mirror of
https://git.eden-emu.dev/eden-emu/eden.git
synced 2026-10-02 12:59:56 +00:00
Compare commits
6 Commits
| Author | SHA1 | Date | |
|---|---|---|---|
| a20c3bd9d9 | |||
| b6a45591d5 | |||
| 29002a17a4 | |||
| 8d83c4a08c | |||
| b55959cf7c | |||
| 269edfd0d4 |
Vendored
+544
-329
File diff suppressed because it is too large
Load Diff
Vendored
+32
-12
@@ -1,23 +1,43 @@
|
||||
// SPDX-License-Identifier: MPL-2.0
|
||||
// Copyright 2022 Skyline Team and Contributors (https://github.com/skyline-emu/)
|
||||
// Copyright © 2022 Skyline Team and Contributors (https://github.com/skyline-emu/)
|
||||
|
||||
#pragma once
|
||||
|
||||
#include <cstdint>
|
||||
|
||||
namespace bcn {
|
||||
/// @brief Decodes a BC1 encoded image to R8G8B8A8
|
||||
void DecodeBc1(const uint8_t *src, uint8_t *dst, size_t x, size_t y, size_t width, size_t height, bool isSigned);
|
||||
/// @brief Decodes a BC2 encoded image to R8G8B8A8
|
||||
void DecodeBc2(const uint8_t *src, uint8_t *dst, size_t x, size_t y, size_t width, size_t height, bool isSigned);
|
||||
//// @brief Decodes a BC3 encoded image to R8G8B8A8
|
||||
void DecodeBc3(const uint8_t *src, uint8_t *dst, size_t x, size_t y, size_t width, size_t height, bool isSigned);
|
||||
/// @brief Decodes a BC4 encoded image to R8
|
||||
/**
|
||||
* @brief Decodes a BC1 encoded image to R8G8B8A8
|
||||
*/
|
||||
void DecodeBc1(const uint8_t *src, uint8_t *dst, size_t x, size_t y, size_t width, size_t height);
|
||||
|
||||
/**
|
||||
* @brief Decodes a BC2 encoded image to R8G8B8A8
|
||||
*/
|
||||
void DecodeBc2(const uint8_t *src, uint8_t *dst, size_t x, size_t y, size_t width, size_t height);
|
||||
|
||||
/**
|
||||
* @brief Decodes a BC3 encoded image to R8G8B8A8
|
||||
*/
|
||||
void DecodeBc3(const uint8_t *src, uint8_t *dst, size_t x, size_t y, size_t width, size_t height);
|
||||
|
||||
/**
|
||||
* @brief Decodes a BC4 encoded image to R8
|
||||
*/
|
||||
void DecodeBc4(const uint8_t *src, uint8_t *dst, size_t x, size_t y, size_t width, size_t height, bool isSigned);
|
||||
//// @brief Decodes a BC5 encoded image to R8G8
|
||||
|
||||
/**
|
||||
* @brief Decodes a BC5 encoded image to R8G8
|
||||
*/
|
||||
void DecodeBc5(const uint8_t *src, uint8_t *dst, size_t x, size_t y, size_t width, size_t height, bool isSigned);
|
||||
//// @brief Decodes a BC6 encoded image to R16G16B16A16
|
||||
|
||||
/**
|
||||
* @brief Decodes a BC6 encoded image to R16G16B16A16
|
||||
*/
|
||||
void DecodeBc6(const uint8_t *src, uint8_t *dst, size_t x, size_t y, size_t width, size_t height, bool isSigned);
|
||||
/// @brief Decodes a BC7 encoded image to R8G8B8A8
|
||||
void DecodeBc7(const uint8_t *src, uint8_t *dst, size_t x, size_t y, size_t width, size_t height, bool isSigned);
|
||||
|
||||
/**
|
||||
* @brief Decodes a BC7 encoded image to R8G8B8A8
|
||||
*/
|
||||
void DecodeBc7(const uint8_t *src, uint8_t *dst, size_t x, size_t y, size_t width, size_t height);
|
||||
}
|
||||
|
||||
@@ -25,42 +25,33 @@ NvMap::Handle::Handle(u64 size_, Id id_)
|
||||
flags.raw = 0;
|
||||
}
|
||||
|
||||
NvResult NvMap::Handle::Alloc(Flags pFlags, u32 pAlign, u8 pKind, u64 pAddress,
|
||||
NvCore::SessionId pSessionId) {
|
||||
std::scoped_lock lock(mutex);
|
||||
NvResult NvMap::Handle::Alloc(Flags pFlags, u32 pAlign, u8 pKind, u64 pAddress, NvCore::SessionId pSessionId) {
|
||||
// Handles cannot be allocated twice
|
||||
if (allocated) {
|
||||
return NvResult::AccessDenied;
|
||||
}
|
||||
|
||||
flags = pFlags;
|
||||
kind = pKind;
|
||||
align = pAlign < YUZU_PAGESIZE ? YUZU_PAGESIZE : pAlign;
|
||||
session_id = pSessionId;
|
||||
|
||||
// This flag is only applicable for handles with an address passed
|
||||
if (pAddress) {
|
||||
flags.keep_uncached_after_free.Assign(0);
|
||||
} else {
|
||||
LOG_CRITICAL(Service_NVDRV,
|
||||
"Mapping nvmap handles without a CPU side address is unimplemented!");
|
||||
LOG_CRITICAL(Service_NVDRV, "Mapping nvmap handles without a CPU side address is unimplemented!");
|
||||
}
|
||||
|
||||
size = Common::AlignUp(size, YUZU_PAGESIZE);
|
||||
aligned_size = Common::AlignUp(size, align);
|
||||
address = pAddress;
|
||||
allocated = true;
|
||||
|
||||
return NvResult::Success;
|
||||
}
|
||||
|
||||
NvResult NvMap::Handle::Duplicate(bool internal_session) {
|
||||
std::scoped_lock lock(mutex);
|
||||
// Unallocated handles cannot be duplicated as duplication requires memory accounting (in HOS)
|
||||
if (!allocated) [[unlikely]] {
|
||||
return NvResult::BadValue;
|
||||
}
|
||||
|
||||
// If we internally use FromId the duplication tracking of handles won't work accurately due to
|
||||
// us not implementing per-process handle refs.
|
||||
if (internal_session) {
|
||||
@@ -68,16 +59,14 @@ NvResult NvMap::Handle::Duplicate(bool internal_session) {
|
||||
} else {
|
||||
dupes++;
|
||||
}
|
||||
|
||||
return NvResult::Success;
|
||||
}
|
||||
|
||||
NvMap::NvMap(Container& core_, Tegra::Host1x::Host1x& host1x_) : host1x{host1x_}, core{core_} {}
|
||||
|
||||
void NvMap::AddHandle(std::shared_ptr<Handle> handle_description) {
|
||||
std::scoped_lock lock(handles_lock);
|
||||
|
||||
handles.emplace(handle_description->id, std::move(handle_description));
|
||||
void NvMap::AddHandle(Handle&& handle_description) {
|
||||
std::scoped_lock l(handles_lock);
|
||||
handles.insert_or_assign(handle_description.id, std::move(handle_description));
|
||||
}
|
||||
|
||||
void NvMap::UnmapHandle(Handle& handle_description) {
|
||||
@@ -116,57 +105,48 @@ void NvMap::UnmapHandle(Handle& handle_description) {
|
||||
bool NvMap::TryRemoveHandle(const Handle& handle_description) {
|
||||
// No dupes left, we can remove from handle map
|
||||
if (handle_description.dupes == 0 && handle_description.internal_dupes == 0) {
|
||||
std::scoped_lock lock(handles_lock);
|
||||
|
||||
auto it{handles.find(handle_description.id)};
|
||||
std::scoped_lock l(handles_lock);
|
||||
auto it = handles.find(handle_description.id);
|
||||
if (it != handles.end()) {
|
||||
handles.erase(it);
|
||||
}
|
||||
|
||||
return true;
|
||||
} else {
|
||||
return false;
|
||||
}
|
||||
}
|
||||
|
||||
NvResult NvMap::CreateHandle(u64 size, std::shared_ptr<NvMap::Handle>& result_out) {
|
||||
if (!size) [[unlikely]] {
|
||||
NvResult NvMap::CreateHandle(u64 size, Handle::Id& out_handle) {
|
||||
if (!Common::AlignUp(size, YUZU_PAGESIZE)) {
|
||||
return NvResult::BadValue;
|
||||
}
|
||||
|
||||
u32 id{next_handle_id.fetch_add(HandleIdIncrement, std::memory_order_relaxed)};
|
||||
auto handle_description{std::make_shared<Handle>(size, id)};
|
||||
AddHandle(handle_description);
|
||||
|
||||
result_out = handle_description;
|
||||
u32 id = next_handle_id.fetch_add(HandleIdIncrement, std::memory_order_relaxed);
|
||||
AddHandle(Handle(size, id));
|
||||
out_handle = id;
|
||||
return NvResult::Success;
|
||||
}
|
||||
|
||||
std::shared_ptr<NvMap::Handle> NvMap::GetHandle(Handle::Id handle) {
|
||||
std::scoped_lock lock(handles_lock);
|
||||
try {
|
||||
return handles.at(handle);
|
||||
} catch (std::out_of_range&) {
|
||||
return nullptr;
|
||||
}
|
||||
std::optional<std::reference_wrapper<NvMap::Handle>> NvMap::GetHandle(Handle::Id handle) {
|
||||
if (auto const it = handles.find(handle); it != handles.end())
|
||||
return {it->second};
|
||||
return std::nullopt;
|
||||
}
|
||||
|
||||
DAddr NvMap::GetHandleAddress(Handle::Id handle) {
|
||||
std::scoped_lock lock(handles_lock);
|
||||
try {
|
||||
return handles.at(handle)->d_address;
|
||||
} catch (std::out_of_range&) {
|
||||
return 0;
|
||||
}
|
||||
if (auto const it = handles.find(handle); it != handles.end())
|
||||
return it->second.d_address;
|
||||
return 0;
|
||||
}
|
||||
|
||||
DAddr NvMap::PinHandle(NvMap::Handle::Id handle, bool low_area_pin) {
|
||||
auto handle_description{GetHandle(handle)};
|
||||
if (!handle_description) [[unlikely]] {
|
||||
std::scoped_lock lock(handles_lock);
|
||||
auto o = GetHandle(handle);
|
||||
if (!o) [[unlikely]] {
|
||||
return 0;
|
||||
}
|
||||
|
||||
std::scoped_lock lock(handle_description->mutex);
|
||||
auto handle_description = &o->get();
|
||||
|
||||
const auto map_low_area = [&] {
|
||||
if (handle_description->pin_virt_address == 0) {
|
||||
u32 address = host1x.Allocator().Allocate(u32(handle_description->aligned_size));
|
||||
@@ -180,17 +160,15 @@ DAddr NvMap::PinHandle(NvMap::Handle::Id handle, bool low_area_pin) {
|
||||
{
|
||||
// Lock now to prevent our queue entry from being removed for allocation in-between the
|
||||
// following check and erase
|
||||
std::scoped_lock queueLock(unmap_queue_lock);
|
||||
std::scoped_lock ql(unmap_queue_lock);
|
||||
if (handle_description->unmap_queue_entry) {
|
||||
unmap_queue.erase(*handle_description->unmap_queue_entry);
|
||||
handle_description->unmap_queue_entry.reset();
|
||||
|
||||
if (low_area_pin) {
|
||||
map_low_area();
|
||||
handle_description->pins++;
|
||||
return static_cast<DAddr>(handle_description->pin_virt_address);
|
||||
return DAddr(handle_description->pin_virt_address);
|
||||
}
|
||||
|
||||
handle_description->pins++;
|
||||
return handle_description->d_address;
|
||||
}
|
||||
@@ -211,12 +189,11 @@ DAddr NvMap::PinHandle(NvMap::Handle::Id handle, bool low_area_pin) {
|
||||
while ((address = smmu.Allocate(aligned_up)) == 0) {
|
||||
// Free handles until the allocation succeeds
|
||||
std::scoped_lock queueLock(unmap_queue_lock);
|
||||
if (auto freeHandleDesc{unmap_queue.front()}) {
|
||||
if (auto free_handle = handles.find(unmap_queue.front()); free_handle != handles.end()) {
|
||||
// Handles in the unmap queue are guaranteed not to be pinned so don't bother
|
||||
// checking if they are before unmapping
|
||||
std::scoped_lock freeLock(freeHandleDesc->mutex);
|
||||
if (handle_description->d_address)
|
||||
UnmapHandle(*freeHandleDesc);
|
||||
UnmapHandle(free_handle->second);
|
||||
} else {
|
||||
LOG_CRITICAL(Service_NVDRV, "Ran out of SMMU address space!");
|
||||
}
|
||||
@@ -234,51 +211,44 @@ DAddr NvMap::PinHandle(NvMap::Handle::Id handle, bool low_area_pin) {
|
||||
|
||||
handle_description->pins++;
|
||||
if (low_area_pin) {
|
||||
return static_cast<DAddr>(handle_description->pin_virt_address);
|
||||
return DAddr(handle_description->pin_virt_address);
|
||||
}
|
||||
return handle_description->d_address;
|
||||
}
|
||||
|
||||
void NvMap::UnpinHandle(Handle::Id handle) {
|
||||
auto handle_description{GetHandle(handle)};
|
||||
if (!handle_description) {
|
||||
return;
|
||||
}
|
||||
|
||||
std::scoped_lock lock(handle_description->mutex);
|
||||
if (--handle_description->pins < 0) {
|
||||
LOG_WARNING(Service_NVDRV, "Pin count imbalance detected!");
|
||||
} else if (!handle_description->pins) {
|
||||
std::scoped_lock queueLock(unmap_queue_lock);
|
||||
|
||||
// Add to the unmap queue allowing this handle's memory to be freed if needed
|
||||
unmap_queue.push_back(handle_description);
|
||||
handle_description->unmap_queue_entry = std::prev(unmap_queue.end());
|
||||
std::scoped_lock lock(handles_lock);
|
||||
if (auto o = GetHandle(handle); o) {
|
||||
auto handle_description = &o->get();
|
||||
if (--handle_description->pins < 0) {
|
||||
LOG_WARNING(Service_NVDRV, "Pin count imbalance detected!");
|
||||
} else if (!handle_description->pins) {
|
||||
std::scoped_lock ql(unmap_queue_lock);
|
||||
// Add to the unmap queue allowing this handle's memory to be freed if needed
|
||||
unmap_queue.push_back(handle);
|
||||
handle_description->unmap_queue_entry = std::prev(unmap_queue.end());
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
void NvMap::DuplicateHandle(Handle::Id handle, bool internal_session) {
|
||||
auto handle_description{GetHandle(handle)};
|
||||
if (!handle_description) {
|
||||
std::scoped_lock lock(handles_lock);
|
||||
auto o = GetHandle(handle);
|
||||
if (!o) {
|
||||
LOG_CRITICAL(Service_NVDRV, "Unregistered handle!");
|
||||
return;
|
||||
}
|
||||
|
||||
auto result = handle_description->Duplicate(internal_session);
|
||||
auto result = o->get().Duplicate(internal_session);
|
||||
if (result != NvResult::Success) {
|
||||
LOG_CRITICAL(Service_NVDRV, "Could not duplicate handle!");
|
||||
}
|
||||
}
|
||||
|
||||
std::optional<NvMap::FreeInfo> NvMap::FreeHandle(Handle::Id handle, bool internal_session) {
|
||||
std::weak_ptr<Handle> hWeak{GetHandle(handle)};
|
||||
FreeInfo freeInfo;
|
||||
|
||||
// We use a weak ptr here so we can tell when the handle has been freed and report that back to
|
||||
// guest
|
||||
if (auto handle_description = hWeak.lock()) {
|
||||
std::scoped_lock lock(handle_description->mutex);
|
||||
|
||||
// We use a weak ptr here so we can tell when the handle has been freed and report that back to guest
|
||||
std::scoped_lock lock(handles_lock);
|
||||
if (auto o = GetHandle(handle); o) {
|
||||
auto handle_description = &o->get();
|
||||
if (internal_session) {
|
||||
if (--handle_description->internal_dupes < 0)
|
||||
LOG_WARNING(Service_NVDRV, "Internal duplicate count imbalance detected!");
|
||||
@@ -288,25 +258,25 @@ std::optional<NvMap::FreeInfo> NvMap::FreeHandle(Handle::Id handle, bool interna
|
||||
} else if (handle_description->dupes == 0) {
|
||||
// Force unmap the handle
|
||||
if (handle_description->d_address) {
|
||||
std::scoped_lock queueLock(unmap_queue_lock);
|
||||
std::scoped_lock ql(unmap_queue_lock);
|
||||
UnmapHandle(*handle_description);
|
||||
}
|
||||
|
||||
handle_description->pins = 0;
|
||||
}
|
||||
}
|
||||
|
||||
// Try to remove the shared ptr to the handle from the map, if nothing else is using the
|
||||
// handle then it will now be freed when `handle_description` goes out of scope
|
||||
if (TryRemoveHandle(*handle_description)) {
|
||||
LOG_DEBUG(Service_NVDRV, "Removed nvmap handle: {}", handle);
|
||||
} else {
|
||||
LOG_DEBUG(Service_NVDRV,
|
||||
"Tried to free nvmap handle: {} but didn't as it still has duplicates",
|
||||
handle);
|
||||
LOG_DEBUG(Service_NVDRV, "Tried to free nvmap handle: {} but didn't as it still has duplicates", handle);
|
||||
}
|
||||
|
||||
freeInfo = {
|
||||
// // If the handle hasn't been freed from memory, mark that
|
||||
// if (!hWeak.expired()) {
|
||||
// LOG_DEBUG(Service_NVDRV, "nvmap handle: {} wasn't freed as it is still in use", handle);
|
||||
// freeInfo.can_unlock = false;
|
||||
// }
|
||||
return FreeInfo{
|
||||
.address = handle_description->address,
|
||||
.size = handle_description->size,
|
||||
.was_uncached = handle_description->flags.map_uncached.Value() != 0,
|
||||
@@ -315,30 +285,15 @@ std::optional<NvMap::FreeInfo> NvMap::FreeHandle(Handle::Id handle, bool interna
|
||||
} else {
|
||||
return std::nullopt;
|
||||
}
|
||||
|
||||
// If the handle hasn't been freed from memory, mark that
|
||||
if (!hWeak.expired()) {
|
||||
LOG_DEBUG(Service_NVDRV, "nvmap handle: {} wasn't freed as it is still in use", handle);
|
||||
freeInfo.can_unlock = false;
|
||||
}
|
||||
|
||||
return freeInfo;
|
||||
}
|
||||
|
||||
void NvMap::UnmapAllHandles(NvCore::SessionId session_id) {
|
||||
auto handles_copy = [&] {
|
||||
std::scoped_lock lk{handles_lock};
|
||||
return handles;
|
||||
}();
|
||||
|
||||
for (auto& [id, handle] : handles_copy) {
|
||||
{
|
||||
std::scoped_lock lk{handle->mutex};
|
||||
if (handle->session_id.id != session_id.id || handle->dupes <= 0) {
|
||||
continue;
|
||||
}
|
||||
std::scoped_lock lk{handles_lock};
|
||||
for (auto it = handles.begin(); it != handles.end(); ++it) {
|
||||
if (it->second.session_id.id != session_id.id || it->second.dupes <= 0) {
|
||||
continue;
|
||||
}
|
||||
FreeHandle(id, false);
|
||||
FreeHandle(it->first, false);
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@@ -12,6 +12,11 @@
|
||||
#include <memory>
|
||||
#include <mutex>
|
||||
#include <optional>
|
||||
#if BOOST_VERSION >= 109000
|
||||
#include <boost/unordered/unordered_node_map.hpp>
|
||||
#else
|
||||
#include <unordered_map>
|
||||
#endif
|
||||
#include <ankerl/unordered_dense.h>
|
||||
#include <assert.h>
|
||||
|
||||
@@ -31,54 +36,36 @@ class Host1x;
|
||||
namespace Service::Nvidia::NvCore {
|
||||
|
||||
class Container;
|
||||
/**
|
||||
* @brief The nvmap core class holds the global state for nvmap and provides methods to manage
|
||||
* handles
|
||||
*/
|
||||
/// @brief The nvmap core class holds the global state for nvmap and provides methods to manage handles
|
||||
class NvMap {
|
||||
public:
|
||||
/**
|
||||
* @brief A handle to a contiguous block of memory in an application's address space
|
||||
*/
|
||||
/// @brief A handle to a contiguous block of memory in an application's address space
|
||||
struct Handle {
|
||||
std::mutex mutex;
|
||||
|
||||
using Id = u32;
|
||||
std::optional<typename std::list<Handle::Id>::iterator> unmap_queue_entry{};
|
||||
u64 align{}; //!< The alignment to use when pinning the handle onto the SMMU
|
||||
u64 size; //!< Page-aligned size of the memory the handle refers to
|
||||
u64 aligned_size; //!< `align`-aligned size of the memory the handle refers to
|
||||
u64 orig_size; //!< Original unaligned size of the memory this handle refers to
|
||||
|
||||
DAddr d_address{}; //!< The memory location in the device's AS that this handle corresponds to, this can also be in the nvdrv tmem
|
||||
VAddr address{}; //!< The memory location in the guest's AS that this handle corresponds to, this can also be in the nvdrv tmem
|
||||
s64 pins{};
|
||||
s32 dupes{1}; //!< How many guest references there are to this handle
|
||||
s32 internal_dupes{0}; //!< How many emulator-internal references there are to this handle
|
||||
|
||||
using Id = u32;
|
||||
Id id; //!< A globally unique identifier for this handle
|
||||
|
||||
s64 pins{};
|
||||
u32 pin_virt_address{};
|
||||
std::optional<typename std::list<std::shared_ptr<Handle>>::iterator> unmap_queue_entry{};
|
||||
|
||||
union Flags {
|
||||
u32 raw;
|
||||
BitField<0, 1, u32> map_uncached; //!< If the handle should be mapped as uncached
|
||||
BitField<2, 1, u32> keep_uncached_after_free; //!< Only applicable when the handle was
|
||||
//!< allocated with a fixed address
|
||||
BitField<4, 1, u32> _unk0_; //!< Passed to IOVMM for pins
|
||||
BitField<2, 1, u32> keep_uncached_after_free; //!< Only applicable when the handle was allocated with a fixed address
|
||||
BitField<4, 1, u32> _unk0_; //!< Passed to IOVMM for pins
|
||||
} flags{};
|
||||
static_assert(sizeof(Flags) == sizeof(u32));
|
||||
|
||||
VAddr address{}; //!< The memory location in the guest's AS that this handle corresponds to,
|
||||
//!< this can also be in the nvdrv tmem
|
||||
bool is_shared_mem_mapped{}; //!< If this nvmap has been mapped with the MapSharedMem IPC
|
||||
//!< call
|
||||
|
||||
u8 kind{}; //!< Used for memory compression
|
||||
bool allocated{}; //!< If the handle has been allocated with `Alloc`
|
||||
bool in_heap{};
|
||||
NvCore::SessionId session_id{};
|
||||
|
||||
DAddr d_address{}; //!< The memory location in the device's AS that this handle corresponds
|
||||
//!< to, this can also be in the nvdrv tmem
|
||||
u8 kind{}; //!< Used for memory compression
|
||||
bool allocated : 1 = false; //!< If the handle has been allocated with `Alloc`
|
||||
bool in_heap : 1 = false;
|
||||
bool is_shared_mem_mapped : 1 = false; //!< If this nvmap has been mapped with the MapSharedMem IPC < call
|
||||
|
||||
Handle(u64 size, Id id);
|
||||
|
||||
@@ -123,9 +110,9 @@ public:
|
||||
/**
|
||||
* @brief Creates an unallocated handle of the given size
|
||||
*/
|
||||
[[nodiscard]] NvResult CreateHandle(u64 size, std::shared_ptr<NvMap::Handle>& result_out);
|
||||
[[nodiscard]] NvResult CreateHandle(u64 size, Handle::Id& out_handle);
|
||||
|
||||
std::shared_ptr<Handle> GetHandle(Handle::Id handle);
|
||||
std::optional<std::reference_wrapper<Handle>> GetHandle(Handle::Id handle);
|
||||
|
||||
DAddr GetHandleAddress(Handle::Id handle);
|
||||
|
||||
@@ -157,20 +144,21 @@ public:
|
||||
|
||||
void UnmapAllHandles(NvCore::SessionId session_id);
|
||||
|
||||
private:
|
||||
std::list<std::shared_ptr<Handle>> unmap_queue{};
|
||||
std::list<Handle::Id> unmap_queue{};
|
||||
/// Main owning map of handles
|
||||
#if BOOST_VERSION >= 109000
|
||||
boost::unordered_node_map<Handle::Id, Handle> handles{};
|
||||
#else
|
||||
std::unordered_map<Handle::Id, Handle> handles{};
|
||||
#endif
|
||||
std::mutex unmap_queue_lock{}; //!< Protects access to `unmap_queue`
|
||||
|
||||
ankerl::unordered_dense::map<Handle::Id, std::shared_ptr<Handle>>
|
||||
handles{}; //!< Main owning map of handles
|
||||
std::mutex handles_lock; //!< Protects access to `handles`
|
||||
|
||||
static constexpr u32 HandleIdIncrement{
|
||||
4}; //!< Each new handle ID is an increment of 4 from the previous
|
||||
static constexpr u32 HandleIdIncrement{4}; //!< Each new handle ID is an increment of 4 from the previous
|
||||
std::atomic<u32> next_handle_id{HandleIdIncrement};
|
||||
Tegra::Host1x::Host1x& host1x;
|
||||
Container& core;
|
||||
|
||||
void AddHandle(std::shared_ptr<Handle> handle);
|
||||
void AddHandle(Handle&& handle);
|
||||
|
||||
/**
|
||||
* @brief Unmaps and frees the SMMU memory region a handle is mapped to
|
||||
@@ -184,7 +172,5 @@ private:
|
||||
* @return If the handle was removed from the map
|
||||
*/
|
||||
bool TryRemoveHandle(const Handle& handle_description);
|
||||
|
||||
Container& core;
|
||||
};
|
||||
} // namespace Service::Nvidia::NvCore
|
||||
|
||||
@@ -328,10 +328,11 @@ NvResult nvhost_as_gpu::MapBufferEx(IoctlMapBufferEx& params) {
|
||||
}
|
||||
}
|
||||
|
||||
auto handle{nvmap.GetHandle(params.handle)};
|
||||
if (!handle) {
|
||||
auto o = nvmap.GetHandle(params.handle);
|
||||
if (!o) {
|
||||
return NvResult::BadValue;
|
||||
}
|
||||
auto handle = &o->get();
|
||||
|
||||
DAddr device_address = DAddr(nvmap.PinHandle(params.handle, false) + params.buffer_offset);
|
||||
u64 size{params.mapping_size ? params.mapping_size : handle->orig_size};
|
||||
|
||||
@@ -103,17 +103,13 @@ NvResult nvhost_nvdec_common::Submit(IoctlSubmit& params, std::span<u8> data, De
|
||||
|
||||
for (std::size_t i = 0; i < syncpt_increments.size(); i++) {
|
||||
const SyncptIncr& syncpt_incr = syncpt_increments[i];
|
||||
fence_thresholds[i] =
|
||||
syncpoint_manager.IncrementSyncpointMaxExt(syncpt_incr.id, syncpt_incr.increments);
|
||||
fence_thresholds[i] = syncpoint_manager.IncrementSyncpointMaxExt(syncpt_incr.id, syncpt_incr.increments);
|
||||
}
|
||||
|
||||
for (const auto& cmd_buffer : command_buffers) {
|
||||
const auto object = nvmap.GetHandle(cmd_buffer.memory_id);
|
||||
ASSERT_OR_EXECUTE(object, return NvResult::InvalidState;);
|
||||
Core::Memory::CpuGuestMemory<Tegra::ChCommandHeader,
|
||||
Core::Memory::GuestMemoryFlags::SafeRead>
|
||||
cmdlist(session->process->GetMemory(), object->address + cmd_buffer.offset,
|
||||
cmd_buffer.word_count);
|
||||
Core::Memory::CpuGuestMemory<Tegra::ChCommandHeader, Core::Memory::GuestMemoryFlags::SafeRead> cmdlist(session->process->GetMemory(), object->get().address + cmd_buffer.offset, cmd_buffer.word_count);
|
||||
host1x.PushEntries(fd, std::move(cmdlist));
|
||||
}
|
||||
|
||||
|
||||
@@ -83,17 +83,14 @@ void nvmap::OnClose(DeviceFD fd) {
|
||||
NvResult nvmap::IocCreate(IocCreateParams& params) {
|
||||
LOG_DEBUG(Service_NVDRV, "called, size={:#08x}", params.size);
|
||||
|
||||
std::shared_ptr<NvCore::NvMap::Handle> handle_description{};
|
||||
auto result =
|
||||
file.CreateHandle(Common::AlignUp(params.size, YUZU_PAGESIZE), handle_description);
|
||||
NvCore::NvMap::Handle handle_description(0, 0);
|
||||
// Orig size is the unaligned size, set the handle to that
|
||||
auto result = file.CreateHandle(params.size, params.handle);
|
||||
if (result != NvResult::Success) {
|
||||
LOG_CRITICAL(Service_NVDRV, "Failed to create Object");
|
||||
return result;
|
||||
}
|
||||
handle_description->orig_size = params.size; // Orig size is the unaligned size
|
||||
params.handle = handle_description->id;
|
||||
LOG_DEBUG(Service_NVDRV, "handle: {}, size: {:#x}", handle_description->id, params.size);
|
||||
|
||||
LOG_DEBUG(Service_NVDRV, "handle: {}, size: {:#x}", params.handle, params.size);
|
||||
return NvResult::Success;
|
||||
}
|
||||
|
||||
@@ -115,30 +112,27 @@ NvResult nvmap::IocAlloc(IocAllocParams& params, DeviceFD fd) {
|
||||
params.align = YUZU_PAGESIZE;
|
||||
}
|
||||
|
||||
auto handle_description{file.GetHandle(params.handle)};
|
||||
if (!handle_description) {
|
||||
std::scoped_lock lock(file.handles_lock);
|
||||
auto o = file.GetHandle(params.handle);
|
||||
if (!o) {
|
||||
LOG_CRITICAL(Service_NVDRV, "Object does not exist, handle={:08X}", params.handle);
|
||||
return NvResult::BadValue;
|
||||
}
|
||||
|
||||
auto handle_description = &o->get();
|
||||
if (handle_description->allocated) {
|
||||
LOG_CRITICAL(Service_NVDRV, "Object is already allocated, handle={:08X}", params.handle);
|
||||
return NvResult::InsufficientMemory;
|
||||
}
|
||||
|
||||
const auto result = handle_description->Alloc(params.flags, params.align, params.kind,
|
||||
params.address, sessions[fd]);
|
||||
const auto result = handle_description->Alloc(params.flags, params.align, params.kind, params.address, sessions[fd]);
|
||||
if (result != NvResult::Success) {
|
||||
LOG_CRITICAL(Service_NVDRV, "Object failed to allocate, handle={:08X}", params.handle);
|
||||
return result;
|
||||
}
|
||||
bool is_out_io{};
|
||||
auto process = container.GetSession(sessions[fd])->process;
|
||||
ASSERT(process->GetPageTable()
|
||||
.LockForMapDeviceAddressSpace(&is_out_io, handle_description->address,
|
||||
handle_description->size,
|
||||
Kernel::KMemoryPermission::None, true, false)
|
||||
.IsSuccess());
|
||||
ASSERT(process->GetPageTable().LockForMapDeviceAddressSpace(&is_out_io, handle_description->address, handle_description->size, Kernel::KMemoryPermission::None, true, false).IsSuccess());
|
||||
return result;
|
||||
}
|
||||
|
||||
@@ -151,13 +145,13 @@ NvResult nvmap::IocGetId(IocGetIdParams& params) {
|
||||
return NvResult::BadValue;
|
||||
}
|
||||
|
||||
auto handle_description{file.GetHandle(params.handle)};
|
||||
if (!handle_description) {
|
||||
std::scoped_lock lock(file.handles_lock);
|
||||
auto o = file.GetHandle(params.handle);
|
||||
if (!o) {
|
||||
LOG_CRITICAL(Service_NVDRV, "Error!");
|
||||
return NvResult::AccessDenied; // This will always return EPERM irrespective of if the
|
||||
// handle exists or not
|
||||
return NvResult::AccessDenied; // This will always return EPERM irrespective of if the handle exists or not
|
||||
}
|
||||
|
||||
auto handle_description = &o->get();
|
||||
params.id = handle_description->id;
|
||||
return NvResult::Success;
|
||||
}
|
||||
@@ -174,12 +168,14 @@ NvResult nvmap::IocFromId(IocFromIdParams& params) {
|
||||
return NvResult::BadValue;
|
||||
}
|
||||
|
||||
auto handle_description{file.GetHandle(params.id)};
|
||||
if (!handle_description) {
|
||||
std::scoped_lock lock(file.handles_lock);
|
||||
auto o = file.GetHandle(params.id);
|
||||
if (!o) {
|
||||
LOG_CRITICAL(Service_NVDRV, "Unregistered handle!");
|
||||
return NvResult::BadValue;
|
||||
}
|
||||
|
||||
auto handle_description = &o->get();
|
||||
auto result = handle_description->Duplicate(false);
|
||||
if (result != NvResult::Success) {
|
||||
LOG_CRITICAL(Service_NVDRV, "Could not duplicate handle!");
|
||||
@@ -199,12 +195,14 @@ NvResult nvmap::IocParam(IocParamParams& params) {
|
||||
return NvResult::BadValue;
|
||||
}
|
||||
|
||||
auto handle_description{file.GetHandle(params.handle)};
|
||||
if (!handle_description) {
|
||||
std::scoped_lock lock(file.handles_lock);
|
||||
auto o = file.GetHandle(params.handle);
|
||||
if (!o) {
|
||||
LOG_CRITICAL(Service_NVDRV, "Not registered handle!");
|
||||
return NvResult::BadValue;
|
||||
}
|
||||
|
||||
auto handle_description = &o->get();
|
||||
switch (params.param) {
|
||||
case HandleParameterType::Size:
|
||||
params.result = static_cast<u32_le>(handle_description->orig_size);
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
|
||||
// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
|
||||
// SPDX-License-Identifier: GPL-3.0-or-later
|
||||
|
||||
// SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
|
||||
@@ -9,7 +9,6 @@
|
||||
#include <span>
|
||||
#include <bc_decoder.h>
|
||||
|
||||
#include "common/assert.h"
|
||||
#include "common/common_types.h"
|
||||
#include "video_core/texture_cache/decode_bc.h"
|
||||
|
||||
@@ -23,8 +22,11 @@ using VideoCore::Surface::PixelFormat;
|
||||
constexpr bool IsSigned(PixelFormat pixel_format) {
|
||||
switch (pixel_format) {
|
||||
case PixelFormat::BC4_SNORM:
|
||||
case PixelFormat::BC4_UNORM:
|
||||
case PixelFormat::BC5_SNORM:
|
||||
case PixelFormat::BC5_UNORM:
|
||||
case PixelFormat::BC6H_SFLOAT:
|
||||
case PixelFormat::BC6H_UFLOAT:
|
||||
return true;
|
||||
default:
|
||||
return false;
|
||||
@@ -60,28 +62,9 @@ u32 ConvertedBytesPerBlock(VideoCore::Surface::PixelFormat pixel_format) {
|
||||
}
|
||||
}
|
||||
|
||||
void DecompressBCn(std::span<const u8> input, std::span<u8> output, BufferImageCopy& copy, VideoCore::Surface::PixelFormat pixel_format) {
|
||||
auto const f = [pixel_format]{
|
||||
switch (pixel_format) {
|
||||
case PixelFormat::BC1_RGBA_UNORM:
|
||||
case PixelFormat::BC1_RGBA_SRGB: return &bcn::DecodeBc1;
|
||||
case PixelFormat::BC2_UNORM:
|
||||
case PixelFormat::BC2_SRGB: return &bcn::DecodeBc2;
|
||||
case PixelFormat::BC3_UNORM:
|
||||
case PixelFormat::BC3_SRGB: return &bcn::DecodeBc3;
|
||||
case PixelFormat::BC4_SNORM:
|
||||
case PixelFormat::BC4_UNORM: return &bcn::DecodeBc4;
|
||||
case PixelFormat::BC5_SNORM:
|
||||
case PixelFormat::BC5_UNORM: return &bcn::DecodeBc5;
|
||||
case PixelFormat::BC6H_SFLOAT:
|
||||
case PixelFormat::BC6H_UFLOAT: return &bcn::DecodeBc6;
|
||||
case PixelFormat::BC7_SRGB:
|
||||
case PixelFormat::BC7_UNORM: return &bcn::DecodeBc7;
|
||||
default:
|
||||
UNREACHABLE_MSG("Unimplemented BCn decompression {}", pixel_format);
|
||||
return &bcn::DecodeBc1;
|
||||
}
|
||||
}();
|
||||
template <auto decompress, PixelFormat pixel_format>
|
||||
void DecompressBlocks(std::span<const u8> input, std::span<u8> output, BufferImageCopy& copy,
|
||||
bool is_signed = false) {
|
||||
const u32 out_bpp = ConvertedBytesPerBlock(pixel_format);
|
||||
const u32 block_size = BlockSize(pixel_format);
|
||||
const u32 width = copy.image_extent.width;
|
||||
@@ -99,7 +82,11 @@ void DecompressBCn(std::span<const u8> input, std::span<u8> output, BufferImageC
|
||||
for (u32 x = 0; x < width; x += block_width) {
|
||||
const u8* src = input.data() + src_offset;
|
||||
u8* const dst = output.data() + dst_offset;
|
||||
f(src, dst, x, y, width, height, IsSigned(pixel_format));
|
||||
if constexpr (IsSigned(pixel_format)) {
|
||||
decompress(src, dst, x, y, width, height, is_signed);
|
||||
} else {
|
||||
decompress(src, dst, x, y, width, height);
|
||||
}
|
||||
src_offset += block_size;
|
||||
dst_offset += block_width * out_bpp;
|
||||
}
|
||||
@@ -109,4 +96,43 @@ void DecompressBCn(std::span<const u8> input, std::span<u8> output, BufferImageC
|
||||
}
|
||||
}
|
||||
|
||||
void DecompressBCn(std::span<const u8> input, std::span<u8> output, BufferImageCopy& copy,
|
||||
VideoCore::Surface::PixelFormat pixel_format) {
|
||||
switch (pixel_format) {
|
||||
case PixelFormat::BC1_RGBA_UNORM:
|
||||
case PixelFormat::BC1_RGBA_SRGB:
|
||||
DecompressBlocks<bcn::DecodeBc1, PixelFormat::BC1_RGBA_UNORM>(input, output, copy);
|
||||
break;
|
||||
case PixelFormat::BC2_UNORM:
|
||||
case PixelFormat::BC2_SRGB:
|
||||
DecompressBlocks<bcn::DecodeBc2, PixelFormat::BC2_UNORM>(input, output, copy);
|
||||
break;
|
||||
case PixelFormat::BC3_UNORM:
|
||||
case PixelFormat::BC3_SRGB:
|
||||
DecompressBlocks<bcn::DecodeBc3, PixelFormat::BC3_UNORM>(input, output, copy);
|
||||
break;
|
||||
case PixelFormat::BC4_SNORM:
|
||||
case PixelFormat::BC4_UNORM:
|
||||
DecompressBlocks<bcn::DecodeBc4, PixelFormat::BC4_UNORM>(
|
||||
input, output, copy, pixel_format == PixelFormat::BC4_SNORM);
|
||||
break;
|
||||
case PixelFormat::BC5_SNORM:
|
||||
case PixelFormat::BC5_UNORM:
|
||||
DecompressBlocks<bcn::DecodeBc5, PixelFormat::BC5_UNORM>(
|
||||
input, output, copy, pixel_format == PixelFormat::BC5_SNORM);
|
||||
break;
|
||||
case PixelFormat::BC6H_SFLOAT:
|
||||
case PixelFormat::BC6H_UFLOAT:
|
||||
DecompressBlocks<bcn::DecodeBc6, PixelFormat::BC6H_UFLOAT>(
|
||||
input, output, copy, pixel_format == PixelFormat::BC6H_SFLOAT);
|
||||
break;
|
||||
case PixelFormat::BC7_SRGB:
|
||||
case PixelFormat::BC7_UNORM:
|
||||
DecompressBlocks<bcn::DecodeBc7, PixelFormat::BC7_UNORM>(input, output, copy);
|
||||
break;
|
||||
default:
|
||||
LOG_WARNING(HW_GPU, "Unimplemented BCn decompression {}", pixel_format);
|
||||
}
|
||||
}
|
||||
|
||||
} // namespace VideoCommon
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
|
||||
// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
|
||||
// SPDX-License-Identifier: GPL-3.0-or-later
|
||||
|
||||
// SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
|
||||
@@ -922,7 +922,8 @@ boost::container::small_vector<BufferImageCopy, 16> UnswizzleImage(Tegra::Memory
|
||||
return copies;
|
||||
}
|
||||
|
||||
void ConvertImage(std::span<const u8> input, const ImageInfo& info, std::span<u8> output, std::span<BufferImageCopy> copies) {
|
||||
void ConvertImage(std::span<const u8> input, const ImageInfo& info, std::span<u8> output,
|
||||
std::span<BufferImageCopy> copies) {
|
||||
u32 output_offset = 0;
|
||||
Common::ScratchBuffer<u8> decode_scratch;
|
||||
|
||||
@@ -954,10 +955,10 @@ void ConvertImage(std::span<const u8> input, const ImageInfo& info, std::span<u8
|
||||
} else if (astc) {
|
||||
// BC1 uses 0.5 bytes per texel
|
||||
// BC3 uses 1 byte per texel
|
||||
auto const compress = recompression_setting == Settings::AstcRecompression::Bc1
|
||||
? Tegra::Texture::BCN::CompressBC1
|
||||
: Tegra::Texture::BCN::CompressBC3;
|
||||
const auto bpp_div = compress == Tegra::Texture::BCN::CompressBC1 ? 2 : 1;
|
||||
const auto compress = recompression_setting == Settings::AstcRecompression::Bc1
|
||||
? Tegra::Texture::BCN::CompressBC1
|
||||
: Tegra::Texture::BCN::CompressBC3;
|
||||
const auto bpp_div = recompression_setting == Settings::AstcRecompression::Bc1 ? 2 : 1;
|
||||
|
||||
const u32 plane_dim = copy.image_extent.width * copy.image_extent.height;
|
||||
const u32 level_size = plane_dim * copy.image_extent.depth *
|
||||
@@ -974,12 +975,18 @@ void ConvertImage(std::span<const u8> input, const ImageInfo& info, std::span<u8
|
||||
copy.image_subresource.num_layers * copy.image_extent.depth,
|
||||
output.subspan(output_offset));
|
||||
|
||||
const u32 aligned_plane_dim = Common::AlignUp(copy.image_extent.width, 4) * Common::AlignUp(copy.image_extent.height, 4);
|
||||
copy.buffer_size = (aligned_plane_dim * copy.image_extent.depth * copy.image_subresource.num_layers) / bpp_div;
|
||||
output_offset += u32(copy.buffer_size);
|
||||
const u32 aligned_plane_dim = Common::AlignUp(copy.image_extent.width, 4) *
|
||||
Common::AlignUp(copy.image_extent.height, 4);
|
||||
|
||||
copy.buffer_size =
|
||||
(aligned_plane_dim * copy.image_extent.depth * copy.image_subresource.num_layers) /
|
||||
bpp_div;
|
||||
output_offset += static_cast<u32>(copy.buffer_size);
|
||||
} else {
|
||||
DecompressBCn(input_offset, output.subspan(output_offset), copy, info.format);
|
||||
output_offset += copy.image_extent.width * copy.image_extent.height * copy.image_subresource.num_layers * ConvertedBytesPerBlock(info.format);
|
||||
output_offset += copy.image_extent.width * copy.image_extent.height *
|
||||
copy.image_subresource.num_layers *
|
||||
ConvertedBytesPerBlock(info.format);
|
||||
}
|
||||
|
||||
copy.buffer_row_length = mip_size.width;
|
||||
|
||||
Reference in New Issue
Block a user