From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001 From: Andrey Ryabinin Date: Thu, 28 Dec 2017 19:06:20 +0300 Subject: [PATCH] x86/mm: Set MODULES_END to 0xffffffffff000000 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit commit f5a40711fa58f1c109165a4fec6078bf2dfd2bdc upstream. Since f06bdd4001c2 ("x86/mm: Adapt MODULES_END based on fixmap section size") kasan_mem_to_shadow(MODULES_END) could be not aligned to a page boundary. So passing page unaligned address to kasan_populate_zero_shadow() have two possible effects: 1) It may leave one page hole in supposed to be populated area. After commit 21506525fb8d ("x86/kasan/64: Teach KASAN about the cpu_entry_area") that hole happens to be in the shadow covering fixmap area and leads to crash: BUG: unable to handle kernel paging request at fffffbffffe8ee04 RIP: 0010:check_memory_region+0x5c/0x190 Call Trace: memcpy+0x1f/0x50 ghes_copy_tofrom_phys+0xab/0x180 ghes_read_estatus+0xfb/0x280 ghes_notify_nmi+0x2b2/0x410 nmi_handle+0x115/0x2c0 default_do_nmi+0x57/0x110 do_nmi+0xf8/0x150 end_repeat_nmi+0x1a/0x1e Note, the crash likely disappeared after commit 92a0f81d8957, which changed kasan_populate_zero_shadow() call the way it was before commit 21506525fb8d. 2) Attempt to load module near MODULES_END will fail, because __vmalloc_node_range() called from kasan_module_alloc() will hit the WARN_ON(!pte_none(*pte)) in the vmap_pte_range() and bail out with error. To fix this we need to make kasan_mem_to_shadow(MODULES_END) page aligned which means that MODULES_END should be 8*PAGE_SIZE aligned. The whole point of commit f06bdd4001c2 was to move MODULES_END down if NR_CPUS is big, so the cpu_entry_area takes a lot of space. But since 92a0f81d8957 ("x86/cpu_entry_area: Move it out of the fixmap") the cpu_entry_area is no longer in fixmap, so we could just set MODULES_END to a fixed 8*PAGE_SIZE aligned address. Fixes: f06bdd4001c2 ("x86/mm: Adapt MODULES_END based on fixmap section size") Reported-by: Jakub Kicinski Signed-off-by: Andrey Ryabinin Signed-off-by: Thomas Gleixner Cc: Andy Lutomirski Cc: Thomas Garnier Link: https://lkml.kernel.org/r/20171228160620.23818-1-aryabinin@virtuozzo.com Signed-off-by: Greg Kroah-Hartman Signed-off-by: Fabian Grünbichler --- Documentation/x86/x86_64/mm.txt | 5 +---- arch/x86/include/asm/pgtable_64_types.h | 2 +- 2 files changed, 2 insertions(+), 5 deletions(-) diff --git a/Documentation/x86/x86_64/mm.txt b/Documentation/x86/x86_64/mm.txt index ad41b3813f0a..ddd5ffd31bd0 100644 --- a/Documentation/x86/x86_64/mm.txt +++ b/Documentation/x86/x86_64/mm.txt @@ -43,7 +43,7 @@ ffffff0000000000 - ffffff7fffffffff (=39 bits) %esp fixup stacks ffffffef00000000 - fffffffeffffffff (=64 GB) EFI region mapping space ... unused hole ... ffffffff80000000 - ffffffff9fffffff (=512 MB) kernel text mapping, from phys 0 -ffffffffa0000000 - [fixmap start] (~1526 MB) module mapping space +ffffffffa0000000 - fffffffffeffffff (1520 MB) module mapping space [fixmap start] - ffffffffff5fffff kernel-internal fixmap range ffffffffff600000 - ffffffffff600fff (=4 kB) legacy vsyscall ABI ffffffffffe00000 - ffffffffffffffff (=2 MB) unused hole @@ -67,9 +67,6 @@ memory window (this size is arbitrary, it can be raised later if needed). The mappings are not part of any other kernel PGD and are only available during EFI runtime calls. -The module mapping space size changes based on the CONFIG requirements for the -following fixmap section. - Note that if CONFIG_RANDOMIZE_MEMORY is enabled, the direct mapping of all physical memory, vmalloc/ioremap space and virtual memory map are randomized. Their order is preserved but their base will be offset early at boot time. diff --git a/arch/x86/include/asm/pgtable_64_types.h b/arch/x86/include/asm/pgtable_64_types.h index e8a809ee0bb6..c92bd73b1e46 100644 --- a/arch/x86/include/asm/pgtable_64_types.h +++ b/arch/x86/include/asm/pgtable_64_types.h @@ -103,7 +103,7 @@ typedef struct { pteval_t pte; } pte_t; #define MODULES_VADDR (__START_KERNEL_map + KERNEL_IMAGE_SIZE) /* The module sections ends with the start of the fixmap */ -#define MODULES_END __fix_to_virt(__end_of_fixed_addresses + 1) +#define MODULES_END _AC(0xffffffffff000000, UL) #define MODULES_LEN (MODULES_END - MODULES_VADDR) #define ESPFIX_PGD_ENTRY _AC(-2, UL) -- 2.14.2