Fw: gdth SCSI driver(?) fails with more than 4GB of memory

Fw: gdth SCSI driver(?) fails with more than 4GB of memory

[Andrew Morton]

Begin forwarded message:

Date: Mon, 17 Jul 2006 13:00:56 +0200
From: Andreas Barth <aba@not.so.argh.org>
To: linux-kernel@vger.kernel.org
Subject: gdth SCSI driver(?) fails with more than 4GB of memory


[please Cc me, I'm currently not subscribed.]

Hi,

I have noticed that one of my boxes stopped to boot correctly after
adding more memory (in total 6 GB) and loading an adjusted kernel for
that.  After some testing around, we noticed that it is enough for the
kernel to boot correctly if we limit the kernel to use 4GB of memory.

If the kernel has 6GB, I directly get error messages like:
SCSI device sda: 143299800 512-byte hdwr sectors (73369 MB)
sda: Write Protect is off
sda: got wrong page
sda: assuming drive cache: write through
SCSI device sda: 143299800 512-byte hdwr sectors (73369 MB)
sda: Write Protect is off
sda: got wrong page

The disk array controller is of type GDT8114RZ and has the most current
firmware version. The box has 4 Xeon CPUs, and physical 6 GB of memory.
The /-device is on the controller in question.

The full log (for 4 and for 6 GB) is put up on
http://neualius.turmzimmer.net/~aba/6GB


Any hints for me how I can use the full 6 GB of memory (and/or what I
should try out to find the bug)?


Cheers,
Andi
-- 
  http://home.arcor.de/andreas-barth/
-
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/

Re: Fw: gdth SCSI driver(?) fails with more than 4GB of memory

[Rolf Eike Beer]

Andreas Barth wrote:

> [please Cc me, I'm currently not subscribed.]

> I have noticed that one of my boxes stopped to boot correctly after
> adding more memory (in total 6 GB) and loading an adjusted kernel for
> that.  After some testing around, we noticed that it is enough for the
> kernel to boot correctly if we limit the kernel to use 4GB of memory.
>
> If the kernel has 6GB, I directly get error messages like:
> SCSI device sda: 143299800 512-byte hdwr sectors (73369 MB)
> sda: Write Protect is off
> sda: got wrong page
> sda: assuming drive cache: write through
> SCSI device sda: 143299800 512-byte hdwr sectors (73369 MB)
> sda: Write Protect is off
> sda: got wrong page

This is from sd.c::sd_read_cache_type

> Any hints for me how I can use the full 6 GB of memory (and/or what I
> should try out to find the bug)?

Please change the printk in sd.c around 1482 that is

printk(KERN_ERR "%s: got wrong page\n", diskname);

to

printk(KERN_ERR "%s: got wrong page (expect: 0x%x, got 0x%x)\n",
diskname, modepage, buffer[offset]);

so we can see at least the reason that made it finally fail.

Eike

Re: Fw: gdth SCSI driver(?) fails with more than 4GB of memory

[Andreas Barth]

* Rolf Eike Beer (eike-kernel@sf-tec.de) [060718 08:46]:
> Andreas Barth wrote:
> 
> > [please Cc me, I'm currently not subscribed.]
> 
> > I have noticed that one of my boxes stopped to boot correctly after
> > adding more memory (in total 6 GB) and loading an adjusted kernel for
> > that.  After some testing around, we noticed that it is enough for the
> > kernel to boot correctly if we limit the kernel to use 4GB of memory.
> >
> > If the kernel has 6GB, I directly get error messages like:
> > SCSI device sda: 143299800 512-byte hdwr sectors (73369 MB)
> > sda: Write Protect is off
> > sda: got wrong page
> > sda: assuming drive cache: write through
> > SCSI device sda: 143299800 512-byte hdwr sectors (73369 MB)
> > sda: Write Protect is off
> > sda: got wrong page
> 
> This is from sd.c::sd_read_cache_type
> 
> > Any hints for me how I can use the full 6 GB of memory (and/or what I
> > should try out to find the bug)?
> 
> Please change the printk in sd.c around 1482 that is
> 
> printk(KERN_ERR "%s: got wrong page\n", diskname);
> 
> to
> 
> printk(KERN_ERR "%s: got wrong page (expect: 0x%x, got 0x%x)\n",
> diskname, modepage, buffer[offset]);

With this patch, it returns now:
SCSI device sda: 143299800 512-byte hdwr sectors (73369 MB)
sda: Write Protect is off
sda: got wrong page (expect: 0x8, got 0x0)
sda: assuming drive cache: write through
SCSI device sda: 143299800 512-byte hdwr sectors (73369 MB)
sda: Write Protect is off
sda: got wrong page (expect: 0x8, got 0x0)

(Full log with your and this patch:
--- a/drivers/scsi/gdth.c
+++ b/drivers/scsi/gdth.c
@@ -1543,7 +1543,7 @@ static int __init gdth_init_pci(gdth_pci
         if (prot_ver < 0x2b)      /* FW < x.43: no 64-bit DMA support */
             ha->dma64_support = 0;
         else
-            ha->dma64_support = 1;
+            ha->dma64_support = 0;
     }

     return 1;
put at http://neualius.turmzimmer.net/~aba/6G/kernel-20060807.log .
If you rather want your patch alone, please tell me.)



Cheers,
Andi
-- 
  http://home.arcor.de/andreas-barth/

Re: Fw: gdth SCSI driver(?) fails with more than 4GB of memory

[Rolf Eike Beer]

[-- Attachment #1: Type: text/plain, Size: 482 bytes --]

> With this patch, it returns now:
> SCSI device sda: 143299800 512-byte hdwr sectors (73369 MB)
> sda: Write Protect is off
> sda: got wrong page (expect: 0x8, got 0x0)
> sda: assuming drive cache: write through
> SCSI device sda: 143299800 512-byte hdwr sectors (73369 MB)
> sda: Write Protect is off
> sda: got wrong page (expect: 0x8, got 0x0)

> If you rather want your patch alone, please tell me.)

No, it's ok this way. This way we just get a little more information.

Eike

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: Fw: gdth SCSI driver(?) fails with more than 4GB of memory

[James Bottomley]

On Mon, 2006-07-17 at 18:46 -0700, Andrew Morton wrote:
> If the kernel has 6GB, I directly get error messages like:
> SCSI device sda: 143299800 512-byte hdwr sectors (73369 MB)
> sda: Write Protect is off
> sda: got wrong page
> sda: assuming drive cache: write through
> SCSI device sda: 143299800 512-byte hdwr sectors (73369 MB)
> sda: Write Protect is off
> sda: got wrong page
> 
> The disk array controller is of type GDT8114RZ and has the most current
> firmware version. The box has 4 Xeon CPUs, and physical 6 GB of memory.
> The /-device is on the controller in question.
> 
> The full log (for 4 and for 6 GB) is put up on
> http://neualius.turmzimmer.net/~aba/6GB

My strongest suspicion here is that the device is lying when it claims
to support 64 bit DMA. Can you try this?  It will turn off 64 bit DMA
and only allow 32 bit DMA on your 6GB platform.  If that works, we'll
beed help from the gdth people to understand the issue.

Thanks,

James

diff --git a/drivers/scsi/gdth.c b/drivers/scsi/gdth.c
index 43afd47..83cf3ca 100644
--- a/drivers/scsi/gdth.c
+++ b/drivers/scsi/gdth.c
@@ -1543,7 +1543,7 @@ static int __init gdth_init_pci(gdth_pci
         if (prot_ver < 0x2b)      /* FW < x.43: no 64-bit DMA support */
             ha->dma64_support = 0;
         else 
-            ha->dma64_support = 1;
+            ha->dma64_support = 0;
     }
 
     return 1;

Re: Fw: gdth SCSI driver(?) fails with more than 4GB of memory

[Andreas Barth]

* James Bottomley (James.Bottomley@SteelEye.com) [060718 15:37]:
> My strongest suspicion here is that the device is lying when it claims
> to support 64 bit DMA. Can you try this?  It will turn off 64 bit DMA
> and only allow 32 bit DMA on your 6GB platform.  If that works, we'll
> beed help from the gdth people to understand the issue.

> diff --git a/drivers/scsi/gdth.c b/drivers/scsi/gdth.c
> index 43afd47..83cf3ca 100644
> --- a/drivers/scsi/gdth.c
> +++ b/drivers/scsi/gdth.c
> @@ -1543,7 +1543,7 @@ static int __init gdth_init_pci(gdth_pci
>          if (prot_ver < 0x2b)      /* FW < x.43: no 64-bit DMA support */
>              ha->dma64_support = 0;
>          else 
> -            ha->dma64_support = 1;
> +            ha->dma64_support = 0;
>      }
>  
>      return 1;


Sorry, this patch didn't help me.


Cheers,
Andi
-- 
  http://home.arcor.de/andreas-barth/