[PATCH] x86_64: Fixed guard page handling again in iounmap

Caused oopses again.  Also fix potential mismatch in checking if
change_page_attr was needed.

To do it without races I needed to change mm/vmalloc.c to export a
__remove_vm_area that does not take vmlist lock.

Noticed by Terence Ripperda and based on a patch of his.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

arch/x86_64/mm/ioremap.c

@@ -133,7 +133,7 @@ ioremap_change_attr(unsigned long phys_addr, unsigned long size,
 					unsigned long flags)
 {
 	int err = 0;
-	if (flags && phys_addr + size - 1 < (end_pfn_map << PAGE_SHIFT)) {
+	if (phys_addr + size - 1 < (end_pfn_map << PAGE_SHIFT)) {
 		unsigned long npages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT;
 		unsigned long vaddr = (unsigned long) __va(phys_addr);
 
@@ -214,7 +214,7 @@ void __iomem * __ioremap(unsigned long phys_addr, unsigned long size, unsigned l
 		remove_vm_area((void *)(PAGE_MASK & (unsigned long) addr));
 		return NULL;
 	}
-	if (ioremap_change_attr(phys_addr, size, flags) < 0) {
+	if (flags && ioremap_change_attr(phys_addr, size, flags) < 0) {
 		area->flags &= 0xffffff;
 		vunmap(addr);
 		return NULL;
@@ -251,7 +251,7 @@ void __iomem *ioremap_nocache (unsigned long phys_addr, unsigned long size)
 
 void iounmap(volatile void __iomem *addr)
 {
-	struct vm_struct *p, **pprev;
+	struct vm_struct *p;
 
 	if (addr <= high_memory) 
 		return; 
@@ -260,24 +260,11 @@ void iounmap(volatile void __iomem *addr)
 		return;
 
 	write_lock(&vmlist_lock);
-	for (p = vmlist, pprev = &vmlist; p != NULL; pprev = &p->next, p = *pprev)
-		if (p->addr == (void *)(PAGE_MASK & (unsigned long)addr))
-			break;
-	if (!p) { 
-		printk("__iounmap: bad address %p\n", addr);
-		goto out_unlock;
-	}
-	*pprev = p->next;
-	unmap_vm_area(p);
-	if ((p->flags >> 20) &&
-		p->phys_addr + p->size - 1 < virt_to_phys(high_memory)) {
-		/* p->size includes the guard page, but cpa doesn't like that */
-		change_page_attr_addr((unsigned long)__va(p->phys_addr),
-				 p->size >> PAGE_SHIFT,
-				 PAGE_KERNEL);
-		global_flush_tlb();
-	} 
-out_unlock:
+	p = __remove_vm_area((void *)((unsigned long)addr & PAGE_MASK));
+	if (!p)
+		printk("iounmap: bad address %p\n", addr);
+	else if (p->flags >> 20)
+		ioremap_change_attr(p->phys_addr, p->size, 0);
 	write_unlock(&vmlist_lock);
 	kfree(p); 
 }

include/linux/vmalloc.h

@@ -41,6 +41,7 @@ extern struct vm_struct *get_vm_area(unsigned long size, unsigned long flags);
 extern struct vm_struct *__get_vm_area(unsigned long size, unsigned long flags,
 					unsigned long start, unsigned long end);
 extern struct vm_struct *remove_vm_area(void *addr);
+extern struct vm_struct *__remove_vm_area(void *addr);
 extern int map_vm_area(struct vm_struct *area, pgprot_t prot,
 			struct page ***pages);
 extern void unmap_vm_area(struct vm_struct *area);

mm/vmalloc.c

@@ -248,31 +248,20 @@ struct vm_struct *get_vm_area(unsigned long size, unsigned long flags)
 	return __get_vm_area(size, flags, VMALLOC_START, VMALLOC_END);
 }
 
-/**
- *	remove_vm_area  -  find and remove a contingous kernel virtual area
- *
- *	@addr:		base address
- *
- *	Search for the kernel VM area starting at @addr, and remove it.
- *	This function returns the found VM area, but using it is NOT safe
- *	on SMP machines.
- */
-struct vm_struct *remove_vm_area(void *addr)
+/* Caller must hold vmlist_lock */
+struct vm_struct *__remove_vm_area(void *addr)
 {
 	struct vm_struct **p, *tmp;
 
-	write_lock(&vmlist_lock);
 	for (p = &vmlist ; (tmp = *p) != NULL ;p = &tmp->next) {
 		 if (tmp->addr == addr)
 			 goto found;
 	}
-	write_unlock(&vmlist_lock);
 	return NULL;
 
 found:
 	unmap_vm_area(tmp);
 	*p = tmp->next;
-	write_unlock(&vmlist_lock);
 
 	/*
 	 * Remove the guard page.
@@ -281,6 +270,24 @@ found:
 	return tmp;
 }
 
+/**
+ *	remove_vm_area  -  find and remove a contingous kernel virtual area
+ *
+ *	@addr:		base address
+ *
+ *	Search for the kernel VM area starting at @addr, and remove it.
+ *	This function returns the found VM area, but using it is NOT safe
+ *	on SMP machines, except for its size or flags.
+ */
+struct vm_struct *remove_vm_area(void *addr)
+{
+	struct vm_struct *v;
+	write_lock(&vmlist_lock);
+	v = __remove_vm_area(addr);
+	write_unlock(&vmlist_lock);
+	return v;
+}
+
 void __vunmap(void *addr, int deallocate_pages)
 {
 	struct vm_struct *area;