memory hotplug: free memmaps allocated by bootmem

This patch is to free memmaps which is allocated by bootmem. Freeing usemap is not necessary. The pages of usemap may be necessary for other sections. If removing section is last section on the node, its section is the final user of usemap page. (usemaps are allocated on its section by previous patch.) But it shouldn't be freed too, because the section must be logical offline state which all pages are isolated against page allocater. If it is freed, page alloctor may use it which will be removed physically soon. It will be disaster. So, this patch keeps it as it is. Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Yinghai Lu <yhlu.kernel@gmail.com> Cc: Yasunori Goto <y-goto@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memory hotplug: free memmaps allocated by bootmem
This patch is to free memmaps which is allocated by bootmem. Freeing usemap is not necessary. The pages of usemap may be necessary for other sections. If removing section is last section on the node, its section is the final user of usemap page. (usemaps are allocated on its section by previous patch.) But it shouldn't be freed too, because the section must be logical offline state which all pages are isolated against page allocater. If it is freed, page alloctor may use it which will be removed physically soon. It will be disaster. So, this patch keeps it as it is. Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Yinghai Lu <yhlu.kernel@gmail.com> Cc: Yasunori Goto <y-goto@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Yasunori Goto · Linus Torvalds
1 parent 86f6dae137
Showing 4 changed files with 60 additions and 7 deletions Side-by-side Diff
mm/internal.h
mm/memory_hotplug.c
mm/page_alloc.c
mm/sparse.c
@@ -34,8 +34,7 @@
 	atomic_dec(&page->_count);
 }
  
-extern void __init __free_pages_bootmem(struct page *page,
-						unsigned int order);
+extern void __free_pages_bootmem(struct page *page, unsigned int order);
  
 /*
  * function for dealing with page's order in buddy system.
@@ -198,8 +198,18 @@
 	return register_new_memory(__pfn_to_section(phys_start_pfn));
 }
  
+#ifdef CONFIG_SPARSEMEM_VMEMMAP
 static int __remove_section(struct zone *zone, struct mem_section *ms)
 {
+	/*
+	 * XXX: Freeing memmap with vmemmap is not implement yet.
+	 *      This should be removed later.
+	 */
+	return -EBUSY;
+}
+#else
+static int __remove_section(struct zone *zone, struct mem_section *ms)
+{
 	unsigned long flags;
 	struct pglist_data *pgdat = zone->zone_pgdat;
 	int ret = -EINVAL;
@@ -216,6 +226,7 @@
 	pgdat_resize_unlock(pgdat, &flags);
 	return 0;
 }
+#endif
  
 /*
  * Reasonably generic function for adding memory.  It is
@@ -546,7 +546,7 @@
 /*
  * permit the bootmem allocator to evade page validation on high-order frees
  */
-void __init __free_pages_bootmem(struct page *page, unsigned int order)
+void __free_pages_bootmem(struct page *page, unsigned int order)
 {
 	if (order == 0) {
 		__ClearPageReserved(page);
@@ -8,6 +8,7 @@
 #include <linux/module.h>
 #include <linux/spinlock.h>
 #include <linux/vmalloc.h>
+#include "internal.h"
 #include <asm/dma.h>
 #include <asm/pgalloc.h>
 #include <asm/pgtable.h>
@@ -376,6 +377,9 @@
 {
 	return; /* XXX: Not implemented yet */
 }
+static void free_map_bootmem(struct page *page, unsigned long nr_pages)
+{
+}
 #else
 static struct page *__kmalloc_section_memmap(unsigned long nr_pages)
 {
  
  
  
@@ -413,17 +417,47 @@
 		free_pages((unsigned long)memmap,
 			   get_order(sizeof(struct page) * nr_pages));
 }
+
+static void free_map_bootmem(struct page *page, unsigned long nr_pages)
+{
+	unsigned long maps_section_nr, removing_section_nr, i;
+	int magic;
+
+	for (i = 0; i < nr_pages; i++, page++) {
+		magic = atomic_read(&page->_mapcount);
+
+		BUG_ON(magic == NODE_INFO);
+
+		maps_section_nr = pfn_to_section_nr(page_to_pfn(page));
+		removing_section_nr = page->private;
+
+		/*
+		 * When this function is called, the removing section is
+		 * logical offlined state. This means all pages are isolated
+		 * from page allocator. If removing section's memmap is placed
+		 * on the same section, it must not be freed.
+		 * If it is freed, page allocator may allocate it which will
+		 * be removed physically soon.
+		 */
+		if (maps_section_nr != removing_section_nr)
+			put_page_bootmem(page);
+	}
+}
 #endif /* CONFIG_SPARSEMEM_VMEMMAP */
  
 static void free_section_usemap(struct page *memmap, unsigned long *usemap)
 {
+	struct page *usemap_page;
+	unsigned long nr_pages;
+
 	if (!usemap)
 		return;
  
+	usemap_page = virt_to_page(usemap);
 	/*
 	 * Check to see if allocation came from hot-plug-add
 	 */
-	if (PageSlab(virt_to_page(usemap))) {
+	if (PageSlab(usemap_page)) {
 		kfree(usemap);
 		if (memmap)
 			__kfree_section_memmap(memmap, PAGES_PER_SECTION);
  
@@ -431,10 +465,19 @@
 	}
  
 	/*
-	 * TODO: Allocations came from bootmem - how do I free up ?
+	 * The usemap came from bootmem. This is packed with other usemaps
+	 * on the section which has pgdat at boot time. Just keep it as is now.
 	 */
-	printk(KERN_WARNING "Not freeing up allocations from bootmem "
-			"- leaking memory\n");
+
+	if (memmap) {
+		struct page *memmap_page;
+		memmap_page = virt_to_page(memmap);
+
+		nr_pages = PAGE_ALIGN(PAGES_PER_SECTION * sizeof(struct page))
+			>> PAGE_SHIFT;
+
+		free_map_bootmem(memmap_page, nr_pages);
+	}
 }
  
 /*
...	...	@@ -34,8 +34,7 @@
34	34	atomic_dec(&page->_count);
35	35	}
36	36
37		-extern void __init __free_pages_bootmem(struct page *page,
38		- unsigned int order);
	37	+extern void __free_pages_bootmem(struct page *page, unsigned int order);
39	38
40	39	/*
41	40	* function for dealing with page's order in buddy system.
...	...	@@ -198,8 +198,18 @@
198	198	return register_new_memory(__pfn_to_section(phys_start_pfn));
199	199	}
200	200
	201	+#ifdef CONFIG_SPARSEMEM_VMEMMAP
201	202	static int __remove_section(struct zone zone, struct mem_section ms)
202	203	{
	204	+ /*
	205	+ * XXX: Freeing memmap with vmemmap is not implement yet.
	206	+ * This should be removed later.
	207	+ */
	208	+ return -EBUSY;
	209	+}
	210	+#else
	211	+static int __remove_section(struct zone zone, struct mem_section ms)
	212	+{
203	213	unsigned long flags;
204	214	struct pglist_data *pgdat = zone->zone_pgdat;
205	215	int ret = -EINVAL;
...	...	@@ -216,6 +226,7 @@
216	226	pgdat_resize_unlock(pgdat, &flags);
217	227	return 0;
218	228	}
	229	+#endif
219	230
220	231	/*
221	232	* Reasonably generic function for adding memory. It is
...	...	@@ -546,7 +546,7 @@
546	546	/*
547	547	* permit the bootmem allocator to evade page validation on high-order frees
548	548	*/
549		-void __init __free_pages_bootmem(struct page *page, unsigned int order)
	549	+void __free_pages_bootmem(struct page *page, unsigned int order)
550	550	{
551	551	if (order == 0) {
552	552	__ClearPageReserved(page);
...	...	@@ -8,6 +8,7 @@
8	8	#include <linux/module.h>
9	9	#include <linux/spinlock.h>
10	10	#include <linux/vmalloc.h>
	11	+#include "internal.h"
11	12	#include <asm/dma.h>
12	13	#include <asm/pgalloc.h>
13	14	#include <asm/pgtable.h>
...	...	@@ -376,6 +377,9 @@
376	377	{
377	378	return; /* XXX: Not implemented yet */
378	379	}
	380	+static void free_map_bootmem(struct page *page, unsigned long nr_pages)
	381	+{
	382	+}
379	383	#else
380	384	static struct page *__kmalloc_section_memmap(unsigned long nr_pages)
381	385	{
382	386
383	387
384	388
...	...	@@ -413,17 +417,47 @@
413	417	free_pages((unsigned long)memmap,
414	418	get_order(sizeof(struct page) * nr_pages));
415	419	}
	420	+
	421	+static void free_map_bootmem(struct page *page, unsigned long nr_pages)
	422	+{
	423	+ unsigned long maps_section_nr, removing_section_nr, i;
	424	+ int magic;
	425	+
	426	+ for (i = 0; i < nr_pages; i++, page++) {
	427	+ magic = atomic_read(&page->_mapcount);
	428	+
	429	+ BUG_ON(magic == NODE_INFO);
	430	+
	431	+ maps_section_nr = pfn_to_section_nr(page_to_pfn(page));
	432	+ removing_section_nr = page->private;
	433	+
	434	+ /*
	435	+ * When this function is called, the removing section is
	436	+ * logical offlined state. This means all pages are isolated
	437	+ * from page allocator. If removing section's memmap is placed
	438	+ * on the same section, it must not be freed.
	439	+ * If it is freed, page allocator may allocate it which will
	440	+ * be removed physically soon.
	441	+ */
	442	+ if (maps_section_nr != removing_section_nr)
	443	+ put_page_bootmem(page);
	444	+ }
	445	+}
416	446	#endif /* CONFIG_SPARSEMEM_VMEMMAP */
417	447
418	448	static void free_section_usemap(struct page memmap, unsigned long usemap)
419	449	{
	450	+ struct page *usemap_page;
	451	+ unsigned long nr_pages;
	452	+
420	453	if (!usemap)
421	454	return;
422	455
	456	+ usemap_page = virt_to_page(usemap);
423	457	/*
424	458	* Check to see if allocation came from hot-plug-add
425	459	*/
426		- if (PageSlab(virt_to_page(usemap))) {
	460	+ if (PageSlab(usemap_page)) {
427	461	kfree(usemap);
428	462	if (memmap)
429	463	__kfree_section_memmap(memmap, PAGES_PER_SECTION);
430	464
...	...	@@ -431,10 +465,19 @@
431	465	}
432	466
433	467	/*
434		- * TODO: Allocations came from bootmem - how do I free up ?
	468	+ * The usemap came from bootmem. This is packed with other usemaps
	469	+ * on the section which has pgdat at boot time. Just keep it as is now.
435	470	*/
436		- printk(KERN_WARNING "Not freeing up allocations from bootmem "
437		- "- leaking memory\n");
	471	+
	472	+ if (memmap) {
	473	+ struct page *memmap_page;
	474	+ memmap_page = virt_to_page(memmap);
	475	+
	476	+ nr_pages = PAGE_ALIGN(PAGES_PER_SECTION * sizeof(struct page))
	477	+ >> PAGE_SHIFT;
	478	+
	479	+ free_map_bootmem(memmap_page, nr_pages);
	480	+ }
438	481	}
439	482
440	483	/*