/*
This file is part of Ingen.
Copyright 2007-2016 David Robillard
Ingen is free software: you can redistribute it and/or modify it under the
terms of the GNU Affero General Public License as published by the Free
Software Foundation, either version 3 of the License, or any later version.
Ingen is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
A PARTICULAR PURPOSE. See the GNU Affero General Public License for details.
You should have received a copy of the GNU Affero General Public License
along with Ingen. If not, see .
*/
#include "ingen/LV2Features.hpp"
#include "ingen/Log.hpp"
#include "lv2/worker/worker.h"
#include "Engine.hpp"
#include "GraphImpl.hpp"
#include "LV2Block.hpp"
#include "Worker.hpp"
namespace ingen {
namespace server {
/// A message in the Worker::_requests ring
struct MessageHeader {
LV2Block* block; ///< Node this message is from
uint32_t size; ///< Size of following data
// `size' bytes of data follow here
};
static LV2_Worker_Status
schedule(LV2_Worker_Schedule_Handle handle,
uint32_t size,
const void* data)
{
LV2Block* block = (LV2Block*)handle;
Engine& engine = block->parent_graph()->engine();
return engine.worker()->request(block, size, data);
}
static LV2_Worker_Status
schedule_sync(LV2_Worker_Schedule_Handle handle,
uint32_t size,
const void* data)
{
LV2Block* block = (LV2Block*)handle;
Engine& engine = block->parent_graph()->engine();
return engine.sync_worker()->request(block, size, data);
}
LV2_Worker_Status
Worker::request(LV2Block* block,
uint32_t size,
const void* data)
{
if (_synchronous) {
return block->work(size, data);
}
Engine& engine = block->parent_graph()->engine();
if (_requests.write_space() < sizeof(MessageHeader) + size) {
engine.log().error("Work request ring overflow\n");
return LV2_WORKER_ERR_NO_SPACE;
}
const MessageHeader msg = { block, size };
if (_requests.write(sizeof(msg), &msg) != sizeof(msg)) {
engine.log().error("Error writing header to work request ring\n");
return LV2_WORKER_ERR_UNKNOWN;
}
if (_requests.write(size, data) != size) {
engine.log().error("Error writing body to work request ring\n");
return LV2_WORKER_ERR_UNKNOWN;
}
_sem.post();
return LV2_WORKER_SUCCESS;
}
SPtr
Worker::Schedule::feature(World* world, Node* n)
{
LV2Block* block = dynamic_cast(n);
if (!block) {
return SPtr();
}
LV2_Worker_Schedule* data = (LV2_Worker_Schedule*)malloc(
sizeof(LV2_Worker_Schedule));
data->handle = block;
data->schedule_work = synchronous ? schedule_sync : schedule;
LV2_Feature* f = (LV2_Feature*)malloc(sizeof(LV2_Feature));
f->URI = LV2_WORKER__schedule;
f->data = data;
return SPtr(f, &free_feature);
}
Worker::Worker(Log& log, uint32_t buffer_size, bool synchronous)
: _schedule(new Schedule(synchronous))
, _log(log)
, _sem(0)
, _requests(buffer_size)
, _responses(buffer_size)
, _buffer((uint8_t*)malloc(buffer_size))
, _buffer_size(buffer_size)
, _thread(nullptr)
, _exit_flag(false)
, _synchronous(synchronous)
{
if (!synchronous) {
_thread = new std::thread(&Worker::run, this);
}
}
Worker::~Worker()
{
_exit_flag = true;
_sem.post();
if (_thread) {
_thread->join();
delete _thread;
}
free(_buffer);
}
void
Worker::run()
{
while (_sem.wait() && !_exit_flag) {
MessageHeader msg;
if (_requests.read_space() > sizeof(msg)) {
if (_requests.read(sizeof(msg), &msg) != sizeof(msg)) {
_log.error("Error reading header from work request ring\n");
continue;
}
if (msg.size >= _buffer_size - sizeof(msg)) {
_log.error("Corrupt work request ring\n");
return;
}
if (_requests.read(msg.size, _buffer) != msg.size) {
_log.error("Error reading body from work request ring\n");
continue;
}
msg.block->work(msg.size, _buffer);
}
}
}
} // namespace server
} // namespace ingen