tmp = tmp->next;}) \
)
+int md_flush_mddev(mddev_t *mddev, sector_t *error_sector)
+{
+ struct list_head *tmp;
+ mdk_rdev_t *rdev;
+ int ret = 0;
+
+ /*
+ * this list iteration is done without any locking in md?!
+ */
+ ITERATE_RDEV(mddev, rdev, tmp) {
+ request_queue_t *r_queue = bdev_get_queue(rdev->bdev);
+ int err;
+
+ if (!r_queue->issue_flush_fn)
+ err = -EOPNOTSUPP;
+ else
+ err = r_queue->issue_flush_fn(r_queue, rdev->bdev->bd_disk, error_sector);
+
+ if (!ret)
+ ret = err;
+ }
+
+ return ret;
+}
+
+static int md_flush_all(request_queue_t *q, struct gendisk *disk,
+ sector_t *error_sector)
+{
+ mddev_t *mddev = q->queuedata;
+
+ return md_flush_mddev(mddev, error_sector);
+}
+
static int md_fail_request (request_queue_t *q, struct bio *bio)
{
bio_io_error(bio, bio->bi_size);
return 0;
fail:
- printk(KERN_ERR "md: disabled device %s, could not read superblock.\n",
+ printk(KERN_WARNING "md: disabled device %s, could not read superblock.\n",
bdevname(rdev->bdev,b));
return -EINVAL;
}
return csum;
}
+/* csum_partial is not consistent between different architectures.
+ * Some (i386) do a 32bit csum. Some (alpha) do 16 bit.
+ * This makes it hard for user-space to know what to do.
+ * So we use calc_sb_csum to set the checksum to allow working
+ * with older kernels, but allow calc_sb_csum_common to
+ * be used when checking if a checksum is correct, to
+ * make life easier for user-space tools that might write
+ * a superblock.
+ */
+static unsigned int calc_sb_csum_common(mdp_super_t *super)
+{
+ unsigned int disk_csum = super->sb_csum;
+ unsigned long long newcsum = 0;
+ unsigned int csum;
+ int i;
+ unsigned int *superc = (int*) super;
+ super->sb_csum = 0;
+
+ for (i=0; i<MD_SB_BYTES/4; i++)
+ newcsum+= superc[i];
+ csum = (newcsum& 0xffffffff) + (newcsum>>32);
+ super->sb_csum = disk_csum;
+ return csum;
+}
+
/*
* Handle superblock details.
* We want to be able to handle multiple superblock formats
if (sb->raid_disks <= 0)
goto abort;
- if (calc_sb_csum(sb) != sb->sb_csum) {
+ if (calc_sb_csum(sb) != sb->sb_csum &&
+ calc_sb_csum_common(sb) != sb->sb_csum) {
printk(KERN_WARNING "md: invalid superblock checksum on %s\n",
b);
goto abort;
static unsigned int calc_sb_1_csum(struct mdp_superblock_1 * sb)
{
unsigned int disk_csum, csum;
+ unsigned long long newcsum;
int size = 256 + sb->max_dev*2;
+ unsigned int *isuper = (unsigned int*)sb;
+ int i;
disk_csum = sb->sb_csum;
sb->sb_csum = 0;
- csum = csum_partial((void *)sb, size, 0);
+ newcsum = 0;
+ for (i=0; size>=4; size -= 4 )
+ newcsum += le32_to_cpu(*isuper++);
+
+ if (size == 2)
+ newcsum += le16_to_cpu(*(unsigned short*) isuper);
+
+ csum = (newcsum & 0xffffffff) + (newcsum >> 32);
sb->sb_csum = disk_csum;
return csum;
}
/*
* prevent the device from being mounted, repartitioned or
* otherwise reused by a RAID array (or any other kernel
- * subsystem), by opening the device. [simply getting an
- * inode is not enough, the SCSI module usage code needs
- * an explicit open() on the device]
+ * subsystem), by bd_claiming the device.
*/
static int lock_rdev(mdk_rdev_t *rdev, dev_t dev)
{
int err = 0;
struct block_device *bdev;
+ char b[BDEVNAME_SIZE];
bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
- if (IS_ERR(bdev))
+ if (IS_ERR(bdev)) {
+ printk(KERN_ERR "md: could not open %s.\n",
+ __bdevname(dev, b));
return PTR_ERR(bdev);
+ }
err = bd_claim(bdev, rdev);
if (err) {
+ printk(KERN_ERR "md: could not bd_claim %s.\n",
+ bdevname(bdev, b));
blkdev_put(bdev);
return err;
}
static void print_desc(mdp_disk_t *desc)
{
- char b[BDEVNAME_SIZE];
-
- printk(" DISK<N:%d,%s(%d,%d),R:%d,S:%d>\n", desc->number,
- __bdevname(MKDEV(desc->major, desc->minor), b),
+ printk(" DISK<N:%d,(%d,%d),R:%d,S:%d>\n", desc->number,
desc->major,desc->minor,desc->raid_disk,desc->state);
}
rdev = (mdk_rdev_t *) kmalloc(sizeof(*rdev), GFP_KERNEL);
if (!rdev) {
- printk(KERN_ERR "md: could not alloc mem for %s!\n",
- __bdevname(newdev, b));
+ printk(KERN_ERR "md: could not alloc mem for new device!\n");
return ERR_PTR(-ENOMEM);
}
memset(rdev, 0, sizeof(*rdev));
goto abort_free;
err = lock_rdev(rdev, newdev);
- if (err) {
- printk(KERN_ERR "md: could not lock %s.\n",
- __bdevname(newdev, b));
+ if (err)
goto abort_free;
- }
+
rdev->desc_nr = -1;
rdev->faulty = 0;
rdev->in_sync = 0;
mddev->pers = pers[pnum];
spin_unlock(&pers_lock);
+ mddev->resync_max_sectors = mddev->size << 1; /* may be over-ridden by personality */
+
err = mddev->pers->run(mddev);
if (err) {
printk(KERN_ERR "md: pers->run() failed ...\n");
*/
mddev->queue->queuedata = mddev;
mddev->queue->make_request_fn = mddev->pers->make_request;
+ mddev->queue->issue_flush_fn = md_flush_all;
mddev->changed = 1;
return 0;
mdk_rdev_t *start_rdev = NULL, *rdev;
start_rdev = md_import_device(startdev, 0, 0);
- if (IS_ERR(start_rdev)) {
- printk(KERN_WARNING "md: could not import %s!\n",
- __bdevname(startdev, b));
+ if (IS_ERR(start_rdev))
return err;
- }
+
/* NOTE: this can only work for 0.90.0 superblocks */
sb = (mdp_super_t*)page_address(start_rdev->sb_page);
if (MAJOR(dev) != desc->major || MINOR(dev) != desc->minor)
continue;
rdev = md_import_device(dev, 0, 0);
- if (IS_ERR(rdev)) {
- printk(KERN_WARNING "md: could not import %s,"
- " trying to run array nevertheless.\n",
- __bdevname(dev, b));
+ if (IS_ERR(rdev))
continue;
- }
+
list_add(&rdev->same_set, &pending_raid_disks);
}
return 0;
}
-static int hot_generate_error(mddev_t * mddev, dev_t dev)
-{
- char b[BDEVNAME_SIZE];
- struct request_queue *q;
- mdk_rdev_t *rdev;
-
- if (!mddev->pers)
- return -ENODEV;
-
- printk(KERN_INFO "md: trying to generate %s error in %s ... \n",
- __bdevname(dev, b), mdname(mddev));
-
- rdev = find_rdev(mddev, dev);
- if (!rdev) {
- /* MD_BUG(); */ /* like hell - it's not a driver bug */
- return -ENXIO;
- }
-
- if (rdev->desc_nr == -1) {
- MD_BUG();
- return -EINVAL;
- }
- if (!rdev->in_sync)
- return -ENODEV;
-
- q = bdev_get_queue(rdev->bdev);
- if (!q) {
- MD_BUG();
- return -ENODEV;
- }
- printk(KERN_INFO "md: okay, generating error!\n");
-// q->oneshot_error = 1; // disabled for now
-
- return 0;
-}
-
static int hot_remove_disk(mddev_t * mddev, dev_t dev)
{
char b[BDEVNAME_SIZE];
if (!mddev->pers)
return -ENODEV;
- printk(KERN_INFO "md: trying to remove %s from %s ... \n",
- __bdevname(dev, b), mdname(mddev));
-
rdev = find_rdev(mddev, dev);
if (!rdev)
return -ENXIO;
if (!mddev->pers)
return -ENODEV;
- printk(KERN_INFO "md: trying to hot-add %s to %s ... \n",
- __bdevname(dev, b), mdname(mddev));
-
if (mddev->major_version != 0) {
printk(KERN_WARNING "%s: HOT_ADD may only be used with"
" version-0 superblocks.\n",
static int md_ioctl(struct inode *inode, struct file *file,
unsigned int cmd, unsigned long arg)
{
- char b[BDEVNAME_SIZE];
int err = 0;
void __user *argp = (void __user *)arg;
struct hd_geometry __user *loc = argp;
}
err = autostart_array(new_decode_dev(arg));
if (err) {
- printk(KERN_WARNING "md: autostart %s failed!\n",
- __bdevname(arg, b));
+ printk(KERN_WARNING "md: autostart failed!\n");
goto abort;
}
goto done;
err = add_new_disk(mddev, &info);
goto done_unlock;
}
- case HOT_GENERATE_ERROR:
- err = hot_generate_error(mddev, new_decode_dev(arg));
- goto done_unlock;
+
case HOT_REMOVE_DISK:
err = hot_remove_disk(mddev, new_decode_dev(arg));
goto done_unlock;
return thread;
}
-void md_interrupt_thread(mdk_thread_t *thread)
+static void md_interrupt_thread(mdk_thread_t *thread)
{
if (!thread->tsk) {
MD_BUG();
if (!mddev->pers->error_handler)
return;
mddev->pers->error_handler(mddev,rdev);
+ set_bit(MD_RECOVERY_INTR, &mddev->recovery);
set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
md_wakeup_thread(mddev->thread);
}
unsigned long max_blocks, resync, res, dt, db, rt;
resync = (mddev->curr_resync - atomic_read(&mddev->recovery_active))/2;
- max_blocks = mddev->size;
+
+ if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery))
+ max_blocks = mddev->resync_max_sectors >> 1;
+ else
+ max_blocks = mddev->size;
/*
* Should not happen.
return 0;
}
-void md_sync_acct(mdk_rdev_t *rdev, unsigned long nr_sectors)
-{
- rdev->bdev->bd_contains->bd_disk->sync_io += nr_sectors;
-}
-
static int is_mddev_idle(mddev_t *mddev)
{
mdk_rdev_t * rdev;
struct gendisk *disk = rdev->bdev->bd_contains->bd_disk;
curr_events = disk_stat_read(disk, read_sectors) +
disk_stat_read(disk, write_sectors) -
- disk->sync_io;
- if ((curr_events - rdev->last_events) > 32) {
+ atomic_read(&disk->sync_io);
+ /* Allow some slack between valud of curr_events and last_events,
+ * as there are some uninteresting races.
+ * Note: the following is an unsigned comparison.
+ */
+ if ((curr_events - rdev->last_events + 32) > 64) {
rdev->last_events = curr_events;
idle = 0;
}
}
} while (mddev->curr_resync < 2);
- max_sectors = mddev->size << 1;
+ if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery))
+ /* resync follows the size requested by the personality,
+ * which default to physical size, but can be virtual size
+ */
+ max_sectors = mddev->resync_max_sectors;
+ else
+ /* recovery follows the physical size of devices */
+ max_sectors = mddev->size << 1;
printk(KERN_INFO "md: syncing RAID array %s\n", mdname(mddev));
printk(KERN_INFO "md: minimum _guaranteed_ reconstruction speed:"
static void autostart_arrays(int part)
{
- char b[BDEVNAME_SIZE];
mdk_rdev_t *rdev;
int i;
dev_t dev = detected_devices[i];
rdev = md_import_device(dev,0, 0);
- if (IS_ERR(rdev)) {
- printk(KERN_ALERT "md: could not import %s!\n",
- __bdevname(dev, b));
+ if (IS_ERR(rdev))
continue;
- }
+
if (rdev->faulty) {
MD_BUG();
continue;
EXPORT_SYMBOL(register_md_personality);
EXPORT_SYMBOL(unregister_md_personality);
EXPORT_SYMBOL(md_error);
-EXPORT_SYMBOL(md_sync_acct);
EXPORT_SYMBOL(md_done_sync);
EXPORT_SYMBOL(md_write_start);
EXPORT_SYMBOL(md_write_end);
EXPORT_SYMBOL(md_unregister_thread);
EXPORT_SYMBOL(md_wakeup_thread);
EXPORT_SYMBOL(md_print_devices);
-EXPORT_SYMBOL(md_interrupt_thread);
EXPORT_SYMBOL(md_check_recovery);
MODULE_LICENSE("GPL");