[RFC] pci: adding new interrupt api to pcie-designware

Commit Message

Joao Pinto May 9, 2017, 12:33 p.m. UTC

This is a proposal for the update of the interrupt API in pcie-designware.

*SoC specific drivers*
All SoC specific drivers that use the common infrastructure (pci-qcom,
pci-artpec6, pci-exynos, pci-imx6) were updated to be compatible.
Other SoC specific drivers like pci-dra7, pci-armada8k, pci-hisi, pci-spear13xx
and pci-layerscape will also work fine.

*Work still to be done - need some inputs*
The pci-keystone driver is the one that I would appreciate some opinions, since
it uses the "old" dw_pcie_msi_chip. I think we have 3 options:

a) Keep the old API + new API in pcie-designware just for pci-keystone to use
the dw_pcie_msi_chip structure
b) Move the old API from pcie-designware to pci-keystone for it to use the
dw_pcie_msi_chip structure
c) Adapt pci-keystone to use the new API also

*Tests*
I made tests with a PCI 4.80 Core IP, using pcie-designware-plat driver.
I used an MSI-only Endpoint and a MSI/MSIX Endpoint and the APi adapted very
well to the situation.

Signed-off-by: Joao Pinto <[e-mail]>
---
 drivers/pci/dwc/pci-exynos.c           |  18 --
 drivers/pci/dwc/pci-imx6.c             |  18 --
 drivers/pci/dwc/pcie-artpec6.c         |  18 --
 drivers/pci/dwc/pcie-designware-host.c | 342 +++++++++++++++++----------------
 drivers/pci/dwc/pcie-designware-plat.c |  15 --
 drivers/pci/dwc/pcie-designware.h      |   6 +-
 drivers/pci/dwc/pcie-qcom.c            |  15 --
 7 files changed, 179 insertions(+), 253 deletions(-)

Comments

Marc Zyngier May 10, 2017, 5 p.m. UTC | #1

On 09/05/17 13:33, Joao Pinto wrote:
> This is a proposal for the update of the interrupt API in pcie-designware.
> 
> *SoC specific drivers*
> All SoC specific drivers that use the common infrastructure (pci-qcom,
> pci-artpec6, pci-exynos, pci-imx6) were updated to be compatible.
> Other SoC specific drivers like pci-dra7, pci-armada8k, pci-hisi, pci-spear13xx
> and pci-layerscape will also work fine.
> 
> *Work still to be done - need some inputs*
> The pci-keystone driver is the one that I would appreciate some opinions, since
> it uses the "old" dw_pcie_msi_chip. I think we have 3 options:
> 
> a) Keep the old API + new API in pcie-designware just for pci-keystone to use
> the dw_pcie_msi_chip structure
> b) Move the old API from pcie-designware to pci-keystone for it to use the
> dw_pcie_msi_chip structure
> c) Adapt pci-keystone to use the new API also

What is the issue with moving Keystone over to this infrastructure?

> 
> *Tests*
> I made tests with a PCI 4.80 Core IP, using pcie-designware-plat driver.
> I used an MSI-only Endpoint and a MSI/MSIX Endpoint and the APi adapted very
> well to the situation.
> 
> Signed-off-by: Joao Pinto <[e-mail]>
> ---
>  drivers/pci/dwc/pci-exynos.c           |  18 --
>  drivers/pci/dwc/pci-imx6.c             |  18 --
>  drivers/pci/dwc/pcie-artpec6.c         |  18 --
>  drivers/pci/dwc/pcie-designware-host.c | 342 +++++++++++++++++----------------
>  drivers/pci/dwc/pcie-designware-plat.c |  15 --
>  drivers/pci/dwc/pcie-designware.h      |   6 +-
>  drivers/pci/dwc/pcie-qcom.c            |  15 --
>  7 files changed, 179 insertions(+), 253 deletions(-)

May I suggest something for your next posting? This patch is extremely
difficult to read (not your fault at all), as it deletes a lot of code
and replaces it by code that is mostly unrelated. It would be a lot
better if you had a series that:

1: adds the new infrastructure in parallel with the old one, with an
opt-in mechanism.
2: convert each individual platform (one patch per SoC)
3: remove the old code entirely.

This will result in a much more readable series, and will give us a good
way to bisect, should a given platform break.

> 
> diff --git a/drivers/pci/dwc/pci-exynos.c b/drivers/pci/dwc/pci-exynos.c
> index 546082a..e49c39a 100644
> --- a/drivers/pci/dwc/pci-exynos.c
> +++ b/drivers/pci/dwc/pci-exynos.c
> @@ -490,15 +490,6 @@ static irqreturn_t exynos_pcie_irq_handler(int irq, void *arg)
>  	return IRQ_HANDLED;
>  }
>  
> -static irqreturn_t exynos_pcie_msi_irq_handler(int irq, void *arg)
> -{
> -	struct exynos_pcie *ep = arg;
> -	struct dw_pcie *pci = ep->pci;
> -	struct pcie_port *pp = &pci->pp;
> -
> -	return dw_handle_msi_irq(pp);
> -}
> -
>  static void exynos_pcie_msi_init(struct exynos_pcie *ep)
>  {
>  	struct dw_pcie *pci = ep->pci;
> @@ -622,15 +613,6 @@ static int __init exynos_add_pcie_port(struct exynos_pcie *ep,
>  			dev_err(dev, "failed to get msi irq\n");
>  			return -ENODEV;
>  		}
> -
> -		ret = devm_request_irq(dev, pp->msi_irq,
> -					exynos_pcie_msi_irq_handler,
> -					IRQF_SHARED | IRQF_NO_THREAD,
> -					"exynos-pcie", ep);
> -		if (ret) {
> -			dev_err(dev, "failed to request msi irq\n");
> -			return ret;
> -		}
>  	}
>  
>  	pp->root_bus_nr = -1;
> diff --git a/drivers/pci/dwc/pci-imx6.c b/drivers/pci/dwc/pci-imx6.c
> index a98cba5..f22ce1f 100644
> --- a/drivers/pci/dwc/pci-imx6.c
> +++ b/drivers/pci/dwc/pci-imx6.c
> @@ -490,15 +490,6 @@ static int imx6_pcie_wait_for_speed_change(struct imx6_pcie *imx6_pcie)
>  	return -EINVAL;
>  }
>  
> -static irqreturn_t imx6_pcie_msi_handler(int irq, void *arg)
> -{
> -	struct imx6_pcie *imx6_pcie = arg;
> -	struct dw_pcie *pci = imx6_pcie->pci;
> -	struct pcie_port *pp = &pci->pp;
> -
> -	return dw_handle_msi_irq(pp);
> -}
> -
>  static int imx6_pcie_establish_link(struct imx6_pcie *imx6_pcie)
>  {
>  	struct dw_pcie *pci = imx6_pcie->pci;
> @@ -620,15 +611,6 @@ static int imx6_add_pcie_port(struct imx6_pcie *imx6_pcie,
>  			dev_err(dev, "failed to get MSI irq\n");
>  			return -ENODEV;
>  		}
> -
> -		ret = devm_request_irq(dev, pp->msi_irq,
> -				       imx6_pcie_msi_handler,
> -				       IRQF_SHARED | IRQF_NO_THREAD,
> -				       "mx6-pcie-msi", imx6_pcie);
> -		if (ret) {
> -			dev_err(dev, "failed to request MSI irq\n");
> -			return ret;
> -		}
>  	}
>  
>  	pp->root_bus_nr = -1;
> diff --git a/drivers/pci/dwc/pcie-artpec6.c b/drivers/pci/dwc/pcie-artpec6.c
> index 82a04ac..d81789b 100644
> --- a/drivers/pci/dwc/pcie-artpec6.c
> +++ b/drivers/pci/dwc/pcie-artpec6.c
> @@ -188,15 +188,6 @@ static struct dw_pcie_host_ops artpec6_pcie_host_ops = {
>  	.host_init = artpec6_pcie_host_init,
>  };
>  
> -static irqreturn_t artpec6_pcie_msi_handler(int irq, void *arg)
> -{
> -	struct artpec6_pcie *artpec6_pcie = arg;
> -	struct dw_pcie *pci = artpec6_pcie->pci;
> -	struct pcie_port *pp = &pci->pp;
> -
> -	return dw_handle_msi_irq(pp);
> -}
> -
>  static int artpec6_add_pcie_port(struct artpec6_pcie *artpec6_pcie,
>  				 struct platform_device *pdev)
>  {
> @@ -211,15 +202,6 @@ static int artpec6_add_pcie_port(struct artpec6_pcie *artpec6_pcie,
>  			dev_err(dev, "failed to get MSI irq\n");
>  			return -ENODEV;
>  		}
> -
> -		ret = devm_request_irq(dev, pp->msi_irq,
> -				       artpec6_pcie_msi_handler,
> -				       IRQF_SHARED | IRQF_NO_THREAD,
> -				       "artpec6-pcie-msi", artpec6_pcie);
> -		if (ret) {
> -			dev_err(dev, "failed to request MSI irq\n");
> -			return ret;
> -		}
>  	}
>  
>  	pp->root_bus_nr = -1;
> diff --git a/drivers/pci/dwc/pcie-designware-host.c b/drivers/pci/dwc/pcie-designware-host.c
> index 28ed32b..4b62315 100644
> --- a/drivers/pci/dwc/pcie-designware-host.c
> +++ b/drivers/pci/dwc/pcie-designware-host.c
> @@ -11,6 +11,7 @@
>   * published by the Free Software Foundation.
>   */
>  
> +#include <linux/irqchip/chained_irq.h>
>  #include <linux/irqdomain.h>
>  #include <linux/of_address.h>
>  #include <linux/of_pci.h>
> @@ -45,40 +46,65 @@ static int dw_pcie_wr_own_conf(struct pcie_port *pp, int where, int size,
>  	return dw_pcie_write(pci->dbi_base + where, size, val);
>  }
>  
> +static void dw_msi_mask_irq(struct irq_data *d) {
> +	pci_msi_mask_irq(d);
> +	irq_chip_mask_parent(d);
> +}
> +
> +static void dw_msi_unmask_irq(struct irq_data *d) {
> +	pci_msi_unmask_irq(d);
> +	irq_chip_unmask_parent(d);
> +}
> +
>  static struct irq_chip dw_msi_irq_chip = {
>  	.name = "PCI-MSI",
> -	.irq_enable = pci_msi_unmask_irq,
> -	.irq_disable = pci_msi_mask_irq,
> -	.irq_mask = pci_msi_mask_irq,
> -	.irq_unmask = pci_msi_unmask_irq,
> +	.irq_mask = dw_msi_mask_irq,
> +	.irq_unmask = dw_msi_unmask_irq,
> +};
> +
> +static struct msi_domain_info dw_pcie_msi_domain_info = {
> +	.flags	= (MSI_FLAG_USE_DEF_DOM_OPS | MSI_FLAG_USE_DEF_CHIP_OPS |
> +		     MSI_FLAG_PCI_MSIX),
> +	.chip	= &dw_msi_irq_chip,
>  };
>  
> -/* MSI int handler */
> -irqreturn_t dw_handle_msi_irq(struct pcie_port *pp)
> +void dw_handle_msi_irq(struct pcie_port *pp)
>  {
> -	u32 val;
> -	int i, pos, irq;
> -	irqreturn_t ret = IRQ_NONE;
> +	int i, pos, virq;
> +	u32 status;
>  
>  	for (i = 0; i < MAX_MSI_CTRLS; i++) {
>  		dw_pcie_rd_own_conf(pp, PCIE_MSI_INTR0_STATUS + i * 12, 4,
> -				    &val);
> -		if (!val)
> +				    &status);
> +		if (!status)
>  			continue;
>  
> -		ret = IRQ_HANDLED;
>  		pos = 0;
> -		while ((pos = find_next_bit((unsigned long *) &val, 32,
> +		while ((pos = find_next_bit((unsigned long *) &status, 32,
>  					    pos)) != 32) {
> -			irq = irq_find_mapping(pp->irq_domain, i * 32 + pos);
> +			virq = irq_find_mapping(pp->irq_domain, i * 32 + pos);
>  			dw_pcie_wr_own_conf(pp, PCIE_MSI_INTR0_STATUS + i * 12,
>  					    4, 1 << pos);
> -			generic_handle_irq(irq);
> +			generic_handle_irq(virq);
>  			pos++;
>  		}
>  	}
> +}
>  
> -	return ret;
> +/* Chained MSI interrupt service routine */
> +static void dw_chained_msi_isr(struct irq_desc *desc)
> +{
> +	struct irq_chip *chip = irq_desc_get_chip(desc);
> +	struct pcie_port *pp;
> +	struct dw_pcie *pci;
> +
> +	chained_irq_enter(chip, desc);
> +	pci = irq_desc_get_handler_data(desc);
> +	pp = &pci->pp;
> +
> +	dw_handle_msi_irq(pp);
> +
> +	chained_irq_exit(chip, desc);
>  }
>  
>  void dw_pcie_msi_init(struct pcie_port *pp)
> @@ -95,183 +121,163 @@ void dw_pcie_msi_init(struct pcie_port *pp)
>  			    (u32)(msi_target >> 32 & 0xffffffff));
>  }
>  
> -static void dw_pcie_msi_clear_irq(struct pcie_port *pp, int irq)
> +static void dw_pci_setup_msi_msg(struct irq_data *data, struct msi_msg *msg)

Please use the word "compose" instead of "setup", as this makes it much
easier to understand (it matches the structure this is assigned to).

>  {
> -	unsigned int res, bit, val;
> +	struct dw_pcie *pci = irq_data_get_irq_chip_data(data);
> +	struct pcie_port *pp = &pci->pp;
> +	u64 msi_target;
>  
> -	res = (irq / 32) * 12;
> -	bit = irq % 32;
> -	dw_pcie_rd_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, &val);
> -	val &= ~(1 << bit);
> -	dw_pcie_wr_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, val);
> -}
> +	if (pp->ops->get_msi_addr)
> +		msi_target = pp->ops->get_msi_addr(pp);
> +	else
> +		msi_target = virt_to_phys((void *)pp->msi_data);

That's quite ugly. You should probably either make the indirection a
requirement, or forbid it altogether.

>  
> -static void clear_irq_range(struct pcie_port *pp, unsigned int irq_base,
> -			    unsigned int nvec, unsigned int pos)
> -{
> -	unsigned int i;
> -
> -	for (i = 0; i < nvec; i++) {
> -		irq_set_msi_desc_off(irq_base, i, NULL);
> -		/* Disable corresponding interrupt on MSI controller */
> -		if (pp->ops->msi_clear_irq)
> -			pp->ops->msi_clear_irq(pp, pos + i);
> -		else
> -			dw_pcie_msi_clear_irq(pp, pos + i);
> -	}
> +	msg->address_lo = (u32)(msi_target & 0xffffffff);
> +	msg->address_hi = (u32)(msi_target >> 32 & 0xffffffff);

Use lower_32_bits/upper_32_bits.

>  
> -	bitmap_release_region(pp->msi_irq_in_use, pos, order_base_2(nvec));
> +	if (pp->ops->get_msi_data)
> +		msg->data = pp->ops->get_msi_data(pp, data->hwirq);
> +	else
> +		msg->data = data->hwirq;
> +
> +	dev_err(pci->dev, "msi#%d address_hi %#x address_lo %#x\n",
> +		(int)data->hwirq, msg->address_hi, msg->address_lo);
>  }
>  
> -static void dw_pcie_msi_set_irq(struct pcie_port *pp, int irq)
> +static int dw_pci_msi_set_affinity(struct irq_data *irq_data,
> +				   const struct cpumask *mask, bool force)
>  {
> -	unsigned int res, bit, val;
> -
> -	res = (irq / 32) * 12;
> -	bit = irq % 32;
> -	dw_pcie_rd_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, &val);
> -	val |= 1 << bit;
> -	dw_pcie_wr_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, val);
> +	return -EINVAL;
>  }
>  
> -static int assign_irq(int no_irqs, struct msi_desc *desc, int *pos)
> +static void dw_pci_bottom_mask(struct irq_data *data)
>  {
> -	int irq, pos0, i;
> -	struct pcie_port *pp;
> -
> -	pp = (struct pcie_port *)msi_desc_to_pci_sysdata(desc);
> -	pos0 = bitmap_find_free_region(pp->msi_irq_in_use, MAX_MSI_IRQS,
> -				       order_base_2(no_irqs));
> -	if (pos0 < 0)
> -		goto no_valid_irq;
> -
> -	irq = irq_find_mapping(pp->irq_domain, pos0);
> -	if (!irq)
> -		goto no_valid_irq;
> -
> -	/*
> -	 * irq_create_mapping (called from dw_pcie_host_init) pre-allocates
> -	 * descs so there is no need to allocate descs here. We can therefore
> -	 * assume that if irq_find_mapping above returns non-zero, then the
> -	 * descs are also successfully allocated.
> -	 */
> +	struct dw_pcie *pci = irq_data_get_irq_chip_data(data);
> +	struct pcie_port *pp = &pci->pp;
> +	unsigned int res, bit, val;
>  
> -	for (i = 0; i < no_irqs; i++) {
> -		if (irq_set_msi_desc_off(irq, i, desc) != 0) {
> -			clear_irq_range(pp, irq, i, pos0);
> -			goto no_valid_irq;
> -		}
> -		/*Enable corresponding interrupt in MSI interrupt controller */
> -		if (pp->ops->msi_set_irq)
> -			pp->ops->msi_set_irq(pp, pos0 + i);
> -		else
> -			dw_pcie_msi_set_irq(pp, pos0 + i);
> -	}
> +	mutex_lock(&pp->lock);

This cannot be a mutex, as mask/unmask can also be called from interrupt
context. Use a spin_lock_irqsave().

>  
> -	*pos = pos0;
> -	desc->nvec_used = no_irqs;
> -	desc->msi_attrib.multiple = order_base_2(no_irqs);
> +	res = (data->hwirq / 32) * 12;
> +	bit = data->hwirq % 32;
>  
> -	return irq;
> +	dw_pcie_rd_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, &val);
> +	val &= ~(1 << bit);
> +	dw_pcie_wr_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, val);

Instead of reading/writing the MMIO register, you'd be better off
keeping a variable in your pcie_port structure, which contains the
current state of that register. You then just update the in-memory
version, and write it back to the HW, avoiding the initial read.

>  
> -no_valid_irq:
> -	*pos = pos0;
> -	return -ENOSPC;
> +	mutex_unlock(&pp->lock);
>  }
>  
> -static void dw_msi_setup_msg(struct pcie_port *pp, unsigned int irq, u32 pos)
> +static void dw_pci_bottom_unmask(struct irq_data *data)
>  {
> -	struct msi_msg msg;
> -	u64 msi_target;
> +	struct dw_pcie *pci = irq_data_get_irq_chip_data(data);
> +	struct pcie_port *pp = &pci->pp;
> +	unsigned int res, bit, val;
>  
> -	if (pp->ops->get_msi_addr)
> -		msi_target = pp->ops->get_msi_addr(pp);
> -	else
> -		msi_target = virt_to_phys((void *)pp->msi_data);
> +	mutex_lock(&pp->lock);

Same here.

>  
> -	msg.address_lo = (u32)(msi_target & 0xffffffff);
> -	msg.address_hi = (u32)(msi_target >> 32 & 0xffffffff);
> +	res = (data->hwirq / 32) * 12;
> +	bit = data->hwirq % 32;
>  
> -	if (pp->ops->get_msi_data)
> -		msg.data = pp->ops->get_msi_data(pp, pos);
> -	else
> -		msg.data = pos;
> +	dw_pcie_rd_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, &val);
> +	val |= 1 << bit;
> +	dw_pcie_wr_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, val);
>  
> -	pci_write_msi_msg(irq, &msg);
> +	mutex_unlock(&pp->lock);
>  }
>  
> -static int dw_msi_setup_irq(struct msi_controller *chip, struct pci_dev *pdev,
> -			    struct msi_desc *desc)
> -{
> -	int irq, pos;
> -	struct pcie_port *pp = pdev->bus->sysdata;
> -
> -	if (desc->msi_attrib.is_msix)
> -		return -EINVAL;
> -
> -	irq = assign_irq(1, desc, &pos);
> -	if (irq < 0)
> -		return irq;
> +static struct irq_chip dw_pci_msi_bottom_irq_chip = {
> +	.name			= "DW MSI",
> +	.irq_compose_msi_msg	= dw_pci_setup_msi_msg,
> +	.irq_set_affinity	= dw_pci_msi_set_affinity,
> +	.irq_mask		= dw_pci_bottom_mask,
> +	.irq_unmask		= dw_pci_bottom_unmask,
> +};
>  
> -	dw_msi_setup_msg(pp, irq, pos);
> +static int dw_pcie_irq_domain_alloc(struct irq_domain *domain,
> +				    unsigned int virq, unsigned int nr_irqs,
> +				    void *args)
> +{
> +	struct dw_pcie *pci = domain->host_data;
> +	struct pcie_port *pp = &pci->pp;
> +	unsigned long bit;
>  
> -	return 0;
> -}
> +	mutex_lock(&pp->lock);
>  
> -static int dw_msi_setup_irqs(struct msi_controller *chip, struct pci_dev *pdev,
> -			     int nvec, int type)
> -{
> -#ifdef CONFIG_PCI_MSI
> -	int irq, pos;
> -	struct msi_desc *desc;
> -	struct pcie_port *pp = pdev->bus->sysdata;
> +	WARN_ON(nr_irqs != 1);
>  
> -	/* MSI-X interrupts are not supported */
> -	if (type == PCI_CAP_ID_MSIX)
> -		return -EINVAL;
> +	bit = find_first_zero_bit(pp->msi_irq_in_use, pp->num_vectors);
> +	if (bit >= pp->num_vectors) {
> +		mutex_unlock(&pp->lock);
> +		return -ENOSPC;
> +	}
>  
> -	WARN_ON(!list_is_singular(&pdev->dev.msi_list));
> -	desc = list_entry(pdev->dev.msi_list.next, struct msi_desc, list);
> +	set_bit(bit, pp->msi_irq_in_use);
>  
> -	irq = assign_irq(nvec, desc, &pos);
> -	if (irq < 0)
> -		return irq;
> +	mutex_unlock(&pp->lock);
>  
> -	dw_msi_setup_msg(pp, irq, pos);
> +	irq_domain_set_info(domain, virq, bit, &dw_pci_msi_bottom_irq_chip,
> +			    domain->host_data, handle_simple_irq,
> +			    NULL, NULL);
>  
>  	return 0;
> -#else
> -	return -EINVAL;
> -#endif
>  }
>  
> -static void dw_msi_teardown_irq(struct msi_controller *chip, unsigned int irq)
> +static void dw_pcie_irq_domain_free(struct irq_domain *domain,
> +				    unsigned int virq, unsigned int nr_irqs)
>  {
> -	struct irq_data *data = irq_get_irq_data(irq);
> -	struct msi_desc *msi = irq_data_get_msi_desc(data);
> -	struct pcie_port *pp = (struct pcie_port *)msi_desc_to_pci_sysdata(msi);
> +	struct irq_data *data = irq_domain_get_irq_data(domain, virq);
> +	struct dw_pcie *pci = irq_data_get_irq_chip_data(data);
> +	struct pcie_port *pp = &pci->pp;
> +
> +	mutex_lock(&pp->lock);
>  
> -	clear_irq_range(pp, irq, 1, data->hwirq);
> +	if (!test_bit(data->hwirq, pp->msi_irq_in_use))
> +		dev_err(pci->dev, "trying to free unused MSI#%lu\n",
> +			data->hwirq);
> +	else
> +		__clear_bit(data->hwirq, pp->msi_irq_in_use);
> +
> +	mutex_unlock(&pp->lock);
>  }
>  
> -static struct msi_controller dw_pcie_msi_chip = {
> -	.setup_irq = dw_msi_setup_irq,
> -	.setup_irqs = dw_msi_setup_irqs,
> -	.teardown_irq = dw_msi_teardown_irq,
> +static const struct irq_domain_ops dw_pcie_msi_domain_ops = {
> +	.alloc	= dw_pcie_irq_domain_alloc,
> +	.free	= dw_pcie_irq_domain_free,
>  };
>  
> -static int dw_pcie_msi_map(struct irq_domain *domain, unsigned int irq,
> -			   irq_hw_number_t hwirq)
> +static int dw_pcie_allocate_domains(struct dw_pcie *pci)
>  {
> -	irq_set_chip_and_handler(irq, &dw_msi_irq_chip, handle_simple_irq);
> -	irq_set_chip_data(irq, domain->host_data);
> +	struct pcie_port *pp = &pci->pp;
> +	struct fwnode_handle *fwnode = of_node_to_fwnode(pci->dev->of_node);
> +
> +	pp->irq_domain = irq_domain_add_linear(NULL, pp->num_vectors,

In general, it is better to use the "create" version and pass the fwnode
to all domains.

> +					       &dw_pcie_msi_domain_ops, pci);
> +	if (!pp->irq_domain) {
> +		dev_err(pci->dev, "failed to create IRQ domain\n");
> +		return -ENOMEM;
> +	}
> +
> +	pp->msi_domain = pci_msi_create_irq_domain(fwnode,
> +						   &dw_pcie_msi_domain_info,
> +						   pp->irq_domain);
> +	if (!pp->msi_domain) {
> +		dev_err(pci->dev, "failed to create MSI domain\n");
> +		irq_domain_remove(pp->irq_domain);
> +		return -ENOMEM;
> +	}
>  
>  	return 0;
>  }
>  
> -static const struct irq_domain_ops msi_domain_ops = {
> -	.map = dw_pcie_msi_map,
> -};
> +void dw_pcie_free_msi(struct pcie_port *pp)
> +{
> +	irq_set_chained_handler(pp->msi_irq, NULL);
> +	irq_set_handler_data(pp->msi_irq, NULL);
> +
> +	irq_domain_remove(pp->msi_domain);
> +	irq_domain_remove(pp->irq_domain);
> +}
>  
>  int dw_pcie_host_init(struct pcie_port *pp)
>  {
> @@ -279,11 +285,13 @@ int dw_pcie_host_init(struct pcie_port *pp)
>  	struct device *dev = pci->dev;
>  	struct device_node *np = dev->of_node;
>  	struct platform_device *pdev = to_platform_device(dev);
> +	struct resource_entry *win, *tmp;
>  	struct pci_bus *bus, *child;
>  	struct resource *cfg_res;
> -	int i, ret;
> +	int ret;
>  	LIST_HEAD(res);
> -	struct resource_entry *win, *tmp;
> +
> +	mutex_init(&pci->pp.lock);
>  
>  	cfg_res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "config");
>  	if (cfg_res) {
> @@ -377,20 +385,22 @@ int dw_pcie_host_init(struct pcie_port *pp)
>  		pci->num_viewport = 2;
>  
>  	if (IS_ENABLED(CONFIG_PCI_MSI)) {
> +		ret = of_property_read_u32(np, "num-vectors",
> +					   &pp->num_vectors);
> +		if (ret)
> +			pp->num_vectors = 1;
> +
>  		if (!pp->ops->msi_host_init) {
> -			pp->irq_domain = irq_domain_add_linear(dev->of_node,
> -						MAX_MSI_IRQS, &msi_domain_ops,
> -						&dw_pcie_msi_chip);
> -			if (!pp->irq_domain) {
> -				dev_err(dev, "irq domain init failed\n");
> -				ret = -ENXIO;
> +			ret = dw_pcie_allocate_domains(pci);
> +			if (ret)
>  				goto error;
> -			}
>  
> -			for (i = 0; i < MAX_MSI_IRQS; i++)
> -				irq_create_mapping(pp->irq_domain, i);
> +			irq_set_chained_handler_and_data(pci->pp.msi_irq,
> +							 dw_chained_msi_isr,
> +							 pci);
>  		} else {
> -			ret = pp->ops->msi_host_init(pp, &dw_pcie_msi_chip);
> +			//TODO
> +			ret = 0;//pp->ops->msi_host_init(pp, &dw_pcie_msi_chip);

What's this TODO for?

>  			if (ret < 0)
>  				goto error;
>  		}
> @@ -400,14 +410,10 @@ int dw_pcie_host_init(struct pcie_port *pp)
>  		pp->ops->host_init(pp);
>  
>  	pp->root_bus_nr = pp->busn->start;
> -	if (IS_ENABLED(CONFIG_PCI_MSI)) {
> -		bus = pci_scan_root_bus_msi(dev, pp->root_bus_nr,
> -					    &dw_pcie_ops, pp, &res,
> -					    &dw_pcie_msi_chip);
> -		dw_pcie_msi_chip.dev = dev;
> -	} else
> -		bus = pci_scan_root_bus(dev, pp->root_bus_nr, &dw_pcie_ops,
> -					pp, &res);
> +
> +	bus = pci_scan_root_bus(dev, pp->root_bus_nr, &dw_pcie_ops,
> +				pp, &res);
> +
>  	if (!bus) {
>  		ret = -ENOMEM;
>  		goto error;
> diff --git a/drivers/pci/dwc/pcie-designware-plat.c b/drivers/pci/dwc/pcie-designware-plat.c
> index 32091b3..3b9717b 100644
> --- a/drivers/pci/dwc/pcie-designware-plat.c
> +++ b/drivers/pci/dwc/pcie-designware-plat.c
> @@ -28,13 +28,6 @@ struct dw_plat_pcie {
>  	struct dw_pcie		*pci;
>  };
>  
> -static irqreturn_t dw_plat_pcie_msi_irq_handler(int irq, void *arg)
> -{
> -	struct pcie_port *pp = arg;
> -
> -	return dw_handle_msi_irq(pp);
> -}
> -
>  static void dw_plat_pcie_host_init(struct pcie_port *pp)
>  {
>  	struct dw_pcie *pci = to_dw_pcie_from_pp(pp);
> @@ -64,14 +57,6 @@ static int dw_plat_add_pcie_port(struct pcie_port *pp,
>  		pp->msi_irq = platform_get_irq(pdev, 0);
>  		if (pp->msi_irq < 0)
>  			return pp->msi_irq;
> -
> -		ret = devm_request_irq(dev, pp->msi_irq,
> -					dw_plat_pcie_msi_irq_handler,
> -					IRQF_SHARED, "dw-plat-pcie-msi", pp);
> -		if (ret) {
> -			dev_err(dev, "failed to request MSI IRQ\n");
> -			return ret;
> -		}
>  	}
>  
>  	pp->root_bus_nr = -1;
> diff --git a/drivers/pci/dwc/pcie-designware.h b/drivers/pci/dwc/pcie-designware.h
> index c6a8405..3abd229 100644
> --- a/drivers/pci/dwc/pcie-designware.h
> +++ b/drivers/pci/dwc/pcie-designware.h
> @@ -165,7 +165,10 @@ struct pcie_port {
>  	struct dw_pcie_host_ops	*ops;
>  	int			msi_irq;
>  	struct irq_domain	*irq_domain;
> +	struct irq_domain	*msi_domain;
>  	unsigned long		msi_data;
> +	u32			num_vectors;
> +	struct mutex		lock;
>  	DECLARE_BITMAP(msi_irq_in_use, MAX_MSI_IRQS);
>  };
>  
> @@ -280,7 +283,8 @@ static inline u32 dw_pcie_readl_dbi2(struct dw_pcie *pci, u32 reg)
>  }
>  
>  #ifdef CONFIG_PCIE_DW_HOST
> -irqreturn_t dw_handle_msi_irq(struct pcie_port *pp);
> +void dw_handle_msi_irq(struct pcie_port *pp);
> +void dw_pcie_free_msi(struct pcie_port *pp);
>  void dw_pcie_msi_init(struct pcie_port *pp);
>  void dw_pcie_setup_rc(struct pcie_port *pp);
>  int dw_pcie_host_init(struct pcie_port *pp);
> diff --git a/drivers/pci/dwc/pcie-qcom.c b/drivers/pci/dwc/pcie-qcom.c
> index 5bf23d4..4e1cc02 100644
> --- a/drivers/pci/dwc/pcie-qcom.c
> +++ b/drivers/pci/dwc/pcie-qcom.c
> @@ -126,13 +126,6 @@ static void qcom_ep_reset_deassert(struct qcom_pcie *pcie)
>  	usleep_range(PERST_DELAY_US, PERST_DELAY_US + 500);
>  }
>  
> -static irqreturn_t qcom_pcie_msi_irq_handler(int irq, void *arg)
> -{
> -	struct pcie_port *pp = arg;
> -
> -	return dw_handle_msi_irq(pp);
> -}
> -
>  static void qcom_pcie_v0_v1_ltssm_enable(struct qcom_pcie *pcie)
>  {
>  	u32 val;
> @@ -724,14 +717,6 @@ static int qcom_pcie_probe(struct platform_device *pdev)
>  		pp->msi_irq = platform_get_irq_byname(pdev, "msi");
>  		if (pp->msi_irq < 0)
>  			return pp->msi_irq;
> -
> -		ret = devm_request_irq(dev, pp->msi_irq,
> -				       qcom_pcie_msi_irq_handler,
> -				       IRQF_SHARED, "qcom-pcie-msi", pp);
> -		if (ret) {
> -			dev_err(dev, "cannot request msi irq\n");
> -			return ret;
> -		}
>  	}
>  
>  	ret = phy_init(pcie->phy);
> 

Thanks,

	M.

Joao Pinto May 11, 2017, 9:17 a.m. UTC | #2

Hello,

Às 6:00 PM de 5/10/2017, Marc Zyngier escreveu:
> On 09/05/17 13:33, Joao Pinto wrote:
>> This is a proposal for the update of the interrupt API in pcie-designware.
>>
>> *SoC specific drivers*
>> All SoC specific drivers that use the common infrastructure (pci-qcom,
>> pci-artpec6, pci-exynos, pci-imx6) were updated to be compatible.
>> Other SoC specific drivers like pci-dra7, pci-armada8k, pci-hisi, pci-spear13xx
>> and pci-layerscape will also work fine.
>>
>> *Work still to be done - need some inputs*
>> The pci-keystone driver is the one that I would appreciate some opinions, since
>> it uses the "old" dw_pcie_msi_chip. I think we have 3 options:
>>
>> a) Keep the old API + new API in pcie-designware just for pci-keystone to use
>> the dw_pcie_msi_chip structure
>> b) Move the old API from pcie-designware to pci-keystone for it to use the
>> dw_pcie_msi_chip structure
>> c) Adapt pci-keystone to use the new API also
> 
> What is the issue with moving Keystone over to this infrastructure?

I don't hardware to test if the Keystone infrastructure porting is right or not
and thats my main concern. Also the Keystone SoC driver maintainer can also have
the option of keeping the current structure and that's why I sent the e-mail
asking for opinions about the subject.

>>
>> *Tests*
>> I made tests with a PCI 4.80 Core IP, using pcie-designware-plat driver.
>> I used an MSI-only Endpoint and a MSI/MSIX Endpoint and the APi adapted very
>> well to the situation.
>>
>> Signed-off-by: Joao Pinto <[e-mail]>
>> ---
>>  drivers/pci/dwc/pci-exynos.c           |  18 --
>>  drivers/pci/dwc/pci-imx6.c             |  18 --
>>  drivers/pci/dwc/pcie-artpec6.c         |  18 --
>>  drivers/pci/dwc/pcie-designware-host.c | 342 +++++++++++++++++----------------
>>  drivers/pci/dwc/pcie-designware-plat.c |  15 --
>>  drivers/pci/dwc/pcie-designware.h      |   6 +-
>>  drivers/pci/dwc/pcie-qcom.c            |  15 --
>>  7 files changed, 179 insertions(+), 253 deletions(-)
> 
> May I suggest something for your next posting? This patch is extremely
> difficult to read (not your fault at all), as it deletes a lot of code
> and replaces it by code that is mostly unrelated. It would be a lot
> better if you had a series that:
> 
> 1: adds the new infrastructure in parallel with the old one, with an
> opt-in mechanism.
> 2: convert each individual platform (one patch per SoC)
> 3: remove the old code entirely.
> 
> This will result in a much more readable series, and will give us a good
> way to bisect, should a given platform break.

Yep indeed! I will do the 3 patch way as you are suggesting.

Thanks!

Marc Zyngier May 11, 2017, 10:05 a.m. UTC | #3

On 11/05/17 10:17, Joao Pinto wrote:
> 
> Hello,
> 
> Às 6:00 PM de 5/10/2017, Marc Zyngier escreveu:
>> On 09/05/17 13:33, Joao Pinto wrote:
>>> This is a proposal for the update of the interrupt API in pcie-designware.
>>>
>>> *SoC specific drivers*
>>> All SoC specific drivers that use the common infrastructure (pci-qcom,
>>> pci-artpec6, pci-exynos, pci-imx6) were updated to be compatible.
>>> Other SoC specific drivers like pci-dra7, pci-armada8k, pci-hisi, pci-spear13xx
>>> and pci-layerscape will also work fine.
>>>
>>> *Work still to be done - need some inputs*
>>> The pci-keystone driver is the one that I would appreciate some opinions, since
>>> it uses the "old" dw_pcie_msi_chip. I think we have 3 options:
>>>
>>> a) Keep the old API + new API in pcie-designware just for pci-keystone to use
>>> the dw_pcie_msi_chip structure
>>> b) Move the old API from pcie-designware to pci-keystone for it to use the
>>> dw_pcie_msi_chip structure
>>> c) Adapt pci-keystone to use the new API also
>>
>> What is the issue with moving Keystone over to this infrastructure?
> 
> I don't hardware to test if the Keystone infrastructure porting is right or not
> and thats my main concern. Also the Keystone SoC driver maintainer can also have
> the option of keeping the current structure and that's why I sent the e-mail
> asking for opinions about the subject.

I think the old infrastructure should disappear, full stop (maintaining
two APIs is hell). It'd be great if you could convert it as well,
leaving the testing responsibility to the TI guys who I'm sure will
gladly help (they've done so in the past for different things).

>>>
>>> *Tests*
>>> I made tests with a PCI 4.80 Core IP, using pcie-designware-plat driver.
>>> I used an MSI-only Endpoint and a MSI/MSIX Endpoint and the APi adapted very
>>> well to the situation.
>>>
>>> Signed-off-by: Joao Pinto <[e-mail]>
>>> ---
>>>  drivers/pci/dwc/pci-exynos.c           |  18 --
>>>  drivers/pci/dwc/pci-imx6.c             |  18 --
>>>  drivers/pci/dwc/pcie-artpec6.c         |  18 --
>>>  drivers/pci/dwc/pcie-designware-host.c | 342 +++++++++++++++++----------------
>>>  drivers/pci/dwc/pcie-designware-plat.c |  15 --
>>>  drivers/pci/dwc/pcie-designware.h      |   6 +-
>>>  drivers/pci/dwc/pcie-qcom.c            |  15 --
>>>  7 files changed, 179 insertions(+), 253 deletions(-)
>>
>> May I suggest something for your next posting? This patch is extremely
>> difficult to read (not your fault at all), as it deletes a lot of code
>> and replaces it by code that is mostly unrelated. It would be a lot
>> better if you had a series that:
>>
>> 1: adds the new infrastructure in parallel with the old one, with an
>> opt-in mechanism.
>> 2: convert each individual platform (one patch per SoC)
>> 3: remove the old code entirely.
>>
>> This will result in a much more readable series, and will give us a good
>> way to bisect, should a given platform break.
> 
> Yep indeed! I will do the 3 patch way as you are suggesting.

Thanks,

	M.

Murali Karicheri May 12, 2017, 4:05 p.m. UTC | #4

On 05/11/2017 06:05 AM, Marc Zyngier wrote:
> On 11/05/17 10:17, Joao Pinto wrote:
>>
>> Hello,
>>
>> Às 6:00 PM de 5/10/2017, Marc Zyngier escreveu:
>>> On 09/05/17 13:33, Joao Pinto wrote:
>>>> This is a proposal for the update of the interrupt API in pcie-designware.
>>>>
>>>> *SoC specific drivers*
>>>> All SoC specific drivers that use the common infrastructure (pci-qcom,
>>>> pci-artpec6, pci-exynos, pci-imx6) were updated to be compatible.
>>>> Other SoC specific drivers like pci-dra7, pci-armada8k, pci-hisi, pci-spear13xx
>>>> and pci-layerscape will also work fine.
>>>>
>>>> *Work still to be done - need some inputs*
>>>> The pci-keystone driver is the one that I would appreciate some opinions, since
>>>> it uses the "old" dw_pcie_msi_chip. I think we have 3 options:
>>>>
>>>> a) Keep the old API + new API in pcie-designware just for pci-keystone to use
>>>> the dw_pcie_msi_chip structure
>>>> b) Move the old API from pcie-designware to pci-keystone for it to use the
>>>> dw_pcie_msi_chip structure
>>>> c) Adapt pci-keystone to use the new API also
>>>
>>> What is the issue with moving Keystone over to this infrastructure?
>>
>> I don't hardware to test if the Keystone infrastructure porting is right or not
>> and thats my main concern. Also the Keystone SoC driver maintainer can also have
>> the option of keeping the current structure and that's why I sent the e-mail
>> asking for opinions about the subject.
> 
> I think the old infrastructure should disappear, full stop (maintaining
> two APIs is hell). It'd be great if you could convert it as well,
> leaving the testing responsibility to the TI guys who I'm sure will
> gladly help (they've done so in the past for different things).
Marc,

Sure! I can help you on the testing part. I haven't had a chance to read this
patch due to my current tasks, but will spend some time on this next week.

Murali

> 
>>>>
>>>> *Tests*
>>>> I made tests with a PCI 4.80 Core IP, using pcie-designware-plat driver.
>>>> I used an MSI-only Endpoint and a MSI/MSIX Endpoint and the APi adapted very
>>>> well to the situation.
>>>>
>>>> Signed-off-by: Joao Pinto <[e-mail]>
>>>> ---
>>>>  drivers/pci/dwc/pci-exynos.c           |  18 --
>>>>  drivers/pci/dwc/pci-imx6.c             |  18 --
>>>>  drivers/pci/dwc/pcie-artpec6.c         |  18 --
>>>>  drivers/pci/dwc/pcie-designware-host.c | 342 +++++++++++++++++----------------
>>>>  drivers/pci/dwc/pcie-designware-plat.c |  15 --
>>>>  drivers/pci/dwc/pcie-designware.h      |   6 +-
>>>>  drivers/pci/dwc/pcie-qcom.c            |  15 --
>>>>  7 files changed, 179 insertions(+), 253 deletions(-)
>>>
>>> May I suggest something for your next posting? This patch is extremely
>>> difficult to read (not your fault at all), as it deletes a lot of code
>>> and replaces it by code that is mostly unrelated. It would be a lot
>>> better if you had a series that:
>>>
>>> 1: adds the new infrastructure in parallel with the old one, with an
>>> opt-in mechanism.
>>> 2: convert each individual platform (one patch per SoC)
>>> 3: remove the old code entirely.
>>>
>>> This will result in a much more readable series, and will give us a good
>>> way to bisect, should a given platform break.
>>
>> Yep indeed! I will do the 3 patch way as you are suggesting.
> 
> Thanks,
> 
> 	M.
>

Marc Zyngier May 12, 2017, 4:11 p.m. UTC | #5

Hi Murali,

On 12/05/17 17:05, Murali Karicheri wrote:
> On 05/11/2017 06:05 AM, Marc Zyngier wrote:
>> On 11/05/17 10:17, Joao Pinto wrote:
>>>
>>> Hello,
>>>
>>> Às 6:00 PM de 5/10/2017, Marc Zyngier escreveu:
>>>> On 09/05/17 13:33, Joao Pinto wrote:
>>>>> This is a proposal for the update of the interrupt API in pcie-designware.
>>>>>
>>>>> *SoC specific drivers*
>>>>> All SoC specific drivers that use the common infrastructure (pci-qcom,
>>>>> pci-artpec6, pci-exynos, pci-imx6) were updated to be compatible.
>>>>> Other SoC specific drivers like pci-dra7, pci-armada8k, pci-hisi, pci-spear13xx
>>>>> and pci-layerscape will also work fine.
>>>>>
>>>>> *Work still to be done - need some inputs*
>>>>> The pci-keystone driver is the one that I would appreciate some opinions, since
>>>>> it uses the "old" dw_pcie_msi_chip. I think we have 3 options:
>>>>>
>>>>> a) Keep the old API + new API in pcie-designware just for pci-keystone to use
>>>>> the dw_pcie_msi_chip structure
>>>>> b) Move the old API from pcie-designware to pci-keystone for it to use the
>>>>> dw_pcie_msi_chip structure
>>>>> c) Adapt pci-keystone to use the new API also
>>>>
>>>> What is the issue with moving Keystone over to this infrastructure?
>>>
>>> I don't hardware to test if the Keystone infrastructure porting is right or not
>>> and thats my main concern. Also the Keystone SoC driver maintainer can also have
>>> the option of keeping the current structure and that's why I sent the e-mail
>>> asking for opinions about the subject.
>>
>> I think the old infrastructure should disappear, full stop (maintaining
>> two APIs is hell). It'd be great if you could convert it as well,
>> leaving the testing responsibility to the TI guys who I'm sure will
>> gladly help (they've done so in the past for different things).
> Marc,
> 
> Sure! I can help you on the testing part. I haven't had a chance to read this
> patch due to my current tasks, but will spend some time on this next week.

Awesome! Please liaise with Joao in order to get this up and running!

Thanks both,

	M.

Murali Karicheri May 12, 2017, 4:21 p.m. UTC | #6

On 05/09/2017 08:33 AM, Joao Pinto wrote:
> This is a proposal for the update of the interrupt API in pcie-designware.
> 
> *SoC specific drivers*
> All SoC specific drivers that use the common infrastructure (pci-qcom,
> pci-artpec6, pci-exynos, pci-imx6) were updated to be compatible.
> Other SoC specific drivers like pci-dra7, pci-armada8k, pci-hisi, pci-spear13xx
> and pci-layerscape will also work fine.
> 
> *Work still to be done - need some inputs*
> The pci-keystone driver is the one that I would appreciate some opinions, since
> it uses the "old" dw_pcie_msi_chip. I think we have 3 options:
> 
> a) Keep the old API + new API in pcie-designware just for pci-keystone to use
> the dw_pcie_msi_chip structure
> b) Move the old API from pcie-designware to pci-keystone for it to use the
> dw_pcie_msi_chip structure
> c) Adapt pci-keystone to use the new API also
> 
> *Tests*
> I made tests with a PCI 4.80 Core IP, using pcie-designware-plat driver.
> I used an MSI-only Endpoint and a MSI/MSIX Endpoint and the APi adapted very
> well to the situation.
> 
> Signed-off-by: Joao Pinto <[e-mail]>
> ---
>  drivers/pci/dwc/pci-exynos.c           |  18 --
>  drivers/pci/dwc/pci-imx6.c             |  18 --
>  drivers/pci/dwc/pcie-artpec6.c         |  18 --
>  drivers/pci/dwc/pcie-designware-host.c | 342 +++++++++++++++++----------------
>  drivers/pci/dwc/pcie-designware-plat.c |  15 --
>  drivers/pci/dwc/pcie-designware.h      |   6 +-
>  drivers/pci/dwc/pcie-qcom.c            |  15 --
>  7 files changed, 179 insertions(+), 253 deletions(-)
> 
> diff --git a/drivers/pci/dwc/pci-exynos.c b/drivers/pci/dwc/pci-exynos.c
> index 546082a..e49c39a 100644
> --- a/drivers/pci/dwc/pci-exynos.c
> +++ b/drivers/pci/dwc/pci-exynos.c
> @@ -490,15 +490,6 @@ static irqreturn_t exynos_pcie_irq_handler(int irq, void *arg)
>  	return IRQ_HANDLED;
>  }
>  
> -static irqreturn_t exynos_pcie_msi_irq_handler(int irq, void *arg)
> -{
> -	struct exynos_pcie *ep = arg;
> -	struct dw_pcie *pci = ep->pci;
> -	struct pcie_port *pp = &pci->pp;
> -
> -	return dw_handle_msi_irq(pp);
> -}
> -
>  static void exynos_pcie_msi_init(struct exynos_pcie *ep)
>  {
>  	struct dw_pcie *pci = ep->pci;
> @@ -622,15 +613,6 @@ static int __init exynos_add_pcie_port(struct exynos_pcie *ep,
>  			dev_err(dev, "failed to get msi irq\n");
>  			return -ENODEV;
>  		}
> -
> -		ret = devm_request_irq(dev, pp->msi_irq,
> -					exynos_pcie_msi_irq_handler,
> -					IRQF_SHARED | IRQF_NO_THREAD,
> -					"exynos-pcie", ep);
> -		if (ret) {
> -			dev_err(dev, "failed to request msi irq\n");
> -			return ret;
> -		}
>  	}
>  
>  	pp->root_bus_nr = -1;
> diff --git a/drivers/pci/dwc/pci-imx6.c b/drivers/pci/dwc/pci-imx6.c
> index a98cba5..f22ce1f 100644
> --- a/drivers/pci/dwc/pci-imx6.c
> +++ b/drivers/pci/dwc/pci-imx6.c
> @@ -490,15 +490,6 @@ static int imx6_pcie_wait_for_speed_change(struct imx6_pcie *imx6_pcie)
>  	return -EINVAL;
>  }
>  
> -static irqreturn_t imx6_pcie_msi_handler(int irq, void *arg)
> -{
> -	struct imx6_pcie *imx6_pcie = arg;
> -	struct dw_pcie *pci = imx6_pcie->pci;
> -	struct pcie_port *pp = &pci->pp;
> -
> -	return dw_handle_msi_irq(pp);
> -}
> -
>  static int imx6_pcie_establish_link(struct imx6_pcie *imx6_pcie)
>  {
>  	struct dw_pcie *pci = imx6_pcie->pci;
> @@ -620,15 +611,6 @@ static int imx6_add_pcie_port(struct imx6_pcie *imx6_pcie,
>  			dev_err(dev, "failed to get MSI irq\n");
>  			return -ENODEV;
>  		}
> -
> -		ret = devm_request_irq(dev, pp->msi_irq,
> -				       imx6_pcie_msi_handler,
> -				       IRQF_SHARED | IRQF_NO_THREAD,
> -				       "mx6-pcie-msi", imx6_pcie);
> -		if (ret) {
> -			dev_err(dev, "failed to request MSI irq\n");
> -			return ret;
> -		}
>  	}
>  
>  	pp->root_bus_nr = -1;
> diff --git a/drivers/pci/dwc/pcie-artpec6.c b/drivers/pci/dwc/pcie-artpec6.c
> index 82a04ac..d81789b 100644
> --- a/drivers/pci/dwc/pcie-artpec6.c
> +++ b/drivers/pci/dwc/pcie-artpec6.c
> @@ -188,15 +188,6 @@ static struct dw_pcie_host_ops artpec6_pcie_host_ops = {
>  	.host_init = artpec6_pcie_host_init,
>  };
>  
> -static irqreturn_t artpec6_pcie_msi_handler(int irq, void *arg)
> -{
> -	struct artpec6_pcie *artpec6_pcie = arg;
> -	struct dw_pcie *pci = artpec6_pcie->pci;
> -	struct pcie_port *pp = &pci->pp;
> -
> -	return dw_handle_msi_irq(pp);
> -}
> -
>  static int artpec6_add_pcie_port(struct artpec6_pcie *artpec6_pcie,
>  				 struct platform_device *pdev)
>  {
> @@ -211,15 +202,6 @@ static int artpec6_add_pcie_port(struct artpec6_pcie *artpec6_pcie,
>  			dev_err(dev, "failed to get MSI irq\n");
>  			return -ENODEV;
>  		}
> -
> -		ret = devm_request_irq(dev, pp->msi_irq,
> -				       artpec6_pcie_msi_handler,
> -				       IRQF_SHARED | IRQF_NO_THREAD,
> -				       "artpec6-pcie-msi", artpec6_pcie);
> -		if (ret) {
> -			dev_err(dev, "failed to request MSI irq\n");
> -			return ret;
> -		}
>  	}
>  
>  	pp->root_bus_nr = -1;
> diff --git a/drivers/pci/dwc/pcie-designware-host.c b/drivers/pci/dwc/pcie-designware-host.c
> index 28ed32b..4b62315 100644
> --- a/drivers/pci/dwc/pcie-designware-host.c
> +++ b/drivers/pci/dwc/pcie-designware-host.c
> @@ -11,6 +11,7 @@
>   * published by the Free Software Foundation.
>   */
>  
> +#include <linux/irqchip/chained_irq.h>
>  #include <linux/irqdomain.h>
>  #include <linux/of_address.h>
>  #include <linux/of_pci.h>
> @@ -45,40 +46,65 @@ static int dw_pcie_wr_own_conf(struct pcie_port *pp, int where, int size,
>  	return dw_pcie_write(pci->dbi_base + where, size, val);
>  }
>  
> +static void dw_msi_mask_irq(struct irq_data *d) {
> +	pci_msi_mask_irq(d);
> +	irq_chip_mask_parent(d);
> +}
> +
> +static void dw_msi_unmask_irq(struct irq_data *d) {
> +	pci_msi_unmask_irq(d);
> +	irq_chip_unmask_parent(d);
> +}
> +
>  static struct irq_chip dw_msi_irq_chip = {
>  	.name = "PCI-MSI",
> -	.irq_enable = pci_msi_unmask_irq,
> -	.irq_disable = pci_msi_mask_irq,
> -	.irq_mask = pci_msi_mask_irq,
> -	.irq_unmask = pci_msi_unmask_irq,
> +	.irq_mask = dw_msi_mask_irq,
> +	.irq_unmask = dw_msi_unmask_irq,
> +};
> +
> +static struct msi_domain_info dw_pcie_msi_domain_info = {
> +	.flags	= (MSI_FLAG_USE_DEF_DOM_OPS | MSI_FLAG_USE_DEF_CHIP_OPS |
> +		     MSI_FLAG_PCI_MSIX),
> +	.chip	= &dw_msi_irq_chip,
>  };
>  
> -/* MSI int handler */
> -irqreturn_t dw_handle_msi_irq(struct pcie_port *pp)
> +void dw_handle_msi_irq(struct pcie_port *pp)
>  {
> -	u32 val;
> -	int i, pos, irq;
> -	irqreturn_t ret = IRQ_NONE;
> +	int i, pos, virq;
> +	u32 status;
>  
>  	for (i = 0; i < MAX_MSI_CTRLS; i++) {
>  		dw_pcie_rd_own_conf(pp, PCIE_MSI_INTR0_STATUS + i * 12, 4,
> -				    &val);
> -		if (!val)
> +				    &status);
> +		if (!status)
>  			continue;
>  
> -		ret = IRQ_HANDLED;
>  		pos = 0;
> -		while ((pos = find_next_bit((unsigned long *) &val, 32,
> +		while ((pos = find_next_bit((unsigned long *) &status, 32,
>  					    pos)) != 32) {
> -			irq = irq_find_mapping(pp->irq_domain, i * 32 + pos);
> +			virq = irq_find_mapping(pp->irq_domain, i * 32 + pos);
>  			dw_pcie_wr_own_conf(pp, PCIE_MSI_INTR0_STATUS + i * 12,
>  					    4, 1 << pos);
> -			generic_handle_irq(irq);
> +			generic_handle_irq(virq);
>  			pos++;
>  		}
>  	}
> +}
>  
> -	return ret;
> +/* Chained MSI interrupt service routine */
> +static void dw_chained_msi_isr(struct irq_desc *desc)
> +{
> +	struct irq_chip *chip = irq_desc_get_chip(desc);
> +	struct pcie_port *pp;
> +	struct dw_pcie *pci;
> +
> +	chained_irq_enter(chip, desc);
> +	pci = irq_desc_get_handler_data(desc);
> +	pp = &pci->pp;
> +
> +	dw_handle_msi_irq(pp);
> +
> +	chained_irq_exit(chip, desc);
>  }
>  
>  void dw_pcie_msi_init(struct pcie_port *pp)
> @@ -95,183 +121,163 @@ void dw_pcie_msi_init(struct pcie_port *pp)
>  			    (u32)(msi_target >> 32 & 0xffffffff));
>  }
>  
> -static void dw_pcie_msi_clear_irq(struct pcie_port *pp, int irq)
> +static void dw_pci_setup_msi_msg(struct irq_data *data, struct msi_msg *msg)
>  {
> -	unsigned int res, bit, val;
> +	struct dw_pcie *pci = irq_data_get_irq_chip_data(data);
> +	struct pcie_port *pp = &pci->pp;
> +	u64 msi_target;
>  
> -	res = (irq / 32) * 12;
> -	bit = irq % 32;
> -	dw_pcie_rd_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, &val);
> -	val &= ~(1 << bit);
> -	dw_pcie_wr_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, val);
> -}
> +	if (pp->ops->get_msi_addr)
> +		msi_target = pp->ops->get_msi_addr(pp);
> +	else
> +		msi_target = virt_to_phys((void *)pp->msi_data);
>  
> -static void clear_irq_range(struct pcie_port *pp, unsigned int irq_base,
> -			    unsigned int nvec, unsigned int pos)
> -{
> -	unsigned int i;
> -
> -	for (i = 0; i < nvec; i++) {
> -		irq_set_msi_desc_off(irq_base, i, NULL);
> -		/* Disable corresponding interrupt on MSI controller */
> -		if (pp->ops->msi_clear_irq)
> -			pp->ops->msi_clear_irq(pp, pos + i);
> -		else
> -			dw_pcie_msi_clear_irq(pp, pos + i);
> -	}
> +	msg->address_lo = (u32)(msi_target & 0xffffffff);
> +	msg->address_hi = (u32)(msi_target >> 32 & 0xffffffff);
>  
> -	bitmap_release_region(pp->msi_irq_in_use, pos, order_base_2(nvec));
> +	if (pp->ops->get_msi_data)
> +		msg->data = pp->ops->get_msi_data(pp, data->hwirq);
> +	else
> +		msg->data = data->hwirq;
> +
> +	dev_err(pci->dev, "msi#%d address_hi %#x address_lo %#x\n",
> +		(int)data->hwirq, msg->address_hi, msg->address_lo);
>  }
>  
> -static void dw_pcie_msi_set_irq(struct pcie_port *pp, int irq)
> +static int dw_pci_msi_set_affinity(struct irq_data *irq_data,
> +				   const struct cpumask *mask, bool force)
>  {
> -	unsigned int res, bit, val;
> -
> -	res = (irq / 32) * 12;
> -	bit = irq % 32;
> -	dw_pcie_rd_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, &val);
> -	val |= 1 << bit;
> -	dw_pcie_wr_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, val);
> +	return -EINVAL;
>  }
>  
> -static int assign_irq(int no_irqs, struct msi_desc *desc, int *pos)
> +static void dw_pci_bottom_mask(struct irq_data *data)
>  {
> -	int irq, pos0, i;
> -	struct pcie_port *pp;
> -
> -	pp = (struct pcie_port *)msi_desc_to_pci_sysdata(desc);
> -	pos0 = bitmap_find_free_region(pp->msi_irq_in_use, MAX_MSI_IRQS,
> -				       order_base_2(no_irqs));
> -	if (pos0 < 0)
> -		goto no_valid_irq;
> -
> -	irq = irq_find_mapping(pp->irq_domain, pos0);
> -	if (!irq)
> -		goto no_valid_irq;
> -
> -	/*
> -	 * irq_create_mapping (called from dw_pcie_host_init) pre-allocates
> -	 * descs so there is no need to allocate descs here. We can therefore
> -	 * assume that if irq_find_mapping above returns non-zero, then the
> -	 * descs are also successfully allocated.
> -	 */
> +	struct dw_pcie *pci = irq_data_get_irq_chip_data(data);
> +	struct pcie_port *pp = &pci->pp;
> +	unsigned int res, bit, val;
>  
> -	for (i = 0; i < no_irqs; i++) {
> -		if (irq_set_msi_desc_off(irq, i, desc) != 0) {
> -			clear_irq_range(pp, irq, i, pos0);
> -			goto no_valid_irq;
> -		}
> -		/*Enable corresponding interrupt in MSI interrupt controller */
> -		if (pp->ops->msi_set_irq)
> -			pp->ops->msi_set_irq(pp, pos0 + i);
> -		else
> -			dw_pcie_msi_set_irq(pp, pos0 + i);
> -	}
> +	mutex_lock(&pp->lock);
>  
> -	*pos = pos0;
> -	desc->nvec_used = no_irqs;
> -	desc->msi_attrib.multiple = order_base_2(no_irqs);
> +	res = (data->hwirq / 32) * 12;
> +	bit = data->hwirq % 32;
>  
> -	return irq;
> +	dw_pcie_rd_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, &val);
> +	val &= ~(1 << bit);
> +	dw_pcie_wr_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, val);
>  
> -no_valid_irq:
> -	*pos = pos0;
> -	return -ENOSPC;
> +	mutex_unlock(&pp->lock);
>  }
>  
> -static void dw_msi_setup_msg(struct pcie_port *pp, unsigned int irq, u32 pos)
> +static void dw_pci_bottom_unmask(struct irq_data *data)
>  {
> -	struct msi_msg msg;
> -	u64 msi_target;
> +	struct dw_pcie *pci = irq_data_get_irq_chip_data(data);
> +	struct pcie_port *pp = &pci->pp;
> +	unsigned int res, bit, val;
>  
> -	if (pp->ops->get_msi_addr)
> -		msi_target = pp->ops->get_msi_addr(pp);
> -	else
> -		msi_target = virt_to_phys((void *)pp->msi_data);
> +	mutex_lock(&pp->lock);
>  
> -	msg.address_lo = (u32)(msi_target & 0xffffffff);
> -	msg.address_hi = (u32)(msi_target >> 32 & 0xffffffff);
> +	res = (data->hwirq / 32) * 12;
> +	bit = data->hwirq % 32;
>  
> -	if (pp->ops->get_msi_data)
> -		msg.data = pp->ops->get_msi_data(pp, pos);
> -	else
> -		msg.data = pos;
> +	dw_pcie_rd_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, &val);
> +	val |= 1 << bit;
> +	dw_pcie_wr_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, val);
>  
> -	pci_write_msi_msg(irq, &msg);
> +	mutex_unlock(&pp->lock);
>  }
>  
> -static int dw_msi_setup_irq(struct msi_controller *chip, struct pci_dev *pdev,
> -			    struct msi_desc *desc)
> -{
> -	int irq, pos;
> -	struct pcie_port *pp = pdev->bus->sysdata;
> -
> -	if (desc->msi_attrib.is_msix)
> -		return -EINVAL;
> -
> -	irq = assign_irq(1, desc, &pos);
> -	if (irq < 0)
> -		return irq;
> +static struct irq_chip dw_pci_msi_bottom_irq_chip = {
> +	.name			= "DW MSI",
> +	.irq_compose_msi_msg	= dw_pci_setup_msi_msg,
> +	.irq_set_affinity	= dw_pci_msi_set_affinity,
> +	.irq_mask		= dw_pci_bottom_mask,
> +	.irq_unmask		= dw_pci_bottom_unmask,
> +};
>  
> -	dw_msi_setup_msg(pp, irq, pos);
> +static int dw_pcie_irq_domain_alloc(struct irq_domain *domain,
> +				    unsigned int virq, unsigned int nr_irqs,
> +				    void *args)
> +{
> +	struct dw_pcie *pci = domain->host_data;
> +	struct pcie_port *pp = &pci->pp;
> +	unsigned long bit;
>  
> -	return 0;
> -}
> +	mutex_lock(&pp->lock);
>  
> -static int dw_msi_setup_irqs(struct msi_controller *chip, struct pci_dev *pdev,
> -			     int nvec, int type)
> -{
> -#ifdef CONFIG_PCI_MSI
> -	int irq, pos;
> -	struct msi_desc *desc;
> -	struct pcie_port *pp = pdev->bus->sysdata;
> +	WARN_ON(nr_irqs != 1);
>  
> -	/* MSI-X interrupts are not supported */
> -	if (type == PCI_CAP_ID_MSIX)
> -		return -EINVAL;
> +	bit = find_first_zero_bit(pp->msi_irq_in_use, pp->num_vectors);
> +	if (bit >= pp->num_vectors) {
> +		mutex_unlock(&pp->lock);
> +		return -ENOSPC;
> +	}
>  
> -	WARN_ON(!list_is_singular(&pdev->dev.msi_list));
> -	desc = list_entry(pdev->dev.msi_list.next, struct msi_desc, list);
> +	set_bit(bit, pp->msi_irq_in_use);
>  
> -	irq = assign_irq(nvec, desc, &pos);
> -	if (irq < 0)
> -		return irq;
> +	mutex_unlock(&pp->lock);
>  
> -	dw_msi_setup_msg(pp, irq, pos);
> +	irq_domain_set_info(domain, virq, bit, &dw_pci_msi_bottom_irq_chip,
> +			    domain->host_data, handle_simple_irq,
> +			    NULL, NULL);
>  
>  	return 0;
> -#else
> -	return -EINVAL;
> -#endif
>  }
>  
> -static void dw_msi_teardown_irq(struct msi_controller *chip, unsigned int irq)
> +static void dw_pcie_irq_domain_free(struct irq_domain *domain,
> +				    unsigned int virq, unsigned int nr_irqs)
>  {
> -	struct irq_data *data = irq_get_irq_data(irq);
> -	struct msi_desc *msi = irq_data_get_msi_desc(data);
> -	struct pcie_port *pp = (struct pcie_port *)msi_desc_to_pci_sysdata(msi);
> +	struct irq_data *data = irq_domain_get_irq_data(domain, virq);
> +	struct dw_pcie *pci = irq_data_get_irq_chip_data(data);
> +	struct pcie_port *pp = &pci->pp;
> +
> +	mutex_lock(&pp->lock);
>  
> -	clear_irq_range(pp, irq, 1, data->hwirq);
> +	if (!test_bit(data->hwirq, pp->msi_irq_in_use))
> +		dev_err(pci->dev, "trying to free unused MSI#%lu\n",
> +			data->hwirq);
> +	else
> +		__clear_bit(data->hwirq, pp->msi_irq_in_use);
> +
> +	mutex_unlock(&pp->lock);
>  }
>  
> -static struct msi_controller dw_pcie_msi_chip = {
> -	.setup_irq = dw_msi_setup_irq,
> -	.setup_irqs = dw_msi_setup_irqs,
> -	.teardown_irq = dw_msi_teardown_irq,
> +static const struct irq_domain_ops dw_pcie_msi_domain_ops = {
> +	.alloc	= dw_pcie_irq_domain_alloc,
> +	.free	= dw_pcie_irq_domain_free,
>  };
>  
> -static int dw_pcie_msi_map(struct irq_domain *domain, unsigned int irq,
> -			   irq_hw_number_t hwirq)
> +static int dw_pcie_allocate_domains(struct dw_pcie *pci)
>  {
> -	irq_set_chip_and_handler(irq, &dw_msi_irq_chip, handle_simple_irq);
> -	irq_set_chip_data(irq, domain->host_data);
> +	struct pcie_port *pp = &pci->pp;
> +	struct fwnode_handle *fwnode = of_node_to_fwnode(pci->dev->of_node);
> +
> +	pp->irq_domain = irq_domain_add_linear(NULL, pp->num_vectors,
> +					       &dw_pcie_msi_domain_ops, pci);
> +	if (!pp->irq_domain) {
> +		dev_err(pci->dev, "failed to create IRQ domain\n");
> +		return -ENOMEM;
> +	}
> +
> +	pp->msi_domain = pci_msi_create_irq_domain(fwnode,
> +						   &dw_pcie_msi_domain_info,
> +						   pp->irq_domain);
> +	if (!pp->msi_domain) {
> +		dev_err(pci->dev, "failed to create MSI domain\n");
> +		irq_domain_remove(pp->irq_domain);
> +		return -ENOMEM;
> +	}
>  
>  	return 0;
>  }
>  
> -static const struct irq_domain_ops msi_domain_ops = {
> -	.map = dw_pcie_msi_map,
> -};
> +void dw_pcie_free_msi(struct pcie_port *pp)
> +{
> +	irq_set_chained_handler(pp->msi_irq, NULL);
> +	irq_set_handler_data(pp->msi_irq, NULL);
> +
> +	irq_domain_remove(pp->msi_domain);
> +	irq_domain_remove(pp->irq_domain);
> +}
>  
>  int dw_pcie_host_init(struct pcie_port *pp)
>  {
> @@ -279,11 +285,13 @@ int dw_pcie_host_init(struct pcie_port *pp)
>  	struct device *dev = pci->dev;
>  	struct device_node *np = dev->of_node;
>  	struct platform_device *pdev = to_platform_device(dev);
> +	struct resource_entry *win, *tmp;
>  	struct pci_bus *bus, *child;
>  	struct resource *cfg_res;
> -	int i, ret;
> +	int ret;
>  	LIST_HEAD(res);
> -	struct resource_entry *win, *tmp;
> +
> +	mutex_init(&pci->pp.lock);
>  
>  	cfg_res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "config");
>  	if (cfg_res) {
> @@ -377,20 +385,22 @@ int dw_pcie_host_init(struct pcie_port *pp)
>  		pci->num_viewport = 2;
>  
>  	if (IS_ENABLED(CONFIG_PCI_MSI)) {
> +		ret = of_property_read_u32(np, "num-vectors",
> +					   &pp->num_vectors);
> +		if (ret)
> +			pp->num_vectors = 1;
> +
>  		if (!pp->ops->msi_host_init) {
> -			pp->irq_domain = irq_domain_add_linear(dev->of_node,
> -						MAX_MSI_IRQS, &msi_domain_ops,
> -						&dw_pcie_msi_chip);
> -			if (!pp->irq_domain) {
> -				dev_err(dev, "irq domain init failed\n");
> -				ret = -ENXIO;
> +			ret = dw_pcie_allocate_domains(pci);
> +			if (ret)
>  				goto error;
> -			}
>  
> -			for (i = 0; i < MAX_MSI_IRQS; i++)
> -				irq_create_mapping(pp->irq_domain, i);
> +			irq_set_chained_handler_and_data(pci->pp.msi_irq,
> +							 dw_chained_msi_isr,
> +							 pci);
>  		} else {
> -			ret = pp->ops->msi_host_init(pp, &dw_pcie_msi_chip);
> +			//TODO
> +			ret = 0;//pp->ops->msi_host_init(pp, &dw_pcie_msi_chip);
>  			if (ret < 0)
>  				goto error;
>  		}
> @@ -400,14 +410,10 @@ int dw_pcie_host_init(struct pcie_port *pp)
>  		pp->ops->host_init(pp);
>  
>  	pp->root_bus_nr = pp->busn->start;
> -	if (IS_ENABLED(CONFIG_PCI_MSI)) {
> -		bus = pci_scan_root_bus_msi(dev, pp->root_bus_nr,
> -					    &dw_pcie_ops, pp, &res,
> -					    &dw_pcie_msi_chip);
> -		dw_pcie_msi_chip.dev = dev;
> -	} else
> -		bus = pci_scan_root_bus(dev, pp->root_bus_nr, &dw_pcie_ops,
> -					pp, &res);
> +
> +	bus = pci_scan_root_bus(dev, pp->root_bus_nr, &dw_pcie_ops,
> +				pp, &res);
> +
>  	if (!bus) {
>  		ret = -ENOMEM;
>  		goto error;
> diff --git a/drivers/pci/dwc/pcie-designware-plat.c b/drivers/pci/dwc/pcie-designware-plat.c
> index 32091b3..3b9717b 100644
> --- a/drivers/pci/dwc/pcie-designware-plat.c
> +++ b/drivers/pci/dwc/pcie-designware-plat.c
> @@ -28,13 +28,6 @@ struct dw_plat_pcie {
>  	struct dw_pcie		*pci;
>  };
>  
> -static irqreturn_t dw_plat_pcie_msi_irq_handler(int irq, void *arg)
> -{
> -	struct pcie_port *pp = arg;
> -
> -	return dw_handle_msi_irq(pp);
> -}
> -
>  static void dw_plat_pcie_host_init(struct pcie_port *pp)
>  {
>  	struct dw_pcie *pci = to_dw_pcie_from_pp(pp);
> @@ -64,14 +57,6 @@ static int dw_plat_add_pcie_port(struct pcie_port *pp,
>  		pp->msi_irq = platform_get_irq(pdev, 0);
>  		if (pp->msi_irq < 0)
>  			return pp->msi_irq;
> -
> -		ret = devm_request_irq(dev, pp->msi_irq,
> -					dw_plat_pcie_msi_irq_handler,
> -					IRQF_SHARED, "dw-plat-pcie-msi", pp);
> -		if (ret) {
> -			dev_err(dev, "failed to request MSI IRQ\n");
> -			return ret;
> -		}
>  	}
>  
>  	pp->root_bus_nr = -1;
> diff --git a/drivers/pci/dwc/pcie-designware.h b/drivers/pci/dwc/pcie-designware.h
> index c6a8405..3abd229 100644
> --- a/drivers/pci/dwc/pcie-designware.h
> +++ b/drivers/pci/dwc/pcie-designware.h
> @@ -165,7 +165,10 @@ struct pcie_port {
>  	struct dw_pcie_host_ops	*ops;
>  	int			msi_irq;
>  	struct irq_domain	*irq_domain;
> +	struct irq_domain	*msi_domain;
>  	unsigned long		msi_data;
> +	u32			num_vectors;
> +	struct mutex		lock;
>  	DECLARE_BITMAP(msi_irq_in_use, MAX_MSI_IRQS);
>  };
>  
> @@ -280,7 +283,8 @@ static inline u32 dw_pcie_readl_dbi2(struct dw_pcie *pci, u32 reg)
>  }
>  
>  #ifdef CONFIG_PCIE_DW_HOST
> -irqreturn_t dw_handle_msi_irq(struct pcie_port *pp);
> +void dw_handle_msi_irq(struct pcie_port *pp);
> +void dw_pcie_free_msi(struct pcie_port *pp);
>  void dw_pcie_msi_init(struct pcie_port *pp);
>  void dw_pcie_setup_rc(struct pcie_port *pp);
>  int dw_pcie_host_init(struct pcie_port *pp);
> diff --git a/drivers/pci/dwc/pcie-qcom.c b/drivers/pci/dwc/pcie-qcom.c
> index 5bf23d4..4e1cc02 100644
> --- a/drivers/pci/dwc/pcie-qcom.c
> +++ b/drivers/pci/dwc/pcie-qcom.c
> @@ -126,13 +126,6 @@ static void qcom_ep_reset_deassert(struct qcom_pcie *pcie)
>  	usleep_range(PERST_DELAY_US, PERST_DELAY_US + 500);
>  }
>  
> -static irqreturn_t qcom_pcie_msi_irq_handler(int irq, void *arg)
> -{
> -	struct pcie_port *pp = arg;
> -
> -	return dw_handle_msi_irq(pp);
> -}
> -
>  static void qcom_pcie_v0_v1_ltssm_enable(struct qcom_pcie *pcie)
>  {
>  	u32 val;
> @@ -724,14 +717,6 @@ static int qcom_pcie_probe(struct platform_device *pdev)
>  		pp->msi_irq = platform_get_irq_byname(pdev, "msi");
>  		if (pp->msi_irq < 0)
>  			return pp->msi_irq;
> -
> -		ret = devm_request_irq(dev, pp->msi_irq,
> -				       qcom_pcie_msi_irq_handler,
> -				       IRQF_SHARED, "qcom-pcie-msi", pp);
> -		if (ret) {
> -			dev_err(dev, "cannot request msi irq\n");
> -			return ret;
> -		}
>  	}
>  
>  	ret = phy_init(pcie->phy);
> 
Joao,

What do you see the challenge in porting the PCI-Keystone driver to the new
interrupt API (option c)? I haven't had a chance to review your patches and recent
changes in Linux kernel for PCI as we are currently using v4.9 kernel for
our platforms. One key difference in the PCI Core used in Keystone SoC is that
the MSI is not implemented using standard registers as in newer PCI Designware
core. It uses application register space for MSI interrupts. Given that
is the case, do you see any issue in porting the driver to the new API?

I should be able to test your patch on Keystone.

Joao Pinto May 15, 2017, 11:13 a.m. UTC | #7

Hi Murali,

Às 5:21 PM de 5/12/2017, Murali Karicheri escreveu:
> On 05/09/2017 08:33 AM, Joao Pinto wrote:
>> This is a proposal for the update of the interrupt API in pcie-designware.
>>
>> *SoC specific drivers*
>> All SoC specific drivers that use the common infrastructure (pci-qcom,
>> pci-artpec6, pci-exynos, pci-imx6) were updated to be compatible.
>> Other SoC specific drivers like pci-dra7, pci-armada8k, pci-hisi, pci-spear13xx
>> and pci-layerscape will also work fine.
>>
>> *Work still to be done - need some inputs*
>> The pci-keystone driver is the one that I would appreciate some opinions, since
>> it uses the "old" dw_pcie_msi_chip. I think we have 3 options:
>>
>> a) Keep the old API + new API in pcie-designware just for pci-keystone to use
>> the dw_pcie_msi_chip structure
>> b) Move the old API from pcie-designware to pci-keystone for it to use the
>> dw_pcie_msi_chip structure
>> c) Adapt pci-keystone to use the new API also
>>
>> *Tests*
>> I made tests with a PCI 4.80 Core IP, using pcie-designware-plat driver.
>> I used an MSI-only Endpoint and a MSI/MSIX Endpoint and the APi adapted very
>> well to the situation.
>>
>> Signed-off-by: Joao Pinto <[e-mail]>
>> ---
>>  drivers/pci/dwc/pci-exynos.c           |  18 --
>>  drivers/pci/dwc/pci-imx6.c             |  18 --
>>  drivers/pci/dwc/pcie-artpec6.c         |  18 --
>>  drivers/pci/dwc/pcie-designware-host.c | 342 +++++++++++++++++----------------
>>  drivers/pci/dwc/pcie-designware-plat.c |  15 --
>>  drivers/pci/dwc/pcie-designware.h      |   6 +-
>>  drivers/pci/dwc/pcie-qcom.c            |  15 --
>>  7 files changed, 179 insertions(+), 253 deletions(-)

[...]

> 
> What do you see the challenge in porting the PCI-Keystone driver to the new
> interrupt API (option c)? I haven't had a chance to review your patches and recent
> changes in Linux kernel for PCI as we are currently using v4.9 kernel for
> our platforms. One key difference in the PCI Core used in Keystone SoC is that
> the MSI is not implemented using standard registers as in newer PCI Designware
> core. It uses application register space for MSI interrupts. Given that
> is the case, do you see any issue in porting the driver to the new API?
> 
> I should be able to test your patch on Keystone.
> 

I don't see any issue in porting it, and I have sent a RFC patch-set in Friday
including the porting of the keystone SoC driver. If you could test it with the
new API it would be great, just to have a sanity check.

A v2 of the RFC will be sent soon, so I would suggest for you to wait until v2
is available, that way you could test with the latest.

Thanks for the help,
Joao

Murali Karicheri May 15, 2017, 2:01 p.m. UTC | #8

On 05/15/2017 07:13 AM, Joao Pinto wrote:
> 
> Hi Murali,
> 
> Às 5:21 PM de 5/12/2017, Murali Karicheri escreveu:
>> On 05/09/2017 08:33 AM, Joao Pinto wrote:
>>> This is a proposal for the update of the interrupt API in pcie-designware.
>>>
>>> *SoC specific drivers*
>>> All SoC specific drivers that use the common infrastructure (pci-qcom,
>>> pci-artpec6, pci-exynos, pci-imx6) were updated to be compatible.
>>> Other SoC specific drivers like pci-dra7, pci-armada8k, pci-hisi, pci-spear13xx
>>> and pci-layerscape will also work fine.
>>>
>>> *Work still to be done - need some inputs*
>>> The pci-keystone driver is the one that I would appreciate some opinions, since
>>> it uses the "old" dw_pcie_msi_chip. I think we have 3 options:
>>>
>>> a) Keep the old API + new API in pcie-designware just for pci-keystone to use
>>> the dw_pcie_msi_chip structure
>>> b) Move the old API from pcie-designware to pci-keystone for it to use the
>>> dw_pcie_msi_chip structure
>>> c) Adapt pci-keystone to use the new API also
>>>
>>> *Tests*
>>> I made tests with a PCI 4.80 Core IP, using pcie-designware-plat driver.
>>> I used an MSI-only Endpoint and a MSI/MSIX Endpoint and the APi adapted very
>>> well to the situation.
>>>
>>> Signed-off-by: Joao Pinto <[e-mail]>
>>> ---
>>>  drivers/pci/dwc/pci-exynos.c           |  18 --
>>>  drivers/pci/dwc/pci-imx6.c             |  18 --
>>>  drivers/pci/dwc/pcie-artpec6.c         |  18 --
>>>  drivers/pci/dwc/pcie-designware-host.c | 342 +++++++++++++++++----------------
>>>  drivers/pci/dwc/pcie-designware-plat.c |  15 --
>>>  drivers/pci/dwc/pcie-designware.h      |   6 +-
>>>  drivers/pci/dwc/pcie-qcom.c            |  15 --
>>>  7 files changed, 179 insertions(+), 253 deletions(-)
> 
> [...]
> 
>>
>> What do you see the challenge in porting the PCI-Keystone driver to the new
>> interrupt API (option c)? I haven't had a chance to review your patches and recent
>> changes in Linux kernel for PCI as we are currently using v4.9 kernel for
>> our platforms. One key difference in the PCI Core used in Keystone SoC is that
>> the MSI is not implemented using standard registers as in newer PCI Designware
>> core. It uses application register space for MSI interrupts. Given that
>> is the case, do you see any issue in porting the driver to the new API?
>>
>> I should be able to test your patch on Keystone.
>>
> 
> I don't see any issue in porting it, and I have sent a RFC patch-set in Friday
> including the porting of the keystone SoC driver. If you could test it with the
> new API it would be great, just to have a sanity check.
> 
> A v2 of the RFC will be sent soon, so I would suggest for you to wait until v2
> is available, that way you could test with the latest.
> 
> Thanks for the help,
> Joao
> 
Ok. Will do

Patch

diff --git a/drivers/pci/dwc/pci-exynos.c b/drivers/pci/dwc/pci-exynos.c
index 546082a..e49c39a 100644
--- a/drivers/pci/dwc/pci-exynos.c
+++ b/drivers/pci/dwc/pci-exynos.c
@@ -490,15 +490,6 @@  static irqreturn_t exynos_pcie_irq_handler(int irq, void *arg)
 	return IRQ_HANDLED;
 }
 
-static irqreturn_t exynos_pcie_msi_irq_handler(int irq, void *arg)
-{
-	struct exynos_pcie *ep = arg;
-	struct dw_pcie *pci = ep->pci;
-	struct pcie_port *pp = &pci->pp;
-
-	return dw_handle_msi_irq(pp);
-}
-
 static void exynos_pcie_msi_init(struct exynos_pcie *ep)
 {
 	struct dw_pcie *pci = ep->pci;
@@ -622,15 +613,6 @@  static int __init exynos_add_pcie_port(struct exynos_pcie *ep,
 			dev_err(dev, "failed to get msi irq\n");
 			return -ENODEV;
 		}
-
-		ret = devm_request_irq(dev, pp->msi_irq,
-					exynos_pcie_msi_irq_handler,
-					IRQF_SHARED | IRQF_NO_THREAD,
-					"exynos-pcie", ep);
-		if (ret) {
-			dev_err(dev, "failed to request msi irq\n");
-			return ret;
-		}
 	}
 
 	pp->root_bus_nr = -1;
diff --git a/drivers/pci/dwc/pci-imx6.c b/drivers/pci/dwc/pci-imx6.c
index a98cba5..f22ce1f 100644
--- a/drivers/pci/dwc/pci-imx6.c
+++ b/drivers/pci/dwc/pci-imx6.c
@@ -490,15 +490,6 @@  static int imx6_pcie_wait_for_speed_change(struct imx6_pcie *imx6_pcie)
 	return -EINVAL;
 }
 
-static irqreturn_t imx6_pcie_msi_handler(int irq, void *arg)
-{
-	struct imx6_pcie *imx6_pcie = arg;
-	struct dw_pcie *pci = imx6_pcie->pci;
-	struct pcie_port *pp = &pci->pp;
-
-	return dw_handle_msi_irq(pp);
-}
-
 static int imx6_pcie_establish_link(struct imx6_pcie *imx6_pcie)
 {
 	struct dw_pcie *pci = imx6_pcie->pci;
@@ -620,15 +611,6 @@  static int imx6_add_pcie_port(struct imx6_pcie *imx6_pcie,
 			dev_err(dev, "failed to get MSI irq\n");
 			return -ENODEV;
 		}
-
-		ret = devm_request_irq(dev, pp->msi_irq,
-				       imx6_pcie_msi_handler,
-				       IRQF_SHARED | IRQF_NO_THREAD,
-				       "mx6-pcie-msi", imx6_pcie);
-		if (ret) {
-			dev_err(dev, "failed to request MSI irq\n");
-			return ret;
-		}
 	}
 
 	pp->root_bus_nr = -1;
diff --git a/drivers/pci/dwc/pcie-artpec6.c b/drivers/pci/dwc/pcie-artpec6.c
index 82a04ac..d81789b 100644
--- a/drivers/pci/dwc/pcie-artpec6.c
+++ b/drivers/pci/dwc/pcie-artpec6.c
@@ -188,15 +188,6 @@  static struct dw_pcie_host_ops artpec6_pcie_host_ops = {
 	.host_init = artpec6_pcie_host_init,
 };
 
-static irqreturn_t artpec6_pcie_msi_handler(int irq, void *arg)
-{
-	struct artpec6_pcie *artpec6_pcie = arg;
-	struct dw_pcie *pci = artpec6_pcie->pci;
-	struct pcie_port *pp = &pci->pp;
-
-	return dw_handle_msi_irq(pp);
-}
-
 static int artpec6_add_pcie_port(struct artpec6_pcie *artpec6_pcie,
 				 struct platform_device *pdev)
 {
@@ -211,15 +202,6 @@  static int artpec6_add_pcie_port(struct artpec6_pcie *artpec6_pcie,
 			dev_err(dev, "failed to get MSI irq\n");
 			return -ENODEV;
 		}
-
-		ret = devm_request_irq(dev, pp->msi_irq,
-				       artpec6_pcie_msi_handler,
-				       IRQF_SHARED | IRQF_NO_THREAD,
-				       "artpec6-pcie-msi", artpec6_pcie);
-		if (ret) {
-			dev_err(dev, "failed to request MSI irq\n");
-			return ret;
-		}
 	}
 
 	pp->root_bus_nr = -1;
diff --git a/drivers/pci/dwc/pcie-designware-host.c b/drivers/pci/dwc/pcie-designware-host.c
index 28ed32b..4b62315 100644
--- a/drivers/pci/dwc/pcie-designware-host.c
+++ b/drivers/pci/dwc/pcie-designware-host.c
@@ -11,6 +11,7 @@ 
  * published by the Free Software Foundation.
  */
 
+#include <linux/irqchip/chained_irq.h>
 #include <linux/irqdomain.h>
 #include <linux/of_address.h>
 #include <linux/of_pci.h>
@@ -45,40 +46,65 @@  static int dw_pcie_wr_own_conf(struct pcie_port *pp, int where, int size,
 	return dw_pcie_write(pci->dbi_base + where, size, val);
 }
 
+static void dw_msi_mask_irq(struct irq_data *d) {
+	pci_msi_mask_irq(d);
+	irq_chip_mask_parent(d);
+}
+
+static void dw_msi_unmask_irq(struct irq_data *d) {
+	pci_msi_unmask_irq(d);
+	irq_chip_unmask_parent(d);
+}
+
 static struct irq_chip dw_msi_irq_chip = {
 	.name = "PCI-MSI",
-	.irq_enable = pci_msi_unmask_irq,
-	.irq_disable = pci_msi_mask_irq,
-	.irq_mask = pci_msi_mask_irq,
-	.irq_unmask = pci_msi_unmask_irq,
+	.irq_mask = dw_msi_mask_irq,
+	.irq_unmask = dw_msi_unmask_irq,
+};
+
+static struct msi_domain_info dw_pcie_msi_domain_info = {
+	.flags	= (MSI_FLAG_USE_DEF_DOM_OPS | MSI_FLAG_USE_DEF_CHIP_OPS |
+		     MSI_FLAG_PCI_MSIX),
+	.chip	= &dw_msi_irq_chip,
 };
 
-/* MSI int handler */
-irqreturn_t dw_handle_msi_irq(struct pcie_port *pp)
+void dw_handle_msi_irq(struct pcie_port *pp)
 {
-	u32 val;
-	int i, pos, irq;
-	irqreturn_t ret = IRQ_NONE;
+	int i, pos, virq;
+	u32 status;
 
 	for (i = 0; i < MAX_MSI_CTRLS; i++) {
 		dw_pcie_rd_own_conf(pp, PCIE_MSI_INTR0_STATUS + i * 12, 4,
-				    &val);
-		if (!val)
+				    &status);
+		if (!status)
 			continue;
 
-		ret = IRQ_HANDLED;
 		pos = 0;
-		while ((pos = find_next_bit((unsigned long *) &val, 32,
+		while ((pos = find_next_bit((unsigned long *) &status, 32,
 					    pos)) != 32) {
-			irq = irq_find_mapping(pp->irq_domain, i * 32 + pos);
+			virq = irq_find_mapping(pp->irq_domain, i * 32 + pos);
 			dw_pcie_wr_own_conf(pp, PCIE_MSI_INTR0_STATUS + i * 12,
 					    4, 1 << pos);
-			generic_handle_irq(irq);
+			generic_handle_irq(virq);
 			pos++;
 		}
 	}
+}
 
-	return ret;
+/* Chained MSI interrupt service routine */
+static void dw_chained_msi_isr(struct irq_desc *desc)
+{
+	struct irq_chip *chip = irq_desc_get_chip(desc);
+	struct pcie_port *pp;
+	struct dw_pcie *pci;
+
+	chained_irq_enter(chip, desc);
+	pci = irq_desc_get_handler_data(desc);
+	pp = &pci->pp;
+
+	dw_handle_msi_irq(pp);
+
+	chained_irq_exit(chip, desc);
 }
 
 void dw_pcie_msi_init(struct pcie_port *pp)
@@ -95,183 +121,163 @@  void dw_pcie_msi_init(struct pcie_port *pp)
 			    (u32)(msi_target >> 32 & 0xffffffff));
 }
 
-static void dw_pcie_msi_clear_irq(struct pcie_port *pp, int irq)
+static void dw_pci_setup_msi_msg(struct irq_data *data, struct msi_msg *msg)
 {
-	unsigned int res, bit, val;
+	struct dw_pcie *pci = irq_data_get_irq_chip_data(data);
+	struct pcie_port *pp = &pci->pp;
+	u64 msi_target;
 
-	res = (irq / 32) * 12;
-	bit = irq % 32;
-	dw_pcie_rd_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, &val);
-	val &= ~(1 << bit);
-	dw_pcie_wr_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, val);
-}
+	if (pp->ops->get_msi_addr)
+		msi_target = pp->ops->get_msi_addr(pp);
+	else
+		msi_target = virt_to_phys((void *)pp->msi_data);
 
-static void clear_irq_range(struct pcie_port *pp, unsigned int irq_base,
-			    unsigned int nvec, unsigned int pos)
-{
-	unsigned int i;
-
-	for (i = 0; i < nvec; i++) {
-		irq_set_msi_desc_off(irq_base, i, NULL);
-		/* Disable corresponding interrupt on MSI controller */
-		if (pp->ops->msi_clear_irq)
-			pp->ops->msi_clear_irq(pp, pos + i);
-		else
-			dw_pcie_msi_clear_irq(pp, pos + i);
-	}
+	msg->address_lo = (u32)(msi_target & 0xffffffff);
+	msg->address_hi = (u32)(msi_target >> 32 & 0xffffffff);
 
-	bitmap_release_region(pp->msi_irq_in_use, pos, order_base_2(nvec));
+	if (pp->ops->get_msi_data)
+		msg->data = pp->ops->get_msi_data(pp, data->hwirq);
+	else
+		msg->data = data->hwirq;
+
+	dev_err(pci->dev, "msi#%d address_hi %#x address_lo %#x\n",
+		(int)data->hwirq, msg->address_hi, msg->address_lo);
 }
 
-static void dw_pcie_msi_set_irq(struct pcie_port *pp, int irq)
+static int dw_pci_msi_set_affinity(struct irq_data *irq_data,
+				   const struct cpumask *mask, bool force)
 {
-	unsigned int res, bit, val;
-
-	res = (irq / 32) * 12;
-	bit = irq % 32;
-	dw_pcie_rd_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, &val);
-	val |= 1 << bit;
-	dw_pcie_wr_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, val);
+	return -EINVAL;
 }
 
-static int assign_irq(int no_irqs, struct msi_desc *desc, int *pos)
+static void dw_pci_bottom_mask(struct irq_data *data)
 {
-	int irq, pos0, i;
-	struct pcie_port *pp;
-
-	pp = (struct pcie_port *)msi_desc_to_pci_sysdata(desc);
-	pos0 = bitmap_find_free_region(pp->msi_irq_in_use, MAX_MSI_IRQS,
-				       order_base_2(no_irqs));
-	if (pos0 < 0)
-		goto no_valid_irq;
-
-	irq = irq_find_mapping(pp->irq_domain, pos0);
-	if (!irq)
-		goto no_valid_irq;
-
-	/*
-	 * irq_create_mapping (called from dw_pcie_host_init) pre-allocates
-	 * descs so there is no need to allocate descs here. We can therefore
-	 * assume that if irq_find_mapping above returns non-zero, then the
-	 * descs are also successfully allocated.
-	 */
+	struct dw_pcie *pci = irq_data_get_irq_chip_data(data);
+	struct pcie_port *pp = &pci->pp;
+	unsigned int res, bit, val;
 
-	for (i = 0; i < no_irqs; i++) {
-		if (irq_set_msi_desc_off(irq, i, desc) != 0) {
-			clear_irq_range(pp, irq, i, pos0);
-			goto no_valid_irq;
-		}
-		/*Enable corresponding interrupt in MSI interrupt controller */
-		if (pp->ops->msi_set_irq)
-			pp->ops->msi_set_irq(pp, pos0 + i);
-		else
-			dw_pcie_msi_set_irq(pp, pos0 + i);
-	}
+	mutex_lock(&pp->lock);
 
-	*pos = pos0;
-	desc->nvec_used = no_irqs;
-	desc->msi_attrib.multiple = order_base_2(no_irqs);
+	res = (data->hwirq / 32) * 12;
+	bit = data->hwirq % 32;
 
-	return irq;
+	dw_pcie_rd_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, &val);
+	val &= ~(1 << bit);
+	dw_pcie_wr_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, val);
 
-no_valid_irq:
-	*pos = pos0;
-	return -ENOSPC;
+	mutex_unlock(&pp->lock);
 }
 
-static void dw_msi_setup_msg(struct pcie_port *pp, unsigned int irq, u32 pos)
+static void dw_pci_bottom_unmask(struct irq_data *data)
 {
-	struct msi_msg msg;
-	u64 msi_target;
+	struct dw_pcie *pci = irq_data_get_irq_chip_data(data);
+	struct pcie_port *pp = &pci->pp;
+	unsigned int res, bit, val;
 
-	if (pp->ops->get_msi_addr)
-		msi_target = pp->ops->get_msi_addr(pp);
-	else
-		msi_target = virt_to_phys((void *)pp->msi_data);
+	mutex_lock(&pp->lock);
 
-	msg.address_lo = (u32)(msi_target & 0xffffffff);
-	msg.address_hi = (u32)(msi_target >> 32 & 0xffffffff);
+	res = (data->hwirq / 32) * 12;
+	bit = data->hwirq % 32;
 
-	if (pp->ops->get_msi_data)
-		msg.data = pp->ops->get_msi_data(pp, pos);
-	else
-		msg.data = pos;
+	dw_pcie_rd_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, &val);
+	val |= 1 << bit;
+	dw_pcie_wr_own_conf(pp, PCIE_MSI_INTR0_ENABLE + res, 4, val);
 
-	pci_write_msi_msg(irq, &msg);
+	mutex_unlock(&pp->lock);
 }
 
-static int dw_msi_setup_irq(struct msi_controller *chip, struct pci_dev *pdev,
-			    struct msi_desc *desc)
-{
-	int irq, pos;
-	struct pcie_port *pp = pdev->bus->sysdata;
-
-	if (desc->msi_attrib.is_msix)
-		return -EINVAL;
-
-	irq = assign_irq(1, desc, &pos);
-	if (irq < 0)
-		return irq;
+static struct irq_chip dw_pci_msi_bottom_irq_chip = {
+	.name			= "DW MSI",
+	.irq_compose_msi_msg	= dw_pci_setup_msi_msg,
+	.irq_set_affinity	= dw_pci_msi_set_affinity,
+	.irq_mask		= dw_pci_bottom_mask,
+	.irq_unmask		= dw_pci_bottom_unmask,
+};
 
-	dw_msi_setup_msg(pp, irq, pos);
+static int dw_pcie_irq_domain_alloc(struct irq_domain *domain,
+				    unsigned int virq, unsigned int nr_irqs,
+				    void *args)
+{
+	struct dw_pcie *pci = domain->host_data;
+	struct pcie_port *pp = &pci->pp;
+	unsigned long bit;
 
-	return 0;
-}
+	mutex_lock(&pp->lock);
 
-static int dw_msi_setup_irqs(struct msi_controller *chip, struct pci_dev *pdev,
-			     int nvec, int type)
-{
-#ifdef CONFIG_PCI_MSI
-	int irq, pos;
-	struct msi_desc *desc;
-	struct pcie_port *pp = pdev->bus->sysdata;
+	WARN_ON(nr_irqs != 1);
 
-	/* MSI-X interrupts are not supported */
-	if (type == PCI_CAP_ID_MSIX)
-		return -EINVAL;
+	bit = find_first_zero_bit(pp->msi_irq_in_use, pp->num_vectors);
+	if (bit >= pp->num_vectors) {
+		mutex_unlock(&pp->lock);
+		return -ENOSPC;
+	}
 
-	WARN_ON(!list_is_singular(&pdev->dev.msi_list));
-	desc = list_entry(pdev->dev.msi_list.next, struct msi_desc, list);
+	set_bit(bit, pp->msi_irq_in_use);
 
-	irq = assign_irq(nvec, desc, &pos);
-	if (irq < 0)
-		return irq;
+	mutex_unlock(&pp->lock);
 
-	dw_msi_setup_msg(pp, irq, pos);
+	irq_domain_set_info(domain, virq, bit, &dw_pci_msi_bottom_irq_chip,
+			    domain->host_data, handle_simple_irq,
+			    NULL, NULL);
 
 	return 0;
-#else
-	return -EINVAL;
-#endif
 }
 
-static void dw_msi_teardown_irq(struct msi_controller *chip, unsigned int irq)
+static void dw_pcie_irq_domain_free(struct irq_domain *domain,
+				    unsigned int virq, unsigned int nr_irqs)
 {
-	struct irq_data *data = irq_get_irq_data(irq);
-	struct msi_desc *msi = irq_data_get_msi_desc(data);
-	struct pcie_port *pp = (struct pcie_port *)msi_desc_to_pci_sysdata(msi);
+	struct irq_data *data = irq_domain_get_irq_data(domain, virq);
+	struct dw_pcie *pci = irq_data_get_irq_chip_data(data);
+	struct pcie_port *pp = &pci->pp;
+
+	mutex_lock(&pp->lock);
 
-	clear_irq_range(pp, irq, 1, data->hwirq);
+	if (!test_bit(data->hwirq, pp->msi_irq_in_use))
+		dev_err(pci->dev, "trying to free unused MSI#%lu\n",
+			data->hwirq);
+	else
+		__clear_bit(data->hwirq, pp->msi_irq_in_use);
+
+	mutex_unlock(&pp->lock);
 }
 
-static struct msi_controller dw_pcie_msi_chip = {
-	.setup_irq = dw_msi_setup_irq,
-	.setup_irqs = dw_msi_setup_irqs,
-	.teardown_irq = dw_msi_teardown_irq,
+static const struct irq_domain_ops dw_pcie_msi_domain_ops = {
+	.alloc	= dw_pcie_irq_domain_alloc,
+	.free	= dw_pcie_irq_domain_free,
 };
 
-static int dw_pcie_msi_map(struct irq_domain *domain, unsigned int irq,
-			   irq_hw_number_t hwirq)
+static int dw_pcie_allocate_domains(struct dw_pcie *pci)
 {
-	irq_set_chip_and_handler(irq, &dw_msi_irq_chip, handle_simple_irq);
-	irq_set_chip_data(irq, domain->host_data);
+	struct pcie_port *pp = &pci->pp;
+	struct fwnode_handle *fwnode = of_node_to_fwnode(pci->dev->of_node);
+
+	pp->irq_domain = irq_domain_add_linear(NULL, pp->num_vectors,
+					       &dw_pcie_msi_domain_ops, pci);
+	if (!pp->irq_domain) {
+		dev_err(pci->dev, "failed to create IRQ domain\n");
+		return -ENOMEM;
+	}
+
+	pp->msi_domain = pci_msi_create_irq_domain(fwnode,
+						   &dw_pcie_msi_domain_info,
+						   pp->irq_domain);
+	if (!pp->msi_domain) {
+		dev_err(pci->dev, "failed to create MSI domain\n");
+		irq_domain_remove(pp->irq_domain);
+		return -ENOMEM;
+	}
 
 	return 0;
 }
 
-static const struct irq_domain_ops msi_domain_ops = {
-	.map = dw_pcie_msi_map,
-};
+void dw_pcie_free_msi(struct pcie_port *pp)
+{
+	irq_set_chained_handler(pp->msi_irq, NULL);
+	irq_set_handler_data(pp->msi_irq, NULL);
+
+	irq_domain_remove(pp->msi_domain);
+	irq_domain_remove(pp->irq_domain);
+}
 
 int dw_pcie_host_init(struct pcie_port *pp)
 {
@@ -279,11 +285,13 @@  int dw_pcie_host_init(struct pcie_port *pp)
 	struct device *dev = pci->dev;
 	struct device_node *np = dev->of_node;
 	struct platform_device *pdev = to_platform_device(dev);
+	struct resource_entry *win, *tmp;
 	struct pci_bus *bus, *child;
 	struct resource *cfg_res;
-	int i, ret;
+	int ret;
 	LIST_HEAD(res);
-	struct resource_entry *win, *tmp;
+
+	mutex_init(&pci->pp.lock);
 
 	cfg_res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "config");
 	if (cfg_res) {
@@ -377,20 +385,22 @@  int dw_pcie_host_init(struct pcie_port *pp)
 		pci->num_viewport = 2;
 
 	if (IS_ENABLED(CONFIG_PCI_MSI)) {
+		ret = of_property_read_u32(np, "num-vectors",
+					   &pp->num_vectors);
+		if (ret)
+			pp->num_vectors = 1;
+
 		if (!pp->ops->msi_host_init) {
-			pp->irq_domain = irq_domain_add_linear(dev->of_node,
-						MAX_MSI_IRQS, &msi_domain_ops,
-						&dw_pcie_msi_chip);
-			if (!pp->irq_domain) {
-				dev_err(dev, "irq domain init failed\n");
-				ret = -ENXIO;
+			ret = dw_pcie_allocate_domains(pci);
+			if (ret)
 				goto error;
-			}
 
-			for (i = 0; i < MAX_MSI_IRQS; i++)
-				irq_create_mapping(pp->irq_domain, i);
+			irq_set_chained_handler_and_data(pci->pp.msi_irq,
+							 dw_chained_msi_isr,
+							 pci);
 		} else {
-			ret = pp->ops->msi_host_init(pp, &dw_pcie_msi_chip);
+			//TODO
+			ret = 0;//pp->ops->msi_host_init(pp, &dw_pcie_msi_chip);
 			if (ret < 0)
 				goto error;
 		}
@@ -400,14 +410,10 @@  int dw_pcie_host_init(struct pcie_port *pp)
 		pp->ops->host_init(pp);
 
 	pp->root_bus_nr = pp->busn->start;
-	if (IS_ENABLED(CONFIG_PCI_MSI)) {
-		bus = pci_scan_root_bus_msi(dev, pp->root_bus_nr,
-					    &dw_pcie_ops, pp, &res,
-					    &dw_pcie_msi_chip);
-		dw_pcie_msi_chip.dev = dev;
-	} else
-		bus = pci_scan_root_bus(dev, pp->root_bus_nr, &dw_pcie_ops,
-					pp, &res);
+
+	bus = pci_scan_root_bus(dev, pp->root_bus_nr, &dw_pcie_ops,
+				pp, &res);
+
 	if (!bus) {
 		ret = -ENOMEM;
 		goto error;
diff --git a/drivers/pci/dwc/pcie-designware-plat.c b/drivers/pci/dwc/pcie-designware-plat.c
index 32091b3..3b9717b 100644
--- a/drivers/pci/dwc/pcie-designware-plat.c
+++ b/drivers/pci/dwc/pcie-designware-plat.c
@@ -28,13 +28,6 @@  struct dw_plat_pcie {
 	struct dw_pcie		*pci;
 };
 
-static irqreturn_t dw_plat_pcie_msi_irq_handler(int irq, void *arg)
-{
-	struct pcie_port *pp = arg;
-
-	return dw_handle_msi_irq(pp);
-}
-
 static void dw_plat_pcie_host_init(struct pcie_port *pp)
 {
 	struct dw_pcie *pci = to_dw_pcie_from_pp(pp);
@@ -64,14 +57,6 @@  static int dw_plat_add_pcie_port(struct pcie_port *pp,
 		pp->msi_irq = platform_get_irq(pdev, 0);
 		if (pp->msi_irq < 0)
 			return pp->msi_irq;
-
-		ret = devm_request_irq(dev, pp->msi_irq,
-					dw_plat_pcie_msi_irq_handler,
-					IRQF_SHARED, "dw-plat-pcie-msi", pp);
-		if (ret) {
-			dev_err(dev, "failed to request MSI IRQ\n");
-			return ret;
-		}
 	}
 
 	pp->root_bus_nr = -1;
diff --git a/drivers/pci/dwc/pcie-designware.h b/drivers/pci/dwc/pcie-designware.h
index c6a8405..3abd229 100644
--- a/drivers/pci/dwc/pcie-designware.h
+++ b/drivers/pci/dwc/pcie-designware.h
@@ -165,7 +165,10 @@  struct pcie_port {
 	struct dw_pcie_host_ops	*ops;
 	int			msi_irq;
 	struct irq_domain	*irq_domain;
+	struct irq_domain	*msi_domain;
 	unsigned long		msi_data;
+	u32			num_vectors;
+	struct mutex		lock;
 	DECLARE_BITMAP(msi_irq_in_use, MAX_MSI_IRQS);
 };
 
@@ -280,7 +283,8 @@  static inline u32 dw_pcie_readl_dbi2(struct dw_pcie *pci, u32 reg)
 }
 
 #ifdef CONFIG_PCIE_DW_HOST
-irqreturn_t dw_handle_msi_irq(struct pcie_port *pp);
+void dw_handle_msi_irq(struct pcie_port *pp);
+void dw_pcie_free_msi(struct pcie_port *pp);
 void dw_pcie_msi_init(struct pcie_port *pp);
 void dw_pcie_setup_rc(struct pcie_port *pp);
 int dw_pcie_host_init(struct pcie_port *pp);
diff --git a/drivers/pci/dwc/pcie-qcom.c b/drivers/pci/dwc/pcie-qcom.c
index 5bf23d4..4e1cc02 100644
--- a/drivers/pci/dwc/pcie-qcom.c
+++ b/drivers/pci/dwc/pcie-qcom.c
@@ -126,13 +126,6 @@  static void qcom_ep_reset_deassert(struct qcom_pcie *pcie)
 	usleep_range(PERST_DELAY_US, PERST_DELAY_US + 500);
 }
 
-static irqreturn_t qcom_pcie_msi_irq_handler(int irq, void *arg)
-{
-	struct pcie_port *pp = arg;
-
-	return dw_handle_msi_irq(pp);
-}
-
 static void qcom_pcie_v0_v1_ltssm_enable(struct qcom_pcie *pcie)
 {
 	u32 val;
@@ -724,14 +717,6 @@  static int qcom_pcie_probe(struct platform_device *pdev)
 		pp->msi_irq = platform_get_irq_byname(pdev, "msi");
 		if (pp->msi_irq < 0)
 			return pp->msi_irq;
-
-		ret = devm_request_irq(dev, pp->msi_irq,
-				       qcom_pcie_msi_irq_handler,
-				       IRQF_SHARED, "qcom-pcie-msi", pp);
-		if (ret) {
-			dev_err(dev, "cannot request msi irq\n");
-			return ret;
-		}
 	}
 
 	ret = phy_init(pcie->phy);