Improve assembler: can now properly assemble lain.com

Makefile

@@ -23,7 +23,8 @@ default: fd$(FLOPPY).img
 
 # COM programs
 %.com: src/%.asm src/*.inc
-	$(NASM) $(NASM_ARGS) -Isrc -l $(@:.com=.lst) -o $@ $<
+	$(AS) $@ $<
+#	$(NASM) $(NASM_ARGS) -Isrc -l $(@:.com=.lst) -o $@ $<
 
 # Bootloader images
 %.bs: boot/%.asm

host/as

@@ -1,4 +1,7 @@
 #!/bin/sh
 outfile="$1"
 infile="$2"
-( awk -f host/asm.awk -v of=sym <"$infile"; cat "$infile") | awk -f host/asm.awk -v of=lst
+listing="${outfile%.*}.lst"
+( awk -f host/asm.awk -v of=sym <"$infile"; cat "$infile") | awk -f host/asm.awk -v of=lst >"$listing"
+grep -A1 '\*\*\*\*' "$listing" && exit 1
+cut -d' ' -f2 "$listing" | xxd -r -p >"$outfile"

host/asm.awk

@@ -83,8 +83,16 @@ BEGIN {
   ops_rel8["jng"]=112+14
   ops_rel8["jg"]=112+15
   ops_rel8["jnle"]=112+15
+  ops_rel8["loopne"]=224
+  ops_rel8["loopnz"]=224
+  ops_rel8["loope"]=225
+  ops_rel8["loopz"]=225
+  ops_rel8["loop"]=226
+  ops_rel8["jmp"]=235
   # single-byte opcodes without operands
   ops_sb["nop"]=128+16
+  ops_sb["movsb"]=164
+  ops_sb["movsw"]=165
   ops_sb["stosb"]=170
   ops_sb["stosw"]=171
   ops_sb["lodsb"]=172
@@ -92,6 +100,17 @@ BEGIN {
   ops_sb["ret"]=195
   ops_sb["retf"]=203
   ops_sb["hlt"]=244
+  # prefix instructions
+  prefix["es"]=38
+  prefix["cs"]=46
+  prefix["ss"]=54
+  prefix["ds"]=64
+  prefix["lock"]=240
+  prefix["rep"]=243
+  prefix["repe"]=242
+  prefix["repz"]=242
+  prefix["repne"]=243
+  prefix["repnz"]=243
 }
 # error string to insert into listing
 function err(str) {
@@ -118,6 +137,10 @@ function getsym(name) {
   if (substr(name,1,1)=="0" || int(name)) {
     return int(name)
   }
+  if (name in r8 || name in r16 || name in sreg) {
+    err("Register name " name "is not a valid immediate")
+    return 0
+  }
   if (substr(name,1,1)==".") name = plabel name
   if (name in prevsym) {
     return prevsym[name]
@@ -126,25 +149,22 @@ function getsym(name) {
   }
   return 0
 }
-# encode a signed byte
-function pushb(val) {
+function push_byte(val) {
   #print("; pushb " val)
   if (val<0 || val>=256) err("Value " val " does not fit in byte")
   hex=hex sprintf("%02X",val)
 }
-# encode a signed byte (for rel8)
-function pushsb(val) {
+function push_signed_byte(val) {
   while (val < 0) val = val + 256
-  pushb(val)
+  push_byte(val)
 }
-# encode a word
-function pushw(val) {
+function push_word(val) {
   while (val<0) val=val+65536
   t=sprintf("%04X",val)
   hex=hex substr(t,3) substr(t,1,2)
 }
-# encode a modrm byte
-function modrm(str, spare, rs) {
+# rs is the register set (r8, r16) that can show up in str
+function push_modrm(str, spare, rs) {
   mod=3
   rm=0
   if (str in rs) {
@@ -173,28 +193,43 @@ function modrm(str, spare, rs) {
     }
   }
   #print("; modR/M:", mod, spare, rm)
-  pushb(mod*64+spare*8+rm)
+  push_byte(mod*64+spare*8+rm)
   if (mod == 1) {
-    pushb(disp)
+    push_byte(disp)
   } else if (mod == 2 || (mod == 0 && rm == 6)) {
-    pushw(disp)
+    push_word(disp)
   }
 }
-# common encoding: two operands, one is modrm, other in spare field
+# common encoding: two operands, one is modrm, other is register via spare field
 # last two bits of opcode specify width (byte/word) and whether modrm operand is first or second
-function enc(opcode) {
-
+function push_op_modrm(opcode) {
+  if (rm1 && byteop && op2 in r8) {
+    push_byte(opcode)
+    push_modrm(op1, r8[op2], r8)
+  } else if (rm1 && wordop && op2 in r16) {
+    push_byte(opcode+1)
+    push_modrm(op1, r16[op2], r16)
+  } else if (rm2 && byteop && op1 in r8) {
+    push_byte(opcode+2)
+    push_modrm(op2, r8[op1], r8)
+  } else if (rm2 && wordop && op1 in r16) {
+    push_byte(opcode+2+1)
+    push_modrm(op2, r16[op1], r16)
+  }
 }
 # common encoding: one operand encoded as modrm with fixed spare field
 # operand can be byte or word, encoded in last bit of opcode
-function encfs(opcode, spare) {
-  if (wordop) {
-    pushb(opcode)
-    modrm(op1, spare, r16)
-  } else if (byteop) {
-    pushb(opcode)
-    modrm(op1, spare, r8)
+function push_op_fixed_spare(opcode, spare) {
+  if (byteop) {
+    push_byte(opcode)
+    push_modrm(op1, spare, r8)
+    return 1
+  } else if (wordop) {
+    push_byte(opcode+1)
+    push_modrm(op1, spare, r16)
+    return 2
   }
+  return 0
 }
 
 # dont process empty lines or comment lines
@@ -233,6 +268,12 @@ $2=="sym" {
   }
   # take note if we got a instruction size specifier
   op=$(opn)
+  if (op in prefix) {
+    push_byte(prefix[op])
+    opn=opn+1
+    op=$(opn)
+  }
+
   if($(opn+1)=="byte" || $(opn+1)=="short") {
     wordop=0
     dwordop=0
@@ -303,115 +344,113 @@ op=="db" {
   for(i=1;i<=c;i++) {
     if (substr(a[i],1,1)=="\"") {
       for(j=2;j<length(a[i]);j++) {
-        pushb(_ord_[substr(a[i],j,1)])
+        push_byte(_ord_[substr(a[i],j,1)])
       }
-    } else pushb(getsym(a[i]))
+    } else push_byte(getsym(a[i]))
   }
 }
 op=="dw" {
-  for(i=1;i<=c;i++) pushw(getsym(a[i]))
+  for(i=1;i<=c;i++) push_word(getsym(a[i]))
 }
 # arithmetics: ADD, SUB, XOR etc
 op in alu {
-  # modrm <- reg
-  if (rm1 && op2 in r8) {
-    pushb(0+alu[op]*8)
-    modrm(op1, r8[op2], r8)
-  } else if (rm1 && op2 in r16) {
-    pushb(1+alu[op]*8)
-    modrm(op1, r16[op2], r16)
-  # reg <- modrm
-  } else if (op1 in r8 && rm2) {
-    pushb(2+alu[op]*8)
-    modrm(op2, r8[op1], r8)
-  } else if (op1 in r16 && rm2) {
-    pushb(3+alu[op]*8)
-    modrm(op2, r16[op1], r16)
-  # modrm <- imm
-  } else if (byteop && rm1 || op1 in r8) {
-    pushb(128) # 80
-    modrm(op1, alu[op], r8)
-    pushb(getsym(op2))
-  } else if (wordop && rm1 || op1 in r16) {
-    pushb(129) # 81
-    modrm(op1, alu[op], r16)
-    pushw(getsym(op2))
+  push_op_modrm(alu[op]*8)
+  if (!hex) {
+    size=push_op_fixed_spare(0x80, alu[op])
+    if (size==1) {
+      push_byte(getsym(op2))
+    } else if (size==2) {
+      push_word(getsym(op2))
+    }
   }
 }
+# no idea why this made this extra, this is a AND without storing the result
+op=="test" {
+  push_op_modrm(132) # 84
+  if (!hex) {
+    size=push_op_fixed_spare(246, 7)
+    if (size==1) {
+      push_byte(getsym(op2))
+    } else if (size==2) {
+      push_word(getsym(op2))
+    }
+  }
+}
+op=="sar" && op2=="1" {
+  push_op_fixed_spare(208, 0) # D0 /0
+}
+op=="inc" && rm1 {
+  push_op_fixed_spare(254, 0) # FE /0
+}
 # MOV variants
 op=="mov" {
   # modrm <-> reg
   if (rm1 && op2 in r8) {
-    pushb(136) # 88
-    modrm(op1, r8[op2], r8)
+    push_byte(136) # 88
+    push_modrm(op1, r8[op2], r8)
   } else if (rm1 && op2 in r16) {
-    pushb(137) # 89
-    modrm(op1, r16[op2], r16)
+    push_byte(137) # 89
+    push_modrm(op1, r16[op2], r16)
   } else if (op1 in r8 && rm2) {
-    pushb(138)
-    modrm(op2, r8[op1], r8)
+    push_byte(138)
+    push_modrm(op2, r8[op1], r8)
   } else if (op1 in r16 && rm2) {
-    pushb(139)
-    modrm(op2, r16[op1], r16)
+    push_byte(139)
+    push_modrm(op2, r16[op1], r16)
   # modrm <-> sreg
   } else if (rm1 && op2 in sreg) {
-    pushb(140)
-    modrm(op1, sreg[op2], r16)
+    push_byte(140)
+    push_modrm(op1, sreg[op2], r16)
   } else if (rm2 && op1 in sreg) {
-    pushb(142)
-    modrm(op2, sreg[op1], r16)
+    push_byte(142)
+    push_modrm(op2, sreg[op1], r16)
   # reg <- imm
   } else if (op1 in r8) {
-    pushb(176+r8[op1])
-    pushb(getsym(op2))
+    push_byte(176+r8[op1])
+    push_byte(getsym(op2))
   } else if (op1 in r16) {
-    pushb(184+r16[op1])
-    pushw(getsym(op2))
+    push_byte(184+r16[op1])
+    push_word(getsym(op2))
   # modrm <- imm
   } else if (byteop && rm1) {
-    pushb(198)
-    modrm(rm1, 0, r16)
+    push_byte(198)
+    push_modrm(rm1, 0, r16)
   } else if (wordop && rm1) {
-    pushb(199)
-    modrm(rm1, 0, r16)
+    push_byte(199)
+    push_modrm(rm1, 0, r16)
   }
 }
 op=="push" && op1 in r16 {
-  pushb(80+r16[op1])
+  push_byte(80+r16[op1])
 }
 op=="pop" && op1 in r16 {
-  pushb(88+r16[op1])
+  push_byte(88+r16[op1])
 }
-op=="test" && op2 in r8 { # 84
-  pushb(132)
-  modrm(op1, r8[op2], r8)
+op=="int" && op1=="3" { # CC breakpoint
+  push_byte(204)
 }
 op=="int" { # CD
-  pushb(205)
-  pushb(getsym(op1))
-}
-op=="jmp" && byteop {
-  pushb(235)
-  pushsb(getsym(op1)-(pos+2))
+  push_byte(205)
+  push_byte(getsym(op1))
 }
 op=="jmp" && wordop {
-  pushb(233)
-  pushw(getsym(op1)-(pos+3))
+  push_byte(233)
+  push_word(getsym(op1)-(pos+3))
 }
-op=="call" {
-  pushb(232)
-  pushw(getsym(op1)-(pos+3))
+op=="call" && wordop {
+  push_byte(232)
+  push_word(getsym(op1)-(pos+3))
 }
 op=="neg" {
-  encfs(246, 3)
+  push_op_fixed_spare(246, 3)
 }
 # opcodes with rel8 encoding
-op in ops_rel8 && byteop {
-  pushb(ops_rel8[op])
-  pushb(getsym(op1)-(pos+2))
+op in ops_rel8 && byteop && c==1 {
+  push_byte(ops_rel8[op])
+  push_signed_byte(getsym(op1)-(pos+2))
 }
 # opcodes without arguments
-op in ops_sb { pushb(ops_sb[op]) }
+op in ops_sb { push_byte(ops_sb[op]) }
 {
   if (!hex) err("no encoding found")
   submit(pos)

host/bin2db.sh

@@ -0,0 +1,2 @@
+#!/bin/sh
+xxd -c8 -g1 "$1"|cut -d' ' -f2-9|sed 's/[0-9A-Fa-f]\{2\}/0x&/g;s/ /, /g;s/^/	db /'