[patch 06/18] ieee1394: nopage

[npiggin]

Convert ieee1394 from nopage to fault.
Remove redundant vma range checks (correct resource range check is retained).

Signed-off-by: Nick Piggin <npiggin@xxxxxxx>
Cc: krh@xxxxxxxxxx
Cc: stefanr@xxxxxxxxxxxxxxxxx
Cc: linux1394-devel@xxxxxxxxxxxxxxxxxxxxx
Cc: linux-kernel@xxxxxxxxxxxxxxx
---
 drivers/ieee1394/dma.c |   39 +++++++++++++++++----------------------
 1 file changed, 17 insertions(+), 22 deletions(-)

Index: linux-2.6/drivers/ieee1394/dma.c
===================================================================
--- linux-2.6.orig/drivers/ieee1394/dma.c
+++ linux-2.6/drivers/ieee1394/dma.c
@@ -231,37 +231,32 @@ void dma_region_sync_for_device(struct d
 
 #ifdef CONFIG_MMU
 
-/* nopage() handler for mmap access */
+/* fault() handler for mmap access */
 
-static struct page *dma_region_pagefault(struct vm_area_struct *area,
-					 unsigned long address, int *type)
+static int dma_region_pagefault(struct vm_area_struct *vma,
+					struct vm_fault *vmf)
 {
-	unsigned long offset;
 	unsigned long kernel_virt_addr;
-	struct page *ret = NOPAGE_SIGBUS;
 
-	struct dma_region *dma = (struct dma_region *)area->vm_private_data;
+	struct dma_region *dma = (struct dma_region *)vma->vm_private_data;
 
 	if (!dma->kvirt)
-		goto out;
+		goto error;
 
-	if ((address < (unsigned long)area->vm_start) ||
-	    (address >
-	     (unsigned long)area->vm_start + (dma->n_pages << PAGE_SHIFT)))
-		goto out;
-
-	if (type)
-		*type = VM_FAULT_MINOR;
-	offset = address - area->vm_start;
-	kernel_virt_addr = (unsigned long)dma->kvirt + offset;
-	ret = vmalloc_to_page((void *)kernel_virt_addr);
-	get_page(ret);
-      out:
-	return ret;
+	if (vmf->pgoff >= dma->n_pages)
+		goto error;
+
+	kernel_virt_addr = (unsigned long)dma->kvirt + (vmf->pgoff << PAGE_SHIFT);
+	vmf->page = vmalloc_to_page((void *)kernel_virt_addr);
+	get_page(vmf->page);
+	return 0;
+
+      error:
+	return VM_FAULT_SIGBUS;
 }
 
 static struct vm_operations_struct dma_region_vm_ops = {
-	.nopage = dma_region_pagefault,
+	.fault = dma_region_pagefault,
 };
 
 /**
@@ -275,7 +270,7 @@ int dma_region_mmap(struct dma_region *d
 	if (!dma->kvirt)
 		return -EINVAL;
 
-	/* must be page-aligned */
+	/* must be page-aligned (XXX: comment is wrong, we could allow pgoff) */
 	if (vma->vm_pgoff != 0)
 		return -EINVAL;
 

-- 

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/