loongarch.lists.linux.dev archive mirror
 help / color / mirror / Atom feed
* [PATCH V3 0/2] DMA: update acpi_dma_get_range to return dma map regions
@ 2022-08-30  3:01 Jianmin Lv
  2022-08-30  3:01 ` [PATCH V3 1/2] ACPI / scan: Support multiple dma windows with different offsets Jianmin Lv
                   ` (2 more replies)
  0 siblings, 3 replies; 10+ messages in thread
From: Jianmin Lv @ 2022-08-30  3:01 UTC (permalink / raw)
  To: lpieralisi, robin.murphy, chenhuacai
  Cc: guohanjun, sudeep.holla, rafael, lenb, robert.moore,
	linux-kernel, linux-acpi, loongarch

The patch series changed acpi_dma_get_range to return dma regions
as of_dma_get_range, so that dev->dma_range_map can be initialized
conveniently.

And acpi_arch_dma_setup for ARM64 is changed wih removing dma_base
and size from it's parameters.

Remove ARCH_HAS_PHYS_TO_DMA for LoongArch and use generic
phys_to_dma/dma_to_phys in include/linux/dma-direct.h.

V1 -> V2
- Removed dma_base and size from acpi_arch_dma_setup' parameters
- Add patch to remove ARCH_HAS_PHYS_TO_DMA for LoongArch

V2 -> V3
- Add kerneldoc for acpi_dma_get_range changing
- Remove redundant code in acpi_arch_dma_setup, and check map


Jianmin Lv (2):
  ACPI / scan: Support multiple dma windows with different offsets
  LoongArch: Remove ARCH_HAS_PHYS_TO_DMA

 arch/loongarch/Kconfig        |  1 -
 arch/loongarch/kernel/dma.c   | 52 +++++++++++++++++-------------------------
 arch/loongarch/kernel/setup.c |  2 +-
 drivers/acpi/arm64/dma.c      | 29 ++++++++++++++---------
 drivers/acpi/scan.c           | 53 +++++++++++++++++++------------------------
 include/acpi/acpi_bus.h       |  3 +--
 include/linux/acpi.h          | 12 ++++++----
 7 files changed, 71 insertions(+), 81 deletions(-)

-- 
1.8.3.1


^ permalink raw reply	[flat|nested] 10+ messages in thread

* [PATCH V3 1/2] ACPI / scan: Support multiple dma windows with different offsets
  2022-08-30  3:01 [PATCH V3 0/2] DMA: update acpi_dma_get_range to return dma map regions Jianmin Lv
@ 2022-08-30  3:01 ` Jianmin Lv
  2022-09-05 12:20   ` Robin Murphy
  2022-08-30  3:01 ` [PATCH V3 2/2] LoongArch: Remove ARCH_HAS_PHYS_TO_DMA Jianmin Lv
  2022-09-05  2:55 ` [PATCH V3 0/2] DMA: update acpi_dma_get_range to return dma map regions Jianmin Lv
  2 siblings, 1 reply; 10+ messages in thread
From: Jianmin Lv @ 2022-08-30  3:01 UTC (permalink / raw)
  To: lpieralisi, robin.murphy, chenhuacai
  Cc: guohanjun, sudeep.holla, rafael, lenb, robert.moore,
	linux-kernel, linux-acpi, loongarch

For DT, of_dma_get_range returns bus_dma_region typed dma regions,
which makes multiple dma windows with different offset available
for translation between dma address and cpu address.

But for ACPI, acpi_dma_get_range doesn't return similar dma regions,
causing no path for setting dev->dma_range_map conveniently. So the
patch changes acpi_dma_get_range and returns bus_dma_region typed
dma regions according to of_dma_get_range.

After changing acpi_dma_get_range, acpi_arch_dma_setup is changed for
ARM64, where original dma_addr and size are removed as these
arguments are now redundant, and pass 0 and U64_MAX for dma_base
and size of arch_setup_dma_ops, so this is a simplification consistent
with what other ACPI architectures also pass to iommu_setup_dma_ops().

Signed-off-by: Jianmin Lv <lvjianmin@loongson.cn>
---
 drivers/acpi/arm64/dma.c | 29 ++++++++++++++++----------
 drivers/acpi/scan.c      | 53 +++++++++++++++++++++---------------------------
 include/acpi/acpi_bus.h  |  3 +--
 include/linux/acpi.h     |  7 +++----
 4 files changed, 45 insertions(+), 47 deletions(-)

diff --git a/drivers/acpi/arm64/dma.c b/drivers/acpi/arm64/dma.c
index f16739a..1ef8e7d 100644
--- a/drivers/acpi/arm64/dma.c
+++ b/drivers/acpi/arm64/dma.c
@@ -4,11 +4,12 @@
 #include <linux/device.h>
 #include <linux/dma-direct.h>
 
-void acpi_arch_dma_setup(struct device *dev, u64 *dma_addr, u64 *dma_size)
+void acpi_arch_dma_setup(struct device *dev)
 {
 	int ret;
 	u64 end, mask;
-	u64 dmaaddr = 0, size = 0, offset = 0;
+	u64 size = 0;
+	const struct bus_dma_region *map = NULL;
 
 	/*
 	 * If @dev is expected to be DMA-capable then the bus code that created
@@ -26,25 +27,31 @@ void acpi_arch_dma_setup(struct device *dev, u64 *dma_addr, u64 *dma_size)
 	else
 		size = 1ULL << 32;
 
-	ret = acpi_dma_get_range(dev, &dmaaddr, &offset, &size);
+	ret = acpi_dma_get_range(dev, &map);
+	if (!ret && map) {
+		const struct bus_dma_region *r = map;
+
+		for (end = 0; r->size; r++) {
+			if (r->dma_start + r->size - 1 > end)
+				end = r->dma_start + r->size - 1;
+		}
+
+		size = end + 1;
+		dev->dma_range_map = map;
+	}
+
 	if (ret == -ENODEV)
 		ret = iort_dma_get_ranges(dev, &size);
+
 	if (!ret) {
 		/*
 		 * Limit coherent and dma mask based on size retrieved from
 		 * firmware.
 		 */
-		end = dmaaddr + size - 1;
+		end = size - 1;
 		mask = DMA_BIT_MASK(ilog2(end) + 1);
 		dev->bus_dma_limit = end;
 		dev->coherent_dma_mask = min(dev->coherent_dma_mask, mask);
 		*dev->dma_mask = min(*dev->dma_mask, mask);
 	}
-
-	*dma_addr = dmaaddr;
-	*dma_size = size;
-
-	ret = dma_direct_set_offset(dev, dmaaddr + offset, dmaaddr, size);
-
-	dev_dbg(dev, "dma_offset(%#08llx)%s\n", offset, ret ? " failed!" : "");
 }
diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
index 762b61f..8c0c2ca 100644
--- a/drivers/acpi/scan.c
+++ b/drivers/acpi/scan.c
@@ -20,6 +20,7 @@
 #include <linux/platform_data/x86/apple.h>
 #include <linux/pgtable.h>
 #include <linux/crc32.h>
+#include <linux/dma-direct.h>
 
 #include "internal.h"
 
@@ -1482,25 +1483,21 @@ enum dev_dma_attr acpi_get_dma_attr(struct acpi_device *adev)
  * acpi_dma_get_range() - Get device DMA parameters.
  *
  * @dev: device to configure
- * @dma_addr: pointer device DMA address result
- * @offset: pointer to the DMA offset result
- * @size: pointer to DMA range size result
+ * @map: pointer to DMA ranges result
  *
- * Evaluate DMA regions and return respectively DMA region start, offset
- * and size in dma_addr, offset and size on parsing success; it does not
- * update the passed in values on failure.
+ * Evaluate DMA regions and return pointer to DMA regions on
+ * parsing success; it does not update the passed in values on failure.
  *
  * Return 0 on success, < 0 on failure.
  */
-int acpi_dma_get_range(struct device *dev, u64 *dma_addr, u64 *offset,
-		       u64 *size)
+int acpi_dma_get_range(struct device *dev, const struct bus_dma_region **map)
 {
 	struct acpi_device *adev;
 	LIST_HEAD(list);
 	struct resource_entry *rentry;
 	int ret;
 	struct device *dma_dev = dev;
-	u64 len, dma_start = U64_MAX, dma_end = 0, dma_offset = 0;
+	struct bus_dma_region *r;
 
 	/*
 	 * Walk the device tree chasing an ACPI companion with a _DMA
@@ -1525,31 +1522,28 @@ int acpi_dma_get_range(struct device *dev, u64 *dma_addr, u64 *offset,
 
 	ret = acpi_dev_get_dma_resources(adev, &list);
 	if (ret > 0) {
+		r = kcalloc(ret + 1, sizeof(*r), GFP_KERNEL);
+		if (!r) {
+			ret = -ENOMEM;
+			goto out;
+		}
+
+		*map = r;
+
 		list_for_each_entry(rentry, &list, node) {
-			if (dma_offset && rentry->offset != dma_offset) {
+			if (rentry->res->start >= rentry->res->end) {
 				ret = -EINVAL;
-				dev_warn(dma_dev, "Can't handle multiple windows with different offsets\n");
+				dev_dbg(dma_dev, "Invalid DMA regions configuration\n");
 				goto out;
 			}
-			dma_offset = rentry->offset;
 
-			/* Take lower and upper limits */
-			if (rentry->res->start < dma_start)
-				dma_start = rentry->res->start;
-			if (rentry->res->end > dma_end)
-				dma_end = rentry->res->end;
-		}
-
-		if (dma_start >= dma_end) {
-			ret = -EINVAL;
-			dev_dbg(dma_dev, "Invalid DMA regions configuration\n");
-			goto out;
+			r->cpu_start = rentry->res->start;
+			r->dma_start = rentry->res->start - rentry->offset;
+			r->size = rentry->res->end - rentry->res->start + 1;
+			r->offset = rentry->offset;
+			r++;
 		}
 
-		*dma_addr = dma_start - dma_offset;
-		len = dma_end - dma_start;
-		*size = max(len, len + 1);
-		*offset = dma_offset;
 	}
  out:
 	acpi_dev_free_resource_list(&list);
@@ -1639,20 +1633,19 @@ int acpi_dma_configure_id(struct device *dev, enum dev_dma_attr attr,
 			  const u32 *input_id)
 {
 	const struct iommu_ops *iommu;
-	u64 dma_addr = 0, size = 0;
 
 	if (attr == DEV_DMA_NOT_SUPPORTED) {
 		set_dma_ops(dev, &dma_dummy_ops);
 		return 0;
 	}
 
-	acpi_arch_dma_setup(dev, &dma_addr, &size);
+	acpi_arch_dma_setup(dev);
 
 	iommu = acpi_iommu_configure_id(dev, input_id);
 	if (PTR_ERR(iommu) == -EPROBE_DEFER)
 		return -EPROBE_DEFER;
 
-	arch_setup_dma_ops(dev, dma_addr, size,
+	arch_setup_dma_ops(dev, 0, U64_MAX,
 				iommu, attr == DEV_DMA_COHERENT);
 
 	return 0;
diff --git a/include/acpi/acpi_bus.h b/include/acpi/acpi_bus.h
index 0dc1ea0b..e106073 100644
--- a/include/acpi/acpi_bus.h
+++ b/include/acpi/acpi_bus.h
@@ -611,8 +611,7 @@ struct acpi_pci_root {
 int acpi_iommu_fwspec_init(struct device *dev, u32 id,
 			   struct fwnode_handle *fwnode,
 			   const struct iommu_ops *ops);
-int acpi_dma_get_range(struct device *dev, u64 *dma_addr, u64 *offset,
-		       u64 *size);
+int acpi_dma_get_range(struct device *dev, const struct bus_dma_region **map);
 int acpi_dma_configure_id(struct device *dev, enum dev_dma_attr attr,
 			   const u32 *input_id);
 static inline int acpi_dma_configure(struct device *dev,
diff --git a/include/linux/acpi.h b/include/linux/acpi.h
index 44975c1..34e0545 100644
--- a/include/linux/acpi.h
+++ b/include/linux/acpi.h
@@ -280,12 +280,12 @@ int acpi_table_parse_madt(enum acpi_madt_type id,
 
 #ifdef CONFIG_ARM64
 void acpi_numa_gicc_affinity_init(struct acpi_srat_gicc_affinity *pa);
-void acpi_arch_dma_setup(struct device *dev, u64 *dma_addr, u64 *dma_size);
+void acpi_arch_dma_setup(struct device *dev);
 #else
 static inline void
 acpi_numa_gicc_affinity_init(struct acpi_srat_gicc_affinity *pa) { }
 static inline void
-acpi_arch_dma_setup(struct device *dev, u64 *dma_addr, u64 *dma_size) { }
+acpi_arch_dma_setup(struct device *dev) { }
 #endif
 
 int acpi_numa_memory_affinity_init (struct acpi_srat_mem_affinity *ma);
@@ -974,8 +974,7 @@ static inline enum dev_dma_attr acpi_get_dma_attr(struct acpi_device *adev)
 	return DEV_DMA_NOT_SUPPORTED;
 }
 
-static inline int acpi_dma_get_range(struct device *dev, u64 *dma_addr,
-				     u64 *offset, u64 *size)
+static inline int acpi_dma_get_range(struct device *dev, const struct bus_dma_region **map)
 {
 	return -ENODEV;
 }
-- 
1.8.3.1


^ permalink raw reply related	[flat|nested] 10+ messages in thread

* [PATCH V3 2/2] LoongArch: Remove ARCH_HAS_PHYS_TO_DMA
  2022-08-30  3:01 [PATCH V3 0/2] DMA: update acpi_dma_get_range to return dma map regions Jianmin Lv
  2022-08-30  3:01 ` [PATCH V3 1/2] ACPI / scan: Support multiple dma windows with different offsets Jianmin Lv
@ 2022-08-30  3:01 ` Jianmin Lv
  2022-09-05 14:42   ` Huacai Chen
  2022-09-05  2:55 ` [PATCH V3 0/2] DMA: update acpi_dma_get_range to return dma map regions Jianmin Lv
  2 siblings, 1 reply; 10+ messages in thread
From: Jianmin Lv @ 2022-08-30  3:01 UTC (permalink / raw)
  To: lpieralisi, robin.murphy, chenhuacai
  Cc: guohanjun, sudeep.holla, rafael, lenb, robert.moore,
	linux-kernel, linux-acpi, loongarch

Use _DMA defined in ACPI spec for translation between
DMA address and CPU address, and implement acpi_arch_dma_setup
for initializing dev->dma_range_map, where acpi_dma_get_range
is called for parsing _DMA.

e.g.
If we have two dma ranges:
cpu address      dma address    size         offset
0x200080000000   0x2080000000   0x400000000  0x1fe000000000
0x400080000000   0x4080000000   0x400000000  0x3fc000000000

_DMA for pci devices should be declared in host bridge as
flowing:

Name (_DMA, ResourceTemplate() {
        QWordMemory (ResourceProducer,
            PosDecode,
            MinFixed,
            MaxFixed,
            NonCacheable,
            ReadWrite,
            0x0,
            0x4080000000,
            0x447fffffff,
            0x3fc000000000,
            0x400000000,
            ,
            ,
            )

        QWordMemory (ResourceProducer,
            PosDecode,
            MinFixed,
            MaxFixed,
            NonCacheable,
            ReadWrite,
            0x0,
            0x2080000000,
            0x247fffffff,
            0x1fe000000000,
            0x400000000,
            ,
            ,
            )
    })

Signed-off-by: Jianmin Lv <lvjianmin@loongson.cn>
---
 arch/loongarch/Kconfig        |  1 -
 arch/loongarch/kernel/dma.c   | 52 +++++++++++++++++--------------------------
 arch/loongarch/kernel/setup.c |  2 +-
 include/linux/acpi.h          |  9 +++++---
 4 files changed, 28 insertions(+), 36 deletions(-)

diff --git a/arch/loongarch/Kconfig b/arch/loongarch/Kconfig
index b57daee..9dedcf9 100644
--- a/arch/loongarch/Kconfig
+++ b/arch/loongarch/Kconfig
@@ -7,7 +7,6 @@ config LOONGARCH
 	select ARCH_ENABLE_MEMORY_HOTPLUG
 	select ARCH_ENABLE_MEMORY_HOTREMOVE
 	select ARCH_HAS_ACPI_TABLE_UPGRADE	if ACPI
-	select ARCH_HAS_PHYS_TO_DMA
 	select ARCH_HAS_PTE_SPECIAL
 	select ARCH_HAS_TICK_BROADCAST if GENERIC_CLOCKEVENTS_BROADCAST
 	select ARCH_INLINE_READ_LOCK if !PREEMPTION
diff --git a/arch/loongarch/kernel/dma.c b/arch/loongarch/kernel/dma.c
index 8c9b531..7a9c6a9 100644
--- a/arch/loongarch/kernel/dma.c
+++ b/arch/loongarch/kernel/dma.c
@@ -2,39 +2,29 @@
 /*
  * Copyright (C) 2020-2022 Loongson Technology Corporation Limited
  */
-#include <linux/init.h>
+#include <linux/acpi.h>
 #include <linux/dma-direct.h>
-#include <linux/dma-mapping.h>
-#include <linux/dma-map-ops.h>
-#include <linux/swiotlb.h>
 
-#include <asm/bootinfo.h>
-#include <asm/dma.h>
-#include <asm/loongson.h>
-
-/*
- * We extract 4bit node id (bit 44~47) from Loongson-3's
- * 48bit physical address space and embed it into 40bit.
- */
-
-static int node_id_offset;
-
-dma_addr_t phys_to_dma(struct device *dev, phys_addr_t paddr)
-{
-	long nid = (paddr >> 44) & 0xf;
-
-	return ((nid << 44) ^ paddr) | (nid << node_id_offset);
-}
-
-phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
+void acpi_arch_dma_setup(struct device *dev)
 {
-	long nid = (daddr >> node_id_offset) & 0xf;
+	int ret;
+	u64 mask, end = 0;
+	const struct bus_dma_region *map = NULL;
+
+	ret = acpi_dma_get_range(dev, &map);
+	if (!ret && map) {
+		const struct bus_dma_region *r = map;
+
+		for (end = 0; r->size; r++) {
+			if (r->dma_start + r->size - 1 > end)
+				end = r->dma_start + r->size - 1;
+		}
+
+		mask = DMA_BIT_MASK(ilog2(end) + 1);
+		dev->bus_dma_limit = end;
+		dev->dma_range_map = map;
+		dev->coherent_dma_mask = min(dev->coherent_dma_mask, mask);
+		*dev->dma_mask = min(*dev->dma_mask, mask);
+	}
 
-	return ((nid << node_id_offset) ^ daddr) | (nid << 44);
-}
-
-void __init plat_swiotlb_setup(void)
-{
-	swiotlb_init(true, SWIOTLB_VERBOSE);
-	node_id_offset = ((readl(LS7A_DMA_CFG) & LS7A_DMA_NODE_MASK) >> LS7A_DMA_NODE_SHF) + 36;
 }
diff --git a/arch/loongarch/kernel/setup.c b/arch/loongarch/kernel/setup.c
index c74860b..974f085 100644
--- a/arch/loongarch/kernel/setup.c
+++ b/arch/loongarch/kernel/setup.c
@@ -247,7 +247,7 @@ static void __init arch_mem_init(char **cmdline_p)
 	sparse_init();
 	memblock_set_bottom_up(true);
 
-	plat_swiotlb_setup();
+	swiotlb_init(true, SWIOTLB_VERBOSE);
 
 	dma_contiguous_reserve(PFN_PHYS(max_low_pfn));
 
diff --git a/include/linux/acpi.h b/include/linux/acpi.h
index 34e0545..33977b87 100644
--- a/include/linux/acpi.h
+++ b/include/linux/acpi.h
@@ -278,14 +278,17 @@ int acpi_table_parse_madt(enum acpi_madt_type id,
 
 void acpi_numa_x2apic_affinity_init(struct acpi_srat_x2apic_cpu_affinity *pa);
 
+#if defined(CONFIG_ARM64) || defined(CONFIG_LOONGARCH)
+void acpi_arch_dma_setup(struct device *dev);
+#else
+static inline void acpi_arch_dma_setup(struct device *dev) { }
+#endif
+
 #ifdef CONFIG_ARM64
 void acpi_numa_gicc_affinity_init(struct acpi_srat_gicc_affinity *pa);
-void acpi_arch_dma_setup(struct device *dev);
 #else
 static inline void
 acpi_numa_gicc_affinity_init(struct acpi_srat_gicc_affinity *pa) { }
-static inline void
-acpi_arch_dma_setup(struct device *dev) { }
 #endif
 
 int acpi_numa_memory_affinity_init (struct acpi_srat_mem_affinity *ma);
-- 
1.8.3.1


^ permalink raw reply related	[flat|nested] 10+ messages in thread

* Re: [PATCH V3 0/2] DMA: update acpi_dma_get_range to return dma map regions
  2022-08-30  3:01 [PATCH V3 0/2] DMA: update acpi_dma_get_range to return dma map regions Jianmin Lv
  2022-08-30  3:01 ` [PATCH V3 1/2] ACPI / scan: Support multiple dma windows with different offsets Jianmin Lv
  2022-08-30  3:01 ` [PATCH V3 2/2] LoongArch: Remove ARCH_HAS_PHYS_TO_DMA Jianmin Lv
@ 2022-09-05  2:55 ` Jianmin Lv
  2 siblings, 0 replies; 10+ messages in thread
From: Jianmin Lv @ 2022-09-05  2:55 UTC (permalink / raw)
  To: lpieralisi, robin.murphy, chenhuacai
  Cc: guohanjun, sudeep.holla, rafael, lenb, robert.moore,
	linux-kernel, linux-acpi, loongarch

Hi, all,

Is there anything else in this patch series that needs to be modified?

Thanks.
Jianmin.


On 2022/8/30 上午11:01, Jianmin Lv wrote:
> The patch series changed acpi_dma_get_range to return dma regions
> as of_dma_get_range, so that dev->dma_range_map can be initialized
> conveniently.
> 
> And acpi_arch_dma_setup for ARM64 is changed wih removing dma_base
> and size from it's parameters.
> 
> Remove ARCH_HAS_PHYS_TO_DMA for LoongArch and use generic
> phys_to_dma/dma_to_phys in include/linux/dma-direct.h.
> 
> V1 -> V2
> - Removed dma_base and size from acpi_arch_dma_setup' parameters
> - Add patch to remove ARCH_HAS_PHYS_TO_DMA for LoongArch
> 
> V2 -> V3
> - Add kerneldoc for acpi_dma_get_range changing
> - Remove redundant code in acpi_arch_dma_setup, and check map
> 
> 
> Jianmin Lv (2):
>    ACPI / scan: Support multiple dma windows with different offsets
>    LoongArch: Remove ARCH_HAS_PHYS_TO_DMA
> 
>   arch/loongarch/Kconfig        |  1 -
>   arch/loongarch/kernel/dma.c   | 52 +++++++++++++++++-------------------------
>   arch/loongarch/kernel/setup.c |  2 +-
>   drivers/acpi/arm64/dma.c      | 29 ++++++++++++++---------
>   drivers/acpi/scan.c           | 53 +++++++++++++++++++------------------------
>   include/acpi/acpi_bus.h       |  3 +--
>   include/linux/acpi.h          | 12 ++++++----
>   7 files changed, 71 insertions(+), 81 deletions(-)
> 


^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH V3 1/2] ACPI / scan: Support multiple dma windows with different offsets
  2022-08-30  3:01 ` [PATCH V3 1/2] ACPI / scan: Support multiple dma windows with different offsets Jianmin Lv
@ 2022-09-05 12:20   ` Robin Murphy
  2022-09-06 12:40     ` Jianmin Lv
  0 siblings, 1 reply; 10+ messages in thread
From: Robin Murphy @ 2022-09-05 12:20 UTC (permalink / raw)
  To: Jianmin Lv, lpieralisi, chenhuacai
  Cc: guohanjun, sudeep.holla, rafael, lenb, robert.moore,
	linux-kernel, linux-acpi, loongarch

On 2022-08-30 04:01, Jianmin Lv wrote:
> For DT, of_dma_get_range returns bus_dma_region typed dma regions,
> which makes multiple dma windows with different offset available
> for translation between dma address and cpu address.
> 
> But for ACPI, acpi_dma_get_range doesn't return similar dma regions,
> causing no path for setting dev->dma_range_map conveniently. So the
> patch changes acpi_dma_get_range and returns bus_dma_region typed
> dma regions according to of_dma_get_range.
> 
> After changing acpi_dma_get_range, acpi_arch_dma_setup is changed for
> ARM64, where original dma_addr and size are removed as these
> arguments are now redundant, and pass 0 and U64_MAX for dma_base
> and size of arch_setup_dma_ops, so this is a simplification consistent
> with what other ACPI architectures also pass to iommu_setup_dma_ops().

Other than a micro-nit that acpi_dma_get_range() could probably use 
resource_size(),

Reviewed-by: Robin Murphy <robin.murphy@arm.com>

It took me longer than I care to admit to figure out where the implicit 
declaration of struct bus_dma_region in the scope of acpi.h and 
acpi_bus.h comes from, but in the end I guess it's sufficiently 
well-defined by the C spec to be reliable.

Thanks for getting this done!

Robin.

> Signed-off-by: Jianmin Lv <lvjianmin@loongson.cn>
> ---
>   drivers/acpi/arm64/dma.c | 29 ++++++++++++++++----------
>   drivers/acpi/scan.c      | 53 +++++++++++++++++++++---------------------------
>   include/acpi/acpi_bus.h  |  3 +--
>   include/linux/acpi.h     |  7 +++----
>   4 files changed, 45 insertions(+), 47 deletions(-)
> 
> diff --git a/drivers/acpi/arm64/dma.c b/drivers/acpi/arm64/dma.c
> index f16739a..1ef8e7d 100644
> --- a/drivers/acpi/arm64/dma.c
> +++ b/drivers/acpi/arm64/dma.c
> @@ -4,11 +4,12 @@
>   #include <linux/device.h>
>   #include <linux/dma-direct.h>
>   
> -void acpi_arch_dma_setup(struct device *dev, u64 *dma_addr, u64 *dma_size)
> +void acpi_arch_dma_setup(struct device *dev)
>   {
>   	int ret;
>   	u64 end, mask;
> -	u64 dmaaddr = 0, size = 0, offset = 0;
> +	u64 size = 0;
> +	const struct bus_dma_region *map = NULL;
>   
>   	/*
>   	 * If @dev is expected to be DMA-capable then the bus code that created
> @@ -26,25 +27,31 @@ void acpi_arch_dma_setup(struct device *dev, u64 *dma_addr, u64 *dma_size)
>   	else
>   		size = 1ULL << 32;
>   
> -	ret = acpi_dma_get_range(dev, &dmaaddr, &offset, &size);
> +	ret = acpi_dma_get_range(dev, &map);
> +	if (!ret && map) {
> +		const struct bus_dma_region *r = map;
> +
> +		for (end = 0; r->size; r++) {
> +			if (r->dma_start + r->size - 1 > end)
> +				end = r->dma_start + r->size - 1;
> +		}
> +
> +		size = end + 1;
> +		dev->dma_range_map = map;
> +	}
> +
>   	if (ret == -ENODEV)
>   		ret = iort_dma_get_ranges(dev, &size);
> +
>   	if (!ret) {
>   		/*
>   		 * Limit coherent and dma mask based on size retrieved from
>   		 * firmware.
>   		 */
> -		end = dmaaddr + size - 1;
> +		end = size - 1;
>   		mask = DMA_BIT_MASK(ilog2(end) + 1);
>   		dev->bus_dma_limit = end;
>   		dev->coherent_dma_mask = min(dev->coherent_dma_mask, mask);
>   		*dev->dma_mask = min(*dev->dma_mask, mask);
>   	}
> -
> -	*dma_addr = dmaaddr;
> -	*dma_size = size;
> -
> -	ret = dma_direct_set_offset(dev, dmaaddr + offset, dmaaddr, size);
> -
> -	dev_dbg(dev, "dma_offset(%#08llx)%s\n", offset, ret ? " failed!" : "");
>   }
> diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
> index 762b61f..8c0c2ca 100644
> --- a/drivers/acpi/scan.c
> +++ b/drivers/acpi/scan.c
> @@ -20,6 +20,7 @@
>   #include <linux/platform_data/x86/apple.h>
>   #include <linux/pgtable.h>
>   #include <linux/crc32.h>
> +#include <linux/dma-direct.h>
>   
>   #include "internal.h"
>   
> @@ -1482,25 +1483,21 @@ enum dev_dma_attr acpi_get_dma_attr(struct acpi_device *adev)
>    * acpi_dma_get_range() - Get device DMA parameters.
>    *
>    * @dev: device to configure
> - * @dma_addr: pointer device DMA address result
> - * @offset: pointer to the DMA offset result
> - * @size: pointer to DMA range size result
> + * @map: pointer to DMA ranges result
>    *
> - * Evaluate DMA regions and return respectively DMA region start, offset
> - * and size in dma_addr, offset and size on parsing success; it does not
> - * update the passed in values on failure.
> + * Evaluate DMA regions and return pointer to DMA regions on
> + * parsing success; it does not update the passed in values on failure.
>    *
>    * Return 0 on success, < 0 on failure.
>    */
> -int acpi_dma_get_range(struct device *dev, u64 *dma_addr, u64 *offset,
> -		       u64 *size)
> +int acpi_dma_get_range(struct device *dev, const struct bus_dma_region **map)
>   {
>   	struct acpi_device *adev;
>   	LIST_HEAD(list);
>   	struct resource_entry *rentry;
>   	int ret;
>   	struct device *dma_dev = dev;
> -	u64 len, dma_start = U64_MAX, dma_end = 0, dma_offset = 0;
> +	struct bus_dma_region *r;
>   
>   	/*
>   	 * Walk the device tree chasing an ACPI companion with a _DMA
> @@ -1525,31 +1522,28 @@ int acpi_dma_get_range(struct device *dev, u64 *dma_addr, u64 *offset,
>   
>   	ret = acpi_dev_get_dma_resources(adev, &list);
>   	if (ret > 0) {
> +		r = kcalloc(ret + 1, sizeof(*r), GFP_KERNEL);
> +		if (!r) {
> +			ret = -ENOMEM;
> +			goto out;
> +		}
> +
> +		*map = r;
> +
>   		list_for_each_entry(rentry, &list, node) {
> -			if (dma_offset && rentry->offset != dma_offset) {
> +			if (rentry->res->start >= rentry->res->end) {
>   				ret = -EINVAL;
> -				dev_warn(dma_dev, "Can't handle multiple windows with different offsets\n");
> +				dev_dbg(dma_dev, "Invalid DMA regions configuration\n");
>   				goto out;
>   			}
> -			dma_offset = rentry->offset;
>   
> -			/* Take lower and upper limits */
> -			if (rentry->res->start < dma_start)
> -				dma_start = rentry->res->start;
> -			if (rentry->res->end > dma_end)
> -				dma_end = rentry->res->end;
> -		}
> -
> -		if (dma_start >= dma_end) {
> -			ret = -EINVAL;
> -			dev_dbg(dma_dev, "Invalid DMA regions configuration\n");
> -			goto out;
> +			r->cpu_start = rentry->res->start;
> +			r->dma_start = rentry->res->start - rentry->offset;
> +			r->size = rentry->res->end - rentry->res->start + 1;
> +			r->offset = rentry->offset;
> +			r++;
>   		}
>   
> -		*dma_addr = dma_start - dma_offset;
> -		len = dma_end - dma_start;
> -		*size = max(len, len + 1);
> -		*offset = dma_offset;
>   	}
>    out:
>   	acpi_dev_free_resource_list(&list);
> @@ -1639,20 +1633,19 @@ int acpi_dma_configure_id(struct device *dev, enum dev_dma_attr attr,
>   			  const u32 *input_id)
>   {
>   	const struct iommu_ops *iommu;
> -	u64 dma_addr = 0, size = 0;
>   
>   	if (attr == DEV_DMA_NOT_SUPPORTED) {
>   		set_dma_ops(dev, &dma_dummy_ops);
>   		return 0;
>   	}
>   
> -	acpi_arch_dma_setup(dev, &dma_addr, &size);
> +	acpi_arch_dma_setup(dev);
>   
>   	iommu = acpi_iommu_configure_id(dev, input_id);
>   	if (PTR_ERR(iommu) == -EPROBE_DEFER)
>   		return -EPROBE_DEFER;
>   
> -	arch_setup_dma_ops(dev, dma_addr, size,
> +	arch_setup_dma_ops(dev, 0, U64_MAX,
>   				iommu, attr == DEV_DMA_COHERENT);
>   
>   	return 0;
> diff --git a/include/acpi/acpi_bus.h b/include/acpi/acpi_bus.h
> index 0dc1ea0b..e106073 100644
> --- a/include/acpi/acpi_bus.h
> +++ b/include/acpi/acpi_bus.h
> @@ -611,8 +611,7 @@ struct acpi_pci_root {
>   int acpi_iommu_fwspec_init(struct device *dev, u32 id,
>   			   struct fwnode_handle *fwnode,
>   			   const struct iommu_ops *ops);
> -int acpi_dma_get_range(struct device *dev, u64 *dma_addr, u64 *offset,
> -		       u64 *size);
> +int acpi_dma_get_range(struct device *dev, const struct bus_dma_region **map);
>   int acpi_dma_configure_id(struct device *dev, enum dev_dma_attr attr,
>   			   const u32 *input_id);
>   static inline int acpi_dma_configure(struct device *dev,
> diff --git a/include/linux/acpi.h b/include/linux/acpi.h
> index 44975c1..34e0545 100644
> --- a/include/linux/acpi.h
> +++ b/include/linux/acpi.h
> @@ -280,12 +280,12 @@ int acpi_table_parse_madt(enum acpi_madt_type id,
>   
>   #ifdef CONFIG_ARM64
>   void acpi_numa_gicc_affinity_init(struct acpi_srat_gicc_affinity *pa);
> -void acpi_arch_dma_setup(struct device *dev, u64 *dma_addr, u64 *dma_size);
> +void acpi_arch_dma_setup(struct device *dev);
>   #else
>   static inline void
>   acpi_numa_gicc_affinity_init(struct acpi_srat_gicc_affinity *pa) { }
>   static inline void
> -acpi_arch_dma_setup(struct device *dev, u64 *dma_addr, u64 *dma_size) { }
> +acpi_arch_dma_setup(struct device *dev) { }
>   #endif
>   
>   int acpi_numa_memory_affinity_init (struct acpi_srat_mem_affinity *ma);
> @@ -974,8 +974,7 @@ static inline enum dev_dma_attr acpi_get_dma_attr(struct acpi_device *adev)
>   	return DEV_DMA_NOT_SUPPORTED;
>   }
>   
> -static inline int acpi_dma_get_range(struct device *dev, u64 *dma_addr,
> -				     u64 *offset, u64 *size)
> +static inline int acpi_dma_get_range(struct device *dev, const struct bus_dma_region **map)
>   {
>   	return -ENODEV;
>   }

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH V3 2/2] LoongArch: Remove ARCH_HAS_PHYS_TO_DMA
  2022-08-30  3:01 ` [PATCH V3 2/2] LoongArch: Remove ARCH_HAS_PHYS_TO_DMA Jianmin Lv
@ 2022-09-05 14:42   ` Huacai Chen
  2022-09-06 12:34     ` Jianmin Lv
  0 siblings, 1 reply; 10+ messages in thread
From: Huacai Chen @ 2022-09-05 14:42 UTC (permalink / raw)
  To: Jianmin Lv
  Cc: lpieralisi, Robin Murphy, Huacai Chen, Hanjun Guo, Sudeep Holla,
	Rafael J . Wysocki, Len Brown, Robert Moore, LKML,
	ACPI Devel Maling List, loongarch

Hi, Jianmin,

The title can be "LoongArch: Use acpi_arch_dma_setup() and remove
ARCH_HAS_PHYS_TO_DMA", and please use resource_size() as arm64.

Acked-by: Huacai Chen <chenhuacai@loongson.cn>


Huacai


On Tue, Aug 30, 2022 at 11:01 AM Jianmin Lv <lvjianmin@loongson.cn> wrote:
>
> Use _DMA defined in ACPI spec for translation between
> DMA address and CPU address, and implement acpi_arch_dma_setup
> for initializing dev->dma_range_map, where acpi_dma_get_range
> is called for parsing _DMA.
>
> e.g.
> If we have two dma ranges:
> cpu address      dma address    size         offset
> 0x200080000000   0x2080000000   0x400000000  0x1fe000000000
> 0x400080000000   0x4080000000   0x400000000  0x3fc000000000
>
> _DMA for pci devices should be declared in host bridge as
> flowing:
>
> Name (_DMA, ResourceTemplate() {
>         QWordMemory (ResourceProducer,
>             PosDecode,
>             MinFixed,
>             MaxFixed,
>             NonCacheable,
>             ReadWrite,
>             0x0,
>             0x4080000000,
>             0x447fffffff,
>             0x3fc000000000,
>             0x400000000,
>             ,
>             ,
>             )
>
>         QWordMemory (ResourceProducer,
>             PosDecode,
>             MinFixed,
>             MaxFixed,
>             NonCacheable,
>             ReadWrite,
>             0x0,
>             0x2080000000,
>             0x247fffffff,
>             0x1fe000000000,
>             0x400000000,
>             ,
>             ,
>             )
>     })
>
> Signed-off-by: Jianmin Lv <lvjianmin@loongson.cn>
> ---
>  arch/loongarch/Kconfig        |  1 -
>  arch/loongarch/kernel/dma.c   | 52 +++++++++++++++++--------------------------
>  arch/loongarch/kernel/setup.c |  2 +-
>  include/linux/acpi.h          |  9 +++++---
>  4 files changed, 28 insertions(+), 36 deletions(-)
>
> diff --git a/arch/loongarch/Kconfig b/arch/loongarch/Kconfig
> index b57daee..9dedcf9 100644
> --- a/arch/loongarch/Kconfig
> +++ b/arch/loongarch/Kconfig
> @@ -7,7 +7,6 @@ config LOONGARCH
>         select ARCH_ENABLE_MEMORY_HOTPLUG
>         select ARCH_ENABLE_MEMORY_HOTREMOVE
>         select ARCH_HAS_ACPI_TABLE_UPGRADE      if ACPI
> -       select ARCH_HAS_PHYS_TO_DMA
>         select ARCH_HAS_PTE_SPECIAL
>         select ARCH_HAS_TICK_BROADCAST if GENERIC_CLOCKEVENTS_BROADCAST
>         select ARCH_INLINE_READ_LOCK if !PREEMPTION
> diff --git a/arch/loongarch/kernel/dma.c b/arch/loongarch/kernel/dma.c
> index 8c9b531..7a9c6a9 100644
> --- a/arch/loongarch/kernel/dma.c
> +++ b/arch/loongarch/kernel/dma.c
> @@ -2,39 +2,29 @@
>  /*
>   * Copyright (C) 2020-2022 Loongson Technology Corporation Limited
>   */
> -#include <linux/init.h>
> +#include <linux/acpi.h>
>  #include <linux/dma-direct.h>
> -#include <linux/dma-mapping.h>
> -#include <linux/dma-map-ops.h>
> -#include <linux/swiotlb.h>
>
> -#include <asm/bootinfo.h>
> -#include <asm/dma.h>
> -#include <asm/loongson.h>
> -
> -/*
> - * We extract 4bit node id (bit 44~47) from Loongson-3's
> - * 48bit physical address space and embed it into 40bit.
> - */
> -
> -static int node_id_offset;
> -
> -dma_addr_t phys_to_dma(struct device *dev, phys_addr_t paddr)
> -{
> -       long nid = (paddr >> 44) & 0xf;
> -
> -       return ((nid << 44) ^ paddr) | (nid << node_id_offset);
> -}
> -
> -phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
> +void acpi_arch_dma_setup(struct device *dev)
>  {
> -       long nid = (daddr >> node_id_offset) & 0xf;
> +       int ret;
> +       u64 mask, end = 0;
> +       const struct bus_dma_region *map = NULL;
> +
> +       ret = acpi_dma_get_range(dev, &map);
> +       if (!ret && map) {
> +               const struct bus_dma_region *r = map;
> +
> +               for (end = 0; r->size; r++) {
> +                       if (r->dma_start + r->size - 1 > end)
> +                               end = r->dma_start + r->size - 1;
> +               }
> +
> +               mask = DMA_BIT_MASK(ilog2(end) + 1);
> +               dev->bus_dma_limit = end;
> +               dev->dma_range_map = map;
> +               dev->coherent_dma_mask = min(dev->coherent_dma_mask, mask);
> +               *dev->dma_mask = min(*dev->dma_mask, mask);
> +       }
>
> -       return ((nid << node_id_offset) ^ daddr) | (nid << 44);
> -}
> -
> -void __init plat_swiotlb_setup(void)
> -{
> -       swiotlb_init(true, SWIOTLB_VERBOSE);
> -       node_id_offset = ((readl(LS7A_DMA_CFG) & LS7A_DMA_NODE_MASK) >> LS7A_DMA_NODE_SHF) + 36;
>  }
> diff --git a/arch/loongarch/kernel/setup.c b/arch/loongarch/kernel/setup.c
> index c74860b..974f085 100644
> --- a/arch/loongarch/kernel/setup.c
> +++ b/arch/loongarch/kernel/setup.c
> @@ -247,7 +247,7 @@ static void __init arch_mem_init(char **cmdline_p)
>         sparse_init();
>         memblock_set_bottom_up(true);
>
> -       plat_swiotlb_setup();
> +       swiotlb_init(true, SWIOTLB_VERBOSE);
>
>         dma_contiguous_reserve(PFN_PHYS(max_low_pfn));
>
> diff --git a/include/linux/acpi.h b/include/linux/acpi.h
> index 34e0545..33977b87 100644
> --- a/include/linux/acpi.h
> +++ b/include/linux/acpi.h
> @@ -278,14 +278,17 @@ int acpi_table_parse_madt(enum acpi_madt_type id,
>
>  void acpi_numa_x2apic_affinity_init(struct acpi_srat_x2apic_cpu_affinity *pa);
>
> +#if defined(CONFIG_ARM64) || defined(CONFIG_LOONGARCH)
> +void acpi_arch_dma_setup(struct device *dev);
> +#else
> +static inline void acpi_arch_dma_setup(struct device *dev) { }
> +#endif
> +
>  #ifdef CONFIG_ARM64
>  void acpi_numa_gicc_affinity_init(struct acpi_srat_gicc_affinity *pa);
> -void acpi_arch_dma_setup(struct device *dev);
>  #else
>  static inline void
>  acpi_numa_gicc_affinity_init(struct acpi_srat_gicc_affinity *pa) { }
> -static inline void
> -acpi_arch_dma_setup(struct device *dev) { }
>  #endif
>
>  int acpi_numa_memory_affinity_init (struct acpi_srat_mem_affinity *ma);
> --
> 1.8.3.1
>
>

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH V3 2/2] LoongArch: Remove ARCH_HAS_PHYS_TO_DMA
  2022-09-05 14:42   ` Huacai Chen
@ 2022-09-06 12:34     ` Jianmin Lv
  0 siblings, 0 replies; 10+ messages in thread
From: Jianmin Lv @ 2022-09-06 12:34 UTC (permalink / raw)
  To: Huacai Chen
  Cc: lpieralisi, Robin Murphy, Huacai Chen, Hanjun Guo, Sudeep Holla,
	Rafael J . Wysocki, Len Brown, Robert Moore, LKML,
	ACPI Devel Maling List, loongarch

Hi, Huacai

Ok, thanks, I'll change title in V4. There is *not* size calculation in
this patch, so the patch does not need resource_size() as in 
acpi_dma_get_range().

On 2022/9/5 下午10:42, Huacai Chen wrote:
> Hi, Jianmin,
> 
> The title can be "LoongArch: Use acpi_arch_dma_setup() and remove
> ARCH_HAS_PHYS_TO_DMA", and please use resource_size() as arm64.
> 
> Acked-by: Huacai Chen <chenhuacai@loongson.cn>
> 
> 
> Huacai
> 
> 
> On Tue, Aug 30, 2022 at 11:01 AM Jianmin Lv <lvjianmin@loongson.cn> wrote:
>>
>> Use _DMA defined in ACPI spec for translation between
>> DMA address and CPU address, and implement acpi_arch_dma_setup
>> for initializing dev->dma_range_map, where acpi_dma_get_range
>> is called for parsing _DMA.
>>
>> e.g.
>> If we have two dma ranges:
>> cpu address      dma address    size         offset
>> 0x200080000000   0x2080000000   0x400000000  0x1fe000000000
>> 0x400080000000   0x4080000000   0x400000000  0x3fc000000000
>>
>> _DMA for pci devices should be declared in host bridge as
>> flowing:
>>
>> Name (_DMA, ResourceTemplate() {
>>          QWordMemory (ResourceProducer,
>>              PosDecode,
>>              MinFixed,
>>              MaxFixed,
>>              NonCacheable,
>>              ReadWrite,
>>              0x0,
>>              0x4080000000,
>>              0x447fffffff,
>>              0x3fc000000000,
>>              0x400000000,
>>              ,
>>              ,
>>              )
>>
>>          QWordMemory (ResourceProducer,
>>              PosDecode,
>>              MinFixed,
>>              MaxFixed,
>>              NonCacheable,
>>              ReadWrite,
>>              0x0,
>>              0x2080000000,
>>              0x247fffffff,
>>              0x1fe000000000,
>>              0x400000000,
>>              ,
>>              ,
>>              )
>>      })
>>
>> Signed-off-by: Jianmin Lv <lvjianmin@loongson.cn>
>> ---
>>   arch/loongarch/Kconfig        |  1 -
>>   arch/loongarch/kernel/dma.c   | 52 +++++++++++++++++--------------------------
>>   arch/loongarch/kernel/setup.c |  2 +-
>>   include/linux/acpi.h          |  9 +++++---
>>   4 files changed, 28 insertions(+), 36 deletions(-)
>>
>> diff --git a/arch/loongarch/Kconfig b/arch/loongarch/Kconfig
>> index b57daee..9dedcf9 100644
>> --- a/arch/loongarch/Kconfig
>> +++ b/arch/loongarch/Kconfig
>> @@ -7,7 +7,6 @@ config LOONGARCH
>>          select ARCH_ENABLE_MEMORY_HOTPLUG
>>          select ARCH_ENABLE_MEMORY_HOTREMOVE
>>          select ARCH_HAS_ACPI_TABLE_UPGRADE      if ACPI
>> -       select ARCH_HAS_PHYS_TO_DMA
>>          select ARCH_HAS_PTE_SPECIAL
>>          select ARCH_HAS_TICK_BROADCAST if GENERIC_CLOCKEVENTS_BROADCAST
>>          select ARCH_INLINE_READ_LOCK if !PREEMPTION
>> diff --git a/arch/loongarch/kernel/dma.c b/arch/loongarch/kernel/dma.c
>> index 8c9b531..7a9c6a9 100644
>> --- a/arch/loongarch/kernel/dma.c
>> +++ b/arch/loongarch/kernel/dma.c
>> @@ -2,39 +2,29 @@
>>   /*
>>    * Copyright (C) 2020-2022 Loongson Technology Corporation Limited
>>    */
>> -#include <linux/init.h>
>> +#include <linux/acpi.h>
>>   #include <linux/dma-direct.h>
>> -#include <linux/dma-mapping.h>
>> -#include <linux/dma-map-ops.h>
>> -#include <linux/swiotlb.h>
>>
>> -#include <asm/bootinfo.h>
>> -#include <asm/dma.h>
>> -#include <asm/loongson.h>
>> -
>> -/*
>> - * We extract 4bit node id (bit 44~47) from Loongson-3's
>> - * 48bit physical address space and embed it into 40bit.
>> - */
>> -
>> -static int node_id_offset;
>> -
>> -dma_addr_t phys_to_dma(struct device *dev, phys_addr_t paddr)
>> -{
>> -       long nid = (paddr >> 44) & 0xf;
>> -
>> -       return ((nid << 44) ^ paddr) | (nid << node_id_offset);
>> -}
>> -
>> -phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
>> +void acpi_arch_dma_setup(struct device *dev)
>>   {
>> -       long nid = (daddr >> node_id_offset) & 0xf;
>> +       int ret;
>> +       u64 mask, end = 0;
>> +       const struct bus_dma_region *map = NULL;
>> +
>> +       ret = acpi_dma_get_range(dev, &map);
>> +       if (!ret && map) {
>> +               const struct bus_dma_region *r = map;
>> +
>> +               for (end = 0; r->size; r++) {
>> +                       if (r->dma_start + r->size - 1 > end)
>> +                               end = r->dma_start + r->size - 1;
>> +               }
>> +
>> +               mask = DMA_BIT_MASK(ilog2(end) + 1);
>> +               dev->bus_dma_limit = end;
>> +               dev->dma_range_map = map;
>> +               dev->coherent_dma_mask = min(dev->coherent_dma_mask, mask);
>> +               *dev->dma_mask = min(*dev->dma_mask, mask);
>> +       }
>>
>> -       return ((nid << node_id_offset) ^ daddr) | (nid << 44);
>> -}
>> -
>> -void __init plat_swiotlb_setup(void)
>> -{
>> -       swiotlb_init(true, SWIOTLB_VERBOSE);
>> -       node_id_offset = ((readl(LS7A_DMA_CFG) & LS7A_DMA_NODE_MASK) >> LS7A_DMA_NODE_SHF) + 36;
>>   }
>> diff --git a/arch/loongarch/kernel/setup.c b/arch/loongarch/kernel/setup.c
>> index c74860b..974f085 100644
>> --- a/arch/loongarch/kernel/setup.c
>> +++ b/arch/loongarch/kernel/setup.c
>> @@ -247,7 +247,7 @@ static void __init arch_mem_init(char **cmdline_p)
>>          sparse_init();
>>          memblock_set_bottom_up(true);
>>
>> -       plat_swiotlb_setup();
>> +       swiotlb_init(true, SWIOTLB_VERBOSE);
>>
>>          dma_contiguous_reserve(PFN_PHYS(max_low_pfn));
>>
>> diff --git a/include/linux/acpi.h b/include/linux/acpi.h
>> index 34e0545..33977b87 100644
>> --- a/include/linux/acpi.h
>> +++ b/include/linux/acpi.h
>> @@ -278,14 +278,17 @@ int acpi_table_parse_madt(enum acpi_madt_type id,
>>
>>   void acpi_numa_x2apic_affinity_init(struct acpi_srat_x2apic_cpu_affinity *pa);
>>
>> +#if defined(CONFIG_ARM64) || defined(CONFIG_LOONGARCH)
>> +void acpi_arch_dma_setup(struct device *dev);
>> +#else
>> +static inline void acpi_arch_dma_setup(struct device *dev) { }
>> +#endif
>> +
>>   #ifdef CONFIG_ARM64
>>   void acpi_numa_gicc_affinity_init(struct acpi_srat_gicc_affinity *pa);
>> -void acpi_arch_dma_setup(struct device *dev);
>>   #else
>>   static inline void
>>   acpi_numa_gicc_affinity_init(struct acpi_srat_gicc_affinity *pa) { }
>> -static inline void
>> -acpi_arch_dma_setup(struct device *dev) { }
>>   #endif
>>
>>   int acpi_numa_memory_affinity_init (struct acpi_srat_mem_affinity *ma);
>> --
>> 1.8.3.1
>>
>>


^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH V3 1/2] ACPI / scan: Support multiple dma windows with different offsets
  2022-09-05 12:20   ` Robin Murphy
@ 2022-09-06 12:40     ` Jianmin Lv
  2022-09-09  8:04       ` Lorenzo Pieralisi
  0 siblings, 1 reply; 10+ messages in thread
From: Jianmin Lv @ 2022-09-06 12:40 UTC (permalink / raw)
  To: Robin Murphy, lpieralisi, chenhuacai
  Cc: guohanjun, sudeep.holla, rafael, lenb, robert.moore,
	linux-kernel, linux-acpi, loongarch



On 2022/9/5 下午8:20, Robin Murphy wrote:
> On 2022-08-30 04:01, Jianmin Lv wrote:
>> For DT, of_dma_get_range returns bus_dma_region typed dma regions,
>> which makes multiple dma windows with different offset available
>> for translation between dma address and cpu address.
>>
>> But for ACPI, acpi_dma_get_range doesn't return similar dma regions,
>> causing no path for setting dev->dma_range_map conveniently. So the
>> patch changes acpi_dma_get_range and returns bus_dma_region typed
>> dma regions according to of_dma_get_range.
>>
>> After changing acpi_dma_get_range, acpi_arch_dma_setup is changed for
>> ARM64, where original dma_addr and size are removed as these
>> arguments are now redundant, and pass 0 and U64_MAX for dma_base
>> and size of arch_setup_dma_ops, so this is a simplification consistent
>> with what other ACPI architectures also pass to iommu_setup_dma_ops().
> 
> Other than a micro-nit that acpi_dma_get_range() could probably use 
> resource_size(),
> 

Ok, thanks, I'll use resource_size() in acpi_dma_get_range().

> Reviewed-by: Robin Murphy <robin.murphy@arm.com>
> 
> It took me longer than I care to admit to figure out where the implicit 
> declaration of struct bus_dma_region in the scope of acpi.h and 
> acpi_bus.h comes from, but in the end I guess it's sufficiently 
> well-defined by the C spec to be reliable.
> 
> Thanks for getting this done!
> 

It's a pleasure!


> Robin.
> 

[...]



^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH V3 1/2] ACPI / scan: Support multiple dma windows with different offsets
  2022-09-06 12:40     ` Jianmin Lv
@ 2022-09-09  8:04       ` Lorenzo Pieralisi
  2022-09-09  9:23         ` Jianmin Lv
  0 siblings, 1 reply; 10+ messages in thread
From: Lorenzo Pieralisi @ 2022-09-09  8:04 UTC (permalink / raw)
  To: Jianmin Lv
  Cc: Robin Murphy, chenhuacai, guohanjun, sudeep.holla, rafael, lenb,
	robert.moore, linux-kernel, linux-acpi, loongarch

On Tue, Sep 06, 2022 at 08:40:48PM +0800, Jianmin Lv wrote:
> 
> 
> On 2022/9/5 下午8:20, Robin Murphy wrote:
> > On 2022-08-30 04:01, Jianmin Lv wrote:
> > > For DT, of_dma_get_range returns bus_dma_region typed dma regions,
> > > which makes multiple dma windows with different offset available
> > > for translation between dma address and cpu address.
> > > 
> > > But for ACPI, acpi_dma_get_range doesn't return similar dma regions,
> > > causing no path for setting dev->dma_range_map conveniently. So the
> > > patch changes acpi_dma_get_range and returns bus_dma_region typed
> > > dma regions according to of_dma_get_range.
> > > 
> > > After changing acpi_dma_get_range, acpi_arch_dma_setup is changed for
> > > ARM64, where original dma_addr and size are removed as these
> > > arguments are now redundant, and pass 0 and U64_MAX for dma_base
> > > and size of arch_setup_dma_ops, so this is a simplification consistent
> > > with what other ACPI architectures also pass to iommu_setup_dma_ops().
> > 
> > Other than a micro-nit that acpi_dma_get_range() could probably use
> > resource_size(),
> > 
> 
> Ok, thanks, I'll use resource_size() in acpi_dma_get_range().

Are you reposting this shortly ? We are almost at -rc5, it would
be good if we can proceed promptly.

Thanks,
Lorenzo

> > Reviewed-by: Robin Murphy <robin.murphy@arm.com>
> > 
> > It took me longer than I care to admit to figure out where the implicit
> > declaration of struct bus_dma_region in the scope of acpi.h and
> > acpi_bus.h comes from, but in the end I guess it's sufficiently
> > well-defined by the C spec to be reliable.
> > 
> > Thanks for getting this done!
> > 
> 
> It's a pleasure!
> 
> 
> > Robin.
> > 
> 
> [...]
> 
> 

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH V3 1/2] ACPI / scan: Support multiple dma windows with different offsets
  2022-09-09  8:04       ` Lorenzo Pieralisi
@ 2022-09-09  9:23         ` Jianmin Lv
  0 siblings, 0 replies; 10+ messages in thread
From: Jianmin Lv @ 2022-09-09  9:23 UTC (permalink / raw)
  To: Lorenzo Pieralisi
  Cc: Robin Murphy, chenhuacai, guohanjun, sudeep.holla, rafael, lenb,
	robert.moore, linux-kernel, linux-acpi, loongarch



On 2022/9/9 下午4:04, Lorenzo Pieralisi wrote:
> On Tue, Sep 06, 2022 at 08:40:48PM +0800, Jianmin Lv wrote:
>>
>>
>> On 2022/9/5 下午8:20, Robin Murphy wrote:
>>> On 2022-08-30 04:01, Jianmin Lv wrote:
>>>> For DT, of_dma_get_range returns bus_dma_region typed dma regions,
>>>> which makes multiple dma windows with different offset available
>>>> for translation between dma address and cpu address.
>>>>
>>>> But for ACPI, acpi_dma_get_range doesn't return similar dma regions,
>>>> causing no path for setting dev->dma_range_map conveniently. So the
>>>> patch changes acpi_dma_get_range and returns bus_dma_region typed
>>>> dma regions according to of_dma_get_range.
>>>>
>>>> After changing acpi_dma_get_range, acpi_arch_dma_setup is changed for
>>>> ARM64, where original dma_addr and size are removed as these
>>>> arguments are now redundant, and pass 0 and U64_MAX for dma_base
>>>> and size of arch_setup_dma_ops, so this is a simplification consistent
>>>> with what other ACPI architectures also pass to iommu_setup_dma_ops().
>>>
>>> Other than a micro-nit that acpi_dma_get_range() could probably use
>>> resource_size(),
>>>
>>
>> Ok, thanks, I'll use resource_size() in acpi_dma_get_range().
> 
> Are you reposting this shortly ? We are almost at -rc5, it would
> be good if we can proceed promptly.
> 
> Thanks,
> Lorenzo
> 
Ok, I'll send V4 today.

Thanks,
Jianmin

>>> Reviewed-by: Robin Murphy <robin.murphy@arm.com>
>>>
>>> It took me longer than I care to admit to figure out where the implicit
>>> declaration of struct bus_dma_region in the scope of acpi.h and
>>> acpi_bus.h comes from, but in the end I guess it's sufficiently
>>> well-defined by the C spec to be reliable.
>>>
>>> Thanks for getting this done!
>>>
>>
>> It's a pleasure!
>>
>>
>>> Robin.
>>>
>>
>> [...]
>>
>>


^ permalink raw reply	[flat|nested] 10+ messages in thread

end of thread, other threads:[~2022-09-09  9:23 UTC | newest]

Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-08-30  3:01 [PATCH V3 0/2] DMA: update acpi_dma_get_range to return dma map regions Jianmin Lv
2022-08-30  3:01 ` [PATCH V3 1/2] ACPI / scan: Support multiple dma windows with different offsets Jianmin Lv
2022-09-05 12:20   ` Robin Murphy
2022-09-06 12:40     ` Jianmin Lv
2022-09-09  8:04       ` Lorenzo Pieralisi
2022-09-09  9:23         ` Jianmin Lv
2022-08-30  3:01 ` [PATCH V3 2/2] LoongArch: Remove ARCH_HAS_PHYS_TO_DMA Jianmin Lv
2022-09-05 14:42   ` Huacai Chen
2022-09-06 12:34     ` Jianmin Lv
2022-09-05  2:55 ` [PATCH V3 0/2] DMA: update acpi_dma_get_range to return dma map regions Jianmin Lv

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).