vmalloc: introduce remap_vmalloc_range_partial

We want to allocate ELF note segment buffer on the 2nd kernel in vmalloc space and remap it to user-space in order to reduce the risk that memory allocation fails on system with huge number of CPUs and so with huge ELF note segment that exceeds 11-order block size. Although there's already remap_vmalloc_range for the purpose of remapping vmalloc memory to user-space, we need to specify user-space range via vma. Mmap on /proc/vmcore needs to remap range across multiple objects, so the interface that requires vma to cover full range is problematic. This patch introduces remap_vmalloc_range_partial that receives user-space range as a pair of base address and size and can be used for mmap on /proc/vmcore case. remap_vmalloc_range is rewritten using remap_vmalloc_range_partial. [akpm@linux-foundation.org: use PAGE_ALIGNED()] Signed-off-by: HATAYAMA Daisuke <d.hatayama@jp.fujitsu.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Vivek Goyal <vgoyal@redhat.com> Cc: Atsushi Kumagai <kumagai-atsushi@mxc.nes.nec.co.jp> Cc: Lisa Mitchell <lisa.mitchell@hp.com> Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

vmalloc: introduce remap_vmalloc_range_partial
We want to allocate ELF note segment buffer on the 2nd kernel in vmalloc space and remap it to user-space in order to reduce the risk that memory allocation fails on system with huge number of CPUs and so with huge ELF note segment that exceeds 11-order block size. Although there's already remap_vmalloc_range for the purpose of remapping vmalloc memory to user-space, we need to specify user-space range via vma. Mmap on /proc/vmcore needs to remap range across multiple objects, so the interface that requires vma to cover full range is problematic. This patch introduces remap_vmalloc_range_partial that receives user-space range as a pair of base address and size and can be used for mmap on /proc/vmcore case. remap_vmalloc_range is rewritten using remap_vmalloc_range_partial. [akpm@linux-foundation.org: use PAGE_ALIGNED()] Signed-off-by: HATAYAMA Daisuke <d.hatayama@jp.fujitsu.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Vivek Goyal <vgoyal@redhat.com> Cc: Atsushi Kumagai <kumagai-atsushi@mxc.nes.nec.co.jp> Cc: Lisa Mitchell <lisa.mitchell@hp.com> Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
HATAYAMA Daisuke · Linus Torvalds
1 parent cef2ac3f6c
Showing 2 changed files with 49 additions and 22 deletions Side-by-side Diff
include/linux/vmalloc.h
mm/vmalloc.c
@@ -82,6 +82,10 @@
 			unsigned long flags, pgprot_t prot);
 extern void vunmap(const void *addr);
  
+extern int remap_vmalloc_range_partial(struct vm_area_struct *vma,
+				       unsigned long uaddr, void *kaddr,
+				       unsigned long size);
+
 extern int remap_vmalloc_range(struct vm_area_struct *vma, void *addr,
 							unsigned long pgoff);
 void vmalloc_sync_all(void);
@@ -1476,10 +1476,9 @@
 	if (!addr)
 		return;
  
-	if ((PAGE_SIZE-1) & (unsigned long)addr) {
-		WARN(1, KERN_ERR "Trying to vfree() bad address (%p)\n", addr);
+	if (WARN(!PAGE_ALIGNED(addr), "Trying to vfree() bad address (%p)\n",
+			addr));
 		return;
-	}
  
 	area = remove_vm_area(addr);
 	if (unlikely(!area)) {
  
  
  
  
  
  
  
  
@@ -2148,42 +2147,43 @@
 }
  
 /**
- *	remap_vmalloc_range  -  map vmalloc pages to userspace
- *	@vma:		vma to cover (map full range of vma)
- *	@addr:		vmalloc memory
- *	@pgoff:		number of pages into addr before first page to map
+ *	remap_vmalloc_range_partial  -  map vmalloc pages to userspace
+ *	@vma:		vma to cover
+ *	@uaddr:		target user address to start at
+ *	@kaddr:		virtual address of vmalloc kernel memory
+ *	@size:		size of map area
  *
  *	Returns:	0 for success, -Exxx on failure
  *
- *	This function checks that addr is a valid vmalloc'ed area, and
- *	that it is big enough to cover the vma. Will return failure if
- *	that criteria isn't met.
+ *	This function checks that @kaddr is a valid vmalloc'ed area,
+ *	and that it is big enough to cover the range starting at
+ *	@uaddr in @vma. Will return failure if that criteria isn't
+ *	met.
  *
  *	Similar to remap_pfn_range() (see mm/memory.c)
  */
-int remap_vmalloc_range(struct vm_area_struct *vma, void *addr,
-						unsigned long pgoff)
+int remap_vmalloc_range_partial(struct vm_area_struct *vma, unsigned long uaddr,
+				void *kaddr, unsigned long size)
 {
 	struct vm_struct *area;
-	unsigned long uaddr = vma->vm_start;
-	unsigned long usize = vma->vm_end - vma->vm_start;
  
-	if ((PAGE_SIZE-1) & (unsigned long)addr)
+	size = PAGE_ALIGN(size);
+
+	if (!PAGE_ALIGNED(uaddr) || !PAGE_ALIGNED(kaddr))
 		return -EINVAL;
  
-	area = find_vm_area(addr);
+	area = find_vm_area(kaddr);
 	if (!area)
 		return -EINVAL;
  
 	if (!(area->flags & VM_USERMAP))
 		return -EINVAL;
  
-	if (usize + (pgoff << PAGE_SHIFT) > area->size - PAGE_SIZE)
+	if (kaddr + size > area->addr + area->size)
 		return -EINVAL;
  
-	addr += pgoff << PAGE_SHIFT;
 	do {
-		struct page *page = vmalloc_to_page(addr);
+		struct page *page = vmalloc_to_page(kaddr);
 		int ret;
  
 		ret = vm_insert_page(vma, uaddr, page);
  
@@ -2191,13 +2191,36 @@
 			return ret;
  
 		uaddr += PAGE_SIZE;
-		addr += PAGE_SIZE;
-		usize -= PAGE_SIZE;
-	} while (usize > 0);
+		kaddr += PAGE_SIZE;
+		size -= PAGE_SIZE;
+	} while (size > 0);
  
 	vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP;
  
 	return 0;
+}
+EXPORT_SYMBOL(remap_vmalloc_range_partial);
+
+/**
+ *	remap_vmalloc_range  -  map vmalloc pages to userspace
+ *	@vma:		vma to cover (map full range of vma)
+ *	@addr:		vmalloc memory
+ *	@pgoff:		number of pages into addr before first page to map
+ *
+ *	Returns:	0 for success, -Exxx on failure
+ *
+ *	This function checks that addr is a valid vmalloc'ed area, and
+ *	that it is big enough to cover the vma. Will return failure if
+ *	that criteria isn't met.
+ *
+ *	Similar to remap_pfn_range() (see mm/memory.c)
+ */
+int remap_vmalloc_range(struct vm_area_struct *vma, void *addr,
+						unsigned long pgoff)
+{
+	return remap_vmalloc_range_partial(vma, vma->vm_start,
+					   addr + (pgoff << PAGE_SHIFT),
+					   vma->vm_end - vma->vm_start);
 }
 EXPORT_SYMBOL(remap_vmalloc_range);
...	...	@@ -82,6 +82,10 @@
82	82	unsigned long flags, pgprot_t prot);
83	83	extern void vunmap(const void *addr);
84	84
	85	+extern int remap_vmalloc_range_partial(struct vm_area_struct *vma,
	86	+ unsigned long uaddr, void *kaddr,
	87	+ unsigned long size);
	88	+
85	89	extern int remap_vmalloc_range(struct vm_area_struct vma, void addr,
86	90	unsigned long pgoff);
87	91	void vmalloc_sync_all(void);
...	...	@@ -1476,10 +1476,9 @@
1476	1476	if (!addr)
1477	1477	return;
1478	1478
1479		- if ((PAGE_SIZE-1) & (unsigned long)addr) {
1480		- WARN(1, KERN_ERR "Trying to vfree() bad address (%p)\n", addr);
	1479	+ if (WARN(!PAGE_ALIGNED(addr), "Trying to vfree() bad address (%p)\n",
	1480	+ addr));
1481	1481	return;
1482		- }
1483	1482
1484	1483	area = remove_vm_area(addr);
1485	1484	if (unlikely(!area)) {
1486	1485
1487	1486
1488	1487
1489	1488
1490	1489
1491	1490
1492	1491
1493	1492
...	...	@@ -2148,42 +2147,43 @@
2148	2147	}
2149	2148
2150	2149	/**
2151		- * remap_vmalloc_range - map vmalloc pages to userspace
2152		- * @vma: vma to cover (map full range of vma)
2153		- * @addr: vmalloc memory
2154		- * @pgoff: number of pages into addr before first page to map
	2150	+ * remap_vmalloc_range_partial - map vmalloc pages to userspace
	2151	+ * @vma: vma to cover
	2152	+ * @uaddr: target user address to start at
	2153	+ * @kaddr: virtual address of vmalloc kernel memory
	2154	+ * @size: size of map area
2155	2155	*
2156	2156	* Returns: 0 for success, -Exxx on failure
2157	2157	*
2158		- * This function checks that addr is a valid vmalloc'ed area, and
2159		- * that it is big enough to cover the vma. Will return failure if
2160		- * that criteria isn't met.
	2158	+ * This function checks that @kaddr is a valid vmalloc'ed area,
	2159	+ * and that it is big enough to cover the range starting at
	2160	+ * @uaddr in @vma. Will return failure if that criteria isn't
	2161	+ * met.
2161	2162	*
2162	2163	* Similar to remap_pfn_range() (see mm/memory.c)
2163	2164	*/
2164		-int remap_vmalloc_range(struct vm_area_struct vma, void addr,
2165		- unsigned long pgoff)
	2165	+int remap_vmalloc_range_partial(struct vm_area_struct *vma, unsigned long uaddr,
	2166	+ void *kaddr, unsigned long size)
2166	2167	{
2167	2168	struct vm_struct *area;
2168		- unsigned long uaddr = vma->vm_start;
2169		- unsigned long usize = vma->vm_end - vma->vm_start;
2170	2169
2171		- if ((PAGE_SIZE-1) & (unsigned long)addr)
	2170	+ size = PAGE_ALIGN(size);
	2171	+
	2172	+ if (!PAGE_ALIGNED(uaddr) \|\| !PAGE_ALIGNED(kaddr))
2172	2173	return -EINVAL;
2173	2174
2174		- area = find_vm_area(addr);
	2175	+ area = find_vm_area(kaddr);
2175	2176	if (!area)
2176	2177	return -EINVAL;
2177	2178
2178	2179	if (!(area->flags & VM_USERMAP))
2179	2180	return -EINVAL;
2180	2181
2181		- if (usize + (pgoff << PAGE_SHIFT) > area->size - PAGE_SIZE)
	2182	+ if (kaddr + size > area->addr + area->size)
2182	2183	return -EINVAL;
2183	2184
2184		- addr += pgoff << PAGE_SHIFT;
2185	2185	do {
2186		- struct page *page = vmalloc_to_page(addr);
	2186	+ struct page *page = vmalloc_to_page(kaddr);
2187	2187	int ret;
2188	2188
2189	2189	ret = vm_insert_page(vma, uaddr, page);
2190	2190
...	...	@@ -2191,13 +2191,36 @@
2191	2191	return ret;
2192	2192
2193	2193	uaddr += PAGE_SIZE;
2194		- addr += PAGE_SIZE;
2195		- usize -= PAGE_SIZE;
2196		- } while (usize > 0);
	2194	+ kaddr += PAGE_SIZE;
	2195	+ size -= PAGE_SIZE;
	2196	+ } while (size > 0);
2197	2197
2198	2198	vma->vm_flags \|= VM_DONTEXPAND \| VM_DONTDUMP;
2199	2199
2200	2200	return 0;
	2201	+}
	2202	+EXPORT_SYMBOL(remap_vmalloc_range_partial);
	2203	+
	2204	+/**
	2205	+ * remap_vmalloc_range - map vmalloc pages to userspace
	2206	+ * @vma: vma to cover (map full range of vma)
	2207	+ * @addr: vmalloc memory
	2208	+ * @pgoff: number of pages into addr before first page to map
	2209	+ *
	2210	+ * Returns: 0 for success, -Exxx on failure
	2211	+ *
	2212	+ * This function checks that addr is a valid vmalloc'ed area, and
	2213	+ * that it is big enough to cover the vma. Will return failure if
	2214	+ * that criteria isn't met.
	2215	+ *
	2216	+ * Similar to remap_pfn_range() (see mm/memory.c)
	2217	+ */
	2218	+int remap_vmalloc_range(struct vm_area_struct vma, void addr,
	2219	+ unsigned long pgoff)
	2220	+{
	2221	+ return remap_vmalloc_range_partial(vma, vma->vm_start,
	2222	+ addr + (pgoff << PAGE_SHIFT),
	2223	+ vma->vm_end - vma->vm_start);
2201	2224	}
2202	2225	EXPORT_SYMBOL(remap_vmalloc_range);
2203	2226