x86_64: add KASan support

This patch adds arch specific code for kernel address sanitizer.

16TB of virtual addressed used for shadow memory. It's located in range
[ffffec0000000000 - fffffc0000000000] between vmemmap and %esp fixup
stacks.

At early stage we map whole shadow region with zero page. Latter, after
pages mapped to direct mapping address range we unmap zero pages from
corresponding shadow (see kasan_map_shadow()) and allocate and map a real
shadow memory reusing vmemmap_populate() function.

Also replace __pa with __pa_nodebug before shadow initialized. __pa with
CONFIG_DEBUG_VIRTUAL=y make external function call (__phys_addr)
__phys_addr is instrumented, so __asan_load could be called before shadow
area initialized.

Signed-off-by: Andrey Ryabinin <[e-mail]>
Cc: Dmitry Vyukov <[e-mail]>
Cc: Konstantin Serebryany <[e-mail]>
Cc: Dmitry Chernenkov <[e-mail]>
Signed-off-by: Andrey Konovalov <[e-mail]>
Cc: Yuri Gribov <[e-mail]>
Cc: Konstantin Khlebnikov <[e-mail]>
Cc: Sasha Levin <[e-mail]>
Cc: Christoph Lameter <[e-mail]>
Cc: Joonsoo Kim <[e-mail]>
Cc: Dave Hansen <[e-mail]>
Cc: Andi Kleen <[e-mail]>
Cc: Ingo Molnar <[e-mail]>
Cc: Thomas Gleixner <[e-mail]>
Cc: "H. Peter Anvin" <[e-mail]>
Cc: Christoph Lameter <[e-mail]>
Cc: Pekka Enberg <[e-mail]>
Cc: David Rientjes <[e-mail]>
Cc: Jim Davis <[e-mail]>
Signed-off-by: Andrew Morton <[e-mail]>
Signed-off-by: Linus Torvalds <[e-mail]>

Documentation/x86/x86_64/mm.txt

···
 ffffe90000000000 - ffffe9ffffffffff (=40 bits) hole
 ffffea0000000000 - ffffeaffffffffff (=40 bits) virtual memory map (1TB)
 ... unused hole ...
+ffffec0000000000 - fffffc0000000000 (=44 bits) kasan shadow memory (16TB)
+... unused hole ...
 ffffff0000000000 - ffffff7fffffffff (=39 bits) %esp fixup stacks
 ... unused hole ...
 ffffffff80000000 - ffffffffa0000000 (=512 MB)  kernel text mapping, from phys 0

arch/x86/Kconfig

···
 	select HAVE_CMPXCHG_LOCAL
 	select HAVE_CMPXCHG_DOUBLE
 	select HAVE_ARCH_KMEMCHECK
+	select HAVE_ARCH_KASAN if X86_64 && SPARSEMEM_VMEMMAP
 	select HAVE_USER_RETURN_NOTIFIER
 	select ARCH_BINFMT_ELF_RANDOMIZE_PIE
 	select HAVE_ARCH_JUMP_LABEL

arch/x86/boot/Makefile

···
 # Set it to -DSVGA_MODE=NORMAL_VGA if you just want the EGA/VGA mode.
 # The number is the same as you would ordinarily press at bootup.
 
+KASAN_SANITIZE := n
+
 SVGA_MODE	:= -DSVGA_MODE=NORMAL_VGA
 
 targets		:= vmlinux.bin setup.bin setup.elf bzImage

arch/x86/boot/compressed/Makefile

···
 #	(see scripts/Makefile.lib size_append)
 #	compressed vmlinux.bin.all + u32 size of vmlinux.bin.all
 
+KASAN_SANITIZE := n
+
 targets := vmlinux vmlinux.bin vmlinux.bin.gz vmlinux.bin.bz2 vmlinux.bin.lzma \
 	vmlinux.bin.xz vmlinux.bin.lzo vmlinux.bin.lz4
 

arch/x86/include/asm/kasan.h

···
+#ifndef _ASM_X86_KASAN_H
+#define _ASM_X86_KASAN_H
+
+/*
+ * Compiler uses shadow offset assuming that addresses start
+ * from 0. Kernel addresses don't start from 0, so shadow
+ * for kernel really starts from compiler's shadow offset +
+ * 'kernel address space start' >> KASAN_SHADOW_SCALE_SHIFT
+ */
+#define KASAN_SHADOW_START      (KASAN_SHADOW_OFFSET + \
+					(0xffff800000000000ULL >> 3))
+/* 47 bits for kernel address -> (47 - 3) bits for shadow */
+#define KASAN_SHADOW_END        (KASAN_SHADOW_START + (1ULL << (47 - 3)))
+
+#ifndef __ASSEMBLY__
+
+extern pte_t kasan_zero_pte[];
+extern pte_t kasan_zero_pmd[];
+extern pte_t kasan_zero_pud[];
+
+#ifdef CONFIG_KASAN
+void __init kasan_map_early_shadow(pgd_t *pgd);
+void __init kasan_init(void);
+#else
+static inline void kasan_map_early_shadow(pgd_t *pgd) { }
+static inline void kasan_init(void) { }
+#endif
+
+#endif
+
+#endif

arch/x86/kernel/Makefile

···
 CFLAGS_REMOVE_early_printk.o = -pg
 endif
 
+KASAN_SANITIZE_head$(BITS).o := n
+
 CFLAGS_irq.o := -I$(src)/../include/asm/trace
 
 obj-y			:= process_$(BITS).o signal.o entry_$(BITS).o

arch/x86/kernel/dumpstack.c

···
 	printk("SMP ");
 #endif
 #ifdef CONFIG_DEBUG_PAGEALLOC
-	printk("DEBUG_PAGEALLOC");
+	printk("DEBUG_PAGEALLOC ");
+#endif
+#ifdef CONFIG_KASAN
+	printk("KASAN");
 #endif
 	printk("\n");
 	if (notify_die(DIE_OOPS, str, regs, err,

arch/x86/kernel/head64.c

···
 #include <asm/bios_ebda.h>
 #include <asm/bootparam_utils.h>
 #include <asm/microcode.h>
+#include <asm/kasan.h>
 
 /*
  * Manage page tables very early on.
···
 
 	next_early_pgt = 0;
 
-	write_cr3(__pa(early_level4_pgt));
+	write_cr3(__pa_nodebug(early_level4_pgt));
 }
 
 /* Create a new PMD entry */
···
 	pmdval_t pmd, *pmd_p;
 
 	/* Invalid address or early pgt is done ?  */
-	if (physaddr >= MAXMEM || read_cr3() != __pa(early_level4_pgt))
+	if (physaddr >= MAXMEM || read_cr3() != __pa_nodebug(early_level4_pgt))
 		return -1;
 
 again:
···
 	/* Kill off the identity-map trampoline */
 	reset_early_page_tables();
 
+	kasan_map_early_shadow(early_level4_pgt);
+
 	/* clear bss before set_intr_gate with early_idt_handler */
 	clear_bss();
 
···
 	clear_page(init_level4_pgt);
 	/* set init_level4_pgt kernel high mapping*/
 	init_level4_pgt[511] = early_level4_pgt[511];
+
+	kasan_map_early_shadow(init_level4_pgt);
 
 	x86_64_start_reservations(real_mode_data);
 }

arch/x86/kernel/head_64.S

···
 	/* This must match the first entry in level2_kernel_pgt */
 	.quad   0x0000000000000000
 
+#ifdef CONFIG_KASAN
+#define FILL(VAL, COUNT)				\
+	.rept (COUNT) ;					\
+	.quad	(VAL) ;					\
+	.endr
+
+NEXT_PAGE(kasan_zero_pte)
+	FILL(kasan_zero_page - __START_KERNEL_map + _KERNPG_TABLE, 512)
+NEXT_PAGE(kasan_zero_pmd)
+	FILL(kasan_zero_pte - __START_KERNEL_map + _KERNPG_TABLE, 512)
+NEXT_PAGE(kasan_zero_pud)
+	FILL(kasan_zero_pmd - __START_KERNEL_map + _KERNPG_TABLE, 512)
+
+#undef FILL
+#endif
+
+
 #include "../../x86/xen/xen-head.S"
 	
 	__PAGE_ALIGNED_BSS
 NEXT_PAGE(empty_zero_page)
 	.skip PAGE_SIZE
+
+#ifdef CONFIG_KASAN
+/*
+ * This page used as early shadow. We don't use empty_zero_page
+ * at early stages, stack instrumentation could write some garbage
+ * to this page.
+ * Latter we reuse it as zero shadow for large ranges of memory
+ * that allowed to access, but not instrumented by kasan
+ * (vmalloc/vmemmap ...).
+ */
+NEXT_PAGE(kasan_zero_page)
+	.skip PAGE_SIZE
+#endif

arch/x86/kernel/setup.c

···
 #include <asm/cacheflush.h>
 #include <asm/processor.h>
 #include <asm/bugs.h>
+#include <asm/kasan.h>
 
 #include <asm/vsyscall.h>
 #include <asm/cpu.h>
···
 #endif
 
 	x86_init.paging.pagetable_init();
+
+	kasan_init();
 
 	if (boot_cpu_data.cpuid_level >= 0) {
 		/* A CPU has %cr4 if and only if it has CPUID */

arch/x86/mm/Makefile

···
 
 obj-$(CONFIG_KMEMCHECK)		+= kmemcheck/
 
+KASAN_SANITIZE_kasan_init_$(BITS).o := n
+obj-$(CONFIG_KASAN)		+= kasan_init_$(BITS).o
+
 obj-$(CONFIG_MMIOTRACE)		+= mmiotrace.o
 mmiotrace-y			:= kmmio.o pf_in.o mmio-mod.o
 obj-$(CONFIG_MMIOTRACE_TEST)	+= testmmiotrace.o

arch/x86/mm/kasan_init_64.c

···
+#include <linux/bootmem.h>
+#include <linux/kasan.h>
+#include <linux/kdebug.h>
+#include <linux/mm.h>
+#include <linux/sched.h>
+#include <linux/vmalloc.h>
+
+#include <asm/tlbflush.h>
+#include <asm/sections.h>
+
+extern pgd_t early_level4_pgt[PTRS_PER_PGD];
+extern struct range pfn_mapped[E820_X_MAX];
+
+extern unsigned char kasan_zero_page[PAGE_SIZE];
+
+static int __init map_range(struct range *range)
+{
+	unsigned long start;
+	unsigned long end;
+
+	start = (unsigned long)kasan_mem_to_shadow(pfn_to_kaddr(range->start));
+	end = (unsigned long)kasan_mem_to_shadow(pfn_to_kaddr(range->end));
+
+	/*
+	 * end + 1 here is intentional. We check several shadow bytes in advance
+	 * to slightly speed up fastpath. In some rare cases we could cross
+	 * boundary of mapped shadow, so we just map some more here.
+	 */
+	return vmemmap_populate(start, end + 1, NUMA_NO_NODE);
+}
+
+static void __init clear_pgds(unsigned long start,
+			unsigned long end)
+{
+	for (; start < end; start += PGDIR_SIZE)
+		pgd_clear(pgd_offset_k(start));
+}
+
+void __init kasan_map_early_shadow(pgd_t *pgd)
+{
+	int i;
+	unsigned long start = KASAN_SHADOW_START;
+	unsigned long end = KASAN_SHADOW_END;
+
+	for (i = pgd_index(start); start < end; i++) {
+		pgd[i] = __pgd(__pa_nodebug(kasan_zero_pud)
+				| _KERNPG_TABLE);
+		start += PGDIR_SIZE;
+	}
+}
+
+static int __init zero_pte_populate(pmd_t *pmd, unsigned long addr,
+				unsigned long end)
+{
+	pte_t *pte = pte_offset_kernel(pmd, addr);
+
+	while (addr + PAGE_SIZE <= end) {
+		WARN_ON(!pte_none(*pte));
+		set_pte(pte, __pte(__pa_nodebug(kasan_zero_page)
+					| __PAGE_KERNEL_RO));
+		addr += PAGE_SIZE;
+		pte = pte_offset_kernel(pmd, addr);
+	}
+	return 0;
+}
+
+static int __init zero_pmd_populate(pud_t *pud, unsigned long addr,
+				unsigned long end)
+{
+	int ret = 0;
+	pmd_t *pmd = pmd_offset(pud, addr);
+
+	while (IS_ALIGNED(addr, PMD_SIZE) && addr + PMD_SIZE <= end) {
+		WARN_ON(!pmd_none(*pmd));
+		set_pmd(pmd, __pmd(__pa_nodebug(kasan_zero_pte)
+					| __PAGE_KERNEL_RO));
+		addr += PMD_SIZE;
+		pmd = pmd_offset(pud, addr);
+	}
+	if (addr < end) {
+		if (pmd_none(*pmd)) {
+			void *p = vmemmap_alloc_block(PAGE_SIZE, NUMA_NO_NODE);
+			if (!p)
+				return -ENOMEM;
+			set_pmd(pmd, __pmd(__pa_nodebug(p) | _KERNPG_TABLE));
+		}
+		ret = zero_pte_populate(pmd, addr, end);
+	}
+	return ret;
+}
+
+
+static int __init zero_pud_populate(pgd_t *pgd, unsigned long addr,
+				unsigned long end)
+{
+	int ret = 0;
+	pud_t *pud = pud_offset(pgd, addr);
+
+	while (IS_ALIGNED(addr, PUD_SIZE) && addr + PUD_SIZE <= end) {
+		WARN_ON(!pud_none(*pud));
+		set_pud(pud, __pud(__pa_nodebug(kasan_zero_pmd)
+					| __PAGE_KERNEL_RO));
+		addr += PUD_SIZE;
+		pud = pud_offset(pgd, addr);
+	}
+
+	if (addr < end) {
+		if (pud_none(*pud)) {
+			void *p = vmemmap_alloc_block(PAGE_SIZE, NUMA_NO_NODE);
+			if (!p)
+				return -ENOMEM;
+			set_pud(pud, __pud(__pa_nodebug(p) | _KERNPG_TABLE));
+		}
+		ret = zero_pmd_populate(pud, addr, end);
+	}
+	return ret;
+}
+
+static int __init zero_pgd_populate(unsigned long addr, unsigned long end)
+{
+	int ret = 0;
+	pgd_t *pgd = pgd_offset_k(addr);
+
+	while (IS_ALIGNED(addr, PGDIR_SIZE) && addr + PGDIR_SIZE <= end) {
+		WARN_ON(!pgd_none(*pgd));
+		set_pgd(pgd, __pgd(__pa_nodebug(kasan_zero_pud)
+					| __PAGE_KERNEL_RO));
+		addr += PGDIR_SIZE;
+		pgd = pgd_offset_k(addr);
+	}
+
+	if (addr < end) {
+		if (pgd_none(*pgd)) {
+			void *p = vmemmap_alloc_block(PAGE_SIZE, NUMA_NO_NODE);
+			if (!p)
+				return -ENOMEM;
+			set_pgd(pgd, __pgd(__pa_nodebug(p) | _KERNPG_TABLE));
+		}
+		ret = zero_pud_populate(pgd, addr, end);
+	}
+	return ret;
+}
+
+
+static void __init populate_zero_shadow(const void *start, const void *end)
+{
+	if (zero_pgd_populate((unsigned long)start, (unsigned long)end))
+		panic("kasan: unable to map zero shadow!");
+}
+
+
+#ifdef CONFIG_KASAN_INLINE
+static int kasan_die_handler(struct notifier_block *self,
+			     unsigned long val,
+			     void *data)
+{
+	if (val == DIE_GPF) {
+		pr_emerg("CONFIG_KASAN_INLINE enabled");
+		pr_emerg("GPF could be caused by NULL-ptr deref or user memory access");
+	}
+	return NOTIFY_OK;
+}
+
+static struct notifier_block kasan_die_notifier = {
+	.notifier_call = kasan_die_handler,
+};
+#endif
+
+void __init kasan_init(void)
+{
+	int i;
+
+#ifdef CONFIG_KASAN_INLINE
+	register_die_notifier(&kasan_die_notifier);
+#endif
+
+	memcpy(early_level4_pgt, init_level4_pgt, sizeof(early_level4_pgt));
+	load_cr3(early_level4_pgt);
+
+	clear_pgds(KASAN_SHADOW_START, KASAN_SHADOW_END);
+
+	populate_zero_shadow((void *)KASAN_SHADOW_START,
+			kasan_mem_to_shadow((void *)PAGE_OFFSET));
+
+	for (i = 0; i < E820_X_MAX; i++) {
+		if (pfn_mapped[i].end == 0)
+			break;
+
+		if (map_range(&pfn_mapped[i]))
+			panic("kasan: unable to allocate shadow!");
+	}
+
+	populate_zero_shadow(kasan_mem_to_shadow((void *)PAGE_OFFSET + MAXMEM),
+				(void *)KASAN_SHADOW_END);
+
+	memset(kasan_zero_page, 0, PAGE_SIZE);
+
+	load_cr3(init_level4_pgt);
+}

arch/x86/realmode/Makefile

···
 # for more details.
 #
 #
-
+KASAN_SANITIZE := n
 subdir- := rm
 
 obj-y += init.o

arch/x86/realmode/rm/Makefile

···
 # for more details.
 #
 #
+KASAN_SANITIZE := n
 
 always := realmode.bin realmode.relocs
 

arch/x86/vdso/Makefile

···
 #
 
 KBUILD_CFLAGS += $(DISABLE_LTO)
+KASAN_SANITIZE := n
 
 VDSO64-$(CONFIG_X86_64)		:= y
 VDSOX32-$(CONFIG_X86_X32_ABI)	:= y

lib/Kconfig.kasan

···
 
 config KASAN_SHADOW_OFFSET
 	hex
+	default 0xdffffc0000000000 if X86_64
 
 choice
 	prompt "Instrumentation type"