static void add_fd(fd_set *fds, int *maxfd, int fd)
{
+ struct stat st;
if (fd < 0)
return;
+ if (fstat(fd, &st) == -1) {
+ dprintf("Invalid fd %d\n", fd);
+ return;
+ }
+ if (st.st_nlink == 0) {
+ dprintf("fd %d was deleted\n", fd);
+ return;
+ }
if (fd > *maxfd)
*maxfd = fd;
FD_SET(fd, fds);
return n;
}
-static unsigned long long read_resync_start(int fd)
+static void read_resync_start(int fd, unsigned long long *v)
{
char buf[30];
int n;
n = read_attr(buf, 30, fd);
- if (n <= 0)
- return 0;
+ if (n <= 0) {
+ dprintf("Failed to read resync_start (%d)\n", fd);
+ return;
+ }
if (strncmp(buf, "none", 4) == 0)
- return MaxSector;
+ *v = MaxSector;
else
- return strtoull(buf, NULL, 10);
+ *v = strtoull(buf, NULL, 10);
}
static unsigned long long read_sync_completed(int fd)
*
*/
+#define ARRAY_DIRTY 1
+#define ARRAY_BUSY 2
static int read_and_act(struct active_array *a)
{
unsigned long long sync_completed;
int check_reshape = 0;
int deactivate = 0;
struct mdinfo *mdi;
- int dirty = 0;
+ int ret = 0;
int count = 0;
+ struct timeval tv;
a->next_state = bad_word;
a->next_action = bad_action;
a->curr_state = read_state(a->info.state_fd);
a->curr_action = read_action(a->action_fd);
- a->info.resync_start = read_resync_start(a->resync_start_fd);
+ if (a->curr_state != clear)
+ /*
+ * In "clear" state, resync_start may wrongly be set to "0"
+ * when the kernel called md_clean but didn't remove the
+ * sysfs attributes yet
+ */
+ read_resync_start(a->resync_start_fd, &a->info.resync_start);
sync_completed = read_sync_completed(a->sync_completed_fd);
for (mdi = a->info.devs; mdi ; mdi = mdi->next) {
mdi->next_state = 0;
mdi->curr_state = 0;
if (mdi->state_fd >= 0) {
- mdi->recovery_start = read_resync_start(mdi->recovery_fd);
+ read_resync_start(mdi->recovery_fd,
+ &mdi->recovery_start);
mdi->curr_state = read_dev_state(mdi->state_fd);
}
}
- if (a->curr_state > inactive &&
- a->prev_state == inactive) {
- /* array has been started
- * possible that container operation has to be completed
- */
- a->container->ss->set_array_state(a, 0);
- }
- if (a->curr_state <= inactive &&
+ gettimeofday(&tv, NULL);
+ dprintf("(%d): %ld.%06ld state:%s prev:%s action:%s prev: %s start:%llu\n",
+ a->info.container_member,
+ tv.tv_sec, tv.tv_usec,
+ array_states[a->curr_state],
+ array_states[a->prev_state],
+ sync_actions[a->curr_action],
+ sync_actions[a->prev_action],
+ a->info.resync_start
+ );
+
+ if ((a->curr_state == bad_word || a->curr_state <= inactive) &&
a->prev_state > inactive) {
/* array has been stopped */
a->container->ss->set_array_state(a, 1);
if (a->curr_state == write_pending) {
a->container->ss->set_array_state(a, 0);
a->next_state = active;
- dirty = 1;
+ ret |= ARRAY_DIRTY;
}
if (a->curr_state == active_idle) {
/* Set array to 'clean' FIRST, then mark clean
* in the metadata
*/
a->next_state = clean;
- dirty = 1;
+ ret |= ARRAY_DIRTY;
}
if (a->curr_state == clean) {
a->container->ss->set_array_state(a, 1);
}
if (a->curr_state == active ||
- a->curr_state == suspended ||
- a->curr_state == bad_word)
- dirty = 1;
+ a->curr_state == suspended)
+ ret |= ARRAY_DIRTY;
if (a->curr_state == readonly) {
/* Well, I'm ready to handle things. If readonly
* wasn't requested, transition to read-auto.
a->next_state = read_auto; /* array is clean */
else {
a->next_state = active; /* Now active for recovery etc */
- dirty = 1;
+ ret |= ARRAY_DIRTY;
}
}
}
a->last_checkpoint = sync_completed;
a->container->ss->sync_metadata(a->container);
- dprintf("%s(%d): state:%s action:%s next(", __func__, a->info.container_member,
+ dprintf("(%d): state:%s action:%s next(", a->info.container_member,
array_states[a->curr_state], sync_actions[a->curr_action]);
/* Effect state changes in the array */
if (a->next_state != bad_word) {
- dprintf(" state:%s", array_states[a->next_state]);
+ dprintf_cont(" state:%s", array_states[a->next_state]);
write_attr(array_states[a->next_state], a->info.state_fd);
}
if (a->next_action != bad_action) {
write_attr(sync_actions[a->next_action], a->action_fd);
- dprintf(" action:%s", sync_actions[a->next_action]);
+ dprintf_cont(" action:%s", sync_actions[a->next_action]);
}
for (mdi = a->info.devs; mdi ; mdi = mdi->next) {
if (mdi->next_state & DS_UNBLOCK) {
- dprintf(" %d:-blocked", mdi->disk.raid_disk);
+ dprintf_cont(" %d:-blocked", mdi->disk.raid_disk);
write_attr("-blocked", mdi->state_fd);
}
if ((mdi->next_state & DS_REMOVE) && mdi->state_fd >= 0) {
int remove_result;
- /* the kernel may not be able to immediately remove the
- * disk, we can simply wait until the next event to try
- * again.
+ /* The kernel may not be able to immediately remove the
+ * disk. In that case we wait a little while and
+ * try again.
*/
remove_result = write_attr("remove", mdi->state_fd);
if (remove_result > 0) {
- dprintf(" %d:removed", mdi->disk.raid_disk);
+ dprintf_cont(" %d:removed", mdi->disk.raid_disk);
close(mdi->state_fd);
close(mdi->recovery_fd);
mdi->state_fd = -1;
- }
+ } else
+ ret |= ARRAY_BUSY;
}
if (mdi->next_state & DS_INSYNC) {
write_attr("+in_sync", mdi->state_fd);
- dprintf(" %d:+in_sync", mdi->disk.raid_disk);
+ dprintf_cont(" %d:+in_sync", mdi->disk.raid_disk);
}
}
- dprintf(" )\n");
+ dprintf_cont(" )\n");
/* move curr_ to prev_ */
a->prev_state = a->curr_state;
if (deactivate)
a->container = NULL;
- return dirty;
+ return ret;
}
static struct mdinfo *
*/
int fd;
if (sigterm)
- fd = open_dev_excl(container->devnum);
+ fd = open_dev_excl(container->devnm);
else
- fd = open_dev_flags(container->devnum, O_RDONLY|O_EXCL);
+ fd = open_dev_flags(container->devnm, O_RDONLY|O_EXCL);
if (fd >= 0 || errno != EBUSY) {
/* OK, we are safe to leave */
if (sigterm && !dirty_arrays)
/* On SIGTERM, someone (the take-over mdmon) will
* clean up
*/
- remove_pidfile(container->devname);
+ remove_pidfile(container->devnm);
exit_now = 1;
signal_manager();
close(fd);
struct timespec ts;
ts.tv_sec = 24*3600;
ts.tv_nsec = 0;
- if (*aap == NULL) {
+ if (*aap == NULL || container->retry_soon) {
/* just waiting to get O_EXCL access */
ts.tv_sec = 0;
ts.tv_nsec = 20000000ULL;
monitor_loop_cnt |= 1;
rv = pselect(maxfd+1, NULL, NULL, &rfds, &ts, &set);
monitor_loop_cnt += 1;
- if (rv == -1 && errno == EINTR)
- rv = 0;
+ if (rv == -1) {
+ if (errno == EINTR) {
+ rv = 0;
+ dprintf("monitor: caught signal\n");
+ } else
+ dprintf("monitor: error %d in pselect\n",
+ errno);
+ }
#ifdef DEBUG
- dprint_wake_reasons(&rfds);
+ else
+ dprint_wake_reasons(&rfds);
#endif
-
+ container->retry_soon = 0;
}
if (update_queue) {
rv = 0;
dirty_arrays = 0;
for (a = *aap; a ; a = a->next) {
- int is_dirty;
if (a->replaces && !discard_this) {
struct active_array **ap;
signal_manager();
}
if (a->container && !a->to_remove) {
- is_dirty = read_and_act(a);
+ int ret = read_and_act(a);
rv |= 1;
- dirty_arrays += is_dirty;
+ dirty_arrays += !!(ret & ARRAY_DIRTY);
/* when terminating stop manipulating the array after it
* is clean, but make sure read_and_act() is given a
* chance to handle 'active_idle'
*/
- if (sigterm && !is_dirty)
+ if (sigterm && !(ret & ARRAY_DIRTY))
a->container = NULL; /* stop touching this array */
+ if (ret & ARRAY_BUSY)
+ container->retry_soon = 1;
}
}
projects / thirdparty / mdadm.git / blobdiff