[PATCH] Fix embedded nic order on Dell PowerEdge systems

[PATCH] Fix embedded nic order on Dell PowerEdge systems

[Phil Oester]

    In commit 6e8af08dfa40b747002207d3ce8e8b43a050d99f ("PCI: enable pci=bfsort
    by default on future Dell systems"), the callback added was placed at the
    top of the list, before the specific PowerEdge models.  This ends up
    breaking the more specific models listed, since the new callback handles
    all Dell systems but the older systems don't have the new 0xB1 smbios
    entry.
    
    Fix this by placing the new check _after_ the more specific PowerEdge
    sytems.
    
    Phil Oester
    
    Signed-off-by: Phil Oester <kernel@linuxace.com>

diff --git a/arch/x86/pci/common.c b/arch/x86/pci/common.c
index 5fe7502..92df322 100644
--- a/arch/x86/pci/common.c
+++ b/arch/x86/pci/common.c
@@ -247,13 +247,6 @@ static const struct dmi_system_id __devinitconst pciprobe_dmi_table[] = {
 	},
 #endif		/* __i386__ */
 	{
-		.callback = find_sort_method,
-		.ident = "Dell System",
-		.matches = {
-			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc"),
-		},
-	},
-	{
 		.callback = set_bf_sort,
 		.ident = "Dell PowerEdge 1950",
 		.matches = {
@@ -294,6 +287,13 @@ static const struct dmi_system_id __devinitconst pciprobe_dmi_table[] = {
 		},
 	},
 	{
+		.callback = find_sort_method,
+		.ident = "Dell System",
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc"),
+		},
+	},
+	{
 		.callback = set_bf_sort,
 		.ident = "HP ProLiant BL20p G3",
 		.matches = {

RE: [PATCH] Fix embedded nic order on Dell PowerEdge systems

[Narendra_K]

> -----Original Message-----
> From: Phil Oester [mailto:kernel@linuxace.com]
> Sent: Thursday, June 09, 2011 1:44 AM
> To: linux-kernel@vger.kernel.org
> Cc: K, Narendra; Hargrave, Jordan; jbarnes@virtuousgeek.org
> Subject: [PATCH] Fix embedded nic order on Dell PowerEdge systems
> 
>     In commit 6e8af08dfa40b747002207d3ce8e8b43a050d99f ("PCI: enable
> pci=bfsort
>     by default on future Dell systems"), the callback added was placed at the
>     top of the list, before the specific PowerEdge models.  This ends up
>     breaking the more specific models listed, since the new callback handles
>     all Dell systems but the older systems don't have the new 0xB1 smbios
>     entry.
> 
>     Fix this by placing the new check _after_ the more specific PowerEdge
>     sytems.
> 
>     Phil Oester
> 
>     Signed-off-by: Phil Oester <kernel@linuxace.com>
> 
> diff --git a/arch/x86/pci/common.c b/arch/x86/pci/common.c index
> 5fe7502..92df322 100644
> --- a/arch/x86/pci/common.c
> +++ b/arch/x86/pci/common.c
> @@ -247,13 +247,6 @@ static const struct dmi_system_id __devinitconst
> pciprobe_dmi_table[] = {
>  	},
>  #endif		/* __i386__ */
>  	{
> -		.callback = find_sort_method,
> -		.ident = "Dell System",
> -		.matches = {
> -			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc"),
> -		},
> -	},
> -	{
>  		.callback = set_bf_sort,
>  		.ident = "Dell PowerEdge 1950",
>  		.matches = {
> @@ -294,6 +287,13 @@ static const struct dmi_system_id __devinitconst
> pciprobe_dmi_table[] = {
>  		},
>  	},
>  	{
> +		.callback = find_sort_method,
> +		.ident = "Dell System",
> +		.matches = {
> +			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc"),
> +		},
> +	},
> +	{
>  		.callback = set_bf_sort,
>  		.ident = "HP ProLiant BL20p G3",
>  		.matches = {
[>] 

Hi, this issue is fixed and the fix is in Jesse's linux-next tree.

http://marc.info/?l=linux-pci&m=130108034729082 (x86/PCI: Preserve existing pci=bfsort whitelist for Dell systems)

With regards,
Narendra K

Re: [PATCH] Fix embedded nic order on Dell PowerEdge systems

[Phil Oester]

On Thu, Jun 09, 2011 at 01:03:00AM -0700, Narendra_K@Dell.com wrote:
> Hi, this issue is fixed and the fix is in Jesse's linux-next tree.

It would be great to get that in -stable.  

Phil

Re: [PATCH] Fix embedded nic order on Dell PowerEdge systems

[Phil Oester]

On Thu, Jun 09, 2011 at 01:03:00AM -0700, Narendra_K@Dell.com wrote:
> > -----Original Message-----
> > From: Phil Oester [mailto:kernel@linuxace.com]
> > Sent: Thursday, June 09, 2011 1:44 AM
> > To: linux-kernel@vger.kernel.org
> > Cc: K, Narendra; Hargrave, Jordan; jbarnes@virtuousgeek.org
> > Subject: [PATCH] Fix embedded nic order on Dell PowerEdge systems
> > 
> >     In commit 6e8af08dfa40b747002207d3ce8e8b43a050d99f ("PCI: enable
> > pci=bfsort
> >     by default on future Dell systems"), the callback added was placed at the
> >     top of the list, before the specific PowerEdge models.  This ends up
> >     breaking the more specific models listed, since the new callback handles
> >     all Dell systems but the older systems don't have the new 0xB1 smbios
> >     entry.
> > 
> >     Fix this by placing the new check _after_ the more specific PowerEdge
> >     sytems.
> > 
> >     Phil Oester
> > 
> >     Signed-off-by: Phil Oester <kernel@linuxace.com>
> > 
> > diff --git a/arch/x86/pci/common.c b/arch/x86/pci/common.c index
> > 5fe7502..92df322 100644
> > --- a/arch/x86/pci/common.c
> > +++ b/arch/x86/pci/common.c
> > @@ -247,13 +247,6 @@ static const struct dmi_system_id __devinitconst
> > pciprobe_dmi_table[] = {
> >  	},
> >  #endif		/* __i386__ */
> >  	{
> > -		.callback = find_sort_method,
> > -		.ident = "Dell System",
> > -		.matches = {
> > -			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc"),
> > -		},
> > -	},
> > -	{
> >  		.callback = set_bf_sort,
> >  		.ident = "Dell PowerEdge 1950",
> >  		.matches = {
> > @@ -294,6 +287,13 @@ static const struct dmi_system_id __devinitconst
> > pciprobe_dmi_table[] = {
> >  		},
> >  	},
> >  	{
> > +		.callback = find_sort_method,
> > +		.ident = "Dell System",
> > +		.matches = {
> > +			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc"),
> > +		},
> > +	},
> > +	{
> >  		.callback = set_bf_sort,
> >  		.ident = "HP ProLiant BL20p G3",
> >  		.matches = {
> [>] 
> 
> Hi, this issue is fixed and the fix is in Jesse's linux-next tree.
> 
> http://marc.info/?l=linux-pci&m=130108034729082 (x86/PCI: Preserve existing pci=bfsort whitelist for Dell systems)
> 
> With regards,
> Narendra K

This patch is still not included in 3.0.0-rc6, nor in any stable series.
Some reason for this?  And did Dell do any testing at all on existing
hardware before submitting 6e8af08dfa4?  Seems even a minor amount of
testing should have shown this regression.

Given 3.0.0 is expected soon, perhaps this can be included?  Jesse?

Phil

RE: [PATCH] Fix embedded nic order on Dell PowerEdge systems

[Narendra_K]

> -----Original Message-----
> From: Phil Oester [mailto:kernel@linuxace.com]
> Sent: Sunday, July 10, 2011 9:03 AM
> To: K, Narendra
> Cc: linux-kernel@vger.kernel.org; Hargrave, Jordan;
> jbarnes@virtuousgeek.org
> Subject: Re: [PATCH] Fix embedded nic order on Dell PowerEdge systems
> 
> On Thu, Jun 09, 2011 at 01:03:00AM -0700, Narendra_K@Dell.com wrote:
> > > -----Original Message-----
> > > From: Phil Oester [mailto:kernel@linuxace.com]
> > > Sent: Thursday, June 09, 2011 1:44 AM
> > > To: linux-kernel@vger.kernel.org
> > > Cc: K, Narendra; Hargrave, Jordan; jbarnes@virtuousgeek.org
> > > Subject: [PATCH] Fix embedded nic order on Dell PowerEdge systems
> > >
> > >     In commit 6e8af08dfa40b747002207d3ce8e8b43a050d99f ("PCI: enable
> > > pci=bfsort
> > >     by default on future Dell systems"), the callback added was placed at
> the
> > >     top of the list, before the specific PowerEdge models.  This ends up
> > >     breaking the more specific models listed, since the new callback
> handles
> > >     all Dell systems but the older systems don't have the new 0xB1 smbios
> > >     entry.
> > >
> > >     Fix this by placing the new check _after_ the more specific PowerEdge
> > >     sytems.
> > >
> > >     Phil Oester
> > >
> > >     Signed-off-by: Phil Oester <kernel@linuxace.com>
> > >
> > > diff --git a/arch/x86/pci/common.c b/arch/x86/pci/common.c index
> > > 5fe7502..92df322 100644
> > > --- a/arch/x86/pci/common.c
> > > +++ b/arch/x86/pci/common.c
> > > @@ -247,13 +247,6 @@ static const struct dmi_system_id
> > > __devinitconst pciprobe_dmi_table[] = {
> > >  	},
> > >  #endif		/* __i386__ */
> > >  	{
> > > -		.callback = find_sort_method,
> > > -		.ident = "Dell System",
> > > -		.matches = {
> > > -			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc"),
> > > -		},
> > > -	},
> > > -	{
> > >  		.callback = set_bf_sort,
> > >  		.ident = "Dell PowerEdge 1950",
> > >  		.matches = {
> > > @@ -294,6 +287,13 @@ static const struct dmi_system_id
> > > __devinitconst pciprobe_dmi_table[] = {
> > >  		},
> > >  	},
> > >  	{
> > > +		.callback = find_sort_method,
> > > +		.ident = "Dell System",
> > > +		.matches = {
> > > +			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc"),
> > > +		},
> > > +	},
> > > +	{
> > >  		.callback = set_bf_sort,
> > >  		.ident = "HP ProLiant BL20p G3",
> > >  		.matches = {
> > [>]
> >
> > Hi, this issue is fixed and the fix is in Jesse's linux-next tree.
> >
> > http://marc.info/?l=linux-pci&m=130108034729082 (x86/PCI: Preserve
> > existing pci=bfsort whitelist for Dell systems)
> >
> > With regards,
> > Narendra K
> 
> This patch is still not included in 3.0.0-rc6, nor in any stable series.
> Some reason for this?  And did Dell do any testing at all on existing hardware
> before submitting 6e8af08dfa4?  Seems even a minor amount of testing
> should have shown this regression.
> 
> Given 3.0.0 is expected soon, perhaps this can be included?  Jesse?
> 

Hi, I think for the patch to be included into stable, it needs to be in the mainline.
Hi Jesse, please consider queuing this patch  for mainline.

With regards,
Narendra K

Re: [PATCH] Fix embedded nic order on Dell PowerEdge systems

[Jesse Barnes]

On Mon, 11 Jul 2011 17:52:01 +0530
<Narendra_K@Dell.com> wrote:
> > This patch is still not included in 3.0.0-rc6, nor in any stable series.
> > Some reason for this?  And did Dell do any testing at all on existing hardware
> > before submitting 6e8af08dfa4?  Seems even a minor amount of testing
> > should have shown this regression.
> > 
> > Given 3.0.0 is expected soon, perhaps this can be included?  Jesse?
> > 
> 
> Hi, I think for the patch to be included into stable, it needs to be in the mainline.
> Hi Jesse, please consider queuing this patch  for mainline.

Ouch, I was sure I had this queued a long time ago but somehow it got
dropped.  Anyway it's queued now and will appear in 3.1.

Thanks,
-- 
Jesse Barnes, Intel Open Source Technology Center