mirror of
https://git.suyu.dev/suyu/suyu.git
synced 2024-11-30 15:26:54 -05:00
5167d1577d
The kernel allows restricting the total size of the handle table through the process capability descriptors. Until now, this functionality wasn't hooked up. With this, the process handle tables become properly restricted. In the case of metadata-less executables, the handle table will assume the maximum size is requested, preserving the behavior that existed before these changes.
355 lines
11 KiB
C++
355 lines
11 KiB
C++
// Copyright 2018 yuzu emulator team
|
|
// Licensed under GPLv2 or any later version
|
|
// Refer to the license.txt file included.
|
|
|
|
#include "common/bit_util.h"
|
|
#include "core/hle/kernel/errors.h"
|
|
#include "core/hle/kernel/handle_table.h"
|
|
#include "core/hle/kernel/process_capability.h"
|
|
#include "core/hle/kernel/vm_manager.h"
|
|
|
|
namespace Kernel {
|
|
namespace {
|
|
|
|
// clang-format off
|
|
|
|
// Shift offsets for kernel capability types.
|
|
enum : u32 {
|
|
CapabilityOffset_PriorityAndCoreNum = 3,
|
|
CapabilityOffset_Syscall = 4,
|
|
CapabilityOffset_MapPhysical = 6,
|
|
CapabilityOffset_MapIO = 7,
|
|
CapabilityOffset_Interrupt = 11,
|
|
CapabilityOffset_ProgramType = 13,
|
|
CapabilityOffset_KernelVersion = 14,
|
|
CapabilityOffset_HandleTableSize = 15,
|
|
CapabilityOffset_Debug = 16,
|
|
};
|
|
|
|
// Combined mask of all parameters that may be initialized only once.
|
|
constexpr u32 InitializeOnceMask = (1U << CapabilityOffset_PriorityAndCoreNum) |
|
|
(1U << CapabilityOffset_ProgramType) |
|
|
(1U << CapabilityOffset_KernelVersion) |
|
|
(1U << CapabilityOffset_HandleTableSize) |
|
|
(1U << CapabilityOffset_Debug);
|
|
|
|
// Packed kernel version indicating 10.4.0
|
|
constexpr u32 PackedKernelVersion = 0x520000;
|
|
|
|
// Indicates possible types of capabilities that can be specified.
|
|
enum class CapabilityType : u32 {
|
|
Unset = 0U,
|
|
PriorityAndCoreNum = (1U << CapabilityOffset_PriorityAndCoreNum) - 1,
|
|
Syscall = (1U << CapabilityOffset_Syscall) - 1,
|
|
MapPhysical = (1U << CapabilityOffset_MapPhysical) - 1,
|
|
MapIO = (1U << CapabilityOffset_MapIO) - 1,
|
|
Interrupt = (1U << CapabilityOffset_Interrupt) - 1,
|
|
ProgramType = (1U << CapabilityOffset_ProgramType) - 1,
|
|
KernelVersion = (1U << CapabilityOffset_KernelVersion) - 1,
|
|
HandleTableSize = (1U << CapabilityOffset_HandleTableSize) - 1,
|
|
Debug = (1U << CapabilityOffset_Debug) - 1,
|
|
Ignorable = 0xFFFFFFFFU,
|
|
};
|
|
|
|
// clang-format on
|
|
|
|
constexpr CapabilityType GetCapabilityType(u32 value) {
|
|
return static_cast<CapabilityType>((~value & (value + 1)) - 1);
|
|
}
|
|
|
|
u32 GetFlagBitOffset(CapabilityType type) {
|
|
const auto value = static_cast<u32>(type);
|
|
return static_cast<u32>(Common::BitSize<u32>() - Common::CountLeadingZeroes32(value));
|
|
}
|
|
|
|
} // Anonymous namespace
|
|
|
|
ResultCode ProcessCapabilities::InitializeForKernelProcess(const u32* capabilities,
|
|
std::size_t num_capabilities,
|
|
VMManager& vm_manager) {
|
|
Clear();
|
|
|
|
// Allow all cores and priorities.
|
|
core_mask = 0xF;
|
|
priority_mask = 0xFFFFFFFFFFFFFFFF;
|
|
kernel_version = PackedKernelVersion;
|
|
|
|
return ParseCapabilities(capabilities, num_capabilities, vm_manager);
|
|
}
|
|
|
|
ResultCode ProcessCapabilities::InitializeForUserProcess(const u32* capabilities,
|
|
std::size_t num_capabilities,
|
|
VMManager& vm_manager) {
|
|
Clear();
|
|
|
|
return ParseCapabilities(capabilities, num_capabilities, vm_manager);
|
|
}
|
|
|
|
void ProcessCapabilities::InitializeForMetadatalessProcess() {
|
|
// Allow all cores and priorities
|
|
core_mask = 0xF;
|
|
priority_mask = 0xFFFFFFFFFFFFFFFF;
|
|
kernel_version = PackedKernelVersion;
|
|
|
|
// Allow all system calls and interrupts.
|
|
svc_capabilities.set();
|
|
interrupt_capabilities.set();
|
|
|
|
// Allow using the maximum possible amount of handles
|
|
handle_table_size = static_cast<s32>(HandleTable::MAX_COUNT);
|
|
|
|
// Allow all debugging capabilities.
|
|
is_debuggable = true;
|
|
can_force_debug = true;
|
|
}
|
|
|
|
ResultCode ProcessCapabilities::ParseCapabilities(const u32* capabilities,
|
|
std::size_t num_capabilities,
|
|
VMManager& vm_manager) {
|
|
u32 set_flags = 0;
|
|
u32 set_svc_bits = 0;
|
|
|
|
for (std::size_t i = 0; i < num_capabilities; ++i) {
|
|
const u32 descriptor = capabilities[i];
|
|
const auto type = GetCapabilityType(descriptor);
|
|
|
|
if (type == CapabilityType::MapPhysical) {
|
|
i++;
|
|
|
|
// The MapPhysical type uses two descriptor flags for its parameters.
|
|
// If there's only one, then there's a problem.
|
|
if (i >= num_capabilities) {
|
|
return ERR_INVALID_COMBINATION;
|
|
}
|
|
|
|
const auto size_flags = capabilities[i];
|
|
if (GetCapabilityType(size_flags) != CapabilityType::MapPhysical) {
|
|
return ERR_INVALID_COMBINATION;
|
|
}
|
|
|
|
const auto result = HandleMapPhysicalFlags(descriptor, size_flags, vm_manager);
|
|
if (result.IsError()) {
|
|
return result;
|
|
}
|
|
} else {
|
|
const auto result =
|
|
ParseSingleFlagCapability(set_flags, set_svc_bits, descriptor, vm_manager);
|
|
if (result.IsError()) {
|
|
return result;
|
|
}
|
|
}
|
|
}
|
|
|
|
return RESULT_SUCCESS;
|
|
}
|
|
|
|
ResultCode ProcessCapabilities::ParseSingleFlagCapability(u32& set_flags, u32& set_svc_bits,
|
|
u32 flag, VMManager& vm_manager) {
|
|
const auto type = GetCapabilityType(flag);
|
|
|
|
if (type == CapabilityType::Unset) {
|
|
return ERR_INVALID_CAPABILITY_DESCRIPTOR;
|
|
}
|
|
|
|
// Bail early on ignorable entries, as one would expect,
|
|
// ignorable descriptors can be ignored.
|
|
if (type == CapabilityType::Ignorable) {
|
|
return RESULT_SUCCESS;
|
|
}
|
|
|
|
// Ensure that the give flag hasn't already been initialized before.
|
|
// If it has been, then bail.
|
|
const u32 flag_length = GetFlagBitOffset(type);
|
|
const u32 set_flag = 1U << flag_length;
|
|
if ((set_flag & set_flags & InitializeOnceMask) != 0) {
|
|
return ERR_INVALID_COMBINATION;
|
|
}
|
|
set_flags |= set_flag;
|
|
|
|
switch (type) {
|
|
case CapabilityType::PriorityAndCoreNum:
|
|
return HandlePriorityCoreNumFlags(flag);
|
|
case CapabilityType::Syscall:
|
|
return HandleSyscallFlags(set_svc_bits, flag);
|
|
case CapabilityType::MapIO:
|
|
return HandleMapIOFlags(flag, vm_manager);
|
|
case CapabilityType::Interrupt:
|
|
return HandleInterruptFlags(flag);
|
|
case CapabilityType::ProgramType:
|
|
return HandleProgramTypeFlags(flag);
|
|
case CapabilityType::KernelVersion:
|
|
return HandleKernelVersionFlags(flag);
|
|
case CapabilityType::HandleTableSize:
|
|
return HandleHandleTableFlags(flag);
|
|
case CapabilityType::Debug:
|
|
return HandleDebugFlags(flag);
|
|
default:
|
|
break;
|
|
}
|
|
|
|
return ERR_INVALID_CAPABILITY_DESCRIPTOR;
|
|
}
|
|
|
|
void ProcessCapabilities::Clear() {
|
|
svc_capabilities.reset();
|
|
interrupt_capabilities.reset();
|
|
|
|
core_mask = 0;
|
|
priority_mask = 0;
|
|
|
|
handle_table_size = 0;
|
|
kernel_version = 0;
|
|
|
|
program_type = ProgramType::SysModule;
|
|
|
|
is_debuggable = false;
|
|
can_force_debug = false;
|
|
}
|
|
|
|
ResultCode ProcessCapabilities::HandlePriorityCoreNumFlags(u32 flags) {
|
|
if (priority_mask != 0 || core_mask != 0) {
|
|
return ERR_INVALID_CAPABILITY_DESCRIPTOR;
|
|
}
|
|
|
|
const u32 core_num_min = (flags >> 16) & 0xFF;
|
|
const u32 core_num_max = (flags >> 24) & 0xFF;
|
|
if (core_num_min > core_num_max) {
|
|
return ERR_INVALID_COMBINATION;
|
|
}
|
|
|
|
const u32 priority_min = (flags >> 10) & 0x3F;
|
|
const u32 priority_max = (flags >> 4) & 0x3F;
|
|
if (priority_min > priority_max) {
|
|
return ERR_INVALID_COMBINATION;
|
|
}
|
|
|
|
// The switch only has 4 usable cores.
|
|
if (core_num_max >= 4) {
|
|
return ERR_INVALID_PROCESSOR_ID;
|
|
}
|
|
|
|
const auto make_mask = [](u64 min, u64 max) {
|
|
const u64 range = max - min + 1;
|
|
const u64 mask = (1ULL << range) - 1;
|
|
|
|
return mask << min;
|
|
};
|
|
|
|
core_mask = make_mask(core_num_min, core_num_max);
|
|
priority_mask = make_mask(priority_min, priority_max);
|
|
return RESULT_SUCCESS;
|
|
}
|
|
|
|
ResultCode ProcessCapabilities::HandleSyscallFlags(u32& set_svc_bits, u32 flags) {
|
|
const u32 index = flags >> 29;
|
|
const u32 svc_bit = 1U << index;
|
|
|
|
// If we've already set this svc before, bail.
|
|
if ((set_svc_bits & svc_bit) != 0) {
|
|
return ERR_INVALID_COMBINATION;
|
|
}
|
|
set_svc_bits |= svc_bit;
|
|
|
|
const u32 svc_mask = (flags >> 5) & 0xFFFFFF;
|
|
for (u32 i = 0; i < 24; ++i) {
|
|
const u32 svc_number = index * 24 + i;
|
|
|
|
if ((svc_mask & (1U << i)) == 0) {
|
|
continue;
|
|
}
|
|
|
|
if (svc_number >= svc_capabilities.size()) {
|
|
return ERR_OUT_OF_RANGE;
|
|
}
|
|
|
|
svc_capabilities[svc_number] = true;
|
|
}
|
|
|
|
return RESULT_SUCCESS;
|
|
}
|
|
|
|
ResultCode ProcessCapabilities::HandleMapPhysicalFlags(u32 flags, u32 size_flags,
|
|
VMManager& vm_manager) {
|
|
// TODO(Lioncache): Implement once the memory manager can handle this.
|
|
return RESULT_SUCCESS;
|
|
}
|
|
|
|
ResultCode ProcessCapabilities::HandleMapIOFlags(u32 flags, VMManager& vm_manager) {
|
|
// TODO(Lioncache): Implement once the memory manager can handle this.
|
|
return RESULT_SUCCESS;
|
|
}
|
|
|
|
ResultCode ProcessCapabilities::HandleInterruptFlags(u32 flags) {
|
|
constexpr u32 interrupt_ignore_value = 0x3FF;
|
|
const u32 interrupt0 = (flags >> 12) & 0x3FF;
|
|
const u32 interrupt1 = (flags >> 22) & 0x3FF;
|
|
|
|
for (u32 interrupt : {interrupt0, interrupt1}) {
|
|
if (interrupt == interrupt_ignore_value) {
|
|
continue;
|
|
}
|
|
|
|
// NOTE:
|
|
// This should be checking a generic interrupt controller value
|
|
// as part of the calculation, however, given we don't currently
|
|
// emulate that, it's sufficient to mark every interrupt as defined.
|
|
|
|
if (interrupt >= interrupt_capabilities.size()) {
|
|
return ERR_OUT_OF_RANGE;
|
|
}
|
|
|
|
interrupt_capabilities[interrupt] = true;
|
|
}
|
|
|
|
return RESULT_SUCCESS;
|
|
}
|
|
|
|
ResultCode ProcessCapabilities::HandleProgramTypeFlags(u32 flags) {
|
|
const u32 reserved = flags >> 17;
|
|
if (reserved != 0) {
|
|
return ERR_RESERVED_VALUE;
|
|
}
|
|
|
|
program_type = static_cast<ProgramType>((flags >> 14) & 0b111);
|
|
return RESULT_SUCCESS;
|
|
}
|
|
|
|
ResultCode ProcessCapabilities::HandleKernelVersionFlags(u32 flags) {
|
|
// Yes, the internal member variable is checked in the actual kernel here.
|
|
// This might look odd for options that are only allowed to be initialized
|
|
// just once, however the kernel has a separate initialization function for
|
|
// kernel processes and userland processes. The kernel variant sets this
|
|
// member variable ahead of time.
|
|
|
|
const u32 major_version = kernel_version >> 19;
|
|
|
|
if (major_version != 0 || flags < 0x80000) {
|
|
return ERR_INVALID_CAPABILITY_DESCRIPTOR;
|
|
}
|
|
|
|
kernel_version = flags;
|
|
return RESULT_SUCCESS;
|
|
}
|
|
|
|
ResultCode ProcessCapabilities::HandleHandleTableFlags(u32 flags) {
|
|
const u32 reserved = flags >> 26;
|
|
if (reserved != 0) {
|
|
return ERR_RESERVED_VALUE;
|
|
}
|
|
|
|
handle_table_size = static_cast<s32>((flags >> 16) & 0x3FF);
|
|
return RESULT_SUCCESS;
|
|
}
|
|
|
|
ResultCode ProcessCapabilities::HandleDebugFlags(u32 flags) {
|
|
const u32 reserved = flags >> 19;
|
|
if (reserved != 0) {
|
|
return ERR_RESERVED_VALUE;
|
|
}
|
|
|
|
is_debuggable = (flags & 0x20000) != 0;
|
|
can_force_debug = (flags & 0x40000) != 0;
|
|
return RESULT_SUCCESS;
|
|
}
|
|
|
|
} // namespace Kernel
|