[PATCH] Fusion: Call pci_set_mwi()

[PATCH] Fusion: Call pci_set_mwi()

[Matthew Wilcox]

Hi Eric,

Even though the 1030 is a PCI-X device, it can still be plugged into a PCI
bus, and end up operating in PCI mode.  According to the documentation,
it will benefit from using MWI and MRM commands, but in order to do so
under Linux, we need to call pci_set_mwi(), which is what this patch does.
Please apply.

P.S. The if condition is just to avoid the warning from __must_check.
Most device drivers do need to check and adjust their internal flags if
it fails, but according to the 1030 manual, the chip automatically
decides when to use MWI.

Signed-off-by: Matthew Wilcox <matthew@wil.cx>

diff --git a/drivers/message/fusion/mptbase.c b/drivers/message/fusion/mptbase.c
index 29d0635..d301a67 100644
--- a/drivers/message/fusion/mptbase.c
+++ b/drivers/message/fusion/mptbase.c
@@ -1153,6 +1153,9 @@ #endif
 	if (pci_enable_device(pdev))
 		return r;
 
+	if (pci_set_mwi(pdev))
+		/* We don't have to do anything if it fails */;
+
 	dinitprintk((KERN_WARNING MYNAM ": mpt_adapter_install\n"));
 
 	if (!pci_set_dma_mask(pdev, DMA_64BIT_MASK)) {

RE: [PATCH] Fusion: Call pci_set_mwi()

[Moore, Eric]

On Friday, October 13, 2006 10:25 AM,Matthew Wilcox wrote:   

> 
> Hi Eric,
> 
> Even though the 1030 is a PCI-X device, it can still be 
> plugged into a PCI
> bus, and end up operating in PCI mode.  According to the 
> documentation,
> it will benefit from using MWI and MRM commands, but in order to do so
> under Linux, we need to call pci_set_mwi(), which is what 
> this patch does.
> Please apply.
> 
> P.S. The if condition is just to avoid the warning from __must_check.
> Most device drivers do need to check and adjust their 
> internal flags if
> it fails, but according to the 1030 manual, the chip automatically
> decides when to use MWI.
> 
> Signed-off-by: Matthew Wilcox <matthew@wil.cx>
> 
> diff --git a/drivers/message/fusion/mptbase.c 
> b/drivers/message/fusion/mptbase.c
> index 29d0635..d301a67 100644
> --- a/drivers/message/fusion/mptbase.c
> +++ b/drivers/message/fusion/mptbase.c
> @@ -1153,6 +1153,9 @@ #endif
>  	if (pci_enable_device(pdev))
>  		return r;
>  
> +	if (pci_set_mwi(pdev))
> +		/* We don't have to do anything if it fails */;
> +
>  	dinitprintk((KERN_WARNING MYNAM ": mpt_adapter_install\n"));
>  
>  	if (!pci_set_dma_mask(pdev, DMA_64BIT_MASK)) {
> 
> 


I just responded to the HP thread, and I remember this patch.

I hadn't had a chance to check this up.  Pls let me know importance of
this,
and I will ask around the office wrt to this patch..  You don't think
this shouldn't break anything else? , such
as pci-express cards, cuase we have 1030 pcie, as well as FC and SAS
that
are express.  

Eric