[PATCH RFC v1 00/26] Add KernelMemorySanitizer infrastructure

[PATCH RFC v1 00/26] Add KernelMemorySanitizer infrastructure

[glider]

KernelMemorySanitizer (KMSAN) is a detector of errors related to uses of
uninitialized memory. It relies on compile-time Clang instrumentation
(similar to MSan: https://clang.llvm.org/docs/MemorySanitizer.html)
and tracks the state of every bit of kernel memory, being able to report
an error if uninitialized value is used in a condition, dereferenced or
copied to userspace, USB or network.

KMSAN has reported more than 200 bugs in the past two years, most of
them with the help of syzkaller (http://syzkaller.appspot.com).

The proposed patchset contains KMSAN runtime implementation together
with small changes to other subsystems needed to make KMSAN work.
The latter changes fall into several categories:
 - nice-to-have features that are independent from KMSAN but simplify
   its implementation (stackdepot changes, CONFIG_GENERIC_CSUM etc.);
 - Kconfig changes that prohibit options incompatible with KMSAN;
 - calls to KMSAN runtime functions that help KMSAN do the bookkeeping
   (e.g. tell it to allocate, copy or delete the metadata);
 - calls to KMSAN runtime functions that tell KMSAN to check memory
   escaping the kernel for uninitialized values. These are required to
   increase the number of true positive error reports;
 - calls to runtime functions that tell KMSAN to ignore certain memory
   ranges to avoid false negative reports. Most certainly there can be
   better ways to deal with every such report.

This patchset allows one to boot and run a defconfig+KMSAN kernel on a QEMU
without known major false positives. It however doesn't guarantee there
are no false positives in drivers of certain devices or less tested
subsystems, although KMSAN is actively tested on syzbot with quite a
rich config.

One may find it handy to review these patches in Gerrit:
https://linux-review.googlesource.com/c/linux/kernel/git/torvalds/linux/+/1081
I've ensured the Change-Id: tags stay away from commit descriptions.

Note: checkpatch.pl complains a lot about the use of BUG_ON in KMSAN
source. I don't have a strong opinion on this, but KMSAN is a debugging
tool, so any runtime invariant violation in it renders the tool useless.
Therefore it doesn't make much sense to not terminate after a bug in
KMSAN.

Alexander Potapenko (26):
  stackdepot: check depot_index before accessing the stack slab
  stackdepot: prevent Clang from optimizing away stackdepot_memcmp()
  kasan: stackdepot: move filter_irq_stacks() to stackdepot.c
  stackdepot: reserve 5 extra bits in depot_stack_handle_t
  printk_safe: externalize printk_context
  kasan: compiler.h: rename __no_kasan_or_inline into
    __no_memory_tool_or_inline
  kmsan: add ReST documentation
  kmsan: gfp: introduce __GFP_NO_KMSAN_SHADOW
  kmsan: introduce __no_sanitize_memory and __SANITIZE_MEMORY__
  kmsan: reduce vmalloc space
  kmsan: add KMSAN runtime
  kmsan: x86: sync metadata pages on page fault
  kmsan: add tests for KMSAN
  kmsan: make READ_ONCE_TASK_STACK() return initialized values
  kmsan: Kconfig changes to disable options incompatible with KMSAN
  kmsan: Changing existing files to enable KMSAN builds
  kmsan: disable KMSAN instrumentation for certain kernel parts
  kmsan: mm: call KMSAN hooks from SLUB code
  kmsan: call KMSAN hooks where needed
  kmsan: disable instrumentation of certain functions
  kmsan: unpoison |tlb| in arch_tlb_gather_mmu()
  kmsan: use __msan_memcpy() where possible.
  kmsan: unpoisoning buffers from devices etc.
  kmsan: hooks for copy_to_user() and friends
  kmsan: disable strscpy() optimization under KMSAN
  net: kasan: kmsan: support CONFIG_GENERIC_CSUM on x86, enable it for
    KASAN/KMSAN

To: Alexander Potapenko <glider@google.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Eric Van Hensbergen <ericvh@gmail.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Martin K. Petersen <martin.petersen@oracle.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Takashi Iwai <tiwai@suse.com>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Wolfram Sang <wsa@the-dreams.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Ilya Leoshkevich <iii@linux.ibm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: linux-mm@kvack.org

 Documentation/dev-tools/index.rst       |   1 +
 Documentation/dev-tools/kmsan.rst       | 418 +++++++++++++++++
 Makefile                                |   3 +-
 arch/Kconfig                            |   2 +-
 arch/s390/include/asm/processor.h       |   4 +-
 arch/x86/Kconfig                        |   7 +
 arch/x86/Kconfig.debug                  |   3 +
 arch/x86/boot/Makefile                  |   2 +
 arch/x86/boot/compressed/Makefile       |   2 +
 arch/x86/boot/compressed/misc.h         |   1 +
 arch/x86/entry/common.c                 |   1 +
 arch/x86/entry/entry_64.S               |  18 +
 arch/x86/entry/vdso/Makefile            |   3 +
 arch/x86/include/asm/checksum.h         |  10 +-
 arch/x86/include/asm/irq_regs.h         |   1 +
 arch/x86/include/asm/kmsan.h            | 129 ++++++
 arch/x86/include/asm/page_64.h          |  13 +
 arch/x86/include/asm/page_64_types.h    |  12 +-
 arch/x86/include/asm/pgtable_64_types.h |  15 +
 arch/x86/include/asm/string_64.h        |   9 +-
 arch/x86/include/asm/syscall_wrapper.h  |   1 +
 arch/x86/include/asm/uaccess.h          |  12 +
 arch/x86/include/asm/unwind.h           |   9 +-
 arch/x86/kernel/Makefile                |   4 +
 arch/x86/kernel/apic/apic.c             |   2 +
 arch/x86/kernel/cpu/Makefile            |   1 +
 arch/x86/kernel/dumpstack_64.c          |   1 +
 arch/x86/kernel/process_64.c            |   5 +
 arch/x86/kernel/traps.c                 |  12 +-
 arch/x86/kernel/uprobes.c               |   7 +-
 arch/x86/lib/Makefile                   |   2 +
 arch/x86/mm/Makefile                    |   2 +
 arch/x86/mm/fault.c                     |  20 +
 arch/x86/mm/ioremap.c                   |   3 +
 arch/x86/realmode/rm/Makefile           |   2 +
 block/bio.c                             |  20 +
 block/blk.h                             |   7 +
 block/partition-generic.c               |   9 +-
 crypto/Kconfig                          |  26 ++
 drivers/char/random.c                   |   2 +
 drivers/firmware/efi/libstub/Makefile   |   1 +
 drivers/gpu/drm/amd/display/Kconfig     |   2 +-
 drivers/i2c/i2c-core-base.c             |   2 +
 drivers/input/serio/libps2.c            |   6 +-
 drivers/scsi/scsi_lib.c                 |   4 +
 drivers/usb/core/message.c              |   6 +-
 drivers/usb/core/urb.c                  |   2 +
 drivers/virtio/virtio_ring.c            |  14 +
 fs/buffer.c                             |   7 +-
 include/asm-generic/cacheflush.h        |   7 +-
 include/asm-generic/uaccess.h           |  12 +-
 include/linux/compiler-clang.h          |   8 +
 include/linux/compiler-gcc.h            |   5 +
 include/linux/compiler.h                |  21 +-
 include/linux/dma-mapping.h             |   2 +
 include/linux/gfp.h                     |   4 +-
 include/linux/highmem.h                 |   4 +
 include/linux/kmsan-checks.h            | 123 +++++
 include/linux/kmsan.h                   | 143 ++++++
 include/linux/mm_types.h                |   9 +
 include/linux/sched.h                   |   5 +
 include/linux/skbuff.h                  |   5 +-
 include/linux/stackdepot.h              |  10 +
 include/linux/string.h                  |   2 +
 include/linux/uaccess.h                 |  32 +-
 init/main.c                             |   3 +
 kernel/Makefile                         |   1 +
 kernel/exit.c                           |   2 +
 kernel/fork.c                           |   2 +
 kernel/kthread.c                        |   2 +
 kernel/printk/printk.c                  |   8 +-
 kernel/printk/printk_safe.c             |   2 +-
 kernel/profile.c                        |   1 +
 kernel/sched/core.c                     |  11 +
 kernel/softirq.c                        |   5 +
 lib/Kconfig.debug                       |   5 +
 lib/Kconfig.kmsan                       |  22 +
 lib/Makefile                            |   2 +
 lib/ioremap.c                           |   5 +
 lib/iov_iter.c                          |   6 +
 lib/stackdepot.c                        |  72 ++-
 lib/string.c                            |   5 +-
 lib/test_kmsan.c                        | 221 +++++++++
 lib/usercopy.c                          |   6 +-
 mm/Makefile                             |   1 +
 mm/compaction.c                         |   9 +
 mm/filemap.c                            |   3 +
 mm/gup.c                                |   3 +
 mm/kasan/common.c                       |  23 -
 mm/kmsan/Makefile                       |   4 +
 mm/kmsan/kmsan.c                        | 570 ++++++++++++++++++++++++
 mm/kmsan/kmsan.h                        | 147 ++++++
 mm/kmsan/kmsan_entry.c                  | 130 ++++++
 mm/kmsan/kmsan_hooks.c                  | 416 +++++++++++++++++
 mm/kmsan/kmsan_init.c                   |  88 ++++
 mm/kmsan/kmsan_instr.c                  | 259 +++++++++++
 mm/kmsan/kmsan_report.c                 | 135 ++++++
 mm/kmsan/kmsan_shadow.c                 | 543 ++++++++++++++++++++++
 mm/kmsan/kmsan_shadow.h                 |  30 ++
 mm/memory.c                             |   2 +
 mm/mmu_gather.c                         |  10 +
 mm/page_alloc.c                         |  16 +
 mm/readahead.c                          |   6 +
 mm/slub.c                               |  37 +-
 mm/vmalloc.c                            |  23 +-
 net/9p/protocol.c                       |   2 +
 net/sched/sch_generic.c                 |   2 +
 scripts/Makefile.kmsan                  |  12 +
 scripts/Makefile.lib                    |   6 +
 sound/core/oss/pcm_oss.c                |   7 +
 110 files changed, 4022 insertions(+), 80 deletions(-)
 create mode 100644 Documentation/dev-tools/kmsan.rst
 create mode 100644 arch/x86/include/asm/kmsan.h
 create mode 100644 include/linux/kmsan-checks.h
 create mode 100644 include/linux/kmsan.h
 create mode 100644 lib/Kconfig.kmsan
 create mode 100644 lib/test_kmsan.c
 create mode 100644 mm/kmsan/Makefile
 create mode 100644 mm/kmsan/kmsan.c
 create mode 100644 mm/kmsan/kmsan.h
 create mode 100644 mm/kmsan/kmsan_entry.c
 create mode 100644 mm/kmsan/kmsan_hooks.c
 create mode 100644 mm/kmsan/kmsan_init.c
 create mode 100644 mm/kmsan/kmsan_instr.c
 create mode 100644 mm/kmsan/kmsan_report.c
 create mode 100644 mm/kmsan/kmsan_shadow.c
 create mode 100644 mm/kmsan/kmsan_shadow.h
 create mode 100644 scripts/Makefile.kmsan

-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 01/26] stackdepot: check depot_index before accessing the stack slab

[glider]

Avoid crashes on corrupted stack ids.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org

---

Change-Id: I0a0b38ed5057090696a2c6ff0be7cfcc24ae6738
---
 lib/stackdepot.c | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

diff --git a/lib/stackdepot.c b/lib/stackdepot.c
index ed717dd08ff3..0d00d2ac0c4b 100644
--- a/lib/stackdepot.c
+++ b/lib/stackdepot.c
@@ -198,9 +198,22 @@ unsigned int stack_depot_fetch(depot_stack_handle_t handle,
 			       unsigned long **entries)
 {
 	union handle_parts parts = { .handle = handle };
-	void *slab = stack_slabs[parts.slabindex];
+	void *slab;
 	size_t offset = parts.offset << STACK_ALLOC_ALIGN;
-	struct stack_record *stack = slab + offset;
+	struct stack_record *stack;
+
+	if (parts.slabindex > depot_index) {
+		WARN(1, "slab index %d out of bounds (%d) for stack id %08x\n",
+			parts.slabindex, depot_index, handle);
+		__memset(trace, 0, sizeof(*trace));
+		return;
+	}
+	slab = stack_slabs[parts.slabindex];
+	stack = slab + offset;
+	if (!stack) {
+		entries = NULL;
+		return 0;
+	}
 
 	*entries = stack->entries;
 	return stack->size;
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 02/26] stackdepot: prevent Clang from optimizing away stackdepot_memcmp()

[glider]

Clang may replace stackdepot_memcmp() with a call to instrumented bcmp(),
which is exactly what we wanted to avoid creating stackdepot_memcmp().
Add a compiler barrier() to prevent optimizations.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org
---

Change-Id: I4495b617b15c0ab003a61c1f0d54d0026fa8b144
---
 lib/stackdepot.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/lib/stackdepot.c b/lib/stackdepot.c
index 0d00d2ac0c4b..785839298e08 100644
--- a/lib/stackdepot.c
+++ b/lib/stackdepot.c
@@ -163,6 +163,11 @@ int stackdepot_memcmp(const unsigned long *u1, const unsigned long *u2,
 			unsigned int n)
 {
 	for ( ; n-- ; u1++, u2++) {
+		/*
+		 * Prevent Clang from replacing this function with a bcmp()
+		 * call.
+		 */
+		barrier();
 		if (*u1 != *u2)
 			return 1;
 	}
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 03/26] kasan: stackdepot: move filter_irq_stacks() to stackdepot.c

[glider]

filter_irq_stacks() can be used by other tools (e.g. KMSAN), so it needs
to be moved to a common location.
lib/stackdepot.c seems a good place, as filter_irq_stacks() is usually
applied to the output of stack_trace_save().

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: linux-mm@kvack.org
---

Change-Id: I65acecf64930a3658e8c2aa7423801082ded8602
---
 include/linux/stackdepot.h |  2 ++
 lib/stackdepot.c           | 23 +++++++++++++++++++++++
 mm/kasan/common.c          | 23 -----------------------
 3 files changed, 25 insertions(+), 23 deletions(-)

diff --git a/include/linux/stackdepot.h b/include/linux/stackdepot.h
index 3efa97d482cb..24d49c732341 100644
--- a/include/linux/stackdepot.h
+++ b/include/linux/stackdepot.h
@@ -19,4 +19,6 @@ depot_stack_handle_t stack_depot_save(unsigned long *entries,
 unsigned int stack_depot_fetch(depot_stack_handle_t handle,
 			       unsigned long **entries);
 
+unsigned int filter_irq_stacks(unsigned long *entries, unsigned int nr_entries);
+
 #endif
diff --git a/lib/stackdepot.c b/lib/stackdepot.c
index 785839298e08..decb1e073b58 100644
--- a/lib/stackdepot.c
+++ b/lib/stackdepot.c
@@ -20,6 +20,7 @@
  */
 
 #include <linux/gfp.h>
+#include <linux/interrupt.h>
 #include <linux/jhash.h>
 #include <linux/kernel.h>
 #include <linux/mm.h>
@@ -319,3 +320,25 @@ depot_stack_handle_t stack_depot_save(unsigned long *entries,
 	return retval;
 }
 EXPORT_SYMBOL_GPL(stack_depot_save);
+
+static inline int in_irqentry_text(unsigned long ptr)
+{
+	return (ptr >= (unsigned long)&__irqentry_text_start &&
+		ptr < (unsigned long)&__irqentry_text_end) ||
+		(ptr >= (unsigned long)&__softirqentry_text_start &&
+		 ptr < (unsigned long)&__softirqentry_text_end);
+}
+
+unsigned int filter_irq_stacks(unsigned long *entries,
+					     unsigned int nr_entries)
+{
+	unsigned int i;
+
+	for (i = 0; i < nr_entries; i++) {
+		if (in_irqentry_text(entries[i])) {
+			/* Include the irqentry function into the stack. */
+			return i + 1;
+		}
+	}
+	return nr_entries;
+}
diff --git a/mm/kasan/common.c b/mm/kasan/common.c
index 6814d6d6a023..154eba5700d8 100644
--- a/mm/kasan/common.c
+++ b/mm/kasan/common.c
@@ -15,7 +15,6 @@
  */
 
 #include <linux/export.h>
-#include <linux/interrupt.h>
 #include <linux/init.h>
 #include <linux/kasan.h>
 #include <linux/kernel.h>
@@ -39,28 +38,6 @@
 #include "kasan.h"
 #include "../slab.h"
 
-static inline int in_irqentry_text(unsigned long ptr)
-{
-	return (ptr >= (unsigned long)&__irqentry_text_start &&
-		ptr < (unsigned long)&__irqentry_text_end) ||
-		(ptr >= (unsigned long)&__softirqentry_text_start &&
-		 ptr < (unsigned long)&__softirqentry_text_end);
-}
-
-static inline unsigned int filter_irq_stacks(unsigned long *entries,
-					     unsigned int nr_entries)
-{
-	unsigned int i;
-
-	for (i = 0; i < nr_entries; i++) {
-		if (in_irqentry_text(entries[i])) {
-			/* Include the irqentry function into the stack. */
-			return i + 1;
-		}
-	}
-	return nr_entries;
-}
-
 static inline depot_stack_handle_t save_stack(gfp_t flags)
 {
 	unsigned long entries[KASAN_STACK_DEPTH];
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 04/26] stackdepot: reserve 5 extra bits in depot_stack_handle_t

[glider]

Some users (currently only KMSAN) may want to use spare bits in
depot_stack_handle_t. Let them do so and provide get_dsh_extra_bits()
and set_dsh_extra_bits() to access those bits.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org
---

Change-Id: I23580dbde85908eeda0bdd8f83a8c3882ab3e012
---
 include/linux/stackdepot.h |  8 ++++++++
 lib/stackdepot.c           | 24 +++++++++++++++++++++++-
 2 files changed, 31 insertions(+), 1 deletion(-)

diff --git a/include/linux/stackdepot.h b/include/linux/stackdepot.h
index 24d49c732341..ac1b5a78d7f6 100644
--- a/include/linux/stackdepot.h
+++ b/include/linux/stackdepot.h
@@ -12,6 +12,11 @@
 #define _LINUX_STACKDEPOT_H
 
 typedef u32 depot_stack_handle_t;
+/*
+ * Number of bits in the handle that stack depot doesn't use. Users may store
+ * information in them.
+ */
+#define STACK_DEPOT_EXTRA_BITS 5
 
 depot_stack_handle_t stack_depot_save(unsigned long *entries,
 				      unsigned int nr_entries, gfp_t gfp_flags);
@@ -20,5 +25,8 @@ unsigned int stack_depot_fetch(depot_stack_handle_t handle,
 			       unsigned long **entries);
 
 unsigned int filter_irq_stacks(unsigned long *entries, unsigned int nr_entries);
+depot_stack_handle_t set_dsh_extra_bits(depot_stack_handle_t handle,
+					unsigned int bits);
+unsigned int get_dsh_extra_bits(depot_stack_handle_t handle);
 
 #endif
diff --git a/lib/stackdepot.c b/lib/stackdepot.c
index decb1e073b58..460efd3ef742 100644
--- a/lib/stackdepot.c
+++ b/lib/stackdepot.c
@@ -40,8 +40,10 @@
 #define STACK_ALLOC_ALIGN 4
 #define STACK_ALLOC_OFFSET_BITS (STACK_ALLOC_ORDER + PAGE_SHIFT - \
 					STACK_ALLOC_ALIGN)
+
 #define STACK_ALLOC_INDEX_BITS (DEPOT_STACK_BITS - \
-		STACK_ALLOC_NULL_PROTECTION_BITS - STACK_ALLOC_OFFSET_BITS)
+		STACK_ALLOC_NULL_PROTECTION_BITS - \
+		STACK_ALLOC_OFFSET_BITS - STACK_DEPOT_EXTRA_BITS)
 #define STACK_ALLOC_SLABS_CAP 8192
 #define STACK_ALLOC_MAX_SLABS \
 	(((1LL << (STACK_ALLOC_INDEX_BITS)) < STACK_ALLOC_SLABS_CAP) ? \
@@ -54,6 +56,7 @@ union handle_parts {
 		u32 slabindex : STACK_ALLOC_INDEX_BITS;
 		u32 offset : STACK_ALLOC_OFFSET_BITS;
 		u32 valid : STACK_ALLOC_NULL_PROTECTION_BITS;
+		u32 extra : STACK_DEPOT_EXTRA_BITS;
 	};
 };
 
@@ -72,6 +75,24 @@ static int next_slab_inited;
 static size_t depot_offset;
 static DEFINE_SPINLOCK(depot_lock);
 
+depot_stack_handle_t set_dsh_extra_bits(depot_stack_handle_t handle,
+					u32 bits)
+{
+	union handle_parts parts = { .handle = handle };
+
+	parts.extra = bits & ((1U << STACK_DEPOT_EXTRA_BITS) - 1);
+	return parts.handle;
+}
+EXPORT_SYMBOL_GPL(set_dsh_extra_bits);
+
+u32 get_dsh_extra_bits(depot_stack_handle_t handle)
+{
+	union handle_parts parts = { .handle = handle };
+
+	return parts.extra;
+}
+EXPORT_SYMBOL_GPL(get_dsh_extra_bits);
+
 static bool init_stack_slab(void **prealloc)
 {
 	if (!*prealloc)
@@ -132,6 +153,7 @@ static struct stack_record *depot_alloc_stack(unsigned long *entries, int size,
 	stack->handle.slabindex = depot_index;
 	stack->handle.offset = depot_offset >> STACK_ALLOC_ALIGN;
 	stack->handle.valid = 1;
+	stack->handle.extra = 0;
 	memcpy(stack->entries, entries, size * sizeof(unsigned long));
 	depot_offset += required_size;
 
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 05/26] printk_safe: externalize printk_context

[glider]

Petr Mladek suggested we use
  this_cpu_read(printk_context) & PRINTK_SAFE_CONTEXT_MASK
instead of checking the spinlock status in kmsan_pr_err()

This appears to be less intrusive, although we'll need to declare
printk_context in some printk header.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org
---

Change-Id: Idb5e7c7cedf6ab6ba635a64423d58d702ad3eca3
---
 kernel/printk/printk_safe.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/printk/printk_safe.c b/kernel/printk/printk_safe.c
index b4045e782743..5ebfc8bc5693 100644
--- a/kernel/printk/printk_safe.c
+++ b/kernel/printk/printk_safe.c
@@ -42,7 +42,7 @@ struct printk_safe_seq_buf {
 };
 
 static DEFINE_PER_CPU(struct printk_safe_seq_buf, safe_print_seq);
-static DEFINE_PER_CPU(int, printk_context);
+DEFINE_PER_CPU(int, printk_context);
 
 #ifdef CONFIG_PRINTK_NMI
 static DEFINE_PER_CPU(struct printk_safe_seq_buf, nmi_print_seq);
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 06/26] kasan: compiler.h: rename __no_kasan_or_inline into __no_memory_tool_or_inline

[glider]

As we use it for KMSAN as well now, let's change the name to be more
generic.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org
---
Note that __no_kasan_or_inline is currently used only on S390, and KMSAN
patches don't introduce uses of __no_memory_tool_or_inline, so this
patch can be landed later if needed.

Change-Id: I842d88b841bf1c75ab659b0016a808c6dcf7cdae
---
 arch/s390/include/asm/processor.h |  4 ++--
 include/linux/compiler.h          | 12 +++++++-----
 2 files changed, 9 insertions(+), 7 deletions(-)

diff --git a/arch/s390/include/asm/processor.h b/arch/s390/include/asm/processor.h
index 51a0e4a2dc96..6bee92331fcd 100644
--- a/arch/s390/include/asm/processor.h
+++ b/arch/s390/include/asm/processor.h
@@ -214,7 +214,7 @@ static inline unsigned long current_stack_pointer(void)
 	return sp;
 }
 
-static __no_kasan_or_inline unsigned short stap(void)
+static __no_memory_tool_or_inline unsigned short stap(void)
 {
 	unsigned short cpu_address;
 
@@ -253,7 +253,7 @@ static inline void __load_psw(psw_t psw)
  * Set PSW mask to specified value, while leaving the
  * PSW addr pointing to the next instruction.
  */
-static __no_kasan_or_inline void __load_psw_mask(unsigned long mask)
+static __no_memory_tool_or_inline void __load_psw_mask(unsigned long mask)
 {
 	unsigned long addr;
 	psw_t psw;
diff --git a/include/linux/compiler.h b/include/linux/compiler.h
index 5e88e7e33abe..4a594bd1bc47 100644
--- a/include/linux/compiler.h
+++ b/include/linux/compiler.h
@@ -199,19 +199,21 @@ void __read_once_size(const volatile void *p, void *res, int size)
 	__READ_ONCE_SIZE;
 }
 
-#ifdef CONFIG_KASAN
+#if defined(CONFIG_KASAN)
 /*
  * We can't declare function 'inline' because __no_sanitize_address confilcts
  * with inlining. Attempt to inline it may cause a build failure.
  * 	https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67368
  * '__maybe_unused' allows us to avoid defined-but-not-used warnings.
  */
-# define __no_kasan_or_inline __no_sanitize_address notrace __maybe_unused
+# define __no_memory_tool_or_inline __no_sanitize_address notrace __maybe_unused
+#elif defined(CONFIG_KMSAN)
+# define __no_memory_tool_or_inline __no_sanitize_memory notrace __maybe_unused
 #else
-# define __no_kasan_or_inline __always_inline
+# define __no_memory_tool_or_inline __always_inline
 #endif
 
-static __no_kasan_or_inline
+static __no_memory_tool_or_inline
 void __read_once_size_nocheck(const volatile void *p, void *res, int size)
 {
 	__READ_ONCE_SIZE;
@@ -274,7 +276,7 @@ static __always_inline void __write_once_size(volatile void *p, void *res, int s
  */
 #define READ_ONCE_NOCHECK(x) __READ_ONCE(x, 0)
 
-static __no_kasan_or_inline
+static __no_memory_tool_or_inline
 unsigned long read_word_at_a_time(const void *addr)
 {
 	kasan_check_read(addr, 1);
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 07/26] kmsan: add ReST documentation

[glider]

Add Documentation/dev-tools/kmsan.rst and reference it in the dev-tools
index.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org

---

Change-Id: Iac6345065e6804ef811f1124fdf779c67ff1530e
---
 Documentation/dev-tools/index.rst |   1 +
 Documentation/dev-tools/kmsan.rst | 418 ++++++++++++++++++++++++++++++
 2 files changed, 419 insertions(+)
 create mode 100644 Documentation/dev-tools/kmsan.rst

diff --git a/Documentation/dev-tools/index.rst b/Documentation/dev-tools/index.rst
index b0522a4dd107..bc5e3fd87efa 100644
--- a/Documentation/dev-tools/index.rst
+++ b/Documentation/dev-tools/index.rst
@@ -19,6 +19,7 @@ whole; patches welcome!
    kcov
    gcov
    kasan
+   kmsan
    ubsan
    kmemleak
    gdb-kernel-debugging
diff --git a/Documentation/dev-tools/kmsan.rst b/Documentation/dev-tools/kmsan.rst
new file mode 100644
index 000000000000..51f9c207cc2c
--- /dev/null
+++ b/Documentation/dev-tools/kmsan.rst
@@ -0,0 +1,418 @@
+=============================
+KernelMemorySanitizer (KMSAN)
+=============================
+
+KMSAN is a dynamic memory error detector aimed at finding uses of uninitialized
+memory.
+It is based on compiler instrumentation, and is quite similar to the userspace
+MemorySanitizer tool (http://clang.llvm.org/docs/MemorySanitizer.html).
+
+KMSAN and Clang
+===============
+
+In order for KMSAN to work the kernel must be
+built with Clang, which is so far the only compiler that has KMSAN support.
+The kernel instrumentation pass is based on the userspace MemorySanitizer tool
+(http://clang.llvm.org/docs/MemorySanitizer.html). Because of the
+instrumentation complexity it's unlikely that any other compiler will support
+KMSAN soon.
+
+Right now the instrumentation pass supports x86_64 only.
+
+How to build
+============
+
+In order to build a kernel with KMSAN you'll need a fresh Clang (10.0.0+, trunk
+version r365008 or greater). Please refer to
+https://llvm.org/docs/GettingStarted.html for the instructions on how to build
+Clang::
+
+  export KMSAN_CLANG_PATH=/path/to/clang
+  # Now configure and build the kernel with CONFIG_KMSAN enabled.
+  make CC=$KMSAN_CLANG_PATH -j64
+
+How KMSAN works
+===============
+
+KMSAN shadow memory
+-------------------
+
+KMSAN associates a so-called shadow byte with every byte of kernel memory.
+A bit in the shadow byte is set iff the corresponding bit of the kernel memory
+byte is uninitialized.
+Marking the memory uninitialized (i.e. setting its shadow bytes to 0xff) is
+called poisoning, marking it initialized (setting the shadow bytes to 0x00) is
+called unpoisoning.
+
+When a new variable is allocated on the stack, it's poisoned by default by
+instrumentation code inserted by the compiler (unless it's a stack variable that
+is immediately initialized). Any new heap allocation done without ``__GFP_ZERO``
+is also poisoned.
+
+Compiler instrumentation also tracks the shadow values with the help from the
+runtime library in ``mm/kmsan/``.
+
+The shadow value of a basic or compound type is an array of bytes of the same
+length.
+When a constant value is written into memory, that memory is unpoisoned.
+When a value is read from memory, its shadow memory is also obtained and
+propagated into all the operations which use that value. For every instruction
+that takes one or more values the compiler generates code that calculates the
+shadow of the result depending on those values and their shadows.
+
+Example::
+
+  int a = 0xff;
+  int b;
+  int c = a | b;
+
+In this case the shadow of ``a`` is ``0``, shadow of ``b`` is ``0xffffffff``,
+shadow of ``c`` is ``0xffffff00``. This means that the upper three bytes of
+``c`` are uninitialized, while the lower byte is initialized.
+
+
+Origin tracking
+---------------
+
+Every four bytes of kernel memory also have a so-called origin assigned to
+them.
+This origin describes the point in program execution at which the uninitialized
+value was created. Every origin is associated with a creation stack, which lets
+the user figure out what's going on.
+
+When an uninitialized variable is allocated on stack or heap, a new origin
+value is created, and that variable's origin is filled with that value.
+When a value is read from memory, its origin is also read and kept together
+with the shadow. For every instruction that takes one or more values the origin
+of the result is one of the origins corresponding to any of the uninitialized
+inputs.
+If a poisoned value is written into memory, its origin is written to the
+corresponding storage as well.
+
+Example 1::
+
+  int a = 0;
+  int b;
+  int c = a + b;
+
+In this case the origin of ``b`` is generated upon function entry, and is
+stored to the origin of ``c`` right before the addition result is written into
+memory.
+
+Several variables may share the same origin address, if they are stored in the
+same four-byte chunk.
+In this case every write to either variable updates the origin for all of them.
+
+Example 2::
+
+  int combine(short a, short b) {
+    union ret_t {
+      int i;
+      short s[2];
+    } ret;
+    ret.s[0] = a;
+    ret.s[1] = b;
+    return ret.i;
+  }
+
+If ``a`` is initialized and ``b`` is not, the shadow of the result would be
+0xffff0000, and the origin of the result would be the origin of ``b``.
+``ret.s[0]`` would have the same origin, but it will be never used, because
+that variable is initialized.
+
+If both function arguments are uninitialized, only the origin of the second
+argument is preserved.
+
+Origin chaining
+~~~~~~~~~~~~~~~
+To ease the debugging, KMSAN creates a new origin for every memory store.
+The new origin references both its creation stack and the previous origin the
+memory location had.
+This may cause increased memory consumption, so we limit the length of origin
+chains in the runtime.
+
+Clang instrumentation API
+-------------------------
+
+Clang instrumentation pass inserts calls to functions defined in
+``mm/kmsan/kmsan_instr.c`` into the kernel code.
+
+Shadow manipulation
+~~~~~~~~~~~~~~~~~~~
+For every memory access the compiler emits a call to a function that returns a
+pair of pointers to the shadow and origin addresses of the given memory::
+
+  typedef struct {
+    void *s, *o;
+  } shadow_origin_ptr_t
+
+  shadow_origin_ptr_t __msan_metadata_ptr_for_load_{1,2,4,8}(void *addr)
+  shadow_origin_ptr_t __msan_metadata_ptr_for_store_{1,2,4,8}(void *addr)
+  shadow_origin_ptr_t __msan_metadata_ptr_for_load_n(void *addr, u64 size)
+  shadow_origin_ptr_t __msan_metadata_ptr_for_store_n(void *addr, u64 size)
+
+The function name depends on the memory access size.
+Each such function also checks if the shadow of the memory in the range
+[``addr``, ``addr + n``) is contiguous and reports an error otherwise.
+
+The compiler makes sure that for every loaded value its shadow and origin
+values are read from memory.
+When a value is stored to memory, its shadow and origin are also stored using
+the metadata pointers.
+
+Origin tracking
+~~~~~~~~~~~~~~~
+A special function is used to create a new origin value for a local variable
+and set the origin of that variable to that value::
+
+  void __msan_poison_alloca(u64 address, u64 size, char *descr)
+
+Access to per-task data
+~~~~~~~~~~~~~~~~~~~~~~~~~
+
+At the beginning of every instrumented function KMSAN inserts a call to
+``__msan_get_context_state()``::
+
+  kmsan_context_state *__msan_get_context_state(void)
+
+``kmsan_context_state`` is declared in ``include/linux/kmsan.h``::
+
+  struct kmsan_context_s {
+    char param_tls[KMSAN_PARAM_SIZE];
+    char retval_tls[RETVAL_SIZE];
+    char va_arg_tls[KMSAN_PARAM_SIZE];
+    char va_arg_origin_tls[KMSAN_PARAM_SIZE];
+    u64 va_arg_overflow_size_tls;
+    depot_stack_handle_t param_origin_tls[PARAM_ARRAY_SIZE];
+    depot_stack_handle_t retval_origin_tls;
+    depot_stack_handle_t origin_tls;
+  };
+
+This structure is used by KMSAN to pass parameter shadows and origins between
+instrumented functions.
+
+String functions
+~~~~~~~~~~~~~~~~
+
+The compiler replaces calls to ``memcpy()``/``memmove()``/``memset()`` with the
+following functions. These functions are also called when data structures are
+initialized or copied, making sure shadow and origin values are copied alongside
+with the data::
+
+  void *__msan_memcpy(void *dst, void *src, u64 n)
+  void *__msan_memmove(void *dst, void *src, u64 n)
+  void *__msan_memset(void *dst, int c, size_t n)
+
+Error reporting
+~~~~~~~~~~~~~~~
+
+For each pointer dereference and each condition the compiler emits a shadow
+check that calls ``__msan_warning()`` in the case a poisoned value is being
+used::
+
+  void __msan_warning(u32 origin)
+
+``__msan_warning()`` causes KMSAN runtime to print an error report.
+
+Inline assembly instrumentation
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+KMSAN instruments every inline assembly output with a call to::
+
+  void __msan_instrument_asm_store(u64 addr, u64 size)
+
+, which unpoisons the memory region.
+
+This approach may mask certain errors, but it also helps to avoid a lot of
+false positives in bitwise operations, atomics etc.
+
+Sometimes the pointers passed into inline assembly don't point to valid memory.
+In such cases they are ignored at runtime.
+
+Disabling the instrumentation
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+A function can be marked with ``__no_sanitize_memory``.
+Doing so doesn't remove KMSAN instrumentation from it, however it makes the
+compiler ignore the uninitialized values coming from the function's inputs,
+and initialize the function's outputs.
+The compiler won't inline functions marked with this attribute into functions
+not marked with it, and vice versa.
+
+It's also possible to disable KMSAN for a single file (e.g. main.o)::
+
+  KMSAN_SANITIZE_main.o := n
+
+or for the whole directory::
+
+  KMSAN_SANITIZE := n
+
+in the Makefile. This comes at a cost however: stack allocations from such files
+and parameters of instrumented functions called from them will have incorrect
+shadow/origin values. As a rule of thumb, avoid using KMSAN_SANITIZE.
+
+Runtime library
+---------------
+The code is located in ``mm/kmsan/``.
+
+Per-task KMSAN state
+~~~~~~~~~~~~~~~~~~~~
+
+Every task_struct has an associated KMSAN task state that holds the KMSAN
+context (see above) and a per-task flag disallowing KMSAN reports::
+
+  struct kmsan_task_state {
+    ...
+    bool allow_reporting;
+    struct kmsan_context_state cstate;
+    ...
+  }
+
+  struct task_struct {
+    ...
+    struct kmsan_task_state kmsan;
+    ...
+  }
+
+
+KMSAN contexts
+~~~~~~~~~~~~~~
+
+When running in a kernel task context, KMSAN uses ``current->kmsan.cstate`` to
+hold the metadata for function parameters and return values.
+
+But in the case the kernel is running in the interrupt, softirq or NMI context,
+where ``current`` is unavailable, KMSAN switches to per-cpu interrupt state::
+
+  DEFINE_PER_CPU(kmsan_context_state[KMSAN_NESTED_CONTEXT_MAX],
+                 kmsan_percpu_cstate);
+
+Metadata allocation
+~~~~~~~~~~~~~~~~~~~
+There are several places in the kernel for which the metadata is stored.
+
+1. Each ``struct page`` instance contains two pointers to its shadow and
+origin pages::
+
+  struct page {
+    ...
+    struct page *shadow, *origin;
+    ...
+  };
+
+Every time a ``struct page`` is allocated, the runtime library allocates two
+additional pages to hold its shadow and origins. This is done by adding hooks
+to ``alloc_pages()``/``free_pages()`` in ``mm/page_alloc.c``.
+To avoid allocating the metadata for non-interesting pages (right now only the
+shadow/origin page themselves and stackdepot storage) the
+``__GFP_NO_KMSAN_SHADOW`` flag is used.
+
+There is a problem related to this allocation algorithm: when two contiguous
+memory blocks are allocated with two different ``alloc_pages()`` calls, their
+shadow pages may not be contiguous. So, if a memory access crosses the boundary
+of a memory block, accesses to shadow/origin memory may potentially corrupt
+other pages or read incorrect values from them.
+
+As a workaround, we check the access size in
+``__msan_metadata_ptr_for_XXX_YYY()`` and return a pointer to a fake shadow
+region in the case of an error::
+
+  char dummy_load_page[PAGE_SIZE] __attribute__((aligned(PAGE_SIZE)));
+  char dummy_store_page[PAGE_SIZE] __attribute__((aligned(PAGE_SIZE)));
+
+``dummy_load_page`` is zero-initialized, so reads from it always yield zeroes.
+All stores to ``dummy_store_page`` are ignored.
+
+Unfortunately at boot time we need to allocate shadow and origin pages for the
+kernel data (``.data``, ``.bss`` etc.) and percpu memory regions, the size of
+which is not a power of 2. As a result, we have to allocate the metadata page by
+page, so that it is also non-contiguous, although it may be perfectly valid to
+access the corresponding kernel memory across page boundaries.
+This can be probably fixed by allocating 1<<N pages at once, splitting them and
+deallocating the rest.
+
+LSB of the ``shadow`` pointer in a ``struct page`` may be set to 1. In this case
+shadow and origin pages are allocated, but KMSAN ignores accesses to them by
+falling back to dummy pages. Allocating the metadata pages is still needed to
+support ``vmap()/vunmap()`` operations on this struct page.
+
+2. For vmalloc memory and modules, there's a direct mapping between the memory
+range, its shadow and origin. KMSAN lessens the vmalloc area by 3/4, making only
+the first quarter available to ``vmalloc()``. The second quarter of the vmalloc
+area contains shadow memory for the first quarter, the third one holds the
+origins. A small part of the fourth quarter contains shadow and origins for the
+kernel modules. Please refer to ``arch/x86/include/asm/pgtable_64_types.h`` for
+more details.
+
+When an array of pages is mapped into a contiguous virtual memory space, their
+shadow and origin pages are similarly mapped into contiguous regions.
+
+3. For CPU entry area there're separate per-CPU arrays that hold its metadata::
+
+  DEFINE_PER_CPU(char[CPU_ENTRY_AREA_SIZE], cpu_entry_area_shadow);
+  DEFINE_PER_CPU(char[CPU_ENTRY_AREA_SIZE], cpu_entry_area_origin);
+
+When calculating shadow and origin addresses for a given memory address, the
+runtime checks whether the address belongs to the physical page range, the
+virtual page range or CPU entry area.
+
+Handling ``pt_regs``
+~~~~~~~~~~~~~~~~~~~
+
+Many functions receive a ``struct pt_regs`` holding the register state at a
+certain point. Registers don't have (easily calculatable) shadow or origin
+associated with them.
+We can assume that the registers are always initialized.
+
+Example report
+--------------
+Here's an example of a real KMSAN report in ``packet_bind_spkt()``::
+
+  ==================================================================
+  BUG: KMSAN: uninit-value in strlen
+  CPU: 0 PID: 1074 Comm: packet Not tainted 4.8.0-rc6+ #1891
+  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
+   0000000000000000 ffff88006b6dfc08 ffffffff82559ae8 ffff88006b6dfb48
+   ffffffff818a7c91 ffffffff85b9c870 0000000000000092 ffffffff85b9c550
+   0000000000000000 0000000000000092 00000000ec400911 0000000000000002
+  Call Trace:
+   [<     inline     >] __dump_stack lib/dump_stack.c:15
+   [<ffffffff82559ae8>] dump_stack+0x238/0x290 lib/dump_stack.c:51
+   [<ffffffff818a6626>] kmsan_report+0x276/0x2e0 mm/kmsan/kmsan.c:1003
+   [<ffffffff818a783b>] __msan_warning+0x5b/0xb0 mm/kmsan/kmsan_instr.c:424
+   [<     inline     >] strlen lib/string.c:484
+   [<ffffffff8259b58d>] strlcpy+0x9d/0x200 lib/string.c:144
+   [<ffffffff84b2eca4>] packet_bind_spkt+0x144/0x230 net/packet/af_packet.c:3132
+   [<ffffffff84242e4d>] SYSC_bind+0x40d/0x5f0 net/socket.c:1370
+   [<ffffffff84242a22>] SyS_bind+0x82/0xa0 net/socket.c:1356
+   [<ffffffff8515991b>] entry_SYSCALL_64_fastpath+0x13/0x8f arch/x86/entry/entry_64.o:?
+  chained origin:
+   [<ffffffff810bb787>] save_stack_trace+0x27/0x50 arch/x86/kernel/stacktrace.c:67
+   [<     inline     >] kmsan_save_stack_with_flags mm/kmsan/kmsan.c:322
+   [<     inline     >] kmsan_save_stack mm/kmsan/kmsan.c:334
+   [<ffffffff818a59f8>] kmsan_internal_chain_origin+0x118/0x1e0 mm/kmsan/kmsan.c:527
+   [<ffffffff818a7773>] __msan_set_alloca_origin4+0xc3/0x130 mm/kmsan/kmsan_instr.c:380
+   [<ffffffff84242b69>] SYSC_bind+0x129/0x5f0 net/socket.c:1356
+   [<ffffffff84242a22>] SyS_bind+0x82/0xa0 net/socket.c:1356
+   [<ffffffff8515991b>] entry_SYSCALL_64_fastpath+0x13/0x8f arch/x86/entry/entry_64.o:?
+  origin description: ----address@SYSC_bind (origin=00000000eb400911)
+  ==================================================================
+
+The report tells that the local variable ``address`` was created uninitialized
+in ``SYSC_bind()`` (the ``bind`` system call implementation). The lower stack
+trace corresponds to the place where this variable was created.
+
+The upper stack shows where the uninit value was used - in ``strlen()``.
+It turned out that the contents of ``address`` were partially copied from the
+userspace, but the buffer wasn't zero-terminated and contained some trailing
+uninitialized bytes.
+``packet_bind_spkt()`` didn't check the length of the buffer, but called
+``strlcpy()`` on it, which called ``strlen()``, which started reading the
+buffer byte by byte till it hit the uninitialized memory.
+
+
+References
+==========
+
+E. Stepanov, K. Serebryany. MemorySanitizer: fast detector of uninitialized
+memory use in C++.
+In Proceedings of CGO 2015.
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 08/26] kmsan: gfp: introduce __GFP_NO_KMSAN_SHADOW

[glider]

This flag is to be used by KMSAN runtime to mark that newly created
memory pages don't need KMSAN metadata backing them.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org

---
We can't decide what to do here:
 - do we need to conditionally define ___GFP_NO_KMSAN_SHADOW depending on
   CONFIG_KMSAN like LOCKDEP does?
 - if KMSAN is defined, and LOCKDEP is not, do we want to "compactify" the GFP
   bits?

Change-Id: If5d0352fd5711ad103328e2c185eb885e826423a
---
 include/linux/gfp.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/include/linux/gfp.h b/include/linux/gfp.h
index fb07b503dc45..b4e7963cd94b 100644
--- a/include/linux/gfp.h
+++ b/include/linux/gfp.h
@@ -44,6 +44,7 @@ struct vm_area_struct;
 #else
 #define ___GFP_NOLOCKDEP	0
 #endif
+#define ___GFP_NO_KMSAN_SHADOW  0x1000000u
 /* If the above are modified, __GFP_BITS_SHIFT may need updating */
 
 /*
@@ -212,12 +213,13 @@ struct vm_area_struct;
 #define __GFP_NOWARN	((__force gfp_t)___GFP_NOWARN)
 #define __GFP_COMP	((__force gfp_t)___GFP_COMP)
 #define __GFP_ZERO	((__force gfp_t)___GFP_ZERO)
+#define __GFP_NO_KMSAN_SHADOW  ((__force gfp_t)___GFP_NO_KMSAN_SHADOW)
 
 /* Disable lockdep for GFP context tracking */
 #define __GFP_NOLOCKDEP ((__force gfp_t)___GFP_NOLOCKDEP)
 
 /* Room for N __GFP_FOO bits */
-#define __GFP_BITS_SHIFT (23 + IS_ENABLED(CONFIG_LOCKDEP))
+#define __GFP_BITS_SHIFT (25)
 #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
 
 /**
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 09/26] kmsan: introduce __no_sanitize_memory and __SANITIZE_MEMORY__

[glider]

__no_sanitize_memory is a function attribute that makes KMSAN
ignore the uninitialized values coming from the function's
inputs, and initialize the function's outputs.

Functions marked with this attribute can't be inlined into functions
not marked with it, and vice versa.

__SANITIZE_MEMORY__ is a macro that's defined iff the file is
instrumented with KMSAN. This is not the same as CONFIG_KMSAN, which is
defined for every file.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org

---

Change-Id: I1f1672652c8392f15f7ca8ac26cd4e71f9cc1e4b
---
 include/linux/compiler-clang.h | 8 ++++++++
 include/linux/compiler-gcc.h   | 5 +++++
 2 files changed, 13 insertions(+)

diff --git a/include/linux/compiler-clang.h b/include/linux/compiler-clang.h
index 333a6695a918..edba13a069a6 100644
--- a/include/linux/compiler-clang.h
+++ b/include/linux/compiler-clang.h
@@ -24,6 +24,14 @@
 #define __no_sanitize_address
 #endif
 
+/* KMSAN is a Clang-only tool, thus putting the defines here */
+#if __has_feature(memory_sanitizer)
+# define __SANITIZE_MEMORY__
+# define __no_sanitize_memory __attribute__((no_sanitize("kernel-memory")))
+#else
+# define __no_sanitize_memory
+#endif
+
 /*
  * Not all versions of clang implement the the type-generic versions
  * of the builtin overflow checkers. Fortunately, clang implements
diff --git a/include/linux/compiler-gcc.h b/include/linux/compiler-gcc.h
index d7ee4c6bad48..e5ebc788dde4 100644
--- a/include/linux/compiler-gcc.h
+++ b/include/linux/compiler-gcc.h
@@ -145,6 +145,11 @@
 #define __no_sanitize_address
 #endif
 
+/*
+ * GCC doesn't support KMSAN.
+ */
+#define __no_sanitize_memory
+
 #if GCC_VERSION >= 50100
 #define COMPILER_HAS_GENERIC_BUILTIN_OVERFLOW 1
 #endif
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 10/26] kmsan: reduce vmalloc space

[glider]

KMSAN is going to use 3/4 of existing vmalloc space to hold the
metadata, therefore we lower VMALLOC_END to make sure vmalloc() doesn't
allocate past the first 1/4.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org

---

Change-Id: Iaa5e8e0fc2aa66c956f937f5a1de6e5ef40d57cc
---
 arch/x86/include/asm/pgtable_64_types.h | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/arch/x86/include/asm/pgtable_64_types.h b/arch/x86/include/asm/pgtable_64_types.h
index 52e5f5f2240d..586629e20436 100644
--- a/arch/x86/include/asm/pgtable_64_types.h
+++ b/arch/x86/include/asm/pgtable_64_types.h
@@ -139,7 +139,22 @@ extern unsigned int ptrs_per_p4d;
 # define VMEMMAP_START		__VMEMMAP_BASE_L4
 #endif /* CONFIG_DYNAMIC_MEMORY_LAYOUT */
 
+#ifndef CONFIG_KMSAN
 #define VMALLOC_END		(VMALLOC_START + (VMALLOC_SIZE_TB << 40) - 1)
+#else
+/*
+ * In KMSAN builds vmalloc area is four times smaller, and the remaining 3/4
+ * are used to keep the metadata for virtual pages.
+ */
+#define VMALLOC_QUARTER_SIZE	((VMALLOC_SIZE_TB << 40) >> 2)
+#define VMALLOC_END		(VMALLOC_START + VMALLOC_QUARTER_SIZE - 1)
+#define VMALLOC_SHADOW_OFFSET	VMALLOC_QUARTER_SIZE
+#define VMALLOC_ORIGIN_OFFSET	(VMALLOC_QUARTER_SIZE * 2)
+#define VMALLOC_META_END	(VMALLOC_END + VMALLOC_ORIGIN_OFFSET)
+#define MODULES_SHADOW_START	(VMALLOC_META_END + 1)
+#define MODULES_ORIGIN_START	(MODULES_SHADOW_START + MODULES_LEN)
+#define MODULES_ORIGIN_END	(MODULES_ORIGIN_START + MODULES_LEN)
+#endif
 
 #define MODULES_VADDR		(__START_KERNEL_map + KERNEL_IMAGE_SIZE)
 /* The module sections ends with the start of the fixmap */
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 11/26] kmsan: add KMSAN runtime

[glider]

This patch adds the KernelMemorySanitizer runtime and associated files:

  - arch/x86/include/asm/kmsan.h: assembly definitions for hooking
    interrupt handlers;
  - include/linux/kmsan-checks.h: user API to enable/disable KMSAN,
    poison/unpoison memory etc.
  - include/linux/kmsan.h: declarations of KMSAN memory hooks to be
    referenced outside KMSAN runtime
  - lib/Kconfig.kmsan: declarations for CONFIG_KMSAN and
    CONFIG_TEST_KMSAN
  - mm/kmsan/Makefile: boilerplate Makefile
  - mm/kmsan/kmsan.h: internal KMSAN declarations
  - mm/kmsan/kmsan.c: core functions that operate with shadow and
    origin memory and perform checks, utility functions
  - mm/kmsan/kmsan_entry.c: KMSAN hooks for entry_64.S
  - mm/kmsan/kmsan_hooks.c: KMSAN hooks for kernel subsystems
  - mm/kmsan/kmsan_init.c: KMSAN initialization routines
  - mm/kmsan/kmsan_instr.c: functions called by KMSAN instrumentation
  - scripts/Makefile.kmsan: CFLAGS_KMSAN

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org

---

Change-Id: I4b3a7aba6d5804afac4f5f7274cadf8675b6e119
---
 arch/x86/Kconfig             |   1 +
 arch/x86/include/asm/kmsan.h | 129 ++++++++
 include/linux/kmsan-checks.h | 123 ++++++++
 include/linux/kmsan.h        | 143 +++++++++
 lib/Kconfig.debug            |   2 +
 lib/Kconfig.kmsan            |  22 ++
 mm/kmsan/Makefile            |   4 +
 mm/kmsan/kmsan.c             | 570 +++++++++++++++++++++++++++++++++++
 mm/kmsan/kmsan.h             | 147 +++++++++
 mm/kmsan/kmsan_entry.c       | 130 ++++++++
 mm/kmsan/kmsan_hooks.c       | 416 +++++++++++++++++++++++++
 mm/kmsan/kmsan_init.c        |  88 ++++++
 mm/kmsan/kmsan_instr.c       | 259 ++++++++++++++++
 mm/kmsan/kmsan_report.c      | 135 +++++++++
 mm/kmsan/kmsan_shadow.c      | 543 +++++++++++++++++++++++++++++++++
 mm/kmsan/kmsan_shadow.h      |  30 ++
 scripts/Makefile.kmsan       |  12 +
 17 files changed, 2754 insertions(+)
 create mode 100644 arch/x86/include/asm/kmsan.h
 create mode 100644 include/linux/kmsan-checks.h
 create mode 100644 include/linux/kmsan.h
 create mode 100644 lib/Kconfig.kmsan
 create mode 100644 mm/kmsan/Makefile
 create mode 100644 mm/kmsan/kmsan.c
 create mode 100644 mm/kmsan/kmsan.h
 create mode 100644 mm/kmsan/kmsan_entry.c
 create mode 100644 mm/kmsan/kmsan_hooks.c
 create mode 100644 mm/kmsan/kmsan_init.c
 create mode 100644 mm/kmsan/kmsan_instr.c
 create mode 100644 mm/kmsan/kmsan_report.c
 create mode 100644 mm/kmsan/kmsan_shadow.c
 create mode 100644 mm/kmsan/kmsan_shadow.h
 create mode 100644 scripts/Makefile.kmsan

diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index d6e1faa28c58..3f83a5c53808 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -135,6 +135,7 @@ config X86
 	select HAVE_ARCH_JUMP_LABEL
 	select HAVE_ARCH_JUMP_LABEL_RELATIVE
 	select HAVE_ARCH_KASAN			if X86_64
+	select HAVE_ARCH_KMSAN			if X86_64
 	select HAVE_ARCH_KGDB
 	select HAVE_ARCH_MMAP_RND_BITS		if MMU
 	select HAVE_ARCH_MMAP_RND_COMPAT_BITS	if MMU && COMPAT
diff --git a/arch/x86/include/asm/kmsan.h b/arch/x86/include/asm/kmsan.h
new file mode 100644
index 000000000000..22322904102b
--- /dev/null
+++ b/arch/x86/include/asm/kmsan.h
@@ -0,0 +1,129 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Assembly bits to safely invoke KMSAN hooks from .S files.
+ *
+ * Adopted from KTSAN assembly hooks implementation by Dmitry Vyukov:
+ * https://github.com/google/ktsan/blob/ktsan/arch/x86/include/asm/ktsan.h
+ *
+ * Copyright (C) 2017-2019 Google LLC
+ * Author: Alexander Potapenko <glider@google.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+#ifndef _ASM_X86_KMSAN_H
+#define _ASM_X86_KMSAN_H
+
+#ifdef CONFIG_KMSAN
+
+#define KMSAN_PUSH_REGS				\
+	pushq	%rax;				\
+	pushq	%rcx;				\
+	pushq	%rdx;				\
+	pushq	%rdi;				\
+	pushq	%rsi;				\
+	pushq	%r8;				\
+	pushq	%r9;				\
+	pushq	%r10;				\
+	pushq	%r11;				\
+/**/
+
+#define KMSAN_POP_REGS				\
+	popq	%r11;				\
+	popq	%r10;				\
+	popq	%r9;				\
+	popq	%r8;				\
+	popq	%rsi;				\
+	popq	%rdi;				\
+	popq	%rdx;				\
+	popq	%rcx;				\
+	popq	%rax;				\
+/**/
+
+#define KMSAN_INTERRUPT_ENTER			\
+	KMSAN_PUSH_REGS				\
+	call	kmsan_interrupt_enter;		\
+	KMSAN_POP_REGS				\
+/**/
+
+#define KMSAN_INTERRUPT_EXIT			\
+	KMSAN_PUSH_REGS				\
+	call	kmsan_interrupt_exit;		\
+	KMSAN_POP_REGS				\
+/**/
+
+#define KMSAN_SOFTIRQ_ENTER			\
+	KMSAN_PUSH_REGS				\
+	call	kmsan_softirq_enter;		\
+	KMSAN_POP_REGS				\
+/**/
+
+#define KMSAN_SOFTIRQ_EXIT			\
+	KMSAN_PUSH_REGS				\
+	call	kmsan_softirq_exit;		\
+	KMSAN_POP_REGS				\
+/**/
+
+#define KMSAN_NMI_ENTER				\
+	KMSAN_PUSH_REGS				\
+	call	kmsan_nmi_enter;		\
+	KMSAN_POP_REGS				\
+/**/
+
+#define KMSAN_NMI_EXIT				\
+	KMSAN_PUSH_REGS				\
+	call	kmsan_nmi_exit;			\
+	KMSAN_POP_REGS				\
+/**/
+
+#define KMSAN_SYSCALL_ENTER			\
+	KMSAN_PUSH_REGS				\
+	call	kmsan_syscall_enter;		\
+	KMSAN_POP_REGS				\
+/**/
+
+#define KMSAN_SYSCALL_EXIT			\
+	KMSAN_PUSH_REGS				\
+	call	kmsan_syscall_exit;		\
+	KMSAN_POP_REGS				\
+/**/
+
+#define KMSAN_IST_ENTER(shift_ist)		\
+	KMSAN_PUSH_REGS				\
+	movq	$shift_ist, %rdi;		\
+	call	kmsan_ist_enter;		\
+	KMSAN_POP_REGS				\
+/**/
+
+#define KMSAN_IST_EXIT(shift_ist)		\
+	KMSAN_PUSH_REGS				\
+	movq	$shift_ist, %rdi;		\
+	call	kmsan_ist_exit;			\
+	KMSAN_POP_REGS				\
+/**/
+
+#define KMSAN_UNPOISON_PT_REGS			\
+	KMSAN_PUSH_REGS				\
+	call	kmsan_unpoison_pt_regs;		\
+	KMSAN_POP_REGS				\
+/**/
+
+
+#else /* ifdef CONFIG_KMSAN */
+
+#define KMSAN_INTERRUPT_ENTER
+#define KMSAN_INTERRUPT_EXIT
+#define KMSAN_SOFTIRQ_ENTER
+#define KMSAN_SOFTIRQ_EXIT
+#define KMSAN_NMI_ENTER
+#define KMSAN_NMI_EXIT
+#define KMSAN_SYSCALL_ENTER
+#define KMSAN_SYSCALL_EXIT
+#define KMSAN_IST_ENTER(shift_ist)
+#define KMSAN_IST_EXIT(shift_ist)
+#define KMSAN_UNPOISON_PT_REGS
+
+#endif /* ifdef CONFIG_KMSAN */
+#endif /* ifndef _ASM_X86_KMSAN_H */
diff --git a/include/linux/kmsan-checks.h b/include/linux/kmsan-checks.h
new file mode 100644
index 000000000000..2b25ced2f13e
--- /dev/null
+++ b/include/linux/kmsan-checks.h
@@ -0,0 +1,123 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * KMSAN checks.
+ * TODO(glider): unite with kmsan.h?
+ *
+ * Copyright (C) 2017-2019 Google LLC
+ * Author: Alexander Potapenko <glider@google.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#ifndef _LINUX_KMSAN_CHECKS_H
+#define _LINUX_KMSAN_CHECKS_H
+
+#include <linux/bug.h>
+#include <linux/types.h>
+
+struct i2c_msg;
+struct page;
+struct sk_buff;
+struct urb;
+
+#ifdef CONFIG_KMSAN
+
+/*
+ * Helper functions that mark the return value initialized.
+ * Note that Clang ignores the inline attribute in the cases when a no_sanitize
+ * function is called from an instrumented one.
+ */
+
+__no_sanitize_memory
+static inline unsigned char KMSAN_INIT_1(unsigned char value)
+{
+	return value;
+}
+
+__no_sanitize_memory
+static inline unsigned short KMSAN_INIT_2(unsigned short value)
+{
+	return value;
+}
+
+__no_sanitize_memory
+static inline unsigned int KMSAN_INIT_4(unsigned int value)
+{
+	return value;
+}
+
+__no_sanitize_memory
+static inline unsigned long KMSAN_INIT_8(unsigned long value)
+{
+	return value;
+}
+
+#define KMSAN_INIT_VALUE(val)		\
+	({				\
+		typeof(val) __ret;	\
+		switch (sizeof(val)) {	\
+		case 1:						\
+			*(unsigned char *)&__ret = KMSAN_INIT_1(	\
+					(unsigned char)val);	\
+			break;					\
+		case 2:						\
+			*(unsigned short *)&__ret = KMSAN_INIT_2(	\
+					(unsigned short)val);	\
+			break;					\
+		case 4:						\
+			*(unsigned int *)&__ret = KMSAN_INIT_4(	\
+					(unsigned int)val);	\
+			break;					\
+		case 8:						\
+			*(unsigned long *)&__ret = KMSAN_INIT_8(	\
+					(unsigned long)val);	\
+			break;					\
+		default:					\
+			BUILD_BUG_ON(1);			\
+		}						\
+		__ret;						\
+	}) /**/
+
+void kmsan_ignore_page(struct page *page, int order);
+void kmsan_poison_shadow(const void *address, size_t size, gfp_t flags);
+void kmsan_unpoison_shadow(const void *address, size_t size);
+void kmsan_check_memory(const void *address, size_t size);
+void kmsan_check_skb(const struct sk_buff *skb);
+void kmsan_handle_urb(const struct urb *urb, bool is_out);
+void kmsan_handle_vprintk(const char **fmt, va_list args);
+void kmsan_handle_i2c_transfer(struct i2c_msg *msgs, int num);
+void kmsan_copy_to_user(const void *to, const void *from, size_t to_copy,
+			size_t left);
+void *__msan_memcpy(void *dst, const void *src, u64 n);
+void kmsan_enter_runtime(unsigned long *flags);
+void kmsan_leave_runtime(unsigned long *flags);
+
+#else
+
+#define KMSAN_INIT_VALUE(value) (value)
+
+static inline void kmsan_ignore_page(struct page *page, int order) {}
+static inline void kmsan_poison_shadow(const void *address, size_t size,
+				       gfp_t flags) {}
+static inline void kmsan_unpoison_shadow(const void *address, size_t size) {}
+static inline void kmsan_check_memory(const void *address, size_t size) {}
+static inline void kmsan_check_skb(const struct sk_buff *skb) {}
+static inline void kmsan_handle_urb(const struct urb *urb, bool is_out) {}
+static inline void kmsan_handle_vprintk(const char **fmt, va_list args) {}
+static inline void kmsan_handle_i2c_transfer(struct i2c_msg *msgs, int num) {}
+static inline void kmsan_copy_to_user(
+	const void *to, const void *from, size_t to_copy, size_t left) {}
+static inline void *__msan_memcpy(void *dst, const void *src, size_t n)
+{
+	return NULL;
+}
+
+static inline void kmsan_enter_runtime(unsigned long *flags) {}
+static inline void kmsan_leave_runtime(unsigned long *flags) {}
+
+#endif
+
+#endif /* _LINUX_KMSAN_CHECKS_H */
diff --git a/include/linux/kmsan.h b/include/linux/kmsan.h
new file mode 100644
index 000000000000..f5638bac368e
--- /dev/null
+++ b/include/linux/kmsan.h
@@ -0,0 +1,143 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * KMSAN API for subsystems.
+ *
+ * Copyright (C) 2017-2019 Google LLC
+ * Author: Alexander Potapenko <glider@google.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+#ifndef LINUX_KMSAN_H
+#define LINUX_KMSAN_H
+
+#include <linux/gfp.h>
+#include <linux/stackdepot.h>
+#include <linux/types.h>
+#include <linux/vmalloc.h>
+
+struct page;
+struct kmem_cache;
+struct task_struct;
+struct vm_struct;
+
+
+extern bool kmsan_ready;
+
+#ifdef CONFIG_KMSAN
+void __init kmsan_initialize_shadow(void);
+void __init kmsan_initialize(void);
+
+/* These constants are defined in the MSan LLVM instrumentation pass. */
+#define RETVAL_SIZE 800
+#define KMSAN_PARAM_SIZE 800
+
+#define PARAM_ARRAY_SIZE (KMSAN_PARAM_SIZE / sizeof(depot_stack_handle_t))
+
+struct kmsan_context_state {
+	char param_tls[KMSAN_PARAM_SIZE];
+	char retval_tls[RETVAL_SIZE];
+	char va_arg_tls[KMSAN_PARAM_SIZE];
+	char va_arg_origin_tls[KMSAN_PARAM_SIZE];
+	u64 va_arg_overflow_size_tls;
+	depot_stack_handle_t param_origin_tls[PARAM_ARRAY_SIZE];
+	depot_stack_handle_t retval_origin_tls;
+	depot_stack_handle_t origin_tls;
+};
+
+struct kmsan_task_state {
+	bool allow_reporting;
+	struct kmsan_context_state cstate;
+};
+
+void kmsan_task_create(struct task_struct *task);
+void kmsan_task_exit(struct task_struct *task);
+void kmsan_alloc_shadow_for_region(void *start, size_t size);
+int kmsan_alloc_page(struct page *page, unsigned int order, gfp_t flags);
+void kmsan_gup_pgd_range(struct page **pages, int nr);
+void kmsan_free_page(struct page *page, unsigned int order);
+void kmsan_split_page(struct page *page, unsigned int order);
+void kmsan_copy_page_meta(struct page *dst, struct page *src);
+
+void kmsan_poison_slab(struct page *page, gfp_t flags);
+void kmsan_kmalloc_large(const void *ptr, size_t size, gfp_t flags);
+void kmsan_kfree_large(const void *ptr);
+void kmsan_kmalloc(struct kmem_cache *s, const void *object, size_t size,
+		   gfp_t flags);
+void kmsan_slab_alloc(struct kmem_cache *s, void *object, gfp_t flags);
+void kmsan_slab_free(struct kmem_cache *s, void *object);
+
+void kmsan_slab_setup_object(struct kmem_cache *s, void *object);
+void kmsan_post_alloc_hook(struct kmem_cache *s, gfp_t flags,
+			size_t size, void *object);
+
+/* vmap */
+void kmsan_vmap_page_range_noflush(unsigned long start, unsigned long end,
+				   pgprot_t prot, struct page **pages);
+void kmsan_vunmap_page_range(unsigned long addr, unsigned long end);
+
+/* ioremap */
+void kmsan_ioremap_page_range(unsigned long addr, unsigned long end,
+			      phys_addr_t phys_addr, pgprot_t prot);
+void kmsan_iounmap_page_range(unsigned long start, unsigned long end);
+
+void kmsan_softirq_enter(void);
+void kmsan_softirq_exit(void);
+
+void kmsan_clear_page(void *page_addr);
+
+#else
+
+static inline void __init kmsan_initialize_shadow(void) { }
+static inline void __init kmsan_initialize(void) { }
+
+static inline void kmsan_task_create(struct task_struct *task) {}
+static inline void kmsan_task_exit(struct task_struct *task) {}
+static inline void kmsan_alloc_shadow_for_region(void *start, size_t size) {}
+static inline int kmsan_alloc_page(struct page *page, unsigned int order,
+				   gfp_t flags)
+{
+	return 0;
+}
+static inline void kmsan_gup_pgd_range(struct page **pages, int nr) {}
+static inline void kmsan_free_page(struct page *page, unsigned int order) {}
+static inline void kmsan_split_page(struct page *page, unsigned int order) {}
+static inline void kmsan_copy_page_meta(struct page *dst, struct page *src) {}
+
+static inline void kmsan_poison_slab(struct page *page, gfp_t flags) {}
+static inline void kmsan_kmalloc_large(const void *ptr, size_t size,
+				       gfp_t flags) {}
+static inline void kmsan_kfree_large(const void *ptr) {}
+static inline void kmsan_kmalloc(struct kmem_cache *s, const void *object,
+				 size_t size, gfp_t flags) {}
+static inline void kmsan_slab_alloc(struct kmem_cache *s, void *object,
+				    gfp_t flags) {}
+static inline void kmsan_slab_free(struct kmem_cache *s, void *object) {}
+
+static inline void kmsan_slab_setup_object(struct kmem_cache *s,
+					   void *object) {}
+static inline void kmsan_post_alloc_hook(struct kmem_cache *s, gfp_t flags,
+					 size_t size, void *object) {}
+
+static inline void kmsan_vmap_page_range_noflush(unsigned long start,
+						 unsigned long end,
+						 pgprot_t prot,
+						 struct page **pages) {}
+static inline void kmsan_vunmap_page_range(unsigned long start,
+					   unsigned long end) {}
+
+static inline void kmsan_ioremap_page_range(unsigned long start,
+					    unsigned long end,
+					    phys_addr_t phys_addr,
+					    pgprot_t prot) {}
+static inline void kmsan_iounmap_page_range(unsigned long start,
+					    unsigned long end) {}
+static inline void kmsan_softirq_enter(void) {}
+static inline void kmsan_softirq_exit(void) {}
+
+static inline void kmsan_clear_page(void *page_addr) {}
+#endif
+
+#endif /* LINUX_KMSAN_H */
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 93d97f9b0157..75c36318943d 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -756,6 +756,8 @@ config DEBUG_STACKOVERFLOW
 
 source "lib/Kconfig.kasan"
 
+source "lib/Kconfig.kmsan"
+
 endmenu # "Memory Debugging"
 
 config ARCH_HAS_KCOV
diff --git a/lib/Kconfig.kmsan b/lib/Kconfig.kmsan
new file mode 100644
index 000000000000..187dddfcf220
--- /dev/null
+++ b/lib/Kconfig.kmsan
@@ -0,0 +1,22 @@
+config HAVE_ARCH_KMSAN
+	bool
+
+if HAVE_ARCH_KMSAN
+
+config KMSAN
+	bool "KMSAN: detector of uninitialized memory use"
+	depends on SLUB && !KASAN
+	select STACKDEPOT
+	help
+	  KMSAN is a dynamic detector of uses of uninitialized memory in the
+	  kernel. It is based on compiler instrumentation provided by Clang
+	  and thus requires Clang 10.0.0+ to build.
+
+config TEST_KMSAN
+	tristate "Module for testing KMSAN for bug detection"
+	depends on m && KMSAN
+	help
+	  Test module that can trigger various uses of uninitialized memory
+	  detectable by KMSAN.
+
+endif
diff --git a/mm/kmsan/Makefile b/mm/kmsan/Makefile
new file mode 100644
index 000000000000..ccf6d2d00a7a
--- /dev/null
+++ b/mm/kmsan/Makefile
@@ -0,0 +1,4 @@
+obj-y := kmsan.o kmsan_instr.o kmsan_init.o kmsan_entry.o kmsan_hooks.o kmsan_report.o kmsan_shadow.o
+
+KMSAN_SANITIZE := n
+KCOV_INSTRUMENT := n
diff --git a/mm/kmsan/kmsan.c b/mm/kmsan/kmsan.c
new file mode 100644
index 000000000000..8009149ee20a
--- /dev/null
+++ b/mm/kmsan/kmsan.c
@@ -0,0 +1,570 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * KMSAN runtime library.
+ *
+ * Copyright (C) 2017-2019 Google LLC
+ * Author: Alexander Potapenko <glider@google.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <asm/page.h>
+#include <linux/compiler.h>
+#include <linux/export.h>
+#include <linux/highmem.h>
+#include <linux/interrupt.h>
+#include <linux/kernel.h>
+#include <linux/kmsan.h>
+#include <linux/memory.h>
+#include <linux/mm.h>
+#include <linux/preempt.h>
+#include <linux/percpu-defs.h>
+#include <linux/mm_types.h>
+#include <linux/slab.h>
+#include <linux/stackdepot.h>
+#include <linux/stacktrace.h>
+#include <linux/types.h>
+#include <linux/vmalloc.h>
+
+#include <linux/mmzone.h>
+
+#include "../slab.h"
+#include "kmsan.h"
+
+/*
+ * Some kernel asm() calls mention the non-existing |__force_order| variable
+ * in the asm constraints to preserve the order of accesses to control
+ * registers. KMSAN turns those mentions into actual memory accesses, therefore
+ * the variable is now required to link the kernel.
+ */
+unsigned long __force_order;
+
+bool kmsan_ready;
+#define KMSAN_STACK_DEPTH 64
+#define MAX_CHAIN_DEPTH 7
+
+/*
+ * According to Documentation/x86/kernel-stacks, kernel code can run on the
+ * following stacks:
+ * - regular task stack - when executing the task code
+ *  - interrupt stack - when handling external hardware interrupts and softirqs
+ *  - NMI stack
+ * 0 is for regular interrupts, 1 for softirqs, 2 for NMI.
+ * Because interrupts may nest, trying to use a new context for every new
+ * interrupt.
+ */
+/* [0] for dummy per-CPU context. */
+DEFINE_PER_CPU(struct kmsan_context_state[KMSAN_NESTED_CONTEXT_MAX],
+	       kmsan_percpu_cstate);
+/* 0 for task context, |i>0| for kmsan_context_state[i]. */
+DEFINE_PER_CPU(int, kmsan_context_level);
+DEFINE_PER_CPU(int, kmsan_in_interrupt);
+DEFINE_PER_CPU(bool, kmsan_in_softirq);
+DEFINE_PER_CPU(bool, kmsan_in_nmi);
+DEFINE_PER_CPU(int, kmsan_in_runtime);
+/* TODO(glider): debug-only. */
+DEFINE_PER_CPU(unsigned long, kmsan_runtime_last_caller);
+
+struct kmsan_context_state *task_kmsan_context_state(void)
+{
+	int cpu = smp_processor_id();
+	int level = this_cpu_read(kmsan_context_level);
+	struct kmsan_context_state *ret;
+
+	if (!kmsan_ready || IN_RUNTIME()) {
+		ret = &per_cpu(kmsan_percpu_cstate[0], cpu);
+		__memset(ret, 0, sizeof(struct kmsan_context_state));
+		return ret;
+	}
+
+	if (!level)
+		ret = &current->kmsan.cstate;
+	else
+		ret = &per_cpu(kmsan_percpu_cstate[level], cpu);
+	return ret;
+}
+
+void kmsan_internal_task_create(struct task_struct *task)
+{
+	struct kmsan_task_state *state = &task->kmsan;
+
+	__memset(state, 0, sizeof(struct kmsan_task_state));
+	state->allow_reporting = true;
+}
+
+void kmsan_internal_memset_shadow(void *addr, int b, size_t size,
+				  bool checked)
+{
+	void *shadow_start;
+	u64 page_offset, address = (u64)addr;
+	size_t to_fill;
+
+	BUG_ON(!metadata_is_contiguous(addr, size, META_SHADOW));
+	while (size) {
+		page_offset = address % PAGE_SIZE;
+		to_fill = min(PAGE_SIZE - page_offset, (u64)size);
+		shadow_start = kmsan_get_metadata((void *)address, to_fill,
+						  META_SHADOW);
+		if (!shadow_start) {
+			if (checked) {
+				kmsan_pr_err("WARNING: not memsetting %d bytes starting at %px, because the shadow is NULL\n", to_fill, address);
+				BUG();
+			}
+			/* Otherwise just move on. */
+		} else {
+			__memset(shadow_start, b, to_fill);
+		}
+		address += to_fill;
+		size -= to_fill;
+	}
+}
+
+void kmsan_internal_poison_shadow(void *address, size_t size,
+				gfp_t flags, unsigned int poison_flags)
+{
+	bool checked = poison_flags & KMSAN_POISON_CHECK;
+	depot_stack_handle_t handle;
+	u32 extra_bits = 0;
+
+	if (poison_flags & KMSAN_POISON_FREE)
+		extra_bits = 1;
+	kmsan_internal_memset_shadow(address, -1, size, checked);
+	handle = kmsan_save_stack_with_flags(flags, extra_bits);
+	kmsan_set_origin_checked(address, size, handle, checked);
+}
+
+void kmsan_internal_unpoison_shadow(void *address, size_t size, bool checked)
+{
+	kmsan_internal_memset_shadow(address, 0, size, checked);
+	kmsan_set_origin_checked(address, size, 0, checked);
+}
+
+depot_stack_handle_t kmsan_save_stack_with_flags(gfp_t flags,
+						 unsigned int reserved)
+{
+	depot_stack_handle_t handle;
+	unsigned long entries[KMSAN_STACK_DEPTH];
+	unsigned int nr_entries;
+
+	nr_entries = stack_trace_save(entries, KMSAN_STACK_DEPTH, 0);
+	filter_irq_stacks(entries, nr_entries);
+
+	/* Don't sleep (see might_sleep_if() in __alloc_pages_nodemask()). */
+	flags &= ~__GFP_DIRECT_RECLAIM;
+
+	handle = stack_depot_save(entries, nr_entries, flags);
+	return set_dsh_extra_bits(handle, reserved);
+}
+
+/*
+ * Depending on the value of is_memmove, this serves as both a memcpy and a
+ * memmove implementation.
+ *
+ * As with the regular memmove, do the following:
+ * - if src and dst don't overlap, use memcpy();
+ * - if src and dst overlap:
+ *   - if src > dst, use memcpy();
+ *   - if src < dst, use reverse-memcpy.
+ * Why this is correct:
+ * - problems may arise if for some part of the overlapping region we
+ *   overwrite its shadow with a new value before copying it somewhere.
+ *   But there's a 1:1 mapping between the kernel memory and its shadow,
+ *   therefore if this doesn't happen with the kernel memory it can't happen
+ *   with the shadow.
+ */
+void kmsan_memcpy_memmove_metadata(void *dst, void *src, size_t n,
+				   bool is_memmove)
+{
+	void *shadow_src, *shadow_dst;
+	depot_stack_handle_t *origin_src, *origin_dst;
+	int src_slots, dst_slots, i, iter, step, skip_bits;
+	depot_stack_handle_t old_origin = 0, chain_origin, new_origin = 0;
+	u32 *align_shadow_src, shadow;
+	bool backwards;
+
+	BUG_ON(!metadata_is_contiguous(dst, n, META_SHADOW));
+	BUG_ON(!metadata_is_contiguous(src, n, META_SHADOW));
+
+	shadow_dst = kmsan_get_metadata(dst, n, META_SHADOW);
+	if (!shadow_dst)
+		return;
+
+	shadow_src = kmsan_get_metadata(src, n, META_SHADOW);
+	if (!shadow_src) {
+		/*
+		 * |src| is untracked: zero out destination shadow, ignore the
+		 * origins, we're done.
+		 */
+		__memset(shadow_dst, 0, n);
+		return;
+	}
+	if (is_memmove)
+		__memmove(shadow_dst, shadow_src, n);
+	else
+		__memcpy(shadow_dst, shadow_src, n);
+
+	origin_dst = kmsan_get_metadata(dst, n, META_ORIGIN);
+	origin_src = kmsan_get_metadata(src, n, META_ORIGIN);
+	BUG_ON(!origin_dst || !origin_src);
+	BUG_ON(!metadata_is_contiguous(dst, n, META_ORIGIN));
+	BUG_ON(!metadata_is_contiguous(src, n, META_ORIGIN));
+	src_slots = (ALIGN((u64)src + n, ORIGIN_SIZE) -
+		     ALIGN_DOWN((u64)src, ORIGIN_SIZE)) / ORIGIN_SIZE;
+	dst_slots = (ALIGN((u64)dst + n, ORIGIN_SIZE) -
+		     ALIGN_DOWN((u64)dst, ORIGIN_SIZE)) / ORIGIN_SIZE;
+	BUG_ON(!src_slots || !dst_slots);
+	BUG_ON((src_slots < 1) || (dst_slots < 1));
+	BUG_ON((src_slots - dst_slots > 1) || (dst_slots - src_slots < -1));
+
+	backwards = is_memmove && (dst > src);
+	i = backwards ? min(src_slots, dst_slots) - 1 : 0;
+	iter = backwards ? -1 : 1;
+
+	align_shadow_src = (u32 *)ALIGN_DOWN((u64)shadow_src, ORIGIN_SIZE);
+	for (step = 0; step < min(src_slots, dst_slots); step++, i += iter) {
+		BUG_ON(i < 0);
+		shadow = align_shadow_src[i];
+		if (i == 0) {
+			/*
+			 * If |src| isn't aligned on ORIGIN_SIZE, don't
+			 * look at the first |src % ORIGIN_SIZE| bytes
+			 * of the first shadow slot.
+			 */
+			skip_bits = ((u64)src % ORIGIN_SIZE) * 8;
+			shadow = (shadow << skip_bits) >> skip_bits;
+		}
+		if (i == src_slots - 1) {
+			/*
+			 * If |src + n| isn't aligned on
+			 * ORIGIN_SIZE, don't look at the last
+			 * |(src + n) % ORIGIN_SIZE| bytes of the
+			 * last shadow slot.
+			 */
+			skip_bits = (((u64)src + n) % ORIGIN_SIZE) * 8;
+			shadow = (shadow >> skip_bits) << skip_bits;
+		}
+		/*
+		 * Overwrite the origin only if the corresponding
+		 * shadow is nonempty.
+		 */
+		if (origin_src[i] && (origin_src[i] != old_origin) && shadow) {
+			old_origin = origin_src[i];
+			chain_origin = kmsan_internal_chain_origin(old_origin);
+			/*
+			 * kmsan_internal_chain_origin() may return
+			 * NULL, but we don't want to lose the previous
+			 * origin value.
+			 */
+			if (chain_origin)
+				new_origin = chain_origin;
+			else
+				new_origin = old_origin;
+		}
+		if (shadow)
+			origin_dst[i] = new_origin;
+		else
+			origin_dst[i] = 0;
+	}
+}
+
+void kmsan_memcpy_metadata(void *dst, void *src, size_t n)
+{
+	kmsan_memcpy_memmove_metadata(dst, src, n, /*is_memmove*/false);
+}
+
+void kmsan_memmove_metadata(void *dst, void *src, size_t n)
+{
+	kmsan_memcpy_memmove_metadata(dst, src, n, /*is_memmove*/true);
+}
+
+depot_stack_handle_t kmsan_internal_chain_origin(depot_stack_handle_t id)
+{
+	depot_stack_handle_t handle;
+	unsigned long entries[3];
+	u64 magic = KMSAN_CHAIN_MAGIC_ORIGIN_FULL;
+	int depth = 0;
+	static int skipped;
+	u32 extra_bits;
+
+	if (!kmsan_ready)
+		return 0;
+
+	if (!id)
+		return id;
+	/*
+	 * Make sure we have enough spare bits in |id| to hold the UAF bit and
+	 * the chain depth.
+	 */
+	BUILD_BUG_ON((1 << STACK_DEPOT_EXTRA_BITS) <= (MAX_CHAIN_DEPTH << 1));
+
+	extra_bits = get_dsh_extra_bits(id);
+
+	depth = extra_bits >> 1;
+	if (depth >= MAX_CHAIN_DEPTH) {
+		skipped++;
+		if (skipped % 10000 == 0) {
+			kmsan_pr_err("not chained %d origins\n", skipped);
+			dump_stack();
+			kmsan_print_origin(id);
+		}
+		return id;
+	}
+	depth++;
+	/* Lowest bit is the UAF flag, higher bits hold the depth. */
+	extra_bits = (depth << 1) | (extra_bits & 1);
+	/* TODO(glider): how do we figure out we've dropped some frames? */
+	entries[0] = magic + depth;
+	entries[1] = kmsan_save_stack_with_flags(GFP_ATOMIC, extra_bits);
+	entries[2] = id;
+	handle = stack_depot_save(entries, ARRAY_SIZE(entries), GFP_ATOMIC);
+	return set_dsh_extra_bits(handle, extra_bits);
+}
+
+void kmsan_write_aligned_origin(void *var, size_t size, u32 origin)
+{
+	u32 *var_cast = (u32 *)var;
+	int i;
+
+	BUG_ON((u64)var_cast % ORIGIN_SIZE);
+	BUG_ON(size % ORIGIN_SIZE);
+	for (i = 0; i < size / ORIGIN_SIZE; i++)
+		var_cast[i] = origin;
+}
+
+/*
+ * TODO(glider): writing an initialized byte shouldn't zero out the origin, if
+ * the remaining three bytes are uninitialized.
+ */
+void kmsan_internal_set_origin(void *addr, int size, u32 origin)
+{
+	void *origin_start;
+	u64 address = (u64)addr, page_offset;
+	size_t to_fill, pad = 0;
+
+	if (!IS_ALIGNED(address, ORIGIN_SIZE)) {
+		pad = address % ORIGIN_SIZE;
+		address -= pad;
+		size += pad;
+	}
+
+	while (size > 0) {
+		page_offset = address % PAGE_SIZE;
+		to_fill = min(PAGE_SIZE - page_offset, (u64)size);
+		/* write at least ORIGIN_SIZE bytes */
+		to_fill = ALIGN(to_fill, ORIGIN_SIZE);
+		BUG_ON(!to_fill);
+		origin_start = kmsan_get_metadata((void *)address, to_fill,
+						  META_ORIGIN);
+		address += to_fill;
+		size -= to_fill;
+		if (!origin_start)
+			/* Can happen e.g. if the memory is untracked. */
+			continue;
+		kmsan_write_aligned_origin(origin_start, to_fill, origin);
+	}
+}
+
+void kmsan_set_origin_checked(void *addr, int size, u32 origin, bool checked)
+{
+	if (checked && !metadata_is_contiguous(addr, size, META_ORIGIN)) {
+		kmsan_pr_err("WARNING: not setting origin for %d bytes starting at %px, because the metadata is incontiguous\n", size, addr);
+		BUG();
+	}
+	kmsan_internal_set_origin(addr, size, origin);
+}
+
+struct page *vmalloc_to_page_or_null(void *vaddr)
+{
+	struct page *page;
+
+	if (!kmsan_internal_is_vmalloc_addr(vaddr) &&
+	    !kmsan_internal_is_module_addr(vaddr))
+		return NULL;
+	page = vmalloc_to_page(vaddr);
+	if (pfn_valid(page_to_pfn(page)))
+		return page;
+	else
+		return NULL;
+}
+
+void kmsan_internal_check_memory(void *addr, size_t size, const void *user_addr,
+				 int reason)
+{
+	unsigned long irq_flags;
+	unsigned long addr64 = (unsigned long)addr;
+	unsigned char *shadow = NULL;
+	depot_stack_handle_t *origin = NULL;
+	depot_stack_handle_t cur_origin = 0, new_origin = 0;
+	int cur_off_start = -1;
+	int i, chunk_size;
+	size_t pos = 0;
+
+	BUG_ON(!metadata_is_contiguous(addr, size, META_SHADOW));
+	if (size <= 0)
+		return;
+	while (pos < size) {
+		chunk_size = min(size - pos,
+				 PAGE_SIZE - ((addr64 + pos) % PAGE_SIZE));
+		shadow = kmsan_get_metadata((void *)(addr64 + pos), chunk_size,
+					    META_SHADOW);
+		if (!shadow) {
+			/*
+			 * This page is untracked. If there were uninitialized
+			 * bytes before, report them.
+			 */
+			if (cur_origin) {
+				ENTER_RUNTIME(irq_flags);
+				kmsan_report(cur_origin, addr, size,
+					     cur_off_start, pos - 1, user_addr,
+					     reason);
+				LEAVE_RUNTIME(irq_flags);
+			}
+			cur_origin = 0;
+			cur_off_start = -1;
+			pos += chunk_size;
+			continue;
+		}
+		for (i = 0; i < chunk_size; i++) {
+			if (!shadow[i]) {
+				/*
+				 * This byte is unpoisoned. If there were
+				 * poisoned bytes before, report them.
+				 */
+				if (cur_origin) {
+					ENTER_RUNTIME(irq_flags);
+					kmsan_report(cur_origin, addr, size,
+						     cur_off_start, pos + i - 1,
+						     user_addr, reason);
+					LEAVE_RUNTIME(irq_flags);
+				}
+				cur_origin = 0;
+				cur_off_start = -1;
+				continue;
+			}
+			origin = kmsan_get_metadata((void *)(addr64 + pos + i),
+						chunk_size - i, META_ORIGIN);
+			BUG_ON(!origin);
+			new_origin = *origin;
+			/*
+			 * Encountered new origin - report the previous
+			 * uninitialized range.
+			 */
+			if (cur_origin != new_origin) {
+				if (cur_origin) {
+					ENTER_RUNTIME(irq_flags);
+					kmsan_report(cur_origin, addr, size,
+						     cur_off_start, pos + i - 1,
+						     user_addr, reason);
+					LEAVE_RUNTIME(irq_flags);
+				}
+				cur_origin = new_origin;
+				cur_off_start = pos + i;
+			}
+		}
+		pos += chunk_size;
+	}
+	BUG_ON(pos != size);
+	if (cur_origin) {
+		ENTER_RUNTIME(irq_flags);
+		kmsan_report(cur_origin, addr, size, cur_off_start, pos - 1,
+			     user_addr, reason);
+		LEAVE_RUNTIME(irq_flags);
+	}
+}
+
+/*
+ * TODO(glider): this check shouldn't be performed for origin pages, because
+ * they're always accessed after the shadow pages.
+ */
+bool metadata_is_contiguous(void *addr, size_t size, bool is_origin)
+{
+	u64 cur_addr = (u64)addr, next_addr;
+	char *cur_meta = NULL, *next_meta = NULL;
+	depot_stack_handle_t *origin_p;
+	bool all_untracked = false;
+	const char *fname = is_origin ? "origin" : "shadow";
+
+	if (!size)
+		return true;
+
+	/* The whole range belongs to the same page. */
+	if (ALIGN_DOWN(cur_addr + size - 1, PAGE_SIZE) ==
+	    ALIGN_DOWN(cur_addr, PAGE_SIZE))
+		return true;
+	cur_meta = kmsan_get_metadata((void *)cur_addr, 1, is_origin);
+	if (!cur_meta)
+		all_untracked = true;
+	for (next_addr = cur_addr + PAGE_SIZE; next_addr < (u64)addr + size;
+		     cur_addr = next_addr,
+		     cur_meta = next_meta,
+		     next_addr += PAGE_SIZE) {
+		next_meta = kmsan_get_metadata((void *)next_addr, 1, is_origin);
+		if (!next_meta) {
+			if (!all_untracked)
+				goto report;
+			continue;
+		}
+		if ((u64)cur_meta == ((u64)next_meta - PAGE_SIZE))
+			continue;
+		goto report;
+	}
+	return true;
+
+report:
+	kmsan_pr_err("BUG: attempting to access two shadow page ranges.\n");
+	dump_stack();
+	kmsan_pr_err("\n");
+	kmsan_pr_err("Access of size %d at %px.\n", size, addr);
+	kmsan_pr_err("Addresses belonging to different ranges: %px and %px\n",
+		     cur_addr, next_addr);
+	kmsan_pr_err("page[0].%s: %px, page[1].%s: %px\n",
+		     fname, cur_meta, fname, next_meta);
+	origin_p = kmsan_get_metadata(addr, 1, META_ORIGIN);
+	if (origin_p) {
+		kmsan_pr_err("Origin: %08x\n", *origin_p);
+		kmsan_print_origin(*origin_p);
+	} else {
+		kmsan_pr_err("Origin: unavailable\n");
+	}
+	return false;
+}
+
+/*
+ * Dummy replacement for __builtin_return_address() which may crash without
+ * frame pointers.
+ */
+void *kmsan_internal_return_address(int arg)
+{
+#ifdef CONFIG_UNWINDER_FRAME_POINTER
+	switch (arg) {
+	case 1:
+		return __builtin_return_address(1);
+	case 2:
+		return __builtin_return_address(2);
+	default:
+		BUG();
+	}
+#else
+	unsigned long entries[1];
+	struct stack_trace trace = {
+		.nr_entries = 0,
+		.entries = entries,
+		.max_entries = 1,
+		.skip = arg
+	};
+	save_stack_trace(&trace);
+	return entries[0];
+#endif
+}
+
+bool kmsan_internal_is_module_addr(void *vaddr)
+{
+	return ((u64)vaddr >= MODULES_VADDR) && ((u64)vaddr < MODULES_END);
+}
+
+bool kmsan_internal_is_vmalloc_addr(void *addr)
+{
+	return ((u64)addr >= VMALLOC_START) && ((u64)addr < VMALLOC_END);
+}
diff --git a/mm/kmsan/kmsan.h b/mm/kmsan/kmsan.h
new file mode 100644
index 000000000000..015e402e9f9d
--- /dev/null
+++ b/mm/kmsan/kmsan.h
@@ -0,0 +1,147 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * KMSAN internal declarations.
+ *
+ * Copyright (C) 2017-2019 Google LLC
+ * Author: Alexander Potapenko <glider@google.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#ifndef __MM_KMSAN_KMSAN_H
+#define __MM_KMSAN_KMSAN_H
+
+#include <asm/pgtable_64_types.h>
+#include <linux/irqflags.h>
+#include <linux/sched.h>
+#include <linux/stackdepot.h>
+#include <linux/stacktrace.h>
+#include <linux/nmi.h>
+#include <linux/mm.h>
+#include <linux/printk.h>
+
+#include "kmsan_shadow.h"
+
+#define KMSAN_MAGIC_MASK 0xffffffffff00
+#define KMSAN_ALLOCA_MAGIC_ORIGIN 0x4110c4071900
+#define KMSAN_CHAIN_MAGIC_ORIGIN_FULL 0xd419170cba00
+
+#define KMSAN_POISON_NOCHECK	0x0
+#define KMSAN_POISON_CHECK	0x1
+#define KMSAN_POISON_FREE	0x2
+
+#define ORIGIN_SIZE 4
+
+#define META_SHADOW	(false)
+#define META_ORIGIN	(true)
+
+#define KMSAN_NESTED_CONTEXT_MAX (8)
+/* [0] for dummy per-CPU context */
+DECLARE_PER_CPU(struct kmsan_context_state[KMSAN_NESTED_CONTEXT_MAX],
+		kmsan_percpu_cstate);
+/* 0 for task context, |i>0| for kmsan_context_state[i]. */
+DECLARE_PER_CPU(int, kmsan_context_level);
+DECLARE_PER_CPU(int, kmsan_in_interrupt);
+DECLARE_PER_CPU(bool, kmsan_in_softirq);
+DECLARE_PER_CPU(bool, kmsan_in_nmi);
+
+extern spinlock_t report_lock;
+
+/* Stolen from kernel/printk/internal.h */
+#define PRINTK_SAFE_CONTEXT_MASK	 0x3fffffff
+/* Defined in kernel/printk/printk_safe.c */
+DECLARE_PER_CPU(int, printk_context);
+
+/* Only print iff printk hasn't acquired the console lock. */
+#define kmsan_pr_err(...) \
+	do { \
+		if (!(this_cpu_read(printk_context) & \
+		      PRINTK_SAFE_CONTEXT_MASK)) \
+			pr_err(__VA_ARGS__); \
+	} while (0)
+
+void kmsan_print_origin(depot_stack_handle_t origin);
+void kmsan_report(depot_stack_handle_t origin,
+		  void *address, int size, int off_first, int off_last,
+		  const void *user_addr, int reason);
+
+
+enum KMSAN_BUG_REASON {
+	REASON_ANY = 0,
+	REASON_COPY_TO_USER = 1,
+	REASON_USE_AFTER_FREE = 2,
+	REASON_SUBMIT_URB = 3,
+};
+
+/*
+ * When a compiler hook is invoked, it may make a call to instrumented code
+ * and eventually call itself recursively. To avoid that, we protect the
+ * runtime entry points with ENTER_RUNTIME()/LEAVE_RUNTIME() macros and exit
+ * the hook if IN_RUNTIME() is true. But when an interrupt occurs inside the
+ * runtime, the hooks won’t run either, which may lead to errors.
+ * Therefore we have to disable interrupts inside the runtime.
+ */
+DECLARE_PER_CPU(int, kmsan_in_runtime);
+DECLARE_PER_CPU(unsigned long, kmsan_runtime_last_caller);
+#define IN_RUNTIME()	(this_cpu_read(kmsan_in_runtime))
+#define ENTER_RUNTIME(irq_flags) \
+	do { \
+		preempt_disable(); \
+		local_irq_save(irq_flags); \
+		stop_nmi();		\
+		this_cpu_inc(kmsan_in_runtime); \
+		this_cpu_write(kmsan_runtime_last_caller, _THIS_IP_); \
+		BUG_ON(this_cpu_read(kmsan_in_runtime) > 1); \
+	} while (0)
+#define LEAVE_RUNTIME(irq_flags)	\
+	do {	\
+		this_cpu_dec(kmsan_in_runtime);	\
+		if (this_cpu_read(kmsan_in_runtime)) { \
+			kmsan_pr_err("kmsan_in_runtime: %d, last_caller: %pS\n", \
+				this_cpu_read(kmsan_in_runtime), \
+				this_cpu_read(kmsan_runtime_last_caller)); \
+			BUG(); \
+		}	\
+		restart_nmi();		\
+		local_irq_restore(irq_flags);	\
+		preempt_enable(); } while (0)
+
+void kmsan_memcpy_metadata(void *dst, void *src, size_t n);
+void kmsan_memmove_metadata(void *dst, void *src, size_t n);
+
+depot_stack_handle_t kmsan_save_stack(void);
+depot_stack_handle_t kmsan_save_stack_with_flags(gfp_t flags,
+						 unsigned int extra_bits);
+void kmsan_internal_poison_shadow(void *address, size_t size, gfp_t flags,
+				  unsigned int poison_flags);
+void kmsan_internal_unpoison_shadow(void *address, size_t size, bool checked);
+void kmsan_internal_memset_shadow(void *address, int b, size_t size,
+				  bool checked);
+depot_stack_handle_t kmsan_internal_chain_origin(depot_stack_handle_t id);
+void kmsan_write_aligned_origin(void *var, size_t size, u32 origin);
+
+void kmsan_internal_task_create(struct task_struct *task);
+void kmsan_internal_set_origin(void *addr, int size, u32 origin);
+void kmsan_set_origin_checked(void *addr, int size, u32 origin, bool checked);
+
+struct kmsan_context_state *task_kmsan_context_state(void);
+
+bool metadata_is_contiguous(void *addr, size_t size, bool is_origin);
+void kmsan_internal_check_memory(void *addr, size_t size, const void *user_addr,
+				 int reason);
+
+struct page *vmalloc_to_page_or_null(void *vaddr);
+
+/* Declared in mm/vmalloc.c */
+void __vunmap_page_range(unsigned long addr, unsigned long end);
+int __vmap_page_range_noflush(unsigned long start, unsigned long end,
+				   pgprot_t prot, struct page **pages);
+
+void *kmsan_internal_return_address(int arg);
+bool kmsan_internal_is_module_addr(void *vaddr);
+bool kmsan_internal_is_vmalloc_addr(void *addr);
+
+#endif  /* __MM_KMSAN_KMSAN_H */
diff --git a/mm/kmsan/kmsan_entry.c b/mm/kmsan/kmsan_entry.c
new file mode 100644
index 000000000000..9511a7dad541
--- /dev/null
+++ b/mm/kmsan/kmsan_entry.c
@@ -0,0 +1,130 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * KMSAN hooks for entry_64.S
+ *
+ * Copyright (C) 2018-2019 Google LLC
+ * Author: Alexander Potapenko <glider@google.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#include "kmsan.h"
+
+static void kmsan_context_enter(void)
+{
+	int level = this_cpu_read(kmsan_context_level) + 1;
+
+	BUG_ON(level >= KMSAN_NESTED_CONTEXT_MAX);
+	this_cpu_write(kmsan_context_level, level);
+}
+
+static void kmsan_context_exit(void)
+{
+	int level = this_cpu_read(kmsan_context_level) - 1;
+
+	BUG_ON(level < 0);
+	this_cpu_write(kmsan_context_level, level);
+}
+
+void kmsan_interrupt_enter(void)
+{
+	int in_interrupt = this_cpu_read(kmsan_in_interrupt);
+
+	/* Turns out it's possible for in_interrupt to be >0 here. */
+	kmsan_context_enter();
+	BUG_ON(in_interrupt > 1);
+	/* Can't check preempt_count() here, it may be zero. */
+	this_cpu_write(kmsan_in_interrupt, in_interrupt + 1);
+}
+EXPORT_SYMBOL(kmsan_interrupt_enter);
+
+void kmsan_interrupt_exit(void)
+{
+	int in_interrupt = this_cpu_read(kmsan_in_interrupt);
+
+	BUG_ON(!in_interrupt);
+	kmsan_context_exit();
+	/* Can't check preempt_count() here, it may be zero. */
+	this_cpu_write(kmsan_in_interrupt, in_interrupt - 1);
+}
+EXPORT_SYMBOL(kmsan_interrupt_exit);
+
+void kmsan_softirq_enter(void)
+{
+	bool in_softirq = this_cpu_read(kmsan_in_softirq);
+
+	BUG_ON(in_softirq);
+	kmsan_context_enter();
+	/* Can't check preempt_count() here, it may be zero. */
+	this_cpu_write(kmsan_in_softirq, true);
+}
+EXPORT_SYMBOL(kmsan_softirq_enter);
+
+void kmsan_softirq_exit(void)
+{
+	bool in_softirq = this_cpu_read(kmsan_in_softirq);
+
+	BUG_ON(!in_softirq);
+	kmsan_context_exit();
+	/* Can't check preempt_count() here, it may be zero. */
+	this_cpu_write(kmsan_in_softirq, false);
+}
+EXPORT_SYMBOL(kmsan_softirq_exit);
+
+void kmsan_nmi_enter(void)
+{
+	bool in_nmi = this_cpu_read(kmsan_in_nmi);
+
+	BUG_ON(in_nmi);
+	BUG_ON(preempt_count() & NMI_MASK);
+	kmsan_context_enter();
+	this_cpu_write(kmsan_in_nmi, true);
+}
+EXPORT_SYMBOL(kmsan_nmi_enter);
+
+void kmsan_nmi_exit(void)
+{
+	bool in_nmi = this_cpu_read(kmsan_in_nmi);
+
+	BUG_ON(!in_nmi);
+	BUG_ON(preempt_count() & NMI_MASK);
+	kmsan_context_exit();
+	this_cpu_write(kmsan_in_nmi, false);
+
+}
+EXPORT_SYMBOL(kmsan_nmi_exit);
+
+void kmsan_syscall_enter(void)
+{
+
+}
+EXPORT_SYMBOL(kmsan_syscall_enter);
+
+void kmsan_syscall_exit(void)
+{
+
+}
+EXPORT_SYMBOL(kmsan_syscall_exit);
+
+void kmsan_ist_enter(u64 shift_ist)
+{
+	kmsan_context_enter();
+}
+EXPORT_SYMBOL(kmsan_ist_enter);
+
+void kmsan_ist_exit(u64 shift_ist)
+{
+	kmsan_context_exit();
+}
+EXPORT_SYMBOL(kmsan_ist_exit);
+
+void kmsan_unpoison_pt_regs(struct pt_regs *regs)
+{
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+	kmsan_internal_unpoison_shadow(regs, sizeof(*regs), /*checked*/true);
+}
+EXPORT_SYMBOL(kmsan_unpoison_pt_regs);
diff --git a/mm/kmsan/kmsan_hooks.c b/mm/kmsan/kmsan_hooks.c
new file mode 100644
index 000000000000..c0b941be528b
--- /dev/null
+++ b/mm/kmsan/kmsan_hooks.c
@@ -0,0 +1,416 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * KMSAN hooks for kernel subsystems.
+ *
+ * These functions handle creation of KMSAN metadata for memory allocations.
+ *
+ * Copyright (C) 2018-2019 Google LLC
+ * Author: Alexander Potapenko <glider@google.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#include <asm/cacheflush.h>
+#include <linux/gfp.h>
+#include <linux/i2c.h>
+#include <linux/mm.h>
+#include <linux/mm_types.h>
+#include <linux/skbuff.h>
+#include <linux/slab.h>
+#include <linux/usb.h>
+
+#include "../slab.h"
+#include "kmsan.h"
+
+/* TODO(glider): do we need to export these symbols? */
+
+/*
+ * The functions may call back to instrumented code, which, in turn, may call
+ * these hooks again. To avoid re-entrancy, we use __GFP_NO_KMSAN_SHADOW.
+ * Instrumented functions shouldn't be called under
+ * ENTER_RUNTIME()/LEAVE_RUNTIME(), because this will lead to skipping
+ * effects of functions like memset() inside instrumented code.
+ */
+/* Called from kernel/kthread.c, kernel/fork.c */
+void kmsan_task_create(struct task_struct *task)
+{
+	unsigned long irq_flags;
+
+	if (!task)
+		return;
+	ENTER_RUNTIME(irq_flags);
+	kmsan_internal_task_create(task);
+	LEAVE_RUNTIME(irq_flags);
+}
+EXPORT_SYMBOL(kmsan_task_create);
+
+
+/* Called from kernel/exit.c */
+void kmsan_task_exit(struct task_struct *task)
+{
+	unsigned long irq_flags;
+	struct kmsan_task_state *state = &task->kmsan;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+
+	ENTER_RUNTIME(irq_flags);
+	state->allow_reporting = false;
+
+	LEAVE_RUNTIME(irq_flags);
+}
+EXPORT_SYMBOL(kmsan_task_exit);
+
+/* Called from mm/slub.c */
+void kmsan_slab_alloc(struct kmem_cache *s, void *object, gfp_t flags)
+{
+	unsigned long irq_flags;
+
+	if (unlikely(object == NULL))
+		return;
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+	/*
+	 * There's a ctor or this is an RCU cache - do nothing. The memory
+	 * status hasn't changed since last use.
+	 */
+	if (s->ctor || (s->flags & SLAB_TYPESAFE_BY_RCU))
+		return;
+
+	ENTER_RUNTIME(irq_flags);
+	if (flags & __GFP_ZERO) {
+		kmsan_internal_unpoison_shadow(object, s->object_size,
+					       KMSAN_POISON_CHECK);
+	} else {
+		kmsan_internal_poison_shadow(object, s->object_size, flags,
+					     KMSAN_POISON_CHECK);
+	}
+	LEAVE_RUNTIME(irq_flags);
+}
+
+/* Called from mm/slub.c */
+void kmsan_slab_free(struct kmem_cache *s, void *object)
+{
+	unsigned long irq_flags;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+	ENTER_RUNTIME(irq_flags);
+
+	/* RCU slabs could be legally used after free within the RCU period */
+	if (unlikely(s->flags & (SLAB_TYPESAFE_BY_RCU | SLAB_POISON)))
+		goto leave;
+	if (s->ctor)
+		goto leave;
+	kmsan_internal_poison_shadow(object, s->object_size,
+				     GFP_KERNEL,
+				     KMSAN_POISON_CHECK | KMSAN_POISON_FREE);
+leave:
+	LEAVE_RUNTIME(irq_flags);
+}
+
+/* Called from mm/slub.c */
+void kmsan_kmalloc_large(const void *ptr, size_t size, gfp_t flags)
+{
+	unsigned long irq_flags;
+
+	if (unlikely(ptr == NULL))
+		return;
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+	ENTER_RUNTIME(irq_flags);
+	if (flags & __GFP_ZERO) {
+		kmsan_internal_unpoison_shadow((void *)ptr, size,
+					       /*checked*/true);
+	} else {
+		kmsan_internal_poison_shadow((void *)ptr, size, flags,
+					     KMSAN_POISON_CHECK);
+	}
+	LEAVE_RUNTIME(irq_flags);
+}
+
+/* Called from mm/slub.c */
+void kmsan_kfree_large(const void *ptr)
+{
+	struct page *page;
+	unsigned long irq_flags;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+	ENTER_RUNTIME(irq_flags);
+	page = virt_to_head_page((void *)ptr);
+	BUG_ON(ptr != page_address(page));
+	kmsan_internal_poison_shadow(
+		(void *)ptr, PAGE_SIZE << compound_order(page), GFP_KERNEL,
+		KMSAN_POISON_CHECK | KMSAN_POISON_FREE);
+	LEAVE_RUNTIME(irq_flags);
+}
+
+
+static unsigned long vmalloc_shadow(unsigned long addr)
+{
+	return (unsigned long)kmsan_get_metadata((void *)addr, 1, META_SHADOW);
+}
+
+static unsigned long vmalloc_origin(unsigned long addr)
+{
+	return (unsigned long)kmsan_get_metadata((void *)addr, 1, META_ORIGIN);
+}
+
+/* Called from mm/vmalloc.c */
+void kmsan_vunmap_page_range(unsigned long start, unsigned long end)
+{
+	__vunmap_page_range(vmalloc_shadow(start), vmalloc_shadow(end));
+	__vunmap_page_range(vmalloc_origin(start), vmalloc_origin(end));
+}
+
+/* Called from lib/ioremap.c */
+/*
+ * This function creates new shadow/origin pages for the physical pages mapped
+ * into the virtual memory. If those physical pages already had shadow/origin,
+ * those are ignored.
+ */
+void kmsan_ioremap_page_range(unsigned long start, unsigned long end,
+	phys_addr_t phys_addr, pgprot_t prot)
+{
+	unsigned long irq_flags;
+	struct page *shadow, *origin;
+	int i, nr;
+	unsigned long off = 0;
+	gfp_t gfp_mask = GFP_KERNEL | __GFP_ZERO | __GFP_NO_KMSAN_SHADOW;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+
+	nr = (end - start) / PAGE_SIZE;
+	ENTER_RUNTIME(irq_flags);
+	for (i = 0; i < nr; i++, off += PAGE_SIZE) {
+		shadow = alloc_pages(gfp_mask, 1);
+		origin = alloc_pages(gfp_mask, 1);
+		__vmap_page_range_noflush(vmalloc_shadow(start + off),
+				vmalloc_shadow(start + off + PAGE_SIZE),
+				prot, &shadow);
+		__vmap_page_range_noflush(vmalloc_origin(start + off),
+				vmalloc_origin(start + off + PAGE_SIZE),
+				prot, &origin);
+	}
+	flush_cache_vmap(vmalloc_shadow(start), vmalloc_shadow(end));
+	flush_cache_vmap(vmalloc_origin(start), vmalloc_origin(end));
+	LEAVE_RUNTIME(irq_flags);
+}
+
+void kmsan_iounmap_page_range(unsigned long start, unsigned long end)
+{
+	int i, nr;
+	struct page *shadow, *origin;
+	unsigned long v_shadow, v_origin;
+	unsigned long irq_flags;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+
+	nr = (end - start) / PAGE_SIZE;
+	ENTER_RUNTIME(irq_flags);
+	v_shadow = (unsigned long)vmalloc_shadow(start);
+	v_origin = (unsigned long)vmalloc_origin(start);
+	for (i = 0; i < nr; i++, v_shadow += PAGE_SIZE, v_origin += PAGE_SIZE) {
+		shadow = vmalloc_to_page_or_null((void *)v_shadow);
+		origin = vmalloc_to_page_or_null((void *)v_origin);
+		__vunmap_page_range(v_shadow, v_shadow + PAGE_SIZE);
+		__vunmap_page_range(v_origin, v_origin + PAGE_SIZE);
+		if (shadow)
+			__free_pages(shadow, 1);
+		if (origin)
+			__free_pages(origin, 1);
+	}
+	LEAVE_RUNTIME(irq_flags);
+}
+
+/* Called from include/linux/uaccess.h, include/linux/uaccess.h */
+void kmsan_copy_to_user(const void *to, const void *from,
+			size_t to_copy, size_t left)
+{
+	void *shadow;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+	/*
+	 * At this point we've copied the memory already. It's hard to check it
+	 * before copying, as the size of actually copied buffer is unknown.
+	 */
+
+	/* copy_to_user() may copy zero bytes. No need to check. */
+	if (!to_copy)
+		return;
+	/* Or maybe copy_to_user() failed to copy anything. */
+	if (to_copy == left)
+		return;
+	if ((u64)to < TASK_SIZE) {
+		/* This is a user memory access, check it. */
+		kmsan_internal_check_memory((void *)from, to_copy - left, to,
+						REASON_COPY_TO_USER);
+		return;
+	}
+	/* Otherwise this is a kernel memory access. This happens when a compat
+	 * syscall passes an argument allocated on the kernel stack to a real
+	 * syscall.
+	 * Don't check anything, just copy the shadow of the copied bytes.
+	 */
+	shadow = kmsan_get_metadata((void *)to, to_copy - left, META_SHADOW);
+	if (shadow)
+		kmsan_memcpy_metadata((void *)to, (void *)from, to_copy - left);
+}
+EXPORT_SYMBOL(kmsan_copy_to_user);
+
+void kmsan_poison_shadow(const void *address, size_t size, gfp_t flags)
+{
+	unsigned long irq_flags;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+	ENTER_RUNTIME(irq_flags);
+	/* The users may want to poison/unpoison random memory. */
+	kmsan_internal_poison_shadow((void *)address, size, flags,
+				     KMSAN_POISON_NOCHECK);
+	LEAVE_RUNTIME(irq_flags);
+}
+EXPORT_SYMBOL(kmsan_poison_shadow);
+
+void kmsan_unpoison_shadow(const void *address, size_t size)
+{
+	unsigned long irq_flags;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+
+	ENTER_RUNTIME(irq_flags);
+	/* The users may want to poison/unpoison random memory. */
+	kmsan_internal_unpoison_shadow((void *)address, size,
+				       KMSAN_POISON_NOCHECK);
+	LEAVE_RUNTIME(irq_flags);
+}
+EXPORT_SYMBOL(kmsan_unpoison_shadow);
+
+void kmsan_check_memory(const void *addr, size_t size)
+{
+	return kmsan_internal_check_memory((void *)addr, size, /*user_addr*/ 0,
+					   REASON_ANY);
+}
+EXPORT_SYMBOL(kmsan_check_memory);
+
+void kmsan_gup_pgd_range(struct page **pages, int nr)
+{
+	int i;
+	void *page_addr;
+
+	/*
+	 * gup_pgd_range() has just created a number of new pages that KMSAN
+	 * treats as uninitialized. In the case they belong to the userspace
+	 * memory, unpoison the corresponding kernel pages.
+	 */
+	for (i = 0; i < nr; i++) {
+		page_addr = page_address(pages[i]);
+		if (((u64)page_addr < TASK_SIZE) &&
+		    ((u64)page_addr + PAGE_SIZE < TASK_SIZE))
+			kmsan_unpoison_shadow(page_addr, PAGE_SIZE);
+	}
+
+}
+EXPORT_SYMBOL(kmsan_gup_pgd_range);
+
+/* Helper function to check an SKB. */
+void kmsan_check_skb(const struct sk_buff *skb)
+{
+	int start = skb_headlen(skb);
+	struct sk_buff *frag_iter;
+	int i, copy = 0;
+	skb_frag_t *f;
+	u32 p_off, p_len, copied;
+	struct page *p;
+	u8 *vaddr;
+
+	if (!skb || !skb->len)
+		return;
+
+	kmsan_internal_check_memory(skb->data, skb_headlen(skb), 0, REASON_ANY);
+	if (skb_is_nonlinear(skb)) {
+		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
+			f = &skb_shinfo(skb)->frags[i];
+
+			skb_frag_foreach_page(f,
+					      skb_frag_off(f)  - start,
+					      copy, p, p_off, p_len, copied) {
+
+				vaddr = kmap_atomic(p);
+				kmsan_internal_check_memory(vaddr + p_off,
+						p_len, /*user_addr*/ 0,
+						REASON_ANY);
+				kunmap_atomic(vaddr);
+			}
+		}
+	}
+	skb_walk_frags(skb, frag_iter)
+		kmsan_check_skb(frag_iter);
+}
+EXPORT_SYMBOL(kmsan_check_skb);
+
+/* Helper function to check an URB. */
+void kmsan_handle_urb(const struct urb *urb, bool is_out)
+{
+	if (!urb)
+		return;
+	if (is_out)
+		kmsan_internal_check_memory(urb->transfer_buffer,
+					    urb->transfer_buffer_length,
+					    /*user_addr*/ 0, REASON_SUBMIT_URB);
+	else
+		kmsan_internal_unpoison_shadow(urb->transfer_buffer,
+					       urb->transfer_buffer_length,
+					       /*checked*/false);
+}
+EXPORT_SYMBOL(kmsan_handle_urb);
+
+/* Helper function to check I2C-transferred data. */
+void kmsan_handle_i2c_transfer(struct i2c_msg *msgs, int num)
+{
+	int i;
+
+	if (!msgs)
+		return;
+	for (i = 0; i < num; i++) {
+		if (msgs[i].flags & I2C_M_RD)
+			kmsan_internal_unpoison_shadow(msgs[i].buf,
+						       msgs[i].len,
+						       /*checked*/false);
+		else
+			kmsan_internal_check_memory(msgs[i].buf, msgs[i].len,
+						    /*user_addr*/0,
+						    REASON_ANY);
+	}
+}
+
+/*
+ * Helper function to check printk.
+ * Some optimizations may cause *fmt to be uninitialized, that's why it's
+ * passed by reference.
+ */
+void kmsan_handle_vprintk(const char **fmt, va_list args)
+{
+	unsigned long irq_flags;
+	size_t fmt_size;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+	ENTER_RUNTIME(irq_flags);
+	fmt_size = strlen(*fmt);
+	LEAVE_RUNTIME(irq_flags);
+	/* TODO(glider): check |args|. */
+	kmsan_internal_check_memory(fmt, sizeof(char *), /*user_addr*/0,
+				    REASON_ANY);
+	kmsan_internal_check_memory((void *)(*fmt), fmt_size, /*user_addr*/0,
+				    REASON_ANY);
+}
+EXPORT_SYMBOL(kmsan_handle_vprintk);
diff --git a/mm/kmsan/kmsan_init.c b/mm/kmsan/kmsan_init.c
new file mode 100644
index 000000000000..606c4b0c90d3
--- /dev/null
+++ b/mm/kmsan/kmsan_init.c
@@ -0,0 +1,88 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * KMSAN initialization routines.
+ *
+ * Copyright (C) 2017-2019 Google LLC
+ * Author: Alexander Potapenko <glider@google.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#include "kmsan.h"
+
+#include <asm/cpu_entry_area.h>
+#include <linux/mm.h>
+#include <linux/memblock.h>
+
+#define NUM_FUTURE_RANGES 128
+struct start_end_pair {
+	void *start, *end;
+};
+
+static struct start_end_pair start_end_pairs[NUM_FUTURE_RANGES] __initdata;
+static int future_index __initdata;
+
+/*
+ * Record a range of memory for which the metadata pages will be created once
+ * the page allocator becomes available.
+ * TODO(glider): squash together ranges belonging to the same page.
+ */
+static void __init kmsan_record_future_shadow_range(void *start, void *end)
+{
+	BUG_ON(future_index == NUM_FUTURE_RANGES);
+	BUG_ON((start >= end) || !start || !end);
+	start_end_pairs[future_index].start = start;
+	start_end_pairs[future_index].end = end;
+	future_index++;
+}
+
+extern char _sdata[], _edata[];
+
+
+
+/*
+ * Initialize the shadow for existing mappings during kernel initialization.
+ * These include kernel text/data sections, NODE_DATA and future ranges
+ * registered while creating other data (e.g. percpu).
+ *
+ * Allocations via memblock can be only done before slab is initialized.
+ */
+void __init kmsan_initialize_shadow(void)
+{
+	int nid;
+	u64 i;
+	const size_t nd_size = roundup(sizeof(pg_data_t), PAGE_SIZE);
+	phys_addr_t p_start, p_end;
+
+	for_each_reserved_mem_region(i, &p_start, &p_end) {
+		kmsan_record_future_shadow_range(phys_to_virt(p_start),
+						 phys_to_virt(p_end+1));
+	}
+	/* Allocate shadow for .data */
+	kmsan_record_future_shadow_range(_sdata, _edata);
+
+	/*
+	 * TODO(glider): alloc_node_data() in arch/x86/mm/numa.c uses
+	 * sizeof(pg_data_t).
+	 */
+	for_each_online_node(nid)
+		kmsan_record_future_shadow_range(
+			NODE_DATA(nid),	(char *)NODE_DATA(nid) + nd_size);
+
+	for (i = 0; i < future_index; i++)
+		kmsan_init_alloc_meta_for_range(start_end_pairs[i].start,
+						start_end_pairs[i].end);
+}
+EXPORT_SYMBOL(kmsan_initialize_shadow);
+
+void __init kmsan_initialize(void)
+{
+	/* Assuming current is init_task */
+	kmsan_internal_task_create(current);
+	kmsan_pr_err("Starting KernelMemorySanitizer\n");
+	kmsan_ready = true;
+}
+EXPORT_SYMBOL(kmsan_initialize);
diff --git a/mm/kmsan/kmsan_instr.c b/mm/kmsan/kmsan_instr.c
new file mode 100644
index 000000000000..74cb7cee7f70
--- /dev/null
+++ b/mm/kmsan/kmsan_instr.c
@@ -0,0 +1,259 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * KMSAN compiler API.
+ *
+ * Copyright (C) 2017-2019 Google LLC
+ * Author: Alexander Potapenko <glider@google.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#include "kmsan.h"
+#include <linux/gfp.h>
+#include <linux/mm.h>
+
+static bool is_bad_asm_addr(void *addr, u64 size, bool is_store)
+{
+	if ((u64)addr < TASK_SIZE)
+		return true;
+	if (!kmsan_get_metadata(addr, size, META_SHADOW))
+		return true;
+	return false;
+}
+
+struct shadow_origin_ptr __msan_metadata_ptr_for_load_n(void *addr, u64 size)
+{
+	return kmsan_get_shadow_origin_ptr(addr, size, /*store*/false);
+}
+EXPORT_SYMBOL(__msan_metadata_ptr_for_load_n);
+
+struct shadow_origin_ptr __msan_metadata_ptr_for_store_n(void *addr, u64 size)
+{
+	return kmsan_get_shadow_origin_ptr(addr, size, /*store*/true);
+}
+EXPORT_SYMBOL(__msan_metadata_ptr_for_store_n);
+
+#define DECLARE_METADATA_PTR_GETTER(size)	\
+struct shadow_origin_ptr __msan_metadata_ptr_for_load_##size(void *addr) \
+{		\
+	return kmsan_get_shadow_origin_ptr(addr, size, /*store*/false);	\
+}		\
+EXPORT_SYMBOL(__msan_metadata_ptr_for_load_##size);			\
+		\
+struct shadow_origin_ptr __msan_metadata_ptr_for_store_##size(void *addr) \
+{									\
+	return kmsan_get_shadow_origin_ptr(addr, size, /*store*/true);	\
+}									\
+EXPORT_SYMBOL(__msan_metadata_ptr_for_store_##size)
+
+DECLARE_METADATA_PTR_GETTER(1);
+DECLARE_METADATA_PTR_GETTER(2);
+DECLARE_METADATA_PTR_GETTER(4);
+DECLARE_METADATA_PTR_GETTER(8);
+
+void __msan_instrument_asm_store(void *addr, u64 size)
+{
+	unsigned long irq_flags;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+	/*
+	 * Most of the accesses are below 32 bytes. The two exceptions so far
+	 * are clwb() (64 bytes) and FPU state (512 bytes).
+	 * It's unlikely that the assembly will touch more than 512 bytes.
+	 */
+	if (size > 512)
+		size = 8;
+	if (is_bad_asm_addr(addr, size, /*is_store*/true))
+		return;
+	ENTER_RUNTIME(irq_flags);
+	/* Unpoisoning the memory on best effort. */
+	kmsan_internal_unpoison_shadow(addr, size, /*checked*/false);
+	LEAVE_RUNTIME(irq_flags);
+}
+EXPORT_SYMBOL(__msan_instrument_asm_store);
+
+void *__msan_memmove(void *dst, void *src, u64 n)
+{
+	void *result;
+	void *shadow_dst;
+
+	result = __memmove(dst, src, n);
+	if (!n)
+		/* Some people call memmove() with zero length. */
+		return result;
+	if (!kmsan_ready || IN_RUNTIME())
+		return result;
+
+	/* Ok to skip address check here, we'll do it later. */
+	shadow_dst = kmsan_get_metadata(dst, n, META_SHADOW);
+
+	if (!shadow_dst)
+		/* Can happen e.g. if the memory is untracked. */
+		return result;
+
+	kmsan_memmove_metadata(dst, src, n);
+
+	return result;
+}
+EXPORT_SYMBOL(__msan_memmove);
+
+void *__msan_memmove_nosanitize(void *dst, void *src, u64 n)
+{
+	return __memmove(dst, src, n);
+}
+EXPORT_SYMBOL(__msan_memmove_nosanitize);
+
+void *__msan_memcpy(void *dst, const void *src, u64 n)
+{
+	void *result;
+	void *shadow_dst;
+
+	result = __memcpy(dst, src, n);
+	if (!n)
+		/* Some people call memcpy() with zero length. */
+		return result;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return result;
+
+	/* Ok to skip address check here, we'll do it later. */
+	shadow_dst = kmsan_get_metadata(dst, n, META_SHADOW);
+	if (!shadow_dst)
+		/* Can happen e.g. if the memory is untracked. */
+		return result;
+
+	kmsan_memcpy_metadata(dst, (void *)src, n);
+
+	return result;
+}
+EXPORT_SYMBOL(__msan_memcpy);
+
+void *__msan_memcpy_nosanitize(void *dst, void *src, u64 n)
+{
+	return __memcpy(dst, src, n);
+}
+EXPORT_SYMBOL(__msan_memcpy_nosanitize);
+
+void *__msan_memset(void *dst, int c, size_t n)
+{
+	void *result;
+	unsigned long irq_flags;
+	depot_stack_handle_t new_origin;
+	unsigned int shadow;
+
+	result = __memset(dst, c, n);
+	if (!kmsan_ready || IN_RUNTIME())
+		return result;
+
+	ENTER_RUNTIME(irq_flags);
+	shadow = 0;
+	kmsan_internal_memset_shadow(dst, shadow, n, /*checked*/false);
+	new_origin = 0;
+	kmsan_internal_set_origin(dst, n, new_origin);
+	LEAVE_RUNTIME(irq_flags);
+
+	return result;
+}
+EXPORT_SYMBOL(__msan_memset);
+
+void *__msan_memset_nosanitize(void *dst, int c, size_t n)
+{
+	return __memset(dst, c, n);
+}
+EXPORT_SYMBOL(__msan_memset_nosanitize);
+
+depot_stack_handle_t __msan_chain_origin(depot_stack_handle_t origin)
+{
+	depot_stack_handle_t ret = 0;
+	unsigned long irq_flags;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return ret;
+
+	/* Creating new origins may allocate memory. */
+	ENTER_RUNTIME(irq_flags);
+	ret = kmsan_internal_chain_origin(origin);
+	LEAVE_RUNTIME(irq_flags);
+	return ret;
+}
+EXPORT_SYMBOL(__msan_chain_origin);
+
+void __msan_poison_alloca(void *address, u64 size, char *descr)
+{
+	depot_stack_handle_t handle;
+	unsigned long entries[4];
+	unsigned long irq_flags;
+	u64 size_copy = size, to_fill;
+	u64 addr_copy = (u64)address;
+	u64 page_offset;
+	void *shadow_start;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+
+	while (size_copy) {
+		page_offset = addr_copy % PAGE_SIZE;
+		to_fill = min(PAGE_SIZE - page_offset, size_copy);
+		shadow_start = kmsan_get_metadata((void *)addr_copy, to_fill,
+						  META_SHADOW);
+		if (!shadow_start)
+			/* Can happen e.g. if the memory is untracked. */
+			continue;
+		__memset(shadow_start, -1, to_fill);
+		addr_copy += to_fill;
+		size_copy -= to_fill;
+	}
+
+	entries[0] = KMSAN_ALLOCA_MAGIC_ORIGIN;
+	entries[1] = (u64)descr;
+	entries[2] = (u64)__builtin_return_address(0);
+	entries[3] = (u64)kmsan_internal_return_address(1);
+
+	/* stack_depot_save() may allocate memory. */
+	ENTER_RUNTIME(irq_flags);
+	handle = stack_depot_save(entries, ARRAY_SIZE(entries), GFP_ATOMIC);
+	LEAVE_RUNTIME(irq_flags);
+	kmsan_internal_set_origin(address, size, handle);
+}
+EXPORT_SYMBOL(__msan_poison_alloca);
+
+void __msan_unpoison_alloca(void *address, u64 size)
+{
+	unsigned long irq_flags;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+
+	ENTER_RUNTIME(irq_flags);
+	/* Assuming the shadow exists. */
+	kmsan_internal_unpoison_shadow(address, size, /*checked*/true);
+	LEAVE_RUNTIME(irq_flags);
+}
+EXPORT_SYMBOL(__msan_unpoison_alloca);
+
+void __msan_warning(u32 origin)
+{
+	unsigned long irq_flags;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+	ENTER_RUNTIME(irq_flags);
+	kmsan_report(origin, /*address*/0, /*size*/0,
+		/*off_first*/0, /*off_last*/0, /*user_addr*/0, REASON_ANY);
+	LEAVE_RUNTIME(irq_flags);
+}
+EXPORT_SYMBOL(__msan_warning);
+
+struct kmsan_context_state *__msan_get_context_state(void)
+{
+	struct kmsan_context_state *ret;
+
+	ret = task_kmsan_context_state();
+	BUG_ON(!ret);
+	return ret;
+}
+EXPORT_SYMBOL(__msan_get_context_state);
diff --git a/mm/kmsan/kmsan_report.c b/mm/kmsan/kmsan_report.c
new file mode 100644
index 000000000000..5dddacb8e6d8
--- /dev/null
+++ b/mm/kmsan/kmsan_report.c
@@ -0,0 +1,135 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * KMSAN error reporting routines.
+ *
+ * Copyright (C) 2019 Google LLC
+ * Author: Alexander Potapenko <glider@google.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#include <linux/console.h>
+#include <linux/stackdepot.h>
+#include <linux/stacktrace.h>
+
+#include "kmsan.h"
+
+DEFINE_SPINLOCK(report_lock);
+
+void kmsan_print_origin(depot_stack_handle_t origin)
+{
+	unsigned long *entries = NULL, *chained_entries = NULL;
+	unsigned long nr_entries, chained_nr_entries, magic;
+	char *descr = NULL;
+	void *pc1 = NULL, *pc2 = NULL;
+	depot_stack_handle_t head;
+
+	if (!origin) {
+		kmsan_pr_err("Origin not found, presumably a false report.\n");
+		return;
+	}
+
+	while (true) {
+		nr_entries = stack_depot_fetch(origin, &entries);
+		magic = nr_entries ? (entries[0] & KMSAN_MAGIC_MASK) : 0;
+		if ((nr_entries == 4) && (magic == KMSAN_ALLOCA_MAGIC_ORIGIN)) {
+			descr = (char *)entries[1];
+			pc1 = (void *)entries[2];
+			pc2 = (void *)entries[3];
+			kmsan_pr_err("Local variable description: %s\n", descr);
+			kmsan_pr_err("Variable was created at:\n");
+			kmsan_pr_err(" %pS\n", pc1);
+			kmsan_pr_err(" %pS\n", pc2);
+			break;
+		}
+		if ((nr_entries == 3) &&
+		    (magic == KMSAN_CHAIN_MAGIC_ORIGIN_FULL)) {
+			head = entries[1];
+			origin = entries[2];
+			kmsan_pr_err("Uninit was stored to memory at:\n");
+			chained_nr_entries =
+				stack_depot_fetch(head, &chained_entries);
+			stack_trace_print(chained_entries, chained_nr_entries,
+					  0);
+			kmsan_pr_err("\n");
+			continue;
+		}
+		kmsan_pr_err("Uninit was created at:\n");
+		if (entries)
+			stack_trace_print(entries, nr_entries, 0);
+		else
+			kmsan_pr_err("No stack\n");
+		break;
+	}
+}
+
+void kmsan_report(depot_stack_handle_t origin,
+		  void *address, int size, int off_first, int off_last,
+		  const void *user_addr, int reason)
+{
+	unsigned long flags;
+	unsigned long *entries;
+	unsigned int nr_entries;
+	bool is_uaf = false;
+	char *bug_type = NULL;
+
+	if (!kmsan_ready)
+		return;
+	if (!current->kmsan.allow_reporting)
+		return;
+	if (this_cpu_read(printk_context) & PRINTK_SAFE_CONTEXT_MASK)
+		return;
+	if (!origin)
+		return;
+
+	nr_entries = stack_depot_fetch(origin, &entries);
+
+	current->kmsan.allow_reporting = false;
+	spin_lock_irqsave(&report_lock, flags);
+	kmsan_pr_err("=====================================================\n");
+	if (get_dsh_extra_bits(origin) & 1)
+		is_uaf = true;
+	switch (reason) {
+	case REASON_ANY:
+		bug_type = is_uaf ? "use-after-free" : "uninit-value";
+		break;
+	case REASON_COPY_TO_USER:
+		bug_type = is_uaf ? "kernel-infoleak-after-free" :
+				    "kernel-infoleak";
+		break;
+	case REASON_SUBMIT_URB:
+		bug_type = is_uaf ? "kernel-usb-infoleak-after-free" :
+				    "kernel-usb-infoleak";
+		break;
+	}
+	kmsan_pr_err("BUG: KMSAN: %s in %pS\n",
+		     bug_type, kmsan_internal_return_address(2));
+	dump_stack();
+	kmsan_pr_err("\n");
+
+	kmsan_print_origin(origin);
+
+	if (size) {
+		kmsan_pr_err("\n");
+		if (off_first == off_last)
+			kmsan_pr_err("Byte %d of %d is uninitialized\n",
+				     off_first, size);
+		else
+			kmsan_pr_err("Bytes %d-%d of %d are uninitialized\n",
+				     off_first, off_last, size);
+	}
+	if (address)
+		kmsan_pr_err("Memory access of size %d starts at %px\n",
+			     size, address);
+	if (user_addr && reason == REASON_COPY_TO_USER)
+		kmsan_pr_err("Data copied to user address %px\n", user_addr);
+	kmsan_pr_err("=====================================================\n");
+	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
+	spin_unlock_irqrestore(&report_lock, flags);
+	if (panic_on_warn)
+		panic("panic_on_warn set ...\n");
+	current->kmsan.allow_reporting = true;
+}
diff --git a/mm/kmsan/kmsan_shadow.c b/mm/kmsan/kmsan_shadow.c
new file mode 100644
index 000000000000..06801d76e6b8
--- /dev/null
+++ b/mm/kmsan/kmsan_shadow.c
@@ -0,0 +1,543 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * KMSAN shadow implementation.
+ *
+ * Copyright (C) 2017-2019 Google LLC
+ * Author: Alexander Potapenko <glider@google.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#include <asm/cpu_entry_area.h>
+#include <asm/page.h>
+#include <asm/pgtable_64_types.h>
+#include <asm/tlbflush.h>
+#include <linux/memblock.h>
+#include <linux/mm_types.h>
+#include <linux/percpu-defs.h>
+#include <linux/slab.h>
+#include <linux/smp.h>
+#include <linux/stddef.h>
+
+#include "kmsan.h"
+#include "kmsan_shadow.h"
+
+#define shadow_page_for(page) \
+	((page)->shadow)
+
+#define origin_page_for(page) \
+	((page)->origin)
+
+#define shadow_ptr_for(page) \
+	(page_address((page)->shadow))
+
+#define origin_ptr_for(page) \
+	(page_address((page)->origin))
+
+#define has_shadow_page(page) \
+	(!!((page)->shadow))
+
+#define has_origin_page(page) \
+	(!!((page)->origin))
+
+#define set_no_shadow_origin_page(page)	\
+	do {				\
+		(page)->shadow = NULL;	\
+		(page)->origin = NULL;	\
+	} while (0) /**/
+
+#define is_ignored_page(page)	\
+	(!!(((u64)((page)->shadow)) % 2))
+
+#define ignore_page(pg)			\
+		((pg)->shadow = (struct page *)((u64)((pg)->shadow) | 1)) \
+
+DEFINE_PER_CPU(char[CPU_ENTRY_AREA_SIZE], cpu_entry_area_shadow);
+DEFINE_PER_CPU(char[CPU_ENTRY_AREA_SIZE], cpu_entry_area_origin);
+
+/*
+ * Dummy load and store pages to be used when the real metadata is unavailable.
+ * There are separate pages for loads and stores, so that every load returns a
+ * zero, and every store doesn't affect other stores.
+ */
+char dummy_load_page[PAGE_SIZE] __aligned(PAGE_SIZE);
+char dummy_store_page[PAGE_SIZE] __aligned(PAGE_SIZE);
+
+/*
+ * Taken from arch/x86/mm/physaddr.h to avoid using an instrumented version.
+ */
+static int kmsan_phys_addr_valid(unsigned long addr)
+{
+#ifdef CONFIG_PHYS_ADDR_T_64BIT
+	return !(addr >> boot_cpu_data.x86_phys_bits);
+#else
+	return 1;
+#endif
+}
+
+/*
+ * Taken from arch/x86/mm/physaddr.c to avoid using an instrumented version.
+ */
+static bool kmsan_virt_addr_valid(void *addr)
+{
+	unsigned long x = (unsigned long)addr;
+	unsigned long y = x - __START_KERNEL_map;
+
+	/* use the carry flag to determine if x was < __START_KERNEL_map */
+	if (unlikely(x > y)) {
+		x = y + phys_base;
+
+		if (y >= KERNEL_IMAGE_SIZE)
+			return false;
+	} else {
+		x = y + (__START_KERNEL_map - PAGE_OFFSET);
+
+		/* carry flag will be set if starting x was >= PAGE_OFFSET */
+		if ((x > y) || !kmsan_phys_addr_valid(x))
+			return false;
+	}
+
+	return pfn_valid(x >> PAGE_SHIFT);
+}
+
+static unsigned long vmalloc_meta(void *addr, bool is_origin)
+{
+	unsigned long addr64 = (unsigned long)addr, off;
+
+	BUG_ON(is_origin && !IS_ALIGNED(addr64, ORIGIN_SIZE));
+	if (kmsan_internal_is_vmalloc_addr(addr)) {
+		return addr64 + (is_origin ? VMALLOC_ORIGIN_OFFSET
+					   : VMALLOC_SHADOW_OFFSET);
+	}
+	if (kmsan_internal_is_module_addr(addr)) {
+		off = addr64 - MODULES_VADDR;
+		return off + (is_origin ? MODULES_ORIGIN_START
+					: MODULES_SHADOW_START);
+	}
+	return 0;
+}
+
+static void *get_cea_meta_or_null(void *addr, bool is_origin)
+{
+	int cpu = smp_processor_id();
+	int off;
+	char *metadata_array;
+
+	if (((u64)addr < CPU_ENTRY_AREA_BASE) ||
+	    ((u64)addr >= (CPU_ENTRY_AREA_BASE + CPU_ENTRY_AREA_MAP_SIZE)))
+		return NULL;
+	off = (char *)addr - (char *)get_cpu_entry_area(cpu);
+	if ((off < 0) || (off >= CPU_ENTRY_AREA_SIZE))
+		return NULL;
+	metadata_array = is_origin ? cpu_entry_area_origin :
+				     cpu_entry_area_shadow;
+	return &per_cpu(metadata_array[off], cpu);
+}
+
+static struct page *virt_to_page_or_null(void *vaddr)
+{
+	if (kmsan_virt_addr_valid(vaddr))
+		return virt_to_page(vaddr);
+	else
+		return NULL;
+}
+
+struct shadow_origin_ptr kmsan_get_shadow_origin_ptr(void *address, u64 size,
+						     bool store)
+{
+	struct shadow_origin_ptr ret;
+	struct page *page;
+	u64 pad, offset, o_offset;
+	const u64 addr64 = (u64)address;
+	u64 o_addr64 = (u64)address;
+	void *shadow;
+
+	if (size > PAGE_SIZE) {
+		WARN(1, "size too big in %s(%px, %d, %d)\n",
+		     __func__, address, size, store);
+		BUG();
+	}
+	if (store) {
+		ret.s = dummy_store_page;
+		ret.o = dummy_store_page;
+	} else {
+		ret.s = dummy_load_page;
+		ret.o = dummy_load_page;
+	}
+	if (!kmsan_ready || IN_RUNTIME())
+		return ret;
+	BUG_ON(!metadata_is_contiguous(address, size, META_SHADOW));
+
+	if (!IS_ALIGNED(addr64, ORIGIN_SIZE)) {
+		pad = addr64 % ORIGIN_SIZE;
+		o_addr64 -= pad;
+	}
+
+	if (kmsan_internal_is_vmalloc_addr(address) ||
+	    kmsan_internal_is_module_addr(address)) {
+		ret.s = (void *)vmalloc_meta(address, META_SHADOW);
+		ret.o = (void *)vmalloc_meta((void *)o_addr64, META_ORIGIN);
+		return ret;
+	}
+
+	if (!kmsan_virt_addr_valid(address)) {
+		page = vmalloc_to_page_or_null(address);
+		if (page)
+			goto next;
+		shadow = get_cea_meta_or_null(address, META_SHADOW);
+		if (shadow) {
+			ret.s = shadow;
+			ret.o = get_cea_meta_or_null((void *)o_addr64,
+						     META_ORIGIN);
+			return ret;
+		}
+	}
+	page = virt_to_page_or_null(address);
+	if (!page)
+		return ret;
+next:
+	if (is_ignored_page(page))
+		return ret;
+
+	if (!has_shadow_page(page) || !has_origin_page(page))
+		return ret;
+	offset = addr64 % PAGE_SIZE;
+	o_offset = o_addr64 % PAGE_SIZE;
+
+	if (offset + size - 1 > PAGE_SIZE) {
+		/*
+		 * The access overflows the current page and touches the
+		 * subsequent ones. Make sure the shadow/origin pages are also
+		 * consequent.
+		 */
+		BUG_ON(!metadata_is_contiguous(address, size, META_SHADOW));
+	}
+
+	ret.s = shadow_ptr_for(page) + offset;
+	ret.o = origin_ptr_for(page) + o_offset;
+	return ret;
+}
+
+/*
+ * Obtain the shadow or origin pointer for the given address, or NULL if there's
+ * none. The caller must check the return value for being non-NULL if needed.
+ * The return value of this function should not depend on whether we're in the
+ * runtime or not.
+ */
+void *kmsan_get_metadata(void *address, size_t size, bool is_origin)
+{
+	struct page *page;
+	void *ret;
+	u64 addr = (u64)address, pad, off;
+
+	if (is_origin && !IS_ALIGNED(addr, ORIGIN_SIZE)) {
+		pad = addr % ORIGIN_SIZE;
+		addr -= pad;
+		size += pad;
+	}
+	address = (void *)addr;
+	if (kmsan_internal_is_vmalloc_addr(address) ||
+	    kmsan_internal_is_module_addr(address)) {
+		return (void *)vmalloc_meta(address, is_origin);
+	}
+
+	if (!kmsan_virt_addr_valid(address)) {
+		page = vmalloc_to_page_or_null(address);
+		if (page)
+			goto next;
+		ret = get_cea_meta_or_null(address, is_origin);
+		if (ret)
+			return ret;
+	}
+	page = virt_to_page_or_null(address);
+	if (!page)
+		return NULL;
+next:
+	if (is_ignored_page(page))
+		return NULL;
+	if (!has_shadow_page(page) || !has_origin_page(page))
+		return NULL;
+	off = addr % PAGE_SIZE;
+
+	ret = (is_origin ? origin_ptr_for(page) : shadow_ptr_for(page)) + off;
+	return ret;
+}
+
+void __init kmsan_init_alloc_meta_for_range(void *start, void *end)
+{
+	u64 addr, size;
+	struct page *page;
+	void *shadow, *origin;
+	struct page *shadow_p, *origin_p;
+
+	start = (void *)ALIGN_DOWN((u64)start, PAGE_SIZE);
+	size = ALIGN((u64)end - (u64)start, PAGE_SIZE);
+	shadow = memblock_alloc(size, PAGE_SIZE);
+	origin = memblock_alloc(size, PAGE_SIZE);
+	for (addr = 0; addr < size; addr += PAGE_SIZE) {
+		page = virt_to_page_or_null((char *)start + addr);
+		shadow_p = virt_to_page_or_null((char *)shadow + addr);
+		set_no_shadow_origin_page(shadow_p);
+		shadow_page_for(page) = shadow_p;
+		origin_p = virt_to_page_or_null((char *)origin + addr);
+		set_no_shadow_origin_page(origin_p);
+		origin_page_for(page) = origin_p;
+	}
+}
+
+/* Called from mm/memory.c */
+void kmsan_copy_page_meta(struct page *dst, struct page *src)
+{
+	unsigned long irq_flags;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+	if (!has_shadow_page(src)) {
+		/* TODO(glider): are we leaking pages here? */
+		set_no_shadow_origin_page(dst);
+		return;
+	}
+	if (!has_shadow_page(dst))
+		return;
+	if (is_ignored_page(src)) {
+		ignore_page(dst);
+		return;
+	}
+
+	ENTER_RUNTIME(irq_flags);
+	__memcpy(shadow_ptr_for(dst), shadow_ptr_for(src),
+		PAGE_SIZE);
+	BUG_ON(!has_origin_page(src) || !has_origin_page(dst));
+	__memcpy(origin_ptr_for(dst), origin_ptr_for(src),
+		PAGE_SIZE);
+	LEAVE_RUNTIME(irq_flags);
+}
+EXPORT_SYMBOL(kmsan_copy_page_meta);
+
+/* Helper function to allocate page metadata. */
+static int kmsan_internal_alloc_meta_for_pages(struct page *page,
+					       unsigned int order,
+					       gfp_t flags, int node)
+{
+	struct page *shadow, *origin;
+	int pages = 1 << order;
+	int i;
+	bool initialized = (flags & __GFP_ZERO) || !kmsan_ready;
+	depot_stack_handle_t handle;
+
+	if (flags & __GFP_NO_KMSAN_SHADOW) {
+		for (i = 0; i < pages; i++)
+			set_no_shadow_origin_page(&page[i]);
+		return 0;
+	}
+
+	/* TODO(glider): must we override the flags? */
+	flags = GFP_ATOMIC;
+	if (initialized)
+		flags |= __GFP_ZERO;
+	shadow = alloc_pages_node(node, flags | __GFP_NO_KMSAN_SHADOW, order);
+	if (!shadow) {
+		for (i = 0; i < pages; i++) {
+			set_no_shadow_origin_page(&page[i]);
+			set_no_shadow_origin_page(&page[i]);
+		}
+		return -ENOMEM;
+	}
+	if (!initialized)
+		__memset(page_address(shadow), -1, PAGE_SIZE * pages);
+
+	origin = alloc_pages_node(node, flags | __GFP_NO_KMSAN_SHADOW, order);
+	/* Assume we've allocated the origin. */
+	if (!origin) {
+		__free_pages(shadow, order);
+		for (i = 0; i < pages; i++)
+			set_no_shadow_origin_page(&page[i]);
+		return -ENOMEM;
+	}
+
+	if (!initialized) {
+		handle = kmsan_save_stack_with_flags(flags, /*extra_bits*/0);
+		/*
+		 * Addresses are page-aligned, pages are contiguous, so it's ok
+		 * to just fill the origin pages with |handle|.
+		 */
+		for (i = 0; i < PAGE_SIZE * pages / sizeof(handle); i++) {
+			((depot_stack_handle_t *)page_address(origin))[i] =
+						handle;
+		}
+	}
+
+	for (i = 0; i < pages; i++) {
+		shadow_page_for(&page[i]) = &shadow[i];
+		set_no_shadow_origin_page(shadow_page_for(&page[i]));
+		origin_page_for(&page[i]) = &origin[i];
+		set_no_shadow_origin_page(origin_page_for(&page[i]));
+	}
+	return 0;
+}
+
+/* Called from mm/page_alloc.c */
+int kmsan_alloc_page(struct page *page, unsigned int order, gfp_t flags)
+{
+	unsigned long irq_flags;
+	int ret;
+
+	if (IN_RUNTIME())
+		return 0;
+	ENTER_RUNTIME(irq_flags);
+	ret = kmsan_internal_alloc_meta_for_pages(page, order, flags, -1);
+	LEAVE_RUNTIME(irq_flags);
+	return ret;
+}
+
+/* Called from mm/page_alloc.c */
+void kmsan_free_page(struct page *page, unsigned int order)
+{
+	struct page *shadow, *origin, *cur_page;
+	int pages = 1 << order;
+	int i;
+	unsigned long irq_flags;
+
+	if (!shadow_page_for(page)) {
+		for (i = 0; i < pages; i++) {
+			cur_page = &page[i];
+			BUG_ON(shadow_page_for(cur_page));
+		}
+		return;
+	}
+
+	if (!kmsan_ready) {
+		for (i = 0; i < pages; i++) {
+			cur_page = &page[i];
+			set_no_shadow_origin_page(cur_page);
+		}
+		return;
+	}
+
+	if (IN_RUNTIME()) {
+		/*
+		 * TODO(glider): looks legit. depot_save_stack() may call
+		 * free_pages().
+		 */
+		return;
+	}
+
+	ENTER_RUNTIME(irq_flags);
+	shadow = shadow_page_for(&page[0]);
+	origin = origin_page_for(&page[0]);
+
+	/* TODO(glider): this is racy. */
+	for (i = 0; i < pages; i++) {
+		BUG_ON(has_shadow_page(shadow_page_for(&page[i])));
+		BUG_ON(has_shadow_page(origin_page_for(&page[i])));
+		set_no_shadow_origin_page(&page[i]);
+	}
+	BUG_ON(has_shadow_page(shadow));
+	__free_pages(shadow, order);
+
+	BUG_ON(has_shadow_page(origin));
+	__free_pages(origin, order);
+	LEAVE_RUNTIME(irq_flags);
+}
+EXPORT_SYMBOL(kmsan_free_page);
+
+/* Called from mm/page_alloc.c */
+void kmsan_split_page(struct page *page, unsigned int order)
+{
+	struct page *shadow, *origin;
+	unsigned long irq_flags;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+
+	ENTER_RUNTIME(irq_flags);
+	if (!has_shadow_page(&page[0])) {
+		BUG_ON(has_origin_page(&page[0]));
+		LEAVE_RUNTIME(irq_flags);
+		return;
+	}
+	shadow = shadow_page_for(&page[0]);
+	split_page(shadow, order);
+
+	origin = origin_page_for(&page[0]);
+	split_page(origin, order);
+	LEAVE_RUNTIME(irq_flags);
+}
+EXPORT_SYMBOL(kmsan_split_page);
+
+/* Called from include/linux/highmem.h */
+void kmsan_clear_page(void *page_addr)
+{
+	struct page *page;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+	BUG_ON(!IS_ALIGNED((u64)page_addr, PAGE_SIZE));
+	page = vmalloc_to_page_or_null(page_addr);
+	if (!page)
+		page = virt_to_page_or_null(page_addr);
+	if (!page || !has_shadow_page(page))
+		return;
+	__memset(shadow_ptr_for(page), 0, PAGE_SIZE);
+	BUG_ON(!has_origin_page(page));
+	__memset(origin_ptr_for(page), 0, PAGE_SIZE);
+}
+EXPORT_SYMBOL(kmsan_clear_page);
+
+/* Called from mm/vmalloc.c */
+void kmsan_vmap_page_range_noflush(unsigned long start, unsigned long end,
+				   pgprot_t prot, struct page **pages)
+{
+	int nr, i, mapped;
+	struct page **s_pages, **o_pages;
+	unsigned long shadow_start, shadow_end, origin_start, origin_end;
+
+	if (!kmsan_ready || IN_RUNTIME())
+		return;
+	shadow_start = vmalloc_meta((void *)start, META_SHADOW);
+	if (!shadow_start)
+		return;
+
+	BUG_ON(start >= end);
+	nr = (end - start) / PAGE_SIZE;
+	s_pages = kcalloc(nr, sizeof(struct page *), GFP_KERNEL);
+	o_pages = kcalloc(nr, sizeof(struct page *), GFP_KERNEL);
+	if (!s_pages || !o_pages)
+		goto ret;
+	for (i = 0; i < nr; i++) {
+		s_pages[i] = shadow_page_for(pages[i]);
+		o_pages[i] = origin_page_for(pages[i]);
+	}
+	prot = __pgprot(pgprot_val(prot) | _PAGE_NX);
+	prot = PAGE_KERNEL;
+
+	shadow_end = vmalloc_meta((void *)end, META_SHADOW);
+	origin_start = vmalloc_meta((void *)start, META_ORIGIN);
+	origin_end = vmalloc_meta((void *)end, META_ORIGIN);
+	mapped = __vmap_page_range_noflush(shadow_start, shadow_end,
+					   prot, s_pages);
+	BUG_ON(mapped != nr);
+	flush_tlb_kernel_range(shadow_start, shadow_end);
+	mapped = __vmap_page_range_noflush(origin_start, origin_end,
+					   prot, o_pages);
+	BUG_ON(mapped != nr);
+	flush_tlb_kernel_range(origin_start, origin_end);
+ret:
+	kfree(s_pages);
+	kfree(o_pages);
+}
+
+void kmsan_ignore_page(struct page *page, int order)
+{
+	int pages = 1 << order;
+	int i;
+	struct page *cp;
+
+	for (i = 0; i < pages; i++) {
+		cp = &page[i];
+		ignore_page(cp);
+	}
+}
diff --git a/mm/kmsan/kmsan_shadow.h b/mm/kmsan/kmsan_shadow.h
new file mode 100644
index 000000000000..eaa7f771b6a5
--- /dev/null
+++ b/mm/kmsan/kmsan_shadow.h
@@ -0,0 +1,30 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * KMSAN shadow API.
+ *
+ * This should be agnostic to shadow implementation details.
+ *
+ * Copyright (C) 2017-2019 Google LLC
+ * Author: Alexander Potapenko <glider@google.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#ifndef __MM_KMSAN_KMSAN_SHADOW_H
+#define __MM_KMSAN_KMSAN_SHADOW_H
+
+#include <asm/cpu_entry_area.h>  /* for CPU_ENTRY_AREA_MAP_SIZE */
+
+struct shadow_origin_ptr {
+	void *s, *o;
+};
+
+struct shadow_origin_ptr kmsan_get_shadow_origin_ptr(void *addr, u64 size,
+						     bool store);
+void *kmsan_get_metadata(void *addr, size_t size, bool is_origin);
+void __init kmsan_init_alloc_meta_for_range(void *start, void *end);
+
+#endif  /* __MM_KMSAN_KMSAN_SHADOW_H */
diff --git a/scripts/Makefile.kmsan b/scripts/Makefile.kmsan
new file mode 100644
index 000000000000..8b3844b66b22
--- /dev/null
+++ b/scripts/Makefile.kmsan
@@ -0,0 +1,12 @@
+ifdef CONFIG_KMSAN
+
+CFLAGS_KMSAN := -fsanitize=kernel-memory
+
+ifeq ($(call cc-option, $(CFLAGS_KMSAN) -Werror),)
+   ifneq ($(CONFIG_COMPILE_TEST),y)
+        $(warning Cannot use CONFIG_KMSAN: \
+            -fsanitize=kernel-memory is not supported by compiler)
+   endif
+endif
+
+endif
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 12/26] kmsan: x86: sync metadata pages on page fault

[glider]

KMSAN assumes shadow and origin pages for every allocated page are
accessible. For pages in vmalloc region those metadata pages reside in
[VMALLOC_END, VMALLOC_META_END), therefore we must sync a bigger memory
region.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org

---

Change-Id: I0d54855489870ef1180b37fe2120b601da464bf7
---
 arch/x86/mm/fault.c | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index 9ceacd1156db..d582337ba45d 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -320,7 +320,17 @@ static void dump_pagetable(unsigned long address)
 
 void vmalloc_sync_all(void)
 {
+#ifdef CONFIG_KMSAN
+	/*
+	 * For KMSAN, make sure metadata pages for vmalloc area and modules are
+	 * also synced.
+	 */
+	sync_global_pgds(VMALLOC_START & PGDIR_MASK, VMALLOC_META_END);
+	sync_global_pgds(MODULES_SHADOW_START & PGDIR_MASK,
+		MODULES_ORIGIN_END);
+#else
 	sync_global_pgds(VMALLOC_START & PGDIR_MASK, VMALLOC_END);
+#endif
 }
 
 /*
@@ -337,7 +347,17 @@ static noinline int vmalloc_fault(unsigned long address)
 	pte_t *pte;
 
 	/* Make sure we are in vmalloc area: */
+#ifdef CONFIG_KMSAN
+	/*
+	 * For KMSAN, make sure metadata pages for vmalloc area and modules are
+	 * also synced.
+	 */
+	if (!(address >= VMALLOC_START && address < VMALLOC_META_END) &&
+		!(address >= MODULES_SHADOW_START &&
+		  address < MODULES_ORIGIN_END))
+#else
 	if (!(address >= VMALLOC_START && address < VMALLOC_END))
+#endif
 		return -1;
 
 	/*
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 13/26] kmsan: add tests for KMSAN

[glider]

The initial commit adds several tests that trigger KMSAN warnings in
simple cases.
To use, build the kernel with CONFIG_TEST_KMSAN and do
`insmod test_kmsan.ko`

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org
---

Change-Id: I287e86ae83a82b770f2baa46e5bbdce1dfa65195
---
 lib/Makefile     |   1 +
 lib/test_kmsan.c | 221 +++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 222 insertions(+)
 create mode 100644 lib/test_kmsan.c

diff --git a/lib/Makefile b/lib/Makefile
index c5892807e06f..cb44262c38ee 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -65,6 +65,7 @@ CFLAGS_test_kasan.o += $(call cc-disable-warning, vla)
 obj-$(CONFIG_TEST_UBSAN) += test_ubsan.o
 CFLAGS_test_ubsan.o += $(call cc-disable-warning, vla)
 UBSAN_SANITIZE_test_ubsan.o := y
+obj-$(CONFIG_TEST_KMSAN) += test_kmsan.o
 obj-$(CONFIG_TEST_KSTRTOX) += test-kstrtox.o
 obj-$(CONFIG_TEST_LIST_SORT) += test_list_sort.o
 obj-$(CONFIG_TEST_LKM) += test_module.o
diff --git a/lib/test_kmsan.c b/lib/test_kmsan.c
new file mode 100644
index 000000000000..97051444c6c4
--- /dev/null
+++ b/lib/test_kmsan.c
@@ -0,0 +1,221 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Module for testing KMSAN.
+ *
+ * Copyright (C) 2017-2019 Google LLC
+ * Author: Alexander Potapenko <glider@google.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+/*
+ * Tests below use noinline and volatile to work around compiler optimizations
+ * that may mask KMSAN bugs.
+ */
+#define pr_fmt(fmt) "kmsan test: %s : " fmt, __func__
+
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/printk.h>
+#include <linux/slab.h>
+#include <linux/kmsan-checks.h>
+
+#define CHECK(x)					\
+	do {						\
+		if (x)					\
+			pr_info(#x " is true\n");	\
+		else					\
+			pr_info(#x " is false\n");	\
+	} while (0)
+
+noinline void use_integer(int cond)
+{
+	CHECK(cond);
+}
+
+int signed_sum3(int a, int b, int c)
+{
+	return a + b + c;
+}
+
+noinline void uninit_kmalloc_test(void)
+{
+	int *ptr;
+
+	pr_info("-----------------------------\n");
+	pr_info("uninitialized kmalloc test (UMR report)\n");
+	ptr = kmalloc(sizeof(int), GFP_KERNEL);
+	pr_info("kmalloc returned %p\n", ptr);
+	CHECK(*ptr);
+}
+noinline void init_kmalloc_test(void)
+{
+	int *ptr;
+
+	pr_info("-----------------------------\n");
+	pr_info("initialized kmalloc test (no reports)\n");
+	ptr = kmalloc(sizeof(int), GFP_KERNEL);
+	memset(ptr, 0, sizeof(int));
+	pr_info("kmalloc returned %p\n", ptr);
+	CHECK(*ptr);
+}
+
+noinline void init_kzalloc_test(void)
+{
+	int *ptr;
+
+	pr_info("-----------------------------\n");
+	pr_info("initialized kzalloc test (no reports)\n");
+	ptr = kzalloc(sizeof(int), GFP_KERNEL);
+	pr_info("kzalloc returned %p\n", ptr);
+	CHECK(*ptr);
+}
+
+noinline void uninit_multiple_args_test(void)
+{
+	volatile int a;
+	volatile char b = 3, c;
+
+	pr_info("-----------------------------\n");
+	pr_info("uninitialized local passed to fn (UMR report)\n");
+	CHECK(signed_sum3(a, b, c));
+}
+
+noinline void uninit_stack_var_test(void)
+{
+	int cond;
+
+	pr_info("-----------------------------\n");
+	pr_info("uninitialized stack variable (UMR report)\n");
+	CHECK(cond);
+}
+
+noinline void init_stack_var_test(void)
+{
+	volatile int cond = 1;
+
+	pr_info("-----------------------------\n");
+	pr_info("initialized stack variable (no reports)\n");
+	CHECK(cond);
+}
+
+noinline void two_param_fn_2(int arg1, int arg2)
+{
+	CHECK(arg1);
+	CHECK(arg2);
+}
+
+noinline void one_param_fn(int arg)
+{
+	two_param_fn_2(arg, arg);
+	CHECK(arg);
+}
+
+noinline void two_param_fn(int arg1, int arg2)
+{
+	int init = 0;
+
+	one_param_fn(init);
+	CHECK(arg1);
+	CHECK(arg2);
+}
+
+void params_test(void)
+{
+	int uninit, init = 1;
+
+	two_param_fn(uninit, init);
+}
+
+noinline void do_uninit_local_array(char *array, int start, int stop)
+{
+	int i;
+	volatile char uninit;
+
+	for (i = start; i < stop; i++)
+		array[i] = uninit;
+}
+
+noinline void uninit_kmsan_check_memory_test(void)
+{
+	volatile char local_array[8];
+
+	pr_info("-----------------------------\n");
+	pr_info("kmsan_check_memory() called on uninit local (UMR report)\n");
+	do_uninit_local_array((char *)local_array, 5, 7);
+
+	kmsan_check_memory((char *)local_array, 8);
+}
+
+noinline void init_kmsan_vmap_vunmap_test(void)
+{
+	const int npages = 2;
+	struct page *pages[npages];
+	void *vbuf;
+	int i;
+
+	pr_info("-----------------------------\n");
+	pr_info("pages initialized via vmap (no reports)\n");
+
+	for (i = 0; i < npages; i++)
+		pages[i] = alloc_page(GFP_KERNEL);
+	vbuf = vmap(pages, npages, VM_MAP, PAGE_KERNEL);
+	memset(vbuf, 0xfe, npages * PAGE_SIZE);
+	for (i = 0; i < npages; i++)
+		kmsan_check_memory(page_address(pages[i]), PAGE_SIZE);
+
+	if (vbuf)
+		vunmap(vbuf);
+	for (i = 0; i < npages; i++)
+		if (pages[i])
+			__free_page(pages[i]);
+}
+
+noinline void init_vmalloc(void)
+{
+	char *buf;
+	int npages = 8, i;
+
+	pr_info("-----------------------------\n");
+	pr_info("pages initialized via vmap (no reports)\n");
+	buf = vmalloc(PAGE_SIZE * npages);
+	buf[0] = 1;
+	memset(buf, 0xfe, PAGE_SIZE * npages);
+	CHECK(buf[0]);
+	for (i = 0; i < npages; i++)
+		kmsan_check_memory(&buf[PAGE_SIZE * i], PAGE_SIZE);
+	vfree(buf);
+}
+
+noinline void uaf_test(void)
+{
+	volatile int *var;
+
+	pr_info("-----------------------------\n");
+	pr_info("use-after-free in kmalloc-ed buffer (UMR report)\n");
+	var = kmalloc(80, GFP_KERNEL);
+	var[3] = 0xfeedface;
+	kfree((int *)var);
+	CHECK(var[3]);
+}
+
+static noinline int __init kmsan_tests_init(void)
+{
+	uninit_kmalloc_test();
+	init_kmalloc_test();
+	init_kzalloc_test();
+	uninit_multiple_args_test();
+	uninit_stack_var_test();
+	init_stack_var_test();
+	uninit_kmsan_check_memory_test();
+	init_kmsan_vmap_vunmap_test();
+	init_vmalloc();
+	uaf_test();
+	return -EAGAIN;
+}
+
+module_init(kmsan_tests_init);
+MODULE_LICENSE("GPL");
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 14/26] kmsan: make READ_ONCE_TASK_STACK() return initialized values

[glider]

To avoid false positives, assume that reading from the task stack
always produces initialized values.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org
---

Change-Id: Ie73e5a41fdc8195699928e65f5cbe0d3d3c9e2fa
---
 arch/x86/include/asm/unwind.h | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/arch/x86/include/asm/unwind.h b/arch/x86/include/asm/unwind.h
index 499578f7e6d7..f60c2bd1ddf2 100644
--- a/arch/x86/include/asm/unwind.h
+++ b/arch/x86/include/asm/unwind.h
@@ -100,9 +100,10 @@ void unwind_module_init(struct module *mod, void *orc_ip, size_t orc_ip_size,
 #endif
 
 /*
- * This disables KASAN checking when reading a value from another task's stack,
- * since the other task could be running on another CPU and could have poisoned
- * the stack in the meantime.
+ * This disables KASAN/KMSAN checking when reading a value from another task's
+ * stack, since the other task could be running on another CPU and could have
+ * poisoned the stack in the meantime. Frame pointers are uninitialized by
+ * default, so for KMSAN we mark the return value initialized unconditionally.
  */
 #define READ_ONCE_TASK_STACK(task, x)			\
 ({							\
@@ -111,7 +112,7 @@ void unwind_module_init(struct module *mod, void *orc_ip, size_t orc_ip_size,
 		val = READ_ONCE(x);			\
 	else						\
 		val = READ_ONCE_NOCHECK(x);		\
-	val;						\
+	KMSAN_INIT_VALUE(val);				\
 })
 
 static inline bool task_on_another_cpu(struct task_struct *task)
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 15/26] kmsan: Kconfig changes to disable options incompatible with KMSAN

[glider]

Certain config options are currently incompatible with KMSAN, disable
them.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org

---

Change-Id: I4521f5c764cfaeba2fae79c954925c302c5e3bc0
---
 arch/Kconfig                        |  2 +-
 arch/x86/Kconfig                    |  1 +
 arch/x86/Kconfig.debug              |  3 +++
 crypto/Kconfig                      | 26 ++++++++++++++++++++++++++
 drivers/gpu/drm/amd/display/Kconfig |  2 +-
 lib/Kconfig.debug                   |  3 +++
 6 files changed, 35 insertions(+), 2 deletions(-)

diff --git a/arch/Kconfig b/arch/Kconfig
index 5f8a5d84dbbe..d5ca838ae8ad 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -843,7 +843,7 @@ config HAVE_ARCH_VMAP_STACK
 config VMAP_STACK
 	default y
 	bool "Use a virtually-mapped stack"
-	depends on HAVE_ARCH_VMAP_STACK && !KASAN
+	depends on HAVE_ARCH_VMAP_STACK && !KASAN && !KMSAN
 	---help---
 	  Enable this if you want the use virtually-mapped kernel stacks
 	  with guard pages.  This causes kernel stack overflows to be
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 3f83a5c53808..6d2b0fb3a871 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -2143,6 +2143,7 @@ config RELOCATABLE
 config RANDOMIZE_BASE
 	bool "Randomize the address of the kernel image (KASLR)"
 	depends on RELOCATABLE
+	depends on !KMSAN
 	default y
 	---help---
 	  In support of Kernel Address Space Layout Randomization (KASLR),
diff --git a/arch/x86/Kconfig.debug b/arch/x86/Kconfig.debug
index bf9cd83de777..db3cd6147829 100644
--- a/arch/x86/Kconfig.debug
+++ b/arch/x86/Kconfig.debug
@@ -292,6 +292,9 @@ choice
 config UNWINDER_ORC
 	bool "ORC unwinder"
 	depends on X86_64
+	# KMSAN doesn't support UNWINDER_ORC yet,
+	# see https://github.com/google/kmsan/issues/48.
+	depends on !KMSAN
 	select STACK_VALIDATION
 	---help---
 	  This option enables the ORC (Oops Rewind Capability) unwinder for
diff --git a/crypto/Kconfig b/crypto/Kconfig
index 9e524044d312..37a22711012c 100644
--- a/crypto/Kconfig
+++ b/crypto/Kconfig
@@ -309,11 +309,13 @@ config CRYPTO_AEGIS128
 config CRYPTO_AEGIS128_SIMD
 	bool "Support SIMD acceleration for AEGIS-128"
 	depends on CRYPTO_AEGIS128 && ((ARM || ARM64) && KERNEL_MODE_NEON)
+	depends on !KMSAN
 	default y
 
 config CRYPTO_AEGIS128_AESNI_SSE2
 	tristate "AEGIS-128 AEAD algorithm (x86_64 AESNI+SSE2 implementation)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_AEAD
 	select CRYPTO_SIMD
 	help
@@ -571,6 +573,7 @@ config CRYPTO_CRC32C
 config CRYPTO_CRC32C_INTEL
 	tristate "CRC32c INTEL hardware acceleration"
 	depends on X86
+	depends on !KMSAN
 	select CRYPTO_HASH
 	help
 	  In Intel processor with SSE4.2 supported, the processor will
@@ -611,6 +614,7 @@ config CRYPTO_CRC32
 config CRYPTO_CRC32_PCLMUL
 	tristate "CRC32 PCLMULQDQ hardware acceleration"
 	depends on X86
+	depends on !KMSAN
 	select CRYPTO_HASH
 	select CRC32
 	help
@@ -649,6 +653,7 @@ config CRYPTO_CRCT10DIF
 config CRYPTO_CRCT10DIF_PCLMUL
 	tristate "CRCT10DIF PCLMULQDQ hardware acceleration"
 	depends on X86 && 64BIT && CRC_T10DIF
+	depends on !KMSAN
 	select CRYPTO_HASH
 	help
 	  For x86_64 processors with SSE4.2 and PCLMULQDQ supported,
@@ -695,6 +700,7 @@ config CRYPTO_POLY1305
 config CRYPTO_POLY1305_X86_64
 	tristate "Poly1305 authenticator algorithm (x86_64/SSE2/AVX2)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_POLY1305
 	help
 	  Poly1305 authenticator algorithm, RFC7539.
@@ -814,6 +820,7 @@ config CRYPTO_SHA1
 config CRYPTO_SHA1_SSSE3
 	tristate "SHA1 digest algorithm (SSSE3/AVX/AVX2/SHA-NI)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_SHA1
 	select CRYPTO_HASH
 	help
@@ -825,6 +832,7 @@ config CRYPTO_SHA1_SSSE3
 config CRYPTO_SHA256_SSSE3
 	tristate "SHA256 digest algorithm (SSSE3/AVX/AVX2/SHA-NI)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_SHA256
 	select CRYPTO_HASH
 	help
@@ -837,6 +845,7 @@ config CRYPTO_SHA256_SSSE3
 config CRYPTO_SHA512_SSSE3
 	tristate "SHA512 digest algorithm (SSSE3/AVX/AVX2)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_SHA512
 	select CRYPTO_HASH
 	help
@@ -1011,6 +1020,7 @@ config CRYPTO_WP512
 config CRYPTO_GHASH_CLMUL_NI_INTEL
 	tristate "GHASH hash function (CLMUL-NI accelerated)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_CRYPTD
 	help
 	  This is the x86_64 CLMUL-NI accelerated implementation of
@@ -1064,6 +1074,7 @@ config CRYPTO_AES_TI
 config CRYPTO_AES_NI_INTEL
 	tristate "AES cipher algorithms (AES-NI)"
 	depends on X86
+	depends on !KMSAN
 	select CRYPTO_AEAD
 	select CRYPTO_LIB_AES
 	select CRYPTO_ALGAPI
@@ -1190,6 +1201,7 @@ config CRYPTO_BLOWFISH_COMMON
 config CRYPTO_BLOWFISH_X86_64
 	tristate "Blowfish cipher algorithm (x86_64)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_BLKCIPHER
 	select CRYPTO_BLOWFISH_COMMON
 	help
@@ -1221,6 +1233,7 @@ config CRYPTO_CAMELLIA_X86_64
 	tristate "Camellia cipher algorithm (x86_64)"
 	depends on X86 && 64BIT
 	depends on CRYPTO
+	depends on !KMSAN
 	select CRYPTO_BLKCIPHER
 	select CRYPTO_GLUE_HELPER_X86
 	help
@@ -1238,6 +1251,7 @@ config CRYPTO_CAMELLIA_AESNI_AVX_X86_64
 	tristate "Camellia cipher algorithm (x86_64/AES-NI/AVX)"
 	depends on X86 && 64BIT
 	depends on CRYPTO
+	depends on !KMSAN
 	select CRYPTO_BLKCIPHER
 	select CRYPTO_CAMELLIA_X86_64
 	select CRYPTO_GLUE_HELPER_X86
@@ -1258,6 +1272,7 @@ config CRYPTO_CAMELLIA_AESNI_AVX2_X86_64
 	tristate "Camellia cipher algorithm (x86_64/AES-NI/AVX2)"
 	depends on X86 && 64BIT
 	depends on CRYPTO
+	depends on !KMSAN
 	select CRYPTO_CAMELLIA_AESNI_AVX_X86_64
 	help
 	  Camellia cipher algorithm module (x86_64/AES-NI/AVX2).
@@ -1303,6 +1318,7 @@ config CRYPTO_CAST5
 config CRYPTO_CAST5_AVX_X86_64
 	tristate "CAST5 (CAST-128) cipher algorithm (x86_64/AVX)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_BLKCIPHER
 	select CRYPTO_CAST5
 	select CRYPTO_CAST_COMMON
@@ -1325,6 +1341,7 @@ config CRYPTO_CAST6
 config CRYPTO_CAST6_AVX_X86_64
 	tristate "CAST6 (CAST-256) cipher algorithm (x86_64/AVX)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_BLKCIPHER
 	select CRYPTO_CAST6
 	select CRYPTO_CAST_COMMON
@@ -1360,6 +1377,7 @@ config CRYPTO_DES_SPARC64
 config CRYPTO_DES3_EDE_X86_64
 	tristate "Triple DES EDE cipher algorithm (x86-64)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_BLKCIPHER
 	select CRYPTO_LIB_DES
 	help
@@ -1426,6 +1444,7 @@ config CRYPTO_CHACHA20
 config CRYPTO_CHACHA20_X86_64
 	tristate "ChaCha stream cipher algorithms (x86_64/SSSE3/AVX2/AVX-512VL)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_BLKCIPHER
 	select CRYPTO_CHACHA20
 	help
@@ -1462,6 +1481,7 @@ config CRYPTO_SERPENT
 config CRYPTO_SERPENT_SSE2_X86_64
 	tristate "Serpent cipher algorithm (x86_64/SSE2)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_BLKCIPHER
 	select CRYPTO_GLUE_HELPER_X86
 	select CRYPTO_SERPENT
@@ -1481,6 +1501,7 @@ config CRYPTO_SERPENT_SSE2_X86_64
 config CRYPTO_SERPENT_SSE2_586
 	tristate "Serpent cipher algorithm (i586/SSE2)"
 	depends on X86 && !64BIT
+	depends on !KMSAN
 	select CRYPTO_BLKCIPHER
 	select CRYPTO_GLUE_HELPER_X86
 	select CRYPTO_SERPENT
@@ -1500,6 +1521,7 @@ config CRYPTO_SERPENT_SSE2_586
 config CRYPTO_SERPENT_AVX_X86_64
 	tristate "Serpent cipher algorithm (x86_64/AVX)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_BLKCIPHER
 	select CRYPTO_GLUE_HELPER_X86
 	select CRYPTO_SERPENT
@@ -1520,6 +1542,7 @@ config CRYPTO_SERPENT_AVX_X86_64
 config CRYPTO_SERPENT_AVX2_X86_64
 	tristate "Serpent cipher algorithm (x86_64/AVX2)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_SERPENT_AVX_X86_64
 	help
 	  Serpent cipher algorithm, by Anderson, Biham & Knudsen.
@@ -1615,6 +1638,7 @@ config CRYPTO_TWOFISH_586
 config CRYPTO_TWOFISH_X86_64
 	tristate "Twofish cipher algorithm (x86_64)"
 	depends on (X86 || UML_X86) && 64BIT
+	depends on !KMSAN
 	select CRYPTO_ALGAPI
 	select CRYPTO_TWOFISH_COMMON
 	help
@@ -1631,6 +1655,7 @@ config CRYPTO_TWOFISH_X86_64
 config CRYPTO_TWOFISH_X86_64_3WAY
 	tristate "Twofish cipher algorithm (x86_64, 3-way parallel)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_BLKCIPHER
 	select CRYPTO_TWOFISH_COMMON
 	select CRYPTO_TWOFISH_X86_64
@@ -1652,6 +1677,7 @@ config CRYPTO_TWOFISH_X86_64_3WAY
 config CRYPTO_TWOFISH_AVX_X86_64
 	tristate "Twofish cipher algorithm (x86_64/AVX)"
 	depends on X86 && 64BIT
+	depends on !KMSAN
 	select CRYPTO_BLKCIPHER
 	select CRYPTO_GLUE_HELPER_X86
 	select CRYPTO_SIMD
diff --git a/drivers/gpu/drm/amd/display/Kconfig b/drivers/gpu/drm/amd/display/Kconfig
index 71991a28a775..99585eaf4dd0 100644
--- a/drivers/gpu/drm/amd/display/Kconfig
+++ b/drivers/gpu/drm/amd/display/Kconfig
@@ -6,7 +6,7 @@ config DRM_AMD_DC
 	bool "AMD DC - Enable new display engine"
 	default y
 	select SND_HDA_COMPONENT if SND_HDA_CORE
-	select DRM_AMD_DC_DCN1_0 if X86 && !(KCOV_INSTRUMENT_ALL && KCOV_ENABLE_COMPARISONS)
+	select DRM_AMD_DC_DCN1_0 if X86 && !(KCOV_INSTRUMENT_ALL && KCOV_ENABLE_COMPARISONS) && !KMSAN
 	help
 	  Choose this option if you want to use the new display engine
 	  support for AMDGPU. This adds required support for Vega and
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 75c36318943d..a3f6f5d68593 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1068,6 +1068,9 @@ menu "Lock Debugging (spinlocks, mutexes, etc...)"
 config LOCK_DEBUGGING_SUPPORT
 	bool
 	depends on TRACE_IRQFLAGS_SUPPORT && STACKTRACE_SUPPORT && LOCKDEP_SUPPORT
+	# KMSAN is incompatible with lockdep,
+	# see https://github.com/google/kmsan/issues/57.
+	depends on !KMSAN
 	default y
 
 config PROVE_LOCKING
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 16/26] kmsan: Changing existing files to enable KMSAN builds

[glider]

Do the following to make KMSAN usable:
 - add KMSAN declarations to struct page and struct task_struct;
 - add Makefile.kmsan to top-level Makefile;
 - call KMSAN initialization from init/main.c
 - add asm hooks to arch/x86/entry/entry_64.S;
 - increase task stack size under KMSAN;
 - disable page merging in block/blk.h;
 - disable CONFIG_KMSAN in arch/x86/boot/compressed/misc.h to use
   default string functions instead of KMSAN ones.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org

---

Change-Id: I37e0b7f2d2f2b0aeac5753ff9d6b411485fc374e
---
 Makefile                             |  3 ++-
 arch/x86/boot/compressed/misc.h      |  1 +
 arch/x86/entry/entry_64.S            | 18 ++++++++++++++++++
 arch/x86/include/asm/page_64_types.h | 12 +++++++-----
 block/blk.h                          |  7 +++++++
 include/linux/mm_types.h             |  9 +++++++++
 include/linux/sched.h                |  5 +++++
 init/main.c                          |  3 +++
 lib/stackdepot.c                     |  7 ++++---
 mm/Makefile                          |  1 +
 scripts/Makefile.lib                 |  6 ++++++
 11 files changed, 63 insertions(+), 9 deletions(-)

diff --git a/Makefile b/Makefile
index ffd7a912fc46..5041e908f051 100644
--- a/Makefile
+++ b/Makefile
@@ -478,7 +478,7 @@ export KBUILD_HOSTCXXFLAGS KBUILD_HOSTLDFLAGS KBUILD_HOSTLDLIBS LDFLAGS_MODULE
 
 export KBUILD_CPPFLAGS NOSTDINC_FLAGS LINUXINCLUDE OBJCOPYFLAGS KBUILD_LDFLAGS
 export KBUILD_CFLAGS CFLAGS_KERNEL CFLAGS_MODULE
-export CFLAGS_KASAN CFLAGS_KASAN_NOSANITIZE CFLAGS_UBSAN
+export CFLAGS_KASAN CFLAGS_KASAN_NOSANITIZE CFLAGS_UBSAN CFLAGS_KMSAN
 export KBUILD_AFLAGS AFLAGS_KERNEL AFLAGS_MODULE
 export KBUILD_AFLAGS_MODULE KBUILD_CFLAGS_MODULE KBUILD_LDFLAGS_MODULE
 export KBUILD_AFLAGS_KERNEL KBUILD_CFLAGS_KERNEL
@@ -898,6 +898,7 @@ KBUILD_CFLAGS += $(call cc-option,-fcf-protection=none)
 endif
 
 include scripts/Makefile.kasan
+include scripts/Makefile.kmsan
 include scripts/Makefile.extrawarn
 include scripts/Makefile.ubsan
 
diff --git a/arch/x86/boot/compressed/misc.h b/arch/x86/boot/compressed/misc.h
index c8181392f70d..dd4bd8c5d97a 100644
--- a/arch/x86/boot/compressed/misc.h
+++ b/arch/x86/boot/compressed/misc.h
@@ -12,6 +12,7 @@
 #undef CONFIG_PARAVIRT_XXL
 #undef CONFIG_PARAVIRT_SPINLOCKS
 #undef CONFIG_KASAN
+#undef CONFIG_KMSAN
 
 /* cpu_feature_enabled() cannot be used this early */
 #define USE_EARLY_PGTABLE_L5
diff --git a/arch/x86/entry/entry_64.S b/arch/x86/entry/entry_64.S
index b7c3ea4cb19d..1218672bad85 100644
--- a/arch/x86/entry/entry_64.S
+++ b/arch/x86/entry/entry_64.S
@@ -35,6 +35,7 @@
 #include <asm/asm.h>
 #include <asm/smap.h>
 #include <asm/pgtable_types.h>
+#include <asm/kmsan.h>
 #include <asm/export.h>
 #include <asm/frame.h>
 #include <asm/nospec-branch.h>
@@ -168,12 +169,14 @@ GLOBAL(entry_SYSCALL_64_after_hwframe)
 	PUSH_AND_CLEAR_REGS rax=$-ENOSYS
 
 	TRACE_IRQS_OFF
+	KMSAN_SYSCALL_ENTER
 
 	/* IRQs are off. */
 	movq	%rax, %rdi
 	movq	%rsp, %rsi
 	call	do_syscall_64		/* returns with IRQs disabled */
 
+	KMSAN_SYSCALL_EXIT
 	TRACE_IRQS_IRETQ		/* we're about to change IF */
 
 	/*
@@ -575,6 +578,7 @@ ENTRY(interrupt_entry)
 
 1:
 	ENTER_IRQ_STACK old_rsp=%rdi save_ret=1
+	KMSAN_INTERRUPT_ENTER
 	/* We entered an interrupt context - irqs are off: */
 	TRACE_IRQS_OFF
 
@@ -604,12 +608,14 @@ common_interrupt:
 	addq	$-0x80, (%rsp)			/* Adjust vector to [-256, -1] range */
 	call	interrupt_entry
 	UNWIND_HINT_REGS indirect=1
+	KMSAN_UNPOISON_PT_REGS
 	call	do_IRQ	/* rdi points to pt_regs */
 	/* 0(%rsp): old RSP */
 ret_from_intr:
 	DISABLE_INTERRUPTS(CLBR_ANY)
 	TRACE_IRQS_OFF
 
+	KMSAN_INTERRUPT_EXIT
 	LEAVE_IRQ_STACK
 
 	testb	$3, CS(%rsp)
@@ -802,6 +808,7 @@ ENTRY(\sym)
 .Lcommon_\sym:
 	call	interrupt_entry
 	UNWIND_HINT_REGS indirect=1
+	KMSAN_UNPOISON_PT_REGS
 	call	\do_sym	/* rdi points to pt_regs */
 	jmp	ret_from_intr
 END(\sym)
@@ -909,15 +916,18 @@ apicinterrupt IRQ_WORK_VECTOR			irq_work_interrupt		smp_irq_work_interrupt
 
 	.if \shift_ist != -1
 	subq	$\ist_offset, CPU_TSS_IST(\shift_ist)
+	KMSAN_IST_ENTER(\shift_ist)
 	.endif
 
 	.if \read_cr2
 	movq	%r12, %rdx			/* Move CR2 into 3rd argument */
 	.endif
 
+	KMSAN_UNPOISON_PT_REGS
 	call	\do_sym
 
 	.if \shift_ist != -1
+	KMSAN_IST_EXIT(\shift_ist)
 	addq	$\ist_offset, CPU_TSS_IST(\shift_ist)
 	.endif
 
@@ -1079,7 +1089,9 @@ ENTRY(do_softirq_own_stack)
 	pushq	%rbp
 	mov	%rsp, %rbp
 	ENTER_IRQ_STACK regs=0 old_rsp=%r11
+	KMSAN_SOFTIRQ_ENTER
 	call	__do_softirq
+	KMSAN_SOFTIRQ_EXIT
 	LEAVE_IRQ_STACK regs=0
 	leaveq
 	ret
@@ -1466,9 +1478,12 @@ ENTRY(nmi)
 	 * done with the NMI stack.
 	 */
 
+	KMSAN_NMI_ENTER
 	movq	%rsp, %rdi
 	movq	$-1, %rsi
+	KMSAN_UNPOISON_PT_REGS
 	call	do_nmi
+	KMSAN_NMI_EXIT
 
 	/*
 	 * Return back to user mode.  We must *not* do the normal exit
@@ -1678,10 +1693,13 @@ end_repeat_nmi:
 	call	paranoid_entry
 	UNWIND_HINT_REGS
 
+	KMSAN_NMI_ENTER
 	/* paranoidentry do_nmi, 0; without TRACE_IRQS_OFF */
 	movq	%rsp, %rdi
 	movq	$-1, %rsi
+	KMSAN_UNPOISON_PT_REGS
 	call	do_nmi
+	KMSAN_NMI_EXIT
 
 	/* Always restore stashed CR3 value (see paranoid_entry) */
 	RESTORE_CR3 scratch_reg=%r15 save_reg=%r14
diff --git a/arch/x86/include/asm/page_64_types.h b/arch/x86/include/asm/page_64_types.h
index 288b065955b7..ea9fbf09f43b 100644
--- a/arch/x86/include/asm/page_64_types.h
+++ b/arch/x86/include/asm/page_64_types.h
@@ -7,18 +7,20 @@
 #endif
 
 #ifdef CONFIG_KASAN
-#define KASAN_STACK_ORDER 1
+#define EXTRA_STACK_ORDER 1
+#elif defined(CONFIG_KMSAN)
+#define EXTRA_STACK_ORDER 2
 #else
-#define KASAN_STACK_ORDER 0
+#define EXTRA_STACK_ORDER 0
 #endif
 
-#define THREAD_SIZE_ORDER	(2 + KASAN_STACK_ORDER)
+#define THREAD_SIZE_ORDER	(2 + EXTRA_STACK_ORDER)
 #define THREAD_SIZE  (PAGE_SIZE << THREAD_SIZE_ORDER)
 
-#define EXCEPTION_STACK_ORDER (0 + KASAN_STACK_ORDER)
+#define EXCEPTION_STACK_ORDER (0 + EXTRA_STACK_ORDER)
 #define EXCEPTION_STKSZ (PAGE_SIZE << EXCEPTION_STACK_ORDER)
 
-#define IRQ_STACK_ORDER (2 + KASAN_STACK_ORDER)
+#define IRQ_STACK_ORDER (2 + EXTRA_STACK_ORDER)
 #define IRQ_STACK_SIZE (PAGE_SIZE << IRQ_STACK_ORDER)
 
 /*
diff --git a/block/blk.h b/block/blk.h
index 47fba9362e60..9ee271a22423 100644
--- a/block/blk.h
+++ b/block/blk.h
@@ -78,6 +78,13 @@ static inline bool biovec_phys_mergeable(struct request_queue *q,
 	phys_addr_t addr1 = page_to_phys(vec1->bv_page) + vec1->bv_offset;
 	phys_addr_t addr2 = page_to_phys(vec2->bv_page) + vec2->bv_offset;
 
+#ifdef CONFIG_KMSAN
+	/*
+	 * Merging consequent physical pages may not work correctly under KMSAN
+	 * if their metadata pages aren't consequent. Just disable merging.
+	 */
+	return false;
+#endif
 	if (addr1 + vec1->bv_len != addr2)
 		return false;
 	if (xen_domain() && !xen_biovec_phys_mergeable(vec1, vec2->bv_page))
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index 2222fa795284..c87c5416a802 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -216,6 +216,15 @@ struct page {
 					   not kmapped, ie. highmem) */
 #endif /* WANT_PAGE_VIRTUAL */
 
+#ifdef CONFIG_KMSAN
+	/*
+	 * Bits in struct page are scarce, so the LSB in *shadow is used to
+	 * indicate whether the page should be ignored by KMSAN or not.
+	 */
+	struct page *shadow;
+	struct page *origin;
+#endif
+
 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
 	int _last_cpupid;
 #endif
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 7417e0207b48..e0f6e95c22ed 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -15,6 +15,7 @@
 #include <linux/sem.h>
 #include <linux/shm.h>
 #include <linux/kcov.h>
+#include <linux/kmsan.h>
 #include <linux/mutex.h>
 #include <linux/plist.h>
 #include <linux/hrtimer.h>
@@ -1172,6 +1173,10 @@ struct task_struct {
 	unsigned int			kasan_depth;
 #endif
 
+#ifdef CONFIG_KMSAN
+	struct kmsan_task_state		kmsan;
+#endif
+
 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
 	/* Index of current stored address in ret_stack: */
 	int				curr_ret_stack;
diff --git a/init/main.c b/init/main.c
index 91f6ebb30ef0..afcca2a38348 100644
--- a/init/main.c
+++ b/init/main.c
@@ -32,6 +32,7 @@
 #include <linux/nmi.h>
 #include <linux/percpu.h>
 #include <linux/kmod.h>
+#include <linux/kmsan.h>
 #include <linux/vmalloc.h>
 #include <linux/kernel_stat.h>
 #include <linux/start_kernel.h>
@@ -554,6 +555,7 @@ static void __init mm_init(void)
 	 */
 	page_ext_init_flatmem();
 	report_meminit();
+	kmsan_initialize_shadow();
 	mem_init();
 	kmem_cache_init();
 	kmemleak_init();
@@ -625,6 +627,7 @@ asmlinkage __visible void __init start_kernel(void)
 	sort_main_extable();
 	trap_init();
 	mm_init();
+	kmsan_initialize();
 
 	ftrace_init();
 
diff --git a/lib/stackdepot.c b/lib/stackdepot.c
index 460efd3ef742..529a86eb1dd1 100644
--- a/lib/stackdepot.c
+++ b/lib/stackdepot.c
@@ -233,8 +233,8 @@ unsigned int stack_depot_fetch(depot_stack_handle_t handle,
 	if (parts.slabindex > depot_index) {
 		WARN(1, "slab index %d out of bounds (%d) for stack id %08x\n",
 			parts.slabindex, depot_index, handle);
-		__memset(trace, 0, sizeof(*trace));
-		return;
+		*entries = NULL;
+		return 0;
 	}
 	slab = stack_slabs[parts.slabindex];
 	stack = slab + offset;
@@ -299,7 +299,8 @@ depot_stack_handle_t stack_depot_save(unsigned long *entries,
 		 * contexts and I/O.
 		 */
 		alloc_flags &= ~GFP_ZONEMASK;
-		alloc_flags &= (GFP_ATOMIC | GFP_KERNEL);
+		alloc_flags &= (GFP_ATOMIC | GFP_KERNEL |
+				__GFP_NO_KMSAN_SHADOW);
 		alloc_flags |= __GFP_NOWARN;
 		page = alloc_pages(alloc_flags, STACK_ALLOC_ORDER);
 		if (page)
diff --git a/mm/Makefile b/mm/Makefile
index d996846697ef..419e6e02dfaf 100644
--- a/mm/Makefile
+++ b/mm/Makefile
@@ -71,6 +71,7 @@ obj-$(CONFIG_PAGE_POISONING) += page_poison.o
 obj-$(CONFIG_SLAB) += slab.o
 obj-$(CONFIG_SLUB) += slub.o
 obj-$(CONFIG_KASAN)	+= kasan/
+obj-$(CONFIG_KMSAN)	+= kmsan/
 obj-$(CONFIG_FAILSLAB) += failslab.o
 obj-$(CONFIG_MEMORY_HOTPLUG) += memory_hotplug.o
 obj-$(CONFIG_MEMTEST)		+= memtest.o
diff --git a/scripts/Makefile.lib b/scripts/Makefile.lib
index 179d55af5852..f9f38f7c1cd4 100644
--- a/scripts/Makefile.lib
+++ b/scripts/Makefile.lib
@@ -140,6 +140,12 @@ _c_flags += $(if $(patsubst n%,, \
 		$(CFLAGS_KASAN), $(CFLAGS_KASAN_NOSANITIZE))
 endif
 
+ifeq ($(CONFIG_KMSAN),y)
+_c_flags += $(if $(patsubst n%,, \
+		$(KMSAN_SANITIZE_$(basetarget).o)$(KMSAN_SANITIZE)y), \
+		$(CFLAGS_KMSAN))
+endif
+
 ifeq ($(CONFIG_UBSAN),y)
 _c_flags += $(if $(patsubst n%,, \
 		$(UBSAN_SANITIZE_$(basetarget).o)$(UBSAN_SANITIZE)$(CONFIG_UBSAN_SANITIZE_ALL)), \
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 17/26] kmsan: disable KMSAN instrumentation for certain kernel parts

[glider]

Instrumenting some files with KMSAN will result in kernel being unable
to link, boot or crashing at runtime for various reasons (e.g. infinite
recursion caused by instrumentation hooks calling instrumented code again).

Disable KMSAN in the following places:
 - arch/x86/boot and arch/x86/realmode/rm, as KMSAN doesn't work for i386;
 - arch/x86/entry/vdso, which isn't linked with KMSAN runtime;
 - three files in arch/x86/kernel - boot problems;
 - arch/x86/mm/cpu_entry_area.c - recursion;
 - EFI stub - build failures;
 - kcov, stackdepot - recursion.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org
---

Change-Id: I90961eabf2dcb9ae992aed259088953bad5e4d6d
---
 arch/x86/boot/Makefile                | 2 ++
 arch/x86/boot/compressed/Makefile     | 2 ++
 arch/x86/entry/vdso/Makefile          | 3 +++
 arch/x86/kernel/Makefile              | 4 ++++
 arch/x86/kernel/cpu/Makefile          | 1 +
 arch/x86/mm/Makefile                  | 2 ++
 arch/x86/realmode/rm/Makefile         | 2 ++
 drivers/firmware/efi/libstub/Makefile | 1 +
 kernel/Makefile                       | 1 +
 lib/Makefile                          | 1 +
 10 files changed, 19 insertions(+)

diff --git a/arch/x86/boot/Makefile b/arch/x86/boot/Makefile
index e2839b5c246c..c039abd4c81f 100644
--- a/arch/x86/boot/Makefile
+++ b/arch/x86/boot/Makefile
@@ -10,6 +10,8 @@
 #
 
 KASAN_SANITIZE			:= n
+# KMSAN doesn't work for i386
+KMSAN_SANITIZE			:= n
 OBJECT_FILES_NON_STANDARD	:= y
 
 # Kernel does not boot with kcov instrumentation here.
diff --git a/arch/x86/boot/compressed/Makefile b/arch/x86/boot/compressed/Makefile
index 6b84afdd7538..9efe2d9fca4c 100644
--- a/arch/x86/boot/compressed/Makefile
+++ b/arch/x86/boot/compressed/Makefile
@@ -18,6 +18,8 @@
 #	compressed vmlinux.bin.all + u32 size of vmlinux.bin.all
 
 KASAN_SANITIZE			:= n
+# KMSAN doesn't work for i386
+KMSAN_SANITIZE			:= n
 OBJECT_FILES_NON_STANDARD	:= y
 
 # Prevents link failures: __sanitizer_cov_trace_pc() is not linked in.
diff --git a/arch/x86/entry/vdso/Makefile b/arch/x86/entry/vdso/Makefile
index 0f2154106d01..000467a1a4f2 100644
--- a/arch/x86/entry/vdso/Makefile
+++ b/arch/x86/entry/vdso/Makefile
@@ -11,6 +11,9 @@ include $(srctree)/lib/vdso/Makefile
 
 KBUILD_CFLAGS += $(DISABLE_LTO)
 KASAN_SANITIZE			:= n
+# Undefined references to KMSAN hooks.
+KMSAN_SANITIZE_vclock_gettime.o := n
+KMSAN_SANITIZE_vgetcpu.o	:= n
 UBSAN_SANITIZE			:= n
 OBJECT_FILES_NON_STANDARD	:= y
 
diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile
index 3578ad248bc9..ce39972a7edf 100644
--- a/arch/x86/kernel/Makefile
+++ b/arch/x86/kernel/Makefile
@@ -28,6 +28,10 @@ KASAN_SANITIZE_dumpstack_$(BITS).o			:= n
 KASAN_SANITIZE_stacktrace.o				:= n
 KASAN_SANITIZE_paravirt.o				:= n
 
+# Work around reboot loop.
+KMSAN_SANITIZE_head$(BITS).o				:= n
+KMSAN_SANITIZE_nmi.o					:= n
+
 OBJECT_FILES_NON_STANDARD_relocate_kernel_$(BITS).o	:= y
 OBJECT_FILES_NON_STANDARD_test_nx.o			:= y
 OBJECT_FILES_NON_STANDARD_paravirt_patch.o		:= y
diff --git a/arch/x86/kernel/cpu/Makefile b/arch/x86/kernel/cpu/Makefile
index d7a1e5a9331c..41f4f8f2f2f0 100644
--- a/arch/x86/kernel/cpu/Makefile
+++ b/arch/x86/kernel/cpu/Makefile
@@ -12,6 +12,7 @@ endif
 # If these files are instrumented, boot hangs during the first second.
 KCOV_INSTRUMENT_common.o := n
 KCOV_INSTRUMENT_perf_event.o := n
+KMSAN_SANITIZE_common.o := n
 
 # Make sure load_percpu_segment has no stackprotector
 nostackp := $(call cc-option, -fno-stack-protector)
diff --git a/arch/x86/mm/Makefile b/arch/x86/mm/Makefile
index 84373dc9b341..42cb3a6409b0 100644
--- a/arch/x86/mm/Makefile
+++ b/arch/x86/mm/Makefile
@@ -7,6 +7,8 @@ KCOV_INSTRUMENT_mem_encrypt_identity.o	:= n
 KASAN_SANITIZE_mem_encrypt.o		:= n
 KASAN_SANITIZE_mem_encrypt_identity.o	:= n
 
+KMSAN_SANITIZE_cpu_entry_area.o := n
+
 ifdef CONFIG_FUNCTION_TRACER
 CFLAGS_REMOVE_mem_encrypt.o		= -pg
 CFLAGS_REMOVE_mem_encrypt_identity.o	= -pg
diff --git a/arch/x86/realmode/rm/Makefile b/arch/x86/realmode/rm/Makefile
index f60501a384f9..27e7bc0bbdde 100644
--- a/arch/x86/realmode/rm/Makefile
+++ b/arch/x86/realmode/rm/Makefile
@@ -7,6 +7,8 @@
 #
 #
 KASAN_SANITIZE			:= n
+# KMSAN doesn't work for i386
+KMSAN_SANITIZE			:= n
 OBJECT_FILES_NON_STANDARD	:= y
 
 # Prevents link failures: __sanitizer_cov_trace_pc() is not linked in.
diff --git a/drivers/firmware/efi/libstub/Makefile b/drivers/firmware/efi/libstub/Makefile
index 0460c7581220..11869c17a64c 100644
--- a/drivers/firmware/efi/libstub/Makefile
+++ b/drivers/firmware/efi/libstub/Makefile
@@ -32,6 +32,7 @@ KBUILD_CFLAGS			:= $(cflags-y) -DDISABLE_BRANCH_PROFILING \
 
 GCOV_PROFILE			:= n
 KASAN_SANITIZE			:= n
+KMSAN_SANITIZE			:= n
 UBSAN_SANITIZE			:= n
 OBJECT_FILES_NON_STANDARD	:= y
 
diff --git a/kernel/Makefile b/kernel/Makefile
index daad787fb795..5fd6fbca2592 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -30,6 +30,7 @@ KCOV_INSTRUMENT_extable.o := n
 # Don't self-instrument.
 KCOV_INSTRUMENT_kcov.o := n
 KASAN_SANITIZE_kcov.o := n
+KMSAN_SANITIZE_kcov.o := n
 CFLAGS_kcov.o := $(call cc-option, -fno-conserve-stack -fno-stack-protector)
 
 # cond_syscall is currently not LTO compatible
diff --git a/lib/Makefile b/lib/Makefile
index cb44262c38ee..a96d1d6e6a34 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -218,6 +218,7 @@ obj-$(CONFIG_IRQ_POLL) += irq_poll.o
 
 obj-$(CONFIG_STACKDEPOT) += stackdepot.o
 KASAN_SANITIZE_stackdepot.o := n
+KMSAN_SANITIZE_stackdepot.o := n
 KCOV_INSTRUMENT_stackdepot.o := n
 
 libfdt_files = fdt.o fdt_ro.o fdt_wip.o fdt_rw.o fdt_sw.o fdt_strerror.o \
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 18/26] kmsan: mm: call KMSAN hooks from SLUB code

[glider]

In order to report uninitialized memory coming from heap allocations
KMSAN has to poison them unless they're created with __GFP_ZERO.

It's handy that we need KMSAN hooks in the places where
init_on_alloc/init_on_free initialization is performed.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org
---

Change-Id: I51103b7981d3aabed747d0c85cbdc85568665871
---
 mm/slub.c | 37 +++++++++++++++++++++++++++++++------
 1 file changed, 31 insertions(+), 6 deletions(-)

diff --git a/mm/slub.c b/mm/slub.c
index 3d63ae320d31..3d6d4c63446e 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -21,6 +21,8 @@
 #include <linux/proc_fs.h>
 #include <linux/seq_file.h>
 #include <linux/kasan.h>
+#include <linux/kmsan.h>
+#include <linux/kmsan-checks.h> /* KMSAN_INIT_VALUE */
 #include <linux/cpu.h>
 #include <linux/cpuset.h>
 #include <linux/mempolicy.h>
@@ -285,17 +287,27 @@ static void prefetch_freepointer(const struct kmem_cache *s, void *object)
 	prefetch(object + s->offset);
 }
 
+/*
+ * When running under KMSAN, get_freepointer_safe() may return an uninitialized
+ * pointer value in the case the current thread loses the race for the next
+ * memory chunk in the freelist. In that case this_cpu_cmpxchg_double() in
+ * slab_alloc_node() will fail, so the uninitialized value won't be used, but
+ * KMSAN will still check all arguments of cmpxchg because of imperfect
+ * handling of inline assembly.
+ * To work around this problem, use KMSAN_INIT_VALUE() to force initialize the
+ * return value of get_freepointer_safe().
+ */
 static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
 {
 	unsigned long freepointer_addr;
 	void *p;
 
 	if (!debug_pagealloc_enabled())
-		return get_freepointer(s, object);
+		return KMSAN_INIT_VALUE(get_freepointer(s, object));
 
 	freepointer_addr = (unsigned long)object + s->offset;
 	probe_kernel_read(&p, (void **)freepointer_addr, sizeof(p));
-	return freelist_ptr(s, p, freepointer_addr);
+	return KMSAN_INIT_VALUE(freelist_ptr(s, p, freepointer_addr));
 }
 
 static inline void set_freepointer(struct kmem_cache *s, void *object, void *fp)
@@ -1390,6 +1402,7 @@ static inline void *kmalloc_large_node_hook(void *ptr, size_t size, gfp_t flags)
 	ptr = kasan_kmalloc_large(ptr, size, flags);
 	/* As ptr might get tagged, call kmemleak hook after KASAN. */
 	kmemleak_alloc(ptr, size, 1, flags);
+	kmsan_kmalloc_large(ptr, size, flags);
 	return ptr;
 }
 
@@ -1397,6 +1410,7 @@ static __always_inline void kfree_hook(void *x)
 {
 	kmemleak_free(x);
 	kasan_kfree_large(x, _RET_IP_);
+	kmsan_kfree_large(x);
 }
 
 static __always_inline bool slab_free_hook(struct kmem_cache *s, void *x)
@@ -1453,6 +1467,12 @@ static inline bool slab_free_freelist_hook(struct kmem_cache *s,
 		} while (object != old_tail);
 	}
 
+	do {
+		object = next;
+		next = get_freepointer(s, object);
+		kmsan_slab_free(s, object);
+	} while (object != old_tail);
+
 /*
  * Compiler cannot detect this function can be removed if slab_free_hook()
  * evaluates to nothing.  Thus, catch all relevant config debug options here.
@@ -2769,6 +2789,7 @@ static __always_inline void *slab_alloc_node(struct kmem_cache *s,
 	if (unlikely(slab_want_init_on_alloc(gfpflags, s)) && object)
 		memset(object, 0, s->object_size);
 
+	kmsan_slab_alloc(s, object, gfpflags);
 	slab_post_alloc_hook(s, gfpflags, 1, &object);
 
 	return object;
@@ -2797,6 +2818,7 @@ void *kmem_cache_alloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size)
 	void *ret = slab_alloc(s, gfpflags, _RET_IP_);
 	trace_kmalloc(_RET_IP_, ret, size, s->size, gfpflags);
 	ret = kasan_kmalloc(s, ret, size, gfpflags);
+
 	return ret;
 }
 EXPORT_SYMBOL(kmem_cache_alloc_trace);
@@ -2809,7 +2831,6 @@ void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t gfpflags, int node)
 
 	trace_kmem_cache_alloc_node(_RET_IP_, ret,
 				    s->object_size, s->size, gfpflags, node);
-
 	return ret;
 }
 EXPORT_SYMBOL(kmem_cache_alloc_node);
@@ -2825,6 +2846,7 @@ void *kmem_cache_alloc_node_trace(struct kmem_cache *s,
 			   size, s->size, gfpflags, node);
 
 	ret = kasan_kmalloc(s, ret, size, gfpflags);
+
 	return ret;
 }
 EXPORT_SYMBOL(kmem_cache_alloc_node_trace);
@@ -3150,7 +3172,7 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
 			  void **p)
 {
 	struct kmem_cache_cpu *c;
-	int i;
+	int i, j;
 
 	/* memcg and kmem_cache debug support */
 	s = slab_pre_alloc_hook(s, flags);
@@ -3188,11 +3210,11 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
 
 	/* Clear memory outside IRQ disabled fastpath loop */
 	if (unlikely(slab_want_init_on_alloc(flags, s))) {
-		int j;
-
 		for (j = 0; j < i; j++)
 			memset(p[j], 0, s->object_size);
 	}
+	for (j = 0; j < i; j++)
+		kmsan_slab_alloc(s, p[j], flags);
 
 	/* memcg and kmem_cache debug support */
 	slab_post_alloc_hook(s, flags, size, p);
@@ -3793,6 +3815,7 @@ static int __init setup_slub_min_objects(char *str)
 
 __setup("slub_min_objects=", setup_slub_min_objects);
 
+__no_sanitize_memory
 void *__kmalloc(size_t size, gfp_t flags)
 {
 	struct kmem_cache *s;
@@ -5698,6 +5721,7 @@ static char *create_unique_id(struct kmem_cache *s)
 	p += sprintf(p, "%07u", s->size);
 
 	BUG_ON(p > name + ID_STR_LENGTH - 1);
+	kmsan_unpoison_shadow(name, p - name);
 	return name;
 }
 
@@ -5847,6 +5871,7 @@ static int sysfs_slab_alias(struct kmem_cache *s, const char *name)
 	al->name = name;
 	al->next = alias_list;
 	alias_list = al;
+	kmsan_unpoison_shadow(al, sizeof(struct saved_alias));
 	return 0;
 }
 
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 19/26] kmsan: call KMSAN hooks where needed

[glider]

Insert KMSAN hooks that check for potential memory errors and/or make
necessary bookkeeping changes:
 - allocate/split/deallocate metadata pages in
   alloc_pages()/split_page()/free_page();
 - clear page shadow and origins in clear_page(), copy_user_highpage();
 - copy page metadata in copy_highpage(), wp_page_copy();
 - handle vmap()/vunmap()/iounmap();
 - handle task creation and deletion;
 - handle vprintk();
 - call softirq entry/exit hooks in kernel/softirq.c;
 - check/initialize memory sent to/read from USB, I2C, and network

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Wolfram Sang <wsa@the-dreams.de>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org
---

Change-Id: I1250a928d9263bf71fdaa067a070bdee686ef47b
---
 arch/x86/include/asm/page_64.h | 13 +++++++++++++
 arch/x86/mm/ioremap.c          |  3 +++
 drivers/i2c/i2c-core-base.c    |  2 ++
 drivers/usb/core/urb.c         |  2 ++
 include/linux/highmem.h        |  4 ++++
 kernel/exit.c                  |  2 ++
 kernel/fork.c                  |  2 ++
 kernel/kthread.c               |  2 ++
 kernel/printk/printk.c         |  8 +++++++-
 kernel/softirq.c               |  5 +++++
 lib/ioremap.c                  |  5 +++++
 mm/compaction.c                |  9 +++++++++
 mm/gup.c                       |  3 +++
 mm/memory.c                    |  2 ++
 mm/page_alloc.c                | 16 ++++++++++++++++
 mm/vmalloc.c                   | 23 +++++++++++++++++++++--
 net/sched/sch_generic.c        |  2 ++
 17 files changed, 100 insertions(+), 3 deletions(-)

diff --git a/arch/x86/include/asm/page_64.h b/arch/x86/include/asm/page_64.h
index 939b1cff4a7b..0ba43d93414f 100644
--- a/arch/x86/include/asm/page_64.h
+++ b/arch/x86/include/asm/page_64.h
@@ -44,14 +44,27 @@ void clear_page_orig(void *page);
 void clear_page_rep(void *page);
 void clear_page_erms(void *page);
 
+/* This is an assembly header, avoid including too much of kmsan.h */
+#ifdef CONFIG_KMSAN
+void kmsan_clear_page(void *page_addr);
+#endif
+__no_sanitize_memory
 static inline void clear_page(void *page)
 {
+#ifdef CONFIG_KMSAN
+	/* alternative_call_2() changes |page|. */
+	void *page_copy = page;
+#endif
 	alternative_call_2(clear_page_orig,
 			   clear_page_rep, X86_FEATURE_REP_GOOD,
 			   clear_page_erms, X86_FEATURE_ERMS,
 			   "=D" (page),
 			   "0" (page)
 			   : "cc", "memory", "rax", "rcx");
+#ifdef CONFIG_KMSAN
+	/* Clear KMSAN shadow for the pages that have it. */
+	kmsan_clear_page(page_copy);
+#endif
 }
 
 void copy_page(void *to, void *from);
diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
index a39dcdb5ae34..fdb2abc11a82 100644
--- a/arch/x86/mm/ioremap.c
+++ b/arch/x86/mm/ioremap.c
@@ -7,6 +7,7 @@
  * (C) Copyright 1995 1996 Linus Torvalds
  */
 
+#include <linux/kmsan.h>
 #include <linux/memblock.h>
 #include <linux/init.h>
 #include <linux/io.h>
@@ -451,6 +452,8 @@ void iounmap(volatile void __iomem *addr)
 		return;
 	}
 
+	kmsan_iounmap_page_range((unsigned long)addr,
+		(unsigned long)addr + get_vm_area_size(p));
 	free_memtype(p->phys_addr, p->phys_addr + get_vm_area_size(p));
 
 	/* Finally remove it */
diff --git a/drivers/i2c/i2c-core-base.c b/drivers/i2c/i2c-core-base.c
index 5f6a4985f2bc..9685d3399c79 100644
--- a/drivers/i2c/i2c-core-base.c
+++ b/drivers/i2c/i2c-core-base.c
@@ -27,6 +27,7 @@
 #include <linux/irqflags.h>
 #include <linux/jump_label.h>
 #include <linux/kernel.h>
+#include <linux/kmsan.h>
 #include <linux/module.h>
 #include <linux/mutex.h>
 #include <linux/of_device.h>
@@ -1975,6 +1976,7 @@ int __i2c_transfer(struct i2c_adapter *adap, struct i2c_msg *msgs, int num)
 				trace_i2c_write(adap, &msgs[i], i);
 	}
 
+	kmsan_handle_i2c_transfer(msgs, num);
 	/* Retry automatically on arbitration loss */
 	orig_jiffies = jiffies;
 	for (ret = 0, try = 0; try <= adap->retries; try++) {
diff --git a/drivers/usb/core/urb.c b/drivers/usb/core/urb.c
index 0eab79f82ce4..5bdb54d71c2e 100644
--- a/drivers/usb/core/urb.c
+++ b/drivers/usb/core/urb.c
@@ -8,6 +8,7 @@
 #include <linux/bitops.h>
 #include <linux/slab.h>
 #include <linux/log2.h>
+#include <linux/kmsan-checks.h>
 #include <linux/usb.h>
 #include <linux/wait.h>
 #include <linux/usb/hcd.h>
@@ -401,6 +402,7 @@ int usb_submit_urb(struct urb *urb, gfp_t mem_flags)
 			URB_SETUP_MAP_SINGLE | URB_SETUP_MAP_LOCAL |
 			URB_DMA_SG_COMBINED);
 	urb->transfer_flags |= (is_out ? URB_DIR_OUT : URB_DIR_IN);
+	kmsan_handle_urb(urb, is_out);
 
 	if (xfertype != USB_ENDPOINT_XFER_CONTROL &&
 			dev->state < USB_STATE_CONFIGURED)
diff --git a/include/linux/highmem.h b/include/linux/highmem.h
index ea5cdbd8c2c3..623b56f48685 100644
--- a/include/linux/highmem.h
+++ b/include/linux/highmem.h
@@ -5,6 +5,7 @@
 #include <linux/fs.h>
 #include <linux/kernel.h>
 #include <linux/bug.h>
+#include <linux/kmsan.h>
 #include <linux/mm.h>
 #include <linux/uaccess.h>
 #include <linux/hardirq.h>
@@ -255,6 +256,8 @@ static inline void copy_user_highpage(struct page *to, struct page *from,
 	vfrom = kmap_atomic(from);
 	vto = kmap_atomic(to);
 	copy_user_page(vto, vfrom, vaddr, to);
+	/* User pages don't have shadow, just clear the destination. */
+	kmsan_clear_page(page_address(to));
 	kunmap_atomic(vto);
 	kunmap_atomic(vfrom);
 }
@@ -270,6 +273,7 @@ static inline void copy_highpage(struct page *to, struct page *from)
 	vfrom = kmap_atomic(from);
 	vto = kmap_atomic(to);
 	copy_page(vto, vfrom);
+	kmsan_copy_page_meta(to, from);
 	kunmap_atomic(vto);
 	kunmap_atomic(vfrom);
 }
diff --git a/kernel/exit.c b/kernel/exit.c
index a46a50d67002..9e3ce929110b 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -60,6 +60,7 @@
 #include <linux/writeback.h>
 #include <linux/shm.h>
 #include <linux/kcov.h>
+#include <linux/kmsan.h>
 #include <linux/random.h>
 #include <linux/rcuwait.h>
 #include <linux/compat.h>
@@ -719,6 +720,7 @@ void __noreturn do_exit(long code)
 
 	profile_task_exit(tsk);
 	kcov_task_exit(tsk);
+	kmsan_task_exit(tsk);
 
 	WARN_ON(blk_needs_flush_plug(tsk));
 
diff --git a/kernel/fork.c b/kernel/fork.c
index bcdf53125210..0f08952a42dc 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -37,6 +37,7 @@
 #include <linux/fdtable.h>
 #include <linux/iocontext.h>
 #include <linux/key.h>
+#include <linux/kmsan.h>
 #include <linux/binfmts.h>
 #include <linux/mman.h>
 #include <linux/mmu_notifier.h>
@@ -931,6 +932,7 @@ static struct task_struct *dup_task_struct(struct task_struct *orig, int node)
 	account_kernel_stack(tsk, 1);
 
 	kcov_task_init(tsk);
+	kmsan_task_create(tsk);
 
 #ifdef CONFIG_FAULT_INJECTION
 	tsk->fail_nth = 0;
diff --git a/kernel/kthread.c b/kernel/kthread.c
index 621467c33fef..9b948559a34a 100644
--- a/kernel/kthread.c
+++ b/kernel/kthread.c
@@ -17,6 +17,7 @@
 #include <linux/unistd.h>
 #include <linux/file.h>
 #include <linux/export.h>
+#include <linux/kmsan.h>
 #include <linux/mutex.h>
 #include <linux/slab.h>
 #include <linux/freezer.h>
@@ -350,6 +351,7 @@ struct task_struct *__kthread_create_on_node(int (*threadfn)(void *data),
 		set_cpus_allowed_ptr(task, cpu_all_mask);
 	}
 	kfree(create);
+	kmsan_task_create(task);
 	return task;
 }
 
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index ca65327a6de8..f77fdcb5f861 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -1914,7 +1914,12 @@ int vprintk_store(int facility, int level,
 	 * The printf needs to come first; we need the syslog
 	 * prefix which might be passed-in as a parameter.
 	 */
-	text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
+	/*
+	 * We've checked the printk arguments in vprintk_emit() already.
+	 * Initialize |text_len| to prevent the errors from spreading.
+	 */
+	text_len = KMSAN_INIT_VALUE(vscnprintf(text, sizeof(textbuf), fmt,
+					       args));
 
 	/* mark and strip a trailing newline */
 	if (text_len && text[text_len-1] == '\n') {
@@ -1972,6 +1977,7 @@ asmlinkage int vprintk_emit(int facility, int level,
 	boot_delay_msec(level);
 	printk_delay();
 
+	kmsan_handle_vprintk(&fmt, args);
 	/* This stops the holder of console_sem just where we want him */
 	logbuf_lock_irqsave(flags);
 	curr_log_seq = log_next_seq;
diff --git a/kernel/softirq.c b/kernel/softirq.c
index 0427a86743a4..6d566dd68b35 100644
--- a/kernel/softirq.c
+++ b/kernel/softirq.c
@@ -11,6 +11,7 @@
 
 #include <linux/export.h>
 #include <linux/kernel_stat.h>
+#include <linux/kmsan.h>
 #include <linux/interrupt.h>
 #include <linux/init.h>
 #include <linux/mm.h>
@@ -370,7 +371,9 @@ static inline void invoke_softirq(void)
 		 * it is the irq stack, because it should be near empty
 		 * at this stage.
 		 */
+		kmsan_softirq_enter();
 		__do_softirq();
+		kmsan_softirq_exit();
 #else
 		/*
 		 * Otherwise, irq_exit() is called on the task stack that can
@@ -600,7 +603,9 @@ static void run_ksoftirqd(unsigned int cpu)
 		 * We can safely run softirq on inline stack, as we are not deep
 		 * in the task stack here.
 		 */
+		kmsan_softirq_enter();
 		__do_softirq();
+		kmsan_softirq_exit();
 		local_irq_enable();
 		cond_resched();
 		return;
diff --git a/lib/ioremap.c b/lib/ioremap.c
index 0a2ffadc6d71..5f830cee5bfc 100644
--- a/lib/ioremap.c
+++ b/lib/ioremap.c
@@ -6,6 +6,7 @@
  *
  * (C) Copyright 1995 1996 Linus Torvalds
  */
+#include <linux/kmsan.h>
 #include <linux/vmalloc.h>
 #include <linux/mm.h>
 #include <linux/sched.h>
@@ -214,6 +215,8 @@ int ioremap_page_range(unsigned long addr,
 	unsigned long start;
 	unsigned long next;
 	int err;
+	unsigned long old_addr = addr;
+	phys_addr_t old_phys_addr = phys_addr;
 
 	might_sleep();
 	BUG_ON(addr >= end);
@@ -228,6 +231,8 @@ int ioremap_page_range(unsigned long addr,
 	} while (pgd++, phys_addr += (next - addr), addr = next, addr != end);
 
 	flush_cache_vmap(start, end);
+	if (!err)
+		kmsan_ioremap_page_range(old_addr, end, old_phys_addr, prot);
 
 	return err;
 }
diff --git a/mm/compaction.c b/mm/compaction.c
index ce08b39d85d4..e9a5898e5fe0 100644
--- a/mm/compaction.c
+++ b/mm/compaction.c
@@ -84,6 +84,15 @@ static void split_map_pages(struct list_head *list)
 
 		for (i = 0; i < nr_pages; i++) {
 			list_add(&page->lru, &tmp_list);
+#ifdef CONFIG_KMSAN
+			/*
+			 * TODO(glider): we may lose the metadata when copying
+			 * something to these pages. Need to allocate shadow
+			 * and origin pages here instead.
+			 */
+			page->shadow = NULL;
+			page->origin = NULL;
+#endif
 			page++;
 		}
 	}
diff --git a/mm/gup.c b/mm/gup.c
index 23a9f9c9d377..c16e4288997d 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -4,6 +4,7 @@
 #include <linux/err.h>
 #include <linux/spinlock.h>
 
+#include <linux/kmsan.h>
 #include <linux/mm.h>
 #include <linux/memremap.h>
 #include <linux/pagemap.h>
@@ -2347,6 +2348,7 @@ int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
 	    gup_fast_permitted(start, end)) {
 		local_irq_save(flags);
 		gup_pgd_range(start, end, write ? FOLL_WRITE : 0, pages, &nr);
+		kmsan_gup_pgd_range(pages, nr);
 		local_irq_restore(flags);
 	}
 
@@ -2416,6 +2418,7 @@ int get_user_pages_fast(unsigned long start, int nr_pages,
 	    gup_fast_permitted(start, end)) {
 		local_irq_disable();
 		gup_pgd_range(addr, end, gup_flags, pages, &nr);
+		kmsan_gup_pgd_range(pages, nr);
 		local_irq_enable();
 		ret = nr;
 	}
diff --git a/mm/memory.c b/mm/memory.c
index b1ca51a079f2..48ceacc06e2d 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -51,6 +51,7 @@
 #include <linux/highmem.h>
 #include <linux/pagemap.h>
 #include <linux/memremap.h>
+#include <linux/kmsan.h>
 #include <linux/ksm.h>
 #include <linux/rmap.h>
 #include <linux/export.h>
@@ -2328,6 +2329,7 @@ static vm_fault_t wp_page_copy(struct vm_fault *vmf)
 		if (!new_page)
 			goto oom;
 		cow_user_page(new_page, old_page, vmf->address, vma);
+		kmsan_copy_page_meta(new_page, old_page);
 	}
 
 	if (mem_cgroup_try_charge_delay(new_page, mm, GFP_KERNEL, &memcg, false))
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index c0b2e0306720..f52494e0dcd9 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -26,6 +26,8 @@
 #include <linux/compiler.h>
 #include <linux/kernel.h>
 #include <linux/kasan.h>
+#include <linux/kmsan.h>
+#include <linux/kmsan-checks.h>
 #include <linux/module.h>
 #include <linux/suspend.h>
 #include <linux/pagevec.h>
@@ -1133,6 +1135,7 @@ static __always_inline bool free_pages_prepare(struct page *page,
 	VM_BUG_ON_PAGE(PageTail(page), page);
 
 	trace_mm_page_free(page, order);
+	kmsan_free_page(page, order);
 
 	/*
 	 * Check tail pages before head page information is cleared to
@@ -3121,6 +3124,7 @@ void split_page(struct page *page, unsigned int order)
 	VM_BUG_ON_PAGE(PageCompound(page), page);
 	VM_BUG_ON_PAGE(!page_count(page), page);
 
+	kmsan_split_page(page, order);
 	for (i = 1; i < (1 << order); i++)
 		set_page_refcounted(page + i);
 	split_page_owner(page, order);
@@ -3253,6 +3257,13 @@ static struct page *rmqueue_pcplist(struct zone *preferred_zone,
 /*
  * Allocate a page from the given zone. Use pcplists for order-0 allocations.
  */
+/*
+ * TODO(glider): rmqueue() may call __msan_poison_alloca() through a call to
+ * set_pfnblock_flags_mask(). If __msan_poison_alloca() attempts to allocate
+ * pages for the stack depot, it may call rmqueue() again, which will result
+ * in a deadlock.
+ */
+__no_sanitize_memory
 static inline
 struct page *rmqueue(struct zone *preferred_zone,
 			struct zone *zone, unsigned int order,
@@ -4779,6 +4790,11 @@ __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order, int preferred_nid,
 
 	trace_mm_page_alloc(page, order, alloc_mask, ac.migratetype);
 
+	if (page)
+		if (kmsan_alloc_page(page, order, gfp_mask)) {
+			__free_pages(page, order);
+			page = NULL;
+		}
 	return page;
 }
 EXPORT_SYMBOL(__alloc_pages_nodemask);
diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index a3c70e275f4e..bdf66ffcf02c 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -29,6 +29,7 @@
 #include <linux/rcupdate.h>
 #include <linux/pfn.h>
 #include <linux/kmemleak.h>
+#include <linux/kmsan.h>
 #include <linux/atomic.h>
 #include <linux/compiler.h>
 #include <linux/llist.h>
@@ -119,7 +120,8 @@ static void vunmap_p4d_range(pgd_t *pgd, unsigned long addr, unsigned long end)
 	} while (p4d++, addr = next, addr != end);
 }
 
-static void vunmap_page_range(unsigned long addr, unsigned long end)
+/* Exported for KMSAN, visible in mm/kmsan/kmsan.h only. */
+void __vunmap_page_range(unsigned long addr, unsigned long end)
 {
 	pgd_t *pgd;
 	unsigned long next;
@@ -133,6 +135,12 @@ static void vunmap_page_range(unsigned long addr, unsigned long end)
 		vunmap_p4d_range(pgd, addr, next);
 	} while (pgd++, addr = next, addr != end);
 }
+EXPORT_SYMBOL(__vunmap_page_range);
+static void vunmap_page_range(unsigned long addr, unsigned long end)
+{
+	kmsan_vunmap_page_range(addr, end);
+	__vunmap_page_range(addr, end);
+}
 
 static int vmap_pte_range(pmd_t *pmd, unsigned long addr,
 		unsigned long end, pgprot_t prot, struct page **pages, int *nr)
@@ -216,8 +224,11 @@ static int vmap_p4d_range(pgd_t *pgd, unsigned long addr,
  * will have pfns corresponding to the "pages" array.
  *
  * Ie. pte at addr+N*PAGE_SIZE shall point to pfn corresponding to pages[N]
+ *
+ * This function is exported for use in KMSAN, but is only declared in KMSAN
+ * headers.
  */
-static int vmap_page_range_noflush(unsigned long start, unsigned long end,
+int __vmap_page_range_noflush(unsigned long start, unsigned long end,
 				   pgprot_t prot, struct page **pages)
 {
 	pgd_t *pgd;
@@ -237,6 +248,14 @@ static int vmap_page_range_noflush(unsigned long start, unsigned long end,
 
 	return nr;
 }
+EXPORT_SYMBOL(__vmap_page_range_noflush);
+
+static int vmap_page_range_noflush(unsigned long start, unsigned long end,
+				   pgprot_t prot, struct page **pages)
+{
+	kmsan_vmap_page_range_noflush(start, end, prot, pages);
+	return __vmap_page_range_noflush(start, end, prot, pages);
+}
 
 static int vmap_page_range(unsigned long start, unsigned long end,
 			   pgprot_t prot, struct page **pages)
diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
index 17bd8f539bc7..fd22c4a4ba42 100644
--- a/net/sched/sch_generic.c
+++ b/net/sched/sch_generic.c
@@ -11,6 +11,7 @@
 #include <linux/module.h>
 #include <linux/types.h>
 #include <linux/kernel.h>
+#include <linux/kmsan-checks.h>
 #include <linux/sched.h>
 #include <linux/string.h>
 #include <linux/errno.h>
@@ -659,6 +660,7 @@ static struct sk_buff *pfifo_fast_dequeue(struct Qdisc *qdisc)
 	} else {
 		qdisc->empty = true;
 	}
+	kmsan_check_skb(skb);
 
 	return skb;
 }
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 20/26] kmsan: disable instrumentation of certain functions

[glider]

Some functions are called from handwritten assembly, and therefore don't
have their arguments' metadata fully set up by the instrumentation code.
Mark them with __no_sanitize_memory to avoid false positives from
spreading further.
Certain functions perform task switching, so that the value of |current|
is different as they proceed. Because KMSAN state pointer is only read
once at the beginning of the function, touching it after |current| has
changed may be dangerous.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org
---

Change-Id: I684d23dac5a22eb0a4cea71993cb934302b17cea
---
 arch/x86/entry/common.c                |  1 +
 arch/x86/include/asm/irq_regs.h        |  1 +
 arch/x86/include/asm/syscall_wrapper.h |  1 +
 arch/x86/kernel/apic/apic.c            |  2 ++
 arch/x86/kernel/dumpstack_64.c         |  1 +
 arch/x86/kernel/process_64.c           |  5 +++++
 arch/x86/kernel/traps.c                | 12 ++++++++++--
 arch/x86/kernel/uprobes.c              |  7 ++++++-
 kernel/profile.c                       |  1 +
 kernel/sched/core.c                    | 11 +++++++++++
 10 files changed, 39 insertions(+), 3 deletions(-)

diff --git a/arch/x86/entry/common.c b/arch/x86/entry/common.c
index 3f8e22615812..0dd5b2acb355 100644
--- a/arch/x86/entry/common.c
+++ b/arch/x86/entry/common.c
@@ -275,6 +275,7 @@ __visible inline void syscall_return_slowpath(struct pt_regs *regs)
 }
 
 #ifdef CONFIG_X86_64
+__no_sanitize_memory
 __visible void do_syscall_64(unsigned long nr, struct pt_regs *regs)
 {
 	struct thread_info *ti;
diff --git a/arch/x86/include/asm/irq_regs.h b/arch/x86/include/asm/irq_regs.h
index 187ce59aea28..d65a00bd6f02 100644
--- a/arch/x86/include/asm/irq_regs.h
+++ b/arch/x86/include/asm/irq_regs.h
@@ -14,6 +14,7 @@
 
 DECLARE_PER_CPU(struct pt_regs *, irq_regs);
 
+__no_sanitize_memory
 static inline struct pt_regs *get_irq_regs(void)
 {
 	return __this_cpu_read(irq_regs);
diff --git a/arch/x86/include/asm/syscall_wrapper.h b/arch/x86/include/asm/syscall_wrapper.h
index e046a405743d..43910ce1b53b 100644
--- a/arch/x86/include/asm/syscall_wrapper.h
+++ b/arch/x86/include/asm/syscall_wrapper.h
@@ -159,6 +159,7 @@
 	ALLOW_ERROR_INJECTION(__x64_sys##name, ERRNO);			\
 	static long __se_sys##name(__MAP(x,__SC_LONG,__VA_ARGS__));	\
 	static inline long __do_sys##name(__MAP(x,__SC_DECL,__VA_ARGS__));\
+	__no_sanitize_memory						\
 	asmlinkage long __x64_sys##name(const struct pt_regs *regs)	\
 	{								\
 		return __se_sys##name(SC_X86_64_REGS_TO_ARGS(x,__VA_ARGS__));\
diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c
index 9e2dd2b296cd..43a1edd919ea 100644
--- a/arch/x86/kernel/apic/apic.c
+++ b/arch/x86/kernel/apic/apic.c
@@ -1118,6 +1118,8 @@ static void local_apic_timer_interrupt(void)
  * [ if a single-CPU system runs an SMP kernel then we call the local
  *   interrupt as well. Thus we cannot inline the local irq ... ]
  */
+/* TODO(glider): |regs| is uninitialized, so is |*regs|. */
+__no_sanitize_memory
 __visible void __irq_entry smp_apic_timer_interrupt(struct pt_regs *regs)
 {
 	struct pt_regs *old_regs = set_irq_regs(regs);
diff --git a/arch/x86/kernel/dumpstack_64.c b/arch/x86/kernel/dumpstack_64.c
index 753b8cfe8b8a..ba883d282a43 100644
--- a/arch/x86/kernel/dumpstack_64.c
+++ b/arch/x86/kernel/dumpstack_64.c
@@ -143,6 +143,7 @@ static bool in_irq_stack(unsigned long *stack, struct stack_info *info)
 	return true;
 }
 
+__no_sanitize_memory
 int get_stack_info(unsigned long *stack, struct task_struct *task,
 		   struct stack_info *info, unsigned long *visit_mask)
 {
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index af64519b2695..27649ee3dbf8 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -500,6 +500,11 @@ void compat_start_thread(struct pt_regs *regs, u32 new_ip, u32 new_sp)
  * Kprobes not supported here. Set the probe on schedule instead.
  * Function graph tracer not supported too.
  */
+/*
+ * TODO(glider): __switch_to() does weird things with tasks, don't report
+ * anything here (also avoid touching the KMSAN state).
+ */
+__no_sanitize_memory
 __visible __notrace_funcgraph struct task_struct *
 __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
 {
diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c
index 4bb0f8447112..a94282d1f60b 100644
--- a/arch/x86/kernel/traps.c
+++ b/arch/x86/kernel/traps.c
@@ -618,7 +618,10 @@ NOKPROBE_SYMBOL(do_int3);
  * Help handler running on a per-cpu (IST or entry trampoline) stack
  * to switch to the normal thread stack if the interrupted code was in
  * user mode. The actual stack switch is done in entry_64.S
+ *
+ * This function switches the registers - don't instrument it with KMSAN!
  */
+__no_sanitize_memory
 asmlinkage __visible notrace struct pt_regs *sync_regs(struct pt_regs *eregs)
 {
 	struct pt_regs *regs = (struct pt_regs *)this_cpu_read(cpu_current_top_of_stack) - 1;
@@ -634,6 +637,11 @@ struct bad_iret_stack {
 };
 
 asmlinkage __visible notrace
+/*
+ * Dark magic happening here, let's not instrument this function.
+ * Also avoid copying any metadata by using raw __memmove().
+ */
+__no_sanitize_memory
 struct bad_iret_stack *fixup_bad_iret(struct bad_iret_stack *s)
 {
 	/*
@@ -648,10 +656,10 @@ struct bad_iret_stack *fixup_bad_iret(struct bad_iret_stack *s)
 		(struct bad_iret_stack *)this_cpu_read(cpu_tss_rw.x86_tss.sp0) - 1;
 
 	/* Copy the IRET target to the new stack. */
-	memmove(&new_stack->regs.ip, (void *)s->regs.sp, 5*8);
+	__memmove(&new_stack->regs.ip, (void *)s->regs.sp, 5*8);
 
 	/* Copy the remainder of the stack from the current stack. */
-	memmove(new_stack, s, offsetof(struct bad_iret_stack, regs.ip));
+	__memmove(new_stack, s, offsetof(struct bad_iret_stack, regs.ip));
 
 	BUG_ON(!user_mode(&new_stack->regs));
 	return new_stack;
diff --git a/arch/x86/kernel/uprobes.c b/arch/x86/kernel/uprobes.c
index 8cd745ef8c7b..bcd4bf5a909f 100644
--- a/arch/x86/kernel/uprobes.c
+++ b/arch/x86/kernel/uprobes.c
@@ -8,6 +8,7 @@
  *	Jim Keniston
  */
 #include <linux/kernel.h>
+#include <linux/kmsan-checks.h>
 #include <linux/sched.h>
 #include <linux/ptrace.h>
 #include <linux/uprobes.h>
@@ -997,9 +998,13 @@ int arch_uprobe_post_xol(struct arch_uprobe *auprobe, struct pt_regs *regs)
 int arch_uprobe_exception_notify(struct notifier_block *self, unsigned long val, void *data)
 {
 	struct die_args *args = data;
-	struct pt_regs *regs = args->regs;
+	struct pt_regs *regs;
 	int ret = NOTIFY_DONE;
 
+	kmsan_unpoison_shadow(args, sizeof(*args));
+	regs = args->regs;
+	if (regs)
+		kmsan_unpoison_shadow(regs, sizeof(*regs));
 	/* We are only interested in userspace traps */
 	if (regs && !user_mode(regs))
 		return NOTIFY_DONE;
diff --git a/kernel/profile.c b/kernel/profile.c
index af7c94bf5fa1..835a5b66d1a4 100644
--- a/kernel/profile.c
+++ b/kernel/profile.c
@@ -399,6 +399,7 @@ void profile_hits(int type, void *__pc, unsigned int nr_hits)
 }
 EXPORT_SYMBOL_GPL(profile_hits);
 
+__no_sanitize_memory
 void profile_tick(int type)
 {
 	struct pt_regs *regs = get_irq_regs();
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index dd05a378631a..951d19d217d7 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -475,6 +475,8 @@ void wake_q_add_safe(struct wake_q_head *head, struct task_struct *task)
 		put_task_struct(task);
 }
 
+/* TODO(glider): context switching here. */
+__no_sanitize_memory
 void wake_up_q(struct wake_q_head *head)
 {
 	struct wake_q_node *node = head->first;
@@ -3180,6 +3182,8 @@ prepare_task_switch(struct rq *rq, struct task_struct *prev,
  * past. prev == current is still correct but we need to recalculate this_rq
  * because prev may have moved to another CPU.
  */
+/* TODO(glider): |current| changes here. */
+__no_sanitize_memory
 static struct rq *finish_task_switch(struct task_struct *prev)
 	__releases(rq->lock)
 {
@@ -3986,6 +3990,8 @@ pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
  *
  * WARNING: must be called with preemption disabled!
  */
+/* TODO(glider): |current| changes here. */
+__no_sanitize_memory
 static void __sched notrace __schedule(bool preempt)
 {
 	struct task_struct *prev, *next;
@@ -4605,6 +4611,8 @@ int task_prio(const struct task_struct *p)
  *
  * Return: 1 if the CPU is currently idle. 0 otherwise.
  */
+/* TODO(glider): nothing to report here. */
+__no_sanitize_memory
 int idle_cpu(int cpu)
 {
 	struct rq *rq = cpu_rq(cpu);
@@ -6544,6 +6552,7 @@ static struct kmem_cache *task_group_cache __read_mostly;
 DECLARE_PER_CPU(cpumask_var_t, load_balance_mask);
 DECLARE_PER_CPU(cpumask_var_t, select_idle_mask);
 
+__no_sanitize_memory
 void __init sched_init(void)
 {
 	unsigned long ptr = 0;
@@ -6716,6 +6725,8 @@ static inline int preempt_count_equals(int preempt_offset)
 	return (nested == preempt_offset);
 }
 
+/* TODO(glider): the args are most certainly initialized. */
+__no_sanitize_memory
 void __might_sleep(const char *file, int line, int preempt_offset)
 {
 	/*
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 21/26] kmsan: unpoison |tlb| in arch_tlb_gather_mmu()

[glider]

This is a hack to reduce stackdepot pressure.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org

---

Change-Id: I22a201e7e4f67ed74f8129072f12e5351b26103a
---
 mm/mmu_gather.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/mm/mmu_gather.c b/mm/mmu_gather.c
index 7d70e5c78f97..8c5ea2d2e7d5 100644
--- a/mm/mmu_gather.c
+++ b/mm/mmu_gather.c
@@ -1,6 +1,7 @@
 #include <linux/gfp.h>
 #include <linux/highmem.h>
 #include <linux/kernel.h>
+#include <linux/kmsan-checks.h>
 #include <linux/mmdebug.h>
 #include <linux/mm_types.h>
 #include <linux/pagemap.h>
@@ -206,6 +207,15 @@ void tlb_flush_mmu(struct mmu_gather *tlb)
 void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm,
 			unsigned long start, unsigned long end)
 {
+	/*
+	 * TODO(glider): struct mmu_gather contains 7 1-bit fields packed into a
+	 * 32-bit unsigned int value. The remaining 25 bits remain uninitialized
+	 * and are never used, but KMSAN updates the origin for them in
+	 * zap_pXX_range() in mm/memory.c, thus creating very long origin
+	 * chains. This is technically correct, but consumes too much memory.
+	 * Unpoisoning the whole structure will prevent creating such chains.
+	 */
+	kmsan_unpoison_shadow(tlb, sizeof(*tlb));
 	tlb->mm = mm;
 
 	/* Is it from 0 to ~0? */
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 22/26] kmsan: use __msan_memcpy() where possible.

[glider]

Unless stated otherwise (by explicitly calling __memcpy()) we want all
memcpy() calls to call __msan_memcpy() so that shadow and origin values
are updated accordingly.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org
---

Change-Id: Ib2512ce5aa8d457453dd38caa12f58f002166813
---
 arch/x86/include/asm/string_64.h | 9 ++++++++-
 include/linux/compiler.h         | 9 ++++++++-
 include/linux/string.h           | 2 ++
 3 files changed, 18 insertions(+), 2 deletions(-)

diff --git a/arch/x86/include/asm/string_64.h b/arch/x86/include/asm/string_64.h
index 75314c3dbe47..d3c76d910c23 100644
--- a/arch/x86/include/asm/string_64.h
+++ b/arch/x86/include/asm/string_64.h
@@ -11,7 +11,13 @@
    function. */
 
 #define __HAVE_ARCH_MEMCPY 1
+#if defined(CONFIG_KMSAN)
+#undef memcpy
+/* __msan_memcpy() is defined in compiler.h */
+#define memcpy(dst, src, len) __msan_memcpy(dst, src, len)
+#else
 extern void *memcpy(void *to, const void *from, size_t len);
+#endif
 extern void *__memcpy(void *to, const void *from, size_t len);
 
 #define __HAVE_ARCH_MEMSET
@@ -64,7 +70,8 @@ char *strcpy(char *dest, const char *src);
 char *strcat(char *dest, const char *src);
 int strcmp(const char *cs, const char *ct);
 
-#if defined(CONFIG_KASAN) && !defined(__SANITIZE_ADDRESS__)
+#if (defined(CONFIG_KASAN) && !defined(__SANITIZE_ADDRESS__)) || \
+	(defined(CONFIG_KMSAN) && !defined(__SANITIZE_MEMORY__))
 
 /*
  * For files that not instrumented (e.g. mm/slub.c) we
diff --git a/include/linux/compiler.h b/include/linux/compiler.h
index 4a594bd1bc47..40b1ab13fec7 100644
--- a/include/linux/compiler.h
+++ b/include/linux/compiler.h
@@ -179,6 +179,13 @@ void ftrace_likely_update(struct ftrace_likely_data *f, int val,
 
 #include <uapi/linux/types.h>
 
+#ifdef CONFIG_KMSAN
+void *__msan_memcpy(void *dst, const void *src, u64 size);
+#define __DO_MEMCPY(res, p, size) __msan_memcpy(res, p, size)
+#else
+#define __DO_MEMCPY(res, p, size) __builtin_memcpy(res, p, size)
+#endif
+
 #define __READ_ONCE_SIZE						\
 ({									\
 	switch (size) {							\
@@ -188,7 +195,7 @@ void ftrace_likely_update(struct ftrace_likely_data *f, int val,
 	case 8: *(__u64 *)res = *(volatile __u64 *)p; break;		\
 	default:							\
 		barrier();						\
-		__builtin_memcpy((void *)res, (const void *)p, size);	\
+		__DO_MEMCPY((void *)res, (const void *)p, size);	\
 		barrier();						\
 	}								\
 })
diff --git a/include/linux/string.h b/include/linux/string.h
index b6ccdc2c7f02..5d8ce09cba2e 100644
--- a/include/linux/string.h
+++ b/include/linux/string.h
@@ -363,6 +363,7 @@ __FORTIFY_INLINE void *memset(void *p, int c, __kernel_size_t size)
 	return __builtin_memset(p, c, size);
 }
 
+#ifndef CONFIG_KMSAN
 __FORTIFY_INLINE void *memcpy(void *p, const void *q, __kernel_size_t size)
 {
 	size_t p_size = __builtin_object_size(p, 0);
@@ -377,6 +378,7 @@ __FORTIFY_INLINE void *memcpy(void *p, const void *q, __kernel_size_t size)
 		fortify_panic(__func__);
 	return __builtin_memcpy(p, q, size);
 }
+#endif
 
 __FORTIFY_INLINE void *memmove(void *p, const void *q, __kernel_size_t size)
 {
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 23/26] kmsan: unpoisoning buffers from devices etc.

[glider]

When data is copied to memory from a device KMSAN should treat it as
initialized. In most cases it's enough to just unpoison the buffer that
is known to come from a device.
In the case with __do_page_cache_readahead() and bio_copy_user_iov() we
have to mark the whole pages as ignored by KMSAN, as it's not obvious
where these pages are read again.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Cc: Martin K. Petersen <martin.petersen@oracle.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Eric Van Hensbergen <ericvh@gmail.com>
Cc: Takashi Iwai <tiwai@suse.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org

---
Suggestions on how to avoid scattering too many annotations over the
code are welcome. Perhaps annotating DMA reads/writes is enough to get
rid of most of these annotations.

Change-Id: Id460e7a86ce564f1357469f53d0c7410ca08f0e9
---
 block/bio.c                  | 20 ++++++++++++++++++++
 block/partition-generic.c    |  9 ++++++++-
 drivers/char/random.c        |  2 ++
 drivers/input/serio/libps2.c |  6 +++++-
 drivers/scsi/scsi_lib.c      |  4 ++++
 drivers/usb/core/message.c   |  6 +++++-
 drivers/virtio/virtio_ring.c | 14 ++++++++++++++
 fs/buffer.c                  |  7 ++++++-
 include/linux/dma-mapping.h  |  2 ++
 include/linux/skbuff.h       |  5 ++++-
 mm/filemap.c                 |  3 +++
 mm/readahead.c               |  6 ++++++
 net/9p/protocol.c            |  2 ++
 sound/core/oss/pcm_oss.c     |  7 +++++++
 14 files changed, 88 insertions(+), 5 deletions(-)

diff --git a/block/bio.c b/block/bio.c
index 8f0ed6228fc5..40773cf4b50c 100644
--- a/block/bio.c
+++ b/block/bio.c
@@ -11,6 +11,7 @@
 #include <linux/slab.h>
 #include <linux/init.h>
 #include <linux/kernel.h>
+#include <linux/kmsan-checks.h>
 #include <linux/export.h>
 #include <linux/mempool.h>
 #include <linux/workqueue.h>
@@ -900,6 +901,13 @@ static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter)
 			__bio_add_page(bio, page, len, offset);
 		}
 		offset = 0;
+		/*
+		 * TODO(glider): these pages will be soon passed to block
+		 * device for reading, so consider them initialized.
+		 */
+		if (iov_iter_rw(iter) == READ)
+			kmsan_unpoison_shadow(page_address(page),
+						PAGE_SIZE);
 	}
 
 	iov_iter_advance(iter, size);
@@ -1274,11 +1282,16 @@ struct bio *bio_copy_user_iov(struct request_queue *q,
 
 			i++;
 		} else {
+			/*
+			 * TODO(glider): KMSAN doesn't track the pages
+			 * allocated for bio here.
+			 */
 			page = alloc_page(q->bounce_gfp | gfp_mask);
 			if (!page) {
 				ret = -ENOMEM;
 				break;
 			}
+			kmsan_ignore_page(page, /*order*/0);
 		}
 
 		if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) {
@@ -1574,6 +1587,13 @@ struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len,
 		if (!page)
 			goto cleanup;
 
+		/*
+		 * TODO(glider): if we're about to read data from a SCSI device,
+		 * assume the page allocated for that is already initialized.
+		 */
+		if (reading)
+			kmsan_unpoison_shadow(page_address(page), PAGE_SIZE);
+
 		if (!reading)
 			memcpy(page_address(page), p, bytes);
 
diff --git a/block/partition-generic.c b/block/partition-generic.c
index aee643ce13d1..87b3ae9e7727 100644
--- a/block/partition-generic.c
+++ b/block/partition-generic.c
@@ -663,13 +663,20 @@ unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p)
 {
 	struct address_space *mapping = bdev->bd_inode->i_mapping;
 	struct page *page;
+	unsigned char *retval;
 
 	page = read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_SHIFT-9)), NULL);
 	if (!IS_ERR(page)) {
 		if (PageError(page))
 			goto fail;
 		p->v = page;
-		return (unsigned char *)page_address(page) +  ((n & ((1 << (PAGE_SHIFT - 9)) - 1)) << 9);
+		retval = (unsigned char *)page_address(page) +
+			 ((n & ((1 << (PAGE_SHIFT - 9)) - 1)) << 9);
+		/*
+		 * Unpoison sector-sized chunk of memory coming from the device.
+		 */
+		kmsan_unpoison_shadow(retval, SECTOR_SIZE);
+		return retval;
 fail:
 		put_page(page);
 	}
diff --git a/drivers/char/random.c b/drivers/char/random.c
index de434feb873a..cc4afc0f1039 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -320,6 +320,7 @@
 #include <linux/fs.h>
 #include <linux/genhd.h>
 #include <linux/interrupt.h>
+#include <linux/kmsan-checks.h>
 #include <linux/mm.h>
 #include <linux/nodemask.h>
 #include <linux/spinlock.h>
@@ -1061,6 +1062,7 @@ static void _extract_crng(struct crng_state *crng,
 	spin_lock_irqsave(&crng->lock, flags);
 	if (arch_get_random_long(&v))
 		crng->state[14] ^= v;
+	kmsan_unpoison_shadow(crng->state, sizeof(crng->state));
 	chacha20_block(&crng->state[0], out);
 	if (crng->state[12] == 0)
 		crng->state[13]++;
diff --git a/drivers/input/serio/libps2.c b/drivers/input/serio/libps2.c
index a8c94a940a79..80a9e0a9d3c3 100644
--- a/drivers/input/serio/libps2.c
+++ b/drivers/input/serio/libps2.c
@@ -8,6 +8,7 @@
 
 
 #include <linux/delay.h>
+#include <linux/kmsan-checks.h>
 #include <linux/module.h>
 #include <linux/sched.h>
 #include <linux/interrupt.h>
@@ -30,6 +31,7 @@ static int ps2_do_sendbyte(struct ps2dev *ps2dev, u8 byte,
 	int error;
 
 	lockdep_assert_held(&ps2dev->serio->lock);
+	kmsan_check_memory(&byte, 1);
 
 	do {
 		ps2dev->nak = 1;
@@ -294,9 +296,11 @@ int __ps2_command(struct ps2dev *ps2dev, u8 *param, unsigned int command)
 
 	serio_pause_rx(ps2dev->serio);
 
-	if (param)
+	if (param) {
 		for (i = 0; i < receive; i++)
 			param[i] = ps2dev->cmdbuf[(receive - 1) - i];
+		kmsan_unpoison_shadow(param, receive);
+	}
 
 	if (ps2dev->cmdcnt &&
 	    (command != PS2_CMD_RESET_BAT || ps2dev->cmdcnt != 1)) {
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index dc210b9d4896..4235b317ac5f 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -14,6 +14,7 @@
 #include <linux/blkdev.h>
 #include <linux/completion.h>
 #include <linux/kernel.h>
+#include <linux/kmsan-checks.h>
 #include <linux/export.h>
 #include <linux/init.h>
 #include <linux/pci.h>
@@ -296,6 +297,9 @@ int __scsi_execute(struct scsi_device *sdev, const unsigned char *cmd,
 	ret = rq->result;
  out:
 	blk_put_request(req);
+	/* TODO(glider): this is a bit rough. */
+	if (data_direction == DMA_FROM_DEVICE)
+		kmsan_unpoison_shadow(buffer, bufflen);
 
 	return ret;
 }
diff --git a/drivers/usb/core/message.c b/drivers/usb/core/message.c
index 5adf489428aa..f1c3f8b3cfae 100644
--- a/drivers/usb/core/message.c
+++ b/drivers/usb/core/message.c
@@ -9,6 +9,7 @@
 #include <linux/usb.h>
 #include <linux/module.h>
 #include <linux/slab.h>
+#include <linux/kmsan-checks.h>
 #include <linux/mm.h>
 #include <linux/timer.h>
 #include <linux/ctype.h>
@@ -101,8 +102,11 @@ static int usb_internal_control_msg(struct usb_device *usb_dev,
 	retv = usb_start_wait_urb(urb, timeout, &length);
 	if (retv < 0)
 		return retv;
-	else
+	else {
+		/* TODO(glider): USB initializes |length| bytes? */
+		kmsan_unpoison_shadow(data, length);
 		return length;
+	}
 }
 
 /**
diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index bdc08244a648..8b7ddc43618f 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -11,6 +11,7 @@
 #include <linux/module.h>
 #include <linux/hrtimer.h>
 #include <linux/dma-mapping.h>
+#include <linux/kmsan-checks.h>
 #include <xen/xen.h>
 
 #ifdef DEBUG
@@ -387,6 +388,12 @@ static void vring_unmap_one_split(const struct vring_virtqueue *vq,
 			       (flags & VRING_DESC_F_WRITE) ?
 			       DMA_FROM_DEVICE : DMA_TO_DEVICE);
 	}
+	/*
+	 * Unmapping DMA memory after a transfer from device requires this
+	 * memory to be unpoisoned.
+	 */
+	if (flags & VRING_DESC_F_WRITE)
+		kmsan_unpoison_shadow((const void *)desc->addr, desc->len);
 }
 
 static struct vring_desc *alloc_indirect_split(struct virtqueue *_vq,
@@ -500,6 +507,13 @@ static inline int virtqueue_add_split(struct virtqueue *_vq,
 			desc[i].flags = cpu_to_virtio16(_vq->vdev, VRING_DESC_F_NEXT | VRING_DESC_F_WRITE);
 			desc[i].addr = cpu_to_virtio64(_vq->vdev, addr);
 			desc[i].len = cpu_to_virtio32(_vq->vdev, sg->length);
+			/*
+			 * It's hard to figure out the buffer's address upon
+			 * receive. Instead we unpoison it once, when exposing
+			 * it to the device, and hope nobody else will write to
+			 * it.
+			 */
+			kmsan_unpoison_shadow(sg_virt(sg), sg->length);
 			prev = i;
 			i = virtio16_to_cpu(_vq->vdev, desc[i].next);
 		}
diff --git a/fs/buffer.c b/fs/buffer.c
index 86a38b979323..c08fb1a77fa1 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -1170,14 +1170,17 @@ static struct buffer_head *__bread_slow(struct buffer_head *bh)
 	lock_buffer(bh);
 	if (buffer_uptodate(bh)) {
 		unlock_buffer(bh);
+		kmsan_unpoison_shadow(bh->b_data, bh->b_size);
 		return bh;
 	} else {
 		get_bh(bh);
 		bh->b_end_io = end_buffer_read_sync;
 		submit_bh(REQ_OP_READ, 0, bh);
 		wait_on_buffer(bh);
-		if (buffer_uptodate(bh))
+		if (buffer_uptodate(bh)) {
+			kmsan_unpoison_shadow(bh->b_data, bh->b_size);
 			return bh;
+		}
 	}
 	brelse(bh);
 	return NULL;
@@ -1320,6 +1323,8 @@ __getblk_gfp(struct block_device *bdev, sector_t block,
 	might_sleep();
 	if (bh == NULL)
 		bh = __getblk_slow(bdev, block, size, gfp);
+	if (bh)
+		kmsan_unpoison_shadow(bh->b_data, bh->b_size);
 	return bh;
 }
 EXPORT_SYMBOL(__getblk_gfp);
diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h
index 4a1c4fca475a..cc7b05eb024f 100644
--- a/include/linux/dma-mapping.h
+++ b/include/linux/dma-mapping.h
@@ -8,6 +8,7 @@
 #include <linux/err.h>
 #include <linux/dma-debug.h>
 #include <linux/dma-direction.h>
+#include <linux/kmsan-checks.h>
 #include <linux/scatterlist.h>
 #include <linux/bug.h>
 #include <linux/mem_encrypt.h>
@@ -281,6 +282,7 @@ static inline dma_addr_t dma_map_page_attrs(struct device *dev,
 	const struct dma_map_ops *ops = get_dma_ops(dev);
 	dma_addr_t addr;
 
+	kmsan_unpoison_shadow(page_address(page), size);
 	BUG_ON(!valid_dma_direction(dir));
 	if (dma_is_direct(ops))
 		addr = dma_direct_map_page(dev, page, offset, size, dir, attrs);
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 4351577b14d7..b8e9cd697248 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -11,6 +11,7 @@
 #define _LINUX_SKBUFF_H
 
 #include <linux/kernel.h>
+#include <linux/kmsan-checks.h>
 #include <linux/compiler.h>
 #include <linux/time.h>
 #include <linux/bug.h>
@@ -2214,6 +2215,8 @@ static inline void *skb_put_data(struct sk_buff *skb, const void *data,
 {
 	void *tmp = skb_put(skb, len);
 
+	/* Unpoison the data received from the network device. */
+	kmsan_unpoison_shadow(data, len);
 	memcpy(tmp, data, len);
 
 	return tmp;
@@ -2221,7 +2224,7 @@ static inline void *skb_put_data(struct sk_buff *skb, const void *data,
 
 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
 {
-	*(u8 *)skb_put(skb, 1) = val;
+	*(u8 *)skb_put(skb, 1) = KMSAN_INIT_VALUE(val);
 }
 
 void *skb_push(struct sk_buff *skb, unsigned int len);
diff --git a/mm/filemap.c b/mm/filemap.c
index 1146fcfa3215..8a17f15806cc 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -18,6 +18,7 @@
 #include <linux/uaccess.h>
 #include <linux/capability.h>
 #include <linux/kernel_stat.h>
+#include <linux/kmsan-checks.h>
 #include <linux/gfp.h>
 #include <linux/mm.h>
 #include <linux/swap.h>
@@ -2810,6 +2811,8 @@ static struct page *do_read_cache_page(struct address_space *mapping,
 		page = wait_on_page_read(page);
 		if (IS_ERR(page))
 			return page;
+		/* Assume all pages in page cache are initialized. */
+		kmsan_unpoison_shadow(page_address(page), PAGE_SIZE);
 		goto out;
 	}
 	if (PageUptodate(page))
diff --git a/mm/readahead.c b/mm/readahead.c
index 2fe72cd29b47..1fe61128095b 100644
--- a/mm/readahead.c
+++ b/mm/readahead.c
@@ -193,7 +193,13 @@ unsigned int __do_page_cache_readahead(struct address_space *mapping,
 			continue;
 		}
 
+		/*
+		 * The easiest way to handle these pages is to mark them
+		 * untracked by KMSAN, assuming they're never used by anything
+		 * else.
+		 */
 		page = __page_cache_alloc(gfp_mask);
+		kmsan_ignore_page(page, /*order*/0);
 		if (!page)
 			break;
 		page->index = page_offset;
diff --git a/net/9p/protocol.c b/net/9p/protocol.c
index 03593eb240d8..5d88ed181422 100644
--- a/net/9p/protocol.c
+++ b/net/9p/protocol.c
@@ -13,6 +13,7 @@
 #include <linux/module.h>
 #include <linux/errno.h>
 #include <linux/kernel.h>
+#include <linux/kmsan-checks.h>
 #include <linux/uaccess.h>
 #include <linux/slab.h>
 #include <linux/sched.h>
@@ -46,6 +47,7 @@ EXPORT_SYMBOL(p9stat_free);
 size_t pdu_read(struct p9_fcall *pdu, void *data, size_t size)
 {
 	size_t len = min(pdu->size - pdu->offset, size);
+	kmsan_unpoison_shadow(&pdu->sdata[pdu->offset], len);
 	memcpy(data, &pdu->sdata[pdu->offset], len);
 	pdu->offset += len;
 	return size - len;
diff --git a/sound/core/oss/pcm_oss.c b/sound/core/oss/pcm_oss.c
index f57c610d7523..c3795e495196 100644
--- a/sound/core/oss/pcm_oss.c
+++ b/sound/core/oss/pcm_oss.c
@@ -12,6 +12,7 @@
 #endif
 
 #include <linux/init.h>
+#include <linux/kmsan-checks.h>	/* for kmsan_unpoison_shadow() */
 #include <linux/slab.h>
 #include <linux/sched/signal.h>
 #include <linux/time.h>
@@ -1054,6 +1055,12 @@ static int snd_pcm_oss_change_params_locked(struct snd_pcm_substream *substream)
 		err = -ENOMEM;
 		goto failure;
 	}
+	/*
+	 * Unpoison the freshly created buffer to prevent KMSAN from reporting
+	 * uninit errors.
+	 * TODO(glider): unpoison it only when it's actually initialized.
+	 */
+	kmsan_unpoison_shadow(runtime->oss.buffer, runtime->oss.period_bytes);
 
 	runtime->oss.params = 0;
 	runtime->oss.prepare = 1;
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 24/26] kmsan: hooks for copy_to_user() and friends

[glider]

Memory that is copied from userspace must be unpoisoned.
Before copying memory to userspace, check it and report an error if it
contains uninitialized bits.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org
---

Change-Id: I38428b9c7d1909b8441dcec1749b080494a7af99
---
 arch/x86/include/asm/uaccess.h   | 12 ++++++++++++
 include/asm-generic/cacheflush.h |  7 ++++++-
 include/asm-generic/uaccess.h    | 12 ++++++++++--
 include/linux/uaccess.h          | 32 +++++++++++++++++++++++++++-----
 lib/iov_iter.c                   |  6 ++++++
 lib/usercopy.c                   |  6 +++++-
 6 files changed, 66 insertions(+), 9 deletions(-)

diff --git a/arch/x86/include/asm/uaccess.h b/arch/x86/include/asm/uaccess.h
index 61d93f062a36..ac4b26583f7c 100644
--- a/arch/x86/include/asm/uaccess.h
+++ b/arch/x86/include/asm/uaccess.h
@@ -6,6 +6,7 @@
  */
 #include <linux/compiler.h>
 #include <linux/kasan-checks.h>
+#include <linux/kmsan-checks.h>
 #include <linux/string.h>
 #include <asm/asm.h>
 #include <asm/page.h>
@@ -174,6 +175,7 @@ __typeof__(__builtin_choose_expr(sizeof(x) > sizeof(0UL), 0ULL, 0UL))
 			ASM_CALL_CONSTRAINT				\
 		     : "0" (ptr), "i" (sizeof(*(ptr))));		\
 	(x) = (__force __typeof__(*(ptr))) __val_gu;			\
+	kmsan_unpoison_shadow(&(x), sizeof(*(ptr)));			\
 	__builtin_expect(__ret_gu, 0);					\
 })
 
@@ -248,6 +250,7 @@ extern void __put_user_8(void);
 	__chk_user_ptr(ptr);					\
 	might_fault();						\
 	__pu_val = x;						\
+	kmsan_check_memory(&(__pu_val), sizeof(*(ptr)));	\
 	switch (sizeof(*(ptr))) {				\
 	case 1:							\
 		__put_user_x(1, __pu_val, ptr, __ret_pu);	\
@@ -270,7 +273,9 @@ extern void __put_user_8(void);
 
 #define __put_user_size(x, ptr, size, label)				\
 do {									\
+	__typeof__(*(ptr)) __pus_val = x;				\
 	__chk_user_ptr(ptr);						\
+	kmsan_check_memory(&(__pus_val), size);				\
 	switch (size) {							\
 	case 1:								\
 		__put_user_goto(x, ptr, "b", "b", "iq", label);	\
@@ -295,7 +300,10 @@ do {									\
  */
 #define __put_user_size_ex(x, ptr, size)				\
 do {									\
+	__typeof__(*(ptr)) __puse_val;					\
 	__chk_user_ptr(ptr);						\
+	__puse_val = x;							\
+	kmsan_check_memory(&(__puse_val), size);			\
 	switch (size) {							\
 	case 1:								\
 		__put_user_asm_ex(x, ptr, "b", "b", "iq");		\
@@ -363,6 +371,7 @@ do {									\
 	default:							\
 		(x) = __get_user_bad();					\
 	}								\
+	kmsan_unpoison_shadow(&(x), size);				\
 } while (0)
 
 #define __get_user_asm(x, addr, err, itype, rtype, ltype, errret)	\
@@ -413,6 +422,7 @@ do {									\
 	default:							\
 		(x) = __get_user_bad();					\
 	}								\
+	kmsan_unpoison_shadow(&(x), size);				\
 } while (0)
 
 #define __get_user_asm_ex(x, addr, itype, rtype, ltype)			\
@@ -428,11 +438,13 @@ do {									\
 #define __put_user_nocheck(x, ptr, size)			\
 ({								\
 	__label__ __pu_label;					\
+	__typeof__(*(ptr)) __pun_val = x;			\
 	int __pu_err = -EFAULT;					\
 	__typeof__(*(ptr)) __pu_val = (x);			\
 	__typeof__(ptr) __pu_ptr = (ptr);			\
 	__typeof__(size) __pu_size = (size);			\
 	__uaccess_begin();					\
+	kmsan_check_memory(&(__pun_val), size);			\
 	__put_user_size(__pu_val, __pu_ptr, __pu_size, __pu_label);	\
 	__pu_err = 0;						\
 __pu_label:							\
diff --git a/include/asm-generic/cacheflush.h b/include/asm-generic/cacheflush.h
index a950a22c4890..707531dccf5e 100644
--- a/include/asm-generic/cacheflush.h
+++ b/include/asm-generic/cacheflush.h
@@ -4,6 +4,7 @@
 
 /* Keep includes the same across arches.  */
 #include <linux/mm.h>
+#include <linux/kmsan-checks.h>
 
 #define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 0
 
@@ -72,10 +73,14 @@ static inline void flush_cache_vunmap(unsigned long start, unsigned long end)
 
 #define copy_to_user_page(vma, page, vaddr, dst, src, len) \
 	do { \
+		kmsan_check_memory(src, len); \
 		memcpy(dst, src, len); \
 		flush_icache_user_range(vma, page, vaddr, len); \
 	} while (0)
 #define copy_from_user_page(vma, page, vaddr, dst, src, len) \
-	memcpy(dst, src, len)
+	do { \
+		memcpy(dst, src, len); \
+		kmsan_unpoison_shadow(dst, len); \
+	} while (0)
 
 #endif /* __ASM_CACHEFLUSH_H */
diff --git a/include/asm-generic/uaccess.h b/include/asm-generic/uaccess.h
index e935318804f8..18a50333ffc0 100644
--- a/include/asm-generic/uaccess.h
+++ b/include/asm-generic/uaccess.h
@@ -142,7 +142,11 @@ static inline int __access_ok(unsigned long addr, unsigned long size)
 
 static inline int __put_user_fn(size_t size, void __user *ptr, void *x)
 {
-	return unlikely(raw_copy_to_user(ptr, x, size)) ? -EFAULT : 0;
+	int n;
+
+	n = raw_copy_to_user(ptr, x, size);
+	kmsan_copy_to_user(ptr, x, size, n);
+	return unlikely(n) ? -EFAULT : 0;
 }
 
 #define __put_user_fn(sz, u, k)	__put_user_fn(sz, u, k)
@@ -203,7 +207,11 @@ extern int __put_user_bad(void) __attribute__((noreturn));
 #ifndef __get_user_fn
 static inline int __get_user_fn(size_t size, const void __user *ptr, void *x)
 {
-	return unlikely(raw_copy_from_user(x, ptr, size)) ? -EFAULT : 0;
+	int copied, to_copy = size;
+
+	copied = raw_copy_from_user(x, ptr, size);
+	kmsan_unpoison_shadow(to, to_copy - res);
+	return unlikely(copied) ? -EFAULT : 0;
 }
 
 #define __get_user_fn(sz, u, k)	__get_user_fn(sz, u, k)
diff --git a/include/linux/uaccess.h b/include/linux/uaccess.h
index d4ee6e942562..4257b5f80689 100644
--- a/include/linux/uaccess.h
+++ b/include/linux/uaccess.h
@@ -5,6 +5,7 @@
 #include <linux/sched.h>
 #include <linux/thread_info.h>
 #include <linux/kasan-checks.h>
+#include <linux/kmsan-checks.h>
 
 #define uaccess_kernel() segment_eq(get_fs(), KERNEL_DS)
 
@@ -58,18 +59,26 @@
 static __always_inline __must_check unsigned long
 __copy_from_user_inatomic(void *to, const void __user *from, unsigned long n)
 {
+	unsigned long to_copy = n;
+
 	kasan_check_write(to, n);
 	check_object_size(to, n, false);
-	return raw_copy_from_user(to, from, n);
+	n = raw_copy_from_user(to, from, n);
+	kmsan_unpoison_shadow(to, to_copy - n);
+	return n;
 }
 
 static __always_inline __must_check unsigned long
 __copy_from_user(void *to, const void __user *from, unsigned long n)
 {
+	unsigned long to_copy = n;
+
 	might_fault();
 	kasan_check_write(to, n);
 	check_object_size(to, n, false);
-	return raw_copy_from_user(to, from, n);
+	n = raw_copy_from_user(to, from, n);
+	kmsan_unpoison_shadow(to, to_copy - n);
+	return n;
 }
 
 /**
@@ -88,29 +97,39 @@ __copy_from_user(void *to, const void __user *from, unsigned long n)
 static __always_inline __must_check unsigned long
 __copy_to_user_inatomic(void __user *to, const void *from, unsigned long n)
 {
+	unsigned long to_copy = n;
+
 	kasan_check_read(from, n);
 	check_object_size(from, n, true);
-	return raw_copy_to_user(to, from, n);
+	n = raw_copy_to_user(to, from, n);
+	kmsan_copy_to_user(to, from, to_copy, n);
+	return n;
 }
 
 static __always_inline __must_check unsigned long
 __copy_to_user(void __user *to, const void *from, unsigned long n)
 {
+	unsigned long to_copy = n;
+
 	might_fault();
 	kasan_check_read(from, n);
 	check_object_size(from, n, true);
-	return raw_copy_to_user(to, from, n);
+	n = raw_copy_to_user(to, from, n);
+	kmsan_copy_to_user(to, from, to_copy, n);
+	return n;
 }
 
 #ifdef INLINE_COPY_FROM_USER
 static inline __must_check unsigned long
 _copy_from_user(void *to, const void __user *from, unsigned long n)
 {
-	unsigned long res = n;
+	unsigned long res = n, to_copy = n;
+
 	might_fault();
 	if (likely(access_ok(from, n))) {
 		kasan_check_write(to, n);
 		res = raw_copy_from_user(to, from, n);
+		kmsan_unpoison_shadow(to, to_copy - res);
 	}
 	if (unlikely(res))
 		memset(to + (n - res), 0, res);
@@ -125,10 +144,13 @@ _copy_from_user(void *, const void __user *, unsigned long);
 static inline __must_check unsigned long
 _copy_to_user(void __user *to, const void *from, unsigned long n)
 {
+	unsigned long to_copy = n;
+
 	might_fault();
 	if (access_ok(to, n)) {
 		kasan_check_read(from, n);
 		n = raw_copy_to_user(to, from, n);
+		kmsan_copy_to_user(to, from, to_copy, n);
 	}
 	return n;
 }
diff --git a/lib/iov_iter.c b/lib/iov_iter.c
index 639d5e7014c1..f038676068b2 100644
--- a/lib/iov_iter.c
+++ b/lib/iov_iter.c
@@ -137,18 +137,24 @@
 
 static int copyout(void __user *to, const void *from, size_t n)
 {
+	size_t to_copy = n;
+
 	if (access_ok(to, n)) {
 		kasan_check_read(from, n);
 		n = raw_copy_to_user(to, from, n);
+		kmsan_copy_to_user(to, from, to_copy, n);
 	}
 	return n;
 }
 
 static int copyin(void *to, const void __user *from, size_t n)
 {
+	size_t to_copy = n;
+
 	if (access_ok(from, n)) {
 		kasan_check_write(to, n);
 		n = raw_copy_from_user(to, from, n);
+		kmsan_unpoison_shadow(to, to_copy - n);
 	}
 	return n;
 }
diff --git a/lib/usercopy.c b/lib/usercopy.c
index cbb4d9ec00f2..abfd93edecba 100644
--- a/lib/usercopy.c
+++ b/lib/usercopy.c
@@ -1,4 +1,5 @@
 // SPDX-License-Identifier: GPL-2.0
+#include <linux/kmsan-checks.h>
 #include <linux/uaccess.h>
 #include <linux/bitops.h>
 
@@ -7,11 +8,12 @@
 #ifndef INLINE_COPY_FROM_USER
 unsigned long _copy_from_user(void *to, const void __user *from, unsigned long n)
 {
-	unsigned long res = n;
+	unsigned long res = n, to_copy = n;
 	might_fault();
 	if (likely(access_ok(from, n))) {
 		kasan_check_write(to, n);
 		res = raw_copy_from_user(to, from, n);
+		kmsan_unpoison_shadow(to, to_copy - res);
 	}
 	if (unlikely(res))
 		memset(to + (n - res), 0, res);
@@ -23,10 +25,12 @@ EXPORT_SYMBOL(_copy_from_user);
 #ifndef INLINE_COPY_TO_USER
 unsigned long _copy_to_user(void __user *to, const void *from, unsigned long n)
 {
+	unsigned long to_copy = n;
 	might_fault();
 	if (likely(access_ok(to, n))) {
 		kasan_check_read(from, n);
 		n = raw_copy_to_user(to, from, n);
+		kmsan_copy_to_user(to, from, to_copy, n);
 	}
 	return n;
 }
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 25/26] kmsan: disable strscpy() optimization under KMSAN

[glider]

Disable the efficient 8-byte reading under KMSAN to avoid false positives.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org

---

Change-Id: I25d1acf5c3df6eff85894cd94f5ddbe93308271c
---
 lib/string.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/lib/string.c b/lib/string.c
index 08ec58cc673b..15efdc51bda6 100644
--- a/lib/string.c
+++ b/lib/string.c
@@ -186,7 +186,10 @@ ssize_t strscpy(char *dest, const char *src, size_t count)
 	if (count == 0 || WARN_ON_ONCE(count > INT_MAX))
 		return -E2BIG;
 
-#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+/**
+ * Disable the efficient 8-byte reading under KMSAN to avoid false positives.
+ */
+#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && !defined(CONFIG_KMSAN)
 	/*
 	 * If src is unaligned, don't cross a page boundary,
 	 * since we don't know if the next page is mapped.
-- 
2.23.0.866.gb869b98d4c-goog

[PATCH RFC v1 26/26] net: kasan: kmsan: support CONFIG_GENERIC_CSUM on x86, enable it for KASAN/KMSAN

[glider]

This is needed to allow memory tools like KASAN and KMSAN see the
memory accesses from the checksum code. Without CONFIG_GENERIC_CSUM the
tools can't see memory accesses originating from handwritten assembly
code.
For KASAN it's a question of detecting more bugs, for KMSAN using the C
implementation also helps avoid false positives originating from
seemingly uninitialized checksum values.

Signed-off-by: Alexander Potapenko <glider@google.com>
To: Alexander Potapenko <glider@google.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: linux-mm@kvack.org
---

Change-Id: I645e2c097253a8d5717ad87e2e2df6f6f67251f3
---
 arch/x86/Kconfig                |  5 +++++
 arch/x86/include/asm/checksum.h | 10 +++++++---
 arch/x86/lib/Makefile           |  2 ++
 3 files changed, 14 insertions(+), 3 deletions(-)

diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 6d2b0fb3a871..dfec655d1a7a 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -272,6 +272,11 @@ config GENERIC_ISA_DMA
 	def_bool y
 	depends on ISA_DMA_API
 
+config GENERIC_CSUM
+	bool
+	default y if KMSAN || KASAN
+	default n
+
 config GENERIC_BUG
 	def_bool y
 	depends on BUG
diff --git a/arch/x86/include/asm/checksum.h b/arch/x86/include/asm/checksum.h
index d79d1e622dcf..ab3464cbce26 100644
--- a/arch/x86/include/asm/checksum.h
+++ b/arch/x86/include/asm/checksum.h
@@ -1,6 +1,10 @@
 /* SPDX-License-Identifier: GPL-2.0 */
-#ifdef CONFIG_X86_32
-# include <asm/checksum_32.h>
+#ifdef CONFIG_GENERIC_CSUM
+# include <asm-generic/checksum.h>
 #else
-# include <asm/checksum_64.h>
+# ifdef CONFIG_X86_32
+#  include <asm/checksum_32.h>
+# else
+#  include <asm/checksum_64.h>
+# endif
 #endif
diff --git a/arch/x86/lib/Makefile b/arch/x86/lib/Makefile
index 5246db42de45..bca9031de9ff 100644
--- a/arch/x86/lib/Makefile
+++ b/arch/x86/lib/Makefile
@@ -55,7 +55,9 @@ endif
         lib-$(CONFIG_X86_USE_3DNOW) += mmx_32.o
 else
         obj-y += iomap_copy_64.o
+ifneq ($(CONFIG_GENERIC_CSUM),y)
         lib-y += csum-partial_64.o csum-copy_64.o csum-wrappers_64.o
+endif
         lib-y += clear_page_64.o copy_page_64.o
         lib-y += memmove_64.o memset_64.o
         lib-y += copy_user_64.o
-- 
2.23.0.866.gb869b98d4c-goog

Re: [PATCH RFC v1 18/26] kmsan: mm: call KMSAN hooks from SLUB code

[Qian Cai]

On Fri, 2019-10-18 at 11:42 +0200, glider@google.com wrote:
> In order to report uninitialized memory coming from heap allocations
> KMSAN has to poison them unless they're created with __GFP_ZERO.
> 
> It's handy that we need KMSAN hooks in the places where
> init_on_alloc/init_on_free initialization is performed.

Well, there is SLUB debug which has red zoning and poisoning checks. What's
value of this patch?

> 
> Signed-off-by: Alexander Potapenko <glider@google.com>
> To: Alexander Potapenko <glider@google.com>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: Vegard Nossum <vegard.nossum@oracle.com>
> Cc: Dmitry Vyukov <dvyukov@google.com>
> Cc: linux-mm@kvack.org
> ---
> 
> Change-Id: I51103b7981d3aabed747d0c85cbdc85568665871
> ---
>  mm/slub.c | 37 +++++++++++++++++++++++++++++++------
>  1 file changed, 31 insertions(+), 6 deletions(-)
> 
> diff --git a/mm/slub.c b/mm/slub.c
> index 3d63ae320d31..3d6d4c63446e 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -21,6 +21,8 @@
>  #include <linux/proc_fs.h>
>  #include <linux/seq_file.h>
>  #include <linux/kasan.h>
> +#include <linux/kmsan.h>
> +#include <linux/kmsan-checks.h> /* KMSAN_INIT_VALUE */
>  #include <linux/cpu.h>
>  #include <linux/cpuset.h>
>  #include <linux/mempolicy.h>
> @@ -285,17 +287,27 @@ static void prefetch_freepointer(const struct kmem_cache *s, void *object)
>  	prefetch(object + s->offset);
>  }
>  
> +/*
> + * When running under KMSAN, get_freepointer_safe() may return an uninitialized
> + * pointer value in the case the current thread loses the race for the next
> + * memory chunk in the freelist. In that case this_cpu_cmpxchg_double() in
> + * slab_alloc_node() will fail, so the uninitialized value won't be used, but
> + * KMSAN will still check all arguments of cmpxchg because of imperfect
> + * handling of inline assembly.
> + * To work around this problem, use KMSAN_INIT_VALUE() to force initialize the
> + * return value of get_freepointer_safe().
> + */
>  static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
>  {
>  	unsigned long freepointer_addr;
>  	void *p;
>  
>  	if (!debug_pagealloc_enabled())
> -		return get_freepointer(s, object);
> +		return KMSAN_INIT_VALUE(get_freepointer(s, object));
>  
>  	freepointer_addr = (unsigned long)object + s->offset;
>  	probe_kernel_read(&p, (void **)freepointer_addr, sizeof(p));
> -	return freelist_ptr(s, p, freepointer_addr);
> +	return KMSAN_INIT_VALUE(freelist_ptr(s, p, freepointer_addr));
>  }
>  
>  static inline void set_freepointer(struct kmem_cache *s, void *object, void *fp)
> @@ -1390,6 +1402,7 @@ static inline void *kmalloc_large_node_hook(void *ptr, size_t size, gfp_t flags)
>  	ptr = kasan_kmalloc_large(ptr, size, flags);
>  	/* As ptr might get tagged, call kmemleak hook after KASAN. */
>  	kmemleak_alloc(ptr, size, 1, flags);
> +	kmsan_kmalloc_large(ptr, size, flags);
>  	return ptr;
>  }
>  
> @@ -1397,6 +1410,7 @@ static __always_inline void kfree_hook(void *x)
>  {
>  	kmemleak_free(x);
>  	kasan_kfree_large(x, _RET_IP_);
> +	kmsan_kfree_large(x);
>  }
>  
>  static __always_inline bool slab_free_hook(struct kmem_cache *s, void *x)
> @@ -1453,6 +1467,12 @@ static inline bool slab_free_freelist_hook(struct kmem_cache *s,
>  		} while (object != old_tail);
>  	}
>  
> +	do {
> +		object = next;
> +		next = get_freepointer(s, object);
> +		kmsan_slab_free(s, object);
> +	} while (object != old_tail);
> +
>  /*
>   * Compiler cannot detect this function can be removed if slab_free_hook()
>   * evaluates to nothing.  Thus, catch all relevant config debug options here.
> @@ -2769,6 +2789,7 @@ static __always_inline void *slab_alloc_node(struct kmem_cache *s,
>  	if (unlikely(slab_want_init_on_alloc(gfpflags, s)) && object)
>  		memset(object, 0, s->object_size);
>  
> +	kmsan_slab_alloc(s, object, gfpflags);
>  	slab_post_alloc_hook(s, gfpflags, 1, &object);
>  
>  	return object;
> @@ -2797,6 +2818,7 @@ void *kmem_cache_alloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size)
>  	void *ret = slab_alloc(s, gfpflags, _RET_IP_);
>  	trace_kmalloc(_RET_IP_, ret, size, s->size, gfpflags);
>  	ret = kasan_kmalloc(s, ret, size, gfpflags);
> +
>  	return ret;
>  }
>  EXPORT_SYMBOL(kmem_cache_alloc_trace);
> @@ -2809,7 +2831,6 @@ void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t gfpflags, int node)
>  
>  	trace_kmem_cache_alloc_node(_RET_IP_, ret,
>  				    s->object_size, s->size, gfpflags, node);
> -
>  	return ret;
>  }
>  EXPORT_SYMBOL(kmem_cache_alloc_node);
> @@ -2825,6 +2846,7 @@ void *kmem_cache_alloc_node_trace(struct kmem_cache *s,
>  			   size, s->size, gfpflags, node);
>  
>  	ret = kasan_kmalloc(s, ret, size, gfpflags);
> +
>  	return ret;
>  }
>  EXPORT_SYMBOL(kmem_cache_alloc_node_trace);
> @@ -3150,7 +3172,7 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
>  			  void **p)
>  {
>  	struct kmem_cache_cpu *c;
> -	int i;
> +	int i, j;
>  
>  	/* memcg and kmem_cache debug support */
>  	s = slab_pre_alloc_hook(s, flags);
> @@ -3188,11 +3210,11 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
>  
>  	/* Clear memory outside IRQ disabled fastpath loop */
>  	if (unlikely(slab_want_init_on_alloc(flags, s))) {
> -		int j;
> -
>  		for (j = 0; j < i; j++)
>  			memset(p[j], 0, s->object_size);
>  	}
> +	for (j = 0; j < i; j++)
> +		kmsan_slab_alloc(s, p[j], flags);
>  
>  	/* memcg and kmem_cache debug support */
>  	slab_post_alloc_hook(s, flags, size, p);
> @@ -3793,6 +3815,7 @@ static int __init setup_slub_min_objects(char *str)
>  
>  __setup("slub_min_objects=", setup_slub_min_objects);
>  
> +__no_sanitize_memory
>  void *__kmalloc(size_t size, gfp_t flags)
>  {
>  	struct kmem_cache *s;
> @@ -5698,6 +5721,7 @@ static char *create_unique_id(struct kmem_cache *s)
>  	p += sprintf(p, "%07u", s->size);
>  
>  	BUG_ON(p > name + ID_STR_LENGTH - 1);
> +	kmsan_unpoison_shadow(name, p - name);
>  	return name;
>  }
>  
> @@ -5847,6 +5871,7 @@ static int sysfs_slab_alias(struct kmem_cache *s, const char *name)
>  	al->name = name;
>  	al->next = alias_list;
>  	alias_list = al;
> +	kmsan_unpoison_shadow(al, sizeof(struct saved_alias));
>  	return 0;
>  }
>  

Re: [PATCH RFC v1 18/26] kmsan: mm: call KMSAN hooks from SLUB code

[Alexander Potapenko]

On Fri, Oct 18, 2019 at 3:22 PM Qian Cai <cai@lca.pw> wrote:
>
> On Fri, 2019-10-18 at 11:42 +0200, glider@google.com wrote:
> > In order to report uninitialized memory coming from heap allocations
> > KMSAN has to poison them unless they're created with __GFP_ZERO.
> >
> > It's handy that we need KMSAN hooks in the places where
> > init_on_alloc/init_on_free initialization is performed.
>
> Well, there is SLUB debug which has red zoning and poisoning checks. What's
> value of this patch?
Sorry, are you talking about the whole patchset or just this patch?
Note that SLUB debug is unable to detect uninitialized values with
bit-to-bit precision, neither have I heard of anyone using it for
detecting uses of uninitialized memory in the kernel at all.
The purpose of SLUB debug is to detect corruptions of freed memory.
> >
> > Signed-off-by: Alexander Potapenko <glider@google.com>
> > To: Alexander Potapenko <glider@google.com>
> > Cc: Andrew Morton <akpm@linux-foundation.org>
> > Cc: Vegard Nossum <vegard.nossum@oracle.com>
> > Cc: Dmitry Vyukov <dvyukov@google.com>
> > Cc: linux-mm@kvack.org
> > ---
> >
> > Change-Id: I51103b7981d3aabed747d0c85cbdc85568665871
> > ---
> >  mm/slub.c | 37 +++++++++++++++++++++++++++++++------
> >  1 file changed, 31 insertions(+), 6 deletions(-)
> >
> > diff --git a/mm/slub.c b/mm/slub.c
> > index 3d63ae320d31..3d6d4c63446e 100644
> > --- a/mm/slub.c
> > +++ b/mm/slub.c
> > @@ -21,6 +21,8 @@
> >  #include <linux/proc_fs.h>
> >  #include <linux/seq_file.h>
> >  #include <linux/kasan.h>
> > +#include <linux/kmsan.h>
> > +#include <linux/kmsan-checks.h> /* KMSAN_INIT_VALUE */
> >  #include <linux/cpu.h>
> >  #include <linux/cpuset.h>
> >  #include <linux/mempolicy.h>
> > @@ -285,17 +287,27 @@ static void prefetch_freepointer(const struct kmem_cache *s, void *object)
> >       prefetch(object + s->offset);
> >  }
> >
> > +/*
> > + * When running under KMSAN, get_freepointer_safe() may return an uninitialized
> > + * pointer value in the case the current thread loses the race for the next
> > + * memory chunk in the freelist. In that case this_cpu_cmpxchg_double() in
> > + * slab_alloc_node() will fail, so the uninitialized value won't be used, but
> > + * KMSAN will still check all arguments of cmpxchg because of imperfect
> > + * handling of inline assembly.
> > + * To work around this problem, use KMSAN_INIT_VALUE() to force initialize the
> > + * return value of get_freepointer_safe().
> > + */
> >  static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
> >  {
> >       unsigned long freepointer_addr;
> >       void *p;
> >
> >       if (!debug_pagealloc_enabled())
> > -             return get_freepointer(s, object);
> > +             return KMSAN_INIT_VALUE(get_freepointer(s, object));
> >
> >       freepointer_addr = (unsigned long)object + s->offset;
> >       probe_kernel_read(&p, (void **)freepointer_addr, sizeof(p));
> > -     return freelist_ptr(s, p, freepointer_addr);
> > +     return KMSAN_INIT_VALUE(freelist_ptr(s, p, freepointer_addr));
> >  }
> >
> >  static inline void set_freepointer(struct kmem_cache *s, void *object, void *fp)
> > @@ -1390,6 +1402,7 @@ static inline void *kmalloc_large_node_hook(void *ptr, size_t size, gfp_t flags)
> >       ptr = kasan_kmalloc_large(ptr, size, flags);
> >       /* As ptr might get tagged, call kmemleak hook after KASAN. */
> >       kmemleak_alloc(ptr, size, 1, flags);
> > +     kmsan_kmalloc_large(ptr, size, flags);
> >       return ptr;
> >  }
> >
> > @@ -1397,6 +1410,7 @@ static __always_inline void kfree_hook(void *x)
> >  {
> >       kmemleak_free(x);
> >       kasan_kfree_large(x, _RET_IP_);
> > +     kmsan_kfree_large(x);
> >  }
> >
> >  static __always_inline bool slab_free_hook(struct kmem_cache *s, void *x)
> > @@ -1453,6 +1467,12 @@ static inline bool slab_free_freelist_hook(struct kmem_cache *s,
> >               } while (object != old_tail);
> >       }
> >
> > +     do {
> > +             object = next;
> > +             next = get_freepointer(s, object);
> > +             kmsan_slab_free(s, object);
> > +     } while (object != old_tail);
> > +
> >  /*
> >   * Compiler cannot detect this function can be removed if slab_free_hook()
> >   * evaluates to nothing.  Thus, catch all relevant config debug options here.
> > @@ -2769,6 +2789,7 @@ static __always_inline void *slab_alloc_node(struct kmem_cache *s,
> >       if (unlikely(slab_want_init_on_alloc(gfpflags, s)) && object)
> >               memset(object, 0, s->object_size);
> >
> > +     kmsan_slab_alloc(s, object, gfpflags);
> >       slab_post_alloc_hook(s, gfpflags, 1, &object);
> >
> >       return object;
> > @@ -2797,6 +2818,7 @@ void *kmem_cache_alloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size)
> >       void *ret = slab_alloc(s, gfpflags, _RET_IP_);
> >       trace_kmalloc(_RET_IP_, ret, size, s->size, gfpflags);
> >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > +
> >       return ret;
> >  }
> >  EXPORT_SYMBOL(kmem_cache_alloc_trace);
> > @@ -2809,7 +2831,6 @@ void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t gfpflags, int node)
> >
> >       trace_kmem_cache_alloc_node(_RET_IP_, ret,
> >                                   s->object_size, s->size, gfpflags, node);
> > -
> >       return ret;
> >  }
> >  EXPORT_SYMBOL(kmem_cache_alloc_node);
> > @@ -2825,6 +2846,7 @@ void *kmem_cache_alloc_node_trace(struct kmem_cache *s,
> >                          size, s->size, gfpflags, node);
> >
> >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > +
> >       return ret;
> >  }
> >  EXPORT_SYMBOL(kmem_cache_alloc_node_trace);
> > @@ -3150,7 +3172,7 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> >                         void **p)
> >  {
> >       struct kmem_cache_cpu *c;
> > -     int i;
> > +     int i, j;
> >
> >       /* memcg and kmem_cache debug support */
> >       s = slab_pre_alloc_hook(s, flags);
> > @@ -3188,11 +3210,11 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> >
> >       /* Clear memory outside IRQ disabled fastpath loop */
> >       if (unlikely(slab_want_init_on_alloc(flags, s))) {
> > -             int j;
> > -
> >               for (j = 0; j < i; j++)
> >                       memset(p[j], 0, s->object_size);
> >       }
> > +     for (j = 0; j < i; j++)
> > +             kmsan_slab_alloc(s, p[j], flags);
> >
> >       /* memcg and kmem_cache debug support */
> >       slab_post_alloc_hook(s, flags, size, p);
> > @@ -3793,6 +3815,7 @@ static int __init setup_slub_min_objects(char *str)
> >
> >  __setup("slub_min_objects=", setup_slub_min_objects);
> >
> > +__no_sanitize_memory
> >  void *__kmalloc(size_t size, gfp_t flags)
> >  {
> >       struct kmem_cache *s;
> > @@ -5698,6 +5721,7 @@ static char *create_unique_id(struct kmem_cache *s)
> >       p += sprintf(p, "%07u", s->size);
> >
> >       BUG_ON(p > name + ID_STR_LENGTH - 1);
> > +     kmsan_unpoison_shadow(name, p - name);
> >       return name;
> >  }
> >
> > @@ -5847,6 +5871,7 @@ static int sysfs_slab_alias(struct kmem_cache *s, const char *name)
> >       al->name = name;
> >       al->next = alias_list;
> >       alias_list = al;
> > +     kmsan_unpoison_shadow(al, sizeof(struct saved_alias));
> >       return 0;
> >  }
> >



-- 
Alexander Potapenko
Software Engineer

Google Germany GmbH
Erika-Mann-Straße, 33
80636 München

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg

Re: [PATCH RFC v1 18/26] kmsan: mm: call KMSAN hooks from SLUB code

[Qian Cai]

On Fri, 2019-10-18 at 15:33 +0200, Alexander Potapenko wrote:
> On Fri, Oct 18, 2019 at 3:22 PM Qian Cai <cai@lca.pw> wrote:
> > 
> > On Fri, 2019-10-18 at 11:42 +0200, glider@google.com wrote:
> > > In order to report uninitialized memory coming from heap allocations
> > > KMSAN has to poison them unless they're created with __GFP_ZERO.
> > > 
> > > It's handy that we need KMSAN hooks in the places where
> > > init_on_alloc/init_on_free initialization is performed.
> > 
> > Well, there is SLUB debug which has red zoning and poisoning checks. What's
> > value of this patch?
> 
> Sorry, are you talking about the whole patchset or just this patch?

Just this patch.

> Note that SLUB debug is unable to detect uninitialized values with
> bit-to-bit precision, neither have I heard of anyone using it for
> detecting uses of uninitialized memory in the kernel at all.
> The purpose of SLUB debug is to detect corruptions of freed memory.

The point is if developers have SLUB debug enabled, all the free objects will be
poisoned, so what's the point of checking uninitialized memory there?

> > > 
> > > Signed-off-by: Alexander Potapenko <glider@google.com>
> > > To: Alexander Potapenko <glider@google.com>
> > > Cc: Andrew Morton <akpm@linux-foundation.org>
> > > Cc: Vegard Nossum <vegard.nossum@oracle.com>
> > > Cc: Dmitry Vyukov <dvyukov@google.com>
> > > Cc: linux-mm@kvack.org
> > > ---
> > > 
> > > Change-Id: I51103b7981d3aabed747d0c85cbdc85568665871
> > > ---
> > >  mm/slub.c | 37 +++++++++++++++++++++++++++++++------
> > >  1 file changed, 31 insertions(+), 6 deletions(-)
> > > 
> > > diff --git a/mm/slub.c b/mm/slub.c
> > > index 3d63ae320d31..3d6d4c63446e 100644
> > > --- a/mm/slub.c
> > > +++ b/mm/slub.c
> > > @@ -21,6 +21,8 @@
> > >  #include <linux/proc_fs.h>
> > >  #include <linux/seq_file.h>
> > >  #include <linux/kasan.h>
> > > +#include <linux/kmsan.h>
> > > +#include <linux/kmsan-checks.h> /* KMSAN_INIT_VALUE */
> > >  #include <linux/cpu.h>
> > >  #include <linux/cpuset.h>
> > >  #include <linux/mempolicy.h>
> > > @@ -285,17 +287,27 @@ static void prefetch_freepointer(const struct kmem_cache *s, void *object)
> > >       prefetch(object + s->offset);
> > >  }
> > > 
> > > +/*
> > > + * When running under KMSAN, get_freepointer_safe() may return an uninitialized
> > > + * pointer value in the case the current thread loses the race for the next
> > > + * memory chunk in the freelist. In that case this_cpu_cmpxchg_double() in
> > > + * slab_alloc_node() will fail, so the uninitialized value won't be used, but
> > > + * KMSAN will still check all arguments of cmpxchg because of imperfect
> > > + * handling of inline assembly.
> > > + * To work around this problem, use KMSAN_INIT_VALUE() to force initialize the
> > > + * return value of get_freepointer_safe().
> > > + */
> > >  static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
> > >  {
> > >       unsigned long freepointer_addr;
> > >       void *p;
> > > 
> > >       if (!debug_pagealloc_enabled())
> > > -             return get_freepointer(s, object);
> > > +             return KMSAN_INIT_VALUE(get_freepointer(s, object));
> > > 
> > >       freepointer_addr = (unsigned long)object + s->offset;
> > >       probe_kernel_read(&p, (void **)freepointer_addr, sizeof(p));
> > > -     return freelist_ptr(s, p, freepointer_addr);
> > > +     return KMSAN_INIT_VALUE(freelist_ptr(s, p, freepointer_addr));
> > >  }
> > > 
> > >  static inline void set_freepointer(struct kmem_cache *s, void *object, void *fp)
> > > @@ -1390,6 +1402,7 @@ static inline void *kmalloc_large_node_hook(void *ptr, size_t size, gfp_t flags)
> > >       ptr = kasan_kmalloc_large(ptr, size, flags);
> > >       /* As ptr might get tagged, call kmemleak hook after KASAN. */
> > >       kmemleak_alloc(ptr, size, 1, flags);
> > > +     kmsan_kmalloc_large(ptr, size, flags);
> > >       return ptr;
> > >  }
> > > 
> > > @@ -1397,6 +1410,7 @@ static __always_inline void kfree_hook(void *x)
> > >  {
> > >       kmemleak_free(x);
> > >       kasan_kfree_large(x, _RET_IP_);
> > > +     kmsan_kfree_large(x);
> > >  }
> > > 
> > >  static __always_inline bool slab_free_hook(struct kmem_cache *s, void *x)
> > > @@ -1453,6 +1467,12 @@ static inline bool slab_free_freelist_hook(struct kmem_cache *s,
> > >               } while (object != old_tail);
> > >       }
> > > 
> > > +     do {
> > > +             object = next;
> > > +             next = get_freepointer(s, object);
> > > +             kmsan_slab_free(s, object);
> > > +     } while (object != old_tail);
> > > +
> > >  /*
> > >   * Compiler cannot detect this function can be removed if slab_free_hook()
> > >   * evaluates to nothing.  Thus, catch all relevant config debug options here.
> > > @@ -2769,6 +2789,7 @@ static __always_inline void *slab_alloc_node(struct kmem_cache *s,
> > >       if (unlikely(slab_want_init_on_alloc(gfpflags, s)) && object)
> > >               memset(object, 0, s->object_size);
> > > 
> > > +     kmsan_slab_alloc(s, object, gfpflags);
> > >       slab_post_alloc_hook(s, gfpflags, 1, &object);
> > > 
> > >       return object;
> > > @@ -2797,6 +2818,7 @@ void *kmem_cache_alloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size)
> > >       void *ret = slab_alloc(s, gfpflags, _RET_IP_);
> > >       trace_kmalloc(_RET_IP_, ret, size, s->size, gfpflags);
> > >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > > +
> > >       return ret;
> > >  }
> > >  EXPORT_SYMBOL(kmem_cache_alloc_trace);
> > > @@ -2809,7 +2831,6 @@ void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t gfpflags, int node)
> > > 
> > >       trace_kmem_cache_alloc_node(_RET_IP_, ret,
> > >                                   s->object_size, s->size, gfpflags, node);
> > > -
> > >       return ret;
> > >  }
> > >  EXPORT_SYMBOL(kmem_cache_alloc_node);
> > > @@ -2825,6 +2846,7 @@ void *kmem_cache_alloc_node_trace(struct kmem_cache *s,
> > >                          size, s->size, gfpflags, node);
> > > 
> > >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > > +
> > >       return ret;
> > >  }
> > >  EXPORT_SYMBOL(kmem_cache_alloc_node_trace);
> > > @@ -3150,7 +3172,7 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> > >                         void **p)
> > >  {
> > >       struct kmem_cache_cpu *c;
> > > -     int i;
> > > +     int i, j;
> > > 
> > >       /* memcg and kmem_cache debug support */
> > >       s = slab_pre_alloc_hook(s, flags);
> > > @@ -3188,11 +3210,11 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> > > 
> > >       /* Clear memory outside IRQ disabled fastpath loop */
> > >       if (unlikely(slab_want_init_on_alloc(flags, s))) {
> > > -             int j;
> > > -
> > >               for (j = 0; j < i; j++)
> > >                       memset(p[j], 0, s->object_size);
> > >       }
> > > +     for (j = 0; j < i; j++)
> > > +             kmsan_slab_alloc(s, p[j], flags);
> > > 
> > >       /* memcg and kmem_cache debug support */
> > >       slab_post_alloc_hook(s, flags, size, p);
> > > @@ -3793,6 +3815,7 @@ static int __init setup_slub_min_objects(char *str)
> > > 
> > >  __setup("slub_min_objects=", setup_slub_min_objects);
> > > 
> > > +__no_sanitize_memory
> > >  void *__kmalloc(size_t size, gfp_t flags)
> > >  {
> > >       struct kmem_cache *s;
> > > @@ -5698,6 +5721,7 @@ static char *create_unique_id(struct kmem_cache *s)
> > >       p += sprintf(p, "%07u", s->size);
> > > 
> > >       BUG_ON(p > name + ID_STR_LENGTH - 1);
> > > +     kmsan_unpoison_shadow(name, p - name);
> > >       return name;
> > >  }
> > > 
> > > @@ -5847,6 +5871,7 @@ static int sysfs_slab_alias(struct kmem_cache *s, const char *name)
> > >       al->name = name;
> > >       al->next = alias_list;
> > >       alias_list = al;
> > > +     kmsan_unpoison_shadow(al, sizeof(struct saved_alias));
> > >       return 0;
> > >  }
> > > 
> 
> 
> 

Re: [PATCH RFC v1 18/26] kmsan: mm: call KMSAN hooks from SLUB code

[Alexander Potapenko]

On Fri, Oct 18, 2019 at 3:42 PM Qian Cai <cai@lca.pw> wrote:
>
> On Fri, 2019-10-18 at 15:33 +0200, Alexander Potapenko wrote:
> > On Fri, Oct 18, 2019 at 3:22 PM Qian Cai <cai@lca.pw> wrote:
> > >
> > > On Fri, 2019-10-18 at 11:42 +0200, glider@google.com wrote:
> > > > In order to report uninitialized memory coming from heap allocations
> > > > KMSAN has to poison them unless they're created with __GFP_ZERO.
> > > >
> > > > It's handy that we need KMSAN hooks in the places where
> > > > init_on_alloc/init_on_free initialization is performed.
> > >
> > > Well, there is SLUB debug which has red zoning and poisoning checks. What's
> > > value of this patch?
> >
> > Sorry, are you talking about the whole patchset or just this patch?
>
> Just this patch.
>
> > Note that SLUB debug is unable to detect uninitialized values with
> > bit-to-bit precision, neither have I heard of anyone using it for
> > detecting uses of uninitialized memory in the kernel at all.
> > The purpose of SLUB debug is to detect corruptions of freed memory.
>
> The point is if developers have SLUB debug enabled, all the free objects will be
> poisoned, so what's the point of checking uninitialized memory there?
You are right, SLUB debug has to be handled separately. If I'm
understanding correctly, right now KMSAN poisons freed memory before
SLUB debug wipes it, therefore the memory will count as initialized.
On the other hand, newly allocated memory is still marked as
uninitialized, so a lot of bugs still remain detectable.
TBH, I haven't tested KMSAN with SLUB debug good enough. Note that
it's anyway a separate build that requires Clang, so right now it
doesn't make much sense to combine KMSAN and SLUB debug together.
> > > >
> > > > Signed-off-by: Alexander Potapenko <glider@google.com>
> > > > To: Alexander Potapenko <glider@google.com>
> > > > Cc: Andrew Morton <akpm@linux-foundation.org>
> > > > Cc: Vegard Nossum <vegard.nossum@oracle.com>
> > > > Cc: Dmitry Vyukov <dvyukov@google.com>
> > > > Cc: linux-mm@kvack.org
> > > > ---
> > > >
> > > > Change-Id: I51103b7981d3aabed747d0c85cbdc85568665871
> > > > ---
> > > >  mm/slub.c | 37 +++++++++++++++++++++++++++++++------
> > > >  1 file changed, 31 insertions(+), 6 deletions(-)
> > > >
> > > > diff --git a/mm/slub.c b/mm/slub.c
> > > > index 3d63ae320d31..3d6d4c63446e 100644
> > > > --- a/mm/slub.c
> > > > +++ b/mm/slub.c
> > > > @@ -21,6 +21,8 @@
> > > >  #include <linux/proc_fs.h>
> > > >  #include <linux/seq_file.h>
> > > >  #include <linux/kasan.h>
> > > > +#include <linux/kmsan.h>
> > > > +#include <linux/kmsan-checks.h> /* KMSAN_INIT_VALUE */
> > > >  #include <linux/cpu.h>
> > > >  #include <linux/cpuset.h>
> > > >  #include <linux/mempolicy.h>
> > > > @@ -285,17 +287,27 @@ static void prefetch_freepointer(const struct kmem_cache *s, void *object)
> > > >       prefetch(object + s->offset);
> > > >  }
> > > >
> > > > +/*
> > > > + * When running under KMSAN, get_freepointer_safe() may return an uninitialized
> > > > + * pointer value in the case the current thread loses the race for the next
> > > > + * memory chunk in the freelist. In that case this_cpu_cmpxchg_double() in
> > > > + * slab_alloc_node() will fail, so the uninitialized value won't be used, but
> > > > + * KMSAN will still check all arguments of cmpxchg because of imperfect
> > > > + * handling of inline assembly.
> > > > + * To work around this problem, use KMSAN_INIT_VALUE() to force initialize the
> > > > + * return value of get_freepointer_safe().
> > > > + */
> > > >  static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
> > > >  {
> > > >       unsigned long freepointer_addr;
> > > >       void *p;
> > > >
> > > >       if (!debug_pagealloc_enabled())
> > > > -             return get_freepointer(s, object);
> > > > +             return KMSAN_INIT_VALUE(get_freepointer(s, object));
> > > >
> > > >       freepointer_addr = (unsigned long)object + s->offset;
> > > >       probe_kernel_read(&p, (void **)freepointer_addr, sizeof(p));
> > > > -     return freelist_ptr(s, p, freepointer_addr);
> > > > +     return KMSAN_INIT_VALUE(freelist_ptr(s, p, freepointer_addr));
> > > >  }
> > > >
> > > >  static inline void set_freepointer(struct kmem_cache *s, void *object, void *fp)
> > > > @@ -1390,6 +1402,7 @@ static inline void *kmalloc_large_node_hook(void *ptr, size_t size, gfp_t flags)
> > > >       ptr = kasan_kmalloc_large(ptr, size, flags);
> > > >       /* As ptr might get tagged, call kmemleak hook after KASAN. */
> > > >       kmemleak_alloc(ptr, size, 1, flags);
> > > > +     kmsan_kmalloc_large(ptr, size, flags);
> > > >       return ptr;
> > > >  }
> > > >
> > > > @@ -1397,6 +1410,7 @@ static __always_inline void kfree_hook(void *x)
> > > >  {
> > > >       kmemleak_free(x);
> > > >       kasan_kfree_large(x, _RET_IP_);
> > > > +     kmsan_kfree_large(x);
> > > >  }
> > > >
> > > >  static __always_inline bool slab_free_hook(struct kmem_cache *s, void *x)
> > > > @@ -1453,6 +1467,12 @@ static inline bool slab_free_freelist_hook(struct kmem_cache *s,
> > > >               } while (object != old_tail);
> > > >       }
> > > >
> > > > +     do {
> > > > +             object = next;
> > > > +             next = get_freepointer(s, object);
> > > > +             kmsan_slab_free(s, object);
> > > > +     } while (object != old_tail);
> > > > +
> > > >  /*
> > > >   * Compiler cannot detect this function can be removed if slab_free_hook()
> > > >   * evaluates to nothing.  Thus, catch all relevant config debug options here.
> > > > @@ -2769,6 +2789,7 @@ static __always_inline void *slab_alloc_node(struct kmem_cache *s,
> > > >       if (unlikely(slab_want_init_on_alloc(gfpflags, s)) && object)
> > > >               memset(object, 0, s->object_size);
> > > >
> > > > +     kmsan_slab_alloc(s, object, gfpflags);
> > > >       slab_post_alloc_hook(s, gfpflags, 1, &object);
> > > >
> > > >       return object;
> > > > @@ -2797,6 +2818,7 @@ void *kmem_cache_alloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size)
> > > >       void *ret = slab_alloc(s, gfpflags, _RET_IP_);
> > > >       trace_kmalloc(_RET_IP_, ret, size, s->size, gfpflags);
> > > >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > > > +
> > > >       return ret;
> > > >  }
> > > >  EXPORT_SYMBOL(kmem_cache_alloc_trace);
> > > > @@ -2809,7 +2831,6 @@ void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t gfpflags, int node)
> > > >
> > > >       trace_kmem_cache_alloc_node(_RET_IP_, ret,
> > > >                                   s->object_size, s->size, gfpflags, node);
> > > > -
> > > >       return ret;
> > > >  }
> > > >  EXPORT_SYMBOL(kmem_cache_alloc_node);
> > > > @@ -2825,6 +2846,7 @@ void *kmem_cache_alloc_node_trace(struct kmem_cache *s,
> > > >                          size, s->size, gfpflags, node);
> > > >
> > > >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > > > +
> > > >       return ret;
> > > >  }
> > > >  EXPORT_SYMBOL(kmem_cache_alloc_node_trace);
> > > > @@ -3150,7 +3172,7 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> > > >                         void **p)
> > > >  {
> > > >       struct kmem_cache_cpu *c;
> > > > -     int i;
> > > > +     int i, j;
> > > >
> > > >       /* memcg and kmem_cache debug support */
> > > >       s = slab_pre_alloc_hook(s, flags);
> > > > @@ -3188,11 +3210,11 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> > > >
> > > >       /* Clear memory outside IRQ disabled fastpath loop */
> > > >       if (unlikely(slab_want_init_on_alloc(flags, s))) {
> > > > -             int j;
> > > > -
> > > >               for (j = 0; j < i; j++)
> > > >                       memset(p[j], 0, s->object_size);
> > > >       }
> > > > +     for (j = 0; j < i; j++)
> > > > +             kmsan_slab_alloc(s, p[j], flags);
> > > >
> > > >       /* memcg and kmem_cache debug support */
> > > >       slab_post_alloc_hook(s, flags, size, p);
> > > > @@ -3793,6 +3815,7 @@ static int __init setup_slub_min_objects(char *str)
> > > >
> > > >  __setup("slub_min_objects=", setup_slub_min_objects);
> > > >
> > > > +__no_sanitize_memory
> > > >  void *__kmalloc(size_t size, gfp_t flags)
> > > >  {
> > > >       struct kmem_cache *s;
> > > > @@ -5698,6 +5721,7 @@ static char *create_unique_id(struct kmem_cache *s)
> > > >       p += sprintf(p, "%07u", s->size);
> > > >
> > > >       BUG_ON(p > name + ID_STR_LENGTH - 1);
> > > > +     kmsan_unpoison_shadow(name, p - name);
> > > >       return name;
> > > >  }
> > > >
> > > > @@ -5847,6 +5871,7 @@ static int sysfs_slab_alias(struct kmem_cache *s, const char *name)
> > > >       al->name = name;
> > > >       al->next = alias_list;
> > > >       alias_list = al;
> > > > +     kmsan_unpoison_shadow(al, sizeof(struct saved_alias));
> > > >       return 0;
> > > >  }
> > > >
> >
> >
> >



-- 
Alexander Potapenko
Software Engineer

Google Germany GmbH
Erika-Mann-Straße, 33
80636 München

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg

Re: [PATCH RFC v1 18/26] kmsan: mm: call KMSAN hooks from SLUB code

[Qian Cai]

On Fri, 2019-10-18 at 15:55 +0200, Alexander Potapenko wrote:
> On Fri, Oct 18, 2019 at 3:42 PM Qian Cai <cai@lca.pw> wrote:
> > 
> > On Fri, 2019-10-18 at 15:33 +0200, Alexander Potapenko wrote:
> > > On Fri, Oct 18, 2019 at 3:22 PM Qian Cai <cai@lca.pw> wrote:
> > > > 
> > > > On Fri, 2019-10-18 at 11:42 +0200, glider@google.com wrote:
> > > > > In order to report uninitialized memory coming from heap allocations
> > > > > KMSAN has to poison them unless they're created with __GFP_ZERO.
> > > > > 
> > > > > It's handy that we need KMSAN hooks in the places where
> > > > > init_on_alloc/init_on_free initialization is performed.
> > > > 
> > > > Well, there is SLUB debug which has red zoning and poisoning checks. What's
> > > > value of this patch?
> > > 
> > > Sorry, are you talking about the whole patchset or just this patch?
> > 
> > Just this patch.
> > 
> > > Note that SLUB debug is unable to detect uninitialized values with
> > > bit-to-bit precision, neither have I heard of anyone using it for
> > > detecting uses of uninitialized memory in the kernel at all.
> > > The purpose of SLUB debug is to detect corruptions of freed memory.
> > 
> > The point is if developers have SLUB debug enabled, all the free objects will be
> > poisoned, so what's the point of checking uninitialized memory there?
> 
> You are right, SLUB debug has to be handled separately. If I'm
> understanding correctly, right now KMSAN poisons freed memory before
> SLUB debug wipes it, therefore the memory will count as initialized.
> On the other hand, newly allocated memory is still marked as
> uninitialized, so a lot of bugs still remain detectable.

Yes, but newly allocated slub objects will be poisoned as well.

> TBH, I haven't tested KMSAN with SLUB debug good enough. Note that
> it's anyway a separate build that requires Clang, so right now it
> doesn't make much sense to combine KMSAN and SLUB debug together.

Can't you just build a debug kernel with SLUB debug enabled but dropping this
patch? If there is an uninitialized memory here leading to data corruption, SLUB
debug should be detected as well as this patch. If not, it needs to understand
why.

> > > > > 
> > > > > Signed-off-by: Alexander Potapenko <glider@google.com>
> > > > > To: Alexander Potapenko <glider@google.com>
> > > > > Cc: Andrew Morton <akpm@linux-foundation.org>
> > > > > Cc: Vegard Nossum <vegard.nossum@oracle.com>
> > > > > Cc: Dmitry Vyukov <dvyukov@google.com>
> > > > > Cc: linux-mm@kvack.org
> > > > > ---
> > > > > 
> > > > > Change-Id: I51103b7981d3aabed747d0c85cbdc85568665871
> > > > > ---
> > > > >  mm/slub.c | 37 +++++++++++++++++++++++++++++++------
> > > > >  1 file changed, 31 insertions(+), 6 deletions(-)
> > > > > 
> > > > > diff --git a/mm/slub.c b/mm/slub.c
> > > > > index 3d63ae320d31..3d6d4c63446e 100644
> > > > > --- a/mm/slub.c
> > > > > +++ b/mm/slub.c
> > > > > @@ -21,6 +21,8 @@
> > > > >  #include <linux/proc_fs.h>
> > > > >  #include <linux/seq_file.h>
> > > > >  #include <linux/kasan.h>
> > > > > +#include <linux/kmsan.h>
> > > > > +#include <linux/kmsan-checks.h> /* KMSAN_INIT_VALUE */
> > > > >  #include <linux/cpu.h>
> > > > >  #include <linux/cpuset.h>
> > > > >  #include <linux/mempolicy.h>
> > > > > @@ -285,17 +287,27 @@ static void prefetch_freepointer(const struct kmem_cache *s, void *object)
> > > > >       prefetch(object + s->offset);
> > > > >  }
> > > > > 
> > > > > +/*
> > > > > + * When running under KMSAN, get_freepointer_safe() may return an uninitialized
> > > > > + * pointer value in the case the current thread loses the race for the next
> > > > > + * memory chunk in the freelist. In that case this_cpu_cmpxchg_double() in
> > > > > + * slab_alloc_node() will fail, so the uninitialized value won't be used, but
> > > > > + * KMSAN will still check all arguments of cmpxchg because of imperfect
> > > > > + * handling of inline assembly.
> > > > > + * To work around this problem, use KMSAN_INIT_VALUE() to force initialize the
> > > > > + * return value of get_freepointer_safe().
> > > > > + */
> > > > >  static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
> > > > >  {
> > > > >       unsigned long freepointer_addr;
> > > > >       void *p;
> > > > > 
> > > > >       if (!debug_pagealloc_enabled())
> > > > > -             return get_freepointer(s, object);
> > > > > +             return KMSAN_INIT_VALUE(get_freepointer(s, object));
> > > > > 
> > > > >       freepointer_addr = (unsigned long)object + s->offset;
> > > > >       probe_kernel_read(&p, (void **)freepointer_addr, sizeof(p));
> > > > > -     return freelist_ptr(s, p, freepointer_addr);
> > > > > +     return KMSAN_INIT_VALUE(freelist_ptr(s, p, freepointer_addr));
> > > > >  }
> > > > > 
> > > > >  static inline void set_freepointer(struct kmem_cache *s, void *object, void *fp)
> > > > > @@ -1390,6 +1402,7 @@ static inline void *kmalloc_large_node_hook(void *ptr, size_t size, gfp_t flags)
> > > > >       ptr = kasan_kmalloc_large(ptr, size, flags);
> > > > >       /* As ptr might get tagged, call kmemleak hook after KASAN. */
> > > > >       kmemleak_alloc(ptr, size, 1, flags);
> > > > > +     kmsan_kmalloc_large(ptr, size, flags);
> > > > >       return ptr;
> > > > >  }
> > > > > 
> > > > > @@ -1397,6 +1410,7 @@ static __always_inline void kfree_hook(void *x)
> > > > >  {
> > > > >       kmemleak_free(x);
> > > > >       kasan_kfree_large(x, _RET_IP_);
> > > > > +     kmsan_kfree_large(x);
> > > > >  }
> > > > > 
> > > > >  static __always_inline bool slab_free_hook(struct kmem_cache *s, void *x)
> > > > > @@ -1453,6 +1467,12 @@ static inline bool slab_free_freelist_hook(struct kmem_cache *s,
> > > > >               } while (object != old_tail);
> > > > >       }
> > > > > 
> > > > > +     do {
> > > > > +             object = next;
> > > > > +             next = get_freepointer(s, object);
> > > > > +             kmsan_slab_free(s, object);
> > > > > +     } while (object != old_tail);
> > > > > +
> > > > >  /*
> > > > >   * Compiler cannot detect this function can be removed if slab_free_hook()
> > > > >   * evaluates to nothing.  Thus, catch all relevant config debug options here.
> > > > > @@ -2769,6 +2789,7 @@ static __always_inline void *slab_alloc_node(struct kmem_cache *s,
> > > > >       if (unlikely(slab_want_init_on_alloc(gfpflags, s)) && object)
> > > > >               memset(object, 0, s->object_size);
> > > > > 
> > > > > +     kmsan_slab_alloc(s, object, gfpflags);
> > > > >       slab_post_alloc_hook(s, gfpflags, 1, &object);
> > > > > 
> > > > >       return object;
> > > > > @@ -2797,6 +2818,7 @@ void *kmem_cache_alloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size)
> > > > >       void *ret = slab_alloc(s, gfpflags, _RET_IP_);
> > > > >       trace_kmalloc(_RET_IP_, ret, size, s->size, gfpflags);
> > > > >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > > > > +
> > > > >       return ret;
> > > > >  }
> > > > >  EXPORT_SYMBOL(kmem_cache_alloc_trace);
> > > > > @@ -2809,7 +2831,6 @@ void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t gfpflags, int node)
> > > > > 
> > > > >       trace_kmem_cache_alloc_node(_RET_IP_, ret,
> > > > >                                   s->object_size, s->size, gfpflags, node);
> > > > > -
> > > > >       return ret;
> > > > >  }
> > > > >  EXPORT_SYMBOL(kmem_cache_alloc_node);
> > > > > @@ -2825,6 +2846,7 @@ void *kmem_cache_alloc_node_trace(struct kmem_cache *s,
> > > > >                          size, s->size, gfpflags, node);
> > > > > 
> > > > >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > > > > +
> > > > >       return ret;
> > > > >  }
> > > > >  EXPORT_SYMBOL(kmem_cache_alloc_node_trace);
> > > > > @@ -3150,7 +3172,7 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> > > > >                         void **p)
> > > > >  {
> > > > >       struct kmem_cache_cpu *c;
> > > > > -     int i;
> > > > > +     int i, j;
> > > > > 
> > > > >       /* memcg and kmem_cache debug support */
> > > > >       s = slab_pre_alloc_hook(s, flags);
> > > > > @@ -3188,11 +3210,11 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> > > > > 
> > > > >       /* Clear memory outside IRQ disabled fastpath loop */
> > > > >       if (unlikely(slab_want_init_on_alloc(flags, s))) {
> > > > > -             int j;
> > > > > -
> > > > >               for (j = 0; j < i; j++)
> > > > >                       memset(p[j], 0, s->object_size);
> > > > >       }
> > > > > +     for (j = 0; j < i; j++)
> > > > > +             kmsan_slab_alloc(s, p[j], flags);
> > > > > 
> > > > >       /* memcg and kmem_cache debug support */
> > > > >       slab_post_alloc_hook(s, flags, size, p);
> > > > > @@ -3793,6 +3815,7 @@ static int __init setup_slub_min_objects(char *str)
> > > > > 
> > > > >  __setup("slub_min_objects=", setup_slub_min_objects);
> > > > > 
> > > > > +__no_sanitize_memory
> > > > >  void *__kmalloc(size_t size, gfp_t flags)
> > > > >  {
> > > > >       struct kmem_cache *s;
> > > > > @@ -5698,6 +5721,7 @@ static char *create_unique_id(struct kmem_cache *s)
> > > > >       p += sprintf(p, "%07u", s->size);
> > > > > 
> > > > >       BUG_ON(p > name + ID_STR_LENGTH - 1);
> > > > > +     kmsan_unpoison_shadow(name, p - name);
> > > > >       return name;
> > > > >  }
> > > > > 
> > > > > @@ -5847,6 +5871,7 @@ static int sysfs_slab_alias(struct kmem_cache *s, const char *name)
> > > > >       al->name = name;
> > > > >       al->next = alias_list;
> > > > >       alias_list = al;
> > > > > +     kmsan_unpoison_shadow(al, sizeof(struct saved_alias));
> > > > >       return 0;
> > > > >  }
> > > > > 
> > > 
> > > 
> > > 
> 
> 
> 

Re: [PATCH RFC v1 18/26] kmsan: mm: call KMSAN hooks from SLUB code

[Alexander Potapenko]

On Fri, Oct 18, 2019 at 4:42 PM Qian Cai <cai@lca.pw> wrote:
>
> On Fri, 2019-10-18 at 15:55 +0200, Alexander Potapenko wrote:
> > On Fri, Oct 18, 2019 at 3:42 PM Qian Cai <cai@lca.pw> wrote:
> > >
> > > On Fri, 2019-10-18 at 15:33 +0200, Alexander Potapenko wrote:
> > > > On Fri, Oct 18, 2019 at 3:22 PM Qian Cai <cai@lca.pw> wrote:
> > > > >
> > > > > On Fri, 2019-10-18 at 11:42 +0200, glider@google.com wrote:
> > > > > > In order to report uninitialized memory coming from heap allocations
> > > > > > KMSAN has to poison them unless they're created with __GFP_ZERO.
> > > > > >
> > > > > > It's handy that we need KMSAN hooks in the places where
> > > > > > init_on_alloc/init_on_free initialization is performed.
> > > > >
> > > > > Well, there is SLUB debug which has red zoning and poisoning checks. What's
> > > > > value of this patch?
> > > >
> > > > Sorry, are you talking about the whole patchset or just this patch?
> > >
> > > Just this patch.
> > >
> > > > Note that SLUB debug is unable to detect uninitialized values with
> > > > bit-to-bit precision, neither have I heard of anyone using it for
> > > > detecting uses of uninitialized memory in the kernel at all.
> > > > The purpose of SLUB debug is to detect corruptions of freed memory.
> > >
> > > The point is if developers have SLUB debug enabled, all the free objects will be
> > > poisoned, so what's the point of checking uninitialized memory there?
> >
> > You are right, SLUB debug has to be handled separately. If I'm
> > understanding correctly, right now KMSAN poisons freed memory before
> > SLUB debug wipes it, therefore the memory will count as initialized.
> > On the other hand, newly allocated memory is still marked as
> > uninitialized, so a lot of bugs still remain detectable.
>
> Yes, but newly allocated slub objects will be poisoned as well.
As far as I can tell, KMSAN hook marking newly allocated objects as
uninitialized is called after slub poisoning.
Therefore these allocations will be treated by KMSAN as uninitialized.
> > TBH, I haven't tested KMSAN with SLUB debug good enough. Note that
> > it's anyway a separate build that requires Clang, so right now it
> > doesn't make much sense to combine KMSAN and SLUB debug together.
>
> Can't you just build a debug kernel with SLUB debug enabled but dropping this
> patch? If there is an uninitialized memory here leading to data corruption, SLUB
> debug should be detected as well as this patch. If not, it needs to understand
> why.
Sorry, there might be some misunderstanding here.
KMSAN keeps the state of heap objects separately by keeping exactly
the same amount of initialized/uninitialized bits for every
allocation.
A call to kmsan_slab_alloc()/kmsan_slab_free() will mark an allocation
as uninitialized for KMSAN. Not doing so will result in false reports.
A call to memset(object, POISON_FREE, object_size) performed by SLAB
debug will actually mark this allocation as initialized from KMSAN
point of view, because we're memsetting a range with initialized data.
Note that use of uninitialized data doesn't necessarily lead to
immediate data corruption, so there might be nothing to detect for
SLUB debug.
>
> > > > > >
> > > > > > Signed-off-by: Alexander Potapenko <glider@google.com>
> > > > > > To: Alexander Potapenko <glider@google.com>
> > > > > > Cc: Andrew Morton <akpm@linux-foundation.org>
> > > > > > Cc: Vegard Nossum <vegard.nossum@oracle.com>
> > > > > > Cc: Dmitry Vyukov <dvyukov@google.com>
> > > > > > Cc: linux-mm@kvack.org
> > > > > > ---
> > > > > >
> > > > > > Change-Id: I51103b7981d3aabed747d0c85cbdc85568665871
> > > > > > ---
> > > > > >  mm/slub.c | 37 +++++++++++++++++++++++++++++++------
> > > > > >  1 file changed, 31 insertions(+), 6 deletions(-)
> > > > > >
> > > > > > diff --git a/mm/slub.c b/mm/slub.c
> > > > > > index 3d63ae320d31..3d6d4c63446e 100644
> > > > > > --- a/mm/slub.c
> > > > > > +++ b/mm/slub.c
> > > > > > @@ -21,6 +21,8 @@
> > > > > >  #include <linux/proc_fs.h>
> > > > > >  #include <linux/seq_file.h>
> > > > > >  #include <linux/kasan.h>
> > > > > > +#include <linux/kmsan.h>
> > > > > > +#include <linux/kmsan-checks.h> /* KMSAN_INIT_VALUE */
> > > > > >  #include <linux/cpu.h>
> > > > > >  #include <linux/cpuset.h>
> > > > > >  #include <linux/mempolicy.h>
> > > > > > @@ -285,17 +287,27 @@ static void prefetch_freepointer(const struct kmem_cache *s, void *object)
> > > > > >       prefetch(object + s->offset);
> > > > > >  }
> > > > > >
> > > > > > +/*
> > > > > > + * When running under KMSAN, get_freepointer_safe() may return an uninitialized
> > > > > > + * pointer value in the case the current thread loses the race for the next
> > > > > > + * memory chunk in the freelist. In that case this_cpu_cmpxchg_double() in
> > > > > > + * slab_alloc_node() will fail, so the uninitialized value won't be used, but
> > > > > > + * KMSAN will still check all arguments of cmpxchg because of imperfect
> > > > > > + * handling of inline assembly.
> > > > > > + * To work around this problem, use KMSAN_INIT_VALUE() to force initialize the
> > > > > > + * return value of get_freepointer_safe().
> > > > > > + */
> > > > > >  static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
> > > > > >  {
> > > > > >       unsigned long freepointer_addr;
> > > > > >       void *p;
> > > > > >
> > > > > >       if (!debug_pagealloc_enabled())
> > > > > > -             return get_freepointer(s, object);
> > > > > > +             return KMSAN_INIT_VALUE(get_freepointer(s, object));
> > > > > >
> > > > > >       freepointer_addr = (unsigned long)object + s->offset;
> > > > > >       probe_kernel_read(&p, (void **)freepointer_addr, sizeof(p));
> > > > > > -     return freelist_ptr(s, p, freepointer_addr);
> > > > > > +     return KMSAN_INIT_VALUE(freelist_ptr(s, p, freepointer_addr));
> > > > > >  }
> > > > > >
> > > > > >  static inline void set_freepointer(struct kmem_cache *s, void *object, void *fp)
> > > > > > @@ -1390,6 +1402,7 @@ static inline void *kmalloc_large_node_hook(void *ptr, size_t size, gfp_t flags)
> > > > > >       ptr = kasan_kmalloc_large(ptr, size, flags);
> > > > > >       /* As ptr might get tagged, call kmemleak hook after KASAN. */
> > > > > >       kmemleak_alloc(ptr, size, 1, flags);
> > > > > > +     kmsan_kmalloc_large(ptr, size, flags);
> > > > > >       return ptr;
> > > > > >  }
> > > > > >
> > > > > > @@ -1397,6 +1410,7 @@ static __always_inline void kfree_hook(void *x)
> > > > > >  {
> > > > > >       kmemleak_free(x);
> > > > > >       kasan_kfree_large(x, _RET_IP_);
> > > > > > +     kmsan_kfree_large(x);
> > > > > >  }
> > > > > >
> > > > > >  static __always_inline bool slab_free_hook(struct kmem_cache *s, void *x)
> > > > > > @@ -1453,6 +1467,12 @@ static inline bool slab_free_freelist_hook(struct kmem_cache *s,
> > > > > >               } while (object != old_tail);
> > > > > >       }
> > > > > >
> > > > > > +     do {
> > > > > > +             object = next;
> > > > > > +             next = get_freepointer(s, object);
> > > > > > +             kmsan_slab_free(s, object);
> > > > > > +     } while (object != old_tail);
> > > > > > +
> > > > > >  /*
> > > > > >   * Compiler cannot detect this function can be removed if slab_free_hook()
> > > > > >   * evaluates to nothing.  Thus, catch all relevant config debug options here.
> > > > > > @@ -2769,6 +2789,7 @@ static __always_inline void *slab_alloc_node(struct kmem_cache *s,
> > > > > >       if (unlikely(slab_want_init_on_alloc(gfpflags, s)) && object)
> > > > > >               memset(object, 0, s->object_size);
> > > > > >
> > > > > > +     kmsan_slab_alloc(s, object, gfpflags);
> > > > > >       slab_post_alloc_hook(s, gfpflags, 1, &object);
> > > > > >
> > > > > >       return object;
> > > > > > @@ -2797,6 +2818,7 @@ void *kmem_cache_alloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size)
> > > > > >       void *ret = slab_alloc(s, gfpflags, _RET_IP_);
> > > > > >       trace_kmalloc(_RET_IP_, ret, size, s->size, gfpflags);
> > > > > >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > > > > > +
> > > > > >       return ret;
> > > > > >  }
> > > > > >  EXPORT_SYMBOL(kmem_cache_alloc_trace);
> > > > > > @@ -2809,7 +2831,6 @@ void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t gfpflags, int node)
> > > > > >
> > > > > >       trace_kmem_cache_alloc_node(_RET_IP_, ret,
> > > > > >                                   s->object_size, s->size, gfpflags, node);
> > > > > > -
> > > > > >       return ret;
> > > > > >  }
> > > > > >  EXPORT_SYMBOL(kmem_cache_alloc_node);
> > > > > > @@ -2825,6 +2846,7 @@ void *kmem_cache_alloc_node_trace(struct kmem_cache *s,
> > > > > >                          size, s->size, gfpflags, node);
> > > > > >
> > > > > >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > > > > > +
> > > > > >       return ret;
> > > > > >  }
> > > > > >  EXPORT_SYMBOL(kmem_cache_alloc_node_trace);
> > > > > > @@ -3150,7 +3172,7 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> > > > > >                         void **p)
> > > > > >  {
> > > > > >       struct kmem_cache_cpu *c;
> > > > > > -     int i;
> > > > > > +     int i, j;
> > > > > >
> > > > > >       /* memcg and kmem_cache debug support */
> > > > > >       s = slab_pre_alloc_hook(s, flags);
> > > > > > @@ -3188,11 +3210,11 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> > > > > >
> > > > > >       /* Clear memory outside IRQ disabled fastpath loop */
> > > > > >       if (unlikely(slab_want_init_on_alloc(flags, s))) {
> > > > > > -             int j;
> > > > > > -
> > > > > >               for (j = 0; j < i; j++)
> > > > > >                       memset(p[j], 0, s->object_size);
> > > > > >       }
> > > > > > +     for (j = 0; j < i; j++)
> > > > > > +             kmsan_slab_alloc(s, p[j], flags);
> > > > > >
> > > > > >       /* memcg and kmem_cache debug support */
> > > > > >       slab_post_alloc_hook(s, flags, size, p);
> > > > > > @@ -3793,6 +3815,7 @@ static int __init setup_slub_min_objects(char *str)
> > > > > >
> > > > > >  __setup("slub_min_objects=", setup_slub_min_objects);
> > > > > >
> > > > > > +__no_sanitize_memory
> > > > > >  void *__kmalloc(size_t size, gfp_t flags)
> > > > > >  {
> > > > > >       struct kmem_cache *s;
> > > > > > @@ -5698,6 +5721,7 @@ static char *create_unique_id(struct kmem_cache *s)
> > > > > >       p += sprintf(p, "%07u", s->size);
> > > > > >
> > > > > >       BUG_ON(p > name + ID_STR_LENGTH - 1);
> > > > > > +     kmsan_unpoison_shadow(name, p - name);
> > > > > >       return name;
> > > > > >  }
> > > > > >
> > > > > > @@ -5847,6 +5871,7 @@ static int sysfs_slab_alias(struct kmem_cache *s, const char *name)
> > > > > >       al->name = name;
> > > > > >       al->next = alias_list;
> > > > > >       alias_list = al;
> > > > > > +     kmsan_unpoison_shadow(al, sizeof(struct saved_alias));
> > > > > >       return 0;
> > > > > >  }
> > > > > >
> > > >
> > > >
> > > >
> >
> >
> >



-- 
Alexander Potapenko
Software Engineer

Google Germany GmbH
Erika-Mann-Straße, 33
80636 München

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg

Re: [PATCH RFC v1 19/26] kmsan: call KMSAN hooks where needed

[Qian Cai]

On Fri, 2019-10-18 at 11:42 +0200, glider@google.com wrote:
> Insert KMSAN hooks that check for potential memory errors and/or make
> necessary bookkeeping changes:
>  - allocate/split/deallocate metadata pages in
>    alloc_pages()/split_page()/free_page();

This also seems unnecessary where there are options like page_poison and
debug_pagealloc should be able detect uninitialized memory access in the page
allocator as well. Even KASAN has some of the functionality.

>  - clear page shadow and origins in clear_page(), copy_user_highpage();
>  - copy page metadata in copy_highpage(), wp_page_copy();
>  - handle vmap()/vunmap()/iounmap();
>  - handle task creation and deletion;
>  - handle vprintk();
>  - call softirq entry/exit hooks in kernel/softirq.c;
>  - check/initialize memory sent to/read from USB, I2C, and network
> 
> Signed-off-by: Alexander Potapenko <glider@google.com>
> To: Alexander Potapenko <glider@google.com>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> Cc: Eric Dumazet <edumazet@google.com>
> Cc: Wolfram Sang <wsa@the-dreams.de>
> Cc: Petr Mladek <pmladek@suse.com>
> Cc: Vegard Nossum <vegard.nossum@oracle.com>
> Cc: Dmitry Vyukov <dvyukov@google.com>
> Cc: linux-mm@kvack.org
> ---
> 
> Change-Id: I1250a928d9263bf71fdaa067a070bdee686ef47b
> ---
>  arch/x86/include/asm/page_64.h | 13 +++++++++++++
>  arch/x86/mm/ioremap.c          |  3 +++
>  drivers/i2c/i2c-core-base.c    |  2 ++
>  drivers/usb/core/urb.c         |  2 ++
>  include/linux/highmem.h        |  4 ++++
>  kernel/exit.c                  |  2 ++
>  kernel/fork.c                  |  2 ++
>  kernel/kthread.c               |  2 ++
>  kernel/printk/printk.c         |  8 +++++++-
>  kernel/softirq.c               |  5 +++++
>  lib/ioremap.c                  |  5 +++++
>  mm/compaction.c                |  9 +++++++++
>  mm/gup.c                       |  3 +++
>  mm/memory.c                    |  2 ++
>  mm/page_alloc.c                | 16 ++++++++++++++++
>  mm/vmalloc.c                   | 23 +++++++++++++++++++++--
>  net/sched/sch_generic.c        |  2 ++
>  17 files changed, 100 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/x86/include/asm/page_64.h b/arch/x86/include/asm/page_64.h
> index 939b1cff4a7b..0ba43d93414f 100644
> --- a/arch/x86/include/asm/page_64.h
> +++ b/arch/x86/include/asm/page_64.h
> @@ -44,14 +44,27 @@ void clear_page_orig(void *page);
>  void clear_page_rep(void *page);
>  void clear_page_erms(void *page);
>  
> +/* This is an assembly header, avoid including too much of kmsan.h */
> +#ifdef CONFIG_KMSAN
> +void kmsan_clear_page(void *page_addr);
> +#endif
> +__no_sanitize_memory
>  static inline void clear_page(void *page)
>  {
> +#ifdef CONFIG_KMSAN
> +	/* alternative_call_2() changes |page|. */
> +	void *page_copy = page;
> +#endif
>  	alternative_call_2(clear_page_orig,
>  			   clear_page_rep, X86_FEATURE_REP_GOOD,
>  			   clear_page_erms, X86_FEATURE_ERMS,
>  			   "=D" (page),
>  			   "0" (page)
>  			   : "cc", "memory", "rax", "rcx");
> +#ifdef CONFIG_KMSAN
> +	/* Clear KMSAN shadow for the pages that have it. */
> +	kmsan_clear_page(page_copy);
> +#endif
>  }
>  
>  void copy_page(void *to, void *from);
> diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
> index a39dcdb5ae34..fdb2abc11a82 100644
> --- a/arch/x86/mm/ioremap.c
> +++ b/arch/x86/mm/ioremap.c
> @@ -7,6 +7,7 @@
>   * (C) Copyright 1995 1996 Linus Torvalds
>   */
>  
> +#include <linux/kmsan.h>
>  #include <linux/memblock.h>
>  #include <linux/init.h>
>  #include <linux/io.h>
> @@ -451,6 +452,8 @@ void iounmap(volatile void __iomem *addr)
>  		return;
>  	}
>  
> +	kmsan_iounmap_page_range((unsigned long)addr,
> +		(unsigned long)addr + get_vm_area_size(p));
>  	free_memtype(p->phys_addr, p->phys_addr + get_vm_area_size(p));
>  
>  	/* Finally remove it */
> diff --git a/drivers/i2c/i2c-core-base.c b/drivers/i2c/i2c-core-base.c
> index 5f6a4985f2bc..9685d3399c79 100644
> --- a/drivers/i2c/i2c-core-base.c
> +++ b/drivers/i2c/i2c-core-base.c
> @@ -27,6 +27,7 @@
>  #include <linux/irqflags.h>
>  #include <linux/jump_label.h>
>  #include <linux/kernel.h>
> +#include <linux/kmsan.h>
>  #include <linux/module.h>
>  #include <linux/mutex.h>
>  #include <linux/of_device.h>
> @@ -1975,6 +1976,7 @@ int __i2c_transfer(struct i2c_adapter *adap, struct i2c_msg *msgs, int num)
>  				trace_i2c_write(adap, &msgs[i], i);
>  	}
>  
> +	kmsan_handle_i2c_transfer(msgs, num);
>  	/* Retry automatically on arbitration loss */
>  	orig_jiffies = jiffies;
>  	for (ret = 0, try = 0; try <= adap->retries; try++) {
> diff --git a/drivers/usb/core/urb.c b/drivers/usb/core/urb.c
> index 0eab79f82ce4..5bdb54d71c2e 100644
> --- a/drivers/usb/core/urb.c
> +++ b/drivers/usb/core/urb.c
> @@ -8,6 +8,7 @@
>  #include <linux/bitops.h>
>  #include <linux/slab.h>
>  #include <linux/log2.h>
> +#include <linux/kmsan-checks.h>
>  #include <linux/usb.h>
>  #include <linux/wait.h>
>  #include <linux/usb/hcd.h>
> @@ -401,6 +402,7 @@ int usb_submit_urb(struct urb *urb, gfp_t mem_flags)
>  			URB_SETUP_MAP_SINGLE | URB_SETUP_MAP_LOCAL |
>  			URB_DMA_SG_COMBINED);
>  	urb->transfer_flags |= (is_out ? URB_DIR_OUT : URB_DIR_IN);
> +	kmsan_handle_urb(urb, is_out);
>  
>  	if (xfertype != USB_ENDPOINT_XFER_CONTROL &&
>  			dev->state < USB_STATE_CONFIGURED)
> diff --git a/include/linux/highmem.h b/include/linux/highmem.h
> index ea5cdbd8c2c3..623b56f48685 100644
> --- a/include/linux/highmem.h
> +++ b/include/linux/highmem.h
> @@ -5,6 +5,7 @@
>  #include <linux/fs.h>
>  #include <linux/kernel.h>
>  #include <linux/bug.h>
> +#include <linux/kmsan.h>
>  #include <linux/mm.h>
>  #include <linux/uaccess.h>
>  #include <linux/hardirq.h>
> @@ -255,6 +256,8 @@ static inline void copy_user_highpage(struct page *to, struct page *from,
>  	vfrom = kmap_atomic(from);
>  	vto = kmap_atomic(to);
>  	copy_user_page(vto, vfrom, vaddr, to);
> +	/* User pages don't have shadow, just clear the destination. */
> +	kmsan_clear_page(page_address(to));
>  	kunmap_atomic(vto);
>  	kunmap_atomic(vfrom);
>  }
> @@ -270,6 +273,7 @@ static inline void copy_highpage(struct page *to, struct page *from)
>  	vfrom = kmap_atomic(from);
>  	vto = kmap_atomic(to);
>  	copy_page(vto, vfrom);
> +	kmsan_copy_page_meta(to, from);
>  	kunmap_atomic(vto);
>  	kunmap_atomic(vfrom);
>  }
> diff --git a/kernel/exit.c b/kernel/exit.c
> index a46a50d67002..9e3ce929110b 100644
> --- a/kernel/exit.c
> +++ b/kernel/exit.c
> @@ -60,6 +60,7 @@
>  #include <linux/writeback.h>
>  #include <linux/shm.h>
>  #include <linux/kcov.h>
> +#include <linux/kmsan.h>
>  #include <linux/random.h>
>  #include <linux/rcuwait.h>
>  #include <linux/compat.h>
> @@ -719,6 +720,7 @@ void __noreturn do_exit(long code)
>  
>  	profile_task_exit(tsk);
>  	kcov_task_exit(tsk);
> +	kmsan_task_exit(tsk);
>  
>  	WARN_ON(blk_needs_flush_plug(tsk));
>  
> diff --git a/kernel/fork.c b/kernel/fork.c
> index bcdf53125210..0f08952a42dc 100644
> --- a/kernel/fork.c
> +++ b/kernel/fork.c
> @@ -37,6 +37,7 @@
>  #include <linux/fdtable.h>
>  #include <linux/iocontext.h>
>  #include <linux/key.h>
> +#include <linux/kmsan.h>
>  #include <linux/binfmts.h>
>  #include <linux/mman.h>
>  #include <linux/mmu_notifier.h>
> @@ -931,6 +932,7 @@ static struct task_struct *dup_task_struct(struct task_struct *orig, int node)
>  	account_kernel_stack(tsk, 1);
>  
>  	kcov_task_init(tsk);
> +	kmsan_task_create(tsk);
>  
>  #ifdef CONFIG_FAULT_INJECTION
>  	tsk->fail_nth = 0;
> diff --git a/kernel/kthread.c b/kernel/kthread.c
> index 621467c33fef..9b948559a34a 100644
> --- a/kernel/kthread.c
> +++ b/kernel/kthread.c
> @@ -17,6 +17,7 @@
>  #include <linux/unistd.h>
>  #include <linux/file.h>
>  #include <linux/export.h>
> +#include <linux/kmsan.h>
>  #include <linux/mutex.h>
>  #include <linux/slab.h>
>  #include <linux/freezer.h>
> @@ -350,6 +351,7 @@ struct task_struct *__kthread_create_on_node(int (*threadfn)(void *data),
>  		set_cpus_allowed_ptr(task, cpu_all_mask);
>  	}
>  	kfree(create);
> +	kmsan_task_create(task);
>  	return task;
>  }
>  
> diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
> index ca65327a6de8..f77fdcb5f861 100644
> --- a/kernel/printk/printk.c
> +++ b/kernel/printk/printk.c
> @@ -1914,7 +1914,12 @@ int vprintk_store(int facility, int level,
>  	 * The printf needs to come first; we need the syslog
>  	 * prefix which might be passed-in as a parameter.
>  	 */
> -	text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
> +	/*
> +	 * We've checked the printk arguments in vprintk_emit() already.
> +	 * Initialize |text_len| to prevent the errors from spreading.
> +	 */
> +	text_len = KMSAN_INIT_VALUE(vscnprintf(text, sizeof(textbuf), fmt,
> +					       args));
>  
>  	/* mark and strip a trailing newline */
>  	if (text_len && text[text_len-1] == '\n') {
> @@ -1972,6 +1977,7 @@ asmlinkage int vprintk_emit(int facility, int level,
>  	boot_delay_msec(level);
>  	printk_delay();
>  
> +	kmsan_handle_vprintk(&fmt, args);
>  	/* This stops the holder of console_sem just where we want him */
>  	logbuf_lock_irqsave(flags);
>  	curr_log_seq = log_next_seq;
> diff --git a/kernel/softirq.c b/kernel/softirq.c
> index 0427a86743a4..6d566dd68b35 100644
> --- a/kernel/softirq.c
> +++ b/kernel/softirq.c
> @@ -11,6 +11,7 @@
>  
>  #include <linux/export.h>
>  #include <linux/kernel_stat.h>
> +#include <linux/kmsan.h>
>  #include <linux/interrupt.h>
>  #include <linux/init.h>
>  #include <linux/mm.h>
> @@ -370,7 +371,9 @@ static inline void invoke_softirq(void)
>  		 * it is the irq stack, because it should be near empty
>  		 * at this stage.
>  		 */
> +		kmsan_softirq_enter();
>  		__do_softirq();
> +		kmsan_softirq_exit();
>  #else
>  		/*
>  		 * Otherwise, irq_exit() is called on the task stack that can
> @@ -600,7 +603,9 @@ static void run_ksoftirqd(unsigned int cpu)
>  		 * We can safely run softirq on inline stack, as we are not deep
>  		 * in the task stack here.
>  		 */
> +		kmsan_softirq_enter();
>  		__do_softirq();
> +		kmsan_softirq_exit();
>  		local_irq_enable();
>  		cond_resched();
>  		return;
> diff --git a/lib/ioremap.c b/lib/ioremap.c
> index 0a2ffadc6d71..5f830cee5bfc 100644
> --- a/lib/ioremap.c
> +++ b/lib/ioremap.c
> @@ -6,6 +6,7 @@
>   *
>   * (C) Copyright 1995 1996 Linus Torvalds
>   */
> +#include <linux/kmsan.h>
>  #include <linux/vmalloc.h>
>  #include <linux/mm.h>
>  #include <linux/sched.h>
> @@ -214,6 +215,8 @@ int ioremap_page_range(unsigned long addr,
>  	unsigned long start;
>  	unsigned long next;
>  	int err;
> +	unsigned long old_addr = addr;
> +	phys_addr_t old_phys_addr = phys_addr;
>  
>  	might_sleep();
>  	BUG_ON(addr >= end);
> @@ -228,6 +231,8 @@ int ioremap_page_range(unsigned long addr,
>  	} while (pgd++, phys_addr += (next - addr), addr = next, addr != end);
>  
>  	flush_cache_vmap(start, end);
> +	if (!err)
> +		kmsan_ioremap_page_range(old_addr, end, old_phys_addr, prot);
>  
>  	return err;
>  }
> diff --git a/mm/compaction.c b/mm/compaction.c
> index ce08b39d85d4..e9a5898e5fe0 100644
> --- a/mm/compaction.c
> +++ b/mm/compaction.c
> @@ -84,6 +84,15 @@ static void split_map_pages(struct list_head *list)
>  
>  		for (i = 0; i < nr_pages; i++) {
>  			list_add(&page->lru, &tmp_list);
> +#ifdef CONFIG_KMSAN
> +			/*
> +			 * TODO(glider): we may lose the metadata when copying
> +			 * something to these pages. Need to allocate shadow
> +			 * and origin pages here instead.
> +			 */
> +			page->shadow = NULL;
> +			page->origin = NULL;
> +#endif
>  			page++;
>  		}
>  	}
> diff --git a/mm/gup.c b/mm/gup.c
> index 23a9f9c9d377..c16e4288997d 100644
> --- a/mm/gup.c
> +++ b/mm/gup.c
> @@ -4,6 +4,7 @@
>  #include <linux/err.h>
>  #include <linux/spinlock.h>
>  
> +#include <linux/kmsan.h>
>  #include <linux/mm.h>
>  #include <linux/memremap.h>
>  #include <linux/pagemap.h>
> @@ -2347,6 +2348,7 @@ int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
>  	    gup_fast_permitted(start, end)) {
>  		local_irq_save(flags);
>  		gup_pgd_range(start, end, write ? FOLL_WRITE : 0, pages, &nr);
> +		kmsan_gup_pgd_range(pages, nr);
>  		local_irq_restore(flags);
>  	}
>  
> @@ -2416,6 +2418,7 @@ int get_user_pages_fast(unsigned long start, int nr_pages,
>  	    gup_fast_permitted(start, end)) {
>  		local_irq_disable();
>  		gup_pgd_range(addr, end, gup_flags, pages, &nr);
> +		kmsan_gup_pgd_range(pages, nr);
>  		local_irq_enable();
>  		ret = nr;
>  	}
> diff --git a/mm/memory.c b/mm/memory.c
> index b1ca51a079f2..48ceacc06e2d 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -51,6 +51,7 @@
>  #include <linux/highmem.h>
>  #include <linux/pagemap.h>
>  #include <linux/memremap.h>
> +#include <linux/kmsan.h>
>  #include <linux/ksm.h>
>  #include <linux/rmap.h>
>  #include <linux/export.h>
> @@ -2328,6 +2329,7 @@ static vm_fault_t wp_page_copy(struct vm_fault *vmf)
>  		if (!new_page)
>  			goto oom;
>  		cow_user_page(new_page, old_page, vmf->address, vma);
> +		kmsan_copy_page_meta(new_page, old_page);
>  	}
>  
>  	if (mem_cgroup_try_charge_delay(new_page, mm, GFP_KERNEL, &memcg, false))
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index c0b2e0306720..f52494e0dcd9 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -26,6 +26,8 @@
>  #include <linux/compiler.h>
>  #include <linux/kernel.h>
>  #include <linux/kasan.h>
> +#include <linux/kmsan.h>
> +#include <linux/kmsan-checks.h>
>  #include <linux/module.h>
>  #include <linux/suspend.h>
>  #include <linux/pagevec.h>
> @@ -1133,6 +1135,7 @@ static __always_inline bool free_pages_prepare(struct page *page,
>  	VM_BUG_ON_PAGE(PageTail(page), page);
>  
>  	trace_mm_page_free(page, order);
> +	kmsan_free_page(page, order);
>  
>  	/*
>  	 * Check tail pages before head page information is cleared to
> @@ -3121,6 +3124,7 @@ void split_page(struct page *page, unsigned int order)
>  	VM_BUG_ON_PAGE(PageCompound(page), page);
>  	VM_BUG_ON_PAGE(!page_count(page), page);
>  
> +	kmsan_split_page(page, order);
>  	for (i = 1; i < (1 << order); i++)
>  		set_page_refcounted(page + i);
>  	split_page_owner(page, order);
> @@ -3253,6 +3257,13 @@ static struct page *rmqueue_pcplist(struct zone *preferred_zone,
>  /*
>   * Allocate a page from the given zone. Use pcplists for order-0 allocations.
>   */
> +/*
> + * TODO(glider): rmqueue() may call __msan_poison_alloca() through a call to
> + * set_pfnblock_flags_mask(). If __msan_poison_alloca() attempts to allocate
> + * pages for the stack depot, it may call rmqueue() again, which will result
> + * in a deadlock.
> + */
> +__no_sanitize_memory
>  static inline
>  struct page *rmqueue(struct zone *preferred_zone,
>  			struct zone *zone, unsigned int order,
> @@ -4779,6 +4790,11 @@ __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order, int preferred_nid,
>  
>  	trace_mm_page_alloc(page, order, alloc_mask, ac.migratetype);
>  
> +	if (page)
> +		if (kmsan_alloc_page(page, order, gfp_mask)) {
> +			__free_pages(page, order);
> +			page = NULL;
> +		}
>  	return page;
>  }
>  EXPORT_SYMBOL(__alloc_pages_nodemask);
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index a3c70e275f4e..bdf66ffcf02c 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -29,6 +29,7 @@
>  #include <linux/rcupdate.h>
>  #include <linux/pfn.h>
>  #include <linux/kmemleak.h>
> +#include <linux/kmsan.h>
>  #include <linux/atomic.h>
>  #include <linux/compiler.h>
>  #include <linux/llist.h>
> @@ -119,7 +120,8 @@ static void vunmap_p4d_range(pgd_t *pgd, unsigned long addr, unsigned long end)
>  	} while (p4d++, addr = next, addr != end);
>  }
>  
> -static void vunmap_page_range(unsigned long addr, unsigned long end)
> +/* Exported for KMSAN, visible in mm/kmsan/kmsan.h only. */
> +void __vunmap_page_range(unsigned long addr, unsigned long end)
>  {
>  	pgd_t *pgd;
>  	unsigned long next;
> @@ -133,6 +135,12 @@ static void vunmap_page_range(unsigned long addr, unsigned long end)
>  		vunmap_p4d_range(pgd, addr, next);
>  	} while (pgd++, addr = next, addr != end);
>  }
> +EXPORT_SYMBOL(__vunmap_page_range);
> +static void vunmap_page_range(unsigned long addr, unsigned long end)
> +{
> +	kmsan_vunmap_page_range(addr, end);
> +	__vunmap_page_range(addr, end);
> +}
>  
>  static int vmap_pte_range(pmd_t *pmd, unsigned long addr,
>  		unsigned long end, pgprot_t prot, struct page **pages, int *nr)
> @@ -216,8 +224,11 @@ static int vmap_p4d_range(pgd_t *pgd, unsigned long addr,
>   * will have pfns corresponding to the "pages" array.
>   *
>   * Ie. pte at addr+N*PAGE_SIZE shall point to pfn corresponding to pages[N]
> + *
> + * This function is exported for use in KMSAN, but is only declared in KMSAN
> + * headers.
>   */
> -static int vmap_page_range_noflush(unsigned long start, unsigned long end,
> +int __vmap_page_range_noflush(unsigned long start, unsigned long end,
>  				   pgprot_t prot, struct page **pages)
>  {
>  	pgd_t *pgd;
> @@ -237,6 +248,14 @@ static int vmap_page_range_noflush(unsigned long start, unsigned long end,
>  
>  	return nr;
>  }
> +EXPORT_SYMBOL(__vmap_page_range_noflush);
> +
> +static int vmap_page_range_noflush(unsigned long start, unsigned long end,
> +				   pgprot_t prot, struct page **pages)
> +{
> +	kmsan_vmap_page_range_noflush(start, end, prot, pages);
> +	return __vmap_page_range_noflush(start, end, prot, pages);
> +}
>  
>  static int vmap_page_range(unsigned long start, unsigned long end,
>  			   pgprot_t prot, struct page **pages)
> diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
> index 17bd8f539bc7..fd22c4a4ba42 100644
> --- a/net/sched/sch_generic.c
> +++ b/net/sched/sch_generic.c
> @@ -11,6 +11,7 @@
>  #include <linux/module.h>
>  #include <linux/types.h>
>  #include <linux/kernel.h>
> +#include <linux/kmsan-checks.h>
>  #include <linux/sched.h>
>  #include <linux/string.h>
>  #include <linux/errno.h>
> @@ -659,6 +660,7 @@ static struct sk_buff *pfifo_fast_dequeue(struct Qdisc *qdisc)
>  	} else {
>  		qdisc->empty = true;
>  	}
> +	kmsan_check_skb(skb);
>  
>  	return skb;
>  }

Re: [PATCH RFC v1 18/26] kmsan: mm: call KMSAN hooks from SLUB code

[Qian Cai]

On Fri, 2019-10-18 at 16:54 +0200, Alexander Potapenko wrote:
> On Fri, Oct 18, 2019 at 4:42 PM Qian Cai <cai@lca.pw> wrote:
> > 
> > On Fri, 2019-10-18 at 15:55 +0200, Alexander Potapenko wrote:
> > > On Fri, Oct 18, 2019 at 3:42 PM Qian Cai <cai@lca.pw> wrote:
> > > > 
> > > > On Fri, 2019-10-18 at 15:33 +0200, Alexander Potapenko wrote:
> > > > > On Fri, Oct 18, 2019 at 3:22 PM Qian Cai <cai@lca.pw> wrote:
> > > > > > 
> > > > > > On Fri, 2019-10-18 at 11:42 +0200, glider@google.com wrote:
> > > > > > > In order to report uninitialized memory coming from heap allocations
> > > > > > > KMSAN has to poison them unless they're created with __GFP_ZERO.
> > > > > > > 
> > > > > > > It's handy that we need KMSAN hooks in the places where
> > > > > > > init_on_alloc/init_on_free initialization is performed.
> > > > > > 
> > > > > > Well, there is SLUB debug which has red zoning and poisoning checks. What's
> > > > > > value of this patch?
> > > > > 
> > > > > Sorry, are you talking about the whole patchset or just this patch?
> > > > 
> > > > Just this patch.
> > > > 
> > > > > Note that SLUB debug is unable to detect uninitialized values with
> > > > > bit-to-bit precision, neither have I heard of anyone using it for
> > > > > detecting uses of uninitialized memory in the kernel at all.
> > > > > The purpose of SLUB debug is to detect corruptions of freed memory.
> > > > 
> > > > The point is if developers have SLUB debug enabled, all the free objects will be
> > > > poisoned, so what's the point of checking uninitialized memory there?
> > > 
> > > You are right, SLUB debug has to be handled separately. If I'm
> > > understanding correctly, right now KMSAN poisons freed memory before
> > > SLUB debug wipes it, therefore the memory will count as initialized.
> > > On the other hand, newly allocated memory is still marked as
> > > uninitialized, so a lot of bugs still remain detectable.
> > 
> > Yes, but newly allocated slub objects will be poisoned as well.
> 
> As far as I can tell, KMSAN hook marking newly allocated objects as
> uninitialized is called after slub poisoning.
> Therefore these allocations will be treated by KMSAN as uninitialized.
> > > TBH, I haven't tested KMSAN with SLUB debug good enough. Note that
> > > it's anyway a separate build that requires Clang, so right now it
> > > doesn't make much sense to combine KMSAN and SLUB debug together.
> > 
> > Can't you just build a debug kernel with SLUB debug enabled but dropping this
> > patch? If there is an uninitialized memory here leading to data corruption, SLUB
> > debug should be detected as well as this patch. If not, it needs to understand
> > why.
> 
> Sorry, there might be some misunderstanding here.
> KMSAN keeps the state of heap objects separately by keeping exactly
> the same amount of initialized/uninitialized bits for every
> allocation.
> A call to kmsan_slab_alloc()/kmsan_slab_free() will mark an allocation
> as uninitialized for KMSAN. Not doing so will result in false reports.
> A call to memset(object, POISON_FREE, object_size) performed by SLAB
> debug will actually mark this allocation as initialized from KMSAN
> point of view, because we're memsetting a range with initialized data.
> Note that use of uninitialized data doesn't necessarily lead to
> immediate data corruption, so there might be nothing to detect for
> SLUB debug.

Well, SLUB debug would mark any access of uninitialized memory as data
corruption with the help of poisoning and red zoning.

It is true that KMSAN might be doing a bit more than SLUB debug here, but the
question is if it is worth the maintenance burden? Do you have any existing bugs
to show that KMSAN would find by this patch but SLUB debug can't?

> > 
> > > > > > > 
> > > > > > > Signed-off-by: Alexander Potapenko <glider@google.com>
> > > > > > > To: Alexander Potapenko <glider@google.com>
> > > > > > > Cc: Andrew Morton <akpm@linux-foundation.org>
> > > > > > > Cc: Vegard Nossum <vegard.nossum@oracle.com>
> > > > > > > Cc: Dmitry Vyukov <dvyukov@google.com>
> > > > > > > Cc: linux-mm@kvack.org
> > > > > > > ---
> > > > > > > 
> > > > > > > Change-Id: I51103b7981d3aabed747d0c85cbdc85568665871
> > > > > > > ---
> > > > > > >  mm/slub.c | 37 +++++++++++++++++++++++++++++++------
> > > > > > >  1 file changed, 31 insertions(+), 6 deletions(-)
> > > > > > > 
> > > > > > > diff --git a/mm/slub.c b/mm/slub.c
> > > > > > > index 3d63ae320d31..3d6d4c63446e 100644
> > > > > > > --- a/mm/slub.c
> > > > > > > +++ b/mm/slub.c
> > > > > > > @@ -21,6 +21,8 @@
> > > > > > >  #include <linux/proc_fs.h>
> > > > > > >  #include <linux/seq_file.h>
> > > > > > >  #include <linux/kasan.h>
> > > > > > > +#include <linux/kmsan.h>
> > > > > > > +#include <linux/kmsan-checks.h> /* KMSAN_INIT_VALUE */
> > > > > > >  #include <linux/cpu.h>
> > > > > > >  #include <linux/cpuset.h>
> > > > > > >  #include <linux/mempolicy.h>
> > > > > > > @@ -285,17 +287,27 @@ static void prefetch_freepointer(const struct kmem_cache *s, void *object)
> > > > > > >       prefetch(object + s->offset);
> > > > > > >  }
> > > > > > > 
> > > > > > > +/*
> > > > > > > + * When running under KMSAN, get_freepointer_safe() may return an uninitialized
> > > > > > > + * pointer value in the case the current thread loses the race for the next
> > > > > > > + * memory chunk in the freelist. In that case this_cpu_cmpxchg_double() in
> > > > > > > + * slab_alloc_node() will fail, so the uninitialized value won't be used, but
> > > > > > > + * KMSAN will still check all arguments of cmpxchg because of imperfect
> > > > > > > + * handling of inline assembly.
> > > > > > > + * To work around this problem, use KMSAN_INIT_VALUE() to force initialize the
> > > > > > > + * return value of get_freepointer_safe().
> > > > > > > + */
> > > > > > >  static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
> > > > > > >  {
> > > > > > >       unsigned long freepointer_addr;
> > > > > > >       void *p;
> > > > > > > 
> > > > > > >       if (!debug_pagealloc_enabled())
> > > > > > > -             return get_freepointer(s, object);
> > > > > > > +             return KMSAN_INIT_VALUE(get_freepointer(s, object));
> > > > > > > 
> > > > > > >       freepointer_addr = (unsigned long)object + s->offset;
> > > > > > >       probe_kernel_read(&p, (void **)freepointer_addr, sizeof(p));
> > > > > > > -     return freelist_ptr(s, p, freepointer_addr);
> > > > > > > +     return KMSAN_INIT_VALUE(freelist_ptr(s, p, freepointer_addr));
> > > > > > >  }
> > > > > > > 
> > > > > > >  static inline void set_freepointer(struct kmem_cache *s, void *object, void *fp)
> > > > > > > @@ -1390,6 +1402,7 @@ static inline void *kmalloc_large_node_hook(void *ptr, size_t size, gfp_t flags)
> > > > > > >       ptr = kasan_kmalloc_large(ptr, size, flags);
> > > > > > >       /* As ptr might get tagged, call kmemleak hook after KASAN. */
> > > > > > >       kmemleak_alloc(ptr, size, 1, flags);
> > > > > > > +     kmsan_kmalloc_large(ptr, size, flags);
> > > > > > >       return ptr;
> > > > > > >  }
> > > > > > > 
> > > > > > > @@ -1397,6 +1410,7 @@ static __always_inline void kfree_hook(void *x)
> > > > > > >  {
> > > > > > >       kmemleak_free(x);
> > > > > > >       kasan_kfree_large(x, _RET_IP_);
> > > > > > > +     kmsan_kfree_large(x);
> > > > > > >  }
> > > > > > > 
> > > > > > >  static __always_inline bool slab_free_hook(struct kmem_cache *s, void *x)
> > > > > > > @@ -1453,6 +1467,12 @@ static inline bool slab_free_freelist_hook(struct kmem_cache *s,
> > > > > > >               } while (object != old_tail);
> > > > > > >       }
> > > > > > > 
> > > > > > > +     do {
> > > > > > > +             object = next;
> > > > > > > +             next = get_freepointer(s, object);
> > > > > > > +             kmsan_slab_free(s, object);
> > > > > > > +     } while (object != old_tail);
> > > > > > > +
> > > > > > >  /*
> > > > > > >   * Compiler cannot detect this function can be removed if slab_free_hook()
> > > > > > >   * evaluates to nothing.  Thus, catch all relevant config debug options here.
> > > > > > > @@ -2769,6 +2789,7 @@ static __always_inline void *slab_alloc_node(struct kmem_cache *s,
> > > > > > >       if (unlikely(slab_want_init_on_alloc(gfpflags, s)) && object)
> > > > > > >               memset(object, 0, s->object_size);
> > > > > > > 
> > > > > > > +     kmsan_slab_alloc(s, object, gfpflags);
> > > > > > >       slab_post_alloc_hook(s, gfpflags, 1, &object);
> > > > > > > 
> > > > > > >       return object;
> > > > > > > @@ -2797,6 +2818,7 @@ void *kmem_cache_alloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size)
> > > > > > >       void *ret = slab_alloc(s, gfpflags, _RET_IP_);
> > > > > > >       trace_kmalloc(_RET_IP_, ret, size, s->size, gfpflags);
> > > > > > >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > > > > > > +
> > > > > > >       return ret;
> > > > > > >  }
> > > > > > >  EXPORT_SYMBOL(kmem_cache_alloc_trace);
> > > > > > > @@ -2809,7 +2831,6 @@ void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t gfpflags, int node)
> > > > > > > 
> > > > > > >       trace_kmem_cache_alloc_node(_RET_IP_, ret,
> > > > > > >                                   s->object_size, s->size, gfpflags, node);
> > > > > > > -
> > > > > > >       return ret;
> > > > > > >  }
> > > > > > >  EXPORT_SYMBOL(kmem_cache_alloc_node);
> > > > > > > @@ -2825,6 +2846,7 @@ void *kmem_cache_alloc_node_trace(struct kmem_cache *s,
> > > > > > >                          size, s->size, gfpflags, node);
> > > > > > > 
> > > > > > >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > > > > > > +
> > > > > > >       return ret;
> > > > > > >  }
> > > > > > >  EXPORT_SYMBOL(kmem_cache_alloc_node_trace);
> > > > > > > @@ -3150,7 +3172,7 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> > > > > > >                         void **p)
> > > > > > >  {
> > > > > > >       struct kmem_cache_cpu *c;
> > > > > > > -     int i;
> > > > > > > +     int i, j;
> > > > > > > 
> > > > > > >       /* memcg and kmem_cache debug support */
> > > > > > >       s = slab_pre_alloc_hook(s, flags);
> > > > > > > @@ -3188,11 +3210,11 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> > > > > > > 
> > > > > > >       /* Clear memory outside IRQ disabled fastpath loop */
> > > > > > >       if (unlikely(slab_want_init_on_alloc(flags, s))) {
> > > > > > > -             int j;
> > > > > > > -
> > > > > > >               for (j = 0; j < i; j++)
> > > > > > >                       memset(p[j], 0, s->object_size);
> > > > > > >       }
> > > > > > > +     for (j = 0; j < i; j++)
> > > > > > > +             kmsan_slab_alloc(s, p[j], flags);
> > > > > > > 
> > > > > > >       /* memcg and kmem_cache debug support */
> > > > > > >       slab_post_alloc_hook(s, flags, size, p);
> > > > > > > @@ -3793,6 +3815,7 @@ static int __init setup_slub_min_objects(char *str)
> > > > > > > 
> > > > > > >  __setup("slub_min_objects=", setup_slub_min_objects);
> > > > > > > 
> > > > > > > +__no_sanitize_memory
> > > > > > >  void *__kmalloc(size_t size, gfp_t flags)
> > > > > > >  {
> > > > > > >       struct kmem_cache *s;
> > > > > > > @@ -5698,6 +5721,7 @@ static char *create_unique_id(struct kmem_cache *s)
> > > > > > >       p += sprintf(p, "%07u", s->size);
> > > > > > > 
> > > > > > >       BUG_ON(p > name + ID_STR_LENGTH - 1);
> > > > > > > +     kmsan_unpoison_shadow(name, p - name);
> > > > > > >       return name;
> > > > > > >  }
> > > > > > > 
> > > > > > > @@ -5847,6 +5871,7 @@ static int sysfs_slab_alias(struct kmem_cache *s, const char *name)
> > > > > > >       al->name = name;
> > > > > > >       al->next = alias_list;
> > > > > > >       alias_list = al;
> > > > > > > +     kmsan_unpoison_shadow(al, sizeof(struct saved_alias));
> > > > > > >       return 0;
> > > > > > >  }
> > > > > > > 
> > > > > 
> > > > > 
> > > > > 
> > > 
> > > 
> > > 
> 
> 
> 

Re: [PATCH RFC v1 23/26] kmsan: unpoisoning buffers from devices etc.

[Christoph Hellwig]

On Fri, Oct 18, 2019 at 11:43:01AM +0200, glider@google.com wrote:
> When data is copied to memory from a device KMSAN should treat it as
> initialized. In most cases it's enough to just unpoison the buffer that
> is known to come from a device.
> In the case with __do_page_cache_readahead() and bio_copy_user_iov() we
> have to mark the whole pages as ignored by KMSAN, as it's not obvious
> where these pages are read again.

I'm not sure how your annotations work because you only Cced me on this
one patch.  Don't do that.

Re: [PATCH RFC v1 18/26] kmsan: mm: call KMSAN hooks from SLUB code

[Alexander Potapenko]

On Fri, Oct 18, 2019 at 5:13 PM Qian Cai <cai@lca.pw> wrote:
>
> On Fri, 2019-10-18 at 16:54 +0200, Alexander Potapenko wrote:
> > On Fri, Oct 18, 2019 at 4:42 PM Qian Cai <cai@lca.pw> wrote:
> > >
> > > On Fri, 2019-10-18 at 15:55 +0200, Alexander Potapenko wrote:
> > > > On Fri, Oct 18, 2019 at 3:42 PM Qian Cai <cai@lca.pw> wrote:
> > > > >
> > > > > On Fri, 2019-10-18 at 15:33 +0200, Alexander Potapenko wrote:
> > > > > > On Fri, Oct 18, 2019 at 3:22 PM Qian Cai <cai@lca.pw> wrote:
> > > > > > >
> > > > > > > On Fri, 2019-10-18 at 11:42 +0200, glider@google.com wrote:
> > > > > > > > In order to report uninitialized memory coming from heap allocations
> > > > > > > > KMSAN has to poison them unless they're created with __GFP_ZERO.
> > > > > > > >
> > > > > > > > It's handy that we need KMSAN hooks in the places where
> > > > > > > > init_on_alloc/init_on_free initialization is performed.
> > > > > > >
> > > > > > > Well, there is SLUB debug which has red zoning and poisoning checks. What's
> > > > > > > value of this patch?
> > > > > >
> > > > > > Sorry, are you talking about the whole patchset or just this patch?
> > > > >
> > > > > Just this patch.
> > > > >
> > > > > > Note that SLUB debug is unable to detect uninitialized values with
> > > > > > bit-to-bit precision, neither have I heard of anyone using it for
> > > > > > detecting uses of uninitialized memory in the kernel at all.
> > > > > > The purpose of SLUB debug is to detect corruptions of freed memory.
> > > > >
> > > > > The point is if developers have SLUB debug enabled, all the free objects will be
> > > > > poisoned, so what's the point of checking uninitialized memory there?
> > > >
> > > > You are right, SLUB debug has to be handled separately. If I'm
> > > > understanding correctly, right now KMSAN poisons freed memory before
> > > > SLUB debug wipes it, therefore the memory will count as initialized.
> > > > On the other hand, newly allocated memory is still marked as
> > > > uninitialized, so a lot of bugs still remain detectable.
> > >
> > > Yes, but newly allocated slub objects will be poisoned as well.
> >
> > As far as I can tell, KMSAN hook marking newly allocated objects as
> > uninitialized is called after slub poisoning.
> > Therefore these allocations will be treated by KMSAN as uninitialized.
> > > > TBH, I haven't tested KMSAN with SLUB debug good enough. Note that
> > > > it's anyway a separate build that requires Clang, so right now it
> > > > doesn't make much sense to combine KMSAN and SLUB debug together.
> > >
> > > Can't you just build a debug kernel with SLUB debug enabled but dropping this
> > > patch? If there is an uninitialized memory here leading to data corruption, SLUB
> > > debug should be detected as well as this patch. If not, it needs to understand
> > > why.
> >
> > Sorry, there might be some misunderstanding here.
> > KMSAN keeps the state of heap objects separately by keeping exactly
> > the same amount of initialized/uninitialized bits for every
> > allocation.
> > A call to kmsan_slab_alloc()/kmsan_slab_free() will mark an allocation
> > as uninitialized for KMSAN. Not doing so will result in false reports.
> > A call to memset(object, POISON_FREE, object_size) performed by SLAB
> > debug will actually mark this allocation as initialized from KMSAN
> > point of view, because we're memsetting a range with initialized data.
> > Note that use of uninitialized data doesn't necessarily lead to
> > immediate data corruption, so there might be nothing to detect for
> > SLUB debug.
>
> Well, SLUB debug would mark any access of uninitialized memory as data
> corruption with the help of poisoning and red zoning.
>
> It is true that KMSAN might be doing a bit more than SLUB debug here, but the
> question is if it is worth the maintenance burden? Do you have any existing bugs
> to show that KMSAN would find by this patch but SLUB debug can't?
Consider the following case:

  int *a = kmalloc(sizeof(int), GFP_KERNEL);
  if (*a) {
    // do something, e.g. become root
  }

IIUC SLUB poisoning will result in the branch being taken, but as long
as there are no invalid accesses the kernel won't crash and no errors
will be reported.
KMSAN however will report an error, because the contents of the buffer
pointed to by |a| are poisoned by kmsan_slab_alloc().
The compiler instrumentation will retrieve the shadow value for *a,
compare it to zero and report a use of uninitialized memory.

Another example would be copying the contents of newly allocated
buffer to the userspace.

> > >
> > > > > > > >
> > > > > > > > Signed-off-by: Alexander Potapenko <glider@google.com>
> > > > > > > > To: Alexander Potapenko <glider@google.com>
> > > > > > > > Cc: Andrew Morton <akpm@linux-foundation.org>
> > > > > > > > Cc: Vegard Nossum <vegard.nossum@oracle.com>
> > > > > > > > Cc: Dmitry Vyukov <dvyukov@google.com>
> > > > > > > > Cc: linux-mm@kvack.org
> > > > > > > > ---
> > > > > > > >
> > > > > > > > Change-Id: I51103b7981d3aabed747d0c85cbdc85568665871
> > > > > > > > ---
> > > > > > > >  mm/slub.c | 37 +++++++++++++++++++++++++++++++------
> > > > > > > >  1 file changed, 31 insertions(+), 6 deletions(-)
> > > > > > > >
> > > > > > > > diff --git a/mm/slub.c b/mm/slub.c
> > > > > > > > index 3d63ae320d31..3d6d4c63446e 100644
> > > > > > > > --- a/mm/slub.c
> > > > > > > > +++ b/mm/slub.c
> > > > > > > > @@ -21,6 +21,8 @@
> > > > > > > >  #include <linux/proc_fs.h>
> > > > > > > >  #include <linux/seq_file.h>
> > > > > > > >  #include <linux/kasan.h>
> > > > > > > > +#include <linux/kmsan.h>
> > > > > > > > +#include <linux/kmsan-checks.h> /* KMSAN_INIT_VALUE */
> > > > > > > >  #include <linux/cpu.h>
> > > > > > > >  #include <linux/cpuset.h>
> > > > > > > >  #include <linux/mempolicy.h>
> > > > > > > > @@ -285,17 +287,27 @@ static void prefetch_freepointer(const struct kmem_cache *s, void *object)
> > > > > > > >       prefetch(object + s->offset);
> > > > > > > >  }
> > > > > > > >
> > > > > > > > +/*
> > > > > > > > + * When running under KMSAN, get_freepointer_safe() may return an uninitialized
> > > > > > > > + * pointer value in the case the current thread loses the race for the next
> > > > > > > > + * memory chunk in the freelist. In that case this_cpu_cmpxchg_double() in
> > > > > > > > + * slab_alloc_node() will fail, so the uninitialized value won't be used, but
> > > > > > > > + * KMSAN will still check all arguments of cmpxchg because of imperfect
> > > > > > > > + * handling of inline assembly.
> > > > > > > > + * To work around this problem, use KMSAN_INIT_VALUE() to force initialize the
> > > > > > > > + * return value of get_freepointer_safe().
> > > > > > > > + */
> > > > > > > >  static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
> > > > > > > >  {
> > > > > > > >       unsigned long freepointer_addr;
> > > > > > > >       void *p;
> > > > > > > >
> > > > > > > >       if (!debug_pagealloc_enabled())
> > > > > > > > -             return get_freepointer(s, object);
> > > > > > > > +             return KMSAN_INIT_VALUE(get_freepointer(s, object));
> > > > > > > >
> > > > > > > >       freepointer_addr = (unsigned long)object + s->offset;
> > > > > > > >       probe_kernel_read(&p, (void **)freepointer_addr, sizeof(p));
> > > > > > > > -     return freelist_ptr(s, p, freepointer_addr);
> > > > > > > > +     return KMSAN_INIT_VALUE(freelist_ptr(s, p, freepointer_addr));
> > > > > > > >  }
> > > > > > > >
> > > > > > > >  static inline void set_freepointer(struct kmem_cache *s, void *object, void *fp)
> > > > > > > > @@ -1390,6 +1402,7 @@ static inline void *kmalloc_large_node_hook(void *ptr, size_t size, gfp_t flags)
> > > > > > > >       ptr = kasan_kmalloc_large(ptr, size, flags);
> > > > > > > >       /* As ptr might get tagged, call kmemleak hook after KASAN. */
> > > > > > > >       kmemleak_alloc(ptr, size, 1, flags);
> > > > > > > > +     kmsan_kmalloc_large(ptr, size, flags);
> > > > > > > >       return ptr;
> > > > > > > >  }
> > > > > > > >
> > > > > > > > @@ -1397,6 +1410,7 @@ static __always_inline void kfree_hook(void *x)
> > > > > > > >  {
> > > > > > > >       kmemleak_free(x);
> > > > > > > >       kasan_kfree_large(x, _RET_IP_);
> > > > > > > > +     kmsan_kfree_large(x);
> > > > > > > >  }
> > > > > > > >
> > > > > > > >  static __always_inline bool slab_free_hook(struct kmem_cache *s, void *x)
> > > > > > > > @@ -1453,6 +1467,12 @@ static inline bool slab_free_freelist_hook(struct kmem_cache *s,
> > > > > > > >               } while (object != old_tail);
> > > > > > > >       }
> > > > > > > >
> > > > > > > > +     do {
> > > > > > > > +             object = next;
> > > > > > > > +             next = get_freepointer(s, object);
> > > > > > > > +             kmsan_slab_free(s, object);
> > > > > > > > +     } while (object != old_tail);
> > > > > > > > +
> > > > > > > >  /*
> > > > > > > >   * Compiler cannot detect this function can be removed if slab_free_hook()
> > > > > > > >   * evaluates to nothing.  Thus, catch all relevant config debug options here.
> > > > > > > > @@ -2769,6 +2789,7 @@ static __always_inline void *slab_alloc_node(struct kmem_cache *s,
> > > > > > > >       if (unlikely(slab_want_init_on_alloc(gfpflags, s)) && object)
> > > > > > > >               memset(object, 0, s->object_size);
> > > > > > > >
> > > > > > > > +     kmsan_slab_alloc(s, object, gfpflags);
> > > > > > > >       slab_post_alloc_hook(s, gfpflags, 1, &object);
> > > > > > > >
> > > > > > > >       return object;
> > > > > > > > @@ -2797,6 +2818,7 @@ void *kmem_cache_alloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size)
> > > > > > > >       void *ret = slab_alloc(s, gfpflags, _RET_IP_);
> > > > > > > >       trace_kmalloc(_RET_IP_, ret, size, s->size, gfpflags);
> > > > > > > >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > > > > > > > +
> > > > > > > >       return ret;
> > > > > > > >  }
> > > > > > > >  EXPORT_SYMBOL(kmem_cache_alloc_trace);
> > > > > > > > @@ -2809,7 +2831,6 @@ void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t gfpflags, int node)
> > > > > > > >
> > > > > > > >       trace_kmem_cache_alloc_node(_RET_IP_, ret,
> > > > > > > >                                   s->object_size, s->size, gfpflags, node);
> > > > > > > > -
> > > > > > > >       return ret;
> > > > > > > >  }
> > > > > > > >  EXPORT_SYMBOL(kmem_cache_alloc_node);
> > > > > > > > @@ -2825,6 +2846,7 @@ void *kmem_cache_alloc_node_trace(struct kmem_cache *s,
> > > > > > > >                          size, s->size, gfpflags, node);
> > > > > > > >
> > > > > > > >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > > > > > > > +
> > > > > > > >       return ret;
> > > > > > > >  }
> > > > > > > >  EXPORT_SYMBOL(kmem_cache_alloc_node_trace);
> > > > > > > > @@ -3150,7 +3172,7 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> > > > > > > >                         void **p)
> > > > > > > >  {
> > > > > > > >       struct kmem_cache_cpu *c;
> > > > > > > > -     int i;
> > > > > > > > +     int i, j;
> > > > > > > >
> > > > > > > >       /* memcg and kmem_cache debug support */
> > > > > > > >       s = slab_pre_alloc_hook(s, flags);
> > > > > > > > @@ -3188,11 +3210,11 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> > > > > > > >
> > > > > > > >       /* Clear memory outside IRQ disabled fastpath loop */
> > > > > > > >       if (unlikely(slab_want_init_on_alloc(flags, s))) {
> > > > > > > > -             int j;
> > > > > > > > -
> > > > > > > >               for (j = 0; j < i; j++)
> > > > > > > >                       memset(p[j], 0, s->object_size);
> > > > > > > >       }
> > > > > > > > +     for (j = 0; j < i; j++)
> > > > > > > > +             kmsan_slab_alloc(s, p[j], flags);
> > > > > > > >
> > > > > > > >       /* memcg and kmem_cache debug support */
> > > > > > > >       slab_post_alloc_hook(s, flags, size, p);
> > > > > > > > @@ -3793,6 +3815,7 @@ static int __init setup_slub_min_objects(char *str)
> > > > > > > >
> > > > > > > >  __setup("slub_min_objects=", setup_slub_min_objects);
> > > > > > > >
> > > > > > > > +__no_sanitize_memory
> > > > > > > >  void *__kmalloc(size_t size, gfp_t flags)
> > > > > > > >  {
> > > > > > > >       struct kmem_cache *s;
> > > > > > > > @@ -5698,6 +5721,7 @@ static char *create_unique_id(struct kmem_cache *s)
> > > > > > > >       p += sprintf(p, "%07u", s->size);
> > > > > > > >
> > > > > > > >       BUG_ON(p > name + ID_STR_LENGTH - 1);
> > > > > > > > +     kmsan_unpoison_shadow(name, p - name);
> > > > > > > >       return name;
> > > > > > > >  }
> > > > > > > >
> > > > > > > > @@ -5847,6 +5871,7 @@ static int sysfs_slab_alias(struct kmem_cache *s, const char *name)
> > > > > > > >       al->name = name;
> > > > > > > >       al->next = alias_list;
> > > > > > > >       alias_list = al;
> > > > > > > > +     kmsan_unpoison_shadow(al, sizeof(struct saved_alias));
> > > > > > > >       return 0;
> > > > > > > >  }
> > > > > > > >
> > > > > >
> > > > > >
> > > > > >
> > > >
> > > >
> > > >
> >
> >
> >



-- 
Alexander Potapenko
Software Engineer

Google Germany GmbH
Erika-Mann-Straße, 33
80636 München

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg

Re: [PATCH RFC v1 18/26] kmsan: mm: call KMSAN hooks from SLUB code

[Qian Cai]

On Fri, 2019-10-18 at 17:30 +0200, Alexander Potapenko wrote:
> On Fri, Oct 18, 2019 at 5:13 PM Qian Cai <cai@lca.pw> wrote:
> > 
> > On Fri, 2019-10-18 at 16:54 +0200, Alexander Potapenko wrote:
> > > On Fri, Oct 18, 2019 at 4:42 PM Qian Cai <cai@lca.pw> wrote:
> > > > 
> > > > On Fri, 2019-10-18 at 15:55 +0200, Alexander Potapenko wrote:
> > > > > On Fri, Oct 18, 2019 at 3:42 PM Qian Cai <cai@lca.pw> wrote:
> > > > > > 
> > > > > > On Fri, 2019-10-18 at 15:33 +0200, Alexander Potapenko wrote:
> > > > > > > On Fri, Oct 18, 2019 at 3:22 PM Qian Cai <cai@lca.pw> wrote:
> > > > > > > > 
> > > > > > > > On Fri, 2019-10-18 at 11:42 +0200, glider@google.com wrote:
> > > > > > > > > In order to report uninitialized memory coming from heap allocations
> > > > > > > > > KMSAN has to poison them unless they're created with __GFP_ZERO.
> > > > > > > > > 
> > > > > > > > > It's handy that we need KMSAN hooks in the places where
> > > > > > > > > init_on_alloc/init_on_free initialization is performed.
> > > > > > > > 
> > > > > > > > Well, there is SLUB debug which has red zoning and poisoning checks. What's
> > > > > > > > value of this patch?
> > > > > > > 
> > > > > > > Sorry, are you talking about the whole patchset or just this patch?
> > > > > > 
> > > > > > Just this patch.
> > > > > > 
> > > > > > > Note that SLUB debug is unable to detect uninitialized values with
> > > > > > > bit-to-bit precision, neither have I heard of anyone using it for
> > > > > > > detecting uses of uninitialized memory in the kernel at all.
> > > > > > > The purpose of SLUB debug is to detect corruptions of freed memory.
> > > > > > 
> > > > > > The point is if developers have SLUB debug enabled, all the free objects will be
> > > > > > poisoned, so what's the point of checking uninitialized memory there?
> > > > > 
> > > > > You are right, SLUB debug has to be handled separately. If I'm
> > > > > understanding correctly, right now KMSAN poisons freed memory before
> > > > > SLUB debug wipes it, therefore the memory will count as initialized.
> > > > > On the other hand, newly allocated memory is still marked as
> > > > > uninitialized, so a lot of bugs still remain detectable.
> > > > 
> > > > Yes, but newly allocated slub objects will be poisoned as well.
> > > 
> > > As far as I can tell, KMSAN hook marking newly allocated objects as
> > > uninitialized is called after slub poisoning.
> > > Therefore these allocations will be treated by KMSAN as uninitialized.
> > > > > TBH, I haven't tested KMSAN with SLUB debug good enough. Note that
> > > > > it's anyway a separate build that requires Clang, so right now it
> > > > > doesn't make much sense to combine KMSAN and SLUB debug together.
> > > > 
> > > > Can't you just build a debug kernel with SLUB debug enabled but dropping this
> > > > patch? If there is an uninitialized memory here leading to data corruption, SLUB
> > > > debug should be detected as well as this patch. If not, it needs to understand
> > > > why.
> > > 
> > > Sorry, there might be some misunderstanding here.
> > > KMSAN keeps the state of heap objects separately by keeping exactly
> > > the same amount of initialized/uninitialized bits for every
> > > allocation.
> > > A call to kmsan_slab_alloc()/kmsan_slab_free() will mark an allocation
> > > as uninitialized for KMSAN. Not doing so will result in false reports.
> > > A call to memset(object, POISON_FREE, object_size) performed by SLAB
> > > debug will actually mark this allocation as initialized from KMSAN
> > > point of view, because we're memsetting a range with initialized data.
> > > Note that use of uninitialized data doesn't necessarily lead to
> > > immediate data corruption, so there might be nothing to detect for
> > > SLUB debug.
> > 
> > Well, SLUB debug would mark any access of uninitialized memory as data
> > corruption with the help of poisoning and red zoning.
> > 
> > It is true that KMSAN might be doing a bit more than SLUB debug here, but the
> > question is if it is worth the maintenance burden? Do you have any existing bugs
> > to show that KMSAN would find by this patch but SLUB debug can't?
> 
> Consider the following case:
> 
>   int *a = kmalloc(sizeof(int), GFP_KERNEL);
>   if (*a) {
>     // do something, e.g. become root
>   }
> 
> IIUC SLUB poisoning will result in the branch being taken, but as long
> as there are no invalid accesses the kernel won't crash and no errors
> will be reported.
> KMSAN however will report an error, because the contents of the buffer
> pointed to by |a| are poisoned by kmsan_slab_alloc().
> The compiler instrumentation will retrieve the shadow value for *a,
> compare it to zero and report a use of uninitialized memory.
> 
> Another example would be copying the contents of newly allocated
> buffer to the userspace.

Make sense. BTW, it might worth testing this patchset with this MM debug config
based on linux-next to edge out some corner cases if you could,

https://raw.githubusercontent.com/cailca/linux-mm/master/x86.config

> 
> > > > 
> > > > > > > > > 
> > > > > > > > > Signed-off-by: Alexander Potapenko <glider@google.com>
> > > > > > > > > To: Alexander Potapenko <glider@google.com>
> > > > > > > > > Cc: Andrew Morton <akpm@linux-foundation.org>
> > > > > > > > > Cc: Vegard Nossum <vegard.nossum@oracle.com>
> > > > > > > > > Cc: Dmitry Vyukov <dvyukov@google.com>
> > > > > > > > > Cc: linux-mm@kvack.org
> > > > > > > > > ---
> > > > > > > > > 
> > > > > > > > > Change-Id: I51103b7981d3aabed747d0c85cbdc85568665871
> > > > > > > > > ---
> > > > > > > > >  mm/slub.c | 37 +++++++++++++++++++++++++++++++------
> > > > > > > > >  1 file changed, 31 insertions(+), 6 deletions(-)
> > > > > > > > > 
> > > > > > > > > diff --git a/mm/slub.c b/mm/slub.c
> > > > > > > > > index 3d63ae320d31..3d6d4c63446e 100644
> > > > > > > > > --- a/mm/slub.c
> > > > > > > > > +++ b/mm/slub.c
> > > > > > > > > @@ -21,6 +21,8 @@
> > > > > > > > >  #include <linux/proc_fs.h>
> > > > > > > > >  #include <linux/seq_file.h>
> > > > > > > > >  #include <linux/kasan.h>
> > > > > > > > > +#include <linux/kmsan.h>
> > > > > > > > > +#include <linux/kmsan-checks.h> /* KMSAN_INIT_VALUE */
> > > > > > > > >  #include <linux/cpu.h>
> > > > > > > > >  #include <linux/cpuset.h>
> > > > > > > > >  #include <linux/mempolicy.h>
> > > > > > > > > @@ -285,17 +287,27 @@ static void prefetch_freepointer(const struct kmem_cache *s, void *object)
> > > > > > > > >       prefetch(object + s->offset);
> > > > > > > > >  }
> > > > > > > > > 
> > > > > > > > > +/*
> > > > > > > > > + * When running under KMSAN, get_freepointer_safe() may return an uninitialized
> > > > > > > > > + * pointer value in the case the current thread loses the race for the next
> > > > > > > > > + * memory chunk in the freelist. In that case this_cpu_cmpxchg_double() in
> > > > > > > > > + * slab_alloc_node() will fail, so the uninitialized value won't be used, but
> > > > > > > > > + * KMSAN will still check all arguments of cmpxchg because of imperfect
> > > > > > > > > + * handling of inline assembly.
> > > > > > > > > + * To work around this problem, use KMSAN_INIT_VALUE() to force initialize the
> > > > > > > > > + * return value of get_freepointer_safe().
> > > > > > > > > + */
> > > > > > > > >  static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
> > > > > > > > >  {
> > > > > > > > >       unsigned long freepointer_addr;
> > > > > > > > >       void *p;
> > > > > > > > > 
> > > > > > > > >       if (!debug_pagealloc_enabled())
> > > > > > > > > -             return get_freepointer(s, object);
> > > > > > > > > +             return KMSAN_INIT_VALUE(get_freepointer(s, object));
> > > > > > > > > 
> > > > > > > > >       freepointer_addr = (unsigned long)object + s->offset;
> > > > > > > > >       probe_kernel_read(&p, (void **)freepointer_addr, sizeof(p));
> > > > > > > > > -     return freelist_ptr(s, p, freepointer_addr);
> > > > > > > > > +     return KMSAN_INIT_VALUE(freelist_ptr(s, p, freepointer_addr));
> > > > > > > > >  }
> > > > > > > > > 
> > > > > > > > >  static inline void set_freepointer(struct kmem_cache *s, void *object, void *fp)
> > > > > > > > > @@ -1390,6 +1402,7 @@ static inline void *kmalloc_large_node_hook(void *ptr, size_t size, gfp_t flags)
> > > > > > > > >       ptr = kasan_kmalloc_large(ptr, size, flags);
> > > > > > > > >       /* As ptr might get tagged, call kmemleak hook after KASAN. */
> > > > > > > > >       kmemleak_alloc(ptr, size, 1, flags);
> > > > > > > > > +     kmsan_kmalloc_large(ptr, size, flags);
> > > > > > > > >       return ptr;
> > > > > > > > >  }
> > > > > > > > > 
> > > > > > > > > @@ -1397,6 +1410,7 @@ static __always_inline void kfree_hook(void *x)
> > > > > > > > >  {
> > > > > > > > >       kmemleak_free(x);
> > > > > > > > >       kasan_kfree_large(x, _RET_IP_);
> > > > > > > > > +     kmsan_kfree_large(x);
> > > > > > > > >  }
> > > > > > > > > 
> > > > > > > > >  static __always_inline bool slab_free_hook(struct kmem_cache *s, void *x)
> > > > > > > > > @@ -1453,6 +1467,12 @@ static inline bool slab_free_freelist_hook(struct kmem_cache *s,
> > > > > > > > >               } while (object != old_tail);
> > > > > > > > >       }
> > > > > > > > > 
> > > > > > > > > +     do {
> > > > > > > > > +             object = next;
> > > > > > > > > +             next = get_freepointer(s, object);
> > > > > > > > > +             kmsan_slab_free(s, object);
> > > > > > > > > +     } while (object != old_tail);
> > > > > > > > > +
> > > > > > > > >  /*
> > > > > > > > >   * Compiler cannot detect this function can be removed if slab_free_hook()
> > > > > > > > >   * evaluates to nothing.  Thus, catch all relevant config debug options here.
> > > > > > > > > @@ -2769,6 +2789,7 @@ static __always_inline void *slab_alloc_node(struct kmem_cache *s,
> > > > > > > > >       if (unlikely(slab_want_init_on_alloc(gfpflags, s)) && object)
> > > > > > > > >               memset(object, 0, s->object_size);
> > > > > > > > > 
> > > > > > > > > +     kmsan_slab_alloc(s, object, gfpflags);
> > > > > > > > >       slab_post_alloc_hook(s, gfpflags, 1, &object);
> > > > > > > > > 
> > > > > > > > >       return object;
> > > > > > > > > @@ -2797,6 +2818,7 @@ void *kmem_cache_alloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size)
> > > > > > > > >       void *ret = slab_alloc(s, gfpflags, _RET_IP_);
> > > > > > > > >       trace_kmalloc(_RET_IP_, ret, size, s->size, gfpflags);
> > > > > > > > >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > > > > > > > > +
> > > > > > > > >       return ret;
> > > > > > > > >  }
> > > > > > > > >  EXPORT_SYMBOL(kmem_cache_alloc_trace);
> > > > > > > > > @@ -2809,7 +2831,6 @@ void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t gfpflags, int node)
> > > > > > > > > 
> > > > > > > > >       trace_kmem_cache_alloc_node(_RET_IP_, ret,
> > > > > > > > >                                   s->object_size, s->size, gfpflags, node);
> > > > > > > > > -
> > > > > > > > >       return ret;
> > > > > > > > >  }
> > > > > > > > >  EXPORT_SYMBOL(kmem_cache_alloc_node);
> > > > > > > > > @@ -2825,6 +2846,7 @@ void *kmem_cache_alloc_node_trace(struct kmem_cache *s,
> > > > > > > > >                          size, s->size, gfpflags, node);
> > > > > > > > > 
> > > > > > > > >       ret = kasan_kmalloc(s, ret, size, gfpflags);
> > > > > > > > > +
> > > > > > > > >       return ret;
> > > > > > > > >  }
> > > > > > > > >  EXPORT_SYMBOL(kmem_cache_alloc_node_trace);
> > > > > > > > > @@ -3150,7 +3172,7 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> > > > > > > > >                         void **p)
> > > > > > > > >  {
> > > > > > > > >       struct kmem_cache_cpu *c;
> > > > > > > > > -     int i;
> > > > > > > > > +     int i, j;
> > > > > > > > > 
> > > > > > > > >       /* memcg and kmem_cache debug support */
> > > > > > > > >       s = slab_pre_alloc_hook(s, flags);
> > > > > > > > > @@ -3188,11 +3210,11 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> > > > > > > > > 
> > > > > > > > >       /* Clear memory outside IRQ disabled fastpath loop */
> > > > > > > > >       if (unlikely(slab_want_init_on_alloc(flags, s))) {
> > > > > > > > > -             int j;
> > > > > > > > > -
> > > > > > > > >               for (j = 0; j < i; j++)
> > > > > > > > >                       memset(p[j], 0, s->object_size);
> > > > > > > > >       }
> > > > > > > > > +     for (j = 0; j < i; j++)
> > > > > > > > > +             kmsan_slab_alloc(s, p[j], flags);
> > > > > > > > > 
> > > > > > > > >       /* memcg and kmem_cache debug support */
> > > > > > > > >       slab_post_alloc_hook(s, flags, size, p);
> > > > > > > > > @@ -3793,6 +3815,7 @@ static int __init setup_slub_min_objects(char *str)
> > > > > > > > > 
> > > > > > > > >  __setup("slub_min_objects=", setup_slub_min_objects);
> > > > > > > > > 
> > > > > > > > > +__no_sanitize_memory
> > > > > > > > >  void *__kmalloc(size_t size, gfp_t flags)
> > > > > > > > >  {
> > > > > > > > >       struct kmem_cache *s;
> > > > > > > > > @@ -5698,6 +5721,7 @@ static char *create_unique_id(struct kmem_cache *s)
> > > > > > > > >       p += sprintf(p, "%07u", s->size);
> > > > > > > > > 
> > > > > > > > >       BUG_ON(p > name + ID_STR_LENGTH - 1);
> > > > > > > > > +     kmsan_unpoison_shadow(name, p - name);
> > > > > > > > >       return name;
> > > > > > > > >  }
> > > > > > > > > 
> > > > > > > > > @@ -5847,6 +5871,7 @@ static int sysfs_slab_alias(struct kmem_cache *s, const char *name)
> > > > > > > > >       al->name = name;
> > > > > > > > >       al->next = alias_list;
> > > > > > > > >       alias_list = al;
> > > > > > > > > +     kmsan_unpoison_shadow(al, sizeof(struct saved_alias));
> > > > > > > > >       return 0;
> > > > > > > > >  }
> > > > > > > > > 
> > > > > > > 
> > > > > > > 
> > > > > > > 
> > > > > 
> > > > > 
> > > > > 
> > > 
> > > 
> > > 
> 
> 
> 

Re: [PATCH RFC v1 23/26] kmsan: unpoisoning buffers from devices etc.

[Matthew Wilcox]

On Fri, Oct 18, 2019 at 11:43:01AM +0200, glider@google.com wrote:
> When data is copied to memory from a device KMSAN should treat it as
> initialized. In most cases it's enough to just unpoison the buffer that
> is known to come from a device.
> In the case with __do_page_cache_readahead() and bio_copy_user_iov() we
> have to mark the whole pages as ignored by KMSAN, as it's not obvious
> where these pages are read again.

...

> +++ b/mm/filemap.c
> @@ -18,6 +18,7 @@
>  #include <linux/uaccess.h>
>  #include <linux/capability.h>
>  #include <linux/kernel_stat.h>
> +#include <linux/kmsan-checks.h>
>  #include <linux/gfp.h>
>  #include <linux/mm.h>
>  #include <linux/swap.h>
> @@ -2810,6 +2811,8 @@ static struct page *do_read_cache_page(struct address_space *mapping,
>  		page = wait_on_page_read(page);
>  		if (IS_ERR(page))
>  			return page;
> +		/* Assume all pages in page cache are initialized. */
> +		kmsan_unpoison_shadow(page_address(page), PAGE_SIZE);

Why would you do that?  The page cache already keeps track of which
pages are initialised -- the PageUptodate flag is set on them.  Indeed,
just adding a kmsan call to SetPageUptodate and __SetPageUptodate would
probably be a very straightforward way of handling things, and probably
means you can get rid of a lot of these other calls.

Re: [PATCH RFC v1 26/26] net: kasan: kmsan: support CONFIG_GENERIC_CSUM on x86, enable it for KASAN/KMSAN

[Randy Dunlap]

Hi,

On 10/18/19 2:43 AM, glider@google.com wrote:
> diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
> index 6d2b0fb3a871..dfec655d1a7a 100644
> --- a/arch/x86/Kconfig
> +++ b/arch/x86/Kconfig
> @@ -272,6 +272,11 @@ config GENERIC_ISA_DMA
>  	def_bool y
>  	depends on ISA_DMA_API
>  
> +config GENERIC_CSUM
> +	bool
> +	default y if KMSAN || KASAN
> +	default n

Drop the "default n".  It's already the default default and we prefer
not to repeat the default default.

> +
>  config GENERIC_BUG
>  	def_bool y
>  	depends on BUG

ta.
-- 
~Randy

Re: [PATCH RFC v1 05/26] printk_safe: externalize printk_context

[Petr Mladek]

On Fri 2019-10-18 11:42:43, glider@google.com wrote:
> Petr Mladek suggested we use
>   this_cpu_read(printk_context) & PRINTK_SAFE_CONTEXT_MASK
> instead of checking the spinlock status in kmsan_pr_err()

I would like to understand why the check is needed.

My guess is that it prevents a infinite recursion.
Is this the case? It might be possible to debug this using
trace_printk().

It is important. If it is the recursion then there are
two possibilities how to prevent it. Either prevent
calling the recursive printk(). Or prevent calling
the memory checker recursive. I am not sure what makes
more sense.

Is printk() the only piece of code where you need to avoid
recursion? I wonder how it is prevented in the other cases.

> This appears to be less intrusive, although we'll need to declare
> printk_context in some printk header.

It is easier than the original approach. But the main motivation
is that it is more reliable. The spinlock is a global lock.
But it seems that it is enough to check the state of the local
CPU.

Finally, could you please CC me to the patch(es) that are using
this variable? I would actually prefer to be in CC of entire
patchsets so that I see the full context.

Best Regards,
Petr

Re: [PATCH RFC v1 19/26] kmsan: call KMSAN hooks where needed

[Petr Mladek]

On Fri 2019-10-18 11:42:57, glider@google.com wrote:
> Insert KMSAN hooks that check for potential memory errors and/or make
> necessary bookkeeping changes:
>  - allocate/split/deallocate metadata pages in
>    alloc_pages()/split_page()/free_page();
>  - clear page shadow and origins in clear_page(), copy_user_highpage();
>  - copy page metadata in copy_highpage(), wp_page_copy();
>  - handle vmap()/vunmap()/iounmap();
>  - handle task creation and deletion;
>  - handle vprintk();

I looked only at the printk part.

>  - call softirq entry/exit hooks in kernel/softirq.c;
>  - check/initialize memory sent to/read from USB, I2C, and network
> 
> Signed-off-by: Alexander Potapenko <glider@google.com>
> To: Alexander Potapenko <glider@google.com>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> Cc: Eric Dumazet <edumazet@google.com>
> Cc: Wolfram Sang <wsa@the-dreams.de>
> Cc: Petr Mladek <pmladek@suse.com>
> Cc: Vegard Nossum <vegard.nossum@oracle.com>
> Cc: Dmitry Vyukov <dvyukov@google.com>
> Cc: linux-mm@kvack.org

Could you please add into CC also the other printk co-maitainers?

  + Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
  + Steven Rostedt <rostedt@goodmis.org>


> diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
> index ca65327a6de8..f77fdcb5f861 100644
> --- a/kernel/printk/printk.c
> +++ b/kernel/printk/printk.c
> @@ -1914,7 +1914,12 @@ int vprintk_store(int facility, int level,
>  	 * The printf needs to come first; we need the syslog
>  	 * prefix which might be passed-in as a parameter.
>  	 */
> -	text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
> +	/*
> +	 * We've checked the printk arguments in vprintk_emit() already.
> +	 * Initialize |text_len| to prevent the errors from spreading.
> +	 */
> +	text_len = KMSAN_INIT_VALUE(vscnprintf(text, sizeof(textbuf), fmt,
> +					       args));

I am a bit confused by the comment.

What is the exact meaning of KMSAN_INIT_VALUE(), please?
Does it prevent checking fmt again?
Does make the text_len variable special? In which way?


>  	/* mark and strip a trailing newline */
>  	if (text_len && text[text_len-1] == '\n') {
> @@ -1972,6 +1977,7 @@ asmlinkage int vprintk_emit(int facility, int level,
>  	boot_delay_msec(level);
>  	printk_delay();
>  
> +	kmsan_handle_vprintk(&fmt, args);

What does this function, please?
Could I find more details in another patch?

>  	/* This stops the holder of console_sem just where we want him */
>  	logbuf_lock_irqsave(flags);
>  	curr_log_seq = log_next_seq;

Best Regards,
Petr

Re: [PATCH RFC v1 15/26] kmsan: Kconfig changes to disable options incompatible with KMSAN

[Harry Wentland]

On 2019-10-18 5:42 a.m., glider@google.com wrote:
> Certain config options are currently incompatible with KMSAN, disable
> them.
> 
> Signed-off-by: Alexander Potapenko <glider@google.com>
> To: Alexander Potapenko <glider@google.com>
> Cc: Herbert Xu <herbert@gondor.apana.org.au>
> Cc: Harry Wentland <harry.wentland@amd.com>
> Cc: Vegard Nossum <vegard.nossum@oracle.com>
> Cc: Dmitry Vyukov <dvyukov@google.com>
> Cc: linux-mm@kvack.org

Acked-by: Harry Wentland <harry.wentland@amd.com>

Harry

> 
> ---
> 
> Change-Id: I4521f5c764cfaeba2fae79c954925c302c5e3bc0
> ---
>  arch/Kconfig                        |  2 +-
>  arch/x86/Kconfig                    |  1 +
>  arch/x86/Kconfig.debug              |  3 +++
>  crypto/Kconfig                      | 26 ++++++++++++++++++++++++++
>  drivers/gpu/drm/amd/display/Kconfig |  2 +-
>  lib/Kconfig.debug                   |  3 +++
>  6 files changed, 35 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/Kconfig b/arch/Kconfig
> index 5f8a5d84dbbe..d5ca838ae8ad 100644
> --- a/arch/Kconfig
> +++ b/arch/Kconfig
> @@ -843,7 +843,7 @@ config HAVE_ARCH_VMAP_STACK
>  config VMAP_STACK
>  	default y
>  	bool "Use a virtually-mapped stack"
> -	depends on HAVE_ARCH_VMAP_STACK && !KASAN
> +	depends on HAVE_ARCH_VMAP_STACK && !KASAN && !KMSAN
>  	---help---
>  	  Enable this if you want the use virtually-mapped kernel stacks
>  	  with guard pages.  This causes kernel stack overflows to be
> diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
> index 3f83a5c53808..6d2b0fb3a871 100644
> --- a/arch/x86/Kconfig
> +++ b/arch/x86/Kconfig
> @@ -2143,6 +2143,7 @@ config RELOCATABLE
>  config RANDOMIZE_BASE
>  	bool "Randomize the address of the kernel image (KASLR)"
>  	depends on RELOCATABLE
> +	depends on !KMSAN
>  	default y
>  	---help---
>  	  In support of Kernel Address Space Layout Randomization (KASLR),
> diff --git a/arch/x86/Kconfig.debug b/arch/x86/Kconfig.debug
> index bf9cd83de777..db3cd6147829 100644
> --- a/arch/x86/Kconfig.debug
> +++ b/arch/x86/Kconfig.debug
> @@ -292,6 +292,9 @@ choice
>  config UNWINDER_ORC
>  	bool "ORC unwinder"
>  	depends on X86_64
> +	# KMSAN doesn't support UNWINDER_ORC yet,
> +	# see https://github.com/google/kmsan/issues/48.
> +	depends on !KMSAN
>  	select STACK_VALIDATION
>  	---help---
>  	  This option enables the ORC (Oops Rewind Capability) unwinder for
> diff --git a/crypto/Kconfig b/crypto/Kconfig
> index 9e524044d312..37a22711012c 100644
> --- a/crypto/Kconfig
> +++ b/crypto/Kconfig
> @@ -309,11 +309,13 @@ config CRYPTO_AEGIS128
>  config CRYPTO_AEGIS128_SIMD
>  	bool "Support SIMD acceleration for AEGIS-128"
>  	depends on CRYPTO_AEGIS128 && ((ARM || ARM64) && KERNEL_MODE_NEON)
> +	depends on !KMSAN
>  	default y
>  
>  config CRYPTO_AEGIS128_AESNI_SSE2
>  	tristate "AEGIS-128 AEAD algorithm (x86_64 AESNI+SSE2 implementation)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_AEAD
>  	select CRYPTO_SIMD
>  	help
> @@ -571,6 +573,7 @@ config CRYPTO_CRC32C
>  config CRYPTO_CRC32C_INTEL
>  	tristate "CRC32c INTEL hardware acceleration"
>  	depends on X86
> +	depends on !KMSAN
>  	select CRYPTO_HASH
>  	help
>  	  In Intel processor with SSE4.2 supported, the processor will
> @@ -611,6 +614,7 @@ config CRYPTO_CRC32
>  config CRYPTO_CRC32_PCLMUL
>  	tristate "CRC32 PCLMULQDQ hardware acceleration"
>  	depends on X86
> +	depends on !KMSAN
>  	select CRYPTO_HASH
>  	select CRC32
>  	help
> @@ -649,6 +653,7 @@ config CRYPTO_CRCT10DIF
>  config CRYPTO_CRCT10DIF_PCLMUL
>  	tristate "CRCT10DIF PCLMULQDQ hardware acceleration"
>  	depends on X86 && 64BIT && CRC_T10DIF
> +	depends on !KMSAN
>  	select CRYPTO_HASH
>  	help
>  	  For x86_64 processors with SSE4.2 and PCLMULQDQ supported,
> @@ -695,6 +700,7 @@ config CRYPTO_POLY1305
>  config CRYPTO_POLY1305_X86_64
>  	tristate "Poly1305 authenticator algorithm (x86_64/SSE2/AVX2)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_POLY1305
>  	help
>  	  Poly1305 authenticator algorithm, RFC7539.
> @@ -814,6 +820,7 @@ config CRYPTO_SHA1
>  config CRYPTO_SHA1_SSSE3
>  	tristate "SHA1 digest algorithm (SSSE3/AVX/AVX2/SHA-NI)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_SHA1
>  	select CRYPTO_HASH
>  	help
> @@ -825,6 +832,7 @@ config CRYPTO_SHA1_SSSE3
>  config CRYPTO_SHA256_SSSE3
>  	tristate "SHA256 digest algorithm (SSSE3/AVX/AVX2/SHA-NI)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_SHA256
>  	select CRYPTO_HASH
>  	help
> @@ -837,6 +845,7 @@ config CRYPTO_SHA256_SSSE3
>  config CRYPTO_SHA512_SSSE3
>  	tristate "SHA512 digest algorithm (SSSE3/AVX/AVX2)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_SHA512
>  	select CRYPTO_HASH
>  	help
> @@ -1011,6 +1020,7 @@ config CRYPTO_WP512
>  config CRYPTO_GHASH_CLMUL_NI_INTEL
>  	tristate "GHASH hash function (CLMUL-NI accelerated)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_CRYPTD
>  	help
>  	  This is the x86_64 CLMUL-NI accelerated implementation of
> @@ -1064,6 +1074,7 @@ config CRYPTO_AES_TI
>  config CRYPTO_AES_NI_INTEL
>  	tristate "AES cipher algorithms (AES-NI)"
>  	depends on X86
> +	depends on !KMSAN
>  	select CRYPTO_AEAD
>  	select CRYPTO_LIB_AES
>  	select CRYPTO_ALGAPI
> @@ -1190,6 +1201,7 @@ config CRYPTO_BLOWFISH_COMMON
>  config CRYPTO_BLOWFISH_X86_64
>  	tristate "Blowfish cipher algorithm (x86_64)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_BLKCIPHER
>  	select CRYPTO_BLOWFISH_COMMON
>  	help
> @@ -1221,6 +1233,7 @@ config CRYPTO_CAMELLIA_X86_64
>  	tristate "Camellia cipher algorithm (x86_64)"
>  	depends on X86 && 64BIT
>  	depends on CRYPTO
> +	depends on !KMSAN
>  	select CRYPTO_BLKCIPHER
>  	select CRYPTO_GLUE_HELPER_X86
>  	help
> @@ -1238,6 +1251,7 @@ config CRYPTO_CAMELLIA_AESNI_AVX_X86_64
>  	tristate "Camellia cipher algorithm (x86_64/AES-NI/AVX)"
>  	depends on X86 && 64BIT
>  	depends on CRYPTO
> +	depends on !KMSAN
>  	select CRYPTO_BLKCIPHER
>  	select CRYPTO_CAMELLIA_X86_64
>  	select CRYPTO_GLUE_HELPER_X86
> @@ -1258,6 +1272,7 @@ config CRYPTO_CAMELLIA_AESNI_AVX2_X86_64
>  	tristate "Camellia cipher algorithm (x86_64/AES-NI/AVX2)"
>  	depends on X86 && 64BIT
>  	depends on CRYPTO
> +	depends on !KMSAN
>  	select CRYPTO_CAMELLIA_AESNI_AVX_X86_64
>  	help
>  	  Camellia cipher algorithm module (x86_64/AES-NI/AVX2).
> @@ -1303,6 +1318,7 @@ config CRYPTO_CAST5
>  config CRYPTO_CAST5_AVX_X86_64
>  	tristate "CAST5 (CAST-128) cipher algorithm (x86_64/AVX)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_BLKCIPHER
>  	select CRYPTO_CAST5
>  	select CRYPTO_CAST_COMMON
> @@ -1325,6 +1341,7 @@ config CRYPTO_CAST6
>  config CRYPTO_CAST6_AVX_X86_64
>  	tristate "CAST6 (CAST-256) cipher algorithm (x86_64/AVX)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_BLKCIPHER
>  	select CRYPTO_CAST6
>  	select CRYPTO_CAST_COMMON
> @@ -1360,6 +1377,7 @@ config CRYPTO_DES_SPARC64
>  config CRYPTO_DES3_EDE_X86_64
>  	tristate "Triple DES EDE cipher algorithm (x86-64)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_BLKCIPHER
>  	select CRYPTO_LIB_DES
>  	help
> @@ -1426,6 +1444,7 @@ config CRYPTO_CHACHA20
>  config CRYPTO_CHACHA20_X86_64
>  	tristate "ChaCha stream cipher algorithms (x86_64/SSSE3/AVX2/AVX-512VL)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_BLKCIPHER
>  	select CRYPTO_CHACHA20
>  	help
> @@ -1462,6 +1481,7 @@ config CRYPTO_SERPENT
>  config CRYPTO_SERPENT_SSE2_X86_64
>  	tristate "Serpent cipher algorithm (x86_64/SSE2)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_BLKCIPHER
>  	select CRYPTO_GLUE_HELPER_X86
>  	select CRYPTO_SERPENT
> @@ -1481,6 +1501,7 @@ config CRYPTO_SERPENT_SSE2_X86_64
>  config CRYPTO_SERPENT_SSE2_586
>  	tristate "Serpent cipher algorithm (i586/SSE2)"
>  	depends on X86 && !64BIT
> +	depends on !KMSAN
>  	select CRYPTO_BLKCIPHER
>  	select CRYPTO_GLUE_HELPER_X86
>  	select CRYPTO_SERPENT
> @@ -1500,6 +1521,7 @@ config CRYPTO_SERPENT_SSE2_586
>  config CRYPTO_SERPENT_AVX_X86_64
>  	tristate "Serpent cipher algorithm (x86_64/AVX)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_BLKCIPHER
>  	select CRYPTO_GLUE_HELPER_X86
>  	select CRYPTO_SERPENT
> @@ -1520,6 +1542,7 @@ config CRYPTO_SERPENT_AVX_X86_64
>  config CRYPTO_SERPENT_AVX2_X86_64
>  	tristate "Serpent cipher algorithm (x86_64/AVX2)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_SERPENT_AVX_X86_64
>  	help
>  	  Serpent cipher algorithm, by Anderson, Biham & Knudsen.
> @@ -1615,6 +1638,7 @@ config CRYPTO_TWOFISH_586
>  config CRYPTO_TWOFISH_X86_64
>  	tristate "Twofish cipher algorithm (x86_64)"
>  	depends on (X86 || UML_X86) && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_ALGAPI
>  	select CRYPTO_TWOFISH_COMMON
>  	help
> @@ -1631,6 +1655,7 @@ config CRYPTO_TWOFISH_X86_64
>  config CRYPTO_TWOFISH_X86_64_3WAY
>  	tristate "Twofish cipher algorithm (x86_64, 3-way parallel)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_BLKCIPHER
>  	select CRYPTO_TWOFISH_COMMON
>  	select CRYPTO_TWOFISH_X86_64
> @@ -1652,6 +1677,7 @@ config CRYPTO_TWOFISH_X86_64_3WAY
>  config CRYPTO_TWOFISH_AVX_X86_64
>  	tristate "Twofish cipher algorithm (x86_64/AVX)"
>  	depends on X86 && 64BIT
> +	depends on !KMSAN
>  	select CRYPTO_BLKCIPHER
>  	select CRYPTO_GLUE_HELPER_X86
>  	select CRYPTO_SIMD
> diff --git a/drivers/gpu/drm/amd/display/Kconfig b/drivers/gpu/drm/amd/display/Kconfig
> index 71991a28a775..99585eaf4dd0 100644
> --- a/drivers/gpu/drm/amd/display/Kconfig
> +++ b/drivers/gpu/drm/amd/display/Kconfig
> @@ -6,7 +6,7 @@ config DRM_AMD_DC
>  	bool "AMD DC - Enable new display engine"
>  	default y
>  	select SND_HDA_COMPONENT if SND_HDA_CORE
> -	select DRM_AMD_DC_DCN1_0 if X86 && !(KCOV_INSTRUMENT_ALL && KCOV_ENABLE_COMPARISONS)
> +	select DRM_AMD_DC_DCN1_0 if X86 && !(KCOV_INSTRUMENT_ALL && KCOV_ENABLE_COMPARISONS) && !KMSAN
>  	help
>  	  Choose this option if you want to use the new display engine
>  	  support for AMDGPU. This adds required support for Vega and
> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
> index 75c36318943d..a3f6f5d68593 100644
> --- a/lib/Kconfig.debug
> +++ b/lib/Kconfig.debug
> @@ -1068,6 +1068,9 @@ menu "Lock Debugging (spinlocks, mutexes, etc...)"
>  config LOCK_DEBUGGING_SUPPORT
>  	bool
>  	depends on TRACE_IRQFLAGS_SUPPORT && STACKTRACE_SUPPORT && LOCKDEP_SUPPORT
> +	# KMSAN is incompatible with lockdep,
> +	# see https://github.com/google/kmsan/issues/57.
> +	depends on !KMSAN
>  	default y
>  
>  config PROVE_LOCKING
> 

Re: [PATCH RFC v1 05/26] printk_safe: externalize printk_context

[Alexander Potapenko]

On Mon, Oct 21, 2019 at 11:09 AM Petr Mladek <pmladek@suse.com> wrote:
>
> On Fri 2019-10-18 11:42:43, glider@google.com wrote:
> > Petr Mladek suggested we use
> >   this_cpu_read(printk_context) & PRINTK_SAFE_CONTEXT_MASK
> > instead of checking the spinlock status in kmsan_pr_err()
>
> I would like to understand why the check is needed.
>
> My guess is that it prevents a infinite recursion.
> Is this the case? It might be possible to debug this using
> trace_printk().
This indeed used to prevent infinite recursion, however looks like the
check isn't needed anymore.
When I remove it, KMSAN doesn't hang, probably because printk is doing
a better job at deadlock prevention.

What I'm seeing now is that e.g. in the following case:
  ptr = kmalloc(sizeof(int), GFP_KERNEL);
  if (ptr)
    pr_info("true\n");
  else
    pr_info("false\n");

KMSAN detects errors within pr_info(), namely in vprintk_store().
If I understand correctly, printing from that point causes printk to
use the per-cpu buffer, which is flushed once we're done with the
original printing:

[   58.984971][ T8390] BUG: KMSAN: uninit-value in
kmsan_handle_vprintk+0xa0/0xf0
...
[   59.061976][    C0] BUG: KMSAN: uninit-value in vsnprintf+0x3276/0x32b0
...
[   59.062457][    C0] BUG: KMSAN: uninit-value in format_decode+0x17f/0x1900
...
[   59.062961][    C0] BUG: KMSAN: uninit-value in format_decode+0x17f/0x1900
...
[   59.063338][    C0] Lost 6207 message(s)!

So it turns out we'd better disable reporting under logbuf_lock,
otherwise these lost messages will confuse the developers.
Because the first pr_info() isn't called by KMSAN itself, the tool
still needs a way to know it's running inside printk.

> It is important. If it is the recursion then there are
> two possibilities how to prevent it. Either prevent
> calling the recursive printk(). Or prevent calling
> the memory checker recursive. I am not sure what makes
> more sense.
>
> Is printk() the only piece of code where you need to avoid
> recursion? I wonder how it is prevented in the other cases.
>
> > This appears to be less intrusive, although we'll need to declare
> > printk_context in some printk header.
>
> It is easier than the original approach. But the main motivation
> is that it is more reliable. The spinlock is a global lock.
> But it seems that it is enough to check the state of the local
> CPU.
>
> Finally, could you please CC me to the patch(es) that are using
> this variable? I would actually prefer to be in CC of entire
> patchsets so that I see the full context.
>
> Best Regards,
> Petr



-- 
Alexander Potapenko
Software Engineer

Google Germany GmbH
Erika-Mann-Straße, 33
80636 München

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg

Re: [PATCH RFC v1 05/26] printk_safe: externalize printk_context

[Alexander Potapenko]

On Wed, Oct 23, 2019 at 7:57 PM Alexander Potapenko <glider@google.com> wrote:
>
> On Mon, Oct 21, 2019 at 11:09 AM Petr Mladek <pmladek@suse.com> wrote:
> >
> > On Fri 2019-10-18 11:42:43, glider@google.com wrote:
> > > Petr Mladek suggested we use
> > >   this_cpu_read(printk_context) & PRINTK_SAFE_CONTEXT_MASK
> > > instead of checking the spinlock status in kmsan_pr_err()
> >
> > I would like to understand why the check is needed.
> >
> > My guess is that it prevents a infinite recursion.
> > Is this the case? It might be possible to debug this using
> > trace_printk().
> This indeed used to prevent infinite recursion, however looks like the
> check isn't needed anymore.
> When I remove it, KMSAN doesn't hang, probably because printk is doing
> a better job at deadlock prevention.
>
> What I'm seeing now is that e.g. in the following case:
>   ptr = kmalloc(sizeof(int), GFP_KERNEL);
>   if (ptr)
>     pr_info("true\n");
>   else
>     pr_info("false\n");
>
> KMSAN detects errors within pr_info(), namely in vprintk_store().
> If I understand correctly, printing from that point causes printk to
> use the per-cpu buffer, which is flushed once we're done with the
> original printing:
>
> [   58.984971][ T8390] BUG: KMSAN: uninit-value in
> kmsan_handle_vprintk+0xa0/0xf0
> ...
> [   59.061976][    C0] BUG: KMSAN: uninit-value in vsnprintf+0x3276/0x32b0
> ...
> [   59.062457][    C0] BUG: KMSAN: uninit-value in format_decode+0x17f/0x1900
> ...
> [   59.062961][    C0] BUG: KMSAN: uninit-value in format_decode+0x17f/0x1900
> ...
> [   59.063338][    C0] Lost 6207 message(s)!
>
> So it turns out we'd better disable reporting under logbuf_lock,
> otherwise these lost messages will confuse the developers.
> Because the first pr_info() isn't called by KMSAN itself, the tool
> still needs a way to know it's running inside printk.
>
> > It is important. If it is the recursion then there are
> > two possibilities how to prevent it. Either prevent
> > calling the recursive printk(). Or prevent calling
> > the memory checker recursive. I am not sure what makes
> > more sense.
> >
> > Is printk() the only piece of code where you need to avoid
> > recursion? I wonder how it is prevented in the other cases.
> >
> > > This appears to be less intrusive, although we'll need to declare
> > > printk_context in some printk header.
> >
> > It is easier than the original approach. But the main motivation
> > is that it is more reliable. The spinlock is a global lock.
> > But it seems that it is enough to check the state of the local
> > CPU.
> >
> > Finally, could you please CC me to the patch(es) that are using
> > this variable? I would actually prefer to be in CC of entire
> > patchsets so that I see the full context.
I am very sorry about this. I'm going to re-send the patches soon and
CC everyone on every patch.
What I thought to be reducing the email burden on every reviewer
turned out to be really inconvenient.
> > Best Regards,
> > Petr
>
>
>
> --
> Alexander Potapenko
> Software Engineer
>
> Google Germany GmbH
> Erika-Mann-Straße, 33
> 80636 München
>
> Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
> Registergericht und -nummer: Hamburg, HRB 86891
> Sitz der Gesellschaft: Hamburg



-- 
Alexander Potapenko
Software Engineer

Google Germany GmbH
Erika-Mann-Straße, 33
80636 München

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg

Re: [PATCH RFC v1 05/26] printk_safe: externalize printk_context

[Petr Mladek]

On Wed 2019-10-23 19:57:39, Alexander Potapenko wrote:
> On Mon, Oct 21, 2019 at 11:09 AM Petr Mladek <pmladek@suse.com> wrote:
> >
> > On Fri 2019-10-18 11:42:43, glider@google.com wrote:
> > > Petr Mladek suggested we use
> > >   this_cpu_read(printk_context) & PRINTK_SAFE_CONTEXT_MASK
> > > instead of checking the spinlock status in kmsan_pr_err()
> >
> > I would like to understand why the check is needed.
> >
> > My guess is that it prevents a infinite recursion.
> > Is this the case? It might be possible to debug this using
> > trace_printk().
> This indeed used to prevent infinite recursion, however looks like the
> check isn't needed anymore.
> When I remove it, KMSAN doesn't hang, probably because printk is doing
> a better job at deadlock prevention.

Great.

> What I'm seeing now is that e.g. in the following case:
>   ptr = kmalloc(sizeof(int), GFP_KERNEL);
>   if (ptr)
>     pr_info("true\n");
>   else
>     pr_info("false\n");
> 
> KMSAN detects errors within pr_info(), namely in vprintk_store().
> If I understand correctly, printing from that point causes printk to
> use the per-cpu buffer, which is flushed once we're done with the
> original printing:
> 
> [   58.984971][ T8390] BUG: KMSAN: uninit-value in
> kmsan_handle_vprintk+0xa0/0xf0
> ...
> [   59.061976][    C0] BUG: KMSAN: uninit-value in vsnprintf+0x3276/0x32b0
> ...
> [   59.062457][    C0] BUG: KMSAN: uninit-value in format_decode+0x17f/0x1900
> ...
> [   59.062961][    C0] BUG: KMSAN: uninit-value in format_decode+0x17f/0x1900

Please, do you have an explanation where the uninitialized values come
from? Is it a false positive? Or is there really a bug how the
printk() parameters are passed?

> ...
> [   59.063338][    C0] Lost 6207 message(s)!

This message comes from report_message_lost(). It means that many
recursive messages did not fit into the limited per-CPU buffers.

> So it turns out we'd better disable reporting under logbuf_lock,
> otherwise these lost messages will confuse the developers.
> Because the first pr_info() isn't called by KMSAN itself, the tool
> still needs a way to know it's running inside printk.

OK. Well, I still would like to understand why the recursive messages
are printed at all. Is it a bug in the printk() code or a false positive?

There should be no recursive messages if we fix the bug or make
some annotations to prevent the false positives.

That said, I am not completely against disabling the recursive
messages. But it would be pity to hide some real bugs.

Best Regards,
Petr

Re: [PATCH RFC v1 05/26] printk_safe: externalize printk_context

[Alexander Potapenko]

On Thu, Oct 24, 2019 at 2:46 PM Petr Mladek <pmladek@suse.com> wrote:
>
> On Wed 2019-10-23 19:57:39, Alexander Potapenko wrote:
> > On Mon, Oct 21, 2019 at 11:09 AM Petr Mladek <pmladek@suse.com> wrote:
> > >
> > > On Fri 2019-10-18 11:42:43, glider@google.com wrote:
> > > > Petr Mladek suggested we use
> > > >   this_cpu_read(printk_context) & PRINTK_SAFE_CONTEXT_MASK
> > > > instead of checking the spinlock status in kmsan_pr_err()
> > >
> > > I would like to understand why the check is needed.
> > >
> > > My guess is that it prevents a infinite recursion.
> > > Is this the case? It might be possible to debug this using
> > > trace_printk().
> > This indeed used to prevent infinite recursion, however looks like the
> > check isn't needed anymore.
> > When I remove it, KMSAN doesn't hang, probably because printk is doing
> > a better job at deadlock prevention.
>
> Great.
>
> > What I'm seeing now is that e.g. in the following case:
> >   ptr = kmalloc(sizeof(int), GFP_KERNEL);
> >   if (ptr)
> >     pr_info("true\n");
> >   else
> >     pr_info("false\n");
> >
> > KMSAN detects errors within pr_info(), namely in vprintk_store().
> > If I understand correctly, printing from that point causes printk to
> > use the per-cpu buffer, which is flushed once we're done with the
> > original printing:
> >
> > [   58.984971][ T8390] BUG: KMSAN: uninit-value in
> > kmsan_handle_vprintk+0xa0/0xf0
> > ...
> > [   59.061976][    C0] BUG: KMSAN: uninit-value in vsnprintf+0x3276/0x32b0
> > ...
> > [   59.062457][    C0] BUG: KMSAN: uninit-value in format_decode+0x17f/0x1900
> > ...
> > [   59.062961][    C0] BUG: KMSAN: uninit-value in format_decode+0x17f/0x1900
>
> Please, do you have an explanation where the uninitialized values come
> from? Is it a false positive? Or is there really a bug how the
> printk() parameters are passed?
I believe these are true bugs.
The problem is, when we pass an uninitialized argument into printk(),
KMSAN will report an error every time this uninitialized argument is
used.
E.g. for an uninitialized format string there'll be at least
strlen(fmt) reports in format_decode(), followed by several reports in
vsnprintf().
Although these reports seem to be real, printing only the first of
them should be more than enough.
In the future we'll actually want KMSAN to check every printk()
argument (which will require parsing the format string to figure out
the arguments' lengths), but disable reports within printk.

> > ...
> > [   59.063338][    C0] Lost 6207 message(s)!
>
> This message comes from report_message_lost(). It means that many
> recursive messages did not fit into the limited per-CPU buffers.
>
> > So it turns out we'd better disable reporting under logbuf_lock,
> > otherwise these lost messages will confuse the developers.
> > Because the first pr_info() isn't called by KMSAN itself, the tool
> > still needs a way to know it's running inside printk.
>
> OK. Well, I still would like to understand why the recursive messages
> are printed at all. Is it a bug in the printk() code or a false positive?
>
> There should be no recursive messages if we fix the bug or make
> some annotations to prevent the false positives.
>
> That said, I am not completely against disabling the recursive
> messages. But it would be pity to hide some real bugs.
>
> Best Regards,
> Petr


--
Alexander Potapenko
Software Engineer

Google Germany GmbH
Erika-Mann-Straße, 33
80636 München

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg

Re: [PATCH RFC v1 05/26] printk_safe: externalize printk_context

[Petr Mladek]

On Mon 2019-10-28 14:09:29, Alexander Potapenko wrote:
> On Thu, Oct 24, 2019 at 2:46 PM Petr Mladek <pmladek@suse.com> wrote:
> > On Wed 2019-10-23 19:57:39, Alexander Potapenko wrote:
> > > What I'm seeing now is that e.g. in the following case:
> > >   ptr = kmalloc(sizeof(int), GFP_KERNEL);
> > >   if (ptr)
> > >     pr_info("true\n");
> > >   else
> > >     pr_info("false\n");
> > >
> > > KMSAN detects errors within pr_info(), namely in vprintk_store().
> > > If I understand correctly, printing from that point causes printk to
> > > use the per-cpu buffer, which is flushed once we're done with the
> > > original printing:
> > >
> > > [   58.984971][ T8390] BUG: KMSAN: uninit-value in
> > > kmsan_handle_vprintk+0xa0/0xf0
> > > ...
> > > [   59.061976][    C0] BUG: KMSAN: uninit-value in vsnprintf+0x3276/0x32b0
> > > ...
> > > [   59.062457][    C0] BUG: KMSAN: uninit-value in format_decode+0x17f/0x1900
> > > ...
> > > [   59.062961][    C0] BUG: KMSAN: uninit-value in format_decode+0x17f/0x1900
> >
> > Please, do you have an explanation where the uninitialized values come
> > from? Is it a false positive? Or is there really a bug how the
> > printk() parameters are passed?
> I believe these are true bugs.
> The problem is, when we pass an uninitialized argument into printk(),
> KMSAN will report an error every time this uninitialized argument is
> used.

I see, thanks for explanation.

> E.g. for an uninitialized format string there'll be at least
> strlen(fmt) reports in format_decode(), followed by several reports in
> vsnprintf().
> Although these reports seem to be real, printing only the first of
> them should be more than enough.

Isn't this a generic problem? I mean that uninitialized values
can be passed and used also in many other locations.

printk() is special because this problem causes infinite loop. But it
would make sense to report also any other problematic value only once.


> In the future we'll actually want KMSAN to check every printk()
> argument (which will require parsing the format string to figure out
> the arguments' lengths), but disable reports within printk.

What is the motivation for this, please?

It looks to me that you want to do very paranoid checks of variables
passed to printk()? Do you want to prevent printk() crashes? Or
do you want to make sure that printk() produces correct values?

From my POV, printk() is debugging tool. It is used to show values
that people are interested in. On one hand, it might make sense to warn
people that a particular value was not initalized. On the other hand,
printk() is not important for the kernel behavior. It just
reads values and does not affect any behavior.

I would like to understand how many printk-related code is
worth the effort.

Best Regards,
Petr

Re: [PATCH RFC v1 05/26] printk_safe: externalize printk_context

[Alexander Potapenko]

On Tue, Oct 29, 2019 at 1:02 PM Petr Mladek <pmladek@suse.com> wrote:
>
> On Mon 2019-10-28 14:09:29, Alexander Potapenko wrote:
> > On Thu, Oct 24, 2019 at 2:46 PM Petr Mladek <pmladek@suse.com> wrote:
> > > On Wed 2019-10-23 19:57:39, Alexander Potapenko wrote:
> > > > What I'm seeing now is that e.g. in the following case:
> > > >   ptr = kmalloc(sizeof(int), GFP_KERNEL);
> > > >   if (ptr)
> > > >     pr_info("true\n");
> > > >   else
> > > >     pr_info("false\n");
> > > >
> > > > KMSAN detects errors within pr_info(), namely in vprintk_store().
> > > > If I understand correctly, printing from that point causes printk to
> > > > use the per-cpu buffer, which is flushed once we're done with the
> > > > original printing:
> > > >
> > > > [   58.984971][ T8390] BUG: KMSAN: uninit-value in
> > > > kmsan_handle_vprintk+0xa0/0xf0
> > > > ...
> > > > [   59.061976][    C0] BUG: KMSAN: uninit-value in vsnprintf+0x3276/0x32b0
> > > > ...
> > > > [   59.062457][    C0] BUG: KMSAN: uninit-value in format_decode+0x17f/0x1900
> > > > ...
> > > > [   59.062961][    C0] BUG: KMSAN: uninit-value in format_decode+0x17f/0x1900
> > >
> > > Please, do you have an explanation where the uninitialized values come
> > > from? Is it a false positive? Or is there really a bug how the
> > > printk() parameters are passed?
> > I believe these are true bugs.
> > The problem is, when we pass an uninitialized argument into printk(),
> > KMSAN will report an error every time this uninitialized argument is
> > used.
>
> I see, thanks for explanation.
>
> > E.g. for an uninitialized format string there'll be at least
> > strlen(fmt) reports in format_decode(), followed by several reports in
> > vsnprintf().
> > Although these reports seem to be real, printing only the first of
> > them should be more than enough.
>
> Isn't this a generic problem? I mean that uninitialized values
> can be passed and used also in many other locations.
More or less, yes.
> printk() is special because this problem causes infinite loop. But it
> would make sense to report also any other problematic value only once.
Yes, printk is a bit different, because finding errors in it causes
other printk() calls which now don't cause infinite loops, but may
produce out-of-order error messages.

>
> > In the future we'll actually want KMSAN to check every printk()
> > argument (which will require parsing the format string to figure out
> > the arguments' lengths), but disable reports within printk.
>
> What is the motivation for this, please?
>
> It looks to me that you want to do very paranoid checks of variables
> passed to printk()? Do you want to prevent printk() crashes? Or
> do you want to make sure that printk() produces correct values?
Simply passing an uninitialized value to printk() may result in
cryptic error messages from printk guts, which may be harder to
understand without knowing how printk works.
The idea is to check the arguments right before printk, so that we can
print a really simple diagnostic message.
> From my POV, printk() is debugging tool. It is used to show values
> that people are interested in. On one hand, it might make sense to warn
> people that a particular value was not initalized. On the other hand,
> printk() is not important for the kernel behavior. It just
> reads values and does not affect any behavior.
>
> I would like to understand how many printk-related code is
> worth the effort.
I think overall you're right.
The feature I'm talking about isn't a critical part of KMSAN
functionality, so in order to keep the code simple I'd better drop it,
as long as we're able to report bugs when uninitialized memory is
passed to printk.

The only change to printk that I'll probably still have to make is to
initialize the result of vscnprintf() here:
https://github.com/google/kmsan/blob/master/kernel/printk/printk.c#L1921
Passing uninitialized data to vscnprintf() causes its result to also
be uninitialized, which causes an avalanche of new reports in
vprintk_store()
I'll send the updated patch (together with other KMSAN patches) your way today.


> Best Regards,
> Petr



-- 
Alexander Potapenko
Software Engineer

Google Germany GmbH
Erika-Mann-Straße, 33
80636 München

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg

Re: [PATCH RFC v1 19/26] kmsan: call KMSAN hooks where needed

[Alexander Potapenko]

On Mon, Oct 21, 2019 at 11:25 AM Petr Mladek <pmladek@suse.com> wrote:
>
> On Fri 2019-10-18 11:42:57, glider@google.com wrote:
> > Insert KMSAN hooks that check for potential memory errors and/or make
> > necessary bookkeeping changes:
> >  - allocate/split/deallocate metadata pages in
> >    alloc_pages()/split_page()/free_page();
> >  - clear page shadow and origins in clear_page(), copy_user_highpage();
> >  - copy page metadata in copy_highpage(), wp_page_copy();
> >  - handle vmap()/vunmap()/iounmap();
> >  - handle task creation and deletion;
> >  - handle vprintk();
>
> I looked only at the printk part.
>
> >  - call softirq entry/exit hooks in kernel/softirq.c;
> >  - check/initialize memory sent to/read from USB, I2C, and network
> >
> > Signed-off-by: Alexander Potapenko <glider@google.com>
> > To: Alexander Potapenko <glider@google.com>
> > Cc: Andrew Morton <akpm@linux-foundation.org>
> > Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> > Cc: Eric Dumazet <edumazet@google.com>
> > Cc: Wolfram Sang <wsa@the-dreams.de>
> > Cc: Petr Mladek <pmladek@suse.com>
> > Cc: Vegard Nossum <vegard.nossum@oracle.com>
> > Cc: Dmitry Vyukov <dvyukov@google.com>
> > Cc: linux-mm@kvack.org
>
> Could you please add into CC also the other printk co-maitainers?
Will do in the next version of this patch.

>   + Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
>   + Steven Rostedt <rostedt@goodmis.org>
>
>
> > diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
> > index ca65327a6de8..f77fdcb5f861 100644
> > --- a/kernel/printk/printk.c
> > +++ b/kernel/printk/printk.c
> > @@ -1914,7 +1914,12 @@ int vprintk_store(int facility, int level,
> >        * The printf needs to come first; we need the syslog
> >        * prefix which might be passed-in as a parameter.
> >        */
> > -     text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
> > +     /*
> > +      * We've checked the printk arguments in vprintk_emit() already.
> > +      * Initialize |text_len| to prevent the errors from spreading.
> > +      */
> > +     text_len = KMSAN_INIT_VALUE(vscnprintf(text, sizeof(textbuf), fmt,
> > +                                            args));
>
> I am a bit confused by the comment.
Fixed in v2.
> What is the exact meaning of KMSAN_INIT_VALUE(), please?
Hope it'll be clear from the "kmsan: add KMSAN runtime" patch.
In fact KMSAN_INIT_VALUE is a wrapper that turns an uninitialized
value into an initialized one:
https://github.com/google/kmsan/blob/3207d604ad68f7e5defdbbd6e63cb97750f380c1/include/linux/kmsan-checks.h#L58
It passes the value into an identity function that always returns
unpoisoned values.
As a result, from now on text_len is considered initialized,
regardless of what vscnprintf() returned.
Without KMSAN_INIT_VALUE the tool will report a huge number of errors
when text_len is used in vprintk_store() and log_output().
> Does it prevent checking fmt again?
> Does make the text_len variable special? In which way?
>
>
> >       /* mark and strip a trailing newline */
> >       if (text_len && text[text_len-1] == '\n') {
> > @@ -1972,6 +1977,7 @@ asmlinkage int vprintk_emit(int facility, int level,
> >       boot_delay_msec(level);
> >       printk_delay();
> >
> > +     kmsan_handle_vprintk(&fmt, args);
>
> What does this function, please?
> Could I find more details in another patch?
>
> >       /* This stops the holder of console_sem just where we want him */
> >       logbuf_lock_irqsave(flags);
> >       curr_log_seq = log_next_seq;
>
> Best Regards,
> Petr



-- 
Alexander Potapenko
Software Engineer

Google Germany GmbH
Erika-Mann-Straße, 33
80636 München

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg

Re: [PATCH RFC v1 19/26] kmsan: call KMSAN hooks where needed

[Alexander Potapenko]

On Fri, Oct 18, 2019 at 5:02 PM Qian Cai <cai@lca.pw> wrote:
>
> On Fri, 2019-10-18 at 11:42 +0200, glider@google.com wrote:
> > Insert KMSAN hooks that check for potential memory errors and/or make
> > necessary bookkeeping changes:
> >  - allocate/split/deallocate metadata pages in
> >    alloc_pages()/split_page()/free_page();
>
> This also seems unnecessary where there are options like page_poison and
> debug_pagealloc should be able detect uninitialized memory access in the page
> allocator as well. Even KASAN has some of the functionality.
I still believe there's some misunderstanding between us because I
didn't CC you on KMSAN runtime or the documentation.
I'll do this and will be happy to answer questions if you still have them.
Without these hooks, KMSAN will be virtually unusable, because it
won't know about heap allocations.

When a buffer is allocated on heap, KMSAN updates the metadata for
that region of memory, writing 0xFF (meaning uninitialized) to every
shadow byte and a stack ID to every 4 region bytes.
Note that shadow and origin bytes are stored separately and are
generally incompatible with what debug pagealloc/page poison do.
> >  - clear page shadow and origins in clear_page(), copy_user_highpage();
> >  - copy page metadata in copy_highpage(), wp_page_copy();
> >  - handle vmap()/vunmap()/iounmap();
> >  - handle task creation and deletion;
> >  - handle vprintk();
> >  - call softirq entry/exit hooks in kernel/softirq.c;
> >  - check/initialize memory sent to/read from USB, I2C, and network
> >
> > Signed-off-by: Alexander Potapenko <glider@google.com>
> > To: Alexander Potapenko <glider@google.com>
> > Cc: Andrew Morton <akpm@linux-foundation.org>
> > Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> > Cc: Eric Dumazet <edumazet@google.com>
> > Cc: Wolfram Sang <wsa@the-dreams.de>
> > Cc: Petr Mladek <pmladek@suse.com>
> > Cc: Vegard Nossum <vegard.nossum@oracle.com>
> > Cc: Dmitry Vyukov <dvyukov@google.com>
> > Cc: linux-mm@kvack.org
> > ---
> >
> > Change-Id: I1250a928d9263bf71fdaa067a070bdee686ef47b
> > ---
> >  arch/x86/include/asm/page_64.h | 13 +++++++++++++
> >  arch/x86/mm/ioremap.c          |  3 +++
> >  drivers/i2c/i2c-core-base.c    |  2 ++
> >  drivers/usb/core/urb.c         |  2 ++
> >  include/linux/highmem.h        |  4 ++++
> >  kernel/exit.c                  |  2 ++
> >  kernel/fork.c                  |  2 ++
> >  kernel/kthread.c               |  2 ++
> >  kernel/printk/printk.c         |  8 +++++++-
> >  kernel/softirq.c               |  5 +++++
> >  lib/ioremap.c                  |  5 +++++
> >  mm/compaction.c                |  9 +++++++++
> >  mm/gup.c                       |  3 +++
> >  mm/memory.c                    |  2 ++
> >  mm/page_alloc.c                | 16 ++++++++++++++++
> >  mm/vmalloc.c                   | 23 +++++++++++++++++++++--
> >  net/sched/sch_generic.c        |  2 ++
> >  17 files changed, 100 insertions(+), 3 deletions(-)
> >
> > diff --git a/arch/x86/include/asm/page_64.h b/arch/x86/include/asm/page_64.h
> > index 939b1cff4a7b..0ba43d93414f 100644
> > --- a/arch/x86/include/asm/page_64.h
> > +++ b/arch/x86/include/asm/page_64.h
> > @@ -44,14 +44,27 @@ void clear_page_orig(void *page);
> >  void clear_page_rep(void *page);
> >  void clear_page_erms(void *page);
> >
> > +/* This is an assembly header, avoid including too much of kmsan.h */
> > +#ifdef CONFIG_KMSAN
> > +void kmsan_clear_page(void *page_addr);
> > +#endif
> > +__no_sanitize_memory
> >  static inline void clear_page(void *page)
> >  {
> > +#ifdef CONFIG_KMSAN
> > +     /* alternative_call_2() changes |page|. */
> > +     void *page_copy = page;
> > +#endif
> >       alternative_call_2(clear_page_orig,
> >                          clear_page_rep, X86_FEATURE_REP_GOOD,
> >                          clear_page_erms, X86_FEATURE_ERMS,
> >                          "=D" (page),
> >                          "0" (page)
> >                          : "cc", "memory", "rax", "rcx");
> > +#ifdef CONFIG_KMSAN
> > +     /* Clear KMSAN shadow for the pages that have it. */
> > +     kmsan_clear_page(page_copy);
> > +#endif
> >  }
> >
> >  void copy_page(void *to, void *from);
> > diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
> > index a39dcdb5ae34..fdb2abc11a82 100644
> > --- a/arch/x86/mm/ioremap.c
> > +++ b/arch/x86/mm/ioremap.c
> > @@ -7,6 +7,7 @@
> >   * (C) Copyright 1995 1996 Linus Torvalds
> >   */
> >
> > +#include <linux/kmsan.h>
> >  #include <linux/memblock.h>
> >  #include <linux/init.h>
> >  #include <linux/io.h>
> > @@ -451,6 +452,8 @@ void iounmap(volatile void __iomem *addr)
> >               return;
> >       }
> >
> > +     kmsan_iounmap_page_range((unsigned long)addr,
> > +             (unsigned long)addr + get_vm_area_size(p));
> >       free_memtype(p->phys_addr, p->phys_addr + get_vm_area_size(p));
> >
> >       /* Finally remove it */
> > diff --git a/drivers/i2c/i2c-core-base.c b/drivers/i2c/i2c-core-base.c
> > index 5f6a4985f2bc..9685d3399c79 100644
> > --- a/drivers/i2c/i2c-core-base.c
> > +++ b/drivers/i2c/i2c-core-base.c
> > @@ -27,6 +27,7 @@
> >  #include <linux/irqflags.h>
> >  #include <linux/jump_label.h>
> >  #include <linux/kernel.h>
> > +#include <linux/kmsan.h>
> >  #include <linux/module.h>
> >  #include <linux/mutex.h>
> >  #include <linux/of_device.h>
> > @@ -1975,6 +1976,7 @@ int __i2c_transfer(struct i2c_adapter *adap, struct i2c_msg *msgs, int num)
> >                               trace_i2c_write(adap, &msgs[i], i);
> >       }
> >
> > +     kmsan_handle_i2c_transfer(msgs, num);
> >       /* Retry automatically on arbitration loss */
> >       orig_jiffies = jiffies;
> >       for (ret = 0, try = 0; try <= adap->retries; try++) {
> > diff --git a/drivers/usb/core/urb.c b/drivers/usb/core/urb.c
> > index 0eab79f82ce4..5bdb54d71c2e 100644
> > --- a/drivers/usb/core/urb.c
> > +++ b/drivers/usb/core/urb.c
> > @@ -8,6 +8,7 @@
> >  #include <linux/bitops.h>
> >  #include <linux/slab.h>
> >  #include <linux/log2.h>
> > +#include <linux/kmsan-checks.h>
> >  #include <linux/usb.h>
> >  #include <linux/wait.h>
> >  #include <linux/usb/hcd.h>
> > @@ -401,6 +402,7 @@ int usb_submit_urb(struct urb *urb, gfp_t mem_flags)
> >                       URB_SETUP_MAP_SINGLE | URB_SETUP_MAP_LOCAL |
> >                       URB_DMA_SG_COMBINED);
> >       urb->transfer_flags |= (is_out ? URB_DIR_OUT : URB_DIR_IN);
> > +     kmsan_handle_urb(urb, is_out);
> >
> >       if (xfertype != USB_ENDPOINT_XFER_CONTROL &&
> >                       dev->state < USB_STATE_CONFIGURED)
> > diff --git a/include/linux/highmem.h b/include/linux/highmem.h
> > index ea5cdbd8c2c3..623b56f48685 100644
> > --- a/include/linux/highmem.h
> > +++ b/include/linux/highmem.h
> > @@ -5,6 +5,7 @@
> >  #include <linux/fs.h>
> >  #include <linux/kernel.h>
> >  #include <linux/bug.h>
> > +#include <linux/kmsan.h>
> >  #include <linux/mm.h>
> >  #include <linux/uaccess.h>
> >  #include <linux/hardirq.h>
> > @@ -255,6 +256,8 @@ static inline void copy_user_highpage(struct page *to, struct page *from,
> >       vfrom = kmap_atomic(from);
> >       vto = kmap_atomic(to);
> >       copy_user_page(vto, vfrom, vaddr, to);
> > +     /* User pages don't have shadow, just clear the destination. */
> > +     kmsan_clear_page(page_address(to));
> >       kunmap_atomic(vto);
> >       kunmap_atomic(vfrom);
> >  }
> > @@ -270,6 +273,7 @@ static inline void copy_highpage(struct page *to, struct page *from)
> >       vfrom = kmap_atomic(from);
> >       vto = kmap_atomic(to);
> >       copy_page(vto, vfrom);
> > +     kmsan_copy_page_meta(to, from);
> >       kunmap_atomic(vto);
> >       kunmap_atomic(vfrom);
> >  }
> > diff --git a/kernel/exit.c b/kernel/exit.c
> > index a46a50d67002..9e3ce929110b 100644
> > --- a/kernel/exit.c
> > +++ b/kernel/exit.c
> > @@ -60,6 +60,7 @@
> >  #include <linux/writeback.h>
> >  #include <linux/shm.h>
> >  #include <linux/kcov.h>
> > +#include <linux/kmsan.h>
> >  #include <linux/random.h>
> >  #include <linux/rcuwait.h>
> >  #include <linux/compat.h>
> > @@ -719,6 +720,7 @@ void __noreturn do_exit(long code)
> >
> >       profile_task_exit(tsk);
> >       kcov_task_exit(tsk);
> > +     kmsan_task_exit(tsk);
> >
> >       WARN_ON(blk_needs_flush_plug(tsk));
> >
> > diff --git a/kernel/fork.c b/kernel/fork.c
> > index bcdf53125210..0f08952a42dc 100644
> > --- a/kernel/fork.c
> > +++ b/kernel/fork.c
> > @@ -37,6 +37,7 @@
> >  #include <linux/fdtable.h>
> >  #include <linux/iocontext.h>
> >  #include <linux/key.h>
> > +#include <linux/kmsan.h>
> >  #include <linux/binfmts.h>
> >  #include <linux/mman.h>
> >  #include <linux/mmu_notifier.h>
> > @@ -931,6 +932,7 @@ static struct task_struct *dup_task_struct(struct task_struct *orig, int node)
> >       account_kernel_stack(tsk, 1);
> >
> >       kcov_task_init(tsk);
> > +     kmsan_task_create(tsk);
> >
> >  #ifdef CONFIG_FAULT_INJECTION
> >       tsk->fail_nth = 0;
> > diff --git a/kernel/kthread.c b/kernel/kthread.c
> > index 621467c33fef..9b948559a34a 100644
> > --- a/kernel/kthread.c
> > +++ b/kernel/kthread.c
> > @@ -17,6 +17,7 @@
> >  #include <linux/unistd.h>
> >  #include <linux/file.h>
> >  #include <linux/export.h>
> > +#include <linux/kmsan.h>
> >  #include <linux/mutex.h>
> >  #include <linux/slab.h>
> >  #include <linux/freezer.h>
> > @@ -350,6 +351,7 @@ struct task_struct *__kthread_create_on_node(int (*threadfn)(void *data),
> >               set_cpus_allowed_ptr(task, cpu_all_mask);
> >       }
> >       kfree(create);
> > +     kmsan_task_create(task);
> >       return task;
> >  }
> >
> > diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
> > index ca65327a6de8..f77fdcb5f861 100644
> > --- a/kernel/printk/printk.c
> > +++ b/kernel/printk/printk.c
> > @@ -1914,7 +1914,12 @@ int vprintk_store(int facility, int level,
> >        * The printf needs to come first; we need the syslog
> >        * prefix which might be passed-in as a parameter.
> >        */
> > -     text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
> > +     /*
> > +      * We've checked the printk arguments in vprintk_emit() already.
> > +      * Initialize |text_len| to prevent the errors from spreading.
> > +      */
> > +     text_len = KMSAN_INIT_VALUE(vscnprintf(text, sizeof(textbuf), fmt,
> > +                                            args));
> >
> >       /* mark and strip a trailing newline */
> >       if (text_len && text[text_len-1] == '\n') {
> > @@ -1972,6 +1977,7 @@ asmlinkage int vprintk_emit(int facility, int level,
> >       boot_delay_msec(level);
> >       printk_delay();
> >
> > +     kmsan_handle_vprintk(&fmt, args);
> >       /* This stops the holder of console_sem just where we want him */
> >       logbuf_lock_irqsave(flags);
> >       curr_log_seq = log_next_seq;
> > diff --git a/kernel/softirq.c b/kernel/softirq.c
> > index 0427a86743a4..6d566dd68b35 100644
> > --- a/kernel/softirq.c
> > +++ b/kernel/softirq.c
> > @@ -11,6 +11,7 @@
> >
> >  #include <linux/export.h>
> >  #include <linux/kernel_stat.h>
> > +#include <linux/kmsan.h>
> >  #include <linux/interrupt.h>
> >  #include <linux/init.h>
> >  #include <linux/mm.h>
> > @@ -370,7 +371,9 @@ static inline void invoke_softirq(void)
> >                * it is the irq stack, because it should be near empty
> >                * at this stage.
> >                */
> > +             kmsan_softirq_enter();
> >               __do_softirq();
> > +             kmsan_softirq_exit();
> >  #else
> >               /*
> >                * Otherwise, irq_exit() is called on the task stack that can
> > @@ -600,7 +603,9 @@ static void run_ksoftirqd(unsigned int cpu)
> >                * We can safely run softirq on inline stack, as we are not deep
> >                * in the task stack here.
> >                */
> > +             kmsan_softirq_enter();
> >               __do_softirq();
> > +             kmsan_softirq_exit();
> >               local_irq_enable();
> >               cond_resched();
> >               return;
> > diff --git a/lib/ioremap.c b/lib/ioremap.c
> > index 0a2ffadc6d71..5f830cee5bfc 100644
> > --- a/lib/ioremap.c
> > +++ b/lib/ioremap.c
> > @@ -6,6 +6,7 @@
> >   *
> >   * (C) Copyright 1995 1996 Linus Torvalds
> >   */
> > +#include <linux/kmsan.h>
> >  #include <linux/vmalloc.h>
> >  #include <linux/mm.h>
> >  #include <linux/sched.h>
> > @@ -214,6 +215,8 @@ int ioremap_page_range(unsigned long addr,
> >       unsigned long start;
> >       unsigned long next;
> >       int err;
> > +     unsigned long old_addr = addr;
> > +     phys_addr_t old_phys_addr = phys_addr;
> >
> >       might_sleep();
> >       BUG_ON(addr >= end);
> > @@ -228,6 +231,8 @@ int ioremap_page_range(unsigned long addr,
> >       } while (pgd++, phys_addr += (next - addr), addr = next, addr != end);
> >
> >       flush_cache_vmap(start, end);
> > +     if (!err)
> > +             kmsan_ioremap_page_range(old_addr, end, old_phys_addr, prot);
> >
> >       return err;
> >  }
> > diff --git a/mm/compaction.c b/mm/compaction.c
> > index ce08b39d85d4..e9a5898e5fe0 100644
> > --- a/mm/compaction.c
> > +++ b/mm/compaction.c
> > @@ -84,6 +84,15 @@ static void split_map_pages(struct list_head *list)
> >
> >               for (i = 0; i < nr_pages; i++) {
> >                       list_add(&page->lru, &tmp_list);
> > +#ifdef CONFIG_KMSAN
> > +                     /*
> > +                      * TODO(glider): we may lose the metadata when copying
> > +                      * something to these pages. Need to allocate shadow
> > +                      * and origin pages here instead.
> > +                      */
> > +                     page->shadow = NULL;
> > +                     page->origin = NULL;
> > +#endif
> >                       page++;
> >               }
> >       }
> > diff --git a/mm/gup.c b/mm/gup.c
> > index 23a9f9c9d377..c16e4288997d 100644
> > --- a/mm/gup.c
> > +++ b/mm/gup.c
> > @@ -4,6 +4,7 @@
> >  #include <linux/err.h>
> >  #include <linux/spinlock.h>
> >
> > +#include <linux/kmsan.h>
> >  #include <linux/mm.h>
> >  #include <linux/memremap.h>
> >  #include <linux/pagemap.h>
> > @@ -2347,6 +2348,7 @@ int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
> >           gup_fast_permitted(start, end)) {
> >               local_irq_save(flags);
> >               gup_pgd_range(start, end, write ? FOLL_WRITE : 0, pages, &nr);
> > +             kmsan_gup_pgd_range(pages, nr);
> >               local_irq_restore(flags);
> >       }
> >
> > @@ -2416,6 +2418,7 @@ int get_user_pages_fast(unsigned long start, int nr_pages,
> >           gup_fast_permitted(start, end)) {
> >               local_irq_disable();
> >               gup_pgd_range(addr, end, gup_flags, pages, &nr);
> > +             kmsan_gup_pgd_range(pages, nr);
> >               local_irq_enable();
> >               ret = nr;
> >       }
> > diff --git a/mm/memory.c b/mm/memory.c
> > index b1ca51a079f2..48ceacc06e2d 100644
> > --- a/mm/memory.c
> > +++ b/mm/memory.c
> > @@ -51,6 +51,7 @@
> >  #include <linux/highmem.h>
> >  #include <linux/pagemap.h>
> >  #include <linux/memremap.h>
> > +#include <linux/kmsan.h>
> >  #include <linux/ksm.h>
> >  #include <linux/rmap.h>
> >  #include <linux/export.h>
> > @@ -2328,6 +2329,7 @@ static vm_fault_t wp_page_copy(struct vm_fault *vmf)
> >               if (!new_page)
> >                       goto oom;
> >               cow_user_page(new_page, old_page, vmf->address, vma);
> > +             kmsan_copy_page_meta(new_page, old_page);
> >       }
> >
> >       if (mem_cgroup_try_charge_delay(new_page, mm, GFP_KERNEL, &memcg, false))
> > diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> > index c0b2e0306720..f52494e0dcd9 100644
> > --- a/mm/page_alloc.c
> > +++ b/mm/page_alloc.c
> > @@ -26,6 +26,8 @@
> >  #include <linux/compiler.h>
> >  #include <linux/kernel.h>
> >  #include <linux/kasan.h>
> > +#include <linux/kmsan.h>
> > +#include <linux/kmsan-checks.h>
> >  #include <linux/module.h>
> >  #include <linux/suspend.h>
> >  #include <linux/pagevec.h>
> > @@ -1133,6 +1135,7 @@ static __always_inline bool free_pages_prepare(struct page *page,
> >       VM_BUG_ON_PAGE(PageTail(page), page);
> >
> >       trace_mm_page_free(page, order);
> > +     kmsan_free_page(page, order);
> >
> >       /*
> >        * Check tail pages before head page information is cleared to
> > @@ -3121,6 +3124,7 @@ void split_page(struct page *page, unsigned int order)
> >       VM_BUG_ON_PAGE(PageCompound(page), page);
> >       VM_BUG_ON_PAGE(!page_count(page), page);
> >
> > +     kmsan_split_page(page, order);
> >       for (i = 1; i < (1 << order); i++)
> >               set_page_refcounted(page + i);
> >       split_page_owner(page, order);
> > @@ -3253,6 +3257,13 @@ static struct page *rmqueue_pcplist(struct zone *preferred_zone,
> >  /*
> >   * Allocate a page from the given zone. Use pcplists for order-0 allocations.
> >   */
> > +/*
> > + * TODO(glider): rmqueue() may call __msan_poison_alloca() through a call to
> > + * set_pfnblock_flags_mask(). If __msan_poison_alloca() attempts to allocate
> > + * pages for the stack depot, it may call rmqueue() again, which will result
> > + * in a deadlock.
> > + */
> > +__no_sanitize_memory
> >  static inline
> >  struct page *rmqueue(struct zone *preferred_zone,
> >                       struct zone *zone, unsigned int order,
> > @@ -4779,6 +4790,11 @@ __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order, int preferred_nid,
> >
> >       trace_mm_page_alloc(page, order, alloc_mask, ac.migratetype);
> >
> > +     if (page)
> > +             if (kmsan_alloc_page(page, order, gfp_mask)) {
> > +                     __free_pages(page, order);
> > +                     page = NULL;
> > +             }
> >       return page;
> >  }
> >  EXPORT_SYMBOL(__alloc_pages_nodemask);
> > diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> > index a3c70e275f4e..bdf66ffcf02c 100644
> > --- a/mm/vmalloc.c
> > +++ b/mm/vmalloc.c
> > @@ -29,6 +29,7 @@
> >  #include <linux/rcupdate.h>
> >  #include <linux/pfn.h>
> >  #include <linux/kmemleak.h>
> > +#include <linux/kmsan.h>
> >  #include <linux/atomic.h>
> >  #include <linux/compiler.h>
> >  #include <linux/llist.h>
> > @@ -119,7 +120,8 @@ static void vunmap_p4d_range(pgd_t *pgd, unsigned long addr, unsigned long end)
> >       } while (p4d++, addr = next, addr != end);
> >  }
> >
> > -static void vunmap_page_range(unsigned long addr, unsigned long end)
> > +/* Exported for KMSAN, visible in mm/kmsan/kmsan.h only. */
> > +void __vunmap_page_range(unsigned long addr, unsigned long end)
> >  {
> >       pgd_t *pgd;
> >       unsigned long next;
> > @@ -133,6 +135,12 @@ static void vunmap_page_range(unsigned long addr, unsigned long end)
> >               vunmap_p4d_range(pgd, addr, next);
> >       } while (pgd++, addr = next, addr != end);
> >  }
> > +EXPORT_SYMBOL(__vunmap_page_range);
> > +static void vunmap_page_range(unsigned long addr, unsigned long end)
> > +{
> > +     kmsan_vunmap_page_range(addr, end);
> > +     __vunmap_page_range(addr, end);
> > +}
> >
> >  static int vmap_pte_range(pmd_t *pmd, unsigned long addr,
> >               unsigned long end, pgprot_t prot, struct page **pages, int *nr)
> > @@ -216,8 +224,11 @@ static int vmap_p4d_range(pgd_t *pgd, unsigned long addr,
> >   * will have pfns corresponding to the "pages" array.
> >   *
> >   * Ie. pte at addr+N*PAGE_SIZE shall point to pfn corresponding to pages[N]
> > + *
> > + * This function is exported for use in KMSAN, but is only declared in KMSAN
> > + * headers.
> >   */
> > -static int vmap_page_range_noflush(unsigned long start, unsigned long end,
> > +int __vmap_page_range_noflush(unsigned long start, unsigned long end,
> >                                  pgprot_t prot, struct page **pages)
> >  {
> >       pgd_t *pgd;
> > @@ -237,6 +248,14 @@ static int vmap_page_range_noflush(unsigned long start, unsigned long end,
> >
> >       return nr;
> >  }
> > +EXPORT_SYMBOL(__vmap_page_range_noflush);
> > +
> > +static int vmap_page_range_noflush(unsigned long start, unsigned long end,
> > +                                pgprot_t prot, struct page **pages)
> > +{
> > +     kmsan_vmap_page_range_noflush(start, end, prot, pages);
> > +     return __vmap_page_range_noflush(start, end, prot, pages);
> > +}
> >
> >  static int vmap_page_range(unsigned long start, unsigned long end,
> >                          pgprot_t prot, struct page **pages)
> > diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
> > index 17bd8f539bc7..fd22c4a4ba42 100644
> > --- a/net/sched/sch_generic.c
> > +++ b/net/sched/sch_generic.c
> > @@ -11,6 +11,7 @@
> >  #include <linux/module.h>
> >  #include <linux/types.h>
> >  #include <linux/kernel.h>
> > +#include <linux/kmsan-checks.h>
> >  #include <linux/sched.h>
> >  #include <linux/string.h>
> >  #include <linux/errno.h>
> > @@ -659,6 +660,7 @@ static struct sk_buff *pfifo_fast_dequeue(struct Qdisc *qdisc)
> >       } else {
> >               qdisc->empty = true;
> >       }
> > +     kmsan_check_skb(skb);
> >
> >       return skb;
> >  }



-- 
Alexander Potapenko
Software Engineer

Google Germany GmbH
Erika-Mann-Straße, 33
80636 München

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg

Re: [PATCH RFC v1 23/26] kmsan: unpoisoning buffers from devices etc.

[Alexander Potapenko]

On Fri, Oct 18, 2019 at 6:22 PM Matthew Wilcox <willy@infradead.org> wrote:
>
> On Fri, Oct 18, 2019 at 11:43:01AM +0200, glider@google.com wrote:
> > When data is copied to memory from a device KMSAN should treat it as
> > initialized. In most cases it's enough to just unpoison the buffer that
> > is known to come from a device.
> > In the case with __do_page_cache_readahead() and bio_copy_user_iov() we
> > have to mark the whole pages as ignored by KMSAN, as it's not obvious
> > where these pages are read again.
>
> ...
>
> > +++ b/mm/filemap.c
> > @@ -18,6 +18,7 @@
> >  #include <linux/uaccess.h>
> >  #include <linux/capability.h>
> >  #include <linux/kernel_stat.h>
> > +#include <linux/kmsan-checks.h>
> >  #include <linux/gfp.h>
> >  #include <linux/mm.h>
> >  #include <linux/swap.h>
> > @@ -2810,6 +2811,8 @@ static struct page *do_read_cache_page(struct address_space *mapping,
> >               page = wait_on_page_read(page);
> >               if (IS_ERR(page))
> >                       return page;
> > +             /* Assume all pages in page cache are initialized. */
> > +             kmsan_unpoison_shadow(page_address(page), PAGE_SIZE);
>
> Why would you do that?  The page cache already keeps track of which
> pages are initialised -- the PageUptodate flag is set on them.  Indeed,
> just adding a kmsan call to SetPageUptodate and __SetPageUptodate would
> probably be a very straightforward way of handling things, and probably
> means you can get rid of a lot of these other calls.
This seems like a very good thing to do, I'll definitely try that.
I however noticed that __SetPageUptodate is used when copying pages,
not during disk I/O. Is that really so?

We basically need the following behavior: if a device writes to a
page, the contents of that page are considered initialized.
However when the kernel copies one page to another, we must explicitly
copy the source shadow page to the destination.

On a related note, there seems to be a PG_dirty bit that indicates the
page is to be flushed to disk.
What's the best place to check such pages for being initialized, so
that we can also report writes of uninitialized data to the disk?


-- 
Alexander Potapenko
Software Engineer

Google Germany GmbH
Erika-Mann-Straße, 33
80636 München

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg

Re: [PATCH RFC v1 19/26] kmsan: call KMSAN hooks where needed

[Qian Cai]

On Tue, 2019-10-29 at 15:09 +0100, Alexander Potapenko wrote:
> On Fri, Oct 18, 2019 at 5:02 PM Qian Cai <cai@lca.pw> wrote:
> > 
> > On Fri, 2019-10-18 at 11:42 +0200, glider@google.com wrote:
> > > Insert KMSAN hooks that check for potential memory errors and/or make
> > > necessary bookkeeping changes:
> > >  - allocate/split/deallocate metadata pages in
> > >    alloc_pages()/split_page()/free_page();
> > 
> > This also seems unnecessary where there are options like page_poison and
> > debug_pagealloc should be able detect uninitialized memory access in the page
> > allocator as well. Even KASAN has some of the functionality.
> 
> I still believe there's some misunderstanding between us because I
> didn't CC you on KMSAN runtime or the documentation.
> I'll do this and will be happy to answer questions if you still have them.
> Without these hooks, KMSAN will be virtually unusable, because it
> won't know about heap allocations.
> 
> When a buffer is allocated on heap, KMSAN updates the metadata for
> that region of memory, writing 0xFF (meaning uninitialized) to every
> shadow byte and a stack ID to every 4 region bytes.
> Note that shadow and origin bytes are stored separately and are
> generally incompatible with what debug pagealloc/page poison do.

That makes more sense to me now. I was a bit worry about some of those options
could stub one's toe with KMSAN, but I had shared with you the MM debug config
in another email before, so you could try it out, so I don't need to spam you
later once this hit the linux-next.

Re: [PATCH RFC v1 23/26] kmsan: unpoisoning buffers from devices etc.

[Alexander Potapenko]

On Tue, Oct 29, 2019 at 3:45 PM Alexander Potapenko <glider@google.com> wrote:
>
> On Fri, Oct 18, 2019 at 6:22 PM Matthew Wilcox <willy@infradead.org> wrote:
> >
> > On Fri, Oct 18, 2019 at 11:43:01AM +0200, glider@google.com wrote:
> > > When data is copied to memory from a device KMSAN should treat it as
> > > initialized. In most cases it's enough to just unpoison the buffer that
> > > is known to come from a device.
> > > In the case with __do_page_cache_readahead() and bio_copy_user_iov() we
> > > have to mark the whole pages as ignored by KMSAN, as it's not obvious
> > > where these pages are read again.
> >
> > ...
> >
> > > +++ b/mm/filemap.c
> > > @@ -18,6 +18,7 @@
> > >  #include <linux/uaccess.h>
> > >  #include <linux/capability.h>
> > >  #include <linux/kernel_stat.h>
> > > +#include <linux/kmsan-checks.h>
> > >  #include <linux/gfp.h>
> > >  #include <linux/mm.h>
> > >  #include <linux/swap.h>
> > > @@ -2810,6 +2811,8 @@ static struct page *do_read_cache_page(struct address_space *mapping,
> > >               page = wait_on_page_read(page);
> > >               if (IS_ERR(page))
> > >                       return page;
> > > +             /* Assume all pages in page cache are initialized. */
> > > +             kmsan_unpoison_shadow(page_address(page), PAGE_SIZE);
> >
> > Why would you do that?  The page cache already keeps track of which
> > pages are initialised -- the PageUptodate flag is set on them.  Indeed,
> > just adding a kmsan call to SetPageUptodate and __SetPageUptodate would
> > probably be a very straightforward way of handling things, and probably
> > means you can get rid of a lot of these other calls.
> This seems like a very good thing to do, I'll definitely try that.
Hm, turns out there's no need for this particular call to
kmsan_unpoison_memory(),
because the errors that it was suppressing are better addressed by
making READ_ONCE_NOCHECK() return initialized memory.
On the other hand, unpoisoning memory in SetPageUptodate doesn't seem
to address the same errors.

 =====================================================
 BUG: KMSAN: uninit-value in[<      none      >] get_wchan+0x2c6/0x410
arch/x86/kernel/process.c:851
 CPU: 0 PID: 5159 Comm: ps Not tainted 5.4.0-rc5+ #3220
 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-1 04/01/2014
 Call Trace:
 [<     inline     >] __dump_stack lib/dump_stack.c:77
 [<      none      >] dump_stack+0x196/0x1f0 lib/dump_stack.c:113
 [<      none      >] kmsan_report+0x127/0x220 mm/kmsan/kmsan_report.c:108
 [<      none      >] __msan_warning+0x73/0xe0 mm/kmsan/kmsan_instr.c:245
 [<      none      >] get_wchan+0x2c6/0x410 arch/x86/kernel/process.c:851
 [<      none      >] do_task_stat+0x16f3/0x30a0 fs/proc/array.c:522
 [<      none      >] proc_tgid_stat+0xbe/0xf0 fs/proc/array.c:632
 [<      none      >] proc_single_show+0x1a8/0x2b0 fs/proc/base.c:756
 [<      none      >] seq_read+0xad1/0x1da0 fs/seq_file.c:229
 [<      none      >] __vfs_read+0x1a9/0xc90 fs/read_write.c:425
 [<      none      >] vfs_read+0x333/0x6a0 fs/read_write.c:461
 [<      none      >] ksys_read+0x281/0x460 fs/read_write.c:587
 [<     inline     >] __do_sys_read fs/read_write.c:597
 [<      none      >] __se_sys_read+0x92/0xb0 fs/read_write.c:595
 [<      none      >] __x64_sys_read+0x4a/0x70 fs/read_write.c:595
 [<      none      >] do_syscall_64+0xb6/0x160 arch/x86/entry/common.c:291
 [<      none      >] entry_SYSCALL_64_after_hwframe+0x63/0xe7
arch/x86/entry/entry_64.S:177
 Local variable description: ----stat.i@__se_sys_newstat
 Variable was created at:
 [<     inline     >] __do_sys_newstat fs/stat.c:337
 [<      none      >] __se_sys_newstat+0x71/0xac0 fs/stat.c:337
 [<     inline     >] __do_sys_newstat fs/stat.c:337
 [<      none      >] __se_sys_newstat+0x71/0xac0 fs/stat.c:337
 =====================================================



> I however noticed that __SetPageUptodate is used when copying pages,
> not during disk I/O. Is that really so?
>
> We basically need the following behavior: if a device writes to a
> page, the contents of that page are considered initialized.
> However when the kernel copies one page to another, we must explicitly
> copy the source shadow page to the destination.
>
> On a related note, there seems to be a PG_dirty bit that indicates the
> page is to be flushed to disk.
> What's the best place to check such pages for being initialized, so
> that we can also report writes of uninitialized data to the disk?
>
>
> --
> Alexander Potapenko
> Software Engineer
>
> Google Germany GmbH
> Erika-Mann-Straße, 33
> 80636 München
>
> Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
> Registergericht und -nummer: Hamburg, HRB 86891
> Sitz der Gesellschaft: Hamburg



--
Alexander Potapenko
Software Engineer

Google Germany GmbH
Erika-Mann-Straße, 33
80636 München

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg