+/*
+ * Copyright (C) 2008 Andre Noll <maan@systemlinux.org>
+ *
+ * Licensed under the GPL v2. For licencing details see COPYING.
+ */
#include <string.h>
#include <stdlib.h>
#include <stdarg.h>
#include "signal.h"
#include "df.h"
#include "time.h"
+#include "snap.h"
-
-struct gengetopt_args_info conf;
-char *dss_error_txt = NULL;
+/** Command line and config file options. */
+static struct gengetopt_args_info conf;
+/** Non-NULL if we log to a file. */
static FILE *logfile;
+/** The read end of the signal pipe */
static int signal_pipe;
-
/** Process id of current rsync process. */
static pid_t rsync_pid;
/** Whether the rsync process is currently stopped */
/** Process id of current rm process. */
static pid_t rm_pid;
/** When the next snapshot is due. */
-struct timeval next_snapshot_time;
-/* Creation time of the snapshot currently being created. */
-int64_t current_snapshot_creation_time;
+static struct timeval next_snapshot_time;
+/** The pid of the pre-create hook. */
+static pid_t pre_create_hook_pid;
+/** The pid of the post-create hook. */
+static pid_t post_create_hook_pid;
+/** Creation time of the snapshot currently being created. */
+static int64_t current_snapshot_creation_time;
+/** Needed by the post-create hook. */
+static char *path_to_last_complete_snapshot;
+/** \sa \ref snap.h for details. */
+static unsigned snapshot_creation_status;
DEFINE_DSS_ERRLIST;
COMMAND(create) \
COMMAND(prune) \
COMMAND(run)
-#define COMMAND(x) int com_ ##x(void);
+#define COMMAND(x) static int com_ ##x(void);
COMMANDS
#undef COMMAND
#define COMMAND(x) if (conf.x ##_given) return com_ ##x();
-int call_command_handler(void)
+static int call_command_handler(void)
{
COMMANDS
DSS_EMERG_LOG("BUG: did not find command handler\n");
- exit(EXIT_FAILURE);
+ return -E_BUG;
}
#undef COMMAND
#undef COMMANDS
-/*
- * complete, not being deleted: 1204565370-1204565371.Sun_Mar_02_2008_14_33-Sun_Mar_02_2008_14_43
- * complete, being deleted: 1204565370-1204565371.being_deleted
- * incomplete, not being deleted: 1204565370-incomplete
- * incomplete, being deleted: 1204565370-incomplete.being_deleted
- */
-enum snapshot_status_flags {
- /** The rsync process terminated successfully. */
- SS_COMPLETE = 1,
- /** The rm process is running to remove this snapshot. */
- SS_BEING_DELETED = 2,
-};
-
-struct snapshot {
- char *name;
- int64_t creation_time;
- int64_t completion_time;
- enum snapshot_status_flags flags;
- unsigned interval;
-};
-
-/*
- * An edge snapshot is either the oldest one or the newest one.
+/**
+ * The log function of dss.
+ *
+ * \param ll Loglevel.
+ * \param fml Usual format string.
*
- * We need to find either of them occasionally: The create code
- * needs to know the newest snapshot because that is the one
- * used as the link destination dir. The pruning code needs to
- * find the oldest one in case disk space becomes low.
+ * All DSS_XXX_LOG() macros use this function.
*/
-struct edge_snapshot_data {
- int64_t now;
- struct snapshot snap;
-};
-
__printf_2_3 void dss_log(int ll, const char* fmt,...)
{
va_list argp;
/**
* Print a message either to stdout or to the log file.
*/
-__printf_1_2 void dss_msg(const char* fmt,...)
+static __printf_1_2 void dss_msg(const char* fmt,...)
{
FILE *outfd = conf.daemon_given? logfile : stdout;
va_list argp;
va_end(argp);
}
-/**
- * Return the desired number of snapshots of an interval.
- */
-unsigned num_snapshots(int interval)
+static int disk_space_low(void)
{
- unsigned n;
+ struct disk_space ds;
+ int ret = get_disk_space(".", &ds);
- assert(interval >= 0);
+ if (ret < 0)
+ return ret;
+ if (conf.min_free_mb_arg)
+ if (ds.free_mb < conf.min_free_mb_arg)
+ return 1;
+ if (conf.min_free_percent_arg)
+ if (ds.percent_free < conf.min_free_percent_arg)
+ return 1;
+ if (conf.min_free_percent_inodes_arg)
+ if (ds.percent_free_inodes < conf.min_free_percent_inodes_arg)
+ return 1;
+ return 0;
+}
- if (interval >= conf.num_intervals_arg)
- return 0;
- n = conf.num_intervals_arg - interval - 1;
- return 1 << n;
+static void dss_get_snapshot_list(struct snapshot_list *sl)
+{
+ get_snapshot_list(sl, conf.unit_interval_arg, conf.num_intervals_arg);
}
-/* return: Whether dirname is a snapshot directory (0: no, 1: yes) */
-int is_snapshot(const char *dirname, int64_t now, struct snapshot *s)
+static void compute_next_snapshot_time(void)
{
+ struct timeval now, unit_interval = {.tv_sec = 24 * 3600 * conf.unit_interval_arg},
+ tmp, diff;
+ int64_t x = 0;
+ unsigned wanted = desired_number_of_snapshots(0, conf.num_intervals_arg),
+ num_complete_snapshots = 0;
int i, ret;
- char *dash, *dot, *tmp;
- int64_t num;
+ struct snapshot *s = NULL;
+ struct snapshot_list sl;
- assert(dirname);
- dash = strchr(dirname, '-');
- if (!dash || !dash[1] || dash == dirname)
- return 0;
- for (i = 0; dirname[i] != '-'; i++)
- if (!isdigit(dirname[i]))
- return 0;
- tmp = dss_strdup(dirname);
- tmp[i] = '\0';
- ret = dss_atoi64(tmp, &num);
- free(tmp);
- if (ret < 0) {
- free(dss_error_txt);
- return 0;
- }
- assert(num >= 0);
- if (num > now)
- return 0;
- s->creation_time = num;
- //DSS_DEBUG_LOG("%s start time: %lli\n", dirname, (long long)s->creation_time);
- s->interval = (long long) ((now - s->creation_time)
- / conf.unit_interval_arg / 24 / 3600);
- if (!strcmp(dash + 1, "incomplete")) {
- s->completion_time = -1;
- s->flags = 0; /* neither complete, nor being deleted */
- goto success;
- }
- if (!strcmp(dash + 1, "incomplete.being_deleted")) {
- s->completion_time = -1;
- s->flags = SS_BEING_DELETED; /* mot cpmplete, being deleted */
- goto success;
- }
- tmp = dash + 1;
- dot = strchr(tmp, '.');
- if (!dot || !dot[1] || dot == tmp)
- return 0;
- for (i = 0; tmp[i] != '.'; i++)
- if (!isdigit(tmp[i]))
- return 0;
- tmp = dss_strdup(dash + 1);
- tmp[i] = '\0';
- ret = dss_atoi64(tmp, &num);
- free(tmp);
- if (ret < 0) {
- free(dss_error_txt);
- return 0;
+ assert(snapshot_creation_status == SCS_READY);
+ current_snapshot_creation_time = 0;
+ dss_get_snapshot_list(&sl);
+ FOR_EACH_SNAPSHOT(s, i, &sl) {
+ if (!(s->flags & SS_COMPLETE))
+ continue;
+ num_complete_snapshots++;
+ x += s->completion_time - s->creation_time;
}
- if (num > now)
- return 0;
- s->completion_time = num;
- s->flags = SS_COMPLETE;
- if (!strcmp(dot + 1, "being_deleted"))
- s->flags |= SS_BEING_DELETED;
-success:
- s->name = dss_strdup(dirname);
- return 1;
+ assert(x >= 0);
+ if (num_complete_snapshots)
+ x /= num_complete_snapshots; /* avg time to create one snapshot */
+ x *= wanted; /* time to create all snapshots in interval 0 */
+ tmp.tv_sec = x;
+ tmp.tv_usec = 0;
+ ret = tv_diff(&unit_interval, &tmp, &diff); /* total sleep time per unit interval */
+ gettimeofday(&now, NULL);
+ if (ret < 0 || !s)
+ goto min_sleep;
+ tv_divide(wanted, &diff, &tmp); /* sleep time betweeen two snapshots */
+ diff.tv_sec = s->completion_time;
+ diff.tv_usec = 0;
+ tv_add(&diff, &tmp, &next_snapshot_time);
+ if (tv_diff(&now, &next_snapshot_time, NULL) < 0)
+ goto out;
+min_sleep:
+ next_snapshot_time = now;
+ next_snapshot_time.tv_sec += 60;
+out:
+ free_snapshot_list(&sl);
}
-int64_t get_current_time(void)
-{
- time_t now;
- time(&now);
- DSS_DEBUG_LOG("now: %lli\n", (long long) now);
- return (int64_t)now;
-}
-char *incomplete_name(int64_t start)
+static int remove_snapshot(struct snapshot *s)
{
- return make_message("%lli-incomplete", (long long)start);
+ int fds[3] = {0, 0, 0};
+ assert(!rm_pid);
+ char *new_name = being_deleted_name(s);
+ int ret = dss_rename(s->name, new_name);
+ char *argv[] = {"rm", "-rf", new_name, NULL};
+
+ if (ret < 0)
+ goto out;
+ DSS_NOTICE_LOG("removing %s (interval = %i)\n", s->name, s->interval);
+ ret = dss_exec(&rm_pid, argv[0], argv, fds);
+out:
+ free(new_name);
+ return ret;
}
-char *being_deleted_name(struct snapshot *s)
+/*
+ * return: 0: no redundant snapshots, 1: rm process started, negative: error
+ */
+static int remove_redundant_snapshot(struct snapshot_list *sl)
{
- if (s->flags & SS_COMPLETE)
- return make_message("%lli-%lli.being_deleted",
- (long long)s->creation_time,
- (long long)s->completion_time);
- return make_message("%lli-incomplete.being_deleted",
- (long long)s->creation_time);
+ int ret, i, interval;
+ struct snapshot *s;
+ unsigned missing = 0;
+
+ DSS_DEBUG_LOG("looking for intervals containing too many snapshots\n");
+ for (interval = conf.num_intervals_arg - 1; interval >= 0; interval--) {
+ unsigned keep = desired_number_of_snapshots(interval, conf.num_intervals_arg);
+ unsigned num = sl->interval_count[interval];
+ struct snapshot *victim = NULL, *prev = NULL;
+ int64_t score = LONG_MAX;
+
+ if (keep >= num)
+ missing += keep - num;
+// DSS_DEBUG_LOG("interval %i: keep: %u, have: %u, missing: %u\n",
+// interval, keep, num, missing);
+ if (keep + missing >= num)
+ continue;
+ /* redundant snapshot in this interval, pick snapshot with lowest score */
+ FOR_EACH_SNAPSHOT(s, i, sl) {
+ int64_t this_score;
+
+ //DSS_DEBUG_LOG("checking %s\n", s->name);
+ if (s->interval > interval) {
+ prev = s;
+ continue;
+ }
+ if (s->interval < interval)
+ break;
+ if (!victim) {
+ victim = s;
+ prev = s;
+ continue;
+ }
+ assert(prev);
+ /* check if s is a better victim */
+ this_score = s->creation_time - prev->creation_time;
+ assert(this_score >= 0);
+ //DSS_DEBUG_LOG("%s: score %lli\n", s->name, (long long)score);
+ if (this_score < score) {
+ score = this_score;
+ victim = s;
+ }
+ prev = s;
+ }
+ assert(victim);
+ if (conf.dry_run_given) {
+ dss_msg("%s would be removed (interval = %i)\n",
+ victim->name, victim->interval);
+ continue;
+ }
+ ret = remove_snapshot(victim);
+ return ret < 0? ret : 1;
+ }
+ return 0;
}
-int complete_name(int64_t start, int64_t end, char **result)
+static int remove_outdated_snapshot(struct snapshot_list *sl)
{
- struct tm start_tm, end_tm;
- time_t *start_seconds = (time_t *) (uint64_t *)&start; /* STFU, gcc */
- time_t *end_seconds = (time_t *) (uint64_t *)&end; /* STFU, gcc */
- char start_str[200], end_str[200];
-
- if (!localtime_r(start_seconds, &start_tm)) {
- make_err_msg("%lli", (long long)start);
- return -E_LOCALTIME;
- }
- if (!localtime_r(end_seconds, &end_tm)) {
- make_err_msg("%lli", (long long)end);
- return -E_LOCALTIME;
- }
- if (!strftime(start_str, sizeof(start_str), "%a_%b_%d_%Y_%H_%M_%S", &start_tm)) {
- make_err_msg("%lli", (long long)start);
- return -E_STRFTIME;
- }
- if (!strftime(end_str, sizeof(end_str), "%a_%b_%d_%Y_%H_%M_%S", &end_tm)) {
- make_err_msg("%lli", (long long)end);
- return -E_STRFTIME;
+ int i, ret;
+ struct snapshot *s;
+
+ DSS_DEBUG_LOG("looking for snapshots belonging to intervals greater than %d\n",
+ conf.num_intervals_arg);
+ FOR_EACH_SNAPSHOT(s, i, sl) {
+ if (s->interval <= conf.num_intervals_arg)
+ continue;
+ if (conf.dry_run_given) {
+ dss_msg("%s would be removed (interval = %i)\n",
+ s->name, s->interval);
+ continue;
+ }
+ ret = remove_snapshot(s);
+ if (ret < 0)
+ return ret;
+ return 1;
}
- *result = make_message("%lli-%lli.%s-%s", (long long) start, (long long) end,
- start_str, end_str);
- return 1;
+ return 0;
}
-struct snapshot_list {
- int64_t now;
- unsigned num_snapshots;
- unsigned array_size;
- struct snapshot **snapshots;
- /**
- * Array of size num_intervals + 1
- *
- * It contains the number of snapshots in each interval. interval_count[num_intervals]
- * is the number of snapshots which belong to any interval greater than num_intervals.
- */
- unsigned *interval_count;
-};
-
-#define FOR_EACH_SNAPSHOT(s, i, sl) \
- for ((i) = 0; (i) < (sl)->num_snapshots && ((s) = (sl)->snapshots[(i)]); (i)++)
+static int remove_oldest_snapshot(struct snapshot_list *sl)
+{
+ struct snapshot *s = get_oldest_snapshot(sl);
+ if (!s) /* no snapshot found */
+ return 0;
+ DSS_INFO_LOG("oldest snapshot: %s\n", s->name);
+ if (s->creation_time == current_snapshot_creation_time)
+ return 0; /* do not remove the snapshot currently being created */
+ return remove_snapshot(s);
+}
+static int rename_incomplete_snapshot(int64_t start)
+{
+ char *old_name;
+ int ret;
-#define NUM_COMPARE(x, y) ((int)((x) < (y)) - (int)((x) > (y)))
+ free(path_to_last_complete_snapshot);
+ ret = complete_name(start, get_current_time(),
+ &path_to_last_complete_snapshot);
+ if (ret < 0)
+ return ret;
+ old_name = incomplete_name(start);
+ ret = dss_rename(old_name, path_to_last_complete_snapshot);
+ if (ret >= 0)
+ DSS_NOTICE_LOG("%s -> %s\n", old_name,
+ path_to_last_complete_snapshot);
+ free(old_name);
+ return ret;
+}
-static int compare_snapshots(const void *a, const void *b)
+static int try_to_free_disk_space(int low_disk_space)
{
- struct snapshot *s1 = *(struct snapshot **)a;
- struct snapshot *s2 = *(struct snapshot **)b;
- return NUM_COMPARE(s2->creation_time, s1->creation_time);
-}
+ int ret;
+ struct snapshot_list sl;
-/** Compute the minimum of \a a and \a b. */
-#define DSS_MIN(a,b) ((a) < (b) ? (a) : (b))
+ if (!low_disk_space && conf.keep_redundant_given)
+ return 0;
+ dss_get_snapshot_list(&sl);
+ ret = remove_outdated_snapshot(&sl);
+ if (ret) /* error, or we are removing something */
+ goto out;
+ /* no outdated snapshot */
+ ret = remove_redundant_snapshot(&sl);
+ if (ret)
+ goto out;
+ ret = 0;
+ if (!low_disk_space)
+ goto out;
+ DSS_WARNING_LOG("disk space low and nothing obvious to remove\n");
+ ret = remove_oldest_snapshot(&sl);
+ if (ret)
+ goto out;
+ DSS_CRIT_LOG("uhuhu: not enough disk space for a single snapshot\n");
+ ret = -ERRNO_TO_DSS_ERROR(ENOSPC);
+out:
+ free_snapshot_list(&sl);
+ return ret;
+}
-int add_snapshot(const char *dirname, void *private)
+static int pre_create_hook(void)
{
- struct snapshot_list *sl = private;
- struct snapshot s;
- int ret = is_snapshot(dirname, sl->now, &s);
+ int ret, fds[3] = {0, 0, 0};
- if (!ret)
- return 1;
- if (sl->num_snapshots >= sl->array_size) {
- sl->array_size = 2 * sl->array_size + 1;
- sl->snapshots = dss_realloc(sl->snapshots,
- sl->array_size * sizeof(struct snapshot *));
+ if (!conf.pre_create_hook_given) {
+ snapshot_creation_status = SCS_PRE_HOOK_SUCCESS;
+ return 0;
}
- sl->snapshots[sl->num_snapshots] = dss_malloc(sizeof(struct snapshot));
- *(sl->snapshots[sl->num_snapshots]) = s;
- sl->interval_count[DSS_MIN(s.interval, conf.num_intervals_arg)]++;
- sl->num_snapshots++;
- return 1;
+ DSS_NOTICE_LOG("executing %s\n", conf.pre_create_hook_arg);
+ ret = dss_exec_cmdline_pid(&pre_create_hook_pid,
+ conf.pre_create_hook_arg, fds);
+ if (ret < 0)
+ return ret;
+ snapshot_creation_status = SCS_PRE_HOOK_RUNNING;
+ return ret;
}
-void get_snapshot_list(struct snapshot_list *sl)
+static int post_create_hook(void)
{
- sl->now = get_current_time();
- sl->num_snapshots = 0;
- sl->array_size = 0;
- sl->snapshots = NULL;
- sl->interval_count = dss_calloc((conf.num_intervals_arg + 1) * sizeof(unsigned));
- for_each_subdir(add_snapshot, sl);
- qsort(sl->snapshots, sl->num_snapshots, sizeof(struct snapshot *),
- compare_snapshots);
+ int ret, fds[3] = {0, 0, 0};
+ char *cmd;
+
+ if (!conf.post_create_hook_given) {
+ snapshot_creation_status = SCS_READY;
+ compute_next_snapshot_time();
+ return 0;
+ }
+ cmd = make_message("%s %s", conf.post_create_hook_arg,
+ path_to_last_complete_snapshot);
+ DSS_NOTICE_LOG("executing %s\n", cmd);
+ ret = dss_exec_cmdline_pid(&post_create_hook_pid, cmd, fds);
+ free(cmd);
+ if (ret < 0)
+ return ret;
+ snapshot_creation_status = SCS_POST_HOOK_RUNNING;
+ return ret;
}
-void free_snapshot_list(struct snapshot_list *sl)
+static void kill_process(pid_t pid)
{
- int i;
- struct snapshot *s;
-
- FOR_EACH_SNAPSHOT(s, i, sl) {
- free(s->name);
- free(s);
- }
- free(sl->interval_count);
- sl->interval_count = NULL;
- free(sl->snapshots);
- sl->snapshots = NULL;
- sl->num_snapshots = 0;
+ if (!pid)
+ return;
+ DSS_WARNING_LOG("sending SIGTERM to pid %d\n", (int)pid);
+ kill(pid, SIGTERM);
}
-void stop_rsync_process(void)
+static void stop_rsync_process(void)
{
if (!rsync_pid || rsync_stopped)
return;
rsync_stopped = 1;
}
-void restart_rsync_process(void)
+static void restart_rsync_process(void)
{
if (!rsync_pid || !rsync_stopped)
return;
rsync_stopped = 0;
}
+
/**
* Print a log message about the exit status of a child.
*/
-void log_termination_msg(pid_t pid, int status)
+static void log_termination_msg(pid_t pid, int status)
{
if (WIFEXITED(status))
DSS_INFO_LOG("child %i exited. Exit status: %i\n", (int)pid,
DSS_WARNING_LOG("child %i terminated abormally\n", (int)pid);
}
-int wait_for_process(pid_t pid, int *status)
+static int wait_for_process(pid_t pid, int *status)
{
int ret;
kill(pid, SIGTERM);
}
if (ret < 0)
- make_err_msg("failed to wait for process %d", (int)pid);
+ DSS_ERROR_LOG("failed to wait for process %d\n", (int)pid);
else
log_termination_msg(pid, *status);
return ret;
}
-int remove_snapshot(struct snapshot *s)
+static int handle_rm_exit(int status)
{
- int fds[3] = {0, 0, 0};
- assert(!rm_pid);
- char *new_name = being_deleted_name(s);
- int ret = dss_rename(s->name, new_name);
- char *argv[] = {"rm", "-rf", new_name, NULL};
-
- if (ret < 0)
- goto out;
- DSS_NOTICE_LOG("removing %s (interval = %i)\n", s->name, s->interval);
- stop_rsync_process();
- ret = dss_exec(&rm_pid, argv[0], argv, fds);
-out:
- free(new_name);
- return ret;
-}
-
-/*
- * return: 0: no redundant snapshots, 1: rm process started, negative: error
- */
-int remove_redundant_snapshot(struct snapshot_list *sl)
-{
- int ret, i, interval;
- struct snapshot *s;
- unsigned missing = 0;
-
- DSS_INFO_LOG("looking for intervals containing too many snapshots\n");
- for (interval = conf.num_intervals_arg - 1; interval >= 0; interval--) {
- unsigned keep = num_snapshots(interval);
- unsigned num = sl->interval_count[interval];
- struct snapshot *victim = NULL, *prev = NULL;
- int64_t score = LONG_MAX;
-
- if (keep >= num)
- missing += keep - num;
- DSS_DEBUG_LOG("interval %i: keep: %u, have: %u, missing: %u\n",
- interval, keep, num, missing);
- if (keep + missing >= num)
- continue;
- /* redundant snapshot in this interval, pick snapshot with lowest score */
- FOR_EACH_SNAPSHOT(s, i, sl) {
- int64_t this_score;
-
- //DSS_DEBUG_LOG("checking %s\n", s->name);
- if (s->interval > interval) {
- prev = s;
- continue;
- }
- if (s->interval < interval)
- break;
- if (!victim) {
- victim = s;
- prev = s;
- continue;
- }
- assert(prev);
- /* check if s is a better victim */
- this_score = s->creation_time - prev->creation_time;
- assert(this_score >= 0);
- //DSS_DEBUG_LOG("%s: score %lli\n", s->name, (long long)score);
- if (this_score < score) {
- score = this_score;
- victim = s;
- }
- prev = s;
- }
- assert(victim);
- if (conf.dry_run_given) {
- dss_msg("%s would be removed (interval = %i)\n",
- victim->name, victim->interval);
- continue;
- }
- ret = remove_snapshot(victim);
- return ret < 0? ret : 1;
- }
- return 0;
+ rm_pid = 0;
+ if (!WIFEXITED(status))
+ return -E_INVOLUNTARY_EXIT;
+ if (WEXITSTATUS(status))
+ return -E_BAD_EXIT_CODE;
+ return 1;
}
-int remove_outdated_snapshot(struct snapshot_list *sl)
+static int wait_for_rm_process(void)
{
- int i, ret;
- struct snapshot *s;
+ int status, ret = wait_for_process(rm_pid, &status);
- DSS_INFO_LOG("looking for snapshots belonging to intervals greater than %d\n",
- conf.num_intervals_arg);
- FOR_EACH_SNAPSHOT(s, i, sl) {
- if (s->interval <= conf.num_intervals_arg)
- continue;
- if (conf.dry_run_given) {
- dss_msg("%s would be removed (interval = %i)\n",
- s->name, s->interval);
- continue;
- }
- ret = remove_snapshot(s);
- if (ret < 0)
- return ret;
- return 1;
+ if (ret < 0)
+ return ret;
+ return handle_rm_exit(status);
+}
+
+static int handle_rsync_exit(int status)
+{
+ int es, ret;
+
+ if (!WIFEXITED(status)) {
+ DSS_ERROR_LOG("rsync process %d died involuntary\n", (int)rsync_pid);
+ ret = -E_INVOLUNTARY_EXIT;
+ snapshot_creation_status = SCS_READY;
+ compute_next_snapshot_time();
+ goto out;
}
- return 0;
+ es = WEXITSTATUS(status);
+ if (es != 0 && es != 23 && es != 24) {
+ DSS_ERROR_LOG("rsync process %d returned %d\n", (int)rsync_pid, es);
+ ret = -E_BAD_EXIT_CODE;
+ snapshot_creation_status = SCS_READY;
+ compute_next_snapshot_time();
+ goto out;
+ }
+ ret = rename_incomplete_snapshot(current_snapshot_creation_time);
+ if (ret < 0)
+ goto out;
+ snapshot_creation_status = SCS_RSYNC_SUCCESS;
+out:
+ rsync_pid = 0;
+ rsync_stopped = 0;
+ return ret;
}
-int handle_rm_exit(int status)
+static int handle_pre_create_hook_exit(int status)
{
int es, ret;
if (!WIFEXITED(status)) {
- make_err_msg("rm process %d died involuntary", (int)rm_pid);
+ snapshot_creation_status = SCS_READY;
+ compute_next_snapshot_time();
ret = -E_INVOLUNTARY_EXIT;
goto out;
}
es = WEXITSTATUS(status);
if (es) {
- make_err_msg("rm process %d returned %d", (int)rm_pid, es);
+ snapshot_creation_status = SCS_READY;
+ compute_next_snapshot_time();
ret = -E_BAD_EXIT_CODE;
goto out;
}
+ snapshot_creation_status = SCS_PRE_HOOK_SUCCESS;
ret = 1;
- rm_pid = 0;
out:
+ pre_create_hook_pid = 0;
return ret;
}
-int wait_for_rm_process(void)
+static int handle_sigchld(void)
{
- int status, ret = wait_for_process(rm_pid, &status);
+ pid_t pid;
+ int status, ret = reap_child(&pid, &status);
- if (ret < 0)
+ if (ret <= 0)
return ret;
- return handle_rm_exit(status);
-}
-
-void kill_process(pid_t pid)
-{
- if (!pid)
- return;
- DSS_WARNING_LOG("sending SIGTERM to pid %d\n", (int)pid);
- kill(pid, SIGTERM);
+ if (pid == rsync_pid)
+ return handle_rsync_exit(status);
+ if (pid == rm_pid)
+ return handle_rm_exit(status);
+ if (pid == pre_create_hook_pid)
+ return handle_pre_create_hook_exit(status);
+ if (pid == post_create_hook_pid) {
+ snapshot_creation_status = SCS_READY;
+ compute_next_snapshot_time();
+ return 1;
+ }
+ DSS_EMERG_LOG("BUG: unknown process %d died\n", (int)pid);
+ return -E_BUG;
}
-int check_config(void)
+static int check_config(void)
{
if (conf.unit_interval_arg <= 0) {
- make_err_msg("bad unit interval: %i", conf.unit_interval_arg);
+ DSS_ERROR_LOG("bad unit interval: %i\n", conf.unit_interval_arg);
return -E_INVALID_NUMBER;
}
DSS_DEBUG_LOG("unit interval: %i day(s)\n", conf.unit_interval_arg);
if (conf.num_intervals_arg <= 0) {
- make_err_msg("bad number of intervals %i", conf.num_intervals_arg);
+ DSS_ERROR_LOG("bad number of intervals %i\n", conf.num_intervals_arg);
return -E_INVALID_NUMBER;
}
DSS_DEBUG_LOG("number of intervals: %i\n", conf.num_intervals_arg);
return 1;
}
-/* exits on errors */
-void parse_config_file(int override)
+static int parse_config_file(int override)
{
int ret;
char *config_file;
ret = stat(config_file, &statbuf);
if (ret && conf.config_file_given) {
ret = -ERRNO_TO_DSS_ERROR(errno);
- make_err_msg("failed to stat config file %s", config_file);
+ DSS_ERROR_LOG("failed to stat config file %s\n", config_file);
goto out;
}
if (!ret) {
.override = override,
.initialize = 0,
.check_required = 1,
- .check_ambiguity = 0
+ .check_ambiguity = 0,
+ .print_errors = 1
};
cmdline_parser_config_file(config_file, &conf, ¶ms);
}
logfile = open_log(conf.logfile_arg);
log_welcome(conf.loglevel_arg);
}
- ret = dss_chdir(conf.dest_dir_arg);
+ DSS_DEBUG_LOG("loglevel: %d\n", conf.loglevel_arg);
+// cmdline_parser_dump(logfile? logfile : stdout, &conf);
out:
free(config_file);
- if (ret >= 0)
- return;
- log_err_msg(EMERG, -ret);
- exit(EXIT_FAILURE);
+ if (ret < 0)
+ DSS_EMERG_LOG("%s\n", dss_strerror(-ret));
+ return ret;
}
-void handle_sighup(void)
+static int change_to_dest_dir(void)
{
- DSS_NOTICE_LOG("SIGHUP\n");
- parse_config_file(1);
+ DSS_INFO_LOG("changing cwd to %s\n", conf.dest_dir_arg);
+ return dss_chdir(conf.dest_dir_arg);
}
-int rename_incomplete_snapshot(int64_t start)
+static int handle_sighup(void)
{
- char *old_name, *new_name;
int ret;
- ret = complete_name(start, get_current_time(), &new_name);
+ DSS_NOTICE_LOG("SIGHUP\n");
+ ret = parse_config_file(1);
if (ret < 0)
return ret;
- old_name = incomplete_name(start);
- ret = dss_rename(old_name, new_name);
- if (ret >= 0)
- DSS_NOTICE_LOG("%s -> %s\n", old_name, new_name);
- free(old_name);
- free(new_name);
- return ret;
+ return change_to_dest_dir();
}
-int handle_rsync_exit(int status)
+static int handle_signal(void)
{
- int es, ret;
+ int sig, ret = next_signal();
- if (!WIFEXITED(status)) {
- make_err_msg("rsync process %d died involuntary", (int)rsync_pid);
- ret = -E_INVOLUNTARY_EXIT;
- goto out;
- }
- es = WEXITSTATUS(status);
- if (es != 0 && es != 23 && es != 24) {
- make_err_msg("rsync process %d returned %d", (int)rsync_pid, es);
- ret = -E_BAD_EXIT_CODE;
+ if (ret <= 0)
goto out;
+ sig = ret;
+ switch (sig) {
+ case SIGINT:
+ case SIGTERM:
+ restart_rsync_process();
+ kill_process(rsync_pid);
+ kill_process(rm_pid);
+ ret = -E_SIGNAL;
+ break;
+ case SIGHUP:
+ ret = handle_sighup();
+ break;
+ case SIGCHLD:
+ ret = handle_sigchld();
+ break;
}
- ret = rename_incomplete_snapshot(current_snapshot_creation_time);
out:
- rsync_pid = 0;
- current_snapshot_creation_time = 0;
- rsync_stopped = 0;
+ if (ret < 0)
+ DSS_ERROR_LOG("%s\n", dss_strerror(-ret));
return ret;
}
-int get_newest_complete(const char *dirname, void *private)
-{
- struct edge_snapshot_data *esd = private;
- struct snapshot s;
- int ret = is_snapshot(dirname, esd->now, &s);
-
- if (ret <= 0)
- return 1;
- if (s.flags != SS_COMPLETE) /* incomplete or being deleted */
- return 1;
- if (s.creation_time < esd->snap.creation_time)
- return 1;
- free(esd->snap.name);
- esd->snap = s;
- return 1;
-}
-
-__malloc char *name_of_newest_complete_snapshot(void)
-{
- struct edge_snapshot_data esd = {
- .now = get_current_time(),
- .snap = {.creation_time = -1}
- };
- for_each_subdir(get_newest_complete, &esd);
- return esd.snap.name;
-}
-
-void create_rsync_argv(char ***argv, int64_t *num)
+static void create_rsync_argv(char ***argv, int64_t *num)
{
- char *logname, *newest = name_of_newest_complete_snapshot();
+ char *logname, *newest;
int i = 0, j;
+ struct snapshot_list sl;
+
+ dss_get_snapshot_list(&sl);
+ newest = name_of_newest_complete_snapshot(&sl);
+ free_snapshot_list(&sl);
*argv = dss_malloc((15 + conf.rsync_option_given) * sizeof(char *));
(*argv)[i++] = dss_strdup("rsync");
free(newest);
} else
DSS_INFO_LOG("no previous snapshot found\n");
- if (conf.exclude_patterns_given) {
- (*argv)[i++] = dss_strdup("--exclude-from");
- (*argv)[i++] = dss_strdup(conf.exclude_patterns_arg);
-
- }
logname = dss_logname();
if (conf.remote_user_given && !strcmp(conf.remote_user_arg, logname))
(*argv)[i++] = dss_strdup(conf.source_dir_arg);
DSS_DEBUG_LOG("argv[%d] = %s\n", j, (*argv)[j]);
}
-void free_rsync_argv(char **argv)
+static void free_rsync_argv(char **argv)
{
int i;
for (i = 0; argv[i]; i++)
free(argv);
}
-int create_snapshot(char **argv)
+static int create_snapshot(char **argv)
{
- int fds[3] = {0, 0, 0};
- char *name = incomplete_name(current_snapshot_creation_time);
+ int ret, fds[3] = {0, 0, 0};
+ char *name;
+ name = incomplete_name(current_snapshot_creation_time);
DSS_NOTICE_LOG("creating new snapshot %s\n", name);
free(name);
- return dss_exec(&rsync_pid, argv[0], argv, fds);
-}
-
-void compute_next_snapshot_time(struct snapshot_list *sl)
-{
- struct timeval now, unit_interval = {.tv_sec = 24 * 3600 * conf.unit_interval_arg},
- tmp, diff;
- int64_t x = 0;
- unsigned wanted = num_snapshots(0), num_complete_snapshots = 0;
- int i, ret;
- struct snapshot *s;
-
- gettimeofday(&now, NULL);
- FOR_EACH_SNAPSHOT(s, i, sl) {
- if (!(s->flags & SS_COMPLETE))
- continue;
- num_complete_snapshots++;
- x += s->completion_time - s->creation_time;
- }
- assert(x >= 0);
- if (num_complete_snapshots)
- x /= num_complete_snapshots; /* avg time to create one snapshot */
- x *= wanted; /* time to create all snapshots in interval 0 */
- tmp.tv_sec = x;
- tmp.tv_usec = 0;
- ret = tv_diff(&unit_interval, &tmp, &diff); /* time between creation */
- if (ret < 0) {
- next_snapshot_time = now;
- return;
- }
- tv_divide(wanted, &diff, &tmp);
- tv_add(&now, &tmp, &next_snapshot_time);
-}
-
-void handle_signal(void)
-{
- int sig, ret = next_signal();
-
- if (ret <= 0)
- goto out;
- sig = ret;
- switch (sig) {
- int status;
- pid_t pid;
- case SIGINT:
- case SIGTERM:
- restart_rsync_process();
- kill_process(rsync_pid);
- kill_process(rm_pid);
- exit(EXIT_FAILURE);
- case SIGHUP:
- handle_sighup();
- ret = 1;
- break;
- case SIGCHLD:
- ret = reap_child(&pid, &status);
- if (ret <= 0)
- break;
- assert(pid == rsync_pid || pid == rm_pid);
- if (pid == rsync_pid)
- ret = handle_rsync_exit(status);
- else
- ret = handle_rm_exit(status);
- }
-out:
- if (ret < 0)
- log_err_msg(ERROR, -ret);
-}
-
-int get_oldest(const char *dirname, void *private)
-{
- struct edge_snapshot_data *esd = private;
- struct snapshot s;
- int ret = is_snapshot(dirname, esd->now, &s);
-
- if (ret <= 0)
- return 1;
- if (s.creation_time > esd->snap.creation_time)
- return 1;
- free(esd->snap.name);
- esd->snap = s;
- return 1;
-}
-
-int remove_oldest_snapshot()
-{
- int ret;
- struct edge_snapshot_data esd = {
- .now = get_current_time(),
- .snap = {.creation_time = LLONG_MAX}
- };
- for_each_subdir(get_oldest, &esd);
- if (!esd.snap.name) /* no snapshot found */
- return 0;
- DSS_INFO_LOG("oldest snapshot: %s\n", esd.snap.name);
- ret = 0;
- if (esd.snap.creation_time == current_snapshot_creation_time)
- goto out; /* do not remove the snapshot currently being created */
- ret = remove_snapshot(&esd.snap);
-out:
- free(esd.snap.name);
- return ret;
-}
-
-/* TODO: Also consider number of inodes. */
-int disk_space_low(void)
-{
- struct disk_space ds;
- int ret = get_disk_space(".", &ds);
-
+ ret = dss_exec(&rsync_pid, argv[0], argv, fds);
if (ret < 0)
return ret;
- if (conf.min_free_mb_arg)
- if (ds.free_mb < conf.min_free_mb_arg)
- return 1;
- if (conf.min_free_percent_arg)
- if (ds.percent_free < conf.min_free_percent_arg)
- return 1;
- return 0;
-}
-
-int try_to_free_disk_space(int low_disk_space, struct snapshot_list *sl)
-{
- int ret;
-
- ret = remove_outdated_snapshot(sl);
- if (ret) /* error, or we are removing something */
- return ret;
- /* no outdated snapshot */
- ret = remove_redundant_snapshot(sl);
- if (ret)
- return ret;
- if (!low_disk_space)
- return 0;
- DSS_WARNING_LOG("disk space low and nothing obvious to remove\n");
- ret = remove_oldest_snapshot();
- if (ret)
- return ret;
- make_err_msg("uhuhu: not enough disk space for a single snapshot");
- return -ENOSPC;
+ snapshot_creation_status = SCS_RSYNC_RUNNING;
+ return ret;
}
-int select_loop(void)
+static int select_loop(void)
{
int ret;
- struct timeval tv = {.tv_sec = 0, .tv_usec = 0};
- struct snapshot_list sl = {.num_snapshots = 0};
+ /* check every 60 seconds for free disk space */
+ struct timeval tv;
for (;;) {
- struct timeval now, *tvp = &tv;
fd_set rfds;
int low_disk_space;
char **rsync_argv;
+ struct timeval now, *tvp;
- free_snapshot_list(&sl);
- get_snapshot_list(&sl);
- compute_next_snapshot_time(&sl);
+ if (rm_pid)
+ tvp = NULL; /* sleep until rm process dies */
+ else { /* sleep one minute */
+ tv.tv_sec = 60;
+ tv.tv_usec = 0;
+ tvp = &tv;
+ }
FD_ZERO(&rfds);
FD_SET(signal_pipe, &rfds);
- if (rsync_pid)
- tv.tv_sec = 60;
- else if (rm_pid)
- tvp = NULL;
+ DSS_DEBUG_LOG("tvp: %p, tv_sec : %lu\n", tvp, (long unsigned) tv.tv_sec);
ret = dss_select(signal_pipe + 1, &rfds, NULL, tvp);
if (ret < 0)
- return ret;
+ goto out;
if (FD_ISSET(signal_pipe, &rfds)) {
- handle_signal();
- continue;
+ ret = handle_signal();
+ if (ret < 0)
+ goto out;
}
if (rm_pid)
continue;
ret = disk_space_low();
if (ret < 0)
- break;
+ goto out;
low_disk_space = ret;
- if (low_disk_space)
- stop_rsync_process();
- ret = try_to_free_disk_space(low_disk_space, &sl);
+ ret = try_to_free_disk_space(low_disk_space);
if (ret < 0)
- break;
- if (rm_pid)
- continue;
- if (rsync_pid) {
- restart_rsync_process();
+ goto out;
+ if (rm_pid) {
+ stop_rsync_process();
continue;
}
- /* neither rsync nor rm are running. Start rsync? */
+ restart_rsync_process();
gettimeofday(&now, NULL);
- if (tv_diff(&next_snapshot_time, &now, &tv) > 0)
+ if (tv_diff(&next_snapshot_time, &now, NULL) > 0)
continue;
- create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
- ret = create_snapshot(rsync_argv);
- free_rsync_argv(rsync_argv);
- if (ret < 0)
- break;
+ switch (snapshot_creation_status) {
+ case SCS_READY:
+ ret = pre_create_hook();
+ if (ret < 0)
+ goto out;
+ continue;
+ case SCS_PRE_HOOK_RUNNING:
+ continue;
+ case SCS_PRE_HOOK_SUCCESS:
+ create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
+ ret = create_snapshot(rsync_argv);
+ free_rsync_argv(rsync_argv);
+ if (ret < 0)
+ goto out;
+ continue;
+ case SCS_RSYNC_RUNNING:
+ continue;
+ case SCS_RSYNC_SUCCESS:
+ ret = post_create_hook();
+ if (ret < 0)
+ goto out;
+ continue;
+ case SCS_POST_HOOK_RUNNING:
+ continue;
+ }
}
- free_snapshot_list(&sl);
+out:
return ret;
}
-int com_run(void)
+static void exit_hook(int exit_code)
+{
+ int fds[3] = {0, 0, 0};
+ char *cmd;
+ pid_t pid;
+
+ if (!conf.exit_hook_given)
+ return;
+ cmd = make_message("%s %s", conf.exit_hook_arg,
+ dss_strerror(-exit_code));
+ DSS_NOTICE_LOG("executing %s\n", cmd);
+ dss_exec_cmdline_pid(&pid, cmd, fds);
+ free(cmd);
+}
+
+static int com_run(void)
{
int ret;
if (conf.dry_run_given) {
- make_err_msg("dry_run not supported by this command");
+ DSS_ERROR_LOG("dry_run not supported by this command\n");
return -E_SYNTAX;
}
ret = install_sighandler(SIGHUP);
if (ret < 0)
return ret;
- return select_loop();
-}
-
-void log_disk_space(struct disk_space *ds)
-{
- DSS_INFO_LOG("free: %uM/%uM (%u%%), %u%% inodes unused\n",
- ds->free_mb, ds->total_mb, ds->percent_free,
- ds->percent_free_inodes);
+ compute_next_snapshot_time();
+ ret = select_loop();
+ if (ret >= 0) /* impossible */
+ ret = -E_BUG;
+ exit_hook(ret);
+ return ret;
}
-int com_prune(void)
+static int com_prune(void)
{
int ret;
struct snapshot_list sl;
return ret;
log_disk_space(&ds);
for (;;) {
- get_snapshot_list(&sl);
+ dss_get_snapshot_list(&sl);
ret = remove_outdated_snapshot(&sl);
free_snapshot_list(&sl);
if (ret < 0)
goto out;
}
for (;;) {
- get_snapshot_list(&sl);
+ dss_get_snapshot_list(&sl);
ret = remove_redundant_snapshot(&sl);
free_snapshot_list(&sl);
if (ret < 0)
return ret;
}
-int com_create(void)
+static int com_create(void)
{
int ret, status;
char **rsync_argv;
- create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
if (conf.dry_run_given) {
int i;
char *msg = NULL;
+ create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
for (i = 0; rsync_argv[i]; i++) {
char *tmp = msg;
msg = make_message("%s%s%s", tmp? tmp : "",
tmp? " " : "", rsync_argv[i]);
free(tmp);
}
+ free_rsync_argv(rsync_argv);
dss_msg("%s\n", msg);
free(msg);
return 1;
}
+ ret = pre_create_hook();
+ if (ret < 0)
+ return ret;
+ if (pre_create_hook_pid) {
+ ret = wait_for_process(pre_create_hook_pid, &status);
+ if (ret < 0)
+ return ret;
+ ret = handle_pre_create_hook_exit(status);
+ if (ret < 0)
+ return ret;
+ }
+ create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
ret = create_snapshot(rsync_argv);
if (ret < 0)
goto out;
if (ret < 0)
goto out;
ret = handle_rsync_exit(status);
+ if (ret < 0)
+ goto out;
+ post_create_hook();
+ if (post_create_hook_pid)
+ ret = wait_for_process(post_create_hook_pid, &status);
out:
free_rsync_argv(rsync_argv);
return ret;
}
-int com_ls(void)
+static int com_ls(void)
{
int i;
struct snapshot_list sl;
struct snapshot *s;
- get_snapshot_list(&sl);
+
+ dss_get_snapshot_list(&sl);
FOR_EACH_SNAPSHOT(s, i, &sl)
dss_msg("%u\t%s\n", s->interval, s->name);
free_snapshot_list(&sl);
return 1;
}
-__noreturn void clean_exit(int status)
-{
- free(dss_error_txt);
- exit(status);
-}
-static void setup_signal_handling(void)
+static int setup_signal_handling(void)
{
int ret;
signal_pipe = signal_init(); /* always successful */
ret = install_sighandler(SIGINT);
if (ret < 0)
- goto err;
+ return ret;
ret = install_sighandler(SIGTERM);
if (ret < 0)
- goto err;
- ret = install_sighandler(SIGCHLD);
- if (ret < 0)
- goto err;
- return;
-err:
- DSS_EMERG_LOG("could not install signal handlers\n");
- exit(EXIT_FAILURE);
+ return ret;
+ return install_sighandler(SIGCHLD);
}
+/**
+ * The main function of dss.
+ *
+ * \param argc Usual argument count.
+ * \param argv Usual argument vector.
+ */
int main(int argc, char **argv)
{
int ret;
.override = 0,
.initialize = 1,
.check_required = 0,
- .check_ambiguity = 0
+ .check_ambiguity = 0,
+ .print_errors = 1
};
cmdline_parser_ext(argc, argv, &conf, ¶ms); /* aborts on errors */
- parse_config_file(0);
-
+ ret = parse_config_file(0);
+ if (ret < 0)
+ goto out;
if (conf.daemon_given)
daemon_init();
- setup_signal_handling();
+ ret = change_to_dest_dir();
+ if (ret < 0)
+ goto out;
+ ret = setup_signal_handling();
+ if (ret < 0)
+ goto out;
ret = call_command_handler();
+out:
if (ret < 0)
- log_err_msg(EMERG, -ret);
- clean_exit(ret >= 0? EXIT_SUCCESS : EXIT_FAILURE);
+ DSS_EMERG_LOG("%s\n", dss_strerror(-ret));
+ exit(ret >= 0? EXIT_SUCCESS : EXIT_FAILURE);
}