Add support for 64-bit Windows builds

mk/platform.mk

@@ -377,7 +377,7 @@ RUSTC_CROSS_FLAGS_arm-unknown-linux-gnueabi :=
 # mipsel-linux configuration
 CC_mipsel-linux=mipsel-linux-gcc
 CXX_mipsel-linux=mipsel-linux-g++
-CPP_mipsel-linux=mipsel-linux-gcc 
+CPP_mipsel-linux=mipsel-linux-gcc
 AR_mipsel-linux=mipsel-linux-ar
 CFG_LIB_NAME_mipsel-linux=lib$(1).so
 CFG_STATIC_LIB_NAME_mipsel-linux=lib$(1).a
@@ -641,7 +641,7 @@ define CFG_MAKE_TOOLCHAIN
 	CXX_$(1)=$(CROSS_PREFIX_$(1))$(CXX_$(1))
 	CPP_$(1)=$(CROSS_PREFIX_$(1))$(CPP_$(1))
 	AR_$(1)=$(CROSS_PREFIX_$(1))$(AR_$(1))
-	RUSTC_CROSS_FLAGS_$(1)=-C linker=$$(call FIND_COMPILER,$$(CXX_$(1))) \
+	RUSTC_CROSS_FLAGS_$(1)=-C linker=$$(call FIND_COMPILER,$$(CC_$(1))) \
 	    -C ar=$$(call FIND_COMPILER,$$(AR_$(1))) $(RUSTC_CROSS_FLAGS_$(1))
 
 	RUSTC_FLAGS_$(1)=$$(RUSTC_CROSS_FLAGS_$(1)) $(RUSTC_FLAGS_$(1))

src/etc/get-snapshot.py

@@ -53,7 +53,7 @@ def unpack_snapshot(triple, dl_path):
   dl_path = sys.argv[2]
 else:
   # There are no 64-bit Windows snapshots yet, so we'll use 32-bit ones instead, for now
-  snap_triple = triple if triple != "x86_64-w64-mingw32" else "i686-pc-mingw32"
+  snap_triple = triple if triple != "x86_64-w64-mingw32" else "i686-w64-mingw32"
   snap = determine_curr_snapshot(snap_triple)
   dl = os.path.join(download_dir_base, snap)
   url = download_url_base + "/" + snap

src/libgreen/context.rs

@@ -13,6 +13,8 @@ use std::uint;
 use std::mem::transmute;
 use std::rt::stack;
 use std::raw;
+#[cfg(target_arch = "x86_64")]
+use std::simd;
 
 // FIXME #7761: Registers is boxed so that it is 16-byte aligned, for storing
 // SSE regs.  It would be marginally better not to do this. In C++ we
@@ -103,11 +105,11 @@ impl Context {
             // invalid for the current task. Lucky for us `rust_swap_registers`
             // is a C function so we don't have to worry about that!
             match in_context.stack_bounds {
-                Some((lo, hi)) => stack::record_stack_bounds(lo, hi),
+                Some((lo, hi)) => stack::record_rust_managed_stack_bounds(lo, hi),
                 // If we're going back to one of the original contexts or
                 // something that's possibly not a "normal task", then reset
                 // the stack limit to 0 to make morestack never fail
-                None => stack::record_stack_bounds(0, uint::MAX),
+                None => stack::record_rust_managed_stack_bounds(0, uint::MAX),
             }
             rust_swap_registers(out_regs, in_regs)
         }
@@ -186,14 +188,30 @@ fn initialize_call_frame(regs: &mut Registers, fptr: InitFn, arg: uint,
 // windows requires saving more registers (both general and XMM), so the windows
 // register context must be larger.
 #[cfg(windows, target_arch = "x86_64")]
-type Registers = [uint, ..34];
+struct Registers {
+    gpr:[uint, ..14],
+    _xmm:[simd::u32x4, ..10]
+}
 #[cfg(not(windows), target_arch = "x86_64")]
-type Registers = [uint, ..22];
+struct Registers {
+    gpr:[uint, ..10],
+    _xmm:[simd::u32x4, ..6]
+}
 
 #[cfg(windows, target_arch = "x86_64")]
-fn new_regs() -> Box<Registers> { box() ([0, .. 34]) }
+fn new_regs() -> Box<Registers> {
+    box() Registers {
+        gpr:[0,..14],
+        _xmm:[simd::u32x4(0,0,0,0),..10]
+    }
+}
 #[cfg(not(windows), target_arch = "x86_64")]
-fn new_regs() -> Box<Registers> { box() ([0, .. 22]) }
+fn new_regs() -> Box<Registers> {
+    box() Registers {
+        gpr:[0,..10],
+        _xmm:[simd::u32x4(0,0,0,0),..6]
+    }
+}
 
 #[cfg(target_arch = "x86_64")]
 fn initialize_call_frame(regs: &mut Registers, fptr: InitFn, arg: uint,
@@ -222,20 +240,20 @@ fn initialize_call_frame(regs: &mut Registers, fptr: InitFn, arg: uint,
 
     // These registers are frobbed by rust_bootstrap_green_task into the right
     // location so we can invoke the "real init function", `fptr`.
-    regs[RUSTRT_R12] = arg as uint;
-    regs[RUSTRT_R13] = procedure.code as uint;
-    regs[RUSTRT_R14] = procedure.env as uint;
-    regs[RUSTRT_R15] = fptr as uint;
+    regs.gpr[RUSTRT_R12] = arg as uint;
+    regs.gpr[RUSTRT_R13] = procedure.code as uint;
+    regs.gpr[RUSTRT_R14] = procedure.env as uint;
+    regs.gpr[RUSTRT_R15] = fptr as uint;
 
     // These registers are picked up by the regular context switch paths. These
     // will put us in "mostly the right context" except for frobbing all the
     // arguments to the right place. We have the small trampoline code inside of
     // rust_bootstrap_green_task to do that.
-    regs[RUSTRT_RSP] = sp as uint;
-    regs[RUSTRT_IP] = rust_bootstrap_green_task as uint;
+    regs.gpr[RUSTRT_RSP] = sp as uint;
+    regs.gpr[RUSTRT_IP] = rust_bootstrap_green_task as uint;
 
     // Last base pointer on the stack should be 0
-    regs[RUSTRT_RBP] = 0;
+    regs.gpr[RUSTRT_RBP] = 0;
 }
 
 #[cfg(target_arch = "arm")]

src/libnative/lib.rs

@@ -137,7 +137,7 @@ pub fn start(argc: int, argv: *const *const u8, main: proc()) -> int {
     task.name = Some(str::Slice("<main>"));
     drop(task.run(|| {
         unsafe {
-            rt::stack::record_stack_bounds(my_stack_bottom, my_stack_top);
+            rt::stack::record_os_managed_stack_bounds(my_stack_bottom, my_stack_top);
         }
         exit_code = Some(run(main.take_unwrap()));
     }).destroy());

src/libnative/task.rs

@@ -84,7 +84,7 @@ pub fn spawn_opts(opts: TaskOpts, f: proc():Send) {
         let addr = &something_around_the_top_of_the_stack as *const int;
         let my_stack = addr as uint;
         unsafe {
-            stack::record_stack_bounds(my_stack - stack + 1024, my_stack);
+            stack::record_os_managed_stack_bounds(my_stack - stack + 1024, my_stack);
         }
         let mut ops = ops;
         ops.stack_bounds = (my_stack - stack + 1024, my_stack);

src/librustrt/libunwind.rs

@@ -60,7 +60,7 @@ pub type _Unwind_Word = libc::uintptr_t;
 pub static unwinder_private_data_size: uint = 5;
 
 #[cfg(target_arch = "x86_64")]
-pub static unwinder_private_data_size: uint = 2;
+pub static unwinder_private_data_size: uint = 6;
 
 #[cfg(target_arch = "arm", not(target_os = "ios"))]
 pub static unwinder_private_data_size: uint = 20;

src/librustrt/stack.rs

@@ -124,8 +124,23 @@ extern fn stack_exhausted() {
     }
 }
 
+// Windows maintains a record of upper and lower stack bounds in the Thread Information
+// Block (TIB), and some syscalls do check that addresses which are supposed to be in
+// the stack, indeed lie between these two values.
+// (See https://github.com/rust-lang/rust/issues/3445#issuecomment-26114839)
+//
+// When using Rust-managed stacks (libgreen), we must maintain these values accordingly.
+// For OS-managed stacks (libnative), we let the OS manage them for us.
+//
+// On all other platforms both variants behave identically.
+
 #[inline(always)]
-pub unsafe fn record_stack_bounds(stack_lo: uint, stack_hi: uint) {
+pub unsafe fn record_os_managed_stack_bounds(stack_lo: uint, _stack_hi: uint) {
+    record_sp_limit(stack_lo + RED_ZONE);
+}
+
+#[inline(always)]
+pub unsafe fn record_rust_managed_stack_bounds(stack_lo: uint, stack_hi: uint) {
     // When the old runtime had segmented stacks, it used a calculation that was
     // "limit + RED_ZONE + FUDGE". The red zone was for things like dynamic
     // symbol resolution, llvm function calls, etc. In theory this red zone
@@ -138,16 +153,17 @@ pub unsafe fn record_stack_bounds(stack_lo: uint, stack_hi: uint) {
 
     return target_record_stack_bounds(stack_lo, stack_hi);
 
-    #[cfg(not(windows))] #[cfg(not(target_arch = "x86_64"))] #[inline(always)]
+    #[cfg(not(windows))] #[inline(always)]
     unsafe fn target_record_stack_bounds(_stack_lo: uint, _stack_hi: uint) {}
+
+    #[cfg(windows, target_arch = "x86")] #[inline(always)]
+    unsafe fn target_record_stack_bounds(stack_lo: uint, stack_hi: uint) {
+        // stack range is at TIB: %fs:0x04 (top) and %fs:0x08 (bottom)
+        asm!("mov $0, %fs:0x04" :: "r"(stack_hi) :: "volatile");
+        asm!("mov $0, %fs:0x08" :: "r"(stack_lo) :: "volatile");
+    }
     #[cfg(windows, target_arch = "x86_64")] #[inline(always)]
     unsafe fn target_record_stack_bounds(stack_lo: uint, stack_hi: uint) {
-        // Windows compiles C functions which may check the stack bounds. This
-        // means that if we want to perform valid FFI on windows, then we need
-        // to ensure that the stack bounds are what they truly are for this
-        // task. More info can be found at:
-        //   https://github.com/rust-lang/rust/issues/3445#issuecomment-26114839
-        //
         // stack range is at TIB: %gs:0x08 (top) and %gs:0x10 (bottom)
         asm!("mov $0, %gs:0x08" :: "r"(stack_hi) :: "volatile");
         asm!("mov $0, %gs:0x10" :: "r"(stack_lo) :: "volatile");

src/librustrt/thread.rs

@@ -44,7 +44,7 @@ static DEFAULT_STACK_SIZE: uint = 1024 * 1024;
 #[no_split_stack]
 extern fn thread_start(main: *mut libc::c_void) -> imp::rust_thread_return {
     unsafe {
-        stack::record_stack_bounds(0, uint::MAX);
+        stack::record_os_managed_stack_bounds(0, uint::MAX);
         let f: Box<proc()> = mem::transmute(main);
         (*f)();
         mem::transmute(0 as imp::rust_thread_return)