ardupilot/libraries/AP_HAL_ChibiOS/shared_dma.cpp
Andrew Tridgell a48f0ef90d HAL_ChibiOS: reduce the impact of UART DMA contention
this changes the heuristics for UART TX DMA allocation to greatly
reduce the chances of DMA contention causing long delays on other
devices

This fixes issues with FETTec driver output and gimbal status messages
as reported by Amilcar and OlliW. The problem is particularly bad when
no GPS is connected to GPS1 on fmuv3 and derived boards (such as
CubeBlack)

key changes:

 - remember the contention_counter across begin() calls, as the GPS
   calls begin with new baudrates regularly

 - added a is_shared() API to Shared_DMA, allowing the UART driver to
   avoid TX DMA on shared streams when at low baudrates.
2021-09-10 14:08:46 +09:00

279 lines
8.6 KiB
C++

/*
* This file is free software: you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This file 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 General Public License for more details.
*
* You should have received a copy of the GNU General Public License along
* with this program. If not, see <http://www.gnu.org/licenses/>.
*
* Code by Andrew Tridgell and Siddharth Bharat Purohit
*/
#include "shared_dma.h"
/*
code to handle sharing of DMA channels between peripherals
*/
#if CH_CFG_USE_MUTEXES == TRUE && !defined(HAL_NO_SHARED_DMA)
#include <AP_Common/ExpandingString.h>
using namespace ChibiOS;
extern const AP_HAL::HAL& hal;
Shared_DMA::dma_lock Shared_DMA::locks[SHARED_DMA_MAX_STREAM_ID+1];
volatile Shared_DMA::dma_stats* Shared_DMA::_contention_stats;
void Shared_DMA::init(void)
{
for (uint8_t i=0; i<SHARED_DMA_MAX_STREAM_ID; i++) {
chMtxObjectInit(&locks[i].mutex);
}
}
// constructor
Shared_DMA::Shared_DMA(uint8_t _stream_id1,
uint8_t _stream_id2,
dma_allocate_fn_t _allocate,
dma_deallocate_fn_t _deallocate)
{
stream_id1 = _stream_id1;
stream_id2 = _stream_id2;
if (stream_id2 < stream_id1) {
stream_id1 = _stream_id2;
stream_id2 = _stream_id1;
}
allocate = _allocate;
deallocate = _deallocate;
}
/*
return true if a stream ID is shared between two peripherals
*/
bool Shared_DMA::is_shared(uint8_t stream_id)
{
return (stream_id < SHARED_DMA_MAX_STREAM_ID) && ((1U<<stream_id) & SHARED_DMA_MASK) != 0;
}
//remove any assigned deallocator or allocator
void Shared_DMA::unregister()
{
if (stream_id1 < SHARED_DMA_MAX_STREAM_ID &&
locks[stream_id1].obj == this) {
locks[stream_id1].deallocate(this);
locks[stream_id1].obj = nullptr;
}
if (stream_id2 < SHARED_DMA_MAX_STREAM_ID &&
locks[stream_id2].obj == this) {
locks[stream_id2].deallocate(this);
locks[stream_id2].obj = nullptr;
}
}
// lock one stream
bool Shared_DMA::lock_stream(uint8_t stream_id)
{
bool cont = false;
if (stream_id < SHARED_DMA_MAX_STREAM_ID) {
const thread_t* curr_owner = locks[stream_id].mutex.owner;
chMtxLock(&locks[stream_id].mutex);
cont = curr_owner != nullptr && curr_owner != locks[stream_id].mutex.owner;
}
return cont;
}
// unlock one stream
void Shared_DMA::unlock_stream(uint8_t stream_id, bool success)
{
if (stream_id < SHARED_DMA_MAX_STREAM_ID) {
chMtxUnlock(&locks[stream_id].mutex);
if (success && _contention_stats != nullptr) {
_contention_stats[stream_id1].transactions++;
}
}
}
// lock one stream, non-blocking
bool Shared_DMA::lock_stream_nonblocking(uint8_t stream_id)
{
if (stream_id < SHARED_DMA_MAX_STREAM_ID) {
return chMtxTryLock(&locks[stream_id].mutex);
}
return true;
}
// lock the DMA channels
void Shared_DMA::lock_core(void)
{
// see if another driver has DMA allocated. If so, call their
// deallocation function
if (stream_id1 < SHARED_DMA_MAX_STREAM_ID &&
locks[stream_id1].obj && locks[stream_id1].obj != this) {
locks[stream_id1].deallocate(locks[stream_id1].obj);
locks[stream_id1].obj = nullptr;
}
if (stream_id2 < SHARED_DMA_MAX_STREAM_ID &&
locks[stream_id2].obj && locks[stream_id2].obj != this) {
locks[stream_id2].deallocate(locks[stream_id2].obj);
locks[stream_id2].obj = nullptr;
}
if ((stream_id1 < SHARED_DMA_MAX_STREAM_ID && locks[stream_id1].obj == nullptr) ||
(stream_id2 < SHARED_DMA_MAX_STREAM_ID && locks[stream_id2].obj == nullptr)) {
// allocate the DMA channels and put our deallocation function in place
allocate(this);
if (stream_id1 < SHARED_DMA_MAX_STREAM_ID) {
locks[stream_id1].deallocate = deallocate;
locks[stream_id1].obj = this;
}
if (stream_id2 < SHARED_DMA_MAX_STREAM_ID) {
locks[stream_id2].deallocate = deallocate;
locks[stream_id2].obj = this;
}
}
#ifdef STM32_DMA_STREAM_ID_ANY
else if (stream_id1 == STM32_DMA_STREAM_ID_ANY ||
stream_id2 == STM32_DMA_STREAM_ID_ANY) {
// call allocator without needing locking
allocate(this);
}
#endif
// update contention stats
if (_contention_stats != nullptr) {
if (stream_id1 < SHARED_DMA_MAX_STREAM_ID) {
if (contention) {
_contention_stats[stream_id1].contended_locks++;
} else {
_contention_stats[stream_id1].uncontended_locks++;
}
}
if (stream_id2 < SHARED_DMA_MAX_STREAM_ID) {
if (contention) {
_contention_stats[stream_id2].contended_locks++;
} else {
_contention_stats[stream_id2].uncontended_locks++;
}
}
}
have_lock = true;
}
// lock the DMA channels, blocking method
void Shared_DMA::lock(void)
{
bool c1 = lock_stream(stream_id1);
bool c2 = lock_stream(stream_id2);
contention = c1 || c2;
lock_core();
}
// lock the DMA channels, non-blocking
bool Shared_DMA::lock_nonblock(void)
{
if (!lock_stream_nonblocking(stream_id1)) {
chSysDisable();
if (locks[stream_id1].obj != nullptr && locks[stream_id1].obj != this) {
locks[stream_id1].obj->contention = true;
if (_contention_stats != nullptr) {
_contention_stats[stream_id1].contended_locks++;
}
}
chSysEnable();
contention = true;
return false;
}
if (_contention_stats != nullptr && stream_id1 < SHARED_DMA_MAX_STREAM_ID) {
_contention_stats[stream_id1].uncontended_locks++;
}
if (!lock_stream_nonblocking(stream_id2)) {
unlock_stream(stream_id1, false);
chSysDisable();
if (locks[stream_id2].obj != nullptr && locks[stream_id2].obj != this) {
locks[stream_id2].obj->contention = true;
if (_contention_stats != nullptr) {
_contention_stats[stream_id2].contended_locks++;
}
}
chSysEnable();
contention = true;
return false;
}
lock_core();
if (_contention_stats != nullptr && stream_id2 < SHARED_DMA_MAX_STREAM_ID) {
_contention_stats[stream_id2].uncontended_locks++;
}
return true;
}
// unlock the DMA channels
#pragma GCC diagnostic push
#pragma GCC diagnostic error "-Wframe-larger-than=128"
void Shared_DMA::unlock(bool success)
{
osalDbgAssert(have_lock, "must have lock");
have_lock = false;
unlock_stream(stream_id2, success);
unlock_stream(stream_id1, success);
}
#pragma GCC diagnostic pop
/*
lock all channels - used on reboot to ensure no sensor DMA is in
progress
*/
void Shared_DMA::lock_all(void)
{
for (uint8_t i=0; i<SHARED_DMA_MAX_STREAM_ID; i++) {
lock_stream(i);
}
}
// display dma contention statistics as text buffer for @SYS/dma.txt
void Shared_DMA::dma_info(ExpandingString &str)
{
// no buffer allocated, start counting
if (_contention_stats == nullptr) {
_contention_stats = new dma_stats[SHARED_DMA_MAX_STREAM_ID+1];
// return zeros on first fetch
}
// a header to allow for machine parsers to determine format
str.printf("DMAV1\n");
for (uint8_t i = 0; i < SHARED_DMA_MAX_STREAM_ID; i++) {
// ignore locks not in use
if (_contention_stats[i].contended_locks == 0
&& _contention_stats[i].uncontended_locks == 0
&& _contention_stats[i].transactions == 0) {
continue;
}
#if STM32_DMA_ADVANCED
#define STREAM_MUX 8
#define STREAM_OFFSET 0
#else
#define STREAM_MUX 7
#define STREAM_OFFSET 1
#endif
const char* fmt = "DMA=%1u:%1u TX=%8u ULCK=%8u CLCK=%8u CONT=%4.1f%%\n";
float cond_per = 100.0f * float(_contention_stats[i].contended_locks)
/ (1 + _contention_stats[i].contended_locks + _contention_stats[i].uncontended_locks);
str.printf(fmt, i / STREAM_MUX + 1, i % STREAM_MUX + STREAM_OFFSET, _contention_stats[i].transactions,
_contention_stats[i].uncontended_locks, _contention_stats[i].contended_locks, cond_per);
_contention_stats[i].contended_locks = 0;
_contention_stats[i].uncontended_locks = 0;
}
}
#endif // CH_CFG_USE_SEMAPHORES