linux-pci.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
To: Vidya Sagar <vidyas@nvidia.com>
Cc: bhelgaas@google.com, treding@nvidia.com, swarren@nvidia.com,
	mperttunen@nvidia.com, jonathanh@nvidia.com, dev@lynxeye.de,
	linux-tegra@vger.kernel.org, linux-pci@vger.kernel.org,
	kthota@nvidia.com, mmaddireddy@nvidia.com, robin.murphy@arm.com
Subject: Re: [PATCH V2] PCI: tegra: Use the DMA-API to get the MSI address
Date: Thu, 11 Apr 2019 10:40:24 +0100	[thread overview]
Message-ID: <20190411094024.GC6227@red-moon> (raw)
In-Reply-To: <1553004121-24606-1-git-send-email-vidyas@nvidia.com>

[+Robin]

I would like Robin to have a look before merging this patch so
that we agree that's the expected approach.

Lorenzo

On Tue, Mar 19, 2019 at 07:32:01PM +0530, Vidya Sagar wrote:
> Since the upstream MSI memory writes are generated by downstream devices,
> it is logically correct to have MSI target memory coming from the DMA pool
> reserved for PCIe than from the general memory pool reserved for CPU
> access. This avoids PCIe DMA addresses coinciding with MSI target address
> thereby raising unwanted MSI interrupts. This patch also enforces to limit
> the MSI target address to 32-bits to make it work for PCIe endponits that
> support only 32-bit MSI target address and those endpoints that support
> 64-bit MSI target address anyway work with 32-bit MSI target address.
> 
> Signed-off-by: Vidya Sagar <vidyas@nvidia.com>
> Reviewed-by: Thierry Reding <treding@nvidia.com>
> Acked-by: Thierry Reding <treding@nvidia.com>
> ---
> v2:
> * changed 'phys' type to 'dma_addr_t' from 'u64'
> * added a comment on why DMA mask is set to 32-bit
> * replaced 'dma' with 'DMA'
> 
>  drivers/pci/controller/pci-tegra.c | 35 ++++++++++++++++++++++++++---------
>  1 file changed, 26 insertions(+), 9 deletions(-)
> 
> diff --git a/drivers/pci/controller/pci-tegra.c b/drivers/pci/controller/pci-tegra.c
> index f4f53d092e00..f8173a5e352d 100644
> --- a/drivers/pci/controller/pci-tegra.c
> +++ b/drivers/pci/controller/pci-tegra.c
> @@ -231,9 +231,9 @@ struct tegra_msi {
>  	struct msi_controller chip;
>  	DECLARE_BITMAP(used, INT_PCI_MSI_NR);
>  	struct irq_domain *domain;
> -	unsigned long pages;
>  	struct mutex lock;
> -	u64 phys;
> +	void *virt;
> +	dma_addr_t phys;
>  	int irq;
>  };
>  
> @@ -1536,7 +1536,7 @@ static int tegra_pcie_msi_setup(struct tegra_pcie *pcie)
>  	err = platform_get_irq_byname(pdev, "msi");
>  	if (err < 0) {
>  		dev_err(dev, "failed to get IRQ: %d\n", err);
> -		goto err;
> +		goto free_irq_domain;
>  	}
>  
>  	msi->irq = err;
> @@ -1545,17 +1545,34 @@ static int tegra_pcie_msi_setup(struct tegra_pcie *pcie)
>  			  tegra_msi_irq_chip.name, pcie);
>  	if (err < 0) {
>  		dev_err(dev, "failed to request IRQ: %d\n", err);
> -		goto err;
> +		goto free_irq_domain;
> +	}
> +
> +	/* Though the PCIe controller can address >32-bit address space, to
> +	 * facilitate endpoints that support only 32-bit MSI target address,
> +	 * the mask is set to 32-bit to make sure that MSI target address is
> +	 * always a 32-bit address
> +	 */
> +	err = dma_set_coherent_mask(dev, DMA_BIT_MASK(32));
> +	if (err < 0) {
> +		dev_err(dev, "failed to set DMA coherent mask: %d\n", err);
> +		goto free_irq;
> +	}
> +
> +	msi->virt = dma_alloc_coherent(dev, PAGE_SIZE, &msi->phys, GFP_KERNEL);
> +	if (!msi->virt) {
> +		dev_err(dev, "failed to allocate DMA memory for MSI\n");
> +		err = -ENOMEM;
> +		goto free_irq;
>  	}
>  
> -	/* setup AFI/FPCI range */
> -	msi->pages = __get_free_pages(GFP_KERNEL, 0);
> -	msi->phys = virt_to_phys((void *)msi->pages);
>  	host->msi = &msi->chip;
>  
>  	return 0;
>  
> -err:
> +free_irq:
> +	free_irq(msi->irq, pcie);
> +free_irq_domain:
>  	irq_domain_remove(msi->domain);
>  	return err;
>  }
> @@ -1592,7 +1609,7 @@ static void tegra_pcie_msi_teardown(struct tegra_pcie *pcie)
>  	struct tegra_msi *msi = &pcie->msi;
>  	unsigned int i, irq;
>  
> -	free_pages(msi->pages, 0);
> +	dma_free_coherent(pcie->dev, PAGE_SIZE, msi->virt, msi->phys);
>  
>  	if (msi->irq > 0)
>  		free_irq(msi->irq, pcie);
> -- 
> 2.7.4
> 

  parent reply	other threads:[~2019-04-11  9:40 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-03-19 14:02 [PATCH V2] PCI: tegra: Use the DMA-API to get the MSI address Vidya Sagar
2019-03-27 11:01 ` vidya sagar
     [not found] ` <CAN7O0+LG8vbB+hApe2tz2LLE_a9XF1iPfD7tKY6=MmhqgoJm8w@mail.gmail.com>
2019-04-01  5:43   ` Vidya Sagar
2019-04-04 21:30     ` Vidya Sagar
2019-04-05 12:46       ` Lorenzo Pieralisi
2019-04-11  9:40 ` Lorenzo Pieralisi [this message]
2019-04-12 15:00   ` Robin Murphy
2019-04-16 10:48     ` Vidya Sagar

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20190411094024.GC6227@red-moon \
    --to=lorenzo.pieralisi@arm.com \
    --cc=bhelgaas@google.com \
    --cc=dev@lynxeye.de \
    --cc=jonathanh@nvidia.com \
    --cc=kthota@nvidia.com \
    --cc=linux-pci@vger.kernel.org \
    --cc=linux-tegra@vger.kernel.org \
    --cc=mmaddireddy@nvidia.com \
    --cc=mperttunen@nvidia.com \
    --cc=robin.murphy@arm.com \
    --cc=swarren@nvidia.com \
    --cc=treding@nvidia.com \
    --cc=vidyas@nvidia.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).