Raid1 and mdadm

Raid1 and mdadm

[Max Booker]

Hello

I am a Linux newbie, using Redhat Linux 7.3 Kernel 2.4.18-17.7.x on an Asus
7V333 with Promise Raid Controller (disabled).

I would like to understand more about Raid in a Linux environment, except
that all the documentation I have read seems not to apply to my case.

I understand that mdadm is better than raidtools, but I still cannot get my
head around the concept of how Linux Raid works.

Neil Brown gave a reply to splee in April, which seems to be relatively
straightforward - however, when I try this method, I get a kernel panic.

My setup is as follows:

GRUB boot loader

    hda1    /boot
    hda2    /
    hda3    swap

    hdd1    (intended) mirror partition of  /

I can create the (degraded) array on hdd1, mkfs on /dev/md0, mount /dev/md0
as /mnt, copy / to /mnt, and edit /mnt/etc/fstab.

Everything works well up to the point where I reboot, with kernel parameters
(md=0,/dev/hdd1 root=/dev/md0). This is where the kernel panics.

I think that I have not edited fstab properly, or GRUB is getting in the
way, or maybe it is not possible to add hda2 to the array because it already
has a filesystem on it.... How do I get hda2 to be part of the Raid array?
Nothing attempted so far has made any impression on the fact that mdadm
insists that /dev/hda2 is not an md component.

I hope that this is enough info.

Best regards

Max Booker

Re: Raid1 and mdadm

[Mads Peter Bach]

Max Booker wrote:
> Hello
> 
> Everything works well up to the point where I reboot, with kernel parameters
> (md=0,/dev/hdd1 root=/dev/md0). This is where the kernel panics.
> 
> I think that I have not edited fstab properly, or GRUB is getting in the
> way, or maybe it is not possible to add hda2 to the array because it already
> has a filesystem on it.... How do I get hda2 to be part of the Raid array?
> Nothing attempted so far has made any impression on the fact that mdadm
> insists that /dev/hda2 is not an md component.
> 
> I hope that this is enough info.

It isn't. We'll need to see the panic, and the things leading up to it, 
to be able to diagnose what's happening.

Does your kernel have the software raid compiled in, or as modules? If 
it's modules, you'll need an initial ramdisk (initrd).

You can add hda2 to the array, after you've successfully booted off the 
array.

-- 
Mads Peter Bach
Systemadministrator,  Det Humanistiske Fakultet, Aalborg Universitet
Kroghstræde 3 - 5.111, DK-9220 Aalborg Øst
# whois MPB1-DK@whois.dk-hostmaster.dk

-
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

Re: Raid1 and mdadm

[Friedrich Lobenstock]

Max Booker wrote:
> 
> GRUB boot loader
> 
>     hda1    /boot
>     hda2    /
>     hda3    swap
> 
>     hdd1    (intended) mirror partition of  /
> 
> I can create the (degraded) array on hdd1, mkfs on /dev/md0, mount /dev/md0
> as /mnt, copy / to /mnt, and edit /mnt/etc/fstab.
> 
> Everything works well up to the point where I reboot, with kernel parameters
> (md=0,/dev/hdd1 root=/dev/md0). This is where the kernel panics.
> 
> I think that I have not edited fstab properly, or GRUB is getting in the
> way, or maybe it is not possible to add hda2 to the array because it already
> has a filesystem on it.... How do I get hda2 to be part of the Raid array?
> Nothing attempted so far has made any impression on the fact that mdadm
> insists that /dev/hda2 is not an md component.
> 

As far as I know you have to use LILO instead of GRUB to boot
from a root raid. Take care that raid1 is compiled into the kernel.

-- 
MfG / Regards
Friedrich Lobenstock

RE: Raid1 and mdadm

[Sarcar, Shourya C (MED)]

Max,

Even if RAID is not compiled into the kernel, you can build an initrd
image (man mkinitrd) and change /etc/lilo.conf to point to the new
initrd image.
This way you will override the pains of a 'proper' kernel compile. 

Thanks
Shourya Sarcar 
_____________________________________________________
Global Recon Engine, Computed Tomography
Edison Engineering Development Program, GE Medical Systems
3000 N Grandview Blvd., MS W 1200, Waukesha, WI 53188, USA
Tel: 262 312 7405 (O) | *320 7405 (DialComm)
 


> -----Original Message-----
> From: Friedrich Lobenstock [mailto:fl@fl.priv.at]
> Sent: Monday, March 31, 2003 5:35 AM
> To: Linux RAID Mailing Liste
> Subject: Re: Raid1 and mdadm
> 
> 
> Max Booker wrote:
> > 
> > GRUB boot loader
> > 
> >     hda1    /boot
> >     hda2    /
> >     hda3    swap
> > 
> >     hdd1    (intended) mirror partition of  /
> > 
> > I can create the (degraded) array on hdd1, mkfs on 
> /dev/md0, mount /dev/md0
> > as /mnt, copy / to /mnt, and edit /mnt/etc/fstab.
> > 
> > Everything works well up to the point where I reboot, with 
> kernel parameters
> > (md=0,/dev/hdd1 root=/dev/md0). This is where the kernel panics.
> > 
> > I think that I have not edited fstab properly, or GRUB is 
> getting in the
> > way, or maybe it is not possible to add hda2 to the array 
> because it already
> > has a filesystem on it.... How do I get hda2 to be part of 
> the Raid array?
> > Nothing attempted so far has made any impression on the 
> fact that mdadm
> > insists that /dev/hda2 is not an md component.
> > 
> 
> As far as I know you have to use LILO instead of GRUB to boot
> from a root raid. Take care that raid1 is compiled into the kernel.
> 
> -- 
> MfG / Regards
> Friedrich Lobenstock
> 
> -
> 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
> 

RE: Raid1 and mdadm

[Cress, Andrew R]

You can use grub for root mirroring (software raid-1), but grub currently
does not write the Master Boot Record to the 2nd disk, so you have to do
that manually the first time.  Sample script below:
# 
tmpa=/tmp/a
tmpg=/tmp/grub.out
cat - <<%%% >$tmpa
install (hd0,0)/grub/stage1 (hd1) (hd1,0)/grub/stage2 (hd0,0)/grub/grub.conf
quit
%%%
grub <$tmpa  >$tmpg 2>&1
# 

However, you MUST link the md driver into the kernel.  It doesn't work right
as a module, even in initrd.
That's probably why it panic'd.

Andy

-----Original Message-----
From: Friedrich Lobenstock [mailto:fl@fl.priv.at] 
Sent: Monday, March 31, 2003 6:35 AM
To: Linux RAID Mailing Liste
Subject: Re: Raid1 and mdadm


Max Booker wrote:
> 
> GRUB boot loader
> 
>     hda1    /boot
>     hda2    /
>     hda3    swap
> 
>     hdd1    (intended) mirror partition of  /
> 
> I can create the (degraded) array on hdd1, mkfs on /dev/md0, mount
/dev/md0
> as /mnt, copy / to /mnt, and edit /mnt/etc/fstab.
> 
> Everything works well up to the point where I reboot, with kernel
parameters
> (md=0,/dev/hdd1 root=/dev/md0). This is where the kernel panics.
> 
> I think that I have not edited fstab properly, or GRUB is getting in the
> way, or maybe it is not possible to add hda2 to the array because it
already
> has a filesystem on it.... How do I get hda2 to be part of the Raid array?
> Nothing attempted so far has made any impression on the fact that mdadm
> insists that /dev/hda2 is not an md component.
> 

As far as I know you have to use LILO instead of GRUB to boot
from a root raid. Take care that raid1 is compiled into the kernel.

-- 
MfG / Regards
Friedrich Lobenstock

-
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

Re: Raid1 and mdadm

[Paul Clements]

"Cress, Andrew R" wrote:

> However, you MUST link the md driver into the kernel.  It doesn't work right
> as a module, even in initrd.

BTW, it would be nice if this were fixed, so that md could be compiled
as a module...would the maintainers be willing to accept patches that
would make "autostart" a module option, so that upon insertion of the md
module it would do the autostart stuff, or something similar? From what
I can tell, we basically need to remove some ifndef MODULE stuff and
make the autostart option call autostart_arrays(), as it is when md is
linked into the kernel...

Or is there a better way to do this?

--
Paul

RE: Raid1 and mdadm

[Cress, Andrew R]

Well, it might be convenient otherwise, but it makes sense that if you want
to boot from the software RAID, that you have to ensure that the md driver
is loaded early enough to recognize that the RAID is configured before it
tries to boot from SCSI alone.

If you aren't booting from the software RAID, md can be a module.  
Also, the other SCSI and RAID components can be modules as long as
CONFIG_BLK_DEV_MD=y.

Andy

-----Original Message-----
From: Paul Clements [mailto:Paul.Clements@SteelEye.com] 
Sent: Tuesday, April 01, 2003 11:08 AM
To: linux-raid@vger.kernel.org
Subject: Re: Raid1 and mdadm


"Cress, Andrew R" wrote:

> However, you MUST link the md driver into the kernel.  It doesn't work
right
> as a module, even in initrd.

BTW, it would be nice if this were fixed, so that md could be compiled
as a module...would the maintainers be willing to accept patches that
would make "autostart" a module option, so that upon insertion of the md
module it would do the autostart stuff, or something similar? From what
I can tell, we basically need to remove some ifndef MODULE stuff and
make the autostart option call autostart_arrays(), as it is when md is
linked into the kernel...

Or is there a better way to do this?

--
Paul
-
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

Re: Raid1 and mdadm

[Neil Brown]

On Tuesday April 1, Paul.Clements@SteelEye.com wrote:
> "Cress, Andrew R" wrote:
> 
> > However, you MUST link the md driver into the kernel.  It doesn't work right
> > as a module, even in initrd.
> 
> BTW, it would be nice if this were fixed, so that md could be compiled
> as a module...would the maintainers be willing to accept patches that
> would make "autostart" a module option, so that upon insertion of the md
> module it would do the autostart stuff, or something similar? From what
> I can tell, we basically need to remove some ifndef MODULE stuff and
> make the autostart option call autostart_arrays(), as it is when md is
> linked into the kernel...

No.  I would not accept patches.
Not because I fundamentally disagree with autostart, but because it
wouldn't work.  I suggest you try it.

> 
> Or is there a better way to do this?

Yes.
initrd/initfs is "the right way".
As far as I know, it *does* work.(*)  You do not need md linked into the
kernel.
You have md.o and whatever personalities you want in the initrd image,
then in the initrd script (linuxrc I think) you load the modules, and
use mdadm to assemble the raid arrays.
You then mount the real root and off you go.

NeilBrown

(*) I say "as far as I know" because I have never tried it, because I
have not yet found the need.  However I believe that others have been
successful. 
I have perfectly happy compiling the raid driver into my kernels so
I have never bothered with initrd.

Re: Raid1 and mdadm

[Paul Clements]

[-- Attachment #1: Type: text/plain, Size: 1556 bytes --]

Hi Neil,

> No.  I would not accept patches.
> Not because I fundamentally disagree with autostart, but because it
> wouldn't work.  I suggest you try it.

Maybe I didn't explain very clearly what my goal is...allow me to
clarify:

As it is now, if you compile md as a module and happen to have a
partition tagged as RAID, you'll panic the kernel. There's code in
fs/partitions (and in 2.5, in init/do_mounts.c) that is calling into
md....yechhh. What I'd like to do is make a clean separation between the
boot up and partitioning code and the md driver, so md can be compiled
as a module. (This would also allow other volume managers to use the
partition tagging to identify devices, if they wanted.) 


> > Or is there a better way to do this?
> 
> Yes.
> initrd/initfs is "the right way".
> As far as I know, it *does* work.(*)  You do not need md linked into the
> kernel.
> You have md.o and whatever personalities you want in the initrd image,
> then in the initrd script (linuxrc I think) you load the modules, and
> use mdadm to assemble the raid arrays.
> You then mount the real root and off you go.

Right, sorry, I wasn't clear. This is basically what I meant. All I want
to do is allow autostart_arrays() to be called when md is compiled as a
module. Then, let the initrd/linuxrc (later initramfs) call
autostart_arrays() (ioctl("/dev/md0", RAID_AUTORUN) is all, really).

I've hacked on 2.5.63. I'll attach the patch, it's fairly short...and
might give you a better idea what I'm proposing...

Any comments would be appreciated.

Thanks,
Paul

[-- Attachment #2: md_modular_patch.diff --]
[-- Type: text/plain, Size: 5543 bytes --]

--- fs/partitions/Makefile.PRISTINE	2003-04-01 12:12:55.000000000 -0500
+++ fs/partitions/Makefile	2003-04-01 12:13:03.000000000 -0500
@@ -2,7 +2,7 @@
 # Makefile for the linux kernel.
 #
 
-obj-y := check.o
+obj-y := check.o raid.o
 
 obj-$(CONFIG_ACORN_PARTITION) += acorn.o
 obj-$(CONFIG_AMIGA_PARTITION) += amiga.o
--- fs/partitions/check.c.PRISTINE	2003-04-01 12:14:12.000000000 -0500
+++ fs/partitions/check.c	2003-04-01 12:15:36.000000000 -0500
@@ -21,6 +21,7 @@
 #include <linux/ctype.h>
 
 #include "check.h"
+#include "raid.h"
 
 #include "acorn.h"
 #include "amiga.h"
@@ -35,10 +36,6 @@
 #include "ultrix.h"
 #include "efi.h"
 
-#if CONFIG_BLK_DEV_MD
-extern void md_autodetect_dev(dev_t dev);
-#endif
-
 int warn_no_part = 1; /*This is ugly: should make genhd removable media aware*/
 
 static int (*check_part[])(struct parsed_partitions *, struct block_device *) = {
@@ -423,7 +420,7 @@ void register_disk(struct gendisk *disk)
 #if CONFIG_BLK_DEV_MD
 			if (!state->parts[j].flags)
 				continue;
-			md_autodetect_dev(bdev->bd_dev+j);
+			raid_autodetect_dev(bdev->bd_dev+j);
 #endif
 		}
 		kfree(state);
@@ -457,7 +454,7 @@ int rescan_partitions(struct gendisk *di
 		add_partition(disk, p, from, size);
 #if CONFIG_BLK_DEV_MD
 		if (state->parts[p].flags)
-			md_autodetect_dev(bdev->bd_dev+p);
+			raid_autodetect_dev(bdev->bd_dev+p);
 #endif
 	}
 	kfree(state);
--- fs/partitions/check.c.PRISTINE	2003-04-01 12:14:12.000000000 -0500
+++ fs/partitions/check.c	2003-04-01 12:15:36.000000000 -0500
@@ -21,6 +21,7 @@
 #include <linux/ctype.h>
 
 #include "check.h"
+#include "raid.h"
 
 #include "acorn.h"
 #include "amiga.h"
@@ -35,10 +36,6 @@
 #include "ultrix.h"
 #include "efi.h"
 
-#if CONFIG_BLK_DEV_MD
-extern void md_autodetect_dev(dev_t dev);
-#endif
-
 int warn_no_part = 1; /*This is ugly: should make genhd removable media aware*/
 
 static int (*check_part[])(struct parsed_partitions *, struct block_device *) = {
@@ -423,7 +420,7 @@ void register_disk(struct gendisk *disk)
 #if CONFIG_BLK_DEV_MD
 			if (!state->parts[j].flags)
 				continue;
-			md_autodetect_dev(bdev->bd_dev+j);
+			raid_autodetect_dev(bdev->bd_dev+j);
 #endif
 		}
 		kfree(state);
@@ -457,7 +454,7 @@ int rescan_partitions(struct gendisk *di
 		add_partition(disk, p, from, size);
 #if CONFIG_BLK_DEV_MD
 		if (state->parts[p].flags)
-			md_autodetect_dev(bdev->bd_dev+p);
+			raid_autodetect_dev(bdev->bd_dev+p);
 #endif
 	}
 	kfree(state);
--- /dev/null	2002-12-05 15:33:16.000000000 -0500
+++ fs/partitions/raid.c	2003-04-01 13:16:53.000000000 -0500
@@ -0,0 +1,28 @@
+/*
+ * Save registered partitions that are marked as autorun RAID arrays.
+ *
+ * check.c saves the partitions by calling raid_autodetect_dev at boot time
+ * md.c uses the partition list to do its autostart_arrays 
+ *
+ * code taken from md.c and new code added - 2003, Paul Clements
+ */
+static dev_t detected_devices[128];
+static int dev_cnt;
+
+void raid_autodetect_dev(dev_t dev)
+{
+        if (dev_cnt >= 0 && dev_cnt < 128)
+                detected_devices[dev_cnt++] = dev;
+}
+
+dev_t raid_get_detected_dev(int i)
+{
+	if (i >= 0 && i < dev_cnt)
+		return detected_devices[i];
+	return (dev_t) 0;
+}
+
+int raid_num_detected_devs(void)
+{
+	return dev_cnt;
+}
--- /dev/null	2002-12-05 15:33:16.000000000 -0500
+++ fs/partitions/raid.h	2003-04-01 12:17:30.000000000 -0500
@@ -0,0 +1,5 @@
+/*
+ *  fs/partitions/raid.h
+ */
+
+void raid_autodetect_dev(dev_t dev);
--- drivers/md/md.c.PRISTINE	2003-04-01 11:37:58.000000000 -0500
+++ drivers/md/md.c	2003-04-02 09:59:10.000000000 -0500
@@ -59,9 +59,7 @@
 #define dprintk(x...) ((void)(DEBUG && printk(x)))
 
 
-#ifndef MODULE
 static void autostart_arrays (void);
-#endif
 
 static mdk_personality_t *pers[MAX_PERSONALITY];
 
@@ -873,7 +871,9 @@ static void unlock_rdev(mdk_rdev_t *rdev
 	blkdev_put(bdev, BDEV_RAW);
 }
 
-void md_autodetect_dev(dev_t dev);
+extern void raid_autodetect_dev(dev_t);
+extern dev_t raid_get_detected_dev(int);
+extern int raid_num_detected_devs(void);
 
 static void export_rdev(mdk_rdev_t * rdev)
 {
@@ -882,9 +882,7 @@ static void export_rdev(mdk_rdev_t * rde
 		MD_BUG();
 	free_disk_sb(rdev);
 	list_del_init(&rdev->same_set);
-#ifndef MODULE
-	md_autodetect_dev(rdev->bdev->bd_dev);
-#endif
+	raid_autodetect_dev(rdev->bdev->bd_dev);
 	unlock_rdev(rdev);
 	kfree(rdev);
 }
@@ -2225,12 +2223,10 @@ static int md_ioctl(struct inode *inode,
 			md_print_devices();
 			goto done;
 
-#ifndef MODULE
 		case RAID_AUTORUN:
 			err = 0;
 			autostart_arrays();
 			goto done;
-#endif
 		default:;
 	}
 
@@ -3237,23 +3233,11 @@ int __init md_init(void)
 	raid_table_header = register_sysctl_table(raid_root_table, 1);
 
 	md_geninit();
-	return (0);
-}
-
 
-#ifndef MODULE
+	if (autostart) 
+		autostart_arrays();
 
-/*
- * Searches all registered partitions for autorun RAID arrays
- * at boot time.
- */
-static dev_t detected_devices[128];
-static int dev_cnt;
-
-void md_autodetect_dev(dev_t dev)
-{
-	if (dev_cnt >= 0 && dev_cnt < 127)
-		detected_devices[dev_cnt++] = dev;
+	return (0);
 }
 
 
@@ -3264,8 +3248,8 @@ static void autostart_arrays(void)
 
 	printk(KERN_INFO "md: Autodetecting RAID arrays.\n");
 
-	for (i = 0; i < dev_cnt; i++) {
-		dev_t dev = detected_devices[i];
+	for (i = 0; i < raid_num_detected_devs(); i++) {
+		dev_t dev = raid_get_detected_dev(i);
 
 		rdev = md_import_device(dev,1);
 		if (IS_ERR(rdev)) {
@@ -3284,8 +3268,6 @@ static void autostart_arrays(void)
 	autorun_devices();
 }
 
-#endif
-
 static __exit void md_exit(void)
 {
 	int i;

Re: Raid1 and mdadm

[Paul Clements]

[-- Attachment #1: Type: text/plain, Size: 212 bytes --]

Paul Clements wrote:
> I've hacked on 2.5.63. I'll attach the patch, it's fairly short...and
> might give you a better idea what I'm proposing...

aghh...that patch is mangled...here's a better one... :)

--
Paul

[-- Attachment #2: md_modular_patch.diff --]
[-- Type: text/plain, Size: 4479 bytes --]

--- fs/partitions/Makefile.PRISTINE	2003-04-01 12:12:55.000000000 -0500
+++ fs/partitions/Makefile	2003-04-01 12:13:03.000000000 -0500
@@ -2,7 +2,7 @@
 # Makefile for the linux kernel.
 #
 
-obj-y := check.o
+obj-y := check.o raid.o
 
 obj-$(CONFIG_ACORN_PARTITION) += acorn.o
 obj-$(CONFIG_AMIGA_PARTITION) += amiga.o
--- fs/partitions/check.c.PRISTINE	2003-04-01 12:14:12.000000000 -0500
+++ fs/partitions/check.c	2003-04-01 12:15:36.000000000 -0500
@@ -21,6 +21,7 @@
 #include <linux/ctype.h>
 
 #include "check.h"
+#include "raid.h"
 
 #include "acorn.h"
 #include "amiga.h"
@@ -35,10 +36,6 @@
 #include "ultrix.h"
 #include "efi.h"
 
-#if CONFIG_BLK_DEV_MD
-extern void md_autodetect_dev(dev_t dev);
-#endif
-
 int warn_no_part = 1; /*This is ugly: should make genhd removable media aware*/
 
 static int (*check_part[])(struct parsed_partitions *, struct block_device *) = {
@@ -423,7 +420,7 @@ void register_disk(struct gendisk *disk)
 #if CONFIG_BLK_DEV_MD
 			if (!state->parts[j].flags)
 				continue;
-			md_autodetect_dev(bdev->bd_dev+j);
+			raid_autodetect_dev(bdev->bd_dev+j);
 #endif
 		}
 		kfree(state);
@@ -457,7 +454,7 @@ int rescan_partitions(struct gendisk *di
 		add_partition(disk, p, from, size);
 #if CONFIG_BLK_DEV_MD
 		if (state->parts[p].flags)
-			md_autodetect_dev(bdev->bd_dev+p);
+			raid_autodetect_dev(bdev->bd_dev+p);
 #endif
 	}
 	kfree(state);
--- /dev/null	2002-12-05 15:33:16.000000000 -0500
+++ fs/partitions/raid.c	2003-04-01 13:16:53.000000000 -0500
@@ -0,0 +1,28 @@
+/*
+ * Save registered partitions that are marked as autorun RAID arrays.
+ *
+ * check.c saves the partitions by calling raid_autodetect_dev at boot time
+ * md.c uses the partition list to do its autostart_arrays 
+ *
+ * code taken from md.c and new code added - 2003, Paul Clements
+ */
+static dev_t detected_devices[128];
+static int dev_cnt;
+
+void raid_autodetect_dev(dev_t dev)
+{
+        if (dev_cnt >= 0 && dev_cnt < 128)
+                detected_devices[dev_cnt++] = dev;
+}
+
+dev_t raid_get_detected_dev(int i)
+{
+	if (i >= 0 && i < dev_cnt)
+		return detected_devices[i];
+	return (dev_t) 0;
+}
+
+int raid_num_detected_devs(void)
+{
+	return dev_cnt;
+}
--- /dev/null	2002-12-05 15:33:16.000000000 -0500
+++ fs/partitions/raid.h	2003-04-01 12:17:30.000000000 -0500
@@ -0,0 +1,5 @@
+/*
+ *  fs/partitions/raid.h
+ */
+
+void raid_autodetect_dev(dev_t dev);
--- drivers/md/md.c.PRISTINE	2003-04-01 11:37:58.000000000 -0500
+++ drivers/md/md.c	2003-04-02 09:59:10.000000000 -0500
@@ -59,9 +59,7 @@
 #define dprintk(x...) ((void)(DEBUG && printk(x)))
 
 
-#ifndef MODULE
 static void autostart_arrays (void);
-#endif
 
 static mdk_personality_t *pers[MAX_PERSONALITY];
 
@@ -873,7 +871,9 @@ static void unlock_rdev(mdk_rdev_t *rdev
 	blkdev_put(bdev, BDEV_RAW);
 }
 
-void md_autodetect_dev(dev_t dev);
+extern void raid_autodetect_dev(dev_t);
+extern dev_t raid_get_detected_dev(int);
+extern int raid_num_detected_devs(void);
 
 static void export_rdev(mdk_rdev_t * rdev)
 {
@@ -882,9 +882,7 @@ static void export_rdev(mdk_rdev_t * rde
 		MD_BUG();
 	free_disk_sb(rdev);
 	list_del_init(&rdev->same_set);
-#ifndef MODULE
-	md_autodetect_dev(rdev->bdev->bd_dev);
-#endif
+	raid_autodetect_dev(rdev->bdev->bd_dev);
 	unlock_rdev(rdev);
 	kfree(rdev);
 }
@@ -2225,12 +2223,10 @@ static int md_ioctl(struct inode *inode,
 			md_print_devices();
 			goto done;
 
-#ifndef MODULE
 		case RAID_AUTORUN:
 			err = 0;
 			autostart_arrays();
 			goto done;
-#endif
 		default:;
 	}
 
@@ -3237,23 +3233,11 @@ int __init md_init(void)
 	raid_table_header = register_sysctl_table(raid_root_table, 1);
 
 	md_geninit();
-	return (0);
-}
-
 
-#ifndef MODULE
+	if (autostart) 
+		autostart_arrays();
 
-/*
- * Searches all registered partitions for autorun RAID arrays
- * at boot time.
- */
-static dev_t detected_devices[128];
-static int dev_cnt;
-
-void md_autodetect_dev(dev_t dev)
-{
-	if (dev_cnt >= 0 && dev_cnt < 127)
-		detected_devices[dev_cnt++] = dev;
+	return (0);
 }
 
 
@@ -3264,8 +3248,8 @@ static void autostart_arrays(void)
 
 	printk(KERN_INFO "md: Autodetecting RAID arrays.\n");
 
-	for (i = 0; i < dev_cnt; i++) {
-		dev_t dev = detected_devices[i];
+	for (i = 0; i < raid_num_detected_devs(); i++) {
+		dev_t dev = raid_get_detected_dev(i);
 
 		rdev = md_import_device(dev,1);
 		if (IS_ERR(rdev)) {
@@ -3284,8 +3268,6 @@ static void autostart_arrays(void)
 	autorun_devices();
 }
 
-#endif
-
 static __exit void md_exit(void)
 {
 	int i;