Skip site navigation (1)Skip section navigation (2) 
Date:      Thu, 29 Jul 2010 02:12:23 +0000 (UTC)
From:      Jeff Roberson <jeff@FreeBSD.org>
To:        src-committers@freebsd.org, svn-src-projects@freebsd.org
Subject:   svn commit: r210586 - projects/ofed/head/sys/ofed/include/linux
Message-ID:  <201007290212.o6T2CN20099343@svn.freebsd.org>
next in thread | raw e-mail | index | archive | help 
Author: jeff
Date: Thu Jul 29 02:12:22 2010
New Revision: 210586
URL: http://svn.freebsd.org/changeset/base/210586

Log:
   - Refine various page allocation methods.  They all essentially override
     the page object pointer to cache the virtual address after allocation.
     This same trick is used in UMA to store the slab address and so it
     should be safe.
  
  Sponsored by:	Isilon Systems, iX Systems, and Panasas.

Modified:
  projects/ofed/head/sys/ofed/include/linux/gfp.h

Modified: projects/ofed/head/sys/ofed/include/linux/gfp.h
==============================================================================
--- projects/ofed/head/sys/ofed/include/linux/gfp.h	Thu Jul 29 02:11:31 2010	(r210585)
+++ projects/ofed/head/sys/ofed/include/linux/gfp.h	Thu Jul 29 02:12:22 2010	(r210586)
@@ -51,12 +51,12 @@
 #define	GFP_IOFS	M_NOWAIT
 
 static inline unsigned long
-get_zeroed_page(gfp_t mask)
+_get_page(gfp_t mask)
 {
 	vm_page_t m;
 	vm_offset_t p;
 
-	p = kmem_malloc(kernel_map, PAGE_SIZE, mask | M_ZERO);
+	p = kmem_malloc(kmem_map, PAGE_SIZE, mask | M_ZERO);
 	if (p) {
 		m = virt_to_page(p);
 		m->flags |= PG_KVA;
@@ -65,6 +65,10 @@ get_zeroed_page(gfp_t mask)
 	return (p);
 }
 
+#define	get_zeroed_page(mask)	_get_page((mask) | M_ZERO)
+#define	alloc_page(mask)	virt_to_page(_get_page((mask)))
+#define	__get_free_page(mask)	_get_page((mask))
+
 static inline void
 free_page(unsigned long page)
 {
@@ -73,29 +77,50 @@ free_page(unsigned long page)
 	m = virt_to_page(page);
 	if (m->flags & PG_KVA) {
 		m->flags &= ~PG_KVA;
-		m->object = kernel_object;
+		m->object = kmem_object;
 	}
-	kmem_free(kernel_map, page, PAGE_SIZE);
+	kmem_free(kmem_map, page, PAGE_SIZE);
+}
+
+static inline void
+__free_page(struct page *m)
+{
+	void *p;
+
+	if ((m->flags & PG_KVA) == 0)
+		panic("__free_page:  Freed page %p not allocated via wrappers.",
+		    m);
+	p = m->object;
+	m->flags &= ~PG_KVA;
+	m->object  = kmem_object;
+	kmem_free(kmem_map, (vm_offset_t)p, PAGE_SIZE);
 }
 
 static inline void
 __free_pages(void *p, unsigned int order)
 {
+	unsigned long start;
 	unsigned long page;
 	vm_page_t m;
 	size_t size;
 
-	size = order << PAGE_SHIFT;
-	for (page = (uintptr_t)p; p < (uintptr_t)p + size; page += PAGE_SIZE) {
+	size = PAGE_SIZE << order;
+	start = (unsigned long)p;
+	for (page = start; page < start + size; page += PAGE_SIZE) {
 		m = virt_to_page(page);
 		if (m->flags & PG_KVA) {
 			m->flags &= ~PG_KVA;
-			m->object = kernel_object;
+			m->object = kmem_object;
 		}
 	}
-	kmem_free(kernel_map, p, size);
+	kmem_free(kmem_map, (vm_offset_t)p, size);
 }
 
+/*
+ * Alloc pages allocates directly from the buddy allocator on linux so
+ * order specifies a power of two bucket of pages and the results
+ * are expected to be aligned on the size as well.
+ */
 static inline struct page *
 alloc_pages(gfp_t gfp_mask, unsigned int order)
 {
@@ -104,9 +129,9 @@ alloc_pages(gfp_t gfp_mask, unsigned int
 	vm_page_t m;
 	size_t size;
 
-	size = order << PAGE_SHIFT;
-	start = kmem_alloc_contig(kernel_map, size, gfp_mask, 0, -1,
-	    PAGE_SIZE, 0, VM_MEMATTR_DEFAULT);
+	size = PAGE_SIZE << order;
+	start = kmem_alloc_contig(kmem_map, size, gfp_mask, 0, -1,
+	    size, 0, VM_MEMATTR_DEFAULT);
 	if (start == 0)
 		return (NULL);
 	for (page = start; page < start + size; page += PAGE_SIZE) {

Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201007290212.o6T2CN20099343>