linux-arm-kernel.lists.infradead.org archive mirror
 help / color / mirror / Atom feed
* [PATCH v2 0/3] PCI: mediatek: Allocate MSI address with dmam_alloc_coherent()
@ 2023-12-11  8:52 Jianjun Wang
  2023-12-11  8:52 ` [PATCH v2 1/3] " Jianjun Wang
                   ` (2 more replies)
  0 siblings, 3 replies; 7+ messages in thread
From: Jianjun Wang @ 2023-12-11  8:52 UTC (permalink / raw)
  To: Lorenzo Pieralisi, Krzysztof Wilczyński, Rob Herring,
	Bjorn Helgaas, Matthias Brugger, AngeloGioacchino Del Regno,
	Marc Zyngier
  Cc: Ryder Lee, Jianjun Wang, linux-pci, linux-mediatek, linux-kernel,
	linux-arm-kernel, jieyy.yang, chuanjia.liu, qizhong.cheng,
	jian.yang, jianguo.zhang

These series of patches change the allocation of MSI address on MediaTek
platform, which uses dmam_alloc_coherent() to allocate the MSI address.

Changes in v2:
1. Add a separate patch for pcie-mediatek-gen3.c to prevent break the
   probe flow when MSI init fails.
2. Change the location of allocate the MSI address to prevent break the
   probe flow.

Jianjun Wang (3):
  PCI: mediatek: Allocate MSI address with dmam_alloc_coherent()
  PCI: mediatek-gen3: Do not break probe flow when MSI init fails
  PCI: mediatek-gen3: Allocate MSI address with dmam_alloc_coherent()

 drivers/pci/controller/pcie-mediatek-gen3.c | 126 +++++++++++---------
 drivers/pci/controller/pcie-mediatek.c      |  24 ++--
 2 files changed, 83 insertions(+), 67 deletions(-)

-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

^ permalink raw reply	[flat|nested] 7+ messages in thread

* [PATCH v2 1/3] PCI: mediatek: Allocate MSI address with dmam_alloc_coherent()
  2023-12-11  8:52 [PATCH v2 0/3] PCI: mediatek: Allocate MSI address with dmam_alloc_coherent() Jianjun Wang
@ 2023-12-11  8:52 ` Jianjun Wang
  2024-06-08  9:01   ` Manivannan Sadhasivam
  2023-12-11  8:52 ` [PATCH v2 2/3] PCI: mediatek-gen3: Do not break probe flow when MSI init fails Jianjun Wang
  2023-12-11  8:52 ` [PATCH v2 3/3] PCI: mediatek-gen3: Allocate MSI address with dmam_alloc_coherent() Jianjun Wang
  2 siblings, 1 reply; 7+ messages in thread
From: Jianjun Wang @ 2023-12-11  8:52 UTC (permalink / raw)
  To: Lorenzo Pieralisi, Krzysztof Wilczyński, Rob Herring,
	Bjorn Helgaas, Matthias Brugger, AngeloGioacchino Del Regno,
	Marc Zyngier
  Cc: Ryder Lee, Jianjun Wang, linux-pci, linux-mediatek, linux-kernel,
	linux-arm-kernel, jieyy.yang, chuanjia.liu, qizhong.cheng,
	jian.yang, jianguo.zhang

Use dmam_alloc_coherent() to allocate the MSI address, instead of using
virt_to_phys().

Signed-off-by: Jianjun Wang <jianjun.wang@mediatek.com>
---
 drivers/pci/controller/pcie-mediatek.c | 24 ++++++++++++++++--------
 1 file changed, 16 insertions(+), 8 deletions(-)

diff --git a/drivers/pci/controller/pcie-mediatek.c b/drivers/pci/controller/pcie-mediatek.c
index 66a8f73296fc..2fb9e44369f8 100644
--- a/drivers/pci/controller/pcie-mediatek.c
+++ b/drivers/pci/controller/pcie-mediatek.c
@@ -178,6 +178,7 @@ struct mtk_pcie_soc {
  * @phy: pointer to PHY control block
  * @slot: port slot
  * @irq: GIC irq
+ * @msg_addr: MSI message address
  * @irq_domain: legacy INTx IRQ domain
  * @inner_domain: inner IRQ domain
  * @msi_domain: MSI IRQ domain
@@ -198,6 +199,7 @@ struct mtk_pcie_port {
 	struct phy *phy;
 	u32 slot;
 	int irq;
+	dma_addr_t msg_addr;
 	struct irq_domain *irq_domain;
 	struct irq_domain *inner_domain;
 	struct irq_domain *msi_domain;
@@ -394,12 +396,10 @@ static struct pci_ops mtk_pcie_ops_v2 = {
 static void mtk_compose_msi_msg(struct irq_data *data, struct msi_msg *msg)
 {
 	struct mtk_pcie_port *port = irq_data_get_irq_chip_data(data);
-	phys_addr_t addr;
 
 	/* MT2712/MT7622 only support 32-bit MSI addresses */
-	addr = virt_to_phys(port->base + PCIE_MSI_VECTOR);
 	msg->address_hi = 0;
-	msg->address_lo = lower_32_bits(addr);
+	msg->address_lo = lower_32_bits(port->msg_addr);
 
 	msg->data = data->hwirq;
 
@@ -494,6 +494,14 @@ static struct msi_domain_info mtk_msi_domain_info = {
 static int mtk_pcie_allocate_msi_domains(struct mtk_pcie_port *port)
 {
 	struct fwnode_handle *fwnode = of_node_to_fwnode(port->pcie->dev->of_node);
+	void *msi_vaddr;
+
+	msi_vaddr = dmam_alloc_coherent(port->pcie->dev, sizeof(dma_addr_t), &port->msg_addr,
+					GFP_KERNEL);
+	if (!msi_vaddr) {
+		dev_err(port->pcie->dev, "failed to alloc and map MSI address\n");
+		return -ENOMEM;
+	}
 
 	mutex_init(&port->lock);
 
@@ -501,6 +509,7 @@ static int mtk_pcie_allocate_msi_domains(struct mtk_pcie_port *port)
 						      &msi_domain_ops, port);
 	if (!port->inner_domain) {
 		dev_err(port->pcie->dev, "failed to create IRQ domain\n");
+		dmam_free_coherent(port->pcie->dev, sizeof(dma_addr_t), msi_vaddr, port->msg_addr);
 		return -ENOMEM;
 	}
 
@@ -508,6 +517,7 @@ static int mtk_pcie_allocate_msi_domains(struct mtk_pcie_port *port)
 						     port->inner_domain);
 	if (!port->msi_domain) {
 		dev_err(port->pcie->dev, "failed to create MSI domain\n");
+		dmam_free_coherent(port->pcie->dev, sizeof(dma_addr_t), msi_vaddr, port->msg_addr);
 		irq_domain_remove(port->inner_domain);
 		return -ENOMEM;
 	}
@@ -518,10 +528,8 @@ static int mtk_pcie_allocate_msi_domains(struct mtk_pcie_port *port)
 static void mtk_pcie_enable_msi(struct mtk_pcie_port *port)
 {
 	u32 val;
-	phys_addr_t msg_addr;
 
-	msg_addr = virt_to_phys(port->base + PCIE_MSI_VECTOR);
-	val = lower_32_bits(msg_addr);
+	val = lower_32_bits(port->msg_addr);
 	writel(val, port->base + PCIE_IMSI_ADDR);
 
 	val = readl(port->base + PCIE_INT_MASK);
@@ -588,7 +596,7 @@ static int mtk_pcie_init_irq_domain(struct mtk_pcie_port *port,
 	if (IS_ENABLED(CONFIG_PCI_MSI)) {
 		ret = mtk_pcie_allocate_msi_domains(port);
 		if (ret)
-			return ret;
+			dev_warn(dev, "no MSI supported, only INTx available\n");
 	}
 
 	return 0;
@@ -732,7 +740,7 @@ static int mtk_pcie_startup_port_v2(struct mtk_pcie_port *port)
 	val &= ~INTX_MASK;
 	writel(val, port->base + PCIE_INT_MASK);
 
-	if (IS_ENABLED(CONFIG_PCI_MSI))
+	if (IS_ENABLED(CONFIG_PCI_MSI) && port->msi_domain)
 		mtk_pcie_enable_msi(port);
 
 	/* Set AHB to PCIe translation windows */
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

^ permalink raw reply related	[flat|nested] 7+ messages in thread

* [PATCH v2 2/3] PCI: mediatek-gen3: Do not break probe flow when MSI init fails
  2023-12-11  8:52 [PATCH v2 0/3] PCI: mediatek: Allocate MSI address with dmam_alloc_coherent() Jianjun Wang
  2023-12-11  8:52 ` [PATCH v2 1/3] " Jianjun Wang
@ 2023-12-11  8:52 ` Jianjun Wang
  2023-12-11  8:52 ` [PATCH v2 3/3] PCI: mediatek-gen3: Allocate MSI address with dmam_alloc_coherent() Jianjun Wang
  2 siblings, 0 replies; 7+ messages in thread
From: Jianjun Wang @ 2023-12-11  8:52 UTC (permalink / raw)
  To: Lorenzo Pieralisi, Krzysztof Wilczyński, Rob Herring,
	Bjorn Helgaas, Matthias Brugger, AngeloGioacchino Del Regno,
	Marc Zyngier
  Cc: Ryder Lee, Jianjun Wang, linux-pci, linux-mediatek, linux-kernel,
	linux-arm-kernel, jieyy.yang, chuanjia.liu, qizhong.cheng,
	jian.yang, jianguo.zhang

Since INTx can still work, the driver probe flow should not be broken by
MSI initialization failures. Additionally, moving the MSI initialization
code into a single function enhances readability.

Fixes: 1bdafba538be ("PCI: mediatek-gen3: Add MSI support")
Signed-off-by: Jianjun Wang <jianjun.wang@mediatek.com>
---
 drivers/pci/controller/pcie-mediatek-gen3.c | 68 ++++++++++-----------
 1 file changed, 33 insertions(+), 35 deletions(-)

diff --git a/drivers/pci/controller/pcie-mediatek-gen3.c b/drivers/pci/controller/pcie-mediatek-gen3.c
index c1ae3d19ec9a..c6a6876d233a 100644
--- a/drivers/pci/controller/pcie-mediatek-gen3.c
+++ b/drivers/pci/controller/pcie-mediatek-gen3.c
@@ -649,59 +649,51 @@ static const struct irq_domain_ops intx_domain_ops = {
 	.map = mtk_pcie_intx_map,
 };
 
-static int mtk_pcie_init_irq_domains(struct mtk_gen3_pcie *pcie)
+static int mtk_pcie_init_msi(struct mtk_gen3_pcie *pcie)
 {
 	struct device *dev = pcie->dev;
-	struct device_node *intc_node, *node = dev->of_node;
-	int ret;
-
-	raw_spin_lock_init(&pcie->irq_lock);
-
-	/* Setup INTx */
-	intc_node = of_get_child_by_name(node, "interrupt-controller");
-	if (!intc_node) {
-		dev_err(dev, "missing interrupt-controller node\n");
-		return -ENODEV;
-	}
-
-	pcie->intx_domain = irq_domain_add_linear(intc_node, PCI_NUM_INTX,
-						  &intx_domain_ops, pcie);
-	if (!pcie->intx_domain) {
-		dev_err(dev, "failed to create INTx IRQ domain\n");
-		ret = -ENODEV;
-		goto out_put_node;
-	}
+	struct device_node *node = dev->of_node;
 
-	/* Setup MSI */
 	mutex_init(&pcie->lock);
 
 	pcie->msi_bottom_domain = irq_domain_add_linear(node, PCIE_MSI_IRQS_NUM,
 				  &mtk_msi_bottom_domain_ops, pcie);
 	if (!pcie->msi_bottom_domain) {
 		dev_err(dev, "failed to create MSI bottom domain\n");
-		ret = -ENODEV;
-		goto err_msi_bottom_domain;
+		return -ENODEV;
 	}
 
-	pcie->msi_domain = pci_msi_create_irq_domain(dev->fwnode,
-						     &mtk_msi_domain_info,
+	pcie->msi_domain = pci_msi_create_irq_domain(dev->fwnode, &mtk_msi_domain_info,
 						     pcie->msi_bottom_domain);
 	if (!pcie->msi_domain) {
 		dev_err(dev, "failed to create MSI domain\n");
-		ret = -ENODEV;
-		goto err_msi_domain;
+		irq_domain_remove(pcie->msi_bottom_domain);
+		return -ENODEV;
 	}
 
-	of_node_put(intc_node);
 	return 0;
+}
 
-err_msi_domain:
-	irq_domain_remove(pcie->msi_bottom_domain);
-err_msi_bottom_domain:
-	irq_domain_remove(pcie->intx_domain);
-out_put_node:
+static int mtk_pcie_init_intx(struct mtk_gen3_pcie *pcie)
+{
+	struct device *dev = pcie->dev;
+	struct device_node *intc_node, *node = dev->of_node;
+
+	intc_node = of_get_child_by_name(node, "interrupt-controller");
+	if (!intc_node) {
+		dev_err(dev, "missing interrupt-controller node\n");
+		return -ENODEV;
+	}
+
+	pcie->intx_domain = irq_domain_add_linear(intc_node, PCI_NUM_INTX,
+						  &intx_domain_ops, pcie);
 	of_node_put(intc_node);
-	return ret;
+	if (!pcie->intx_domain) {
+		dev_err(dev, "failed to create INTx IRQ domain\n");
+		return -ENODEV;
+	}
+
+	return 0;
 }
 
 static void mtk_pcie_irq_teardown(struct mtk_gen3_pcie *pcie)
@@ -774,10 +766,16 @@ static int mtk_pcie_setup_irq(struct mtk_gen3_pcie *pcie)
 	struct platform_device *pdev = to_platform_device(dev);
 	int err;
 
-	err = mtk_pcie_init_irq_domains(pcie);
+	raw_spin_lock_init(&pcie->irq_lock);
+
+	err = mtk_pcie_init_intx(pcie);
 	if (err)
 		return err;
 
+	err = mtk_pcie_init_msi(pcie);
+	if (err)
+		dev_warn(dev, "no MSI supported, only INTx available\n");
+
 	pcie->irq = platform_get_irq(pdev, 0);
 	if (pcie->irq < 0)
 		return pcie->irq;
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

^ permalink raw reply related	[flat|nested] 7+ messages in thread

* [PATCH v2 3/3] PCI: mediatek-gen3: Allocate MSI address with dmam_alloc_coherent()
  2023-12-11  8:52 [PATCH v2 0/3] PCI: mediatek: Allocate MSI address with dmam_alloc_coherent() Jianjun Wang
  2023-12-11  8:52 ` [PATCH v2 1/3] " Jianjun Wang
  2023-12-11  8:52 ` [PATCH v2 2/3] PCI: mediatek-gen3: Do not break probe flow when MSI init fails Jianjun Wang
@ 2023-12-11  8:52 ` Jianjun Wang
  2 siblings, 0 replies; 7+ messages in thread
From: Jianjun Wang @ 2023-12-11  8:52 UTC (permalink / raw)
  To: Lorenzo Pieralisi, Krzysztof Wilczyński, Rob Herring,
	Bjorn Helgaas, Matthias Brugger, AngeloGioacchino Del Regno,
	Marc Zyngier
  Cc: Ryder Lee, Jianjun Wang, linux-pci, linux-mediatek, linux-kernel,
	linux-arm-kernel, jieyy.yang, chuanjia.liu, qizhong.cheng,
	jian.yang, jianguo.zhang

Use dmam_alloc_coherent() to allocate the MSI address, instead of using
static physical address.

Signed-off-by: Jianjun Wang <jianjun.wang@mediatek.com>
---
 drivers/pci/controller/pcie-mediatek-gen3.c | 72 ++++++++++++---------
 1 file changed, 41 insertions(+), 31 deletions(-)

diff --git a/drivers/pci/controller/pcie-mediatek-gen3.c b/drivers/pci/controller/pcie-mediatek-gen3.c
index c6a6876d233a..7cfd7ef9ad95 100644
--- a/drivers/pci/controller/pcie-mediatek-gen3.c
+++ b/drivers/pci/controller/pcie-mediatek-gen3.c
@@ -120,7 +120,6 @@ struct mtk_msi_set {
  * struct mtk_gen3_pcie - PCIe port information
  * @dev: pointer to PCIe device
  * @base: IO mapped register base
- * @reg_base: physical register base
  * @mac_reset: MAC reset control
  * @phy_reset: PHY reset control
  * @phy: PHY controller block
@@ -139,7 +138,6 @@ struct mtk_msi_set {
 struct mtk_gen3_pcie {
 	struct device *dev;
 	void __iomem *base;
-	phys_addr_t reg_base;
 	struct reset_control *mac_reset;
 	struct reset_control *phy_reset;
 	struct phy *phy;
@@ -309,24 +307,8 @@ static int mtk_pcie_set_trans_table(struct mtk_gen3_pcie *pcie,
 
 static void mtk_pcie_enable_msi(struct mtk_gen3_pcie *pcie)
 {
-	int i;
 	u32 val;
 
-	for (i = 0; i < PCIE_MSI_SET_NUM; i++) {
-		struct mtk_msi_set *msi_set = &pcie->msi_sets[i];
-
-		msi_set->base = pcie->base + PCIE_MSI_SET_BASE_REG +
-				i * PCIE_MSI_SET_OFFSET;
-		msi_set->msg_addr = pcie->reg_base + PCIE_MSI_SET_BASE_REG +
-				    i * PCIE_MSI_SET_OFFSET;
-
-		/* Configure the MSI capture address */
-		writel_relaxed(lower_32_bits(msi_set->msg_addr), msi_set->base);
-		writel_relaxed(upper_32_bits(msi_set->msg_addr),
-			       pcie->base + PCIE_MSI_SET_ADDR_HI_BASE +
-			       i * PCIE_MSI_SET_ADDR_HI_OFFSET);
-	}
-
 	val = readl_relaxed(pcie->base + PCIE_MSI_SET_ENABLE_REG);
 	val |= PCIE_MSI_SET_ENABLE;
 	writel_relaxed(val, pcie->base + PCIE_MSI_SET_ENABLE_REG);
@@ -653,6 +635,29 @@ static int mtk_pcie_init_msi(struct mtk_gen3_pcie *pcie)
 {
 	struct device *dev = pcie->dev;
 	struct device_node *node = dev->of_node;
+	struct mtk_msi_set *msi_set;
+	void *msg_vaddr[PCIE_MSI_SET_NUM];
+	int i, j, ret = -ENODEV;
+
+	for (i = 0; i < PCIE_MSI_SET_NUM; i++) {
+		msi_set = &pcie->msi_sets[i];
+
+		msi_set->base = pcie->base + PCIE_MSI_SET_BASE_REG +
+				i * PCIE_MSI_SET_OFFSET;
+
+		msg_vaddr[i] = dmam_alloc_coherent(dev, sizeof(dma_addr_t),
+							 &msi_set->msg_addr, GFP_KERNEL);
+		if (!msg_vaddr[i]) {
+			dev_err(dev, "failed to alloc and map MSI address for set %d\n", i);
+			ret = -ENOMEM;
+			goto err_alloc_addr;
+		}
+
+		/* Configure the MSI capture address */
+		writel_relaxed(lower_32_bits(msi_set->msg_addr), msi_set->base);
+		writel_relaxed(upper_32_bits(msi_set->msg_addr), pcie->base +
+			       PCIE_MSI_SET_ADDR_HI_BASE + i * PCIE_MSI_SET_ADDR_HI_OFFSET);
+	}
 
 	mutex_init(&pcie->lock);
 
@@ -660,18 +665,24 @@ static int mtk_pcie_init_msi(struct mtk_gen3_pcie *pcie)
 				  &mtk_msi_bottom_domain_ops, pcie);
 	if (!pcie->msi_bottom_domain) {
 		dev_err(dev, "failed to create MSI bottom domain\n");
-		return -ENODEV;
+		goto err_alloc_addr;
 	}
 
 	pcie->msi_domain = pci_msi_create_irq_domain(dev->fwnode, &mtk_msi_domain_info,
 						     pcie->msi_bottom_domain);
-	if (!pcie->msi_domain) {
-		dev_err(dev, "failed to create MSI domain\n");
-		irq_domain_remove(pcie->msi_bottom_domain);
-		return -ENODEV;
+	if (pcie->msi_domain)
+		return 0;
+
+	dev_err(dev, "failed to create MSI domain\n");
+	irq_domain_remove(pcie->msi_bottom_domain);
+
+err_alloc_addr:
+	for (j = 0; j < i; j++) {
+		msi_set = &pcie->msi_sets[j];
+		dmam_free_coherent(dev, sizeof(dma_addr_t), msg_vaddr[j], msi_set->msg_addr);
 	}
 
-	return 0;
+	return ret;
 }
 
 static int mtk_pcie_init_intx(struct mtk_gen3_pcie *pcie)
@@ -789,20 +800,14 @@ static int mtk_pcie_parse_port(struct mtk_gen3_pcie *pcie)
 {
 	struct device *dev = pcie->dev;
 	struct platform_device *pdev = to_platform_device(dev);
-	struct resource *regs;
 	int ret;
 
-	regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "pcie-mac");
-	if (!regs)
-		return -EINVAL;
-	pcie->base = devm_ioremap_resource(dev, regs);
+	pcie->base = devm_platform_ioremap_resource_byname(pdev, "pcie-mac");
 	if (IS_ERR(pcie->base)) {
 		dev_err(dev, "failed to map register base\n");
 		return PTR_ERR(pcie->base);
 	}
 
-	pcie->reg_base = regs->start;
-
 	pcie->phy_reset = devm_reset_control_get_optional_exclusive(dev, "phy");
 	if (IS_ERR(pcie->phy_reset)) {
 		ret = PTR_ERR(pcie->phy_reset);
@@ -1013,6 +1018,11 @@ static void mtk_pcie_irq_restore(struct mtk_gen3_pcie *pcie)
 	for (i = 0; i < PCIE_MSI_SET_NUM; i++) {
 		struct mtk_msi_set *msi_set = &pcie->msi_sets[i];
 
+		/* Configure the MSI capture address */
+		writel_relaxed(lower_32_bits(msi_set->msg_addr), msi_set->base);
+		writel_relaxed(upper_32_bits(msi_set->msg_addr), pcie->base +
+			       PCIE_MSI_SET_ADDR_HI_BASE + i * PCIE_MSI_SET_ADDR_HI_OFFSET);
+
 		writel_relaxed(msi_set->saved_irq_state,
 			       msi_set->base + PCIE_MSI_SET_ENABLE_OFFSET);
 	}
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

^ permalink raw reply related	[flat|nested] 7+ messages in thread

* Re: [PATCH v2 1/3] PCI: mediatek: Allocate MSI address with dmam_alloc_coherent()
  2023-12-11  8:52 ` [PATCH v2 1/3] " Jianjun Wang
@ 2024-06-08  9:01   ` Manivannan Sadhasivam
  2024-06-09 12:32     ` Marc Zyngier
  0 siblings, 1 reply; 7+ messages in thread
From: Manivannan Sadhasivam @ 2024-06-08  9:01 UTC (permalink / raw)
  To: Jianjun Wang
  Cc: Lorenzo Pieralisi, Krzysztof Wilczyński, Rob Herring,
	Bjorn Helgaas, Matthias Brugger, AngeloGioacchino Del Regno,
	Marc Zyngier, Ryder Lee, linux-pci, linux-mediatek, linux-kernel,
	linux-arm-kernel, jieyy.yang, chuanjia.liu, qizhong.cheng,
	jian.yang, jianguo.zhang

On Mon, Dec 11, 2023 at 04:52:54PM +0800, Jianjun Wang wrote:
> Use dmam_alloc_coherent() to allocate the MSI address, instead of using
> virt_to_phys().
> 

What is the reason for this change? So now PCIE_MSI_VECTOR becomes unused?

- Mani

> Signed-off-by: Jianjun Wang <jianjun.wang@mediatek.com>
> ---
>  drivers/pci/controller/pcie-mediatek.c | 24 ++++++++++++++++--------
>  1 file changed, 16 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/pci/controller/pcie-mediatek.c b/drivers/pci/controller/pcie-mediatek.c
> index 66a8f73296fc..2fb9e44369f8 100644
> --- a/drivers/pci/controller/pcie-mediatek.c
> +++ b/drivers/pci/controller/pcie-mediatek.c
> @@ -178,6 +178,7 @@ struct mtk_pcie_soc {
>   * @phy: pointer to PHY control block
>   * @slot: port slot
>   * @irq: GIC irq
> + * @msg_addr: MSI message address
>   * @irq_domain: legacy INTx IRQ domain
>   * @inner_domain: inner IRQ domain
>   * @msi_domain: MSI IRQ domain
> @@ -198,6 +199,7 @@ struct mtk_pcie_port {
>  	struct phy *phy;
>  	u32 slot;
>  	int irq;
> +	dma_addr_t msg_addr;
>  	struct irq_domain *irq_domain;
>  	struct irq_domain *inner_domain;
>  	struct irq_domain *msi_domain;
> @@ -394,12 +396,10 @@ static struct pci_ops mtk_pcie_ops_v2 = {
>  static void mtk_compose_msi_msg(struct irq_data *data, struct msi_msg *msg)
>  {
>  	struct mtk_pcie_port *port = irq_data_get_irq_chip_data(data);
> -	phys_addr_t addr;
>  
>  	/* MT2712/MT7622 only support 32-bit MSI addresses */
> -	addr = virt_to_phys(port->base + PCIE_MSI_VECTOR);
>  	msg->address_hi = 0;
> -	msg->address_lo = lower_32_bits(addr);
> +	msg->address_lo = lower_32_bits(port->msg_addr);
>  
>  	msg->data = data->hwirq;
>  
> @@ -494,6 +494,14 @@ static struct msi_domain_info mtk_msi_domain_info = {
>  static int mtk_pcie_allocate_msi_domains(struct mtk_pcie_port *port)
>  {
>  	struct fwnode_handle *fwnode = of_node_to_fwnode(port->pcie->dev->of_node);
> +	void *msi_vaddr;
> +
> +	msi_vaddr = dmam_alloc_coherent(port->pcie->dev, sizeof(dma_addr_t), &port->msg_addr,
> +					GFP_KERNEL);
> +	if (!msi_vaddr) {
> +		dev_err(port->pcie->dev, "failed to alloc and map MSI address\n");
> +		return -ENOMEM;
> +	}
>  
>  	mutex_init(&port->lock);
>  
> @@ -501,6 +509,7 @@ static int mtk_pcie_allocate_msi_domains(struct mtk_pcie_port *port)
>  						      &msi_domain_ops, port);
>  	if (!port->inner_domain) {
>  		dev_err(port->pcie->dev, "failed to create IRQ domain\n");
> +		dmam_free_coherent(port->pcie->dev, sizeof(dma_addr_t), msi_vaddr, port->msg_addr);
>  		return -ENOMEM;
>  	}
>  
> @@ -508,6 +517,7 @@ static int mtk_pcie_allocate_msi_domains(struct mtk_pcie_port *port)
>  						     port->inner_domain);
>  	if (!port->msi_domain) {
>  		dev_err(port->pcie->dev, "failed to create MSI domain\n");
> +		dmam_free_coherent(port->pcie->dev, sizeof(dma_addr_t), msi_vaddr, port->msg_addr);
>  		irq_domain_remove(port->inner_domain);
>  		return -ENOMEM;
>  	}
> @@ -518,10 +528,8 @@ static int mtk_pcie_allocate_msi_domains(struct mtk_pcie_port *port)
>  static void mtk_pcie_enable_msi(struct mtk_pcie_port *port)
>  {
>  	u32 val;
> -	phys_addr_t msg_addr;
>  
> -	msg_addr = virt_to_phys(port->base + PCIE_MSI_VECTOR);
> -	val = lower_32_bits(msg_addr);
> +	val = lower_32_bits(port->msg_addr);
>  	writel(val, port->base + PCIE_IMSI_ADDR);
>  
>  	val = readl(port->base + PCIE_INT_MASK);
> @@ -588,7 +596,7 @@ static int mtk_pcie_init_irq_domain(struct mtk_pcie_port *port,
>  	if (IS_ENABLED(CONFIG_PCI_MSI)) {
>  		ret = mtk_pcie_allocate_msi_domains(port);
>  		if (ret)
> -			return ret;
> +			dev_warn(dev, "no MSI supported, only INTx available\n");
>  	}
>  
>  	return 0;
> @@ -732,7 +740,7 @@ static int mtk_pcie_startup_port_v2(struct mtk_pcie_port *port)
>  	val &= ~INTX_MASK;
>  	writel(val, port->base + PCIE_INT_MASK);
>  
> -	if (IS_ENABLED(CONFIG_PCI_MSI))
> +	if (IS_ENABLED(CONFIG_PCI_MSI) && port->msi_domain)
>  		mtk_pcie_enable_msi(port);
>  
>  	/* Set AHB to PCIe translation windows */
> -- 
> 2.18.0
> 
> 

-- 
மணிவண்ணன் சதாசிவம்

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [PATCH v2 1/3] PCI: mediatek: Allocate MSI address with dmam_alloc_coherent()
  2024-06-08  9:01   ` Manivannan Sadhasivam
@ 2024-06-09 12:32     ` Marc Zyngier
  2024-06-11 17:21       ` Bjorn Helgaas
  0 siblings, 1 reply; 7+ messages in thread
From: Marc Zyngier @ 2024-06-09 12:32 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: Jianjun Wang, Lorenzo Pieralisi, Krzysztof Wilczyński,
	Rob Herring, Bjorn Helgaas, Matthias Brugger,
	AngeloGioacchino Del Regno, Ryder Lee, linux-pci, linux-mediatek,
	linux-kernel, linux-arm-kernel, jieyy.yang, chuanjia.liu,
	qizhong.cheng, jian.yang, jianguo.zhang

On Sat, 08 Jun 2024 10:01:52 +0100,
Manivannan Sadhasivam <mani@kernel.org> wrote:
> 
> On Mon, Dec 11, 2023 at 04:52:54PM +0800, Jianjun Wang wrote:
> > Use dmam_alloc_coherent() to allocate the MSI address, instead of using
> > virt_to_phys().
> > 
> 
> What is the reason for this change? So now PCIE_MSI_VECTOR becomes unused?

More importantly, this is yet another example of the DW reference
driver nonsense, where memory is allocated for *MSI*, while the whole
point of MSIs is that it is a write that doesn't target memory, making
any form of RAM allocation absolutely pointless.

This silly approach has been cargo-culted for years, and while I
caught a few in my time, you can't beat copy-paste.

IMO, this patch is only making things worse instead of fixing things.

	M.

-- 
Without deviation from the norm, progress is not possible.

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [PATCH v2 1/3] PCI: mediatek: Allocate MSI address with dmam_alloc_coherent()
  2024-06-09 12:32     ` Marc Zyngier
@ 2024-06-11 17:21       ` Bjorn Helgaas
  0 siblings, 0 replies; 7+ messages in thread
From: Bjorn Helgaas @ 2024-06-11 17:21 UTC (permalink / raw)
  To: Marc Zyngier
  Cc: Manivannan Sadhasivam, Jianjun Wang, Lorenzo Pieralisi,
	Krzysztof Wilczyński, Rob Herring, Bjorn Helgaas,
	Matthias Brugger, AngeloGioacchino Del Regno, Ryder Lee,
	linux-pci, linux-mediatek, linux-kernel, linux-arm-kernel,
	jieyy.yang, chuanjia.liu, qizhong.cheng, jian.yang, jianguo.zhang

On Sun, Jun 09, 2024 at 01:32:38PM +0100, Marc Zyngier wrote:
> On Sat, 08 Jun 2024 10:01:52 +0100,
> Manivannan Sadhasivam <mani@kernel.org> wrote:
> > 
> > On Mon, Dec 11, 2023 at 04:52:54PM +0800, Jianjun Wang wrote:
> > > Use dmam_alloc_coherent() to allocate the MSI address, instead of using
> > > virt_to_phys().
> > 
> > What is the reason for this change? So now PCIE_MSI_VECTOR becomes unused?
> 
> More importantly, this is yet another example of the DW reference
> driver nonsense, where memory is allocated for *MSI*, while the whole
> point of MSIs is that it is a write that doesn't target memory, making
> any form of RAM allocation absolutely pointless.
> 
> This silly approach has been cargo-culted for years, and while I
> caught a few in my time, you can't beat copy-paste.
> 
> IMO, this patch is only making things worse instead of fixing things.

Probably partly my fault.  I think there are two pieces here:

  1) allocating the MSI address
  2) computing the PCI bus address

I don't know how to do 1), but I do encourage people not to use
virt_to_phys() for 2), since (in general) CPU physical addresses are
not the same as PCI bus addresses.


^ permalink raw reply	[flat|nested] 7+ messages in thread

end of thread, other threads:[~2024-06-11 17:22 UTC | newest]

Thread overview: 7+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2023-12-11  8:52 [PATCH v2 0/3] PCI: mediatek: Allocate MSI address with dmam_alloc_coherent() Jianjun Wang
2023-12-11  8:52 ` [PATCH v2 1/3] " Jianjun Wang
2024-06-08  9:01   ` Manivannan Sadhasivam
2024-06-09 12:32     ` Marc Zyngier
2024-06-11 17:21       ` Bjorn Helgaas
2023-12-11  8:52 ` [PATCH v2 2/3] PCI: mediatek-gen3: Do not break probe flow when MSI init fails Jianjun Wang
2023-12-11  8:52 ` [PATCH v2 3/3] PCI: mediatek-gen3: Allocate MSI address with dmam_alloc_coherent() Jianjun Wang

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).