ladybird/Kernel/init.cpp
Andreas Kling 37304203dd Kernel: Lock thread list while in Thread::unref()
This patch does three things:

- Convert the global thread list from a HashMap to an IntrusiveList
- Combine the thread list and its lock into a SpinLockProtectedValue
- Customize Thread::unref() so it locks the list while unreffing

This closes the same race window for Thread as @sin-ack's recent changes
did for Process.

Note that the HashMap->IntrusiveList conversion means that we lose O(1)
lookups, but the majority of clients of this list are doing traversal,
not lookup. Once we have an intrusive hashing solution, we should port
this to use that, but for now, this gets rid of heap allocations during
a sensitive time.
2021-08-15 12:44:35 +02:00

381 lines
14 KiB
C++

/*
* Copyright (c) 2018-2020, Andreas Kling <kling@serenityos.org>
*
* SPDX-License-Identifier: BSD-2-Clause
*/
#include <AK/Types.h>
#include <Kernel/ACPI/DynamicParser.h>
#include <Kernel/ACPI/Initialize.h>
#include <Kernel/ACPI/MultiProcessorParser.h>
#include <Kernel/Arch/PC/BIOS.h>
#include <Kernel/Arch/x86/Processor.h>
#include <Kernel/BootInfo.h>
#include <Kernel/Bus/PCI/Access.h>
#include <Kernel/Bus/PCI/Initializer.h>
#include <Kernel/Bus/USB/USBManagement.h>
#include <Kernel/Bus/VirtIO/VirtIO.h>
#include <Kernel/CMOS.h>
#include <Kernel/CommandLine.h>
#include <Kernel/Devices/FullDevice.h>
#include <Kernel/Devices/HID/HIDManagement.h>
#include <Kernel/Devices/KCOVDevice.h>
#include <Kernel/Devices/MemoryDevice.h>
#include <Kernel/Devices/NullDevice.h>
#include <Kernel/Devices/PCISerialDevice.h>
#include <Kernel/Devices/RandomDevice.h>
#include <Kernel/Devices/SB16.h>
#include <Kernel/Devices/SerialDevice.h>
#include <Kernel/Devices/VMWareBackdoor.h>
#include <Kernel/Devices/ZeroDevice.h>
#include <Kernel/FileSystem/Ext2FileSystem.h>
#include <Kernel/FileSystem/SysFS.h>
#include <Kernel/FileSystem/VirtualFileSystem.h>
#include <Kernel/Graphics/GraphicsManagement.h>
#include <Kernel/Heap/SlabAllocator.h>
#include <Kernel/Heap/kmalloc.h>
#include <Kernel/Interrupts/APIC.h>
#include <Kernel/Interrupts/InterruptManagement.h>
#include <Kernel/Interrupts/PIC.h>
#include <Kernel/KSyms.h>
#include <Kernel/Memory/MemoryManager.h>
#include <Kernel/Multiboot.h>
#include <Kernel/Net/NetworkTask.h>
#include <Kernel/Net/NetworkingManagement.h>
#include <Kernel/Panic.h>
#include <Kernel/Prekernel/Prekernel.h>
#include <Kernel/Process.h>
#include <Kernel/ProcessExposed.h>
#include <Kernel/RTC.h>
#include <Kernel/Random.h>
#include <Kernel/Scheduler.h>
#include <Kernel/Sections.h>
#include <Kernel/Storage/StorageManagement.h>
#include <Kernel/TTY/ConsoleManagement.h>
#include <Kernel/TTY/PTYMultiplexer.h>
#include <Kernel/TTY/VirtualConsole.h>
#include <Kernel/Tasks/FinalizerTask.h>
#include <Kernel/Tasks/SyncTask.h>
#include <Kernel/Time/TimeManagement.h>
#include <Kernel/WorkQueue.h>
#include <Kernel/kstdio.h>
// Defined in the linker script
typedef void (*ctor_func_t)();
extern ctor_func_t start_heap_ctors[];
extern ctor_func_t end_heap_ctors[];
extern ctor_func_t start_ctors[];
extern ctor_func_t end_ctors[];
extern size_t __stack_chk_guard;
READONLY_AFTER_INIT size_t __stack_chk_guard;
extern "C" u8 start_of_safemem_text[];
extern "C" u8 end_of_safemem_text[];
extern "C" u8 start_of_safemem_atomic_text[];
extern "C" u8 end_of_safemem_atomic_text[];
extern "C" u8 end_of_kernel_image[];
multiboot_module_entry_t multiboot_copy_boot_modules_array[16];
size_t multiboot_copy_boot_modules_count;
READONLY_AFTER_INIT bool g_in_early_boot;
namespace Kernel {
[[noreturn]] static void init_stage2(void*);
static void setup_serial_debug();
// boot.S expects these functions to exactly have the following signatures.
// We declare them here to ensure their signatures don't accidentally change.
extern "C" void init_finished(u32 cpu) __attribute__((used));
extern "C" [[noreturn]] void init_ap(FlatPtr cpu, Processor* processor_info);
extern "C" [[noreturn]] void init(BootInfo const&);
READONLY_AFTER_INIT VirtualConsole* tty0;
static Processor s_bsp_processor; // global but let's keep it "private"
// SerenityOS Kernel C++ entry point :^)
//
// This is where C++ execution begins, after boot.S transfers control here.
//
// The purpose of init() is to start multi-tasking. It does the bare minimum
// amount of work needed to start the scheduler.
//
// Once multi-tasking is ready, we spawn a new thread that starts in the
// init_stage2() function. Initialization continues there.
extern "C" {
READONLY_AFTER_INIT PhysicalAddress start_of_prekernel_image;
READONLY_AFTER_INIT PhysicalAddress end_of_prekernel_image;
READONLY_AFTER_INIT size_t physical_to_virtual_offset;
READONLY_AFTER_INIT FlatPtr kernel_mapping_base;
READONLY_AFTER_INIT FlatPtr kernel_load_base;
#if ARCH(X86_64)
READONLY_AFTER_INIT PhysicalAddress boot_pml4t;
#endif
READONLY_AFTER_INIT PhysicalAddress boot_pdpt;
READONLY_AFTER_INIT PhysicalAddress boot_pd0;
READONLY_AFTER_INIT PhysicalAddress boot_pd_kernel;
READONLY_AFTER_INIT PageTableEntry* boot_pd_kernel_pt1023;
READONLY_AFTER_INIT const char* kernel_cmdline;
READONLY_AFTER_INIT u32 multiboot_flags;
READONLY_AFTER_INIT multiboot_memory_map_t* multiboot_memory_map;
READONLY_AFTER_INIT size_t multiboot_memory_map_count;
READONLY_AFTER_INIT multiboot_module_entry_t* multiboot_modules;
READONLY_AFTER_INIT size_t multiboot_modules_count;
READONLY_AFTER_INIT PhysicalAddress multiboot_framebuffer_addr;
READONLY_AFTER_INIT u32 multiboot_framebuffer_pitch;
READONLY_AFTER_INIT u32 multiboot_framebuffer_width;
READONLY_AFTER_INIT u32 multiboot_framebuffer_height;
READONLY_AFTER_INIT u8 multiboot_framebuffer_bpp;
READONLY_AFTER_INIT u8 multiboot_framebuffer_type;
}
extern "C" [[noreturn]] UNMAP_AFTER_INIT void init(BootInfo const& boot_info)
{
g_in_early_boot = true;
start_of_prekernel_image = PhysicalAddress { boot_info.start_of_prekernel_image };
end_of_prekernel_image = PhysicalAddress { boot_info.end_of_prekernel_image };
physical_to_virtual_offset = boot_info.physical_to_virtual_offset;
kernel_mapping_base = boot_info.kernel_mapping_base;
kernel_load_base = boot_info.kernel_load_base;
#if ARCH(X86_64)
gdt64ptr = boot_info.gdt64ptr;
code64_sel = boot_info.code64_sel;
boot_pml4t = PhysicalAddress { boot_info.boot_pml4t };
#endif
boot_pdpt = PhysicalAddress { boot_info.boot_pdpt };
boot_pd0 = PhysicalAddress { boot_info.boot_pd0 };
boot_pd_kernel = PhysicalAddress { boot_info.boot_pd_kernel };
boot_pd_kernel_pt1023 = (PageTableEntry*)boot_info.boot_pd_kernel_pt1023;
kernel_cmdline = (char const*)boot_info.kernel_cmdline;
multiboot_flags = boot_info.multiboot_flags;
multiboot_memory_map = (multiboot_memory_map_t*)boot_info.multiboot_memory_map;
multiboot_memory_map_count = boot_info.multiboot_memory_map_count;
multiboot_modules = (multiboot_module_entry_t*)boot_info.multiboot_modules;
multiboot_modules_count = boot_info.multiboot_modules_count;
multiboot_framebuffer_addr = PhysicalAddress { boot_info.multiboot_framebuffer_addr };
multiboot_framebuffer_pitch = boot_info.multiboot_framebuffer_pitch;
multiboot_framebuffer_width = boot_info.multiboot_framebuffer_width;
multiboot_framebuffer_height = boot_info.multiboot_framebuffer_height;
multiboot_framebuffer_bpp = boot_info.multiboot_framebuffer_bpp;
multiboot_framebuffer_type = boot_info.multiboot_framebuffer_type;
setup_serial_debug();
// We need to copy the command line before kmalloc is initialized,
// as it may overwrite parts of multiboot!
CommandLine::early_initialize(kernel_cmdline);
memcpy(multiboot_copy_boot_modules_array, multiboot_modules, multiboot_modules_count * sizeof(multiboot_module_entry_t));
multiboot_copy_boot_modules_count = multiboot_modules_count;
s_bsp_processor.early_initialize(0);
// Invoke the constructors needed for the kernel heap
for (ctor_func_t* ctor = start_heap_ctors; ctor < end_heap_ctors; ctor++)
(*ctor)();
kmalloc_init();
slab_alloc_init();
load_kernel_symbol_table();
ConsoleDevice::initialize();
s_bsp_processor.initialize(0);
CommandLine::initialize();
Memory::MemoryManager::initialize(0);
// Ensure that the safemem sections are not empty. This could happen if the linker accidentally discards the sections.
VERIFY(+start_of_safemem_text != +end_of_safemem_text);
VERIFY(+start_of_safemem_atomic_text != +end_of_safemem_atomic_text);
// Invoke all static global constructors in the kernel.
// Note that we want to do this as early as possible.
for (ctor_func_t* ctor = start_ctors; ctor < end_ctors; ctor++)
(*ctor)();
APIC::initialize();
InterruptManagement::initialize();
ACPI::initialize();
// Initialize TimeManagement before using randomness!
TimeManagement::initialize(0);
__stack_chk_guard = get_fast_random<size_t>();
ProcFSComponentRegistry::initialize();
Process::initialize();
Scheduler::initialize();
dmesgln("Starting SerenityOS...");
{
RefPtr<Thread> init_stage2_thread;
Process::create_kernel_process(init_stage2_thread, "init_stage2", init_stage2, nullptr, THREAD_AFFINITY_DEFAULT, Process::RegisterProcess::No);
// We need to make sure we drop the reference for init_stage2_thread
// before calling into Scheduler::start, otherwise we will have a
// dangling Thread that never gets cleaned up
}
Scheduler::start();
VERIFY_NOT_REACHED();
}
//
// This is where C++ execution begins for APs, after boot.S transfers control here.
//
// The purpose of init_ap() is to initialize APs for multi-tasking.
//
extern "C" [[noreturn]] UNMAP_AFTER_INIT void init_ap(FlatPtr cpu, Processor* processor_info)
{
processor_info->early_initialize(cpu);
processor_info->initialize(cpu);
Memory::MemoryManager::initialize(cpu);
Scheduler::set_idle_thread(APIC::the().get_idle_thread(cpu));
Scheduler::start();
VERIFY_NOT_REACHED();
}
//
// This method is called once a CPU enters the scheduler and its idle thread
// At this point the initial boot stack can be freed
//
extern "C" UNMAP_AFTER_INIT void init_finished(u32 cpu)
{
if (cpu == 0) {
// TODO: we can reuse the boot stack, maybe for kmalloc()?
} else {
APIC::the().init_finished(cpu);
TimeManagement::initialize(cpu);
}
}
void init_stage2(void*)
{
// This is a little bit of a hack. We can't register our process at the time we're
// creating it, but we need to be registered otherwise finalization won't be happy.
// The colonel process gets away without having to do this because it never exits.
Process::register_new(*Process::current());
WorkQueue::initialize();
if (APIC::initialized() && APIC::the().enabled_processor_count() > 1) {
// We can't start the APs until we have a scheduler up and running.
// We need to be able to process ICI messages, otherwise another
// core may send too many and end up deadlocking once the pool is
// exhausted
APIC::the().boot_aps();
}
// Initialize the PCI Bus as early as possible, for early boot (PCI based) serial logging
SysFSComponentRegistry::initialize();
PCI::initialize();
PCISerialDevice::detect();
VirtualFileSystem::initialize();
NullDevice::initialize();
if (!get_serial_debug())
(void)SerialDevice::must_create(0).leak_ref();
(void)SerialDevice::must_create(1).leak_ref();
(void)SerialDevice::must_create(2).leak_ref();
(void)SerialDevice::must_create(3).leak_ref();
VMWareBackdoor::the(); // don't wait until first mouse packet
HIDManagement::initialize();
GraphicsManagement::the().initialize();
ConsoleManagement::the().initialize();
SyncTask::spawn();
FinalizerTask::spawn();
auto boot_profiling = kernel_command_line().is_boot_profiling_enabled();
USB::USBManagement::initialize();
BIOSSysFSDirectory::initialize();
ACPI::ACPISysFSDirectory::initialize();
VirtIO::detect();
NetworkingManagement::the().initialize();
Syscall::initialize();
#ifdef ENABLE_KERNEL_COVERAGE_COLLECTION
(void)KCOVDevice::must_create().leak_ref();
#endif
(void)MemoryDevice::must_create().leak_ref();
(void)ZeroDevice::must_create().leak_ref();
(void)FullDevice::must_create().leak_ref();
(void)RandomDevice::must_create().leak_ref();
PTYMultiplexer::initialize();
SB16::detect();
StorageManagement::initialize(kernel_command_line().root_device(), kernel_command_line().is_force_pio());
if (!VirtualFileSystem::the().mount_root(StorageManagement::the().root_filesystem())) {
PANIC("VirtualFileSystem::mount_root failed");
}
// Switch out of early boot mode.
g_in_early_boot = false;
// NOTE: Everything marked READONLY_AFTER_INIT becomes non-writable after this point.
MM.protect_readonly_after_init_memory();
// NOTE: Everything marked UNMAP_AFTER_INIT becomes inaccessible after this point.
MM.unmap_text_after_init();
// NOTE: Everything in the .ksyms section becomes inaccessible after this point.
MM.unmap_ksyms_after_init();
int error;
// FIXME: It would be nicer to set the mode from userspace.
// FIXME: It would be smarter to not hardcode that the first tty is the only graphical one
ConsoleManagement::the().first_tty()->set_graphical(GraphicsManagement::the().framebuffer_devices_exist());
RefPtr<Thread> thread;
auto userspace_init = kernel_command_line().userspace_init();
auto init_args = kernel_command_line().userspace_init_args();
Process::create_user_process(thread, userspace_init, (uid_t)0, (gid_t)0, ProcessID(0), error, move(init_args), {}, tty0);
if (error != 0) {
PANIC("init_stage2: Error spawning SystemServer: {}", error);
}
thread->set_priority(THREAD_PRIORITY_HIGH);
if (boot_profiling) {
dbgln("Starting full system boot profiling");
MutexLocker mutex_locker(Process::current()->big_lock());
auto result = Process::current()->sys$profiling_enable(-1, ~0ull);
VERIFY(!result.is_error());
}
NetworkTask::spawn();
Process::current()->sys$exit(0);
VERIFY_NOT_REACHED();
}
UNMAP_AFTER_INIT void setup_serial_debug()
{
// serial_debug will output all the dbgln() data to COM1 at
// 8-N-1 57600 baud. this is particularly useful for debugging the boot
// process on live hardware.
if (StringView(kernel_cmdline).contains("serial_debug")) {
set_serial_debug(true);
}
}
// Define some Itanium C++ ABI methods to stop the linker from complaining.
// If we actually call these something has gone horribly wrong
void* __dso_handle __attribute__((visibility("hidden")));
}