mirror of
https://github.com/ArduPilot/ardupilot
synced 2025-01-07 00:18:29 -04:00
ccb583d092
add support for log time to block logger refactor rotation into backed. Don't start logs when erasing correct log start logic separate read and write points so that requesting log information does not corrupt the current log when starting a new log stop logging first clear the write buffer when starting a new log insert utc time when requesting info for the current log stop logging and request formats again when starting a new log cope with erase happening while we are logging keep pushing out startup messages even when format messages are done don't log to the gcs in the io thread don't start new logs in the io thread don't validate logs while erasing flush logs when stopping logging account for page header when calculating logs sizes don't return data when asked for more data than in the log optimize locking and use separate semaphore to mediate ring buffer access stop logging when the chip is full and send a notification calculate logs sizes correctly even when they wrap read log data correctly even when it wraps add stats support to block logger reset dropped when starting a new log fail logging when the chip is full refactor critical bufferspace checks increase messagewriter budget to 250us and to 300us for FMT
1009 lines
28 KiB
C++
1009 lines
28 KiB
C++
/*
|
|
AP_Logger logging - file oriented variant
|
|
|
|
This uses posix file IO to create log files called logs/NN.bin in the
|
|
given directory
|
|
|
|
SD Card Rates on PixHawk:
|
|
- deletion rate seems to be ~50 files/second.
|
|
- stat seems to be ~150/second
|
|
- readdir loop of 511 entry directory ~62,000 microseconds
|
|
*/
|
|
|
|
#include <AP_HAL/AP_HAL.h>
|
|
#include <AP_Filesystem/AP_Filesystem.h>
|
|
|
|
#if HAVE_FILESYSTEM_SUPPORT
|
|
#include "AP_Logger_File.h"
|
|
|
|
#include <AP_Common/AP_Common.h>
|
|
#include <AP_InternalError/AP_InternalError.h>
|
|
#include <AP_RTC/AP_RTC.h>
|
|
|
|
#include <AP_Math/AP_Math.h>
|
|
#include <GCS_MAVLink/GCS.h>
|
|
#include <stdio.h>
|
|
|
|
|
|
extern const AP_HAL::HAL& hal;
|
|
|
|
#define LOGGER_PAGE_SIZE 1024UL
|
|
|
|
#ifndef HAL_LOGGER_WRITE_CHUNK_SIZE
|
|
#define HAL_LOGGER_WRITE_CHUNK_SIZE 4096
|
|
#endif
|
|
|
|
#define MB_to_B 1000000
|
|
#define B_to_MB 0.000001
|
|
|
|
/*
|
|
constructor
|
|
*/
|
|
AP_Logger_File::AP_Logger_File(AP_Logger &front,
|
|
LoggerMessageWriter_DFLogStart *writer,
|
|
const char *log_directory) :
|
|
AP_Logger_Backend(front, writer),
|
|
_write_fd(-1),
|
|
_read_fd(-1),
|
|
_log_directory(log_directory),
|
|
_writebuf(0),
|
|
_writebuf_chunk(HAL_LOGGER_WRITE_CHUNK_SIZE),
|
|
_perf_write(hal.util->perf_alloc(AP_HAL::Util::PC_ELAPSED, "DF_write")),
|
|
_perf_fsync(hal.util->perf_alloc(AP_HAL::Util::PC_ELAPSED, "DF_fsync")),
|
|
_perf_errors(hal.util->perf_alloc(AP_HAL::Util::PC_COUNT, "DF_errors")),
|
|
_perf_overruns(hal.util->perf_alloc(AP_HAL::Util::PC_COUNT, "DF_overruns"))
|
|
{
|
|
df_stats_clear();
|
|
}
|
|
|
|
|
|
void AP_Logger_File::ensure_log_directory_exists()
|
|
{
|
|
int ret;
|
|
struct stat st;
|
|
|
|
const char* custom_dir = hal.util->get_custom_log_directory();
|
|
if (custom_dir != nullptr){
|
|
_log_directory = custom_dir;
|
|
}
|
|
|
|
EXPECT_DELAY_MS(3000);
|
|
ret = AP::FS().stat(_log_directory, &st);
|
|
if (ret == -1) {
|
|
ret = AP::FS().mkdir(_log_directory);
|
|
}
|
|
if (ret == -1 && errno != EEXIST) {
|
|
printf("Failed to create log directory %s : %s\n", _log_directory, strerror(errno));
|
|
}
|
|
}
|
|
|
|
void AP_Logger_File::Init()
|
|
{
|
|
// determine and limit file backend buffersize
|
|
uint32_t bufsize = _front._params.file_bufsize;
|
|
if (bufsize > 64) {
|
|
bufsize = 64; // PixHawk has DMA limitations.
|
|
}
|
|
bufsize *= 1024;
|
|
|
|
const uint32_t desired_bufsize = bufsize;
|
|
|
|
// If we can't allocate the full size, try to reduce it until we can allocate it
|
|
while (!_writebuf.set_size(bufsize) && bufsize >= _writebuf_chunk) {
|
|
bufsize *= 0.9;
|
|
}
|
|
if (bufsize >= _writebuf_chunk && bufsize != desired_bufsize) {
|
|
hal.console->printf("AP_Logger: reduced buffer %u/%u\n", (unsigned)bufsize, (unsigned)desired_bufsize);
|
|
}
|
|
|
|
if (!_writebuf.get_size()) {
|
|
hal.console->printf("Out of memory for logging\n");
|
|
return;
|
|
}
|
|
|
|
hal.console->printf("AP_Logger_File: buffer size=%u\n", (unsigned)bufsize);
|
|
|
|
_initialised = true;
|
|
hal.scheduler->register_io_process(FUNCTOR_BIND_MEMBER(&AP_Logger_File::_io_timer, void));
|
|
}
|
|
|
|
bool AP_Logger_File::file_exists(const char *filename) const
|
|
{
|
|
struct stat st;
|
|
EXPECT_DELAY_MS(3000);
|
|
if (AP::FS().stat(filename, &st) == -1) {
|
|
// hopefully errno==ENOENT. If some error occurs it is
|
|
// probably better to assume this file exists.
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
bool AP_Logger_File::log_exists(const uint16_t lognum) const
|
|
{
|
|
char *filename = _log_file_name(lognum);
|
|
if (filename == nullptr) {
|
|
return false; // ?!
|
|
}
|
|
bool ret = file_exists(filename);
|
|
free(filename);
|
|
return ret;
|
|
}
|
|
|
|
void AP_Logger_File::periodic_1Hz()
|
|
{
|
|
AP_Logger_Backend::periodic_1Hz();
|
|
|
|
if (!io_thread_alive()) {
|
|
if (io_thread_warning_decimation_counter == 0 && _initialised) {
|
|
// we don't print this error unless we did initialise. When _initialised is set to true
|
|
// we register the IO timer callback
|
|
gcs().send_text(MAV_SEVERITY_CRITICAL, "AP_Logger: stuck thread (%s)", last_io_operation);
|
|
}
|
|
if (io_thread_warning_decimation_counter++ > 57) {
|
|
io_thread_warning_decimation_counter = 0;
|
|
}
|
|
// If you try to close the file here then it will almost
|
|
// certainly block. Since this is the main thread, this is
|
|
// likely to cause a crash.
|
|
|
|
// semaphore_write_fd not taken here as if the io thread is
|
|
// dead it may not release lock...
|
|
_write_fd = -1;
|
|
_initialised = false;
|
|
}
|
|
}
|
|
|
|
void AP_Logger_File::periodic_fullrate()
|
|
{
|
|
AP_Logger_Backend::push_log_blocks();
|
|
}
|
|
|
|
uint32_t AP_Logger_File::bufferspace_available()
|
|
{
|
|
const uint32_t space = _writebuf.space();
|
|
const uint32_t crit = critical_message_reserved_space(_writebuf.get_size());
|
|
|
|
return (space > crit) ? space - crit : 0;
|
|
}
|
|
|
|
// return true for CardInserted() if we successfully initialized
|
|
bool AP_Logger_File::CardInserted(void) const
|
|
{
|
|
return _initialised && !_open_error;
|
|
}
|
|
|
|
// returns the amount of disk space available in _log_directory (in bytes)
|
|
// returns -1 on error
|
|
int64_t AP_Logger_File::disk_space_avail()
|
|
{
|
|
return AP::FS().disk_free(_log_directory);
|
|
}
|
|
|
|
// returns the total amount of disk space (in use + available) in
|
|
// _log_directory (in bytes).
|
|
// returns -1 on error
|
|
int64_t AP_Logger_File::disk_space()
|
|
{
|
|
return AP::FS().disk_space(_log_directory);
|
|
}
|
|
|
|
// find_oldest_log - find oldest log in _log_directory
|
|
// returns 0 if no log was found
|
|
uint16_t AP_Logger_File::find_oldest_log()
|
|
{
|
|
if (_cached_oldest_log != 0) {
|
|
return _cached_oldest_log;
|
|
}
|
|
|
|
uint16_t last_log_num = find_last_log();
|
|
if (last_log_num == 0) {
|
|
return 0;
|
|
}
|
|
|
|
uint16_t current_oldest_log = 0; // 0 is invalid
|
|
|
|
// We could count up to find_last_log(), but if people start
|
|
// relying on the min_avail_space_percent feature we could end up
|
|
// doing a *lot* of asprintf()s and stat()s
|
|
EXPECT_DELAY_MS(3000);
|
|
auto *d = AP::FS().opendir(_log_directory);
|
|
if (d == nullptr) {
|
|
// SD card may have died? On linux someone may have rm-rf-d
|
|
return 0;
|
|
}
|
|
|
|
// we only remove files which look like xxx.BIN
|
|
EXPECT_DELAY_MS(3000);
|
|
for (struct dirent *de=AP::FS().readdir(d); de; de=AP::FS().readdir(d)) {
|
|
EXPECT_DELAY_MS(3000);
|
|
uint8_t length = strlen(de->d_name);
|
|
if (length < 5) {
|
|
// not long enough for \d+[.]BIN
|
|
continue;
|
|
}
|
|
if (strncmp(&de->d_name[length-4], ".BIN", 4)) {
|
|
// doesn't end in .BIN
|
|
continue;
|
|
}
|
|
|
|
uint16_t thisnum = strtoul(de->d_name, nullptr, 10);
|
|
if (thisnum > MAX_LOG_FILES) {
|
|
// ignore files above our official maximum...
|
|
continue;
|
|
}
|
|
if (current_oldest_log == 0) {
|
|
current_oldest_log = thisnum;
|
|
} else {
|
|
if (current_oldest_log <= last_log_num) {
|
|
if (thisnum > last_log_num) {
|
|
current_oldest_log = thisnum;
|
|
} else if (thisnum < current_oldest_log) {
|
|
current_oldest_log = thisnum;
|
|
}
|
|
} else { // current_oldest_log > last_log_num
|
|
if (thisnum > last_log_num) {
|
|
if (thisnum < current_oldest_log) {
|
|
current_oldest_log = thisnum;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
AP::FS().closedir(d);
|
|
_cached_oldest_log = current_oldest_log;
|
|
|
|
return current_oldest_log;
|
|
}
|
|
|
|
void AP_Logger_File::Prep_MinSpace()
|
|
{
|
|
if (hal.util->was_watchdog_reset()) {
|
|
// don't clear space if watchdog reset, it takes too long
|
|
return;
|
|
}
|
|
const uint16_t first_log_to_remove = find_oldest_log();
|
|
if (first_log_to_remove == 0) {
|
|
// no files to remove
|
|
return;
|
|
}
|
|
|
|
const int64_t target_free = (int64_t)_front._params.min_MB_free * MB_to_B;
|
|
|
|
_cached_oldest_log = 0;
|
|
|
|
uint16_t log_to_remove = first_log_to_remove;
|
|
|
|
uint16_t count = 0;
|
|
do {
|
|
int64_t avail = disk_space_avail();
|
|
if (avail == -1) {
|
|
break;
|
|
}
|
|
if (avail >= target_free) {
|
|
break;
|
|
}
|
|
if (count++ > MAX_LOG_FILES+10) {
|
|
// *way* too many deletions going on here. Possible internal error.
|
|
INTERNAL_ERROR(AP_InternalError::error_t::logger_too_many_deletions);
|
|
break;
|
|
}
|
|
char *filename_to_remove = _log_file_name(log_to_remove);
|
|
if (filename_to_remove == nullptr) {
|
|
INTERNAL_ERROR(AP_InternalError::error_t::logger_bad_getfilename);
|
|
break;
|
|
}
|
|
if (file_exists(filename_to_remove)) {
|
|
hal.console->printf("Removing (%s) for minimum-space requirements (%.0fMB < %.0fMB)\n",
|
|
filename_to_remove, (double)avail*B_to_MB, (double)target_free*B_to_MB);
|
|
EXPECT_DELAY_MS(2000);
|
|
if (AP::FS().unlink(filename_to_remove) == -1) {
|
|
hal.console->printf("Failed to remove %s: %s\n", filename_to_remove, strerror(errno));
|
|
free(filename_to_remove);
|
|
if (errno == ENOENT) {
|
|
// corruption - should always have a continuous
|
|
// sequence of files... however, there may be still
|
|
// files out there, so keep going.
|
|
} else {
|
|
break;
|
|
}
|
|
} else {
|
|
free(filename_to_remove);
|
|
}
|
|
}
|
|
log_to_remove++;
|
|
if (log_to_remove > MAX_LOG_FILES) {
|
|
log_to_remove = 1;
|
|
}
|
|
} while (log_to_remove != first_log_to_remove);
|
|
}
|
|
|
|
void AP_Logger_File::Prep() {
|
|
if (!NeedPrep()) {
|
|
return;
|
|
}
|
|
if (hal.util->get_soft_armed()) {
|
|
// do not want to do any filesystem operations while we are e.g. flying
|
|
return;
|
|
}
|
|
Prep_MinSpace();
|
|
}
|
|
|
|
bool AP_Logger_File::NeedPrep()
|
|
{
|
|
if (!CardInserted()) {
|
|
// should not have been called?!
|
|
return false;
|
|
}
|
|
|
|
const int64_t actual = disk_space_avail();
|
|
if (actual == -1) {
|
|
return false;
|
|
}
|
|
|
|
if (actual < (int64_t)_front._params.min_MB_free * MB_to_B) {
|
|
return true;
|
|
}
|
|
|
|
return false;
|
|
}
|
|
|
|
/*
|
|
construct a log file name given a log number.
|
|
The number in the log filename will *not* be zero-padded.
|
|
Note: Caller must free.
|
|
*/
|
|
char *AP_Logger_File::_log_file_name_short(const uint16_t log_num) const
|
|
{
|
|
char *buf = nullptr;
|
|
if (asprintf(&buf, "%s/%u.BIN", _log_directory, (unsigned)log_num) == -1) {
|
|
return nullptr;
|
|
}
|
|
return buf;
|
|
}
|
|
|
|
/*
|
|
construct a log file name given a log number.
|
|
The number in the log filename will be zero-padded.
|
|
Note: Caller must free.
|
|
*/
|
|
char *AP_Logger_File::_log_file_name_long(const uint16_t log_num) const
|
|
{
|
|
char *buf = nullptr;
|
|
if (asprintf(&buf, "%s/%08u.BIN", _log_directory, (unsigned)log_num) == -1) {
|
|
return nullptr;
|
|
}
|
|
return buf;
|
|
}
|
|
|
|
/*
|
|
return a log filename appropriate for the supplied log_num if a
|
|
filename exists with the short (not-zero-padded name) then it is the
|
|
appropirate name, otherwise the long (zero-padded) version is.
|
|
Note: Caller must free.
|
|
*/
|
|
char *AP_Logger_File::_log_file_name(const uint16_t log_num) const
|
|
{
|
|
char *filename = _log_file_name_short(log_num);
|
|
if (filename == nullptr) {
|
|
return nullptr;
|
|
}
|
|
if (file_exists(filename)) {
|
|
return filename;
|
|
}
|
|
free(filename);
|
|
return _log_file_name_long(log_num);
|
|
}
|
|
|
|
/*
|
|
return path name of the lastlog.txt marker file
|
|
Note: Caller must free.
|
|
*/
|
|
char *AP_Logger_File::_lastlog_file_name(void) const
|
|
{
|
|
char *buf = nullptr;
|
|
if (asprintf(&buf, "%s/LASTLOG.TXT", _log_directory) == -1) {
|
|
return nullptr;
|
|
}
|
|
return buf;
|
|
}
|
|
|
|
|
|
// remove all log files
|
|
void AP_Logger_File::EraseAll()
|
|
{
|
|
if (hal.util->get_soft_armed()) {
|
|
// do not want to do any filesystem operations while we are e.g. flying
|
|
return;
|
|
}
|
|
if (!_initialised) {
|
|
return;
|
|
}
|
|
|
|
const bool was_logging = (_write_fd != -1);
|
|
stop_logging();
|
|
|
|
for (uint16_t log_num=1; log_num<=MAX_LOG_FILES; log_num++) {
|
|
char *fname = _log_file_name(log_num);
|
|
if (fname == nullptr) {
|
|
break;
|
|
}
|
|
EXPECT_DELAY_MS(3000);
|
|
AP::FS().unlink(fname);
|
|
free(fname);
|
|
}
|
|
char *fname = _lastlog_file_name();
|
|
if (fname != nullptr) {
|
|
AP::FS().unlink(fname);
|
|
free(fname);
|
|
}
|
|
|
|
_cached_oldest_log = 0;
|
|
|
|
if (was_logging) {
|
|
start_new_log();
|
|
}
|
|
}
|
|
|
|
bool AP_Logger_File::WritesOK() const
|
|
{
|
|
if (_write_fd == -1) {
|
|
return false;
|
|
}
|
|
if (_open_error) {
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
|
|
bool AP_Logger_File::StartNewLogOK() const
|
|
{
|
|
if (_open_error) {
|
|
return false;
|
|
}
|
|
return AP_Logger_Backend::StartNewLogOK();
|
|
}
|
|
|
|
/* Write a block of data at current offset */
|
|
bool AP_Logger_File::_WritePrioritisedBlock(const void *pBuffer, uint16_t size, bool is_critical)
|
|
{
|
|
if (! WriteBlockCheckStartupMessages()) {
|
|
_dropped++;
|
|
return false;
|
|
}
|
|
|
|
if (!semaphore.take(1)) {
|
|
return false;
|
|
}
|
|
|
|
uint32_t space = _writebuf.space();
|
|
|
|
if (_writing_startup_messages &&
|
|
_startup_messagewriter->fmt_done()) {
|
|
// the state machine has called us, and it has finished
|
|
// writing format messages out. It can always get back to us
|
|
// with more messages later, so let's leave room for other
|
|
// things:
|
|
const uint32_t now = AP_HAL::millis();
|
|
const bool must_dribble = (now - last_messagewrite_message_sent) > 100;
|
|
if (!must_dribble &&
|
|
space < non_messagewriter_message_reserved_space(_writebuf.get_size())) {
|
|
// this message isn't dropped, it will be sent again...
|
|
semaphore.give();
|
|
return false;
|
|
}
|
|
last_messagewrite_message_sent = now;
|
|
} else {
|
|
// we reserve some amount of space for critical messages:
|
|
if (!is_critical && space < critical_message_reserved_space(_writebuf.get_size())) {
|
|
_dropped++;
|
|
semaphore.give();
|
|
return false;
|
|
}
|
|
}
|
|
|
|
// if no room for entire message - drop it:
|
|
if (space < size) {
|
|
hal.util->perf_count(_perf_overruns);
|
|
_dropped++;
|
|
semaphore.give();
|
|
return false;
|
|
}
|
|
|
|
_writebuf.write((uint8_t*)pBuffer, size);
|
|
df_stats_gather(size, _writebuf.space());
|
|
semaphore.give();
|
|
return true;
|
|
}
|
|
|
|
/*
|
|
find the highest log number
|
|
*/
|
|
uint16_t AP_Logger_File::find_last_log()
|
|
{
|
|
unsigned ret = 0;
|
|
char *fname = _lastlog_file_name();
|
|
if (fname == nullptr) {
|
|
return ret;
|
|
}
|
|
EXPECT_DELAY_MS(3000);
|
|
int fd = AP::FS().open(fname, O_RDONLY);
|
|
free(fname);
|
|
if (fd != -1) {
|
|
char buf[10];
|
|
memset(buf, 0, sizeof(buf));
|
|
if (AP::FS().read(fd, buf, sizeof(buf)-1) > 0) {
|
|
ret = strtol(buf, NULL, 10);
|
|
}
|
|
AP::FS().close(fd);
|
|
}
|
|
return ret;
|
|
}
|
|
|
|
uint32_t AP_Logger_File::_get_log_size(const uint16_t log_num)
|
|
{
|
|
char *fname = _log_file_name(log_num);
|
|
if (fname == nullptr) {
|
|
return 0;
|
|
}
|
|
if (_write_fd != -1 && write_fd_semaphore.take_nonblocking()) {
|
|
if (_write_filename != nullptr && strcmp(_write_filename, fname) == 0) {
|
|
// it is the file we are currently writing
|
|
free(fname);
|
|
write_fd_semaphore.give();
|
|
return _write_offset;
|
|
}
|
|
write_fd_semaphore.give();
|
|
}
|
|
struct stat st;
|
|
EXPECT_DELAY_MS(3000);
|
|
if (AP::FS().stat(fname, &st) != 0) {
|
|
printf("Unable to fetch Log File Size: %s\n", strerror(errno));
|
|
free(fname);
|
|
return 0;
|
|
}
|
|
free(fname);
|
|
return st.st_size;
|
|
}
|
|
|
|
uint32_t AP_Logger_File::_get_log_time(const uint16_t log_num)
|
|
{
|
|
char *fname = _log_file_name(log_num);
|
|
if (fname == nullptr) {
|
|
return 0;
|
|
}
|
|
if (_write_fd != -1 && write_fd_semaphore.take_nonblocking()) {
|
|
if (_write_filename != nullptr && strcmp(_write_filename, fname) == 0) {
|
|
// it is the file we are currently writing
|
|
free(fname);
|
|
write_fd_semaphore.give();
|
|
uint64_t utc_usec;
|
|
if (!AP::rtc().get_utc_usec(utc_usec)) {
|
|
return 0;
|
|
}
|
|
return utc_usec / 1000000U;
|
|
}
|
|
write_fd_semaphore.give();
|
|
}
|
|
struct stat st;
|
|
EXPECT_DELAY_MS(3000);
|
|
if (AP::FS().stat(fname, &st) != 0) {
|
|
free(fname);
|
|
return 0;
|
|
}
|
|
free(fname);
|
|
return st.st_mtime;
|
|
}
|
|
|
|
/*
|
|
find the number of pages in a log
|
|
*/
|
|
void AP_Logger_File::get_log_boundaries(const uint16_t list_entry, uint32_t & start_page, uint32_t & end_page)
|
|
{
|
|
const uint16_t log_num = log_num_from_list_entry(list_entry);
|
|
if (log_num == 0) {
|
|
// that failed - probably no logs
|
|
start_page = 0;
|
|
end_page = 0;
|
|
return;
|
|
}
|
|
|
|
start_page = 0;
|
|
end_page = _get_log_size(log_num) / LOGGER_PAGE_SIZE;
|
|
}
|
|
|
|
/*
|
|
retrieve data from a log file
|
|
*/
|
|
int16_t AP_Logger_File::get_log_data(const uint16_t list_entry, const uint16_t page, const uint32_t offset, const uint16_t len, uint8_t *data)
|
|
{
|
|
if (!_initialised || _open_error) {
|
|
return -1;
|
|
}
|
|
|
|
const uint16_t log_num = log_num_from_list_entry(list_entry);
|
|
if (log_num == 0) {
|
|
// that failed - probably no logs
|
|
return -1;
|
|
}
|
|
|
|
if (_read_fd != -1 && log_num != _read_fd_log_num) {
|
|
AP::FS().close(_read_fd);
|
|
_read_fd = -1;
|
|
}
|
|
if (_read_fd == -1) {
|
|
char *fname = _log_file_name(log_num);
|
|
if (fname == nullptr) {
|
|
return -1;
|
|
}
|
|
stop_logging();
|
|
EXPECT_DELAY_MS(3000);
|
|
_read_fd = AP::FS().open(fname, O_RDONLY);
|
|
if (_read_fd == -1) {
|
|
_open_error = true;
|
|
int saved_errno = errno;
|
|
::printf("Log read open fail for %s - %s\n",
|
|
fname, strerror(saved_errno));
|
|
hal.console->printf("Log read open fail for %s - %s\n",
|
|
fname, strerror(saved_errno));
|
|
free(fname);
|
|
return -1;
|
|
}
|
|
free(fname);
|
|
_read_offset = 0;
|
|
_read_fd_log_num = log_num;
|
|
}
|
|
uint32_t ofs = page * (uint32_t)LOGGER_PAGE_SIZE + offset;
|
|
|
|
if (ofs != _read_offset) {
|
|
if (AP::FS().lseek(_read_fd, ofs, SEEK_SET) == (off_t)-1) {
|
|
AP::FS().close(_read_fd);
|
|
_read_fd = -1;
|
|
return -1;
|
|
}
|
|
_read_offset = ofs;
|
|
}
|
|
int16_t ret = (int16_t)AP::FS().read(_read_fd, data, len);
|
|
if (ret > 0) {
|
|
_read_offset += ret;
|
|
}
|
|
return ret;
|
|
}
|
|
|
|
/*
|
|
find size and date of a log
|
|
*/
|
|
void AP_Logger_File::get_log_info(const uint16_t list_entry, uint32_t &size, uint32_t &time_utc)
|
|
{
|
|
uint16_t log_num = log_num_from_list_entry(list_entry);
|
|
if (log_num == 0) {
|
|
// that failed - probably no logs
|
|
size = 0;
|
|
time_utc = 0;
|
|
return;
|
|
}
|
|
|
|
size = _get_log_size(log_num);
|
|
time_utc = _get_log_time(log_num);
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
get the number of logs - note that the log numbers must be consecutive
|
|
*/
|
|
uint16_t AP_Logger_File::get_num_logs()
|
|
{
|
|
uint16_t ret = 0;
|
|
uint16_t high = find_last_log();
|
|
uint16_t i;
|
|
for (i=high; i>0; i--) {
|
|
if (! log_exists(i)) {
|
|
break;
|
|
}
|
|
ret++;
|
|
}
|
|
if (i == 0) {
|
|
for (i=MAX_LOG_FILES; i>high; i--) {
|
|
if (! log_exists(i)) {
|
|
break;
|
|
}
|
|
ret++;
|
|
}
|
|
}
|
|
return ret;
|
|
}
|
|
|
|
/*
|
|
stop logging
|
|
*/
|
|
void AP_Logger_File::stop_logging(void)
|
|
{
|
|
// best-case effort to avoid annoying the IO thread
|
|
const bool have_sem = write_fd_semaphore.take(hal.util->get_soft_armed()?1:20);
|
|
if (_write_fd != -1) {
|
|
int fd = _write_fd;
|
|
_write_fd = -1;
|
|
AP::FS().close(fd);
|
|
}
|
|
if (have_sem) {
|
|
write_fd_semaphore.give();
|
|
}
|
|
}
|
|
|
|
/*
|
|
start writing to a new log file
|
|
*/
|
|
void AP_Logger_File::start_new_log(void)
|
|
{
|
|
if (_open_error) {
|
|
// we have previously failed to open a file - don't try again
|
|
// to prevent us trying to open files while in flight
|
|
return;
|
|
}
|
|
|
|
// set _open_error here to avoid infinite recursion. Simply
|
|
// writing a prioritised block may try to open a log - which means
|
|
// if anything in the start_new_log path does a gcs().send_text()
|
|
// (for example), you will end up recursing if we don't take
|
|
// precautions. We will reset _open_error if we actually manage
|
|
// to open the log...
|
|
_open_error = true;
|
|
|
|
stop_logging();
|
|
|
|
start_new_log_reset_variables();
|
|
|
|
if (_read_fd != -1) {
|
|
AP::FS().close(_read_fd);
|
|
_read_fd = -1;
|
|
}
|
|
|
|
if (disk_space_avail() < _free_space_min_avail && disk_space() > 0) {
|
|
hal.console->printf("Out of space for logging\n");
|
|
return;
|
|
}
|
|
|
|
uint16_t log_num = find_last_log();
|
|
// re-use empty logs if possible
|
|
if (_get_log_size(log_num) > 0 || log_num == 0) {
|
|
log_num++;
|
|
}
|
|
if (log_num > MAX_LOG_FILES) {
|
|
log_num = 1;
|
|
}
|
|
if (!write_fd_semaphore.take(1)) {
|
|
return;
|
|
}
|
|
if (_write_filename) {
|
|
free(_write_filename);
|
|
_write_filename = nullptr;
|
|
}
|
|
_write_filename = _log_file_name(log_num);
|
|
if (_write_filename == nullptr) {
|
|
write_fd_semaphore.give();
|
|
return;
|
|
}
|
|
|
|
#if CONFIG_HAL_BOARD == HAL_BOARD_CHIBIOS
|
|
// remember if we had utc time when we opened the file
|
|
uint64_t utc_usec;
|
|
_need_rtc_update = !AP::rtc().get_utc_usec(utc_usec);
|
|
#endif
|
|
|
|
// create the log directory if need be
|
|
ensure_log_directory_exists();
|
|
|
|
EXPECT_DELAY_MS(3000);
|
|
_write_fd = AP::FS().open(_write_filename, O_WRONLY|O_CREAT|O_TRUNC);
|
|
_cached_oldest_log = 0;
|
|
|
|
if (_write_fd == -1) {
|
|
_initialised = false;
|
|
write_fd_semaphore.give();
|
|
int saved_errno = errno;
|
|
::printf("Log open fail for %s - %s\n",
|
|
_write_filename, strerror(saved_errno));
|
|
hal.console->printf("Log open fail for %s - %s\n",
|
|
_write_filename, strerror(saved_errno));
|
|
return;
|
|
}
|
|
_last_write_ms = AP_HAL::millis();
|
|
_write_offset = 0;
|
|
_writebuf.clear();
|
|
write_fd_semaphore.give();
|
|
|
|
// now update lastlog.txt with the new log number
|
|
char *fname = _lastlog_file_name();
|
|
|
|
EXPECT_DELAY_MS(3000);
|
|
int fd = AP::FS().open(fname, O_WRONLY|O_CREAT);
|
|
free(fname);
|
|
if (fd == -1) {
|
|
return;
|
|
}
|
|
|
|
char buf[30];
|
|
snprintf(buf, sizeof(buf), "%u\r\n", (unsigned)log_num);
|
|
const ssize_t to_write = strlen(buf);
|
|
const ssize_t written = AP::FS().write(fd, buf, to_write);
|
|
AP::FS().close(fd);
|
|
|
|
if (written < to_write) {
|
|
return;
|
|
}
|
|
_open_error = false;
|
|
|
|
return;
|
|
}
|
|
|
|
|
|
#if CONFIG_HAL_BOARD == HAL_BOARD_SITL || CONFIG_HAL_BOARD == HAL_BOARD_LINUX
|
|
void AP_Logger_File::flush(void)
|
|
#if APM_BUILD_TYPE(APM_BUILD_Replay) || APM_BUILD_TYPE(APM_BUILD_UNKNOWN)
|
|
{
|
|
uint32_t tnow = AP_HAL::millis();
|
|
while (_write_fd != -1 && _initialised && !_open_error && _writebuf.available()) {
|
|
// convince the IO timer that it really is OK to write out
|
|
// less than _writebuf_chunk bytes:
|
|
if (tnow > 2001) { // avoid resetting _last_write_time to 0
|
|
_last_write_time = tnow - 2001;
|
|
}
|
|
_io_timer();
|
|
}
|
|
if (write_fd_semaphore.take(1)) {
|
|
if (_write_fd != -1) {
|
|
::fsync(_write_fd);
|
|
}
|
|
write_fd_semaphore.give();
|
|
} else {
|
|
INTERNAL_ERROR(AP_InternalError::error_t::logger_flushing_without_sem);
|
|
}
|
|
}
|
|
#else
|
|
{
|
|
// flush is for replay and examples only
|
|
}
|
|
#endif // APM_BUILD_TYPE(APM_BUILD_Replay) || APM_BUILD_TYPE(APM_BUILD_UNKNOWN)
|
|
#endif
|
|
|
|
void AP_Logger_File::_io_timer(void)
|
|
{
|
|
uint32_t tnow = AP_HAL::millis();
|
|
_io_timer_heartbeat = tnow;
|
|
if (_write_fd == -1 || !_initialised || _open_error) {
|
|
return;
|
|
}
|
|
|
|
uint32_t nbytes = _writebuf.available();
|
|
if (nbytes == 0) {
|
|
return;
|
|
}
|
|
if (nbytes < _writebuf_chunk &&
|
|
tnow - _last_write_time < 2000UL) {
|
|
// write in _writebuf_chunk-sized chunks, but always write at
|
|
// least once per 2 seconds if data is available
|
|
return;
|
|
}
|
|
if (tnow - _free_space_last_check_time > _free_space_check_interval) {
|
|
_free_space_last_check_time = tnow;
|
|
last_io_operation = "disk_space_avail";
|
|
if (disk_space_avail() < _free_space_min_avail && disk_space() > 0) {
|
|
hal.console->printf("Out of space for logging\n");
|
|
stop_logging();
|
|
_open_error = true; // prevent logging starting again
|
|
last_io_operation = "";
|
|
return;
|
|
}
|
|
last_io_operation = "";
|
|
}
|
|
|
|
hal.util->perf_begin(_perf_write);
|
|
|
|
_last_write_time = tnow;
|
|
if (nbytes > _writebuf_chunk) {
|
|
// be kind to the filesystem layer
|
|
nbytes = _writebuf_chunk;
|
|
}
|
|
|
|
uint32_t size;
|
|
const uint8_t *head = _writebuf.readptr(size);
|
|
nbytes = MIN(nbytes, size);
|
|
|
|
// try to align writes on a 512 byte boundary to avoid filesystem reads
|
|
if ((nbytes + _write_offset) % 512 != 0) {
|
|
uint32_t ofs = (nbytes + _write_offset) % 512;
|
|
if (ofs < nbytes) {
|
|
nbytes -= ofs;
|
|
}
|
|
}
|
|
|
|
last_io_operation = "write";
|
|
if (!write_fd_semaphore.take(1)) {
|
|
return;
|
|
}
|
|
if (_write_fd == -1) {
|
|
write_fd_semaphore.give();
|
|
return;
|
|
}
|
|
ssize_t nwritten = AP::FS().write(_write_fd, head, nbytes);
|
|
last_io_operation = "";
|
|
if (nwritten <= 0) {
|
|
if ((tnow - _last_write_ms)/1000U > unsigned(_front._params.file_timeout)) {
|
|
// if we can't write for LOG_FILE_TIMEOUT seconds we give up and close
|
|
// the file. This allows us to cope with temporary write
|
|
// failures caused by directory listing
|
|
hal.util->perf_count(_perf_errors);
|
|
last_io_operation = "close";
|
|
AP::FS().close(_write_fd);
|
|
last_io_operation = "";
|
|
_write_fd = -1;
|
|
_initialised = false;
|
|
printf("Failed to write to File: %s\n", strerror(errno));
|
|
}
|
|
_last_write_failed = true;
|
|
} else {
|
|
_last_write_failed = false;
|
|
_last_write_ms = tnow;
|
|
_write_offset += nwritten;
|
|
_writebuf.advance(nwritten);
|
|
/*
|
|
the best strategy for minimizing corruption on microSD cards
|
|
seems to be to write in 4k chunks and fsync the file on each
|
|
chunk, ensuring the directory entry is updated after each
|
|
write.
|
|
*/
|
|
#if CONFIG_HAL_BOARD != HAL_BOARD_SITL && CONFIG_HAL_BOARD_SUBTYPE != HAL_BOARD_SUBTYPE_LINUX_NONE
|
|
last_io_operation = "fsync";
|
|
AP::FS().fsync(_write_fd);
|
|
last_io_operation = "";
|
|
#endif
|
|
|
|
#if CONFIG_HAL_BOARD == HAL_BOARD_CHIBIOS
|
|
// ChibiOS does not update mtime on writes, so if we opened
|
|
// without knowing the time we should update it later
|
|
if (_need_rtc_update) {
|
|
uint64_t utc_usec;
|
|
if (AP::rtc().get_utc_usec(utc_usec)) {
|
|
AP::FS().set_mtime(_write_filename, utc_usec/(1000U*1000U));
|
|
_need_rtc_update = false;
|
|
}
|
|
}
|
|
#endif
|
|
}
|
|
|
|
write_fd_semaphore.give();
|
|
hal.util->perf_end(_perf_write);
|
|
}
|
|
|
|
bool AP_Logger_File::io_thread_alive() const
|
|
{
|
|
// if the io thread hasn't had a heartbeat in a full seconds then it is dead
|
|
// this is enough time for a sdcard remount
|
|
return (AP_HAL::millis() - _io_timer_heartbeat) < 3000U;
|
|
}
|
|
|
|
bool AP_Logger_File::logging_failed() const
|
|
{
|
|
if (!_initialised) {
|
|
return true;
|
|
}
|
|
if (_open_error) {
|
|
return true;
|
|
}
|
|
if (!io_thread_alive()) {
|
|
// No heartbeat in a second. IO thread is dead?! Very Not
|
|
// Good.
|
|
return true;
|
|
}
|
|
if (_last_write_failed) {
|
|
return true;
|
|
}
|
|
|
|
return false;
|
|
}
|
|
|
|
#endif // HAVE_FILESYSTEM_SUPPORT
|
|
|