Re: [PATCH] x86/PCI: sta2x11: use default DMA address translation ops

[Nicolas Saenz Julienne]

Hi Bjorn,
thanks for having a look at it.

On Thu, 2019-10-17 at 17:41 -0500, Bjorn Helgaas wrote:
> Hi Nicolas,
> 
> I'm hoping Christoph will chime in and one of the x86 guys will merge
> this, since I'm not a DMA expert.  Trivial comments/questions below.
> 
> On Wed, Oct 16, 2019 at 06:51:37PM +0200, Nicolas Saenz Julienne wrote:
> > The devices found behind this PCIe chip have unusual DMA mapping
> > constraints as there is an AMBA interconnect placed in between them and
> > the different PCI endpoints. The offset between physical memory
> > addresses and AMBA's view is provided by reading a PCI config register,
> > which is saved and used whenever DMA mapping is needed.
> > 
> > It turns out that this DMA setup can be represented by properly setting
> > 'dma_pfn_offset', 'dma_bus_mask' and 'dma_mask' during the PCI device
> > enable fixup. And ultimately allows us to get rid of this device's
> > custom DMA functions.
> > 
> > Aside from the code deletion and DMA setup, sta2x11_pdev_to_mapping() is
> > moved to avoid warnings whenever CONFIG_PM is not enabled.
> > 
> > Signed-off-by: Nicolas Saenz Julienne <nsaenzjulienne@xxxxxxx>
> > -	pci_read_config_dword(pdev, AHB_BASE(0), &map->amba_base);
> > +
> > +	pci_read_config_dword(pdev, AHB_BASE(0), &amba_base);
> > +	dev->dma_pfn_offset = PFN_DOWN(-amba_base);
> > +	dev->bus_dma_mask = amba_base + STA2X11_AMBA_SIZE - 1;
> 
> I think of a mask as typically being one less than a power of two, but
> that's not the case here, e.g., STA2X11_AMBA_SIZE - 1 == 0x1fffffff
> (512MB-1), so if amba_size is 1G, the mask will be 0x5fffffff.
> 
> Just double-checking to be sure that's what you intend.

Yes I'm aware of it.

I know it's counter-intuitive name wise but dma-direct uses it more like a top
address than a mask (see for example 'dma_capable()' in dma-direct.h). There is
at least another driver using 'dev->bus_dma_mask' as such (see
arch/powerpc/sysdev/fsl_pci.c).

That said, I agree it something Christoph should ratify for this to get merged.

> > +	pci_set_consistent_dma_mask(pdev, amba_base + STA2X11_AMBA_SIZE - 1);
> > +	pci_set_dma_mask(pdev, amba_base + STA2X11_AMBA_SIZE - 1);
> 
> Maybe add a local variable instead of repeating the "amba_base + ..."
> expression three times?

Noted.

> >  	/* Configure AHB mapping */
> >  	pci_write_config_dword(pdev, AHB_PEXLBASE(0), 0);
> > @@ -252,14 +156,25 @@ static void sta2x11_map_ep(struct pci_dev *pdev)
> >  		pci_write_config_dword(pdev, AHB_CRW(i), 0);
> >  
> >  	dev_info(&pdev->dev,
> > -		 "sta2x11: Map EP %i: AMBA address %#8x-%#8x\n",
> > -		 sta2x11_pdev_to_ep(pdev),  map->amba_base,
> > -		 map->amba_base + STA2X11_AMBA_SIZE - 1);
> > +		 "sta2x11: Map EP %i: AMBA address %#8x-%#8llx\n",
> > +		 sta2x11_pdev_to_ep(pdev), amba_base, dev->bus_dma_mask);
> 
> This would read better as
> 
>   amba_base, amba_base + STA2X11_AMBA_SIZE - 1
> 
> I know that's the same dev->bus_dma_mask, but a "mask" is not the
> obvious name for the end of a range.

Ok.

I just got a notification from my mail server that it failed to deliver the
mail to Christoph, it seems I misspelled it.

I'll fix all this and send a v2.

Regards,
Nicolas

Attachment: signature.asc
Description: This is a digitally signed message part