static LIST_HEAD(all_mddevs);
static DEFINE_SPINLOCK(all_mddevs_lock);
-/*
- * iterates through all used mddevs in the system.
- * We take care to grab the all_mddevs_lock whenever navigating
- * the list, and to always hold a refcount when unlocked.
- * Any code which breaks out of this loop while own
- * a reference to the current mddev and must mddev_put it.
- */
-#define for_each_mddev(_mddev,_tmp) \
- \
- for (({ spin_lock(&all_mddevs_lock); \
- _tmp = all_mddevs.next; \
- _mddev = NULL;}); \
- ({ if (_tmp != &all_mddevs) \
- mddev_get(list_entry(_tmp, struct mddev, all_mddevs));\
- spin_unlock(&all_mddevs_lock); \
- if (_mddev) mddev_put(_mddev); \
- _mddev = list_entry(_tmp, struct mddev, all_mddevs); \
- _tmp != &all_mddevs;}); \
- ({ spin_lock(&all_mddevs_lock); \
- _tmp = _tmp->next;}) \
- )
-
/* Rather than calling directly into the personality make_request function,
* IO requests come here first so that we can check if the device is
* being suspended pending a reconfiguration.
return sprintf(page, "%llu\n", (unsigned long long)rdev->sectors / 2);
}
-static int overlaps(sector_t s1, sector_t l1, sector_t s2, sector_t l2)
+static int md_rdevs_overlap(struct md_rdev *a, struct md_rdev *b)
{
/* check if two start/length pairs overlap */
- if (s1+l1 <= s2)
- return 0;
- if (s2+l2 <= s1)
- return 0;
- return 1;
+ if (a->data_offset + a->sectors <= b->data_offset)
+ return false;
+ if (b->data_offset + b->sectors <= a->data_offset)
+ return false;
+ return true;
+}
+
+static bool md_rdev_overlaps(struct md_rdev *rdev)
+{
+ struct mddev *mddev;
+ struct md_rdev *rdev2;
+
+ spin_lock(&all_mddevs_lock);
+ list_for_each_entry(mddev, &all_mddevs, all_mddevs) {
+ rdev_for_each(rdev2, mddev) {
+ if (rdev != rdev2 && rdev->bdev == rdev2->bdev &&
+ md_rdevs_overlap(rdev, rdev2)) {
+ spin_unlock(&all_mddevs_lock);
+ return true;
+ }
+ }
+ }
+ spin_unlock(&all_mddevs_lock);
+ return false;
}
static int strict_blocks_to_sectors(const char *buf, sector_t *sectors)
return -EINVAL; /* component must fit device */
rdev->sectors = sectors;
- if (sectors > oldsectors && my_mddev->external) {
- /* Need to check that all other rdevs with the same
- * ->bdev do not overlap. 'rcu' is sufficient to walk
- * the rdev lists safely.
- * This check does not provide a hard guarantee, it
- * just helps avoid dangerous mistakes.
- */
- struct mddev *mddev;
- int overlap = 0;
- struct list_head *tmp;
- rcu_read_lock();
- for_each_mddev(mddev, tmp) {
- struct md_rdev *rdev2;
-
- rdev_for_each(rdev2, mddev)
- if (rdev->bdev == rdev2->bdev &&
- rdev != rdev2 &&
- overlaps(rdev->data_offset, rdev->sectors,
- rdev2->data_offset,
- rdev2->sectors)) {
- overlap = 1;
- break;
- }
- if (overlap) {
- mddev_put(mddev);
- break;
- }
- }
- rcu_read_unlock();
- if (overlap) {
- /* Someone else could have slipped in a size
- * change here, but doing so is just silly.
- * We put oldsectors back because we *know* it is
- * safe, and trust userspace not to race with
- * itself
- */
- rdev->sectors = oldsectors;
- return -EBUSY;
- }
+ /*
+ * Check that all other rdevs with the same bdev do not overlap. This
+ * check does not provide a hard guarantee, it just helps avoid
+ * dangerous mistakes.
+ */
+ if (sectors > oldsectors && my_mddev->external &&
+ md_rdev_overlaps(rdev)) {
+ /*
+ * Someone else could have slipped in a size change here, but
+ * doing so is just silly. We put oldsectors back because we
+ * know it is safe, and trust userspace not to race with itself.
+ */
+ rdev->sectors = oldsectors;
+ return -EBUSY;
}
return len;
}
unsigned long update_time;
sector_t mark_cnt[SYNC_MARKS];
int last_mark,m;
- struct list_head *tmp;
sector_t last_check;
int skipped = 0;
struct md_rdev *rdev;
try_again:
if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
goto skip;
- for_each_mddev(mddev2, tmp) {
+ spin_lock(&all_mddevs_lock);
+ list_for_each_entry(mddev2, &all_mddevs, all_mddevs) {
if (mddev2 == mddev)
continue;
if (!mddev->parallel_resync
desc, mdname(mddev),
mdname(mddev2));
}
- mddev_put(mddev2);
+ spin_unlock(&all_mddevs_lock);
+
if (signal_pending(current))
flush_signals(current);
schedule();
finish_wait(&resync_wait, &wq);
}
}
+ spin_unlock(&all_mddevs_lock);
} while (mddev->curr_resync < MD_RESYNC_DELAYED);
j = 0;
static int md_notify_reboot(struct notifier_block *this,
unsigned long code, void *x)
{
- struct list_head *tmp;
- struct mddev *mddev;
+ struct mddev *mddev, *n;
int need_delay = 0;
- for_each_mddev(mddev, tmp) {
+ spin_lock(&all_mddevs_lock);
+ list_for_each_entry_safe(mddev, n, &all_mddevs, all_mddevs) {
+ mddev_get(mddev);
+ spin_unlock(&all_mddevs_lock);
if (mddev_trylock(mddev)) {
if (mddev->pers)
__md_stop_writes(mddev);
mddev_unlock(mddev);
}
need_delay = 1;
+ mddev_put(mddev);
+ spin_lock(&all_mddevs_lock);
}
+ spin_unlock(&all_mddevs_lock);
+
/*
* certain more exotic SCSI devices are known to be
* volatile wrt too early system reboots. While the
static __exit void md_exit(void)
{
- struct mddev *mddev;
- struct list_head *tmp;
+ struct mddev *mddev, *n;
int delay = 1;
unregister_blkdev(MD_MAJOR,"md");
}
remove_proc_entry("mdstat", NULL);
- for_each_mddev(mddev, tmp) {
+ spin_lock(&all_mddevs_lock);
+ list_for_each_entry_safe(mddev, n, &all_mddevs, all_mddevs) {
+ mddev_get(mddev);
+ spin_unlock(&all_mddevs_lock);
export_array(mddev);
mddev->ctime = 0;
mddev->hold_active = 0;
/*
- * for_each_mddev() will call mddev_put() at the end of each
- * iteration. As the mddev is now fully clear, this will
- * schedule the mddev for destruction by a workqueue, and the
+ * As the mddev is now fully clear, mddev_put will schedule
+ * the mddev for destruction by a workqueue, and the
* destroy_workqueue() below will wait for that to complete.
*/
+ mddev_put(mddev);
+ spin_lock(&all_mddevs_lock);
}
+ spin_unlock(&all_mddevs_lock);
+
destroy_workqueue(md_rdev_misc_wq);
destroy_workqueue(md_misc_wq);
destroy_workqueue(md_wq);
projects / people / ms / linux.git / blobdiff