Overhaul memory allocation model

This commit makes several fundamental changes to memory handling:

- the frame allocator is now only an allocator for free frames, and does
  not track used frames.
- the frame allocator now stores its free list inside the free frames
  themselves, as a hybrid stack/span model.
  - This has the implication that all frames must currently fit within
    the offset area.
- kutil has a new allocator interface, which is the only allowed way for
  any code outside of src/kernel to allocate. Code under src/kernel
  _may_ use new/delete, but should prefer the allocator interface.
- the heap manager has become heap_allocator, which is merely an
  implementation of kutil::allocator which doles out sections of a given
  address range.
- the heap manager now only writes block headers when necessary,
  avoiding page faults until they're actually needed
- page_manager now has a page fault handler, which checks with the
  address_manager to see if the address is known, and provides a frame
  mapping if it is, allowing heap manager to work with its entire
  address size from the start. (Currently 32GiB.)

.gitignore

@@ -2,7 +2,6 @@
 /build*
 *.bak
 tags
-.gdbinit
 popcorn.log
 *.o
 *.a

LICENSE.md

@@ -81,6 +81,33 @@ MIT license:
 > IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 > CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
+## printf
+
+Popcorn uses 's tiny [printf](https://github.com/mpaland/printf) library for its
+`*printf()` functions, which is also released under the terms of the MIT license:
+
+> The MIT License (MIT)
+>
+> Copyright (c) 2014 Marco Paland
+>
+> Permission is hereby granted, free of charge, to any person obtaining a copy
+> of this software and associated documentation files (the "Software"), to deal
+> in the Software without restriction, including without limitation the rights
+> to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+> copies of the Software, and to permit persons to whom the Software is
+> furnished to do so, subject to the following conditions:
+>
+> The above copyright notice and this permission notice shall be included in all
+> copies or substantial portions of the Software.
+>
+> THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+> IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+> FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+> AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+> LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+> OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+> SOFTWARE.
+
 ## GNU-EFI
 
 Popcorn's UEFI bootloader initially used

README.md

@@ -27,11 +27,21 @@ The design goals of the project are:
 
 ## Building
 
-Popcorn uses the `ninja` build tool, and generates the build files for it with
-the `generate_build.py` script.  The other requirements are:
+Popcorn uses the [Ninja][] build tool, and generates the build files for it
+with a custom tool called [Bonnibel][]. Bonnibel requires [Python 3][] and can
+be downloaded with `pip`:
 
-* python 3 for generating the build config
-  * The Jinja2 package is also required
+```
+pip3 install bonnibel
+```
+
+[Ninja]:    https://ninja-build.org
+[Bonnibel]: https://github.com/justinian/bonnibel
+[Python 3]: https://python.org
+
+Requrirements:
+
+* python 3 (for installing and running Bonnibel)
 * clang
 * mtools
 * ninja
@@ -45,7 +55,7 @@ Popcorn host binaries.
 
 ### Building and running Popcorn
 
-Once the toolchain has been set up, running `generate_build.py` will set up the
+Once the toolchain has been set up, running Bonnibel's `pb` command will set up the
 build configuration, and `ninja -C build` will actually run the build.  If you
 have `qemu-system-x86_64` installed, the `qemu.sh` script will to run Popcorn
 in QEMU `-nographic` mode.
@@ -54,7 +64,8 @@ I personally run this either from a real debian amd64 testing/buster machine or
 a windows WSL debian testing/buster installation. The following should be
 enough to set up such a system to build the kernel:
 
-    sudo apt install qemu-system-x86 nasm clang-6.0 mtools
-	sudo update-alternatives /usr/bin/clang clang /usr/bin/clang-6.0 1000
-	sudo update-alternatives /usr/bin/clang++ clang++ /usr/bin/clang++-6.0 1000
+    sudo apt install qemu-system-x86 nasm clang-6.0 mtools python3-pip curl
+    sudo update-alternatives /usr/bin/clang clang /usr/bin/clang-6.0 1000
+    sudo update-alternatives /usr/bin/clang++ clang++ /usr/bin/clang++-6.0 1000
+    sudo pip3 install bonnibel
 

assets/debugging/popcorn.elf-gdb.py

@@ -0,0 +1,31 @@
+import gdb
+
+class PrintStackCommand(gdb.Command):
+    def __init__(self):
+        super().__init__("popc_stack", gdb.COMMAND_DATA)
+
+    def invoke(self, arg, from_tty):
+        args = gdb.string_to_argv(arg)
+
+        base = "$rsp"
+        if len(args) > 0:
+            base = args[0]
+
+        depth = 22
+        if len(args) > 1:
+            depth = int(args[1])
+
+        for i in range(depth-1, -1, -1):
+            offset = i * 8
+            base_addr = gdb.parse_and_eval(base)
+            value = gdb.parse_and_eval(f"*(uint64_t*)({base} + {offset})")
+            print("{:016x} (+{:04x}): {:016x}".format(int(base_addr) + offset, offset, int(value)))
+
+
+PrintStackCommand()
+
+import time
+time.sleep(3.5)
+gdb.execute("target remote :1234")
+gdb.execute("set waiting = false")
+gdb.execute("display/i $rip")

assets/initrd.toml

@@ -14,10 +14,10 @@ source = "../assets/fonts/tamsyn8x16r.psf"
 
 [[files]]
 dest = "nulldrv1"
-source = "host/nulldrv"
+source = "user/nulldrv"
 executable = true
 
 [[files]]
 dest = "nulldrv2"
-source = "host/nulldrv"
+source = "user/nulldrv"
 executable = true

generate_build.py

@@ -1,199 +0,0 @@
-#!/usr/bin/env python3
-
-from collections import namedtuple
-
-library = namedtuple('library', ['path', 'deps'])
-program = namedtuple('program', ['path', 'deps', 'output', 'targets'])
-source = namedtuple('source', ['name', 'input', 'output', 'action'])
-version = namedtuple('version', ['major', 'minor', 'patch', 'sha', 'dirty'])
-
-MODULES = {}
-
-class Source:
-    Actions = {'.c': 'cc', '.cpp': 'cxx', '.s': 'nasm'}
-
-    def __init__(self, path, root, modroot):
-        from os.path import relpath, splitext
-        self.input = path
-        self.name = relpath(path, root)
-        self.output = relpath(path, modroot) + ".o"
-        self.action = self.Actions.get(splitext(path)[1], None)
-
-    def __str__(self):
-        return "{} {}:{}:{}".format(self.action, self.output, self.name, self.input)
-
-
-class Module:
-    def __init__(self, name, output, root, **kwargs):
-        from os.path import commonpath, dirname, isdir, join
-
-        self.name = name
-        self.output = output
-        self.kind = kwargs.get("kind", "exe")
-        self.target = kwargs.get("target", None)
-        self.deps = kwargs.get("deps", tuple())
-        self.includes = kwargs.get("includes", tuple())
-        self.defines = kwargs.get("defines", tuple())
-        self.depmods = []
-
-        sources = [join(root, f) for f in kwargs.get("source", tuple())]
-        modroot = commonpath(sources)
-        while not isdir(modroot):
-            modroot = dirname(modroot)
-
-        self.sources = [Source(f, root, modroot) for f in sources]
-
-    def __str__(self):
-        return "Module {} {}\n\t".format(self.kind, self.name)
-
-    def __find_depmods(self, modules):
-        self.depmods = set()
-        open_list = set(self.deps)
-        closed_list = set()
-
-        while open_list:
-            dep = modules[open_list.pop()]
-            open_list |= (set(dep.deps) - closed_list)
-            self.depmods.add(dep)
-
-        self.libdeps = [d for d in self.depmods if d.kind == "lib"]
-        self.exedeps = [d for d in self.depmods if d.kind != "lib"]
-
-    @classmethod
-    def load(cls, filename):
-        from os.path import abspath, dirname
-        from yaml import load
-
-        root = dirname(filename)
-        modules = {}
-        moddata = load(open(filename, "r"))
-
-        for name, data in moddata.items():
-            modules[name] = cls(name, root=root, **data)
-
-        for mod in modules.values():
-            mod.__find_depmods(modules)
-
-        targets = {}
-        for mod in modules.values():
-            if mod.target is None: continue
-            if mod.target not in targets:
-                targets[mod.target] = set()
-            targets[mod.target].add(mod)
-            targets[mod.target] |= mod.depmods
-
-        return modules.values(), targets
-
-
-def get_template(env, typename, name):
-    from jinja2.exceptions import TemplateNotFound
-    try:
-        return env.get_template("{}.{}.j2".format(typename, name))
-    except TemplateNotFound:
-        return env.get_template("{}.default.j2".format(typename))
-
-
-def get_git_version():
-    from subprocess import run
-    cp = run(['git', 'describe', '--dirty'],
-            check=True, capture_output=True)
-    full_version = cp.stdout.decode('utf-8').strip()
-
-    cp = run(['git', 'rev-parse', 'HEAD'],
-            check=True, capture_output=True)
-    full_sha = cp.stdout.decode('utf-8').strip()
-
-    dirty = False
-    parts1 = full_version.split('-')
-    if parts1[-1] == "dirty":
-        dirty = True
-        parts1 = parts1[:-1]
-
-    parts2 = parts1[0].split('.')
-
-    return version(
-        parts2[0],
-        parts2[1],
-        parts2[2],
-        full_sha[:7],
-        dirty)
-
-
-def main(buildroot="build", modulefile="modules.yaml"):
-    import os
-    from os.path import abspath, dirname, isabs, isdir, join
-
-    generator = abspath(__file__)
-    srcroot = dirname(generator)
-    if not isabs(modulefile):
-        modulefile = join(srcroot, modulefile)
-
-    if not isabs(buildroot):
-        buildroot = join(srcroot, buildroot)
-
-    if not isdir(buildroot):
-        os.mkdir(buildroot)
-
-    git_version = get_git_version()
-    print("Generating build files for Popcorn {}.{}.{}-{}...".format(
-        git_version.major, git_version.minor, git_version.patch, git_version.sha))
-
-    from jinja2 import Environment, FileSystemLoader
-    template_dir = join(srcroot, "scripts", "templates")
-    env = Environment(loader=FileSystemLoader(template_dir))
-
-    buildfiles = []
-    templates = set()
-    modules, targets = Module.load(modulefile)
-
-    for mod in modules:
-        buildfile = join(buildroot, mod.name + ".ninja")
-        buildfiles.append(buildfile)
-        with open(buildfile, 'w') as out:
-            template = get_template(env, mod.kind, mod.name)
-            templates.add(template.filename)
-            out.write(template.render(
-                module=mod,
-                buildfile=buildfile,
-                version=git_version))
-
-    for target, mods in targets.items():
-        root = join(buildroot, target)
-        if not isdir(root):
-            os.mkdir(root)
-
-        buildfile = join(root, "target.ninja")
-        buildfiles.append(buildfile)
-        with open(buildfile, 'w') as out:
-            template = get_template(env, "target", target)
-            templates.add(template.filename)
-            out.write(template.render(
-                target=target,
-                modules=mods,
-                buildfile=buildfile,
-                version=git_version))
-
-    # Top level buildfile cannot use an absolute path or ninja won't
-    # reload itself properly on changes.
-    # See: https://github.com/ninja-build/ninja/issues/1240
-    buildfile = "build.ninja"
-    buildfiles.append(buildfile)
-
-    with open(join(buildroot, buildfile), 'w') as out:
-        template = env.get_template("build.ninja.j2")
-        templates.add(template.filename)
-
-        out.write(template.render(
-            targets=targets,
-            buildroot=buildroot,
-            srcroot=srcroot,
-            buildfile=buildfile,
-            buildfiles=buildfiles,
-            templates=[abspath(f) for f in templates],
-            generator=generator,
-            modulefile=modulefile,
-            version=git_version))
-
-if __name__ == "__main__":
-    import sys
-    main(*sys.argv[1:])

modules.yaml

@@ -1,116 +1,132 @@
-kernel:
-    output: popcorn.elf
-    target: host
-    deps:
-        - elf
-        - initrd
-        - kutil
-    includes:
-        - src/kernel
-    source:
-        - src/kernel/crti.s
-        - src/kernel/apic.cpp
-        - src/kernel/assert.cpp
-        - src/kernel/boot.s
-        - src/kernel/console.cpp
-        - src/kernel/cpprt.cpp
-        - src/kernel/cpu.cpp
-        - src/kernel/debug.cpp
-        - src/kernel/debug.s
-        - src/kernel/device_manager.cpp
-        - src/kernel/font.cpp
-        - src/kernel/fs/gpt.cpp
-        - src/kernel/gdt.cpp
-        - src/kernel/gdt.s
-        - src/kernel/interrupts.cpp
-        - src/kernel/interrupts.s
-        - src/kernel/io.cpp
-        - src/kernel/loader.s
-        - src/kernel/log.cpp
-        - src/kernel/main.cpp
-        - src/kernel/memory_bootstrap.cpp
-        - src/kernel/msr.cpp
-        - src/kernel/page_manager.cpp
-        - src/kernel/pci.cpp
-        - src/kernel/process.cpp
-        - src/kernel/scheduler.cpp
-        - src/kernel/screen.cpp
-        - src/kernel/serial.cpp
-        - src/kernel/syscall.cpp
-        - src/kernel/syscall.s
-        - src/kernel/crtn.s
+name: Popcorn
+templates: scripts/templates
+modules:
+    kernel:
+        output: popcorn.elf
+        target: host
+        deps:
+            - elf
+            - initrd
+            - kutil
+        includes:
+            - src/kernel
+        source:
+            - src/kernel/crti.s
+            - src/kernel/apic.cpp
+            - src/kernel/assert.cpp
+            - src/kernel/boot.s
+            - src/kernel/console.cpp
+            - src/kernel/cpprt.cpp
+            - src/kernel/cpu.cpp
+            - src/kernel/debug.cpp
+            - src/kernel/debug.s
+            - src/kernel/device_manager.cpp
+            - src/kernel/font.cpp
+            - src/kernel/frame_allocator.cpp
+            - src/kernel/fs/gpt.cpp
+            - src/kernel/gdt.cpp
+            - src/kernel/gdt.s
+            - src/kernel/interrupts.cpp
+            - src/kernel/interrupts.s
+            - src/kernel/io.cpp
+            - src/kernel/loader.s
+            - src/kernel/log.cpp
+            - src/kernel/main.cpp
+            - src/kernel/memory_bootstrap.cpp
+            - src/kernel/msr.cpp
+            - src/kernel/page_manager.cpp
+            - src/kernel/pci.cpp
+            - src/kernel/process.cpp
+            - src/kernel/scheduler.cpp
+            - src/kernel/screen.cpp
+            - src/kernel/serial.cpp
+            - src/kernel/syscall.cpp
+            - src/kernel/syscall.s
+            - src/kernel/syscalls/exit.cpp
+            - src/kernel/syscalls/fork.cpp
+            - src/kernel/syscalls/getpid.cpp
+            - src/kernel/syscalls/message.cpp
+            - src/kernel/syscalls/noop.cpp
+            - src/kernel/syscalls/pause.cpp
+            - src/kernel/syscalls/sleep.cpp
+            - src/kernel/task.s
+            - src/kernel/crtn.s
 
-boot:
-    kind: exe
-    target: boot
-    output: boot.elf
-    source:
-        - src/boot/crt0.s
-        - src/boot/console.cpp
-        - src/boot/guids.cpp
-        - src/boot/loader.cpp
-        - src/boot/main.cpp
-        - src/boot/memory.cpp
-        - src/boot/reloc.cpp
-        - src/boot/utility.cpp
+    boot:
+        kind: exe
+        target: boot
+        output: boot.elf
+        source:
+            - src/boot/crt0.s
+            - src/boot/console.cpp
+            - src/boot/guids.cpp
+            - src/boot/loader.cpp
+            - src/boot/main.cpp
+            - src/boot/memory.cpp
+            - src/boot/reloc.cpp
+            - src/boot/utility.cpp
 
-nulldrv:
-    kind: exe
-    target: host
-    output: nulldrv
-    source:
-        - src/drivers/nulldrv/main.s
+    nulldrv:
+        kind: exe
+        target: user
+        output: nulldrv
+        source:
+            - src/drivers/nulldrv/main.cpp
+            - src/drivers/nulldrv/main.s
 
-elf:
-    kind: lib
-    output: libelf.a
-    deps:
-        - kutil
-    includes:
-        - src/libraries/elf/include
-    source:
-        - src/libraries/elf/elf.cpp
+    elf:
+        kind: lib
+        output: libelf.a
+        deps:
+            - kutil
+        includes:
+            - src/libraries/elf/include
+        source:
+            - src/libraries/elf/elf.cpp
 
-initrd:
-    kind: lib
-    output: libinitrd.a
-    deps:
-        - kutil
-    includes:
-        - src/libraries/initrd/include
-    source:
-        - src/libraries/initrd/initrd.cpp
+    initrd:
+        kind: lib
+        output: libinitrd.a
+        deps:
+            - kutil
+        includes:
+            - src/libraries/initrd/include
+        source:
+            - src/libraries/initrd/initrd.cpp
 
-kutil:
-    kind: lib
-    output: libkutil.a
-    includes:
-        - src/libraries/kutil/include
-    source:
-        - src/libraries/kutil/assert.cpp
-        - src/libraries/kutil/frame_allocator.cpp
-        - src/libraries/kutil/heap_manager.cpp
-        - src/libraries/kutil/memory.cpp
+    kutil:
+        kind: lib
+        output: libkutil.a
+        includes:
+            - src/libraries/kutil/include
+        source:
+            - src/libraries/kutil/assert.cpp
+            - src/libraries/kutil/bip_buffer.cpp
+            - src/libraries/kutil/heap_allocator.cpp
+            - src/libraries/kutil/logger.cpp
+            - src/libraries/kutil/memory.cpp
+            - src/libraries/kutil/printf.c
 
-makerd:
-    kind: exe
-    target: native
-    output: makerd
-    deps:
-        - initrd
-    source:
-        - src/tools/makerd/entry.cpp
-        - src/tools/makerd/main.cpp
+    makerd:
+        kind: exe
+        target: native
+        output: makerd
+        deps:
+            - initrd
+        source:
+            - src/tools/makerd/entry.cpp
+            - src/tools/makerd/main.cpp
 
-tests:
-    kind: exe
-    target: native
-    output: tests
-    deps:
-        - kutil
-    source:
-        - src/tests/address_manager.cpp
-        - src/tests/frame_allocator.cpp
-        - src/tests/linked_list.cpp
-        - src/tests/heap_manager.cpp
-        - src/tests/main.cpp
+    tests:
+        kind: exe
+        target: native
+        output: tests
+        deps:
+            - kutil
+        source:
+            - src/tests/address_manager.cpp
+            - src/tests/constexpr_hash.cpp
+            - src/tests/linked_list.cpp
+            - src/tests/logger.cpp
+            - src/tests/heap_allocator.cpp
+            - src/tests/main.cpp

qemu.sh

@@ -1,33 +1,52 @@
 #!/usr/bin/env bash
 
 build="$(dirname $0)/build"
+assets="$(dirname $0)/assets"
 debug=""
+flash_name="ovmf_vars"
 gfx="-nographic"
+kvm=""
 
 for arg in $@; do
 	case "${arg}" in
 		--debug)
 			debug="-s"
+			flash_name="ovmf_vars_d"
 			;;
 		--gfx)
 			gfx=""
 			;;
+		--kvm)
+			kvm="-enable-kvm"
+			;;
 		*)
 			build="${arg}"
 			;;
 	esac
 done
 
-kvm=""
-if [[ -c /dev/kvm ]]; then
-	kvm="-enable-kvm"
+if [[ ! -c /dev/kvm ]]; then
+	kvm=""
+fi
+
+if ! ninja -C "${build}"; then
+	exit 1
+fi
+
+if [[ -n $TMUX ]]; then
+	if [[ -n $debug ]]; then
+		tmux split-window "gdb ${build}/popcorn.elf" &
+	else
+		tmux split-window -l 10 "sleep 1; telnet localhost 45454" &
+	fi
 fi
 
-ninja -C "${build}" && \
 exec qemu-system-x86_64 \
-	-drive "if=pflash,format=raw,file=${build}/flash.img" \
+	-drive "if=pflash,format=raw,readonly,file=${assets}/ovmf/x64/ovmf_code.fd" \
+	-drive "if=pflash,format=raw,file=${build}/${flash_name}.fd" \
 	-drive "format=raw,file=${build}/popcorn.img" \
-	-smp 1 \
+	-monitor telnet:localhost:45454,server,nowait \
+	-smp 4 \
 	-m 512 \
 	-d mmu,int,guest_errors \
 	-D popcorn.log \

scripts/build_sysroot_clang.sh

@@ -3,6 +3,7 @@
 TARGET="x86_64-elf"
 NASM_VERSION="2.13.03"
 BINUTILS_VERSION="2.31.1"
+LLVM_BRANCH="release_80"
 
 TOOLS="clang" # lld libunwind libcxxabi libcxx"
 PROJECTS="compiler-rt libcxxabi libcxx libunwind"
@@ -77,7 +78,7 @@ function build_llvm() {
 	if [[ ! -d "${WORK}/llvm" ]]; then
 		echo "Downloading LLVM..."
 		git clone -q \
-			--branch release_70 \
+			--branch "${LLVM_BRANCH}" \
 			--depth 1 \
 			"https://git.llvm.org/git/llvm.git" "${WORK}/llvm"
 	fi
@@ -86,7 +87,7 @@ function build_llvm() {
 		if [[ ! -d "${WORK}/llvm/tools/${tool}" ]]; then
 			echo "Downloading ${tool}..."
 			git clone -q \
-				--branch release_70 \
+				--branch "${LLVM_BRANCH}" \
 				--depth 1 \
 				"https://git.llvm.org/git/${tool}.git" "${WORK}/llvm/tools/${tool}"
 		fi
@@ -95,7 +96,7 @@ function build_llvm() {
 	if [[ ! -d "${WORK}/llvm/tools/clang/tools/extra" ]]; then
 		echo "Downloading clang-tools-extra..."
 		git clone -q \
-			--branch release_70 \
+			--branch "${LLVM_BRANCH}" \
 			--depth 1 \
 			"https://git.llvm.org/git/clang-tools-extra.git" "${WORK}/llvm/tools/clang/tools/extra"
 	fi
@@ -104,7 +105,7 @@ function build_llvm() {
 		if [[ ! -d "${WORK}/llvm/projects/${proj}" ]]; then
 			echo "Downloading ${proj}..."
 			git clone -q \
-				--branch release_70 \
+				--branch "${LLVM_BRANCH}" \
 				--depth 1 \
 				"https://git.llvm.org/git/${proj}.git" "${WORK}/llvm/projects/${proj}"
 		fi
@@ -114,7 +115,7 @@ function build_llvm() {
 		if [[ ! -d "${WORK}/llvm/runtimes/${proj}" ]]; then
 			echo "Downloading ${proj}..."
 			git clone -q \
-				--branch release_70 \
+				--branch "${LLVM_BRANCH}" \
 				--depth 1 \
 				"https://git.llvm.org/git/${proj}.git" "${WORK}/llvm/runtime/${proj}"
 		fi

scripts/templates/build.ninja.j2

@@ -1,109 +1,14 @@
-ninja_required_version = 1.3
-builddir = {{ buildroot }}
-srcroot = {{ srcroot }}
-modulefile = {{ modulefile }}
+{% extends "build.base.j2" %}
 
-warnflags = $
-    -Wformat=2 $
-    -Winit-self $
-    -Wfloat-equal $
-    -Winline $
-    -Wmissing-format-attribute $
-    -Wmissing-include-dirs $
-    -Wswitch $
-    -Wundef $
-    -Wdisabled-optimization $
-    -Wpointer-arith $
-    -Wno-attributes $
-    -Wno-sign-compare $
-    -Wno-multichar $
-    -Wno-div-by-zero $
-    -Wno-endif-labels $
-    -Wno-pragmas $
-    -Wno-format-extra-args $
-    -Wno-unused-result $
-    -Wno-deprecated-declarations $
-    -Wno-unused-function $
-    -Werror
-
-ccflags = $
+{% block variables %}
+{{ super() }}
+ccflags = $ccflags $
     -I${srcroot}/src/include $
-    -I${srcroot}/src/include/x86_64 $
-    -DVERSION_MAJOR={{ version.major }} $
-    -DVERSION_MINOR={{ version.minor }} $
-    -DVERSION_PATCH={{ version.patch }} $
-    -DVERSION_GITSHA=0x{% if version.dirty %}1{% else %}0{% endif %}{{ version.sha }} $
-    -DGIT_VERSION=\"{{ version.major }}.{{ version.minor }}.{{ version.patch }}-{{ version.sha }}\" $
-    -DGIT_VERSION_WIDE=L\"{{ version.major }}.{{ version.minor }}.{{ version.patch }}-{{ version.sha }}\" $
-    $warnflags
-
-asflags = $
-    -DVERSION_MAJOR={{ version.major }} $
-    -DVERSION_MINOR={{ version.minor }} $
-    -DVERSION_PATCH={{ version.patch }} $
-    -DVERSION_GITSHA=0x{% if version.dirty %}1{% else %}0{% endif %}{{ version.sha }}
-
-cflags = -std=c11
-cxxflags = -std=c++14
-libs =
-
-rule cc
-    deps = gcc
-    depfile = $out.d
-    description = Compiling $name
-    command = $cc -MMD -MF $out.d $ccflags $cflags -o $out -c $in
-
-rule dump_cc_defs
-    description = Dumping CC defines for $target
-    command = echo "" | $cc $ccflags $cflags -dM -E - > $out
-
-rule dump_cc_run
-    description = Dumping CC arguments for $target
-    command = $
-        echo "#!/bin/bash" > $out; $
-        echo '$cc $ccflags $cflags $$*' > $out; $
-        chmod a+x $out
-
-rule cxx
-    deps = gcc
-    depfile = $out.d
-    description = Compiling $name
-    command = $cxx -MMD -MF $out.d $cxxflags $ccflags -o $out -c $in
-
-rule dump_cxx_defs
-    description = Dumping C++ defines for $target
-    command = echo "" | $cxx -x c++ $cxxflags $ccflags -dM -E - > $out
-
-rule dump_cxx_run
-    description = Dumping C++ arguments for $target
-    command = $
-        echo "#!/bin/bash" > $out; $
-        echo '$cc $cxxflags $ccflags $$*' > $out; $
-        chmod a+x $out
-
-rule nasm
-    deps = gcc
-    depfile = $out.d
-    description = Assembling $name
-    command = $nasm -o $out -felf64 -MD $out.d $asflags $in
-
-rule exe
-    description = Linking $name
-    command = $ld $ldflags -o $out $in $libs
-
-rule lib
-    description = Archiving $name
-    command = $ar qcs $out $in
-
-rule regen
-    generator = true
-    description = Regenrating build files
-    command = {{ generator }} $builddir $modulefile
-
-rule cp
-    description = Copying $name
-    command = cp $in $out
+    -I${srcroot}/src/include/x86_64 
+{% endblock %}
 
+{% block baserules %}
+{{ super() }}
 rule makerd
     description = Making init ramdisk
     command = $builddir/native/makerd $in $out
@@ -135,29 +40,24 @@ rule strip
         objcopy --only-keep-debug $out $out.debug; $
         strip -g $out; $
         objcopy --add-gnu-debuglink=$out.debug $out
+{% endblock %}
 
+{% block extra %}
+build $builddir/ovmf_vars.fd : cp $srcroot/assets/ovmf/x64/ovmf_vars.fd
+    name = ovmf_vars.fd
 
-{% for target in targets %}
-subninja {{ target }}/target.ninja
-{% endfor %}
-
-build $
-    {%- for buildfile in buildfiles %}
-    {{ buildfile }} $
-    {%- endfor %}
-    : regen | $
-    {%- for template in templates %}
-    {{ template }} $
-    {%- endfor %}
-    $modulefile $
-    {{ generator }}
-
-build $builddir/flash.img : cp $srcroot/assets/ovmf/x64/OVMF.fd
-    name = flash.img
+build $builddir/ovmf_vars_d.fd : cp $srcroot/assets/ovmf/x64/ovmf_vars_d.fd
+    name = ovmf_vars_d.fd
 
 build $builddir/popcorn.elf | $builddir/popcorn.elf.debug : strip $builddir/host/popcorn.elf
     name = kernel
 
+build $builddir/popcorn.dump : dump $builddir/host/popcorn.elf
+    name = kernel
+
+build $builddir/popcorn.elf-gdb.py : cp ${srcroot}/assets/debugging/popcorn.elf-gdb.py
+    name = kernel debug python scripts
+
 build $builddir/fatroot/popcorn.elf : cp $builddir/popcorn.elf
     name = kernel to FAT image
 
@@ -166,7 +66,7 @@ build $builddir/fatroot/efi/boot/bootx64.efi : cp $builddir/boot/boot.efi
 
 build $builddir/fatroot/initrd.img : makerd ${srcroot}/assets/initrd.toml | $
     ${builddir}/native/makerd $
-    ${builddir}/host/nulldrv
+    ${builddir}/user/nulldrv
 
 build $builddir/popcorn.img : makefat | $
     $builddir/fatroot/initrd.img $
@@ -174,5 +74,13 @@ build $builddir/popcorn.img : makefat | $
     $builddir/fatroot/efi/boot/bootx64.efi
     name = popcorn.img
 
-# vim: et ts=4 sts=4 sw=4
+default $
+    $builddir/ovmf_vars.fd $
+    $builddir/ovmf_vars_d.fd $
+    $builddir/popcorn.dump $
+    $builddir/popcorn.elf-gdb.py $
+    $builddir/popcorn.img
+{% endblock %}
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
 

scripts/templates/exe.boot.j2

@@ -2,18 +2,13 @@
 {% block variables %}
 {{ super() }}
 
-ld = ld
-cc = clang
-cxx = clang++
-
 ccflags = $ccflags $
     -DKERNEL_FILENAME=L\"popcorn.elf\" $
     -DGNU_EFI_USE_MS_ABI $
     -DHAVE_USE_MS_ABI $
     -DEFI_DEBUG=0 $
     -DEFI_DEBUG_CLEAR_MEMORY=0 $
-	-DBOOTLOADER_DEBUG $
-    -fPIC
+    -DBOOTLOADER_DEBUG
 
 ldflags = $ldflags $
     -T ${srcroot}/src/arch/x86_64/boot.ld $
@@ -27,3 +22,6 @@ build $builddir/boot.efi : makeefi ${builddir}/{{ module.output }}
     name = boot.efi
 
 {% endblock %}
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
+

scripts/templates/exe.default.j2

@@ -1,12 +0,0 @@
-{% extends "module.base.j2" %}
-{% block variables %}
-{{ super() }}
-
-libs = $
-    -L${builddir} $
-    {%- for dep in module.libdeps %}
-    -l{{ dep.name }} $
-    {%- endfor %}
-    $libs
-
-{% endblock %}

scripts/templates/exe.kernel.j2

@@ -7,3 +7,6 @@ libs = $libs
 ldflags = $ldflags -T ${srcroot}/src/arch/x86_64/kernel.ld
 
 {% endblock %}
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
+

scripts/templates/exe.tests.j2

@@ -5,3 +5,6 @@
 ccflags = $ccflags -ggdb
 
 {% endblock %}
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
+

scripts/templates/lib.default.j2

@@ -1 +0,0 @@
-{% extends "module.base.j2" %}

scripts/templates/module.base.j2

@@ -1,41 +0,0 @@
-moddir = ${builddir}/{{ module.name }}.dir
-
-{% block variables %}
-ccflags = $ccflags $
-    {%- for dep in module.depmods %}
-    {%- for inc in dep.includes %}
-    -I${srcroot}/{{ inc }} $
-    {%- endfor %}
-    {%- endfor %}
-    {%- for inc in module.includes %}
-    -I${srcroot}/{{ inc }} $
-    {%- endfor %}
-    {%- for define in module.defines %}
-    -D{{ define }} $
-    {%- endfor %}
-{% endblock %}
-
-{% for source in module.sources %}
-build ${moddir}/{{ source.output }} : {{ source.action }} {{ source.input }} || {{ buildfile }}
-    name = {{ source.name }}
-{% endfor %}
-
-build ${builddir}/{{ module.output }} : {{ module.kind }} $
-{%- for source in module.sources %}
-    ${moddir}/{{ source.output }} $
-{%- endfor -%}
-{%- for dep in module.libdeps %}
-    ${builddir}/{{ dep.output }} $
-{%- endfor %}
-    | $
-{% for dep in module.exedeps %}
-    ${builddir}/{{ dep.output }} $
-{%- endfor -%}
-    {{ buildfile }}
-    name = {{ module.name }}
-
-{% block extra %}
-{% endblock %}
-
-# vim: ft=ninja et ts=4 sts=4 sw=4
-

scripts/templates/target.boot.j2

@@ -1,35 +1,41 @@
 {% extends "target.default.j2" %}
 
 {% block binaries %}
-ld = ld
 cc = clang
 cxx = clang++
+ld = ld
+ar = ar
 nasm = nasm
+objcopy = objcopy
 {% endblock %}
 
 {% block variables %}
 
 ccflags = $ccflags $
-    -ggdb $
-    -nostdlib $
-    -ffreestanding $
-    -nodefaultlibs $
-    -fno-builtin $
-    -mno-sse $
-    -fno-omit-frame-pointer $
-    -mno-red-zone $
-    -fshort-wchar
+	-ggdb $
+	-nostdlib $
+	-ffreestanding $
+	-nodefaultlibs $
+	-fno-builtin $
+	-mno-sse $
+	-fno-omit-frame-pointer $
+	-mno-red-zone $
+	-fshort-wchar $
+	-D__ELF__ $
+	-fPIC
 
 cxxflags = $cxxflags $
-    -nostdlibinc $
-    -fno-exceptions $
-    -fno-rtti
+	-fno-exceptions $
+	-fno-rtti $
 
 ldflags = $ldflags $
-    -g $
-    -nostdlib $
-    -znocombreloc $
+	-g $
+	-nostdlib $
+	-znocombreloc $
 	-Bsymbolic $
-    -nostartfiles
+	-nostartfiles
 
 {% endblock %}
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
+

scripts/templates/target.default.j2

@@ -1,24 +0,0 @@
-builddir = $builddir/{{ target }}
-target = {{ target }}
-
-{% block variables %}
-{% endblock %}
-
-{% block binaries %}
-cc = clang
-cxx = clang++
-ld = clang++
-ar = ar
-nasm = nasm
-objcopy = objcopy
-{% endblock %}
-
-{% for module in modules %}
-subninja {{ module.name }}.ninja
-{% endfor %}
-
-build ${builddir}/cc.defs : dump_cc_defs | {{ buildfile }}
-build ${builddir}/cxx.defs : dump_cxx_defs | {{ buildfile }}
-build ${builddir}/cc.run : dump_cc_run | {{ buildfile }}
-build ${builddir}/cxx.run : dump_cxx_run | {{ buildfile }}
-

scripts/templates/target.host.j2

@@ -40,4 +40,6 @@ ldflags = $ldflags $
     -Bstatic
 
 {% endblock %}
-# vim: et ts=4 sts=4 sw=4
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
+

scripts/templates/target.native.j2

@@ -1,8 +1,15 @@
 {% extends "target.default.j2" %}
 
+{% block binaries %}
+{{ super() }}
+ld = clang++
+{% endblock %}
+
 {% block variables %}
 
 ccflags = $ccflags -g -ggdb
 
 {% endblock %}
-# vim: et ts=4 sts=4 sw=4
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
+

scripts/templates/target.user.j2

@@ -0,0 +1,49 @@
+{% extends "target.default.j2" %}
+
+{% block binaries %}
+cc = ${srcroot}/sysroot/bin/clang
+cxx = ${srcroot}/sysroot/bin/clang++
+ld = ${srcroot}/sysroot/bin/x86_64-elf-ld
+ar = ${srcroot}/sysroot/bin/x86_64-elf-ar
+nasm = ${srcroot}/sysroot/bin/nasm
+objcopy = ${srcroot}/sysroot/bin/x86_64-elf-objcopy
+{% endblock %}
+
+{% block variables %}
+
+ccflags = $ccflags $
+    -nostdlib $
+    -nodefaultlibs $
+    -fno-builtin $
+    -mno-sse $
+    -fno-omit-frame-pointer $
+    -mno-red-zone $
+    -g $
+    -mcmodel=large $
+    -D__ELF__ $
+    -D__POPCORN__ $
+    -isystem${srcroot}/sysroot/include $
+    --sysroot="${srcroot}/sysroot"
+
+cxxflags = $cxxflags $
+    -fno-exceptions $
+    -fno-rtti $
+    -isystem${srcroot}/sysroot/include/c++/v1
+
+ldflags = $ldflags $
+    -g $
+    -nostdlib $
+    -znocombreloc $
+    -Bsymbolic $
+    -nostartfiles $
+    -Bstatic $
+    --sysroot="${srcroot}/sysroot" $
+	-L "${srcroot}/sysroot/lib" $
+
+libs = $libs $
+	-lc
+
+{% endblock %}
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
+

src/boot/guids.inc

@@ -3,5 +3,9 @@ GUID(0x8868e871,0xe4f1,0x11d3,0xbc,0x22,0x00,0x80,0xc7,0x3c,0x88,0x81, guid_acpi
 GUID(0x09576e92,0x6d3f,0x11d2,0x8e,0x39,0x00,0xa0,0xc9,0x69,0x72,0x3b, guid_file_info);
 GUID(0x9042a9de,0x23dc,0x4a38,0x96,0xfb,0x7a,0xde,0xd0,0x80,0x51,0x6a, guid_gfx_out);
 GUID(0x964e5b22,0x6459,0x11d2,0x8e,0x39,0x00,0xa0,0xc9,0x69,0x72,0x3b, guid_simple_filesystem);
+GUID(0x09576e91,0x6d3f,0x11d2,0x8e,0x39,0x00,0xa0,0xc9,0x69,0x72,0x3b, guid_device_path);
+GUID(0x8b843e20,0x8132,0x4852,0x90,0xcc,0x55,0x1a,0x4e,0x4a,0x7f,0x1c, guid_device_path_to_text);
+
+GUID(0x10d0669c,0x9ec6,0x4268,0xbc,0x48,0xff,0x74,0x75,0x21,0xfe,0x07, guid_popcorn_vendor);
 
 // vim: ft=c

src/boot/loader.cpp

@@ -9,31 +9,73 @@
 static wchar_t kernel_name[] = KERNEL_FILENAME;
 static wchar_t initrd_name[] = INITRD_FILENAME;
 
+EFI_STATUS
+loader_alloc_aligned(
+	EFI_BOOT_SERVICES *bootsvc,
+	EFI_MEMORY_TYPE mem_type,
+	size_t *length,
+	void **pages)
+{
+	EFI_STATUS status;
+	EFI_PHYSICAL_ADDRESS addr;
+
+	size_t alignment = PAGE_SIZE;
+	while (alignment < *length)
+		alignment *= 2;
+
+	size_t page_count = alignment / PAGE_SIZE;
+	*length = alignment;
+
+	con_debug(L"Trying to find %d aligned pages for %x\n", page_count, mem_type);
+
+	status = bootsvc->AllocatePages(AllocateAnyPages, mem_type, page_count * 2, &addr);
+	CHECK_EFI_STATUS_OR_RETURN(status, L"Allocating %d pages for alignment", page_count * 2);
+	con_debug(L"    Found %d pages at %lx\n", page_count * 2, addr);
+
+	EFI_PHYSICAL_ADDRESS aligned = addr;
+	aligned = ((aligned - 1) & ~(alignment - 1)) + alignment;
+	con_debug(L"    Aligning %lx to %lx\n", addr, aligned);
+
+	size_t before = 
+		(reinterpret_cast<uint64_t>(aligned) -
+		reinterpret_cast<uint64_t>(addr)) /
+		PAGE_SIZE;
+
+	if (before) {
+		con_debug(L"    Freeing %d initial pages\n", before);
+		bootsvc->FreePages(addr, before);
+	}
+
+	size_t after = page_count - before;
+	if (after) {
+		EFI_PHYSICAL_ADDRESS end = 
+			reinterpret_cast<EFI_PHYSICAL_ADDRESS>(
+				reinterpret_cast<uint64_t>(aligned) +
+				page_count * PAGE_SIZE);
+		con_debug(L"    Freeing %d remaining pages\n", after);
+		bootsvc->FreePages(end, after);
+	}
+
+	*pages = (void *)aligned;
+	return EFI_SUCCESS;
+}
+
 EFI_STATUS
 loader_alloc_pages(
 	EFI_BOOT_SERVICES *bootsvc,
 	EFI_MEMORY_TYPE mem_type,
 	size_t *length,
-	void **pages,
-	bool align)
+	void **pages)
 {
 	EFI_STATUS status;
 
 	size_t page_count = ((*length - 1) / PAGE_SIZE) + 1;
 	EFI_PHYSICAL_ADDRESS addr = (EFI_PHYSICAL_ADDRESS)*pages;
 
-	if (align) {
-		// Align addr to the next multiple of N pages
-		size_t align_size = page_count * PAGE_SIZE;
-		addr = ((addr - 1) & ~(align_size - 1)) + align_size;
-	}
+	con_debug(L"Trying to find %d non-aligned pages for %x at %lx\n",
+			page_count, mem_type, addr);
 
 	status = bootsvc->AllocatePages(AllocateAddress, mem_type, page_count, &addr);
-	if (status == EFI_NOT_FOUND || status == EFI_OUT_OF_RESOURCES) {
-		// couldn't get the address we wanted, try loading the kernel anywhere
-		status =
-			bootsvc->AllocatePages(AllocateAnyPages, mem_type, page_count, &addr);
-	}
 	CHECK_EFI_STATUS_OR_RETURN(status,
 		L"Allocating %d kernel pages type %x",
 		page_count, mem_type);
@@ -69,12 +111,11 @@ loader_load_initrd(
 
 	data->initrd_length = ((EFI_FILE_INFO *)info)->FileSize;
 
-	status = loader_alloc_pages(
+	status = loader_alloc_aligned(
 			bootsvc,
 			memtype_initrd,
 			&data->initrd_length,
-			&data->initrd,
-			true);
+			&data->initrd);
 	CHECK_EFI_STATUS_OR_RETURN(status, L"Allocating pages");
 
 	status = file->Read(file, &data->initrd_length, data->initrd);
@@ -143,7 +184,7 @@ loader_load_elf(
 		header.machine != 0x3e)
 		CHECK_EFI_STATUS_OR_RETURN(EFI_LOAD_ERROR, L"ELF load error: wrong machine architecture");
 
-	con_debug(L"ELF is valid, entrypoint %lu\r\n", header.entrypoint);
+	con_debug(L"ELF is valid, entrypoint %lx\r\n", header.entrypoint);
 
 	data->kernel_entry = (void *)header.entrypoint;
 
@@ -161,7 +202,7 @@ loader_load_elf(
 
 		length = prog_header.mem_size;
 		void *addr = (void *)(prog_header.vaddr - KERNEL_VIRT_ADDRESS);
-		status = loader_alloc_pages(bootsvc, memtype_kernel, &length, &addr, false);
+		status = loader_alloc_pages(bootsvc, memtype_kernel, &length, &addr);
 		CHECK_EFI_STATUS_OR_RETURN(status, L"Allocating kernel pages");
 
 		if (data->kernel == 0)
@@ -236,20 +277,19 @@ loader_load_kernel(
 
 		CHECK_EFI_STATUS_OR_RETURN(status, L"loader_load_elf: %s", kernel_name);
 
-		data->initrd = (void *)((uint64_t)data->kernel + data->kernel_length);
-		status = loader_load_initrd(bootsvc, root, data);
-
-		CHECK_EFI_STATUS_OR_RETURN(status, L"loader_load_file: %s", initrd_name);
-
-		data->data = (void *)((uint64_t)data->initrd + data->initrd_length);
+		data->data = (void *)((uint64_t)data->kernel + data->kernel_length);
 		data->data_length += PAGE_SIZE; // extra page for map growth
-		status = loader_alloc_pages(
+
+		status = loader_alloc_aligned(
 				bootsvc,
 				memtype_data,
 				&data->data_length,
-				&data->data,
-				true);
-		CHECK_EFI_STATUS_OR_RETURN(status, L"loader_alloc_pages: kernel data");
+				&data->data);
+		CHECK_EFI_STATUS_OR_RETURN(status, L"loader_alloc_aligned: kernel data");
+
+		data->initrd = (void *)((uint64_t)data->data + data->data_length);
+		status = loader_load_initrd(bootsvc, root, data);
+		CHECK_EFI_STATUS_OR_RETURN(status, L"loader_load_file: %s", initrd_name);
 
 		return EFI_SUCCESS;
 	}

src/boot/main.cpp

@@ -5,7 +5,7 @@
 
 #include "console.h"
 #include "guids.h"
-#include "kernel_data.h"
+#include "kernel_args.h"
 #include "loader.h"
 #include "memory.h"
 #include "utility.h"
@@ -34,7 +34,51 @@ struct kernel_header {
 };
 #pragma pack(pop)
 
-using kernel_entry = void (*)(popcorn_data *);
+using kernel_entry = void (*)(kernel_args *);
+
+static void
+type_to_wchar(wchar_t *into, uint32_t type)
+{
+	for (int j=0; j<4; ++j)
+		into[j] = static_cast<wchar_t>(reinterpret_cast<char *>(&type)[j]);
+}
+
+EFI_STATUS
+detect_debug_mode(EFI_RUNTIME_SERVICES *run, kernel_args *header) {
+	wchar_t var_name[] = L"debug";
+
+	EFI_STATUS status;
+	uint8_t debug = 0;
+	UINTN var_size = sizeof(debug);
+
+#ifdef __POPCORN_SET_DEBUG_UEFI_VAR__
+	debug = __POPCORN_SET_DEBUG_UEFI_VAR__;
+	uint32_t attrs =
+		EFI_VARIABLE_NON_VOLATILE |
+		EFI_VARIABLE_BOOTSERVICE_ACCESS |
+		EFI_VARIABLE_RUNTIME_ACCESS;
+	status = run->SetVariable(
+			var_name,
+			&guid_popcorn_vendor,
+			attrs,
+			var_size,
+			&debug);
+	CHECK_EFI_STATUS_OR_RETURN(status, "detect_debug_mode::SetVariable");
+#endif
+
+	status = run->GetVariable(
+			var_name,
+			&guid_popcorn_vendor,
+			nullptr,
+			&var_size,
+			&debug);
+	CHECK_EFI_STATUS_OR_RETURN(status, "detect_debug_mode::GetVariable");
+
+	if (debug)
+		header->flags |= POPCORN_FLAG_DEBUG;
+
+	return EFI_SUCCESS;
+}
 
 extern "C" EFI_STATUS
 efi_main(EFI_HANDLE image_handle, EFI_SYSTEM_TABLE *system_table)
@@ -73,14 +117,13 @@ efi_main(EFI_HANDLE image_handle, EFI_SYSTEM_TABLE *system_table)
 	status = memory_get_map_length(bootsvc, &data_length);
 	CHECK_EFI_STATUS_OR_FAIL(status);
 
-	size_t header_size = sizeof(popcorn_data);
-	const size_t header_align = alignof(popcorn_data);
+	size_t header_size = sizeof(kernel_args);
+	const size_t header_align = alignof(kernel_args);
 	if (header_size % header_align)
 		header_size += header_align - (header_size % header_align);
 
 	data_length += header_size;
 
-
 	// Load the kernel image from disk and check it
 	//
 	console::print(L"Loading kernel into memory...\r\n");
@@ -90,9 +133,9 @@ efi_main(EFI_HANDLE image_handle, EFI_SYSTEM_TABLE *system_table)
 	status = loader_load_kernel(bootsvc, &load);
 	CHECK_EFI_STATUS_OR_FAIL(status);
 
-	console::print(L"    %u image bytes at 0x%x\r\n", load.kernel_length, load.kernel);
-	console::print(L"    %u initrd bytes at 0x%x\r\n", load.initrd_length, load.initrd);
-	console::print(L"    %u data bytes at 0x%x\r\n", load.data_length, load.data);
+	console::print(L"    %x image bytes  at 0x%x\r\n", load.kernel_length, load.kernel);
+	console::print(L"    %x data bytes   at 0x%x\r\n", load.data_length, load.data);
+	console::print(L"    %x initrd bytes at 0x%x\r\n", load.initrd_length, load.initrd);
 
 	struct kernel_header *version = (struct kernel_header *)load.kernel;
 	if (version->magic != KERNEL_HEADER_MAGIC) {
@@ -102,7 +145,7 @@ efi_main(EFI_HANDLE image_handle, EFI_SYSTEM_TABLE *system_table)
 
 	console::print(L"    Kernel version %d.%d.%d %x%s\r\n",
 			version->major, version->minor, version->patch, version->gitsha & 0x0fffffff,
-			version->gitsha & 0xf0000000 ? "*" : "");
+			version->gitsha & 0xf0000000 ? L"*" : L"");
 	console::print(L"    Entrypoint 0x%x\r\n", load.kernel_entry);
 
 	kernel_entry kernel_main =
@@ -111,12 +154,12 @@ efi_main(EFI_HANDLE image_handle, EFI_SYSTEM_TABLE *system_table)
 
 	// Set up the kernel data pages to pass to the kernel
 	//
-	struct popcorn_data *data_header = (struct popcorn_data *)load.data; 
+	struct kernel_args *data_header = (struct kernel_args *)load.data;
 	memory_mark_pointer_fixup((void **)&data_header);
 
 	data_header->magic = DATA_HEADER_MAGIC;
 	data_header->version = DATA_HEADER_VERSION;
-	data_header->length = sizeof(struct popcorn_data);
+	data_header->length = sizeof(struct kernel_args);
 
 	data_header->scratch_pages = SCRATCH_PAGES;
 	data_header->flags = 0;
@@ -168,6 +211,8 @@ efi_main(EFI_HANDLE image_handle, EFI_SYSTEM_TABLE *system_table)
 	data_header->memory_map_length = map.length;
 	data_header->memory_map_desc_size = map.size;
 
+	detect_debug_mode(runsvc, data_header);
+
 	// bootsvc->Stall(5000000);
 
 	status = bootsvc->ExitBootServices(image_handle, map.key);

src/drivers/nulldrv/main.cpp

@@ -1 +1,24 @@
-int main(int argc, const char **argv) { return 0; }
+#include <stdint.h>
+#include <stdlib.h>
+
+extern "C" {
+	int32_t getpid();
+	int32_t fork();
+	void sleep(uint64_t til);
+	void debug();
+	void message(const char *msg);
+
+	int main(int, const char **);
+}
+
+
+int
+main(int argc, const char **argv)
+{
+	int32_t pid = getpid();
+	int32_t child = fork();
+	message("hello from nulldrv!");
+	for (int i = 1; i < 5; ++i)
+		sleep(i*10);
+	return 0;
+}

src/drivers/nulldrv/main.s

@@ -1,61 +1,79 @@
 section .bss
-mypid: resq 1
-mychild: resq 1
+mymessage:
+	resq 1024
+
+extern main
+extern exit
 
 section .text
+global getpid
+getpid:
+	push rbp
+	mov rbp, rsp
+
+	mov rax, 0x02  ; getpid syscall
+	syscall        ; pid is now already in rax, so just return
+
+	pop rbp
+	ret
+
+global debug
+debug:
+	push rbp
+	mov rbp, rsp
+
+	mov rax, 0x00  ; debug syscall
+	syscall
+
+	pop rbp
+	ret
+
+global sleep
+sleep:
+	push rbp
+	mov rbp, rsp
+
+	mov rax, 0x21  ; sleep syscall
+	syscall
+
+	pop rbp
+	ret
+
+global fork
+fork:
+	push rbp
+	mov rbp, rsp
+
+	mov rax, 0x03
+	syscall    ; pid left in rax
+
+	pop rbp
+	ret
+	
+
+global message
+message:
+	push rbp
+	mov rbp, rsp
+
+	; message should already be in rdi
+	mov rax, 0x10
+	syscall
+
+	pop rbp
+	ret
+
+
 global _start
 _start:
 	xor rbp, rbp	; Sentinel rbp
+	push rbp
+	push rbp
+	mov rbp, rsp
 
-	mov rax, 5		; GETPID syscall
-	int 0xee
-	mov [mypid], rax
+	mov rdi, 0
+	mov rsi, 0
+	call main
 
-	mov rax, 8		; FORK syscall
-	int 0xee
-	mov [mychild], rax
-
-	mov r12, [mypid]
-	mov r13, [mychild]
-	mov rax, 1		; DEBUG syscall
-	int 0xee
-
-	cmp r12, 1
-	je .dosend
-	jne .doreceive
-
-.preloop:
-	mov r11, 0		; counter
-	mov rbx, 20		; sleep timeout
-
-.loop:
-	mov rax, 1		; MESSAGE syscall
-	;mov rax, 0		; NOOP syscall
-	;syscall
-	int 0xee
-
-	inc r11
-	cmp r11, 2
-
-	jle .loop
-
-	mov rax, 4		; SLEEP syscall
-	; syscall
-	int 0xee
-
-	add rbx, 20
-
-	mov r11, 0
-	jmp .loop
-
-.dosend:
-	mov rax, 6      ; SEND syscall
-	mov rdi, 2      ; target is pid 2
-	int 0xee
-	jmp .preloop
-
-.doreceive:
-	mov rax, 7      ; RECEIVE syscall
-	mov rdi, 1      ; source is pid 2
-	int 0xee
-	jmp .preloop
+	mov rdi, rax
+	call exit

src/include/kernel_args.h

@@ -7,8 +7,10 @@
 #define DATA_HEADER_MAGIC     0x600dda7a
 #define DATA_HEADER_VERSION   1
 
+#define POPCORN_FLAG_DEBUG    0x00000001
+
 #pragma pack(push, 1)
-struct popcorn_data {
+struct kernel_args {
 	uint32_t magic;
 	uint16_t version;
 	uint16_t length;

src/include/kernel_memory.h

@@ -2,6 +2,9 @@
 /// \file kernel_memory.h
 /// Constants related to the kernel's memory layout
 
+#include <stddef.h>
+#include <stdint.h>
+
 namespace memory {
 
 	/// Size of a single page frame.
@@ -19,4 +22,11 @@ namespace memory {
 	/// Initial process thread's stack size, in pages
 	static const unsigned initial_stack_pages = 1;
 
+	/// Max size of the kernel heap
+	static const size_t kernel_max_heap = 0x800000000; // 32GiB
+
+	/// Helper to determine if a physical address can be accessed
+	/// through the page_offset area.
+	inline bool page_mappable(uintptr_t a) { return (a & page_offset) == 0; }
+
 } // namespace memory

src/include/log_areas.inc

@@ -0,0 +1,9 @@
+LOG(apic,   info);
+LOG(device, info);
+LOG(paging, warn);
+LOG(driver, info);
+LOG(memory, info);
+LOG(fs,     info);
+LOG(task,   info);
+LOG(boot,   info);
+LOG(syscall,info);

src/kernel/apic.cpp

@@ -46,6 +46,7 @@ apic::apic(uint32_t *base) :
 lapic::lapic(uint32_t *base, isr spurious) :
 	apic(base)
 {
+	// TODO: This causes a "reserved" page fault under KVM
 	apic_write(m_base, 0xf0, static_cast<uint32_t>(spurious));
 	log::info(logs::apic, "LAPIC created, base %lx", m_base);
 }

src/kernel/assert.cpp

@@ -8,19 +8,14 @@ __kernel_assert(const char *file, unsigned line, const char *message)
 	if (cons) {
 		cons->set_color(9 , 0);
 		cons->puts("\n\n       ERROR: ");
+		cons->puts(message);
+		cons->puts("\n              ");
 		cons->puts(file);
 		cons->puts(":");
 		cons->put_dec(line);
-		cons->puts(":  ");
-		cons->puts(message);
+		cons->puts("\n");
 	}
 
-	__asm__ ( "int $0e7h" );
 	while (1) __asm__ ("hlt");
-}
-
-extern "C" [[noreturn]] void
-__assert_fail(const char *message, const char *file, unsigned int line, const char *function)
-{
-	__kernel_assert(file, line, message);
+	__asm__ ( "int $0xe4" );
 }

src/kernel/boot.s

@@ -18,11 +18,11 @@ global _start:function (_start.end - _start)
 _start:
 	cli
 
-	mov rsp, stack_end
-	push 0 ; signal end of stack with 0 return address
-	push 0 ; and a few extra entries in case of stack
-	push 0 ; problems
-	push 0
+	mov rsp, idle_stack_end
+	mov qword [rsp + 0x00], 0 ; signal end of stack with 0 return address
+	mov qword [rsp + 0x08], 0 ; and a few extra entries in case of stack
+	mov qword [rsp + 0x10], 0 ; problems
+	mov qword [rsp + 0x18], 0
 
 	mov rbp, rsp
 	extern kernel_main
@@ -47,6 +47,9 @@ interrupts_disable:
 
 section .bss
 align 0x100
-stack_begin:
-	resb 0x4000 ; 16KiB stack space
-stack_end:
+idle_stack_begin:
+	resb 0x1000 ; 4KiB stack space
+
+global idle_stack_end
+idle_stack_end:
+	resq 4

src/kernel/console.cpp

@@ -1,6 +1,7 @@
 #include "kutil/coord.h"
 #include "kutil/guid.h"
 #include "kutil/memory.h"
+#include "kutil/printf.h"
 #include "console.h"
 #include "font.h"
 #include "screen.h"
@@ -278,100 +279,9 @@ console::putc(char c)
 void console::vprintf(const char *fmt, va_list args)
 {
 	static const unsigned buf_size = 256;
-	char buffer[256];
-
-	const char *r = fmt;
-	char *w = buffer;
-	char *wend = buffer + buf_size;
-
-#define flush() do { *w = 0; puts(buffer); w = buffer; } while(0)
-
-	while (r && *r) {
-		if (w == wend) flush();
-
-		if (*r != '%') {
-			*w++ = *r++;
-			continue;
-		}
-
-		r++; // chomp the %
-		flush();
-
-		bool done = false;
-		bool right = true;
-		int width = 0;
-		char pad = ' ';
-
-		while (!done) {
-			char c = *r++;
-			switch (c) {
-				case '%': *w = '%'; done = true; break;
-
-				case '0':
-					if (width == 0) pad = '0';
-					// else fall through
-				case '1':
-				case '2':
-				case '3':
-				case '4':
-				case '5':
-				case '6':
-				case '7':
-				case '8':
-				case '9': width = width * 10 + (c - '0'); break;
-
-				case '-': right = !right; break;
-
-				case 'x': put_hex<uint32_t>(va_arg(args, uint32_t), right ? width : -width, pad); done = true; break;
-
-				case 'd':
-				case 'u': put_dec<uint32_t>(va_arg(args, uint32_t), right ? width : -width, pad); done = true; break;
-
-				case 's': {
-						const char *s = va_arg(args, const char*);
-						if (s) puts(s);
-					}
-					done = true;
-					break;
-
-				case 'G': {
-						// Special: GUID type
-						kutil::guid g = va_arg(args, kutil::guid);
-						put_hex<uint32_t>(g.a, 8, '0');
-						putc('-');
-						put_hex<uint16_t>((g.a >> 32) & 0xffff, 4, '0');
-						putc('-');
-						put_hex<uint16_t>((g.a >> 48) & 0xffff, 4, '0');
-						putc('-');
-						put_hex<uint16_t>((kutil::byteswap(g.b) >> 16) & 0xffff, 4, '0');
-						putc('-');
-						put_hex<uint16_t>(kutil::byteswap(g.b) & 0xffff, 4, '0');
-						put_hex<uint32_t>(kutil::byteswap(g.b >> 32), 8, '0');
-					}
-					done = true;
-					break;
-
-				case 'l':
-					switch (*r++) {
-						case 'x': put_hex<uint64_t>(va_arg(args, uint64_t), right ? width : -width, pad); done = true; break;
-
-						case 'd':
-						case 'u': put_dec<uint32_t>(va_arg(args, uint64_t), right ? width : -width, pad); done = true; break;
-
-						default:
-							done = true;
-							break;
-					}
-					break;
-
-				default:
-					done = true;
-					break;
-			}
-		}
-	}
-
-	flush();
+	char buffer[buf_size];
+	vsnprintf_(buffer, buf_size, fmt, args);
+	puts(buffer);
 }
 
 void

src/kernel/cpu.cpp

@@ -3,6 +3,8 @@
 #include "cpu.h"
 #include "log.h"
 
+cpu_data bsp_cpu_data;
+
 inline static void
 __cpuid(
 	uint32_t leaf,

src/kernel/cpu.h

@@ -2,15 +2,27 @@
 
 #include <stdint.h>
 
+struct process;
+
 struct cpu_state
 {
-	uint64_t ds;
 	uint64_t r15, r14, r13, r12, r11, r10, r9,  r8;
 	uint64_t rdi, rsi, rbp, rbx, rdx, rcx, rax;
 	uint64_t interrupt, errorcode;
 	uint64_t rip, cs, rflags, user_rsp, ss;
 };
 
+/// Per-cpu state data. If you change this, remember to update the assembly
+/// version in 'tasking.inc'
+struct cpu_data
+{
+	uintptr_t rsp0;
+	uintptr_t rsp3;
+	process *tcb;
+};
+
+extern cpu_data bsp_cpu_data;
+
 class cpu_id
 {
 public:

src/kernel/debug.cpp

@@ -4,56 +4,65 @@
 #include "gdt.h"
 #include "page_manager.h"
 
-#define print_reg(name, value) cons->printf("         %s: %016lx\n", name, (value));
+size_t __counter_syscall_enter = 0;
+size_t __counter_syscall_sysret = 0;
 
 void
 print_regs(const cpu_state &regs)
 {
 	console *cons = console::get();
 
-	print_reg("rax", regs.rax);
-	print_reg("rbx", regs.rbx);
-	print_reg("rcx", regs.rcx);
-	print_reg("rdx", regs.rdx);
-	print_reg("rdi", regs.rdi);
-	print_reg("rsi", regs.rsi);
+	uint64_t cr2 = 0;
+	__asm__ __volatile__ ("mov %%cr2, %0" : "=r"(cr2));
+
+	print_regL("rax", regs.rax);
+	print_regM("rbx", regs.rbx);
+	print_regR("rcx", regs.rcx);
+	print_regL("rdx", regs.rdx);
+	print_regM("rdi", regs.rdi);
+	print_regR("rsi", regs.rsi);
 
 	cons->puts("\n");
-	print_reg(" r8", regs.r8);
-	print_reg(" r9", regs.r9);
-	print_reg("r10", regs.r10);
-	print_reg("r11", regs.r11);
-	print_reg("r12", regs.r12);
-	print_reg("r13", regs.r13);
-	print_reg("r14", regs.r14);
-	print_reg("r15", regs.r15);
+	print_regL(" r8", regs.r8);
+	print_regM(" r9", regs.r9);
+	print_regR("r10", regs.r10);
+	print_regL("r11", regs.r11);
+	print_regM("r12", regs.r12);
+	print_regR("r13", regs.r13);
+	print_regL("r14", regs.r14);
+	print_regM("r15", regs.r15);
+
+	cons->puts("\n\n");
+	print_regL("rbp", regs.rbp);
+	print_regM("rsp", regs.user_rsp);
+	print_regR("sp0", bsp_cpu_data.rsp0);
+
+	print_regL("rip", regs.rip);
+	print_regM("cr3", page_manager::get()->get_pml4());
+	print_regR("cr2", cr2);
 
 	cons->puts("\n");
-	print_reg("rbp", regs.rbp);
-	print_reg("rsp", regs.user_rsp);
-	print_reg("sp0", tss_get_stack(0));
-
-	cons->puts("\n");
-	print_reg(" ds", regs.ds);
-	print_reg(" cs", regs.cs);
-	print_reg(" ss", regs.ss);
-
-	cons->puts("\n");
-	print_reg("rip", regs.rip);
-
-	cons->puts("\n");
-	print_reg("cr3", page_manager::get()->get_pml4());
 }
 
+struct frame
+{
+	frame *prev;
+	uintptr_t return_addr;
+};
+
 void
 print_stacktrace(int skip)
 {
 	console *cons = console::get();
-	int frame = 0;
-	uint64_t bp = get_frame(skip);
-	while (bp) {
-		cons->printf("    frame %2d: %lx\n", frame, bp);
-		bp = get_frame(++frame + skip);
+
+	frame *fp = nullptr;
+	int fi = -skip;
+	__asm__ __volatile__ ( "mov %%rbp, %0" : "=r" (fp) );
+
+	while (fp && fp->return_addr) {
+		if (fi++ >= 0)
+			cons->printf("  frame %2d: %lx\n", fi-1, fp->return_addr);
+		fp = fp->prev;
 	}
 }
 

src/kernel/debug.h

@@ -8,12 +8,18 @@ extern "C" {
 	uintptr_t get_rsp();
 	uintptr_t get_rip();
 	uintptr_t get_frame(int frame);
-
+	uintptr_t get_gsbase();
+	void _halt();
 }
 
+extern size_t __counter_syscall_enter;
+extern size_t __counter_syscall_sysret;
+
 void print_regs(const cpu_state &regs);
 void print_stack(const cpu_state &regs);
-void print_stacktrace(int skip = 0);
+void print_stacktrace(int skip);
 
-#define print_reg(name, value) cons->printf("         %s: %016lx\n", name, (value));
+#define print_regL(name, value) cons->printf("       %s: %016lx", name, (value));
+#define print_regM(name, value) cons->printf("    %s: %016lx", name, (value));
+#define print_regR(name, value) cons->printf("    %s: %016lx\n", name, (value));
 

src/kernel/debug.s

@@ -8,24 +8,13 @@ get_rip:
 	pop rax ; do the same thing as 'ret', except with 'jmp'
 	jmp rax ; with the return address still in rax
 
+global get_gsbase
+get_gsbase:
+	rdgsbase rax
+	ret
+
 global _halt
 _halt:
 	hlt
 	jmp _halt
 
-
-global get_frame
-get_frame:
-	mov rcx, rbp
-
-.loop:
-	mov rax, [rcx + 8]
-	mov rcx, [rcx]
-	cmp rdi, 0
-	je .done
-
-	sub rdi, 1
-	jmp .loop
-
-.done:
-	ret

src/kernel/device_manager.cpp

@@ -14,7 +14,7 @@
 
 static const char expected_signature[] = "RSD PTR ";
 
-device_manager device_manager::s_instance(nullptr);
+device_manager device_manager::s_instance(nullptr, kutil::allocator::invalid);
 
 struct acpi1_rsdp
 {
@@ -59,8 +59,13 @@ void irq4_callback(void *)
 }
 
 
-device_manager::device_manager(const void *root_table) :
-	m_lapic(nullptr)
+device_manager::device_manager(const void *root_table, kutil::allocator &alloc) :
+	m_lapic(nullptr),
+	m_ioapics(alloc),
+	m_pci(alloc),
+	m_devices(alloc),
+	m_irqs(alloc),
+	m_blockdevs(alloc)
 {
 	kassert(root_table != 0, "ACPI root table pointer is null.");
 
@@ -93,7 +98,7 @@ device_manager::device_manager(const void *root_table) :
 ioapic *
 device_manager::get_ioapic(int i)
 {
-	return (i < m_ioapics.count()) ? m_ioapics[i] : nullptr;
+	return (i < m_ioapics.count()) ? &m_ioapics[i] : nullptr;
 }
 
 static void
@@ -148,26 +153,44 @@ device_manager::load_apic(const acpi_apic *apic)
 	uint8_t const *p = apic->controller_data;
 	uint8_t const *end = p + count;
 
-	// Pass one: set up IOAPIC objcts
+	// Pass one: count IOAPIC objcts
+	int num_ioapics = 0;
+	while (p < end) {
+		const uint8_t type = p[0];
+		const uint8_t length = p[1];
+		if (type == 1) num_ioapics++;
+		p += length;
+	}
+
+	m_ioapics.set_capacity(num_ioapics);
+
+	// Pass two: set up IOAPIC objcts
+	p = apic->controller_data;
 	while (p < end) {
 		const uint8_t type = p[0];
 		const uint8_t length = p[1];
 		if (type == 1) {
 			uint32_t *base = reinterpret_cast<uint32_t *>(kutil::read_from<uint32_t>(p+4));
 			uint32_t base_gsr = kutil::read_from<uint32_t>(p+8);
-			m_ioapics.append(new ioapic(base, base_gsr));
+			m_ioapics.emplace(base, base_gsr);
 		}
 		p += length;
 	}
 
-	// Pass two: configure APIC objects
+	// Pass three: configure APIC objects
 	p = apic->controller_data;
 	while (p < end) {
 		const uint8_t type = p[0];
 		const uint8_t length = p[1];
 
 		switch (type) {
-		case 0: // Local APIC
+		case 0: { // Local APIC
+				uint8_t uid = kutil::read_from<uint8_t>(p+2);
+				uint8_t id = kutil::read_from<uint8_t>(p+3);
+				log::debug(logs::device, "    Local APIC uid %x id %x", id);
+			}
+			break;
+
 		case 1: // I/O APIC
 			break;
 
@@ -180,7 +203,7 @@ device_manager::load_apic(const acpi_apic *apic)
 						source, gsi, (flags & 0x3), ((flags >> 2) & 0x3));
 
 				// TODO: in a multiple-IOAPIC system this might be elsewhere
-				m_ioapics[0]->redirect(source, static_cast<isr>(gsi), flags, true);
+				m_ioapics[0].redirect(source, static_cast<isr>(gsi), flags, true);
 			}
 			break;
 
@@ -206,10 +229,10 @@ device_manager::load_apic(const acpi_apic *apic)
 		p += length;
 	}
 
-	for (uint8_t i = 0; i < m_ioapics[0]->get_num_gsi(); ++i) {
+	for (uint8_t i = 0; i < m_ioapics[0].get_num_gsi(); ++i) {
 		switch (i) {
 			case 2: break;
-			default: m_ioapics[0]->mask(i, false);
+			default: m_ioapics[0].mask(i, false);
 		}
 	}
 

src/kernel/device_manager.h

@@ -2,14 +2,13 @@
 /// \file device_manager.h
 /// The device manager definition
 #include "kutil/vector.h"
+#include "apic.h"
 #include "pci.h"
 
 struct acpi_xsdt;
 struct acpi_apic;
 struct acpi_mcfg;
 class block_device;
-class lapic;
-class ioapic;
 
 using irq_callback = void (*)(void *);
 
@@ -19,8 +18,9 @@ class device_manager
 {
 public:
 	/// Constructor.
-	/// \arg root_table   Pointer to the ACPI RSDP
-	device_manager(const void *root_table);
+	/// \arg root_table  Pointer to the ACPI RSDP
+	/// \arg alloc       Allocator for device arrays
+	device_manager(const void *root_table, kutil::allocator &alloc);
 
 	/// Get the system global device manager.
 	/// \returns  A reference to the system device manager
@@ -105,7 +105,7 @@ private:
 	void bad_irq(uint8_t irq);
 
 	lapic *m_lapic;
-	kutil::vector<ioapic *> m_ioapics;
+	kutil::vector<ioapic> m_ioapics;
 
 	kutil::vector<pci_group> m_pci;
 	kutil::vector<pci_device> m_devices;

src/kernel/frame_allocator.cpp

@@ -0,0 +1,93 @@
+#include "kutil/assert.h"
+#include "kutil/memory.h"
+#include "frame_allocator.h"
+
+using memory::frame_size;
+using memory::page_offset;
+using frame_block_node = kutil::list_node<frame_block>;
+
+frame_allocator g_frame_allocator;
+
+int
+frame_block::compare(const frame_block *rhs) const
+{
+	if (address < rhs->address)
+		return -1;
+	else if (address > rhs->address)
+		return 1;
+	return 0;
+}
+
+
+frame_allocator::raw_alloc::raw_alloc(frame_allocator &fa) : m_fa(fa) {}
+
+void *
+frame_allocator::raw_alloc::allocate(size_t size)
+{
+	kassert(size <= frame_size, "Raw allocator only allocates a single page");
+
+	uintptr_t addr = 0;
+	if (size <= frame_size)
+		m_fa.allocate(1, &addr);
+	return reinterpret_cast<void*>(addr + page_offset);
+}
+
+void
+frame_allocator::raw_alloc::free(void *p)
+{
+	m_fa.free(reinterpret_cast<uintptr_t>(p), 1);
+}
+
+
+frame_allocator::frame_allocator() :
+	m_raw_alloc(*this)
+{
+}
+
+size_t
+frame_allocator::allocate(size_t count, uintptr_t *address)
+{
+	kassert(!m_free.empty(), "frame_allocator::pop_frames ran out of free frames!");
+	if (m_free.empty())
+		return 0;
+
+	auto *first = m_free.front();
+
+	if (count >= first->count) {
+		*address = first->address;
+		m_free.remove(first);
+		return first->count;
+	} else {
+		first->count -= count;
+		*address = first->address + (first->count * frame_size);
+		return count;
+	}
+}
+
+inline uintptr_t end(frame_block *node) { return node->address + node->count * frame_size; }
+
+void
+frame_allocator::free(uintptr_t address, size_t count)
+{
+	frame_block_node *node =
+		reinterpret_cast<frame_block_node*>(address + page_offset);
+
+	kutil::memset(node, 0, sizeof(frame_block_node));
+	node->address = address;
+	node->count = count;
+
+	m_free.sorted_insert(node);
+
+	frame_block_node *next = node->next();
+	if (next && end(node) == next->address) {
+		node->count += next->count;
+		m_free.remove(next);
+	}
+
+	frame_block_node *prev = node->prev();
+	if (prev && end(prev) == address) {
+		prev->count += node->count;
+		m_free.remove(node);
+	}
+}
+

src/kernel/frame_allocator.h

@@ -0,0 +1,70 @@
+#pragma once
+/// \file frame_allocator.h
+/// Allocator for physical memory frames
+
+#include <stdint.h>
+
+#include "kutil/allocator.h"
+#include "kutil/linked_list.h"
+
+struct frame_block;
+using frame_block_list = kutil::linked_list<frame_block>;
+
+/// Allocator for physical memory frames
+class frame_allocator
+{
+public:
+	/// Default constructor
+	frame_allocator();
+
+	/// Get free frames from the free list. Only frames from the first free block
+	/// are returned, so the number may be less than requested, but they will
+	/// be contiguous.
+	/// \arg count    The maximum number of frames to get
+	/// \arg address  [out] The physical address of the first frame
+	/// \returns      The number of frames retrieved
+	size_t allocate(size_t count, uintptr_t *address);
+
+	/// Free previously allocated frames.
+	/// \arg address  The physical address of the first frame to free
+	/// \arg count    The number of frames to be freed
+	void free(uintptr_t address, size_t count);
+
+	/// Get a memory allocator that allocates raw pages
+	/// \returns  The allocator ojbect
+	kutil::allocator & raw_allocator() { return m_raw_alloc; }
+
+private:
+	class raw_alloc :
+		public kutil::allocator
+	{
+	public:
+		raw_alloc(frame_allocator &fa);
+		virtual void * allocate(size_t size) override;
+		virtual void free(void *p) override;
+	private:
+		frame_allocator &m_fa;
+	};
+
+	raw_alloc m_raw_alloc;
+	frame_block_list m_free; ///< Free frames list
+
+	frame_allocator(const frame_allocator &) = delete;
+};
+
+
+/// A block of contiguous frames. Each `frame_block` represents contiguous
+/// physical frames with the same attributes.
+struct frame_block
+{
+	uintptr_t address;
+	uint32_t count;
+
+	/// Compare two blocks by address.
+	/// \arg rhs   The right-hand comparator
+	/// \returns   <0 if this is sorts earlier, >0 if this sorts later, 0 for equal
+	int compare(const frame_block *rhs) const;
+};
+
+
+extern frame_allocator g_frame_allocator;

src/kernel/gdt.cpp

@@ -184,18 +184,25 @@ gdt_init()
 }
 
 void
-gdt_dump()
+gdt_dump(int index)
 {
 	const table_ptr &table = g_gdtr;
 
 	console *cons = console::get();
-	cons->printf("         GDT: loc:%lx size:%d\n", table.base, table.limit+1);
 
+	int start = 0;
 	int count = (table.limit + 1) / sizeof(gdt_descriptor);
+	if (index != -1) {
+		start = index;
+		count = 1;
+	} else {
+		cons->printf("         GDT: loc:%lx size:%d\n", table.base, table.limit+1);
+	}
+
 	const gdt_descriptor *gdt =
 		reinterpret_cast<const gdt_descriptor *>(table.base);
 
-	for (int i = 0; i < count; ++i) {
+	for (int i = start; i < start+count; ++i) {
 		uint32_t base =
 			(gdt[i].base_high << 24) |
 			(gdt[i].base_mid << 16) |
@@ -233,17 +240,25 @@ gdt_dump()
 }
 
 void
-idt_dump()
+idt_dump(int index)
 {
 	const table_ptr &table = g_idtr;
 
-	log::info(logs::boot, "Loaded IDT at: %lx size: %d bytes", table.base, table.limit+1);
 
+	int start = 0;
 	int count = (table.limit + 1) / sizeof(idt_descriptor);
+	if (index != -1) {
+		start = index;
+		count = 1;
+		log::info(logs::boot, "IDT FOR INDEX %02x", index);
+	} else {
+		log::info(logs::boot, "Loaded IDT at: %lx size: %d bytes", table.base, table.limit+1);
+	}
+
 	const idt_descriptor *idt =
 		reinterpret_cast<const idt_descriptor *>(table.base);
 
-	for (int i = 0; i < count; ++i) {
+	for (int i = start; i < start+count; ++i) {
 		uint64_t base =
 			(static_cast<uint64_t>(idt[i].base_high) << 32) |
 			(static_cast<uint64_t>(idt[i].base_mid)  << 16) |

src/kernel/gdt.h

@@ -25,7 +25,9 @@ void tss_set_stack(int ring, uintptr_t rsp);
 uintptr_t tss_get_stack(int ring);
 
 /// Dump information about the current GDT to the screen
-void gdt_dump();
+/// \arg index  Which entry to print, or -1 for all entries
+void gdt_dump(int index = -1);
 
 /// Dump information about the current IDT to the screen
-void idt_dump();
+/// \arg index  Which entry to print, or -1 for all entries
+void idt_dump(int index = -1);

src/kernel/interrupts.cpp

@@ -18,9 +18,8 @@ static const uint16_t PIC2 = 0xa0;
 extern "C" {
 	void _halt();
 
-	uintptr_t isr_handler(uintptr_t, cpu_state*);
-	uintptr_t irq_handler(uintptr_t, cpu_state*);
-	uintptr_t syscall_handler(uintptr_t, cpu_state);
+	void isr_handler(cpu_state*);
+	void irq_handler(cpu_state*);
 
 #define ISR(i, name)     extern void name ();
 #define EISR(i, name)    extern void name ();
@@ -104,13 +103,45 @@ interrupts_init()
 	log::info(logs::boot, "Interrupts enabled.");
 }
 
-uintptr_t
-isr_handler(uintptr_t return_rsp, cpu_state *regs)
+void
+isr_handler(cpu_state *regs)
 {
 	console *cons = console::get();
 
 	switch (static_cast<isr>(regs->interrupt & 0xff)) {
 
+	case isr::isrDebug: {
+			cons->set_color(11);
+			cons->puts("\nDebug Exception:\n");
+			cons->set_color();
+
+			uint64_t dr = 0;
+
+			__asm__ __volatile__ ("mov %%dr0, %0" : "=r"(dr));
+			print_regL("dr0", dr);
+
+			__asm__ __volatile__ ("mov %%dr1, %0" : "=r"(dr));
+			print_regM("dr1", dr);
+
+			__asm__ __volatile__ ("mov %%dr2, %0" : "=r"(dr));
+			print_regM("dr2", dr);
+
+			__asm__ __volatile__ ("mov %%dr3, %0" : "=r"(dr));
+			print_regR("dr3", dr);
+
+			__asm__ __volatile__ ("mov %%dr6, %0" : "=r"(dr));
+			print_regL("dr6", dr);
+
+			__asm__ __volatile__ ("mov %%dr7, %0" : "=r"(dr));
+			print_regR("dr7", dr);
+
+			print_regL("rip", regs->rip);
+			print_regM("rsp", regs->user_rsp);
+			print_regM("fla", regs->rflags);
+			_halt();
+		}
+		break;
+
 	case isr::isrGPFault: {
 			cons->set_color(9);
 			cons->puts("\nGeneral Protection Fault:\n");
@@ -124,13 +155,13 @@ isr_handler(uintptr_t return_rsp, cpu_state *regs)
 				switch ((regs->errorcode & 0x07) >> 1) {
 				case 0:
 					cons->printf(" GDT[%x]\n", index);
-					gdt_dump();
+					gdt_dump(index);
 					break;
 
 				case 1:
 				case 3:
 					cons->printf(" IDT[%x]\n", index);
-					idt_dump();
+					idt_dump(index);
 					break;
 
 				default:
@@ -145,41 +176,35 @@ isr_handler(uintptr_t return_rsp, cpu_state *regs)
 			print_stacktrace(2);
 			print_stack(*regs);
 			*/
-
 		}
 		_halt();
 		break;
 
 	case isr::isrPageFault: {
-			cons->set_color(11);
-			cons->puts("\nPage Fault:\n");
-			cons->set_color();
-
-			cons->puts("       flags:");
-			if (regs->errorcode & 0x01) cons->puts(" present");
-			if (regs->errorcode & 0x02) cons->puts(" write");
-			if (regs->errorcode & 0x04) cons->puts(" user");
-			if (regs->errorcode & 0x08) cons->puts(" reserved");
-			if (regs->errorcode & 0x10) cons->puts(" ip");
-			cons->puts("\n");
-
-			uint64_t cr2 = 0;
+			uintptr_t cr2 = 0;
 			__asm__ __volatile__ ("mov %%cr2, %0" : "=r"(cr2));
-			print_reg("cr2", cr2);
 
-			print_reg("rsp", regs->user_rsp);
-			print_reg("rip", regs->rip);
+			if (!page_manager::get()->fault_handler(cr2)) {
+				cons->set_color(11);
+				cons->puts("\nPage Fault:\n");
+				cons->set_color();
 
-			cons->puts("\n");
-			//print_stacktrace(2);
+				cons->puts("       flags:");
+				if (regs->errorcode & 0x01) cons->puts(" present");
+				if (regs->errorcode & 0x02) cons->puts(" write");
+				if (regs->errorcode & 0x04) cons->puts(" user");
+				if (regs->errorcode & 0x08) cons->puts(" reserved");
+				if (regs->errorcode & 0x10) cons->puts(" ip");
+				cons->puts("\n");
+				print_regs(*regs);
+				print_stacktrace(2);
+				_halt();
+			}
 		}
-		_halt();
 		break;
 
-	case isr::isrTimer: {
-			scheduler &s = scheduler::get();
-			return_rsp = s.tick(return_rsp);
-		}
+	case isr::isrTimer:
+		scheduler::get().tick();
 		break;
 
 	case isr::isrLINT0:
@@ -198,14 +223,15 @@ isr_handler(uintptr_t return_rsp, cpu_state *regs)
 		_halt();
 		break;
 
-	case isr::isrSyscall: {
-			return_rsp = syscall_dispatch(return_rsp, *regs);
-		}
+	/*
+	case isr::isrSyscall:
+		syscall_dispatch(regs);
 		break;
+	*/
 
 	case isr::isrSpurious:
 		// No EOI for the spurious interrupt
-		return return_rsp;
+		return;
 
 	case isr::isrIgnore0:
 	case isr::isrIgnore1:
@@ -242,16 +268,14 @@ isr_handler(uintptr_t return_rsp, cpu_state *regs)
 				regs->interrupt, regs->errorcode);
 
 		print_regs(*regs);
-		//print_stacktrace(2);
+		print_stacktrace(2);
 		_halt();
 	}
 	*reinterpret_cast<uint32_t *>(0xffffff80fee000b0) = 0;
-
-	return return_rsp;
 }
 
-uintptr_t
-irq_handler(uintptr_t return_rsp, cpu_state *regs)
+void
+irq_handler(cpu_state *regs)
 {
 	console *cons = console::get();
 	uint8_t irq = get_irq(regs->interrupt);
@@ -265,11 +289,4 @@ irq_handler(uintptr_t return_rsp, cpu_state *regs)
 	}
 
 	*reinterpret_cast<uint32_t *>(0xffffff80fee000b0) = 0;
-	return return_rsp;
-}
-
-uintptr_t
-syscall_handler(uintptr_t return_rsp, cpu_state regs)
-{
-	return syscall_dispatch(return_rsp, regs);
 }

src/kernel/interrupts.s

@@ -3,39 +3,36 @@
 extern isr_handler
 global isr_handler_prelude
 isr_handler_prelude:
-	push_all_and_segments
+	push_all
+	check_swap_gs
 
 	mov rdi, rsp
 	mov rsi, rsp
 	call isr_handler
-	mov rsp, rax
-
-	pop_all_and_segments
-
-	add rsp, 16		; because the ISRs added err/num
-	sti
-	iretq
+	jmp isr_handler_return
 
 extern irq_handler
 global irq_handler_prelude
 irq_handler_prelude:
-	push_all_and_segments
+	push_all
+	check_swap_gs
 
 	mov rdi, rsp
 	mov rsi, rsp
 	call irq_handler
-	mov rsp, rax
+	; fall through to isr_handler_return
 
-	pop_all_and_segments
+global isr_handler_return
+isr_handler_return:
+	check_swap_gs
+	pop_all
 
 	add rsp, 16		; because the ISRs added err/num
-	sti
 	iretq
 
 %macro EMIT_ISR 2
 	global %1
 	%1:
-		cli
 		push 0
 		push %2
 		jmp isr_handler_prelude
@@ -44,7 +41,6 @@ irq_handler_prelude:
 %macro EMIT_EISR 2
 	global %1
 	%1:
-		cli
 		push %2
 		jmp isr_handler_prelude
 %endmacro
@@ -52,7 +48,6 @@ irq_handler_prelude:
 %macro EMIT_IRQ 2
 	global %1
 	%1:
-		cli
 		push 0
 		push %2
 		jmp irq_handler_prelude

src/kernel/loader.s

@@ -1,25 +1,22 @@
 %include "push_all.inc"
 
-extern load_process
+extern load_process_image
 
 global ramdisk_process_loader
 ramdisk_process_loader:
 
 	; create_process already pushed a cpu_state onto the stack for us, this
-	; acts both as the cpu_state parameter to load_process, and the saved
-	; state for the following iretq
-	;
-	; Additional parameters:
-	;  rax - the address of the program image
-	;  rbx - the size of the program image
-	;  rcx - the address of this process' process structure
-	mov rdi, rax
-	mov rsi, rbx
-	mov rdx, rcx
-	call load_process
+	; acts both as the cpu_state parameter to load_process_image, and the
+	; saved state for the following iretq
 
-	pop_all_and_segments
-	add rsp, 16		; because the ISRs add err/num
+	pop rdi ; the address of the program image
+	pop rsi ; the size of the program image
+	pop rdx ; the address of this process' process structure
+
+	call load_process_image
+
+	push rax ; load_process_image returns the process entrypoint
+
+	swapgs
 	iretq
 
-

src/kernel/log.cpp

@@ -1,107 +1,55 @@
-#include <type_traits>
-#include "kutil/assert.h"
 #include "kutil/memory.h"
 #include "console.h"
 #include "log.h"
+#include "scheduler.h"
 
 
-static const uint64_t default_enabled[] = {0x00, 0xff, 0xff, 0xff};
-static const uint8_t level_colors[] = {0x07, 0x0f, 0x0b, 0x09};
+static uint8_t log_buffer[0x10000];
+static log::logger g_logger(log_buffer, sizeof(log_buffer));
 
-static const char *levels[] = {"debug", " info", " warn", "error", "fatal"};
-static const char *areas[] = {
-	"boot  ",
-	"memory",
-	"apic  ",
-	"device",
-	"driver",
-	"file  ",
-	"task  ",
-	"paging",
+static const uint8_t level_colors[] = {0x07, 0x07, 0x0f, 0x0b, 0x09};
 
-	nullptr
-};
-
-
-log log::s_log;
-
-log::log() : m_cons(nullptr)
+static void
+output_log(log::area_t area, log::level severity, const char *message)
 {
-	kassert(0, "Invalid log constructor");
-}
-
-log::log(console *cons) :
-	m_cons(cons)
-{
-	const int num_levels = static_cast<int>(level::max) - 1;
-	for (int i = 0; i < num_levels; ++i)
-		m_enabled[i] = default_enabled[i];
+	auto *cons = console::get();
+	cons->set_color(level_colors[static_cast<int>(severity)]);
+	cons->printf("%7s %5s: %s\n",
+			g_logger.area_name(area),
+			g_logger.level_name(severity),
+			message);
+	cons->set_color();
 }
 
 void
-log::init(console *cons)
+logger_task()
 {
-	new (&s_log) log(cons);
-	log::info(logs::boot, "Logging system initialized.");
-}
+	uint8_t buffer[257];
+	auto *ent = reinterpret_cast<log::logger::entry *>(buffer);
+	auto *cons = console::get();
 
-static inline uint64_t
-bit_mask(logs area) { return 1 << static_cast<uint64_t>(area); }
+	g_logger.set_immediate(nullptr);
+	log::info(logs::task, "Starting kernel logger task");
 
-void
-log::enable(logs type, level at_level)
-{
-	using under_t = std::underlying_type<level>::type;
-	under_t at = static_cast<under_t>(at_level);
-	under_t max = sizeof(m_enabled) / sizeof(m_enabled[0]);
+	scheduler &s = scheduler::get();
 
-	for (under_t i = 0; i < max; ++i) {
-		if (i >= at)
-			s_log.m_enabled[i] |= bit_mask(type);
-		else
-			s_log.m_enabled[i] &= ~bit_mask(type);
+	while (true) {
+		if(g_logger.get_entry(buffer, sizeof(buffer))) {
+			buffer[ent->bytes] = 0;
+			cons->set_color(level_colors[static_cast<int>(ent->severity)]);
+			cons->printf("%7s %5s: %s\n",
+				g_logger.area_name(ent->area),
+				g_logger.level_name(ent->severity),
+				ent->message);
+			cons->set_color();
+		} else {
+			s.schedule();
+		}
 	}
 }
 
-template <>
-void
-log::trylog<log::level::fatal>(logs area, const char *fmt, ...)
+void logger_init()
 {
-	va_list args;
-	va_start(args, fmt);
-	s_log.output(level::fatal, area, fmt, args);
-	va_end(args);
-	while(1) __asm__ ("hlt");
+	new (&g_logger) log::logger(log_buffer, sizeof(log_buffer));
+	g_logger.set_immediate(output_log);
 }
-
-template <log::level L>
-void
-log::trylog(logs area, const char *fmt, ...)
-{
-	auto i = static_cast<std::underlying_type<level>::type>(L);
-	if ((s_log.m_enabled[i] & bit_mask(area)) == 0) return;
-
-	va_list args;
-	va_start(args, fmt);
-	s_log.output(L, area, fmt, args);
-	va_end(args);
-}
-
-void
-log::output(level severity, logs area, const char *fmt, va_list args)
-{
-	m_cons->set_color(level_colors[static_cast<int>(severity)]);
-
-	m_cons->printf("%s %s: ",
-			areas[static_cast<int>(area)],
-			levels[static_cast<int>(severity)]);
-	m_cons->vprintf(fmt, args);
-	m_cons->set_color();
-	m_cons->puts("\n");
-}
-
-const log::trylog_p log::debug = &trylog<level::debug>;
-const log::trylog_p log::info = &trylog<level::info>;
-const log::trylog_p log::warn = &trylog<level::warn>;
-const log::trylog_p log::error = &trylog<level::error>;
-const log::trylog_p log::fatal = &trylog<level::fatal>;

src/kernel/log.h

@@ -1,52 +1,9 @@
 #pragma once
-#include <stdarg.h>
-#include <stdint.h>
 
-class console;
+#include "kutil/logger.h"
 
+namespace log = kutil::log;
+namespace logs = kutil::logs;
 
-enum class logs
-{
-	boot,
-	memory,
-	apic,
-	device,
-	driver,
-	fs,
-	task,
-	paging,
-
-	max
-};
-
-
-class log
-{
-public:
-	enum class level {debug, info, warn, error, fatal, max};
-
-	static void init(console *cons);
-	static void enable(logs type, level at_level);
-
-	template <level L>
-	static void trylog(logs area, const char *fmt, ...);
-	using trylog_p = void (*)(logs area, const char *fmt, ...);
-
-	static const trylog_p debug;
-	static const trylog_p info;
-	static const trylog_p warn;
-	static const trylog_p error;
-	static const trylog_p fatal;
-
-private:
-	void output(level severity, logs area, const char *fmt, va_list args);
-
-	/// Bitmasks for what categories are enabled. fatal is
-	/// always enabled, so leave it out.
-	uint64_t m_enabled[static_cast<uint64_t>(level::max) - 1];
-	console *m_cons;
-
-	log();
-	log(console *cons);
-	static log s_log;
-};
+void logger_init();
+void logger_task();

src/kernel/main.cpp

@@ -11,16 +11,15 @@
 #include "gdt.h"
 #include "interrupts.h"
 #include "io.h"
-#include "kernel_data.h"
+#include "kernel_args.h"
 #include "log.h"
 #include "page_manager.h"
 #include "scheduler.h"
-#include "screen.h"
 #include "serial.h"
 #include "syscall.h"
 
 extern "C" {
-	void kernel_main(popcorn_data *header);
+	void kernel_main(kernel_args *header);
 	void *__bss_start, *__bss_end;
 }
 
@@ -37,37 +36,21 @@ init_console()
 	cons->set_color(0x08, 0x00);
 	cons->puts(GIT_VERSION " booting...\n");
 
-	log::init(cons);
-	log::enable(logs::apic, log::level::info);
-	log::enable(logs::device, log::level::info);
-
-	log::enable(logs::driver, log::level::debug);
-	log::enable(logs::memory, log::level::debug);
-	log::enable(logs::fs, log::level::debug);
-	log::enable(logs::task, log::level::debug);
-	log::enable(logs::boot, log::level::debug);
-	log::enable(logs::paging, log::level::debug);
+	logger_init();
 }
 
 void
-kernel_main(popcorn_data *header)
+kernel_main(kernel_args *header)
 {
-#ifdef DEBUG
-	// Run `waf configure --debug` to enable compiling with DEBUG turned on.
-	// Then attach to QEMU's gdb server and `set waiting = false` to start
-	// the kernel. This compensates for GDB's poor handling of QEMU going
-	// through the x86 PC startup and switching to 64 bit mode when you
-	// attach to qemu with the -S option.
-	bool waiting = true;
+	bool waiting = header && (header->flags && POPCORN_FLAG_DEBUG);
 	while (waiting);
-#endif
 
 	kutil::assert_set_callback(__kernel_assert);
 
 	gdt_init();
 	interrupts_init();
 
-	memory_initialize(
+	kutil::allocator &heap = memory_initialize(
 			header->scratch_pages,
 			header->memory_map,
 			header->memory_map_length,
@@ -88,7 +71,7 @@ kernel_main(popcorn_data *header)
 	log::debug(logs::boot, "ACPI root table is at: %016lx", header->acpi_table);
 	log::debug(logs::boot, "Runtime service is at: %016lx", header->runtime);
 
-	initrd::disk ird(header->initrd);
+	initrd::disk ird(header->initrd, heap);
 	log::info(logs::boot, "initrd loaded with %d files.", ird.files().count());
 	for (auto &f : ird.files())
 		log::info(logs::boot, "  %s%s (%d bytes).", f.executable() ? "*" : "", f.name(), f.size());
@@ -99,7 +82,7 @@ kernel_main(popcorn_data *header)
 	*/
 
 	device_manager *devices =
-		new (&device_manager::get()) device_manager(header->acpi_table);
+		new (&device_manager::get()) device_manager(header->acpi_table, heap);
 
 	interrupts_enable();
 
@@ -146,7 +129,9 @@ kernel_main(popcorn_data *header)
 	devices->get_lapic()->calibrate_timer();
 
 	syscall_enable();
-	scheduler *sched = new (&scheduler::get()) scheduler(devices->get_lapic());
+	scheduler *sched = new (&scheduler::get()) scheduler(devices->get_lapic(), heap);
+
+	sched->create_kernel_task(-1, logger_task);
 
 	for (auto &f : ird.files()) {
 		if (f.executable())

src/kernel/memory_bootstrap.cpp

@@ -2,72 +2,26 @@
 #include <utility>
 #include "kutil/address_manager.h"
 #include "kutil/assert.h"
-#include "kutil/frame_allocator.h"
-#include "kutil/heap_manager.h"
+#include "kutil/heap_allocator.h"
+#include "frame_allocator.h"
 #include "io.h"
 #include "log.h"
 #include "page_manager.h"
 
-using kutil::frame_block;
-using kutil::frame_block_flags;
-using kutil::frame_block_list;
 using memory::frame_size;
+using memory::kernel_max_heap;
 using memory::kernel_offset;
 using memory::page_offset;
 
 static const unsigned ident_page_flags = 0xb;
 
-kutil::frame_allocator g_frame_allocator;
 kutil::address_manager g_kernel_address_manager;
-kutil::heap_manager g_kernel_heap_manager;
+kutil::heap_allocator g_kernel_heap;
 
-void * mm_grow_callback(size_t length)
-{
-	kassert(length % frame_size == 0,
-			"Heap manager requested a fractional page.");
-
-	size_t pages = length / frame_size;
-	log::info(logs::memory, "Heap manager growing heap by %d pages.", pages);
-
-	uintptr_t addr = g_kernel_address_manager.allocate(length);
-	g_page_manager.map_pages(addr, pages);
-
-	return reinterpret_cast<void *>(addr);
-}
-
-
-namespace {
-	// Page-by-page initial allocator for the initial frame_block allocator
-	struct page_consumer
-	{
-		page_consumer(uintptr_t start, unsigned count, unsigned used = 0) :
-			current(start + used * frame_size),
-			used(used),
-			max(count) {}
-
-		void * get_page() {
-			kassert(used++ < max, "page_consumer ran out of pages");
-			void *retval = reinterpret_cast<void *>(current);
-			current += frame_size;
-			return retval;
-		}
-
-		void * operator()(size_t size) {
-			kassert(size == frame_size, "page_consumer used with non-page size!");
-			return get_page();
-		}
-
-		unsigned left() const { return max - used; }
-
-		uintptr_t current;
-		unsigned used, max;
-	};
-
-	using block_allocator =
-		kutil::slab_allocator<kutil::frame_block, page_consumer &>;
-	using region_allocator =
-		kutil::slab_allocator<kutil::buddy_region, page_consumer &>;
-}
+void * operator new(size_t size)           { return g_kernel_heap.allocate(size); }
+void * operator new [] (size_t size)       { return g_kernel_heap.allocate(size); }
+void operator delete (void *p) noexcept    { return g_kernel_heap.free(p); }
+void operator delete [] (void *p) noexcept { return g_kernel_heap.free(p); }
 
 enum class efi_memory_type : uint32_t
 {
@@ -107,100 +61,114 @@ struct efi_memory_descriptor
 	uint64_t flags;
 };
 
-static const efi_memory_descriptor *
-desc_incr(const efi_memory_descriptor *d, size_t desc_length)
+struct memory_map
 {
-	return reinterpret_cast<const efi_memory_descriptor *>(
-			reinterpret_cast<const uint8_t *>(d) + desc_length);
-}
+	memory_map(const void *efi_map, size_t map_length, size_t desc_length) :
+		efi_map(efi_map), map_length(map_length), desc_length(desc_length) {}
 
-void
-gather_block_lists(
-		block_allocator &allocator,
-		frame_block_list &used,
-		frame_block_list &free,
-		const void *memory_map,
-		size_t map_length,
-		size_t desc_length)
-{
-	efi_memory_descriptor const *desc = reinterpret_cast<efi_memory_descriptor const *>(memory_map);
-	efi_memory_descriptor const *end = desc_incr(desc, map_length);
+	class iterator
+	{
+	public:
+		iterator(const memory_map &map, efi_memory_descriptor const *item) :
+			map(map), item(item) {}
 
-	while (desc < end) {
-		auto *block = allocator.pop();
-		block->address = desc->physical_start;
-		block->count = desc->pages;
-		bool block_used;
-
-		switch (desc->type) {
-		case efi_memory_type::loader_code:
-		case efi_memory_type::loader_data:
-			block_used = true;
-			block->flags = frame_block_flags::pending_free;
-			break;
-
-		case efi_memory_type::boot_services_code:
-		case efi_memory_type::boot_services_data:
-		case efi_memory_type::available:
-			block_used = false;
-			break;
-
-		case efi_memory_type::acpi_reclaim:
-			block_used = true;
-			block->flags =
-				frame_block_flags::acpi_wait |
-				frame_block_flags::map_ident;
-			break;
-
-		case efi_memory_type::persistent:
-			block_used = false;
-			block->flags = frame_block_flags::nonvolatile;
-			break;
-
-		case efi_memory_type::popcorn_kernel:
-			block_used = true;
-			block->flags =
-				frame_block_flags::permanent |
-				frame_block_flags::map_kernel;
-			break;
-
-		case efi_memory_type::popcorn_data:
-		case efi_memory_type::popcorn_initrd:
-			block_used = true;
-			block->flags =
-				frame_block_flags::pending_free |
-				frame_block_flags::map_kernel;
-			break;
-
-		case efi_memory_type::popcorn_scratch:
-			block_used = true;
-			block->flags = frame_block_flags::map_offset;
-			break;
-
-		default:
-			block_used = true;
-			block->flags = frame_block_flags::permanent;
-			break;
+		inline efi_memory_descriptor const * operator*() const { return item; }
+		inline bool operator!=(const iterator &other) { return item != other.item; }
+		inline iterator & operator++() {
+			item = kutil::offset_pointer(item, map.desc_length);
+			return *this;
 		}
 
-		if (block_used) 
-			used.push_back(block);
-		else
-			free.push_back(block);
+	private:
+		const memory_map &map;
+		efi_memory_descriptor const *item;
+	};
 
-		desc = desc_incr(desc, desc_length);
+	iterator begin() const {
+		return iterator(*this, reinterpret_cast<efi_memory_descriptor const *>(efi_map));
 	}
-}
 
-void
+	iterator end() const {
+		const void *end = kutil::offset_pointer(efi_map, map_length);
+		return iterator(*this, reinterpret_cast<efi_memory_descriptor const *>(end));
+	}
+
+	const void *efi_map;
+	size_t map_length;
+	size_t desc_length;
+};
+
+class memory_bootstrap
+{
+public:
+	memory_bootstrap(const void *memory_map, size_t map_length, size_t desc_length) :
+		map(memory_map, map_length, desc_length) {}
+
+	void add_free_frames(frame_allocator &fa) {
+		for (auto *desc : map) {
+			if (desc->type == efi_memory_type::loader_code ||
+				desc->type == efi_memory_type::loader_data ||
+				desc->type == efi_memory_type::boot_services_code ||
+				desc->type == efi_memory_type::boot_services_data ||
+				desc->type == efi_memory_type::available)
+			{
+				fa.free(desc->physical_start, desc->pages);
+			}
+		}
+	}
+
+	void add_used_frames(kutil::address_manager &am) {
+		for (auto *desc : map) {
+			if (desc->type == efi_memory_type::popcorn_data ||
+				desc->type == efi_memory_type::popcorn_initrd)
+			{
+				uintptr_t virt_addr = desc->physical_start + kernel_offset;
+				am.mark(virt_addr, desc->pages * frame_size);
+			}
+			else if (desc->type == efi_memory_type::popcorn_kernel)
+			{
+				uintptr_t virt_addr = desc->physical_start + kernel_offset;
+				am.mark_permanent(virt_addr, desc->pages * frame_size);
+			}
+		}
+	}
+
+	void page_in_kernel(page_manager &pm, page_table *pml4) {
+		for (auto *desc : map) {
+			if (desc->type == efi_memory_type::popcorn_kernel ||
+				desc->type == efi_memory_type::popcorn_data ||
+				desc->type == efi_memory_type::popcorn_initrd)
+			{
+				uintptr_t virt_addr = desc->physical_start + kernel_offset;
+				pm.page_in(pml4, desc->physical_start, virt_addr, desc->pages);
+			}
+
+			if (desc->type == efi_memory_type::acpi_reclaim) {
+				pm.page_in(pml4, desc->physical_start, desc->physical_start, desc->pages);
+			}
+		}
+
+		// Put our new PML4 into CR3 to start using it
+		page_manager::set_pml4(pml4);
+		pm.m_kernel_pml4 = pml4;
+	}
+
+private:
+	const memory_map map;
+};
+
+kutil::allocator &
 memory_initialize(uint16_t scratch_pages, const void *memory_map, size_t map_length, size_t desc_length)
 {
 	// make sure the options we want in CR4 are set
 	uint64_t cr4;
 	__asm__ __volatile__ ( "mov %%cr4, %0" : "=r" (cr4) );
-	cr4 |= 0x00080; // Enable global pages
-	cr4 |= 0x00200; // Enable FXSAVE/FXRSTOR
-	cr4 |= 0x20000; // Enable PCIDs
+	cr4 |=
+		0x000080 | // Enable global pages
+		0x000200 | // Enable FXSAVE/FXRSTOR
+		0x010000 | // Enable FSGSBASE
+		0x020000 | // Enable PCIDs
+		0;
 	__asm__ __volatile__ ( "mov %0, %%cr4" :: "r" (cr4) );
 
 	// The bootloader reserved "scratch_pages" pages for page tables and
@@ -224,81 +192,51 @@ memory_initialize(uint16_t scratch_pages, const void *memory_map, size_t map_len
 	__sync_synchronize();
 	io_wait();
 
-	// We now have pages starting at "scratch_virt" to bootstrap ourselves. Start by
-	// taking inventory of free pages.
 	uintptr_t scratch_virt = scratch_phys + page_offset;
-	uint64_t used_pages = 2; // starts with PML4 + offset PDP
-	page_consumer allocator(scratch_virt, scratch_pages, used_pages);
+	memory_bootstrap bootstrap {memory_map, map_length, desc_length};
 
-	block_allocator block_slab(frame_size, allocator);
-	frame_block_list used;
-	frame_block_list free;
+	// Now tell the frame allocator what's free
+	frame_allocator *fa = new (&g_frame_allocator) frame_allocator;
+	bootstrap.add_free_frames(*fa);
 
-	gather_block_lists(block_slab, used, free, memory_map, map_length, desc_length);
-	block_slab.allocate(); // Make sure we have extra
+	// Build an initial address manager that we'll copy into the real
+	// address manager later (so that we can use a raw allocator now)
+	kutil::allocator &alloc = fa->raw_allocator();
+	kutil::address_manager init_am(alloc);
+	init_am.add_regions(kernel_offset, page_offset - kernel_offset);
+	bootstrap.add_used_frames(init_am);
 
-	// Now go back through these lists and consolidate
-	block_slab.append(frame_block::consolidate(free));
+	// Add the heap into the address manager
+	uintptr_t heap_start = page_offset - kernel_max_heap;
+	init_am.mark(heap_start, kernel_max_heap);
 
-	region_allocator region_slab(frame_size, allocator);
-	region_slab.allocate(); // Allocate some buddy regions for the address_manager
+	kutil::allocator *heap_alloc =
+		new (&g_kernel_heap) kutil::heap_allocator(heap_start, kernel_max_heap);
 
+	// Copy everything into the real address manager
 	kutil::address_manager *am =
-		new (&g_kernel_address_manager) kutil::address_manager(std::move(region_slab));
+		new (&g_kernel_address_manager) kutil::address_manager(
+				std::move(init_am), *heap_alloc);
 
-	am->add_regions(kernel_offset, page_offset - kernel_offset);
+	// Create the page manager
+	page_manager *pm = new (&g_page_manager) page_manager(*fa, *am);
 
-	// Finally, build an acutal set of kernel page tables that just contains
+	// Give the frame_allocator back the rest of the scratch pages
+	fa->free(scratch_phys + (3 * frame_size), scratch_pages - 3);
+
+	// Finally, build an acutal set of kernel page tables where we'll only add
 	// what the kernel actually has mapped, but making everything writable
 	// (especially the page tables themselves)
-	page_table *pml4 = reinterpret_cast<page_table *>(allocator.get_page());
+	page_table *pml4 = &tables[2];
+	pml4 = kutil::offset_pointer(pml4, page_offset);
+
 	kutil::memset(pml4, 0, sizeof(page_table));
 	pml4->entries[511] = reinterpret_cast<uintptr_t>(id_pdp) | 0x10b;
 
-	kutil::frame_allocator *fa =
-		new (&g_frame_allocator) kutil::frame_allocator(std::move(block_slab));
-	page_manager *pm = new (&g_page_manager) page_manager(*fa, *am);
+	bootstrap.page_in_kernel(*pm, pml4);
 
-	// Give the rest to the page_manager's cache for use in page_in
-	pm->free_table_pages(
-			reinterpret_cast<void *>(allocator.current),
-			allocator.left());
+	// Reclaim the old PML4
+	fa->free(scratch_phys, 1);
 
-	for (auto *block : used) {
-		uintptr_t virt_addr = 0;
-
-		switch (block->flags & frame_block_flags::map_mask) {
-			case frame_block_flags::map_ident:
-				virt_addr = block->address;
-				break;
-
-			case frame_block_flags::map_kernel:
-				virt_addr = block->address + kernel_offset;
-				if (block->flags && frame_block_flags::permanent)
-					am->mark_permanent(virt_addr, block->count * frame_size);
-				else
-					am->mark(virt_addr, block->count * frame_size);
-				break;
-
-			default:
-				break;
-		}
-
-		block->flags -= frame_block_flags::map_mask;
-		if (virt_addr)
-			pm->page_in(pml4, block->address, virt_addr, block->count);
-	}
-
-	fa->init(std::move(free), std::move(used));
-
-	// Put our new PML4 into CR3 to start using it
-	page_manager::set_pml4(pml4);
-	pm->m_kernel_pml4 = pml4;
-
-	// Give the old pml4 back to the page_manager to recycle
-	pm->free_table_pages(reinterpret_cast<void *>(scratch_virt), 1);
-
-	// Set the heap manager
-	new (&g_kernel_heap_manager) kutil::heap_manager(mm_grow_callback);
-	kutil::setup::set_heap(&g_kernel_heap_manager);
+	return *heap_alloc;
 }

src/kernel/page_manager.cpp

@@ -2,14 +2,15 @@
 
 #include "kutil/assert.h"
 #include "console.h"
+#include "io.h"
 #include "log.h"
 #include "page_manager.h"
 
 using memory::frame_size;
 using memory::kernel_offset;
 using memory::page_offset;
+using memory::page_mappable;
 
-extern kutil::frame_allocator g_frame_allocator;
 extern kutil::address_manager g_kernel_address_manager;
 page_manager g_page_manager(
 	g_frame_allocator,
@@ -38,7 +39,7 @@ struct free_page_header
 
 
 page_manager::page_manager(
-		kutil::frame_allocator &frames,
+		frame_allocator &frames,
 		kutil::address_manager &addrs) :
 	m_page_cache(nullptr),
 	m_frames(frames),
@@ -68,28 +69,60 @@ page_manager::create_process_map()
 uintptr_t
 page_manager::copy_page(uintptr_t orig)
 {
-	uintptr_t virt = m_addrs.allocate(2 * frame_size);
-	uintptr_t copy = 0;
+	bool paged_orig = false;
+	bool paged_copy = false;
 
+	uintptr_t orig_virt;
+
+	if (page_mappable(orig)) {
+		orig_virt = orig + page_offset;
+	} else {
+		orig_virt = m_addrs.allocate(frame_size);
+		page_in(get_pml4(), orig, orig_virt, 1);
+		paged_orig = true;
+	}
+
+	uintptr_t copy = 0;
+	uintptr_t copy_virt;
 	size_t n = m_frames.allocate(1, &copy);
 	kassert(n, "copy_page could not allocate page");
 
-	page_in(get_pml4(), orig, virt, 1);
-	page_in(get_pml4(), copy, virt + frame_size, 1);
+	if (page_mappable(copy)) {
+		copy_virt = copy + page_offset;
+	} else {
+		copy_virt = m_addrs.allocate(frame_size);
+		page_in(get_pml4(), copy, copy_virt, 1);
+		paged_copy = true;
+	}
+
+	// TODO: multiple page copies at a time, so that we don't have to keep
+	// paying this mapping penalty
+	if (paged_orig || paged_copy) {
+		set_pml4(get_pml4());
+		__sync_synchronize();
+		io_wait();
+	}
 
 	kutil::memcpy(
-			reinterpret_cast<void *>(virt + frame_size),
-			reinterpret_cast<void *>(virt),
+			reinterpret_cast<void *>(copy_virt),
+			reinterpret_cast<void *>(orig_virt),
 			frame_size);
 
-	page_out(get_pml4(), virt, 2);
+	if (paged_orig) {
+		page_out(get_pml4(), orig_virt, 1);
+		m_addrs.free(orig_virt);
+	}
+
+	if (paged_copy) {
+		page_out(get_pml4(), copy_virt, 1);
+		m_addrs.free(copy_virt);
+	}
 
-	m_addrs.free(virt);
 	return copy;
 }
 
 page_table *
-page_manager::copy_table(page_table *from, page_table::level lvl)
+page_manager::copy_table(page_table *from, page_table::level lvl, page_table_indices index)
 {
 	page_table *to = get_table_page();
 	log::debug(logs::paging, "Page manager copying level %d table at %016lx to %016lx.", lvl, from, to);
@@ -105,12 +138,17 @@ page_manager::copy_table(page_table *from, page_table::level lvl)
 			512;
 
 	unsigned pages_copied = 0;
+	uintptr_t from_addr = 0;
+	uintptr_t to_addr = 0;
+
 	for (int i = 0; i < max; ++i) {
 		if (!from->is_present(i)) {
 			to->entries[i] = 0;
 			continue;
 		}
 
+		index[lvl] = i;
+
 		bool is_page =
 			lvl == page_table::level::pt ||
 			from->is_large_page(lvl, i);
@@ -119,17 +157,20 @@ page_manager::copy_table(page_table *from, page_table::level lvl)
 			uint16_t flags = from->entries[i] &  0xfffull;
 			uintptr_t orig = from->entries[i] & ~0xfffull;
 			to->entries[i] = copy_page(orig) | flags;
-			pages_copied++;
+			if (!pages_copied++)
+				from_addr = index.addr();
+			to_addr = index.addr();
 		} else {
 			uint16_t flags = 0;
 			page_table *next_from = from->get(i, &flags);
-			page_table *next_to = copy_table(next_from, page_table::deeper(lvl));
+			page_table *next_to = copy_table(next_from, page_table::deeper(lvl), index);
 			to->set(i, next_to, flags);
 		}
 	}
 
 	if (pages_copied)
-		log::debug(logs::paging, "   copied %3u pages", pages_copied);
+		log::debug(logs::paging, "   copied %3u pages %016lx - %016lx",
+			pages_copied, from_addr, to_addr + frame_size);
 
 	return to;
 }
@@ -137,13 +178,20 @@ page_manager::copy_table(page_table *from, page_table::level lvl)
 void
 page_manager::delete_process_map(page_table *pml4)
 {
+	bool was_pml4 = (pml4 == get_pml4());
+	if (was_pml4)
+		set_pml4(m_kernel_pml4);
+
+	log::info(logs::paging, "Deleting process pml4 at %016lx%s",
+			pml4, was_pml4 ? " (was current)" : "");
+
 	unmap_table(pml4, page_table::level::pml4, true);
 }
 
 void
 page_manager::map_offset_pointer(void **pointer, size_t length)
 {
-	log::info(logs::paging, "Mapping offset pointer region at %016lx size 0x%lx", *pointer, length);
+	log::debug(logs::paging, "Mapping offset pointer region at %016lx size 0x%lx", *pointer, length);
 	*pointer = kutil::offset_pointer(*pointer, page_offset);
 }
 
@@ -178,6 +226,7 @@ page_manager::get_table_page()
 
 	free_page_header *page = m_page_cache;
 	m_page_cache = page->next;
+
 	return reinterpret_cast<page_table *>(page);
 }
 
@@ -197,6 +246,11 @@ page_manager::free_table_pages(void *pages, size_t count)
 void *
 page_manager::map_pages(uintptr_t address, size_t count, bool user, page_table *pml4)
 {
+	if (!address) {
+		kassert(!user, "Cannot call map_pages with 0 address for user mapping");
+		address = m_addrs.allocate(count * frame_size);
+	}
+
 	void *ret = reinterpret_cast<void *>(address);
 	if (!pml4) pml4 = get_pml4();
 
@@ -217,7 +271,7 @@ page_manager::map_pages(uintptr_t address, size_t count, bool user, page_table *
 }
 
 void
-page_manager::unmap_table(page_table *table, page_table::level lvl, bool free)
+page_manager::unmap_table(page_table *table, page_table::level lvl, bool free, page_table_indices index)
 {
 	const int max =
 		lvl == page_table::level::pml4 ?
@@ -225,6 +279,7 @@ page_manager::unmap_table(page_table *table, page_table::level lvl, bool free)
 			512;
 
 	uintptr_t free_start = 0;
+	uintptr_t free_start_virt = 0;
 	uintptr_t free_count = 0;
 
 	size_t size =
@@ -236,34 +291,78 @@ page_manager::unmap_table(page_table *table, page_table::level lvl, bool free)
 	for (int i = 0; i < max; ++i) {
 		if (!table->is_present(i)) continue;
 
+		index[lvl] = i;
+
 		bool is_page =
 			lvl == page_table::level::pt ||
 			table->is_large_page(lvl, i);
 
 		if (is_page) {
 			uintptr_t frame = table->entries[i] & ~0xfffull;
-			if (!free_count || free_start != frame + free_count * size) {
-				if (free_count && free)
-					m_frames.free(free_start, free_count * size / frame_size);
-				free_start = frame;
-				free_count = 1;
+			if (!free_count || frame != free_start + free_count * size) {
+				if (free_count && free) {
+					log::debug(logs::paging,
+							"  freeing v:%016lx-%016lx p:%016lx-%016lx",
+							free_start_virt, free_start_virt + free_count * frame_size,
+							free_start, free_start + free_count * frame_size);
+
+					m_frames.free(free_start, (free_count * size) / frame_size);
+					free_count = 0;
+				}
+
+				if (!free_count) {
+					free_start = frame;
+					free_start_virt = index.addr();
+				}
 			}
+
+			free_count += 1;
 		} else {
 			page_table *next = table->get(i);
-			unmap_table(next, page_table::deeper(lvl), free);
+			unmap_table(next, page_table::deeper(lvl), free, index);
 		}
 	}
 
-	if (free_count && free)
-		m_frames.free(free_start, free_count * size / frame_size);
+	if (free_count && free) {
+		log::debug(logs::paging,
+				"  freeing v:%016lx-%016lx p:%016lx-%016lx",
+				free_start_virt, free_start_virt + free_count * frame_size,
+				free_start, free_start + free_count * frame_size);
+
+		m_frames.free(free_start, (free_count * size) / frame_size);
+	}
 	free_table_pages(table, 1);
+
+	log::debug(logs::paging, "Unmapped%s lv %d table at %016lx",
+			free ? " (and freed)" : "", lvl, table);
 }
 
 void
 page_manager::unmap_pages(void* address, size_t count, page_table *pml4)
 {
-	if (!pml4) pml4 = get_pml4();
-	page_out(pml4, reinterpret_cast<uintptr_t>(address), count, true);
+	if (!pml4)
+		pml4 = get_pml4();
+
+	uintptr_t iaddr = reinterpret_cast<uintptr_t>(address);
+
+	page_out(pml4, iaddr, count, true);
+	if (iaddr >= kernel_offset) {
+		// TODO
+		// m_addrs.free(address, count);
+	}
+}
+
+bool
+page_manager::fault_handler(uintptr_t addr)
+{
+	if (!m_addrs.contains(addr))
+		return false;
+
+	uintptr_t page = addr & ~0xfffull;
+	bool user = addr < kernel_offset;
+	map_pages(page, 1, user);
+
+	return true;
 }
 
 void
@@ -279,8 +378,10 @@ page_manager::check_needs_page(page_table *table, unsigned index, bool user)
 void
 page_manager::page_in(page_table *pml4, uintptr_t phys_addr, uintptr_t virt_addr, size_t count, bool user, bool large)
 {
+	/*
 	log::debug(logs::paging, "page_in for table %016lx p:%016lx v:%016lx c:%4d u:%d l:%d",
 			pml4, phys_addr, virt_addr, count, user, large);
+	*/
 
 	page_table_indices idx{virt_addr};
 	page_table *tables[4] = {pml4, nullptr, nullptr, nullptr};

src/kernel/page_manager.h

@@ -7,9 +7,9 @@
 
 #include "kutil/address_manager.h"
 #include "kutil/enum_bitfields.h"
-#include "kutil/frame_allocator.h"
 #include "kutil/linked_list.h"
 #include "kutil/slab_allocator.h"
+#include "frame_allocator.h"
 #include "kernel_memory.h"
 #include "page_table.h"
 
@@ -20,7 +20,7 @@ class page_manager
 {
 public:
 	page_manager(
-		kutil::frame_allocator &frames,
+		frame_allocator &frames,
 		kutil::address_manager &addrs);
 
 	/// Helper to get the number of pages needed for a given number of bytes.
@@ -44,8 +44,9 @@ public:
 	/// \arg pml4  A pointer to the PML4 table to install.
 	static inline void set_pml4(page_table *pml4)
 	{
-		uintptr_t p = reinterpret_cast<uintptr_t>(pml4) - memory::page_offset;
-		__asm__ __volatile__ ( "mov %0, %%cr3" :: "r" (p & ~0xfffull) );
+		constexpr uint64_t phys_mask = ~memory::page_offset & ~0xfffull;
+		uintptr_t p = reinterpret_cast<uintptr_t>(pml4) & phys_mask;
+		__asm__ __volatile__ ( "mov %0, %%cr3" :: "r" (p) );
 	}
 
 	/// Allocate but don't switch to a new PML4 table. This table
@@ -62,10 +63,12 @@ public:
 	/// \arg lvl   Level of the given tables (default is PML4)
 	/// \returns   The new page table
 	page_table * copy_table(page_table *from,
-			page_table::level lvl = page_table::level::pml4);
+			page_table::level lvl = page_table::level::pml4,
+			page_table_indices index = {});
 
 	/// Allocate and map pages into virtual memory.
-	/// \arg address  The virtual address at which to map the pages
+	/// \arg address  The virtual address at which to map the pages, or zero
+	///               for any free kernel space.
 	/// \arg count    The number of pages to map
 	/// \arg user     True is this memory is user-accessible
 	/// \arg pml4     The pml4 to map into - null for the current one
@@ -108,6 +111,14 @@ public:
 	/// \returns  A pointer to the system page manager
 	static page_manager * get();
 
+	/// Get a pointer to the kernel's PML4
+	inline page_table * get_kernel_pml4() { return m_kernel_pml4; }
+
+	/// Attempt to handle a page fault.
+	/// \arg addr  Address that triggered the fault
+	/// \returns   True if the fault was handled
+	bool fault_handler(uintptr_t addr);
+
 private:
 	/// Copy a physical page
 	/// \arg orig  Physical address of the page to copy
@@ -158,15 +169,16 @@ private:
 			bool free = false);
 
 	/// Low-level routine for unmapping an entire table of memory at once
-	void unmap_table(page_table *table, page_table::level lvl, bool free);
+	void unmap_table(page_table *table, page_table::level lvl, bool free,
+			page_table_indices index = {});
 
 	page_table *m_kernel_pml4; ///< The PML4 of just kernel pages
 	free_page_header *m_page_cache; ///< Cache of free pages to use for tables
 
-	kutil::frame_allocator &m_frames;
+	frame_allocator &m_frames;
 	kutil::address_manager &m_addrs;
 
-	friend void memory_initialize(uint16_t, const void *, size_t, size_t);
+	friend class memory_bootstrap;
 	page_manager(const page_manager &) = delete;
 };
 
@@ -199,4 +211,8 @@ page_table_align(T p)
 
 
 /// Bootstrap the memory managers.
-void memory_initialize(uint16_t scratch_pages, const void *memory_map, size_t map_length, size_t desc_length);
+kutil::allocator & memory_initialize(
+		uint16_t scratch_pages,
+		const void *memory_map,
+		size_t map_length,
+		size_t desc_length);

src/kernel/process.cpp

@@ -1,11 +1,23 @@
+#include "kutil/heap_allocator.h"
 #include "cpu.h"
+#include "debug.h"
 #include "log.h"
 #include "process.h"
 #include "scheduler.h"
 
+extern "C" void task_fork_return_thunk();
+extern kutil::heap_allocator g_kernel_heap; // TODO: this is a bad hack to get access to the heap
+
+void
+process::exit(uint32_t code)
+{
+	return_code = code;
+	flags -= process_flags::running;
+	page_manager::get()->delete_process_map(pml4);
+}
 
 pid_t
-process::fork(uintptr_t in_rsp)
+process::fork()
 {
 	auto &sched = scheduler::get();
 	auto *child = sched.create_process();
@@ -18,40 +30,79 @@ process::fork(uintptr_t in_rsp)
 
 	sched.m_runlists[child->priority].push_back(child);
 
-	child->rsp = in_rsp;
-
 	child->pml4 = page_manager::get()->copy_table(pml4);
 	kassert(child->pml4, "process::fork() got null pml4");
 
-	child->setup_kernel_stack(kernel_stack_size, kernel_stack);
-	child->rsp = child->kernel_stack + (in_rsp - kernel_stack);
+	child->rsp3 = bsp_cpu_data.rsp3;
+	child->setup_kernel_stack();
 
-	log::debug(logs::task, "Copied process %d to %d, new PML4 %016lx.",
-			pid, child->pid, child->pml4);
-	log::debug(logs::task, "  copied stack %016lx to %016lx, rsp %016lx to %016lx.",
-			kernel_stack, child->kernel_stack, in_rsp, child->rsp);
+	log::debug(logs::task, "Copied process %d to %d",
+			pid, child->pid);
 
-	// Add in the faked fork return value
-	cpu_state *regs = reinterpret_cast<cpu_state *>(child->rsp);
-	regs->rax = 0;
+	log::debug(logs::task, "    PML4 %016lx", child->pml4);
+	log::debug(logs::task, "    RSP3 %016lx", child->rsp3);
+	log::debug(logs::task, "    RSP0 %016lx", child->rsp0);
+
+	// Initialize a new empty stack with a fake saved state
+	// for returning out of syscall_handler_prelude
+	size_t ret_seg_size = sizeof(uintptr_t) * 8;
+	child->rsp -= ret_seg_size;
+
+	void *this_ret_seg =
+		reinterpret_cast<void*>(rsp0 - ret_seg_size);
+	void *child_ret_seg =
+		reinterpret_cast<void*>(child->rsp);
+	kutil::memcpy(child_ret_seg, this_ret_seg, ret_seg_size);
+
+	child->add_fake_task_return(
+		reinterpret_cast<uintptr_t>(task_fork_return_thunk));
+
+	log::debug(logs::task, "    RSP  %016lx", child->rsp);
 
 	return child->pid;
 }
 
 void *
-process::setup_kernel_stack(size_t size, uintptr_t orig)
+process::setup_kernel_stack()
 {
-	void *stack0 = kutil::malloc(size);
+	constexpr unsigned null_frame_entries = 2;
+	constexpr size_t null_frame_size = null_frame_entries * sizeof(uint64_t);
 
-	if (orig)
-		kutil::memcpy(stack0, reinterpret_cast<void*>(orig), size);
-	else
-		kutil::memset(stack0, 0, size);
+	void *stack_bottom = g_kernel_heap.allocate(initial_stack_size);
+	kutil::memset(stack_bottom, 0, initial_stack_size);
 
-	kernel_stack_size = size;
-	kernel_stack = reinterpret_cast<uintptr_t>(stack0);
+	log::debug(logs::memory, "Created kernel stack at %016lx size 0x%lx",
+			stack_bottom, initial_stack_size);
 
-	return stack0;
+	void *stack_top =
+		kutil::offset_pointer(stack_bottom,
+				initial_stack_size - null_frame_size);
+
+	uint64_t *null_frame = reinterpret_cast<uint64_t*>(stack_top);
+	for (unsigned i = 0; i < null_frame_entries; ++i)
+		null_frame[i] = 0;
+
+	kernel_stack_size = initial_stack_size;
+	kernel_stack = reinterpret_cast<uintptr_t>(stack_bottom);
+	rsp0 = reinterpret_cast<uintptr_t>(stack_top);
+	rsp = rsp0;
+
+	return stack_top;
+}
+
+void
+process::add_fake_task_return(uintptr_t rip)
+{
+	rsp -= sizeof(uintptr_t) * 7;
+	uintptr_t *stack = reinterpret_cast<uintptr_t*>(rsp);
+
+	stack[6] = rip;        // return rip
+	stack[5] = rsp0;       // rbp
+	stack[4] = 0xbbbbbbbb; // rbx
+	stack[3] = 0x12121212; // r12
+	stack[2] = 0x13131313; // r13
+	stack[1] = 0x14141414; // r14
+	stack[0] = 0x15151515; // r15
 }
 
 bool

src/kernel/process.h

@@ -7,7 +7,8 @@
 #include "kutil/linked_list.h"
 #include "page_manager.h"
 
-typedef uint32_t pid_t;
+typedef int32_t pid_t;
+struct cpu_state;
 
 
 enum class process_flags : uint32_t
@@ -32,9 +33,20 @@ enum class process_wait : uint8_t
 	receive
 };
 
-/// A process
+/// A process.
+/// 
 struct process
 {
+	static const size_t initial_stack_size = 0x1000;
+
+	// Fields used by assembly routines go first.  If you change any of these,
+	// be sure to change the assembly definitions in 'tasking.inc'
+	uintptr_t rsp;
+	uintptr_t rsp0;
+	uintptr_t rsp3;
+	page_table *pml4;
+	// End of assembly fields
+
 	pid_t pid;
 	pid_t ppid;
 
@@ -51,17 +63,17 @@ struct process
 
 	uint32_t reserved1;
 
-	uintptr_t rsp;
-	page_table *pml4;
-
 	uintptr_t kernel_stack;
 	size_t kernel_stack_size;
 
+	/// Terminate this process.
+	/// \arg code   The return code to exit with.
+	void exit(unsigned code);
+
 	/// Copy this process.
-	/// \arg in_rsp The RSP of the calling process
 	/// \returns    Returns the child's pid to the parent, and
 	///             0 to the child.
-	pid_t fork(uint64_t in_rsp);
+	pid_t fork();
 
 	/// Unready this process until it gets a signal
 	/// \arg sigmask  A bitfield of signals to wake on
@@ -118,13 +130,15 @@ struct process
 private:
 	friend class scheduler;
 
-	/// Set up a new kernel stack for this process, optionally copying the
-	/// given stack. Sets the kernel stack on the process object, but also
-	/// returns it.
-	/// \arg size  Size of the stack to allocate
-	/// \arg orig  Address of a stack to copy, or 0 for no copying.
-	/// \returns   The address of the new stack as a pointer
-	void * setup_kernel_stack(size_t size, uintptr_t orig);
+	/// Set up a new empty kernel stack for this process. Sets rsp0 on this
+	/// process object, but also returns it.
+	/// \returns   The new rsp0 as a pointer
+	void * setup_kernel_stack();
+
+	/// Initialize this process' kenrel stack with a fake return segment for
+	/// returning out of task_switch.
+	/// \arg rip  The rip to return to
+	void add_fake_task_return(uintptr_t rip);
 };
 
 using process_list = kutil::linked_list<process>;

src/kernel/push_all.inc

@@ -1,48 +1,82 @@
-%macro push_all_and_segments 0
-	push rax
-	push rcx
-	push rdx
-	push rbx
-	push rbp
-	push rsi
-	push rdi
+struc REGS
+.r15           resq 1  ; 0x00
+.r14           resq 1  ; 0x08
+.r13           resq 1  ; 0x10
+.r12           resq 1  ; 0x18
+.r11           resq 1  ; 0x20
+.r10           resq 1  ; 0x28
+.r9            resq 1  ; 0x30
+.r8            resq 1  ; 0x38
 
-	push r8
-	push r9
-	push r10
-	push r11
-	push r12
-	push r13
-	push r14
-	push r15
+.rdi           resq 1  ; 0x40
+.rsi           resq 1  ; 0x48
+.rbp           resq 1  ; 0x50
+.rbx           resq 1  ; 0x58
+.rdx           resq 1  ; 0x60
+.rcx           resq 1  ; 0x68
+.rax           resq 1  ; 0x70
 
-	mov ax, ds
-	push rax
+.int           resq 1  ; 0x78
+.err           resq 1  ; 0x80
+.rip           resq 1  ; 0x88
+.cs3           resq 1  ; 0x90
+.rflags        resq 1  ; 0x98
+.rsp3          resq 1  ; 0xa0
+.ss3           resq 1  ; 0xa8
+endstruc
+
+regs_total_size equ 0xb0
+regs_extra_size equ 0x78
+
+%macro push_all 0
+	sub rsp, regs_extra_size
+
+	mov [rsp + REGS.rax], rax
+	mov [rsp + REGS.rcx], rcx
+	mov [rsp + REGS.rdx], rdx
+	mov [rsp + REGS.rbx], rbx
+	mov [rsp + REGS.rbp], rbp
+	mov [rsp + REGS.rsi], rsi
+	mov [rsp + REGS.rdi], rdi
+
+	mov [rsp + REGS.r8 ], r8
+	mov [rsp + REGS.r9 ], r9
+	mov [rsp + REGS.r10], r10
+	mov [rsp + REGS.r11], r11
+	mov [rsp + REGS.r12], r12
+	mov [rsp + REGS.r13], r13
+	mov [rsp + REGS.r14], r14
+	mov [rsp + REGS.r15], r15
 %endmacro
 
-%macro pop_all_and_segments 0
-	pop rax
-	mov ds, ax
-	mov es, ax
-	mov fs, ax
-	mov gs, ax
+%macro pop_all 0
+	mov rax, [rsp + REGS.rax]
+	mov rcx, [rsp + REGS.rcx]
+	mov rdx, [rsp + REGS.rdx]
+	mov rbx, [rsp + REGS.rbx]
+	mov rbp, [rsp + REGS.rbp]
+	mov rsi, [rsp + REGS.rsi]
+	mov rdi, [rsp + REGS.rdi]
 
-	pop r15
-	pop r14
-	pop r13
-	pop r12
-	pop r11
-	pop r10
-	pop r9
-	pop r8
+	mov r8,  [rsp + REGS.r8 ]
+	mov r9,  [rsp + REGS.r9 ]
+	mov r10, [rsp + REGS.r10]
+	mov r11, [rsp + REGS.r11]
+	mov r12, [rsp + REGS.r12]
+	mov r13, [rsp + REGS.r13]
+	mov r14, [rsp + REGS.r14]
+	mov r15, [rsp + REGS.r15]
 
-	pop rdi
-	pop rsi
-	pop rbp
-	pop rbx
-	pop rdx
-	pop rcx
-	pop rax
+	add rsp, regs_extra_size
+%endmacro
+
+%macro check_swap_gs 0
+	mov rax, [rsp+0x90]
+	and rax, 0x03 ; mask out the RPL
+	cmp rax, 0x03
+	jne %%noswapgs
+	swapgs
+%%noswapgs:
 %endmacro
 
 ; vim: ft=asm

src/kernel/scheduler.cpp

@@ -1,6 +1,7 @@
 #include "apic.h"
 #include "console.h"
 #include "cpu.h"
+#include "debug.h"
 #include "gdt.h"
 #include "interrupts.h"
 #include "io.h"
@@ -15,30 +16,33 @@
 
 using memory::initial_stack;
 
-scheduler scheduler::s_instance(nullptr);
+scheduler scheduler::s_instance(nullptr, kutil::allocator::invalid);
 
-const int stack_size = 0x1000;
 const uint64_t rflags_noint = 0x002;
 const uint64_t rflags_int = 0x202;
 
 extern "C" {
 	void ramdisk_process_loader();
-	void load_process(const void *image_start, size_t bytes, process *proc, cpu_state state);
+	uintptr_t load_process_image(const void *image_start, size_t bytes, process *proc);
 };
 
-scheduler::scheduler(lapic *apic) :
+extern uint64_t idle_stack_end;
+
+scheduler::scheduler(lapic *apic, kutil::allocator &alloc) :
 	m_apic(apic),
-	m_next_pid(1)
+	m_next_pid(1),
+	m_process_allocator(alloc)
 {
 	auto *idle = m_process_allocator.pop();
-
 	uint8_t last_pri = num_priorities - 1;
 
 	// The kernel idle task, also the thread we're in now
 	idle->pid = 0;
 	idle->ppid = 0;
 	idle->priority = last_pri;
-	idle->rsp = 0; // This will get set when we switch away
+	idle->rsp = 0;  // This will get set when we switch away
+	idle->rsp3 = 0; // Never used for the idle task
+	idle->rsp0 = reinterpret_cast<uintptr_t>(&idle_stack_end);
 	idle->pml4 = page_manager::get_pml4();
 	idle->quanta = process_quanta;
 	idle->flags =
@@ -48,10 +52,13 @@ scheduler::scheduler(lapic *apic) :
 
 	m_runlists[last_pri].push_back(idle);
 	m_current = idle;
+
+	bsp_cpu_data.rsp0 = idle->rsp0;
+	bsp_cpu_data.tcb = idle;
 }
 
-void
-load_process(const void *image_start, size_t bytes, process *proc, cpu_state state)
+uintptr_t
+load_process_image(const void *image_start, size_t bytes, process *proc)
 {
 	// We're now in the process space for this process, allocate memory for the
 	// process code and load it
@@ -61,7 +68,7 @@ load_process(const void *image_start, size_t bytes, process *proc, cpu_state sta
 
 	// TODO: Handle bad images gracefully
 	elf::elf image(image_start, bytes);
-	kassert(image.valid(), "Invalid ELF passed to load_process");
+	kassert(image.valid(), "Invalid ELF passed to load_process_image");
 
 	const unsigned program_count = image.program_count();
 	for (unsigned i = 0; i < program_count; ++i) {
@@ -102,17 +109,20 @@ load_process(const void *image_start, size_t bytes, process *proc, cpu_state sta
 		kutil::memcpy(dest, src, header->size);
 	}
 
-	state.rip = image.entrypoint();
 	proc->flags &= ~process_flags::loading;
 
-	log::debug(logs::task, "  Loaded! New process rip: %016lx", state.rip);
+	uintptr_t entrypoint = image.entrypoint();
+	log::debug(logs::task, "  Loaded! New process rip: %016lx", entrypoint);
+	return entrypoint;
 }
 
 process_node *
-scheduler::create_process()
+scheduler::create_process(pid_t pid)
 {
+	kassert(pid <= 0, "Cannot specify a positive pid in create_process");
+
 	auto *proc = m_process_allocator.pop();
-	proc->pid = m_next_pid++;
+	proc->pid = pid ? pid : m_next_pid++;
 	proc->priority = default_priority;
 	return proc;
 }
@@ -129,39 +139,28 @@ scheduler::load_process(const char *name, const void *data, size_t size)
 	uint16_t ss = (4 << 3) | 3;  // User SS is GDT entry 4, ring 3
 
 	// Set up the page tables - this also allocates an initial user stack
-	page_table *pml4 = page_manager::get()->create_process_map();
+	proc->pml4 = page_manager::get()->create_process_map();
 
-	// Create a one-page kernel stack space
-	void *stack0 = proc->setup_kernel_stack(stack_size, 0);
+	// Create an initial kernel stack space
+	void *sp0 = proc->setup_kernel_stack();
+	uintptr_t *stack = reinterpret_cast<uintptr_t *>(sp0) - 7;
 
-	// Stack grows down, point to the end
-	void *sp0 = kutil::offset_pointer(stack0, stack_size);
+	// Pass args to ramdisk_process_loader on the stack
+	stack[0] = reinterpret_cast<uintptr_t>(data);
+	stack[1] = reinterpret_cast<uintptr_t>(size);
+	stack[2] = reinterpret_cast<uintptr_t>(proc);
 
-	cpu_state *state = reinterpret_cast<cpu_state *>(sp0) - 1;
+	proc->rsp = reinterpret_cast<uintptr_t>(stack);
+	proc->add_fake_task_return(
+			reinterpret_cast<uintptr_t>(ramdisk_process_loader));
 
-	// Highest state in the stack is the process' kernel stack for the loader
-	// to iret to:
-	state->ds = state->ss = ss;
-	state->cs = cs;
-	state->rflags = rflags_int;
-	state->rip = 0; // to be filled by the loader
-	state->user_rsp = initial_stack;
+	// Arguments for iret - rip will be pushed on before these
+	stack[3] = cs;
+	stack[4] = rflags_int;
+	stack[5] = initial_stack;
+	stack[6] = ss;
 
-	// Next state in the stack is the loader's kernel stack. The scheduler will
-	// iret to this which will kick off the loading:
-	cpu_state *loader_state = reinterpret_cast<cpu_state *>(sp0) - 2;
-
-	loader_state->ds = loader_state->ss = kss;
-	loader_state->cs = kcs;
-	loader_state->rflags = rflags_noint;
-	loader_state->rip = reinterpret_cast<uint64_t>(ramdisk_process_loader);
-	loader_state->user_rsp = reinterpret_cast<uint64_t>(state);
-
-	loader_state->rax = reinterpret_cast<uint64_t>(data);
-	loader_state->rbx = size;
-
-	proc->rsp = reinterpret_cast<uintptr_t>(loader_state);
-	proc->pml4 = pml4;
+	proc->rsp3 = initial_stack;
 	proc->quanta = process_quanta;
 	proc->flags =
 		process_flags::running |
@@ -170,17 +169,44 @@ scheduler::load_process(const char *name, const void *data, size_t size)
 
 	m_runlists[default_priority].push_back(proc);
 
-	loader_state->rcx = reinterpret_cast<uint64_t>(proc);
-
 	log::debug(logs::task, "Creating process %s: pid %d  pri %d", name, proc->pid, proc->priority);
-	log::debug(logs::task, "     RSP0 %016lx", state);
-	log::debug(logs::task, "     PML4 %016lx", pml4);
+	log::debug(logs::task, "     RSP  %016lx", proc->rsp);
+	log::debug(logs::task, "     RSP0 %016lx", proc->rsp0);
+	log::debug(logs::task, "     PML4 %016lx", proc->pml4);
+}
+
+void
+scheduler::create_kernel_task(pid_t pid, void (*task)())
+{
+	auto *proc = create_process(pid);
+
+	uint16_t kcs = (1 << 3) | 0; // Kernel CS is GDT entry 1, ring 0
+	uint16_t kss = (2 << 3) | 0; // Kernel SS is GDT entry 2, ring 0
+
+	// Create an initial kernel stack space
+	proc->setup_kernel_stack();
+	proc->add_fake_task_return(
+			reinterpret_cast<uintptr_t>(task));
+
+	proc->pml4 = page_manager::get()->get_kernel_pml4();
+	proc->quanta = process_quanta;
+	proc->flags =
+		process_flags::running |
+		process_flags::ready;
+
+	m_runlists[default_priority].push_back(proc);
+
+	log::debug(logs::task, "Creating kernel task: pid %d  pri %d", proc->pid, proc->priority);
+	log::debug(logs::task, "    RSP0 %016lx", proc->rsp0);
+	log::debug(logs::task, "     RSP %016lx", proc->rsp);
+	log::debug(logs::task, "    PML4 %016lx", proc->pml4);
 }
 
 void
 scheduler::start()
 {
 	log::info(logs::task, "Starting scheduler.");
+	wrmsr(msr::ia32_gs_base, reinterpret_cast<uintptr_t>(&bsp_cpu_data));
 	m_tick_count = m_apic->enable_timer(isr::isrTimer, quantum_micros, false);
 }
 
@@ -233,16 +259,15 @@ void scheduler::prune(uint64_t now)
 	}
 }
 
-uintptr_t
-scheduler::schedule(uintptr_t rsp0)
+void
+scheduler::schedule()
 {
-
 	// TODO: lol a real clock
 	static uint64_t now = 0;
 
-	m_current->rsp = rsp0;
-	m_runlists[m_current->priority].remove(m_current);
+	pid_t lastpid = m_current->pid;
 
+	m_runlists[m_current->priority].remove(m_current);
 	if (m_current->flags && process_flags::ready) {
 		m_runlists[m_current->priority].push_back(m_current);
 	} else {
@@ -258,32 +283,24 @@ scheduler::schedule(uintptr_t rsp0)
 	}
 
 	m_current = m_runlists[pri].pop_front();
-	rsp0 = m_current->rsp;
 
-	// Set rsp0 to after the end of the about-to-be-popped cpu state
-	tss_set_stack(0, rsp0 + sizeof(cpu_state));
-	wrmsr(msr::ia32_kernel_gs_base, rsp0);
+	if (lastpid != m_current->pid) {
+		task_switch(m_current);
 
-	// Swap page tables
-	page_table *pml4 = m_current->pml4;
-	page_manager::set_pml4(pml4);
-
-	bool loading = m_current->flags && process_flags::loading;
-	log::debug(logs::task, "Scheduler switched to process %d, priority %d%s.",
-			m_current->pid, m_current->priority, loading ? " (loading)" : "");
-
-	return rsp0;
+		bool loading = m_current->flags && process_flags::loading;
+		log::debug(logs::task, "Scheduler switched to process %d, priority %d%s.",
+				m_current->pid, m_current->priority, loading ? " (loading)" : "");
+	}
 }
 
-uintptr_t
-scheduler::tick(uintptr_t rsp0)
+void
+scheduler::tick()
 {
 	if (--m_current->quanta == 0) {
 		m_current->quanta = process_quanta;
-		rsp0 = schedule(rsp0);
+		schedule();
 	}
 	m_apic->reset_timer(m_tick_count);
-	return rsp0;
 }
 
 process_node *

src/kernel/scheduler.h

@@ -3,6 +3,7 @@
 /// The task scheduler and related definitions
 
 #include <stdint.h>
+#include "kutil/allocator.h"
 #include "kutil/slab_allocator.h"
 #include "process.h"
 
@@ -10,7 +11,9 @@ class lapic;
 struct page_table;
 struct cpu_state;
 
-extern "C" uintptr_t isr_handler(uintptr_t, cpu_state*);
+extern "C" void isr_handler(cpu_state*);
+extern "C" void task_switch(process *next);
+extern "C" void task_fork(process *child);
 
 
 /// The task scheduler
@@ -21,14 +24,15 @@ public:
 	static const uint8_t default_priority = num_priorities / 2;
 
 	/// How long the timer quantum is
-	static const uint64_t quantum_micros = 100000;
+	static const uint64_t quantum_micros = 1000;
 
 	/// How many quantums a process gets before being rescheduled
-	static const uint16_t process_quanta = 10;
+	static const uint16_t process_quanta = 100;
 
 	/// Constructor.
 	/// \arg apic  Pointer to the local APIC object
-	scheduler(lapic *apic);
+	/// \arg alloc Allocator to use for TCBs
+	scheduler(lapic *apic, kutil::allocator &alloc);
 
 	/// Create a new process from a program image in memory.
 	/// \arg name  Name of the program image
@@ -36,14 +40,17 @@ public:
 	/// \arg size  Size of the program image, in bytes
 	void load_process(const char *name, const void *data, size_t size);
 
+	/// Create a new kernel task
+	/// \arg pid   Pid to use for this task, must be negative
+	/// \arg proc  Function to run as a kernel task
+	void create_kernel_task(pid_t pid, void (*task)());
+
 	/// Start the scheduler working. This may involve starting
 	/// timer interrupts or other preemption methods.
 	void start();
 
 	/// Run the scheduler, possibly switching to a new task
-	/// \arg rsp0  The stack pointer of the current interrupt handler
-	/// \returns   The stack pointer to switch to
-	uintptr_t schedule(uintptr_t rsp0);
+	void schedule();
 
 	/// Get the current process.
 	/// \returns  A pointer to the current process' process struct
@@ -60,18 +67,17 @@ public:
 
 private:
 	friend uintptr_t syscall_dispatch(uintptr_t, cpu_state &);
-	friend uintptr_t isr_handler(uintptr_t, cpu_state*);
+	friend void isr_handler(cpu_state*);
 	friend class process;
 
 	/// Create a new process object. This process will have its pid
 	/// set but nothing else.
+	/// \arg pid  The pid to give the process (0 for automatic)
 	/// \returns  The new process object
-	process_node * create_process();
+	process_node * create_process(pid_t pid = 0);
 
 	/// Handle a timer tick
-	/// \arg rsp0  The stack pointer of the current interrupt handler
-	/// \returns   The stack pointer to switch to
-	uintptr_t tick(uintptr_t rsp0);
+	void tick();
 
 	void prune(uint64_t now);
 

src/kernel/syscall.cpp

@@ -1,16 +1,125 @@
 #include "console.h"
 #include "cpu.h"
 #include "debug.h"
+#include "log.h"
 #include "msr.h"
 #include "process.h"
 #include "scheduler.h"
 #include "syscall.h"
 
 extern "C" {
-	void _halt();
+	void syscall_invalid(uint64_t call);
 	void syscall_handler_prelude();
 }
 
+namespace syscalls {
+
+
+void send() {}
+void receive() {}
+
+} // namespace syscalls
+
+uintptr_t syscall_registry[static_cast<unsigned>(syscall::MAX)];
+const char * syscall_names[static_cast<unsigned>(syscall::MAX)];
+
+void
+syscall_invalid(uint64_t call)
+{
+	console *cons = console::get();
+	cons->set_color(9);
+	cons->printf("\nReceived unknown syscall: %02x\n", call);
+
+	const unsigned num_calls =
+		static_cast<unsigned>(syscall::MAX);
+
+	cons->printf("  Known syscalls:\n");
+		cons->printf("          invalid %016lx\n", syscall_invalid);
+
+	for (unsigned i = 0; i < num_calls; ++i) {
+		const char *name = syscall_names[i];
+		uintptr_t handler = syscall_registry[i];
+		if (name)
+			cons->printf("    %02x %10s %016lx\n", i, name, handler);
+	}
+
+	cons->set_color();
+	_halt();
+}
+
+/*
+void
+syscall_dispatch(cpu_state *regs)
+{
+	console *cons = console::get();
+	syscall call = static_cast<syscall>(regs->rax);
+
+	auto &s = scheduler::get();
+	auto *p = s.current();
+
+	switch (call) {
+	case syscall::noop:
+		break;
+
+	case syscall::debug:
+		cons->set_color(11);
+		cons->printf("\nProcess %d: Received DEBUG syscall\n", p->pid);
+		cons->set_color();
+		print_regs(*regs);
+		cons->printf("\n         Syscall enters: %8d\n", __counter_syscall_enter);
+		cons->printf("         Syscall sysret: %8d\n", __counter_syscall_sysret);
+		break;
+
+	case syscall::send:
+		{
+			pid_t target = regs->rdi;
+			uintptr_t data = regs->rsi;
+
+			cons->set_color(11);
+			cons->printf("\nProcess %d: Received SEND syscall, target %d, data %016lx\n", p->pid, target, data);
+			cons->set_color();
+
+			if (p->wait_on_send(target))
+				s.schedule();
+		}
+		break;
+
+	case syscall::receive:
+		{
+			pid_t source = regs->rdi;
+			uintptr_t data = regs->rsi;
+
+			cons->set_color(11);
+			cons->printf("\nProcess %d: Received RECEIVE syscall, source %d, dat %016lx\n", p->pid, source, data);
+			cons->set_color();
+
+			if (p->wait_on_receive(source))
+				s.schedule();
+		}
+		break;
+
+	case syscall::fork:
+		{
+			cons->set_color(11);
+			cons->printf("\nProcess %d: Received FORK syscall\n", p->pid);
+			cons->set_color();
+
+			pid_t pid = p->fork(regs);
+			cons->printf("\n   fork returning %d\n", pid);
+			regs->rax = pid;
+		}
+		break;
+
+	default:
+		cons->set_color(9);
+		cons->printf("\nReceived unknown syscall: %02x\n", call);
+		cons->set_color();
+		_halt();
+		break;
+	}
+}
+*/
+
 void
 syscall_enable()
 {
@@ -32,113 +141,20 @@ syscall_enable()
 
 	// IA32_FMASK - FLAGS mask inside syscall
 	wrmsr(msr::ia32_fmask, 0x200);
-}
 
-uintptr_t
-syscall_dispatch(uintptr_t return_rsp, cpu_state &regs)
-{
-	console *cons = console::get();
-	syscall call = static_cast<syscall>(regs.rax);
+	static constexpr unsigned num_calls =
+		static_cast<unsigned>(syscall::MAX);
 
-	auto &s = scheduler::get();
-	auto *p = s.current();
-
-	switch (call) {
-	case syscall::noop:
-		break;
-
-	case syscall::debug:
-		cons->set_color(11);
-		cons->printf("\nProcess %u: Received DEBUG syscall\n", p->pid);
-		cons->set_color();
-		print_regs(regs);
-		break;
-
-	case syscall::message:
-		cons->set_color(11);
-		cons->printf("\nProcess %u: Received MESSAGE syscall\n", p->pid);
-		cons->set_color();
-		break;
-
-	case syscall::pause:
-		{
-			cons->set_color(11);
-
-			auto &s = scheduler::get();
-			auto *p = s.current();
-			p->wait_on_signal(-1ull);
-			cons->printf("\nProcess %u: Received PAUSE syscall\n", p->pid);
-			cons->set_color();
-			return_rsp = s.schedule(return_rsp);
-		}
-		break;
-
-	case syscall::sleep:
-		{
-			cons->set_color(11);
-			cons->printf("\nProcess %u: Received SLEEP syscall\n", p->pid);
-			cons->printf("Sleeping until %lu\n", regs.rbx);
-			cons->set_color();
-
-			p->wait_on_time(regs.rbx);
-			return_rsp = s.schedule(return_rsp);
-		}
-		break;
-
-	case syscall::getpid:
-		cons->set_color(11);
-		cons->printf("\nProcess %u: Received GETPID syscall\n", p->pid);
-		cons->set_color();
-		regs.rax = p->pid;
-		break;
-
-	case syscall::send:
-		{
-			uint32_t target = regs.rdi;
-
-			cons->set_color(11);
-			cons->printf("\nProcess %u: Received SEND syscall, target %u\n", p->pid, target);
-			cons->set_color();
-
-			if (p->wait_on_send(target))
-				return_rsp = s.schedule(return_rsp);
-		}
-		break;
-
-	case syscall::receive:
-		{
-			uint32_t source = regs.rdi;
-
-			cons->set_color(11);
-			cons->printf("\nProcess %u: Received RECEIVE syscall, source %u\n", p->pid, source);
-			cons->set_color();
-
-			if (p->wait_on_receive(source))
-				return_rsp = s.schedule(return_rsp);
-		}
-		break;
-
-	case syscall::fork:
-		{
-			cons->set_color(11);
-			cons->printf("\nProcess %u: Received FORK syscall\n", p->pid);
-			cons->set_color();
-
-			pid_t pid = p->fork(return_rsp);
-			if (pid == scheduler::get().current()->pid)
-				pid = 0;
-			regs.rax = pid;
-		}
-		break;
-
-	default:
-		cons->set_color(9);
-		cons->printf("\nReceived unknown syscall: %02x\n", call);
-		cons->set_color();
-		_halt();
-		break;
+	for (unsigned i = 0; i < num_calls; ++i) {
+		syscall_registry[i] = reinterpret_cast<uintptr_t>(syscall_invalid);
+		syscall_names[i] = nullptr;
 	}
 
-	return return_rsp;
+#define SYSCALL(id, name, result, ...) \
+	syscall_registry[id] = reinterpret_cast<uintptr_t>(syscalls::name); \
+	syscall_names[id] = #name; \
+	static_assert( id <= num_calls, "Syscall " #name " has id > syscall::MAX" );
+#include "syscalls.inc"
+#undef SYSCALL
 }
 

src/kernel/syscall.h

@@ -6,19 +6,20 @@ struct cpu_state;
 
 enum class syscall : uint64_t
 {
-	noop           = 0x0000,
-	debug          = 0x0001,
-	message        = 0x0002,
-	pause          = 0x0003,
-	sleep          = 0x0004,
-	getpid         = 0x0005,
-	send           = 0x0006,
-	receive        = 0x0007,
-	fork           = 0x0008,
+#define SYSCALL(id, name, result, ...) name = id,
+#include "syscalls.inc"
+#undef SYSCALL
 
-	last_syscall
+	// Maximum syscall id. If you change this, also change
+	// MAX_SYSCALLS in syscall.s
+	MAX = 0x40
 };
 
 void syscall_enable();
-uintptr_t syscall_dispatch(uintptr_t, cpu_state &);
 
+namespace syscalls
+{
+#define SYSCALL(id, name, result, ...) result name (__VA_ARGS__);
+#include "syscalls.inc"
+#undef SYSCALL
+}

src/kernel/syscall.s

@@ -1,24 +1,62 @@
 %include "push_all.inc"
+%include "tasking.inc"
+
+; Make sure to keep MAX_SYSCALLS in sync with
+; syscall::MAX in syscall.h
+MAX_SYSCALLS equ 0x40
+
+extern __counter_syscall_enter
+extern __counter_syscall_sysret
+
+extern syscall_registry
+extern syscall_invalid
 
-extern syscall_handler
 global syscall_handler_prelude
+global syscall_handler_prelude.return
 syscall_handler_prelude:
-	push 0		; ss, doesn't matter here
-	push rsp
-	pushf
-	push 0		; cs, doesn't matter here
-	push rcx	; user rip
-	push 0		; bogus interrupt
-	push 0		; bogus errorcode
-	push_all_and_segments
+	swapgs
+	mov [gs:CPU_DATA.rsp3], rsp
+	mov rsp, [gs:CPU_DATA.rsp0]
 
-	mov rdi, rsp
-	call syscall_handler
-	mov rsp, rax
+	push rcx
+	push rbp
+	mov rbp, rsp
 
-	pop_all_and_segments
-	add rsp, 16	; ignore bogus interrupt / error
-	pop rcx		; user rip
-	add rsp, 32 ; ignore cs, flags, rsp, ss
+	push rbx
+	push r11
+	push r12
+	push r13
+	push r14
+	push r15
 
+	inc qword [rel __counter_syscall_enter]
+
+	cmp rax, MAX_SYSCALLS
+	jle .ok_syscall
+
+	mov rdi, rax
+	call syscall_invalid
+
+.ok_syscall:
+	lea r11, [rel syscall_registry]
+	mov r11, [r11 + rax * 8]
+	call r11
+
+	inc qword [rel __counter_syscall_sysret]
+
+.return:
+	pop r15
+	pop r14
+	pop r13
+	pop r12
+	pop r11
+	pop rbx
+
+	pop rbp
+	pop rcx
+
+	mov [gs:CPU_DATA.rsp0], rsp
+	mov rsp, [gs:CPU_DATA.rsp3]
+
+	swapgs
 	o64 sysret

src/kernel/syscalls.inc

@@ -0,0 +1,12 @@
+SYSCALL(0x00, noop,     void)
+SYSCALL(0x01, exit,     void, int64_t)
+SYSCALL(0x02, getpid,   pid_t)
+SYSCALL(0x03, fork,     pid_t)
+
+SYSCALL(0x10, message,  void, const char *)
+
+SYSCALL(0x20, pause,    void)
+SYSCALL(0x21, sleep,    void, uint64_t)
+
+SYSCALL(0x30, send,     void)
+SYSCALL(0x31, receive,  void)

src/kernel/syscalls/exit.cpp

@@ -0,0 +1,17 @@
+#include "log.h"
+#include "scheduler.h"
+
+namespace syscalls {
+
+void
+exit(int64_t status)
+{
+	auto &s = scheduler::get();
+	auto *p = s.current();
+	log::debug(logs::syscall, "Process %d exiting with code %d", p->pid, status);
+
+	p->exit(status);
+	s.schedule();
+}
+
+} // namespace syscalls

src/kernel/syscalls/fork.cpp

@@ -0,0 +1,24 @@
+
+#include "log.h"
+#include "scheduler.h"
+
+namespace syscalls {
+
+pid_t
+fork()
+{
+	auto &s = scheduler::get();
+	auto *p = s.current();
+	pid_t ppid = p->pid;
+
+	log::debug(logs::syscall, "Process %d calling fork()", ppid);
+
+	pid_t pid = p->fork();
+
+	p = s.current();
+	log::debug(logs::syscall, "Process %d's fork: returning %d from process %d", ppid, pid, p->pid);
+
+	return pid;
+}
+
+} // namespace syscalls

src/kernel/syscalls/getpid.cpp

@@ -0,0 +1,14 @@
+#include "log.h"
+#include "scheduler.h"
+
+namespace syscalls {
+
+pid_t
+getpid()
+{
+	auto &s = scheduler::get();
+	auto *p = s.current();
+	return p->pid;
+}
+
+} // namespace syscalls

src/kernel/syscalls/message.cpp

@@ -0,0 +1,14 @@
+#include "log.h"
+#include "scheduler.h"
+
+namespace syscalls {
+
+void
+message(const char *message)
+{
+	auto &s = scheduler::get();
+	auto *p = s.current();
+	log::info(logs::syscall, "Message[%d]: %s", p->pid, message);
+}
+
+} // namespace syscalls

src/kernel/syscalls/noop.cpp

@@ -0,0 +1,14 @@
+#include "log.h"
+#include "scheduler.h"
+
+namespace syscalls {
+
+void
+noop()
+{
+	auto &s = scheduler::get();
+	auto *p = s.current();
+	log::debug(logs::syscall, "Process %d called noop syscall.", p->pid);
+}
+
+} // namespace syscalls

src/kernel/syscalls/pause.cpp

@@ -0,0 +1,15 @@
+#include "log.h"
+#include "scheduler.h"
+
+namespace syscalls {
+
+void
+pause()
+{
+	auto &s = scheduler::get();
+	auto *p = s.current();
+	p->wait_on_signal(-1ull);
+	s.schedule();
+}
+
+} // namespace syscalls

src/kernel/syscalls/sleep.cpp

@@ -0,0 +1,17 @@
+#include "log.h"
+#include "scheduler.h"
+
+namespace syscalls {
+
+void
+sleep(uint64_t til)
+{
+	auto &s = scheduler::get();
+	auto *p = s.current();
+	log::debug(logs::syscall, "Process %d sleeping until %d", p->pid, til);
+
+	p->wait_on_time(til);
+	s.schedule();
+}
+
+} // namespace syscalls

src/kernel/task.s

@@ -0,0 +1,63 @@
+%include "tasking.inc"
+
+extern g_tss
+global task_switch
+task_switch:
+	push rbp
+	mov rbp, rsp
+
+	; Save the rest of the callee-saved regs
+	push rbx
+	push r12
+	push r13
+	push r14
+	push r15
+
+	; Update previous task's TCB
+	mov rax, [gs:CPU_DATA.tcb]     ; rax: current task TCB
+	mov [rax + TCB.rsp], rsp
+
+	; Copy off saved user rsp
+	mov rcx, [gs:CPU_DATA.rsp3]    ; rcx: curretn task's saved user rsp
+	mov [rax + TCB.rsp3], rcx
+
+	; Install next task's TCB
+	mov [gs:CPU_DATA.tcb], rdi     ; rdi: next TCB (function param)
+	mov rsp, [rdi + TCB.rsp]       ; next task's stack pointer
+	mov rax, 0x0000007fffffffff
+	and rax, [rdi + TCB.pml4]      ; rax: next task's pml4 (phys portion of address)
+
+	; Update syscall/interrupt rsp
+	mov rcx, [rdi + TCB.rsp0]      ; rcx: top of next task's kernel stack
+	mov [gs:CPU_DATA.rsp0], rcx
+
+	; Update saved user rsp
+	mov rcx, [rdi + TCB.rsp3]      ; rcx: new task's saved user rsp
+	mov [gs:CPU_DATA.rsp3], rcx
+
+	lea rdx, [rel g_tss]           ; rdx: address of TSS
+	mov [rdx + TSS.rsp0], rcx
+
+	; check if we need to update CR3
+	mov rdx, cr3                   ; rdx: old CR3
+	cmp rax, rdx
+	je .no_cr3
+	mov cr3, rax
+.no_cr3:
+
+	pop r15
+	pop r14
+	pop r13
+	pop r12
+	pop rbx
+
+	pop rbp
+	ret
+
+
+extern syscall_handler_prelude.return
+global task_fork_return_thunk
+task_fork_return_thunk:
+	mov rax, 0
+	jmp syscall_handler_prelude.return
+

src/kernel/tasking.inc

@@ -0,0 +1,33 @@
+struc TCB
+.rsp:          resq 1
+.rsp0:		   resq 1
+.rsp3:		   resq 1
+.pml4:         resq 1
+endstruc
+
+struc CPU_DATA
+.rsp0:         resq 1
+.rsp3:         resq 1
+.tcb:          resq 1
+endstruc
+
+struc TSS
+.res0:         resd 1
+.rsp0:         resq 1
+.rsp1:         resq 1
+.rsp2:         resq 1
+.ist0:         resq 1
+.ist1:         resq 1
+.ist2:         resq 1
+.ist3:         resq 1
+.ist4:         resq 1
+.ist5:         resq 1
+.ist6:         resq 1
+.ist7:         resq 1
+.res1:         resq 1
+.res2:         resw 1
+.iomap:        resw 1
+endstruc
+
+
+; vim: ft=asm

src/libraries/initrd/include/initrd/initrd.h

@@ -49,7 +49,7 @@ class disk
 public:
 	/// Constructor.
 	/// \arg start  The start of the initrd in memory
-	disk(const void *start);
+	disk(const void *start, kutil::allocator &alloc);
 
 	/// Get the vector of files on the disk
 	const kutil::vector<file> & files() const { return m_files; }

src/libraries/initrd/initrd.cpp

@@ -23,7 +23,8 @@ file::executable() const {
 }
 
 
-disk::disk(const void *start)
+disk::disk(const void *start, kutil::allocator &alloc) :
+	m_files(alloc)
 {
 	auto *header = reinterpret_cast<const disk_header *>(start);
 	size_t length = header->length;

src/libraries/kutil/bip_buffer.cpp

@@ -0,0 +1,95 @@
+#include "kutil/assert.h"
+#include "kutil/bip_buffer.h"
+
+namespace kutil {
+
+bip_buffer::bip_buffer() :
+		m_start_a(0),
+		m_start_b(0),
+		m_size_a(0),
+		m_size_b(0),
+		m_buffer_size(0),
+		m_buffer(nullptr)
+	{}
+
+bip_buffer::bip_buffer(uint8_t *buffer, size_t size) :
+		m_start_a(0),
+		m_start_b(0),
+		m_size_a(0),
+		m_size_b(0),
+		m_buffer_size(size),
+		m_buffer(buffer)
+	{}
+
+size_t bip_buffer::reserve(size_t size, void **area)
+{
+	if (m_size_r) {
+		*area = nullptr;
+		return 0;
+	}
+
+	size_t remaining = 0;
+	if (m_size_b) {
+		// If B exists, we're appending there. Get space between
+		// the end of B and start of A.
+		remaining = m_start_a - m_start_b - m_size_b;
+		m_start_r = m_start_b + m_size_b;
+	} else {
+		// B doesn't exist, check the space both before and after A.
+		// If the end of A has enough room for this write, put it there.
+		remaining = m_buffer_size - m_start_a - m_size_a;
+		m_start_r = m_start_a + m_size_a;
+
+		// Otherwise use the bigger of the areas in front of and after A
+		if (remaining < size && m_start_a > remaining) {
+			remaining = m_start_a;
+			m_start_r = 0;
+		}
+	}
+
+	if (!remaining) {
+		*area = nullptr;
+		return 0;
+	}
+
+	m_size_r = (remaining < size) ? remaining : size;
+	*area = &m_buffer[m_start_r];
+	return m_size_r;
+}
+
+void bip_buffer::commit(size_t size)
+{
+	kassert(size <= m_size_r, "Tried to commit more than reserved");
+
+	if (m_start_r == m_start_a + m_size_a) {
+		// We were adding to A
+		m_size_a += size;
+	} else {
+		// We were adding to B
+		kassert(m_start_r == m_start_b + m_size_b, "Bad m_start_r!");
+		m_size_b += size;
+	}
+
+	m_start_r = m_size_r = 0;
+}
+
+size_t bip_buffer::get_block(void **area) const
+{
+	*area = m_size_a ? &m_buffer[m_start_a] : nullptr;
+	return m_size_a;
+}
+
+void bip_buffer::consume(size_t size)
+{
+	kassert(size <= m_size_a, "Consumed more bytes than exist in A");
+	if (size >= m_size_a) {
+		m_size_a = m_size_b;
+		m_start_a = m_start_b;
+		m_size_b = m_start_b = 0;
+	} else {
+		m_size_a -= size;
+		m_start_a += size;
+	}
+}
+
+} // namespace kutil

src/libraries/kutil/frame_allocator.cpp

@@ -1,159 +0,0 @@
-#include <algorithm>
-
-#include "kutil/frame_allocator.h"
-
-namespace kutil {
-
-int
-frame_block::compare(const frame_block *rhs) const
-{
-	if (address < rhs->address)
-		return -1;
-	else if (address > rhs->address)
-		return 1;
-
-	return 0;
-}
-
-frame_block_list
-frame_block::consolidate(frame_block_list &list)
-{
-	frame_block_list freed;
-
-	for (auto *cur : list) {
-		auto *next = cur->next();
-
-		while ( next &&
-				cur->flags == next->flags &&
-				cur->end() == next->address) {
-			cur->count += next->count;
-			list.remove(next);
-			freed.push_back(next);
-		}
-	}
-
-	return freed;
-}
-
-void
-frame_block::zero()
-{
-	address = 0;
-	count = 0;
-	flags = frame_block_flags::none;
-}
-
-void
-frame_block::copy(frame_block *other)
-{
-	address = other->address;
-	count = other->count;
-	flags = other->flags;
-}
-
-
-frame_allocator::frame_allocator(
-	frame_block_list cache)
-{
-	m_block_slab.append(cache);
-}
-
-void
-frame_allocator::init(
-	frame_block_list free,
-	frame_block_list used)
-{
-	m_free.append(free);
-	m_used.append(used);
-}
-
-void
-frame_allocator::consolidate_blocks()
-{
-	m_block_slab.append(frame_block::consolidate(m_free));
-	m_block_slab.append(frame_block::consolidate(m_used));
-}
-
-size_t
-frame_allocator::allocate(size_t count, uintptr_t *address)
-{
-	kassert(!m_free.empty(), "frame_allocator::pop_frames ran out of free frames!");
-
-	auto *first = m_free.front();
-
-	unsigned n = std::min(count, static_cast<size_t>(first->count));
-	*address = first->address;
-
-	if (count >= first->count) {
-		m_free.remove(first);
-		m_used.sorted_insert(first);
-	} else {
-		auto *used = m_block_slab.pop();
-		used->copy(first);
-		used->count = n;
-		m_used.sorted_insert(used);
-
-		first->address += n * frame_size;
-		first->count -= n;
-	}
-
-	m_block_slab.append(frame_block::consolidate(m_used));
-	return n;
-}
-
-void
-frame_allocator::free(uintptr_t address, size_t count)
-{
-	size_t block_count = 0;
-
-	for (auto *block : m_used) {
-		if (!block->contains(address)) continue;
-
-		size_t size = frame_size * count;
-		uintptr_t end = address + size;
-
-		size_t leading = address - block->address;
-		size_t trailing =
-			end > block->end() ?
-			0 : (block->end() - end);
-
-		if (leading) {
-			size_t frames = leading / frame_size;
-
-			auto *lead_block = m_block_slab.pop();
-
-			lead_block->copy(block);
-			lead_block->count = frames;
-
-			block->count -= frames;
-			block->address += leading;
-
-			m_used.insert_before(block, lead_block);
-		}
-
-		if (trailing) {
-			size_t frames = trailing / frame_size;
-
-			auto *trail_block = m_block_slab.pop();
-
-			trail_block->copy(block);
-			trail_block->count = frames;
-			trail_block->address += size;
-			trail_block->address += size;
-
-			block->count -= frames;
-
-			m_used.insert_after(block, trail_block);
-		}
-
-		address += block->count * frame_size;
-
-		m_used.remove(block);
-		m_free.sorted_insert(block);
-		++block_count;
-	}
-
-	kassert(block_count, "Couldn't find existing allocated frames to free");
-}
-
-} // namespace kutil

src/libraries/kutil/heap_allocator.cpp

@@ -1,12 +1,11 @@
 #include <stdint.h>
 #include "kutil/assert.h"
 #include "kutil/memory.h"
-#include "kutil/heap_manager.h"
+#include "kutil/heap_allocator.h"
 
 namespace kutil {
 
-
-struct heap_manager::mem_header
+struct heap_allocator::mem_header
 {
 	mem_header(mem_header *prev, mem_header *next, uint8_t size) :
 		m_prev(prev), m_next(next)
@@ -14,34 +13,29 @@ struct heap_manager::mem_header
 		set_size(size);
 	}
 
-	inline void set_size(uint8_t size)
-	{
+	inline void set_size(uint8_t size) {
 		m_prev = reinterpret_cast<mem_header *>(
 			reinterpret_cast<uintptr_t>(prev()) | (size & 0x3f));
 	}
 
-	inline void set_used(bool used)
-	{
+	inline void set_used(bool used) {
 		m_next = reinterpret_cast<mem_header *>(
 			reinterpret_cast<uintptr_t>(next()) | (used ? 1 : 0));
 	}
 
-	inline void set_next(mem_header *next)
-	{
+	inline void set_next(mem_header *next) {
 		bool u = used();
 		m_next = next;
 		set_used(u);
 	}
 
-	inline void set_prev(mem_header *prev)
-	{
+	inline void set_prev(mem_header *prev) {
 		uint8_t s = size();
 		m_prev = prev;
 		set_size(s);
 	}
 
-	void remove()
-	{
+	void remove() {
 		if (next()) next()->set_prev(prev());
 		if (prev()) prev()->set_next(next());
 		set_prev(nullptr);
@@ -67,25 +61,24 @@ private:
 };
 
 
-heap_manager::heap_manager() :
-	m_grow(nullptr)
-{
-}
+heap_allocator::heap_allocator() : m_next(0), m_size(0) {}
 
-heap_manager::heap_manager(grow_callback grow_cb) :
-	m_grow(grow_cb)
+heap_allocator::heap_allocator(uintptr_t start, size_t size) :
+	m_next(start), m_size(size)
 {
 	kutil::memset(m_free, 0, sizeof(m_free));
-	grow_memory();
 }
 
 void *
-heap_manager::allocate(size_t length)
+heap_allocator::allocate(size_t length)
 {
 	size_t total = length + sizeof(mem_header);
 	unsigned size = min_size;
 	while (total > (1 << size)) size++;
+
 	kassert(size <= max_size, "Tried to allocate a block bigger than max_size");
+	if (size > max_size)
+		return nullptr;
 
 	mem_header *header = pop_free(size);
 	header->set_used(true);
@@ -93,18 +86,28 @@ heap_manager::allocate(size_t length)
 }
 
 void
-heap_manager::free(void *p)
+heap_allocator::free(void *p)
 {
+	if (!p) return;
+
 	mem_header *header = reinterpret_cast<mem_header *>(p);
 	header -= 1; // p points after the header
 	header->set_used(false);
 
 	while (header->size() != max_size) {
+		auto size = header->size();
+
 		mem_header *buddy = header->buddy();
-		if (buddy->used() || buddy->size() != header->size()) break;
+		if (buddy->used() || buddy->size() != size)
+			break;
+
+		if (get_free(size) == buddy)
+			get_free(size) = buddy->next();
+
 		buddy->remove();
+
 		header = header->eldest() ? header : buddy;
-		header->set_size(header->size() + 1);
+		header->set_size(size + 1);
 	}
 
 	uint8_t size = header->size();
@@ -115,47 +118,60 @@ heap_manager::free(void *p)
 }
 
 void
-heap_manager::grow_memory()
+heap_allocator::ensure_block(unsigned size)
 {
-	size_t length = (1 << max_size);
-
-	kassert(m_grow, "Tried to grow heap without a growth callback");
-	void *next = m_grow(length);
-
-	mem_header *block = new (next) mem_header(nullptr, get_free(max_size), max_size);
-	get_free(max_size) = block;
-	if (block->next())
-		block->next()->set_prev(block);
-}
-
-void
-heap_manager::ensure_block(unsigned size)
-{
-	if (get_free(size) != nullptr) return;
-	else if (size == max_size) {
-		grow_memory();
+	if (get_free(size) != nullptr)
 		return;
+
+	if (size == max_size) {
+		size_t bytes = (1 << max_size);
+		if (bytes <= m_size) {
+			mem_header *next = reinterpret_cast<mem_header *>(m_next);
+			new (next) mem_header(nullptr, nullptr, size);
+			get_free(size) = next;
+			m_next += bytes;
+			m_size -= bytes;
+		}
+	} else {
+		mem_header *orig = pop_free(size + 1);
+		if (orig) {
+			mem_header *next = kutil::offset_pointer(orig, 1 << size);
+			new (next) mem_header(orig, nullptr, size);
+
+			orig->set_next(next);
+			orig->set_size(size);
+			get_free(size) = orig;
+		}
 	}
-
-	mem_header *orig = pop_free(size + 1);
-
-	mem_header *next = kutil::offset_pointer(orig, 1 << size);
-	new (next) mem_header(orig, nullptr, size);
-
-	orig->set_next(next);
-	orig->set_size(size);
-	get_free(size) = orig;
 }
 
-heap_manager::mem_header *
-heap_manager::pop_free(unsigned size)
+heap_allocator::mem_header *
+heap_allocator::pop_free(unsigned size)
 {
 	ensure_block(size);
 	mem_header *block = get_free(size);
-	get_free(size) = block->next();
-
-	block->remove();
+	if (block) {
+		get_free(size) = block->next();
+		block->remove();
+	}
 	return block;
 }
 
+
+class invalid_allocator :
+	public allocator
+{
+public:
+	virtual void * allocate(size_t) override {
+		kassert(false, "Attempting to allocate from allocator::invalid");
+		return nullptr;
+	}
+
+	virtual void free(void *) override {
+		kassert(false, "Attempting to free from allocator::invalid");
+	}
+} _invalid_allocator;
+
+allocator &allocator::invalid = _invalid_allocator;
+
 } // namespace kutil

src/libraries/kutil/include/kutil/allocator.h

@@ -0,0 +1,31 @@
+#pragma once
+/// \file allocator.h
+/// Allocator interface
+
+#include <stdint.h>
+#include "kernel_memory.h"
+
+namespace kutil {
+
+class allocator
+{
+public:
+	/// Allocate memory.
+	/// \arg length  The amount of memory to allocate, in bytes
+	/// \returns     A pointer to the allocated memory, or nullptr if
+	///              allocation failed.
+	virtual void * allocate(size_t size) = 0;
+
+	/// Free a previous allocation.
+	/// \arg p  A pointer previously retuned by allocate()
+	virtual void free(void *p) = 0;
+
+	template <typename T>
+	inline T * allocate(unsigned count) {
+		return reinterpret_cast<T*>(allocate(count * sizeof(T)));
+	}
+
+	static allocator &invalid;
+};
+
+} // namespace kutil

src/libraries/kutil/include/kutil/bip_buffer.h

@@ -0,0 +1,59 @@
+#pragma once
+/// \file bip_buffer.h
+/// A Bip Buffer (bipartite circular buffer). For more on the Bip Buffer structure, see
+/// https://www.codeproject.com/Articles/3479/The-Bip-Buffer-The-Circular-Buffer-with-a-Twist
+
+#include <stddef.h>
+#include <stdint.h>
+
+namespace kutil {
+
+class bip_buffer
+{
+public:
+	/// Default constructor. Creates a zero-size buffer.
+	bip_buffer();
+
+	/// Constructor.
+	bip_buffer(uint8_t *buffer, size_t size);
+
+	/// Reserve an area of buffer for a write.
+	/// \arg size  Requested size, in bytes
+	/// \arg area  [out] Pointer to returned area
+	/// \returns   Size of returned area, in bytes, or 0 on failure
+	size_t reserve(size_t size, void **area);
+
+	/// Commit a pending write started by reserve()
+	/// \arg size  Amount of data used, in bytes
+	void commit(size_t size);
+
+	/// Get a pointer to a block of data in the buffer.
+	/// \arg area  [out] Pointer to the retuned area
+	/// \returns   Size of the returned area, in bytes
+	size_t get_block(void **area) const;
+
+	/// Mark a number of bytes as consumed, freeing buffer space
+	/// \arg size  Number of bytes to consume
+	void consume(size_t size);
+
+	/// Get total amount of data in the buffer.
+	/// \returns  Number of bytes committed to the buffer
+	inline size_t size() const { return m_size_a + m_size_b; }
+
+	/// Get total amount of free buffer remaining
+	/// \returns  Number of bytes of buffer that are free
+	inline size_t free_space() const { return m_buffer_size - size(); }
+
+private:
+	size_t m_start_a;
+	size_t m_start_b;
+	size_t m_start_r;
+	size_t m_size_a;
+	size_t m_size_b;
+	size_t m_size_r;
+
+	const size_t m_buffer_size;
+	uint8_t * const m_buffer;
+};
+
+} // namespace kutil

src/libraries/kutil/include/kutil/buddy_allocator.h

@@ -3,6 +3,7 @@
 /// Helper base class for buddy allocators with external node storage.
 
 #include <stdint.h>
+#include <utility>
 #include "kutil/assert.h"
 #include "kutil/linked_list.h"
 #include "kutil/slab_allocator.h"
@@ -25,15 +26,21 @@ public:
 	static const size_t min_alloc = (1 << size_min);
 	static const size_t max_alloc = (1 << size_max);
 
-	/// Constructor.
-	buddy_allocator() {}
+	/// Default constructor creates an invalid object.
+	buddy_allocator() : m_alloc(allocator::invalid) {}
 
-	/// Constructor with an initial cache of region structs from bootstrapped
-	/// memory.
-	/// \arg cache  List of pre-allocated ununused region_type structures
-	buddy_allocator(region_list cache)
+	/// Constructor.
+	/// \arg alloc  Allocator to use for region nodes
+	buddy_allocator(allocator &alloc) : m_alloc(alloc) {}
+
+	/// Move-like constructor. Takes ownership of existing regions.
+	buddy_allocator(buddy_allocator &&other, allocator &alloc) :
+		m_alloc(alloc)
 	{
-		m_alloc.append(cache);
+		for (unsigned i = 0; i < buckets; ++i) {
+			m_free[i] = std::move(other.m_free[i]);
+			m_used[i] = std::move(other.m_used[i]);
+		}
 	}
 
 	/// Add address space to be managed.
@@ -173,6 +180,23 @@ public:
 		}
 	}
 
+	/// Check if an allocation exists
+	/// \arg addr  Address within the managed space
+	/// \returns   True if the address is in a region currently allocated
+	bool contains(uintptr_t addr)
+	{
+		for (unsigned i = size_max; i >= size_min; --i) {
+			for (auto *r : used_bucket(i)) {
+				if (r->contains(addr))
+					return true;
+				else if (r->address < addr)
+					break;
+			}
+		}
+
+		return false;
+	}
+
 protected:
 	/// Split a region of the given size into two smaller regions, returning
 	/// the new latter half
@@ -266,6 +290,8 @@ struct buddy_region
 
 	inline uintptr_t end() const { return address + (1ull << size); }
 	inline uintptr_t half() const { return address + (1ull << (size - 1)); }
+	inline bool contains(uintptr_t p) const { return p >= address && p < end(); }
+
 	inline uintptr_t buddy() const { return address ^ (1ull << size); }
 	inline bool elder() const { return address < buddy(); }
 

src/libraries/kutil/include/kutil/constexpr_hash.h

@@ -0,0 +1,38 @@
+#pragma once
+/// \file constexpr_hash.h
+/// A complile-time hashing function
+
+#include <stdint.h>
+#include <stddef.h>
+
+namespace kutil {
+
+constexpr static const uint8_t pearson_hash_table[256] = {
+	0x76,0x07,0xbe,0x47,0xcf,0x41,0x0a,0xe8,0x01,0x5c,0x9f,0xc5,0x24,0x63,0x9a,0x85,
+	0x39,0x2c,0xe2,0x34,0xb9,0xf2,0xae,0x40,0x10,0x90,0x94,0xd1,0x98,0x2d,0x16,0xfd,
+	0xc6,0x48,0x0d,0xce,0x74,0x43,0x28,0xf9,0x61,0x12,0xd0,0xcd,0xd8,0xd7,0xa8,0x78,
+	0x73,0x70,0xcc,0x1e,0x17,0xa7,0x87,0x38,0x68,0x91,0xc1,0x04,0x3f,0xf5,0xde,0xa3,
+	0x8a,0xe5,0x9b,0xec,0x97,0xd5,0x71,0x4a,0x20,0xca,0xc8,0xc4,0x83,0x53,0xe7,0x7b,
+	0x64,0x31,0x06,0xe0,0x7a,0xb6,0x52,0x8c,0xba,0x58,0xcb,0xb5,0x37,0x51,0x59,0xa1,
+	0x11,0xe3,0x5a,0xdb,0xe1,0x6d,0x46,0x62,0xaf,0xbd,0x57,0xb8,0x0e,0xf4,0xdd,0xa6,
+	0x45,0xf8,0x35,0x42,0x56,0xdf,0xad,0x80,0xb2,0x0b,0x5b,0xd4,0x86,0xb3,0xf0,0xc9,
+	0x3c,0xa5,0xc0,0x8e,0x55,0x77,0xeb,0x36,0x79,0xab,0x4c,0x25,0xed,0xa9,0x75,0x8f,
+	0xee,0xc2,0x72,0x8b,0x60,0x2a,0xfa,0x32,0xe9,0xda,0x03,0x1b,0x27,0x69,0x18,0x9e,
+	0x88,0x96,0x54,0x81,0x30,0x22,0x7c,0x4f,0xc7,0xef,0x5d,0xa4,0x67,0x44,0xc3,0x99,
+	0xbb,0xd3,0x8d,0x65,0xb1,0x82,0x09,0x1a,0x13,0xd9,0x9c,0x4d,0xb0,0xfc,0xac,0xbc,
+	0x6a,0x29,0x95,0x19,0x92,0xaa,0x49,0x7d,0x3b,0xfb,0x50,0xb7,0xf3,0x5e,0x3e,0x6b,
+	0x3a,0x14,0x2b,0xb4,0xfe,0xe6,0x93,0x23,0xd6,0x1f,0xd2,0x0c,0x1d,0x9d,0x6c,0x66,
+	0x1c,0x89,0xbf,0xf6,0xff,0x6f,0x84,0x6e,0x2e,0xea,0x21,0xf7,0x7f,0x33,0xf1,0xe4,
+	0x3d,0x0f,0x05,0x08,0x4e,0xa2,0xa0,0x2f,0xdc,0x00,0x5f,0x15,0x7e,0x02,0x4b,0x26
+};
+
+constexpr inline uint8_t pearson_hash_8(const char *s, uint8_t inv) {
+	return (*s) ? pearson_hash_8(s + 1, pearson_hash_table[inv ^ *s]) : inv;
+}
+
+} // namespace kutil
+
+constexpr inline uint8_t operator "" _h (const char *s, size_t len) {
+	return kutil::pearson_hash_8(s, static_cast<uint8_t>(len & 0xff));
+}
+

src/libraries/kutil/include/kutil/frame_allocator.h

@@ -1,122 +0,0 @@
-#pragma once
-/// \file frame_allocator.h
-/// Allocator for physical memory frames
-
-#include <stdint.h>
-
-#include "kutil/enum_bitfields.h"
-#include "kutil/linked_list.h"
-#include "kutil/slab_allocator.h"
-
-namespace kutil {
-
-struct frame_block;
-using frame_block_list = kutil::linked_list<frame_block>;
-using frame_block_slab = kutil::slab_allocator<frame_block>;
-
-/// Allocator for physical memory frames
-class frame_allocator
-{
-public:
-	/// Size of a single page frame.
-	static const size_t frame_size = 0x1000;
-
-	/// Default constructor
-	frame_allocator() = default;
-
-	/// Constructor with a provided initial frame_block cache.
-	/// \arg cache  List of pre-allocated but unused frame_block structures
-	frame_allocator(frame_block_list cache);
-
-	/// Initialize the frame allocator from bootstraped memory.
-	/// \arg free   List of free blocks
-	/// \arg used   List of currently used blocks
-	void init(
-			frame_block_list free,
-			frame_block_list used);
-
-	/// Get free frames from the free list. Only frames from the first free block
-	/// are returned, so the number may be less than requested, but they will
-	/// be contiguous.
-	/// \arg count    The maximum number of frames to get
-	/// \arg address  [out] The physical address of the first frame
-	/// \returns      The number of frames retrieved
-	size_t allocate(size_t count, uintptr_t *address);
-
-	/// Free previously allocated frames.
-	/// \arg address  The physical address of the first frame to free
-	/// \arg count    The number of frames to be freed
-	void free(uintptr_t address, size_t count);
-
-	/// Consolidate the free and used block lists. Return freed blocks
-	/// to the cache.
-	void consolidate_blocks();
-
-private:
-	frame_block_list m_free; ///< Free frames list
-	frame_block_list m_used; ///< In-use frames list
-	frame_block_slab m_block_slab; ///< frame_block slab allocator
-
-	frame_allocator(const frame_allocator &) = delete;
-};
-
-
-/// Flags used by `frame_block`.
-enum class frame_block_flags : uint32_t
-{
-	none         = 0x0000,
-
-	mmio         = 0x0001,  ///< Memory is a MMIO region
-	nonvolatile  = 0x0002,  ///< Memory is non-volatile storage
-
-	pending_free = 0x0020,  ///< Memory should be freed
-	acpi_wait    = 0x0040,  ///< Memory should be freed after ACPI init
-	permanent    = 0x0080,  ///< Memory is permanently unusable
-
-	// The following are used only during the memory bootstraping
-	// process, and tell the page manager where to initially map
-	// the given block.
-	map_ident    = 0x0100,  ///< Identity map
-	map_kernel   = 0x0200,  ///< Map into normal kernel space
-	map_offset   = 0x0400,  ///< Map into offset kernel space
-	map_mask     = 0x0700,  ///< Mask of all map_* values
-};
-
-} // namespace kutil
-
-IS_BITFIELD(kutil::frame_block_flags);
-
-namespace kutil {
-
-/// A block of contiguous frames. Each `frame_block` represents contiguous
-/// physical frames with the same attributes.
-struct frame_block
-{
-	uintptr_t address;
-	uint32_t count;
-	frame_block_flags flags;
-
-	inline bool has_flag(frame_block_flags f) const { return bitfield_has(flags, f); }
-	inline uintptr_t end() const { return address + (count * frame_allocator::frame_size); }
-	inline bool contains(uintptr_t addr) const { return addr >= address && addr < end(); }
-
-	/// Helper to zero out a block and optionally set the next pointer.
-	void zero();
-
-	/// Helper to copy a bock from another block
-	/// \arg other  The block to copy from
-	void copy(frame_block *other);
-
-	/// Compare two blocks by address.
-	/// \arg rhs   The right-hand comparator
-	/// \returns   <0 if this is sorts earlier, >0 if this sorts later, 0 for equal
-	int compare(const frame_block *rhs) const;
-
-	/// Traverse the list, joining adjacent blocks where possible.
-	/// \arg list  The list to consolidate
-	/// \returns   A linked list of freed frame_block structures.
-	static frame_block_list consolidate(frame_block_list &list);
-};
-
-} // namespace kutil
-

src/libraries/kutil/include/kutil/heap_allocator.h

@@ -1,49 +1,45 @@
 #pragma once
-/// \file heap_manager.h
-/// A buddy allocator and related definitions.
+/// \file heap_allocator.h
+/// A buddy allocator for a memory heap
 
 #include <stddef.h>
+#include "kutil/allocator.h"
 
 namespace kutil {
 
 
-/// Manager for allocation of heap memory.
-class heap_manager
+/// Allocator for a given heap range
+class heap_allocator :
+	public allocator
 {
 public:
-	/// Callback signature for growth function. Memory returned does not need
-	/// to be contiguous, but needs to be alined to the length requested.
-	using grow_callback = void * (*)(size_t length);
+	/// Default constructor creates a valid but empty heap.
+	heap_allocator();
 
-	/// Default constructor. Creates an invalid manager.
-	heap_manager();
-
-	/// Constructor.
-	/// \arg grow_cb  Function pointer to grow the heap size
-	heap_manager(grow_callback grow_cb);
+	/// Constructor. The given memory area must already have been reserved.
+	/// \arg start  Starting address of the heap
+	/// \arg size   Size of the heap in bytes
+	heap_allocator(uintptr_t start, size_t size);
 
 	/// Allocate memory from the area managed.
 	/// \arg length  The amount of memory to allocate, in bytes
 	/// \returns     A pointer to the allocated memory, or nullptr if
 	///              allocation failed.
-	void * allocate(size_t length);
+	virtual void * allocate(size_t length) override;
 
 	/// Free a previous allocation.
 	/// \arg p  A pointer previously retuned by allocate()
-	void free(void *p);
+	virtual void free(void *p) override;
 
 	/// Minimum block size is (2^min_size). Must be at least 6.
 	static const unsigned min_size = 6;
 
 	/// Maximum block size is (2^max_size). Must be less than 64.
-	static const unsigned max_size = 16;
+	static const unsigned max_size = 22;
 
 protected:
 	class mem_header;
 
-	/// Expand the size of memory
-	void grow_memory();
-
 	/// Ensure there is a block of a given size, recursively splitting
 	/// \arg size   Size category of the block we want
 	void ensure_block(unsigned size);
@@ -58,11 +54,11 @@ protected:
 	/// \returns    A detached block of the given size
 	mem_header * pop_free(unsigned size);
 
+	uintptr_t m_next;
+	size_t m_size;
 	mem_header *m_free[max_size - min_size + 1];
 
-	grow_callback m_grow;
-
-	heap_manager(const heap_manager &) = delete;
+	heap_allocator(const heap_allocator &) = delete;
 };
 
 } // namespace kutil

src/libraries/kutil/include/kutil/linked_list.h

@@ -116,27 +116,47 @@ public:
 	/// Constructor. Creates an empty list.
 	linked_list() :
 		m_head(nullptr),
-		m_tail(nullptr)
+		m_tail(nullptr),
+		m_count(0)
 	{}
 
 	/// Move constructor. Takes ownership of list elements.
 	linked_list(linked_list<T> &&other) :
 		m_head(other.m_head),
-		m_tail(other.m_tail)
+		m_tail(other.m_tail),
+		m_count(other.m_count)
 	{
 		other.m_head = other.m_tail = nullptr;
+		other.m_count = 0;
+	}
+
+	/// Assignment operator. Takes ownership of list elements.
+	/// Destructive towards current data!
+	linked_list & operator=(linked_list &&other)
+	{
+		m_head = other.m_head;
+		m_tail = other.m_tail;
+		m_count = other.m_count;
+		other.m_head = other.m_tail = nullptr;
+		other.m_count = 0;
+		return *this;
 	}
 
 	/// Check if the list is empty.
 	/// \returns  true if the list is empty
 	bool empty() const { return m_head == nullptr; }
 
+	/// Get the cached length of the list.
+	/// \returns  The number of entries in the list.
+	size_t length() const { return m_count; }
+
 	/// Count the items in the list.
 	/// \returns  The number of entries in the list.
-	size_t length() const
+	size_t count_length()
 	{
 		size_t len = 0;
 		for (item_type *cur = m_head; cur; cur = cur->m_next) ++len;
+		m_count = len;
 		return len;
 	}
 
@@ -164,6 +184,8 @@ public:
 			item->m_prev = nullptr;
 			m_head = item;
 		}
+
+		m_count += 1;
 	}
 
 	/// Append an item to the end of this list.
@@ -182,6 +204,8 @@ public:
 			item->m_next = nullptr;
 			m_tail = item;
 		}
+
+		m_count += 1;
 	}
 
 	/// Remove an item from the front of this list.
@@ -217,6 +241,8 @@ public:
 			m_tail = list.m_tail;
 		}
 
+		m_count += list.m_count;
+		list.m_count = 0;
 		list.m_head = list.m_tail = nullptr;
 	}
 
@@ -235,6 +261,8 @@ public:
 			m_tail = list.m_tail;
 		}
 
+		m_count += list.m_count;
+		list.m_count = 0;
 		list.m_head = list.m_tail = nullptr;
 	}
 
@@ -248,6 +276,7 @@ public:
 		if (item == m_tail)
 			m_tail = item->m_prev;
 		item->remove();
+		m_count -= 1;
 	}
 
 	/// Inserts an item into the list before another given item.
@@ -257,12 +286,14 @@ public:
 	{
 		if (!item) return;
 
-		if (!existing)
+		if (!existing) {
 			push_back(item);
-		else if (existing == m_head)
+		} else if (existing == m_head) {
 			push_front(item);
-		else
+		} else {
 			existing->insert_before(item);
+			m_count += 1;
+		}
 	}
 
 	/// Inserts an item into the list after another given item.
@@ -272,12 +303,14 @@ public:
 	{
 		if (!item) return;
 
-		if (!existing)
+		if (!existing) {
 			push_front(item);
-		else if (existing == m_tail)
+		} else if (existing == m_tail) {
 			push_back(item);
-		else
+		} else {
 			existing->insert_after(item);
+			m_count += 1;
+		}
 	}
 
 	/// Insert an item into the list in a sorted position. Depends on T
@@ -309,6 +342,7 @@ public:
 private:
 	item_type *m_head;
 	item_type *m_tail;
+	size_t m_count;
 };
 
 

src/libraries/kutil/include/kutil/logger.h

@@ -0,0 +1,118 @@
+#pragma once
+/// \file logger.h
+/// Kernel logging facility.
+
+#include <stdarg.h>
+#include <stdint.h>
+
+#include "kutil/bip_buffer.h"
+
+namespace kutil {
+namespace log {
+
+using area_t = uint8_t;
+enum class level : uint8_t {
+	none, debug, info, warn, error, fatal, max
+};
+
+class logger
+{
+public:
+	/// Default constructor. Creates a logger without a backing store.
+	logger();
+
+	/// Constructor. Logs are written to the given buffer.
+	logger(uint8_t *buffer, size_t size);
+
+	/// Register a log area for future use.
+	/// \arg area      The key for the new area
+	/// \arg name      The area name
+	/// \arg verbosity What level of logs to print for this area
+	void register_area(area_t area, const char *name, level verbosity);
+
+	/// Callback type for immediate-mode logging
+	typedef void (*immediate)(area_t, level, const char *);
+
+	/// Register an immediate-mode log callback
+	inline void set_immediate(immediate cb) { m_immediate = cb; }
+
+	/// Get the default logger.
+	inline logger & get() { return *s_log; }
+
+	/// Get the registered name for a given area
+	inline const char * area_name(area_t area) const { return m_names[area]; }
+
+	/// Get the name of a level
+	inline const char * level_name(level l) const { return s_level_names[static_cast<unsigned>(l)]; }
+
+	/// Write to the log
+	/// \arg severity  The severity of the message
+	/// \arg area      The log area to write to
+	/// \arg fmt       A printf-like format string
+	inline void log(level severity, area_t area, const char *fmt, ...)
+	{
+		level limit = get_level(area);
+		if (limit == level::none || severity < limit)
+			return;
+
+		va_list args;
+		va_start(args, fmt);
+		output(severity, area, fmt, args);
+		va_end(args);
+	}
+
+	struct entry
+	{
+		uint8_t bytes;
+		area_t area;
+		level severity;
+		uint8_t sequence;
+		char message[0];
+	};
+
+	/// Get the next log entry from the buffer
+	/// \arg buffer  The buffer to copy the log message into
+	/// \arg size    Size of the passed-in buffer, in bytes
+	/// \returns     Number of bytes copied into the buffer
+	size_t get_entry(void *buffer, size_t size);
+
+private:
+	friend void debug(area_t area, const char *fmt, ...);
+	friend void info (area_t area, const char *fmt, ...);
+	friend void warn (area_t area, const char *fmt, ...);
+	friend void error(area_t area, const char *fmt, ...);
+	friend void fatal(area_t area, const char *fmt, ...);
+
+	void output(level severity, area_t area, const char *fmt, va_list args);
+
+	void set_level(area_t area, level l);
+	level get_level(area_t area);
+
+	static const unsigned num_areas = 1 << (sizeof(area_t) * 8);
+	uint8_t m_levels[num_areas / 2];
+	const char *m_names[num_areas];
+	immediate m_immediate;
+
+	uint8_t m_sequence;
+
+	kutil::bip_buffer m_buffer;
+
+	static logger *s_log;
+	static const char *s_level_names[static_cast<unsigned>(level::max)];
+};
+
+void debug(area_t area, const char *fmt, ...);
+void info (area_t area, const char *fmt, ...);
+void warn (area_t area, const char *fmt, ...);
+void error(area_t area, const char *fmt, ...);
+void fatal(area_t area, const char *fmt, ...);
+
+} // namespace log
+
+namespace logs {
+#define LOG(name, lvl) extern log::area_t name;
+#include "log_areas.inc"
+#undef LOG
+} // namespace logs
+
+} // namespace kutil

src/libraries/kutil/include/kutil/memory.h

@@ -9,15 +9,6 @@ void * operator new (size_t, void *p) noexcept;
 
 namespace kutil {
 
-/// Allocate memory.
-/// \arg n   The number of bytes to allocate
-/// \returns The allocated memory
-void * malloc(size_t n);
-
-/// Free memory allocated by malloc().
-/// \arg p   A pointer previously returned by malloc()
-void free(void *p);
-
 /// Fill memory with the given value.
 /// \arg p   The beginning of the memory area to fill
 /// \arg v   The byte value to fill memory with
@@ -67,14 +58,4 @@ inline T* mask_pointer(T *p, uintptr_t mask)
 /// \arg off  An optional offset into the region
 uint8_t checksum(const void *p, size_t len, size_t off = 0);
 
-
-class heap_manager;
-
-namespace setup {
-
-/// Set the heap that malloc() / free() will use.
-/// \arg mm  The heap manager for the heap to use.
-void set_heap(heap_manager *mm);
-
-} // namespace kutil::setup
 } // namespace kutil

src/libraries/kutil/include/kutil/printf.h

@@ -0,0 +1,96 @@
+///////////////////////////////////////////////////////////////////////////////
+// \author (c) Marco Paland (info@paland.com)
+//             2014-2018, PALANDesign Hannover, Germany
+//
+// \license The MIT License (MIT)
+//
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files (the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+// copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions:
+// 
+// The above copyright notice and this permission notice shall be included in
+// all copies or substantial portions of the Software.
+// 
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+// THE SOFTWARE.
+//
+// \brief Tiny printf, sprintf and snprintf implementation, optimized for speed on
+//        embedded systems with a very limited resources.
+//        Use this instead of bloated standard/newlib printf.
+//        These routines are thread safe and reentrant.
+//
+///////////////////////////////////////////////////////////////////////////////
+
+#ifndef _PRINTF_H_
+#define _PRINTF_H_
+
+#include <stdarg.h>
+#include <stddef.h>
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * Tiny printf implementation
+ * You have to implement _putchar if you use printf()
+ * To avoid conflicts with the regular printf() API it is overridden by macro defines
+ * and internal underscore-appended functions like printf_() are used
+ * \param format A string that specifies the format of the output
+ * \return The number of characters that are written into the array, not counting the terminating null character
+ */
+#define printf printf_
+int printf_(const char* format, ...);
+
+
+/**
+ * Tiny sprintf implementation
+ * Due to security reasons (buffer overflow) YOU SHOULD CONSIDER USING (V)SNPRINTF INSTEAD!
+ * \param buffer A pointer to the buffer where to store the formatted string. MUST be big enough to store the output!
+ * \param format A string that specifies the format of the output
+ * \return The number of characters that are WRITTEN into the buffer, not counting the terminating null character
+ */
+#define sprintf sprintf_
+int sprintf_(char* buffer, const char* format, ...);
+
+
+/**
+ * Tiny snprintf/vsnprintf implementation
+ * \param buffer A pointer to the buffer where to store the formatted string
+ * \param count The maximum number of characters to store in the buffer, including a terminating null character
+ * \param format A string that specifies the format of the output
+ * \return The number of characters that are WRITTEN into the buffer, not counting the terminating null character
+ *         If the formatted string is truncated the buffer size (count) is returned
+ */
+#define snprintf  snprintf_
+#define vsnprintf vsnprintf_
+int  snprintf_(char* buffer, size_t count, const char* format, ...);
+int vsnprintf_(char* buffer, size_t count, const char* format, va_list va);
+
+
+/**
+ * printf with output function
+ * You may use this as dynamic alternative to printf() with its fixed _putchar() output
+ * \param out An output function which takes one character and an argument pointer
+ * \param arg An argument pointer for user data passed to output function
+ * \param format A string that specifies the format of the output
+ * \return The number of characters that are sent to the output function, not counting the terminating null character
+ */
+int fctprintf(void (*out)(char character, void* arg), void* arg, const char* format, ...);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+
+#endif  // _PRINTF_H_

src/libraries/kutil/include/kutil/slab_allocator.h

@@ -1,6 +1,7 @@
 #pragma once
 /// \file slab_allocator.h
 /// A slab allocator and related definitions
+#include "kutil/allocator.h"
 #include "kutil/assert.h"
 #include "kutil/linked_list.h"
 #include "kutil/memory.h"
@@ -9,7 +10,7 @@ namespace kutil {
 
 
 /// A slab allocator for small structures kept in a linked list
-template <typename T, typename Alloc = void * (*)(size_t)>
+template <typename T, size_t N = memory::frame_size>
 class slab_allocator :
 	public linked_list<T>
 {
@@ -19,8 +20,7 @@ public:
 	/// Default constructor.
 	/// \arg chunk_size The size of chunk to allocate, in bytes. 0 means default.
 	/// \arg alloc      The allocator to use to allocate chunks. Defaults to malloc().
-	slab_allocator(size_t chunk_size = 0, Alloc alloc = malloc) :
-		m_chunk_size(chunk_size),
+	slab_allocator(allocator &alloc) :
 		m_alloc(alloc)
 	{
 	}
@@ -45,18 +45,16 @@ public:
 
 	void allocate()
 	{
-		size_t size = m_chunk_size ? m_chunk_size : 10 * sizeof(item_type);
-		void *memory = m_alloc(size);
-		size_t count = size / sizeof(item_type);
+		constexpr unsigned count = N / sizeof(item_type);
 
+		void *memory = m_alloc.allocate(N);
 		item_type *items = reinterpret_cast<item_type *>(memory); 
 		for (size_t i = 0; i < count; ++i)
 			this->push_back(&items[i]);
 	}
 
 private:
-	size_t m_chunk_size;
-	Alloc m_alloc;
+	allocator& m_alloc;
 };
 
 } // namespace kutil

src/libraries/kutil/include/kutil/vector.h

@@ -4,6 +4,7 @@
 
 #include <algorithm>
 #include <utility>
+#include "kutil/allocator.h"
 #include "kutil/memory.h"
 
 namespace kutil {
@@ -14,18 +15,20 @@ class vector
 {
 public:
 	/// Default constructor. Creates an empty vector with no capacity.
-	vector() :
+	vector(kutil::allocator &alloc = allocator::invalid) :
 		m_size(0),
 		m_capacity(0),
-		m_elements(nullptr)
+		m_elements(nullptr),
+		m_alloc(alloc)
 	{}
 
 	/// Constructor. Creates an empty array with capacity.
 	/// \arg capacity  Initial capacity to allocate
-	vector(size_t capacity) :
+	vector(size_t capacity, allocator &alloc) :
 		m_size(0),
 		m_capacity(0),
-		m_elements(nullptr)
+		m_elements(nullptr),
+		m_alloc(alloc)
 	{
 		set_capacity(capacity);
 	}
@@ -34,7 +37,8 @@ public:
 	vector(const vector& other) :
 		m_size(0),
 		m_capacity(0),
-		m_elements(nullptr)
+		m_elements(nullptr),
+		m_alloc(other.m_alloc)
 	{
 		set_capacity(other.m_capacity);
 		kutil::memcpy(m_elements, other.m_elements, other.m_size * sizeof(T));
@@ -45,7 +49,8 @@ public:
 	vector(vector&& other) :
 		m_size(other.m_size),
 		m_capacity(other.m_capacity),
-		m_elements(other.m_elements)
+		m_elements(other.m_elements),
+		m_alloc(other.m_alloc)
 	{
 		other.m_size = 0;
 		other.m_capacity = 0;
@@ -142,7 +147,7 @@ public:
 	/// \arg capacity  Number of elements to allocate
 	void set_capacity(size_t capacity)
 	{
-		T *new_array = reinterpret_cast<T *>(malloc(capacity * sizeof(T)));
+		T *new_array = m_alloc.allocate<T>(capacity);
 		size_t size = std::min(capacity, m_size);
 
 		kutil::memcpy(new_array, m_elements, size * sizeof(T));
@@ -151,7 +156,7 @@ public:
 		m_size = size;
 		m_capacity = capacity;
 
-		delete [] m_elements;
+		m_alloc.free(m_elements);
 		m_elements = new_array;
 	}
 
@@ -159,6 +164,7 @@ private:
 	size_t m_size;
 	size_t m_capacity;
 	T *m_elements;
+	allocator &m_alloc;
 };
 
 } // namespace kutil

src/libraries/kutil/logger.cpp

@@ -0,0 +1,165 @@
+#include "kutil/assert.h"
+#include "kutil/constexpr_hash.h"
+#include "kutil/logger.h"
+#include "kutil/memory.h"
+#include "kutil/printf.h"
+
+namespace kutil {
+namespace logs {
+#define LOG(name, lvl) \
+	log::area_t name = #name ## _h; \
+	const char * name ## _name = #name;
+#include "log_areas.inc"
+#undef LOG
+}
+
+namespace log {
+
+using kutil::memset;
+using kutil::memcpy;
+
+logger *logger::s_log = nullptr;
+const char *logger::s_level_names[] = {"", "debug", " info", " warn", "error", "fatal"};
+
+logger::logger() :
+	m_buffer(nullptr, 0),
+	m_immediate(nullptr),
+	m_sequence(0)
+{
+	memset(&m_levels, 0, sizeof(m_levels));
+	memset(&m_names, 0, sizeof(m_names));
+	s_log = this;
+}
+
+logger::logger(uint8_t *buffer, size_t size) :
+	m_buffer(buffer, size),
+	m_immediate(nullptr),
+	m_sequence(0)
+{
+	memset(&m_levels, 0, sizeof(m_levels));
+	memset(&m_names, 0, sizeof(m_names));
+	s_log = this;
+
+#define LOG(name, lvl) \
+	register_area(logs::name, logs::name ## _name, log::level::lvl);
+#include "log_areas.inc"
+#undef LOG
+}
+
+void
+logger::set_level(area_t area, level l)
+{
+	unsigned uarea = static_cast<unsigned>(area);
+	uint8_t ulevel = static_cast<uint8_t>(l) & 0x0f;
+	uint8_t &flags = m_levels[uarea / 2];
+	if (uarea & 1)
+		flags = (flags & 0x0f) | (ulevel << 4);
+	else
+		flags = (flags & 0xf0) | ulevel;
+}
+
+level
+logger::get_level(area_t area)
+{
+	unsigned uarea = static_cast<unsigned>(area);
+	uint8_t &flags = m_levels[uarea / 2];
+	if (uarea & 1)
+		return static_cast<level>((flags & 0xf0) >> 4);
+	else
+		return static_cast<level>(flags & 0x0f);
+}
+
+void
+logger::register_area(area_t area, const char *name, level verbosity)
+{
+	m_names[area] = name;
+	set_level(area, verbosity);
+}
+
+void
+logger::output(level severity, area_t area, const char *fmt, va_list args)
+{
+	uint8_t buffer[256];
+	entry *header = reinterpret_cast<entry *>(buffer);
+	header->bytes = sizeof(entry);
+	header->area = area;
+	header->severity = severity;
+	header->sequence = m_sequence++;
+
+	header->bytes +=
+		vsnprintf(header->message, sizeof(buffer) - sizeof(entry), fmt, args);
+
+	if (m_immediate) {
+		buffer[header->bytes] = 0;
+		m_immediate(area, severity, header->message);
+		return;
+	}
+
+	uint8_t *out;
+	size_t n = m_buffer.reserve(header->bytes, reinterpret_cast<void**>(&out));
+	if (n < sizeof(entry)) {
+		m_buffer.commit(0); // Cannot even write the header, abort
+		return;
+	}
+
+	if (n < header->bytes)
+		header->bytes = n;
+
+	memcpy(out, buffer, n);
+	m_buffer.commit(n);
+}
+
+size_t
+logger::get_entry(void *buffer, size_t size)
+{
+	void *out;
+	size_t out_size = m_buffer.get_block(&out);
+	entry *ent = reinterpret_cast<entry *>(out);
+	if (out_size == 0 || out == 0)
+		return 0;
+
+	kassert(out_size >= sizeof(entry), "Couldn't read a full entry");
+	if (out_size < sizeof(entry))
+		return 0;
+
+	kassert(size >= ent->bytes, "Didn't pass a big enough buffer");
+	if (size < ent->bytes)
+		return 0;
+
+	memcpy(buffer, out, ent->bytes);
+	m_buffer.consume(ent->bytes);
+	return ent->bytes;
+}
+
+#define LOG_LEVEL_FUNCTION(name) \
+	void name (area_t area, const char *fmt, ...) { \
+		logger *l = logger::s_log; \
+		if (!l) return; \
+		level limit = l->get_level(area); \
+		if (limit == level::none || level::name < limit) return; \
+		va_list args; \
+		va_start(args, fmt); \
+		l->output(level::name, area, fmt, args); \
+		va_end(args); \
+	}
+
+LOG_LEVEL_FUNCTION(debug);
+LOG_LEVEL_FUNCTION(info);
+LOG_LEVEL_FUNCTION(warn);
+LOG_LEVEL_FUNCTION(error);
+
+void fatal(area_t area, const char *fmt, ...)
+{
+	logger *l = logger::s_log;
+	if (!l) return;
+
+	va_list args;
+	va_start(args, fmt);
+	l->output(level::fatal, area, fmt, args);
+	va_end(args);
+
+	kassert(false, "log::fatal");
+}
+
+} // namespace log
+} // namespace kutil

src/libraries/kutil/memory.cpp

@@ -1,46 +1,11 @@
 #include "kutil/memory.h"
-#include "kutil/heap_manager.h"
 
 namespace std {
 	enum class __attribute__ ((__type_visibility("default"))) align_val_t : size_t { };
 }
 
-#ifdef __POPCORN__
-void * operator new(size_t n, std::align_val_t) { return kutil::malloc(n); }
-void * operator new (size_t n)					{ return kutil::malloc(n); }
-void * operator new[] (size_t n)				{ return kutil::malloc(n); }
-void operator delete (void *p) noexcept			{ return kutil::free(p); }
-void operator delete[] (void *p) noexcept		{ return kutil::free(p); }
-#endif
-
 namespace kutil {
 
-namespace setup {
-
-static heap_manager *heap_memory_manager;
-
-void
-set_heap(heap_manager *mm)
-{
-	setup::heap_memory_manager = mm;
-}
-
-} // namespace kutil::setup
-
-
-void *
-malloc(size_t n)
-{
-	return setup::heap_memory_manager->allocate(n);
-}
-
-void
-free(void *p)
-{
-	setup::heap_memory_manager->free(p);
-}
-
-
 void *
 memset(void *s, uint8_t v, size_t n)
 {

src/libraries/kutil/printf.c

@@ -0,0 +1,766 @@
+///////////////////////////////////////////////////////////////////////////////
+// \author (c) Marco Paland (info@paland.com)
+//             2014-2019, PALANDesign Hannover, Germany
+//
+// \license The MIT License (MIT)
+//
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files (the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+// copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions:
+//
+// The above copyright notice and this permission notice shall be included in
+// all copies or substantial portions of the Software.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+// THE SOFTWARE.
+//
+// \brief Tiny printf, sprintf and (v)snprintf implementation, optimized for speed on
+//        embedded systems with a very limited resources. These routines are thread
+//        safe and reentrant!
+//        Use this instead of the bloated standard/newlib printf cause these use
+//        malloc for printf (and may not be thread safe).
+//
+///////////////////////////////////////////////////////////////////////////////
+
+#include <stdbool.h>
+#include <stdint.h>
+
+#include "kutil/printf.h"
+
+
+// define this globally (e.g. gcc -DPRINTF_INCLUDE_CONFIG_H ...) to include the
+// printf_config.h header file
+// default: undefined
+#ifdef PRINTF_INCLUDE_CONFIG_H
+#include "printf_config.h"
+#endif
+
+
+// 'ntoa' conversion buffer size, this must be big enough to hold one converted
+// numeric number including padded zeros (dynamically created on stack)
+// default: 32 byte
+#ifndef PRINTF_NTOA_BUFFER_SIZE
+#define PRINTF_NTOA_BUFFER_SIZE    32U
+#endif
+
+// 'ftoa' conversion buffer size, this must be big enough to hold one converted
+// float number including padded zeros (dynamically created on stack)
+// default: 32 byte
+#ifndef PRINTF_FTOA_BUFFER_SIZE
+#define PRINTF_FTOA_BUFFER_SIZE    32U
+#endif
+
+// support for the floating point type (%f)
+// default: activated
+#ifndef PRINTF_DISABLE_SUPPORT_FLOAT
+#define PRINTF_SUPPORT_FLOAT
+#endif
+
+// support for the long long types (%llu or %p)
+// default: activated
+#ifndef PRINTF_DISABLE_SUPPORT_LONG_LONG
+#define PRINTF_SUPPORT_LONG_LONG
+#endif
+
+// support for the ptrdiff_t type (%t)
+// ptrdiff_t is normally defined in <stddef.h> as long or long long type
+// default: activated
+#ifndef PRINTF_DISABLE_SUPPORT_PTRDIFF_T
+#define PRINTF_SUPPORT_PTRDIFF_T
+#endif
+
+///////////////////////////////////////////////////////////////////////////////
+
+// internal flag definitions
+#define FLAGS_ZEROPAD   (1U <<  0U)
+#define FLAGS_LEFT      (1U <<  1U)
+#define FLAGS_PLUS      (1U <<  2U)
+#define FLAGS_SPACE     (1U <<  3U)
+#define FLAGS_HASH      (1U <<  4U)
+#define FLAGS_UPPERCASE (1U <<  5U)
+#define FLAGS_CHAR      (1U <<  6U)
+#define FLAGS_SHORT     (1U <<  7U)
+#define FLAGS_LONG      (1U <<  8U)
+#define FLAGS_LONG_LONG (1U <<  9U)
+#define FLAGS_PRECISION (1U << 10U)
+
+
+// output function type
+typedef void (*out_fct_type)(char character, void* buffer, size_t idx, size_t maxlen);
+
+
+// wrapper (used as buffer) for output function type
+typedef struct {
+  void  (*fct)(char character, void* arg);
+  void* arg;
+} out_fct_wrap_type;
+
+
+// internal buffer output
+static inline void _out_buffer(char character, void* buffer, size_t idx, size_t maxlen)
+{
+  if (idx < maxlen) {
+    ((char*)buffer)[idx] = character;
+  }
+}
+
+
+// internal null output
+static inline void _out_null(char character, void* buffer, size_t idx, size_t maxlen)
+{
+  (void)character; (void)buffer; (void)idx; (void)maxlen;
+}
+
+#ifdef PRINTF_PUTCHAR_FUNC
+// internal screen ouput wrapper
+static inline void _out_char(char character, void* buffer, size_t idx, size_t maxlen)
+{
+  (void)buffer; (void)idx; (void)maxlen;
+  if (character) {
+    PRINTF_PUTCHAR_FUNC (character);
+  }
+}
+#endif
+
+// internal output function wrapper
+static inline void _out_fct(char character, void* buffer, size_t idx, size_t maxlen)
+{
+  (void)idx; (void)maxlen;
+  if (character) {
+    // buffer is the output fct pointer
+    ((out_fct_wrap_type*)buffer)->fct(character, ((out_fct_wrap_type*)buffer)->arg);
+  }
+}
+
+
+// internal secure strlen
+// \return The length of the string (excluding the terminating 0) limited by 'maxsize'
+static inline unsigned int _strnlen_s(const char* str, size_t maxsize)
+{
+  const char* s;
+  for (s = str; *s && maxsize--; ++s);
+  return (unsigned int)(s - str);
+}
+
+
+// internal test if char is a digit (0-9)
+// \return true if char is a digit
+static inline bool _is_digit(char ch)
+{
+  return (ch >= '0') && (ch <= '9');
+}
+
+
+// internal ASCII string to unsigned int conversion
+static unsigned int _atoi(const char** str)
+{
+  unsigned int i = 0U;
+  while (_is_digit(**str)) {
+    i = i * 10U + (unsigned int)(*((*str)++) - '0');
+  }
+  return i;
+}
+
+
+// internal itoa format
+static size_t _ntoa_format(out_fct_type out, char* buffer, size_t idx, size_t maxlen, char* buf, size_t len, bool negative, unsigned int base, unsigned int prec, unsigned int width, unsigned int flags)
+{
+  const size_t start_idx = idx;
+
+  // pad leading zeros
+  if (!(flags & FLAGS_LEFT)) {
+    if (width && (flags & FLAGS_ZEROPAD) && (negative || (flags & (FLAGS_PLUS | FLAGS_SPACE)))) {
+      width--;
+    }
+    while ((len < prec) && (len < PRINTF_NTOA_BUFFER_SIZE)) {
+      buf[len++] = '0';
+    }
+    while ((flags & FLAGS_ZEROPAD) && (len < width) && (len < PRINTF_NTOA_BUFFER_SIZE)) {
+      buf[len++] = '0';
+    }
+  }
+
+  // handle hash
+  if (flags & FLAGS_HASH) {
+    if (!(flags & FLAGS_PRECISION) && len && ((len == prec) || (len == width))) {
+      len--;
+      if (len && (base == 16U)) {
+        len--;
+      }
+    }
+    if ((base == 16U) && !(flags & FLAGS_UPPERCASE) && (len < PRINTF_NTOA_BUFFER_SIZE)) {
+      buf[len++] = 'x';
+    }
+    else if ((base == 16U) && (flags & FLAGS_UPPERCASE) && (len < PRINTF_NTOA_BUFFER_SIZE)) {
+      buf[len++] = 'X';
+    }
+    else if ((base == 2U) && (len < PRINTF_NTOA_BUFFER_SIZE)) {
+      buf[len++] = 'b';
+    }
+    if (len < PRINTF_NTOA_BUFFER_SIZE) {
+      buf[len++] = '0';
+    }
+  }
+
+  if (len < PRINTF_NTOA_BUFFER_SIZE) {
+    if (negative) {
+      buf[len++] = '-';
+    }
+    else if (flags & FLAGS_PLUS) {
+      buf[len++] = '+';  // ignore the space if the '+' exists
+    }
+    else if (flags & FLAGS_SPACE) {
+      buf[len++] = ' ';
+    }
+  }
+
+  // pad spaces up to given width
+  if (!(flags & FLAGS_LEFT) && !(flags & FLAGS_ZEROPAD)) {
+    for (size_t i = len; i < width; i++) {
+      out(' ', buffer, idx++, maxlen);
+    }
+  }
+
+  // reverse string
+  for (size_t i = 0U; i < len; i++) {
+    out(buf[len - i - 1U], buffer, idx++, maxlen);
+  }
+
+  // append pad spaces up to given width
+  if (flags & FLAGS_LEFT) {
+    while (idx - start_idx < width) {
+      out(' ', buffer, idx++, maxlen);
+    }
+  }
+
+  return idx;
+}
+
+
+// internal itoa for 'long' type
+static size_t _ntoa_long(out_fct_type out, char* buffer, size_t idx, size_t maxlen, unsigned long value, bool negative, unsigned long base, unsigned int prec, unsigned int width, unsigned int flags)
+{
+  char buf[PRINTF_NTOA_BUFFER_SIZE];
+  size_t len = 0U;
+
+  // no hash for 0 values
+  if (!value) {
+    flags &= ~FLAGS_HASH;
+  }
+
+  // write if precision != 0 and value is != 0
+  if (!(flags & FLAGS_PRECISION) || value) {
+    do {
+      const char digit = (char)(value % base);
+      buf[len++] = digit < 10 ? '0' + digit : (flags & FLAGS_UPPERCASE ? 'A' : 'a') + digit - 10;
+      value /= base;
+    } while (value && (len < PRINTF_NTOA_BUFFER_SIZE));
+  }
+
+  return _ntoa_format(out, buffer, idx, maxlen, buf, len, negative, (unsigned int)base, prec, width, flags);
+}
+
+
+// internal itoa for 'long long' type
+#if defined(PRINTF_SUPPORT_LONG_LONG)
+static size_t _ntoa_long_long(out_fct_type out, char* buffer, size_t idx, size_t maxlen, unsigned long long value, bool negative, unsigned long long base, unsigned int prec, unsigned int width, unsigned int flags)
+{
+  char buf[PRINTF_NTOA_BUFFER_SIZE];
+  size_t len = 0U;
+
+  // no hash for 0 values
+  if (!value) {
+    flags &= ~FLAGS_HASH;
+  }
+
+  // write if precision != 0 and value is != 0
+  if (!(flags & FLAGS_PRECISION) || value) {
+    do {
+      const char digit = (char)(value % base);
+      buf[len++] = digit < 10 ? '0' + digit : (flags & FLAGS_UPPERCASE ? 'A' : 'a') + digit - 10;
+      value /= base;
+    } while (value && (len < PRINTF_NTOA_BUFFER_SIZE));
+  }
+
+  return _ntoa_format(out, buffer, idx, maxlen, buf, len, negative, (unsigned int)base, prec, width, flags);
+}
+#endif  // PRINTF_SUPPORT_LONG_LONG
+
+
+#if defined(PRINTF_SUPPORT_FLOAT)
+static size_t _ftoa(out_fct_type out, char* buffer, size_t idx, size_t maxlen, double value, unsigned int prec, unsigned int width, unsigned int flags)
+{
+  const size_t start_idx = idx;
+
+  char buf[PRINTF_FTOA_BUFFER_SIZE];
+  size_t len  = 0U;
+  double diff = 0.0;
+
+  // if input is larger than thres_max, revert to exponential
+  const double thres_max = (double)0x7FFFFFFF;
+
+  // powers of 10
+  static const double pow10[] = { 1, 10, 100, 1000, 10000, 100000, 1000000, 10000000, 100000000, 1000000000 };
+
+  // test for NaN
+  if (value != value) {
+    out('n', buffer, idx++, maxlen);
+    out('a', buffer, idx++, maxlen);
+    out('n', buffer, idx++, maxlen);
+    return idx;
+  }
+
+  // test for negative
+  bool negative = false;
+  if (value < 0) {
+    negative = true;
+    value = 0 - value;
+  }
+
+  // set default precision to 6, if not set explicitly
+  if (!(flags & FLAGS_PRECISION)) {
+    prec = 6U;
+  }
+  // limit precision to 9, cause a prec >= 10 can lead to overflow errors
+  while ((len < PRINTF_FTOA_BUFFER_SIZE) && (prec > 9U)) {
+    buf[len++] = '0';
+    prec--;
+  }
+
+  int whole = (int)value;
+  double tmp = (value - whole) * pow10[prec];
+  unsigned long frac = (unsigned long)tmp;
+  diff = tmp - frac;
+
+  if (diff > 0.5) {
+    ++frac;
+    // handle rollover, e.g. case 0.99 with prec 1 is 1.0
+    if (frac >= pow10[prec]) {
+      frac = 0;
+      ++whole;
+    }
+  }
+  else if (diff < 0.5) {
+  }
+  else if ((frac == 0U) || (frac & 1U)) {
+    // if halfway, round up if odd OR if last digit is 0
+    ++frac;
+  }
+
+  // TBD: for very large numbers switch back to native sprintf for exponentials. Anyone want to write code to replace this?
+  // Normal printf behavior is to print EVERY whole number digit which can be 100s of characters overflowing your buffers == bad
+  if (value > thres_max) {
+    return 0U;
+  }
+
+  if (prec == 0U) {
+    diff = value - (double)whole;
+    if ((!(diff < 0.5) || (diff > 0.5)) && (whole & 1)) {
+      // exactly 0.5 and ODD, then round up
+      // 1.5 -> 2, but 2.5 -> 2
+      ++whole;
+    }
+  }
+  else {
+    unsigned int count = prec;
+    // now do fractional part, as an unsigned number
+    while (len < PRINTF_FTOA_BUFFER_SIZE) {
+      --count;
+      buf[len++] = (char)(48U + (frac % 10U));
+      if (!(frac /= 10U)) {
+        break;
+      }
+    }
+    // add extra 0s
+    while ((len < PRINTF_FTOA_BUFFER_SIZE) && (count-- > 0U)) {
+      buf[len++] = '0';
+    }
+    if (len < PRINTF_FTOA_BUFFER_SIZE) {
+      // add decimal
+      buf[len++] = '.';
+    }
+  }
+
+  // do whole part, number is reversed
+  while (len < PRINTF_FTOA_BUFFER_SIZE) {
+    buf[len++] = (char)(48 + (whole % 10));
+    if (!(whole /= 10)) {
+      break;
+    }
+  }
+
+  // pad leading zeros
+  if (!(flags & FLAGS_LEFT) && (flags & FLAGS_ZEROPAD)) {
+    if (width && (negative || (flags & (FLAGS_PLUS | FLAGS_SPACE)))) {
+      width--;
+    }
+    while ((len < width) && (len < PRINTF_FTOA_BUFFER_SIZE)) {
+      buf[len++] = '0';
+    }
+  }
+
+  if (len < PRINTF_FTOA_BUFFER_SIZE) {
+    if (negative) {
+      buf[len++] = '-';
+    }
+    else if (flags & FLAGS_PLUS) {
+      buf[len++] = '+';  // ignore the space if the '+' exists
+    }
+    else if (flags & FLAGS_SPACE) {
+      buf[len++] = ' ';
+    }
+  }
+
+  // pad spaces up to given width
+  if (!(flags & FLAGS_LEFT) && !(flags & FLAGS_ZEROPAD)) {
+    for (size_t i = len; i < width; i++) {
+      out(' ', buffer, idx++, maxlen);
+    }
+  }
+
+  // reverse string
+  for (size_t i = 0U; i < len; i++) {
+    out(buf[len - i - 1U], buffer, idx++, maxlen);
+  }
+
+  // append pad spaces up to given width
+  if (flags & FLAGS_LEFT) {
+    while (idx - start_idx < width) {
+      out(' ', buffer, idx++, maxlen);
+    }
+  }
+
+  return idx;
+}
+#endif  // PRINTF_SUPPORT_FLOAT
+
+
+// internal vsnprintf
+static int _vsnprintf(out_fct_type out, char* buffer, const size_t maxlen, const char* format, va_list va)
+{
+  unsigned int flags, width, precision, n;
+  size_t idx = 0U;
+
+  if (!buffer) {
+    // use null output function
+    out = _out_null;
+  }
+
+  while (*format)
+  {
+    // format specifier?  %[flags][width][.precision][length]
+    if (*format != '%') {
+      // no
+      out(*format, buffer, idx++, maxlen);
+      format++;
+      continue;
+    }
+    else {
+      // yes, evaluate it
+      format++;
+    }
+
+    // evaluate flags
+    flags = 0U;
+    do {
+      switch (*format) {
+        case '0': flags |= FLAGS_ZEROPAD; format++; n = 1U; break;
+        case '-': flags |= FLAGS_LEFT;    format++; n = 1U; break;
+        case '+': flags |= FLAGS_PLUS;    format++; n = 1U; break;
+        case ' ': flags |= FLAGS_SPACE;   format++; n = 1U; break;
+        case '#': flags |= FLAGS_HASH;    format++; n = 1U; break;
+        default :                                   n = 0U; break;
+      }
+    } while (n);
+
+    // evaluate width field
+    width = 0U;
+    if (_is_digit(*format)) {
+      width = _atoi(&format);
+    }
+    else if (*format == '*') {
+      const int w = va_arg(va, int);
+      if (w < 0) {
+        flags |= FLAGS_LEFT;    // reverse padding
+        width = (unsigned int)-w;
+      }
+      else {
+        width = (unsigned int)w;
+      }
+      format++;
+    }
+
+    // evaluate precision field
+    precision = 0U;
+    if (*format == '.') {
+      flags |= FLAGS_PRECISION;
+      format++;
+      if (_is_digit(*format)) {
+        precision = _atoi(&format);
+      }
+      else if (*format == '*') {
+        const int prec = (int)va_arg(va, int);
+        precision = prec > 0 ? (unsigned int)prec : 0U;
+        format++;
+      }
+    }
+
+    // evaluate length field
+    switch (*format) {
+      case 'l' :
+        flags |= FLAGS_LONG;
+        format++;
+        if (*format == 'l') {
+          flags |= FLAGS_LONG_LONG;
+          format++;
+        }
+        break;
+      case 'h' :
+        flags |= FLAGS_SHORT;
+        format++;
+        if (*format == 'h') {
+          flags |= FLAGS_CHAR;
+          format++;
+        }
+        break;
+#if defined(PRINTF_SUPPORT_PTRDIFF_T)
+      case 't' :
+        flags |= (sizeof(ptrdiff_t) == sizeof(long) ? FLAGS_LONG : FLAGS_LONG_LONG);
+        format++;
+        break;
+#endif
+      case 'j' :
+        flags |= (sizeof(intmax_t) == sizeof(long) ? FLAGS_LONG : FLAGS_LONG_LONG);
+        format++;
+        break;
+      case 'z' :
+        flags |= (sizeof(size_t) == sizeof(long) ? FLAGS_LONG : FLAGS_LONG_LONG);
+        format++;
+        break;
+      default :
+        break;
+    }
+
+    // evaluate specifier
+    switch (*format) {
+      case 'd' :
+      case 'i' :
+      case 'u' :
+      case 'x' :
+      case 'X' :
+      case 'o' :
+      case 'b' : {
+        // set the base
+        unsigned int base;
+        if (*format == 'x' || *format == 'X') {
+          base = 16U;
+        }
+        else if (*format == 'o') {
+          base =  8U;
+        }
+        else if (*format == 'b') {
+          base =  2U;
+        }
+        else {
+          base = 10U;
+          flags &= ~FLAGS_HASH;   // no hash for dec format
+        }
+        // uppercase
+        if (*format == 'X') {
+          flags |= FLAGS_UPPERCASE;
+        }
+
+        // no plus or space flag for u, x, X, o, b
+        if ((*format != 'i') && (*format != 'd')) {
+          flags &= ~(FLAGS_PLUS | FLAGS_SPACE);
+        }
+
+        // ignore '0' flag when precision is given
+        if (flags & FLAGS_PRECISION) {
+          flags &= ~FLAGS_ZEROPAD;
+        }
+
+        // convert the integer
+        if ((*format == 'i') || (*format == 'd')) {
+          // signed
+          if (flags & FLAGS_LONG_LONG) {
+#if defined(PRINTF_SUPPORT_LONG_LONG)
+            const long long value = va_arg(va, long long);
+            idx = _ntoa_long_long(out, buffer, idx, maxlen, (unsigned long long)(value > 0 ? value : 0 - value), value < 0, base, precision, width, flags);
+#endif
+          }
+          else if (flags & FLAGS_LONG) {
+            const long value = va_arg(va, long);
+            idx = _ntoa_long(out, buffer, idx, maxlen, (unsigned long)(value > 0 ? value : 0 - value), value < 0, base, precision, width, flags);
+          }
+          else {
+            const int value = (flags & FLAGS_CHAR) ? (char)va_arg(va, int) : (flags & FLAGS_SHORT) ? (short int)va_arg(va, int) : va_arg(va, int);
+            idx = _ntoa_long(out, buffer, idx, maxlen, (unsigned int)(value > 0 ? value : 0 - value), value < 0, base, precision, width, flags);
+          }
+        }
+        else {
+          // unsigned
+          if (flags & FLAGS_LONG_LONG) {
+#if defined(PRINTF_SUPPORT_LONG_LONG)
+            idx = _ntoa_long_long(out, buffer, idx, maxlen, va_arg(va, unsigned long long), false, base, precision, width, flags);
+#endif
+          }
+          else if (flags & FLAGS_LONG) {
+            idx = _ntoa_long(out, buffer, idx, maxlen, va_arg(va, unsigned long), false, base, precision, width, flags);
+          }
+          else {
+            const unsigned int value = (flags & FLAGS_CHAR) ? (unsigned char)va_arg(va, unsigned int) : (flags & FLAGS_SHORT) ? (unsigned short int)va_arg(va, unsigned int) : va_arg(va, unsigned int);
+            idx = _ntoa_long(out, buffer, idx, maxlen, value, false, base, precision, width, flags);
+          }
+        }
+        format++;
+        break;
+      }
+#if defined(PRINTF_SUPPORT_FLOAT)
+      case 'f' :
+      case 'F' :
+        idx = _ftoa(out, buffer, idx, maxlen, va_arg(va, double), precision, width, flags);
+        format++;
+        break;
+#endif  // PRINTF_SUPPORT_FLOAT
+      case 'c' : {
+        unsigned int l = 1U;
+        // pre padding
+        if (!(flags & FLAGS_LEFT)) {
+          while (l++ < width) {
+            out(' ', buffer, idx++, maxlen);
+          }
+        }
+        // char output
+        out((char)va_arg(va, int), buffer, idx++, maxlen);
+        // post padding
+        if (flags & FLAGS_LEFT) {
+          while (l++ < width) {
+            out(' ', buffer, idx++, maxlen);
+          }
+        }
+        format++;
+        break;
+      }
+
+      case 's' : {
+        const char* p = va_arg(va, char*);
+        unsigned int l = _strnlen_s(p, precision ? precision : (size_t)-1);
+        // pre padding
+        if (flags & FLAGS_PRECISION) {
+          l = (l < precision ? l : precision);
+        }
+        if (!(flags & FLAGS_LEFT)) {
+          while (l++ < width) {
+            out(' ', buffer, idx++, maxlen);
+          }
+        }
+        // string output
+        while ((*p != 0) && (!(flags & FLAGS_PRECISION) || precision--)) {
+          out(*(p++), buffer, idx++, maxlen);
+        }
+        // post padding
+        if (flags & FLAGS_LEFT) {
+          while (l++ < width) {
+            out(' ', buffer, idx++, maxlen);
+          }
+        }
+        format++;
+        break;
+      }
+
+      case 'p' : {
+        width = sizeof(void*) * 2U;
+        flags |= FLAGS_ZEROPAD | FLAGS_UPPERCASE;
+#if defined(PRINTF_SUPPORT_LONG_LONG)
+        const bool is_ll = sizeof(uintptr_t) == sizeof(long long);
+        if (is_ll) {
+          idx = _ntoa_long_long(out, buffer, idx, maxlen, (uintptr_t)va_arg(va, void*), false, 16U, precision, width, flags);
+        }
+        else {
+#endif
+          idx = _ntoa_long(out, buffer, idx, maxlen, (unsigned long)((uintptr_t)va_arg(va, void*)), false, 16U, precision, width, flags);
+#if defined(PRINTF_SUPPORT_LONG_LONG)
+        }
+#endif
+        format++;
+        break;
+      }
+
+      case '%' :
+        out('%', buffer, idx++, maxlen);
+        format++;
+        break;
+
+      default :
+        out(*format, buffer, idx++, maxlen);
+        format++;
+        break;
+    }
+  }
+
+  // termination
+  out((char)0, buffer, idx < maxlen ? idx : maxlen - 1U, maxlen);
+
+  // return written chars without terminating \0
+  return (int)idx;
+}
+
+
+///////////////////////////////////////////////////////////////////////////////
+
+#ifdef PRINTF_PUTCHAR_FUNC
+int printf_(const char* format, ...)
+{
+  va_list va;
+  va_start(va, format);
+  char buffer[1];
+  const int ret = _vsnprintf(_out_char, buffer, (size_t)-1, format, va);
+  va_end(va);
+  return ret;
+}
+#endif
+
+int sprintf_(char* buffer, const char* format, ...)
+{
+  va_list va;
+  va_start(va, format);
+  const int ret = _vsnprintf(_out_buffer, buffer, (size_t)-1, format, va);
+  va_end(va);
+  return ret;
+}
+
+
+int snprintf_(char* buffer, size_t count, const char* format, ...)
+{
+  va_list va;
+  va_start(va, format);
+  const int ret = _vsnprintf(_out_buffer, buffer, count, format, va);
+  va_end(va);
+  return ret;
+}
+
+
+int vsnprintf_(char* buffer, size_t count, const char* format, va_list va)
+{
+  return _vsnprintf(_out_buffer, buffer, count, format, va);
+}
+
+
+int fctprintf(void (*out)(char character, void* arg), void* arg, const char* format, ...)
+{
+  va_list va;
+  va_start(va, format);
+  const out_fct_wrap_type out_fct_wrap = { out, arg };
+  const int ret = _vsnprintf(_out_fct, (char*)(uintptr_t)&out_fct_wrap, (size_t)-1, format, va);
+  va_end(va);
+  return ret;
+}

src/tests/address_manager.cpp

@@ -6,6 +6,7 @@
 #include <stdint.h>
 
 #include "kutil/address_manager.h"
+#include "kutil/allocator.h"
 #include "catch.hpp"
 
 using namespace kutil;
@@ -14,9 +15,18 @@ static const size_t max_block = 1ull << 36;
 static const size_t start = max_block;
 static const size_t GB = 1ull << 30;
 
+class malloc_allocator :
+	public kutil::allocator
+{
+public:
+	virtual void * allocate(size_t n) override { return malloc(n); }
+	virtual void free(void *p) override { free(p); }
+};
+
 TEST_CASE( "Buddy addresses tests", "[address buddy]" )
 {
-	address_manager am;
+	malloc_allocator alloc;
+	address_manager am(alloc);
 	am.add_regions(start, max_block * 2);
 
 	// Blocks should be:

src/tests/constexpr_hash.cpp

@@ -0,0 +1,21 @@
+#include "kutil/constexpr_hash.h"
+#include "catch.hpp"
+
+using namespace kutil;
+
+TEST_CASE( "constexpr hash", "[hash]" )
+{
+	const unsigned hash1 = static_cast<unsigned>("hash1!"_h);
+	CHECK(hash1 == 210);
+
+	const unsigned hash2 = static_cast<unsigned>("hash1!"_h);
+	CHECK(hash1 == hash2);
+
+	const unsigned hash3 = static_cast<unsigned>("not hash1!"_h);
+	CHECK(hash1 != hash3);
+	CHECK(hash3 == 37);
+
+	const unsigned hash4 = static_cast<unsigned>("another thing that's longer"_h);
+	CHECK(hash1 != hash4);
+	CHECK(hash4 == 212);
+}

src/tests/frame_allocator.cpp

@@ -1,45 +0,0 @@
-#include "kutil/frame_allocator.h"
-#include "catch.hpp"
-
-using namespace kutil;
-
-TEST_CASE( "Frame allocator tests", "[memory frame]" )
-{
-	frame_block_list free;
-	frame_block_list used;
-	frame_block_list cache;
-
-	auto *f = new frame_block_list::item_type;
-	f->address = 0x1000;
-	f->count = 1;
-	f->flags = kutil::frame_block_flags::none;
-	free.sorted_insert(f);
-
-	auto *g = new frame_block_list::item_type;
-	g->address = 0x2000;
-	g->count = 1;
-	g->flags = kutil::frame_block_flags::none;
-	free.sorted_insert(g);
-
-	frame_allocator fa(std::move(cache));
-	fa.init(std::move(free), std::move(used));
-
-	fa.consolidate_blocks();
-
-	uintptr_t a = 0;
-	size_t c = fa.allocate(2, &a);
-	CHECK( a == 0x1000 );
-	CHECK( c == 2 );
-
-	fa.free(a, 2);
-	a = 0;
-
-	fa.consolidate_blocks();
-
-	c = fa.allocate(2, &a);
-	CHECK( a == 0x1000 );
-	CHECK( c == 2 );
-
-	delete f;
-	delete g;
-}

src/tests/heap_allocator.cpp

@@ -0,0 +1,170 @@
+#include <chrono>
+#include <random>
+#include <vector>
+#include <signal.h>
+#include <stddef.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <sys/mman.h>
+
+#include "kutil/memory.h"
+#include "kutil/heap_allocator.h"
+#include "catch.hpp"
+
+using namespace kutil;
+
+const size_t hs = 0x10; // header size
+const size_t max_block = 1 << 22;
+
+int signalled = 0;
+void *signalled_at = nullptr;
+
+void *mem_base = nullptr;
+
+std::vector<size_t> sizes = {
+	16000, 8000, 4000, 4000, 1000, 1000, 1000, 1000, 1000, 1000, 1000, 150,
+	150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 48, 48, 48, 13 };
+
+void segfault_handler(int signum, siginfo_t *info, void *ctxp)
+{
+	signalled += 1;
+	signalled_at = info->si_addr;
+	mprotect(signalled_at, max_block, PROT_READ|PROT_WRITE);
+}
+
+TEST_CASE( "Buddy blocks tests", "[memory buddy]" )
+{
+	using clock = std::chrono::system_clock;
+	unsigned seed = clock::now().time_since_epoch().count();
+	std::default_random_engine rng(seed);
+
+	mem_base = aligned_alloc(max_block, max_block * 4);
+
+	// Catch segfaults so we can track memory access
+	struct sigaction sigact;
+	memset(&sigact, 0, sizeof(sigact));
+	sigemptyset(&sigact.sa_mask);
+	sigact.sa_flags = SA_NODEFER|SA_SIGINFO;
+	sigact.sa_sigaction = segfault_handler;
+	sigaction(SIGSEGV, &sigact, nullptr);
+
+	// Protect our memory arena so we trigger out fault handler
+	REQUIRE( mprotect(mem_base, max_block*4, PROT_NONE) == 0 );
+
+	heap_allocator mm(
+			reinterpret_cast<uintptr_t>(mem_base),
+			max_block * 4);
+
+	// Initial creation should not have allocated
+	CHECK( signalled == 0 );
+	signalled = 0;
+
+	// Allocating should signal just at the first page.
+	void *p = mm.allocate(max_block - hs);
+	CHECK( p == offset_pointer(mem_base, hs) );
+	CHECK( signalled == 1 );
+	CHECK( signalled_at == mem_base );
+	signalled = 0;
+
+	// Freeing and allocating should not allocate
+	mm.free(p);
+	p = mm.allocate(max_block - hs);
+	CHECK( p == offset_pointer(mem_base, hs) );
+	CHECK( signalled == 0 );
+	signalled = 0;
+
+	mm.free(p);
+	CHECK( signalled == 0 );
+	signalled = 0;
+
+	// Blocks should be:
+	// 22: 0-4M
+
+	std::vector<void *> allocs(6);
+	for (int i = 0; i < 6; ++i)
+		allocs[i] = mm.allocate(150); // size 8
+
+	// Should not have grown
+	CHECK( signalled == 0 );
+	signalled = 0;
+
+	// Blocks should be:
+	// 22: [0-4M]
+	// 21: [0-2M], 2-4M
+	// 20: [0-1M], 1-2M
+	// 19: [0-512K], 512K-1M
+	// 18: [0-256K], 256-512K
+	// 17: [0-128K], 128-256K
+	// 16: [0-64K], 64-128K
+	// 15: [0-32K], 32K-64K
+	// 14: [0-16K], 16K-32K
+	// 13: [0-8K], 8K-16K
+	// 12: [0-4K], 4K-8K
+	// 11: [0-2K], 2K-4K
+	// 10: [0-1K, 1-2K]
+	//  9: [0, 512, 1024], 1536
+	//  8: [0, 256, 512, 768, 1024, 1280]
+
+	// We have free memory at 1526 and 2K, but we should get 4K
+	void *big = mm.allocate(4000); // size 12
+
+	CHECK( signalled == 0 );
+	signalled = 0;
+
+	REQUIRE( big == offset_pointer(mem_base, 4096 + hs) );
+	mm.free(big);
+
+	// free up 512
+	mm.free(allocs[3]);
+	mm.free(allocs[4]);
+
+	// Blocks should be:
+	// ...
+	//  9: [0, 512, 1024], 1536
+	//  8: [0, 256, 512], 768, 1024, [1280]
+
+	// A request for a 512-block should not cross the buddy divide
+	big = mm.allocate(500); // size 9
+	REQUIRE( big >= offset_pointer(mem_base, 1536 + hs) );
+	mm.free(big);
+
+	mm.free(allocs[0]);
+	mm.free(allocs[1]);
+	mm.free(allocs[2]);
+	mm.free(allocs[5]);
+	allocs.clear();
+
+	std::shuffle(sizes.begin(), sizes.end(), rng);
+
+	allocs.reserve(sizes.size());
+	for (size_t size : sizes)
+		allocs.push_back(mm.allocate(size));
+
+	std::shuffle(allocs.begin(), allocs.end(), rng);
+	for (void *p: allocs)
+		mm.free(p);
+	allocs.clear();
+
+	big = mm.allocate(max_block / 2 + 1);
+
+	// If everything was freed / joined correctly, that should not have allocated
+	CHECK( signalled == 0 );
+	signalled = 0;
+
+	// And we should have gotten back the start of memory
+	CHECK( big == offset_pointer(mem_base, hs) );
+
+	// Allocating again should signal at the next page.
+	void *p2 = mm.allocate(max_block - hs);
+	CHECK( p2 == offset_pointer(mem_base, max_block + hs) );
+	CHECK( signalled == 1 );
+	CHECK( signalled_at == offset_pointer(mem_base, max_block) );
+	signalled = 0;
+
+	mm.free(p2);
+	CHECK( signalled == 0 );
+	signalled = 0;
+
+	free(mem_base);
+}
+

src/tests/heap_manager.cpp

@@ -1,191 +0,0 @@
-#include <chrono>
-#include <random>
-#include <vector>
-#include <stddef.h>
-#include <stdlib.h>
-#include <stdint.h>
-
-#include "kutil/memory.h"
-#include "kutil/heap_manager.h"
-#include "catch.hpp"
-
-using namespace kutil;
-
-static std::vector<void *> memory;
-
-static size_t total_alloc_size = 0;
-static size_t total_alloc_calls = 0;
-
-const size_t hs = 0x10; // header size
-const size_t max_block = 1 << 16;
-
-std::vector<size_t> sizes = {
-	16000, 8000, 4000, 4000, 1000, 1000, 1000, 1000, 1000, 1000, 1000, 150,
-	150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 48, 48, 48, 13 };
-
-void * grow_callback(size_t length)
-{
-	total_alloc_calls += 1;
-	total_alloc_size += length;
-
-	void *p = aligned_alloc(max_block, length * 2);
-	memory.push_back(p);
-	return p;
-}
-
-void free_memory()
-{
-	for (void *p : memory) ::free(p);
-	memory.clear();
-	total_alloc_size = 0;
-	total_alloc_calls = 0;
-}
-
-TEST_CASE( "Buddy blocks tests", "[memory buddy]" )
-{
-	using clock = std::chrono::system_clock;
-	unsigned seed = clock::now().time_since_epoch().count();
-	std::default_random_engine rng(seed);
-
-	heap_manager mm(grow_callback);
-
-	// The ctor should have allocated an initial block
-	CHECK( total_alloc_size == max_block );
-	CHECK( total_alloc_calls == 1 );
-
-	// Blocks should be:
-	// 16: 0-64K
-
-	std::vector<void *> allocs(6);
-	for (int i = 0; i < 6; ++i)
-		allocs[i] = mm.allocate(150); // size 8
-
-	// Should not have grown
-	CHECK( total_alloc_size == max_block );
-	CHECK( total_alloc_calls == 1 );
-	CHECK( memory[0] != nullptr );
-
-	// Blocks should be:
-	// 16: [0-64K]
-	// 15: [0-32K], 32K-64K
-	// 14: [0-16K], 16K-32K
-	// 13: [0-8K], 8K-16K
-	// 12: [0-4K], 4K-8K
-	// 11: [0-2K], 2K-4K
-	// 10: [0-1K, 1-2K]
-	//  9: [0, 512, 1024], 1536
-	//  8: [0, 256, 512, 768, 1024, 1280]
-
-	// We have free memory at 1526 and 2K, but we should get 4K
-	void *big = mm.allocate(4000); // size 12
-	REQUIRE( big == offset_pointer(memory[0], 4096 + hs) );
-	mm.free(big);
-
-	// free up 512
-	mm.free(allocs[3]);
-	mm.free(allocs[4]);
-
-	// Blocks should be:
-	// ...
-	//  9: [0, 512, 1024], 1536
-	//  8: [0, 256, 512], 768, 1024, [1280]
-
-	// A request for a 512-block should not cross the buddy divide
-	big = mm.allocate(500); // size 9
-	REQUIRE( big >= offset_pointer(memory[0], 1536 + hs) );
-	mm.free(big);
-
-	mm.free(allocs[0]);
-	mm.free(allocs[1]);
-	mm.free(allocs[2]);
-	mm.free(allocs[5]);
-	allocs.clear();
-
-	std::shuffle(sizes.begin(), sizes.end(), rng);
-
-	allocs.reserve(sizes.size());
-	for (size_t size : sizes)
-		allocs.push_back(mm.allocate(size));
-
-	std::shuffle(allocs.begin(), allocs.end(), rng);
-	for (void *p: allocs)
-		mm.free(p);
-	allocs.clear();
-
-	big = mm.allocate(64000);
-
-	// If everything was freed / joined correctly, that should not have allocated
-	CHECK( total_alloc_size == max_block );
-	CHECK( total_alloc_calls == 1 );
-
-	// And we should have gotten back the start of memory
-	CHECK( big == offset_pointer(memory[0], hs) );
-
-	free_memory();
-}
-
-bool check_in_memory(void *p)
-{
-	for (void *mem : memory)
-		if (p >= mem && p <= offset_pointer(mem, max_block))
-			return true;
-	return false;
-}
-
-TEST_CASE( "Non-contiguous blocks tests", "[memory buddy]" )
-{
-	using clock = std::chrono::system_clock;
-	unsigned seed = clock::now().time_since_epoch().count();
-	std::default_random_engine rng(seed);
-
-	heap_manager mm(grow_callback);
-	std::vector<void *> allocs;
-
-	const int blocks = 3;
-	for (int i = 0; i < blocks; ++i) {
-		void *p = mm.allocate(64000);
-		REQUIRE( memory[i] != nullptr );
-		REQUIRE( p == offset_pointer(memory[i], hs) );
-		allocs.push_back(p);
-	}
-
-	CHECK( total_alloc_size == max_block * blocks );
-	CHECK( total_alloc_calls == blocks );
-
-	for (void *p : allocs)
-		mm.free(p);
-	allocs.clear();
-
-	allocs.reserve(sizes.size() * blocks);
-
-	for (int i = 0; i < blocks; ++i) {
-		std::shuffle(sizes.begin(), sizes.end(), rng);
-
-		for (size_t size : sizes)
-			allocs.push_back(mm.allocate(size));
-	}
-
-	for (void *p : allocs)
-		CHECK( check_in_memory(p) );
-
-	std::shuffle(allocs.begin(), allocs.end(), rng);
-	for (void *p: allocs)
-		mm.free(p);
-	allocs.clear();
-
-	CHECK( total_alloc_size == max_block * blocks );
-	CHECK( total_alloc_calls == blocks );
-
-	for (int i = 0; i < blocks; ++i)
-		allocs.push_back(mm.allocate(64000));
-
-	// If everything was freed / joined correctly, that should not have allocated
-	CHECK( total_alloc_size == max_block * blocks );
-	CHECK( total_alloc_calls == blocks );
-
-	for (void *p : allocs)
-		CHECK( check_in_memory(p) );
-
-	free_memory();
-}
-