diff options
author | Richard Henderson <richard.henderson@linaro.org> | 2021-01-30 14:24:25 -0800 |
---|---|---|
committer | Richard Henderson <richard.henderson@linaro.org> | 2021-06-19 08:51:11 -0700 |
commit | 7b7d8b2d9a7fd68de821f96267e224c1a6256af1 (patch) | |
tree | 0de572bc83a96bf0793c3b27605c44c7c8704f0d | |
parent | bcb81061dc8173076d9669f969d132b998cd2af9 (diff) | |
download | qemu-7b7d8b2d9a7fd68de821f96267e224c1a6256af1.zip qemu-7b7d8b2d9a7fd68de821f96267e224c1a6256af1.tar.gz qemu-7b7d8b2d9a7fd68de821f96267e224c1a6256af1.tar.bz2 |
tcg/tci: Use ffi for calls
This requires adjusting where arguments are stored.
Place them on the stack at left-aligned positions.
Adjust the stack frame to be at entirely positive offsets.
Tested-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
-rw-r--r-- | include/tcg/tcg.h | 1 | ||||
-rw-r--r-- | tcg/tcg.c | 64 | ||||
-rw-r--r-- | tcg/tci.c | 138 | ||||
-rw-r--r-- | tcg/tci/tcg-target.c.inc | 50 | ||||
-rw-r--r-- | tcg/tci/tcg-target.h | 2 |
5 files changed, 151 insertions, 104 deletions
diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h index 064dab3..236315b 100644 --- a/include/tcg/tcg.h +++ b/include/tcg/tcg.h @@ -52,6 +52,7 @@ #define MAX_OPC_PARAM (4 + (MAX_OPC_PARAM_PER_ARG * MAX_OPC_PARAM_ARGS)) #define CPU_TEMP_BUF_NLONGS 128 +#define TCG_STATIC_FRAME_SIZE (CPU_TEMP_BUF_NLONGS * sizeof(long)) /* Default target word size to pointer size. */ #ifndef TCG_TARGET_REG_BITS @@ -147,7 +147,12 @@ static void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg, TCGReg arg1, intptr_t arg2); static bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val, TCGReg base, intptr_t ofs); +#ifdef CONFIG_TCG_INTERPRETER +static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target, + ffi_cif *cif); +#else static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target); +#endif static bool tcg_target_const_match(int64_t val, TCGType type, int ct); #ifdef TCG_TARGET_NEED_LDST_LABELS static int tcg_out_ldst_finalize(TCGContext *s); @@ -1554,25 +1559,37 @@ void tcg_gen_callN(void *func, TCGTemp *ret, int nargs, TCGTemp **args) for (i = 0; i < nargs; i++) { int argtype = extract32(typemask, (i + 1) * 3, 3); bool is_64bit = (argtype & ~1) == dh_typecode_i64; + bool want_align = false; - if (TCG_TARGET_REG_BITS < 64 && is_64bit) { -#ifdef TCG_TARGET_CALL_ALIGN_ARGS - /* some targets want aligned 64 bit args */ - if (real_args & 1) { - op->args[pi++] = TCG_CALL_DUMMY_ARG; - real_args++; - } +#if defined(CONFIG_TCG_INTERPRETER) + /* + * Align all arguments, so that they land in predictable places + * for passing off to ffi_call. + */ + want_align = true; +#elif defined(TCG_TARGET_CALL_ALIGN_ARGS) + /* Some targets want aligned 64 bit args */ + want_align = is_64bit; #endif - /* If stack grows up, then we will be placing successive - arguments at lower addresses, which means we need to - reverse the order compared to how we would normally - treat either big or little-endian. For those arguments - that will wind up in registers, this still works for - HPPA (the only current STACK_GROWSUP target) since the - argument registers are *also* allocated in decreasing - order. If another such target is added, this logic may - have to get more complicated to differentiate between - stack arguments and register arguments. */ + + if (TCG_TARGET_REG_BITS < 64 && want_align && (real_args & 1)) { + op->args[pi++] = TCG_CALL_DUMMY_ARG; + real_args++; + } + + if (TCG_TARGET_REG_BITS < 64 && is_64bit) { + /* + * If stack grows up, then we will be placing successive + * arguments at lower addresses, which means we need to + * reverse the order compared to how we would normally + * treat either big or little-endian. For those arguments + * that will wind up in registers, this still works for + * HPPA (the only current STACK_GROWSUP target) since the + * argument registers are *also* allocated in decreasing + * order. If another such target is added, this logic may + * have to get more complicated to differentiate between + * stack arguments and register arguments. + */ #if defined(HOST_WORDS_BIGENDIAN) != defined(TCG_TARGET_STACK_GROWSUP) op->args[pi++] = temp_arg(args[i] + 1); op->args[pi++] = temp_arg(args[i]); @@ -3836,6 +3853,7 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op) const int nb_oargs = TCGOP_CALLO(op); const int nb_iargs = TCGOP_CALLI(op); const TCGLifeData arg_life = op->life; + const TCGHelperInfo *info; int flags, nb_regs, i; TCGReg reg; TCGArg arg; @@ -3847,7 +3865,8 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op) TCGRegSet allocated_regs; func_addr = tcg_call_func(op); - flags = tcg_call_flags(op); + info = tcg_call_info(op); + flags = info->flags; nb_regs = ARRAY_SIZE(tcg_target_call_iarg_regs); if (nb_regs > nb_iargs) { @@ -3939,7 +3958,16 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op) save_globals(s, allocated_regs); } +#ifdef CONFIG_TCG_INTERPRETER + { + gpointer hash = (gpointer)(uintptr_t)info->typemask; + ffi_cif *cif = g_hash_table_lookup(ffi_table, hash); + assert(cif != NULL); + tcg_out_call(s, func_addr, cif); + } +#else tcg_out_call(s, func_addr); +#endif /* assign output registers and emit moves if needed */ for(i = 0; i < nb_oargs; i++) { @@ -18,45 +18,26 @@ */ #include "qemu/osdep.h" - -/* Enable TCI assertions only when debugging TCG (and without NDEBUG defined). - * Without assertions, the interpreter runs much faster. */ -#if defined(CONFIG_DEBUG_TCG) -# define tci_assert(cond) assert(cond) -#else -# define tci_assert(cond) ((void)(cond)) -#endif - #include "qemu-common.h" #include "tcg/tcg.h" /* MAX_OPC_PARAM_IARGS */ #include "exec/cpu_ldst.h" #include "tcg/tcg-op.h" #include "qemu/compiler.h" +#include <ffi.h> -#if MAX_OPC_PARAM_IARGS != 6 -# error Fix needed, number of supported input arguments changed! -#endif -#if TCG_TARGET_REG_BITS == 32 -typedef uint64_t (*helper_function)(tcg_target_ulong, tcg_target_ulong, - tcg_target_ulong, tcg_target_ulong, - tcg_target_ulong, tcg_target_ulong, - tcg_target_ulong, tcg_target_ulong, - tcg_target_ulong, tcg_target_ulong, - tcg_target_ulong, tcg_target_ulong); + +/* + * Enable TCI assertions only when debugging TCG (and without NDEBUG defined). + * Without assertions, the interpreter runs much faster. + */ +#if defined(CONFIG_DEBUG_TCG) +# define tci_assert(cond) assert(cond) #else -typedef uint64_t (*helper_function)(tcg_target_ulong, tcg_target_ulong, - tcg_target_ulong, tcg_target_ulong, - tcg_target_ulong, tcg_target_ulong); +# define tci_assert(cond) ((void)(cond)) #endif __thread uintptr_t tci_tb_ptr; -static tcg_target_ulong tci_read_reg(const tcg_target_ulong *regs, TCGReg index) -{ - tci_assert(index < TCG_TARGET_NB_REGS); - return regs[index]; -} - static void tci_write_reg(tcg_target_ulong *regs, TCGReg index, tcg_target_ulong value) { @@ -133,6 +114,7 @@ static tcg_target_ulong tci_read_label(const uint8_t **tb_ptr) * I = immediate (tcg_target_ulong) * l = label or pointer * m = immediate (TCGMemOpIdx) + * n = immediate (call return length) * r = register * s = signed ldst offset */ @@ -153,6 +135,18 @@ static void tci_args_l(const uint8_t **tb_ptr, void **l0) check_size(start, tb_ptr); } +static void tci_args_nll(const uint8_t **tb_ptr, uint8_t *n0, + void **l1, void **l2) +{ + const uint8_t *start = *tb_ptr; + + *n0 = tci_read_b(tb_ptr); + *l1 = (void *)tci_read_label(tb_ptr); + *l2 = (void *)tci_read_label(tb_ptr); + + check_size(start, tb_ptr); +} + static void tci_args_rr(const uint8_t **tb_ptr, TCGReg *r0, TCGReg *r1) { @@ -487,11 +481,14 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env, { const uint8_t *tb_ptr = v_tb_ptr; tcg_target_ulong regs[TCG_TARGET_NB_REGS]; - long tcg_temps[CPU_TEMP_BUF_NLONGS]; - uintptr_t sp_value = (uintptr_t)(tcg_temps + CPU_TEMP_BUF_NLONGS); + uint64_t stack[(TCG_STATIC_CALL_ARGS_SIZE + TCG_STATIC_FRAME_SIZE) + / sizeof(uint64_t)]; + void *call_slots[TCG_STATIC_CALL_ARGS_SIZE / sizeof(uint64_t)]; regs[TCG_AREG0] = (tcg_target_ulong)env; - regs[TCG_REG_CALL_STACK] = sp_value; + regs[TCG_REG_CALL_STACK] = (uintptr_t)stack; + /* Other call_slots entries initialized at first use (see below). */ + call_slots[0] = NULL; tci_assert(tb_ptr); for (;;) { @@ -509,40 +506,58 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env, #endif TCGMemOpIdx oi; int32_t ofs; - void *ptr; + void *ptr, *cif; /* Skip opcode and size entry. */ tb_ptr += 2; switch (opc) { case INDEX_op_call: - tci_args_l(&tb_ptr, &ptr); + /* + * Set up the ffi_avalue array once, delayed until now + * because many TB's do not make any calls. In tcg_gen_callN, + * we arranged for every real argument to be "left-aligned" + * in each 64-bit slot. + */ + if (unlikely(call_slots[0] == NULL)) { + for (int i = 0; i < ARRAY_SIZE(call_slots); ++i) { + call_slots[i] = &stack[i]; + } + } + + tci_args_nll(&tb_ptr, &len, &ptr, &cif); + + /* Helper functions may need to access the "return address" */ tci_tb_ptr = (uintptr_t)tb_ptr; -#if TCG_TARGET_REG_BITS == 32 - tmp64 = ((helper_function)ptr)(tci_read_reg(regs, TCG_REG_R0), - tci_read_reg(regs, TCG_REG_R1), - tci_read_reg(regs, TCG_REG_R2), - tci_read_reg(regs, TCG_REG_R3), - tci_read_reg(regs, TCG_REG_R4), - tci_read_reg(regs, TCG_REG_R5), - tci_read_reg(regs, TCG_REG_R6), - tci_read_reg(regs, TCG_REG_R7), - tci_read_reg(regs, TCG_REG_R8), - tci_read_reg(regs, TCG_REG_R9), - tci_read_reg(regs, TCG_REG_R10), - tci_read_reg(regs, TCG_REG_R11)); - tci_write_reg(regs, TCG_REG_R0, tmp64); - tci_write_reg(regs, TCG_REG_R1, tmp64 >> 32); -#else - tmp64 = ((helper_function)ptr)(tci_read_reg(regs, TCG_REG_R0), - tci_read_reg(regs, TCG_REG_R1), - tci_read_reg(regs, TCG_REG_R2), - tci_read_reg(regs, TCG_REG_R3), - tci_read_reg(regs, TCG_REG_R4), - tci_read_reg(regs, TCG_REG_R5)); - tci_write_reg(regs, TCG_REG_R0, tmp64); -#endif + + ffi_call(cif, ptr, stack, call_slots); + + /* Any result winds up "left-aligned" in the stack[0] slot. */ + switch (len) { + case 0: /* void */ + break; + case 1: /* uint32_t */ + /* + * Note that libffi has an odd special case in that it will + * always widen an integral result to ffi_arg. + */ + if (sizeof(ffi_arg) == 4) { + regs[TCG_REG_R0] = *(uint32_t *)stack; + break; + } + /* fall through */ + case 2: /* uint64_t */ + if (TCG_TARGET_REG_BITS == 32) { + tci_write_reg64(regs, TCG_REG_R1, TCG_REG_R0, stack[0]); + } else { + regs[TCG_REG_R0] = stack[0]; + } + break; + default: + g_assert_not_reached(); + } break; + case INDEX_op_br: tci_args_l(&tb_ptr, &ptr); tb_ptr = ptr; @@ -1119,7 +1134,7 @@ int print_insn_tci(bfd_vma addr, disassemble_info *info) TCGCond c; TCGMemOpIdx oi; uint8_t pos, len; - void *ptr; + void *ptr, *cif; const uint8_t *tb_ptr; status = info->read_memory_func(addr, buf, 2, info); @@ -1147,13 +1162,18 @@ int print_insn_tci(bfd_vma addr, disassemble_info *info) switch (op) { case INDEX_op_br: - case INDEX_op_call: case INDEX_op_exit_tb: case INDEX_op_goto_tb: tci_args_l(&tb_ptr, &ptr); info->fprintf_func(info->stream, "%-12s %p", op_name, ptr); break; + case INDEX_op_call: + tci_args_nll(&tb_ptr, &len, &ptr, &cif); + info->fprintf_func(info->stream, "%-12s %d, %p, %p", + op_name, len, ptr, cif); + break; + case INDEX_op_brcond_i32: case INDEX_op_brcond_i64: tci_args_rrcl(&tb_ptr, &r0, &r1, &c, &ptr); diff --git a/tcg/tci/tcg-target.c.inc b/tcg/tci/tcg-target.c.inc index d54c01b..fa3de99 100644 --- a/tcg/tci/tcg-target.c.inc +++ b/tcg/tci/tcg-target.c.inc @@ -192,23 +192,8 @@ static const int tcg_target_reg_alloc_order[] = { # error Fix needed, number of supported input arguments changed! #endif -static const int tcg_target_call_iarg_regs[] = { - TCG_REG_R0, - TCG_REG_R1, - TCG_REG_R2, - TCG_REG_R3, - TCG_REG_R4, - TCG_REG_R5, -#if TCG_TARGET_REG_BITS == 32 - /* 32 bit hosts need 2 * MAX_OPC_PARAM_IARGS registers. */ - TCG_REG_R6, - TCG_REG_R7, - TCG_REG_R8, - TCG_REG_R9, - TCG_REG_R10, - TCG_REG_R11, -#endif -}; +/* No call arguments via registers. All will be stored on the "stack". */ +static const int tcg_target_call_iarg_regs[] = { }; static const int tcg_target_call_oarg_regs[] = { TCG_REG_R0, @@ -292,8 +277,9 @@ static void tci_out_label(TCGContext *s, TCGLabel *label) static void stack_bounds_check(TCGReg base, target_long offset) { if (base == TCG_REG_CALL_STACK) { - tcg_debug_assert(offset < 0); - tcg_debug_assert(offset >= -(CPU_TEMP_BUF_NLONGS * sizeof(long))); + tcg_debug_assert(offset >= 0); + tcg_debug_assert(offset < (TCG_STATIC_CALL_ARGS_SIZE + + TCG_STATIC_FRAME_SIZE)); } } @@ -593,11 +579,25 @@ static void tcg_out_movi(TCGContext *s, TCGType type, } } -static inline void tcg_out_call(TCGContext *s, const tcg_insn_unit *arg) +static void tcg_out_call(TCGContext *s, const tcg_insn_unit *func, + ffi_cif *cif) { uint8_t *old_code_ptr = s->code_ptr; + uint8_t which; + + if (cif->rtype == &ffi_type_void) { + which = 0; + } else if (cif->rtype->size == 4) { + which = 1; + } else { + tcg_debug_assert(cif->rtype->size == 8); + which = 2; + } tcg_out_op_t(s, INDEX_op_call); - tcg_out_i(s, (uintptr_t)arg); + tcg_out8(s, which); + tcg_out_i(s, (uintptr_t)func); + tcg_out_i(s, (uintptr_t)cif); + old_code_ptr[1] = s->code_ptr - old_code_ptr; } @@ -822,11 +822,9 @@ static void tcg_target_init(TCGContext *s) s->reserved_regs = 0; tcg_regset_set_reg(s->reserved_regs, TCG_REG_CALL_STACK); - /* We use negative offsets from "sp" so that we can distinguish - stores that might pretend to be call arguments. */ - tcg_set_frame(s, TCG_REG_CALL_STACK, - -CPU_TEMP_BUF_NLONGS * sizeof(long), - CPU_TEMP_BUF_NLONGS * sizeof(long)); + /* The call arguments come first, followed by the temp storage. */ + tcg_set_frame(s, TCG_REG_CALL_STACK, TCG_STATIC_CALL_ARGS_SIZE, + TCG_STATIC_FRAME_SIZE); } /* Generate global QEMU prologue and epilogue code. */ diff --git a/tcg/tci/tcg-target.h b/tcg/tci/tcg-target.h index d0b5f3f..f2e5cba 100644 --- a/tcg/tci/tcg-target.h +++ b/tcg/tci/tcg-target.h @@ -162,7 +162,7 @@ typedef enum { /* Used for function call generation. */ #define TCG_TARGET_CALL_STACK_OFFSET 0 -#define TCG_TARGET_STACK_ALIGN 16 +#define TCG_TARGET_STACK_ALIGN 8 #define HAVE_TCG_QEMU_TB_EXEC |