Merge branch 'x86-mm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull x86 mm changes from Ingo Molnar:
 "The main changes in this cycle were:

   - reduce the x86/32 PAE per task PGD allocation overhead from 4K to
     0.032k (Fenghua Yu)

   - early_ioremap/memunmap() usage cleanups (Juergen Gross)

   - gbpages support cleanups (Luis R Rodriguez)

   - improve AMD Bulldozer (family 0x15) ASLR I$ aliasing workaround to
     increase randomization by 3 bits (per bootup) (Hector
     Marco-Gisbert)

   - misc fixlets"

* 'x86-mm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/mm: Improve AMD Bulldozer ASLR workaround
  x86/mm/pat: Initialize __cachemode2pte_tbl[] and __pte2cachemode_tbl[] in a bit more readable fashion
  init.h: Clean up the __setup()/early_param() macros
  x86/mm: Simplify probe_page_size_mask()
  x86/mm: Further simplify 1 GB kernel linear mappings handling
  x86/mm: Use early_param_on_off() for direct_gbpages
  init.h: Add early_param_on_off()
  x86/mm: Simplify enabling direct_gbpages
  x86/mm: Use IS_ENABLED() for direct_gbpages
  x86/mm: Unexport set_memory_ro() and set_memory_rw()
  x86/mm, efi: Use early_ioremap() in arch/x86/platform/efi/efi-bgrt.c
  x86/mm: Use early_memunmap() instead of early_iounmap()
  x86/mm/pat: Ensure different messages in STRICT_DEVMEM and PAT cases
  x86/mm: Reduce PAE-mode per task pgd allocation overhead from 4K to 32 bytes

arch/x86/Kconfig

@@ -1295,14 +1295,14 @@ config ARCH_DMA_ADDR_T_64BIT
 	def_bool y
 	depends on X86_64 || HIGHMEM64G
 
-config DIRECT_GBPAGES
-	bool "Enable 1GB pages for kernel pagetables" if EXPERT
-	default y
-	depends on X86_64
+config X86_DIRECT_GBPAGES
+	def_bool y
+	depends on X86_64 && !DEBUG_PAGEALLOC && !KMEMCHECK
 	---help---
-	  Allow the kernel linear mapping to use 1GB pages on CPUs that
-	  support it. This can improve the kernel's performance a tiny bit by
-	  reducing TLB pressure. If in doubt, say "Y".
+	  Certain kernel features effectively disable kernel
+	  linear 1 GB mappings (even if the CPU otherwise
+	  supports them), so don't confuse the user by printing
+	  that we have them enabled.
 
 # Common NUMA Features
 config NUMA

arch/x86/include/asm/elf.h

@@ -366,6 +366,7 @@ enum align_flags {
 struct va_alignment {
 	int flags;
 	unsigned long mask;
+	unsigned long bits;
 } ____cacheline_aligned;
 
 extern struct va_alignment va_align;

arch/x86/kernel/cpu/amd.c

@@ -5,6 +5,7 @@
 
 #include <linux/io.h>
 #include <linux/sched.h>
+#include <linux/random.h>
 #include <asm/processor.h>
 #include <asm/apic.h>
 #include <asm/cpu.h>
@@ -488,6 +489,9 @@ static void bsp_init_amd(struct cpuinfo_x86 *c)
 
 		va_align.mask	  = (upperbit - 1) & PAGE_MASK;
 		va_align.flags    = ALIGN_VA_32 | ALIGN_VA_64;
+
+		/* A random value per boot for bit slice [12:upper_bit) */
+		va_align.bits = get_random_int() & va_align.mask;
 	}
 }
 

arch/x86/kernel/devicetree.c

@@ -286,13 +286,13 @@ static void __init x86_flattree_get_config(void)
 	initial_boot_params = dt = early_memremap(initial_dtb, map_len);
 	size = of_get_flat_dt_size();
 	if (map_len < size) {
-		early_iounmap(dt, map_len);
+		early_memunmap(dt, map_len);
 		initial_boot_params = dt = early_memremap(initial_dtb, size);
 		map_len = size;
 	}
 
 	unflatten_and_copy_device_tree();
-	early_iounmap(dt, map_len);
+	early_memunmap(dt, map_len);
 }
 #else
 static inline void x86_flattree_get_config(void) { }

arch/x86/kernel/e820.c

@@ -661,7 +661,7 @@ void __init parse_e820_ext(u64 phys_addr, u32 data_len)
 	extmap = (struct e820entry *)(sdata->data);
 	__append_e820_map(extmap, entries);
 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
-	early_iounmap(sdata, data_len);
+	early_memunmap(sdata, data_len);
 	printk(KERN_INFO "e820: extended physical RAM map:\n");
 	e820_print_map("extended");
 }

arch/x86/kernel/setup.c

@@ -354,7 +354,7 @@ static void __init relocate_initrd(void)
 		mapaddr = ramdisk_image & PAGE_MASK;
 		p = early_memremap(mapaddr, clen+slop);
 		memcpy(q, p+slop, clen);
-		early_iounmap(p, clen+slop);
+		early_memunmap(p, clen+slop);
 		q += clen;
 		ramdisk_image += clen;
 		ramdisk_size  -= clen;
@@ -438,7 +438,7 @@ static void __init parse_setup_data(void)
 		data_len = data->len + sizeof(struct setup_data);
 		data_type = data->type;
 		pa_next = data->next;
-		early_iounmap(data, sizeof(*data));
+		early_memunmap(data, sizeof(*data));
 
 		switch (data_type) {
 		case SETUP_E820_EXT:
@@ -470,7 +470,7 @@ static void __init e820_reserve_setup_data(void)
 			 E820_RAM, E820_RESERVED_KERN);
 		found = 1;
 		pa_data = data->next;
-		early_iounmap(data, sizeof(*data));
+		early_memunmap(data, sizeof(*data));
 	}
 	if (!found)
 		return;
@@ -491,7 +491,7 @@ static void __init memblock_x86_reserve_range_setup_data(void)
 		data = early_memremap(pa_data, sizeof(*data));
 		memblock_reserve(pa_data, sizeof(*data) + data->len);
 		pa_data = data->next;
-		early_iounmap(data, sizeof(*data));
+		early_memunmap(data, sizeof(*data));
 	}
 }
 

arch/x86/kernel/sys_x86_64.c

@@ -34,10 +34,26 @@ static unsigned long get_align_mask(void)
 	return va_align.mask;
 }
 
+/*
+ * To avoid aliasing in the I$ on AMD F15h, the bits defined by the
+ * va_align.bits, [12:upper_bit), are set to a random value instead of
+ * zeroing them. This random value is computed once per boot. This form
+ * of ASLR is known as "per-boot ASLR".
+ *
+ * To achieve this, the random value is added to the info.align_offset
+ * value before calling vm_unmapped_area() or ORed directly to the
+ * address.
+ */
+static unsigned long get_align_bits(void)
+{
+	return va_align.bits & get_align_mask();
+}
+
 unsigned long align_vdso_addr(unsigned long addr)
 {
 	unsigned long align_mask = get_align_mask();
-	return (addr + align_mask) & ~align_mask;
+	addr = (addr + align_mask) & ~align_mask;
+	return addr | get_align_bits();
 }
 
 static int __init control_va_addr_alignment(char *str)
@@ -135,8 +151,12 @@ arch_get_unmapped_area(struct file *filp, unsigned long addr,
 	info.length = len;
 	info.low_limit = begin;
 	info.high_limit = end;
-	info.align_mask = filp ? get_align_mask() : 0;
+	info.align_mask = 0;
 	info.align_offset = pgoff << PAGE_SHIFT;
+	if (filp) {
+		info.align_mask = get_align_mask();
+		info.align_offset += get_align_bits();
+	}
 	return vm_unmapped_area(&info);
 }
 
@@ -174,8 +194,12 @@ arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
 	info.length = len;
 	info.low_limit = PAGE_SIZE;
 	info.high_limit = mm->mmap_base;
-	info.align_mask = filp ? get_align_mask() : 0;
+	info.align_mask = 0;
 	info.align_offset = pgoff << PAGE_SHIFT;
+	if (filp) {
+		info.align_mask = get_align_mask();
+		info.align_offset += get_align_bits();
+	}
 	addr = vm_unmapped_area(&info);
 	if (!(addr & ~PAGE_MASK))
 		return addr;

arch/x86/mm/init.c

@@ -29,29 +29,33 @@
 
 /*
  * Tables translating between page_cache_type_t and pte encoding.
- * Minimal supported modes are defined statically, modified if more supported
- * cache modes are available.
- * Index into __cachemode2pte_tbl is the cachemode.
- * Index into __pte2cachemode_tbl are the caching attribute bits of the pte
+ *
+ * Minimal supported modes are defined statically, they are modified
+ * during bootup if more supported cache modes are available.
+ *
+ *   Index into __cachemode2pte_tbl[] is the cachemode.
+ *
+ *   Index into __pte2cachemode_tbl[] are the caching attribute bits of the pte
  *   (_PAGE_PWT, _PAGE_PCD, _PAGE_PAT) at index bit positions 0, 1, 2.
  */
 uint16_t __cachemode2pte_tbl[_PAGE_CACHE_MODE_NUM] = {
-	[_PAGE_CACHE_MODE_WB]		= 0,
-	[_PAGE_CACHE_MODE_WC]		= _PAGE_PWT,
-	[_PAGE_CACHE_MODE_UC_MINUS]	= _PAGE_PCD,
-	[_PAGE_CACHE_MODE_UC]		= _PAGE_PCD | _PAGE_PWT,
-	[_PAGE_CACHE_MODE_WT]		= _PAGE_PCD,
-	[_PAGE_CACHE_MODE_WP]		= _PAGE_PCD,
+	[_PAGE_CACHE_MODE_WB      ]	= 0         | 0        ,
+	[_PAGE_CACHE_MODE_WC      ]	= _PAGE_PWT | 0        ,
+	[_PAGE_CACHE_MODE_UC_MINUS]	= 0         | _PAGE_PCD,
+	[_PAGE_CACHE_MODE_UC      ]	= _PAGE_PWT | _PAGE_PCD,
+	[_PAGE_CACHE_MODE_WT      ]	= 0         | _PAGE_PCD,
+	[_PAGE_CACHE_MODE_WP      ]	= 0         | _PAGE_PCD,
 };
 EXPORT_SYMBOL(__cachemode2pte_tbl);
+
 uint8_t __pte2cachemode_tbl[8] = {
-	[__pte2cm_idx(0)] = _PAGE_CACHE_MODE_WB,
-	[__pte2cm_idx(_PAGE_PWT)] = _PAGE_CACHE_MODE_WC,
-	[__pte2cm_idx(_PAGE_PCD)] = _PAGE_CACHE_MODE_UC_MINUS,
-	[__pte2cm_idx(_PAGE_PWT | _PAGE_PCD)] = _PAGE_CACHE_MODE_UC,
-	[__pte2cm_idx(_PAGE_PAT)] = _PAGE_CACHE_MODE_WB,
-	[__pte2cm_idx(_PAGE_PWT | _PAGE_PAT)] = _PAGE_CACHE_MODE_WC,
-	[__pte2cm_idx(_PAGE_PCD | _PAGE_PAT)] = _PAGE_CACHE_MODE_UC_MINUS,
+	[__pte2cm_idx( 0        | 0         | 0        )] = _PAGE_CACHE_MODE_WB,
+	[__pte2cm_idx(_PAGE_PWT | 0         | 0        )] = _PAGE_CACHE_MODE_WC,
+	[__pte2cm_idx( 0        | _PAGE_PCD | 0        )] = _PAGE_CACHE_MODE_UC_MINUS,
+	[__pte2cm_idx(_PAGE_PWT | _PAGE_PCD | 0        )] = _PAGE_CACHE_MODE_UC,
+	[__pte2cm_idx( 0        | 0         | _PAGE_PAT)] = _PAGE_CACHE_MODE_WB,
+	[__pte2cm_idx(_PAGE_PWT | 0         | _PAGE_PAT)] = _PAGE_CACHE_MODE_WC,
+	[__pte2cm_idx(0         | _PAGE_PCD | _PAGE_PAT)] = _PAGE_CACHE_MODE_UC_MINUS,
 	[__pte2cm_idx(_PAGE_PWT | _PAGE_PCD | _PAGE_PAT)] = _PAGE_CACHE_MODE_UC,
 };
 EXPORT_SYMBOL(__pte2cachemode_tbl);
@@ -131,21 +135,7 @@ void  __init early_alloc_pgt_buf(void)
 
 int after_bootmem;
 
-int direct_gbpages
-#ifdef CONFIG_DIRECT_GBPAGES
-				= 1
-#endif
-;
-
-static void __init init_gbpages(void)
-{
-#ifdef CONFIG_X86_64
-	if (direct_gbpages && cpu_has_gbpages)
-		printk(KERN_INFO "Using GB pages for direct mapping\n");
-	else
-		direct_gbpages = 0;
-#endif
-}
+early_param_on_off("gbpages", "nogbpages", direct_gbpages, CONFIG_X86_DIRECT_GBPAGES);
 
 struct map_range {
 	unsigned long start;
@@ -157,16 +147,12 @@ static int page_size_mask;
 
 static void __init probe_page_size_mask(void)
 {
-	init_gbpages();
-
 #if !defined(CONFIG_DEBUG_PAGEALLOC) && !defined(CONFIG_KMEMCHECK)
 	/*
 	 * For CONFIG_DEBUG_PAGEALLOC, identity mapping will use small pages.
 	 * This will simplify cpa(), which otherwise needs to support splitting
 	 * large pages into small in interrupt context, etc.
 	 */
-	if (direct_gbpages)
-		page_size_mask |= 1 << PG_LEVEL_1G;
 	if (cpu_has_pse)
 		page_size_mask |= 1 << PG_LEVEL_2M;
 #endif
@@ -181,6 +167,14 @@ static void __init probe_page_size_mask(void)
 		__supported_pte_mask |= _PAGE_GLOBAL;
 	} else
 		__supported_pte_mask &= ~_PAGE_GLOBAL;
+
+	/* Enable 1 GB linear kernel mappings if available: */
+	if (direct_gbpages && cpu_has_gbpages) {
+		printk(KERN_INFO "Using GB pages for direct mapping\n");
+		page_size_mask |= 1 << PG_LEVEL_1G;
+	} else {
+		direct_gbpages = 0;
+	}
 }
 
 #ifdef CONFIG_X86_32

arch/x86/mm/init_64.c

@@ -130,20 +130,6 @@ int kernel_ident_mapping_init(struct x86_mapping_info *info, pgd_t *pgd_page,
 	return 0;
 }
 
-static int __init parse_direct_gbpages_off(char *arg)
-{
-	direct_gbpages = 0;
-	return 0;
-}
-early_param("nogbpages", parse_direct_gbpages_off);
-
-static int __init parse_direct_gbpages_on(char *arg)
-{
-	direct_gbpages = 1;
-	return 0;
-}
-early_param("gbpages", parse_direct_gbpages_on);
-
 /*
  * NOTE: pagetable_init alloc all the fixmap pagetables contiguous on the
  * physical space so we can cache the place of the first one and move

arch/x86/mm/pageattr.c

@@ -81,11 +81,9 @@ void arch_report_meminfo(struct seq_file *m)
 	seq_printf(m, "DirectMap4M:    %8lu kB\n",
 			direct_pages_count[PG_LEVEL_2M] << 12);
 #endif
-#ifdef CONFIG_X86_64
 	if (direct_gbpages)
 		seq_printf(m, "DirectMap1G:    %8lu kB\n",
 			direct_pages_count[PG_LEVEL_1G] << 20);
-#endif
 }
 #else
 static inline void split_page_count(int level) { }
@@ -1654,13 +1652,11 @@ int set_memory_ro(unsigned long addr, int numpages)
 {
 	return change_page_attr_clear(&addr, numpages, __pgprot(_PAGE_RW), 0);
 }
-EXPORT_SYMBOL_GPL(set_memory_ro);
 
 int set_memory_rw(unsigned long addr, int numpages)
 {
 	return change_page_attr_set(&addr, numpages, __pgprot(_PAGE_RW), 0);
 }
-EXPORT_SYMBOL_GPL(set_memory_rw);
 
 int set_memory_np(unsigned long addr, int numpages)
 {

arch/x86/mm/pat.c

@@ -610,7 +610,7 @@ pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
 }
 
 #ifdef CONFIG_STRICT_DEVMEM
-/* This check is done in drivers/char/mem.c in case of STRICT_DEVMEM*/
+/* This check is done in drivers/char/mem.c in case of STRICT_DEVMEM */
 static inline int range_is_allowed(unsigned long pfn, unsigned long size)
 {
 	return 1;
@@ -628,7 +628,7 @@ static inline int range_is_allowed(unsigned long pfn, unsigned long size)
 
 	while (cursor < to) {
 		if (!devmem_is_allowed(pfn)) {
-			printk(KERN_INFO "Program %s tried to access /dev/mem between [mem %#010Lx-%#010Lx]\n",
+			printk(KERN_INFO "Program %s tried to access /dev/mem between [mem %#010Lx-%#010Lx], PAT prevents it\n",
 			       current->comm, from, to - 1);
 			return 0;
 		}

arch/x86/mm/pgtable.c

@@ -275,12 +275,87 @@ static void pgd_prepopulate_pmd(struct mm_struct *mm, pgd_t *pgd, pmd_t *pmds[])
 	}
 }
 
+/*
+ * Xen paravirt assumes pgd table should be in one page. 64 bit kernel also
+ * assumes that pgd should be in one page.
+ *
+ * But kernel with PAE paging that is not running as a Xen domain
+ * only needs to allocate 32 bytes for pgd instead of one page.
+ */
+#ifdef CONFIG_X86_PAE
+
+#include <linux/slab.h>
+
+#define PGD_SIZE	(PTRS_PER_PGD * sizeof(pgd_t))
+#define PGD_ALIGN	32
+
+static struct kmem_cache *pgd_cache;
+
+static int __init pgd_cache_init(void)
+{
+	/*
+	 * When PAE kernel is running as a Xen domain, it does not use
+	 * shared kernel pmd. And this requires a whole page for pgd.
+	 */
+	if (!SHARED_KERNEL_PMD)
+		return 0;
+
+	/*
+	 * when PAE kernel is not running as a Xen domain, it uses
+	 * shared kernel pmd. Shared kernel pmd does not require a whole
+	 * page for pgd. We are able to just allocate a 32-byte for pgd.
+	 * During boot time, we create a 32-byte slab for pgd table allocation.
+	 */
+	pgd_cache = kmem_cache_create("pgd_cache", PGD_SIZE, PGD_ALIGN,
+				      SLAB_PANIC, NULL);
+	if (!pgd_cache)
+		return -ENOMEM;
+
+	return 0;
+}
+core_initcall(pgd_cache_init);
+
+static inline pgd_t *_pgd_alloc(void)
+{
+	/*
+	 * If no SHARED_KERNEL_PMD, PAE kernel is running as a Xen domain.
+	 * We allocate one page for pgd.
+	 */
+	if (!SHARED_KERNEL_PMD)
+		return (pgd_t *)__get_free_page(PGALLOC_GFP);
+
+	/*
+	 * Now PAE kernel is not running as a Xen domain. We can allocate
+	 * a 32-byte slab for pgd to save memory space.
+	 */
+	return kmem_cache_alloc(pgd_cache, PGALLOC_GFP);
+}
+
+static inline void _pgd_free(pgd_t *pgd)
+{
+	if (!SHARED_KERNEL_PMD)
+		free_page((unsigned long)pgd);
+	else
+		kmem_cache_free(pgd_cache, pgd);
+}
+#else
+static inline pgd_t *_pgd_alloc(void)
+{
+	return (pgd_t *)__get_free_page(PGALLOC_GFP);
+}
+
+static inline void _pgd_free(pgd_t *pgd)
+{
+	free_page((unsigned long)pgd);
+}
+#endif /* CONFIG_X86_PAE */
+
 pgd_t *pgd_alloc(struct mm_struct *mm)
 {
 	pgd_t *pgd;
 	pmd_t *pmds[PREALLOCATED_PMDS];
 
-	pgd = (pgd_t *)__get_free_page(PGALLOC_GFP);
+	pgd = _pgd_alloc();
 
 	if (pgd == NULL)
 		goto out;
@@ -310,7 +385,7 @@ pgd_t *pgd_alloc(struct mm_struct *mm)
 out_free_pmds:
 	free_pmds(mm, pmds);
 out_free_pgd:
-	free_page((unsigned long)pgd);
+	_pgd_free(pgd);
 out:
 	return NULL;
 }
@@ -320,7 +395,7 @@ void pgd_free(struct mm_struct *mm, pgd_t *pgd)
 	pgd_mop_up_pmds(mm, pgd);
 	pgd_dtor(pgd);
 	paravirt_pgd_free(mm, pgd);
-	free_page((unsigned long)pgd);
+	_pgd_free(pgd);
 }
 
 /*

arch/x86/platform/efi/efi-bgrt.c

@@ -67,7 +67,7 @@ void __init efi_bgrt_init(void)
 
 	image = efi_lookup_mapped_addr(bgrt_tab->image_address);
 	if (!image) {
-		image = early_memremap(bgrt_tab->image_address,
+		image = early_ioremap(bgrt_tab->image_address,
 				       sizeof(bmp_header));
 		ioremapped = true;
 		if (!image) {
@@ -89,7 +89,7 @@ void __init efi_bgrt_init(void)
 	}
 
 	if (ioremapped) {
-		image = early_memremap(bgrt_tab->image_address,
+		image = early_ioremap(bgrt_tab->image_address,
 				       bmp_header.size);
 		if (!image) {
 			pr_err("Ignoring BGRT: failed to map image memory\n");

include/linux/init.h

@@ -263,11 +263,31 @@ struct obs_kernel_param {
 #define __setup(str, fn)						\
 	__setup_param(str, fn, fn, 0)
 
-/* NOTE: fn is as per module_param, not __setup!  Emits warning if fn
- * returns non-zero. */
+/*
+ * NOTE: fn is as per module_param, not __setup!
+ * Emits warning if fn returns non-zero.
+ */
 #define early_param(str, fn)						\
 	__setup_param(str, fn, fn, 1)
 
+#define early_param_on_off(str_on, str_off, var, config)		\
+									\
+	int var = IS_ENABLED(config);					\
+									\
+	static int __init parse_##var##_on(char *arg)			\
+	{								\
+		var = 1;						\
+		return 0;						\
+	}								\
+	__setup_param(str_on, parse_##var##_on, parse_##var##_on, 1);	\
+									\
+	static int __init parse_##var##_off(char *arg)			\
+	{								\
+		var = 0;						\
+		return 0;						\
+	}								\
+	__setup_param(str_off, parse_##var##_off, parse_##var##_off, 1)
+
 /* Relies on boot_command_line being set */
 void __init parse_early_param(void);
 void __init parse_early_options(char *cmdline);