diff --git a/src/libextra/test.rs b/src/libextra/test.rs
index c01c619b0c285..5cf63d958544b 100644
--- a/src/libextra/test.rs
+++ b/src/libextra/test.rs
@@ -739,7 +739,7 @@ fn get_concurrency() -> uint {
             }
         }
         None => {
-            rt::util::default_sched_threads()
+            rt::default_sched_threads()
         }
     }
 }
diff --git a/src/libstd/cleanup.rs b/src/libstd/cleanup.rs
index 1c9944664eee6..38925621f1200 100644
--- a/src/libstd/cleanup.rs
+++ b/src/libstd/cleanup.rs
@@ -123,10 +123,10 @@ pub unsafe fn annihilate() {
 
     if debug_mem() {
         // We do logging here w/o allocation.
-        rterrln!("annihilator stats:\n  \
-                     total boxes: {}\n  \
-                    unique boxes: {}\n  \
-                     bytes freed: {}",
-                 stats.n_total_boxes, stats.n_unique_boxes, stats.n_bytes_freed);
+        debug2!("annihilator stats:\n  \
+                       total boxes: {}\n  \
+                      unique boxes: {}\n  \
+                       bytes freed: {}",
+                stats.n_total_boxes, stats.n_unique_boxes, stats.n_bytes_freed);
     }
 }
diff --git a/src/libstd/os.rs b/src/libstd/os.rs
index 4de7f32248db0..e7cba0895e451 100644
--- a/src/libstd/os.rs
+++ b/src/libstd/os.rs
@@ -1185,7 +1185,7 @@ pub fn last_os_error() -> ~str {
  */
 pub fn set_exit_status(code: int) {
     use rt;
-    rt::util::set_exit_status(code);
+    rt::set_exit_status(code);
 }
 
 unsafe fn load_argc_and_argv(argc: c_int, argv: **c_char) -> ~[~str] {
diff --git a/src/libstd/rt/borrowck.rs b/src/libstd/rt/borrowck.rs
index 6be23a983abb2..3c2000c522c4d 100644
--- a/src/libstd/rt/borrowck.rs
+++ b/src/libstd/rt/borrowck.rs
@@ -15,10 +15,10 @@ use option::{Option, None, Some};
 use ptr::RawPtr;
 use rt::env;
 use rt::local::Local;
+use rt::task;
 use rt::task::Task;
 use str::{OwnedStr, StrSlice};
 use str;
-use sys;
 use uint;
 use unstable::raw;
 use vec::ImmutableVector;
@@ -64,7 +64,7 @@ unsafe fn fail_borrowed(box: *mut raw::Box<()>, file: *c_char, line: size_t) {
         None => { // not recording borrows
             let msg = "borrowed";
             do msg.with_c_str |msg_p| {
-                sys::begin_unwind_(msg_p, file, line);
+                task::begin_unwind(msg_p, file, line);
             }
         }
         Some(borrow_list) => { // recording borrows
@@ -80,7 +80,7 @@ unsafe fn fail_borrowed(box: *mut raw::Box<()>, file: *c_char, line: size_t) {
                 }
             }
             do msg.with_c_str |msg_p| {
-                sys::begin_unwind_(msg_p, file, line)
+                task::begin_unwind(msg_p, file, line)
             }
         }
     }
@@ -179,7 +179,7 @@ pub unsafe fn unrecord_borrow(a: *u8, old_ref_count: uint,
             if br.box != a || br.file != file || br.line != line {
                 let err = format!("wrong borrow found, br={:?}", br);
                 do err.with_c_str |msg_p| {
-                    sys::begin_unwind_(msg_p, file, line)
+                    task::begin_unwind(msg_p, file, line)
                 }
             }
             borrow_list
diff --git a/src/libstd/rt/local_heap.rs b/src/libstd/rt/local_heap.rs
index f863ccacaf5c9..262da9f3b8e5e 100644
--- a/src/libstd/rt/local_heap.rs
+++ b/src/libstd/rt/local_heap.rs
@@ -86,6 +86,17 @@ impl Drop for LocalHeap {
     }
 }
 
+pub unsafe fn local_malloc(td: *libc::c_char, size: libc::uintptr_t) -> *libc::c_char {
+    // XXX: Unsafe borrow for speed. Lame.
+    let task: Option<*mut Task> = Local::try_unsafe_borrow();
+    match task {
+        Some(task) => {
+            (*task).heap.alloc(td as *libc::c_void, size as uint) as *libc::c_char
+        }
+        None => rtabort!("local malloc outside of task")
+    }
+}
+
 // A little compatibility function
 pub unsafe fn local_free(ptr: *libc::c_char) {
     // XXX: Unsafe borrow for speed. Lame.
diff --git a/src/libstd/rt/local_ptr.rs b/src/libstd/rt/local_ptr.rs
index 33384594c575f..c9534413c53e4 100644
--- a/src/libstd/rt/local_ptr.rs
+++ b/src/libstd/rt/local_ptr.rs
@@ -178,5 +178,5 @@ pub fn maybe_tls_key() -> Option<tls::Key> {
 #[inline]
 #[cfg(test)]
 pub fn maybe_tls_key() -> Option<tls::Key> {
-    unsafe { ::cast::transmute(::realstd::rt::local_ptr::maybe_tls_key()) }
+    unsafe { ::cast::transmute(::realstd::rt::shouldnt_be_public::maybe_tls_key()) }
 }
diff --git a/src/libstd/macros.rs b/src/libstd/rt/macros.rs
similarity index 84%
rename from src/libstd/macros.rs
rename to src/libstd/rt/macros.rs
index 17dc03d0098a8..c6ff3427c15f6 100644
--- a/src/libstd/macros.rs
+++ b/src/libstd/rt/macros.rs
@@ -8,8 +8,12 @@
 // option. This file may not be copied, modified, or distributed
 // except according to those terms.
 
+//! Macros used by the runtime.
+//!
+//! These macros call functions which are only accessible in the `rt` module, so
+//! they aren't defined anywhere outside of the `rt` module.
+
 #[macro_escape];
-#[doc(hidden)];
 
 macro_rules! rterrln (
     ($($arg:tt)*) => ( {
@@ -37,7 +41,7 @@ macro_rules! rtassert (
 )
 
 
-macro_rules! rtabort(
+macro_rules! rtabort (
     ($($msg:tt)*) => ( {
         ::rt::util::abort(format!($($msg)*));
     } )
diff --git a/src/libstd/rt/mod.rs b/src/libstd/rt/mod.rs
index 2db973b602ee8..78f0bb0a07ca6 100644
--- a/src/libstd/rt/mod.rs
+++ b/src/libstd/rt/mod.rs
@@ -76,6 +76,16 @@ use vec::{OwnedVector, MutableVector, ImmutableVector};
 use self::thread::Thread;
 use self::work_queue::WorkQueue;
 
+// the os module needs to reach into this helper, so allow general access
+// through this reexport.
+pub use self::util::set_exit_status;
+
+// this is somewhat useful when a program wants to spawn a "reasonable" number
+// of workers based on the constraints of the system that it's running on.
+// Perhaps this shouldn't be a `pub use` though and there should be another
+// method...
+pub use self::util::default_sched_threads;
+
 // XXX: these probably shouldn't be public...
 #[doc(hidden)]
 pub mod shouldnt_be_public {
@@ -86,8 +96,12 @@ pub mod shouldnt_be_public {
     pub use super::select::SelectInner;
     pub use super::rtio::EventLoop;
     pub use super::select::{SelectInner, SelectPortInner};
+    pub use super::local_ptr::maybe_tls_key;
 }
 
+// Internal macros used by the runtime.
+mod macros;
+
 /// The global (exchange) heap.
 pub mod global_heap;
 
@@ -158,17 +172,14 @@ pub mod comm;
 
 mod select;
 
-// FIXME #5248 shouldn't be pub
 /// The runtime needs to be able to put a pointer into thread-local storage.
-pub mod local_ptr;
+mod local_ptr;
 
-// FIXME #5248: The import in `sched` doesn't resolve unless this is pub!
 /// Bindings to pthread/windows thread-local storage.
-pub mod thread_local_storage;
+mod thread_local_storage;
 
-// FIXME #5248 shouldn't be pub
 /// Just stuff
-pub mod util;
+mod util;
 
 // Global command line argument storage
 pub mod args;
diff --git a/src/libstd/rt/task.rs b/src/libstd/rt/task.rs
index 71ab3b571c4c6..d5278975d8dfe 100644
--- a/src/libstd/rt/task.rs
+++ b/src/libstd/rt/task.rs
@@ -17,7 +17,7 @@ use borrow;
 use cast::transmute;
 use cleanup;
 use local_data;
-use libc::{c_void, uintptr_t};
+use libc::{c_void, uintptr_t, c_char, size_t};
 use prelude::*;
 use option::{Option, Some, None};
 use rt::borrowck;
@@ -465,6 +465,48 @@ impl Unwinder {
     }
 }
 
+/// This is the entry point of unwinding for things like lang items and such.
+/// The arguments are normally generated by the compiler.
+pub fn begin_unwind(msg: *c_char, file: *c_char, line: size_t) -> ! {
+    use rt::in_green_task_context;
+    use rt::task::Task;
+    use rt::local::Local;
+    use rt::logging::Logger;
+    use str::Str;
+    use c_str::CString;
+
+    unsafe {
+        let msg = CString::new(msg, false);
+        let file = CString::new(file, false);
+        let msg = match msg.as_str() {
+            Some(s) => s, None => rtabort!("message wasn't utf8?")
+        };
+        let file = match file.as_str() {
+            Some(s) => s, None => rtabort!("message wasn't utf8?")
+        };
+
+        if in_green_task_context() {
+            // Be careful not to allocate in this block, if we're failing we may
+            // have been failing due to a lack of memory in the first place...
+            do Local::borrow |task: &mut Task| {
+                let n = task.name.as_ref().map(|n| n.as_slice()).unwrap_or("<unnamed>");
+                format_args!(|args| { task.logger.log(args) },
+                             "task '{}' failed at '{}', {}:{}",
+                             n, msg, file, line);
+            }
+        } else {
+            rterrln!("failed in non-task context at '{}', {}:{}",
+                     msg, file, line as int);
+        }
+
+        let task: *mut Task = Local::unsafe_borrow();
+        if (*task).unwinder.unwinding {
+            rtabort!("unwinding again");
+        }
+        (*task).unwinder.begin_unwind();
+    }
+}
+
 #[cfg(test)]
 mod test {
     use rt::test::*;
diff --git a/src/libstd/rt/util.rs b/src/libstd/rt/util.rs
index 727bdb782d2c8..647d88c26f2f5 100644
--- a/src/libstd/rt/util.rs
+++ b/src/libstd/rt/util.rs
@@ -16,8 +16,6 @@ use option::{Some, None, Option};
 use os;
 use str::StrSlice;
 use unstable::atomics::{AtomicInt, INIT_ATOMIC_INT, SeqCst};
-
-#[cfg(target_os="macos")]
 use unstable::running_on_valgrind;
 
 // Indicates whether we should perform expensive sanity checks, including rtassert!
@@ -37,21 +35,17 @@ pub fn num_cpus() -> uint {
     }
 }
 
-/// Valgrind has a fixed-sized array (size around 2000) of segment descriptors wired into it; this
-/// is a hard limit and requires rebuilding valgrind if you want to go beyond it. Normally this is
-/// not a problem, but in some tests, we produce a lot of threads casually. Making lots of threads
-/// alone might not be a problem _either_, except on OSX, the segments produced for new threads
-/// _take a while_ to get reclaimed by the OS. Combined with the fact that libuv schedulers fork off
-/// a separate thread for polling fsevents on OSX, we get a perfect storm of creating "too many
-/// mappings" for valgrind to handle when running certain stress tests in the runtime.
-#[cfg(target_os="macos")]
-pub fn limit_thread_creation_due_to_osx_and_valgrind() -> bool {
-    running_on_valgrind()
-}
-
-#[cfg(not(target_os="macos"))]
+/// Valgrind has a fixed-sized array (size around 2000) of segment descriptors
+/// wired into it; this is a hard limit and requires rebuilding valgrind if you
+/// want to go beyond it. Normally this is not a problem, but in some tests, we
+/// produce a lot of threads casually.  Making lots of threads alone might not
+/// be a problem _either_, except on OSX, the segments produced for new threads
+/// _take a while_ to get reclaimed by the OS. Combined with the fact that libuv
+/// schedulers fork off a separate thread for polling fsevents on OSX, we get a
+/// perfect storm of creating "too many mappings" for valgrind to handle when
+/// running certain stress tests in the runtime.
 pub fn limit_thread_creation_due_to_osx_and_valgrind() -> bool {
-    false
+    (cfg!(target_os="macos")) && running_on_valgrind()
 }
 
 /// Get's the number of scheduler threads requested by the environment
diff --git a/src/libstd/std.rs b/src/libstd/std.rs
index 53837f965938a..5a3f40f280fff 100644
--- a/src/libstd/std.rs
+++ b/src/libstd/std.rs
@@ -88,9 +88,6 @@ pub mod linkhack {
     }
 }
 
-// Internal macros
-mod macros;
-
 /* The Prelude. */
 
 pub mod prelude;
diff --git a/src/libstd/sys.rs b/src/libstd/sys.rs
index ce1be592d79c8..10c3fed1d54db 100644
--- a/src/libstd/sys.rs
+++ b/src/libstd/sys.rs
@@ -14,9 +14,10 @@
 
 use c_str::ToCStr;
 use cast;
+use libc::size_t;
 use libc;
-use libc::{c_char, size_t};
 use repr;
+use rt::task;
 use str;
 use unstable::intrinsics;
 
@@ -109,7 +110,7 @@ impl FailWithCause for ~str {
     fn fail_with(cause: ~str, file: &'static str, line: uint) -> ! {
         do cause.with_c_str |msg_buf| {
             do file.with_c_str |file_buf| {
-                begin_unwind_(msg_buf, file_buf, line as libc::size_t)
+                task::begin_unwind(msg_buf, file_buf, line as libc::size_t)
             }
         }
     }
@@ -119,47 +120,12 @@ impl FailWithCause for &'static str {
     fn fail_with(cause: &'static str, file: &'static str, line: uint) -> ! {
         do cause.with_c_str |msg_buf| {
             do file.with_c_str |file_buf| {
-                begin_unwind_(msg_buf, file_buf, line as libc::size_t)
+                task::begin_unwind(msg_buf, file_buf, line as libc::size_t)
             }
         }
     }
 }
 
-// FIXME #4427: Temporary until rt::rt_fail_ goes away
-pub fn begin_unwind_(msg: *c_char, file: *c_char, line: size_t) -> ! {
-    use rt::in_green_task_context;
-    use rt::task::Task;
-    use rt::local::Local;
-    use rt::logging::Logger;
-    use str::Str;
-
-    unsafe {
-        // XXX: Bad re-allocations. fail2! needs some refactoring
-        let msg = str::raw::from_c_str(msg);
-        let file = str::raw::from_c_str(file);
-
-        if in_green_task_context() {
-            // Be careful not to allocate in this block, if we're failing we may
-            // have been failing due to a lack of memory in the first place...
-            do Local::borrow |task: &mut Task| {
-                let n = task.name.as_ref().map(|n| n.as_slice()).unwrap_or("<unnamed>");
-                format_args!(|args| { task.logger.log(args) },
-                             "task '{}' failed at '{}', {}:{}",
-                             n, msg.as_slice(), file.as_slice(), line);
-            }
-        } else {
-            rterrln!("failed in non-task context at '{}', {}:{}",
-                     msg, file, line as int);
-        }
-
-        let task: *mut Task = Local::unsafe_borrow();
-        if (*task).unwinder.unwinding {
-            rtabort!("unwinding again");
-        }
-        (*task).unwinder.begin_unwind();
-    }
-}
-
 #[cfg(test)]
 mod tests {
     use cast;
diff --git a/src/libstd/task/spawn.rs b/src/libstd/task/spawn.rs
index d26f8eb10248e..611d2f1fdb67b 100644
--- a/src/libstd/task/spawn.rs
+++ b/src/libstd/task/spawn.rs
@@ -448,7 +448,7 @@ impl RuntimeGlue {
     }
 
     fn with_task_handle_and_failing(blk: &fn(&KillHandle, bool)) {
-        rtassert!(in_green_task_context());
+        assert!(in_green_task_context());
         unsafe {
             // Can't use safe borrow, because the taskgroup destructor needs to
             // access the scheduler again to send kill signals to other tasks.
@@ -458,7 +458,7 @@ impl RuntimeGlue {
     }
 
     fn with_my_taskgroup<U>(blk: &fn(&Taskgroup) -> U) -> U {
-        rtassert!(in_green_task_context());
+        assert!(in_green_task_context());
         unsafe {
             // Can't use safe borrow, because creating new hashmaps for the
             // tasksets requires an rng, which needs to borrow the sched.
@@ -553,7 +553,7 @@ fn enlist_many(child: &KillHandle, child_arc: &TaskGroupArc,
 }
 
 pub fn spawn_raw(mut opts: TaskOpts, f: ~fn()) {
-    rtassert!(in_green_task_context());
+    assert!(in_green_task_context());
 
     let child_data = Cell::new(gen_child_taskgroup(opts.linked, opts.supervised));
     let indestructible = opts.indestructible;
@@ -631,7 +631,7 @@ pub fn spawn_raw(mut opts: TaskOpts, f: ~fn()) {
             let (thread_port, thread_chan) = oneshot();
             let thread_port_cell = Cell::new(thread_port);
             let join_task = do Task::build_child(None) {
-                rtdebug!("running join task");
+                debug2!("running join task");
                 let thread_port = thread_port_cell.take();
                 let thread: Thread = thread_port.recv();
                 thread.join();
@@ -648,11 +648,11 @@ pub fn spawn_raw(mut opts: TaskOpts, f: ~fn()) {
                 let join_task = join_task_cell.take();
 
                 let bootstrap_task = ~do Task::new_root(&mut new_sched.stack_pool, None) || {
-                    rtdebug!("boostrapping a 1:1 scheduler");
+                    debug2!("boostrapping a 1:1 scheduler");
                 };
                 new_sched.bootstrap(bootstrap_task);
 
-                rtdebug!("enqueing join_task");
+                debug2!("enqueing join_task");
                 // Now tell the original scheduler to join with this thread
                 // by scheduling a thread-joining task on the original scheduler
                 orig_sched_handle.send_task_from_friend(join_task);
@@ -684,7 +684,7 @@ pub fn spawn_raw(mut opts: TaskOpts, f: ~fn()) {
     }
 
     task.name = opts.name.take();
-    rtdebug!("spawn calling run_task");
+    debug2!("spawn calling run_task");
     Scheduler::run_task(task);
 
 }
diff --git a/src/libstd/unstable/lang.rs b/src/libstd/unstable/lang.rs
index fca477763c5e6..6f2928e99fed7 100644
--- a/src/libstd/unstable/lang.rs
+++ b/src/libstd/unstable/lang.rs
@@ -12,16 +12,13 @@
 
 use c_str::ToCStr;
 use cast::transmute;
-use libc::{c_char, c_void, size_t, uintptr_t};
-use option::{Option, None, Some};
-use sys;
-use rt::task::Task;
-use rt::local::Local;
+use libc::{c_char, size_t, uintptr_t};
+use rt::task;
 use rt::borrowck;
 
 #[lang="fail_"]
 pub fn fail_(expr: *c_char, file: *c_char, line: size_t) -> ! {
-    sys::begin_unwind_(expr, file, line);
+    task::begin_unwind(expr, file, line);
 }
 
 #[lang="fail_bounds_check"]
@@ -36,14 +33,7 @@ pub fn fail_bounds_check(file: *c_char, line: size_t,
 
 #[lang="malloc"]
 pub unsafe fn local_malloc(td: *c_char, size: uintptr_t) -> *c_char {
-    // XXX: Unsafe borrow for speed. Lame.
-    let task: Option<*mut Task> = Local::try_unsafe_borrow();
-    match task {
-        Some(task) => {
-            (*task).heap.alloc(td as *c_void, size as uint) as *c_char
-        }
-        None => rtabort!("local malloc outside of task")
-    }
+    ::rt::local_heap::local_malloc(td, size)
 }
 
 // NB: Calls to free CANNOT be allowed to fail, as throwing an exception from