[RFC] [PATCH] madvise() does not always return -EBADF on non-file mapped area

[suzuki <suzuki@in.ibm.com>]

[-- Attachment #1: Type: text/plain, Size: 536 bytes --]

hi,

I came across the following problem. The madvise() system call returns 
-EBADF for areas which does not map to files, only for *behaviour* 
request MADV_WILLNEED.

Is this the intended behaviour of madvise() ?

According to man pages, madvise returns :

EBADF - the map exists, but the area maps something that isn?t a file.

I have attached a patch which could resolve the issue.

[ There is already a bug reported in OSDL regarding this issue: Bug # 
2995 ].


-- 
regards,

Suzuki K P
Linux Technology Centre
IBM Software Labs


[-- Attachment #2: madvise-ebadf-fix.patch --]
[-- Type: text/plain, Size: 920 bytes --]

Patch to fix madvise() syscall to return -EBADF on non-file mapped regions.

Signed Off by: Suzuki K P <suzuki@in.ibm.com>

--- mm/madvise.c	2005-07-01 16:08:26.000000000 +0530
+++ mm/madvise.c.new	2005-07-01 16:07:45.000000000 +0530
@@ -62,9 +62,6 @@ static long madvise_willneed(struct vm_a
 {
 	struct file *file = vma->vm_file;
 
-	if (!file)
-		return -EBADF;
-
 	start = ((start - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
 	if (end > vma->vm_end)
 		end = vma->vm_end;
@@ -114,8 +111,12 @@ static long madvise_dontneed(struct vm_a
 static long madvise_vma(struct vm_area_struct * vma, unsigned long start,
 			unsigned long end, int behavior)
 {
+	struct file* filp = vma->vm_file;
 	long error = -EBADF;
 
+	if(!filp)
+		goto  out;
+
 	switch (behavior) {
 	case MADV_NORMAL:
 	case MADV_SEQUENTIAL:
@@ -136,6 +137,7 @@ static long madvise_vma(struct vm_area_s
 		break;
 	}
 		
+out:	
 	return error;
 }