Using the redzone means any interrupt can clobber local variables, so disable
it.  This isn't a full solution since the payload is not built this way,
but it's better than nothing for now.

* 'master' of github.com:cloudius-systems/osv:
  libc: implement putenv()
  libc: implement getpwuid()
  libc: implement uname()
  pthread: increase default stack size
  pthread: fix pthread_self()
  main: switch payload loading to be in a pthread

Pretend we're another OS to avoid confusing the payload.

Java stomps all over a 64k stack.

The pthread_self() implementation was inconsistent with pthread_create(),
leading to corruption.  Switch to a simpler implementation using
thread-local storage.

This way, the payload may use the pthread APIs in the thread it started in.

* 'master' of github.com:cloudius-systems/osv: (30 commits)
  elf: relocate R_X86_64_64 correctly
  sched: make thread stack be controlled by the thread creator
  main: don't destroy DriverFactory for now
  bootfs: disable read01.so temporarily
  mmu: don't map NULL
  libc: fix confstr(NULL, ...)
  elf: fix segfault on unmatched symbol
  mmu: move physical memory map to ffff800000000000
  implement positional I/O (pread/pwrite/preadv/pwritev)
  implement vectored I/O (readv/writev)
  uio-ify the read/write path
  clean up the vnode operation defintion a bit
  import basic uio support code from FreeBSD
  add a real strerror() implementation from musl
  add misc routines from musl
  more stat variants
  add getuid/getgid stubs
  implement lxstat
  implement a real console layer
  simplify the console abstractions
  ...

host side reset calls

Fixes generated code going nuts over incorrect JVM_NanoTime.

Needed for pthread_attr_*() (and in general to make sure the stack is
a regular vma as expected by the payload).

segfaults.

breaks the build

Keep the first 2MB unmapped for null pointer refernences.

confstr(NULL) is defined to return the size of required storage; use a
temporary string to make it not fault.

lookup_symbol() checks if the symbol is undefined, before checking if we
actually looked it up.  Check for a missing symbol first.

This allows us to clear the area around the null pointer, and so trap errors.

The switch is tricky wrt setting up memory, since we need the memory to
be mapped in order to free it, but we need free memory to map (for the page
tables).  What we do is set up a temporart 1:1 map at ffff800000000000
that mirrors the first 1G map we already have from bootstreap, use that
to free all of memory under 1G, then map and free everything else.

This will allow for easy addition of the vectored and positional I/O
syscalls, as well as making the structure of the I/O path more similar
to *BSD and Solaris.

until we support symlinks lxstat is identical to stat

Create a central place the managed the console driver (just isa-serial for
now), and exposes them as a device file and to the debug output routines.

This is the low-level console abstraction and should deal in the
char * hardware format.  The upper layer (debug.cc) already handles
all conversion from std::string and boost formats.

Destroy the drivers on OS reset (currently just call it from the loader directly) It fixes the issue of gdb breakage w/ the drivers

Initial memory is physical; the mmu converts it to virtual addresses, and
then it can be added to the memory pool.  Right now there is not much
difference, but the 1:1 mapping is moving soon.