* Added a DMAResource::Init() method that gets a device node - for now, it will

reuse the block_io attributes in the node to build the dma_restrictions.
* DMAResource::Init() now dumps the dma_resources (should be done with the
  TRACE() macro later).
* Turned off IOScheduler debug output.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26825 a95241bf-73f2-0310-859d-f6bbb57e9c96

src/system/kernel/device_manager/IOScheduler.cpp

@@ -20,7 +20,7 @@
 #include <util/AutoLock.h>
 
 
-#define TRACE_IO_SCHEDULER
+//#define TRACE_IO_SCHEDULER
 #ifdef TRACE_IO_SCHEDULER
 #	define TRACE(x...) dprintf(x)
 #else

src/system/kernel/device_manager/dma_resources.cpp

@@ -6,6 +6,8 @@
 
 #include "dma_resources.h"
 
+#include <block_io.h>
+
 #include <kernel.h>
 #include <util/AutoLock.h>
 
@@ -20,6 +22,8 @@
 #endif
 
 
+extern device_manager_info gDeviceManagerModule;
+
 const size_t kMaxBounceBufferSize = 4 * B_PAGE_SIZE;
 
 
@@ -94,6 +98,40 @@ DMAResource::~DMAResource()
 }
 
 
+status_t
+DMAResource::Init(device_node* node, size_t blockSize)
+{
+	dma_restrictions restrictions;
+	memset(&restrictions, 0, sizeof(dma_restrictions));
+
+	// TODO: add DMA attributes instead of reusing block_io's
+
+	uint32 value;
+	if (gDeviceManagerModule.get_attr_uint32(node,
+			B_BLOCK_DEVICE_DMA_ALIGNMENT, &value, true) == B_OK)
+		restrictions.alignment = value + 1;
+
+	if (gDeviceManagerModule.get_attr_uint32(node,
+			B_BLOCK_DEVICE_DMA_BOUNDARY, &value, true) == B_OK)
+		restrictions.boundary = value + 1;
+
+	if (gDeviceManagerModule.get_attr_uint32(node,
+			B_BLOCK_DEVICE_MAX_SG_BLOCK_SIZE, &value, true) == B_OK)
+		restrictions.max_segment_size = value;
+
+	if (gDeviceManagerModule.get_attr_uint32(node,
+			B_BLOCK_DEVICE_MAX_BLOCKS_ITEM, &value, true) == B_OK)
+		restrictions.max_transfer_size = value * blockSize;
+
+	uint32 bufferCount;
+	if (gDeviceManagerModule.get_attr_uint32(node,
+			B_BLOCK_DEVICE_MAX_SG_BLOCKS, &bufferCount, true) != B_OK)
+		bufferCount = 16;
+
+	return Init(restrictions, blockSize, bufferCount);
+}
+
+
 status_t
 DMAResource::Init(const dma_restrictions& restrictions, size_t blockSize,
 	uint32 bufferCount)
@@ -123,6 +161,13 @@ DMAResource::Init(const dma_restrictions& restrictions, size_t blockSize,
 			fBounceBufferSize);
 	}
 
+	dprintf("DMAResource@%p: low/high %lx/%lx, max segment count %lu, align %lu, "
+		"boundary %lu, max transfer %lu, max segment size %lu\n", this,
+		fRestrictions.low_address, fRestrictions.high_address,
+		fRestrictions.max_segment_count, fRestrictions.alignment,
+		fRestrictions.boundary, fRestrictions.max_transfer_size,
+		fRestrictions.max_segment_size);
+
 	fScratchVecs = (iovec*)malloc(
 		sizeof(iovec) * fRestrictions.max_segment_count);
 	if (fScratchVecs == NULL)
@@ -150,9 +195,10 @@ DMAResource::CreateBuffer(size_t size, DMABuffer** _buffer)
 	area_id area = -1;
 
 	if (size != 0) {
-		if (fRestrictions.alignment > B_PAGE_SIZE
-			|| fRestrictions.boundary > B_PAGE_SIZE)
-			panic("not yet implemented");
+		if (fRestrictions.alignment > B_PAGE_SIZE)
+			dprintf("dma buffer restrictions not yet implemented: alignment %lu\n", fRestrictions.alignment);
+		if (fRestrictions.boundary > B_PAGE_SIZE)
+			dprintf("dma buffer restrictions not yet implemented: boundary %lu\n", fRestrictions.boundary);
 
 		size = ROUNDUP(size, B_PAGE_SIZE);
 

src/system/kernel/device_manager/dma_resources.h

@@ -12,6 +12,7 @@
 #include <util/DoublyLinkedList.h>
 
 
+struct device_node;
 struct IOOperation;
 struct IORequest;
 
@@ -69,6 +70,7 @@ public:
 
 			status_t			Init(const dma_restrictions& restrictions,
 									size_t blockSize, uint32 bufferCount);
+			status_t			Init(device_node* node, size_t blockSize);
 
 			status_t			CreateBuffer(DMABuffer** _buffer)
 									{ return CreateBuffer(0, _buffer); }