Re: + mpt-fusion-dont-oops-if-numphys==0.patch added to -mm tree

Re: + mpt-fusion-dont-oops-if-numphys==0.patch added to -mm tree

[James Bottomley]

On Tue, 2008-02-19 at 15:34 -0800, akpm@linux-foundation.org wrote:
> Subject: mpt fusion: don't oops if NumPhys==0
> From: Krzysztof Oledzki <olel@ans.pl>
> 
> Don't oops if NumPhys==0, instead return -ENODEV.
> This patch fixes http://bugzilla.kernel.org/show_bug.cgi?id=9909
> 
> Signed-off-by: Krzysztof Piotr Oledzki <ole@ans.pl>
> Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
> Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
> Cc: Eric Moore <Eric.Moore@lsi.com>
> Cc: Sathya Prakash <sathya.prakash@lsi.com>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> ---
> 
>  drivers/message/fusion/mptsas.c |    5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff -puN drivers/message/fusion/mptsas.c~mpt-fusion-dont-oops-if-numphys==0 drivers/message/fusion/mptsas.c
> --- a/drivers/message/fusion/mptsas.c~mpt-fusion-dont-oops-if-numphys==0
> +++ a/drivers/message/fusion/mptsas.c
> @@ -1701,6 +1701,11 @@ mptsas_sas_expander_pg0(MPT_ADAPTER *ioc
>  	if (error)
>  		goto out_free_consistent;
>  
> +	if (!buffer->NumPhys) {
> +		error = -ENODEV;
> +		goto out_free_consistent;
> +	}
> +
>  	/* save config data */
>  	port_info->num_phys = buffer->NumPhys;
>  	port_info->phy_info = kcalloc(port_info->num_phys,
> _

This is a bit odd ... It looks like a work around for a fusion firmware
problem.  What's the topology this thing is connected to?

James

Re: + mpt-fusion-dont-oops-if-numphys==0.patch added to -mm tree

[Krzysztof Oledzki]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 1802 bytes --]



On Tue, 19 Feb 2008, James Bottomley wrote:

>
> On Tue, 2008-02-19 at 15:34 -0800, akpm@linux-foundation.org wrote:
>> Subject: mpt fusion: don't oops if NumPhys==0
>> From: Krzysztof Oledzki <olel@ans.pl>
>>
>> Don't oops if NumPhys==0, instead return -ENODEV.
>> This patch fixes http://bugzilla.kernel.org/show_bug.cgi?id=9909
>>
>> Signed-off-by: Krzysztof Piotr Oledzki <ole@ans.pl>
>> Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
>> Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
>> Cc: Eric Moore <Eric.Moore@lsi.com>
>> Cc: Sathya Prakash <sathya.prakash@lsi.com>
>> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
>> ---
>>
>>  drivers/message/fusion/mptsas.c |    5 +++++
>>  1 file changed, 5 insertions(+)
>>
>> diff -puN drivers/message/fusion/mptsas.c~mpt-fusion-dont-oops-if-numphys==0 drivers/message/fusion/mptsas.c
>> --- a/drivers/message/fusion/mptsas.c~mpt-fusion-dont-oops-if-numphys==0
>> +++ a/drivers/message/fusion/mptsas.c
>> @@ -1701,6 +1701,11 @@ mptsas_sas_expander_pg0(MPT_ADAPTER *ioc
>>  	if (error)
>>  		goto out_free_consistent;
>>
>> +	if (!buffer->NumPhys) {
>> +		error = -ENODEV;
>> +		goto out_free_consistent;
>> +	}
>> +
>>  	/* save config data */
>>  	port_info->num_phys = buffer->NumPhys;
>>  	port_info->phy_info = kcalloc(port_info->num_phys,
>> _
>
> This is a bit odd ... It looks like a work around for a fusion firmware
> problem.

Workaround or not, the driver should not oops by allocating 0 size memory 
block and dereferencing it. It used to work in <2.6.23 but not now.

> What's the topology this thing is connected to?

It is a Dell-1950III server with two 400GB SAS disks (RAID1) installed in 
a dwo disk SAS backplane.

Best regrds,

 			Krzysztof Olędzki

Re: + mpt-fusion-dont-oops-if-numphys==0.patch added to -mm tree

[Maximilian Wilhelm]

Am Wednesday, den 20 February hub Krzysztof Oledzki folgendes in die Tasten:

Hi Krzysztof, *

Thanks for CC'ing me.

> >On Tue, 2008-02-19 at 15:34 -0800, akpm@linux-foundation.org wrote:
> >>Subject: mpt fusion: don't oops if NumPhys==0
> >>From: Krzysztof Oledzki <olel@ans.pl>
> >>
> >>Don't oops if NumPhys==0, instead return -ENODEV.
> >>This patch fixes http://bugzilla.kernel.org/show_bug.cgi?id=9909
> >>
> >>Signed-off-by: Krzysztof Piotr Oledzki <ole@ans.pl>
> >>Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
> >>Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
> >>Cc: Eric Moore <Eric.Moore@lsi.com>
> >>Cc: Sathya Prakash <sathya.prakash@lsi.com>
> >>Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> >>---
> >>
> >> drivers/message/fusion/mptsas.c |    5 +++++
> >> 1 file changed, 5 insertions(+)
> >>
> >>diff -puN 
> >>drivers/message/fusion/mptsas.c~mpt-fusion-dont-oops-if-numphys==0 
> >>drivers/message/fusion/mptsas.c
> >>--- a/drivers/message/fusion/mptsas.c~mpt-fusion-dont-oops-if-numphys==0
> >>+++ a/drivers/message/fusion/mptsas.c
> >>@@ -1701,6 +1701,11 @@ mptsas_sas_expander_pg0(MPT_ADAPTER *ioc
> >> 	if (error)
> >> 		goto out_free_consistent;
> >>
> >>+	if (!buffer->NumPhys) {
> >>+		error = -ENODEV;
> >>+		goto out_free_consistent;
> >>+	}
> >>+
> >> 	/* save config data */
> >> 	port_info->num_phys = buffer->NumPhys;
> >> 	port_info->phy_info = kcalloc(port_info->num_phys,
> >>_
> >
> >This is a bit odd ... It looks like a work around for a fusion firmware
> >problem.

> Workaround or not, the driver should not oops by allocating 0 size memory 
> block and dereferencing it. It used to work in <2.6.23 but not now.

ACK.
Nevertheless my server is working absolutely well with this
fix/workaround. No problems at all.

> >What's the topology this thing is connected to?

> It is a Dell-1950III server with two 400GB SAS disks (RAID1) installed in 
> a dwo disk SAS backplane.

Here it's a Dell 1950 with two 73G SAS disks (RAID1) but having four
SAS disk slots.

If you need further information about the hardware, just let me know.

Ciao
Max
-- 
	Follow the white penguin.

Re: + mpt-fusion-dont-oops-if-numphys==0.patch added to -mm tree

[James Bottomley]

On Wed, 2008-02-20 at 04:17 +0100, Maximilian Wilhelm wrote:
> Am Wednesday, den 20 February hub Krzysztof Oledzki folgendes in die Tasten:
> > >What's the topology this thing is connected to?
> 
> > It is a Dell-1950III server with two 400GB SAS disks (RAID1) installed in 
> > a dwo disk SAS backplane.
> 
> Here it's a Dell 1950 with two 73G SAS disks (RAID1) but having four
> SAS disk slots.
> 
> If you need further information about the hardware, just let me know.

Is that directly connected or via an internal expander?

James

Re: + mpt-fusion-dont-oops-if-numphys==0.patch added to -mm tree

[Maximilian Wilhelm]

Am Tuesday, den 19 February hub James Bottomley folgendes in die Tasten:

> On Wed, 2008-02-20 at 04:17 +0100, Maximilian Wilhelm wrote:
> > Am Wednesday, den 20 February hub Krzysztof Oledzki folgendes in die Tasten:
> > > >What's the topology this thing is connected to?

> > > It is a Dell-1950III server with two 400GB SAS disks (RAID1) installed in 
> > > a dwo disk SAS backplane.

> > Here it's a Dell 1950 with two 73G SAS disks (RAID1) but having four
> > SAS disk slots.

> > If you need further information about the hardware, just let me know.

> Is that directly connected or via an internal expander?

The disks are connected to an expander.
Have a look at http://files.rfc2324.org/mptsas_panic/DSC_2133.jpg for
a picture.

Further questions?

Ciao
Max
-- 
	Follow the white penguin.