+/* SPDX-License-Identifier: GPL-2.0 */
#include <string.h>
#include <stdlib.h>
+#include <stdio.h>
#include <stdarg.h>
#include <assert.h>
#include <errno.h>
#include <sys/types.h>
#include <signal.h>
#include <ctype.h>
+#include <stdbool.h>
#include <sys/stat.h>
#include <unistd.h>
#include <inttypes.h>
#include <sys/wait.h>
#include <fnmatch.h>
#include <limits.h>
-
+#include <fcntl.h>
+#include <lopsub.h>
+#include <sys/mman.h>
#include "gcc-compat.h"
-#include "cmdline.h"
#include "log.h"
-#include "string.h"
-#include "error.h"
-#include "fd.h"
+#include "str.h"
+#include "err.h"
+#include "file.h"
#include "exec.h"
#include "daemon.h"
-#include "signal.h"
+#include "sig.h"
#include "df.h"
-#include "time.h"
-
+#include "tv.h"
+#include "snap.h"
+#include "ipc.h"
+#include "dss.lsg.h"
+
+#define CMD_PTR(_cname) lls_cmd(LSG_DSS_CMD_ ## _cname, dss_suite)
+#define OPT_RESULT(_cname, _oname) (lls_opt_result(\
+ LSG_DSS_ ## _cname ## _OPT_ ## _oname, (CMD_PTR(_cname) == CMD_PTR(DSS))? lpr : sublpr))
+#define OPT_GIVEN(_cname, _oname) (lls_opt_given(OPT_RESULT(_cname, _oname)))
+#define OPT_STRING_VAL(_cname, _oname) (lls_string_val(0, \
+ OPT_RESULT(_cname, _oname)))
+#define OPT_UINT32_VAL(_cname, _oname) (lls_uint32_val(0, \
+ OPT_RESULT(_cname, _oname)))
+
+struct dss_user_data {int (*handler)(void);};
+#define EXPORT_CMD_HANDLER(_cmd) const struct dss_user_data \
+ lsg_dss_com_ ## _cmd ## _user_data = { \
+ .handler = com_ ## _cmd \
+ };
-struct gengetopt_args_info conf;
-char *dss_error_txt = NULL;
+/*
+ * Command line and active options. We need to keep a copy of the parsed
+ * command line options for the SIGHUP case where we merge the command line
+ * options and the new config file options.
+ */
+static struct lls_parse_result *cmdline_lpr, *lpr;
+
+/** Parsed subcommand options. */
+static struct lls_parse_result *cmdline_sublpr, *sublpr;
+/* The executing subcommand (NULL at startup). */
+static const struct lls_command *subcmd;
+/** Wether daemon_init() was called. */
+static bool daemonized;
+/** 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 */
-static int rsync_stopped;
-/** Process id of current rm process. */
-static pid_t rm_pid;
+/** Process id of current pre-create-hook/rsync/post-create-hook process. */
+static pid_t create_pid;
+/** Whether the pre-create-hook/rsync/post-create-hook is currently stopped. */
+static int create_process_stopped;
+/** How many times in a row the rsync command failed. */
+static int num_consecutive_rsync_errors;
+/** Process id of current pre-remove/rm/post-remove process. */
+static pid_t remove_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 int64_t next_snapshot_time;
+/** When to try to remove something. */
+static struct timeval next_removal_check;
+/** Creation time of the snapshot currently being created. */
+static int64_t current_snapshot_creation_time;
+/* Set by the pre-rm hook, cleared by handle_remove_exit(). */
+struct snapshot *snapshot_currently_being_removed;
+/** Needed by the post-create hook. */
+static char *path_to_last_complete_snapshot;
+static char *name_of_reference_snapshot;
+/** \sa \ref snap.h for details. */
+enum hook_status snapshot_creation_status;
+/** \sa \ref snap.h for details. */
+enum hook_status snapshot_removal_status;
DEFINE_DSS_ERRLIST;
+static const char *hook_status_description[] = {HOOK_STATUS_ARRAY};
+
+/* may be called with ds == NULL. */
+static int disk_space_low(struct disk_space *ds)
+{
+ struct disk_space ds_struct;
+ uint32_t val;
+ if (!ds) {
+ int ret = get_disk_space(".", &ds_struct);
+ if (ret < 0)
+ return ret;
+ ds = &ds_struct;
+ }
+ val = OPT_UINT32_VAL(DSS, MIN_FREE_MB);
+ if (val != 0)
+ if (ds->free_mb < val)
+ return 1;
+ val = OPT_UINT32_VAL(DSS, MIN_FREE_PERCENT);
+ if (val != 0)
+ if (ds->percent_free < val)
+ return 1;
+ val = OPT_UINT32_VAL(DSS, MIN_FREE_PERCENT_INODES);
+ if (val != 0)
+ if (ds->percent_free_inodes < val)
+ return 1;
+ return 0;
+}
-/* a litte cpp magic helps to DRY */
-#define COMMANDS \
- COMMAND(ls) \
- COMMAND(create) \
- COMMAND(prune) \
- COMMAND(run)
-#define COMMAND(x) int com_ ##x(void);
-COMMANDS
-#undef COMMAND
-#define COMMAND(x) if (conf.x ##_given) return com_ ##x();
-int call_command_handler(void)
+static void dump_dss_config(const char *msg)
{
- COMMANDS
- DSS_EMERG_LOG("BUG: did not find command handler\n");
- exit(EXIT_FAILURE);
+ const char dash[] = "-----------------------------";
+ char *lopsub_dump;
+ int ret;
+ FILE *log = logfile? logfile : stderr;
+ struct disk_space ds;
+ int64_t now = get_current_time();
+
+ if (OPT_UINT32_VAL(DSS, LOGLEVEL) > INFO)
+ return;
+
+ fprintf(log, "%s <%s config> %s\n", dash, msg, dash);
+ fprintf(log, "\n*** disk space ***\n\n");
+ ret = get_disk_space(".", &ds);
+ if (ret >= 0) {
+ DSS_INFO_LOG(("disk space low: %s\n", disk_space_low(&ds)?
+ "yes" : "no"));
+ log_disk_space(&ds);
+ } else
+ DSS_ERROR_LOG(("can not get free disk space: %s\n",
+ dss_strerror(-ret)));
+
+ /* we continue on errors from get_disk_space */
+
+ fprintf(log, "\n*** non-default options ***\n\n");
+ lopsub_dump = lls_dump_parse_result(lpr, CMD_PTR(DSS), true);
+ fprintf(log, "%s", lopsub_dump);
+ free(lopsub_dump);
+ fprintf(log, "\n*** non-default options for \"run\" ***\n\n");
+ lopsub_dump = lls_dump_parse_result(lpr, CMD_PTR(RUN), true);
+ fprintf(log, "%s", lopsub_dump);
+ free(lopsub_dump);
+ fprintf(log, "\n*** internal state ***\n\n");
+ fprintf(log,
+ "pid: %d\n"
+ "logfile: %s\n"
+ "snapshot_currently_being_removed: %s\n"
+ "path_to_last_complete_snapshot: %s\n"
+ "reference_snapshot: %s\n"
+ "snapshot_creation_status: %s\n"
+ "snapshot_removal_status: %s\n"
+ "num_consecutive_rsync_errors: %d\n"
+ ,
+ (int) getpid(),
+ logfile? OPT_STRING_VAL(RUN, LOGFILE) : "stderr",
+ snapshot_currently_being_removed?
+ snapshot_currently_being_removed->name : "(none)",
+ path_to_last_complete_snapshot?
+ path_to_last_complete_snapshot : "(none)",
+ name_of_reference_snapshot?
+ name_of_reference_snapshot : "(none)",
+ hook_status_description[snapshot_creation_status],
+ hook_status_description[snapshot_removal_status],
+ num_consecutive_rsync_errors
+ );
+ if (create_pid != 0)
+ fprintf(log,
+ "create_pid: %" PRId32 "\n"
+ "create process is %sstopped\n"
+ ,
+ create_pid,
+ create_process_stopped? "" : "not "
+ );
+ if (remove_pid != 0)
+ fprintf(log, "remove_pid: %" PRId32 "\n", remove_pid);
+ if (next_snapshot_time != 0)
+ fprintf(log, "next snapshot due in %" PRId64 " seconds\n",
+ next_snapshot_time - now);
+ if (current_snapshot_creation_time != 0)
+ fprintf(log, "current_snapshot_creation_time: %"
+ PRId64 " (%" PRId64 " seconds ago)\n",
+ current_snapshot_creation_time,
+ now - current_snapshot_creation_time
+ );
+ if (next_removal_check.tv_sec != 0) {
+ fprintf(log, "next removal check: %llu (%llu seconds ago)\n",
+ (long long unsigned)next_removal_check.tv_sec,
+ now - (long long unsigned)next_removal_check.tv_sec
+ );
+
+ }
+ fprintf(log, "%s </%s config> %s\n", dash, msg, dash);
}
-#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,
-};
+static int loglevel = -1;
+static const char *location_file = NULL;
+static int location_line = -1;
+static const char *location_func = NULL;
-struct snapshot {
- char *name;
- int64_t creation_time;
- int64_t completion_time;
- enum snapshot_status_flags flags;
- unsigned interval;
-};
+void dss_log_set_params(int ll, const char *file, int line, const char *func)
+{
+ loglevel = ll;
+ location_file = file;
+ location_line = line;
+ location_func = func;
+}
-/*
- * 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,...)
+__printf_1_2 void dss_log(const char* fmt,...)
{
va_list argp;
FILE *outfd;
struct tm *tm;
time_t t1;
char str[255] = "";
+ int lpr_ll = lpr? OPT_UINT32_VAL(DSS, LOGLEVEL) : WARNING;
- if (ll < conf.loglevel_arg)
+ if (loglevel < lpr_ll)
return;
outfd = logfile? logfile : stderr;
- time(&t1);
- tm = localtime(&t1);
- strftime(str, sizeof(str), "%b %d %H:%M:%S", tm);
- fprintf(outfd, "%s ", str);
- if (conf.loglevel_arg <= INFO)
- fprintf(outfd, "%i: ", ll);
+ if (subcmd == CMD_PTR(RUN)) {
+ time(&t1);
+ tm = localtime(&t1);
+ strftime(str, sizeof(str), "%b %d %H:%M:%S", tm);
+ fprintf(outfd, "%s ", str);
+ if (lpr_ll <= INFO)
+ fprintf(outfd, "%i: ", loglevel);
+ }
+ if (subcmd == CMD_PTR(RUN))
+#ifdef DSS_NO_FUNC_NAMES
+ fprintf(outfd, "%s:%d: ", location_file, location_line);
+#else
+ fprintf(outfd, "%s: ", location_func);
+#endif
va_start(argp, fmt);
vfprintf(outfd, fmt, argp);
va_end(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;
+ FILE *outfd = logfile? logfile : stdout;
va_list argp;
va_start(argp, fmt);
vfprintf(outfd, fmt, argp);
va_end(argp);
}
-/**
- * Return the desired number of snapshots of an interval.
- */
-unsigned num_snapshots(int interval)
+static char *get_config_file_name(void)
{
- unsigned n;
-
- assert(interval >= 0);
-
- if (interval >= conf.num_intervals_arg)
- return 0;
- n = conf.num_intervals_arg - interval - 1;
- return 1 << n;
+ char *home, *config_file;
+
+ if (OPT_GIVEN(DSS, CONFIG_FILE))
+ return dss_strdup(OPT_STRING_VAL(DSS, CONFIG_FILE));
+ home = get_homedir();
+ config_file = make_message("%s/.dssrc", home);
+ free(home);
+ return config_file;
}
-/* return: Whether dirname is a snapshot directory (0: no, 1: yes) */
-int is_snapshot(const char *dirname, int64_t now, struct snapshot *s)
+static int send_signal(int sig, bool wait)
{
- int i, ret;
- char *dash, *dot, *tmp;
- int64_t num;
+ pid_t pid;
+ char *config_file = get_config_file_name();
+ int ret = get_dss_pid(config_file, &pid);
+ unsigned ms = 32;
+ struct timespec ts;
- 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)
+ free(config_file);
+ if (ret < 0)
+ return ret;
+ if (OPT_GIVEN(DSS, DRY_RUN)) {
+ dss_msg("%d\n", (int)pid);
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;
+ DSS_NOTICE_LOG(("sending signal %d to pid %d\n", sig, (int)pid));
+ ret = kill(pid, sig);
+ if (ret < 0)
+ return -ERRNO_TO_DSS_ERROR(errno);
+ if (!wait)
+ return 1;
+ while (ms < 5000) {
+ ts.tv_sec = ms / 1000;
+ ts.tv_nsec = (ms % 1000) * 1000 * 1000;
+ ret = nanosleep(&ts, NULL);
+ if (ret < 0)
+ return -ERRNO_TO_DSS_ERROR(errno);
+ ret = kill(pid, 0);
+ if (ret < 0) {
+ if (errno != ESRCH)
+ return -ERRNO_TO_DSS_ERROR(errno);
+ return 1;
+ }
+ ms *= 2;
}
- 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;
+ return -E_KILL_TIMEOUT;
}
-int64_t get_current_time(void)
-{
- time_t now;
- time(&now);
- DSS_DEBUG_LOG("now: %lli\n", (long long) now);
- return (int64_t)now;
-}
+struct signal_info {
+ const char * const name;
+ int num;
+};
-char *incomplete_name(int64_t start)
-{
- return make_message("%lli-incomplete", (long long)start);
-}
+/*
+ * The table below was taken 2016 from proc/sig.c of procps-3.2.8. Copyright
+ * 1998-2003 by Albert Cahalan, GPLv2.
+ */
+static const struct signal_info signal_table[] = {
+ {"ABRT", SIGABRT}, /* IOT */
+ {"ALRM", SIGALRM},
+ {"BUS", SIGBUS},
+ {"CHLD", SIGCHLD}, /* CLD */
+ {"CONT", SIGCONT},
+ {"FPE", SIGFPE},
+ {"HUP", SIGHUP},
+ {"ILL", SIGILL},
+ {"INT", SIGINT},
+ {"KILL", SIGKILL},
+ {"PIPE", SIGPIPE},
+#ifdef SIGPOLL
+ {"POLL", SIGPOLL}, /* IO */
+#endif
+ {"PROF", SIGPROF},
+#ifdef SIGPWR
+ {"PWR", SIGPWR},
+#endif
+ {"QUIT", SIGQUIT},
+ {"SEGV", SIGSEGV},
+#ifdef SIGSTKFLT
+ {"STKFLT", SIGSTKFLT},
+#endif
+ {"STOP", SIGSTOP},
+ {"SYS", SIGSYS}, /* UNUSED */
+ {"TERM", SIGTERM},
+ {"TRAP", SIGTRAP},
+ {"TSTP", SIGTSTP},
+ {"TTIN", SIGTTIN},
+ {"TTOU", SIGTTOU},
+ {"URG", SIGURG},
+ {"USR1", SIGUSR1},
+ {"USR2", SIGUSR2},
+ {"VTALRM", SIGVTALRM},
+ {"WINCH", SIGWINCH},
+ {"XCPU", SIGXCPU},
+ {"XFSZ", SIGXFSZ}
+};
-char *being_deleted_name(struct snapshot *s)
-{
- 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);
-}
+#define SIGNAL_TABLE_SIZE (sizeof(signal_table) / sizeof(signal_table[0]))
+#ifndef SIGRTMAX
+#define SIGRTMAX 64
+#endif
-int complete_name(int64_t start, int64_t end, char **result)
+static int com_kill(void)
{
- 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;
+ bool w_given = OPT_GIVEN(KILL, WAIT);
+ const char *arg = OPT_STRING_VAL(KILL, SIGNAL);
+ int ret, i;
+
+ if (*arg >= '0' && *arg <= '9') {
+ int64_t val;
+ ret = dss_atoi64(arg, &val);
+ if (ret < 0)
+ return ret;
+ if (val < 0 || val > SIGRTMAX)
+ return -ERRNO_TO_DSS_ERROR(EINVAL);
+ return send_signal(val, w_given);
}
- *result = make_message("%lli-%lli.%s-%s", (long long) start, (long long) end,
- start_str, end_str);
- return 1;
+ if (strncasecmp(arg, "sig", 3) == 0)
+ arg += 3;
+ if (strcasecmp(arg, "CLD") == 0)
+ return send_signal(SIGCHLD, w_given);
+ if (strcasecmp(arg, "IOT") == 0)
+ return send_signal(SIGABRT, w_given);
+ for (i = 0; i < SIGNAL_TABLE_SIZE; i++)
+ if (strcasecmp(arg, signal_table[i].name) == 0)
+ return send_signal(signal_table[i].num, w_given);
+ DSS_ERROR_LOG(("invalid sigspec: %s\n", arg));
+ return -ERRNO_TO_DSS_ERROR(EINVAL);
}
+EXPORT_CMD_HANDLER(kill);
-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 void dss_get_snapshot_list(struct snapshot_list *sl)
+{
+ get_snapshot_list(sl, OPT_UINT32_VAL(DSS, UNIT_INTERVAL),
+ OPT_UINT32_VAL(DSS, NUM_INTERVALS));
+}
+static int64_t compute_next_snapshot_time(void)
+{
+ int64_t x = 0, now = get_current_time(), unit_interval
+ = 24 * 3600 * OPT_UINT32_VAL(DSS, UNIT_INTERVAL), ret,
+ last_completion_time;
+ unsigned wanted = desired_number_of_snapshots(0,
+ OPT_UINT32_VAL(DSS, NUM_INTERVALS)),
+ num_complete = 0;
+ int i;
+ struct snapshot *s = NULL;
+ struct snapshot_list sl;
+ dss_get_snapshot_list(&sl);
+ FOR_EACH_SNAPSHOT(s, i, &sl) {
+ if (!(s->flags & SS_COMPLETE))
+ continue;
+ num_complete++;
+ x += s->completion_time - s->creation_time;
+ last_completion_time = s->completion_time;
+ }
+ assert(x >= 0);
-#define NUM_COMPARE(x, y) ((int)((x) < (y)) - (int)((x) > (y)))
+ ret = now;
+ if (num_complete == 0)
+ goto out;
+ x /= num_complete; /* avg time to create one snapshot */
+ if (unit_interval < x * wanted) /* oops, no sleep at all */
+ goto out;
+ ret = last_completion_time + unit_interval / wanted - x;
+out:
+ free_snapshot_list(&sl);
+ return ret;
+}
-static int compare_snapshots(const void *a, const void *b)
+static inline void invalidate_next_snapshot_time(void)
{
- struct snapshot *s1 = *(struct snapshot **)a;
- struct snapshot *s2 = *(struct snapshot **)b;
- return NUM_COMPARE(s2->creation_time, s1->creation_time);
+ next_snapshot_time = 0;
}
-/** Compute the minimum of \a a and \a b. */
-#define DSS_MIN(a,b) ((a) < (b) ? (a) : (b))
+static inline int next_snapshot_time_is_valid(void)
+{
+ return next_snapshot_time != 0;
+}
-int add_snapshot(const char *dirname, void *private)
+static int next_snapshot_is_due(void)
{
- struct snapshot_list *sl = private;
- struct snapshot s;
- int ret = is_snapshot(dirname, sl->now, &s);
+ int64_t now = get_current_time();
- if (!ret)
+ if (!next_snapshot_time_is_valid())
+ next_snapshot_time = compute_next_snapshot_time();
+ if (next_snapshot_time <= now) {
+ DSS_DEBUG_LOG(("next snapshot: now\n"));
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 *));
}
- 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_DEBUG_LOG(("next snapshot due in %" PRId64 " seconds\n",
+ next_snapshot_time - now));
+ return 0;
}
-void get_snapshot_list(struct snapshot_list *sl)
+static void pre_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);
+ assert(snapshot_creation_status == HS_READY);
+ /* make sure that the next snapshot time will be recomputed */
+ invalidate_next_snapshot_time();
+ DSS_DEBUG_LOG(("executing %s\n", OPT_STRING_VAL(DSS, PRE_CREATE_HOOK)));
+ dss_exec_cmdline_pid(&create_pid, OPT_STRING_VAL(DSS, PRE_CREATE_HOOK));
+ snapshot_creation_status = HS_PRE_RUNNING;
}
-void free_snapshot_list(struct snapshot_list *sl)
+static void pre_remove_hook(struct snapshot *s, const char *why)
{
- int i;
- struct snapshot *s;
+ char *cmd;
- 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;
-}
-
-void stop_rsync_process(void)
-{
- if (!rsync_pid || rsync_stopped)
+ if (!s)
return;
- kill(SIGSTOP, rsync_pid);
- rsync_stopped = 1;
+ DSS_DEBUG_LOG(("%s snapshot %s\n", why, s->name));
+ assert(snapshot_removal_status == HS_READY);
+ assert(remove_pid == 0);
+ assert(!snapshot_currently_being_removed);
+
+ snapshot_currently_being_removed = dss_malloc(sizeof(struct snapshot));
+ *snapshot_currently_being_removed = *s;
+ snapshot_currently_being_removed->name = dss_strdup(s->name);
+
+ cmd = make_message("%s %s/%s", OPT_STRING_VAL(DSS, PRE_REMOVE_HOOK),
+ OPT_STRING_VAL(DSS, DEST_DIR), s->name);
+ DSS_DEBUG_LOG(("executing %s\n", cmd));
+ dss_exec_cmdline_pid(&remove_pid, cmd);
+ free(cmd);
+ snapshot_removal_status = HS_PRE_RUNNING;
}
-void restart_rsync_process(void)
+static int exec_rm(void)
{
- if (!rsync_pid || !rsync_stopped)
- return;
- kill (SIGCONT, rsync_pid);
- rsync_stopped = 0;
+ struct snapshot *s = snapshot_currently_being_removed;
+ char *new_name = being_deleted_name(s);
+ char *argv[4];
+ int ret;
+
+ argv[0] = "rm";
+ argv[1] = "-rf";
+ argv[2] = new_name;
+ argv[3] = NULL;
+
+ assert(snapshot_removal_status == HS_PRE_SUCCESS);
+ assert(remove_pid == 0);
+
+ DSS_NOTICE_LOG(("removing %s (interval = %i)\n", s->name, s->interval));
+ ret = dss_rename(s->name, new_name);
+ if (ret < 0)
+ goto out;
+ dss_exec(&remove_pid, argv[0], argv);
+ snapshot_removal_status = HS_RUNNING;
+out:
+ free(new_name);
+ return ret;
}
-/**
- * Print a log message about the exit status of a child.
- */
-void log_termination_msg(pid_t pid, int status)
+static int snapshot_is_being_created(struct snapshot *s)
{
- if (WIFEXITED(status))
- DSS_INFO_LOG("child %i exited. Exit status: %i\n", (int)pid,
- WEXITSTATUS(status));
- else if (WIFSIGNALED(status))
- DSS_NOTICE_LOG("child %i was killed by signal %i\n", (int)pid,
- WTERMSIG(status));
- else
- DSS_WARNING_LOG("child %i terminated abormally\n", (int)pid);
+ return s->creation_time == current_snapshot_creation_time;
}
-int wait_for_process(pid_t pid, int *status)
+static struct snapshot *find_orphaned_snapshot(struct snapshot_list *sl)
{
- int ret;
-
- DSS_DEBUG_LOG("Waiting for process %d to terminate\n", (int)pid);
- for (;;) {
- fd_set rfds;
+ struct snapshot *s;
+ int i;
- FD_ZERO(&rfds);
- FD_SET(signal_pipe, &rfds);
- ret = dss_select(signal_pipe + 1, &rfds, NULL, NULL);
- if (ret < 0)
- break;
- ret = next_signal();
- if (!ret)
+ DSS_DEBUG_LOG(("looking for old incomplete snapshots\n"));
+ FOR_EACH_SNAPSHOT(s, i, sl) {
+ if (snapshot_is_being_created(s))
continue;
- if (ret == SIGCHLD) {
- ret = waitpid(pid, status, 0);
- if (ret >= 0)
- break;
- if (errno != EINTR) { /* error */
- ret = -ERRNO_TO_DSS_ERROR(errno);
- break;
- }
- }
- /* SIGINT or SIGTERM */
- DSS_WARNING_LOG("sending SIGTERM to pid %d\n", (int)pid);
- kill(pid, SIGTERM);
+ /*
+ * We know that no rm is currently running, so if s is marked
+ * as being deleted, a previously started rm must have failed.
+ */
+ if (s->flags & SS_BEING_DELETED)
+ return s;
+
+ if (s->flags & SS_COMPLETE) /* good snapshot */
+ continue;
+ /*
+ * This snapshot is incomplete and it is not the snapshot
+ * currently being created. However, we must not remove it if
+ * rsync is about to be restarted. As only the newest snapshot
+ * can be restarted, this snapshot is orphaned if it is not the
+ * newest snapshot or if we are not about to restart rsync.
+ */
+ if (get_newest_snapshot(sl) != s)
+ return s;
+ if (snapshot_creation_status != HS_NEEDS_RESTART)
+ return s;
}
- if (ret < 0)
- make_err_msg("failed to wait for process %d", (int)pid);
- else
- log_termination_msg(pid, *status);
- return ret;
+ /* no orphaned snapshots */
+ return NULL;
}
-int remove_snapshot(struct snapshot *s)
+static int is_reference_snapshot(struct snapshot *s)
{
- 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;
+ if (!name_of_reference_snapshot)
+ return 0;
+ return strcmp(s->name, name_of_reference_snapshot)? 0 : 1;
}
/*
* return: 0: no redundant snapshots, 1: rm process started, negative: error
*/
-int remove_redundant_snapshot(struct snapshot_list *sl)
+static struct snapshot *find_redundant_snapshot(struct snapshot_list *sl)
{
- int ret, i, interval;
+ int i, interval;
struct snapshot *s;
unsigned missing = 0;
+ uint32_t N = OPT_UINT32_VAL(DSS, NUM_INTERVALS);
- 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);
+ DSS_DEBUG_LOG(("looking for intervals containing too many snapshots\n"));
+ for (interval = N - 1; interval >= 0; interval--) {
+ unsigned keep = desired_number_of_snapshots(interval, N);
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 (snapshot_is_being_created(s))
+ continue;
+ if (is_reference_snapshot(s))
+ continue;
if (s->interval > interval) {
prev = s;
continue;
/* 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 victim;
}
- return 0;
+ return NULL;
}
-int remove_outdated_snapshot(struct snapshot_list *sl)
+static struct snapshot *find_outdated_snapshot(struct snapshot_list *sl)
{
- int i, ret;
+ int i;
struct snapshot *s;
- DSS_INFO_LOG("looking for snapshots belonging to intervals greater than %d\n",
- conf.num_intervals_arg);
+ DSS_DEBUG_LOG(("looking for snapshots belonging to intervals >= %d\n",
+ OPT_UINT32_VAL(DSS, NUM_INTERVALS)));
FOR_EACH_SNAPSHOT(s, i, sl) {
- if (s->interval <= conf.num_intervals_arg)
+ if (snapshot_is_being_created(s))
continue;
- if (conf.dry_run_given) {
- dss_msg("%s would be removed (interval = %i)\n",
- s->name, s->interval);
+ if (is_reference_snapshot(s))
continue;
- }
- ret = remove_snapshot(s);
- if (ret < 0)
- return ret;
- return 1;
+ if (s->interval < OPT_UINT32_VAL(DSS, NUM_INTERVALS))
+ continue;
+ return s;
}
- return 0;
+ return NULL;
}
-int handle_rm_exit(int status)
+static struct snapshot *find_oldest_removable_snapshot(struct snapshot_list *sl)
{
- int es, ret;
+ int i, num_complete;
+ struct snapshot *s, *ref = NULL;
- if (!WIFEXITED(status)) {
- make_err_msg("rm process %d died involuntary", (int)rm_pid);
- ret = -E_INVOLUNTARY_EXIT;
- goto out;
- }
- es = WEXITSTATUS(status);
- if (es) {
- make_err_msg("rm process %d returned %d", (int)rm_pid, es);
- ret = -E_BAD_EXIT_CODE;
- goto out;
+ num_complete = num_complete_snapshots(sl);
+ if (num_complete <= OPT_UINT32_VAL(DSS, MIN_COMPLETE))
+ return NULL;
+ FOR_EACH_SNAPSHOT(s, i, sl) {
+ if (snapshot_is_being_created(s))
+ continue;
+ if (is_reference_snapshot(s)) { /* avoid this one */
+ ref = s;
+ continue;
+ }
+ DSS_INFO_LOG(("oldest removable snapshot: %s\n", s->name));
+ return s;
}
- ret = 1;
- rm_pid = 0;
-out:
+ assert(ref);
+ DSS_WARNING_LOG(("removing reference snapshot %s\n", ref->name));
+ return ref;
+}
+
+static int rename_incomplete_snapshot(int64_t start)
+{
+ char *old_name;
+ int ret;
+ int64_t now;
+
+ /*
+ * We don't want the dss_rename() below to fail with EEXIST because the
+ * last complete snapshot was created (and completed) in the same
+ * second as this one.
+ */
+ while ((now = get_current_time()) == start)
+ sleep(1);
+ free(path_to_last_complete_snapshot);
+ ret = complete_name(start, now, &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;
}
-int wait_for_rm_process(void)
+static int try_to_free_disk_space(void)
{
- int status, ret = wait_for_process(rm_pid, &status);
+ int ret;
+ struct snapshot_list sl;
+ struct snapshot *victim;
+ struct timeval now;
+ const char *why;
+ int low_disk_space;
+ ret = disk_space_low(NULL);
if (ret < 0)
return ret;
- return handle_rm_exit(status);
+ low_disk_space = ret;
+ gettimeofday(&now, NULL);
+ if (tv_diff(&next_removal_check, &now, NULL) > 0)
+ return 0;
+ if (!low_disk_space) {
+ if (OPT_GIVEN(DSS, KEEP_REDUNDANT))
+ return 0;
+ if (snapshot_creation_status != HS_READY)
+ return 0;
+ if (next_snapshot_is_due())
+ return 0;
+ }
+ /*
+ * Idle and --keep_redundant not given, or low disk space. Look at
+ * existing snapshots.
+ */
+ dss_get_snapshot_list(&sl);
+ ret = 0;
+ /*
+ * Don't remove anything if there is free space and we have fewer
+ * snapshots than configured, plus one. This way there is always one
+ * snapshot that can be recycled.
+ */
+ if (!low_disk_space && sl.num_snapshots <=
+ 1 << OPT_UINT32_VAL(DSS, NUM_INTERVALS))
+ goto out;
+ why = "outdated";
+ victim = find_outdated_snapshot(&sl);
+ if (victim)
+ goto remove;
+ why = "redundant";
+ victim = find_redundant_snapshot(&sl);
+ if (victim)
+ goto remove;
+ why = "orphaned";
+ victim = find_orphaned_snapshot(&sl);
+ if (victim)
+ goto remove;
+ /* try harder only if disk space is low */
+ if (!low_disk_space)
+ goto out;
+ DSS_WARNING_LOG(("disk space low and nothing obvious to remove\n"));
+ why = "oldest";
+ victim = find_oldest_removable_snapshot(&sl);
+ if (victim)
+ goto remove;
+ DSS_CRIT_LOG(("uhuhu: disk space low and nothing to remove\n"));
+ ret = -ERRNO_TO_DSS_ERROR(ENOSPC);
+ goto out;
+remove:
+ pre_remove_hook(victim, why);
+out:
+ free_snapshot_list(&sl);
+ return ret;
}
-void kill_process(pid_t pid)
+static void post_create_hook(void)
{
- if (!pid)
- return;
- DSS_WARNING_LOG("sending SIGTERM to pid %d\n", (int)pid);
- kill(pid, SIGTERM);
+ char *cmd = make_message("%s %s/%s",
+ OPT_STRING_VAL(DSS, POST_CREATE_HOOK),
+ OPT_STRING_VAL(DSS, DEST_DIR), path_to_last_complete_snapshot);
+ DSS_NOTICE_LOG(("executing %s\n", cmd));
+ dss_exec_cmdline_pid(&create_pid, cmd);
+ free(cmd);
+ snapshot_creation_status = HS_POST_RUNNING;
}
-int check_config(void)
+static void post_remove_hook(void)
{
- if (conf.unit_interval_arg <= 0) {
- make_err_msg("bad unit interval: %i", 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);
- return -E_INVALID_NUMBER;
- }
- DSS_DEBUG_LOG("number of intervals: %i\n", conf.num_intervals_arg);
- return 1;
+ char *cmd;
+ struct snapshot *s = snapshot_currently_being_removed;
+
+ assert(s);
+
+ cmd = make_message("%s %s/%s", OPT_STRING_VAL(DSS, POST_REMOVE_HOOK),
+ OPT_STRING_VAL(DSS, DEST_DIR), s->name);
+ DSS_NOTICE_LOG(("executing %s\n", cmd));
+ dss_exec_cmdline_pid(&remove_pid, cmd);
+ free(cmd);
+ snapshot_removal_status = HS_POST_RUNNING;
}
-/* exits on errors */
-void parse_config_file(int override)
+static void dss_kill(pid_t pid, int sig, const char *msg)
{
- int ret;
- char *config_file;
- struct stat statbuf;
- char *old_logfile_arg = NULL;
- int old_daemon_given = 0;
-
- if (conf.config_file_given)
- config_file = dss_strdup(conf.config_file_arg);
- else {
- char *home = get_homedir();
- config_file = make_message("%s/.dssrc", home);
- free(home);
- }
- if (override) { /* SIGHUP */
- if (conf.logfile_given)
- old_logfile_arg = dss_strdup(conf.logfile_arg);
- old_daemon_given = conf.daemon_given;
- }
+ const char *signame, *process_name;
- 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);
- goto out;
- }
- if (!ret) {
- struct cmdline_parser_params params = {
- .override = override,
- .initialize = 0,
- .check_required = 1,
- .check_ambiguity = 0
- };
- cmdline_parser_config_file(config_file, &conf, ¶ms);
- }
- ret = check_config();
- if (ret < 0)
- goto out;
- if (override) {
- /* don't change daemon mode on SIGHUP */
- conf.daemon_given = old_daemon_given;
- close_log(logfile);
- logfile = NULL;
- if (conf.logfile_given)
- free(old_logfile_arg);
- else if (conf.daemon_given) { /* re-use old logfile */
- conf.logfile_arg = old_logfile_arg;
- conf.logfile_given = 1;
- }
- }
- if (conf.logfile_given) {
- logfile = open_log(conf.logfile_arg);
- log_welcome(conf.loglevel_arg);
+ if (pid == 0)
+ return;
+ switch (sig) {
+ case SIGTERM: signame = "TERM"; break;
+ case SIGSTOP: signame = "STOP"; break;
+ case SIGCONT: signame = "CONT"; break;
+ default: signame = "????";
}
- ret = dss_chdir(conf.dest_dir_arg);
-out:
- free(config_file);
- if (ret >= 0)
+
+ if (pid == create_pid)
+ process_name = "create";
+ else if (pid == remove_pid)
+ process_name = "remove";
+ else process_name = "??????";
+
+ if (msg)
+ DSS_INFO_LOG(("%s\n", msg));
+ DSS_DEBUG_LOG(("sending signal %d (%s) to pid %d (%s process)\n",
+ sig, signame, (int)pid, process_name));
+ if (kill(pid, sig) >= 0)
return;
- log_err_msg(EMERG, -ret);
- exit(EXIT_FAILURE);
+ DSS_INFO_LOG(("failed to send signal %d (%s) to pid %d (%s process)\n",
+ sig, signame, (int)pid, process_name));
}
-void handle_sighup(void)
+static void stop_create_process(void)
{
- DSS_NOTICE_LOG("SIGHUP\n");
- parse_config_file(1);
+ if (create_process_stopped)
+ return;
+ dss_kill(create_pid, SIGSTOP, "suspending create process");
+ create_process_stopped = 1;
+}
+
+static void restart_create_process(void)
+{
+ if (!create_process_stopped)
+ return;
+ dss_kill(create_pid, SIGCONT, "resuming create process");
+ create_process_stopped = 0;
+}
+
+/**
+ * Print a log message about the exit status of a child.
+ */
+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,
+ WEXITSTATUS(status)));
+ else if (WIFSIGNALED(status))
+ DSS_NOTICE_LOG(("child %i was killed by signal %i\n", (int)pid,
+ WTERMSIG(status)));
+ else
+ DSS_WARNING_LOG(("child %i terminated abormally\n", (int)pid));
}
-int rename_incomplete_snapshot(int64_t start)
+static int wait_for_process(pid_t pid, int *status)
{
- char *old_name, *new_name;
int ret;
- ret = complete_name(start, get_current_time(), &new_name);
+ DSS_DEBUG_LOG(("Waiting for process %d to terminate\n", (int)pid));
+ for (;;) {
+ fd_set rfds;
+
+ FD_ZERO(&rfds);
+ FD_SET(signal_pipe, &rfds);
+ ret = dss_select(signal_pipe + 1, &rfds, NULL, NULL);
+ if (ret < 0)
+ break;
+ ret = next_signal();
+ if (!ret)
+ continue;
+ if (ret == SIGCHLD) {
+ ret = waitpid(pid, status, 0);
+ if (ret >= 0)
+ break;
+ if (errno != EINTR) { /* error */
+ ret = -ERRNO_TO_DSS_ERROR(errno);
+ break;
+ }
+ }
+ /* SIGINT or SIGTERM */
+ dss_kill(pid, SIGTERM, "killing child process");
+ }
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);
+ DSS_ERROR_LOG(("failed to wait for process %d\n", (int)pid));
+ else
+ log_termination_msg(pid, *status);
return ret;
}
-int handle_rsync_exit(int status)
+static void handle_pre_remove_exit(int status)
+{
+ if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {
+ snapshot_removal_status = HS_READY;
+ gettimeofday(&next_removal_check, NULL);
+ next_removal_check.tv_sec += 60;
+ return;
+ }
+ snapshot_removal_status = HS_PRE_SUCCESS;
+}
+
+static int handle_rm_exit(int status)
+{
+ if (!WIFEXITED(status)) {
+ snapshot_removal_status = HS_READY;
+ return -E_INVOLUNTARY_EXIT;
+ }
+ if (WEXITSTATUS(status)) {
+ snapshot_removal_status = HS_READY;
+ return -E_BAD_EXIT_CODE;
+ }
+ snapshot_removal_status = HS_SUCCESS;
+ return 1;
+}
+
+static void handle_post_remove_exit(void)
+{
+ snapshot_removal_status = HS_READY;
+}
+
+static int handle_remove_exit(int status)
+{
+ int ret;
+ struct snapshot *s = snapshot_currently_being_removed;
+
+ assert(s);
+ switch (snapshot_removal_status) {
+ case HS_PRE_RUNNING:
+ handle_pre_remove_exit(status);
+ ret = 1;
+ break;
+ case HS_RUNNING:
+ ret = handle_rm_exit(status);
+ break;
+ case HS_POST_RUNNING:
+ handle_post_remove_exit();
+ ret = 1;
+ break;
+ default:
+ ret = -E_BUG;
+ }
+ if (snapshot_removal_status == HS_READY) {
+ free(s->name);
+ free(s);
+ snapshot_currently_being_removed = NULL;
+ }
+ remove_pid = 0;
+ return ret;
+}
+
+static int wait_for_remove_process(void)
+{
+ int status, ret;
+
+ assert(remove_pid);
+ assert(
+ snapshot_removal_status == HS_PRE_RUNNING ||
+ snapshot_removal_status == HS_RUNNING ||
+ snapshot_removal_status == HS_POST_RUNNING
+ );
+ ret = wait_for_process(remove_pid, &status);
+ if (ret < 0)
+ return ret;
+ return handle_remove_exit(status);
+}
+
+static int handle_rsync_exit(int status)
{
int es, ret;
if (!WIFEXITED(status)) {
- make_err_msg("rsync process %d died involuntary", (int)rsync_pid);
+ DSS_ERROR_LOG(("rsync process %d died involuntary\n", (int)create_pid));
ret = -E_INVOLUNTARY_EXIT;
+ snapshot_creation_status = HS_READY;
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;
+ /*
+ * Restart rsync on non-fatal errors:
+ * 24: Partial transfer due to vanished source files
+ */
+ if (es != 0 && es != 24) {
+ DSS_WARNING_LOG(("rsync exit code %d, error count %d\n",
+ es, ++num_consecutive_rsync_errors));
+ if (!logfile) { /* called by com_run() */
+ ret = -E_BAD_EXIT_CODE;
+ goto out;
+ }
+ if (num_consecutive_rsync_errors >
+ OPT_UINT32_VAL(RUN, MAX_RSYNC_ERRORS)) {
+ ret = -E_TOO_MANY_RSYNC_ERRORS;
+ snapshot_creation_status = HS_READY;
+ goto out;
+ }
+ DSS_WARNING_LOG(("restarting rsync process\n"));
+ snapshot_creation_status = HS_NEEDS_RESTART;
+ next_snapshot_time = get_current_time() + 60;
+ ret = 1;
goto out;
}
+ num_consecutive_rsync_errors = 0;
ret = rename_incomplete_snapshot(current_snapshot_creation_time);
+ if (ret < 0)
+ goto out;
+ snapshot_creation_status = HS_SUCCESS;
+ free(name_of_reference_snapshot);
+ name_of_reference_snapshot = NULL;
+out:
+ create_process_stopped = 0;
+ return ret;
+}
+
+static int handle_pre_create_hook_exit(int status)
+{
+ int es, ret;
+ static int warn_count;
+
+ if (!WIFEXITED(status)) {
+ snapshot_creation_status = HS_READY;
+ ret = -E_INVOLUNTARY_EXIT;
+ goto out;
+ }
+ es = WEXITSTATUS(status);
+ if (es) {
+ if (!warn_count--) {
+ DSS_NOTICE_LOG(("pre_create_hook %s returned %d\n",
+ OPT_STRING_VAL(DSS, PRE_CREATE_HOOK), es));
+ DSS_NOTICE_LOG(("deferring snapshot creation...\n"));
+ warn_count = 60; /* warn only once per hour */
+ }
+ next_snapshot_time = get_current_time() + 60;
+ snapshot_creation_status = HS_READY;
+ ret = 0;
+ goto out;
+ }
+ warn_count = 0;
+ snapshot_creation_status = HS_PRE_SUCCESS;
+ ret = 1;
out:
- rsync_pid = 0;
- current_snapshot_creation_time = 0;
- rsync_stopped = 0;
return ret;
}
-int get_newest_complete(const char *dirname, void *private)
+static int handle_sigchld(void)
{
- struct edge_snapshot_data *esd = private;
- struct snapshot s;
- int ret = is_snapshot(dirname, esd->now, &s);
+ pid_t pid;
+ int status, ret = reap_child(&pid, &status);
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;
+ return ret;
+
+ if (pid == create_pid) {
+ switch (snapshot_creation_status) {
+ case HS_PRE_RUNNING:
+ ret = handle_pre_create_hook_exit(status);
+ break;
+ case HS_RUNNING:
+ ret = handle_rsync_exit(status);
+ break;
+ case HS_POST_RUNNING:
+ snapshot_creation_status = HS_READY;
+ ret = 1;
+ break;
+ default:
+ DSS_EMERG_LOG(("BUG: create can't die in status %d\n",
+ snapshot_creation_status));
+ return -E_BUG;
+ }
+ create_pid = 0;
+ return ret;
+ }
+ if (pid == remove_pid) {
+ ret = handle_remove_exit(status);
+ if (ret < 0)
+ return ret;
+ return ret;
+ }
+ DSS_EMERG_LOG(("BUG: unknown process %d died\n", (int)pid));
+ return -E_BUG;
}
-__malloc char *name_of_newest_complete_snapshot(void)
+/* also checks if . is a mountpoint, if --mountpoint was given */
+static int change_to_dest_dir(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;
+ int ret;
+ const char *dd = OPT_STRING_VAL(DSS, DEST_DIR);
+ struct stat dot, dotdot;
+
+ DSS_INFO_LOG(("changing cwd to %s\n", dd));
+ if (chdir(dd) < 0) {
+ ret = -ERRNO_TO_DSS_ERROR(errno);
+ DSS_ERROR_LOG(("could not change cwd to %s\n", dd));
+ return ret;
+ }
+ if (!OPT_GIVEN(DSS, MOUNTPOINT))
+ return 0;
+ if (stat(".", &dot) < 0) {
+ ret = -ERRNO_TO_DSS_ERROR(errno);
+ DSS_ERROR_LOG(("could not stat .\n"));
+ return ret;
+ }
+ if (stat("..", &dotdot) < 0) {
+ ret = -ERRNO_TO_DSS_ERROR(errno);
+ DSS_ERROR_LOG(("could not stat ..\n"));
+ return ret;
+ }
+ if (dot.st_dev == dotdot.st_dev && dot.st_ino != dotdot.st_ino) {
+ DSS_ERROR_LOG(("mountpoint check failed for %s\n", dd));
+ return -E_MOUNTPOINT;
+ }
+ return 1;
}
-void create_rsync_argv(char ***argv, int64_t *num)
+static int check_config(void)
{
- char *logname, *newest = name_of_newest_complete_snapshot();
- int i = 0, j;
+ int ret;
+ uint32_t unit_interval = OPT_UINT32_VAL(DSS, UNIT_INTERVAL);
+ uint32_t num_intervals = OPT_UINT32_VAL(DSS, NUM_INTERVALS);
- *argv = dss_malloc((15 + conf.rsync_option_given) * sizeof(char *));
- (*argv)[i++] = dss_strdup("rsync");
- (*argv)[i++] = dss_strdup("-aq");
- (*argv)[i++] = dss_strdup("--delete");
- for (j = 0; j < conf.rsync_option_given; j++)
- (*argv)[i++] = dss_strdup(conf.rsync_option_arg[j]);
- if (newest) {
- DSS_INFO_LOG("using %s as reference snapshot\n", newest);
- (*argv)[i++] = make_message("--link-dest=../%s", newest);
- 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);
+ if (unit_interval == 0) {
+ DSS_ERROR_LOG(("bad unit interval: %i\n", unit_interval));
+ return -E_INVALID_NUMBER;
+ }
+ DSS_DEBUG_LOG(("unit interval: %i day(s)\n", unit_interval));
+ if (num_intervals == 0 || num_intervals > 30) {
+ DSS_ERROR_LOG(("bad number of intervals: %i\n", num_intervals));
+ return -E_INVALID_NUMBER;
}
- logname = dss_logname();
- if (conf.remote_user_given && !strcmp(conf.remote_user_arg, logname))
- (*argv)[i++] = dss_strdup(conf.source_dir_arg);
- else
- (*argv)[i++] = make_message("%s@%s:%s/", conf.remote_user_given?
- conf.remote_user_arg : logname,
- conf.remote_host_arg, conf.source_dir_arg);
- free(logname);
- *num = get_current_time();
- (*argv)[i++] = incomplete_name(*num);
- (*argv)[i++] = NULL;
- for (j = 0; j < i; j++)
- DSS_DEBUG_LOG("argv[%d] = %s\n", j, (*argv)[j]);
+ if (subcmd == CMD_PTR(RUN) || subcmd == CMD_PTR(CREATE))
+ if (!OPT_GIVEN(DSS, SOURCE_DIR)) {
+ DSS_ERROR_LOG(("--source-dir required\n"));
+ return -E_SYNTAX;
+ }
+ if (subcmd == CMD_PTR(RUN) || subcmd == CMD_PTR(CREATE)
+ || subcmd == CMD_PTR(LS) || subcmd == CMD_PTR(PRUNE)) {
+ if (!OPT_GIVEN(DSS, DEST_DIR)) {
+ DSS_ERROR_LOG(("--dest-dir required\n"));
+ return -E_SYNTAX;
+ }
+ ret = change_to_dest_dir();
+ if (ret < 0)
+ return ret;
+ }
+ DSS_DEBUG_LOG(("number of intervals: %i\n", num_intervals));
+ return 1;
}
-void free_rsync_argv(char **argv)
+static int lopsub_error(int lopsub_ret, char **errctx)
{
- int i;
- for (i = 0; argv[i]; i++)
- free(argv[i]);
- free(argv);
+ const char *msg = lls_strerror(-lopsub_ret);
+ if (*errctx)
+ DSS_ERROR_LOG(("%s: %s\n", *errctx, msg));
+ else
+ DSS_ERROR_LOG(("%s\n", msg));
+ free(*errctx);
+ *errctx = NULL;
+ return -E_LOPSUB;
}
-int create_snapshot(char **argv)
+static int parse_config_file(bool sighup, const struct lls_command *cmd)
{
- int fds[3] = {0, 0, 0};
- char *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);
+ int ret, fd = -1;
+ char *config_file = get_config_file_name();
+ struct stat statbuf;
+ void *map;
+ size_t sz;
+ int cf_argc;
+ char **cf_argv, *errctx = NULL;
+ struct lls_parse_result *cf_lpr, *merged_lpr, *clpr;
+ const char *subcmd_name;
+
+ ret = open(config_file, O_RDONLY);
+ if (ret < 0) {
+ if (errno != ENOENT || OPT_GIVEN(DSS, CONFIG_FILE)) {
+ ret = -ERRNO_TO_DSS_ERROR(errno);
+ DSS_ERROR_LOG(("config file %s can not be opened\n",
+ config_file));
+ goto out;
+ }
+ /* no config file -- nothing to do */
+ ret = 0;
+ goto success;
+ }
+ fd = ret;
+ ret = fstat(fd, &statbuf);
+ if (ret < 0) {
+ ret = -ERRNO_TO_DSS_ERROR(errno);
+ DSS_ERROR_LOG(("failed to stat config file %s\n", config_file));
+ goto close_fd;
+ }
+ sz = statbuf.st_size;
+ if (sz == 0) { /* config file is empty -- nothing to do */
+ ret = 0;
+ goto success;
+ }
+ map = mmap(NULL, sz, PROT_READ, MAP_PRIVATE, fd, 0);
+ if (map == MAP_FAILED) {
+ ret = -ERRNO_TO_DSS_ERROR(errno);
+ DSS_ERROR_LOG(("failed to mmap config file %s\n",
+ config_file));
+ goto close_fd;
+ }
+ if (cmd == CMD_PTR(DSS))
+ subcmd_name = NULL;
+ else
+ subcmd_name = lls_command_name(cmd);
+ ret = lls_convert_config(map, sz, subcmd_name, &cf_argv, &errctx);
+ munmap(map, sz);
+ if (ret < 0) {
+ DSS_ERROR_LOG(("failed to convert config file %s\n",
+ config_file));
+ ret = lopsub_error(ret, &errctx);
+ goto close_fd;
+ }
+ cf_argc = ret;
+ ret = lls_parse(cf_argc, cf_argv, cmd, &cf_lpr, &errctx);
+ lls_free_argv(cf_argv);
+ if (ret < 0) {
+ ret = lopsub_error(ret, &errctx);
+ goto close_fd;
+ }
+ clpr = cmd == CMD_PTR(DSS)? cmdline_lpr : cmdline_sublpr;
+ if (sighup) /* config file overrides command line */
+ ret = lls_merge(cf_lpr, clpr, cmd, &merged_lpr, &errctx);
+ else /* command line options overrride config file options */
+ ret = lls_merge(clpr, cf_lpr, cmd, &merged_lpr, &errctx);
+ lls_free_parse_result(cf_lpr, cmd);
+ if (ret < 0) {
+ ret = lopsub_error(ret, &errctx);
+ goto close_fd;
+ }
+ ret = 1;
+success:
+ assert(ret >= 0);
+ DSS_DEBUG_LOG(("loglevel: %d\n", OPT_UINT32_VAL(DSS, LOGLEVEL)));
+ if (cmd != CMD_PTR(DSS)) {
+ if (ret > 0) {
+ if (sublpr != cmdline_sublpr)
+ lls_free_parse_result(sublpr, cmd);
+ sublpr = merged_lpr;
+ } else
+ sublpr = cmdline_sublpr;
+ } else {
+ if (ret > 0) {
+ if (lpr != cmdline_lpr)
+ lls_free_parse_result(lpr, cmd);
+ lpr = merged_lpr;
+ } else
+ lpr = cmdline_lpr;
+ }
+close_fd:
+ if (fd >= 0)
+ close(fd);
+out:
+ free(config_file);
+ if (ret < 0)
+ DSS_EMERG_LOG(("%s\n", dss_strerror(-ret)));
+ return ret;
}
-void compute_next_snapshot_time(struct snapshot_list *sl)
+static int handle_sighup(void)
{
- 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;
+ int ret;
- 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;
+ DSS_NOTICE_LOG(("SIGHUP, re-reading config\n"));
+ dump_dss_config("old");
+ ret = parse_config_file(true /* SIGHUP */, CMD_PTR(DSS));
+ if (ret < 0)
+ return ret;
+ ret = parse_config_file(true /* SIGHUP */, CMD_PTR(RUN));
+ if (ret < 0)
+ return ret;
+ ret = check_config();
+ if (ret < 0)
+ return ret;
+ close_log(logfile);
+ logfile = NULL;
+ if (OPT_GIVEN(RUN, DAEMON) || daemonized) {
+ logfile = open_log(OPT_STRING_VAL(RUN, LOGFILE));
+ log_welcome(OPT_UINT32_VAL(DSS, LOGLEVEL));
+ daemonized = true;
}
- tv_divide(wanted, &diff, &tmp);
- tv_add(&now, &tmp, &next_snapshot_time);
+ dump_dss_config("reloaded");
+ invalidate_next_snapshot_time();
+ return 1;
}
-void handle_signal(void)
+static void kill_children(void)
+{
+ restart_create_process();
+ dss_kill(create_pid, SIGTERM, NULL);
+ dss_kill(remove_pid, SIGTERM, NULL);
+}
+
+static int handle_signal(void)
{
int sig, ret = next_signal();
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);
+ return -E_SIGNAL;
case SIGHUP:
- handle_sighup();
- ret = 1;
+ ret = handle_sighup();
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);
+ ret = handle_sigchld();
+ break;
}
out:
if (ret < 0)
- log_err_msg(ERROR, -ret);
+ DSS_ERROR_LOG(("%s\n", dss_strerror(-ret)));
+ return ret;
}
-int get_oldest(const char *dirname, void *private)
+/*
+ * We can not use rsync locally if the local user is different from the remote
+ * user or if the src dir is not on the local host (or both).
+ */
+static int use_rsync_locally(char *logname)
{
- struct edge_snapshot_data *esd = private;
- struct snapshot s;
- int ret = is_snapshot(dirname, esd->now, &s);
+ const char *h = OPT_STRING_VAL(DSS, REMOTE_HOST);
- if (ret <= 0)
- return 1;
- if (s.creation_time > esd->snap.creation_time)
- return 1;
- free(esd->snap.name);
- esd->snap = s;
+ if (strcmp(h, "localhost") && strcmp(h, "127.0.0.1"))
+ return 0;
+ if (OPT_GIVEN(DSS, REMOTE_USER) &&
+ strcmp(OPT_STRING_VAL(DSS, REMOTE_USER), logname))
+ return 0;
return 1;
}
-int remove_oldest_snapshot()
+static int rename_resume_snap(int64_t creation_time)
{
+ struct snapshot_list sl;
+ struct snapshot *s = NULL;
+ char *new_name = incomplete_name(creation_time);
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);
+ const char *why;
+
+ sl.num_snapshots = 0;
+
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);
+ dss_get_snapshot_list(&sl);
+ /*
+ * Snapshot recycling: We first look at the newest snapshot. If this
+ * snapshot happens to be incomplete, the last rsync process was
+ * aborted and we reuse this one. Otherwise we look at snapshots which
+ * could be removed (outdated and redundant snapshots) as candidates
+ * for recycling. If no outdated/redundant snapshot exists, we check if
+ * there is an orphaned snapshot, which likely is useless anyway.
+ *
+ * Only if no existing snapshot is suitable for recycling, we bite the
+ * bullet and create a new one.
+ */
+ s = get_newest_snapshot(&sl);
+ if (!s) /* no snapshots at all */
+ goto out;
+ /* re-use last snapshot if it is incomplete */
+ why = "aborted";
+ if ((s->flags & SS_COMPLETE) == 0)
+ goto out;
+ why = "outdated";
+ s = find_outdated_snapshot(&sl);
+ if (s)
+ goto out;
+ why = "redundant";
+ s = find_redundant_snapshot(&sl);
+ if (s)
+ goto out;
+ why = "orphaned";
+ s = find_orphaned_snapshot(&sl);
out:
- free(esd.snap.name);
+ if (s) {
+ DSS_NOTICE_LOG(("recycling %s snapshot %s\n", why, s->name));
+ ret = dss_rename(s->name, new_name);
+ }
+ if (ret >= 0)
+ DSS_NOTICE_LOG(("creating %s\n", new_name));
+ free(new_name);
+ free_snapshot_list(&sl);
return ret;
}
-/* TODO: Also consider number of inodes. */
-int disk_space_low(void)
+static void create_rsync_argv(char ***argv, int64_t *num)
{
- struct disk_space ds;
- int ret = get_disk_space(".", &ds);
+ char *logname;
+ int i = 0, j, N = OPT_GIVEN(DSS, RSYNC_OPTION);
+ struct snapshot_list sl;
+ static bool seeded;
- 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;
+ dss_get_snapshot_list(&sl);
+ assert(!name_of_reference_snapshot);
+ name_of_reference_snapshot = name_of_newest_complete_snapshot(&sl);
+ free_snapshot_list(&sl);
+
+ *argv = dss_malloc((15 + N) * sizeof(char *));
+ (*argv)[i++] = dss_strdup("rsync");
+ (*argv)[i++] = dss_strdup("-a");
+ (*argv)[i++] = dss_strdup("--delete");
+ if (!seeded) {
+ srandom((unsigned)time(NULL)); /* no need to be fancy here */
+ seeded = true;
+ }
+ if (1000 * (random() / (RAND_MAX + 1.0)) < OPT_UINT32_VAL(DSS, CHECKSUM)) {
+ DSS_NOTICE_LOG(("adding --checksum to rsync options\n"));
+ (*argv)[i++] = dss_strdup("--checksum");
+ }
+ for (j = 0; j < N; j++)
+ (*argv)[i++] = dss_strdup(lls_string_val(j,
+ OPT_RESULT(DSS, RSYNC_OPTION)));
+ if (name_of_reference_snapshot) {
+ DSS_INFO_LOG(("using %s as reference\n", name_of_reference_snapshot));
+ (*argv)[i++] = make_message("--link-dest=../%s",
+ name_of_reference_snapshot);
+ } else
+ DSS_INFO_LOG(("no suitable reference snapshot found\n"));
+ logname = dss_logname();
+ if (use_rsync_locally(logname))
+ (*argv)[i++] = dss_strdup(OPT_STRING_VAL(DSS, SOURCE_DIR));
+ else
+ (*argv)[i++] = make_message("%s@%s:%s/",
+ OPT_GIVEN(DSS, REMOTE_USER)?
+ OPT_STRING_VAL(DSS, REMOTE_USER) : logname,
+ OPT_STRING_VAL(DSS, REMOTE_HOST),
+ OPT_STRING_VAL(DSS, SOURCE_DIR));
+ free(logname);
+ *num = get_current_time();
+ (*argv)[i++] = incomplete_name(*num);
+ (*argv)[i++] = NULL;
+ for (j = 0; j < i; j++)
+ DSS_DEBUG_LOG(("argv[%d] = %s\n", j, (*argv)[j]));
}
-int try_to_free_disk_space(int low_disk_space, struct snapshot_list *sl)
+static void free_rsync_argv(char **argv)
+{
+ int i;
+
+ if (!argv)
+ return;
+ for (i = 0; argv[i]; i++)
+ free(argv[i]);
+ free(argv);
+}
+
+static int create_snapshot(char **argv)
{
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)
+ assert(argv);
+ ret = rename_resume_snap(current_snapshot_creation_time);
+ if (ret < 0)
return ret;
- make_err_msg("uhuhu: not enough disk space for a single snapshot");
- return -ENOSPC;
+ dss_exec(&create_pid, argv[0], argv);
+ snapshot_creation_status = HS_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;
+ char **rsync_argv = NULL;
for (;;) {
- struct timeval now, *tvp = &tv;
fd_set rfds;
- int low_disk_space;
- char **rsync_argv;
+ struct timeval *tvp;
- free_snapshot_list(&sl);
- get_snapshot_list(&sl);
- compute_next_snapshot_time(&sl);
+ if (remove_pid)
+ tvp = NULL; /* sleep until rm hook/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;
ret = dss_select(signal_pipe + 1, &rfds, NULL, tvp);
if (ret < 0)
- return ret;
+ goto out;
if (FD_ISSET(signal_pipe, &rfds)) {
- handle_signal();
+ ret = handle_signal();
+ if (ret < 0)
+ goto out;
+ }
+ if (remove_pid)
+ continue;
+ if (snapshot_removal_status == HS_PRE_SUCCESS) {
+ ret = exec_rm();
+ if (ret < 0)
+ goto out;
continue;
}
- if (rm_pid)
+ if (snapshot_removal_status == HS_SUCCESS) {
+ post_remove_hook();
continue;
- ret = disk_space_low();
- if (ret < 0)
- break;
- 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();
if (ret < 0)
- break;
- if (rm_pid)
- continue;
- if (rsync_pid) {
- restart_rsync_process();
+ goto out;
+ if (snapshot_removal_status != HS_READY) {
+ stop_create_process();
continue;
}
- /* neither rsync nor rm are running. Start rsync? */
- gettimeofday(&now, NULL);
- if (tv_diff(&next_snapshot_time, &now, &tv) > 0)
+ restart_create_process();
+ switch (snapshot_creation_status) {
+ case HS_READY:
+ if (!next_snapshot_is_due())
+ continue;
+ pre_create_hook();
continue;
- create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
- ret = create_snapshot(rsync_argv);
- free_rsync_argv(rsync_argv);
- if (ret < 0)
- break;
+ case HS_PRE_RUNNING:
+ case HS_RUNNING:
+ case HS_POST_RUNNING:
+ continue;
+ case HS_PRE_SUCCESS:
+ if (!name_of_reference_snapshot) {
+ free_rsync_argv(rsync_argv);
+ create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
+ }
+ ret = create_snapshot(rsync_argv);
+ if (ret < 0)
+ goto out;
+ continue;
+ case HS_NEEDS_RESTART:
+ if (!next_snapshot_is_due())
+ continue;
+ ret = create_snapshot(rsync_argv);
+ if (ret < 0)
+ goto out;
+ continue;
+ case HS_SUCCESS:
+ post_create_hook();
+ continue;
+ }
}
- free_snapshot_list(&sl);
+out:
return ret;
}
-int com_run(void)
+static void exit_hook(int exit_code)
{
- int ret;
+ pid_t pid;
+ char **argv, *tmp = dss_strdup(OPT_STRING_VAL(DSS, EXIT_HOOK));
+ unsigned n = split_args(tmp, &argv, " \t");
+
+ n++;
+ argv = dss_realloc(argv, (n + 1) * sizeof(char *));
+ argv[n - 1] = dss_strdup(dss_strerror(-exit_code));
+ argv[n] = NULL;
+ dss_exec(&pid, argv[0], argv);
+ free(argv[n - 1]);
+ free(argv);
+ free(tmp);
+}
- if (conf.dry_run_given) {
- make_err_msg("dry_run not supported by this command");
+static void lock_dss_or_die(void)
+{
+ char *config_file = get_config_file_name();
+ int ret = lock_dss(config_file);
+
+ free(config_file);
+ if (ret < 0) {
+ DSS_EMERG_LOG(("failed to lock: %s\n", dss_strerror(-ret)));
+ exit(EXIT_FAILURE);
+ }
+}
+
+static int com_run(void)
+{
+ int ret, fd = -1;
+ char *config_file;
+ pid_t pid;
+
+ if (OPT_GIVEN(DSS, DRY_RUN)) {
+ DSS_ERROR_LOG(("dry run not supported by this command\n"));
return -E_SYNTAX;
}
+ config_file = get_config_file_name();
+ ret = get_dss_pid(config_file, &pid);
+ free(config_file);
+ if (ret >= 0) {
+ DSS_ERROR_LOG(("pid %d\n", (int)pid));
+ return -E_ALREADY_RUNNING;
+ }
+ if (OPT_GIVEN(RUN, DAEMON)) {
+ fd = daemon_init();
+ daemonized = true;
+ logfile = open_log(OPT_STRING_VAL(RUN, LOGFILE));
+ }
+ lock_dss_or_die();
+ dump_dss_config("startup");
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);
+ if (fd >= 0) {
+ ret = write(fd, "\0", 1);
+ if (ret != 1) {
+ DSS_ERROR_LOG(("write to daemon pipe returned %d\n",
+ ret));
+ if (ret < 0)
+ return -ERRNO_TO_DSS_ERROR(errno);
+ return -E_BUG;
+ }
+ }
+ ret = select_loop();
+ if (ret >= 0) /* impossible */
+ ret = -E_BUG;
+ kill_children();
+ exit_hook(ret);
+ while (wait(NULL) >= 0 || errno != ECHILD)
+ ; /* still have children to wait for */
+ return ret;
}
+EXPORT_CMD_HANDLER(run);
-int com_prune(void)
+static int com_prune(void)
{
int ret;
struct snapshot_list sl;
+ struct snapshot *victim;
struct disk_space ds;
+ const char *why;
+ lock_dss_or_die();
ret = get_disk_space(".", &ds);
if (ret < 0)
return ret;
log_disk_space(&ds);
- for (;;) {
- get_snapshot_list(&sl);
- ret = remove_outdated_snapshot(&sl);
- free_snapshot_list(&sl);
- if (ret < 0)
- return ret;
- if (!ret)
- break;
- ret = wait_for_rm_process();
- if (ret < 0)
- goto out;
+ dss_get_snapshot_list(&sl);
+ why = "outdated";
+ victim = find_outdated_snapshot(&sl);
+ if (victim)
+ goto rm;
+ why = "redundant";
+ victim = find_redundant_snapshot(&sl);
+ if (victim)
+ goto rm;
+ ret = 0;
+ goto out;
+rm:
+ if (OPT_GIVEN(DSS, DRY_RUN)) {
+ dss_msg("%s snapshot %s (interval = %i)\n",
+ why, victim->name, victim->interval);
+ ret = 0;
+ goto out;
}
- for (;;) {
- get_snapshot_list(&sl);
- ret = remove_redundant_snapshot(&sl);
- free_snapshot_list(&sl);
- if (ret < 0)
- return ret;
- if (!ret)
- break;
- ret = wait_for_rm_process();
+ pre_remove_hook(victim, why);
+ if (snapshot_removal_status == HS_PRE_RUNNING) {
+ ret = wait_for_remove_process();
if (ret < 0)
goto out;
+ if (snapshot_removal_status != HS_PRE_SUCCESS)
+ goto out;
}
- return 1;
+ ret = exec_rm();
+ if (ret < 0)
+ goto out;
+ ret = wait_for_remove_process();
+ if (ret < 0)
+ goto out;
+ if (snapshot_removal_status != HS_SUCCESS)
+ goto out;
+ post_remove_hook();
+ if (snapshot_removal_status != HS_POST_RUNNING)
+ goto out;
+ ret = wait_for_remove_process();
+ if (ret < 0)
+ goto out;
+ ret = 1;
out:
+ free_snapshot_list(&sl);
return ret;
}
+EXPORT_CMD_HANDLER(prune);
-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) {
+ lock_dss_or_die();
+ if (OPT_GIVEN(DSS, DRY_RUN)) {
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;
}
+ pre_create_hook();
+ if (create_pid) {
+ ret = wait_for_process(create_pid, &status);
+ if (ret < 0)
+ return ret;
+ ret = handle_pre_create_hook_exit(status);
+ if (ret <= 0) /* error, or pre-create failed */
+ return ret;
+ }
+ create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
ret = create_snapshot(rsync_argv);
if (ret < 0)
goto out;
- ret = wait_for_process(rsync_pid, &status);
+ ret = wait_for_process(create_pid, &status);
if (ret < 0)
goto out;
ret = handle_rsync_exit(status);
+ if (ret < 0)
+ goto out;
+ post_create_hook();
+ if (create_pid)
+ ret = wait_for_process(create_pid, &status);
out:
free_rsync_argv(rsync_argv);
return ret;
}
+EXPORT_CMD_HANDLER(create);
-int com_ls(void)
+static int com_ls(void)
{
int i;
struct snapshot_list sl;
struct snapshot *s;
- get_snapshot_list(&sl);
- FOR_EACH_SNAPSHOT(s, i, &sl)
- dss_msg("%u\t%s\n", s->interval, s->name);
+ int64_t now = get_current_time();
+
+ dss_get_snapshot_list(&sl);
+ FOR_EACH_SNAPSHOT(s, i, &sl) {
+ int64_t d;
+ if (s->flags & SS_COMPLETE)
+ d = (s->completion_time - s->creation_time) / 60;
+ else
+ d = (now - s->creation_time) / 60;
+ dss_msg("%u\t%s\t%3" PRId64 ":%02" PRId64 "\n", s->interval,
+ s->name, d / 60, d % 60);
+ }
free_snapshot_list(&sl);
return 1;
}
+EXPORT_CMD_HANDLER(ls);
-__noreturn void clean_exit(int status)
+static int com_configtest(void)
{
- free(dss_error_txt);
- exit(status);
+ printf("Syntax Ok\n");
+ return 0;
}
-static void setup_signal_handling(void)
+EXPORT_CMD_HANDLER(configtest);
+
+static int setup_signal_handling(void)
{
int ret;
- DSS_INFO_LOG("setting up signal handlers\n");
+ DSS_INFO_LOG(("setting up signal handlers\n"));
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);
+}
+
+static void handle_version_and_help(void)
+{
+ char *txt;
+
+ if (OPT_GIVEN(DSS, DETAILED_HELP))
+ txt = lls_long_help(CMD_PTR(DSS));
+ else if (OPT_GIVEN(DSS, HELP))
+ txt = lls_short_help(CMD_PTR(DSS));
+ else if (OPT_GIVEN(DSS, VERSION))
+ txt = make_message("%s\n", VERSION_STRING);
+ else
+ return;
+ printf("%s", txt);
+ free(txt);
+ exit(EXIT_SUCCESS);
+}
+
+static void show_subcommand_summary(void)
+{
+ const struct lls_command *cmd;
+ int i;
+
+ printf("Available subcommands:\n");
+ for (i = 1; (cmd = lls_cmd(i, dss_suite)); i++) {
+ const char *name = lls_command_name(cmd);
+ const char *purpose = lls_purpose(cmd);
+ printf("%-11s%s\n", name, purpose);
+ }
+ exit(EXIT_SUCCESS);
}
int main(int argc, char **argv)
{
int ret;
- struct cmdline_parser_params params = {
- .override = 0,
- .initialize = 1,
- .check_required = 0,
- .check_ambiguity = 0
- };
+ char *errctx = NULL;
+ unsigned num_inputs;
+ const struct dss_user_data *ud;
- cmdline_parser_ext(argc, argv, &conf, ¶ms); /* aborts on errors */
- parse_config_file(0);
-
- if (conf.daemon_given)
- daemon_init();
- setup_signal_handling();
- ret = call_command_handler();
+ ret = lls_parse(argc, argv, CMD_PTR(DSS), &cmdline_lpr, &errctx);
+ if (ret < 0) {
+ ret = lopsub_error(ret, &errctx);
+ goto out;
+ }
+ lpr = cmdline_lpr;
+ ret = parse_config_file(false /* no SIGHUP */, CMD_PTR(DSS));
if (ret < 0)
- log_err_msg(EMERG, -ret);
- clean_exit(ret >= 0? EXIT_SUCCESS : EXIT_FAILURE);
+ goto out;
+ handle_version_and_help();
+ num_inputs = lls_num_inputs(lpr);
+ if (num_inputs == 0)
+ show_subcommand_summary();
+ ret = lls_lookup_subcmd(argv[argc - num_inputs], dss_suite, &errctx);
+ if (ret < 0) {
+ ret = lopsub_error(ret, &errctx);
+ goto out;
+ }
+ subcmd = lls_cmd(ret, dss_suite);
+ ret = lls_parse(num_inputs, argv + argc - num_inputs, subcmd,
+ &cmdline_sublpr, &errctx);
+ if (ret < 0) {
+ ret = lopsub_error(ret, &errctx);
+ goto out;
+ }
+ sublpr = cmdline_sublpr;
+ ret = parse_config_file(false /* no SIGHUP */, subcmd);
+ if (ret < 0)
+ goto out;
+ ret = check_config();
+ if (ret < 0)
+ goto out;
+ ret = setup_signal_handling();
+ if (ret < 0)
+ goto out;
+ ud = lls_user_data(subcmd);
+ ret = ud->handler();
+ signal_shutdown();
+out:
+ if (ret < 0) {
+ if (errctx)
+ DSS_ERROR_LOG(("%s\n", errctx));
+ DSS_EMERG_LOG(("%s\n", dss_strerror(-ret)));
+ }
+ free(errctx);
+ lls_free_parse_result(lpr, CMD_PTR(DSS));
+ if (lpr != cmdline_lpr)
+ lls_free_parse_result(cmdline_lpr, CMD_PTR(DSS));
+ lls_free_parse_result(sublpr, subcmd);
+ if (sublpr != cmdline_sublpr)
+ lls_free_parse_result(cmdline_sublpr, subcmd);
+ exit(ret >= 0? EXIT_SUCCESS : EXIT_FAILURE);
}
projects / dss.git / blobdiff