Merge tag 's390-6.11-4' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux

Pull s390 fixes from Vasily Gorbik:

 - Fix KASLR base offset to account for symbol offsets in the vmlinux
   ELF file, preventing tool breakages like the drgn debugger

 - Fix potential memory corruption of physmem_info during kernel
   physical address randomization

 - Fix potential memory corruption due to overlap between the relocated
   lowcore and identity mapping by correctly reserving lowcore memory

 - Fix performance regression and avoid randomizing identity mapping
   base by default

 - Fix unnecessary delay of AP bus binding complete uevent to prevent
   startup lag in KVM guests using AP

* tag 's390-6.11-4' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390/boot: Fix KASLR base offset off by __START_KERNEL bytes
  s390/boot: Avoid possible physmem_info segment corruption
  s390/ap: Refine AP bus bindings complete processing
  s390/mm: Pin identity mapping base to zero
  s390/mm: Prevent lowcore vs identity mapping overlap

arch/s390/Kconfig

@@ -604,6 +604,19 @@ config RANDOMIZE_BASE
 	  as a security feature that deters exploit attempts relying on
 	  knowledge of the location of kernel internals.
 
+config RANDOMIZE_IDENTITY_BASE
+	bool "Randomize the address of the identity mapping base"
+	depends on RANDOMIZE_BASE
+	default DEBUG_VM
+	help
+	  The identity mapping base address is pinned to zero by default.
+	  Allow randomization of that base to expose otherwise missed
+	  notion of physical and virtual addresses of data structures.
+	  That does not have any impact on the base address at which the
+	  kernel image is loaded.
+
+	  If unsure, say N
+
 config KERNEL_IMAGE_BASE
 	hex "Kernel image base address"
 	range 0x100000 0x1FFFFFE0000000 if !KASAN

arch/s390/boot/startup.c

@@ -162,7 +162,7 @@ static void kaslr_adjust_relocs(unsigned long min_addr, unsigned long max_addr,
 		loc = (long)*reloc + phys_offset;
 		if (loc < min_addr || loc > max_addr)
 			error("64-bit relocation outside of kernel!\n");
-		*(u64 *)loc += offset - __START_KERNEL;
+		*(u64 *)loc += offset;
 	}
 }
 
@@ -177,7 +177,7 @@ static void kaslr_adjust_got(unsigned long offset)
 	 */
 	for (entry = (u64 *)vmlinux.got_start; entry < (u64 *)vmlinux.got_end; entry++) {
 		if (*entry)
-			*entry += offset - __START_KERNEL;
+			*entry += offset;
 	}
 }
 
@@ -252,7 +252,7 @@ static unsigned long setup_kernel_memory_layout(unsigned long kernel_size)
 	vmemmap_size = SECTION_ALIGN_UP(pages) * sizeof(struct page);
 
 	/* choose kernel address space layout: 4 or 3 levels. */
-	BUILD_BUG_ON(!IS_ALIGNED(__START_KERNEL, THREAD_SIZE));
+	BUILD_BUG_ON(!IS_ALIGNED(TEXT_OFFSET, THREAD_SIZE));
 	BUILD_BUG_ON(!IS_ALIGNED(__NO_KASLR_START_KERNEL, THREAD_SIZE));
 	BUILD_BUG_ON(__NO_KASLR_END_KERNEL > _REGION1_SIZE);
 	vsize = get_vmem_size(ident_map_size, vmemmap_size, vmalloc_size, _REGION3_SIZE);
@@ -341,7 +341,8 @@ static unsigned long setup_kernel_memory_layout(unsigned long kernel_size)
 	BUILD_BUG_ON(MAX_DCSS_ADDR > (1UL << MAX_PHYSMEM_BITS));
 	max_mappable = max(ident_map_size, MAX_DCSS_ADDR);
 	max_mappable = min(max_mappable, vmemmap_start);
-	__identity_base = round_down(vmemmap_start - max_mappable, rte_size);
+	if (IS_ENABLED(CONFIG_RANDOMIZE_IDENTITY_BASE))
+		__identity_base = round_down(vmemmap_start - max_mappable, rte_size);
 
 	return asce_limit;
 }
@@ -388,31 +389,25 @@ static void kaslr_adjust_vmlinux_info(long offset)
 #endif
 }
 
-static void fixup_vmlinux_info(void)
-{
-	vmlinux.entry -= __START_KERNEL;
-	kaslr_adjust_vmlinux_info(-__START_KERNEL);
-}
-
 void startup_kernel(void)
 {
-	unsigned long kernel_size = vmlinux.image_size + vmlinux.bss_size;
-	unsigned long nokaslr_offset_phys, kaslr_large_page_offset;
-	unsigned long amode31_lma = 0;
+	unsigned long vmlinux_size = vmlinux.image_size + vmlinux.bss_size;
+	unsigned long nokaslr_text_lma, text_lma = 0, amode31_lma = 0;
+	unsigned long kernel_size = TEXT_OFFSET + vmlinux_size;
+	unsigned long kaslr_large_page_offset;
 	unsigned long max_physmem_end;
 	unsigned long asce_limit;
 	unsigned long safe_addr;
 	psw_t psw;
 
-	fixup_vmlinux_info();
 	setup_lpp();
 
 	/*
 	 * Non-randomized kernel physical start address must be _SEGMENT_SIZE
 	 * aligned (see blow).
 	 */
-	nokaslr_offset_phys = ALIGN(mem_safe_offset(), _SEGMENT_SIZE);
-	safe_addr = PAGE_ALIGN(nokaslr_offset_phys + kernel_size);
+	nokaslr_text_lma = ALIGN(mem_safe_offset(), _SEGMENT_SIZE);
+	safe_addr = PAGE_ALIGN(nokaslr_text_lma + vmlinux_size);
 
 	/*
 	 * Reserve decompressor memory together with decompression heap,
@@ -456,16 +451,27 @@ void startup_kernel(void)
 	 */
 	kaslr_large_page_offset = __kaslr_offset & ~_SEGMENT_MASK;
 	if (kaslr_enabled()) {
-		unsigned long end = ident_map_size - kaslr_large_page_offset;
+		unsigned long size = vmlinux_size + kaslr_large_page_offset;
 
-		__kaslr_offset_phys = randomize_within_range(kernel_size, _SEGMENT_SIZE, 0, end);
+		text_lma = randomize_within_range(size, _SEGMENT_SIZE, TEXT_OFFSET, ident_map_size);
 	}
-	if (!__kaslr_offset_phys)
-		__kaslr_offset_phys = nokaslr_offset_phys;
-	__kaslr_offset_phys |= kaslr_large_page_offset;
+	if (!text_lma)
+		text_lma = nokaslr_text_lma;
+	text_lma |= kaslr_large_page_offset;
+
+	/*
+	 * [__kaslr_offset_phys..__kaslr_offset_phys + TEXT_OFFSET] region is
+	 * never accessed via the kernel image mapping as per the linker script:
+	 *
+	 *	. = TEXT_OFFSET;
+	 *
+	 * Therefore, this region could be used for something else and does
+	 * not need to be reserved. See how it is skipped in setup_vmem().
+	 */
+	__kaslr_offset_phys = text_lma - TEXT_OFFSET;
 	kaslr_adjust_vmlinux_info(__kaslr_offset_phys);
-	physmem_reserve(RR_VMLINUX, __kaslr_offset_phys, kernel_size);
-	deploy_kernel((void *)__kaslr_offset_phys);
+	physmem_reserve(RR_VMLINUX, text_lma, vmlinux_size);
+	deploy_kernel((void *)text_lma);
 
 	/* vmlinux decompression is done, shrink reserved low memory */
 	physmem_reserve(RR_DECOMPRESSOR, 0, (unsigned long)_decompressor_end);
@@ -488,7 +494,7 @@ void startup_kernel(void)
 		amode31_lma = randomize_within_range(vmlinux.amode31_size, PAGE_SIZE, amode31_min, SZ_2G);
 	}
 	if (!amode31_lma)
-		amode31_lma = __kaslr_offset_phys - vmlinux.amode31_size;
+		amode31_lma = text_lma - vmlinux.amode31_size;
 	physmem_reserve(RR_AMODE31, amode31_lma, vmlinux.amode31_size);
 
 	/*
@@ -504,8 +510,8 @@ void startup_kernel(void)
 	 * - copy_bootdata() must follow setup_vmem() to propagate changes
 	 *   to bootdata made by setup_vmem()
 	 */
-	clear_bss_section(__kaslr_offset_phys);
-	kaslr_adjust_relocs(__kaslr_offset_phys, __kaslr_offset_phys + vmlinux.image_size,
+	clear_bss_section(text_lma);
+	kaslr_adjust_relocs(text_lma, text_lma + vmlinux.image_size,
 			    __kaslr_offset, __kaslr_offset_phys);
 	kaslr_adjust_got(__kaslr_offset);
 	setup_vmem(__kaslr_offset, __kaslr_offset + kernel_size, asce_limit);

arch/s390/boot/vmem.c

@@ -90,7 +90,7 @@ static void kasan_populate_shadow(unsigned long kernel_start, unsigned long kern
 		}
 		memgap_start = end;
 	}
-	kasan_populate(kernel_start, kernel_end, POPULATE_KASAN_MAP_SHADOW);
+	kasan_populate(kernel_start + TEXT_OFFSET, kernel_end, POPULATE_KASAN_MAP_SHADOW);
 	kasan_populate(0, (unsigned long)__identity_va(0), POPULATE_KASAN_ZERO_SHADOW);
 	kasan_populate(AMODE31_START, AMODE31_END, POPULATE_KASAN_ZERO_SHADOW);
 	if (IS_ENABLED(CONFIG_KASAN_VMALLOC)) {
@@ -475,7 +475,17 @@ void setup_vmem(unsigned long kernel_start, unsigned long kernel_end, unsigned l
 				 (unsigned long)__identity_va(end),
 				 POPULATE_IDENTITY);
 	}
-	pgtable_populate(kernel_start, kernel_end, POPULATE_KERNEL);
+
+	/*
+	 * [kernel_start..kernel_start + TEXT_OFFSET] region is never
+	 * accessed as per the linker script:
+	 *
+	 *	. = TEXT_OFFSET;
+	 *
+	 * Therefore, skip mapping TEXT_OFFSET bytes to prevent access to
+	 * [__kaslr_offset_phys..__kaslr_offset_phys + TEXT_OFFSET] region.
+	 */
+	pgtable_populate(kernel_start + TEXT_OFFSET, kernel_end, POPULATE_KERNEL);
 	pgtable_populate(AMODE31_START, AMODE31_END, POPULATE_DIRECT);
 	pgtable_populate(__abs_lowcore, __abs_lowcore + sizeof(struct lowcore),
 			 POPULATE_ABS_LOWCORE);

arch/s390/boot/vmlinux.lds.S

@@ -109,7 +109,12 @@ SECTIONS
 #ifdef CONFIG_KERNEL_UNCOMPRESSED
 	. = ALIGN(PAGE_SIZE);
 	. += AMODE31_SIZE;		/* .amode31 section */
-	. = ALIGN(1 << 20);		/* _SEGMENT_SIZE */
+
+	/*
+	 * Make sure the location counter is not less than TEXT_OFFSET.
+	 * _SEGMENT_SIZE is not available, use ALIGN(1 << 20) instead.
+	 */
+	. = MAX(TEXT_OFFSET, ALIGN(1 << 20));
 #else
 	. = ALIGN(8);
 #endif

arch/s390/include/asm/page.h

@@ -279,8 +279,9 @@ static inline unsigned long virt_to_pfn(const void *kaddr)
 #define AMODE31_SIZE		(3 * PAGE_SIZE)
 
 #define KERNEL_IMAGE_SIZE	(512 * 1024 * 1024)
-#define __START_KERNEL		0x100000
 #define __NO_KASLR_START_KERNEL	CONFIG_KERNEL_IMAGE_BASE
 #define __NO_KASLR_END_KERNEL	(__NO_KASLR_START_KERNEL + KERNEL_IMAGE_SIZE)
 
+#define TEXT_OFFSET		0x100000
+
 #endif /* _S390_PAGE_H */

arch/s390/kernel/setup.c

@@ -734,7 +734,23 @@ static void __init memblock_add_physmem_info(void)
 }
 
 /*
- * Reserve memory used for lowcore/command line/kernel image.
+ * Reserve memory used for lowcore.
+ */
+static void __init reserve_lowcore(void)
+{
+	void *lowcore_start = get_lowcore();
+	void *lowcore_end = lowcore_start + sizeof(struct lowcore);
+	void *start, *end;
+
+	if ((void *)__identity_base < lowcore_end) {
+		start = max(lowcore_start, (void *)__identity_base);
+		end = min(lowcore_end, (void *)(__identity_base + ident_map_size));
+		memblock_reserve(__pa(start), __pa(end));
+	}
+}
+
+/*
+ * Reserve memory used for absolute lowcore/command line/kernel image.
  */
 static void __init reserve_kernel(void)
 {
@@ -918,6 +934,7 @@ void __init setup_arch(char **cmdline_p)
 
 	/* Do some memory reservations *before* memory is added to memblock */
 	reserve_pgtables();
+	reserve_lowcore();
 	reserve_kernel();
 	reserve_initrd();
 	reserve_certificate_list();

arch/s390/kernel/vmlinux.lds.S

@@ -39,7 +39,7 @@ PHDRS {
 
 SECTIONS
 {
-	. = __START_KERNEL;
+	. = TEXT_OFFSET;
 	.text : {
 		_stext = .;		/* Start of text section */
 		_text = .;		/* Text and read-only data */

arch/s390/tools/relocs.c

@@ -280,7 +280,7 @@ static int do_reloc(struct section *sec, Elf_Rel *rel)
 	case R_390_GOTOFF64:
 		break;
 	case R_390_64:
-		add_reloc(&relocs64, offset - ehdr.e_entry);
+		add_reloc(&relocs64, offset);
 		break;
 	default:
 		die("Unsupported relocation type: %d\n", r_type);

drivers/s390/crypto/ap_bus.c

@@ -971,11 +971,16 @@ int ap_driver_register(struct ap_driver *ap_drv, struct module *owner,
 		       char *name)
 {
 	struct device_driver *drv = &ap_drv->driver;
+	int rc;
 
 	drv->bus = &ap_bus_type;
 	drv->owner = owner;
 	drv->name = name;
-	return driver_register(drv);
+	rc = driver_register(drv);
+
+	ap_check_bindings_complete();
+
+	return rc;
 }
 EXPORT_SYMBOL(ap_driver_register);