RGBCube/serenity
1
Fork 0
mirror of https://github.com/RGBCube/serenity synced 2025-07-26 01:17:34 +00:00
Code Issues Activity Actions

Kernel+AudioServer: Use interrupts for Intel HDA audio buffer completion

Browse source 
We used to not care about stopping an audio output stream for Intel HDA
since AudioServer would continuously send new buffers to play. Since
707f5ac150ef858760eb9faa52b9ba80c50c4262 however, that has changed.

Intel HDA now uses interrupts to detect when each buffer was completed
by the device, and uses a simple heuristic to detect whether a buffer
underrun has occurred so it can stop the output stream.

This was tested on Qemu's Intel HDA (Linux x86_64) and a bare metal MSI
Starship/Matisse HD Audio Controller.
This commit is contained in:
Jelle Raaijmakers 2023-06-27 16:07:05 +02:00
parent 2e474e8c18
commit 5c64686666
5 changed files with 120 additions and 25 deletions
Download patch file Download diff file Expand all files Collapse all files

56
Kernel/Devices/Audio/IntelHDA/Stream.cpp
View file

@ -73,7 +73,6 @@ ErrorOr<void> Stream::initialize_buffer()
m_stream_io_window->write32(StreamRegisterOffset::CyclicBufferLength, buffers->size());
// 3.3.39: Input/Output/Bidirectional Stream Descriptor Last Valid Index
VERIFY(number_of_buffers_required_for_cyclic_buffer_size <= 256);
m_stream_io_window->write16(StreamRegisterOffset::LastValidIndex, number_of_buffers_required_for_cyclic_buffer_size - 1);
// 3.6.2: Buffer Descriptor List
@ -83,12 +82,12 @@ ErrorOr<void> Stream::initialize_buffer()
m_stream_io_window->write32(StreamRegisterOffset::BDLUpperBaseAddress, bdl_physical_address >> 32);
// 3.6.3: Buffer Descriptor List Entry
auto* buffer_descriptor_entry = m_buffer_descriptor_list->vaddr().as_ptr();
for (u8 buffer_index = 0; buffer_index < buffers->page_count(); ++buffer_index) {
auto* entry = buffer_descriptor_entry + buffer_index * 0x10;
*bit_cast<u64*>(entry) = buffers->physical_page(buffer_index)->paddr().get();
*bit_cast<u32*>(entry + 8) = PAGE_SIZE;
*bit_cast<u32*>(entry + 12) = 0;
auto* buffer_descriptors = bit_cast<BufferDescriptorEntry*>(m_buffer_descriptor_list->vaddr().as_ptr());
for (size_t buffer_index = 0; buffer_index < buffers->page_count(); ++buffer_index) {
auto* entry = &buffer_descriptors[buffer_index];
entry->address = buffers->physical_page(buffer_index)->paddr().get();
entry->size = PAGE_SIZE;
entry->flags = BufferDescriptorEntryFlag::InterruptOnCompletion;
}
return {};
}));
@ -135,6 +134,7 @@ void Stream::start()
auto control = read_control();
control |= StreamControlFlag::StreamRun;
control |= StreamControlFlag::InterruptOnCompletionEnable;
write_control(control);
m_running = true;
}
@ -176,23 +176,45 @@ ErrorOr<void> Stream::set_format(FormatParameters format)
return {};
}
ErrorOr<void> OutputStream::handle_interrupt(Badge<Controller>)
{
auto interrupt_status = m_stream_io_window->read8(StreamRegisterOffset::Status);
if ((interrupt_status & StreamStatusFlag::BufferCompletionInterruptStatus) > 0) {
// 3.3.36: BCIS remains active until software clears it by writing a 1 to this bit position.
m_stream_io_window->write8(StreamRegisterOffset::Status, interrupt_status);
// Wake up thread waiting for new buffers to write to
m_irq_queue.wake_all();
// If the read head is past our last written buffer, stop the stream. There are three possible
// condition combinations of last & new link, and current buffer position for our circular buffer.
auto new_link_position = m_stream_io_window->read32(StreamRegisterOffset::LinkPosition);
if ((m_last_link_position < m_buffer_position && m_buffer_position < new_link_position)
|| (new_link_position < m_last_link_position && m_last_link_position < m_buffer_position)
|| (m_buffer_position < new_link_position && new_link_position < m_last_link_position)) {
dbgln_if(INTEL_HDA_DEBUG, "OutputStream::{}: Stopping because of stream underrun (link position: {} → {}, buffer position: {})",
__FUNCTION__, m_last_link_position, new_link_position, m_buffer_position);
TRY(stop());
}
m_last_link_position = new_link_position;
}
return {};
}
ErrorOr<size_t> OutputStream::write(UserOrKernelBuffer const& data, size_t length)
{
auto wait_until_buffer_index_can_be_written = [&](u8 buffer_index) {
while (m_running) {
auto link_position = m_stream_io_window->read32(StreamRegisterOffset::LinkPosition);
auto read_buffer_index = link_position / PAGE_SIZE;
m_last_link_position = m_stream_io_window->read32(StreamRegisterOffset::LinkPosition);
auto read_buffer_index = m_last_link_position / PAGE_SIZE;
if (read_buffer_index != buffer_index)
return;
auto microseconds_to_wait = ((read_buffer_index + 1) * PAGE_SIZE - link_position)
/ m_format_parameters.number_of_channels
* 8 / m_format_parameters.pcm_bits
* 1'000'000 / m_format_parameters.sample_rate;
dbgln_if(INTEL_HDA_DEBUG, "IntelHDA: Waiting {} µs until buffer {} becomes writeable", microseconds_to_wait, buffer_index);
dbgln_if(INTEL_HDA_DEBUG, "IntelHDA: Waiting until buffer {} becomes writeable", buffer_index);
// NOTE: we don't care about the reason for interruption - we simply calculate the next delay
[[maybe_unused]] auto block_result = Thread::current()->sleep(Duration::from_microseconds(microseconds_to_wait));
m_irq_queue.wait_forever("IntelHDA"sv);
}
};
@ -203,7 +225,7 @@ ErrorOr<size_t> OutputStream::write(UserOrKernelBuffer const& data, size_t lengt
wait_until_buffer_index_can_be_written(buffer_index);
TRY(m_buffers.with([&](auto& buffers) -> ErrorOr<void> {
// NOTE: if the buffers were reinitialized, we might point to an out of bounds page
// NOTE: if the buffers were reinitialized, we might point to an out-of-bounds page
if (buffer_index >= buffers->page_count())
return EAGAIN;