv4 demux of the frame in wNetRxBuf (wNetRxLen bytes).
+process_frame:
+    ld a, [wNetRxBuf]
+    and $F0
+    cp $40
+    ret nz                      ; not IPv4
+    ld a, [wNetRxBuf]
+    and $0F
+    add a, a
+    add a, a
+    ld [wNetIhl], a             ; ihl in bytes
+    ld a, [wNetRxBuf+9]         ; protocol
+    cp 1
+    jr z, icmp_in
+    ret
+
+; -----------------------------------------------------------------------------
+; icmp_in - handle an inbound ICMP message.
+icmp_in:
+    ld a, [wNetIhl]
+    ld e, a
+    ld d, 0
+    ld hl, wNetRxBuf
+    add hl, de                  ; HL = &icmp
+    ld a, [hl]                  ; type
+    cp 8
+    jr z, icmp_request
+    or a                        ; type 0 = echo reply
+    jr z, icmp_reply
+    ret
+
+; incoming echo request -> answer it in place
+icmp_request:
+    ld a, [wNetIhl]
+    ld e, a
+    ld d, 0
+    ld hl, wNetRxBuf
+    add hl, de                  ; HL = &icmp
+    xor a
+    ld [hl+], a                 ; type = 0
+    inc hl                      ; skip code
+    xor a
+    ld [hl+], a
+    ld [hl], a                  ; cksum = 0
+    call net_swap_ip
+    ; icmp cksum over (rxlen - ihl)
+    ld a, [wNetIhl]
+    ld e, a
+    ld d, 0
+    ld hl, wNetRxBuf
+    add hl, de                  ; HL = &icmp
+    ld a, [wNetRxLen]
+    sub e
+    ld b, a
+    ld de, 0
+    push hl
+    call net_cksum              ; HL = cksum
+    pop de                      ; DE = &icmp
+    ld a, e
+    add 2
+    ld e, a
+    ld a, d
+    adc 0
+    ld d, a                     ; DE = &icmp+2
+    ld a, h
+    ld [de], a
+    inc de
+    ld a, l
+    ld [de], a
+    ; IP cksum
+    xor a
+    ld [wNetRxBuf+10], a
+    ld [wNetRxBuf+11], a
+    ld a, [wNetIhl]
+    ld b, a
+    ld hl, wNetRxBuf
+    ld de, 0
+    call net_cksum
+    ld a, h
+    ld [wNetRxBuf+10], a
+    ld a, l
+    ld [wNetRxBuf+11], a
+    ; send it back
+    ld a, [wNetRxLen]
+    ld b, a
+    ld hl, wNetRxBuf
+    call net_slip_send
+    ret
+
+; incoming echo reply -> deliver to the first ICMP socket
+icmp_reply:
+    ld c, 0
+.scan
+    ld a, c
+    call net_sock_ptr           ; HL = &sock
+    ld a, [hl]
+    cp SOCK_ICMP
+    jr z, .deliver
+    inc c
+    ld a, c
+    cp MAX_SOCKS
+    jr c, .scan
+    ret                         ; no ICMP socket
+.deliver
+    ; HL = &sock. payload = wNetRxBuf + ihl + 8, len = rxlen - ihl - 8
+    push hl
+    ; set HASRX = 1
+    ld a, l
+    add SK_HASRX
+    ld l, a
+    ld a, h
+    adc 0
+    ld h, a
+    ld a, 1
+    ld [hl], a
+    pop hl
+    ; SRCIP = wNetRxBuf[12..15]
+    push hl
+    ld a, l
+    add SK_SRCIP
+    ld l, a
+    ld a, h
+    adc 0
+    ld h, a
+    ld de, wNetRxBuf+12
+    ld b, 4
+.sip
+    ld a, [de]
+    ld [hl+], a
+    inc de
+    dec b
+    jr nz, .sip
+    pop hl
+    ; RXLEN + RXBUF
+    ld a, [wNetIhl]
+    add 8
+    ld e, a
+    ld d, 0                     ; DE = ihl+8 (payload offset)
+    ld a, [wNetRxLen]
+    sub e
+    ld c, a                     ; C = payload length
+    push hl
+    ld a, l
+    add SK_RXLEN
+    ld l, a
+    ld a, h
+    adc 0
+    ld h, a
+    ld a, c
+    ld [hl], a                  ; RXLEN
+    pop hl
+    ; copy payload
+    ld a, l
+    add SK_RXBUF
+    ld l, a
+    ld a, h
+    adc 0
+    ld h, a                     ; HL = &sock.RXBUF (dest)
+    push hl
+    ld hl, wNetRxBuf
+    add hl, de                  ; HL = &payload (src)
+    pop de                      ; DE = dest
+    ld a, c
+    or a
+    ret z
+.cp
+    ld a, [hl+]
+    ld [de], a
+    inc de
+    dec c
+    jr nz, .cp
+    ret
+
+; net_swap_ip - swap the src/dst IP fields of the frame in wNetRxBuf
+net_swap_ip:
+    ld hl, wNetRxBuf+12
+    ld de, wNetRxBuf+16
+    ld b, 4
+.l
+    ld a, [hl]
+    ld c, a
+    ld a, [de]
+    ld [hl+], a
+    ld a, c
+    ld [de], a
+    inc de
+    dec b
+    jr nz, .l
+    ret
diff --git a/src/syscall.asm b/src/syscall.asm
index f926d89..5b40ef0 100644
--- a/src/syscall.asm
+++ b/src/syscall.asm
@@ -67,6 +67,7 @@ SyscallTable:
     dw sys_srecv                ; 30 SRECV
     dw sys_srecv_nb             ; 31 SRECVNB
     dw sys_pollin               ; 32 POLLIN
+    dw sys_net                  ; 33 NET
 
 ; -----------------------------------------------------------------------------
 sys_nosys: