s390/kasan: add initialization code and enable it

Kasan needs 1/8 of kernel virtual address space to be reserved as the
shadow area. And eventually it requires the shadow memory offset to be
known at compile time (passed to the compiler when full instrumentation
is enabled).  Any value picked as the shadow area offset for 3-level
paging would eat up identity mapping on 4-level paging (with 1PB
shadow area size). So, the kernel sticks to 3-level paging when kasan
is enabled. 3TB border is picked as the shadow offset.  The memory
layout is adjusted so, that physical memory border does not exceed
KASAN_SHADOW_START and vmemmap does not go below KASAN_SHADOW_END.

Due to the fact that on s390 paging is set up very late and to cover
more code with kasan instrumentation, temporary identity mapping and
final shadow memory are set up early. The shadow memory mapping is
later carried over to init_mm.pgd during paging_init.

For the needs of paging structures allocation and shadow memory
population a primitive allocator is used, which simply chops off
memory blocks from the end of the physical memory.

Kasan currenty doesn't track vmemmap and vmalloc areas.

Current memory layout (for 3-level paging, 2GB physical memory).

---[ Identity Mapping ]---
0x0000000000000000-0x0000000000100000
---[ Kernel Image Start ]---
0x0000000000100000-0x0000000002b00000
---[ Kernel Image End ]---
0x0000000002b00000-0x0000000080000000        2G <- physical memory border
0x0000000080000000-0x0000030000000000     3070G PUD I
---[ Kasan Shadow Start ]---
0x0000030000000000-0x0000030010000000      256M PMD RW X  <- shadow for 2G memory
0x0000030010000000-0x0000037ff0000000   523776M PTE RO NX <- kasan zero ro page
0x0000037ff0000000-0x0000038000000000      256M PMD RW X  <- shadow for 2G modules
---[ Kasan Shadow End ]---
0x0000038000000000-0x000003d100000000      324G PUD I
---[ vmemmap Area ]---
0x000003d100000000-0x000003e080000000
---[ vmalloc Area ]---
0x000003e080000000-0x000003ff80000000
---[ Modules Area ]---
0x000003ff80000000-0x0000040000000000        2G
Acked-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: NVasily Gorbik <gor@linux.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/Kconfig

@@ -120,6 +120,7 @@ config S390
 	select HAVE_ALIGNED_STRUCT_PAGE if SLUB
 	select HAVE_ARCH_AUDITSYSCALL
 	select HAVE_ARCH_JUMP_LABEL
+	select HAVE_ARCH_KASAN
 	select CPU_NO_EFFICIENT_FFS if !HAVE_MARCH_Z9_109_FEATURES
 	select HAVE_ARCH_SECCOMP_FILTER
 	select HAVE_ARCH_SOFT_DIRTY

arch/s390/include/asm/kasan.h

+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef __ASM_KASAN_H
+#define __ASM_KASAN_H
+
+#include <asm/pgtable.h>
+
+#ifdef CONFIG_KASAN
+
+#define KASAN_SHADOW_SCALE_SHIFT 3
+#define KASAN_SHADOW_SIZE						       \
+	(_AC(1, UL) << (_REGION2_SHIFT - KASAN_SHADOW_SCALE_SHIFT))
+#define KASAN_SHADOW_OFFSET	_AC(0x30000000000, UL)
+#define KASAN_SHADOW_START	KASAN_SHADOW_OFFSET
+#define KASAN_SHADOW_END	(KASAN_SHADOW_START + KASAN_SHADOW_SIZE)
+
+extern void kasan_early_init(void);
+extern void kasan_copy_shadow(pgd_t *dst);
+#else
+static inline void kasan_early_init(void) { }
+static inline void kasan_copy_shadow(pgd_t *dst) { }
+#endif
+
+#endif

arch/s390/include/asm/pgtable.h

@@ -1181,6 +1181,7 @@ static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
 
 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
+#define pgd_offset_raw(pgd, addr) ((pgd) + pgd_index(addr))
 
 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
 #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)

arch/s390/kernel/early_nobss.c

@@ -14,6 +14,7 @@
 #include <asm/sections.h>
 #include <asm/lowcore.h>
 #include <asm/timex.h>
+#include <asm/kasan.h>
 #include "entry.h"
 
 static void __init reset_tod_clock(void)
@@ -40,4 +41,5 @@ void __init startup_init_nobss(void)
 {
 	reset_tod_clock();
 	clear_bss_section();
+	kasan_early_init();
 }

arch/s390/kernel/setup.c

@@ -531,14 +531,19 @@ static void __init setup_memory_end(void)
 {
 	unsigned long vmax, vmalloc_size, tmp;
 
-	/* Choose kernel address space layout: 2, 3, or 4 levels. */
+	/* Choose kernel address space layout: 3 or 4 levels. */
 	vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
-	tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
-	tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
-	if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
+	if (IS_ENABLED(CONFIG_KASAN)) {
 		vmax = _REGION2_SIZE; /* 3-level kernel page table */
-	else
-		vmax = _REGION1_SIZE; /* 4-level kernel page table */
+	} else {
+		tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
+		tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
+		if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
+			vmax = _REGION2_SIZE; /* 3-level kernel page table */
+		else
+			vmax = _REGION1_SIZE; /* 4-level kernel page table */
+	}
+
 	/* module area is at the end of the kernel address space. */
 	MODULES_END = vmax;
 	MODULES_VADDR = MODULES_END - MODULES_LEN;
@@ -556,6 +561,11 @@ static void __init setup_memory_end(void)
 
 	/* Take care that memory_end is set and <= vmemmap */
 	memory_end = min(memory_end ?: max_physmem_end, tmp);
+#ifdef CONFIG_KASAN
+	/* fit in kasan shadow memory region between 1:1 and vmemmap */
+	memory_end = min(memory_end, KASAN_SHADOW_START);
+	vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
+#endif
 	max_pfn = max_low_pfn = PFN_DOWN(memory_end);
 	memblock_remove(memory_end, ULONG_MAX);
 

arch/s390/mm/Makefile

@@ -10,3 +10,6 @@ obj-$(CONFIG_CMM)		+= cmm.o
 obj-$(CONFIG_HUGETLB_PAGE)	+= hugetlbpage.o
 obj-$(CONFIG_S390_PTDUMP)	+= dump_pagetables.o
 obj-$(CONFIG_PGSTE)		+= gmap.o
+
+KASAN_SANITIZE_kasan_init.o	:= n
+obj-$(CONFIG_KASAN)		+= kasan_init.o

arch/s390/mm/init.c

@@ -42,6 +42,7 @@
 #include <asm/ctl_reg.h>
 #include <asm/sclp.h>
 #include <asm/set_memory.h>
+#include <asm/kasan.h>
 
 pgd_t swapper_pg_dir[PTRS_PER_PGD] __section(.bss..swapper_pg_dir);
 
@@ -98,8 +99,9 @@ void __init paging_init(void)
 	S390_lowcore.user_asce = S390_lowcore.kernel_asce;
 	crst_table_init((unsigned long *) init_mm.pgd, pgd_type);
 	vmem_map_init();
+	kasan_copy_shadow(init_mm.pgd);
 
-        /* enable virtual mapping in kernel mode */
+	/* enable virtual mapping in kernel mode */
 	__ctl_load(S390_lowcore.kernel_asce, 1, 1);
 	__ctl_load(S390_lowcore.kernel_asce, 7, 7);
 	__ctl_load(S390_lowcore.kernel_asce, 13, 13);

arch/s390/mm/kasan_init.c

+// SPDX-License-Identifier: GPL-2.0
+#include <linux/kasan.h>
+#include <linux/sched/task.h>
+#include <linux/memblock.h>
+#include <asm/pgalloc.h>
+#include <asm/pgtable.h>
+#include <asm/kasan.h>
+#include <asm/processor.h>
+#include <asm/sclp.h>
+#include <asm/sections.h>
+#include <asm/setup.h>
+
+static unsigned long pgalloc_pos __initdata;
+static unsigned long pgalloc_low __initdata;
+
+#define __sha(x) ((unsigned long)kasan_mem_to_shadow((void *)x))
+
+static pgd_t early_pg_dir[PTRS_PER_PGD] __initdata __aligned(PAGE_SIZE);
+
+static void __init kasan_early_panic(const char *reason)
+{
+	sclp_early_printk("The Linux kernel failed to boot with the KernelAddressSanitizer:\n");
+	sclp_early_printk(reason);
+	disabled_wait(0);
+}
+
+static void * __init kasan_early_alloc_pages(unsigned int order)
+{
+	pgalloc_pos -= (PAGE_SIZE << order);
+
+	if (pgalloc_pos < pgalloc_low)
+		kasan_early_panic("out of memory during initialisation\n");
+
+	return (void *)pgalloc_pos;
+}
+
+static void * __init kasan_early_crst_alloc(unsigned long val)
+{
+	unsigned long *table;
+
+	table = kasan_early_alloc_pages(CRST_ALLOC_ORDER);
+	if (table)
+		crst_table_init(table, val);
+	return table;
+}
+
+static pte_t * __init kasan_early_pte_alloc(void)
+{
+	static void *pte_leftover;
+	pte_t *pte;
+
+	BUILD_BUG_ON(_PAGE_TABLE_SIZE * 2 != PAGE_SIZE);
+
+	if (!pte_leftover) {
+		pte_leftover = kasan_early_alloc_pages(0);
+		pte = pte_leftover + _PAGE_TABLE_SIZE;
+	} else {
+		pte = pte_leftover;
+		pte_leftover = NULL;
+	}
+	memset64((u64 *)pte, _PAGE_INVALID, PTRS_PER_PTE);
+	return pte;
+}
+
+enum populate_mode {
+	POPULATE_ONE2ONE,
+	POPULATE_MAP,
+	POPULATE_ZERO_SHADOW
+};
+static void __init kasan_early_vmemmap_populate(unsigned long address,
+						unsigned long end,
+						enum populate_mode mode)
+{
+	unsigned long pgt_prot_zero, pgt_prot;
+	pgd_t *pg_dir;
+	p4d_t *p4_dir;
+	pud_t *pu_dir;
+	pmd_t *pm_dir;
+	pte_t *pt_dir;
+
+	pgt_prot_zero = pgprot_val(PAGE_KERNEL_RO);
+	pgt_prot_zero &= ~_PAGE_NOEXEC;
+	pgt_prot = pgprot_val(PAGE_KERNEL_EXEC);
+
+	while (address < end) {
+		pg_dir = pgd_offset_k(address);
+		if (pgd_none(*pg_dir)) {
+			if (mode == POPULATE_ZERO_SHADOW &&
+			    IS_ALIGNED(address, PGDIR_SIZE) &&
+			    end - address >= PGDIR_SIZE) {
+				pgd_populate(&init_mm, pg_dir, kasan_zero_p4d);
+				address = (address + PGDIR_SIZE) & PGDIR_MASK;
+				continue;
+			}
+			p4_dir = kasan_early_crst_alloc(_REGION2_ENTRY_EMPTY);
+			pgd_populate(&init_mm, pg_dir, p4_dir);
+		}
+
+		p4_dir = p4d_offset(pg_dir, address);
+		if (p4d_none(*p4_dir)) {
+			if (mode == POPULATE_ZERO_SHADOW &&
+			    IS_ALIGNED(address, P4D_SIZE) &&
+			    end - address >= P4D_SIZE) {
+				p4d_populate(&init_mm, p4_dir, kasan_zero_pud);
+				address = (address + P4D_SIZE) & P4D_MASK;
+				continue;
+			}
+			pu_dir = kasan_early_crst_alloc(_REGION3_ENTRY_EMPTY);
+			p4d_populate(&init_mm, p4_dir, pu_dir);
+		}
+
+		pu_dir = pud_offset(p4_dir, address);
+		if (pud_none(*pu_dir)) {
+			if (mode == POPULATE_ZERO_SHADOW &&
+			    IS_ALIGNED(address, PUD_SIZE) &&
+			    end - address >= PUD_SIZE) {
+				pud_populate(&init_mm, pu_dir, kasan_zero_pmd);
+				address = (address + PUD_SIZE) & PUD_MASK;
+				continue;
+			}
+			pm_dir = kasan_early_crst_alloc(_SEGMENT_ENTRY_EMPTY);
+			pud_populate(&init_mm, pu_dir, pm_dir);
+		}
+
+		pm_dir = pmd_offset(pu_dir, address);
+		if (pmd_none(*pm_dir)) {
+			if (mode == POPULATE_ZERO_SHADOW &&
+			    IS_ALIGNED(address, PMD_SIZE) &&
+			    end - address >= PMD_SIZE) {
+				pmd_populate(&init_mm, pm_dir, kasan_zero_pte);
+				address = (address + PMD_SIZE) & PMD_MASK;
+				continue;
+			}
+			pt_dir = kasan_early_pte_alloc();
+			pmd_populate(&init_mm, pm_dir, pt_dir);
+		}
+
+		pt_dir = pte_offset_kernel(pm_dir, address);
+		if (pte_none(*pt_dir)) {
+			void *page;
+
+			switch (mode) {
+			case POPULATE_ONE2ONE:
+				page = (void *)address;
+				pte_val(*pt_dir) = __pa(page) | pgt_prot;
+				break;
+			case POPULATE_MAP:
+				page = kasan_early_alloc_pages(0);
+				memset(page, 0, PAGE_SIZE);
+				pte_val(*pt_dir) = __pa(page) | pgt_prot;
+				break;
+			case POPULATE_ZERO_SHADOW:
+				page = kasan_zero_page;
+				pte_val(*pt_dir) = __pa(page) | pgt_prot_zero;
+				break;
+			}
+		}
+		address += PAGE_SIZE;
+	}
+}
+
+static void __init kasan_set_pgd(pgd_t *pgd, unsigned long asce_type)
+{
+	unsigned long asce_bits;
+
+	asce_bits = asce_type | _ASCE_TABLE_LENGTH;
+	S390_lowcore.kernel_asce = (__pa(pgd) & PAGE_MASK) | asce_bits;
+	S390_lowcore.user_asce = S390_lowcore.kernel_asce;
+
+	__ctl_load(S390_lowcore.kernel_asce, 1, 1);
+	__ctl_load(S390_lowcore.kernel_asce, 7, 7);
+	__ctl_load(S390_lowcore.kernel_asce, 13, 13);
+}
+
+static void __init kasan_enable_dat(void)
+{
+	psw_t psw;
+
+	psw.mask = __extract_psw();
+	psw_bits(psw).dat = 1;
+	psw_bits(psw).as = PSW_BITS_AS_HOME;
+	__load_psw_mask(psw.mask);
+}
+
+void __init kasan_early_init(void)
+{
+	unsigned long untracked_mem_end;
+	unsigned long shadow_alloc_size;
+	unsigned long initrd_end;
+	unsigned long asce_type;
+	unsigned long memsize;
+	unsigned long vmax;
+	unsigned long pgt_prot = pgprot_val(PAGE_KERNEL_RO);
+	pte_t pte_z;
+	pmd_t pmd_z = __pmd(__pa(kasan_zero_pte) | _SEGMENT_ENTRY);
+	pud_t pud_z = __pud(__pa(kasan_zero_pmd) | _REGION3_ENTRY);
+	p4d_t p4d_z = __p4d(__pa(kasan_zero_pud) | _REGION2_ENTRY);
+
+	pgt_prot &= ~_PAGE_NOEXEC;
+	pte_z = __pte(__pa(kasan_zero_page) | pgt_prot);
+
+	/* 3 level paging */
+	BUILD_BUG_ON(!IS_ALIGNED(KASAN_SHADOW_START, PUD_SIZE));
+	BUILD_BUG_ON(!IS_ALIGNED(KASAN_SHADOW_END, PUD_SIZE));
+	crst_table_init((unsigned long *)early_pg_dir, _REGION3_ENTRY_EMPTY);
+	untracked_mem_end = vmax = _REGION2_SIZE;
+	asce_type = _ASCE_TYPE_REGION3;
+
+	/* init kasan zero shadow */
+	crst_table_init((unsigned long *)kasan_zero_p4d, p4d_val(p4d_z));
+	crst_table_init((unsigned long *)kasan_zero_pud, pud_val(pud_z));
+	crst_table_init((unsigned long *)kasan_zero_pmd, pmd_val(pmd_z));
+	memset64((u64 *)kasan_zero_pte, pte_val(pte_z), PTRS_PER_PTE);
+
+	memsize = min(max_physmem_end, KASAN_SHADOW_START);
+	shadow_alloc_size = memsize >> KASAN_SHADOW_SCALE_SHIFT;
+	if (IS_ENABLED(CONFIG_MODULES))
+		shadow_alloc_size += MODULES_LEN >> KASAN_SHADOW_SCALE_SHIFT;
+	pgalloc_low = round_up((unsigned long)_end, _SEGMENT_SIZE);
+	if (IS_ENABLED(CONFIG_BLK_DEV_INITRD)) {
+		initrd_end =
+		    round_up(INITRD_START + INITRD_SIZE, _SEGMENT_SIZE);
+		pgalloc_low = max(pgalloc_low, initrd_end);
+	}
+
+	if (pgalloc_low + shadow_alloc_size > memsize)
+		kasan_early_panic("out of memory during initialisation\n");
+
+	pgalloc_pos = memsize;
+	init_mm.pgd = early_pg_dir;
+	/*
+	 * Current memory layout:
+	 * +- 0 -------------+	 +- shadow start -+
+	 * | 1:1 ram mapping |	/| 1/8 ram	  |
+	 * +- end of ram ----+ / +----------------+
+	 * | ... gap ...     |/  |	kasan	  |
+	 * +- shadow start --+	 |	zero	  |
+	 * | 1/8 addr space  |	 |	page	  |
+	 * +- shadow end    -+	 |	mapping	  |
+	 * | ... gap ...     |\  |    (untracked) |
+	 * +- modules vaddr -+ \ +----------------+
+	 * | 2Gb	     |	\| 256Mb	  |
+	 * +-----------------+	 +- shadow end ---+
+	 */
+	/* populate identity mapping */
+	kasan_early_vmemmap_populate(0, memsize, POPULATE_ONE2ONE);
+	/* populate kasan shadow (for identity mapping / modules / zero page) */
+	kasan_early_vmemmap_populate(__sha(0), __sha(memsize), POPULATE_MAP);
+	if (IS_ENABLED(CONFIG_MODULES)) {
+		untracked_mem_end = vmax - MODULES_LEN;
+		kasan_early_vmemmap_populate(__sha(untracked_mem_end),
+					     __sha(vmax), POPULATE_MAP);
+	}
+	kasan_early_vmemmap_populate(__sha(memsize), __sha(untracked_mem_end),
+				     POPULATE_ZERO_SHADOW);
+	kasan_set_pgd(early_pg_dir, asce_type);
+	kasan_enable_dat();
+	/* enable kasan */
+	init_task.kasan_depth = 0;
+	memblock_reserve(pgalloc_pos, memsize - pgalloc_pos);
+	sclp_early_printk("KernelAddressSanitizer initialized\n");
+}
+
+void __init kasan_copy_shadow(pgd_t *pg_dir)
+{
+	/*
+	 * At this point we are still running on early pages setup early_pg_dir,
+	 * while swapper_pg_dir has just been initialized with identity mapping.
+	 * Carry over shadow memory region from early_pg_dir to swapper_pg_dir.
+	 */
+
+	pgd_t *pg_dir_src;
+	pgd_t *pg_dir_dst;
+	p4d_t *p4_dir_src;
+	p4d_t *p4_dir_dst;
+	pud_t *pu_dir_src;
+	pud_t *pu_dir_dst;
+
+	pg_dir_src = pgd_offset_raw(early_pg_dir, KASAN_SHADOW_START);
+	pg_dir_dst = pgd_offset_raw(pg_dir, KASAN_SHADOW_START);
+	p4_dir_src = p4d_offset(pg_dir_src, KASAN_SHADOW_START);
+	p4_dir_dst = p4d_offset(pg_dir_dst, KASAN_SHADOW_START);
+	if (!p4d_folded(*p4_dir_src)) {
+		/* 4 level paging */
+		memcpy(p4_dir_dst, p4_dir_src,
+		       (KASAN_SHADOW_SIZE >> P4D_SHIFT) * sizeof(p4d_t));
+		return;
+	}
+	/* 3 level paging */
+	pu_dir_src = pud_offset(p4_dir_src, KASAN_SHADOW_START);
+	pu_dir_dst = pud_offset(p4_dir_dst, KASAN_SHADOW_START);
+	memcpy(pu_dir_dst, pu_dir_src,
+	       (KASAN_SHADOW_SIZE >> PUD_SHIFT) * sizeof(pud_t));
+}