[PATCH 1/2] PCI: Provide sensible irq vector alloc/free routines

Bjorn Helgaas helgaas at kernel.org
Fri May 6 09:04:08 PDT 2016


On Thu, May 05, 2016 at 04:04:55PM +0200, Christoph Hellwig wrote:
> Add a new pci_alloc_irq_vectors helper that allocates MSI-X or multi-MSI
> vectors for PCI device while isolating the driver from the arcane details.
> 
> This include handling both MSI-X, MSI and legacy interrupt fallbacks
> transparently, automatic capping to the available vectors as well as storing
> the information needed for request_irq in the PCI device itself so that
> a lot of boiler plate code in the driver can be removed.
> 
> In the future this will also allow us to automatically set up spreading
> for interrupt vectors without having to duplicate it in all the drivers.

I like this a lot.  One relatively minor comment below.

> Signed-off-by: Christoph Hellwig <hch at lst.de>
> ---
>  drivers/pci/msi.c   | 108 ++++++++++++++++++++++++++++++++++++++++++++++++++++
>  include/linux/pci.h |  19 +++++++++
>  2 files changed, 127 insertions(+)
> 
> diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c
> index a080f44..a510484 100644
> --- a/drivers/pci/msi.c
> +++ b/drivers/pci/msi.c
> @@ -4,6 +4,7 @@
>   *
>   * Copyright (C) 2003-2004 Intel
>   * Copyright (C) Tom Long Nguyen (tom.l.nguyen at intel.com)
> + * Copyright (c) 2016 Christoph Hellwig.
>   */
>  
>  #include <linux/err.h>
> @@ -1120,6 +1121,113 @@ int pci_enable_msix_range(struct pci_dev *dev, struct msix_entry *entries,
>  }
>  EXPORT_SYMBOL(pci_enable_msix_range);
>  
> +static int __pci_enable_msix(struct pci_dev *dev, int nr_vecs)
> +{
> +	struct msix_entry *msix_entries;
> +	int ret, i;
> +
> +	msix_entries = kcalloc(nr_vecs, sizeof(struct msix_entry), GFP_KERNEL);
> +	if (!msix_entries)
> +		return -ENOMEM;
> +
> +	for (i = 0; i < nr_vecs; i++)
> +		msix_entries[i].entry = i;
> +
> +	ret = msix_capability_init(dev, msix_entries, nr_vecs);
> +	if (ret == 0) {
> +		for (i = 0; i < nr_vecs; i++)
> +			dev->irqs[i] = msix_entries[i].vector;
> +	}
> +
> +	kfree(msix_entries);
> +	return ret;
> +}
> +
> +static int __pci_enable_msi(struct pci_dev *dev, int nr_vecs)
> +{
> +	int ret, i;
> +
> +	ret = msi_capability_init(dev, nr_vecs);
> +	if (ret == 0) {
> +		for (i = 0; i < nr_vecs; i++)
> +			dev->irqs[i] = dev->irq + i;
> +	}
> +
> +	return ret;
> +}
> +
> +/**
> + * pci_alloc_irq_vectors - allocate multiple IRQs for a device
> + * @dev:		PCI device to operate on
> + * @nr_vecs:		number of vectors to operate on
> + * @flags:		flags or quirks for the allocation
> + *
> + * Allocate @nr_vecs interrupt vectors for @dev, using MSI-X or MSI
> + * vectors if available, and fall back to a single legacy vector
> + * if neither is available.  Return the number of vectors allocated
> + * (which might be smaller than @nr_vecs) if successful, or a negative
> + * error code on error.  The Linux irq numbers for the allocated
> + * vectors are stored in pdev->irqs.

I think the "flags" argument penalizes working devices unnecessarily.
Everybody that implements MSI-X correctly has to pass that zero
argument:

  pci_alloc_irq_vectors(pdev, nr_io_queues, 0);

instead of putting the burden on the broken folks like this:

  pdev->msix_broken = 1;
  pci_alloc_irq_vectors(pdev, nr_io_queues);

If we remember this via a bit in struct pci_dev, we can also make sure
pci_enable_msix(), pci_enable_msix_range(), etc. fail as they should.

> + */
> +int pci_alloc_irq_vectors(struct pci_dev *dev, unsigned int nr_vecs,
> +		unsigned int flags)
> +{
> +	unsigned int ret;
> +
> +	if (WARN_ON_ONCE(dev->msi_enabled || dev->msix_enabled))
> +		return -EINVAL;
> +
> +	if (!pci_msi_supported(dev, 1))
> +		goto use_legacy_irq;
> +
> +	if (dev->msix_cap && !(flags & PCI_IRQ_NOMSIX))
> +		nr_vecs = min_t(unsigned int, nr_vecs, pci_msix_vec_count(dev));
> +	else if (dev->msi_cap)
> +		nr_vecs = min_t(unsigned int, nr_vecs, pci_msi_vec_count(dev));
> +	else
> +		goto use_legacy_irq;
> +
> +	dev->irqs = kcalloc(nr_vecs, sizeof(u32), GFP_KERNEL);
> +	if (!dev->irqs)
> +		return -ENOMEM;
> +
> +	if (dev->msix_cap && !(flags & PCI_IRQ_NOMSIX))
> +		ret = __pci_enable_msix(dev, nr_vecs);
> +	else
> +		ret = __pci_enable_msi(dev, nr_vecs);
> +	if (ret)
> +		goto out_free_irqs;
> +
> +	return 0;
> +
> +out_free_irqs:
> +	kfree(dev->irqs);
> +use_legacy_irq:
> +	dev->irqs = &dev->irq;
> +	return 1;
> +}
> +EXPORT_SYMBOL(pci_alloc_irq_vectors);
> +
> +/**
> + * pci_free_irq_vectors - free previously allocated IRQs for a device
> + * @dev:		PCI device to operate on
> + *
> + * Undoes the allocations and enabling in pci_alloc_irq_vectors().
> + */
> +void pci_free_irq_vectors(struct pci_dev *dev)
> +{
> +	if (dev->msix_enabled)
> +		pci_disable_msix(dev);
> +	else if (dev->msi_enabled)
> +		pci_disable_msi(dev);
> +
> +	if (dev->irqs != &dev->irq)
> +		kfree(dev->irqs);
> +	dev->irqs = NULL;
> +}
> +EXPORT_SYMBOL(pci_free_irq_vectors);
> +
> +
>  struct pci_dev *msi_desc_to_pci_dev(struct msi_desc *desc)
>  {
>  	return to_pci_dev(desc->dev);
> diff --git a/include/linux/pci.h b/include/linux/pci.h
> index 932ec74..e201d0d 100644
> --- a/include/linux/pci.h
> +++ b/include/linux/pci.h
> @@ -322,6 +322,7 @@ struct pci_dev {
>  	 * directly, use the values stored here. They might be different!
>  	 */
>  	unsigned int	irq;
> +	unsigned int	*irqs;
>  	struct resource resource[DEVICE_COUNT_RESOURCE]; /* I/O and memory regions + expansion ROMs */
>  
>  	bool match_driver;		/* Skip attaching driver */
> @@ -1255,6 +1256,8 @@ struct msix_entry {
>  	u16	entry;	/* driver uses to specify entry, OS writes */
>  };
>  
> +#define PCI_IRQ_NOMSIX		(1 << 0) /* don't try to use MSI-X interrupts */
> +
>  #ifdef CONFIG_PCI_MSI
>  int pci_msi_vec_count(struct pci_dev *dev);
>  void pci_msi_shutdown(struct pci_dev *dev);
> @@ -1283,6 +1286,10 @@ static inline int pci_enable_msix_exact(struct pci_dev *dev,
>  		return rc;
>  	return 0;
>  }
> +
> +int pci_alloc_irq_vectors(struct pci_dev *dev, unsigned int nr_vecs,
> +		unsigned int flags);
> +void pci_free_irq_vectors(struct pci_dev *dev);
>  #else
>  static inline int pci_msi_vec_count(struct pci_dev *dev) { return -ENOSYS; }
>  static inline void pci_msi_shutdown(struct pci_dev *dev) { }
> @@ -1306,6 +1313,18 @@ static inline int pci_enable_msix_range(struct pci_dev *dev,
>  static inline int pci_enable_msix_exact(struct pci_dev *dev,
>  		      struct msix_entry *entries, int nvec)
>  { return -ENOSYS; }
> +
> +static inline int pci_alloc_irq_vectors(struct pci_dev *dev,
> +		unsigned int nr_vecs, unsigned int flags)
> +{
> +	dev->irqs = &dev->irq;
> +	return 1;
> +}
> +
> +static inline void pci_free_irq_vectors(struct pci_dev *dev)
> +{
> +	dev->irqs = NULL;
> +}
>  #endif
>  
>  #ifdef CONFIG_PCIEPORTBUS
> -- 
> 2.1.4
> 



More information about the Linux-nvme mailing list