xen/balloon: Don't rely on the page granularity is the same for Xen and Linux
authorJulien Grall <julien.grall@citrix.com>
Mon, 4 May 2015 14:39:08 +0000 (15:39 +0100)
committerDavid Vrabel <david.vrabel@citrix.com>
Fri, 23 Oct 2015 13:20:38 +0000 (14:20 +0100)
For ARM64 guests, Linux is able to support either 64K or 4K page
granularity. Although, the hypercall interface is always based on 4K
page granularity.

With 64K page granularity, a single page will be spread over multiple
Xen frame.

To avoid splitting the page into 4K frame, take advantage of the
extent_order field to directly allocate/free chunk of the Linux page
size.

Note that PVMMU is only used for PV guest (which is x86) and the page
granularity is always 4KB. Some BUILD_BUG_ON has been added to ensure
that because the code has not been modified.

Signed-off-by: Julien Grall <julien.grall@citrix.com>
Reviewed-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
drivers/xen/balloon.c

index f56662324a47176a5184c6703e788951469d2d54..b50d22960ce7cc3e3217e7b933a837b45a36a2dd 100644 (file)
@@ -112,6 +112,12 @@ static struct ctl_table xen_root[] = {
 
 #endif
 
+/*
+ * Use one extent per PAGE_SIZE to avoid to break down the page into
+ * multiple frame.
+ */
+#define EXTENT_ORDER (fls(XEN_PFN_PER_PAGE) - 1)
+
 /*
  * balloon_process() state:
  *
@@ -304,6 +310,12 @@ static enum bp_state reserve_additional_memory(void)
        nid = memory_add_physaddr_to_nid(resource->start);
 
 #ifdef CONFIG_XEN_HAVE_PVMMU
+       /*
+        * We don't support PV MMU when Linux and Xen is using
+        * different page granularity.
+        */
+       BUILD_BUG_ON(XEN_PAGE_SIZE != PAGE_SIZE);
+
         /*
          * add_memory() will build page tables for the new memory so
          * the p2m must contain invalid entries so the correct
@@ -384,11 +396,11 @@ static bool balloon_is_inflated(void)
 static enum bp_state increase_reservation(unsigned long nr_pages)
 {
        int rc;
-       unsigned long  pfn, i;
+       unsigned long i;
        struct page   *page;
        struct xen_memory_reservation reservation = {
                .address_bits = 0,
-               .extent_order = 0,
+               .extent_order = EXTENT_ORDER,
                .domid        = DOMID_SELF
        };
 
@@ -401,7 +413,11 @@ static enum bp_state increase_reservation(unsigned long nr_pages)
                        nr_pages = i;
                        break;
                }
-               frame_list[i] = page_to_pfn(page);
+
+               /* XENMEM_populate_physmap requires a PFN based on Xen
+                * granularity.
+                */
+               frame_list[i] = page_to_xen_pfn(page);
                page = balloon_next_page(page);
        }
 
@@ -415,10 +431,16 @@ static enum bp_state increase_reservation(unsigned long nr_pages)
                page = balloon_retrieve(false);
                BUG_ON(page == NULL);
 
-               pfn = page_to_pfn(page);
-
 #ifdef CONFIG_XEN_HAVE_PVMMU
+               /*
+                * We don't support PV MMU when Linux and Xen is using
+                * different page granularity.
+                */
+               BUILD_BUG_ON(XEN_PAGE_SIZE != PAGE_SIZE);
+
                if (!xen_feature(XENFEAT_auto_translated_physmap)) {
+                       unsigned long pfn = page_to_pfn(page);
+
                        set_phys_to_machine(pfn, frame_list[i]);
 
                        /* Link back into the page tables if not highmem. */
@@ -445,14 +467,15 @@ static enum bp_state increase_reservation(unsigned long nr_pages)
 static enum bp_state decrease_reservation(unsigned long nr_pages, gfp_t gfp)
 {
        enum bp_state state = BP_DONE;
-       unsigned long  pfn, i;
-       struct page   *page;
+       unsigned long i;
+       struct page *page, *tmp;
        int ret;
        struct xen_memory_reservation reservation = {
                .address_bits = 0,
-               .extent_order = 0,
+               .extent_order = EXTENT_ORDER,
                .domid        = DOMID_SELF
        };
+       LIST_HEAD(pages);
 
        if (nr_pages > ARRAY_SIZE(frame_list))
                nr_pages = ARRAY_SIZE(frame_list);
@@ -465,8 +488,7 @@ static enum bp_state decrease_reservation(unsigned long nr_pages, gfp_t gfp)
                        break;
                }
                scrub_page(page);
-
-               frame_list[i] = page_to_pfn(page);
+               list_add(&page->lru, &pages);
        }
 
        /*
@@ -478,14 +500,25 @@ static enum bp_state decrease_reservation(unsigned long nr_pages, gfp_t gfp)
         */
        kmap_flush_unused();
 
-       /* Update direct mapping, invalidate P2M, and add to balloon. */
-       for (i = 0; i < nr_pages; i++) {
-               pfn = frame_list[i];
-               frame_list[i] = pfn_to_gfn(pfn);
-               page = pfn_to_page(pfn);
+       /*
+        * Setup the frame, update direct mapping, invalidate P2M,
+        * and add to balloon.
+        */
+       i = 0;
+       list_for_each_entry_safe(page, tmp, &pages, lru) {
+               /* XENMEM_decrease_reservation requires a GFN */
+               frame_list[i++] = xen_page_to_gfn(page);
 
 #ifdef CONFIG_XEN_HAVE_PVMMU
+               /*
+                * We don't support PV MMU when Linux and Xen is using
+                * different page granularity.
+                */
+               BUILD_BUG_ON(XEN_PAGE_SIZE != PAGE_SIZE);
+
                if (!xen_feature(XENFEAT_auto_translated_physmap)) {
+                       unsigned long pfn = page_to_pfn(page);
+
                        if (!PageHighMem(page)) {
                                ret = HYPERVISOR_update_va_mapping(
                                                (unsigned long)__va(pfn << PAGE_SHIFT),
@@ -495,6 +528,7 @@ static enum bp_state decrease_reservation(unsigned long nr_pages, gfp_t gfp)
                        __set_phys_to_machine(pfn, INVALID_P2M_ENTRY);
                }
 #endif
+               list_del(&page->lru);
 
                balloon_append(page);
        }
@@ -603,6 +637,12 @@ int alloc_xenballooned_pages(int nr_pages, struct page **pages)
                if (page) {
                        pages[pgno++] = page;
 #ifdef CONFIG_XEN_HAVE_PVMMU
+                       /*
+                        * We don't support PV MMU when Linux and Xen is using
+                        * different page granularity.
+                        */
+                       BUILD_BUG_ON(XEN_PAGE_SIZE != PAGE_SIZE);
+
                        ret = xen_alloc_p2m_entry(page_to_pfn(page));
                        if (ret < 0)
                                goto out_undo;