fix lightud for 48bit virtual address

Set 2 type values (~3 and ~4) to lightud type. It is enabled only for
ARM64, whose kernel uses 48bit virtual address.

Change-Id: I331eb1997af35c3ed63d6b15d6d59cfc5bc355e0

src/Makefile

@@ -581,6 +581,14 @@ endif
 endif
 endif
 
+ifeq (arm64,$(TARGET_LJARCH))
+  GENLIBBC_FLAGS= --arm64
+  LIBBC_H= host/buildvm_libbc_arm64.h
+else
+  GENLIBBC_FLAGS=
+  LIBBC_H= host/buildvm_libbc.h
+endif
+
 Q= @
 E= @echo
 #Q=
@@ -600,7 +608,7 @@ clean:
 	$(HOST_RM) $(ALL_RM)
 
 libbc:
-	./$(LUAJIT_T) host/genlibbc.lua -o host/buildvm_libbc.h $(LJLIB_C)
+	./$(LUAJIT_T) host/genlibbc.lua $(GENLIBBC_FLAGS) -o $(LIBBC_H) $(LJLIB_C)
 	$(MAKE) all
 
 depend:

src/host/buildvm_lib.c

@@ -7,7 +7,11 @@
 #include "lj_obj.h"
 #include "lj_bc.h"
 #include "lj_lib.h"
+#if LJ_TARGET_ARM64
+#include "buildvm_libbc_arm64.h"
+#else
 #include "buildvm_libbc.h"
+#endif
 
 /* Context for library definitions. */
 static uint8_t obuf[8192];
@@ -385,6 +389,8 @@ void emit_lib(BuildCtx *ctx)
 	  ok = LJ_HASJIT;
 	else if (!strcmp(buf, "#if LJ_HASFFI\n"))
 	  ok = LJ_HASFFI;
+        else if (!strcmp(buf, "#if LJ_TARGET_ARM64\n"))
+          ok = (LUAJIT_ARCH_ARM64 == LUAJIT_TARGET);
 	if (!ok) {
 	  int lvl = 1;
 	  while (fgets(buf, sizeof(buf), fp) != NULL) {

src/host/buildvm_libbc_arm64.h

@@ -0,0 +1,30 @@
+/* This is a generated file. DO NOT EDIT! */
+
+static const int libbc_endian = 0;
+
+static const uint8_t libbc_code[] = {
+0,1,2,0,0,1,2,24,1,0,0,76,1,2,0,241,135,158,166,3,220,203,178,130,4,0,1,2,0,
+0,1,2,24,1,0,0,76,1,2,0,243,244,148,165,20,198,190,199,252,3,0,1,2,0,0,0,3,
+16,0,6,0,21,1,0,0,76,1,2,0,0,2,10,0,0,0,15,16,0,13,0,16,1,10,0,41,2,1,0,21,
+3,0,0,41,4,1,0,77,2,8,128,18,6,1,0,18,8,5,0,59,9,5,0,66,6,3,2,10,6,0,0,88,7,
+1,128,76,6,2,0,79,2,248,127,75,0,1,0,0,2,11,0,0,0,16,16,0,13,0,16,1,10,0,43,
+2,0,0,18,3,0,0,41,4,0,0,88,5,7,128,18,7,1,0,18,9,5,0,18,10,6,0,66,7,3,2,10,
+7,0,0,88,8,1,128,76,7,2,0,70,5,3,3,82,5,247,127,75,0,1,0,0,1,2,0,0,0,3,16,0,
+13,0,21,1,0,0,76,1,2,0,0,2,10,0,0,2,30,16,0,13,0,21,2,0,0,11,1,0,0,88,3,7,128,
+8,2,0,0,88,3,23,128,59,3,2,0,43,4,0,0,64,4,2,0,76,3,2,0,88,3,18,128,16,1,15,
+0,41,3,1,0,3,3,1,0,88,3,14,128,3,1,2,0,88,3,12,128,59,3,1,0,22,4,1,1,18,5,2,
+0,41,6,1,0,77,4,4,128,23,8,1,7,59,9,7,0,64,9,8,0,79,4,252,127,43,4,0,0,64,4,
+2,0,76,3,2,0,75,0,1,0,0,2,0
+};
+
+static const struct { const char *name; int ofs; } libbc_map[] = {
+{"math_deg",0},
+{"math_rad",25},
+{"string_len",50},
+{"table_foreachi",69},
+{"table_foreach",136},
+{"table_getn",207},
+{"table_remove",226},
+{NULL,355}
+};
+

src/host/genlibbc.lua

@@ -15,9 +15,11 @@ local format = string.format
 
 local isbe = (string.byte(string.dump(function() end), 5) % 2 == 1)
 
+local arm64_target = false
+
 local function usage(arg)
   io.stderr:write("Usage: ", arg and arg[0] or "genlibbc",
-		  " [-o buildvm_libbc.h] lib_*.c\n")
+		  " [-o buildvm_libbc.h] [--arm64] lib_*.c\n")
   os.exit(1)
 end
 
@@ -26,11 +28,16 @@ local function parse_arg(arg)
   if not (arg and arg[1]) then
     usage(arg)
   end
+  while arg[1] == "-o" or arg[1] == "--arm64" do
     if arg[1] == "-o" then
       outfile = arg[2]
       if not outfile then usage(arg) end
       table.remove(arg, 1)
       table.remove(arg, 1)
+    elseif arg[1] == "--arm64" then
+      arm64_target = true
+      table.remove(arg, 1)
+    end
   end
   return outfile
 end
@@ -74,11 +81,6 @@ local function read_uleb128(p)
   return p, v
 end
 
--- ORDER LJ_T
-local name2itype = {
-  str = 5, func = 9, tab = 12, int = 14, num = 15
-}
-
 local BC = {}
 for i=0,#bcnames/6-1 do
   BC[string.gsub(string.sub(bcnames, i*6+1, i*6+6), " ", "")] = i
@@ -97,6 +99,18 @@ local function fixup_dump(dump, fixup)
   p = read_uleb128(p)
   p, sizebc = read_uleb128(p)
   local rawtab = {}
+  -- ORDER LJ_T
+  local name2itype
+  if arm64_target == true then
+    -- Because CHECK_num is not found anywhere, num type is omitted.
+    name2itype = {
+      str = 6, func = 10, tab = 13, int = 15
+    }
+  else
+    name2itype = {
+      str = 5, func = 9, tab = 12, int = 14, num = 15
+    }
+  end
   for i=0,sizebc-1 do
     local op = p[xop]
     if op == BC.KSHORT then

src/lib_base.c

@@ -55,6 +55,9 @@ LJLIB_ASM(assert)		LJLIB_REC(.)
 LJLIB_PUSH("nil")
 LJLIB_PUSH("boolean")
 LJLIB_PUSH(top-1)  /* boolean */
+#if LJ_TARGET_ARM64
+LJLIB_PUSH("userdata")
+#endif
 LJLIB_PUSH("userdata")
 LJLIB_PUSH("string")
 LJLIB_PUSH("upval")

src/lj_api.c

@@ -189,7 +189,7 @@ LUA_API int lua_type(lua_State *L, int idx)
   cTValue *o = index2adr(L, idx);
   if (tvisnumber(o)) {
     return LUA_TNUMBER;
-#if LJ_64 && !LJ_GC64
+#if LJ_64 && !LJ_GC64 || LJ_TARGET_ARM64
   } else if (tvislightud(o)) {
     return LUA_TLIGHTUSERDATA;
 #endif
@@ -198,7 +198,11 @@ LUA_API int lua_type(lua_State *L, int idx)
   } else {  /* Magic internal/external tag conversion. ORDER LJ_T */
     uint32_t t = ~itype(o);
 #if LJ_64
+#if LJ_TARGET_ARM64
+    int tt = (int)((U64x(75a069,80400110) >> 4*t) & 15u);
+#else
     int tt = (int)((U64x(75a06,98042110) >> 4*t) & 15u);
+#endif
 #else
     int tt = (int)(((t < 8 ? 0x98042110u : 0x75a06u) >> 4*(t&7)) & 15u);
 #endif

src/lj_ffrecord.c

@@ -196,8 +196,10 @@ static void LJ_FASTCALL recff_type(jit_State *J, RecordFFData *rd)
   uint32_t t;
   if (tvisnumber(&rd->argv[0]))
     t = ~LJ_TNUMX;
+#if !LJ_TARGET_ARM64
   else if (LJ_64 && !LJ_GC64 && tvislightud(&rd->argv[0]))
     t = ~LJ_TLIGHTUD;
+#endif
   else
     t = ~itype(&rd->argv[0]);
   J->base[0] = lj_ir_kstr(J, strV(&J->fn->c.upvalue[t]));

src/lj_ir.h

@@ -303,6 +303,17 @@ LJ_DATA const uint8_t lj_ir_mode[IR__MAX+1];
 ** a TValue after implicit or explicit conversion. Their types must be
 ** contiguous and next to IRT_NUM (see the typerange macros below).
 */
+#if LJ_TARGET_ARM64
+#define IRTDEF(_) \
+  _(NIL, 4) _(FALSE, 4) _(TRUE, 4) _(LIGHTUD1, LJ_64 ? 8 : 4) \
+  _(LIGHTUD2, LJ_64 ? 8 : 4) \
+  _(STR, IRTSIZE_PGC) _(P32, 4) _(THREAD, IRTSIZE_PGC) _(PROTO, IRTSIZE_PGC) \
+  _(FUNC, IRTSIZE_PGC) _(P64, 8) _(CDATA, IRTSIZE_PGC) _(TAB, IRTSIZE_PGC) \
+  _(UDATA, IRTSIZE_PGC) \
+  _(FLOAT, 4) _(NUM, 8) _(I8, 1) _(U8, 1) _(I16, 2) _(U16, 2) \
+  _(INT, 4) _(U32, 4) _(I64, 8) _(U64, 8) \
+  _(SOFTFP, 4)  /* There is room for 5 more types. */
+#else
 #define IRTDEF(_) \
   _(NIL, 4) _(FALSE, 4) _(TRUE, 4) _(LIGHTUD, LJ_64 ? 8 : 4) \
   _(STR, IRTSIZE_PGC) _(P32, 4) _(THREAD, IRTSIZE_PGC) _(PROTO, IRTSIZE_PGC) \
@@ -311,6 +322,7 @@ LJ_DATA const uint8_t lj_ir_mode[IR__MAX+1];
   _(FLOAT, 4) _(NUM, 8) _(I8, 1) _(U8, 1) _(I16, 2) _(U16, 2) \
   _(INT, 4) _(U32, 4) _(I64, 8) _(U64, 8) \
   _(SOFTFP, 4)  /* There is room for 8 more types. */
+#endif
 
 /* IR result type and flags (8 bit). */
 typedef enum {
@@ -373,14 +385,25 @@ typedef struct IRType1 { uint8_t irt; } IRType1;
 #define irt_isfp(t)		(irt_isnum(t) || irt_isfloat(t))
 #define irt_isinteger(t)	(irt_typerange((t), IRT_I8, IRT_INT))
 #define irt_isgcv(t)		(irt_typerange((t), IRT_STR, IRT_UDATA))
+#if LJ_TARGET_ARM64
+#define irt_isaddr(t)		(irt_typerange((t), IRT_LIGHTUD1, IRT_UDATA))
+#else
 #define irt_isaddr(t)		(irt_typerange((t), IRT_LIGHTUD, IRT_UDATA))
+#endif
 #define irt_isint64(t)		(irt_typerange((t), IRT_I64, IRT_U64))
 
 #if LJ_GC64
+#if LJ_TARGET_ARM64
 #define IRT_IS64 \
   ((1u<<IRT_NUM)|(1u<<IRT_I64)|(1u<<IRT_U64)|(1u<<IRT_P64)|\
-   (1u<<IRT_LIGHTUD)|(1u<<IRT_STR)|(1u<<IRT_THREAD)|(1u<<IRT_PROTO)|\
+   (1u<<IRT_LIGHTUD1)|(1u<<IRT_LIGHTUD2)|(1u<<IRT_STR)|(1u<<IRT_THREAD)|\
-   (1u<<IRT_FUNC)|(1u<<IRT_CDATA)|(1u<<IRT_TAB)|(1u<<IRT_UDATA))
+   (1u<<IRT_PROTO)|(1u<<IRT_FUNC)|(1u<<IRT_CDATA)|(1u<<IRT_TAB)|(1u<<IRT_UDATA))
+#else
+#define IRT_IS64 \
+  ((1u<<IRT_NUM)|(1u<<IRT_I64)|(1u<<IRT_U64)|(1u<<IRT_P64)|\
+   (1u<<IRT_LIGHTUD)|(1u<<IRT_STR)|(1u<<IRT_THREAD)|\
+   (1u<<IRT_PROTO)|(1u<<IRT_FUNC)|(1u<<IRT_CDATA)|(1u<<IRT_TAB)|(1u<<IRT_UDATA))
+#endif
 #elif LJ_64
 #define IRT_IS64 \
   ((1u<<IRT_NUM)|(1u<<IRT_I64)|(1u<<IRT_U64)|(1u<<IRT_P64)|(1u<<IRT_LIGHTUD))
@@ -412,7 +435,9 @@ static LJ_AINLINE IRType itype2irt(const TValue *tv)
 
 static LJ_AINLINE uint32_t irt_toitype_(IRType t)
 {
+#if !LJ_TARGET_ARM64
   lua_assert(!LJ_64 || LJ_GC64 || t != IRT_LIGHTUD);
+#endif
   if (LJ_DUALNUM && t > IRT_NUM) {
     return LJ_TISNUM;
   } else {
@@ -492,7 +517,12 @@ typedef uint32_t TRef;
 #define tref_isnil(tr)		(tref_istype((tr), IRT_NIL))
 #define tref_isfalse(tr)	(tref_istype((tr), IRT_FALSE))
 #define tref_istrue(tr)		(tref_istype((tr), IRT_TRUE))
+#if LJ_TARGET_ARM64
+#define tref_islightud(tr)	(tref_istype((tr), IRT_LIGHTUD1) ||\
+                                 tref_istype((tr), IRT_LIGHTUD2))
+#else
 #define tref_islightud(tr)	(tref_istype((tr), IRT_LIGHTUD))
+#endif
 #define tref_isstr(tr)		(tref_istype((tr), IRT_STR))
 #define tref_isfunc(tr)		(tref_istype((tr), IRT_FUNC))
 #define tref_iscdata(tr)	(tref_istype((tr), IRT_CDATA))

src/lj_obj.c

@@ -13,12 +13,17 @@ LJ_DATADEF const char *const lj_obj_typename[] = {  /* ORDER LUA_T */
   "no value", "nil", "boolean", "userdata", "number", "string",
   "table", "function", "userdata", "thread", "proto", "cdata"
 };
-
+#if LJ_TARGET_ARM64
+LJ_DATADEF const char *const lj_obj_itypename[] = {  /* ORDER LJ_T */
+  "nil", "boolean", "boolean", "userdata", "userdata", "string", "upval", "thread",
+  "proto", "function", "trace", "cdata", "table", "userdata", "number"
+};
+#else
 LJ_DATADEF const char *const lj_obj_itypename[] = {  /* ORDER LJ_T */
   "nil", "boolean", "boolean", "userdata", "string", "upval", "thread",
   "proto", "function", "trace", "cdata", "table", "userdata", "number"
 };
-
+#endif
 /* Compare two objects without calling metamethods. */
 int LJ_FASTCALL lj_obj_equal(cTValue *o1, cTValue *o2)
 {

src/lj_obj.h

@@ -254,6 +254,24 @@ typedef const TValue cTValue;
 ** GC objects are at the end, table/userdata must be lowest.
 ** Also check lj_ir.h for similar ordering constraints.
 */
+#if LJ_TARGET_ARM64
+#define LJ_TNIL			(~0u)
+#define LJ_TFALSE		(~1u)
+#define LJ_TTRUE		(~2u)
+#define LJ_TLIGHTUD1		(~3u)
+#define LJ_TLIGHTUD2		(~4u)
+#define LJ_TSTR			(~5u)
+#define LJ_TUPVAL		(~6u)
+#define LJ_TTHREAD		(~7u)
+#define LJ_TPROTO		(~8u)
+#define LJ_TFUNC		(~9u)
+#define LJ_TTRACE		(~10u)
+#define LJ_TCDATA		(~11u)
+#define LJ_TTAB			(~12u)
+#define LJ_TUDATA		(~13u)
+/* This is just the canonical number type used in some places. */
+#define LJ_TNUMX		(~14u)
+#else
 #define LJ_TNIL			(~0u)
 #define LJ_TFALSE		(~1u)
 #define LJ_TTRUE		(~2u)
@@ -269,7 +287,7 @@ typedef const TValue cTValue;
 #define LJ_TUDATA		(~12u)
 /* This is just the canonical number type used in some places. */
 #define LJ_TNUMX		(~13u)
-
+#endif
 /* Integers have itype == LJ_TISNUM doubles have itype < LJ_TISNUM */
 #if LJ_64 && !LJ_GC64
 #define LJ_TISNUM		0xfffeffffu
@@ -285,6 +303,10 @@ typedef const TValue cTValue;
 #define LJ_GCVMASK		(((uint64_t)1 << 47) - 1)
 #endif
 
+#if LJ_TARGET_ARM64
+#define LJ_SET47BIT              ((uint64_t)1 << 47)
+#endif
+
 /* -- String object ------------------------------------------------------- */
 
 /* String object header. String payload follows. */
@@ -749,8 +771,12 @@ typedef union GCobj {
 #if LJ_64 && !LJ_GC64
 #define tvislightud(o)	(((int32_t)itype(o) >> 15) == -2)
 #else
+#if LJ_TARGET_ARM64
+#define tvislightud(o)	(itype(o) == LJ_TLIGHTUD1 || itype(o) == LJ_TLIGHTUD2)
+#else
 #define tvislightud(o)	(itype(o) == LJ_TLIGHTUD)
 #endif
+#endif
 #define tvisstr(o)	(itype(o) == LJ_TSTR)
 #define tvisfunc(o)	(itype(o) == LJ_TFUNC)
 #define tvisthread(o)	(itype(o) == LJ_TTHREAD)
@@ -795,8 +821,13 @@ typedef union GCobj {
 #endif
 #define boolV(o)	check_exp(tvisbool(o), (LJ_TFALSE - itype(o)))
 #if LJ_64
+#if LJ_TARGET_ARM64
+#define lightudV(o) \
+  check_exp(tvislightud(o), (void *)((o)->u64 & U64x(0000ffff,ffffffff)))
+#else
 #define lightudV(o) \
   check_exp(tvislightud(o), (void *)((o)->u64 & U64x(00007fff,ffffffff)))
+#endif
 #else
 #define lightudV(o)	check_exp(tvislightud(o), gcrefp((o)->gcr, void))
 #endif
@@ -827,7 +858,14 @@ typedef union GCobj {
 static LJ_AINLINE void setlightudV(TValue *o, void *p)
 {
 #if LJ_GC64
+#if LJ_TARGET_ARM64
+  if ((uint64_t)p & ((uint64_t)0x1<<47))
+    o->u64 = (uint64_t)p | (((uint64_t)LJ_TLIGHTUD2) << 47);
+  else
+    o->u64 = (uint64_t)p | (((uint64_t)LJ_TLIGHTUD1) << 47);
+#else
   o->u64 = (uint64_t)p | (((uint64_t)LJ_TLIGHTUD) << 47);
+#endif
 #elif LJ_64
   o->u64 = (uint64_t)p | (((uint64_t)0xffff) << 48);
 #else
@@ -836,8 +874,13 @@ static LJ_AINLINE void setlightudV(TValue *o, void *p)
 }
 
 #if LJ_64
+#if LJ_TARGET_ARM64
+#define checklightudptr(L, p) \
+  (((uint64_t)(p) >> 48) ? (lj_err_msg(L, LJ_ERR_BADLU), NULL) : (p))
+#else
 #define checklightudptr(L, p) \
   (((uint64_t)(p) >> 47) ? (lj_err_msg(L, LJ_ERR_BADLU), NULL) : (p))
+#endif
 #else
 #define checklightudptr(L, p)	(p)
 #endif

src/vm_arm64.dasc

@@ -381,8 +381,9 @@ static void build_subroutines(BuildCtx *ctx)
   |  and sp, CARG1, #CFRAME_RAWMASK
   |->vm_unwind_ff_eh:			// Landing pad for external unwinder.
   |  ldr L, SAVE_L
-  |    movz TISNUM, #(LJ_TISNUM>>1)&0xffff, lsl #48
+  |    movz TISNUM, #(LJ_TISNUM)&0xffff, lsl #48
-  |    movz TISNUMhi, #(LJ_TISNUM>>1)&0xffff, lsl #16
+  |    asr  TISNUM, TISNUM, #1
+  |    lsr  TISNUMhi, TISNUM, #32
   |    movn TISNIL, #0
   |    mov RC, #16			// 2 results: false + error message.
   |  ldr BASE, L->base
@@ -447,8 +448,9 @@ static void build_subroutines(BuildCtx *ctx)
   |  str L, GL->cur_L
   |  mov RA, BASE
   |   ldp BASE, CARG1, L->base
-  |    movz TISNUM, #(LJ_TISNUM>>1)&0xffff, lsl #48
+  |    movz TISNUM, #(LJ_TISNUM)&0xffff, lsl #48
-  |    movz TISNUMhi, #(LJ_TISNUM>>1)&0xffff, lsl #16
+  |    asr  TISNUM, TISNUM, #1
+  |    lsr  TISNUMhi, TISNUM, #32
   |  ldr PC, [BASE, FRAME_PC]
   |     strb wzr, L->status
   |    movn TISNIL, #0
@@ -486,8 +488,9 @@ static void build_subroutines(BuildCtx *ctx)
   |3:  // Entry point for vm_cpcall/vm_resume (BASE = base, PC = ftype).
   |  str L, GL->cur_L
   |  ldp RB, CARG1, L->base		// RB = old base (for vmeta_call).
-  |    movz TISNUM, #(LJ_TISNUM>>1)&0xffff, lsl #48
+  |    movz TISNUM, #(LJ_TISNUM)&0xffff, lsl #48
-  |    movz TISNUMhi, #(LJ_TISNUM>>1)&0xffff, lsl #16
+  |    asr  TISNUM, TISNUM, #1
+  |    lsr  TISNUMhi, TISNUM, #32
   |  add PC, PC, BASE
   |    movn TISNIL, #0
   |  sub PC, PC, RB			// PC = frame delta + frame type
@@ -589,6 +592,7 @@ static void build_subroutines(BuildCtx *ctx)
   |
   |->vmeta_tgets:
   |  movk CARG2, #(LJ_TTAB>>1)&0xffff, lsl #48
+  |  orr CARG2, CARG2, #LJ_SET47BIT
   |  str CARG2, GL->tmptv
   |  add CARG2, GL, #offsetof(global_State, tmptv)
   |2:
@@ -647,6 +651,7 @@ static void build_subroutines(BuildCtx *ctx)
   |
   |->vmeta_tsets:
   |  movk CARG2, #(LJ_TTAB>>1)&0xffff, lsl #48
+  |  orr CARG2, CARG2, #LJ_SET47BIT
   |  str CARG2, GL->tmptv
   |  add CARG2, GL, #offsetof(global_State, tmptv)
   |2:
@@ -988,6 +993,7 @@ static void build_subroutines(BuildCtx *ctx)
   |4:
   |  mov CARG1, RB			// Use metatable as default result.
   |  movk CARG1, #(LJ_TTAB>>1)&0xffff, lsl #48
+  |  orr CARG1, CARG1, #LJ_SET47BIT
   |  b ->fff_restv
   |5:
   |  cmp TMP0, TISNIL
@@ -1882,8 +1888,9 @@ static void build_subroutines(BuildCtx *ctx)
   |  bl extern lj_ccallback_enter	// (CTState *cts, void *cf)
   |  // Returns lua_State *.
   |  ldp BASE, RC, L:CRET1->base
-  |   movz TISNUM, #(LJ_TISNUM>>1)&0xffff, lsl #48
+  |   movz TISNUM, #(LJ_TISNUM)&0xffff, lsl #48
-  |   movz TISNUMhi, #(LJ_TISNUM>>1)&0xffff, lsl #16
+  |   asr  TISNUM, TISNUM, #1
+  |   lsr  TISNUMhi, TISNUM, #32
   |   movn TISNIL, #0
   |   mov L, CRET1
   |  ldr LFUNC:CARG3, [BASE, FRAME_FUNC]
@@ -2660,6 +2667,7 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop)
     }
     |  ldr BASE, L->base
     |   movk CRET1, #(LJ_TTAB>>1)&0xffff, lsl #48
+    |   orr CRET1, CRET1, #LJ_SET47BIT
     |  str CRET1, [BASE, RA, lsl #3]
     |  ins_next
     |