Get --ftrace working on aarch64

This change implements a new approach to function call logging, that's
based on the GCC flag: -fpatchable-function-entry. Read the commentary
in build/config.mk to learn how it works.

ape/aarch64.lds

@@ -71,6 +71,8 @@ SECTIONS {
     KEEP(*(SORT_NONE(.fini)))
   } =0x1f2003d5
 
+  . = ALIGN(__privileged_end > __privileged_start ? 0x4000 : 1);
+
   .privileged : {
     PROVIDE_HIDDEN(__privileged_start = .);
     *(.privileged*)
@@ -271,6 +273,10 @@ SECTIONS {
   }
 }
 
+HIDDEN(__privileged_addr = ROUNDDOWN(__privileged_start, PAGESIZE));
+HIDDEN(__privileged_size = (ROUNDUP(__privileged_end, PAGESIZE) -
+                            ROUNDDOWN(__privileged_start, PAGESIZE)));
+
 PROVIDE(ape_stack_memsz = DEFINED(ape_stack_memsz) ? ape_stack_memsz : STACKSIZE);
 PROVIDE_HIDDEN(_tls_size = _tbss_end - _tdata_start);
 PROVIDE_HIDDEN(_tdata_size = _tdata_end - _tdata_start);

ape/ape.mk

@@ -89,7 +89,7 @@ APE_SRCS = $(APE_SRCS_C) $(APE_SRCS_S)
 APE_OBJS = $(APE_SRCS_S:%.S=o/$(MODE)/%.o)
 APE_CHECKS = $(APE_HDRS:%=o/%.ok)
 
-o/$(MODE)/ape/public/ape.lds: OVERRIDE_CPPFLAGS += -UCOSMO
+o/$(MODE)/ape/public/ape.lds: CPPFLAGS += -UCOSMO
 o/$(MODE)/ape/public/ape.lds:			\
 		ape/public/ape.lds		\
 		ape/ape.lds			\

build/config.mk

@@ -12,13 +12,15 @@
 #   - Reasonably small
 #
 ifeq ($(MODE),)
-CONFIG_CCFLAGS += $(BACKTRACES) $(FTRACE) -O2
+ENABLE_FTRACE = 1
+CONFIG_CCFLAGS += $(BACKTRACES) -O2
 CONFIG_CPPFLAGS += -DSYSDEBUG
 TARGET_ARCH ?= -msse3
 endif
 
 ifeq ($(MODE), aarch64)
-CONFIG_CCFLAGS += -O2 $(BACKTRACES) #$(FTRACE)
+ENABLE_FTRACE = 1
+CONFIG_CCFLAGS += -O2 $(BACKTRACES)
 CONFIG_CPPFLAGS += -DSYSDEBUG
 endif
 
@@ -33,7 +35,8 @@ endif
 #   - Compiles 28% faster
 #
 ifeq ($(MODE), fastbuild)
-CONFIG_CCFLAGS += $(BACKTRACES) $(FTRACE) -O
+ENABLE_FTRACE = 1
+CONFIG_CCFLAGS += $(BACKTRACES) -O
 CONFIG_CPPFLAGS += -DSYSDEBUG -DDWARFLESS
 CONFIG_OFLAGS += -g0
 CONFIG_LDFLAGS += -S
@@ -53,8 +56,9 @@ endif
 #   - GCC 8+ hoists check fails into .text.cold, thus minimizing impact
 #
 ifeq ($(MODE), opt)
+ENABLE_FTRACE = 1
 CONFIG_CPPFLAGS += -DNDEBUG -DSYSDEBUG -msse2avx -Wa,-msse2avx
-CONFIG_CCFLAGS += $(BACKTRACES) $(FTRACE) -O3 -fmerge-all-constants
+CONFIG_CCFLAGS += $(BACKTRACES) -O3 -fmerge-all-constants
 TARGET_ARCH ?= -march=native
 endif
 
@@ -123,16 +127,18 @@ endif
 #   - Enormous binaries
 #
 ifeq ($(MODE), dbg)
+ENABLE_FTRACE = 1
 CONFIG_CPPFLAGS += -DMODE_DBG
-CONFIG_CCFLAGS += $(BACKTRACES) $(FTRACE) -DSYSDEBUG -O -fno-inline
+CONFIG_CCFLAGS += $(BACKTRACES) -DSYSDEBUG -O -fno-inline
 CONFIG_COPTS += -fsanitize=address -fsanitize=undefined
 TARGET_ARCH ?= -msse3
 OVERRIDE_CCFLAGS += -fno-pie
 endif
 
 ifeq ($(MODE), aarch64-dbg)
+ENABLE_FTRACE = 1
 CONFIG_CPPFLAGS += -DMODE_DBG
-CONFIG_CCFLAGS += $(BACKTRACES) $(FTRACE) -DSYSDEBUG -O -fno-inline
+CONFIG_CCFLAGS += $(BACKTRACES) -DSYSDEBUG -O -fno-inline
 CONFIG_COPTS += -fsanitize=undefined
 endif
 
@@ -147,7 +153,8 @@ endif
 #   - No Windows bloat!
 #
 ifeq ($(MODE), sysv)
-CONFIG_CCFLAGS += $(BACKTRACES) $(FTRACE) -O2
+ENABLE_FTRACE = 1
+CONFIG_CCFLAGS += $(BACKTRACES) -O2
 CONFIG_CPPFLAGS += -DSYSDEBUG -DSUPPORT_VECTOR=121
 TARGET_ARCH ?= -msse3
 endif
@@ -363,8 +370,9 @@ endif
 # GCC11 Mode
 # https://justine.lol/compilers/x86_64-linux-musl__x86_64-linux-musl__g++-11.2.0.tar.xz
 ifeq ($(MODE), gcc11)
+ENABLE_FTRACE = 1
 .UNVEIL += rx:/opt/gcc11
-CONFIG_CCFLAGS += $(BACKTRACES) $(FTRACE) -DSYSDEBUG -O2
+CONFIG_CCFLAGS += $(BACKTRACES) -DSYSDEBUG -O2
 AS = /opt/gcc11/bin/x86_64-linux-musl-as
 CC = /opt/gcc11/bin/x86_64-linux-musl-gcc
 CXX = /opt/gcc11/bin/x86_64-linux-musl-g++
@@ -376,7 +384,7 @@ STRIP = /opt/gcc11/bin/x86_64-linux-musl-strip
 OBJCOPY = /opt/gcc11/bin/x86_64-linux-musl-objcopy
 OBJDUMP = /opt/gcc11/bin/x86_64-linux-musl-objdump
 ADDR2LINE = /opt/gcc11/bin/x86_64-linux-musl-addr2line
-CONFIG_CCFLAGS += $(BACKTRACES) $(FTRACE) -O2 -Wno-stringop-overread
+CONFIG_CCFLAGS += $(BACKTRACES) -O2 -Wno-stringop-overread
 CONFIG_CFLAGS += -Wno-old-style-definition
 CONFIG_CPPFLAGS += -DNDEBUG -DSYSDEBUG
 TARGET_ARCH ?= -msse3
@@ -385,7 +393,7 @@ endif
 # LLVM Mode
 ifeq ($(MODE), llvm)
 TARGET_ARCH ?= -msse3
-CONFIG_CCFLAGS += $(BACKTRACES) $(FTRACE) -DSYSDEBUG -O2
+CONFIG_CCFLAGS += $(BACKTRACES) -DSYSDEBUG -O2
 AS = clang
 CC = clang
 CXX = clang++
@@ -432,3 +440,59 @@ CONFIG_CXXFLAGS += -std=c++11
 TARGET_ARCH ?= -msse3
 
 endif
+
+ifneq ($(ENABLE_FTRACE),)
+CONFIG_CPPFLAGS += -DFTRACE
+FTRACE_CCFLAGS = -fno-inline-functions-called-once
+OVERRIDE_CFLAGS += $(FTRACE_CCFLAGS)
+OVERRIDE_CXXFLAGS += $(FTRACE_CCFLAGS)
+# function prologue nops for --ftrace
+ifeq ($(ARCH), x86_64)
+# this flag causes gcc to generate functions like this
+#
+#       nop nop nop nop nop nop nop nop nop
+#     func:
+#       nop nop
+#       ...
+#
+# which tool/build/fixupobj.c improves at build time like this
+#
+#       nop nop nop nop nop nop nop nop nop
+#     func:
+#       xchg %ax,%ax
+#       ...
+#
+# which --ftrace morphs at runtime like this
+#
+#       ud2                # 2 bytes
+#       call ftrace_hook   # 5 bytes
+#       jmp +2             # 2 bytes
+#     func:
+#       jmp -7             # 2 bytes
+#       ...
+#
+CONFIG_CCFLAGS += -fpatchable-function-entry=11,9
+endif
+ifeq ($(ARCH), aarch64)
+# this flag causes gcc to generate functions like this
+#
+#       nop nop nop nop nop nop
+#     func:
+#       nop
+#       ...
+#
+# which --ftrace morphs at runtime like this
+#
+#       brk #31337
+#       stp x29,x30,[sp,#-16]!
+#       mov x29,sp
+#       bl  ftrace_hook
+#       ldp x29,x30,[sp],#16
+#       b   +1
+#     func:
+#       b   -5
+#       ...
+#
+CONFIG_CCFLAGS += -fpatchable-function-entry=7,6
+endif
+endif

build/definitions.mk

@@ -137,9 +137,6 @@ else
 IMAGE_BASE_VIRTUAL ?= 0x400000
 endif
 
-FTRACE =								\
-	-pg
-
 BACKTRACES =								\
 	-fno-omit-frame-pointer						\
 	-fno-optimize-sibling-calls					\

dsp/bmp/bmp.mk

@@ -41,11 +41,11 @@ $(DSP_BMP_A).pkg:				\
 
 o/$(MODE)/dsp/bmp/float2short.o			\
 o/$(MODE)/dsp/bmp/scalevolume.o: private	\
-		OVERRIDE_CFLAGS +=		\
+		CFLAGS +=			\
 			$(MATHEMATICAL)
 
 o/tiny/dsp/bmp/scalevolume.o: private		\
-		OVERRIDE_CFLAGS +=		\
+		CFLAGS +=			\
 			-Os
 
 DSP_BMP_LIBS = $(foreach x,$(DSP_BMP_ARTIFACTS),$($(x)))

dsp/core/core.mk

@@ -40,15 +40,15 @@ o/$(MODE)/dsp/core/magikarp.o			\
 o/$(MODE)/dsp/core/c93654369.o			\
 o/$(MODE)/dsp/core/float2short.o		\
 o/$(MODE)/dsp/core/scalevolume.o: private	\
-		OVERRIDE_CFLAGS +=		\
+		CFLAGS +=			\
 			$(MATHEMATICAL)
 
 o/tiny/dsp/core/scalevolume.o: private		\
-		OVERRIDE_CFLAGS +=		\
+		CFLAGS +=			\
 			-Os
 
 o/$(MODE)/dsp/core/det3.o: private		\
-		OVERRIDE_CFLAGS +=		\
+		CFLAGS +=			\
 			-ffast-math
 
 DSP_CORE_LIBS = $(foreach x,$(DSP_CORE_ARTIFACTS),$($(x)))

dsp/mpeg/mpeg.mk

@@ -52,7 +52,7 @@ $(DSP_MPEG_A).pkg:				\
 		$(foreach x,$(DSP_MPEG_A_DIRECTDEPS),$($(x)_A).pkg)
 
 o/$(MODE)/dsp/mpeg/clamp4int256-k8.o: private	\
-		OVERRIDE_CFLAGS +=		\
+		CFLAGS +=			\
 			-Os
 
 DSP_MPEG_LIBS = $(foreach x,$(DSP_MPEG_ARTIFACTS),$($(x)))

dsp/scale/scale.mk

@@ -52,7 +52,7 @@ o/$(MODE)/dsp/scale/cdecimate2xuint8x8.o	\
 o/$(MODE)/dsp/scale/gyarados.o			\
 o/$(MODE)/dsp/scale/magikarp.o			\
 o/$(MODE)/dsp/scale/scale.o: private		\
-		OVERRIDE_CFLAGS +=		\
+		CFLAGS +=			\
 			$(MATHEMATICAL)
 
 DSP_SCALE_LIBS = $(foreach x,$(DSP_SCALE_ARTIFACTS),$($(x)))

dsp/tty/tty.mk

@@ -54,12 +54,12 @@ $(DSP_TTY_A).pkg:				\
 		$(foreach x,$(DSP_TTY_A_DIRECTDEPS),$($(x)_A).pkg)
 
 o/$(MODE)/dsp/tty/ttyraster.o: private		\
-		OVERRIDE_CFLAGS +=		\
+		CFLAGS +=			\
 			$(MATHEMATICAL)
 
 ifeq ($(ARCH), aarch64)
 # takes 14 seconds to compile with aarch64 gcc
-o/$(MODE)/dsp/tty/ttyraster.o: private OVERRIDE_CFLAGS += -O1
+o/$(MODE)/dsp/tty/ttyraster.o: private CFLAGS += -O1
 o/$(MODE)/dsp/tty/ttyraster.o: private QUOTA += -C128
 endif
 

dsp/tty/windex-avx2.S

@@ -23,10 +23,11 @@
 //	@param	rdi points to nonempty array
 //	@param	rsi is item count divisible by 16
 //	@note	needs avx2 (haswell+)
+	.ftrace1
 windex_avx2:
+	.ftrace2
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	and	$-32,%rsp
 	sub	$32,%rsp
 	vmovdqa	(%rdi),%ymm1

dsp/tty/windex-sse4.S

@@ -23,10 +23,11 @@
 //	@param	rdi points to nonempty array
 //	@param	esi is 16-byte aligned 8+ / 8 multiple array item count
 //	@note	needs sse4 (nehalem+)
+	.ftrace1
 windex_sse4:
+	.ftrace2
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	$8,%eax
 	sub	$32,%rsp
 	movdqa	(%rdi),%xmm2

examples/clock.c

@@ -8,6 +8,7 @@
 ╚─────────────────────────────────────────────────────────────────*/
 #endif
 #include "libc/calls/struct/timespec.h"
+#include "libc/log/log.h"
 #include "libc/stdio/stdio.h"
 #include "libc/time/time.h"
 
@@ -16,6 +17,7 @@
  */
 
 int main(int argc, char *argv[]) {
+  ShowCrashReports();
   unsigned long i;
   volatile unsigned long x;
   struct timespec now, start, next, interval;

examples/examples.mk

@@ -103,7 +103,7 @@ o/$(MODE)/examples/examples.pkg:						\
 		$(foreach x,$(EXAMPLES_DIRECTDEPS),$($(x)_A).pkg)
 
 o/$(MODE)/examples/unbourne.o: private						\
-		OVERRIDE_CPPFLAGS +=						\
+		CPPFLAGS +=							\
 			-DSTACK_FRAME_UNLIMITED					\
 			-fpie
 
@@ -169,7 +169,7 @@ o/$(MODE)/examples/symtab.com:							\
 	@$(MAKE_SYMTAB_ZIP)
 
 o/$(MODE)/examples/picol.o: private				\
-		OVERRIDE_CPPFLAGS +=				\
+		CPPFLAGS +=					\
 			-DSTACK_FRAME_UNLIMITED
 
 o/$(MODE)/examples/picol.com.dbg:				\

libc/calls/calls.mk

@@ -69,21 +69,21 @@ $(LIBC_CALLS_A).pkg:					\
 # we can't use asan because:
 #   siginfo_t memory is owned by kernels
 o/$(MODE)/libc/calls/siginfo2cosmo.o: private		\
-		OVERRIDE_COPTS +=			\
+		COPTS +=				\
 			-ffreestanding			\
 			-fno-sanitize=address
 
 # we can't use asan because:
 #   ucontext_t memory is owned by xnu kernel
 o/$(MODE)/libc/calls/sigenter-xnu.o: private		\
-		OVERRIDE_COPTS +=			\
+		COPTS +=				\
 			-ffreestanding			\
 			-fno-sanitize=address
 
 # we can't use asan because:
 #   vdso memory is owned by linux kernel
 o/$(MODE)/libc/calls/vdsofunc.greg.o: private		\
-		OVERRIDE_COPTS +=			\
+		COPTS +=				\
 			-ffreestanding			\
 			-fno-sanitize=address
 
@@ -92,7 +92,7 @@ o/$(MODE)/libc/calls/vdsofunc.greg.o: private		\
 o/$(MODE)/libc/calls/ntspawn.o				\
 o/$(MODE)/libc/calls/mkntcmdline.o			\
 o/$(MODE)/libc/calls/mkntenvblock.o: private		\
-		OVERRIDE_COPTS +=			\
+		COPTS +=				\
 			-ffreestanding			\
 			-fno-sanitize=address
 
@@ -100,7 +100,7 @@ o/$(MODE)/libc/calls/mkntenvblock.o: private		\
 #   windows owns the data structure
 o/$(MODE)/libc/calls/wincrash.o				\
 o/$(MODE)/libc/calls/ntcontext2linux.o: private		\
-		OVERRIDE_COPTS +=			\
+		COPTS +=				\
 			-fno-sanitize=all
 
 # we always want -O3 because:
@@ -111,7 +111,7 @@ o/$(MODE)/libc/calls/sigenter-netbsd.o			\
 o/$(MODE)/libc/calls/sigenter-openbsd.o			\
 o/$(MODE)/libc/calls/sigenter-xnu.o			\
 o/$(MODE)/libc/calls/ntcontext2linux.o: private		\
-		OVERRIDE_COPTS +=			\
+		COPTS +=				\
 			-O3
 
 # we must disable static stack safety because:
@@ -125,14 +125,14 @@ o/$(MODE)/libc/calls/execve-sysv.o			\
 o/$(MODE)/libc/calls/readlinkat-nt.o			\
 o/$(MODE)/libc/calls/execve-nt.greg.o			\
 o/$(MODE)/libc/calls/mkntenvblock.o: private		\
-		OVERRIDE_CPPFLAGS +=			\
+		CPPFLAGS +=				\
 			-DSTACK_FRAME_UNLIMITED
 
 # we must segregate codegen because:
 #   file contains multiple independently linkable apis
 o/$(MODE)/libc/calls/ioctl-siocgifconf.o		\
 o/$(MODE)/libc/calls/ioctl-siocgifconf-nt.o: private	\
-		OVERRIDE_COPTS +=			\
+		COPTS +=				\
 			-ffunction-sections		\
 			-fdata-sections
 
@@ -156,7 +156,7 @@ o//libc/calls/ioctl_tiocgwinsz.o			\
 o//libc/calls/ioctl_tiocswinsz-nt.o			\
 o//libc/calls/ioctl_tiocswinsz.o			\
 o//libc/calls/fcntl.o: private				\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-Os
 
 # we always want -Os because:
@@ -165,7 +165,7 @@ o//libc/calls/getcwd.greg.o				\
 o//libc/calls/getcwd-nt.greg.o				\
 o//libc/calls/getcwd-xnu.greg.o				\
 o//libc/calls/statfs2cosmo.o: private			\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-Os
 
 # we always want -O2 because:
@@ -180,17 +180,17 @@ o/$(MODE)/libc/calls/timespec_frommicros.o		\
 o/$(MODE)/libc/calls/timeval_tomillis.o			\
 o/$(MODE)/libc/calls/timeval_frommillis.o		\
 o/$(MODE)/libc/calls/timeval_frommicros.o: private	\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-O2
 
 o/$(MODE)/libc/calls/pledge-linux.o			\
 o/$(MODE)/libc/calls/unveil.o: private			\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-DSTACK_FRAME_UNLIMITED
 
 ifeq ($(ARCH), aarch64)
-o/$(MODE)/libc/calls/sigaction.o: private OVERRIDE_CFLAGS += -mcmodel=large
+o/$(MODE)/libc/calls/sigaction.o: private CFLAGS += -mcmodel=large
-o/$(MODE)/libc/calls/getloadavg-nt.o: private OVERRIDE_CFLAGS += -ffreestanding
+o/$(MODE)/libc/calls/getloadavg-nt.o: private CFLAGS += -ffreestanding
 endif
 
 # we want -Os because:
@@ -198,7 +198,7 @@ endif
 # we need pic because:
 #   so it can be an LD_PRELOAD payload
 o/$(MODE)/libc/calls/pledge-linux.o: private		\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-Os				\
 			-fPIC
 

libc/calls/fmodl.S

@@ -26,9 +26,10 @@
 //	@return	remainder ∈ (-|𝑦|,|𝑦|) in %st
 //	@define	𝑥-truncl(𝑥/𝑦)*𝑦
 //	@see	emod()
-fmodl:	push	%rbp
+	.ftrace1
+fmodl:	.ftrace2
+	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	fldt	32(%rbp)
 	fldt	16(%rbp)
 1:	fprem

libc/elf/struct/sym.h

@@ -12,7 +12,9 @@ typedef struct Elf64_Sym {
   uint8_t st_other;
   /* SHN_UNDEF, <section index>, SHN_ABS, SHN_COMMON, etc. */
   Elf64_Section st_shndx;
+  /* byte offset into GetElfSectionAddress(st_shndx) */
   Elf64_Addr st_value;
+  /* byte length optionally set by .size directive */
   Elf64_Xword st_size;
 } Elf64_Sym;
 

libc/fmt/fmt.mk

@@ -55,14 +55,14 @@ $(LIBC_FMT_A).pkg:				\
 		$(foreach x,$(LIBC_FMT_A_DIRECTDEPS),$($(x)_A).pkg)
 
 $(LIBC_FMT_A_OBJS): private			\
-		OVERRIDE_CFLAGS +=		\
+		CFLAGS +=			\
 			-fno-jump-tables
 
 o/$(MODE)/libc/fmt/formatint64.o		\
 o/$(MODE)/libc/fmt/formatint64thousands.o	\
 o/$(MODE)/libc/fmt/dosdatetimetounix.o		\
 o/$(MODE)/libc/fmt/itoa64radix10.greg.o: private\
-		OVERRIDE_CFLAGS +=		\
+		CFLAGS +=			\
 			-O3
 
 o/$(MODE)/libc/fmt/atoi.o			\
@@ -74,7 +74,7 @@ o/$(MODE)/libc/fmt/strtoimax.o			\
 o/$(MODE)/libc/fmt/strtoumax.o			\
 o/$(MODE)/libc/fmt/wcstoimax.o			\
 o/$(MODE)/libc/fmt/wcstoumax.o: private		\
-		OVERRIDE_CFLAGS +=		\
+		CFLAGS +=			\
 			-Os
 
 # we can't use compiler magic because:
@@ -82,7 +82,7 @@ o/$(MODE)/libc/fmt/wcstoumax.o: private		\
 o/$(MODE)/libc/fmt/strerrno.greg.o		\
 o/$(MODE)/libc/fmt/strerrdoc.greg.o		\
 o/$(MODE)/libc/fmt/strerror_wr.greg.o: private	\
-		OVERRIDE_CFLAGS +=		\
+		CFLAGS +=			\
 			-fpie			\
 			-ffreestanding		\
 			$(NO_MAGIC)

libc/intrin/aarch64/asmdefs.internal.h

@@ -1,5 +1,6 @@
 #ifndef COSMOPOLITAN_LIBC_INTRIN_AARCH64_ASMDEFS_H_
 #define COSMOPOLITAN_LIBC_INTRIN_AARCH64_ASMDEFS_H_
+#include "libc/macros.internal.h"
 #ifdef __ASSEMBLER__
 // clang-format off
 
@@ -44,7 +45,9 @@ GNU_PROPERTY (FEATURE_1_AND, FEATURE_1_BTI|FEATURE_1_PAC)
   .global name;		\
   .type name,%function;	\
   .align alignment;		\
+  .ftrace1;                     \
   name:			\
+  .ftrace2;		\
   .cfi_startproc;	\
   BTI_C;
 

libc/intrin/aarch64/memchr.S

@@ -25,7 +25,7 @@
 │  SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                      │
 │                                                                              │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/intrin/aarch64/asmdefs.h"
+#include "libc/intrin/aarch64/asmdefs.internal.h"
 
 #define __memchr_aarch64 memchr
 

libc/intrin/aarch64/memcmp.S

@@ -25,7 +25,7 @@
 │  SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                      │
 │                                                                              │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/intrin/aarch64/asmdefs.h"
+#include "libc/intrin/aarch64/asmdefs.internal.h"
 
 #define __memcmp_aarch64 memcmp
 

libc/intrin/aarch64/memcpy.S

@@ -25,7 +25,7 @@
 │  SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                      │
 │                                                                              │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/intrin/aarch64/asmdefs.h"
+#include "libc/intrin/aarch64/asmdefs.internal.h"
 
 #define __memcpy_aarch64_simd memcpy
 #define __memmove_aarch64_simd memmove

libc/intrin/aarch64/memrchr.S

@@ -25,7 +25,7 @@
 │  SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                      │
 │                                                                              │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/intrin/aarch64/asmdefs.h"
+#include "libc/intrin/aarch64/asmdefs.internal.h"
 
 #define __memrchr_aarch64 memrchr
 

libc/intrin/aarch64/memset.S

@@ -25,7 +25,7 @@
 │  SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                      │
 │                                                                              │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/intrin/aarch64/asmdefs.h"
+#include "libc/intrin/aarch64/asmdefs.internal.h"
 
 #define __memset_aarch64 memset
 

libc/intrin/aarch64/stpcpy.S

@@ -25,7 +25,7 @@
 │  SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                      │
 │                                                                              │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/intrin/aarch64/asmdefs.h"
+#include "libc/intrin/aarch64/asmdefs.internal.h"
 
 #define __stpcpy_aarch64 stpcpy
 

libc/intrin/aarch64/strchr.S

@@ -25,7 +25,7 @@
 │  SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                      │
 │                                                                              │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/intrin/aarch64/asmdefs.h"
+#include "libc/intrin/aarch64/asmdefs.internal.h"
 
 #define __strchr_aarch64 strchr
 

libc/intrin/aarch64/strchrnul.S

@@ -25,7 +25,7 @@
 │  SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                      │
 │                                                                              │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/intrin/aarch64/asmdefs.h"
+#include "libc/intrin/aarch64/asmdefs.internal.h"
 
 #define __strchrnul_aarch64 strchrnul
 

libc/intrin/aarch64/strcmp.S

@@ -25,7 +25,7 @@
 │  SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                      │
 │                                                                              │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/intrin/aarch64/asmdefs.h"
+#include "libc/intrin/aarch64/asmdefs.internal.h"
 
 #define __strcmp_aarch64 strcmp
 

libc/intrin/aarch64/strcpy.S

@@ -25,7 +25,7 @@
 │  SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                      │
 │                                                                              │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/intrin/aarch64/asmdefs.h"
+#include "libc/intrin/aarch64/asmdefs.internal.h"
 
 #define __strcpy_aarch64 strcpy
 

libc/intrin/aarch64/strlen.S

@@ -25,7 +25,7 @@
 │  SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                      │
 │                                                                              │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/intrin/aarch64/asmdefs.h"
+#include "libc/intrin/aarch64/asmdefs.internal.h"
 
 #define __strlen_aarch64 strlen
 

libc/intrin/aarch64/strncmp.S

@@ -25,7 +25,7 @@
 │  SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                      │
 │                                                                              │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/intrin/aarch64/asmdefs.h"
+#include "libc/intrin/aarch64/asmdefs.internal.h"
 
 #define __strncmp_aarch64 strncmp
 

libc/intrin/aarch64/strnlen.S

@@ -25,7 +25,7 @@
 │  SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                      │
 │                                                                              │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/intrin/aarch64/asmdefs.h"
+#include "libc/intrin/aarch64/asmdefs.internal.h"
 
 #define __strnlen_aarch64 strnlen
 

libc/intrin/aarch64/strrchr.S

@@ -25,7 +25,7 @@
 │  SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                      │
 │                                                                              │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/intrin/aarch64/asmdefs.h"
+#include "libc/intrin/aarch64/asmdefs.internal.h"
 
 #define __strrchr_aarch64 strrchr
 

libc/intrin/assertfail.greg.c

@@ -26,6 +26,7 @@
 #include "libc/intrin/weaken.h"
 #include "libc/log/backtrace.internal.h"
 #include "libc/log/internal.h"
+#include "libc/runtime/internal.h"
 #include "libc/runtime/runtime.h"
 #include "libc/runtime/symbols.internal.h"
 #include "libc/thread/thread.h"
@@ -38,7 +39,7 @@ relegated void __assert_fail(const char *expr, const char *file, int line) {
     strace_enabled(-1);
     ftrace_enabled(-1);
     owner = 0;
-    me = __tls_enabled ? __get_tls()->tib_tid : sys_gettid();
+    me = __tls_enabled ? __get_tls()->tib_tid : __pid;
     pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, 0);
     kprintf("%s:%d: assert(%s) failed (tid %d) %m\n", file, line, expr, me);
     if (__vforked ||

libc/intrin/bsr128.S

@@ -33,9 +33,10 @@
 //	@param	rsi:rdi is 128-bit unsigned 𝑥 value
 //	@return	eax number in range [0,128) or undef if 𝑥 is 0
 //	@see	also treasure trove of nearly identical functions
+	.ftrace1
 _bsr128:
+	.ftrace2
 	.leafprologue
-	.profilable
 	bsr	%rsi,%rax
 	jnz	2f
 	bsr	%rdi,%rax

libc/intrin/bzero.c

@@ -26,7 +26,7 @@
 typedef char xmm_t __attribute__((__vector_size__(16), __aligned__(1)));
 typedef long long xmm_a __attribute__((__vector_size__(16), __aligned__(16)));
 
-static void bzero_sse(char *p, size_t n) {
+static void bzero128(char *p, size_t n) {
   xmm_t v = {0};
   if (IsAsan()) __asan_verify(p, n);
   if (n <= 32) {
@@ -162,6 +162,6 @@ void bzero(void *p, size_t n) {
     bzero_avx(b, n);
 #endif
   } else {
-    bzero_sse(b, n);
+    bzero128(b, n);
   }
 }

libc/intrin/intrin.mk

@@ -16,6 +16,7 @@ LIBC_INTRIN_A_CHECKS = $(LIBC_INTRIN_A).pkg
 
 ifeq ($(ARCH), aarch64)
 LIBC_INTRIN_A_SRCS_S += $(wildcard libc/intrin/aarch64/*.S)
+LIBC_INTRIN_A_HDRS += libc/intrin/aarch64/asmdefs.internal.h
 endif
 
 LIBC_INTRIN_A_OBJS =					\
@@ -49,39 +50,39 @@ $(LIBC_INTRIN_A).pkg:					\
 # we can't use asan because:
 #   __strace_init() calls this before asan is initialized
 o/$(MODE)/libc/intrin/strace_enabled.o: private		\
-		OVERRIDE_COPTS +=			\
+		COPTS +=				\
 			-fno-sanitize=address
 
 # we can't use asan because:
 #   asan guard pages haven't been allocated yet
 o/$(MODE)/libc/intrin/directmap.o			\
 o/$(MODE)/libc/intrin/directmap-nt.o: private		\
-		OVERRIDE_COPTS +=			\
+		COPTS +=				\
 			-ffreestanding			\
 			-fno-sanitize=address
 
 # we want small code size because:
 #   to keep .text.head under 4096 bytes
 o/$(MODE)/libc/intrin/mman.greg.o: private		\
-		OVERRIDE_COPTS +=			\
+		COPTS +=				\
 			-Os
 
 # we can't use asan and ubsan because:
 #   this is asan and ubsan
 o/$(MODE)/libc/intrin/asan.o				\
 o/$(MODE)/libc/intrin/ubsan.o: private			\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-fno-sanitize=all		\
 			-fno-stack-protector
 
 o/$(MODE)/libc/intrin/asan.o: private			\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-O2				\
 			-finline			\
 			-finline-functions
 
 o/$(MODE)/libc/intrin/asanthunk.o: private		\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-x-no-pg			\
 			$(MNO_FENTRY)			\
 			-ffreestanding			\
@@ -95,7 +96,7 @@ o/$(MODE)/libc/intrin/strerrno.greg.o			\
 o/$(MODE)/libc/intrin/strerrdoc.greg.o			\
 o/$(MODE)/libc/intrin/strerror_wr.greg.o		\
 o/$(MODE)/libc/intrin/kprintf.greg.o: private		\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-fpie				\
 			-fwrapv				\
 			-x-no-pg			\
@@ -109,10 +110,9 @@ o/$(MODE)/libc/intrin/kprintf.greg.o: private		\
 o/$(MODE)/libc/intrin/futex_wait.o			\
 o/$(MODE)/libc/intrin/futex_wake.o			\
 o/$(MODE)/libc/intrin/gettid.greg.o			\
-o/$(MODE)/libc/intrin/sys_gettid.greg.o			\
 o/$(MODE)/libc/intrin/_trylock_debug_4.o		\
 o/$(MODE)/libc/intrin/_spinlock_debug_4.o: private	\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-fwrapv				\
 			-x-no-pg			\
 			$(MNO_FENTRY)			\
@@ -125,13 +125,13 @@ o/$(MODE)/libc/intrin/_spinlock_debug_4.o: private	\
 #   global gone could be raised
 o/$(MODE)/libc/intrin/exit.o				\
 o/$(MODE)/libc/intrin/restorewintty.o: private		\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-fno-sanitize=all
 
 # we can't use -ftrapv because:
 #   this file implements it
 o/$(MODE)/libc/intrin/ftrapv.o: private			\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-ffunction-sections		\
 			-ffreestanding			\
 			-fwrapv
@@ -142,7 +142,7 @@ o/$(MODE)/libc/intrin/describeflags.o			\
 o/$(MODE)/libc/intrin/describeframe.o			\
 o/$(MODE)/libc/intrin/describemapflags.o		\
 o/$(MODE)/libc/intrin/describeprotflags.o: private	\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-fno-sanitize=address
 
 o/$(MODE)/libc/intrin/exit1.greg.o			\
@@ -180,7 +180,7 @@ o/$(MODE)/libc/intrin/createfilemappingnuma.o		\
 o/$(MODE)/libc/intrin/waitformultipleobjects.o		\
 o/$(MODE)/libc/intrin/generateconsolectrlevent.o	\
 o/$(MODE)/libc/intrin/wsawaitformultipleevents.o: private\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-Os				\
 			-fwrapv				\
 			-ffreestanding			\
@@ -188,20 +188,20 @@ o/$(MODE)/libc/intrin/wsawaitformultipleevents.o: private\
 			-fno-sanitize=all
 
 o//libc/intrin/memmove.o: private			\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-fno-toplevel-reorder
 
 o//libc/intrin/bzero.o					\
 o//libc/intrin/memcmp.o					\
 o//libc/intrin/memset.o					\
 o//libc/intrin/memmove.o: private			\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-O2 -finline
 
 o/$(MODE)/libc/intrin/bzero.o				\
 o/$(MODE)/libc/intrin/memcmp.o				\
 o/$(MODE)/libc/intrin/memmove.o: private		\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-fpie
 
 # these assembly files are safe to build on aarch64

libc/intrin/kprintf.greg.c

@@ -382,7 +382,7 @@ privileged static size_t kformat(char *b, size_t n, const char *fmt,
             if (tib) {
               x = atomic_load_explicit(&tib->tib_tid, memory_order_relaxed);
             } else {
-              x = sys_gettid();
+              x = __pid;
             }
             if (!__nocolor && p + 7 <= e) {
               *p++ = '\e';

libc/intrin/sched_yield.S

@@ -19,13 +19,14 @@
 #include "libc/dce.h"
 #include "libc/sysv/consts/nr.h"
 #include "libc/macros.internal.h"
-.privileged
 
 //	Relinquishes scheduled quantum.
 //
 //	@return	0 on success, or -1 w/ errno
 //	@norestart
+	.ftrace1
 sched_yield:
+	.ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp

libc/intrin/strcmp.c

@@ -19,6 +19,7 @@
 #include "libc/dce.h"
 #include "libc/intrin/asan.internal.h"
 #include "libc/str/str.h"
+#ifndef __aarch64__
 
 /**
  * Compares NUL-terminated strings.
@@ -28,7 +29,7 @@
  * @return is <0, 0, or >0 based on uint8_t comparison
  * @asyncsignalsafe
  */
-noasan int strcmp(const char *a, const char *b) {
+int strcmp(const char *a, const char *b) {
   int c;
   size_t i = 0;
   uint64_t v, w, d;
@@ -58,3 +59,5 @@ noasan int strcmp(const char *a, const char *b) {
   }
   return (a[i] & 255) - (b[i] & 255);
 }
+
+#endif /* __aarch64__ */

libc/intrin/sys_gettid.greg.c

@@ -24,7 +24,7 @@
 
 __msabi extern typeof(GetCurrentThreadId) *const __imp_GetCurrentThreadId;
 
-privileged int sys_gettid(void) {
+int sys_gettid(void) {
 #ifdef __x86_64__
   int tid;
   int64_t wut;

libc/intrin/tpenc.S

@@ -23,8 +23,9 @@
 //	@param	edi is int to encode
 //	@return	rax is word-encoded byte buffer
 //	@note	invented on a napkin in a new jersey diner
-_tpenc:	.leafprologue
+	.ftrace1
-	.profilable
+_tpenc:	.ftrace2
+	.leafprologue
 	mov	%edi,%edi
 	xor	%eax,%eax
 	cmp	$127,%edi
@@ -47,20 +48,20 @@ _tpenc:	.leafprologue
 	.rodata
 	.balign	4
 	.underrun
-kTpenc:	.rept	4			# MSB≤10 (0x7FF)
+kTpenc:	.rept	4			// MSB≤10 (0x7FF)
-	.byte	1,0b11000000		# len,mark
+	.byte	1,0b11000000		// len,mark
 	.endr
-	.rept	5			# MSB≤15 (0xFFFF)
+	.rept	5			// MSB≤15 (0xFFFF)
-	.byte	2,0b11100000		# len,mark
+	.byte	2,0b11100000		// len,mark
 	.endr
-	.rept	5			# MSB≤20 (0x1FFFFF)
+	.rept	5			// MSB≤20 (0x1FFFFF)
-	.byte	3,0b11110000		# len,mark
+	.byte	3,0b11110000		// len,mark
 	.endr
-	.rept	5			# MSB≤25 (0x3FFFFFF)
+	.rept	5			// MSB≤25 (0x3FFFFFF)
-	.byte	4,0b11111000		# len,mark
+	.byte	4,0b11111000		// len,mark
 	.endr
-	.rept	6			# MSB≤31 (0xffffffff)
+	.rept	6			// MSB≤31 (0xffffffff)
-	.byte	5,0b11111100		# len,mark
+	.byte	5,0b11111100		// len,mark
 	.endr
 	.zero	2
 	.endobj	kTpenc

libc/log/log.mk

@@ -58,15 +58,15 @@ $(LIBC_LOG_A).pkg:					\
 
 o/$(MODE)/libc/log/backtrace2.o				\
 o/$(MODE)/libc/log/backtrace3.o: private		\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-fno-sanitize=all
 
 o/$(MODE)/libc/log/checkfail.o: private			\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-mgeneral-regs-only
 
 o/$(MODE)/libc/log/watch.o: private			\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			-ffreestanding
 
 o/$(MODE)/libc/log/watch.o				\
@@ -82,7 +82,7 @@ o/$(MODE)/libc/log/startfatal.o				\
 o/$(MODE)/libc/log/startfatal_ndebug.o			\
 o/$(MODE)/libc/log/ubsan.o				\
 o/$(MODE)/libc/log/die.o: private			\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			$(NO_MAGIC)
 
 LIBC_LOG_LIBS = $(foreach x,$(LIBC_LOG_ARTIFACTS),$($(x)))

libc/macros.internal.h

@@ -213,6 +213,36 @@
  .org	1347b+\fieldsize,0x00
 .endm
 
+//	Inserts --ftrace pre-prologue.
+//	This goes immediately before the function symbol.
+//	@see	.ftrace2
+.macro	.ftrace1
+#ifdef FTRACE
+#ifdef __x86_64__
+	.rept	9
+	nop
+	.endr
+#elif defined(__aarch64__)
+	.rept	6
+	nop
+	.endr
+#endif /* __x86_64__ */
+#endif /* FTRACE */
+.endm
+
+//	Inserts --ftrace prologue.
+//	This goes immediately after the function symbol.
+//	@see	.ftrace1
+.macro	.ftrace2
+#ifdef FTRACE
+#ifdef __x86_64__
+	xchg	%ax,%ax
+#elif defined(__aarch64__)
+	nop
+#endif /* __x86_64__ */
+#endif /* FTRACE */
+.endm
+
 #ifdef __x86_64__
 
 #if __MNO_VZEROUPPER__ + 0
@@ -346,7 +376,7 @@
 
 //	Pads function prologue unconditionally for runtime hooking.
 //	@cost	≥0.3 cycles, 5 bytes
-//	@see	.profilable
+//	@see	.ftrace1
 .macro	.hookable
 	.byte	0x0f
 	.byte	0x1f
@@ -473,31 +503,6 @@
 #endif
 .endm
 
-//	Inserts profiling hook in prologue if cc wants it.
-//
-//	Cosmopolitan does this in a slightly different way from normal
-//	GNU toolchains. We always use the -mnop-mcount behavior, since
-//	the runtime is able to morph the binary at runtime. It is good
-//	since we can put hooks for profiling and function tracing into
-//	most builds, without any impact on performance.
-//
-//	@cost	≥0.3 cycles, 5 bytes
-//	@see	build/compile
-.macro	.profilable
-#ifdef __PG__
-1382:
-#if defined(__MFENTRY__)
-	call	__fentry__
-#elif defined(__PIC__) || defined(__PIE__)
-//	nopw 0x00(%rax,%rax,1)
-	.byte	0x66,0x0f,0x1f,0x44,0x00,0x00
-#else
-//	nopl 0x00(%rax,%rax,1)
-	.byte	0x0f,0x1f,0x44,0x00,0x00
-#endif
-#endif
-.endm
-
 //	Pushes RVA on stack of linktime mergeable string literal.
 //	@see	popstr
 .macro	pushstr	text

libc/nexgen32e/gclongjmp.S

@@ -30,11 +30,12 @@
 //	@see	examples/ctrlc.c
 //	@threadsafe
 //	@noreturn
+	.ftrace1
 _gclongjmp:
+	.ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	%fs:0,%r12			# __get_tls()
 	mov	0x18(%r12),%r12			# Tls::garbages
 	test	%r12,%r12

libc/nexgen32e/longjmp.S

@@ -25,7 +25,9 @@
 //	@noreturn
 //	@see	_gclongjmp()
 //	@see	siglongjmp()
+	.ftrace1
 longjmp:
+	.ftrace2
 #ifdef __x86_64__
 	mov	%esi,%eax
 	test	%eax,%eax

libc/nexgen32e/mul4x4adx.S

@@ -32,10 +32,11 @@
 //	@param	rdx is right hand side which must have 4 quadwords
 //	@note	words are host endian while array is little endian
 //	@mayalias
+	.ftrace1
 Mul4x4Adx:
+	.ftrace2
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	sub	$56,%rsp
 	mov	%r15,-8(%rbp)
 	mov	%r14,-16(%rbp)

libc/nexgen32e/mul6x6adx.S

@@ -32,10 +32,11 @@
 //	@param	rdx is right hand side which must have 4 quadwords
 //	@note	words are host endian while array is little endian
 //	@mayalias
+	.ftrace1
 Mul6x6Adx:
+	.ftrace2
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	sub	$64,%rsp
 	mov	%r15,-8(%rbp)
 	mov	%r14,-16(%rbp)

libc/nexgen32e/mul8x8adx.S

@@ -32,10 +32,11 @@
 //	@param	rdx is right hand side which must have 8 quadwords
 //	@note	words are host endian while array is little endian
 //	@mayalias
+	.ftrace1
 Mul8x8Adx:
+	.ftrace2
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	sub	$104,%rsp
 	mov	%r15,-8(%rbp)
 	mov	%r14,-16(%rbp)

libc/nexgen32e/nexgen32e.mk

@@ -56,7 +56,7 @@ o/$(MODE)/libc/nexgen32e/ktoupper.o			\
 o/$(MODE)/libc/nexgen32e/pid.o				\
 o/$(MODE)/libc/nexgen32e/program_invocation_name2.o	\
 o/$(MODE)/libc/nexgen32e/threaded.o: private		\
-		OVERRIDE_CFLAGS +=			\
+		CFLAGS +=				\
 			$(NO_MAGIC)
 
 # these assembly files are safe to build on aarch64

libc/nexgen32e/nt2sysv.S

@@ -29,12 +29,13 @@
 //	@return	%rax,%xmm0
 //	@note	slower than __sysv2nt
 //	@see	NT2SYSV() macro
+	.ftrace1
 __nt2sysv:
+	.ftrace2
 	push	%rbp
 	mov	%rsp,%rbp
 //	TODO(jart): We should probably find some way to use our own
 //	            stack when Windows delivers signals ;_;
-	.profilable
 	sub	$0x100,%rsp
 	push	%rbx
 	push	%rdi

libc/nexgen32e/rldecode.S

@@ -24,9 +24,10 @@
 //	@param	di points to output buffer
 //	@param	si points to uint8_t {len₁,byte₁}, ..., {0,0}
 //	@mode	long,legacy,real
+	.ftrace1
 rldecode:
+	.ftrace2
 	.leafprologue
-	.profilable
 	xor	%ecx,%ecx
 0:	lodsb
 	xchg	%al,%cl

libc/nexgen32e/setjmp.S

@@ -26,7 +26,8 @@
 //	@assume	system five nexgen32e abi conformant
 //	@note	code built w/ microsoft abi compiler can't call this
 //	@see	longjmp(), _gclongjmp()
-setjmp:
+	.ftrace1
+setjmp:	.ftrace2
 #ifdef __x86_64__
 	lea	8(%rsp),%rax
 	mov	%rax,(%rdi)

libc/nexgen32e/sha1.S

@@ -613,10 +613,11 @@ BSWAP_SHUFB_CTL:
 //	@param	%rsi points to input data
 //	@param	%rdx is number of 64-byte blocks to process
 //	@see	X86_HAVE(SHA)
+	.ftrace1
 sha1_transform_avx2:
+	.ftrace2
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	push	%rbx
 	push	%r12
 	push	%r13

libc/nexgen32e/sha1ni.S

@@ -77,10 +77,11 @@ Tim Chen <tim.c.chen@linux.intel.com>\n"
 //	@param	%rsi points to input data
 //	@param	%rdx is number of 64-byte blocks to process
 //	@see	X86_HAVE(SHA)
+	.ftrace1
 sha1_transform_ni:
+	.ftrace2
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	sub	$FRAME_SIZE,%rsp
 	shl	$6,NUM_BLKS		# convert to bytes
 	jz	.Ldone_hash

libc/nexgen32e/sha256.S

@@ -530,10 +530,11 @@ STACK_SIZE	= _RSP      + _RSP_SIZE
 ########################################################################
 .text
 .balign 32
+	.ftrace1
 sha256_transform_rorx:
+	.ftrace2
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	pushq	%rbx
 	pushq	%r12
 	pushq	%r13

libc/nexgen32e/sha256ni.S

@@ -80,9 +80,10 @@ Tim Chen <tim.c.chen@linux.intel.com>\n"
 //	@param	%rsi points to input data
 //	@param	%rdx is number of blocks to process
 //	@see	X86_HAVE(SHA)
+	.ftrace1
 sha256_transform_ni:
+	.ftrace2
 	.leafprologue
-	.profilable
 	shl	$6,NUM_BLKS			# convert to bytes
 	jz	.Ldone_hash
 	add	DATA_PTR,NUM_BLKS		# pointer to end of data

libc/nexgen32e/sha512.S

@@ -571,10 +571,11 @@ frame_size = frame_GPRSAVE + GPRSAVE_SIZE
 # of SHA512 message blocks.
 # "blocks" is the message length in SHA512 blocks
 ########################################################################
+	.ftrace1
 sha512_transform_rorx:
+	.ftrace2
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	# Allocate Stack Space
 	mov	%rsp, %rax
 	sub	$frame_size, %rsp

libc/nexgen32e/tinywcslen.greg.S

@@ -23,9 +23,10 @@
 //	@param	RDI is wchar_t *s
 //	@param	EAX is unsigned length
 //	@see	libc/nexgen32e/strsak32.S
+	.ftrace1
 tinywcslen:
+	.ftrace2
 	.leafprologue
-	.profilable
 	xor	%eax,%eax
 1:	cmpl	$0,(%rdi,%rax,4)
 	jz	2f

libc/nexgen32e/tinywcsnlen.greg.S

@@ -24,9 +24,10 @@
 //	@param	RSI is size_t n
 //	@param	EAX is unsigned length
 //	@see	libc/nexgen32e/strsak32.S
+	.ftrace1
 tinywcsnlen:
+	.ftrace2
 	.leafprologue
-	.profilable
 	xor	%eax,%eax
 1:	cmp	%esi,%eax
 	jae	2f

libc/nt/API-MS-Win-Core-Synch-l1-2-0/WaitOnAddress.S

@@ -2,11 +2,12 @@
 .imp	API-MS-Win-Core-Synch-l1-2-0,__imp_WaitOnAddress,WaitOnAddress,111
 
 	.text.windows
+        .ftrace1
 WaitOnAddress:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_WaitOnAddress(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/API-MS-Win-Core-Synch-l1-2-0/WakeByAddressAll.S

@@ -2,11 +2,12 @@
 .imp	API-MS-Win-Core-Synch-l1-2-0,__imp_WakeByAddressAll,WakeByAddressAll,113
 
 	.text.windows
+        .ftrace1
 WakeByAddressAll:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	%rdi,%rcx
 	sub	$32,%rsp
 	call	*__imp_WakeByAddressAll(%rip)

libc/nt/API-MS-Win-Core-Synch-l1-2-0/WakeByAddressSingle.S

@@ -2,11 +2,12 @@
 .imp	API-MS-Win-Core-Synch-l1-2-0,__imp_WakeByAddressSingle,WakeByAddressSingle,116
 
 	.text.windows
+        .ftrace1
 WakeByAddressSingle:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	%rdi,%rcx
 	sub	$32,%rsp
 	call	*__imp_WakeByAddressSingle(%rip)

libc/nt/MsWSock/AcceptEx.S

@@ -2,11 +2,12 @@
 .imp	MsWSock,__imp_AcceptEx,AcceptEx,0
 
 	.text.windows
+        .ftrace1
 AcceptEx:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_AcceptEx(%rip),%rax
 	jmp	__sysv2nt8
 #elif defined(__aarch64__)

libc/nt/MsWSock/DisconnectEx.S

@@ -2,11 +2,12 @@
 .imp	MsWSock,__imp_DisconnectEx,DisconnectEx,0
 
 	.text.windows
+        .ftrace1
 DisconnectEx:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_DisconnectEx(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/MsWSock/GetAcceptExSockaddrs.S

@@ -2,11 +2,12 @@
 .imp	MsWSock,__imp_GetAcceptExSockaddrs,GetAcceptExSockaddrs,0
 
 	.text.windows
+        .ftrace1
 GetAcceptExSockaddrs:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_GetAcceptExSockaddrs(%rip),%rax
 	jmp	__sysv2nt8
 #elif defined(__aarch64__)

libc/nt/MsWSock/TransmitFile.S

@@ -2,11 +2,12 @@
 .imp	MsWSock,__imp_TransmitFile,TransmitFile,0
 
 	.text.windows
+        .ftrace1
 TransmitFile:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_TransmitFile(%rip),%rax
 	jmp	__sysv2nt8
 #elif defined(__aarch64__)

libc/nt/MsWSock/WSARecvEx.S

@@ -2,11 +2,12 @@
 .imp	MsWSock,__imp_WSARecvEx,WSARecvEx,0
 
 	.text.windows
+        .ftrace1
 WSARecvEx:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_WSARecvEx(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/PowrProf/SetSuspendState.S

@@ -2,11 +2,12 @@
 .imp	PowrProf,__imp_SetSuspendState,SetSuspendState,0
 
 	.text.windows
+        .ftrace1
 SetSuspendState:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_SetSuspendState(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/advapi32/AccessCheck.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_AccessCheck,AccessCheck,0
 
 	.text.windows
+        .ftrace1
 AccessCheck:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_AccessCheck(%rip),%rax
 	jmp	__sysv2nt8
 #elif defined(__aarch64__)

libc/nt/advapi32/AdjustTokenPrivileges.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_AdjustTokenPrivileges,AdjustTokenPrivileges,0
 
 	.text.windows
+        .ftrace1
 AdjustTokenPrivileges:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_AdjustTokenPrivileges(%rip),%rax
 	jmp	__sysv2nt6
 #elif defined(__aarch64__)

libc/nt/advapi32/CreateProcessAsUserW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_CreateProcessAsUserW,CreateProcessAsUserW,0
 
 	.text.windows
+        .ftrace1
 CreateProcessAsUser:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_CreateProcessAsUserW(%rip),%rax
 	jmp	__sysv2nt12
 #elif defined(__aarch64__)

libc/nt/advapi32/DeregisterEventSource.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_DeregisterEventSource,DeregisterEventSource,1239
 
 	.text.windows
+        .ftrace1
 DeregisterEventSource:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	%rdi,%rcx
 	sub	$32,%rsp
 	call	*__imp_DeregisterEventSource(%rip)

libc/nt/advapi32/DuplicateToken.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_DuplicateToken,DuplicateToken,0
 
 	.text.windows
+        .ftrace1
 DuplicateToken:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_DuplicateToken(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/advapi32/DuplicateTokenEx.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_DuplicateTokenEx,DuplicateTokenEx,0
 
 	.text.windows
+        .ftrace1
 DuplicateTokenEx:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_DuplicateTokenEx(%rip),%rax
 	jmp	__sysv2nt6
 #elif defined(__aarch64__)

libc/nt/advapi32/GetFileSecurityW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_GetFileSecurityW,GetFileSecurityW,0
 
 	.text.windows
+        .ftrace1
 GetFileSecurity:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_GetFileSecurityW(%rip),%rax
 	jmp	__sysv2nt6
 #elif defined(__aarch64__)

libc/nt/advapi32/GetUserNameW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_GetUserNameW,GetUserNameW,1381
 
 	.text.windows
+        .ftrace1
 GetUserName:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_GetUserNameW(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/advapi32/ImpersonateSelf.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_ImpersonateSelf,ImpersonateSelf,0
 
 	.text.windows
+        .ftrace1
 ImpersonateSelf:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	%rdi,%rcx
 	sub	$32,%rsp
 	call	*__imp_ImpersonateSelf(%rip)

libc/nt/advapi32/InitiateShutdownW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_InitiateShutdownW,InitiateShutdownW,1403
 
 	.text.windows
+        .ftrace1
 InitiateShutdown:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_InitiateShutdownW(%rip),%rax
 	jmp	__sysv2nt6
 #elif defined(__aarch64__)

libc/nt/advapi32/LookupPrivilegeValueW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_LookupPrivilegeValueW,LookupPrivilegeValueW,1432
 
 	.text.windows
+        .ftrace1
 LookupPrivilegeValue:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_LookupPrivilegeValueW(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/advapi32/MapGenericMask.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_MapGenericMask,MapGenericMask,0
 
 	.text.windows
+        .ftrace1
 MapGenericMask:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_MapGenericMask(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/advapi32/OpenProcessToken.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_OpenProcessToken,OpenProcessToken,0
 
 	.text.windows
+        .ftrace1
 OpenProcessToken:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_OpenProcessToken(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/advapi32/OpenThreadToken.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_OpenThreadToken,OpenThreadToken,0
 
 	.text.windows
+        .ftrace1
 OpenThreadToken:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_OpenThreadToken(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/advapi32/RegCloseKey.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegCloseKey,RegCloseKey,0
 
 	.text.windows
+        .ftrace1
 RegCloseKey:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	%rdi,%rcx
 	sub	$32,%rsp
 	call	*__imp_RegCloseKey(%rip)

libc/nt/advapi32/RegConnectRegistryW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegConnectRegistryW,RegConnectRegistryW,1608
 
 	.text.windows
+        .ftrace1
 RegConnectRegistry:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_RegConnectRegistryW(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/advapi32/RegCreateKeyExW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegCreateKeyExW,RegCreateKeyExW,0
 
 	.text.windows
+        .ftrace1
 RegCreateKeyEx:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_RegCreateKeyExW(%rip),%rax
 	jmp	__sysv2nt10
 #elif defined(__aarch64__)

libc/nt/advapi32/RegCreateKeyW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegCreateKeyW,RegCreateKeyW,1616
 
 	.text.windows
+        .ftrace1
 RegCreateKey:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_RegCreateKeyW(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/advapi32/RegDeleteKeyExW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegDeleteKeyExW,RegDeleteKeyExW,0
 
 	.text.windows
+        .ftrace1
 RegDeleteKeyEx:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_RegDeleteKeyExW(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/advapi32/RegDeleteKeyW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegDeleteKeyW,RegDeleteKeyW,1624
 
 	.text.windows
+        .ftrace1
 RegDeleteKey:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_RegDeleteKeyW(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/advapi32/RegDeleteTreeW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegDeleteTreeW,RegDeleteTreeW,0
 
 	.text.windows
+        .ftrace1
 RegDeleteTree:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_RegDeleteTreeW(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/advapi32/RegDeleteValueW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegDeleteValueW,RegDeleteValueW,0
 
 	.text.windows
+        .ftrace1
 RegDeleteValue:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_RegDeleteValueW(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/advapi32/RegDisablePredefinedCache.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegDisablePredefinedCache,RegDisablePredefinedCache,1629
 
 	.text.windows
+        .ftrace1
 RegDisablePredefinedCache:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	%rdi,%rcx
 	sub	$32,%rsp
 	call	*__imp_RegDisablePredefinedCache(%rip)

libc/nt/advapi32/RegDisableReflectionKey.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegDisableReflectionKey,RegDisableReflectionKey,1631
 
 	.text.windows
+        .ftrace1
 RegDisableReflectionKey:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	%rdi,%rcx
 	sub	$32,%rsp
 	call	*__imp_RegDisableReflectionKey(%rip)

libc/nt/advapi32/RegEnableReflectionKey.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegEnableReflectionKey,RegEnableReflectionKey,1632
 
 	.text.windows
+        .ftrace1
 RegEnableReflectionKey:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	%rdi,%rcx
 	sub	$32,%rsp
 	call	*__imp_RegEnableReflectionKey(%rip)

libc/nt/advapi32/RegEnumKeyExW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegEnumKeyExW,RegEnumKeyExW,0
 
 	.text.windows
+        .ftrace1
 RegEnumKeyEx:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_RegEnumKeyExW(%rip),%rax
 	jmp	__sysv2nt8
 #elif defined(__aarch64__)

libc/nt/advapi32/RegEnumKeyW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegEnumKeyW,RegEnumKeyW,1636
 
 	.text.windows
+        .ftrace1
 RegEnumKey:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_RegEnumKeyW(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/advapi32/RegEnumValueW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegEnumValueW,RegEnumValueW,0
 
 	.text.windows
+        .ftrace1
 RegEnumValue:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_RegEnumValueW(%rip),%rax
 	jmp	__sysv2nt8
 #elif defined(__aarch64__)

libc/nt/advapi32/RegFlushKey.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegFlushKey,RegFlushKey,0
 
 	.text.windows
+        .ftrace1
 RegFlushKey:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	%rdi,%rcx
 	sub	$32,%rsp
 	call	*__imp_RegFlushKey(%rip)

libc/nt/advapi32/RegGetKeySecurity.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegGetKeySecurity,RegGetKeySecurity,0
 
 	.text.windows
+        .ftrace1
 RegGetKeySecurity:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_RegGetKeySecurity(%rip),%rax
 	jmp	__sysv2nt
 #elif defined(__aarch64__)

libc/nt/advapi32/RegGetValueW.S

@@ -2,11 +2,12 @@
 .imp	advapi32,__imp_RegGetValueW,RegGetValueW,0
 
 	.text.windows
+        .ftrace1
 RegGetValue:
+        .ftrace2
 #ifdef __x86_64__
 	push	%rbp
 	mov	%rsp,%rbp
-	.profilable
 	mov	__imp_RegGetValueW(%rip),%rax
 	jmp	__sysv2nt8
 #elif defined(__aarch64__)