max_hw_sectors error caused by recent NVMe driver commit

[Michael Kelley (LINUX)]

From: Christoph Hellwig <hch at lst.de> Sent: Sunday, February 12, 2023 10:00 PM
> 
> Hi Michael,
> 
> does this fix the problem for you?

Yes, basic smoke test works.  The underlying NVMe device
capabilities now show through instead of being replaced by
swiotlb values.  I did not do extensive functional testing.

See one code comment below.

Michael

> 
> commit 0467e96bafd2e84b67ba5c122bbbbac5e3f267e9
> Author: Christoph Hellwig <hch at lst.de>
> Date:   Mon Feb 13 06:58:33 2023 +0100
> 
>     nvme dma_mask
> 
> diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
> index c734934c407ccf..cb068f92bd597e 100644
> --- a/drivers/nvme/host/pci.c
> +++ b/drivers/nvme/host/pci.c
> @@ -2509,18 +2509,12 @@ static int nvme_pci_enable(struct nvme_dev *dev)
>  {
>  	int result = -ENOMEM;
>  	struct pci_dev *pdev = to_pci_dev(dev->dev);
> -	int dma_address_bits = 64;
> 
>  	if (pci_enable_device_mem(pdev))
>  		return result;
> 
>  	pci_set_master(pdev);
> 
> -	if (dev->ctrl.quirks & NVME_QUIRK_DMA_ADDRESS_BITS_48)
> -		dma_address_bits = 48;
> -	if (dma_set_mask_and_coherent(dev->dev, DMA_BIT_MASK(dma_address_bits)))
> -		goto disable;
> -
>  	if (readl(dev->bar + NVME_REG_CSTS) == -1) {
>  		result = -ENODEV;
>  		goto disable;
> @@ -2998,7 +2992,11 @@ static struct nvme_dev *nvme_pci_alloc_dev(struct
> pci_dev *pdev,
>  			     quirks);
>  	if (ret)
>  		goto out_put_device;
> -
> +
> +	if (dev->ctrl.quirks & NVME_QUIRK_DMA_ADDRESS_BITS_48)
> +		dma_set_mask_and_coherent(dev->dev, DMA_BIT_MASK(48));
> +	else
> +		dma_set_mask_and_coherent(dev->dev, DMA_BIT_MASK(64));

Any reason to use dev->dev instead of &pdev->dev?  They should be
the same, but with everything else in this function using &pdev->dev,
the lack of consistency suggests there might be a reason.

>  	dma_set_min_align_mask(&pdev->dev, NVME_CTRL_PAGE_SIZE - 1);
>  	dma_set_max_seg_size(&pdev->dev, 0xffffffff);
>