On Monday May 19, 3tcdgwg3@prodigy.net wrote:
> Hi,
>
> I am trying to simulate a case that two drives
> in an array fail ad same time.
> I use two ide drives, I try to create a
> raid 5 array with 4 arms, created as following:
>
> /dev/hdc1
> /dev/hde1
> /dev/hdc2
> /dev/hde2
>
> This is just for test, I know create two arms on
> one hard drive doesn't make much sense.
>
>
> Anyway, when I run this array, if I power off one
> of hard drive (/dev/hde) to simulate two arms failing
> at same time in an array, I got system Oops. I am using
> 2.4-18 kernel.
>
> Anyone can tell me if this is normal? or if there is a fix for this?
>
Congratulations and thanks. You have managed to trigger a bug that
no-one else has found.
The following patch (against 2.4.20) should fix it. If you can test
and confirm I would really appreciate it.
NeilBrown
------------------------------------------------------------
Handle concurrent failure of two drives in raid5
If two drives both fail during a write request, raid5 doesn't
cope properly and will eventually oops.
With this patch, blocks that have already been 'written'
are failed when double drive failure is noticed, as well as
blocks that are about to be written.
----------- Diffstat output ------------
./drivers/md/raid5.c | 10 +++++++++-
1 files changed, 9 insertions(+), 1 deletion(-)
diff ./drivers/md/raid5.c~current~ ./drivers/md/raid5.c
--- ./drivers/md/raid5.c~current~ 2003-05-21 12:42:07.000000000 +1000
+++ ./drivers/md/raid5.c 2003-05-21 12:37:37.000000000 +1000
@@ -882,7 +882,7 @@ static void handle_stripe(struct stripe_
/* check if the array has lost two devices and, if so, some requests might
* need to be failed
*/
- if (failed > 1 && to_read+to_write) {
+ if (failed > 1 && to_read+to_write+written) {
for (i=disks; i--; ) {
/* fail all writes first */
if (sh->bh_write[i]) to_write--;
@@ -891,6 +891,14 @@ static void handle_stripe(struct stripe_
bh->b_reqnext = return_fail;
return_fail = bh;
}
+ /* and fail all 'written' */
+ if (sh->bh_written[i]) written--;
+ while ((bh = sh->bh_written[i])) {
+ sh->bh_written[i] = bh->b_reqnext;
+ bh->b_reqnext = return_fail;
+ return_fail = bh;
+ }
+
/* fail any reads if this device is non-operational */
if (!conf->disks[i].operational) {
spin_lock_irq(&conf->device_lock);
-
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
