Marcelo,
This patch fixes a little bug in md/raid1 that I noticed during my
hack/rip/tear at the 2.5 code.
Thanks,
NeilBrown
=================
Make sure raid1 superblock changes get written out promptly.
mark_disk_bad (called from raid1_error) sets "sb_dirty"
and wakes-up raid1d to get the superblock written.
However, raid1d only writes the superblock if there are
pending failed read requests to be retried. So if a device
fails on a write or an ioctl (raidsetfaulty), then the superblock
wont be updated.
This patch moves the call to md_update_sb outside of the loop
in raid1d. This is consistant with the pattern in raid5.
----------- Diffstat output ------------
./drivers/md/raid1.c | 8 ++++----
1 files changed, 4 insertions(+), 4 deletions(-)
--- ./drivers/md/raid1.c 2002/06/27 15:36:23 1.1
+++ ./drivers/md/raid1.c 2002/06/27 16:17:41 1.2
@@ -1152,9 +1152,12 @@ static void raid1d (void *data)
struct raid1_bh *r1_bh;
struct buffer_head *bh;
unsigned long flags;
- mddev_t *mddev;
+ raid1_conf_t *conf = data;
+ mddev_t *mddev = conf->mddev;
kdev_t dev;
+ if (mddev->sb_dirty)
+ md_update_sb(mddev);
for (;;) {
md_spin_lock_irqsave(&retry_list_lock, flags);
@@ -1165,8 +1168,6 @@ static void raid1d (void *data)
md_spin_unlock_irqrestore(&retry_list_lock, flags);
mddev = r1_bh->mddev;
- if (mddev->sb_dirty)
- md_update_sb(mddev);
bh = &r1_bh->bh_req;
switch(r1_bh->cmd) {
case SPECIAL:
@@ -1177,7 +1178,6 @@ static void raid1d (void *data)
int i, sum_bhs = 0;
int disks = MD_SB_DISKS;
struct buffer_head *bhl, *mbh;
- raid1_conf_t *conf;
conf = mddev_to_conf(mddev);
bhl = raid1_alloc_bh(conf, conf->raid_disks); /* don't really need this many */
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
