dac960 broken ?

dac960 broken ?

[Pascal Lengard]

Hello,

I have several "servers" using old mylex DAC960P scsi raid adapter.
So I though clever to install redhat 7.2 on them ...

redhat 7.2 does install and run well with the 2.4.7-10 kernel from the
distribution, but when I try upgrading to 2.4.9-13 (via rpm) it does not boot, 
there is a problem with resolving ext3fs symbols ... this is more a RedHat
problem, but read on :-)

I choosed to compile a customized kernel with only what I need inside kernel
(ext3fs, dac960, ..) plus some modules I might need some day.

I tried compiling 2.4.14 => my mylex card is not detected !
	(driver dac960 version 2.4.11 from 11 october 2001)
I tried with 2.4.9-13 from redhat => same problem
	(driver dac960 version 2.4.10 from 23 july 2001)
I tried custom 2.4.7-10 from redhat => works like a charm
	(driver dac960 version 2.4.10 from 1 february 2001)
all these kernels were compiled with the same .config (make oldconfig)

hardware used:
	DAC960P-2,  D040351-0-IBM REV.E firmware 3.51-0-04

detected like this by kernel 2.4.7-10:
DAC960: ***** DAC960 RAID Driver Version 2.4.10 of 1 February 2001 *****
DAC960: Copyright 1998-2001 by Leonard N. Zubkoff <lnz@dandelion.com>
DAC960#0: Configuring Mylex DAC960PD PCI RAID Controller
DAC960#0:   Firmware Version: 3.51-0-04, Channels: 2, Memory Size: 4MB
DAC960#0:   PCI Bus: 1, Device: 10, Function: 0, I/O Address: 0x6200
DAC960#0:   PCI Address: 0xBF800C00 mapped at 0xC482DC00, IRQ Channel: 11
DAC960#0:   Controller Queue Depth: 64, Maximum Blocks per Command: 128
DAC960#0:   Driver Queue Depth: 63, Scatter/Gather Limit: 17 of 17 Segments
DAC960#0:   Stripe Size: 64KB, Segment Size: 8KB, BIOS Geometry: 128/32


I am insterested in any clue, since I am stuck to 2.4.7 for now ...
linux-kernel readers, please cc me on replies since I am not subscribed
to the list.

Pascal Lengard

Re: dac960 broken ?

[Alan Cox]

> distribution, but when I try upgrading to 2.4.9-13 (via rpm) it does not boot, 
> there is a problem with resolving ext3fs symbols ... this is more a RedHat
> problem, but read on :-)

You don't provide enough information. Precisely what symbols, precisely what
error message

Alan

Re: dac960 broken ?

[Leonard N. Zubkoff]

  Date: Wed, 28 Nov 2001 15:19:38 +0100 (CET)
  From: Pascal Lengard <pascal.lengard@wanadoo.fr>

  Hello,

  I have several "servers" using old mylex DAC960P scsi raid adapter.
  So I though clever to install redhat 7.2 on them ...

  redhat 7.2 does install and run well with the 2.4.7-10 kernel from the
  distribution, but when I try upgrading to 2.4.9-13 (via rpm) it does not boot, 
  there is a problem with resolving ext3fs symbols ... this is more a RedHat
  problem, but read on :-)

  I choosed to compile a customized kernel with only what I need inside kernel
  (ext3fs, dac960, ..) plus some modules I might need some day.

  I tried compiling 2.4.14 => my mylex card is not detected !
	  (driver dac960 version 2.4.11 from 11 october 2001)
  I tried with 2.4.9-13 from redhat => same problem
	  (driver dac960 version 2.4.10 from 23 july 2001)
  I tried custom 2.4.7-10 from redhat => works like a charm
	  (driver dac960 version 2.4.10 from 1 february 2001)
  all these kernels were compiled with the same .config (make oldconfig)

  hardware used:
	  DAC960P-2,  D040351-0-IBM REV.E firmware 3.51-0-04

  detected like this by kernel 2.4.7-10:
  DAC960: ***** DAC960 RAID Driver Version 2.4.10 of 1 February 2001 *****
  DAC960: Copyright 1998-2001 by Leonard N. Zubkoff <lnz@dandelion.com>
  DAC960#0: Configuring Mylex DAC960PD PCI RAID Controller
  DAC960#0:   Firmware Version: 3.51-0-04, Channels: 2, Memory Size: 4MB
  DAC960#0:   PCI Bus: 1, Device: 10, Function: 0, I/O Address: 0x6200
  DAC960#0:   PCI Address: 0xBF800C00 mapped at 0xC482DC00, IRQ Channel: 11
  DAC960#0:   Controller Queue Depth: 64, Maximum Blocks per Command: 128
  DAC960#0:   Driver Queue Depth: 63, Scatter/Gather Limit: 17 of 17 Segments
  DAC960#0:   Stripe Size: 64KB, Segment Size: 8KB, BIOS Geometry: 128/32


  I am insterested in any clue, since I am stuck to 2.4.7 for now ...
  linux-kernel readers, please cc me on replies since I am not subscribed
  to the list.

  Pascal Lengard

Hmmm.  Nothing you've described makes any sense to me as I don't believe the
driver has changed in a way that would break the basic detection of the boards.
When you say that the card is not detected, precisely what do you mean?  Does
the driver report anything at all?

		Leonard

Re: dac960 broken ?

[Doug Ledford]

Leonard N. Zubkoff wrote:

>   Date: Wed, 28 Nov 2001 15:19:38 +0100 (CET)
>   From: Pascal Lengard <pascal.lengard@wanadoo.fr>
> 
>   Hello,
> 
>   I have several "servers" using old mylex DAC960P scsi raid adapter.
>   So I though clever to install redhat 7.2 on them ...
> 
>   redhat 7.2 does install and run well with the 2.4.7-10 kernel from the
>   distribution, but when I try upgrading to 2.4.9-13 (via rpm) it does not boot, 
>   there is a problem with resolving ext3fs symbols ... this is more a RedHat
>   problem, but read on :-)


Sounds pretty simple to me.  If you are using lilo, then you might not 
be rerunning lilo after installing the new kernel.  Also, after 
installing a new kernel you need a new initrd image.  Your unresolved 
symbols problem is because you are trying to load the ext3 module from a 
2.4.7-10 kernel in the 2.4.9-13 errata kernel.  The general answer here 
is that there is significant documentation on what must be done to 
upgrade a Red Hat kernel.  Read it.


>   I choosed to compile a customized kernel with only what I need inside kernel
>   (ext3fs, dac960, ..) plus some modules I might need some day.
> 
>   I tried compiling 2.4.14 => my mylex card is not detected !
> 	  (driver dac960 version 2.4.11 from 11 october 2001)
>   I tried with 2.4.9-13 from redhat => same problem
> 	  (driver dac960 version 2.4.10 from 23 july 2001)
>   I tried custom 2.4.7-10 from redhat => works like a charm
> 	  (driver dac960 version 2.4.10 from 1 february 2001)
>   all these kernels were compiled with the same .config (make oldconfig)
> 
>   hardware used:
> 	  DAC960P-2,  D040351-0-IBM REV.E firmware 3.51-0-04
> 
>   detected like this by kernel 2.4.7-10:
>   DAC960: ***** DAC960 RAID Driver Version 2.4.10 of 1 February 2001 *****
>   DAC960: Copyright 1998-2001 by Leonard N. Zubkoff <lnz@dandelion.com>
>   DAC960#0: Configuring Mylex DAC960PD PCI RAID Controller
>   DAC960#0:   Firmware Version: 3.51-0-04, Channels: 2, Memory Size: 4MB
>   DAC960#0:   PCI Bus: 1, Device: 10, Function: 0, I/O Address: 0x6200
>   DAC960#0:   PCI Address: 0xBF800C00 mapped at 0xC482DC00, IRQ Channel: 11
>   DAC960#0:   Controller Queue Depth: 64, Maximum Blocks per Command: 128
>   DAC960#0:   Driver Queue Depth: 63, Scatter/Gather Limit: 17 of 17 Segments
>   DAC960#0:   Stripe Size: 64KB, Segment Size: 8KB, BIOS Geometry: 128/32
> 
> 
>   I am insterested in any clue, since I am stuck to 2.4.7 for now ...
>   linux-kernel readers, please cc me on replies since I am not subscribed
>   to the list.
> 
>   Pascal Lengard
> 
> Hmmm.  Nothing you've described makes any sense to me as I don't believe the
> driver has changed in a way that would break the basic detection of the boards.
> When you say that the card is not detected, precisely what do you mean?  Does
> the driver report anything at all?
> 
> 		Leonard
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 
> 



-- 

  Doug Ledford <dledford@redhat.com>  http://people.redhat.com/dledford
       Please check my web site for aic7xxx updates/answers before
                       e-mailing me about problems

Re: dac960 broken ?

[Pascal Lengard]

Hello Alan,

Sorry, I must apology here. I just rerun the redhat kernel and the error
is not linked to ext3fs (anyway it made no sens since the user base of
ext3 is HUGE and no problem was reported ...).

The problem is linked to DAC960 also:
here are the messages:

	Loading scsi_mod module
	Loading DAC960 module
	/lib/DAC960.o: init_module: Operation not permitted
	Hint: insmod errors can be caused by incorrect module parameters, including invalid IO or IRQ parameters
	ERROR /bin/insmod exited abnormally!
	Loading jbd module
	Journalled Block Device driver loaded
	Loading ext3 module
	Mounting /proc filesystem
	Creaing root device
	Mounting root filesystem
	mount: error 19 mounting ext3
	pivotroot: pivot_root(/sysroot,/sysroot/initrd) failed: 2
	Freeing unused kernel memory: 216k freed
	Kernel panic: No init found. Try passing init= option to kernel

Pascal Lengard

On Wed, 28 Nov 2001, Alan Cox wrote:
> You don't provide enough information. Precisely what symbols, precisely what
> error message
> 
> Alan
> 

Re: dac960 broken ?

[John Jasen]

Redhat's bugzilla does have a few entries regarding the DAC960 and RH 7.1.
Strangely, I can't see the one my friend recently reported ...

--
-- John E. Jasen (jjasen1@umbc.edu)
-- In theory, theory and practise are the same. In practise, they aren't.

Re: dac960 broken ?

[Pascal Lengard]

On Wed, 28 Nov 2001, Leonard N. Zubkoff wrote:

> Hmmm.  Nothing you've described makes any sense to me as I don't believe the
> driver has changed in a way that would break the basic detection of the boards.
> When you say that the card is not detected, precisely what do you mean?  Does
> the driver report anything at all?

OK, I did not write down the messages, but while testing, with DAC960
compiled as module, It did load and initialise but said "no such peripheral"
or something equivalent once (I know it's not precise enough ...).

I ran some tests to record precise messages today:

Here are the messages from the kernel 2.4.9-13 shipped by redhat:
-----------------------------------------------------------------
	Loading scsi_mod module
	Loading DAC960 module
	/lib/DAC960.o: init_module: Operation not permitted
	Hint: insmod errors can be caused by incorrect module parameters, including invalid IO or IRQ parameters
	ERROR /bin/insmod exited abnormally!
	Loading jbd module
	Journalled Block Device driver loaded
	Loading ext3 module
	Mounting /proc filesystem
	Creaing root device
	Mounting root filesystem
	mount: error 19 mounting ext3
	pivotroot: pivot_root(/sysroot,/sysroot/initrd) failed: 2
	Freeing unused kernel memory: 216k freed
	Kernel panic: No init found. Try passing init= option to kernel


Here is dmesg of 2.4.7-10 with dac960 compiled INSIDE kernel (WORKS):
---------------------------------------------------------------------
	Journalled Block Device driver loaded
	pty: 512 Unix98 ptys configured
	Serial driver version 5.05c (2001-07-08) with MANY_PORTS SHARE_IRQ SERIAL_PCI enabled
	ttyS00 at 0x03f8 (irq = 4) is a 16550A
	ttyS01 at 0x02f8 (irq = 3) is a 16550A
	Real Time Clock Driver v1.10d
	block: queued sectors max/low 41557kB/13852kB, 128 slots per queue
	RAMDISK driver initialized: 16 RAM disks of 4096K size 1024 blocksize
	Floppy drive(s): fd0 is 2.88M
	FDC 0 is an 82078.
	DAC960: ***** DAC960 RAID Driver Version 2.4.10 of 1 February 2001 *****
	DAC960: Copyright 1998-2001 by Leonard N. Zubkoff <lnz@dandelion.com>
	DAC960#0: Configuring Mylex DAC960PD PCI RAID Controller
	DAC960#0:   Firmware Version: 3.51-0-04, Channels: 2, Memory Size: 4MB
	DAC960#0:   PCI Bus: 1, Device: 10, Function: 0, I/O Address: 0x6200
	DAC960#0:   PCI Address: 0xBF800C00 mapped at 0x45002C00, IRQ Channel: 11
	DAC960#0:   Controller Queue Depth: 64, Maximum Blocks per Command: 128
	DAC960#0:   Driver Queue Depth: 63, Scatter/Gather Limit: 17 of 17 Segments
	DAC960#0:   Stripe Size: 64KB, Segment Size: 8KB, BIOS Geometry: 128/32
	DAC960#0:   Physical Devices:
	<snip the devices>
	DAC960#0:   Logical Drives:
	<snip logical drives>
	Partition check:
	<snip>
	loop: loaded (max 8 devices)
	SCSI subsystem driver Revision: 1.00
	request_module[scsi_hostadapter]: Root fs not mounted
	NET4: Linux TCP/IP 1.0 for NET4.0

Here are messages from 2.4.16 compiled INSIDE kernel (BROKEN)
with this method:
copy .config from 2.4.7-10 that worked (dmesg just above)
and "make oldconfig"
------------------------------------------------------------
	Journalled Block device driver loaded
	pty 512 Unix98 ptys configured
	Serial driver version 5.05c ....
	ttyS0 at ...
	ttyS1 at ...
	Real Time Clock Driver 1.10e
	block: 128 slots per queue, batch=32
	RAMDISK driver initilized ...
	Floppy drives ...
	FDC0 is an 82078

here DAC960 should talk!!

	loop: loaded (max 8 devices)
	SCSI subsystem driver Revision: 1.00
	request_module [scsi_hostadapter]: Root fs not mounted
	NET4: Linux TCP/IP ...
	IP Protocols ...
	IP: Routing ...
	TCP: Hash ...
	NET4: Unix domain ...
	request_module[block-major-48]: Root fs not mounted
	VFS: Cannot open root device "300b" or 30:0b
	please append a correct "root=" boot option
	Kernel Panic: VFS: Unable to mount root fs on 30:0b


When it works, DAC960 is initialized and "talks" just after FDC0,
You have seen that with 2.4.16 it does not.
the sentence "please append a correct root= boot option" makes no sens
to me since in lilo.conf both kernels are started with same options:
	image=/boot/vmlinuz-2.4.16-dac960
		label=linuxdac
		read-only
		root=/dev/rd/c0d1p3
	image=/boot/vmlinuz-2.4.7-10custom
		label=linux247
		read-only
		root=/dev/rd/c0d1p3

I can tell you DAC960 REALLY is compiled inside the kernel since
a "grep DAC960" System.map-2.4.16-dac960 shows lots of entries.

 
Now I am lost ... If anyone has an idea of the problem ...


Pascal Lengard

Re: dac960 broken ?

[Leonard N. Zubkoff]

  Date: Thu, 29 Nov 2001 19:42:01 +0100 (CET)
  From: Pascal Lengard <pascal.lengard@wanadoo.fr>

  On Wed, 28 Nov 2001, Leonard N. Zubkoff wrote:

  > Hmmm.  Nothing you've described makes any sense to me as I don't believe
  > the driver has changed in a way that would break the basic detection of the
  > boards.  When you say that the card is not detected, precisely what do you
  > mean?  Does the driver report anything at all?

  OK, I did not write down the messages, but while testing, with DAC960
  compiled as module, It did load and initialise but said "no such peripheral"
  or something equivalent once (I know it's not precise enough ...).

  I ran some tests to record precise messages today:

  Here are the messages from the kernel 2.4.9-13 shipped by redhat:
  -----------------------------------------------------------------
	  Loading scsi_mod module
	  Loading DAC960 module
	  /lib/DAC960.o: init_module: Operation not permitted
	  Hint: insmod errors can be caused by incorrect module parameters, including invalid IO or IRQ parameters
	  ERROR /bin/insmod exited abnormally!
	  Loading jbd module
	  Journalled Block Device driver loaded
	  Loading ext3 module
	  Mounting /proc filesystem
	  Creaing root device
	  Mounting root filesystem
	  mount: error 19 mounting ext3
	  pivotroot: pivot_root(/sysroot,/sysroot/initrd) failed: 2
	  Freeing unused kernel memory: 216k freed
	  Kernel panic: No init found. Try passing init= option to kernel


Hmmm.  Can you check if the following patch is present in the Linux kernel
you're using:

--- linux/init/main.c-	Sat Oct  6 08:49:16 2001
+++ linux/init/main.c	Wed Oct 10 09:06:07 2001
@@ -221,6 +221,24 @@
 	{ "dasdg", (DASD_MAJOR << MINORBITS) + (6 << 2) },
 	{ "dasdh", (DASD_MAJOR << MINORBITS) + (7 << 2) },
 #endif
+#if defined(CONFIG_BLK_DEV_DAC960) || defined(CONFIG_BLK_DEV_DAC960_MODULE)
+	{ "rd/c0d0p",0x3000 },
+	{ "rd/c0d1p",0x3008 },
+	{ "rd/c0d2p",0x3010 },
+	{ "rd/c0d3p",0x3018 },
+	{ "rd/c0d4p",0x3020 },
+	{ "rd/c0d5p",0x3028 },
+	{ "rd/c0d6p",0x3030 },
+	{ "rd/c0d7p",0x3038 },
+	{ "rd/c0d8p",0x3040 },
+	{ "rd/c0d9p",0x3048 },
+	{ "rd/c0d10p",0x3050 },
+	{ "rd/c0d11p",0x3058 },
+	{ "rd/c0d12p",0x3060 },
+	{ "rd/c0d13p",0x3068 },
+	{ "rd/c0d14p",0x3070 },
+	{ "rd/c0d15p",0x3078 },
+#endif
 #if defined(CONFIG_BLK_CPQ_DA) || defined(CONFIG_BLK_CPQ_DA_MODULE)
 	{ "ida/c0d0p",0x4800 },
 	{ "ida/c0d1p",0x4810 },

Without this patch, which Linux has repeatedly refused to include in the
standard sources, it is entirely possible that the root= processing won't work
correctly.

		Leonard

Re: dac960 broken ?

[Pascal Lengard]

working 2.4.7-10 from redhat includes definitions for devices rd/c0...
(not in #define(CONFIG_BLK_DEV_DAC960), patch is different)

broken 2.4.9-13 from redhat includes these also in the same way.
(so ...)

broken 2.4.16 does not define these devices in linux/init/main.c
I just tried your patch on 2.4.16. It went smoothly but the 2.4.16+patch
behaves exactly as before, same messages as without patch when booting ...
and kernel panic.

Pascal Lengard


On Thu, 29 Nov 2001, Leonard N. Zubkoff wrote:
> Hmmm.  Can you check if the following patch is present in the Linux kernel
> you're using:
> 
> --- linux/init/main.c-	Sat Oct  6 08:49:16 2001
> +++ linux/init/main.c	Wed Oct 10 09:06:07 2001
> @@ -221,6 +221,24 @@
>  	{ "dasdg", (DASD_MAJOR << MINORBITS) + (6 << 2) },
>  	{ "dasdh", (DASD_MAJOR << MINORBITS) + (7 << 2) },
>  #endif
> +#if defined(CONFIG_BLK_DEV_DAC960) || defined(CONFIG_BLK_DEV_DAC960_MODULE)
> +	{ "rd/c0d0p",0x3000 },
> +	{ "rd/c0d1p",0x3008 },
> +	{ "rd/c0d2p",0x3010 },
> +	{ "rd/c0d3p",0x3018 },
> +	{ "rd/c0d4p",0x3020 },
> +	{ "rd/c0d5p",0x3028 },
> +	{ "rd/c0d6p",0x3030 },
> +	{ "rd/c0d7p",0x3038 },
> +	{ "rd/c0d8p",0x3040 },
> +	{ "rd/c0d9p",0x3048 },
> +	{ "rd/c0d10p",0x3050 },
> +	{ "rd/c0d11p",0x3058 },
> +	{ "rd/c0d12p",0x3060 },
> +	{ "rd/c0d13p",0x3068 },
> +	{ "rd/c0d14p",0x3070 },
> +	{ "rd/c0d15p",0x3078 },
> +#endif
>  #if defined(CONFIG_BLK_CPQ_DA) || defined(CONFIG_BLK_CPQ_DA_MODULE)
>  	{ "ida/c0d0p",0x4800 },
>  	{ "ida/c0d1p",0x4810 },
> 
> Without this patch, which Linux has repeatedly refused to include in the
> standard sources, it is entirely possible that the root= processing won't work
> correctly.
> 
> 		Leonard
>