On Wednesday April 9, 76306.1226@compuserve.com wrote:
> Neil Brown wrote:
> >>
> >> This could only happen if a second disk failed during rebuild of a 3+
> >> disk array.
> >
> >True, but if a disk fails during a rebuild the rebuild is aborted, so
> >it doesn't really matter.
>
>
> Even on a 3-disk mirror where there's a good data disk left to rebuild
> from?
> That sounds bad, and it looked to me like the code I quoted was looking for
> another working drive if one failed during rebuild/resync.
I'm faily sure that the resync is aborted. I think it then restarts
if there is useful work to be done.
This is probably not a good thing and should be fixed.
Also it is quite possible that the code is inconsistant and tries to
cope with a situation that doesn't happen.
>
>
> >However there are more cosmetic changes than real content, making it
> >harder for me to find the significant change. It is much easier to
> >read a patch if it just does one thing.
>
> diff -u --exclude-from=/home/me/.exclude -r linux-2.4.21-pre6-ref/drivers/md/raid1.c pre6-raid1d-a/drivers/md/raid1.c
> --- linux-2.4.21-pre6-ref/drivers/md/raid1.c Thu Apr 3 13:21:37 2003
> +++ pre6-raid1d-a/drivers/md/raid1.c Wed Apr 9 01:24:06 2003
> @ -1188,7 +1188,7 @
> for (i = 0; i < disks ; i++) {
> if (!conf->mirrors[i].operational)
> continue;
> - if (i==conf->last_used)
> + if (conf->mirrors[i].dev == bh->b_dev)
> /* we read from here, no need to write */
> continue;
> if (i < conf->raid_disks
>
Thanks. I don't think there is much point in making this change in
2.4 without fixing the rest of the related problems. However the
equivalent patch for 2.5:
--- ./drivers/md/raid1.c~current~ 2003-04-14 16:39:48.000000000 +1000
+++ ./drivers/md/raid1.c 2003-04-14 16:40:33.000000000 +1000
@@ -826,7 +826,7 @@ static void sync_request_write(mddev_t *
if (!conf->mirrors[i].rdev ||
conf->mirrors[i].rdev->faulty)
continue;
- if (i == conf->last_used)
+ if (conf->mirrors[i].rdev->bdev == bio->bi_bdev)
/*
* we read from here, no need to write
*/
is worth while and I will make sure it gets to Linus.
NeilBrown
-
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
