diff --git a/src/libtools/signal32.c b/src/libtools/signal32.c
index 064fd3a0f..7d92f91f1 100644
--- a/src/libtools/signal32.c
+++ b/src/libtools/signal32.c
@@ -863,8 +863,8 @@ EXPORT int my32_getcontext(x64emu_t* emu, void* ucp)
     u->uc_mcontext.gregs[I386_CS] = R_CS;
     u->uc_mcontext.gregs[I386_SS] = R_SS;
     // get FloatPoint status
-    if(u->uc_mcontext.fpregs)
-        save_fpreg(emu, from_ptrv(u->uc_mcontext.fpregs));
+    u->uc_mcontext.fpregs = to_ptrv(ucp + 236);    // magic offset of fpregs in an actual i386 u_context
+    fpu_savenv(emu, from_ptrv(u->uc_mcontext.fpregs), 1);   // it seems getcontext only save fpu env, not fpu regs
     // get signal mask
     sigprocmask(SIG_SETMASK, NULL, (sigset_t*)&u->uc_sigmask);
     // ensure uc_link is properly initialized
@@ -898,8 +898,7 @@ EXPORT int my32_setcontext(x64emu_t* emu, void* ucp)
     R_CS = u->uc_mcontext.gregs[I386_CS];
     R_SS = u->uc_mcontext.gregs[I386_SS];
     // set FloatPoint status
-    if(u->uc_mcontext.fpregs)
-        load_fpreg(emu, from_ptrv(u->uc_mcontext.fpregs));
+    fpu_loadenv(emu, from_ptrv(u->uc_mcontext.fpregs), 1);
     // set signal mask
     sigprocmask(SIG_SETMASK, (sigset_t*)&u->uc_sigmask, NULL);
     // set uc_link
diff --git a/src/libtools/signals.c b/src/libtools/signals.c
index a4602bc6a..6bebf6d23 100644
--- a/src/libtools/signals.c
+++ b/src/libtools/signals.c
@@ -2243,8 +2243,10 @@ EXPORT int my_getcontext(x64emu_t* emu, void* ucp)
     // get segments
     u->uc_mcontext.gregs[X64_CSGSFS] = ((uint64_t)(R_CS)) | (((uint64_t)(R_GS))<<16) | (((uint64_t)(R_FS))<<32);
     // get FloatPoint status
-    u->uc_mcontext.fpregs = &u->xstate;
-    fpu_fxsave64(emu, &u->xstate);
+    u->uc_mcontext.fpregs = ucp + 408;
+    fpu_savenv(emu, (void*)u->uc_mcontext.fpregs, 1);
+    *(uint32_t*)(ucp + 432) = emu->mxcsr.x32;
+
     // get signal mask
     sigprocmask(SIG_SETMASK, NULL, (sigset_t*)&u->uc_sigmask);
     // ensure uc_link is properly initialized
@@ -2283,7 +2285,8 @@ EXPORT int my_setcontext(x64emu_t* emu, void* ucp)
     R_GS = (u->uc_mcontext.gregs[X64_CSGSFS]>>16)&0xffff;
     R_FS = (u->uc_mcontext.gregs[X64_CSGSFS]>>32)&0xffff;
     // set FloatPoint status
-    fpu_fxrstor64(emu, &u->xstate);
+    fpu_loadenv(emu, (void*)u->uc_mcontext.fpregs, 1);
+    emu->mxcsr.x32 = *(uint32_t*)(ucp + 432);
     // set signal mask
     sigprocmask(SIG_SETMASK, (sigset_t*)&u->uc_sigmask, NULL);
     // set uc_link