On Thursday June 15, [EMAIL PROTECTED] wrote:
> I am getting the same error as eric did you find a solution to the problem
> 

Apply the following patch to 2.4.0-test1-ac18 or later.  

Ingo: could you pease add this to your tree?


The problem occurs if you try to mkraid an array for that some reason
fails to run. The mddev structure gets freed, and then later unlocked ->
oops.

If you can figure out why it is failing to make your array, you could
work around the problem:-)  A possible reason is that it cannot find
the personality module.

The important change is in the second chunk.  The other two chunks are
related and partly cosmetic.

NeilBrown


--- ./drivers/block/md.c        2000/06/15 00:00:53     1.1
+++ ./drivers/block/md.c        2000/06/15 00:05:24     1.2
@@ -2730,11 +2730,9 @@
                        goto done_unlock;
 
                case STOP_ARRAY:
-                       err = do_md_stop (mddev, 0);
-                       if (err)
-                               goto done_unlock;
-                       else
-                               goto done;
+                       if (!(err = do_md_stop (mddev, 0)))
+                               mddev = NULL;
+                       goto done_unlock;
 
                case STOP_ARRAY_RO:
                        err = do_md_stop (mddev, 1);
@@ -2836,7 +2834,8 @@
                         */
                        if (err) {
                                mddev->sb_dirty = 0;
-                               do_md_stop (mddev, 0);
+                               if (!do_md_stop (mddev, 0))
+                                       mddev = NULL;
                        }
                        goto done_unlock;
                }
@@ -2851,8 +2850,6 @@
 abort_unlock:
        if (mddev)
                unlock_mddev(mddev);
-       else
-               printk("huh11?\n");
 
        return err;
 done:


>       -----Original Message-----
>       From:   Neil Brown [SMTP:[EMAIL PROTECTED]]
>       Sent:   Wednesday, June 07, 2000 11:15 PM
>       To:     Erik Petersen
>       Cc:     [EMAIL PROTECTED]
>       Subject:        Re: mkraid & 2.4.0-test1-ac10 fails
> 
>       On Wednesday June 7, [EMAIL PROTECTED] wrote:
>       > Is there some secret I need to know to get MD running on
> 2.4.0-test1?
>       > I'm using the tools from
> ftp.kernel.org/pub/linux/daemons/raid/alpha
>       > (Aug '99) but to no avail.
>       > 
>       > When I try to build a RAID5 with 3 SCSI disk partitions, I get a
> forced
>       > oops with a complaint about a bug in asm/semaphore.h line 185
> (magic
>       > number mismatch). If I #undef WAITQUEUE_DEBUG in asm/semaphore.h
> (which
>       > checks the magic number), mkraid complains: /dev/md0: Invalid
> argument.
>       > 
>       > Does MD work on 2.4.0-test1? Am I using the right tools? Is there
>       > someting I am missing?
>       > 
>       > --
>       > Erik Petersen
>       > <[EMAIL PROTECTED]>
> 
>       Hmmm. works for me.
> 
>       line 185 is inside "up()", and I have checked all the "up" calls in
>       md.c and raid5.c and convinced myself that the semaphore is properly
>       initialised, so something more obscure is going on.
> 
>       Can you get me a "ksymoops" of the Oops message that you get, and
> maybe
>       a copy of your .config?
> 
>       Thanks,
>       NeilBrown

Reply via email to