mlock hanging on non-ram with VM_IO set

[Andrea Arcangeli]

I've got a report of mlock deadlocking on non-ram mappings (precisely on
a mmap of /dev/mem outside the mem_map array range)

I believe this should fix it (I verified the mlock is the only one
passing down pages == NULL, and infact mlock is the only one hanging on
VM_IO vmas since the pfn isn't valid and handle_mm_fault sure can't
istantiate a valid pfn there since the ptes are filled synchronously at
mmap time), comments? (I don't see any reason why mlock should be
allowed to trigger page faults on VM_IO regions)

This is not a bad bug, mlock is privilegied, and /dev/mem is privilegied
too, and mlock makes no sense on a non-ram mappings (noop), but I guess
it's ok to add it mostly for things like mlockall that would otherwise
deadlock the box and I guess it's nicer if they work transparently, no
matter what mappings are in the task.

btw, mlock traps the error and ignores it as it should.

this is untested but I guess it should fix it (and it applied cleanly to
kernel CVS).

Signed-off-by: Andrea Arcangeli <andrea@xxxxxxxxxx>

--- sles/mm/memory.c.~1~	2004-11-12 12:30:25.000000000 +0100
+++ sles/mm/memory.c	2004-11-16 17:58:02.752131952 +0100
@@ -753,7 +753,7 @@ int get_user_pages(struct task_struct *t
 			continue;
 		}
 
-		if (!vma || (pages && (vma->vm_flags & VM_IO))
+		if (!vma || (vma->vm_flags & VM_IO)
 				|| !(flags & vma->vm_flags))
 			return i ? : -EFAULT;
 
-
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/