[PATCH 0/4] iommu/amd: Enable support for up to 2K interrupts per

[PATCH 0/4] iommu/amd: Enable support for up to 2K interrupts per

[Sairaj Kodilkar]

Newer versions of the AMD platform can support up to 2K interrupts per
function. This support can be enabled via field NumIntRemapMode in IOMMU
control register (MMIO offset 0x0018).

This series adds support to enable 2K interrupts per function on top of
v6.14-rc5.

Please see more details in [1]

[1] https://www.amd.com/content/dam/amd/en/documents/processor-tech-docs/specifications/48882_IOMMU.pdf

Kishon Vijay Abraham I (1):
  iommu/amd: Enable support for up to 2K interrupts per function

Sairaj Kodilkar (3):
  iommu/amd: Introduce generic function to set multibit feature value
  iommu/amd: Replace slab cache allocator with page allocator
  iommu/amd: Rename DTE_INTTABLEN* and MAX_IRQS_PER_TABLE macro

 drivers/iommu/amd/amd_iommu_types.h | 27 ++++++-----
 drivers/iommu/amd/init.c            | 65 +++++++++++---------------
 drivers/iommu/amd/iommu.c           | 72 +++++++++++++++++++++--------
 3 files changed, 94 insertions(+), 70 deletions(-)

-- 
2.34.1

[PATCH 1/4] iommu/amd: Introduce generic function to set multibit feature value

[Sairaj Kodilkar]

Define generic function `iommu_feature_set()` to set the values
in the feature control register and replace `iommu_set_inv_tlb_timeout()`
with it.

Signed-off-by: Sairaj Kodilkar <sarunkod@amd.com>
---
 drivers/iommu/amd/amd_iommu_types.h |  2 +-
 drivers/iommu/amd/init.c            | 27 ++++++++++-----------------
 2 files changed, 11 insertions(+), 18 deletions(-)

diff --git a/drivers/iommu/amd/amd_iommu_types.h b/drivers/iommu/amd/amd_iommu_types.h
index 23caea22f8dc..dcc6fbfa62e2 100644
--- a/drivers/iommu/amd/amd_iommu_types.h
+++ b/drivers/iommu/amd/amd_iommu_types.h
@@ -181,7 +181,7 @@
 #define CONTROL_IRTCACHEDIS	59
 #define CONTROL_SNPAVIC_EN	61
 
-#define CTRL_INV_TO_MASK	(7 << CONTROL_INV_TIMEOUT)
+#define CTRL_INV_TO_MASK	7
 #define CTRL_INV_TO_NONE	0
 #define CTRL_INV_TO_1MS		1
 #define CTRL_INV_TO_10MS	2
diff --git a/drivers/iommu/amd/init.c b/drivers/iommu/amd/init.c
index cb536d372b12..eaea9dc8af9d 100644
--- a/drivers/iommu/amd/init.c
+++ b/drivers/iommu/amd/init.c
@@ -412,33 +412,26 @@ static void iommu_set_device_table(struct amd_iommu *iommu)
 			&entry, sizeof(entry));
 }
 
-/* Generic functions to enable/disable certain features of the IOMMU. */
-void iommu_feature_enable(struct amd_iommu *iommu, u8 bit)
+static void iommu_feature_set(struct amd_iommu *iommu, u64 val, u64 mask, u8 shift)
 {
 	u64 ctrl;
 
 	ctrl = readq(iommu->mmio_base +  MMIO_CONTROL_OFFSET);
-	ctrl |= (1ULL << bit);
+	mask <<= shift;
+	ctrl &= ~mask;
+	ctrl |= (val << shift) & mask;
 	writeq(ctrl, iommu->mmio_base +  MMIO_CONTROL_OFFSET);
 }
 
-static void iommu_feature_disable(struct amd_iommu *iommu, u8 bit)
+/* Generic functions to enable/disable certain features of the IOMMU. */
+void iommu_feature_enable(struct amd_iommu *iommu, u8 bit)
 {
-	u64 ctrl;
-
-	ctrl = readq(iommu->mmio_base + MMIO_CONTROL_OFFSET);
-	ctrl &= ~(1ULL << bit);
-	writeq(ctrl, iommu->mmio_base + MMIO_CONTROL_OFFSET);
+	iommu_feature_set(iommu, 1ULL, 1ULL, bit);
 }
 
-static void iommu_set_inv_tlb_timeout(struct amd_iommu *iommu, int timeout)
+static void iommu_feature_disable(struct amd_iommu *iommu, u8 bit)
 {
-	u64 ctrl;
-
-	ctrl = readq(iommu->mmio_base + MMIO_CONTROL_OFFSET);
-	ctrl &= ~CTRL_INV_TO_MASK;
-	ctrl |= (timeout << CONTROL_INV_TIMEOUT) & CTRL_INV_TO_MASK;
-	writeq(ctrl, iommu->mmio_base + MMIO_CONTROL_OFFSET);
+	iommu_feature_set(iommu, 0ULL, 1ULL, bit);
 }
 
 /* Function to enable the hardware */
@@ -2652,7 +2645,7 @@ static void iommu_init_flags(struct amd_iommu *iommu)
 	iommu_feature_enable(iommu, CONTROL_COHERENT_EN);
 
 	/* Set IOTLB invalidation timeout to 1s */
-	iommu_set_inv_tlb_timeout(iommu, CTRL_INV_TO_1S);
+	iommu_feature_set(iommu, CTRL_INV_TO_1S, CTRL_INV_TO_MASK, CONTROL_INV_TIMEOUT);
 
 	/* Enable Enhanced Peripheral Page Request Handling */
 	if (check_feature(FEATURE_EPHSUP))
-- 
2.34.1

[PATCH 2/4] iommu/amd: Replace slab cache allocator with page allocator

[Sairaj Kodilkar]

Commit 05152a049444 ("iommu/amd: Add slab-cache for irq remapping tables")
introduces slab cache allocator. But slab cache allocator provides benefit
only when the allocation and deallocation of many identical objects is
frequent. The AMD IOMMU driver allocates Interrupt remapping table (IRT)
when device driver requests IRQ for the first time and never frees it.
Hence the slab allocator does not provide any benefit here.

Signed-off-by: Sairaj Kodilkar <sarunkod@amd.com>
---
 drivers/iommu/amd/amd_iommu_types.h |  4 ----
 drivers/iommu/amd/init.c            | 21 +--------------------
 drivers/iommu/amd/iommu.c           | 26 ++++++++++++++------------
 3 files changed, 15 insertions(+), 36 deletions(-)

diff --git a/drivers/iommu/amd/amd_iommu_types.h b/drivers/iommu/amd/amd_iommu_types.h
index dcc6fbfa62e2..a1e2b8317d11 100644
--- a/drivers/iommu/amd/amd_iommu_types.h
+++ b/drivers/iommu/amd/amd_iommu_types.h
@@ -313,7 +313,6 @@
  * AMD IOMMU hardware only support 512 IRTEs despite
  * the architectural limitation of 2048 entries.
  */
-#define DTE_INTTAB_ALIGNMENT    128
 #define DTE_INTTABLEN_VALUE     9ULL
 #define DTE_INTTABLEN           (DTE_INTTABLEN_VALUE << 1)
 #define DTE_INTTABLEN_MASK      (0xfULL << 1)
@@ -492,9 +491,6 @@ extern const struct iommu_ops amd_iommu_ops;
 /* IVRS indicates that pre-boot remapping was enabled */
 extern bool amdr_ivrs_remap_support;
 
-/* kmem_cache to get tables with 128 byte alignement */
-extern struct kmem_cache *amd_iommu_irq_cache;
-
 #define PCI_SBDF_TO_SEGID(sbdf)		(((sbdf) >> 16) & 0xffff)
 #define PCI_SBDF_TO_DEVID(sbdf)		((sbdf) & 0xffff)
 #define PCI_SEG_DEVID_TO_SBDF(seg, devid)	((((u32)(seg) & 0xffff) << 16) | \
diff --git a/drivers/iommu/amd/init.c b/drivers/iommu/amd/init.c
index eaea9dc8af9d..0fbe92a0b48a 100644
--- a/drivers/iommu/amd/init.c
+++ b/drivers/iommu/amd/init.c
@@ -12,7 +12,6 @@
 #include <linux/acpi.h>
 #include <linux/list.h>
 #include <linux/bitmap.h>
-#include <linux/slab.h>
 #include <linux/syscore_ops.h>
 #include <linux/interrupt.h>
 #include <linux/msi.h>
@@ -2932,9 +2931,6 @@ static struct syscore_ops amd_iommu_syscore_ops = {
 
 static void __init free_iommu_resources(void)
 {
-	kmem_cache_destroy(amd_iommu_irq_cache);
-	amd_iommu_irq_cache = NULL;
-
 	free_iommu_all();
 	free_pci_segments();
 }
@@ -3033,7 +3029,7 @@ static void __init ivinfo_init(void *ivrs)
 static int __init early_amd_iommu_init(void)
 {
 	struct acpi_table_header *ivrs_base;
-	int remap_cache_sz, ret;
+	int ret;
 	acpi_status status;
 
 	if (!amd_iommu_detected)
@@ -3095,22 +3091,7 @@ static int __init early_amd_iommu_init(void)
 
 	if (amd_iommu_irq_remap) {
 		struct amd_iommu_pci_seg *pci_seg;
-		/*
-		 * Interrupt remapping enabled, create kmem_cache for the
-		 * remapping tables.
-		 */
 		ret = -ENOMEM;
-		if (!AMD_IOMMU_GUEST_IR_GA(amd_iommu_guest_ir))
-			remap_cache_sz = MAX_IRQS_PER_TABLE * sizeof(u32);
-		else
-			remap_cache_sz = MAX_IRQS_PER_TABLE * (sizeof(u64) * 2);
-		amd_iommu_irq_cache = kmem_cache_create("irq_remap_cache",
-							remap_cache_sz,
-							DTE_INTTAB_ALIGNMENT,
-							0, NULL);
-		if (!amd_iommu_irq_cache)
-			goto out;
-
 		for_each_pci_segment(pci_seg) {
 			if (alloc_irq_lookup_table(pci_seg))
 				goto out;
diff --git a/drivers/iommu/amd/iommu.c b/drivers/iommu/amd/iommu.c
index cd5116d8c3b2..0960b95656e6 100644
--- a/drivers/iommu/amd/iommu.c
+++ b/drivers/iommu/amd/iommu.c
@@ -75,8 +75,6 @@ struct iommu_cmd {
  */
 DEFINE_IDA(pdom_ids);
 
-struct kmem_cache *amd_iommu_irq_cache;
-
 static int amd_iommu_attach_device(struct iommu_domain *dom,
 				   struct device *dev);
 
@@ -3121,7 +3119,7 @@ static struct irq_remap_table *get_irq_table(struct amd_iommu *iommu, u16 devid)
 	return table;
 }
 
-static struct irq_remap_table *__alloc_irq_table(void)
+static struct irq_remap_table *__alloc_irq_table(int nid, int order)
 {
 	struct irq_remap_table *table;
 
@@ -3129,19 +3127,13 @@ static struct irq_remap_table *__alloc_irq_table(void)
 	if (!table)
 		return NULL;
 
-	table->table = kmem_cache_alloc(amd_iommu_irq_cache, GFP_KERNEL);
+	table->table = iommu_alloc_pages_node(nid, GFP_KERNEL, order);
 	if (!table->table) {
 		kfree(table);
 		return NULL;
 	}
 	raw_spin_lock_init(&table->lock);
 
-	if (!AMD_IOMMU_GUEST_IR_GA(amd_iommu_guest_ir))
-		memset(table->table, 0,
-		       MAX_IRQS_PER_TABLE * sizeof(u32));
-	else
-		memset(table->table, 0,
-		       (MAX_IRQS_PER_TABLE * (sizeof(u64) * 2)));
 	return table;
 }
 
@@ -3173,6 +3165,14 @@ static int set_remap_table_entry_alias(struct pci_dev *pdev, u16 alias,
 	return 0;
 }
 
+static inline size_t get_irq_table_size(unsigned int max_irqs)
+{
+	if (!AMD_IOMMU_GUEST_IR_GA(amd_iommu_guest_ir))
+		return max_irqs * sizeof(u32);
+
+	return max_irqs * (sizeof(u64) * 2);
+}
+
 static struct irq_remap_table *alloc_irq_table(struct amd_iommu *iommu,
 					       u16 devid, struct pci_dev *pdev)
 {
@@ -3180,6 +3180,8 @@ static struct irq_remap_table *alloc_irq_table(struct amd_iommu *iommu,
 	struct irq_remap_table *new_table = NULL;
 	struct amd_iommu_pci_seg *pci_seg;
 	unsigned long flags;
+	int order = get_order(get_irq_table_size(MAX_IRQS_PER_TABLE));
+	int nid = iommu && iommu->dev ? dev_to_node(&iommu->dev->dev) : NUMA_NO_NODE;
 	u16 alias;
 
 	spin_lock_irqsave(&iommu_table_lock, flags);
@@ -3198,7 +3200,7 @@ static struct irq_remap_table *alloc_irq_table(struct amd_iommu *iommu,
 	spin_unlock_irqrestore(&iommu_table_lock, flags);
 
 	/* Nothing there yet, allocate new irq remapping table */
-	new_table = __alloc_irq_table();
+	new_table = __alloc_irq_table(nid, order);
 	if (!new_table)
 		return NULL;
 
@@ -3233,7 +3235,7 @@ static struct irq_remap_table *alloc_irq_table(struct amd_iommu *iommu,
 	spin_unlock_irqrestore(&iommu_table_lock, flags);
 
 	if (new_table) {
-		kmem_cache_free(amd_iommu_irq_cache, new_table->table);
+		iommu_free_pages(new_table->table, order);
 		kfree(new_table);
 	}
 	return table;
-- 
2.34.1

[PATCH 3/4] iommu/amd: Rename DTE_INTTABLEN* and MAX_IRQS_PER_TABLE macro

[Sairaj Kodilkar]

AMD iommu can support both 512 and 2K interrupts on newer platform. Hence
add suffix "512" to the existing macros.

Signed-off-by: Sairaj Kodilkar <sarunkod@amd.com>
---
 drivers/iommu/amd/amd_iommu_types.h | 6 +++---
 drivers/iommu/amd/init.c            | 2 +-
 drivers/iommu/amd/iommu.c           | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/iommu/amd/amd_iommu_types.h b/drivers/iommu/amd/amd_iommu_types.h
index a1e2b8317d11..2efb47dba270 100644
--- a/drivers/iommu/amd/amd_iommu_types.h
+++ b/drivers/iommu/amd/amd_iommu_types.h
@@ -313,10 +313,10 @@
  * AMD IOMMU hardware only support 512 IRTEs despite
  * the architectural limitation of 2048 entries.
  */
-#define DTE_INTTABLEN_VALUE     9ULL
-#define DTE_INTTABLEN           (DTE_INTTABLEN_VALUE << 1)
 #define DTE_INTTABLEN_MASK      (0xfULL << 1)
-#define MAX_IRQS_PER_TABLE      (1 << DTE_INTTABLEN_VALUE)
+#define DTE_INTTABLEN_VALUE_512 9ULL
+#define DTE_INTTABLEN_512       (DTE_INTTABLEN_VALUE_512 << 1)
+#define MAX_IRQS_PER_TABLE_512  BIT(DTE_INTTABLEN_VALUE_512)
 
 #define PAGE_MODE_NONE    0x00
 #define PAGE_MODE_1_LEVEL 0x01
diff --git a/drivers/iommu/amd/init.c b/drivers/iommu/amd/init.c
index 0fbe92a0b48a..9a311c627090 100644
--- a/drivers/iommu/amd/init.c
+++ b/drivers/iommu/amd/init.c
@@ -1061,7 +1061,7 @@ static bool __copy_device_table(struct amd_iommu *iommu)
 		int_tab_len = old_devtb[devid].data[2] & DTE_INTTABLEN_MASK;
 		if (irq_v && (int_ctl || int_tab_len)) {
 			if ((int_ctl != DTE_IRQ_REMAP_INTCTL) ||
-			    (int_tab_len != DTE_INTTABLEN)) {
+			    (int_tab_len != DTE_INTTABLEN_512)) {
 				pr_err("Wrong old irq remapping flag: %#x\n", devid);
 				memunmap(old_devtb);
 				return false;
diff --git a/drivers/iommu/amd/iommu.c b/drivers/iommu/amd/iommu.c
index 0960b95656e6..1063c4988dca 100644
--- a/drivers/iommu/amd/iommu.c
+++ b/drivers/iommu/amd/iommu.c
@@ -3093,7 +3093,7 @@ static void set_dte_irq_entry(struct amd_iommu *iommu, u16 devid,
 	new &= ~DTE_IRQ_PHYS_ADDR_MASK;
 	new |= iommu_virt_to_phys(table->table);
 	new |= DTE_IRQ_REMAP_INTCTL;
-	new |= DTE_INTTABLEN;
+	new |= DTE_INTTABLEN_512;
 	new |= DTE_IRQ_REMAP_ENABLE;
 	WRITE_ONCE(dte->data[2], new);
 
-- 
2.34.1

[PATCH 4/4] iommu/amd: Enable support for up to 2K interrupts per function

[Sairaj Kodilkar]

From: Kishon Vijay Abraham I <kvijayab@amd.com>

AMD IOMMU optionally supports up to 2K interrupts per function on newer
platforms. Support for this feature is indicated through Extended
Feature 2 Register (MMIO Offset 01A0h[NumIntRemapSup]). Allocate 2K IRTEs
per device when this support is available.

Co-developed-by: Sairaj Kodilkar <sarunkod@amd.com>
Signed-off-by: Sairaj Kodilkar <sarunkod@amd.com>
Signed-off-by: Kishon Vijay Abraham I <kvijayab@amd.com>
---
 drivers/iommu/amd/amd_iommu_types.h | 15 ++++++---
 drivers/iommu/amd/init.c            | 16 +++++++++-
 drivers/iommu/amd/iommu.c           | 48 +++++++++++++++++++++++------
 3 files changed, 65 insertions(+), 14 deletions(-)

diff --git a/drivers/iommu/amd/amd_iommu_types.h b/drivers/iommu/amd/amd_iommu_types.h
index 2efb47dba270..78821d45669d 100644
--- a/drivers/iommu/amd/amd_iommu_types.h
+++ b/drivers/iommu/amd/amd_iommu_types.h
@@ -112,6 +112,10 @@
 #define FEATURE_SNPAVICSUP_GAM(x) \
 	(FIELD_GET(FEATURE_SNPAVICSUP, x) == 0x1)
 
+#define FEATURE_NUM_INT_REMAP_SUP	GENMASK_ULL(9, 8)
+#define FEATURE_NUM_INT_REMAP_SUP_2K(x) \
+	(FIELD_GET(FEATURE_NUM_INT_REMAP_SUP, x) == 0x1)
+
 /* Note:
  * The current driver only support 16-bit PASID.
  * Currently, hardware only implement upto 16-bit PASID
@@ -175,6 +179,9 @@
 #define CONTROL_GAM_EN		25
 #define CONTROL_GALOG_EN	28
 #define CONTROL_GAINT_EN	29
+#define CONTROL_NUM_INT_REMAP_MODE	43
+#define CONTROL_NUM_INT_REMAP_MODE_MASK	0x03
+#define CONTROL_NUM_INT_REMAP_MODE_2K	0x01
 #define CONTROL_EPH_EN		45
 #define CONTROL_XT_EN		50
 #define CONTROL_INTCAPXT_EN	51
@@ -309,14 +316,13 @@
 #define DTE_IRQ_REMAP_INTCTL    (2ULL << 60)
 #define DTE_IRQ_REMAP_ENABLE    1ULL
 
-/*
- * AMD IOMMU hardware only support 512 IRTEs despite
- * the architectural limitation of 2048 entries.
- */
 #define DTE_INTTABLEN_MASK      (0xfULL << 1)
 #define DTE_INTTABLEN_VALUE_512 9ULL
 #define DTE_INTTABLEN_512       (DTE_INTTABLEN_VALUE_512 << 1)
 #define MAX_IRQS_PER_TABLE_512  BIT(DTE_INTTABLEN_VALUE_512)
+#define DTE_INTTABLEN_VALUE_2K	11ULL
+#define DTE_INTTABLEN_2K	(DTE_INTTABLEN_VALUE_2K << 1)
+#define MAX_IRQS_PER_TABLE_2K	BIT(DTE_INTTABLEN_VALUE_2K)
 
 #define PAGE_MODE_NONE    0x00
 #define PAGE_MODE_1_LEVEL 0x01
@@ -847,6 +853,7 @@ struct iommu_dev_data {
 	struct device *dev;
 	u16 devid;			  /* PCI Device ID */
 
+	unsigned int max_irqs;		  /* Maximum IRQs supported by device */
 	u32 max_pasids;			  /* Max supported PASIDs */
 	u32 flags;			  /* Holds AMD_IOMMU_DEVICE_FLAG_<*> */
 	int ats_qdep;
diff --git a/drivers/iommu/amd/init.c b/drivers/iommu/amd/init.c
index 9a311c627090..c99b2e34f3b4 100644
--- a/drivers/iommu/amd/init.c
+++ b/drivers/iommu/amd/init.c
@@ -1061,7 +1061,8 @@ static bool __copy_device_table(struct amd_iommu *iommu)
 		int_tab_len = old_devtb[devid].data[2] & DTE_INTTABLEN_MASK;
 		if (irq_v && (int_ctl || int_tab_len)) {
 			if ((int_ctl != DTE_IRQ_REMAP_INTCTL) ||
-			    (int_tab_len != DTE_INTTABLEN_512)) {
+			    (int_tab_len != DTE_INTTABLEN_512 &&
+			     int_tab_len != DTE_INTTABLEN_2K)) {
 				pr_err("Wrong old irq remapping flag: %#x\n", devid);
 				memunmap(old_devtb);
 				return false;
@@ -2737,6 +2738,17 @@ static void iommu_enable_irtcachedis(struct amd_iommu *iommu)
 		iommu->irtcachedis_enabled ? "disabled" : "enabled");
 }
 
+static void iommu_enable_2k_int(struct amd_iommu *iommu)
+{
+	if (!FEATURE_NUM_INT_REMAP_SUP_2K(amd_iommu_efr2))
+		return;
+
+	iommu_feature_set(iommu,
+			  CONTROL_NUM_INT_REMAP_MODE_2K,
+			  CONTROL_NUM_INT_REMAP_MODE_MASK,
+			  CONTROL_NUM_INT_REMAP_MODE);
+}
+
 static void early_enable_iommu(struct amd_iommu *iommu)
 {
 	iommu_disable(iommu);
@@ -2749,6 +2761,7 @@ static void early_enable_iommu(struct amd_iommu *iommu)
 	iommu_enable_ga(iommu);
 	iommu_enable_xt(iommu);
 	iommu_enable_irtcachedis(iommu);
+	iommu_enable_2k_int(iommu);
 	iommu_enable(iommu);
 	amd_iommu_flush_all_caches(iommu);
 }
@@ -2805,6 +2818,7 @@ static void early_enable_iommus(void)
 			iommu_enable_ga(iommu);
 			iommu_enable_xt(iommu);
 			iommu_enable_irtcachedis(iommu);
+			iommu_enable_2k_int(iommu);
 			iommu_set_device_table(iommu);
 			amd_iommu_flush_all_caches(iommu);
 		}
diff --git a/drivers/iommu/amd/iommu.c b/drivers/iommu/amd/iommu.c
index 1063c4988dca..fc420fdf8539 100644
--- a/drivers/iommu/amd/iommu.c
+++ b/drivers/iommu/amd/iommu.c
@@ -2392,8 +2392,14 @@ static struct iommu_device *amd_iommu_probe_device(struct device *dev)
 	}
 
 out_err:
+
 	iommu_completion_wait(iommu);
 
+	if (FEATURE_NUM_INT_REMAP_SUP_2K(amd_iommu_efr2))
+		dev_data->max_irqs = MAX_IRQS_PER_TABLE_2K;
+	else
+		dev_data->max_irqs = MAX_IRQS_PER_TABLE_512;
+
 	if (dev_is_pci(dev))
 		pci_prepare_ats(to_pci_dev(dev), PAGE_SHIFT);
 
@@ -3079,6 +3085,13 @@ static void iommu_flush_irt_and_complete(struct amd_iommu *iommu, u16 devid)
 	raw_spin_unlock_irqrestore(&iommu->lock, flags);
 }
 
+static inline u8 iommu_get_int_tablen(struct iommu_dev_data *dev_data)
+{
+	if (dev_data && dev_data->max_irqs == MAX_IRQS_PER_TABLE_2K)
+		return DTE_INTTABLEN_2K;
+	return DTE_INTTABLEN_512;
+}
+
 static void set_dte_irq_entry(struct amd_iommu *iommu, u16 devid,
 			      struct irq_remap_table *table)
 {
@@ -3093,7 +3106,7 @@ static void set_dte_irq_entry(struct amd_iommu *iommu, u16 devid,
 	new &= ~DTE_IRQ_PHYS_ADDR_MASK;
 	new |= iommu_virt_to_phys(table->table);
 	new |= DTE_IRQ_REMAP_INTCTL;
-	new |= DTE_INTTABLEN_512;
+	new |= iommu_get_int_tablen(dev_data);
 	new |= DTE_IRQ_REMAP_ENABLE;
 	WRITE_ONCE(dte->data[2], new);
 
@@ -3174,13 +3187,14 @@ static inline size_t get_irq_table_size(unsigned int max_irqs)
 }
 
 static struct irq_remap_table *alloc_irq_table(struct amd_iommu *iommu,
-					       u16 devid, struct pci_dev *pdev)
+					       u16 devid, struct pci_dev *pdev,
+					       unsigned int max_irqs)
 {
 	struct irq_remap_table *table = NULL;
 	struct irq_remap_table *new_table = NULL;
 	struct amd_iommu_pci_seg *pci_seg;
 	unsigned long flags;
-	int order = get_order(get_irq_table_size(MAX_IRQS_PER_TABLE));
+	int order = get_order(get_irq_table_size(max_irqs));
 	int nid = iommu && iommu->dev ? dev_to_node(&iommu->dev->dev) : NUMA_NO_NODE;
 	u16 alias;
 
@@ -3242,13 +3256,14 @@ static struct irq_remap_table *alloc_irq_table(struct amd_iommu *iommu,
 }
 
 static int alloc_irq_index(struct amd_iommu *iommu, u16 devid, int count,
-			   bool align, struct pci_dev *pdev)
+			   bool align, struct pci_dev *pdev,
+			   unsigned long max_irqs)
 {
 	struct irq_remap_table *table;
 	int index, c, alignment = 1;
 	unsigned long flags;
 
-	table = alloc_irq_table(iommu, devid, pdev);
+	table = alloc_irq_table(iommu, devid, pdev, max_irqs);
 	if (!table)
 		return -ENODEV;
 
@@ -3259,7 +3274,7 @@ static int alloc_irq_index(struct amd_iommu *iommu, u16 devid, int count,
 
 	/* Scan table for free entries */
 	for (index = ALIGN(table->min_index, alignment), c = 0;
-	     index < MAX_IRQS_PER_TABLE;) {
+	     index < max_irqs;) {
 		if (!iommu->irte_ops->is_allocated(table, index)) {
 			c += 1;
 		} else {
@@ -3529,6 +3544,14 @@ static void fill_msi_msg(struct msi_msg *msg, u32 index)
 	msg->data = index;
 	msg->address_lo = 0;
 	msg->arch_addr_lo.base_address = X86_MSI_BASE_ADDRESS_LOW;
+	/*
+	 * The struct msi_msg.dest_mode_logical is used to set the DM bit
+	 * in MSI Message Address Register. For device w/ 2K int-remap support,
+	 * this is bit must be set to 1 regardless of the actual destination
+	 * mode, which is signified by the IRTE[DM].
+	 */
+	if (FEATURE_NUM_INT_REMAP_SUP_2K(amd_iommu_efr2))
+		msg->arch_addr_lo.dest_mode_logical = true;
 	msg->address_hi = X86_MSI_BASE_ADDRESS_HIGH;
 }
 
@@ -3591,6 +3614,8 @@ static int irq_remapping_alloc(struct irq_domain *domain, unsigned int virq,
 	struct amd_ir_data *data = NULL;
 	struct amd_iommu *iommu;
 	struct irq_cfg *cfg;
+	struct iommu_dev_data *dev_data;
+	unsigned long max_irqs;
 	int i, ret, devid, seg, sbdf;
 	int index;
 
@@ -3609,6 +3634,9 @@ static int irq_remapping_alloc(struct irq_domain *domain, unsigned int virq,
 	if (!iommu)
 		return -EINVAL;
 
+	dev_data = search_dev_data(iommu, devid);
+	max_irqs = dev_data ? dev_data->max_irqs : MAX_IRQS_PER_TABLE_512;
+
 	ret = irq_domain_alloc_irqs_parent(domain, virq, nr_irqs, arg);
 	if (ret < 0)
 		return ret;
@@ -3616,7 +3644,7 @@ static int irq_remapping_alloc(struct irq_domain *domain, unsigned int virq,
 	if (info->type == X86_IRQ_ALLOC_TYPE_IOAPIC) {
 		struct irq_remap_table *table;
 
-		table = alloc_irq_table(iommu, devid, NULL);
+		table = alloc_irq_table(iommu, devid, NULL, max_irqs);
 		if (table) {
 			if (!table->min_index) {
 				/*
@@ -3637,9 +3665,11 @@ static int irq_remapping_alloc(struct irq_domain *domain, unsigned int virq,
 		bool align = (info->type == X86_IRQ_ALLOC_TYPE_PCI_MSI);
 
 		index = alloc_irq_index(iommu, devid, nr_irqs, align,
-					msi_desc_to_pci_dev(info->desc));
+					msi_desc_to_pci_dev(info->desc),
+					max_irqs);
 	} else {
-		index = alloc_irq_index(iommu, devid, nr_irqs, false, NULL);
+		index = alloc_irq_index(iommu, devid, nr_irqs, false, NULL,
+					max_irqs);
 	}
 
 	if (index < 0) {
-- 
2.34.1

Re: [PATCH 0/4] iommu/amd: Enable support for up to 2K interrupts per

[Vasant Hegde]

On 3/7/2025 3:28 PM, Sairaj Kodilkar wrote:
> Newer versions of the AMD platform can support up to 2K interrupts per
> function. This support can be enabled via field NumIntRemapMode in IOMMU
> control register (MMIO offset 0x0018).
> 
> This series adds support to enable 2K interrupts per function on top of
> v6.14-rc5.

I have reviewed/tested this series. It looks good to me.

For entire series :

Reviewed-by: Vasant Hegde <vasant.hegde@amd.com>


-Vasant

Re: [PATCH 0/4] iommu/amd: Enable support for up to 2K interrupts per

[Joerg Roedel]

On Fri, Mar 07, 2025 at 03:28:18PM +0530, Sairaj Kodilkar wrote:
> Kishon Vijay Abraham I (1):
>   iommu/amd: Enable support for up to 2K interrupts per function
> 
> Sairaj Kodilkar (3):
>   iommu/amd: Introduce generic function to set multibit feature value
>   iommu/amd: Replace slab cache allocator with page allocator
>   iommu/amd: Rename DTE_INTTABLEN* and MAX_IRQS_PER_TABLE macro
> 
>  drivers/iommu/amd/amd_iommu_types.h | 27 ++++++-----
>  drivers/iommu/amd/init.c            | 65 +++++++++++---------------
>  drivers/iommu/amd/iommu.c           | 72 +++++++++++++++++++++--------
>  3 files changed, 94 insertions(+), 70 deletions(-)

Applied, thanks.