Google Git
Sign in
fuchsia / fuchsia / cfef5042b7b9976bc7e0ed8c11f005ee41d1b7b9 / . / src / devices / sysmem / drivers / sysmem / logical_buffer_collection.cc
blob: 3cc109797fbf598bbeaacbb6f0ab1501ada8e016 [file] [log] [blame]
// Copyright 2019 The Fuchsia Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include "logical_buffer_collection.h"
#include <fuchsia/sysmem2/llcpp/fidl.h>
#include <inttypes.h>
#include <lib/image-format/image_format.h>
#include <lib/sysmem-make-tracking/make_tracking.h>
#include <lib/sysmem-version/sysmem-version.h>
#include <limits.h> // PAGE_SIZE
#include <zircon/assert.h>
#include <zircon/errors.h>
#include <limits> // std::numeric_limits
#include <ddk/trace/event.h>
#include <fbl/algorithm.h>
#include <fbl/string_printf.h>
#include "buffer_collection.h"
#include "buffer_collection_token.h"
#include "device.h"
#include "koid_util.h"
#include "macros.h"
#include "usage_pixel_format_cost.h"
#include "utils.h"
namespace sysmem_driver {
namespace {
// Sysmem is creating the VMOs, so sysmem can have all the rights and just not
// mis-use any rights. Remove ZX_RIGHT_EXECUTE though.
const uint32_t kSysmemVmoRights = ZX_DEFAULT_VMO_RIGHTS & ~ZX_RIGHT_EXECUTE;
// 1 GiB cap for now.
const uint64_t kMaxTotalSizeBytesPerCollection = 1ull * 1024 * 1024 * 1024;
// 256 MiB cap for now.
const uint64_t kMaxSizeBytesPerBuffer = 256ull * 1024 * 1024;
// Zero-initialized, so it shouldn't take up space on-disk.
constexpr uint64_t kFlushThroughBytes = 8192;
const uint8_t kZeroes[kFlushThroughBytes] = {};
template <typename T>
bool IsNonZeroPowerOf2(T value) {
if (!value) {
return false;
}
if (value & (value - 1)) {
return false;
}
return true;
}
// TODO(fxbug.dev/50590): It'd be nice if this could be a function template over FIDL scalar field
// types.
#define FIELD_DEFAULT_1(builder_ptr_name, field_name) \
do { \
auto builder_ptr = (builder_ptr_name); \
static_assert(fidl::IsTableBuilder<std::remove_pointer<decltype(builder_ptr)>::type>::value); \
using FieldType = std::remove_reference<decltype((builder_ptr->field_name()))>::type; \
if (!builder_ptr->has_##field_name()) { \
builder_ptr->set_##field_name(sysmem::MakeTracking(&allocator_, static_cast<FieldType>(1))); \
ZX_DEBUG_ASSERT(builder_ptr->field_name() == 1); \
} \
ZX_DEBUG_ASSERT(builder_ptr->has_##field_name()); \
} while (false)
// TODO(fxbug.dev/50590): It'd be nice if this could be a function template over FIDL scalar field
// types.
#define FIELD_DEFAULT_MAX(builder_ptr_name, field_name) \
do { \
auto builder_ptr = (builder_ptr_name); \
static_assert(fidl::IsTableBuilder<std::remove_pointer<decltype(builder_ptr)>::type>::value); \
using FieldType = std::remove_reference<decltype((builder_ptr->field_name()))>::type; \
if (!builder_ptr->has_##field_name()) { \
builder_ptr->set_##field_name( \
sysmem::MakeTracking(&allocator_, std::numeric_limits<FieldType>::max())); \
ZX_DEBUG_ASSERT(builder_ptr->field_name() == std::numeric_limits<FieldType>::max()); \
} \
ZX_DEBUG_ASSERT(builder_ptr->has_##field_name()); \
} while (false)
// TODO(fxbug.dev/50590): It'd be nice if this could be a function template over FIDL scalar field
// types.
#define FIELD_DEFAULT_ZERO(builder_ptr_name, field_name) \
do { \
auto builder_ptr = (builder_ptr_name); \
static_assert(fidl::IsTableBuilder<std::remove_pointer_t<decltype(builder_ptr)>>::value); \
using FieldType = std::remove_reference<decltype((builder_ptr->field_name()))>::type; \
if (!builder_ptr->has_##field_name()) { \
builder_ptr->set_##field_name(sysmem::MakeTracking(&allocator_, static_cast<FieldType>(0))); \
ZX_DEBUG_ASSERT(!static_cast<bool>(builder_ptr->field_name())); \
} \
ZX_DEBUG_ASSERT(builder_ptr->has_##field_name()); \
} while (false)
#define FIELD_DEFAULT_FALSE(builder_ptr_name, field_name) \
do { \
auto builder_ptr = (builder_ptr_name); \
static_assert(fidl::IsTableBuilder<std::remove_pointer_t<decltype(builder_ptr)>>::value); \
using FieldType = std::remove_reference<decltype((builder_ptr->field_name()))>::type; \
static_assert(std::is_same<FieldType, bool>::value); \
if (!builder_ptr->has_##field_name()) { \
builder_ptr->set_##field_name(sysmem::MakeTracking(&allocator_, false)); \
ZX_DEBUG_ASSERT(!builder_ptr->field_name()); \
} \
ZX_DEBUG_ASSERT(builder_ptr->has_##field_name()); \
} while (false)
#define FIELD_DEFAULT(builder_ptr_name, field_name, value_name) \
do { \
auto builder_ptr = (builder_ptr_name); \
static_assert(fidl::IsTableBuilder<std::remove_pointer_t<decltype(builder_ptr)>>::value); \
using FieldType = std::remove_reference<decltype((builder_ptr->field_name()))>::type; \
static_assert(!fidl::IsFidlObject<FieldType>::value); \
static_assert(!fidl::IsVectorView<FieldType>::value); \
static_assert(!fidl::IsStringView<FieldType>::value); \
if (!builder_ptr->has_##field_name()) { \
auto field_value = (value_name); \
builder_ptr->set_##field_name(sysmem::MakeTracking(&allocator_, field_value)); \
ZX_DEBUG_ASSERT(builder_ptr->field_name() == field_value); \
} \
ZX_DEBUG_ASSERT(builder_ptr->has_##field_name()); \
} while (false)
template <typename FieldRefType, typename Enable = void>
struct FieldDefaultCreator : std::false_type {};
template <typename TableType>
struct FieldDefaultCreator<TableType, std::enable_if_t<fidl::IsTable<TableType>::value>> {
static auto Create(fidl::Allocator* allocator) {
return sysmem::MakeTracking<TableType>(allocator);
}
};
template <typename VectorItemType>
struct FieldDefaultCreator<fidl::VectorView<VectorItemType>, void> {
static auto Create(fidl::Allocator* allocator, size_t count, size_t capacity) {
return sysmem::MakeTracking<VectorItemType[]>(allocator, count, capacity);
}
};
#define FIELD_DEFAULT_SET(builder_ptr_name, field_name) \
do { \
auto builder_ptr = (builder_ptr_name); \
static_assert(fidl::IsTableBuilder<std::remove_pointer_t<decltype(builder_ptr)>>::value); \
using TableType = std::remove_reference_t<decltype((builder_ptr->field_name()))>; \
static_assert(fidl::IsTable<TableType>::value); \
if (!builder_ptr->has_##field_name()) { \
builder_ptr->set_##field_name( \
sysmem::MakeTracking(&allocator_, allocator_.make_table_builder<TableType>())); \
} \
ZX_DEBUG_ASSERT(builder_ptr->has_##field_name()); \
} while (false)
// regardless of capacity, initial count is always 0
#define FIELD_DEFAULT_SET_VECTOR(builder_ptr_name, field_name, capacity_param) \
do { \
auto builder_ptr = (builder_ptr_name); \
static_assert(fidl::IsTableBuilder<std::remove_pointer_t<decltype(builder_ptr)>>::value); \
using VectorFieldType = std::remove_reference_t<decltype((builder_ptr->field_name()))>; \
static_assert(fidl::IsVectorView<VectorFieldType>::value); \
using ElementType = typename VectorFieldType::elem_type; \
if (!builder_ptr->has_##field_name()) { \
size_t capacity = (capacity_param); \
builder_ptr->set_##field_name(allocator_.make_vec_ptr<ElementType>(0, capacity)); \
} \
ZX_DEBUG_ASSERT(builder_ptr->has_##field_name()); \
} while (false)
template <typename T>
T AlignUp(T value, T divisor) {
return (value + divisor - 1) / divisor * divisor;
}
void BarrierAfterFlush() {
#if defined(__aarch64__)
// According to the ARMv8 ARM K11.5.4 it's better to use DSB instead of DMB
// for ordering with respect to MMIO (DMB is ok if all agents are just
// observing memory). The system shareability domain is used because that's
// the only domain the video decoder is guaranteed to be in. SY is used
// instead of LD or ST because section B2.3.5 says that the barrier needs both
// read and write access types to be effective with regards to cache
// operations.
asm __volatile__("dsb sy");
#elif defined(__x86_64__)
// This is here just in case we both (a) don't need to flush cache on x86 due to cache coherent
// DMA (CLFLUSH not needed), and (b) we have code using non-temporal stores or "string
// operations" whose surrounding code didn't itself take care of doing an SFENCE. After returning
// from this function, we may write to MMIO to start DMA - we want any previous (program order)
// non-temporal stores to be visible to HW before that MMIO write that starts DMA. The MFENCE
// instead of SFENCE is mainly paranoia, though one could hypothetically create HW that starts or
// continues DMA based on an MMIO read (please don't), in which case MFENCE might be needed here
// before that read.
asm __volatile__("mfence");
#else
ZX_PANIC("logical_buffer_collection.cc missing BarrierAfterFlush() impl for this platform");
#endif
}
} // namespace
// static
void LogicalBufferCollection::Create(zx::channel buffer_collection_token_request,
Device* parent_device) {
fbl::RefPtr<LogicalBufferCollection> logical_buffer_collection =
fbl::AdoptRef<LogicalBufferCollection>(new LogicalBufferCollection(parent_device));
// The existence of a channel-owned BufferCollectionToken adds a
// fbl::RefPtr<> ref to LogicalBufferCollection.
LogInfo("LogicalBufferCollection::Create()");
logical_buffer_collection->CreateBufferCollectionToken(
logical_buffer_collection, std::numeric_limits<uint32_t>::max(),
std::move(buffer_collection_token_request), nullptr);
}
// static
//
// The buffer_collection_token is the client end of the BufferCollectionToken
// which the client is exchanging for the BufferCollection (which the client is
// passing the server end of in buffer_collection_request).
//
// However, before we convert the client's token into a BufferCollection and
// start processing the messages the client may have already sent toward the
// BufferCollection, we want to process all the messages the client may have
// already sent toward the BufferCollectionToken. This comes up because the
// BufferCollectionToken and Allocator are separate channels.
//
// We know that fidl_server will process all messages before it processes the
// close - it intentionally delays noticing the close until no messages are
// available to read.
//
// So this method will close the buffer_collection_token and when it closes via
// normal FIDL processing path, the token will remember the
// buffer_collection_request to essentially convert itself into.
void LogicalBufferCollection::BindSharedCollection(Device* parent_device,
zx::channel buffer_collection_token,
zx::channel buffer_collection_request,
const ClientInfo* client_info) {
ZX_DEBUG_ASSERT(buffer_collection_token);
ZX_DEBUG_ASSERT(buffer_collection_request);
zx_koid_t token_client_koid;
zx_koid_t token_server_koid;
zx_status_t status =
get_channel_koids(buffer_collection_token, &token_client_koid, &token_server_koid);
if (status != ZX_OK) {
LogErrorStatic(client_info, "Failed to get channel koids");
// ~buffer_collection_token
// ~buffer_collection_request
return;
}
BufferCollectionToken* token = parent_device->FindTokenByServerChannelKoid(token_server_koid);
if (!token) {
// The most likely scenario for why the token was not found is that Sync() was not called on
// either the BufferCollectionToken or the BufferCollection.
LogErrorStatic(client_info,
"BindSharedCollection could not find token from server channel koid %ld; "
"perhaps BufferCollectionToken.Sync() was not called",
token_server_koid);
// ~buffer_collection_token
// ~buffer_collection_request
return;
}
// This will token->FailAsync() if the token has already got one, or if the
// token already saw token->Close().
token->SetBufferCollectionRequest(std::move(buffer_collection_request));
if (client_info) {
// The info will be propagated into the logcial buffer collection when the token closes.
token->SetDebugClientInfo(client_info->name, client_info->id);
}
// At this point, the token will process the rest of its previously queued
// messages (from client to server), and then will convert the token into
// a BufferCollection (view). That conversion happens async shortly in
// BindSharedCollectionInternal() (unless the LogicalBufferCollection fails
// before then, in which case everything just gets deleted).
//
// ~buffer_collection_token here closes the client end of the token, but we
// still process the rest of the queued messages before we process the
// close.
//
// ~buffer_collection_token
}
zx_status_t LogicalBufferCollection::ValidateBufferCollectionToken(Device* parent_device,
zx_koid_t token_server_koid) {
BufferCollectionToken* token = parent_device->FindTokenByServerChannelKoid(token_server_koid);
return token ? ZX_OK : ZX_ERR_NOT_FOUND;
}
void LogicalBufferCollection::CreateBufferCollectionToken(
fbl::RefPtr<LogicalBufferCollection> self, uint32_t rights_attenuation_mask,
zx::channel buffer_collection_token_request, const ClientInfo* client_info) {
ZX_DEBUG_ASSERT(buffer_collection_token_request.get());
auto token = BufferCollectionToken::Create(parent_device_, self, rights_attenuation_mask);
token->SetErrorHandler([this, token_ptr = token.get()](zx_status_t status) {
// Clean close from FIDL channel point of view is ZX_ERR_PEER_CLOSED,
// and ZX_OK is never passed to the error handler.
ZX_DEBUG_ASSERT(status != ZX_OK);
// The dispatcher shut down before we were able to Bind(...)
if (status == ZX_ERR_BAD_STATE) {
Fail("sysmem dispatcher shutting down - status: %d", status);
return;
}
// We know |this| is alive because the token is alive and the token has
// a fbl::RefPtr<LogicalBufferCollection>. The token is alive because
// the token is still in token_views_.
//
// Any other deletion of the token_ptr out of token_views_ (outside of
// this error handler) doesn't run this error handler.
//
// TODO(dustingreen): Switch to contains() when C++20.
ZX_DEBUG_ASSERT(token_views_.find(token_ptr) != token_views_.end());
zx::channel buffer_collection_request = token_ptr->TakeBufferCollectionRequest();
if (!(status == ZX_ERR_PEER_CLOSED && (token_ptr->is_done() || buffer_collection_request))) {
// We don't have to explicitly remove token from token_views_
// because Fail() will token_views_.clear().
//
// A token whose error handler sees anything other than clean close
// with is_done() implies LogicalBufferCollection failure. The
// ability to detect unexpected closure of a token is a main reason
// we use a channel for BufferCollectionToken instead of an
// eventpair.
//
// If a participant for some reason finds itself with an extra token it doesn't need, the
// participant should use Close() to avoid triggering this failure.
Fail("Token failure causing LogicalBufferCollection failure - status: %d", status);
return;
}
// At this point we know the token channel was closed cleanly, and that
// before the client's closing the channel, the client did a
// token::Close() or allocator::BindSharedCollection().
ZX_DEBUG_ASSERT(status == ZX_ERR_PEER_CLOSED &&
(token_ptr->is_done() || buffer_collection_request));
// BufferCollectionToken enforces that these never both become true; the
// BufferCollectionToken will fail instead.
ZX_DEBUG_ASSERT(!(token_ptr->is_done() && buffer_collection_request));
if (!buffer_collection_request) {
// This was a token::Close(). In this case we want to stop tracking the token now that we've
// processed all its previously-queued inbound messages. This might be the last token, so we
// MaybeAllocate(). This path isn't a failure (unless there are also zero BufferCollection
// views in which case MaybeAllocate() calls Fail()).
auto self = token_ptr->parent_shared();
ZX_DEBUG_ASSERT(self.get() == this);
token_views_.erase(token_ptr);
MaybeAllocate();
// ~self may delete "this"
} else {
// At this point we know that this was a BindSharedCollection(). We
// need to convert the BufferCollectionToken into a BufferCollection.
//
// ~token_ptr during this call
BindSharedCollectionInternal(token_ptr, std::move(buffer_collection_request));
}
});
auto token_ptr = token.get();
token_views_.insert({token_ptr, std::move(token)});
zx_koid_t server_koid;
zx_koid_t client_koid;
zx_status_t status =
get_channel_koids(buffer_collection_token_request, &server_koid, &client_koid);
if (status != ZX_OK) {
Fail("get_channel_koids() failed - status: %d", status);
return;
}
token_ptr->SetServerKoid(server_koid);
if (token_ptr->was_unfound_token()) {
LogClientError(client_info,
"BufferCollectionToken.Duplicate() received for creating token with server koid"
"%ld after BindSharedCollection() previously received attempting to use same"
"token. Client sequence should be Duplicate(), Sync(), BindSharedCollection()."
"Missing Sync()?",
server_koid);
}
LogInfo("CreateBufferCollectionToken() - server_koid: %lu", token_ptr->server_koid());
token_ptr->Bind(std::move(buffer_collection_token_request));
}
void LogicalBufferCollection::OnSetConstraints() {
MaybeAllocate();
return;
}
void LogicalBufferCollection::SetName(uint32_t priority, std::string name) {
if (!name_ || (priority > name_->priority)) {
name_ = CollectionName{priority, name};
name_property_ = node_.CreateString("name", name);
}
}
void LogicalBufferCollection::SetDebugTimeoutLogDeadline(int64_t deadline) {
creation_timer_.Cancel();
zx_status_t status =
creation_timer_.PostForTime(parent_device_->dispatcher(), zx::time(deadline));
ZX_ASSERT(status == ZX_OK);
}
LogicalBufferCollection::AllocationResult LogicalBufferCollection::allocation_result() {
ZX_DEBUG_ASSERT(has_allocation_result_ ||
(allocation_result_status_ == ZX_OK && !allocation_result_info_));
// If this assert fails, it mean we've already done ::Fail(). This should be impossible since
// Fail() clears all BufferCollection views so they shouldn't be able to call
// ::allocation_result().
ZX_DEBUG_ASSERT(
!(has_allocation_result_ && allocation_result_status_ == ZX_OK && !allocation_result_info_));
return {
.buffer_collection_info =
allocation_result_info_ ? &allocation_result_info_.value() : nullptr,
.status = allocation_result_status_,
};
}
LogicalBufferCollection::LogicalBufferCollection(Device* parent_device)
: parent_device_(parent_device) {
TRACE_DURATION("gfx", "LogicalBufferCollection::LogicalBufferCollection", "this", this);
LogInfo("LogicalBufferCollection::LogicalBufferCollection()");
parent_device_->AddLogicalBufferCollection(this);
node_ = parent_device_->collections_node().CreateChild(CreateUniqueName("logical-collection-"));
zx_status_t status = creation_timer_.PostDelayed(parent_device_->dispatcher(), zx::sec(5));
ZX_ASSERT(status == ZX_OK);
// nothing else to do here
}
LogicalBufferCollection::~LogicalBufferCollection() {
TRACE_DURATION("gfx", "LogicalBufferCollection::~LogicalBufferCollection", "this", this);
LogInfo("~LogicalBufferCollection");
// Every entry in these collections keeps a
// fbl::RefPtr<LogicalBufferCollection>, so these should both already be
// empty.
ZX_DEBUG_ASSERT(token_views_.empty());
ZX_DEBUG_ASSERT(collection_views_.empty());
// Cancel all TrackedParentVmo waits to avoid a use-after-free of |this|
for (auto& tracked : parent_vmos_) {
tracked.second->CancelWait();
}
if (memory_allocator_) {
memory_allocator_->RemoveDestroyCallback(reinterpret_cast<intptr_t>(this));
}
parent_device_->RemoveLogicalBufferCollection(this);
// LOG(INFO, "~LogicalBufferCollection allocator_.debug_needed_buffer_size(): %zu\n",
// allocator_.inner_allocator().debug_needed_buffer_size());
}
void LogicalBufferCollection::Fail(const char* format, ...) {
if (format) {
va_list args;
va_start(args, format);
vLog(true, "LogicalBufferCollection", "fail", format, args);
va_end(args);
}
// Close all the associated channels. We do this by swapping into local
// collections and clearing those, since deleting the items in the
// collections will delete |this|.
TokenMap local_token_views;
token_views_.swap(local_token_views);
CollectionMap local_collection_views;
collection_views_.swap(local_collection_views);
// Since all the token views and collection views will shortly be gone, there
// will be no way for any client to be sent the VMOs again, so we can close
// the handles to the VMOs here. This is necessary in order to get
// ZX_VMO_ZERO_CHILDREN to happen in TrackedParentVmo, but not sufficient
// alone (clients must also close their VMO(s)).
//
// We can't just allocation_result_info_.reset() here, because we're using a
// BufferThenHeapAllocator<> that'll delay close of the VMOs until during
// ~LogicalBufferCollection (a deadlock) unless we dig into the structure and
// close these VMOs directly.
if (allocation_result_info_) {
for (uint32_t i = 0; i < allocation_result_info_->buffers().count(); ++i) {
if (allocation_result_info_->buffers()[i].has_vmo()) {
allocation_result_info_->buffers()[i].vmo().reset();
}
if (allocation_result_info_->buffers()[i].has_aux_vmo()) {
allocation_result_info_->buffers()[i].aux_vmo().reset();
}
}
}
allocation_result_info_.reset();
// |this| can be deleted during these calls to clear(), unless parent_vmos_
// isn't empty yet, or unless the caller of Fail() has its own temporary
// fbl::RefPtr<LogicalBufferCollection> on the stack.
//
// These clear() calls will close the channels, which in turn will inform
// the participants to close their child VMO handles. We don't revoke the
// child VMOs, so the LogicalBufferCollection will stick around until
// parent_vmo_map_ becomes empty thanks to participants closing their child
// VMOs.
local_token_views.clear();
local_collection_views.clear();
}
void LogicalBufferCollection::LogInfo(const char* format, ...) {
va_list args;
va_start(args, format);
vLog(false, "LogicalBufferCollection", "info", format, args);
va_end(args);
}
void LogicalBufferCollection::LogErrorStatic(const ClientInfo* client_info, const char* format,
...) {
va_list args;
va_start(args, format);
fbl::String formatted = fbl::StringVPrintf(format, args);
if (client_info && !client_info->name.empty()) {
fbl::String client_name =
fbl::StringPrintf(" - client \"%s\" id %ld", client_info->name.c_str(), client_info->id);
formatted = fbl::String::Concat({formatted, client_name});
}
zxlogf(ERROR, "[LogicalBufferCollection error] %s", formatted.c_str());
va_end(args);
}
void LogicalBufferCollection::VLogClientError(const ClientInfo* client_info, const char* format,
va_list args) {
const char* collection_name = name_ ? name_->name.c_str() : "Unknown";
fbl::String formatted = fbl::StringVPrintf(format, args);
if (client_info && !client_info->name.empty()) {
fbl::String client_name =
fbl::StringPrintf(" - collection \"%s\" - client \"%s\" id %ld", collection_name,
client_info->name.c_str(), client_info->id);
formatted = fbl::String::Concat({formatted, client_name});
} else {
fbl::String client_name = fbl::StringPrintf(" - collection \"%s\"", collection_name);
formatted = fbl::String::Concat({formatted, client_name});
}
zxlogf(ERROR, "[LogicalBufferCollection error] %s", formatted.c_str());
va_end(args);
}
void LogicalBufferCollection::LogClientError(const ClientInfo* client_info, const char* format,
...) {
va_list args;
va_start(args, format);
VLogClientError(client_info, format, args);
va_end(args);
}
void LogicalBufferCollection::LogError(const char* format, ...) {
va_list args;
va_start(args, format);
VLogError(format, args);
va_end(args);
}
void LogicalBufferCollection::VLogError(const char* format, va_list args) {
VLogClientError(current_client_info_, format, args);
}
void LogicalBufferCollection::InitializeConstraintSnapshots(
const ConstraintsList& constraints_list) {
ZX_DEBUG_ASSERT(constraints_at_allocation_.empty());
ZX_DEBUG_ASSERT(!constraints_list.empty());
for (auto& constraints : constraints_list) {
ConstraintInfoSnapshot snapshot;
snapshot.node = node().CreateChild(CreateUniqueName("collection-at-allocation-"));
if (constraints.builder.has_min_buffer_count_for_camping()) {
snapshot.node.CreateUint("min_buffer_count_for_camping",
constraints.builder.min_buffer_count_for_camping(),
&snapshot.node_constraints);
}
if (constraints.builder.has_min_buffer_count_for_shared_slack()) {
snapshot.node.CreateUint("min_buffer_count_for_shared_slack",
constraints.builder.min_buffer_count_for_shared_slack(),
&snapshot.node_constraints);
}
if (constraints.builder.has_min_buffer_count_for_dedicated_slack()) {
snapshot.node.CreateUint("min_buffer_count_for_dedicated_slack",
constraints.builder.min_buffer_count_for_dedicated_slack(),
&snapshot.node_constraints);
}
if (constraints.builder.has_min_buffer_count()) {
snapshot.node.CreateUint("min_buffer_count", constraints.builder.min_buffer_count(),
&snapshot.node_constraints);
}
snapshot.node.CreateUint("debug_id", constraints.client.id, &snapshot.node_constraints);
snapshot.node.CreateString("debug_name", constraints.client.name, &snapshot.node_constraints);
constraints_at_allocation_.push_back(std::move(snapshot));
}
}
void LogicalBufferCollection::MaybeAllocate() {
if (!token_views_.empty()) {
// All tokens must be converted into BufferCollection views or Close()ed
// before allocation will happen.
return;
}
if (collection_views_.empty()) {
// The LogicalBufferCollection should be failed because there are no clients left, despite only
// getting here if all of the clients did a clean Close().
if (is_allocate_attempted_) {
// Only log as info because this is a normal way to destroy the buffer collection.
LogInfo("All clients called Close(), but now zero clients remain (after allocation).");
Fail(nullptr);
} else {
Fail("All clients called Close(), but now zero clients remain (before allocation).");
}
return;
}
if (is_allocate_attempted_) {
// Allocate was already attempted.
return;
}
// Sweep looking for any views that don't have constraints.
for (auto& [key, value] : collection_views_) {
if (!key->has_constraints()) {
return;
}
}
// All the views have seen SetConstraints(), and there are no tokens left.
// Regardless of whether allocation succeeds or fails, we remember we've
// started an attempt to allocate so we don't attempt again.
is_allocate_attempted_ = true;
TryAllocate();
return;
}
// This only runs on a clean stack.
void LogicalBufferCollection::TryAllocate() {
TRACE_DURATION("gfx", "LogicalBufferCollection::TryAllocate", "this", this);
// If we're here it means we still have collection_views_, because if the
// last collection view disappeared we would have run ~this which would have
// cleared the Post() canary so this method woudn't be running.
ZX_DEBUG_ASSERT(!collection_views_.empty());
// Currently only BufferCollection(s) that have already done a clean Close()
// have their constraints in constraints_list_. The rest of the constraints
// are still with collection_views_. Move all constraints into
// constraints_list_.
for (auto& [key, value] : collection_views_) {
ZX_DEBUG_ASSERT(key->has_constraints());
constraints_list_.emplace_back(key->TakeConstraints(),
ClientInfo{key->debug_name(), key->debug_id()});
ZX_DEBUG_ASSERT(!key->has_constraints());
}
InitializeConstraintSnapshots(constraints_list_);
if (!CombineConstraints()) {
// It's impossible to combine the constraints due to incompatible
// constraints, or all participants set null constraints.
SetFailedAllocationResult(ZX_ERR_NOT_SUPPORTED);
return;
}
ZX_DEBUG_ASSERT(!!constraints_);
fit::result<llcpp::fuchsia::sysmem2::BufferCollectionInfo, zx_status_t> result = Allocate();
if (!result.is_ok()) {
ZX_DEBUG_ASSERT(result.error() != ZX_OK);
SetFailedAllocationResult(result.error());
return;
}
ZX_DEBUG_ASSERT(result.is_ok());
SetAllocationResult(result.take_value());
return;
}
void LogicalBufferCollection::SetFailedAllocationResult(zx_status_t status) {
ZX_DEBUG_ASSERT(status != ZX_OK);
// Only set result once.
ZX_DEBUG_ASSERT(!has_allocation_result_);
// allocation_result_status_ is initialized to ZX_OK, so should still be set
// that way.
ZX_DEBUG_ASSERT(allocation_result_status_ == ZX_OK);
creation_timer_.Cancel();
allocation_result_status_ = status;
// Was initialized to nullptr.
ZX_DEBUG_ASSERT(!allocation_result_info_);
has_allocation_result_ = true;
SendAllocationResult();
return;
}
void LogicalBufferCollection::SetAllocationResult(
llcpp::fuchsia::sysmem2::BufferCollectionInfo&& info) {
// Setting empty constraints as the success case isn't allowed. That's
// considered a failure. At least one participant must specify non-empty
// constraints.
ZX_DEBUG_ASSERT(!info.IsEmpty());
// Only set result once.
ZX_DEBUG_ASSERT(!has_allocation_result_);
// allocation_result_status_ is initialized to ZX_OK, so should still be set
// that way.
ZX_DEBUG_ASSERT(allocation_result_status_ == ZX_OK);
creation_timer_.Cancel();
allocation_result_status_ = ZX_OK;
allocation_result_info_ = std::move(info);
has_allocation_result_ = true;
SendAllocationResult();
return;
}
void LogicalBufferCollection::SendAllocationResult() {
ZX_DEBUG_ASSERT(has_allocation_result_);
ZX_DEBUG_ASSERT(token_views_.empty());
ZX_DEBUG_ASSERT(!collection_views_.empty());
for (auto& [key, value] : collection_views_) {
key->OnBuffersAllocated();
}
if (allocation_result_status_ != ZX_OK) {
Fail(
"LogicalBufferCollection::SendAllocationResult() done sending "
"allocation failure - now auto-failing self.");
return;
}
}
void LogicalBufferCollection::BindSharedCollectionInternal(BufferCollectionToken* token,
zx::channel buffer_collection_request) {
ZX_DEBUG_ASSERT(buffer_collection_request.get());
auto self = token->parent_shared();
ZX_DEBUG_ASSERT(self.get() == this);
auto collection = BufferCollection::Create(self);
collection->SetDebugClientInfo(token->debug_name().data(), token->debug_name().size(),
token->debug_id());
collection->SetErrorHandler([this, collection_ptr = collection.get()](zx_status_t status) {
// status passed to an error handler is never ZX_OK. Clean close is
// ZX_ERR_PEER_CLOSED.
ZX_DEBUG_ASSERT(status != ZX_OK);
// The dispatcher shut down before we were able to Bind(...)
if (status == ZX_ERR_BAD_STATE) {
Fail("sysmem dispatcher shutting down - status: %d", status);
return;
}
// We know collection_ptr is still alive because collection_ptr is
// still in collection_views_. We know this is still alive because
// this has a RefPtr<> ref from collection_ptr.
//
// TODO(dustingreen): Switch to contains() when C++20.
ZX_DEBUG_ASSERT(collection_views_.find(collection_ptr) != collection_views_.end());
// The BufferCollection may have had Close() called on it, in which
// case closure of the BufferCollection doesn't cause
// LogicalBufferCollection failure. Or, Close() wasn't called and
// the LogicalBufferCollection is out of here.
if (!(status == ZX_ERR_PEER_CLOSED && collection_ptr->is_done())) {
// We don't have to explicitly remove collection from collection_views_ because Fail() will
// collection_views_.clear().
//
// A BufferCollection view whose error handler runs implies LogicalBufferCollection failure.
//
// A LogicalBufferCollection intentionally treats any error that might be triggered by a
// client failure as a LogicalBufferCollection failure, because a LogicalBufferCollection can
// use a lot of RAM and can tend to block creating a replacement LogicalBufferCollection.
//
// If a participant is cleanly told to be done with a BufferCollection, the participant can
// send Close() before BufferCollection channel close to avoid triggering this failure, in
// case the initiator might want to continue using the BufferCollection without the
// participant.
//
// TODO(fxbug.dev/33670): Provide a way to mark a BufferCollection view as expendable without
// implying that the channel is closing, so that the client can still detect when the
// BufferCollection VMOs need to be closed based on BufferCollection channel closure by
// sysmem.
//
// In rare cases, an initiator might choose to use Close() to avoid this failure, but more
// typically initiators will just close their BufferCollection view without Close() first, and
// this failure results. This is considered acceptable partly because it helps exercise code
// in participants that may see BufferCollection channel closure before closure of related
// channels, and it helps get the VMO handles closed ASAP to avoid letting those continue to
// use space of a MemoryAllocator's pool of pre-reserved space (for example).
//
// TODO(fxbug.dev/45878): Provide a way to distinguish between BufferCollection clean/unclean
// close so that we print an error if participant closes before initiator
Fail(nullptr);
return;
}
// At this point we know the collection_ptr is cleanly done (Close()
// was sent from client) and can be removed from the set of tracked
// collections. We keep the collection's constraints (if any), as
// those are still relevant - this lets a participant do
// SetConstraints() followed by Close() followed by closing the
// participant's BufferCollection channel, which is convenient for
// some participants.
//
// If this causes collection_tokens_.empty() and collection_views_.empty(),
// MaybeAllocate() takes care of calling Fail().
if (collection_ptr->has_constraints()) {
constraints_list_.emplace_back(
collection_ptr->TakeConstraints(),
ClientInfo{collection_ptr->debug_name(), collection_ptr->debug_id()});
}
auto self = collection_ptr->parent_shared();
ZX_DEBUG_ASSERT(self.get() == this);
collection_views_.erase(collection_ptr);
MaybeAllocate();
// ~self may delete "this"
return;
});
auto collection_ptr = collection.get();
collection_views_.insert({collection_ptr, std::move(collection)});
// ~BufferCollectionToken calls UntrackTokenKoid().
token_views_.erase(token);
collection_ptr->Bind(std::move(buffer_collection_request));
}
bool LogicalBufferCollection::IsMinBufferSizeSpecifiedByAnyParticipant() {
ZX_DEBUG_ASSERT(!collection_views_.empty());
ZX_DEBUG_ASSERT(collection_views_.end() ==
std::find_if(collection_views_.begin(), collection_views_.end(),
[](auto& item_pair) { return item_pair.first->has_constraints(); }));
ZX_DEBUG_ASSERT(!constraints_list_.empty());
for (auto& [constraints, client_info_unused] : constraints_list_) {
if (constraints.has_buffer_memory_constraints() &&
constraints.buffer_memory_constraints().has_min_size_bytes() &&
constraints.buffer_memory_constraints().min_size_bytes() > 0) {
return true;
}
if (constraints.has_image_format_constraints()) {
for (auto& image_format_constraints : constraints.image_format_constraints()) {
if (image_format_constraints.has_min_coded_width() &&
image_format_constraints.has_min_coded_height() &&
image_format_constraints.min_coded_width() > 0 &&
image_format_constraints.min_coded_height() > 0) {
return true;
}
if (image_format_constraints.has_required_max_coded_width() &&
image_format_constraints.has_required_max_coded_height() &&
image_format_constraints.required_max_coded_width() > 0 &&
image_format_constraints.required_max_coded_height() > 0) {
return true;
}
}
}
}
return false;
}
bool LogicalBufferCollection::CombineConstraints() {
// This doesn't necessarily mean that any of the collection_views_ have
// set non-empty constraints though. We do require that at least one
// participant (probably the initiator) retains an open channel to its
// BufferCollection until allocation is done, else allocation won't be
// attempted.
ZX_DEBUG_ASSERT(!collection_views_.empty());
// Caller is supposed to move all constraints into constraints_list_ before calling.
ZX_DEBUG_ASSERT(collection_views_.end() ==
std::find_if(collection_views_.begin(), collection_views_.end(),
[](auto& item_pair) { return item_pair.first->has_constraints(); }));
// We also know that all the constraints are in constraints_list_ now,
// including all constraints from collection_views_.
ZX_DEBUG_ASSERT(!constraints_list_.empty());
// At least one participant must specify min buffer size (in terms of non-zero min buffer size or
// non-zero min image size or non-zero potential max image size).
//
// This also enforces that at least one participant must specify non-empty constraints.
if (!IsMinBufferSizeSpecifiedByAnyParticipant()) {
// Too unconstrained... We refuse to allocate buffers without any min size
// bounds from any participant. At least one particpant must provide
// some form of size bounds (in terms of buffer size bounds or in terms
// of image size bounds).
LogError(
"At least one participant must specify buffer_memory_constraints or "
"image_format_constraints that implies non-zero min buffer size.");
return false;
}
// Start with empty constraints / unconstrained.
auto acc = allocator_.make_table_builder<llcpp::fuchsia::sysmem2::BufferCollectionConstraints>();
// Sanitize initial accumulation target to keep accumulation simpler. This is guaranteed to
// succeed; the input is always the same.
bool result = CheckSanitizeBufferCollectionConstraints(CheckSanitizeStage::kInitial, &acc);
ZX_DEBUG_ASSERT(result);
// Accumulate each participant's constraints.
while (!constraints_list_.empty()) {
Constraints constraints_entry = std::move(constraints_list_.front());
constraints_list_.pop_front();
current_client_info_ = &constraints_entry.client;
auto defer_reset = fit::defer([this] { current_client_info_ = nullptr; });
if (!CheckSanitizeBufferCollectionConstraints(CheckSanitizeStage::kNotAggregated,
&constraints_entry.builder)) {
return false;
}
auto constraints = constraints_entry.builder.build();
if (!AccumulateConstraintBufferCollection(&acc, &constraints)) {
// This is a failure. The space of permitted settings contains no
// points.
return false;
}
}
if (!CheckSanitizeBufferCollectionConstraints(CheckSanitizeStage::kAggregated, &acc)) {
return false;
}
constraints_ = std::move(acc);
return true;
}
// TODO(dustingreen): Consider rejecting secure_required + any non-secure heaps, including the
// potentially-implicit SYSTEM_RAM heap.
//
// TODO(dustingreen): From a particular participant, CPU usage without
// IsCpuAccessibleHeapPermitted() should fail.
//
// TODO(dustingreen): From a particular participant, be more picky about which domains are supported
// vs. which heaps are supported.
static bool IsHeapPermitted(
const llcpp::fuchsia::sysmem2::BufferMemoryConstraints::Builder& constraints,
llcpp::fuchsia::sysmem2::HeapType heap) {
if (constraints.heap_permitted().count()) {
auto begin = constraints.heap_permitted().begin();
auto end = constraints.heap_permitted().end();
return std::find(begin, end, heap) != end;
}
// Zero heaps in heap_permitted() means any heap is ok.
return true;
}
static bool IsSecurePermitted(
const llcpp::fuchsia::sysmem2::BufferMemoryConstraints::Builder& constraints) {
// TODO(fxbug.dev/37452): Generalize this by finding if there's a heap that maps to secure
// MemoryAllocator in the permitted heaps.
return constraints.inaccessible_domain_supported() &&
(IsHeapPermitted(constraints, llcpp::fuchsia::sysmem2::HeapType::AMLOGIC_SECURE) ||
IsHeapPermitted(constraints, llcpp::fuchsia::sysmem2::HeapType::AMLOGIC_SECURE_VDEC));
}
static bool IsCpuAccessSupported(
const llcpp::fuchsia::sysmem2::BufferMemoryConstraints& constraints) {
return constraints.cpu_domain_supported() || constraints.ram_domain_supported();
}
bool LogicalBufferCollection::CheckSanitizeBufferUsage(
CheckSanitizeStage stage, llcpp::fuchsia::sysmem2::BufferUsage::Builder* buffer_usage) {
FIELD_DEFAULT_ZERO(buffer_usage, none);
FIELD_DEFAULT_ZERO(buffer_usage, cpu);
FIELD_DEFAULT_ZERO(buffer_usage, vulkan);
FIELD_DEFAULT_ZERO(buffer_usage, display);
FIELD_DEFAULT_ZERO(buffer_usage, video);
switch (stage) {
case CheckSanitizeStage::kInitial:
// empty usage is allowed for kInitial
break;
case CheckSanitizeStage::kNotAggregated:
// At least one usage bit must be specified by any participant that
// specifies constraints. The "none" usage bit can be set by a participant
// that doesn't directly use the buffers, so we know that the participant
// didn't forget to set usage.
if (buffer_usage->none() == 0 && buffer_usage->cpu() == 0 && buffer_usage->vulkan() == 0 &&
buffer_usage->display() == 0 && buffer_usage->video() == 0) {
LogError("At least one usage bit must be set by a participant.");
return false;
}
if (buffer_usage->none() != 0) {
if (buffer_usage->cpu() != 0 || buffer_usage->vulkan() != 0 ||
buffer_usage->display() != 0 || buffer_usage->video() != 0) {
LogError("A participant indicating 'none' usage can't specify any other usage.");
return false;
}
}
break;
case CheckSanitizeStage::kAggregated:
if (buffer_usage->cpu() == 0 && buffer_usage->vulkan() == 0 && buffer_usage->display() == 0 &&
buffer_usage->video() == 0) {
LogError("At least one non-'none' usage bit must be set across all participants.");
return false;
}
break;
}
return true;
}
size_t LogicalBufferCollection::InitialCapacityOrZero(CheckSanitizeStage stage,
size_t initial_capacity) {
return (stage == CheckSanitizeStage::kInitial) ? initial_capacity : 0;
}
// Nearly all constraint checks must go under here or under ::Allocate() (not in
// the Accumulate* methods), else we could fail to notice a single participant
// providing unsatisfiable constraints, where no Accumulate* happens. The
// constraint checks that are present under Accumulate* are commented explaining
// why it's ok for them to be there.
bool LogicalBufferCollection::CheckSanitizeBufferCollectionConstraints(
CheckSanitizeStage stage,
llcpp::fuchsia::sysmem2::BufferCollectionConstraints::Builder* constraints) {
bool was_empty = constraints->IsEmpty();
FIELD_DEFAULT_SET(constraints, usage);
if (was_empty) {
// Completely empty constraints are permitted, so convert to NONE_USAGE to avoid triggering the
// check applied to non-empty constraints where at least one usage bit must be set (NONE_USAGE
// counts for that check, and doesn't constrain anything).
FIELD_DEFAULT(&constraints->get_builder_usage(), none, llcpp::fuchsia::sysmem2::NONE_USAGE);
}
FIELD_DEFAULT_ZERO(constraints, min_buffer_count_for_camping);
FIELD_DEFAULT_ZERO(constraints, min_buffer_count_for_dedicated_slack);
FIELD_DEFAULT_ZERO(constraints, min_buffer_count_for_shared_slack);
FIELD_DEFAULT_ZERO(constraints, min_buffer_count);
FIELD_DEFAULT_MAX(constraints, max_buffer_count);
ZX_DEBUG_ASSERT(constraints->has_buffer_memory_constraints() ||
stage != CheckSanitizeStage::kAggregated);
FIELD_DEFAULT_SET(constraints, buffer_memory_constraints);
ZX_DEBUG_ASSERT(constraints->has_buffer_memory_constraints());
FIELD_DEFAULT_SET_VECTOR(constraints, image_format_constraints, InitialCapacityOrZero(stage, 64));
FIELD_DEFAULT_FALSE(constraints, need_clear_aux_buffers_for_secure);
FIELD_DEFAULT(constraints, allow_clear_aux_buffers_for_secure,
!IsWriteUsage(constraints->usage()));
if (!CheckSanitizeBufferUsage(stage, &constraints->get_builder_usage())) {
LogError("CheckSanitizeBufferUsage() failed");
return false;
}
if (constraints->max_buffer_count() == 0) {
LogError("max_buffer_count == 0");
return false;
}
if (constraints->min_buffer_count() > constraints->max_buffer_count()) {
LogError("min_buffer_count > max_buffer_count");
return false;
}
if (!CheckSanitizeBufferMemoryConstraints(
stage, constraints->usage(), &constraints->get_builder_buffer_memory_constraints())) {
return false;
}
if (stage != CheckSanitizeStage::kAggregated) {
if (IsCpuUsage(constraints->usage())) {
if (!IsCpuAccessSupported(constraints->buffer_memory_constraints())) {
LogError("IsCpuUsage() && !IsCpuAccessSupported()");
return false;
}
// From a single participant, reject secure_required in combination with CPU usage, since CPU
// usage isn't possible given secure memory.
if (constraints->buffer_memory_constraints().secure_required()) {
LogError("IsCpuUsage() && secure_required");
return false;
}
// It's fine if a participant sets CPU usage but also permits inaccessible domain and possibly
// IsSecurePermitted(). In that case the participant is expected to pay attetion to the
// coherency domain and is_secure and realize that it shouldn't attempt to read/write the
// VMOs.
}
if (constraints->buffer_memory_constraints().secure_required() &&
IsCpuAccessSupported(constraints->buffer_memory_constraints())) {
// This is a little picky, but easier to be less picky later than more picky later.
LogError("secure_required && IsCpuAccessSupported()");
return false;
}
}
for (uint32_t i = 0; i < constraints->image_format_constraints().count(); ++i) {
if (!CheckSanitizeImageFormatConstraints(
stage, &constraints->get_builders_image_format_constraints()[i])) {
return false;
}
}
if (stage == CheckSanitizeStage::kNotAggregated) {
// As an optimization, only check the unaggregated inputs.
for (uint32_t i = 0; i < constraints->image_format_constraints().count(); ++i) {
for (uint32_t j = i + 1; j < constraints->image_format_constraints().count(); ++j) {
if (ImageFormatIsPixelFormatEqual(
constraints->image_format_constraints()[i].pixel_format(),
constraints->image_format_constraints()[j].pixel_format())) {
LogError("image format constraints %d and %d have identical formats", i, j);
return false;
}
}
}
}
return true;
}
bool LogicalBufferCollection::CheckSanitizeBufferMemoryConstraints(
CheckSanitizeStage stage, const llcpp::fuchsia::sysmem2::BufferUsage& buffer_usage,
llcpp::fuchsia::sysmem2::BufferMemoryConstraints::Builder* constraints) {
FIELD_DEFAULT_ZERO(constraints, min_size_bytes);
FIELD_DEFAULT_MAX(constraints, max_size_bytes);
FIELD_DEFAULT_FALSE(constraints, physically_contiguous_required);
FIELD_DEFAULT_FALSE(constraints, secure_required);
// The CPU domain is supported by default.
FIELD_DEFAULT(constraints, cpu_domain_supported, true);
// If !usage.cpu, then participant doesn't care what domain, so indicate support
// for RAM and inaccessible domains in that case.
FIELD_DEFAULT(constraints, ram_domain_supported, !buffer_usage.cpu());
FIELD_DEFAULT(constraints, inaccessible_domain_supported, !buffer_usage.cpu());
if (stage != CheckSanitizeStage::kAggregated) {
if (constraints->has_heap_permitted() && !constraints->heap_permitted().count()) {
LogError("constraints->has_heap_permitted() && !constraints->heap_permitted().count()");
return false;
}
}
// TODO(dustingreen): When 0 heaps specified, constrain heap list based on other constraints.
// For now 0 heaps means any heap.
FIELD_DEFAULT_SET_VECTOR(constraints, heap_permitted, 0);
ZX_DEBUG_ASSERT(stage != CheckSanitizeStage::kInitial ||
constraints->heap_permitted().count() == 0);
if (constraints->min_size_bytes() > constraints->max_size_bytes()) {
LogError("min_size_bytes > max_size_bytes");
return false;
}
if (constraints->secure_required() && !IsSecurePermitted(*constraints)) {
LogError("secure memory required but not permitted");
return false;
}
return true;
}
bool LogicalBufferCollection::CheckSanitizeImageFormatConstraints(
CheckSanitizeStage stage,
llcpp::fuchsia::sysmem2::ImageFormatConstraints::Builder* constraints) {
// We never CheckSanitizeImageFormatConstraints() on empty (aka initial) constraints.
ZX_DEBUG_ASSERT(stage != CheckSanitizeStage::kInitial);
FIELD_DEFAULT_SET(constraints, pixel_format);
FIELD_DEFAULT_ZERO(&constraints->get_builder_pixel_format(), type);
FIELD_DEFAULT_ZERO(&constraints->get_builder_pixel_format(), format_modifier_value);
FIELD_DEFAULT_SET_VECTOR(constraints, color_spaces, 0);
FIELD_DEFAULT_ZERO(constraints, min_coded_width);
FIELD_DEFAULT_MAX(constraints, max_coded_width);
FIELD_DEFAULT_ZERO(constraints, min_coded_height);
FIELD_DEFAULT_MAX(constraints, max_coded_height);
FIELD_DEFAULT_ZERO(constraints, min_bytes_per_row);
FIELD_DEFAULT_MAX(constraints, max_bytes_per_row);
FIELD_DEFAULT_MAX(constraints, max_coded_width_times_coded_height);
FIELD_DEFAULT_1(constraints, coded_width_divisor);
FIELD_DEFAULT_1(constraints, coded_height_divisor);
FIELD_DEFAULT_1(constraints, bytes_per_row_divisor);
FIELD_DEFAULT_1(constraints, start_offset_divisor);
FIELD_DEFAULT_1(constraints, display_width_divisor);
FIELD_DEFAULT_1(constraints, display_height_divisor);
FIELD_DEFAULT_MAX(constraints, required_min_coded_width);
FIELD_DEFAULT_ZERO(constraints, required_max_coded_width);
FIELD_DEFAULT_MAX(constraints, required_min_coded_height);
FIELD_DEFAULT_ZERO(constraints, required_max_coded_height);
FIELD_DEFAULT_MAX(constraints, required_min_bytes_per_row);
FIELD_DEFAULT_ZERO(constraints, required_max_bytes_per_row);
if (constraints->pixel_format().type() == llcpp::fuchsia::sysmem2::PixelFormatType::INVALID) {
LogError("PixelFormatType INVALID not allowed");
return false;
}
if (!ImageFormatIsSupported(constraints->pixel_format())) {
LogError("Unsupported pixel format");
return false;
}
uint32_t min_bytes_per_row_given_min_width =
ImageFormatStrideBytesPerWidthPixel(constraints->pixel_format()) *
constraints->min_coded_width();
constraints->min_bytes_per_row() =
std::max(constraints->min_bytes_per_row(), min_bytes_per_row_given_min_width);
if (!constraints->color_spaces().count()) {
LogError("color_spaces.count() == 0 not allowed");
return false;
}
if (constraints->min_coded_width() > constraints->max_coded_width()) {
LogError("min_coded_width > max_coded_width");
return false;
}
if (constraints->min_coded_height() > constraints->max_coded_height()) {
LogError("min_coded_height > max_coded_height");
return false;
}
if (constraints->min_bytes_per_row() > constraints->max_bytes_per_row()) {
LogError("min_bytes_per_row > max_bytes_per_row");
return false;
}
if (constraints->min_coded_width() * constraints->min_coded_height() >
constraints->max_coded_width_times_coded_height()) {
LogError(
"min_coded_width * min_coded_height > "
"max_coded_width_times_coded_height");
return false;
}
if (!IsNonZeroPowerOf2(constraints->coded_width_divisor())) {
LogError("non-power-of-2 coded_width_divisor not supported");
return false;
}
if (!IsNonZeroPowerOf2(constraints->coded_height_divisor())) {
LogError("non-power-of-2 coded_width_divisor not supported");
return false;
}
if (!IsNonZeroPowerOf2(constraints->bytes_per_row_divisor())) {
LogError("non-power-of-2 bytes_per_row_divisor not supported");
return false;
}
if (!IsNonZeroPowerOf2(constraints->start_offset_divisor())) {
LogError("non-power-of-2 start_offset_divisor not supported");
return false;
}
if (constraints->start_offset_divisor() > PAGE_SIZE) {
LogError("support for start_offset_divisor > PAGE_SIZE not yet implemented");
return false;
}
if (!IsNonZeroPowerOf2(constraints->display_width_divisor())) {
LogError("non-power-of-2 display_width_divisor not supported");
return false;
}
if (!IsNonZeroPowerOf2(constraints->display_height_divisor())) {
LogError("non-power-of-2 display_height_divisor not supported");
return false;
}
for (uint32_t i = 0; i < constraints->color_spaces().count(); ++i) {
if (!ImageFormatIsSupportedColorSpaceForPixelFormat(constraints->color_spaces()[i],
constraints->pixel_format())) {
auto colorspace_type = constraints->color_spaces()[i].has_type()
? constraints->color_spaces()[i].type()
: llcpp::fuchsia::sysmem2::ColorSpaceType::INVALID;
LogError(
"!ImageFormatIsSupportedColorSpaceForPixelFormat() "
"color_space.type: %u "
"pixel_format.type: %u",
colorspace_type, constraints->pixel_format().type());
return false;
}
}
if (constraints->required_min_coded_width() == 0) {
LogError("required_min_coded_width == 0");
return false;
}
ZX_DEBUG_ASSERT(constraints->required_min_coded_width() != 0);
if (constraints->required_min_coded_width() < constraints->min_coded_width()) {
LogError("required_min_coded_width < min_coded_width");
return false;
}
if (constraints->required_max_coded_width() > constraints->max_coded_width()) {
LogError("required_max_coded_width > max_coded_width");
return false;
}
if (constraints->required_min_coded_height() == 0) {
LogError("required_min_coded_height == 0");
return false;
}
ZX_DEBUG_ASSERT(constraints->required_min_coded_height() != 0);
if (constraints->required_min_coded_height() < constraints->min_coded_height()) {
LogError("required_min_coded_height < min_coded_height");
return false;
}
if (constraints->required_max_coded_height() > constraints->max_coded_height()) {
LogError("required_max_coded_height > max_coded_height");
return false;
}
if (constraints->required_min_bytes_per_row() == 0) {
LogError("required_min_bytes_per_row == 0");
return false;
}
ZX_DEBUG_ASSERT(constraints->required_min_bytes_per_row() != 0);
if (constraints->required_min_bytes_per_row() < constraints->min_bytes_per_row()) {
LogError("required_min_bytes_per_row < min_bytes_per_row");
return false;
}
if (constraints->required_max_bytes_per_row() > constraints->max_bytes_per_row()) {
LogError("required_max_bytes_per_row > max_bytes_per_row");
return false;
}
// TODO(dustingreen): Check compatibility of color_space[] entries vs. the
// pixel_format. In particular, 2020 and 2100 don't have 8 bpp, only 10 or
// 12 bpp, while a given PixelFormat.type is a specific bpp. There's
// probably no reason to allow 2020 or 2100 to be specified along with a
// PixelFormat.type that's 8 bpp for example.
return true;
}
bool LogicalBufferCollection::AccumulateConstraintsBufferUsage(
llcpp::fuchsia::sysmem2::BufferUsage::Builder* acc, llcpp::fuchsia::sysmem2::BufferUsage* c) {
// We accumulate "none" usage just like other usages, to make aggregation and CheckSanitize
// consistent/uniform.
acc->none() |= c->none();
acc->cpu() |= c->cpu();
acc->vulkan() |= c->vulkan();
acc->display() |= c->display();
acc->video() |= c->video();
return true;
}
// |acc| accumulated constraints so far
//
// |c| additional constraint to aggregate into acc
bool LogicalBufferCollection::AccumulateConstraintBufferCollection(
llcpp::fuchsia::sysmem2::BufferCollectionConstraints::Builder* acc,
llcpp::fuchsia::sysmem2::BufferCollectionConstraints* c) {
if (!AccumulateConstraintsBufferUsage(&acc->get_builder_usage(), &c->usage())) {
return false;
}
acc->min_buffer_count_for_camping() += c->min_buffer_count_for_camping();
acc->min_buffer_count_for_dedicated_slack() += c->min_buffer_count_for_dedicated_slack();
acc->min_buffer_count_for_shared_slack() =
std::max(acc->min_buffer_count_for_shared_slack(), c->min_buffer_count_for_shared_slack());
acc->min_buffer_count() = std::max(acc->min_buffer_count(), c->min_buffer_count());
// 0 is replaced with 0xFFFFFFFF in
// CheckSanitizeBufferCollectionConstraints.
ZX_DEBUG_ASSERT(acc->max_buffer_count() != 0);
ZX_DEBUG_ASSERT(c->max_buffer_count() != 0);
acc->max_buffer_count() = std::min(acc->max_buffer_count(), c->max_buffer_count());
// CheckSanitizeBufferCollectionConstraints() takes care of setting a default
// buffer_collection_constraints, so we can assert that both acc and c "has_" one.
ZX_DEBUG_ASSERT(acc->has_buffer_memory_constraints());
ZX_DEBUG_ASSERT(c->has_buffer_memory_constraints());
if (!AccumulateConstraintBufferMemory(&acc->get_builder_buffer_memory_constraints(),
&c->buffer_memory_constraints())) {
return false;
}
if (!acc->image_format_constraints().count()) {
// Take the whole VectorView<>, as the count() can only go down later, so the capacity of
// c.image_format_constraints() is fine.
acc->set_image_format_constraints(
sysmem::MakeTracking(&allocator_, std::move(c->image_format_constraints())));
} else {
ZX_DEBUG_ASSERT(acc->image_format_constraints().count());
if (c->image_format_constraints().count()) {
if (!AccumulateConstraintImageFormats(&acc->get_builders_image_format_constraints(),
&c->image_format_constraints())) {
// We return false if we've seen non-zero
// image_format_constraint_count from at least one participant
// but among non-zero image_format_constraint_count participants
// since then the overlap has dropped to empty set.
//
// This path is taken when there are completely non-overlapping
// PixelFormats and also when PixelFormat(s) overlap but none
// of those have any non-empty settings space remaining. In
// that case we've removed the PixelFormat from consideration
// despite it being common among participants (so far).
return false;
}
ZX_DEBUG_ASSERT(acc->image_format_constraints().count());
}
}
acc->need_clear_aux_buffers_for_secure() =
acc->need_clear_aux_buffers_for_secure() || c->need_clear_aux_buffers_for_secure();
acc->allow_clear_aux_buffers_for_secure() =
acc->allow_clear_aux_buffers_for_secure() && c->allow_clear_aux_buffers_for_secure();
// We check for consistency of these later only if we're actually attempting to allocate secure
// buffers.
// acc->image_format_constraints().count() == 0 is allowed here, when all
// participants had image_format_constraints().count() == 0.
return true;
}
bool LogicalBufferCollection::AccumulateConstraintHeapPermitted(
fidl::VectorView<llcpp::fuchsia::sysmem2::HeapType>* acc,
fidl::VectorView<llcpp::fuchsia::sysmem2::HeapType>* c) {
// Remove any heap in acc that's not in c. If zero heaps
// remain in acc, return false.
ZX_DEBUG_ASSERT(acc->count() > 0);
for (uint32_t ai = 0; ai < acc->count(); ++ai) {
uint32_t ci;
for (ci = 0; ci < c->count(); ++ci) {
if ((*acc)[ai] == (*c)[ci]) {
// We found heap in c. Break so we can move on to
// the next heap.
break;
}
}
if (ci == c->count()) {
// Remove from acc because not found in c.
//
// Copy formerly last item on top of the item being removed, if not the same item.
if (ai != acc->count() - 1) {
(*acc)[ai] = (*acc)[acc->count() - 1];
}
// remove last item
acc->set_count(acc->count() - 1);
// adjust ai to force current index to be processed again as it's
// now a different item
--ai;
}
}
if (!acc->count()) {
LogError("Zero heap permitted overlap");
return false;
}
return true;
}
bool LogicalBufferCollection::AccumulateConstraintBufferMemory(
llcpp::fuchsia::sysmem2::BufferMemoryConstraints::Builder* acc,
llcpp::fuchsia::sysmem2::BufferMemoryConstraints* c) {
acc->min_size_bytes() = std::max(acc->min_size_bytes(), c->min_size_bytes());
// Don't permit 0 as the overall min_size_bytes; that would be nonsense. No
// particular initiator should feel that it has to specify 1 in this field;
// that's just built into sysmem instead. While a VMO will have a minimum
// actual size of page size, we do permit treating buffers as if they're 1
// byte, mainly for testing reasons, and to avoid any unnecessary dependence
// or assumptions re. page size.
acc->min_size_bytes() = std::max(acc->min_size_bytes(), 1u);
acc->max_size_bytes() = std::min(acc->max_size_bytes(), c->max_size_bytes());
acc->physically_contiguous_required() =
acc->physically_contiguous_required() || c->physically_contiguous_required();
acc->secure_required() = acc->secure_required() || c->secure_required();
acc->ram_domain_supported() = acc->ram_domain_supported() && c->ram_domain_supported();
acc->cpu_domain_supported() = acc->cpu_domain_supported() && c->cpu_domain_supported();
acc->inaccessible_domain_supported() =
acc->inaccessible_domain_supported() && c->inaccessible_domain_supported();
if (!acc->heap_permitted().count()) {
acc->set_heap_permitted(sysmem::MakeTracking(&allocator_, std::move(c->heap_permitted())));
} else {
if (c->heap_permitted().count()) {
if (!AccumulateConstraintHeapPermitted(&acc->heap_permitted(), &c->heap_permitted())) {
return false;
}
}
}
return true;
}
bool LogicalBufferCollection::AccumulateConstraintImageFormats(
fidl::VectorView<llcpp::fuchsia::sysmem2::ImageFormatConstraints::Builder>* acc,
fidl::VectorView<llcpp::fuchsia::sysmem2::ImageFormatConstraints>* c) {
// Remove any pixel_format in acc that's not in c. Process any format
// that's in both. If processing the format results in empty set for that
// format, pretend as if the format wasn't in c and remove that format from
// acc. If acc ends up with zero formats, return false.
// This method doesn't get called unless there's at least one format in
// acc.
ZX_DEBUG_ASSERT(acc->count());
for (uint32_t ai = 0; ai < acc->count(); ++ai) {
size_t ci;
for (ci = 0; ci < c->count(); ++ci) {
if (ImageFormatIsPixelFormatEqual((*acc)[ai].pixel_format(), (*c)[ci].pixel_format())) {
if (!AccumulateConstraintImageFormat(&(*acc)[ai], &(*c)[ci])) {
// Pretend like the format wasn't in c to begin with, so
// this format gets removed from acc. Only if this results
// in zero formats in acc do we end up returning false.
ci = c->count();
break;
}
// We found the format in c and processed the format without
// that resulting in empty set; break so we can move on to the
// next format.
break;
}
}
if (ci == c->count()) {
// Remove from acc because not found in c.
//
// Move last item on top of the item being removed, if not the same item.
if (ai != acc->count() - 1) {
(*acc)[ai] = std::move((*acc)[acc->count() - 1]);
} else {
// Stuff under this item would get deleted later anyway, but delete now to avoid keeping
// cruft we don't need.
(*acc)[ai] = llcpp::fuchsia::sysmem2::ImageFormatConstraints::Builder(nullptr);
}
// remove last item
acc->set_count(acc->count() - 1);
// adjust ai to force current index to be processed again as it's
// now a different item
--ai;
}
}
if (!acc->count()) {
// It's ok for this check to be under Accumulate* because it's permitted
// for a given participant to have zero image_format_constraints_count.
// It's only when the count becomes non-zero then drops back to zero
// (checked here), or if we end up with no image format constraints and
// no buffer constraints (checked in ::Allocate()), that we care.
LogError("all pixel_format(s) eliminated");
return false;
}
return true;
}
bool LogicalBufferCollection::AccumulateConstraintImageFormat(
llcpp::fuchsia::sysmem2::ImageFormatConstraints::Builder* acc,
llcpp::fuchsia::sysmem2::ImageFormatConstraints* c) {
ZX_DEBUG_ASSERT(ImageFormatIsPixelFormatEqual(acc->pixel_format(), c->pixel_format()));
// Checked previously.
ZX_DEBUG_ASSERT(acc->color_spaces().count());
// Checked previously.
ZX_DEBUG_ASSERT(c->color_spaces().count());
if (!AccumulateConstraintColorSpaces(&acc->get_builders_color_spaces(), &c->color_spaces())) {
return false;
}
// Else AccumulateConstraintColorSpaces() would have returned false.
ZX_DEBUG_ASSERT(acc->color_spaces().count());
acc->min_coded_width() = std::max(acc->min_coded_width(), c->min_coded_width());
acc->max_coded_width() = std::min(acc->max_coded_width(), c->max_coded_width());
acc->min_coded_height() = std::max(acc->min_coded_height(), c->min_coded_height());
acc->max_coded_height() = std::min(acc->max_coded_height(), c->max_coded_height());
acc->min_bytes_per_row() = std::max(acc->min_bytes_per_row(), c->min_bytes_per_row());
acc->max_bytes_per_row() = std::min(acc->max_bytes_per_row(), c->max_bytes_per_row());
acc->max_coded_width_times_coded_height() =
std::min(acc->max_coded_width_times_coded_height(), c->max_coded_width_times_coded_height());
acc->coded_width_divisor() = std::max(acc->coded_width_divisor(), c->coded_width_divisor());
acc->coded_width_divisor() =
std::max(acc->coded_width_divisor(), ImageFormatCodedWidthMinDivisor(acc->pixel_format()));
acc->coded_height_divisor() = std::max(acc->coded_height_divisor(), c->coded_height_divisor());
acc->coded_height_divisor() =
std::max(acc->coded_height_divisor(), ImageFormatCodedHeightMinDivisor(acc->pixel_format()));
acc->bytes_per_row_divisor() = std::max(acc->bytes_per_row_divisor(), c->bytes_per_row_divisor());
acc->bytes_per_row_divisor() =
std::max(acc->bytes_per_row_divisor(), ImageFormatSampleAlignment(acc->pixel_format()));
acc->start_offset_divisor() = std::max(acc->start_offset_divisor(), c->start_offset_divisor());
acc->start_offset_divisor() =
std::max(acc->start_offset_divisor(), ImageFormatSampleAlignment(acc->pixel_format()));
acc->display_width_divisor() = std::max(acc->display_width_divisor(), c->display_width_divisor());
acc->display_height_divisor() =
std::max(acc->display_height_divisor(), c->display_height_divisor());
// The required_ space is accumulated by taking the union, and must be fully
// within the non-required_ space, else fail. For example, this allows a
// video decoder to indicate that it's capable of outputting a wide range of
// output dimensions, but that it has specific current dimensions that are
// presently required_ (min == max) for decode to proceed.
ZX_DEBUG_ASSERT(acc->required_min_coded_width() != 0);
ZX_DEBUG_ASSERT(c->required_min_coded_width() != 0);
acc->required_min_coded_width() =
std::min(acc->required_min_coded_width(), c->required_min_coded_width());
acc->required_max_coded_width() =
std::max(acc->required_max_coded_width(), c->required_max_coded_width());
ZX_DEBUG_ASSERT(acc->required_min_coded_height() != 0);
ZX_DEBUG_ASSERT(c->required_min_coded_height() != 0);
acc->required_min_coded_height() =
std::min(acc->required_min_coded_height(), c->required_min_coded_height());
acc->required_max_coded_height() =
std::max(acc->required_max_coded_height(), c->required_max_coded_height());
ZX_DEBUG_ASSERT(acc->required_min_bytes_per_row() != 0);
ZX_DEBUG_ASSERT(c->required_min_bytes_per_row() != 0);
acc->required_min_bytes_per_row() =
std::min(acc->required_min_bytes_per_row(), c->required_min_bytes_per_row());
acc->required_max_bytes_per_row() =
std::max(acc->required_max_bytes_per_row(), c->required_max_bytes_per_row());
return true;
}
bool LogicalBufferCollection::AccumulateConstraintColorSpaces(
fidl::VectorView<llcpp::fuchsia::sysmem2::ColorSpace::Builder>* acc,
fidl::VectorView<llcpp::fuchsia::sysmem2::ColorSpace>* c) {
// Remove any color_space in acc that's not in c. If zero color spaces
// remain in acc, return false.
for (uint32_t ai = 0; ai < acc->count(); ++ai) {
uint32_t ci;
for (ci = 0; ci < c->count(); ++ci) {
if (IsColorSpaceEqual((*acc)[ai], (*c)[ci])) {
// We found the color space in c. Break so we can move on to
// the next color space.
break;
}
}
if (ci == c->count()) {
// Remove from acc because not found in c.
//
// Move formerly last item on top of the item being removed, if not same item.
if (ai != acc->count() - 1) {
(*acc)[ai] = std::move((*acc)[acc->count() - 1]);
} else {
// Stuff under this item would get deleted later anyway, but delete now to avoid keeping
// cruft we don't need.
(*acc)[ai] = llcpp::fuchsia::sysmem2::ColorSpace::Builder(nullptr);
}
// remove last item
acc->set_count(acc->count() - 1);
// adjust ai to force current index to be processed again as it's
// now a different item
--ai;
}
}
if (!acc->count()) {
// It's ok for this check to be under Accumulate* because it's also
// under CheckSanitize(). It's fine to provide a slightly more helpful
// error message here and early out here.
LogError("Zero color_space overlap");
return false;
}
return true;
}
bool LogicalBufferCollection::IsColorSpaceEqual(
const llcpp::fuchsia::sysmem2::ColorSpace::Builder& a,
const llcpp::fuchsia::sysmem2::ColorSpace& b) {
return a.type() == b.type();
}
static fit::result<llcpp::fuchsia::sysmem2::HeapType, zx_status_t> GetHeap(
const llcpp::fuchsia::sysmem2::BufferMemoryConstraints::Builder& constraints, Device* device) {
if (constraints.secure_required()) {
// TODO(fxbug.dev/37452): Generalize this.
//
// checked previously
ZX_DEBUG_ASSERT(!constraints.secure_required() || IsSecurePermitted(constraints));
if (IsHeapPermitted(constraints, llcpp::fuchsia::sysmem2::HeapType::AMLOGIC_SECURE)) {
return fit::ok(llcpp::fuchsia::sysmem2::HeapType::AMLOGIC_SECURE);
} else {
ZX_DEBUG_ASSERT(
IsHeapPermitted(constraints, llcpp::fuchsia::sysmem2::HeapType::AMLOGIC_SECURE_VDEC));
return fit::ok(llcpp::fuchsia::sysmem2::HeapType::AMLOGIC_SECURE_VDEC);
}
}
if (IsHeapPermitted(constraints, llcpp::fuchsia::sysmem2::HeapType::SYSTEM_RAM)) {
return fit::ok(llcpp::fuchsia::sysmem2::HeapType::SYSTEM_RAM);
}
for (size_t i = 0; i < constraints.heap_permitted().count(); ++i) {
auto heap = constraints.heap_permitted()[i];
const auto& heap_properties = device->GetHeapProperties(heap);
if (heap_properties.has_coherency_domain_support() &&
((heap_properties.coherency_domain_support().cpu_supported() &&
constraints.cpu_domain_supported()) ||
(heap_properties.coherency_domain_support().ram_supported() &&
constraints.ram_domain_supported()) ||
(heap_properties.coherency_domain_support().inaccessible_supported() &&
constraints.inaccessible_domain_supported()))) {
return fit::ok(heap);
}
}
return fit::error(ZX_ERR_NOT_FOUND);
}
static fit::result<llcpp::fuchsia::sysmem2::CoherencyDomain> GetCoherencyDomain(
const llcpp::fuchsia::sysmem2::BufferCollectionConstraints::Builder& constraints,
MemoryAllocator* memory_allocator) {
ZX_DEBUG_ASSERT(constraints.has_buffer_memory_constraints());
using llcpp::fuchsia::sysmem2::CoherencyDomain;
const auto& heap_properties = memory_allocator->heap_properties();
ZX_DEBUG_ASSERT(heap_properties.has_coherency_domain_support());
// Display prefers RAM coherency domain for now.
if (constraints.usage().display() != 0) {
if (constraints.buffer_memory_constraints().ram_domain_supported()) {
// Display controllers generally aren't cache coherent, so prefer
// RAM coherency domain.
//
// TODO - base on the system in use.
return fit::ok(llcpp::fuchsia::sysmem2::CoherencyDomain::RAM);
}
}
if (heap_properties.coherency_domain_support().cpu_supported() &&
constraints.buffer_memory_constraints().cpu_domain_supported()) {
return fit::ok(CoherencyDomain::CPU);
}
if (heap_properties.coherency_domain_support().ram_supported() &&
constraints.buffer_memory_constraints().ram_domain_supported()) {
return fit::ok(CoherencyDomain::RAM);
}
if (heap_properties.coherency_domain_support().inaccessible_supported() &&
constraints.buffer_memory_constraints().inaccessible_domain_supported()) {
// Intentionally permit treating as Inaccessible if we reach here, even
// if the heap permits CPU access. Only domain in common among
// participants is Inaccessible.
return fit::ok(llcpp::fuchsia::sysmem2::CoherencyDomain::INACCESSIBLE);
}
return fit::error();
}
fit::result<llcpp::fuchsia::sysmem2::BufferCollectionInfo, zx_status_t>
LogicalBufferCollection::Allocate() {
TRACE_DURATION("gfx", "LogicalBufferCollection:Allocate", "this", this);
ZX_DEBUG_ASSERT(constraints_);
llcpp::fuchsia::sysmem2::BufferCollectionInfo::Builder result =
allocator_.make_table_builder<llcpp::fuchsia::sysmem2::BufferCollectionInfo>();
uint32_t min_buffer_count = constraints_->min_buffer_count_for_camping() +
constraints_->min_buffer_count_for_dedicated_slack() +
constraints_->min_buffer_count_for_shared_slack();
min_buffer_count = std::max(min_buffer_count, constraints_->min_buffer_count());
uint32_t max_buffer_count = constraints_->max_buffer_count();
if (min_buffer_count > max_buffer_count) {
LogError(
"aggregate min_buffer_count > aggregate max_buffer_count - "
"min: %u max: %u",
min_buffer_count, max_buffer_count);
return fit::error(ZX_ERR_NOT_SUPPORTED);
}
if (min_buffer_count > llcpp::fuchsia::sysmem::MAX_COUNT_BUFFER_COLLECTION_INFO_BUFFERS) {
LogError("aggregate min_buffer_count (%d) > MAX_COUNT_BUFFER_COLLECTION_INFO_BUFFERS (%d)",
min_buffer_count, llcpp::fuchsia::sysmem::MAX_COUNT_BUFFER_COLLECTION_INFO_BUFFERS);
return fit::error(ZX_ERR_NOT_SUPPORTED);
}
result.set_buffers(allocator_.make_vec_ptr<llcpp::fuchsia::sysmem2::VmoBuffer>(min_buffer_count));
ZX_DEBUG_ASSERT(result.buffers().count() == min_buffer_count);
ZX_DEBUG_ASSERT(result.buffers().count() <= max_buffer_count);
uint64_t min_size_bytes = 0;
uint64_t max_size_bytes = std::numeric_limits<uint64_t>::max();
result.set_settings(
sysmem::MakeTracking<llcpp::fuchsia::sysmem2::SingleBufferSettings>(&allocator_));
llcpp::fuchsia::sysmem2::SingleBufferSettings::Builder& settings = result.get_builder_settings();
settings.set_buffer_settings(
sysmem::MakeTracking<llcpp::fuchsia::sysmem2::BufferMemorySettings>(&allocator_));
llcpp::fuchsia::sysmem2::BufferMemorySettings::Builder& buffer_settings =
settings.get_builder_buffer_settings();
ZX_DEBUG_ASSERT(constraints_->has_buffer_memory_constraints());
const llcpp::fuchsia::sysmem2::BufferMemoryConstraints::Builder& buffer_constraints =
constraints_->get_builder_buffer_memory_constraints();
buffer_settings.set_is_physically_contiguous(
sysmem::MakeTracking(&allocator_, buffer_constraints.physically_contiguous_required()));
// checked previously
ZX_DEBUG_ASSERT(IsSecurePermitted(buffer_constraints) || !buffer_constraints.secure_required());
buffer_settings.set_is_secure(
sysmem::MakeTracking(&allocator_, buffer_constraints.secure_required()));
if (buffer_settings.is_secure()) {
if (constraints_->need_clear_aux_buffers_for_secure() &&
!constraints_->allow_clear_aux_buffers_for_secure()) {
LogError(
"is_secure && need_clear_aux_buffers_for_secure && !allow_clear_aux_buffers_for_secure");
return fit::error(ZX_ERR_NOT_SUPPORTED);
}
}
auto result_get_heap = GetHeap(buffer_constraints, parent_device_);
if (!result_get_heap.is_ok()) {
LogError("Can not find a heap permitted by buffer constraints, error %d",
result_get_heap.error());
return fit::error(result_get_heap.error());
}
buffer_settings.set_heap(sysmem::MakeTracking(&allocator_, result_get_heap.value()));
// We can't fill out buffer_settings yet because that also depends on
// ImageFormatConstraints. We do need the min and max from here though.
min_size_bytes = buffer_constraints.min_size_bytes();
max_size_bytes = buffer_constraints.max_size_bytes();
// Get memory allocator for settings.
MemoryAllocator* allocator = parent_device_->GetAllocator(buffer_settings);
if (!allocator) {
LogError("No memory allocator for buffer settings");
return fit::error(ZX_ERR_NO_MEMORY);
}
auto coherency_domain_result = GetCoherencyDomain(*constraints_, allocator);
if (!coherency_domain_result.is_ok()) {
LogError("No coherency domain found for buffer constraints");
return fit::error(ZX_ERR_NOT_SUPPORTED);
}
buffer_settings.set_coherency_domain(
sysmem::MakeTracking(&allocator_, coherency_domain_result.value()));
// It's allowed for zero participants to have any ImageFormatConstraint(s),
// in which case the combined constraints_ will have zero (and that's fine,
// when allocating raw buffers that don't need any ImageFormatConstraint).
//
// At least for now, we pick which PixelFormat to use before determining if
// the constraints associated with that PixelFormat imply a buffer size
// range in min_size_bytes..max_size_bytes.
if (constraints_->image_format_constraints().count()) {
// Pick the best ImageFormatConstraints.
uint32_t best_index = UINT32_MAX;
bool found_unsupported_when_protected = false;
for (uint32_t i = 0; i < constraints_->image_format_constraints().count(); ++i) {
if (buffer_settings.is_secure() &&
!ImageFormatCompatibleWithProtectedMemory(
constraints_->image_format_constraints()[i].pixel_format())) {
found_unsupported_when_protected = true;
continue;
}
if (best_index == UINT32_MAX) {
best_index = i;
} else {
if (CompareImageFormatConstraintsByIndex(i, best_index) < 0) {
best_index = i;
}
}
}
if (best_index == UINT32_MAX) {
ZX_DEBUG_ASSERT(found_unsupported_when_protected);
LogError("No formats were compatible with protected memory.");
return fit::error(ZX_ERR_NOT_SUPPORTED);
}
// move from constraints_ to settings.
settings.set_image_format_constraints(sysmem::MakeTracking(
&allocator_, std::move(constraints_->image_format_constraints()[best_index])));
}
// Compute the min buffer size implied by image_format_constraints, so we ensure the buffers can
// hold the min-size image.
if (settings.has_image_format_constraints()) {
const llcpp::fuchsia::sysmem2::ImageFormatConstraints& image_format_constraints =
settings.image_format_constraints();
llcpp::fuchsia::sysmem2::ImageFormat::Builder min_image =
allocator_.make_table_builder<llcpp::fuchsia::sysmem2::ImageFormat>();
min_image.set_pixel_format(sysmem::MakeTracking(
&allocator_,
sysmem::V2ClonePixelFormat(&allocator_, image_format_constraints.pixel_format()).build()));
// We use required_max_coded_width because that's the max width that the producer (or
// initiator) wants these buffers to be able to hold.
min_image.set_coded_width(sysmem::MakeTracking(
&allocator_, AlignUp(std::max(image_format_constraints.min_coded_width(),
image_format_constraints.required_max_coded_width()),
image_format_constraints.coded_width_divisor())));
if (min_image.coded_width() > image_format_constraints.max_coded_width()) {
LogError("coded_width_divisor caused coded_width > max_coded_width");
return fit::error(ZX_ERR_NOT_SUPPORTED);
}
// We use required_max_coded_height because that's the max height that the producer (or
// initiator) wants these buffers to be able to hold.
min_image.set_coded_height(sysmem::MakeTracking(
&allocator_, AlignUp(std::max(image_format_constraints.min_coded_height(),
image_format_constraints.required_max_coded_height()),
image_format_constraints.coded_height_divisor())));
if (min_image.coded_height() > image_format_constraints.max_coded_height()) {
LogError("coded_height_divisor caused coded_height > max_coded_height");
return fit::error(ZX_ERR_NOT_SUPPORTED);
}
min_image.set_bytes_per_row(sysmem::MakeTracking(
&allocator_, AlignUp(std::max(image_format_constraints.min_bytes_per_row(),
ImageFormatStrideBytesPerWidthPixel(
image_format_constraints.pixel_format()) *
min_image.coded_width()),
image_format_constraints.bytes_per_row_divisor())));
if (min_image.bytes_per_row() > image_format_constraints.max_bytes_per_row()) {
LogError(
"bytes_per_row_divisor caused bytes_per_row > "
"max_bytes_per_row");
return fit::error(ZX_ERR_NOT_SUPPORTED);
}
if (min_image.coded_width() * min_image.coded_height() >
image_format_constraints.max_coded_width_times_coded_height()) {
LogError(
"coded_width * coded_height > "
"max_coded_width_times_coded_height");
return fit::error(ZX_ERR_NOT_SUPPORTED);
}
// These don't matter for computing size in bytes.
ZX_DEBUG_ASSERT(!min_image.has_display_width());
ZX_DEBUG_ASSERT(!min_image.has_display_height());
// Checked previously.
ZX_DEBUG_ASSERT(image_format_constraints.color_spaces().count() >= 1);
// This doesn't matter for computing size in bytes, as we trust the pixel_format to fully
// specify the image size. But set it to the first ColorSpace anyway, just so the
// color_space.type is a valid value.
min_image.set_color_space(sysmem::MakeTracking(
&allocator_,
sysmem::V2CloneColorSpace(&allocator_, image_format_constraints.color_spaces()[0])
.build()));
uint64_t image_min_size_bytes = ImageFormatImageSize(min_image.build());
if (image_min_size_bytes > min_size_bytes) {
if (image_min_size_bytes > max_size_bytes) {
LogError("image_min_size_bytes > max_size_bytes");
return fit::error(ZX_ERR_NOT_SUPPORTED);
}
min_size_bytes = image_min_size_bytes;
ZX_DEBUG_ASSERT(min_size_bytes <= max_size_bytes);
}
}
// Currently redundant with earlier checks, but just in case...
if (min_size_bytes == 0) {
LogError("min_size_bytes == 0");
return fit::error(ZX_ERR_NOT_SUPPORTED);
}
ZX_DEBUG_ASSERT(min_size_bytes != 0);
// For purposes of enforcing max_size_bytes, we intentionally don't care that a VMO can only be a
// multiple of page size.
uint64_t total_size_bytes = min_size_bytes * result.buffers().count();
if (total_size_bytes > kMaxTotalSizeBytesPerCollection) {
LogError("total_size_bytes > kMaxTotalSizeBytesPerCollection");
return fit::error(ZX_ERR_NO_MEMORY);
}
if (min_size_bytes > kMaxSizeBytesPerBuffer) {
LogError("min_size_bytes > kMaxSizeBytesPerBuffer");
return fit::error(ZX_ERR_NO_MEMORY);
}
ZX_DEBUG_ASSERT(min_size_bytes <= std::numeric_limits<uint32_t>::max());
if (settings.has_image_format_constraints()) {
const llcpp::fuchsia::sysmem2::ImageFormatConstraints& image_format_constraints =
settings.image_format_constraints();
node_.CreateUint("pixel_format",
static_cast<uint64_t>(image_format_constraints.pixel_format().type()),
&vmo_properties_);
if (image_format_constraints.pixel_format().has_format_modifier_value()) {
node_.CreateUint("pixel_format_modifier",
image_format_constraints.pixel_format().format_modifier_value(),
&vmo_properties_);
}
if (image_format_constraints.min_coded_width() > 0) {
node_.CreateUint("min_coded_width", image_format_constraints.min_coded_width(),
&vmo_properties_);
}
if (image_format_constraints.min_coded_height() > 0) {
node_.CreateUint("min_coded_height", image_format_constraints.min_coded_height(),
&vmo_properties_);
}
if (image_format_constraints.required_max_coded_width() > 0) {
node_.CreateUint("required_max_coded_width",
image_format_constraints.required_max_coded_width(), &vmo_properties_);
}
if (image_format_constraints.required_max_coded_height() > 0) {
node_.CreateUint("required_max_coded_height",
image_format_constraints.required_max_coded_height(), &vmo_properties_);
}
}
// Now that min_size_bytes accounts for any ImageFormatConstraints, we can just allocate
// min_size_bytes buffers.
//
// If an initiator (or a participant) wants to force buffers to be larger than the size implied by
// minimum image dimensions, the initiator can use BufferMemorySettings.min_size_bytes to force
// allocated buffers to be large enough.
buffer_settings.set_size_bytes(
sysmem::MakeTracking(&allocator_, static_cast<uint32_t>(min_size_bytes)));
// Get memory allocator for aux buffers, if needed.
MemoryAllocator* maybe_aux_allocator = nullptr;
std::optional<llcpp::fuchsia::sysmem2::SingleBufferSettings::Builder> maybe_aux_settings;
if (buffer_settings.is_secure() && constraints_->need_clear_aux_buffers_for_secure()) {
maybe_aux_settings.emplace(
allocator_.make_table_builder<llcpp::fuchsia::sysmem2::SingleBufferSettings>());
maybe_aux_settings->set_buffer_settings(sysmem::MakeTracking(
&allocator_,
allocator_.make_table_builder<llcpp::fuchsia::sysmem2::BufferMemorySettings>().build()));
auto& aux_buffer_settings = maybe_aux_settings->get_builder_buffer_settings();
aux_buffer_settings.set_size_bytes(
sysmem::MakeTracking(&allocator_, buffer_settings.size_bytes()));
aux_buffer_settings.set_is_physically_contiguous(sysmem::MakeTracking(&allocator_, false));
aux_buffer_settings.set_is_secure(sysmem::MakeTracking(&allocator_, false));
aux_buffer_settings.set_coherency_domain(
sysmem::MakeTracking(&allocator_, llcpp::fuchsia::sysmem2::CoherencyDomain::CPU));
aux_buffer_settings.set_heap(
sysmem::MakeTracking(&allocator_, llcpp::fuchsia::sysmem2::HeapType::SYSTEM_RAM));
maybe_aux_allocator = parent_device_->GetAllocator(aux_buffer_settings);
ZX_DEBUG_ASSERT(maybe_aux_allocator);
}
if (buffer_settings.size_bytes() > parent_device_->settings().max_allocation_size) {
// This is different than max_size_bytes. While max_size_bytes is part of the constraints,
// max_allocation_size isn't part of the constraints. The latter is used for simulating OOM or
// preventing unpredictable memory pressure caused by a fuzzer or similar source of
// unpredictability in tests.
LogError("AllocateVmo() failed because size %u > max_allocation_size %ld",
buffer_settings.size_bytes(), parent_device_->settings().max_allocation_size);
return fit::error(ZX_ERR_NO_MEMORY);
}
for (uint32_t i = 0; i < result.buffers().count(); ++i) {
auto allocate_result = AllocateVmo(allocator, settings, i);
if (!allocate_result.is_ok()) {
LogError("AllocateVmo() failed");
return fit::error(ZX_ERR_NO_MEMORY);
}
zx::vmo vmo = allocate_result.take_value();
auto vmo_buffer = allocator_.make_table_builder<llcpp::fuchsia::sysmem2::VmoBuffer>();
vmo_buffer.set_vmo(sysmem::MakeTracking(&allocator_, std::move(vmo)));
vmo_buffer.set_vmo_usable_start(sysmem::MakeTracking(&allocator_, 0ul));
if (maybe_aux_allocator) {
ZX_DEBUG_ASSERT(maybe_aux_settings);
auto aux_allocate_result = AllocateVmo(maybe_aux_allocator, maybe_aux_settings.value(), i);
if (!aux_allocate_result.is_ok()) {
LogError("AllocateVmo() failed (aux)");
return fit::error(ZX_ERR_NO_MEMORY);
}
zx::vmo aux_vmo = aux_allocate_result.take_value();
vmo_buffer.set_aux_vmo(sysmem::MakeTracking(&allocator_, std::move(aux_vmo)));
}
result.buffers()[i] = vmo_buffer.build();
}
vmo_count_property_ = node_.CreateUint("vmo_count", result.buffers().count());
// Make sure we have sufficient barrier after allocating/clearing/flushing any VMO newly allocated
// by allocator above.
BarrierAfterFlush();
// Register failure handler with memory allocator.
allocator->AddDestroyCallback(reinterpret_cast<intptr_t>(this), [this]() {
Fail("LogicalBufferCollection memory allocator gone - now auto-failing self.");
});
memory_allocator_ = allocator;
return fit::ok(result.build());
}
fit::result<zx::vmo> LogicalBufferCollection::AllocateVmo(
MemoryAllocator* allocator,
const llcpp::fuchsia::sysmem2::SingleBufferSettings::Builder& settings, uint32_t index) {
TRACE_DURATION("gfx", "LogicalBufferCollection::AllocateVmo", "size_bytes",
settings.buffer_settings().size_bytes());
zx::vmo child_vmo;
// Physical VMOs only support slices where the size (and offset) are page_size aligned,
// so we should also round up when allocating.
size_t rounded_size_bytes = fbl::round_up(settings.buffer_settings().size_bytes(), ZX_PAGE_SIZE);
if (rounded_size_bytes < settings.buffer_settings().size_bytes()) {
LogError("size_bytes overflows when rounding to multiple of page_size");
return fit::error();
}
// raw_vmo may itself be a child VMO of an allocator's overall contig VMO,
// but that's an internal detail of the allocator. The ZERO_CHILDREN signal
// will only be set when all direct _and indirect_ child VMOs are fully
// gone (not just handles closed, but the kernel object is deleted, which
// avoids races with handle close, and means there also aren't any
// mappings left).
zx::vmo raw_parent_vmo;
std::optional<std::string> name;
if (name_) {
name = fbl::StringPrintf("%s:%d", name_->name.c_str(), index).c_str();
}
zx_status_t status = allocator->Allocate(rounded_size_bytes, name, &raw_parent_vmo);
if (status != ZX_OK) {
LogError(
"allocator->Allocate failed - size_bytes: %zu "
"status: %d",
rounded_size_bytes, status);
return fit::error();
}
zx_info_vmo_t info;
status = raw_parent_vmo.get_info(ZX_INFO_VMO, &info, sizeof(info), nullptr, nullptr);
if (status != ZX_OK) {
LogError("raw_parent_vmo.get_info(ZX_INFO_VMO) failed - status %d", status);
return fit::error();
}
auto node = node_.CreateChild(fbl::StringPrintf("vmo-%ld", info.koid).c_str());
node.CreateUint("koid", info.koid, &vmo_properties_);
vmo_properties_.emplace(std::move(node));
// Write zeroes to the VMO, so that the allocator doesn't need to. Also flush those zeroes to
// RAM so the newly-allocated VMO is fully zeroed in both RAM and CPU coherency domains.
//
// This is measured to be significantly more than half the overall time cost when repeatedly
// allocating and deallocating a buffer collection with 4MiB buffer space per collection. On
// astro this was measured to be ~2100us out of ~2550us per-cycle duration. Larger buffer space
// per collection would take longer here.
//
// If we find this is taking too long, we could ask the allocator if it's already providing
// pre-zeroed VMOs. And/or zero allocator backing space async during deallocation, but wait on
// deallocations to be done before failing an new allocation.
//
// TODO(fxbug.dev/34590): Zero secure/protected VMOs.
const auto& heap_properties = allocator->heap_properties();
ZX_DEBUG_ASSERT(heap_properties.has_coherency_domain_support());
ZX_DEBUG_ASSERT(heap_properties.has_need_clear());
if (heap_properties.need_clear()) {
uint64_t offset = 0;
while (offset < info.size_bytes) {
uint64_t bytes_to_write = std::min(sizeof(kZeroes), info.size_bytes - offset);
// TODO(fxbug.dev/59796): Use ZX_VMO_OP_ZERO instead.
status = raw_parent_vmo.write(kZeroes, offset, bytes_to_write);
if (status != ZX_OK) {
LogError("raw_parent_vmo.write() failed - status: %d", status);
return fit::error();
}
offset += bytes_to_write;
}
// Flush out the zeroes.
status = raw_parent_vmo.op_range(ZX_VMO_OP_CACHE_CLEAN, 0, info.size_bytes, nullptr, 0);
if (status != ZX_OK) {
LogError("raw_parent_vmo.op_range(ZX_VMO_OP_CACHE_CLEAN) failed - status: %d", status);
return fit::error();
}
}
// We immediately create the ParentVmo instance so it can take care of calling allocator->Delete()
// if this method returns early. We intentionally don't emplace into parent_vmos_ until
// StartWait() has succeeded. In turn, StartWait() requires a child VMO to have been created
// already (else ZX_VMO_ZERO_CHILDREN would trigger too soon).
//
// We need to keep the raw_parent_vmo around so we can wait for ZX_VMO_ZERO_CHILDREN, and so we
// can call allocator->Delete(raw_parent_vmo).
//
// Until that happens, we can't let LogicalBufferCollection itself go away, because it needs to
// stick around to tell allocator that the allocator's VMO can be deleted/reclaimed.
//
// We let cooked_parent_vmo go away before returning from this method, since it's only purpose
// was to attenuate the rights of local_child_vmo. The local_child_vmo counts as a child of
// raw_parent_vmo for ZX_VMO_ZERO_CHILDREN.
//
// The fbl::RefPtr(this) is fairly similar (in this usage) to shared_from_this().
auto tracked_parent_vmo = std::unique_ptr<TrackedParentVmo>(new TrackedParentVmo(
fbl::RefPtr(this), std::move(raw_parent_vmo),
[this, allocator](TrackedParentVmo* tracked_parent_vmo) mutable {
auto node_handle = parent_vmos_.extract(tracked_parent_vmo->vmo().get());
ZX_DEBUG_ASSERT(!node_handle || node_handle.mapped().get() == tracked_parent_vmo);
allocator->Delete(tracked_parent_vmo->TakeVmo());
// ~node_handle may delete "this".
}));
zx::vmo cooked_parent_vmo;
status = tracked_parent_vmo->vmo().duplicate(kSysmemVmoRights, &cooked_parent_vmo);
if (status != ZX_OK) {
LogError("zx::object::duplicate() failed - status: %d", status);
return fit::error();
}
zx::vmo local_child_vmo;
status =
cooked_parent_vmo.create_child(ZX_VMO_CHILD_SLICE, 0, rounded_size_bytes, &local_child_vmo);
if (status != ZX_OK) {
LogError("zx::vmo::create_child() failed - status: %d", status);
return fit::error();
}
zx_info_handle_basic_t child_info{};
local_child_vmo.get_info(ZX_INFO_HANDLE_BASIC, &child_info, sizeof(child_info), nullptr, nullptr);
tracked_parent_vmo->set_child_koid(child_info.koid);
TRACE_INSTANT("gfx", "Child VMO created", TRACE_SCOPE_THREAD, "koid", child_info.koid);
// Now that we know at least one child of raw_parent_vmo exists, we can StartWait() and add to
// map. From this point, ZX_VMO_ZERO_CHILDREN is the only way that allocator->Delete() gets
// called.
status = tracked_parent_vmo->StartWait(parent_device_->dispatcher());
if (status != ZX_OK) {
LogError("tracked_parent->StartWait() failed - status: %d", status);
// ~tracked_parent_vmo calls allocator->Delete().
return fit::error();
}
zx_handle_t raw_parent_vmo_handle = tracked_parent_vmo->vmo().get();
TrackedParentVmo& parent_vmo_ref = *tracked_parent_vmo;
auto emplace_result = parent_vmos_.emplace(raw_parent_vmo_handle, std::move(tracked_parent_vmo));
ZX_DEBUG_ASSERT(emplace_result.second);
// Now inform the allocator about the child VMO before we return it.
status = allocator->SetupChildVmo(
parent_vmo_ref.vmo(), local_child_vmo,
sysmem::V2CloneSingleBufferSettingsBuilder(&allocator_, settings).build());
if (status != ZX_OK) {
LogError("allocator->SetupChildVmo() failed - status: %d", status);
// In this path, the ~local_child_vmo will async trigger parent_vmo_ref::OnZeroChildren()
// which will call allocator->Delete() via above do_delete lambda passed to
// ParentVmo::ParentVmo().
return fit::error();
}
if (name) {
local_child_vmo.set_property(ZX_PROP_NAME, name->c_str(), name->size());
}
// ~cooked_parent_vmo is fine, since local_child_vmo counts as a child of raw_parent_vmo for
// ZX_VMO_ZERO_CHILDREN purposes.
return fit::ok(std::move(local_child_vmo));
}
void LogicalBufferCollection::CreationTimedOut(async_dispatcher_t* dispatcher,
async::TaskBase* task, zx_status_t status) {
if (status != ZX_OK)
return;
std::string name = name_ ? name_->name : "Unknown";
LogError("Allocation of %s timed out. Waiting for tokens: ", name.c_str());
for (auto& token : token_views_) {
if (token.second->debug_name() != "") {
LogError("Name %s id %ld", token.second->debug_name().c_str(), token.second->debug_id());
} else {
LogError("Unknown token");
}
}
LogError("Collections:");
for (auto& collection : collection_views_) {
const char* constraints_state = collection.second->has_constraints() ? "set" : "unset";
if (collection.second->debug_name() != "") {
LogError("Name \"%s\" id %ld (constraints %s)", collection.second->debug_name().c_str(),
collection.second->debug_id(), constraints_state);
} else {
LogError("Name unknown (constraints %s)", constraints_state);
}
}
}
static int32_t clamp_difference(int32_t a, int32_t b) {
int32_t raw_result = a - b;
int32_t cooked_result = raw_result;
if (cooked_result > 0) {
cooked_result = 1;
} else if (cooked_result < 0) {
cooked_result = -1;
}
ZX_DEBUG_ASSERT(cooked_result == 0 || cooked_result == 1 || cooked_result == -1);
return cooked_result;
}
// 1 means a > b, 0 means ==, -1 means a < b.
//
// TODO(dustingreen): Pay attention to constraints_->usage, by checking any
// overrides that prefer particular PixelFormat based on a usage / usage
// combination.
int32_t LogicalBufferCollection::CompareImageFormatConstraintsTieBreaker(
const llcpp::fuchsia::sysmem2::ImageFormatConstraints& a,
const llcpp::fuchsia::sysmem2::ImageFormatConstraints& b) {
// If there's not any cost difference, fall back to choosing the
// pixel_format that has the larger type enum value as a tie-breaker.
int32_t result = clamp_difference(static_cast<int32_t>(a.pixel_format().type()),
static_cast<int32_t>(b.pixel_format().type()));
if (result != 0)
return result;
result = clamp_difference(static_cast<int32_t>(a.pixel_format().has_format_modifier_value()),
static_cast<int32_t>(b.pixel_format().has_format_modifier_value()));
if (result != 0)
return result;
if (a.pixel_format().has_format_modifier_value() &&
b.pixel_format().has_format_modifier_value()) {
result = clamp_difference(static_cast<int32_t>(a.pixel_format().format_modifier_value()),
static_cast<int32_t>(b.pixel_format().format_modifier_value()));
}
return result;
}
int32_t LogicalBufferCollection::CompareImageFormatConstraintsByIndex(uint32_t index_a,
uint32_t index_b) {
// This method is allowed to look at constraints_.
ZX_DEBUG_ASSERT(constraints_);
int32_t cost_compare = UsagePixelFormatCost::Compare(parent_device_->pdev_device_info_vid(),
parent_device_->pdev_device_info_pid(),
*constraints_, index_a, index_b);
if (cost_compare != 0) {
return cost_compare;
}
// If we get this far, there's no known reason to choose one PixelFormat
// over another, so just pick one based on a tie-breaker that'll distinguish
// between PixelFormat(s).
int32_t tie_breaker_compare =
CompareImageFormatConstraintsTieBreaker(constraints_->image_format_constraints()[index_a],
constraints_->image_format_constraints()[index_b]);
return tie_breaker_compare;
}
LogicalBufferCollection::TrackedParentVmo::TrackedParentVmo(
fbl::RefPtr<LogicalBufferCollection> buffer_collection, zx::vmo vmo,
LogicalBufferCollection::TrackedParentVmo::DoDelete do_delete)
: buffer_collection_(std::move(buffer_collection)),
vmo_(std::move(vmo)),
do_delete_(std::move(do_delete)),
zero_children_wait_(this, vmo_.get(), ZX_VMO_ZERO_CHILDREN) {
ZX_DEBUG_ASSERT(buffer_collection_);
ZX_DEBUG_ASSERT(vmo_);
ZX_DEBUG_ASSERT(do_delete_);
}
LogicalBufferCollection::TrackedParentVmo::~TrackedParentVmo() {
ZX_DEBUG_ASSERT(!waiting_);
if (do_delete_) {
do_delete_(this);
}
}
zx_status_t LogicalBufferCollection::TrackedParentVmo::StartWait(async_dispatcher_t* dispatcher) {
LogInfo("LogicalBufferCollection::TrackedParentVmo::StartWait()");
// The current thread is the dispatcher thread.
ZX_DEBUG_ASSERT(!waiting_);
zx_status_t status = zero_children_wait_.Begin(dispatcher);
if (status != ZX_OK) {
LogErrorStatic(nullptr, "zero_children_wait_.Begin() failed - status: %d", status);
return status;
}
waiting_ = true;
return ZX_OK;
}
zx_status_t LogicalBufferCollection::TrackedParentVmo::CancelWait() {
waiting_ = false;
return zero_children_wait_.Cancel();
}
zx::vmo LogicalBufferCollection::TrackedParentVmo::TakeVmo() {
ZX_DEBUG_ASSERT(!waiting_);
ZX_DEBUG_ASSERT(vmo_);
return std::move(vmo_);
}
const zx::vmo& LogicalBufferCollection::TrackedParentVmo::vmo() const {
ZX_DEBUG_ASSERT(vmo_);
return vmo_;
}
void LogicalBufferCollection::TrackedParentVmo::OnZeroChildren(async_dispatcher_t* dispatcher,
async::WaitBase* wait,
zx_status_t status,
const zx_packet_signal_t* signal) {
TRACE_DURATION("gfx", "LogicalBufferCollection::TrackedParentVmo::OnZeroChildren",
"buffer_collection", buffer_collection_.get(), "child_koid", child_koid_);
LogInfo("LogicalBufferCollection::TrackedParentVmo::OnZeroChildren()");
ZX_DEBUG_ASSERT(waiting_);
waiting_ = false;
if (status == ZX_ERR_CANCELED) {
// The collection canceled all of these waits as part of destruction, do nothing.
return;
}
ZX_DEBUG_ASSERT(status == ZX_OK);
ZX_DEBUG_ASSERT(signal->trigger & ZX_VMO_ZERO_CHILDREN);
ZX_DEBUG_ASSERT(do_delete_);
LogicalBufferCollection::TrackedParentVmo::DoDelete local_do_delete = std::move(do_delete_);
ZX_DEBUG_ASSERT(!do_delete_);
// will delete "this"
local_do_delete(this);
ZX_DEBUG_ASSERT(!local_do_delete);
}
} // namespace sysmem_driver