* [PATCH v2 01/10] mm/kasan: implement kasan_poison_range
2025-09-19 14:57 [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework Ethan Graham
@ 2025-09-19 14:57 ` Ethan Graham
2025-09-23 16:46 ` Andrey Konovalov
2025-09-19 14:57 ` [PATCH v2 02/10] kfuzztest: add user-facing API and data structures Ethan Graham
` (10 subsequent siblings)
11 siblings, 1 reply; 30+ messages in thread
From: Ethan Graham @ 2025-09-19 14:57 UTC (permalink / raw)
To: ethangraham, glider
Cc: andreyknvl, andy, brauner, brendan.higgins, davem, davidgow,
dhowells, dvyukov, elver, herbert, ignat, jack, jannh, johannes,
kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel, linux-mm,
lukas, rmoar, shuah, sj, tarasmadan
From: Ethan Graham <ethangraham@google.com>
Introduce a new helper function, kasan_poison_range(), to encapsulate
the logic for poisoning an arbitrary memory range of a given size, and
expose it publically in <include/linux/kasan.h>.
This is a preparatory change for the upcoming KFuzzTest patches, which
requires the ability to poison the inter-region padding in its input
buffers.
No functional change to any other subsystem is intended by this commit.
Signed-off-by: Ethan Graham <ethangraham@google.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
---
PR v1:
- Enforce KASAN_GRANULE_SIZE alignment for the end of the range in
kasan_poison_range(), and return -EINVAL when this isn't respected.
---
---
include/linux/kasan.h | 11 +++++++++++
mm/kasan/shadow.c | 34 ++++++++++++++++++++++++++++++++++
2 files changed, 45 insertions(+)
diff --git a/include/linux/kasan.h b/include/linux/kasan.h
index 890011071f2b..cd6cdf732378 100644
--- a/include/linux/kasan.h
+++ b/include/linux/kasan.h
@@ -102,6 +102,16 @@ static inline bool kasan_has_integrated_init(void)
}
#ifdef CONFIG_KASAN
+
+/**
+ * kasan_poison_range - poison the memory range [@addr, @addr + @size)
+ *
+ * The exact behavior is subject to alignment with KASAN_GRANULE_SIZE, defined
+ * in <mm/kasan/kasan.h>: if @start is unaligned, the initial partial granule
+ * at the beginning of the range is only poisoned if CONFIG_KASAN_GENERIC=y.
+ */
+int kasan_poison_range(const void *addr, size_t size);
+
void __kasan_unpoison_range(const void *addr, size_t size);
static __always_inline void kasan_unpoison_range(const void *addr, size_t size)
{
@@ -402,6 +412,7 @@ static __always_inline bool kasan_check_byte(const void *addr)
#else /* CONFIG_KASAN */
+static inline int kasan_poison_range(const void *start, size_t size) { return 0; }
static inline void kasan_unpoison_range(const void *address, size_t size) {}
static inline void kasan_poison_pages(struct page *page, unsigned int order,
bool init) {}
diff --git a/mm/kasan/shadow.c b/mm/kasan/shadow.c
index d2c70cd2afb1..7faed02264f2 100644
--- a/mm/kasan/shadow.c
+++ b/mm/kasan/shadow.c
@@ -147,6 +147,40 @@ void kasan_poison(const void *addr, size_t size, u8 value, bool init)
}
EXPORT_SYMBOL_GPL(kasan_poison);
+int kasan_poison_range(const void *addr, size_t size)
+{
+ uintptr_t start_addr = (uintptr_t)addr;
+ uintptr_t head_granule_start;
+ uintptr_t poison_body_start;
+ uintptr_t poison_body_end;
+ size_t head_prefix_size;
+ uintptr_t end_addr;
+
+ if ((start_addr + size) % KASAN_GRANULE_SIZE)
+ return -EINVAL;
+
+ end_addr = ALIGN_DOWN(start_addr + size, KASAN_GRANULE_SIZE);
+ if (start_addr >= end_addr)
+ return -EINVAL;
+
+ head_granule_start = ALIGN_DOWN(start_addr, KASAN_GRANULE_SIZE);
+ head_prefix_size = start_addr - head_granule_start;
+
+ if (IS_ENABLED(CONFIG_KASAN_GENERIC) && head_prefix_size > 0)
+ kasan_poison_last_granule((void *)head_granule_start,
+ head_prefix_size);
+
+ poison_body_start = ALIGN(start_addr, KASAN_GRANULE_SIZE);
+ poison_body_end = ALIGN_DOWN(end_addr, KASAN_GRANULE_SIZE);
+
+ if (poison_body_start < poison_body_end)
+ kasan_poison((void *)poison_body_start,
+ poison_body_end - poison_body_start,
+ KASAN_SLAB_REDZONE, false);
+ return 0;
+}
+EXPORT_SYMBOL(kasan_poison_range);
+
#ifdef CONFIG_KASAN_GENERIC
void kasan_poison_last_granule(const void *addr, size_t size)
{
--
2.51.0.470.ga7dc726c21-goog
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH v2 01/10] mm/kasan: implement kasan_poison_range
2025-09-19 14:57 ` [PATCH v2 01/10] mm/kasan: implement kasan_poison_range Ethan Graham
@ 2025-09-23 16:46 ` Andrey Konovalov
0 siblings, 0 replies; 30+ messages in thread
From: Andrey Konovalov @ 2025-09-23 16:46 UTC (permalink / raw)
To: Ethan Graham
Cc: ethangraham, glider, andy, brauner, brendan.higgins, davem,
davidgow, dhowells, dvyukov, elver, herbert, ignat, jack, jannh,
johannes, kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel,
linux-mm, lukas, rmoar, shuah, sj, tarasmadan
On Fri, Sep 19, 2025 at 4:58 PM Ethan Graham <ethan.w.s.graham@gmail.com> wrote:
>
> From: Ethan Graham <ethangraham@google.com>
>
> Introduce a new helper function, kasan_poison_range(), to encapsulate
> the logic for poisoning an arbitrary memory range of a given size, and
> expose it publically in <include/linux/kasan.h>.
>
> This is a preparatory change for the upcoming KFuzzTest patches, which
> requires the ability to poison the inter-region padding in its input
> buffers.
>
> No functional change to any other subsystem is intended by this commit.
>
> Signed-off-by: Ethan Graham <ethangraham@google.com>
> Reviewed-by: Alexander Potapenko <glider@google.com>
>
> ---
> PR v1:
> - Enforce KASAN_GRANULE_SIZE alignment for the end of the range in
> kasan_poison_range(), and return -EINVAL when this isn't respected.
> ---
> ---
> include/linux/kasan.h | 11 +++++++++++
> mm/kasan/shadow.c | 34 ++++++++++++++++++++++++++++++++++
> 2 files changed, 45 insertions(+)
>
> diff --git a/include/linux/kasan.h b/include/linux/kasan.h
> index 890011071f2b..cd6cdf732378 100644
> --- a/include/linux/kasan.h
> +++ b/include/linux/kasan.h
> @@ -102,6 +102,16 @@ static inline bool kasan_has_integrated_init(void)
> }
>
> #ifdef CONFIG_KASAN
> +
> +/**
> + * kasan_poison_range - poison the memory range [@addr, @addr + @size)
> + *
> + * The exact behavior is subject to alignment with KASAN_GRANULE_SIZE, defined
> + * in <mm/kasan/kasan.h>: if @start is unaligned, the initial partial granule
> + * at the beginning of the range is only poisoned if CONFIG_KASAN_GENERIC=y.
> + */
> +int kasan_poison_range(const void *addr, size_t size);
> +
> void __kasan_unpoison_range(const void *addr, size_t size);
> static __always_inline void kasan_unpoison_range(const void *addr, size_t size)
> {
> @@ -402,6 +412,7 @@ static __always_inline bool kasan_check_byte(const void *addr)
>
> #else /* CONFIG_KASAN */
>
> +static inline int kasan_poison_range(const void *start, size_t size) { return 0; }
> static inline void kasan_unpoison_range(const void *address, size_t size) {}
> static inline void kasan_poison_pages(struct page *page, unsigned int order,
> bool init) {}
> diff --git a/mm/kasan/shadow.c b/mm/kasan/shadow.c
> index d2c70cd2afb1..7faed02264f2 100644
> --- a/mm/kasan/shadow.c
> +++ b/mm/kasan/shadow.c
> @@ -147,6 +147,40 @@ void kasan_poison(const void *addr, size_t size, u8 value, bool init)
> }
> EXPORT_SYMBOL_GPL(kasan_poison);
>
> +int kasan_poison_range(const void *addr, size_t size)
This should go into common.c, otherwise this won't be built with the
HW_TAGS mode enabled.
Also, you need a wrapper with a kasan_enabled() check; see how
kasan_unpoison_range() is defined.
> +{
> + uintptr_t start_addr = (uintptr_t)addr;
> + uintptr_t head_granule_start;
> + uintptr_t poison_body_start;
> + uintptr_t poison_body_end;
> + size_t head_prefix_size;
> + uintptr_t end_addr;
> +
> + if ((start_addr + size) % KASAN_GRANULE_SIZE)
> + return -EINVAL;
Other similar KASAN functions do a WARN_ON(bad alignment). I think
printing a warning is fair for this to force the caller to enforce
proper alignment.
> +
> + end_addr = ALIGN_DOWN(start_addr + size, KASAN_GRANULE_SIZE);
I don't think we need to ALIGN_DOWN(): we already checked that
(start_addr + size) % KASAN_GRANULE_SIZE == 0.
> + if (start_addr >= end_addr)
> + return -EINVAL;
Can also do a WARN_ON().
> +
> + head_granule_start = ALIGN_DOWN(start_addr, KASAN_GRANULE_SIZE);
> + head_prefix_size = start_addr - head_granule_start;
> +
> + if (IS_ENABLED(CONFIG_KASAN_GENERIC) && head_prefix_size > 0)
> + kasan_poison_last_granule((void *)head_granule_start,
> + head_prefix_size);
Let's rename kasan_poison_last_granule() to kasan_poison_granule()
then. Here the granule being poisoned is not the last one.
> +
> + poison_body_start = ALIGN(start_addr, KASAN_GRANULE_SIZE);
> + poison_body_end = ALIGN_DOWN(end_addr, KASAN_GRANULE_SIZE);
end_addr is already aligned.
> +
> + if (poison_body_start < poison_body_end)
> + kasan_poison((void *)poison_body_start,
> + poison_body_end - poison_body_start,
> + KASAN_SLAB_REDZONE, false);
> + return 0;
> +}
> +EXPORT_SYMBOL(kasan_poison_range);
> +
> #ifdef CONFIG_KASAN_GENERIC
> void kasan_poison_last_granule(const void *addr, size_t size)
> {
> --
> 2.51.0.470.ga7dc726c21-goog
>
^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH v2 02/10] kfuzztest: add user-facing API and data structures
2025-09-19 14:57 [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework Ethan Graham
2025-09-19 14:57 ` [PATCH v2 01/10] mm/kasan: implement kasan_poison_range Ethan Graham
@ 2025-09-19 14:57 ` Ethan Graham
2025-09-19 15:05 ` Alexander Potapenko
2025-09-24 8:44 ` Johannes Berg
2025-09-19 14:57 ` [PATCH v2 03/10] kfuzztest: implement core module and input processing Ethan Graham
` (9 subsequent siblings)
11 siblings, 2 replies; 30+ messages in thread
From: Ethan Graham @ 2025-09-19 14:57 UTC (permalink / raw)
To: ethangraham, glider
Cc: andreyknvl, andy, brauner, brendan.higgins, davem, davidgow,
dhowells, dvyukov, elver, herbert, ignat, jack, jannh, johannes,
kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel, linux-mm,
lukas, rmoar, shuah, sj, tarasmadan
From: Ethan Graham <ethangraham@google.com>
Add the foundational user-facing components for the KFuzzTest framework.
This includes the main API header <linux/kfuzztest.h>, the Kconfig
option to enable the feature, and the required linker script changes
which introduce three new ELF sections in vmlinux.
Note that KFuzzTest is intended strictly for debug builds only, and
should never be enabled in a production build. The fact that it exposes
internal kernel functions and state directly to userspace may constitute
a serious security vulnerability if used for any reason other than
testing.
The header defines:
- The FUZZ_TEST() macro for creating test targets.
- The data structures required for the binary serialization format,
which allows passing complex inputs from userspace.
- The metadata structures for test targets, constraints and annotations,
which are placed in dedicated ELF sections (.kfuzztest_*) for
discovery.
This patch only adds the public interface and build integration; no
runtime logic is included.
Signed-off-by: Ethan Graham <ethangraham@google.com>
---
PR v1:
- Move KFuzzTest metadata definitions to generic vmlinux linkage so that
the framework isn't bound to x86_64.
- Return -EFAULT when simple_write_to_buffer returns a value not equal
to the input length in the main FUZZ_TEST macro.
- Enforce a maximum input size of 64KiB in the main FUZZ_TEST macro,
returning -EINVAL when it isn't respected.
- Refactor KFUZZTEST_ANNOTATION_* macros.
- Taint the kernel with TAINT_TEST inside the FUZZ_TEST macro when a
fuzz target is invoked for the first time.
---
---
include/asm-generic/vmlinux.lds.h | 22 +-
include/linux/kfuzztest.h | 493 ++++++++++++++++++++++++++++++
lib/Kconfig.debug | 1 +
lib/kfuzztest/Kconfig | 20 ++
4 files changed, 535 insertions(+), 1 deletion(-)
create mode 100644 include/linux/kfuzztest.h
create mode 100644 lib/kfuzztest/Kconfig
diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h
index ae2d2359b79e..9afe569d013b 100644
--- a/include/asm-generic/vmlinux.lds.h
+++ b/include/asm-generic/vmlinux.lds.h
@@ -373,7 +373,8 @@ defined(CONFIG_AUTOFDO_CLANG) || defined(CONFIG_PROPELLER_CLANG)
TRACE_PRINTKS() \
BPF_RAW_TP() \
TRACEPOINT_STR() \
- KUNIT_TABLE()
+ KUNIT_TABLE() \
+ KFUZZTEST_TABLE()
/*
* Data section helpers
@@ -966,6 +967,25 @@ defined(CONFIG_AUTOFDO_CLANG) || defined(CONFIG_PROPELLER_CLANG)
BOUNDED_SECTION_POST_LABEL(.kunit_init_test_suites, \
__kunit_init_suites, _start, _end)
+#ifdef CONFIG_KFUZZTEST
+#define KFUZZTEST_TABLE() \
+ . = ALIGN(PAGE_SIZE); \
+ __kfuzztest_targets_start = .; \
+ KEEP(*(.kfuzztest_target)); \
+ __kfuzztest_targets_end = .; \
+ . = ALIGN(PAGE_SIZE); \
+ __kfuzztest_constraints_start = .; \
+ KEEP(*(.kfuzztest_constraint)); \
+ __kfuzztest_constraints_end = .; \
+ . = ALIGN(PAGE_SIZE); \
+ __kfuzztest_annotations_start = .; \
+ KEEP(*(.kfuzztest_annotation)); \
+ __kfuzztest_annotations_end = .;
+
+#else /* CONFIG_KFUZZTEST */
+#define KFUZZTEST_TABLE()
+#endif /* CONFIG_KFUZZTEST */
+
#ifdef CONFIG_BLK_DEV_INITRD
#define INIT_RAM_FS \
. = ALIGN(4); \
diff --git a/include/linux/kfuzztest.h b/include/linux/kfuzztest.h
new file mode 100644
index 000000000000..38970dea8fa5
--- /dev/null
+++ b/include/linux/kfuzztest.h
@@ -0,0 +1,493 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * The Kernel Fuzz Testing Framework (KFuzzTest) API for defining fuzz targets
+ * for internal kernel functions.
+ *
+ * For more information please see Documentation/dev-tools/kfuzztest.rst.
+ *
+ * Copyright 2025 Google LLC
+ */
+#ifndef KFUZZTEST_H
+#define KFUZZTEST_H
+
+#include <linux/fs.h>
+#include <linux/printk.h>
+#include <linux/types.h>
+
+#define KFUZZTEST_HEADER_MAGIC (0xBFACE)
+#define KFUZZTEST_V0 (0)
+
+/**
+ * @brief The KFuzzTest Input Serialization Format
+ *
+ * KFuzzTest receives its input from userspace as a single binary blob. This
+ * format allows for the serialization of complex, pointer-rich C structures
+ * into a flat buffer that can be safely passed into the kernel. This format
+ * requires only a single copy from userspace into a kernel buffer, and no
+ * further kernel allocations. Pointers are patched internally using a "region"
+ * system where each region corresponds to some pointed-to data.
+ *
+ * Regions should be padded to respect alignment constraints of their underlying
+ * types, and should be followed by at least 8 bytes of padding. These padded
+ * regions are poisoned by KFuzzTest to ensure that KASAN catches OOB accesses.
+ *
+ * The format consists of a header and three main components:
+ * 1. An 8-byte header: Contains KFUZZTEST_MAGIC in the first 4 bytes, and the
+ * version number in the subsequent 4 bytes. This ensures backwards
+ * compatibility in the event of future format changes.
+ * 2. A reloc_region_array: Defines the memory layout of the target structure
+ * by partitioning the payload into logical regions. Each logical region
+ * should contain the byte representation of the type that it represents,
+ * including any necessary padding. The region descriptors should be
+ * ordered by offset ascending.
+ * 3. A reloc_table: Provides "linking" instructions that tell the kernel how
+ * to patch pointer fields to point to the correct regions. By design,
+ * the first region (index 0) is passed as input into a FUZZ_TEST.
+ * 4. A Payload: The raw binary data for the target structure and its associated
+ * buffers. This should be aligned to the maximum alignment of all
+ * regions to satisfy alignment requirements of the input types, but this
+ * isn't checked by the parser.
+ *
+ * For a detailed specification of the binary layout see the full documentation
+ * at: Documentation/dev-tools/kfuzztest.rst
+ */
+
+/**
+ * struct reloc_region - single contiguous memory region in the payload
+ *
+ * @offset: The byte offset of this region from the start of the payload, which
+ * should be aligned to the alignment requirements of the region's
+ * underlying type.
+ * @size: The size of this region in bytes.
+ */
+struct reloc_region {
+ uint32_t offset;
+ uint32_t size;
+};
+
+/**
+ * struct reloc_region_array - array of regions in an input
+ *
+ * @num_regions: The total number of regions defined.
+ * @regions: A flexible array of `num_regions` region descriptors.
+ */
+struct reloc_region_array {
+ uint32_t num_regions;
+ struct reloc_region regions[];
+};
+
+/**
+ * struct reloc_entry - a single pointer to be patched in an input
+ *
+ * @region_id: The index of the region in the `reloc_region_array` that
+ * contains the pointer.
+ * @region_offset: The start offset of the pointer inside of the region.
+ * @value: contains the index of the pointee region, or KFUZZTEST_REGIONID_NULL
+ * if the pointer is NULL.
+ */
+struct reloc_entry {
+ uint32_t region_id;
+ uint32_t region_offset;
+ uint32_t value;
+};
+
+/**
+ * struct reloc_table - array of relocations required by an input
+ *
+ * @num_entries: the number of pointer relocations.
+ * @padding_size: the number of padded bytes between the last relocation in
+ * entries, and the start of the payload data. This should be at least
+ * 8 bytes, as it is used for poisoning.
+ * @entries: array of relocations.
+ */
+struct reloc_table {
+ uint32_t num_entries;
+ uint32_t padding_size;
+ struct reloc_entry entries[];
+};
+
+/**
+ * kfuzztest_parse_and_relocate - validate and relocate a KFuzzTest input
+ *
+ * @input: A buffer containing the serialized input for a fuzz target.
+ * @input_size: the size in bytes of the @input buffer.
+ * @arg_ret: return pointer for the test case's input structure.
+ */
+int kfuzztest_parse_and_relocate(void *input, size_t input_size, void **arg_ret);
+
+/*
+ * Dump some information on the parsed headers and payload. Can be useful for
+ * debugging inputs when writing an encoder for the KFuzzTest input format.
+ */
+__attribute__((unused)) static inline void kfuzztest_debug_header(struct reloc_region_array *regions,
+ struct reloc_table *rt, void *payload_start,
+ void *payload_end)
+{
+ uint32_t i;
+
+ pr_info("regions: { num_regions = %u } @ %px", regions->num_regions, regions);
+ for (i = 0; i < regions->num_regions; i++) {
+ pr_info(" region_%u: { start: 0x%x, size: 0x%x }", i, regions->regions[i].offset,
+ regions->regions[i].size);
+ }
+
+ pr_info("reloc_table: { num_entries = %u, padding = %u } @ offset 0x%tx", rt->num_entries, rt->padding_size,
+ (char *)rt - (char *)regions);
+ for (i = 0; i < rt->num_entries; i++) {
+ pr_info(" reloc_%u: { src: %u, offset: 0x%x, dst: %u }", i, rt->entries[i].region_id,
+ rt->entries[i].region_offset, rt->entries[i].value);
+ }
+
+ pr_info("payload: [0x%lx, 0x%tx)", (char *)payload_start - (char *)regions,
+ (char *)payload_end - (char *)regions);
+}
+
+struct kfuzztest_target {
+ const char *name;
+ const char *arg_type_name;
+ ssize_t (*write_input_cb)(struct file *filp, const char __user *buf, size_t len, loff_t *off);
+} __aligned(32);
+
+#define KFUZZTEST_MAX_INPUT_SIZE (PAGE_SIZE * 16)
+
+/**
+ * FUZZ_TEST - defines a KFuzzTest target
+ *
+ * @test_name: The unique identifier for the fuzz test, which is used to name
+ * the debugfs entry, e.g., /sys/kernel/debug/kfuzztest/@test_name.
+ * @test_arg_type: The struct type that defines the inputs for the test. This
+ * must be the full struct type (e.g., "struct my_inputs"), not a typedef.
+ *
+ * Context:
+ * This macro is the primary entry point for the KFuzzTest framework. It
+ * generates all the necessary boilerplate for a fuzz test, including:
+ * - A static `struct kfuzztest_target` instance that is placed in a
+ * dedicated ELF section for discovery by userspace tools.
+ * - A `debugfs` write callback that handles receiving serialized data from
+ * a fuzzer, parsing it, and "hydrating" it into a valid C struct.
+ * - A function stub where the developer places the test logic.
+ *
+ * User-Provided Logic:
+ * The developer must provide the body of the fuzz test logic within the curly
+ * braces following the macro invocation. Within this scope, the framework
+ * provides the `arg` variable, which is a pointer of type `@test_arg_type *`
+ * to the fully hydrated input structure. All pointer fields within this struct
+ * have been relocated and are valid kernel pointers. This is the primary
+ * variable to use for accessing fuzzing inputs.
+ *
+ * Example Usage:
+ *
+ * // 1. The kernel function we want to fuzz.
+ * int process_data(const char *data, size_t len);
+ *
+ * // 2. Define a struct to hold all inputs for the function.
+ * struct process_data_inputs {
+ * const char *data;
+ * size_t len;
+ * };
+ *
+ * // 3. Define the fuzz test using the FUZZ_TEST macro.
+ * FUZZ_TEST(process_data_fuzzer, struct process_data_inputs)
+ * {
+ * int ret;
+ * // Use KFUZZTEST_EXPECT_* to enforce preconditions.
+ * // The test will exit early if data is NULL.
+ * KFUZZTEST_EXPECT_NOT_NULL(process_data_inputs, data);
+ *
+ * // Use KFUZZTEST_ANNOTATE_* to provide hints to the fuzzer.
+ * // This links the 'len' field to the 'data' buffer.
+ * KFUZZTEST_ANNOTATE_LEN(process_data_inputs, len, data);
+ *
+ * // Call the function under test using the 'arg' variable. OOB memory
+ * // accesses will be caught by KASAN, but the user can also choose to
+ * // validate the return value and log any failures.
+ * ret = process_data(arg->data, arg->len);
+ * }
+ */
+#define FUZZ_TEST(test_name, test_arg_type) \
+ static ssize_t kfuzztest_write_cb_##test_name(struct file *filp, const char __user *buf, size_t len, \
+ loff_t *off); \
+ static void kfuzztest_logic_##test_name(test_arg_type *arg); \
+ static const struct kfuzztest_target __fuzz_test__##test_name __section(".kfuzztest_target") __used = { \
+ .name = #test_name, \
+ .arg_type_name = #test_arg_type, \
+ .write_input_cb = kfuzztest_write_cb_##test_name, \
+ }; \
+ static ssize_t kfuzztest_write_cb_##test_name(struct file *filp, const char __user *buf, size_t len, \
+ loff_t *off) \
+ { \
+ test_arg_type *arg; \
+ void *buffer; \
+ int ret; \
+ \
+ /*
+ * Taint the kernel on the first fuzzing invocation. The debugfs
+ * interface provides a high-risk entry point for userspace to
+ * call kernel functions with untrusted input.
+ */ \
+ if (!test_taint(TAINT_TEST)) \
+ add_taint(TAINT_TEST, LOCKDEP_STILL_OK); \
+ if (len >= KFUZZTEST_MAX_INPUT_SIZE) { \
+ pr_warn(#test_name ": user input of size %zu is too large", len); \
+ return -EINVAL; \
+ } \
+ buffer = kmalloc(len, GFP_KERNEL); \
+ if (!buffer) \
+ return -ENOMEM; \
+ ret = simple_write_to_buffer(buffer, len, off, buf, len); \
+ if (ret != len){ \
+ ret = -EFAULT; \
+ goto out; \
+ }; \
+ ret = kfuzztest_parse_and_relocate(buffer, len, (void **)&arg); \
+ if (ret < 0) \
+ goto out; \
+ kfuzztest_logic_##test_name(arg); \
+ ret = len; \
+out: \
+ kfree(buffer); \
+ return ret; \
+ } \
+ static void kfuzztest_logic_##test_name(test_arg_type *arg)
+
+enum kfuzztest_constraint_type {
+ EXPECT_EQ,
+ EXPECT_NE,
+ EXPECT_LT,
+ EXPECT_LE,
+ EXPECT_GT,
+ EXPECT_GE,
+ EXPECT_IN_RANGE,
+};
+
+/**
+ * struct kfuzztest_constraint - a metadata record for a domain constraint
+ *
+ * Domain constraints are rules about the input data that must be satisfied for
+ * a fuzz test to proceed. While they are enforced in the kernel with a runtime
+ * check, they are primarily intended as a discoverable contract for userspace
+ * fuzzers.
+ *
+ * Instances of this struct are generated by the KFUZZTEST_EXPECT_* macros
+ * and placed into the read-only ".kfuzztest_constraint" ELF section of the
+ * vmlinux binary. A fuzzer can parse this section to learn about the
+ * constraints and generate valid inputs more intelligently.
+ *
+ * For an example of how these constraints are used within a fuzz test, see the
+ * documentation for the FUZZ_TEST() macro.
+ *
+ * @input_type: The name of the input struct type, without the leading
+ * "struct ".
+ * @field_name: The name of the field within the struct that this constraint
+ * applies to.
+ * @value1: The primary value used in the comparison (e.g., the upper
+ * bound for EXPECT_LE).
+ * @value2: The secondary value, used only for multi-value comparisons
+ * (e.g., the upper bound for EXPECT_IN_RANGE).
+ * @type: The type of the constraint.
+ */
+struct kfuzztest_constraint {
+ const char *input_type;
+ const char *field_name;
+ uintptr_t value1;
+ uintptr_t value2;
+ enum kfuzztest_constraint_type type;
+} __aligned(64);
+
+#define __KFUZZTEST_DEFINE_CONSTRAINT(arg_type, field, val1, val2, tpe, predicate) \
+ do { \
+ static struct kfuzztest_constraint __constraint_##arg_type##_##field \
+ __section(".kfuzztest_constraint") __used = { \
+ .input_type = "struct " #arg_type, \
+ .field_name = #field, \
+ .value1 = (uintptr_t)val1, \
+ .value2 = (uintptr_t)val2, \
+ .type = tpe, \
+ }; \
+ if (!(predicate)) \
+ return; \
+ } while (0)
+
+/**
+ * KFUZZTEST_EXPECT_EQ - constrain a field to be equal to a value
+ *
+ * @arg_type: name of the input structure, without the leading "struct ".
+ * @field: some field that is comparable
+ * @val: a value of the same type as @arg_type.@field
+ */
+#define KFUZZTEST_EXPECT_EQ(arg_type, field, val) \
+ __KFUZZTEST_DEFINE_CONSTRAINT(arg_type, field, val, 0x0, EXPECT_EQ, arg->field == val)
+
+/**
+ * KFUZZTEST_EXPECT_NE - constrain a field to be not equal to a value
+ *
+ * @arg_type: name of the input structure, without the leading "struct ".
+ * @field: some field that is comparable.
+ * @val: a value of the same type as @arg_type.@field.
+ */
+#define KFUZZTEST_EXPECT_NE(arg_type, field, val) \
+ __KFUZZTEST_DEFINE_CONSTRAINT(arg_type, field, val, 0x0, EXPECT_NE, arg->field != val)
+
+/**
+ * KFUZZTEST_EXPECT_LT - constrain a field to be less than a value
+ *
+ * @arg_type: name of the input structure, without the leading "struct ".
+ * @field: some field that is comparable.
+ * @val: a value of the same type as @arg_type.@field.
+ */
+#define KFUZZTEST_EXPECT_LT(arg_type, field, val) \
+ __KFUZZTEST_DEFINE_CONSTRAINT(arg_type, field, val, 0x0, EXPECT_LT, arg->field < val)
+
+/**
+ * KFUZZTEST_EXPECT_LE - constrain a field to be less than or equal to a value
+ *
+ * @arg_type: name of the input structure, without the leading "struct ".
+ * @field: some field that is comparable.
+ * @val: a value of the same type as @arg_type.@field.
+ */
+#define KFUZZTEST_EXPECT_LE(arg_type, field, val) \
+ __KFUZZTEST_DEFINE_CONSTRAINT(arg_type, field, val, 0x0, EXPECT_LE, arg->field <= val)
+
+/**
+ * KFUZZTEST_EXPECT_GT - constrain a field to be greater than a value
+ *
+ * @arg_type: name of the input structure, without the leading "struct ".
+ * @field: some field that is comparable.
+ * @val: a value of the same type as @arg_type.@field.
+ */
+#define KFUZZTEST_EXPECT_GT(arg_type, field, val) \
+ __KFUZZTEST_DEFINE_CONSTRAINT(arg_type, field, val, 0x0, EXPECT_GT, arg->field > val)
+
+/**
+ * KFUZZTEST_EXPECT_GE - constrain a field to be greater than or equal to a value
+ *
+ * @arg_type: name of the input structure, without the leading "struct ".
+ * @field: some field that is comparable.
+ * @val: a value of the same type as @arg_type.@field.
+ */
+#define KFUZZTEST_EXPECT_GE(arg_type, field, val) \
+ __KFUZZTEST_DEFINE_CONSTRAINT(arg_type, field, val, 0x0, EXPECT_GE, arg->field >= val)
+
+/**
+ * KFUZZTEST_EXPECT_NOT_NULL - constrain a pointer field to be non-NULL
+ *
+ * @arg_type: name of the input structure, without the leading "struct ".
+ * @field: a pointer field.
+ */
+#define KFUZZTEST_EXPECT_NOT_NULL(arg_type, field) KFUZZTEST_EXPECT_NE(arg_type, field, NULL)
+
+/**
+ * KFUZZTEST_EXPECT_IN_RANGE - constrain a field to be within a range
+ *
+ * @arg_type: name of the input structure, without the leading "struct ".
+ * @field: some field that is comparable.
+ * @lower_bound: a lower bound of the same type as @arg_type.@field.
+ * @upper_bound: an upper bound of the same type as @arg_type.@field.
+ */
+#define KFUZZTEST_EXPECT_IN_RANGE(arg_type, field, lower_bound, upper_bound) \
+ __KFUZZTEST_DEFINE_CONSTRAINT(arg_type, field, lower_bound, upper_bound, \
+ EXPECT_IN_RANGE, arg->field >= lower_bound && arg->field <= upper_bound)
+
+/**
+ * Annotations express attributes about structure fields that can't be easily
+ * or safely verified at runtime. They are intended as hints to the fuzzing
+ * engine to help it generate more semantically correct and effective inputs.
+ * Unlike constraints, annotations do not add any runtime checks and do not
+ * cause a test to exit early.
+ *
+ * For example, a `char *` field could be a raw byte buffer or a C-style
+ * null-terminated string. A fuzzer that is aware of this distinction can avoid
+ * creating inputs that would cause trivial, uninteresting crashes from reading
+ * past the end of a non-null-terminated buffer.
+ */
+enum kfuzztest_annotation_attribute {
+ ATTRIBUTE_LEN,
+ ATTRIBUTE_STRING,
+ ATTRIBUTE_ARRAY,
+};
+
+/**
+ * struct kfuzztest_annotation - a metadata record for a fuzzer hint
+ *
+ * This struct captures a single hint about a field in the input structure.
+ * Instances are generated by the KFUZZTEST_ANNOTATE_* macros and are placed
+ * into the read-only ".kfuzztest_annotation" ELF section of the vmlinux binary.
+ *
+ * A userspace fuzzer can parse this section to understand the semantic
+ * relationships between fields (e.g., which field is a length for which
+ * buffer) and the expected format of the data (e.g., a null-terminated
+ * string). This allows the fuzzer to be much more intelligent during input
+ * generation and mutation.
+ *
+ * For an example of how annotations are used within a fuzz test, see the
+ * documentation for the FUZZ_TEST() macro.
+ *
+ * @input_type: The name of the input struct type.
+ * @field_name: The name of the field being annotated (e.g., the data
+ * buffer field).
+ * @linked_field_name: For annotations that link two fields (like
+ * ATTRIBUTE_LEN), this is the name of the related field (e.g., the
+ * length field). For others, this may be unused.
+ * @attrib: The type of the annotation hint.
+ */
+struct kfuzztest_annotation {
+ const char *input_type;
+ const char *field_name;
+ const char *linked_field_name;
+ enum kfuzztest_annotation_attribute attrib;
+} __aligned(32);
+
+#define __KFUZZTEST_ANNOTATE(arg_type, field, linked_field, attribute) \
+ static struct kfuzztest_annotation __annotation_##arg_type##_##field __section(".kfuzztest_annotation") \
+ __used = { \
+ .input_type = "struct " #arg_type, \
+ .field_name = #field, \
+ .linked_field_name = #linked_field, \
+ .attrib = attribute, \
+ }
+
+/**
+ * KFUZZTEST_ANNOTATE_STRING - annotate a char* field as a C string
+ *
+ * We define a C string as a sequence of non-zero characters followed by exactly
+ * one null terminator.
+ *
+ * @arg_type: name of the input structure, without the leading "struct ".
+ * @field: the name of the field to annotate.
+ */
+#define KFUZZTEST_ANNOTATE_STRING(arg_type, field) __KFUZZTEST_ANNOTATE(arg_type, field, NULL, ATTRIBUTE_STRING)
+
+/**
+ * KFUZZTEST_ANNOTATE_ARRAY - annotate a pointer as an array
+ *
+ * We define an array as a contiguous memory region containing zero or more
+ * elements of the same type.
+ *
+ * @arg_type: name of the input structure, without the leading "struct ".
+ * @field: the name of the field to annotate.
+ */
+#define KFUZZTEST_ANNOTATE_ARRAY(arg_type, field) __KFUZZTEST_ANNOTATE(arg_type, field, NULL, ATTRIBUTE_ARRAY)
+
+/**
+ * KFUZZTEST_ANNOTATE_LEN - annotate a field as the length of another
+ *
+ * This expresses the relationship `arg_type.field == len(linked_field)`, where
+ * `linked_field` is an array.
+ *
+ * @arg_type: name of the input structure, without the leading "struct ".
+ * @field: the name of the field to annotate.
+ * @linked_field: the name of an array field with length @field.
+ */
+#define KFUZZTEST_ANNOTATE_LEN(arg_type, field, linked_field) \
+ __KFUZZTEST_ANNOTATE(arg_type, field, linked_field, ATTRIBUTE_LEN)
+
+#define KFUZZTEST_REGIONID_NULL U32_MAX
+
+/**
+ * The end of the input should be padded by at least this number of bytes as
+ * it is poisoned to detect out of bounds accesses at the end of the last
+ * region.
+ */
+#define KFUZZTEST_POISON_SIZE 0x8
+
+#endif /* KFUZZTEST_H */
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index dc0e0c6ed075..49a1748b9f24 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1947,6 +1947,7 @@ endmenu
menu "Kernel Testing and Coverage"
source "lib/kunit/Kconfig"
+source "lib/kfuzztest/Kconfig"
config NOTIFIER_ERROR_INJECTION
tristate "Notifier error injection"
diff --git a/lib/kfuzztest/Kconfig b/lib/kfuzztest/Kconfig
new file mode 100644
index 000000000000..f9fb5abf8d27
--- /dev/null
+++ b/lib/kfuzztest/Kconfig
@@ -0,0 +1,20 @@
+# SPDX-License-Identifier: GPL-2.0-only
+
+config KFUZZTEST
+ bool "KFuzzTest - enable support for internal fuzz targets"
+ depends on DEBUG_FS && DEBUG_KERNEL
+ help
+ Enables support for the kernel fuzz testing framework (KFuzzTest), an
+ interface for exposing internal kernel functions to a userspace fuzzing
+ engine. KFuzzTest targets are exposed via a debugfs interface that
+ accepts serialized userspace inputs, and is designed to make it easier
+ to fuzz deeply nested kernel code that is hard to reach from the system
+ call boundary. Using a simple macro-based API, developers can add a new
+ fuzz target with minimal boilerplate code.
+
+ It is strongly recommended to also enable CONFIG_KASAN for byte-accurate
+ out-of-bounds detection, as KFuzzTest was designed with this in mind. It
+ is also recommended to enable CONFIG_KCOV for coverage guided fuzzing.
+
+ WARNING: This exposes internal kernel functions directly to userspace
+ and must NEVER be enabled in production builds.
--
2.51.0.470.ga7dc726c21-goog
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH v2 02/10] kfuzztest: add user-facing API and data structures
2025-09-19 14:57 ` [PATCH v2 02/10] kfuzztest: add user-facing API and data structures Ethan Graham
@ 2025-09-19 15:05 ` Alexander Potapenko
2025-09-24 8:44 ` Johannes Berg
1 sibling, 0 replies; 30+ messages in thread
From: Alexander Potapenko @ 2025-09-19 15:05 UTC (permalink / raw)
To: Ethan Graham
Cc: ethangraham, andreyknvl, andy, brauner, brendan.higgins, davem,
davidgow, dhowells, dvyukov, elver, herbert, ignat, jack, jannh,
johannes, kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel,
linux-mm, lukas, rmoar, shuah, sj, tarasmadan
On Fri, Sep 19, 2025 at 4:58 PM Ethan Graham <ethan.w.s.graham@gmail.com> wrote:
>
> From: Ethan Graham <ethangraham@google.com>
>
> Add the foundational user-facing components for the KFuzzTest framework.
> This includes the main API header <linux/kfuzztest.h>, the Kconfig
> option to enable the feature, and the required linker script changes
> which introduce three new ELF sections in vmlinux.
>
> Note that KFuzzTest is intended strictly for debug builds only, and
> should never be enabled in a production build. The fact that it exposes
> internal kernel functions and state directly to userspace may constitute
> a serious security vulnerability if used for any reason other than
> testing.
>
> The header defines:
> - The FUZZ_TEST() macro for creating test targets.
> - The data structures required for the binary serialization format,
> which allows passing complex inputs from userspace.
> - The metadata structures for test targets, constraints and annotations,
> which are placed in dedicated ELF sections (.kfuzztest_*) for
> discovery.
>
> This patch only adds the public interface and build integration; no
> runtime logic is included.
>
> Signed-off-by: Ethan Graham <ethangraham@google.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH v2 02/10] kfuzztest: add user-facing API and data structures
2025-09-19 14:57 ` [PATCH v2 02/10] kfuzztest: add user-facing API and data structures Ethan Graham
2025-09-19 15:05 ` Alexander Potapenko
@ 2025-09-24 8:44 ` Johannes Berg
1 sibling, 0 replies; 30+ messages in thread
From: Johannes Berg @ 2025-09-24 8:44 UTC (permalink / raw)
To: Ethan Graham, ethangraham, glider
Cc: andreyknvl, andy, brauner, brendan.higgins, davem, davidgow,
dhowells, dvyukov, elver, herbert, ignat, jack, jannh, kasan-dev,
kees, kunit-dev, linux-crypto, linux-kernel, linux-mm, lukas,
rmoar, shuah, sj, tarasmadan
On Fri, 2025-09-19 at 14:57 +0000, Ethan Graham wrote:
>
> + * User-Provided Logic:
> + * The developer must provide the body of the fuzz test logic within the curly
> + * braces following the macro invocation. Within this scope, the framework
> + * provides the `arg` variable, which is a pointer of type `@test_arg_type *`
FWIW, git complained about trailing whitespace on this line.
I'm trying to apply this and integrate it with ARCH=um and honggfuzz
(because afl++ doesn't work with -fsanitize-coverage=trace-pc and I have
issues with clang right now ...). Fingers crossed :)
johannes
^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH v2 03/10] kfuzztest: implement core module and input processing
2025-09-19 14:57 [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework Ethan Graham
2025-09-19 14:57 ` [PATCH v2 01/10] mm/kasan: implement kasan_poison_range Ethan Graham
2025-09-19 14:57 ` [PATCH v2 02/10] kfuzztest: add user-facing API and data structures Ethan Graham
@ 2025-09-19 14:57 ` Ethan Graham
2025-09-19 15:05 ` Alexander Potapenko
2025-09-19 14:57 ` [PATCH v2 04/10] tools: add kfuzztest-bridge utility Ethan Graham
` (8 subsequent siblings)
11 siblings, 1 reply; 30+ messages in thread
From: Ethan Graham @ 2025-09-19 14:57 UTC (permalink / raw)
To: ethangraham, glider
Cc: andreyknvl, andy, brauner, brendan.higgins, davem, davidgow,
dhowells, dvyukov, elver, herbert, ignat, jack, jannh, johannes,
kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel, linux-mm,
lukas, rmoar, shuah, sj, tarasmadan
From: Ethan Graham <ethangraham@google.com>
Add the core runtime implementation for KFuzzTest. This includes the
module initialization, and the logic for receiving and processing
user-provided inputs through debugfs.
On module load, the framework discovers all test targets by iterating
over the .kfuzztest_target section, creating a corresponding debugfs
directory with a write-only 'input' file for each of them.
Writing to an 'input' file triggers the main fuzzing sequence:
1. The serialized input is copied from userspace into a kernel buffer.
2. The buffer is parsed to validate the region array and relocation
table.
3. Pointers are patched based on the relocation entries, and in KASAN
builds the inter-region padding is poisoned.
4. The resulting struct is passed to the user-defined test logic.
Signed-off-by: Ethan Graham <ethangraham@google.com>
---
PR v2:
- Fix build issues identified by the kernel test robot <lkp@intel.com>.
- Address some nits pointed out by Alexander Potapenko.
PR v1:
- Update kfuzztest/parse.c interfaces to take `unsigned char *` instead
of `void *`, reducing the number of pointer casts.
- Expose minimum region alignment via a new debugfs file.
- Expose number of successful invocations via a new debugfs file.
- Refactor module init function, add _config directory with entries
containing KFuzzTest state information.
- Account for kasan_poison_range() return value in input parsing logic.
- Validate alignment of payload end.
- Move static sizeof assertions into /lib/kfuzztest/main.c.
- Remove the taint in kfuzztest/main.c. We instead taint the kernel as
soon as a fuzz test is invoked for the first time, which is done in
the primary FUZZ_TEST macro.
RFC v2:
- The module's init function now taints the kernel with TAINT_TEST.
---
---
include/linux/kfuzztest.h | 4 +
lib/Makefile | 2 +
lib/kfuzztest/Makefile | 4 +
lib/kfuzztest/main.c | 242 ++++++++++++++++++++++++++++++++++++++
lib/kfuzztest/parse.c | 204 ++++++++++++++++++++++++++++++++
5 files changed, 456 insertions(+)
create mode 100644 lib/kfuzztest/Makefile
create mode 100644 lib/kfuzztest/main.c
create mode 100644 lib/kfuzztest/parse.c
diff --git a/include/linux/kfuzztest.h b/include/linux/kfuzztest.h
index 38970dea8fa5..2620e48bb620 100644
--- a/include/linux/kfuzztest.h
+++ b/include/linux/kfuzztest.h
@@ -150,6 +150,9 @@ struct kfuzztest_target {
#define KFUZZTEST_MAX_INPUT_SIZE (PAGE_SIZE * 16)
+/* Increments a global counter after a successful invocation. */
+void record_invocation(void);
+
/**
* FUZZ_TEST - defines a KFuzzTest target
*
@@ -243,6 +246,7 @@ struct kfuzztest_target {
if (ret < 0) \
goto out; \
kfuzztest_logic_##test_name(arg); \
+ record_invocation(); \
ret = len; \
out: \
kfree(buffer); \
diff --git a/lib/Makefile b/lib/Makefile
index 392ff808c9b9..02789bf88499 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -325,6 +325,8 @@ obj-$(CONFIG_GENERIC_LIB_CMPDI2) += cmpdi2.o
obj-$(CONFIG_GENERIC_LIB_UCMPDI2) += ucmpdi2.o
obj-$(CONFIG_OBJAGG) += objagg.o
+obj-$(CONFIG_KFUZZTEST) += kfuzztest/
+
# pldmfw library
obj-$(CONFIG_PLDMFW) += pldmfw/
diff --git a/lib/kfuzztest/Makefile b/lib/kfuzztest/Makefile
new file mode 100644
index 000000000000..142d16007eea
--- /dev/null
+++ b/lib/kfuzztest/Makefile
@@ -0,0 +1,4 @@
+# SPDX-License-Identifier: GPL-2.0
+
+obj-$(CONFIG_KFUZZTEST) += kfuzztest.o
+kfuzztest-objs := main.o parse.o
diff --git a/lib/kfuzztest/main.c b/lib/kfuzztest/main.c
new file mode 100644
index 000000000000..c36a7a0b7602
--- /dev/null
+++ b/lib/kfuzztest/main.c
@@ -0,0 +1,242 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * KFuzzTest core module initialization and debugfs interface.
+ *
+ * Copyright 2025 Google LLC
+ */
+#include <linux/atomic.h>
+#include <linux/debugfs.h>
+#include <linux/fs.h>
+#include <linux/kfuzztest.h>
+#include <linux/module.h>
+#include <linux/printk.h>
+#include <linux/kasan.h>
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Ethan Graham <ethangraham@google.com>");
+MODULE_DESCRIPTION("Kernel Fuzz Testing Framework (KFuzzTest)");
+
+/*
+ * Enforce a fixed struct size to ensure a consistent stride when iterating over
+ * the array of these structs in the dedicated ELF section.
+ */
+static_assert(sizeof(struct kfuzztest_target) == 32, "struct kfuzztest_target should have size 32");
+static_assert(sizeof(struct kfuzztest_constraint) == 64, "struct kfuzztest_constraint should have size 64");
+static_assert(sizeof(struct kfuzztest_annotation) == 32, "struct kfuzztest_annotation should have size 32");
+
+extern const struct kfuzztest_target __kfuzztest_targets_start[];
+extern const struct kfuzztest_target __kfuzztest_targets_end[];
+
+/**
+ * struct kfuzztest_state - global state for the KFuzzTest module
+ *
+ * @kfuzztest_dir: The root debugfs directory, /sys/kernel/debug/kfuzztest/.
+ * @num_invocations: total number of target invocations.
+ * @num_targets: number of registered targets.
+ * @target_fops: array of file operations for each registered target.
+ * @minalign_fops: file operations for the /_config/minalign file.
+ * @num_invocations_fops: file operations for the /_config/num_invocations file.
+ */
+struct kfuzztest_state {
+ struct dentry *kfuzztest_dir;
+ atomic_t num_invocations;
+ size_t num_targets;
+
+ struct file_operations *target_fops;
+ struct file_operations minalign_fops;
+ struct file_operations num_invocations_fops;
+};
+
+static struct kfuzztest_state state;
+
+void record_invocation(void)
+{
+ atomic_inc(&state.num_invocations);
+}
+
+static void cleanup_kfuzztest_state(struct kfuzztest_state *st)
+{
+ debugfs_remove_recursive(st->kfuzztest_dir);
+ st->num_targets = 0;
+ st->num_invocations = (atomic_t)ATOMIC_INIT(0);
+ kfree(st->target_fops);
+ st->target_fops = NULL;
+}
+
+static const umode_t KFUZZTEST_INPUT_PERMS = 0222;
+static const umode_t KFUZZTEST_MINALIGN_PERMS = 0444;
+
+static ssize_t read_cb_integer(struct file *filp, char __user *buf, size_t count, loff_t *f_pos, size_t value)
+{
+ char buffer[64];
+ int len;
+
+ len = scnprintf(buffer, sizeof(buffer), "%zu\n", value);
+ return simple_read_from_buffer(buf, count, f_pos, buffer, len);
+}
+
+/*
+ * Callback for /sys/kernel/debug/kfuzztest/_config/minalign. Minalign
+ * corresponds to the minimum alignment that regions in a KFuzzTest input must
+ * satisfy. This callback returns that value in string format.
+ */
+static ssize_t minalign_read_cb(struct file *filp, char __user *buf, size_t count, loff_t *f_pos)
+{
+ int minalign = MAX(KFUZZTEST_POISON_SIZE, ARCH_KMALLOC_MINALIGN);
+ return read_cb_integer(filp, buf, count, f_pos, minalign);
+}
+
+/*
+ * Callback for /sys/kernel/debug/kfuzztest/_config/num_invocations, which
+ * returns the value in string format.
+ */
+static ssize_t num_invocations_read_cb(struct file *filp, char __user *buf, size_t count, loff_t *f_pos)
+{
+ return read_cb_integer(filp, buf, count, f_pos, atomic_read(&state.num_invocations));
+}
+
+static int create_read_only_file(struct dentry *parent, const char *name, struct file_operations *fops)
+{
+ struct dentry *file;
+ int err = 0;
+
+ file = debugfs_create_file(name, KFUZZTEST_MINALIGN_PERMS, parent, NULL, fops);
+ if (!file)
+ err = -ENOMEM;
+ else if (IS_ERR(file))
+ err = PTR_ERR(file);
+ return err;
+}
+
+static int initialize_config_dir(struct kfuzztest_state *st)
+{
+ struct dentry *dir;
+ int err = 0;
+
+ dir = debugfs_create_dir("_config", st->kfuzztest_dir);
+ if (!dir)
+ err = -ENOMEM;
+ else if (IS_ERR(dir))
+ err = PTR_ERR(dir);
+ if (err) {
+ pr_info("kfuzztest: failed to create /_config dir");
+ goto out;
+ }
+
+ st->minalign_fops = (struct file_operations){
+ .owner = THIS_MODULE,
+ .read = minalign_read_cb,
+ };
+ err = create_read_only_file(dir, "minalign", &st->minalign_fops);
+ if (err) {
+ pr_info("kfuzztest: failed to create /_config/minalign");
+ goto out;
+ }
+
+ st->num_invocations_fops = (struct file_operations){
+ .owner = THIS_MODULE,
+ .read = num_invocations_read_cb,
+ };
+ err = create_read_only_file(dir, "num_invocations", &st->num_invocations_fops);
+ if (err)
+ pr_info("kfuzztest: failed to create /_config/num_invocations");
+out:
+ return err;
+}
+
+static int initialize_target_dir(struct kfuzztest_state *st, const struct kfuzztest_target *targ,
+ struct file_operations *fops)
+{
+ struct dentry *dir, *input;
+ int err = 0;
+
+ dir = debugfs_create_dir(targ->name, st->kfuzztest_dir);
+ if (!dir)
+ err = -ENOMEM;
+ else if (IS_ERR(dir))
+ err = PTR_ERR(dir);
+ if (err) {
+ pr_info("kfuzztest: failed to create /kfuzztest/%s dir", targ->name);
+ goto out;
+ }
+
+ input = debugfs_create_file("input", KFUZZTEST_INPUT_PERMS, dir, NULL, fops);
+ if (!input)
+ err = -ENOMEM;
+ else if (IS_ERR(input))
+ err = PTR_ERR(input);
+ if (err)
+ pr_info("kfuzztest: failed to create /kfuzztest/%s/input", targ->name);
+out:
+ return err;
+}
+
+/**
+ * kfuzztest_init - initializes the debug filesystem for KFuzzTest
+ *
+ * Each registered target in the ".kfuzztest_targets" section gets its own
+ * subdirectory under "/sys/kernel/debug/kfuzztest/<test-name>" containing one
+ * write-only "input" file used for receiving inputs from userspace.
+ * Furthermore, a directory "/sys/kernel/debug/kfuzztest/_config" is created,
+ * containing two read-only files "minalign" and "num_invocations", that return
+ * the minimum required region alignment and number of target invocations
+ * respectively.
+ *
+ * @return 0 on success or an error
+ */
+static int __init kfuzztest_init(void)
+{
+ const struct kfuzztest_target *targ;
+ int err = 0;
+ int i = 0;
+
+ state.num_targets = __kfuzztest_targets_end - __kfuzztest_targets_start;
+ state.target_fops = kzalloc(sizeof(struct file_operations) * state.num_targets, GFP_KERNEL);
+ if (!state.target_fops)
+ return -ENOMEM;
+
+ /* Create the main "kfuzztest" directory in /sys/kernel/debug. */
+ state.kfuzztest_dir = debugfs_create_dir("kfuzztest", NULL);
+ if (!state.kfuzztest_dir) {
+ pr_warn("kfuzztest: could not create 'kfuzztest' debugfs directory");
+ return -ENOMEM;
+ }
+ if (IS_ERR(state.kfuzztest_dir)) {
+ pr_warn("kfuzztest: could not create 'kfuzztest' debugfs directory");
+ err = PTR_ERR(state.kfuzztest_dir);
+ state.kfuzztest_dir = NULL;
+ return err;
+ }
+
+ err = initialize_config_dir(&state);
+ if (err)
+ goto cleanup_failure;
+
+ for (targ = __kfuzztest_targets_start; targ < __kfuzztest_targets_end; targ++, i++) {
+ state.target_fops[i] = (struct file_operations){
+ .owner = THIS_MODULE,
+ .write = targ->write_input_cb,
+ };
+ err = initialize_target_dir(&state, targ, &state.target_fops[i]);
+ /* Bail out if a single target fails to initialize. This avoids
+ * partial setup, and a failure here likely indicates an issue
+ * with debugfs. */
+ if (err)
+ goto cleanup_failure;
+ pr_info("kfuzztest: registered target %s", targ->name);
+ }
+ return 0;
+
+cleanup_failure:
+ cleanup_kfuzztest_state(&state);
+ return err;
+}
+
+static void __exit kfuzztest_exit(void)
+{
+ pr_info("kfuzztest: exiting");
+ cleanup_kfuzztest_state(&state);
+}
+
+module_init(kfuzztest_init);
+module_exit(kfuzztest_exit);
diff --git a/lib/kfuzztest/parse.c b/lib/kfuzztest/parse.c
new file mode 100644
index 000000000000..5aaeca6a7fde
--- /dev/null
+++ b/lib/kfuzztest/parse.c
@@ -0,0 +1,204 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * KFuzzTest input parsing and validation.
+ *
+ * Copyright 2025 Google LLC
+ */
+#include <linux/kfuzztest.h>
+#include <linux/kasan.h>
+
+static int kfuzztest_relocate_v0(struct reloc_region_array *regions, struct reloc_table *rt,
+ unsigned char *payload_start, unsigned char *payload_end)
+{
+ unsigned char *poison_start, *poison_end;
+ struct reloc_region reg, src, dst;
+ uintptr_t *ptr_location;
+ struct reloc_entry re;
+ size_t i;
+ int ret;
+
+ /* Patch pointers. */
+ for (i = 0; i < rt->num_entries; i++) {
+ re = rt->entries[i];
+ src = regions->regions[re.region_id];
+ ptr_location = (uintptr_t *)(payload_start + src.offset + re.region_offset);
+ if (re.value == KFUZZTEST_REGIONID_NULL)
+ *ptr_location = (uintptr_t)NULL;
+ else if (re.value < regions->num_regions) {
+ dst = regions->regions[re.value];
+ *ptr_location = (uintptr_t)(payload_start + dst.offset);
+ } else {
+ return -EINVAL;
+ }
+ }
+
+ /* Poison the padding between regions. */
+ for (i = 0; i < regions->num_regions; i++) {
+ reg = regions->regions[i];
+
+ /* Points to the beginning of the inter-region padding */
+ poison_start = payload_start + reg.offset + reg.size;
+ if (i < regions->num_regions - 1)
+ poison_end = payload_start + regions->regions[i + 1].offset;
+ else
+ poison_end = payload_end;
+
+ if (poison_end > payload_end)
+ return -EINVAL;
+
+ ret = kasan_poison_range(poison_start, poison_end - poison_start);
+ if (ret)
+ return ret;
+ }
+
+ /* Poison the padded area preceding the payload. */
+ return kasan_poison_range(payload_start - rt->padding_size, rt->padding_size);
+}
+
+static bool kfuzztest_input_is_valid(struct reloc_region_array *regions, struct reloc_table *rt,
+ unsigned char *payload_start, unsigned char *payload_end)
+{
+ size_t payload_size = payload_end - payload_start;
+ struct reloc_region reg, next_reg;
+ size_t usable_payload_size;
+ uint32_t region_end_offset;
+ struct reloc_entry reloc;
+ uint32_t i;
+
+ if (payload_start > payload_end)
+ return false;
+ if (payload_size < KFUZZTEST_POISON_SIZE)
+ return false;
+ if ((uintptr_t)payload_end % KFUZZTEST_POISON_SIZE)
+ return false;
+ usable_payload_size = payload_size - KFUZZTEST_POISON_SIZE;
+
+ for (i = 0; i < regions->num_regions; i++) {
+ reg = regions->regions[i];
+ if (check_add_overflow(reg.offset, reg.size, ®ion_end_offset))
+ return false;
+ if ((size_t)region_end_offset > usable_payload_size)
+ return false;
+
+ if (i < regions->num_regions - 1) {
+ next_reg = regions->regions[i + 1];
+ if (reg.offset > next_reg.offset)
+ return false;
+ /* Enforce the minimum poisonable gap between
+ * consecutive regions. */
+ if (reg.offset + reg.size + KFUZZTEST_POISON_SIZE > next_reg.offset)
+ return false;
+ }
+ }
+
+ if (rt->padding_size < KFUZZTEST_POISON_SIZE) {
+ pr_info("validation failed because rt->padding_size = %u", rt->padding_size);
+ return false;
+ }
+
+ for (i = 0; i < rt->num_entries; i++) {
+ reloc = rt->entries[i];
+ if (reloc.region_id >= regions->num_regions)
+ return false;
+ if (reloc.value != KFUZZTEST_REGIONID_NULL && reloc.value >= regions->num_regions)
+ return false;
+
+ reg = regions->regions[reloc.region_id];
+ if (reloc.region_offset % (sizeof(uintptr_t)) || reloc.region_offset + sizeof(uintptr_t) > reg.size)
+ return false;
+ }
+
+ return true;
+}
+
+static int kfuzztest_parse_input_v0(unsigned char *input, size_t input_size, struct reloc_region_array **ret_regions,
+ struct reloc_table **ret_reloc_table, unsigned char **ret_payload_start,
+ unsigned char **ret_payload_end)
+{
+ size_t reloc_entries_size, reloc_regions_size;
+ unsigned char *payload_end, *payload_start;
+ size_t reloc_table_size, regions_size;
+ struct reloc_region_array *regions;
+ struct reloc_table *rt;
+ size_t curr_offset = 0;
+
+ if (input_size < sizeof(struct reloc_region_array) + sizeof(struct reloc_table))
+ return -EINVAL;
+
+ regions = (struct reloc_region_array *)input;
+ if (check_mul_overflow(regions->num_regions, sizeof(struct reloc_region), &reloc_regions_size))
+ return -EINVAL;
+ if (check_add_overflow(sizeof(*regions), reloc_regions_size, ®ions_size))
+ return -EINVAL;
+
+ curr_offset = regions_size;
+ if (curr_offset > input_size)
+ return -EINVAL;
+ if (input_size - curr_offset < sizeof(struct reloc_table))
+ return -EINVAL;
+
+ rt = (struct reloc_table *)(input + curr_offset);
+
+ if (check_mul_overflow((size_t)rt->num_entries, sizeof(struct reloc_entry), &reloc_entries_size))
+ return -EINVAL;
+ if (check_add_overflow(sizeof(*rt), reloc_entries_size, &reloc_table_size))
+ return -EINVAL;
+ if (check_add_overflow(reloc_table_size, rt->padding_size, &reloc_table_size))
+ return -EINVAL;
+
+ if (check_add_overflow(curr_offset, reloc_table_size, &curr_offset))
+ return -EINVAL;
+ if (curr_offset > input_size)
+ return -EINVAL;
+
+ payload_start = input + curr_offset;
+ payload_end = input + input_size;
+
+ if (!kfuzztest_input_is_valid(regions, rt, payload_start, payload_end))
+ return -EINVAL;
+
+ *ret_regions = regions;
+ *ret_reloc_table = rt;
+ *ret_payload_start = payload_start;
+ *ret_payload_end = payload_end;
+ return 0;
+}
+
+static int kfuzztest_parse_and_relocate_v0(unsigned char *input, size_t input_size, void **arg_ret)
+{
+ unsigned char *payload_start, *payload_end;
+ struct reloc_region_array *regions;
+ struct reloc_table *reloc_table;
+ int ret;
+
+ ret = kfuzztest_parse_input_v0(input, input_size, ®ions, &reloc_table, &payload_start, &payload_end);
+ if (ret < 0)
+ return ret;
+
+ ret = kfuzztest_relocate_v0(regions, reloc_table, payload_start, payload_end);
+ if (ret < 0)
+ return ret;
+ *arg_ret = (void *)payload_start;
+ return 0;
+}
+
+int kfuzztest_parse_and_relocate(void *input, size_t input_size, void **arg_ret)
+{
+ size_t header_size = 2 * sizeof(u32);
+ u32 version, magic;
+
+ if (input_size < sizeof(u32) + sizeof(u32))
+ return -EINVAL;
+
+ magic = *(u32 *)input;
+ if (magic != KFUZZTEST_HEADER_MAGIC)
+ return -EINVAL;
+
+ version = *(u32 *)(input + sizeof(u32));
+ switch (version) {
+ case KFUZZTEST_V0:
+ return kfuzztest_parse_and_relocate_v0(input + header_size, input_size - header_size, arg_ret);
+ }
+
+ return -EINVAL;
+}
--
2.51.0.470.ga7dc726c21-goog
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH v2 03/10] kfuzztest: implement core module and input processing
2025-09-19 14:57 ` [PATCH v2 03/10] kfuzztest: implement core module and input processing Ethan Graham
@ 2025-09-19 15:05 ` Alexander Potapenko
0 siblings, 0 replies; 30+ messages in thread
From: Alexander Potapenko @ 2025-09-19 15:05 UTC (permalink / raw)
To: Ethan Graham
Cc: ethangraham, andreyknvl, andy, brauner, brendan.higgins, davem,
davidgow, dhowells, dvyukov, elver, herbert, ignat, jack, jannh,
johannes, kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel,
linux-mm, lukas, rmoar, shuah, sj, tarasmadan
On Fri, Sep 19, 2025 at 4:58 PM Ethan Graham <ethan.w.s.graham@gmail.com> wrote:
>
> From: Ethan Graham <ethangraham@google.com>
>
> Add the core runtime implementation for KFuzzTest. This includes the
> module initialization, and the logic for receiving and processing
> user-provided inputs through debugfs.
>
> On module load, the framework discovers all test targets by iterating
> over the .kfuzztest_target section, creating a corresponding debugfs
> directory with a write-only 'input' file for each of them.
>
> Writing to an 'input' file triggers the main fuzzing sequence:
> 1. The serialized input is copied from userspace into a kernel buffer.
> 2. The buffer is parsed to validate the region array and relocation
> table.
> 3. Pointers are patched based on the relocation entries, and in KASAN
> builds the inter-region padding is poisoned.
> 4. The resulting struct is passed to the user-defined test logic.
>
> Signed-off-by: Ethan Graham <ethangraham@google.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH v2 04/10] tools: add kfuzztest-bridge utility
2025-09-19 14:57 [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework Ethan Graham
` (2 preceding siblings ...)
2025-09-19 14:57 ` [PATCH v2 03/10] kfuzztest: implement core module and input processing Ethan Graham
@ 2025-09-19 14:57 ` Ethan Graham
2025-09-19 15:05 ` Alexander Potapenko
2025-09-19 14:57 ` [PATCH v2 05/10] kfuzztest: add ReST documentation Ethan Graham
` (7 subsequent siblings)
11 siblings, 1 reply; 30+ messages in thread
From: Ethan Graham @ 2025-09-19 14:57 UTC (permalink / raw)
To: ethangraham, glider
Cc: andreyknvl, andy, brauner, brendan.higgins, davem, davidgow,
dhowells, dvyukov, elver, herbert, ignat, jack, jannh, johannes,
kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel, linux-mm,
lukas, rmoar, shuah, sj, tarasmadan
From: Ethan Graham <ethangraham@google.com>
Introduce the kfuzztest-bridge tool, a userspace utility for sending
structured inputs to KFuzzTest harnesses via debugfs.
The bridge takes a textual description of the expected input format, a
file containing random bytes, and the name of the target fuzz test. It
parses the description, encodes the random data into the binary format
expected by the kernel, and writes the result to the corresponding
debugfs entry.
This allows for both simple manual testing and integration with
userspace fuzzing engines. For example, it can be used for smoke testing
by providing data from /dev/urandom, or act as a bridge for blob-based
fuzzers (e.g., AFL) to target KFuzzTest harnesses.
Signed-off-by: Ethan Graham <ethangraham@google.com>
---
PR v2:
- Move kfuzztest-bridge tool under tools/testing, as suggested by
SeongJae Park.
- Cleanup several resource leaks that were pointed out by Alexander
Potapenko.
PR v1:
- Add additional context in header comment of kfuzztest-bridge/parser.c.
- Add some missing NULL checks.
- Refactor skip_whitespace() function in input_lexer.c.
- Use ctx->minalign to compute correct region alignment, which is read
from /sys/kernel/debug/kfuzztest/_config/minalign.
---
---
tools/Makefile | 18 +-
tools/testing/kfuzztest-bridge/.gitignore | 2 +
tools/testing/kfuzztest-bridge/Build | 6 +
tools/testing/kfuzztest-bridge/Makefile | 49 ++
tools/testing/kfuzztest-bridge/bridge.c | 115 +++++
tools/testing/kfuzztest-bridge/byte_buffer.c | 85 ++++
tools/testing/kfuzztest-bridge/byte_buffer.h | 31 ++
tools/testing/kfuzztest-bridge/encoder.c | 390 ++++++++++++++++
tools/testing/kfuzztest-bridge/encoder.h | 16 +
tools/testing/kfuzztest-bridge/input_lexer.c | 256 +++++++++++
tools/testing/kfuzztest-bridge/input_lexer.h | 58 +++
tools/testing/kfuzztest-bridge/input_parser.c | 423 ++++++++++++++++++
tools/testing/kfuzztest-bridge/input_parser.h | 82 ++++
tools/testing/kfuzztest-bridge/rand_stream.c | 77 ++++
tools/testing/kfuzztest-bridge/rand_stream.h | 57 +++
15 files changed, 1662 insertions(+), 3 deletions(-)
create mode 100644 tools/testing/kfuzztest-bridge/.gitignore
create mode 100644 tools/testing/kfuzztest-bridge/Build
create mode 100644 tools/testing/kfuzztest-bridge/Makefile
create mode 100644 tools/testing/kfuzztest-bridge/bridge.c
create mode 100644 tools/testing/kfuzztest-bridge/byte_buffer.c
create mode 100644 tools/testing/kfuzztest-bridge/byte_buffer.h
create mode 100644 tools/testing/kfuzztest-bridge/encoder.c
create mode 100644 tools/testing/kfuzztest-bridge/encoder.h
create mode 100644 tools/testing/kfuzztest-bridge/input_lexer.c
create mode 100644 tools/testing/kfuzztest-bridge/input_lexer.h
create mode 100644 tools/testing/kfuzztest-bridge/input_parser.c
create mode 100644 tools/testing/kfuzztest-bridge/input_parser.h
create mode 100644 tools/testing/kfuzztest-bridge/rand_stream.c
create mode 100644 tools/testing/kfuzztest-bridge/rand_stream.h
diff --git a/tools/Makefile b/tools/Makefile
index c31cbbd12c45..dfb0cd19aeb9 100644
--- a/tools/Makefile
+++ b/tools/Makefile
@@ -21,6 +21,7 @@ help:
@echo ' hv - tools used when in Hyper-V clients'
@echo ' iio - IIO tools'
@echo ' intel-speed-select - Intel Speed Select tool'
+ @echo ' kfuzztest-bridge - KFuzzTest userspace utility'
@echo ' kvm_stat - top-like utility for displaying kvm statistics'
@echo ' leds - LEDs tools'
@echo ' nolibc - nolibc headers testing and installation'
@@ -98,6 +99,9 @@ sched_ext: FORCE
selftests: FORCE
$(call descend,testing/$@)
+kfuzztest-bridge: FORCE
+ $(call descend,testing/kfuzztest-bridge)
+
thermal: FORCE
$(call descend,lib/$@)
@@ -126,7 +130,8 @@ all: acpi counter cpupower gpio hv firewire \
perf selftests bootconfig spi turbostat usb \
virtio mm bpf x86_energy_perf_policy \
tmon freefall iio objtool kvm_stat wmi \
- debugging tracing thermal thermometer thermal-engine ynl
+ debugging tracing thermal thermometer thermal-engine ynl \
+ kfuzztest-bridge
acpi_install:
$(call descend,power/$(@:_install=),install)
@@ -140,6 +145,9 @@ counter_install firewire_install gpio_install hv_install iio_install perf_instal
selftests_install:
$(call descend,testing/$(@:_install=),install)
+kfuzztest-bridge_install:
+ $(call descend,testing/kfuzztest-bridge,install)
+
thermal_install:
$(call descend,lib/$(@:_install=),install)
@@ -170,7 +178,8 @@ install: acpi_install counter_install cpupower_install gpio_install \
virtio_install mm_install bpf_install x86_energy_perf_policy_install \
tmon_install freefall_install objtool_install kvm_stat_install \
wmi_install debugging_install intel-speed-select_install \
- tracing_install thermometer_install thermal-engine_install ynl_install
+ tracing_install thermometer_install thermal-engine_install ynl_install \
+ kfuzztest-bridge_install
acpi_clean:
$(call descend,power/acpi,clean)
@@ -200,6 +209,9 @@ sched_ext_clean:
selftests_clean:
$(call descend,testing/$(@:_clean=),clean)
+kfuzztest-bridge_clean:
+ $(call descend,testing/kfuzztest-bridge,clean)
+
thermal_clean:
$(call descend,lib/thermal,clean)
@@ -230,6 +242,6 @@ clean: acpi_clean counter_clean cpupower_clean hv_clean firewire_clean \
freefall_clean build_clean libbpf_clean libsubcmd_clean \
gpio_clean objtool_clean leds_clean wmi_clean firmware_clean debugging_clean \
intel-speed-select_clean tracing_clean thermal_clean thermometer_clean thermal-engine_clean \
- sched_ext_clean ynl_clean
+ sched_ext_clean ynl_clean kfuzztest-bridge_clean
.PHONY: FORCE
diff --git a/tools/testing/kfuzztest-bridge/.gitignore b/tools/testing/kfuzztest-bridge/.gitignore
new file mode 100644
index 000000000000..4aa9fb0d44e2
--- /dev/null
+++ b/tools/testing/kfuzztest-bridge/.gitignore
@@ -0,0 +1,2 @@
+# SPDX-License-Identifier: GPL-2.0-only
+kfuzztest-bridge
diff --git a/tools/testing/kfuzztest-bridge/Build b/tools/testing/kfuzztest-bridge/Build
new file mode 100644
index 000000000000..d07341a226d6
--- /dev/null
+++ b/tools/testing/kfuzztest-bridge/Build
@@ -0,0 +1,6 @@
+kfuzztest-bridge-y += bridge.o
+kfuzztest-bridge-y += byte_buffer.o
+kfuzztest-bridge-y += encoder.o
+kfuzztest-bridge-y += input_lexer.o
+kfuzztest-bridge-y += input_parser.o
+kfuzztest-bridge-y += rand_stream.o
diff --git a/tools/testing/kfuzztest-bridge/Makefile b/tools/testing/kfuzztest-bridge/Makefile
new file mode 100644
index 000000000000..6e110bdeaee5
--- /dev/null
+++ b/tools/testing/kfuzztest-bridge/Makefile
@@ -0,0 +1,49 @@
+# SPDX-License-Identifier: GPL-2.0
+# Makefile for KFuzzTest-Bridge
+include ../../scripts/Makefile.include
+
+bindir ?= /usr/bin
+
+ifeq ($(srctree),)
+srctree := $(patsubst %/,%,$(dir $(CURDIR)))
+srctree := $(patsubst %/,%,$(dir $(srctree)))
+srctree := $(patsubst %/,%,$(dir $(srctree)))
+endif
+
+MAKEFLAGS += -r
+
+override CFLAGS += -O2 -g
+override CFLAGS += -Wall -Wextra
+override CFLAGS += -D_GNU_SOURCE
+override CFLAGS += -I$(OUTPUT)include -I$(srctree)/tools/include
+
+ALL_TARGETS := kfuzztest-bridge
+ALL_PROGRAMS := $(patsubst %,$(OUTPUT)%,$(ALL_TARGETS))
+
+KFUZZTEST_BRIDGE_IN := $(OUTPUT)kfuzztest-bridge-in.o
+KFUZZTEST_BRIDGE := $(OUTPUT)kfuzztest-bridge
+
+all: $(ALL_PROGRAMS)
+
+export srctree OUTPUT CC LD CFLAGS
+include $(srctree)/tools/build/Makefile.include
+
+$(KFUZZTEST_BRIDGE_IN): FORCE
+ $(Q)$(MAKE) $(build)=kfuzztest-bridge
+
+$(KFUZZTEST_BRIDGE): $(KFUZZTEST_BRIDGE_IN)
+ $(QUIET_LINK)$(CC) $(CFLAGS) $< -o $@ $(LDFLAGS)
+
+clean:
+ rm -f $(ALL_PROGRAMS)
+ find $(or $(OUTPUT),.) -name '*.o' -delete -o -name '\.*.d' -delete -o -name '\.*.o.cmd' -delete
+
+install: $(ALL_PROGRAMS)
+ install -d -m 755 $(DESTDIR)$(bindir); \
+ for program in $(ALL_PROGRAMS); do \
+ install $$program $(DESTDIR)$(bindir); \
+ done
+
+FORCE:
+
+.PHONY: all install clean FORCE prepare
diff --git a/tools/testing/kfuzztest-bridge/bridge.c b/tools/testing/kfuzztest-bridge/bridge.c
new file mode 100644
index 000000000000..aec0eb4e9ff7
--- /dev/null
+++ b/tools/testing/kfuzztest-bridge/bridge.c
@@ -0,0 +1,115 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * KFuzzTest tool for sending inputs into a KFuzzTest harness
+ *
+ * Copyright 2025 Google LLC
+ */
+#include <errno.h>
+#include <fcntl.h>
+#include <stdio.h>
+#include <string.h>
+#include <unistd.h>
+
+#include "byte_buffer.h"
+#include "encoder.h"
+#include "input_lexer.h"
+#include "input_parser.h"
+#include "rand_stream.h"
+
+static int invoke_kfuzztest_target(const char *target_name, const char *data, ssize_t data_size)
+{
+ ssize_t bytes_written;
+ char *buf = NULL;
+ int ret;
+ int fd;
+
+ if (asprintf(&buf, "/sys/kernel/debug/kfuzztest/%s/input", target_name) < 0)
+ return -ENOMEM;
+
+ fd = openat(AT_FDCWD, buf, O_WRONLY, 0);
+ if (fd < 0) {
+ ret = -errno;
+ goto out_free;
+ }
+
+ /*
+ * A KFuzzTest target's debugfs handler expects the entire input to be
+ * written in a single contiguous blob. Treat partial writes as errors.
+ */
+ bytes_written = write(fd, data, data_size);
+ if (bytes_written != data_size) {
+ ret = (bytes_written < 0) ? -errno : -EIO;
+ goto out_close;
+ }
+ ret = 0;
+
+out_close:
+ if (close(fd) != 0 && ret == 0)
+ ret = -errno;
+out_free:
+ free(buf);
+ return ret;
+}
+
+static int invoke_one(const char *input_fmt, const char *fuzz_target, const char *input_filepath)
+{
+ struct ast_node *ast_prog;
+ struct byte_buffer *bb;
+ struct rand_stream *rs;
+ struct token **tokens;
+ size_t num_tokens;
+ size_t num_bytes;
+ int err;
+
+ err = tokenize(input_fmt, &tokens, &num_tokens);
+ if (err) {
+ fprintf(stderr, "tokenization failed: %s\n", strerror(-err));
+ return err;
+ }
+
+ err = parse(tokens, num_tokens, &ast_prog);
+ if (err) {
+ fprintf(stderr, "parsing failed: %s\n", strerror(-err));
+ goto cleanup_tokens;
+ }
+
+ rs = new_rand_stream(input_filepath, 1024);
+ if (!rs) {
+ err = -ENOMEM;
+ goto cleanup_ast;
+ }
+
+ err = encode(ast_prog, rs, &num_bytes, &bb);
+ if (err == STREAM_EOF) {
+ fprintf(stderr, "encoding failed: reached EOF in %s\n", input_filepath);
+ err = -EINVAL;
+ goto cleanup_rs;
+ } else if (err) {
+ fprintf(stderr, "encoding failed: %s\n", strerror(-err));
+ goto cleanup_rs;
+ }
+
+ err = invoke_kfuzztest_target(fuzz_target, bb->buffer, (ssize_t)num_bytes);
+ if (err)
+ fprintf(stderr, "invocation failed: %s\n", strerror(-err));
+
+ destroy_byte_buffer(bb);
+cleanup_rs:
+ destroy_rand_stream(rs);
+cleanup_ast:
+ destroy_ast_node(ast_prog);
+cleanup_tokens:
+ destroy_tokens(tokens, num_tokens);
+ return err;
+}
+
+int main(int argc, char *argv[])
+{
+ if (argc != 4) {
+ printf("Usage: %s <input-description> <fuzz-target-name> <input-file>\n", argv[0]);
+ printf("For more detailed information see Documentation/dev-tools/kfuzztest.rst\n");
+ return 1;
+ }
+
+ return invoke_one(argv[1], argv[2], argv[3]);
+}
diff --git a/tools/testing/kfuzztest-bridge/byte_buffer.c b/tools/testing/kfuzztest-bridge/byte_buffer.c
new file mode 100644
index 000000000000..1974dbf3862e
--- /dev/null
+++ b/tools/testing/kfuzztest-bridge/byte_buffer.c
@@ -0,0 +1,85 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * A simple byte buffer implementation for encoding binary data
+ *
+ * Copyright 2025 Google LLC
+ */
+#include <asm-generic/errno-base.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "byte_buffer.h"
+
+struct byte_buffer *new_byte_buffer(size_t initial_size)
+{
+ struct byte_buffer *ret;
+ size_t alloc_size = initial_size >= 8 ? initial_size : 8;
+
+ ret = malloc(sizeof(*ret));
+ if (!ret)
+ return NULL;
+
+ ret->alloc_size = alloc_size;
+ ret->buffer = malloc(alloc_size);
+ if (!ret->buffer) {
+ free(ret);
+ return NULL;
+ }
+ ret->num_bytes = 0;
+ return ret;
+}
+
+void destroy_byte_buffer(struct byte_buffer *buf)
+{
+ free(buf->buffer);
+ free(buf);
+}
+
+int append_bytes(struct byte_buffer *buf, const char *bytes, size_t num_bytes)
+{
+ size_t req_size;
+ size_t new_size;
+ char *new_ptr;
+
+ req_size = buf->num_bytes + num_bytes;
+ new_size = buf->alloc_size;
+
+ while (req_size > new_size)
+ new_size *= 2;
+ if (new_size != buf->alloc_size) {
+ new_ptr = realloc(buf->buffer, new_size);
+ if (!new_ptr)
+ return -ENOMEM;
+ buf->buffer = new_ptr;
+ buf->alloc_size = new_size;
+ }
+ memcpy(buf->buffer + buf->num_bytes, bytes, num_bytes);
+ buf->num_bytes += num_bytes;
+ return 0;
+}
+
+int append_byte(struct byte_buffer *buf, char c)
+{
+ return append_bytes(buf, &c, 1);
+}
+
+int encode_le(struct byte_buffer *buf, uint64_t value, size_t byte_width)
+{
+ size_t i;
+ int ret;
+
+ for (i = 0; i < byte_width; ++i)
+ if ((ret = append_byte(buf, (uint8_t)((value >> (i * 8)) & 0xFF))))
+ return ret;
+ return 0;
+}
+
+int pad(struct byte_buffer *buf, size_t num_padding)
+{
+ int ret;
+ size_t i;
+ for (i = 0; i < num_padding; i++)
+ if ((ret = append_byte(buf, 0)))
+ return ret;
+ return 0;
+}
diff --git a/tools/testing/kfuzztest-bridge/byte_buffer.h b/tools/testing/kfuzztest-bridge/byte_buffer.h
new file mode 100644
index 000000000000..6a31bfb5e78f
--- /dev/null
+++ b/tools/testing/kfuzztest-bridge/byte_buffer.h
@@ -0,0 +1,31 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * A simple byte buffer implementation for encoding binary data
+ *
+ * Copyright 2025 Google LLC
+ */
+#ifndef KFUZZTEST_BRIDGE_BYTE_BUFFER_H
+#define KFUZZTEST_BRIDGE_BYTE_BUFFER_H
+
+#include <stdint.h>
+#include <stdlib.h>
+
+struct byte_buffer {
+ char *buffer;
+ size_t num_bytes;
+ size_t alloc_size;
+};
+
+struct byte_buffer *new_byte_buffer(size_t initial_size);
+
+void destroy_byte_buffer(struct byte_buffer *buf);
+
+int append_bytes(struct byte_buffer *buf, const char *bytes, size_t num_bytes);
+
+int append_byte(struct byte_buffer *buf, char c);
+
+int encode_le(struct byte_buffer *buf, uint64_t value, size_t byte_width);
+
+int pad(struct byte_buffer *buf, size_t num_padding);
+
+#endif /* KFUZZTEST_BRIDGE_BYTE_BUFFER_H */
diff --git a/tools/testing/kfuzztest-bridge/encoder.c b/tools/testing/kfuzztest-bridge/encoder.c
new file mode 100644
index 000000000000..11ff5bd589d3
--- /dev/null
+++ b/tools/testing/kfuzztest-bridge/encoder.c
@@ -0,0 +1,390 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Encoder for KFuzzTest binary input format
+ *
+ * Copyright 2025 Google LLC
+ */
+#include <errno.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "byte_buffer.h"
+#include "input_parser.h"
+#include "rand_stream.h"
+
+#define KFUZZTEST_MAGIC 0xBFACE
+#define KFUZZTEST_PROTO_VERSION 0
+
+/*
+ * The KFuzzTest binary input format requires at least 8 bytes of padding
+ * at the head and tail of every region.
+ */
+#define KFUZZTEST_POISON_SIZE 8
+
+#define BUFSIZE_SMALL 32
+#define BUFSIZE_LARGE 128
+
+struct region_info {
+ const char *name;
+ uint32_t offset;
+ uint32_t size;
+};
+
+struct reloc_info {
+ uint32_t src_reg;
+ uint32_t offset;
+ uint32_t dst_reg;
+};
+
+struct encoder_ctx {
+ struct byte_buffer *payload;
+ struct rand_stream *rand;
+
+ struct region_info *regions;
+ size_t num_regions;
+
+ struct reloc_info *relocations;
+ size_t num_relocations;
+
+ size_t minalign;
+ size_t reg_offset;
+ int curr_reg;
+};
+
+static void cleanup_ctx(struct encoder_ctx *ctx)
+{
+ if (ctx->regions)
+ free(ctx->regions);
+ if (ctx->relocations)
+ free(ctx->relocations);
+ if (ctx->payload)
+ destroy_byte_buffer(ctx->payload);
+}
+
+static int read_minalign(struct encoder_ctx *ctx)
+{
+ const char *minalign_file = "/sys/kernel/debug/kfuzztest/_config/minalign";
+ char buffer[64 + 1] = { 0 };
+ int ret = 0;
+
+ FILE *f = fopen(minalign_file, "r");
+ if (!f)
+ return -ENOENT;
+
+ fread(&buffer, 1, sizeof(buffer) - 1, f);
+ if (ferror(f))
+ return ferror(f);
+
+ /*
+ * atoi returns 0 on error. Since we expect a strictly positive
+ * minalign value on all architectures, any non-positive value
+ * represents an error.
+ */
+ ret = atoi(buffer);
+ if (ret <= 0) {
+ fclose(f);
+ return -EINVAL;
+ }
+ ctx->minalign = ret;
+ fclose(f);
+ return 0;
+}
+
+static int pad_payload(struct encoder_ctx *ctx, size_t amount)
+{
+ int ret;
+
+ if ((ret = pad(ctx->payload, amount)))
+ return ret;
+ ctx->reg_offset += amount;
+ return 0;
+}
+
+static int align_payload(struct encoder_ctx *ctx, size_t alignment)
+{
+ size_t pad_amount = ROUND_UP_TO_MULTIPLE(ctx->payload->num_bytes, alignment) - ctx->payload->num_bytes;
+ return pad_payload(ctx, pad_amount);
+}
+
+static int lookup_reg(struct encoder_ctx *ctx, const char *name)
+{
+ size_t i;
+
+ for (i = 0; i < ctx->num_regions; i++) {
+ if (strcmp(ctx->regions[i].name, name) == 0)
+ return i;
+ }
+ return -ENOENT;
+}
+
+static int add_reloc(struct encoder_ctx *ctx, struct reloc_info reloc)
+{
+ void *new_ptr = realloc(ctx->relocations, (ctx->num_relocations + 1) * sizeof(struct reloc_info));
+ if (!new_ptr)
+ return -ENOMEM;
+
+ ctx->relocations = new_ptr;
+ ctx->relocations[ctx->num_relocations] = reloc;
+ ctx->num_relocations++;
+ return 0;
+}
+
+static int build_region_map(struct encoder_ctx *ctx, struct ast_node *top_level)
+{
+ struct ast_program *prog;
+ struct ast_node *reg;
+ size_t i;
+
+ if (top_level->type != NODE_PROGRAM)
+ return -EINVAL;
+
+ prog = &top_level->data.program;
+ ctx->regions = malloc(prog->num_members * sizeof(struct region_info));
+ if (!ctx->regions)
+ return -ENOMEM;
+
+ ctx->num_regions = prog->num_members;
+ for (i = 0; i < ctx->num_regions; i++) {
+ reg = prog->members[i];
+ /* Offset is determined after the second pass. */
+ ctx->regions[i] = (struct region_info){
+ .name = reg->data.region.name,
+ .size = node_size(reg),
+ };
+ }
+ return 0;
+}
+/**
+ * Encodes a value node as little-endian. A value node is one that has no
+ * children, and can therefore be directly written into the payload.
+ */
+static int encode_value_le(struct encoder_ctx *ctx, struct ast_node *node)
+{
+ size_t array_size;
+ char rand_char;
+ size_t length;
+ size_t i;
+ int reg;
+ int ret;
+
+ switch (node->type) {
+ case NODE_ARRAY:
+ array_size = node->data.array.num_elems * node->data.array.elem_size;
+ for (i = 0; i < array_size; i++) {
+ if ((ret = next_byte(ctx->rand, &rand_char)))
+ return ret;
+ if ((ret = append_byte(ctx->payload, rand_char)))
+ return ret;
+ }
+ ctx->reg_offset += array_size;
+ if (node->data.array.null_terminated) {
+ if ((ret = pad_payload(ctx, 1)))
+ return ret;
+ ctx->reg_offset++;
+ }
+ break;
+ case NODE_LENGTH:
+ reg = lookup_reg(ctx, node->data.length.length_of);
+ if (reg < 0)
+ return reg;
+ length = ctx->regions[reg].size;
+ if ((ret = encode_le(ctx->payload, length, node->data.length.byte_width)))
+ return ret;
+ ctx->reg_offset += node->data.length.byte_width;
+ break;
+ case NODE_PRIMITIVE:
+ for (i = 0; i < node->data.primitive.byte_width; i++) {
+ if ((ret = next_byte(ctx->rand, &rand_char)))
+ return ret;
+ if ((ret = append_byte(ctx->payload, rand_char)))
+ return ret;
+ }
+ ctx->reg_offset += node->data.primitive.byte_width;
+ break;
+ case NODE_POINTER:
+ reg = lookup_reg(ctx, node->data.pointer.points_to);
+ if (reg < 0)
+ return reg;
+ if ((ret = add_reloc(ctx, (struct reloc_info){ .src_reg = ctx->curr_reg,
+ .offset = ctx->reg_offset,
+ .dst_reg = reg })))
+ return ret;
+ /* Placeholder pointer value, as pointers are patched by KFuzzTest anyways. */
+ if ((ret = encode_le(ctx->payload, UINTPTR_MAX, sizeof(uintptr_t))))
+ return ret;
+ ctx->reg_offset += sizeof(uintptr_t);
+ break;
+ case NODE_PROGRAM:
+ case NODE_REGION:
+ default:
+ return -EINVAL;
+ }
+ return 0;
+}
+
+static int encode_region(struct encoder_ctx *ctx, struct ast_region *reg)
+{
+ struct ast_node *child;
+ size_t i;
+ int ret;
+
+ ctx->reg_offset = 0;
+ for (i = 0; i < reg->num_members; i++) {
+ child = reg->members[i];
+ if ((ret = align_payload(ctx, node_alignment(child))))
+ return ret;
+ if ((ret = encode_value_le(ctx, child)))
+ return ret;
+ }
+ return 0;
+}
+
+static int encode_payload(struct encoder_ctx *ctx, struct ast_node *top_level)
+{
+ struct ast_node *reg;
+ size_t i;
+ int ret;
+
+ for (i = 0; i < ctx->num_regions; i++) {
+ reg = top_level->data.program.members[i];
+ if ((ret = align_payload(ctx, MAX(ctx->minalign, node_alignment(reg)))))
+ return ret;
+
+ ctx->curr_reg = i;
+ ctx->regions[i].offset = ctx->payload->num_bytes;
+ if ((ret = encode_region(ctx, ®->data.region)))
+ return ret;
+ if ((ret = pad_payload(ctx, KFUZZTEST_POISON_SIZE)))
+ return ret;
+ }
+ return align_payload(ctx, ctx->minalign);
+}
+
+static int encode_region_array(struct encoder_ctx *ctx, struct byte_buffer **ret)
+{
+ struct byte_buffer *reg_array;
+ struct region_info info;
+ int retcode;
+ size_t i;
+
+ reg_array = new_byte_buffer(BUFSIZE_SMALL);
+ if (!reg_array)
+ return -ENOMEM;
+
+ if ((retcode = encode_le(reg_array, ctx->num_regions, sizeof(uint32_t))))
+ goto fail;
+
+ for (i = 0; i < ctx->num_regions; i++) {
+ info = ctx->regions[i];
+ if ((retcode = encode_le(reg_array, info.offset, sizeof(uint32_t))))
+ goto fail;
+ if ((retcode = encode_le(reg_array, info.size, sizeof(uint32_t))))
+ goto fail;
+ }
+ *ret = reg_array;
+ return 0;
+
+fail:
+ destroy_byte_buffer(reg_array);
+ return retcode;
+}
+
+static int encode_reloc_table(struct encoder_ctx *ctx, size_t padding_amount, struct byte_buffer **ret)
+{
+ struct byte_buffer *reloc_table;
+ struct reloc_info info;
+ int retcode;
+ size_t i;
+
+ reloc_table = new_byte_buffer(BUFSIZE_SMALL);
+ if (!reloc_table)
+ return -ENOMEM;
+
+ if ((retcode = encode_le(reloc_table, ctx->num_relocations, sizeof(uint32_t))) ||
+ (retcode = encode_le(reloc_table, padding_amount, sizeof(uint32_t))))
+ goto fail;
+
+ for (i = 0; i < ctx->num_relocations; i++) {
+ info = ctx->relocations[i];
+ if ((retcode = encode_le(reloc_table, info.src_reg, sizeof(uint32_t))) ||
+ (retcode = encode_le(reloc_table, info.offset, sizeof(uint32_t))) ||
+ (retcode = encode_le(reloc_table, info.dst_reg, sizeof(uint32_t))))
+ goto fail;
+ }
+ pad(reloc_table, padding_amount);
+ *ret = reloc_table;
+ return 0;
+
+fail:
+ destroy_byte_buffer(reloc_table);
+ return retcode;
+}
+
+static size_t reloc_table_size(struct encoder_ctx *ctx)
+{
+ return 2 * sizeof(uint32_t) + 3 * ctx->num_relocations * sizeof(uint32_t);
+}
+
+int encode(struct ast_node *top_level, struct rand_stream *r, size_t *num_bytes, struct byte_buffer **ret)
+{
+ struct byte_buffer *region_array = NULL;
+ struct byte_buffer *final_buffer = NULL;
+ struct byte_buffer *reloc_table = NULL;
+ size_t header_size;
+ int alignment;
+ int retcode;
+
+ struct encoder_ctx ctx = { 0 };
+ if ((retcode = read_minalign(&ctx)))
+ return retcode;
+
+ if ((retcode = build_region_map(&ctx, top_level)))
+ goto fail;
+
+ ctx.rand = r;
+ ctx.payload = new_byte_buffer(BUFSIZE_SMALL);
+ if (!ctx.payload) {
+ retcode = -ENOMEM;
+ goto fail;
+ }
+ if ((retcode = encode_payload(&ctx, top_level)))
+ goto fail;
+
+ if ((retcode = encode_region_array(&ctx, ®ion_array)))
+ goto fail;
+
+ header_size = sizeof(uint64_t) + region_array->num_bytes + reloc_table_size(&ctx);
+ alignment = node_alignment(top_level);
+ if ((retcode = encode_reloc_table(
+ &ctx, ROUND_UP_TO_MULTIPLE(header_size + KFUZZTEST_POISON_SIZE, alignment) - header_size,
+ &reloc_table)))
+ goto fail;
+
+ final_buffer = new_byte_buffer(BUFSIZE_LARGE);
+ if (!final_buffer) {
+ retcode = -ENOMEM;
+ goto fail;
+ }
+
+ if ((retcode = encode_le(final_buffer, KFUZZTEST_MAGIC, sizeof(uint32_t))) ||
+ (retcode = encode_le(final_buffer, KFUZZTEST_PROTO_VERSION, sizeof(uint32_t))) ||
+ (retcode = append_bytes(final_buffer, region_array->buffer, region_array->num_bytes)) ||
+ (retcode = append_bytes(final_buffer, reloc_table->buffer, reloc_table->num_bytes)) ||
+ (retcode = append_bytes(final_buffer, ctx.payload->buffer, ctx.payload->num_bytes))) {
+ destroy_byte_buffer(final_buffer);
+ goto fail;
+ }
+
+ *num_bytes = final_buffer->num_bytes;
+ *ret = final_buffer;
+
+fail:
+ if (region_array)
+ destroy_byte_buffer(region_array);
+ if (reloc_table)
+ destroy_byte_buffer(reloc_table);
+ cleanup_ctx(&ctx);
+ return retcode;
+}
diff --git a/tools/testing/kfuzztest-bridge/encoder.h b/tools/testing/kfuzztest-bridge/encoder.h
new file mode 100644
index 000000000000..73f8c4b7893c
--- /dev/null
+++ b/tools/testing/kfuzztest-bridge/encoder.h
@@ -0,0 +1,16 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Encoder for KFuzzTest binary input format
+ *
+ * Copyright 2025 Google LLC
+ */
+#ifndef KFUZZTEST_BRIDGE_ENCODER_H
+#define KFUZZTEST_BRIDGE_ENCODER_H
+
+#include "input_parser.h"
+#include "rand_stream.h"
+#include "byte_buffer.h"
+
+int encode(struct ast_node *top_level, struct rand_stream *r, size_t *num_bytes, struct byte_buffer **ret);
+
+#endif /* KFUZZTEST_BRIDGE_ENCODER_H */
diff --git a/tools/testing/kfuzztest-bridge/input_lexer.c b/tools/testing/kfuzztest-bridge/input_lexer.c
new file mode 100644
index 000000000000..d0a3e352a265
--- /dev/null
+++ b/tools/testing/kfuzztest-bridge/input_lexer.c
@@ -0,0 +1,256 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Parser for KFuzzTest textual input format
+ *
+ * Copyright 2025 Google LLC
+ */
+#include <errno.h>
+#include <stdbool.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "input_lexer.h"
+
+struct keyword_map {
+ const char *keyword;
+ enum token_type type;
+};
+
+static struct keyword_map keywords[] = {
+ { "ptr", TOKEN_KEYWORD_PTR }, { "arr", TOKEN_KEYWORD_ARR },
+ { "len", TOKEN_KEYWORD_LEN }, { "str", TOKEN_KEYWORD_STR },
+ { "u8", TOKEN_KEYWORD_U8 }, { "u16", TOKEN_KEYWORD_U16 },
+ { "u32", TOKEN_KEYWORD_U32 }, { "u64", TOKEN_KEYWORD_U64 },
+};
+
+static struct token *make_token(enum token_type type, size_t position)
+{
+ struct token *ret = calloc(1, sizeof(*ret));
+ ret->position = position;
+ ret->type = type;
+ return ret;
+}
+
+void destroy_tokens(struct token **tokens, size_t num_tokens)
+{
+ size_t i;
+
+ if (!tokens)
+ return;
+
+ for (i = 0; i < num_tokens; i++)
+ if (tokens[i])
+ free(tokens[i]);
+ free(tokens);
+}
+
+struct lexer {
+ const char *start;
+ const char *current;
+ size_t position;
+};
+
+static char advance(struct lexer *l)
+{
+ l->current++;
+ l->position++;
+ return l->current[-1];
+}
+
+static void retreat(struct lexer *l)
+{
+ l->position--;
+ l->current--;
+}
+
+static char peek(struct lexer *l)
+{
+ return *l->current;
+}
+
+static bool is_digit(char c)
+{
+ return c >= '0' && c <= '9';
+}
+
+static bool is_alpha(char c)
+{
+ return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z');
+}
+
+static bool is_whitespace(char c)
+{
+ switch (c) {
+ case ' ':
+ case '\r':
+ case '\t':
+ case '\n':
+ return true;
+ default:
+ return false;
+ }
+}
+
+static void skip_whitespace(struct lexer *l)
+{
+ while (is_whitespace(peek(l)))
+ advance(l);
+}
+
+static struct token *number(struct lexer *l)
+{
+ struct token *tok;
+ uint64_t value;
+ while (is_digit(peek(l)))
+ advance(l);
+ value = strtoull(l->start, NULL, 10);
+ tok = make_token(TOKEN_INTEGER, l->position);
+ tok->data.integer = value;
+ return tok;
+}
+
+static enum token_type check_keyword(struct lexer *l, const char *keyword,
+ enum token_type type)
+{
+ size_t len = strlen(keyword);
+
+ if (((size_t)(l->current - l->start) == len) &&
+ strncmp(l->start, keyword, len) == 0)
+ return type;
+ return TOKEN_IDENTIFIER;
+}
+
+static struct token *identifier(struct lexer *l)
+{
+ enum token_type type = TOKEN_IDENTIFIER;
+ struct token *tok;
+ size_t i;
+
+ while (is_digit(peek(l)) || is_alpha(peek(l)) || peek(l) == '_')
+ advance(l);
+
+ for (i = 0; i < ARRAY_SIZE(keywords); i++) {
+ if (check_keyword(l, keywords[i].keyword, keywords[i].type) !=
+ TOKEN_IDENTIFIER) {
+ type = keywords[i].type;
+ break;
+ }
+ }
+
+ tok = make_token(type, l->position);
+ if (!tok)
+ return NULL;
+ if (type == TOKEN_IDENTIFIER) {
+ tok->data.identifier.start = l->start;
+ tok->data.identifier.length = l->current - l->start;
+ }
+ return tok;
+}
+
+static struct token *scan_token(struct lexer *l)
+{
+ char c;
+ skip_whitespace(l);
+
+ l->start = l->current;
+ c = peek(l);
+
+ if (c == '\0')
+ return make_token(TOKEN_EOF, l->position);
+
+ advance(l);
+ switch (c) {
+ case '{':
+ return make_token(TOKEN_LBRACE, l->position);
+ case '}':
+ return make_token(TOKEN_RBRACE, l->position);
+ case '[':
+ return make_token(TOKEN_LBRACKET, l->position);
+ case ']':
+ return make_token(TOKEN_RBRACKET, l->position);
+ case ',':
+ return make_token(TOKEN_COMMA, l->position);
+ case ';':
+ return make_token(TOKEN_SEMICOLON, l->position);
+ default:
+ retreat(l);
+ if (is_digit(c))
+ return number(l);
+ if (is_alpha(c) || c == '_')
+ return identifier(l);
+ return make_token(TOKEN_ERROR, l->position);
+ }
+}
+
+int primitive_byte_width(enum token_type type)
+{
+ switch (type) {
+ case TOKEN_KEYWORD_U8:
+ return 1;
+ case TOKEN_KEYWORD_U16:
+ return 2;
+ case TOKEN_KEYWORD_U32:
+ return 4;
+ case TOKEN_KEYWORD_U64:
+ return 8;
+ default:
+ return 0;
+ }
+}
+
+int tokenize(const char *input, struct token ***tokens, size_t *num_tokens)
+{
+ struct lexer l = { .start = input, .current = input };
+ struct token **ret_tokens;
+ size_t token_arr_size;
+ size_t token_count;
+ struct token *tok;
+ void *tmp;
+ int err;
+
+ token_arr_size = 128;
+ ret_tokens = calloc(token_arr_size, sizeof(struct token *));
+ if (!ret_tokens)
+ return -ENOMEM;
+
+ token_count = 0;
+ do {
+ tok = scan_token(&l);
+ if (!tok) {
+ err = -ENOMEM;
+ goto failure;
+ }
+
+ if (token_count == token_arr_size) {
+ token_arr_size *= 2;
+ tmp = realloc(ret_tokens, token_arr_size);
+ if (!tmp) {
+ err = -ENOMEM;
+ goto failure;
+ }
+ ret_tokens = tmp;
+ }
+
+ ret_tokens[token_count] = tok;
+ if (tok->type == TOKEN_ERROR) {
+ err = -EINVAL;
+ goto failure;
+ }
+ token_count++;
+ } while (tok->type != TOKEN_EOF);
+
+ *tokens = ret_tokens;
+ *num_tokens = token_count;
+ return 0;
+
+failure:
+ destroy_tokens(ret_tokens, token_count);
+ return err;
+}
+
+bool is_primitive(struct token *tok)
+{
+ return tok->type >= TOKEN_KEYWORD_U8 && tok->type <= TOKEN_KEYWORD_U64;
+}
diff --git a/tools/testing/kfuzztest-bridge/input_lexer.h b/tools/testing/kfuzztest-bridge/input_lexer.h
new file mode 100644
index 000000000000..40814493c24d
--- /dev/null
+++ b/tools/testing/kfuzztest-bridge/input_lexer.h
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Lexer for KFuzzTest textual input format
+ *
+ * Copyright 2025 Google LLC
+ */
+#ifndef KFUZZTEST_BRIDGE_INPUT_LEXER_H
+#define KFUZZTEST_BRIDGE_INPUT_LEXER_H
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <stdbool.h>
+
+#define ARRAY_SIZE(x) (sizeof(x) / sizeof(x[0]))
+
+enum token_type {
+ TOKEN_LBRACE,
+ TOKEN_RBRACE,
+ TOKEN_LBRACKET,
+ TOKEN_RBRACKET,
+ TOKEN_COMMA,
+ TOKEN_SEMICOLON,
+
+ TOKEN_KEYWORD_PTR,
+ TOKEN_KEYWORD_ARR,
+ TOKEN_KEYWORD_LEN,
+ TOKEN_KEYWORD_STR,
+ TOKEN_KEYWORD_U8,
+ TOKEN_KEYWORD_U16,
+ TOKEN_KEYWORD_U32,
+ TOKEN_KEYWORD_U64,
+
+ TOKEN_IDENTIFIER,
+ TOKEN_INTEGER,
+
+ TOKEN_EOF,
+ TOKEN_ERROR,
+};
+
+struct token {
+ enum token_type type;
+ union {
+ uint64_t integer;
+ struct {
+ const char *start;
+ size_t length;
+ } identifier;
+ } data;
+ int position;
+};
+
+int tokenize(const char *input, struct token ***tokens, size_t *num_tokens);
+void destroy_tokens(struct token **tokens, size_t num_tokens);
+
+bool is_primitive(struct token *tok);
+int primitive_byte_width(enum token_type type);
+
+#endif /* KFUZZTEST_BRIDGE_INPUT_LEXER_H */
diff --git a/tools/testing/kfuzztest-bridge/input_parser.c b/tools/testing/kfuzztest-bridge/input_parser.c
new file mode 100644
index 000000000000..b1fd8ba5217e
--- /dev/null
+++ b/tools/testing/kfuzztest-bridge/input_parser.c
@@ -0,0 +1,423 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Parser for the KFuzzTest textual input format
+ *
+ * This file implements a parser for a simple DSL used to describe C-like data
+ * structures. This format allows the kfuzztest-bridge tool to encode a random
+ * byte stream into the structured binary format expected by a KFuzzTest
+ * harness.
+ *
+ * The format consists of semicolon-separated "regions," which are analogous to
+ * C structs. For example:
+ *
+ * "my_struct { ptr[buf] len[buf, u64] }; buf { arr[u8, 42] };"
+ *
+ * This describes a `my_struct` region that contains a pointer to a `buf` region
+ * and its corresponding length encoded over 8 bytes, where `buf` itself
+ * contains a 42-byte array.
+ *
+ * Copyright 2025 Google LLC
+ */
+#include <errno.h>
+#include <stdio.h>
+#include <string.h>
+
+#include "input_lexer.h"
+#include "input_parser.h"
+
+static struct token *peek(struct parser *p)
+{
+ return p->tokens[p->curr_token];
+}
+
+static struct token *advance(struct parser *p)
+{
+ struct token *tok;
+ if (p->curr_token >= p->token_count)
+ return NULL;
+ tok = peek(p);
+ p->curr_token++;
+ return tok;
+}
+
+static struct token *consume(struct parser *p, enum token_type type, const char *err_msg)
+{
+ if (peek(p)->type != type) {
+ printf("parser failure at position %d: %s\n", peek(p)->position, err_msg);
+ return NULL;
+ }
+ return advance(p);
+}
+
+static bool match(struct parser *p, enum token_type t)
+{
+ struct token *tok = peek(p);
+ return tok->type == t;
+}
+
+static int parse_primitive(struct parser *p, struct ast_node **node_ret)
+{
+ struct ast_node *ret;
+ struct token *tok;
+ int byte_width;
+
+ tok = advance(p);
+ byte_width = primitive_byte_width(tok->type);
+ if (!byte_width)
+ return -EINVAL;
+
+ ret = malloc(sizeof(*ret));
+ if (!ret)
+ return -ENOMEM;
+
+ ret->type = NODE_PRIMITIVE;
+ ret->data.primitive.byte_width = byte_width;
+ *node_ret = ret;
+ return 0;
+}
+
+static int parse_ptr(struct parser *p, struct ast_node **node_ret)
+{
+ const char *points_to;
+ struct ast_node *ret;
+ struct token *tok;
+ if (!consume(p, TOKEN_KEYWORD_PTR, "expected 'ptr'"))
+ return -EINVAL;
+ if (!consume(p, TOKEN_LBRACKET, "expected '['"))
+ return -EINVAL;
+
+ tok = consume(p, TOKEN_IDENTIFIER, "expected identifier");
+ if (!tok)
+ return -EINVAL;
+
+ if (!consume(p, TOKEN_RBRACKET, "expected ']'"))
+ return -EINVAL;
+
+ ret = malloc(sizeof(*ret));
+ ret->type = NODE_POINTER;
+
+ points_to = strndup(tok->data.identifier.start, tok->data.identifier.length);
+ if (!points_to) {
+ free(ret);
+ return -EINVAL;
+ }
+
+ ret->data.pointer.points_to = points_to;
+ *node_ret = ret;
+ return 0;
+}
+
+static int parse_arr(struct parser *p, struct ast_node **node_ret)
+{
+ struct token *type, *num_elems;
+ struct ast_node *ret;
+
+ if (!consume(p, TOKEN_KEYWORD_ARR, "expected 'arr'") || !consume(p, TOKEN_LBRACKET, "expected '['"))
+ return -EINVAL;
+
+ type = advance(p);
+ if (!is_primitive(type))
+ return -EINVAL;
+
+ if (!consume(p, TOKEN_COMMA, "expected ','"))
+ return -EINVAL;
+
+ num_elems = consume(p, TOKEN_INTEGER, "expected integer");
+ if (!num_elems)
+ return -EINVAL;
+
+ if (!consume(p, TOKEN_RBRACKET, "expected ']'"))
+ return -EINVAL;
+
+ ret = malloc(sizeof(*ret));
+ if (!ret)
+ return -ENOMEM;
+
+ ret->type = NODE_ARRAY;
+ ret->data.array.num_elems = num_elems->data.integer;
+ ret->data.array.elem_size = primitive_byte_width(type->type);
+ ret->data.array.null_terminated = false;
+ *node_ret = ret;
+ return 0;
+}
+
+static int parse_str(struct parser *p, struct ast_node **node_ret)
+{
+ struct ast_node *ret;
+ struct token *len;
+
+ if (!consume(p, TOKEN_KEYWORD_STR, "expected 'str'") || !consume(p, TOKEN_LBRACKET, "expected '['"))
+ return -EINVAL;
+
+ len = consume(p, TOKEN_INTEGER, "expected integer");
+ if (!len)
+ return -EINVAL;
+
+ if (!consume(p, TOKEN_RBRACKET, "expected ']'"))
+ return -EINVAL;
+
+ ret = malloc(sizeof(*ret));
+ if (!ret)
+ return -ENOMEM;
+
+ /* A string is the susbet of byte arrays that are null-terminated. */
+ ret->type = NODE_ARRAY;
+ ret->data.array.num_elems = len->data.integer;
+ ret->data.array.elem_size = sizeof(char);
+ ret->data.array.null_terminated = true;
+ *node_ret = ret;
+ return 0;
+}
+
+static int parse_len(struct parser *p, struct ast_node **node_ret)
+{
+ struct token *type, *len;
+ const char *length_of;
+ struct ast_node *ret;
+
+ if (!consume(p, TOKEN_KEYWORD_LEN, "expected 'len'") || !consume(p, TOKEN_LBRACKET, "expected '['"))
+ return -EINVAL;
+
+ len = advance(p);
+ if (len->type != TOKEN_IDENTIFIER)
+ return -EINVAL;
+
+ if (!consume(p, TOKEN_COMMA, "expected ','"))
+ return -EINVAL;
+
+ type = advance(p);
+ if (!is_primitive(type))
+ return -EINVAL;
+
+ if (!consume(p, TOKEN_RBRACKET, "expected ']'"))
+ return -EINVAL;
+
+ ret = malloc(sizeof(*ret));
+ if (!ret)
+ return -ENOMEM;
+
+ length_of = strndup(len->data.identifier.start, len->data.identifier.length);
+ if (!length_of) {
+ free(ret);
+ return -ENOMEM;
+ }
+
+ ret->type = NODE_LENGTH;
+ ret->data.length.length_of = length_of;
+ ret->data.length.byte_width = primitive_byte_width(type->type);
+
+ *node_ret = ret;
+ return 0;
+}
+
+static int parse_type(struct parser *p, struct ast_node **node_ret)
+{
+ if (is_primitive(peek(p)))
+ return parse_primitive(p, node_ret);
+
+ if (peek(p)->type == TOKEN_KEYWORD_PTR)
+ return parse_ptr(p, node_ret);
+
+ if (peek(p)->type == TOKEN_KEYWORD_ARR)
+ return parse_arr(p, node_ret);
+
+ if (peek(p)->type == TOKEN_KEYWORD_STR)
+ return parse_str(p, node_ret);
+
+ if (peek(p)->type == TOKEN_KEYWORD_LEN)
+ return parse_len(p, node_ret);
+
+ return -EINVAL;
+}
+
+static int parse_region(struct parser *p, struct ast_node **node_ret)
+{
+ struct token *tok, *identifier;
+ struct ast_region *region;
+ struct ast_node *node;
+ struct ast_node *ret;
+ void *new_ptr;
+ int err;
+
+ identifier = consume(p, TOKEN_IDENTIFIER, "expected identifier");
+ if (!identifier)
+ return -EINVAL;
+
+ ret = malloc(sizeof(*ret));
+ if (!ret)
+ return -ENOMEM;
+
+ tok = consume(p, TOKEN_LBRACE, "expected '{'");
+ if (!tok) {
+ err = -EINVAL;
+ goto fail_early;
+ }
+
+ region = &ret->data.region;
+ region->name = strndup(identifier->data.identifier.start, identifier->data.identifier.length);
+ if (!region->name) {
+ err = -ENOMEM;
+ goto fail_early;
+ }
+
+ region->num_members = 0;
+ while (!match(p, TOKEN_RBRACE)) {
+ err = parse_type(p, &node);
+ if (err)
+ goto fail;
+ new_ptr = realloc(region->members, (region->num_members + 1) * sizeof(struct ast_node *));
+ if (!new_ptr) {
+ err = -ENOMEM;
+ goto fail;
+ }
+ region->num_members++;
+ region->members = new_ptr;
+ region->members[region->num_members - 1] = node;
+ }
+
+ if (!consume(p, TOKEN_RBRACE, "expected '}'") || !consume(p, TOKEN_SEMICOLON, "expected ';'")) {
+ err = -EINVAL;
+ goto fail;
+ }
+
+ ret->type = NODE_REGION;
+ *node_ret = ret;
+ return 0;
+
+fail:
+ destroy_ast_node(ret);
+ return err;
+
+fail_early:
+ free(ret);
+ return err;
+}
+
+static int parse_program(struct parser *p, struct ast_node **node_ret)
+{
+ struct ast_program *prog;
+ struct ast_node *reg;
+ struct ast_node *ret;
+ void *new_ptr;
+ int err;
+
+ ret = malloc(sizeof(*ret));
+ if (!ret)
+ return -ENOMEM;
+ ret->type = NODE_PROGRAM;
+
+ prog = &ret->data.program;
+ prog->num_members = 0;
+ prog->members = NULL;
+ while (!match(p, TOKEN_EOF)) {
+ err = parse_region(p, ®);
+ if (err)
+ goto fail;
+
+ new_ptr = realloc(prog->members, ++prog->num_members * sizeof(struct ast_node *));
+ if (!new_ptr) {
+ err = -ENOMEM;
+ goto fail;
+ }
+ prog->members = new_ptr;
+ prog->members[prog->num_members - 1] = reg;
+ }
+
+ *node_ret = ret;
+ return 0;
+
+fail:
+ destroy_ast_node(ret);
+ return err;
+}
+
+size_t node_alignment(struct ast_node *node)
+{
+ size_t max_alignment = 1;
+ size_t i;
+
+ switch (node->type) {
+ case NODE_PROGRAM:
+ for (i = 0; i < node->data.program.num_members; i++)
+ max_alignment = MAX(max_alignment, node_alignment(node->data.program.members[i]));
+ return max_alignment;
+ case NODE_REGION:
+ for (i = 0; i < node->data.region.num_members; i++)
+ max_alignment = MAX(max_alignment, node_alignment(node->data.region.members[i]));
+ return max_alignment;
+ case NODE_ARRAY:
+ return node->data.array.elem_size;
+ case NODE_LENGTH:
+ return node->data.length.byte_width;
+ case NODE_PRIMITIVE:
+ /* Primitives are aligned to their size. */
+ return node->data.primitive.byte_width;
+ case NODE_POINTER:
+ return sizeof(uintptr_t);
+ }
+
+ /* Anything should be at least 1-byte-aligned. */
+ return 1;
+}
+
+size_t node_size(struct ast_node *node)
+{
+ size_t total = 0;
+ size_t i;
+
+ switch (node->type) {
+ case NODE_PROGRAM:
+ for (i = 0; i < node->data.program.num_members; i++)
+ total += node_size(node->data.program.members[i]);
+ return total;
+ case NODE_REGION:
+ for (i = 0; i < node->data.region.num_members; i++) {
+ /* Account for padding within region. */
+ total = ROUND_UP_TO_MULTIPLE(total, node_alignment(node->data.region.members[i]));
+ total += node_size(node->data.region.members[i]);
+ }
+ return total;
+ case NODE_ARRAY:
+ return node->data.array.elem_size * node->data.array.num_elems +
+ (node->data.array.null_terminated ? 1 : 0);
+ case NODE_LENGTH:
+ return node->data.length.byte_width;
+ case NODE_PRIMITIVE:
+ return node->data.primitive.byte_width;
+ case NODE_POINTER:
+ return sizeof(uintptr_t);
+ }
+ return 0;
+}
+
+int parse(struct token **tokens, size_t token_count, struct ast_node **node_ret)
+{
+ struct parser p = { .tokens = tokens, .token_count = token_count, .curr_token = 0 };
+ return parse_program(&p, node_ret);
+}
+
+void destroy_ast_node(struct ast_node *node)
+{
+ size_t i;
+
+ switch (node->type) {
+ case NODE_PROGRAM:
+ for (i = 0; i < node->data.program.num_members; i++)
+ destroy_ast_node(node->data.program.members[i]);
+ break;
+ case NODE_REGION:
+ for (i = 0; i < node->data.region.num_members; i++)
+ destroy_ast_node(node->data.region.members[i]);
+ free((void *)node->data.region.name);
+ break;
+ case NODE_LENGTH:
+ free((void *)node->data.length.length_of);
+ break;
+ case NODE_POINTER:
+ free((void *)node->data.pointer.points_to);
+ break;
+ default:
+ break;
+ }
+ free(node);
+}
diff --git a/tools/testing/kfuzztest-bridge/input_parser.h b/tools/testing/kfuzztest-bridge/input_parser.h
new file mode 100644
index 000000000000..5f444b40f672
--- /dev/null
+++ b/tools/testing/kfuzztest-bridge/input_parser.h
@@ -0,0 +1,82 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Parser for KFuzzTest textual input format
+ *
+ * Copyright 2025 Google LLC
+ */
+#ifndef KFUZZTEST_BRIDGE_INPUT_PARSER_H
+#define KFUZZTEST_BRIDGE_INPUT_PARSER_H
+
+#include <stdlib.h>
+
+/* Rounds x up to the nearest multiple of n. */
+#define ROUND_UP_TO_MULTIPLE(x, n) (((n) == 0) ? (0) : (((x) + (n) - 1) / (n)) * (n))
+
+#define MAX(a, b) ((a) > (b) ? (a) : (b))
+
+enum ast_node_type {
+ NODE_PROGRAM,
+ NODE_REGION,
+ NODE_ARRAY,
+ NODE_LENGTH,
+ NODE_PRIMITIVE,
+ NODE_POINTER,
+};
+
+struct ast_node; /* Forward declaration. */
+
+struct ast_program {
+ struct ast_node **members;
+ size_t num_members;
+};
+
+struct ast_region {
+ const char *name;
+ struct ast_node **members;
+ size_t num_members;
+};
+
+struct ast_array {
+ int elem_size;
+ int null_terminated; /* True iff the array should always end with 0. */
+ size_t num_elems;
+};
+
+struct ast_length {
+ size_t byte_width;
+ const char *length_of;
+};
+
+struct ast_primitive {
+ size_t byte_width;
+};
+
+struct ast_pointer {
+ const char *points_to;
+};
+
+struct ast_node {
+ enum ast_node_type type;
+ union {
+ struct ast_program program;
+ struct ast_region region;
+ struct ast_array array;
+ struct ast_length length;
+ struct ast_primitive primitive;
+ struct ast_pointer pointer;
+ } data;
+};
+
+struct parser {
+ struct token **tokens;
+ size_t token_count;
+ size_t curr_token;
+};
+
+int parse(struct token **tokens, size_t token_count, struct ast_node **node_ret);
+void destroy_ast_node(struct ast_node *node);
+
+size_t node_size(struct ast_node *node);
+size_t node_alignment(struct ast_node *node);
+
+#endif /* KFUZZTEST_BRIDGE_INPUT_PARSER_H */
diff --git a/tools/testing/kfuzztest-bridge/rand_stream.c b/tools/testing/kfuzztest-bridge/rand_stream.c
new file mode 100644
index 000000000000..bca6b3de5aad
--- /dev/null
+++ b/tools/testing/kfuzztest-bridge/rand_stream.c
@@ -0,0 +1,77 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Implements a cached file-reader for iterating over a byte stream of
+ * pseudo-random data
+ *
+ * Copyright 2025 Google LLC
+ */
+#include "rand_stream.h"
+
+static int refill(struct rand_stream *rs)
+{
+ rs->valid_bytes = fread(rs->buffer, sizeof(char), rs->buffer_size, rs->source);
+ rs->buffer_pos = 0;
+ if (rs->valid_bytes != rs->buffer_size && ferror(rs->source))
+ return ferror(rs->source);
+ return 0;
+}
+
+struct rand_stream *new_rand_stream(const char *path_to_file, size_t cache_size)
+{
+ struct rand_stream *rs;
+
+ rs = malloc(sizeof(*rs));
+ if (!rs)
+ return NULL;
+
+ rs->valid_bytes = 0;
+ rs->source = fopen(path_to_file, "rb");
+ if (!rs->source) {
+ free(rs);
+ return NULL;
+ }
+
+ if (fseek(rs->source, 0, SEEK_END)) {
+ fclose(rs->source);
+ free(rs);
+ return NULL;
+ }
+ rs->source_size = ftell(rs->source);
+
+ if (fseek(rs->source, 0, SEEK_SET)) {
+ fclose(rs->source);
+ free(rs);
+ return NULL;
+ }
+
+ rs->buffer = malloc(cache_size);
+ if (!rs->buffer) {
+ fclose(rs->source);
+ free(rs);
+ return NULL;
+ }
+ rs->buffer_size = cache_size;
+ return rs;
+}
+
+void destroy_rand_stream(struct rand_stream *rs)
+{
+ fclose(rs->source);
+ free(rs->buffer);
+ free(rs);
+}
+
+int next_byte(struct rand_stream *rs, char *ret)
+{
+ int res;
+
+ if (rs->buffer_pos >= rs->valid_bytes) {
+ res = refill(rs);
+ if (res)
+ return res;
+ if (rs->valid_bytes == 0)
+ return STREAM_EOF;
+ }
+ *ret = rs->buffer[rs->buffer_pos++];
+ return 0;
+}
diff --git a/tools/testing/kfuzztest-bridge/rand_stream.h b/tools/testing/kfuzztest-bridge/rand_stream.h
new file mode 100644
index 000000000000..acb3271d30ca
--- /dev/null
+++ b/tools/testing/kfuzztest-bridge/rand_stream.h
@@ -0,0 +1,57 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Implements a cached file-reader for iterating over a byte stream of
+ * pseudo-random data
+ *
+ * Copyright 2025 Google LLC
+ */
+#ifndef KFUZZTEST_BRIDGE_RAND_STREAM_H
+#define KFUZZTEST_BRIDGE_RAND_STREAM_H
+
+#include <stdlib.h>
+#include <stdio.h>
+
+#define STREAM_EOF 1
+
+/**
+ * struct rand_stream - a buffered bytestream reader
+ *
+ * Reads and returns bytes from a file, using buffered pre-fetching to amortize
+ * the cost of reads.
+ */
+struct rand_stream {
+ FILE *source;
+ size_t source_size;
+ char *buffer;
+ size_t buffer_size;
+ size_t buffer_pos;
+ size_t valid_bytes;
+};
+
+/**
+ * new_rand_stream - return a new struct rand_stream
+ *
+ * @path_to_file: source of the output byte stream.
+ * @cache_size: size of the read-ahead cache in bytes.
+ */
+struct rand_stream *new_rand_stream(const char *path_to_file, size_t cache_size);
+
+/**
+ * destroy_rand_stream - clean up a rand stream's resources
+ *
+ * @rs: a struct rand_stream
+ */
+void destroy_rand_stream(struct rand_stream *rs);
+
+/**
+ * next_byte - return the next byte from a struct rand_stream
+ *
+ * @rs: an initialized struct rand_stream.
+ * @ret: return pointer.
+ *
+ * @return 0 on success or a negative value on failure.
+ *
+ */
+int next_byte(struct rand_stream *rs, char *ret);
+
+#endif /* KFUZZTEST_BRIDGE_RAND_STREAM_H */
--
2.51.0.470.ga7dc726c21-goog
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH v2 04/10] tools: add kfuzztest-bridge utility
2025-09-19 14:57 ` [PATCH v2 04/10] tools: add kfuzztest-bridge utility Ethan Graham
@ 2025-09-19 15:05 ` Alexander Potapenko
0 siblings, 0 replies; 30+ messages in thread
From: Alexander Potapenko @ 2025-09-19 15:05 UTC (permalink / raw)
To: Ethan Graham
Cc: ethangraham, andreyknvl, andy, brauner, brendan.higgins, davem,
davidgow, dhowells, dvyukov, elver, herbert, ignat, jack, jannh,
johannes, kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel,
linux-mm, lukas, rmoar, shuah, sj, tarasmadan
On Fri, Sep 19, 2025 at 4:58 PM Ethan Graham <ethan.w.s.graham@gmail.com> wrote:
>
> From: Ethan Graham <ethangraham@google.com>
>
> Introduce the kfuzztest-bridge tool, a userspace utility for sending
> structured inputs to KFuzzTest harnesses via debugfs.
>
> The bridge takes a textual description of the expected input format, a
> file containing random bytes, and the name of the target fuzz test. It
> parses the description, encodes the random data into the binary format
> expected by the kernel, and writes the result to the corresponding
> debugfs entry.
>
> This allows for both simple manual testing and integration with
> userspace fuzzing engines. For example, it can be used for smoke testing
> by providing data from /dev/urandom, or act as a bridge for blob-based
> fuzzers (e.g., AFL) to target KFuzzTest harnesses.
>
> Signed-off-by: Ethan Graham <ethangraham@google.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH v2 05/10] kfuzztest: add ReST documentation
2025-09-19 14:57 [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework Ethan Graham
` (3 preceding siblings ...)
2025-09-19 14:57 ` [PATCH v2 04/10] tools: add kfuzztest-bridge utility Ethan Graham
@ 2025-09-19 14:57 ` Ethan Graham
2025-09-19 14:57 ` [PATCH v2 06/10] kfuzztest: add KFuzzTest sample fuzz targets Ethan Graham
` (6 subsequent siblings)
11 siblings, 0 replies; 30+ messages in thread
From: Ethan Graham @ 2025-09-19 14:57 UTC (permalink / raw)
To: ethangraham, glider
Cc: andreyknvl, andy, brauner, brendan.higgins, davem, davidgow,
dhowells, dvyukov, elver, herbert, ignat, jack, jannh, johannes,
kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel, linux-mm,
lukas, rmoar, shuah, sj, tarasmadan
From: Ethan Graham <ethangraham@google.com>
Add Documentation/dev-tools/kfuzztest.rst and reference it in the
dev-tools index.
Signed-off-by: Ethan Graham <ethangraham@google.com>
Acked-by: Alexander Potapenko <glider@google.com>
---
PR v2:
- Update documentation to reflect new location of kfuzztest-bridge,
under tools/testing.
PR v1:
- Fix some typos and reword some sections.
- Correct kfuzztest-bridge grammar description.
- Reference documentation in kfuzztest-bridge/input_parser.c header
comment.
RFC v2:
- Add documentation for kfuzztest-bridge tool introduced in patch 4.
---
---
Documentation/dev-tools/index.rst | 1 +
Documentation/dev-tools/kfuzztest.rst | 385 ++++++++++++++++++
tools/testing/kfuzztest-bridge/input_parser.c | 2 +
3 files changed, 388 insertions(+)
create mode 100644 Documentation/dev-tools/kfuzztest.rst
diff --git a/Documentation/dev-tools/index.rst b/Documentation/dev-tools/index.rst
index 65c54b27a60b..00ccc4da003b 100644
--- a/Documentation/dev-tools/index.rst
+++ b/Documentation/dev-tools/index.rst
@@ -32,6 +32,7 @@ Documentation/process/debugging/index.rst
kfence
kselftest
kunit/index
+ kfuzztest
ktap
checkuapi
gpio-sloppy-logic-analyzer
diff --git a/Documentation/dev-tools/kfuzztest.rst b/Documentation/dev-tools/kfuzztest.rst
new file mode 100644
index 000000000000..0c74732ecf21
--- /dev/null
+++ b/Documentation/dev-tools/kfuzztest.rst
@@ -0,0 +1,385 @@
+.. SPDX-License-Identifier: GPL-2.0
+.. Copyright 2025 Google LLC
+
+=========================================
+Kernel Fuzz Testing Framework (KFuzzTest)
+=========================================
+
+Overview
+========
+
+The Kernel Fuzz Testing Framework (KFuzzTest) is a framework designed to expose
+internal kernel functions to a userspace fuzzing engine.
+
+It is intended for testing stateless or low-state functions that are difficult
+to reach from the system call interface, such as routines involved in file
+format parsing or complex data transformations. This provides a method for
+in-situ fuzzing of kernel code without requiring that it be built as a separate
+userspace library or that its dependencies be stubbed out.
+
+The framework consists of four main components:
+
+1. An API, based on the ``FUZZ_TEST`` macro, for defining test targets
+ directly in the kernel tree.
+2. A binary serialization format for passing complex, pointer-rich data
+ structures from userspace to the kernel.
+3. A ``debugfs`` interface through which a userspace fuzzer submits
+ serialized test inputs.
+4. Metadata embedded in dedicated ELF sections of the ``vmlinux`` binary to
+ allow for the discovery of available fuzz targets by external tooling.
+
+.. warning::
+ KFuzzTest is a debugging and testing tool. It exposes internal kernel
+ functions to userspace with minimal sanitization and is designed for
+ use in controlled test environments only. It must **NEVER** be enabled
+ in production kernels.
+
+Supported Architectures
+=======================
+
+KFuzzTest is designed for generic architecture support. It has only been
+explicitly tested on x86_64.
+
+Usage
+=====
+
+To enable KFuzzTest, configure the kernel with::
+
+ CONFIG_KFUZZTEST=y
+
+which depends on ``CONFIG_DEBUGFS`` for receiving userspace inputs, and
+``CONFIG_DEBUG_KERNEL`` as an additional guardrail for preventing KFuzzTest
+from finding its way into a production build accidentally.
+
+The KFuzzTest sample fuzz targets can be built in with
+``CONFIG_SAMPLE_KFUZZTEST``.
+
+KFuzzTest currently only supports targets that are built into the kernel, as the
+core module's startup process discovers fuzz targets from a dedicated ELF
+section during startup. Furthermore, constraints and annotations emit metadata
+that can be scanned from a ``vmlinux`` binary by a userspace fuzzing engine.
+
+Declaring a KFuzzTest target
+----------------------------
+
+A fuzz target should be defined in a .c file. The recommended place to define
+this is under the subsystem's ``/tests`` directory in a ``<file-name>_kfuzz.c``
+file, following the convention used by KUnit. The only strict requirement is
+that the function being fuzzed is visible to the fuzz target.
+
+Defining a fuzz target involves three main parts: defining an input structure,
+writing the test body using the ``FUZZ_TEST`` macro, and optionally adding
+metadata for the fuzzer.
+
+The following example illustrates how to create a fuzz target for a function
+``int process_data(const char *data, size_t len)``.
+
+.. code-block:: c
+
+ /*
+ * 1. Define a struct to model the inputs for the function under test.
+ * Each field corresponds to an argument needed by the function.
+ */
+ struct process_data_inputs {
+ const char *data;
+ size_t len;
+ };
+
+ /*
+ * 2. Define the fuzz target using the FUZZ_TEST macro.
+ * The first parameter is a unique name for the target.
+ * The second parameter is the input struct defined above.
+ */
+ FUZZ_TEST(test_process_data, struct process_data_inputs)
+ {
+ /*
+ * Within this body, the 'arg' variable is a pointer to a
+ * fully initialized 'struct process_data_inputs'.
+ */
+
+ /*
+ * 3. (Optional) Add constraints to define preconditions.
+ * This check ensures 'arg->data' is not NULL. If the condition
+ * is not met, the test exits early. This also creates metadata
+ * to inform the fuzzer.
+ */
+ KFUZZTEST_EXPECT_NOT_NULL(process_data_inputs, data);
+
+ /*
+ * 4. (Optional) Add annotations to provide semantic hints to the
+ * fuzzer. This annotation informs the fuzzer that the 'len' field is
+ * the length of the buffer pointed to by 'data'. Annotations do not
+ * add any runtime checks.
+ */
+ KFUZZTEST_ANNOTATE_LEN(process_data_inputs, len, data);
+
+ /*
+ * 5. Call the kernel function with the provided inputs.
+ * Memory errors like out-of-bounds accesses on 'arg->data' will
+ * be detected by KASAN or other memory error detection tools.
+ */
+ process_data(arg->data, arg->len);
+ }
+
+KFuzzTest provides two families of macros to improve the quality of fuzzing:
+
+- ``KFUZZTEST_EXPECT_*``: These macros define constraints, which are
+ preconditions that must be true for the test to proceed. They are enforced
+ with a runtime check in the kernel. If a check fails, the current test run is
+ aborted. This metadata helps the userspace fuzzer avoid generating invalid
+ inputs.
+
+- ``KFUZZTEST_ANNOTATE_*``: These macros define annotations, which are purely
+ semantic hints for the fuzzer. They do not add any runtime checks and exist
+ only to help the fuzzer generate more intelligent and structurally correct
+ inputs. For example, KFUZZTEST_ANNOTATE_LEN links a size field to a pointer
+ field, which is a common pattern in C APIs.
+
+Metadata
+--------
+
+Macros ``FUZZ_TEST``, ``KFUZZTEST_EXPECT_*`` and ``KFUZZTEST_ANNOTATE_*`` embed
+metadata into several sections within the main ``.data`` section of the final
+``vmlinux`` binary; ``.kfuzztest_target``, ``.kfuzztest_constraint`` and
+``.kfuzztest_annotation`` respectively.
+
+This serves two purposes:
+
+1. The core module uses the ``.kfuzztest_target`` section at boot to discover
+ every ``FUZZ_TEST`` instance and create its ``debugfs`` directory and
+ ``input`` file.
+2. Userspace fuzzers can read this metadata from the ``vmlinux`` binary to
+ discover targets and learn about their rules and structure in order to
+ generate correct and effective inputs.
+
+The metadata in the ``.kfuzztest_*`` sections consists of arrays of fixed-size C
+structs (e.g., ``struct kfuzztest_target``). Fields within these structs that
+are pointers, such as ``name`` or ``arg_type_name``, contain addresses that
+point to other locations in the ``vmlinux`` binary. A userspace tool that
+parsing the ELF file must resolve these pointers to read the data that they
+reference. For example, to get a target's name, a tool must:
+
+1. Read the ``struct kfuzztest_target`` from the ``.kfuzztest_target`` section.
+2. Read the address in the ``.name`` field.
+3. Use that address to locate and read null-terminated string from its position
+ elsewhere in the binary (e.g., ``.rodata``).
+
+Tooling Dependencies
+--------------------
+
+For userspace tools to parse the ``vmlinux`` binary and make use of emitted
+KFuzzTest metadata, the kernel must be compiled with DWARF debug information.
+This is required for tools to understand the layout of C structs, resolve type
+information, and correctly interpret constraints and annotations.
+
+When using KFuzzTest with automated fuzzing tools, either
+``CONFIG_DEBUG_INFO_DWARF4`` or ``CONFIG_DEBUG_INFO_DWARF5`` should be enabled.
+
+Input Format
+============
+
+KFuzzTest targets receive their inputs from userspace via a write to a dedicated
+debugfs file ``/sys/kernel/debug/kfuzztest/<test-name>/input``.
+
+The data written to this file must be a single binary blob that follows a
+specific serialization format. This format is designed to allow complex,
+pointer-rich C structures to be represented in a flat buffer, requiring only a
+single kernel allocation and copy from userspace.
+
+An input is first prefixed by an 8-byte header containing a magic value in the
+first four bytes, defined as ``KFUZZTEST_HEADER_MAGIC`` in
+`<include/linux/kfuzztest.h>``, and a version number in the subsequent four
+bytes.
+
+Version 0
+---------
+
+In version 0 (i.e., when the version number in the 8-byte header is equal to 0),
+the input format consists of three main parts laid out sequentially: a region
+array, a relocation table, and the payload.::
+
+ +----------------+---------------------+-----------+----------------+
+ | region array | relocation table | padding | payload |
+ +----------------+---------------------+-----------+----------------+
+
+Region Array
+^^^^^^^^^^^^
+
+This component is a header that describes how the raw data in the Payload is
+partitioned into logical memory regions. It consists of a count of regions
+followed by an array of ``struct reloc_region``, where each entry defines a
+single region with its size and offset from the start of the payload.
+
+.. code-block:: c
+
+ struct reloc_region {
+ uint32_t offset;
+ uint32_t size;
+ };
+
+ struct reloc_region_array {
+ uint32_t num_regions;
+ struct reloc_region regions[];
+ };
+
+By convention, region 0 represents the top-level input struct that is passed
+as the arg variable to the ``FUZZ_TEST`` body. Subsequent regions typically
+represent data buffers or structs pointed to by fields within that struct.
+Region array entries must be ordered by ascending offset, and must not overlap
+with one another.
+
+Relocation Table
+^^^^^^^^^^^^^^^^
+
+The relocation table contains the instructions for the kernel to "hydrate" the
+payload by patching pointer fields. It contains an array of
+``struct reloc_entry`` items. Each entry acts as a linking instruction,
+specifying:
+
+- The location of a pointer that needs to be patched (identified by a region
+ ID and an offset within that region).
+
+- The target region that the pointer should point to (identified by the
+ target's region ID) or ``KFUZZTEST_REGIONID_NULL`` if the pointer is ``NULL``.
+
+This table also specifies the amount of padding between its end and the start
+of the payload, which should be at least 8 bytes.
+
+.. code-block:: c
+
+ struct reloc_entry {
+ uint32_t region_id;
+ uint32_t region_offset;
+ uint32_t value;
+ };
+
+ struct reloc_table {
+ uint32_t num_entries;
+ uint32_t padding_size;
+ struct reloc_entry entries[];
+ };
+
+Payload
+^^^^^^^
+
+The payload contains the raw binary data for all regions, concatenated together
+according to their specified offsets.
+
+- Region specific alignment: The data for each individual region must start at
+ an offset that is aligned to its own C type's requirements. For example, a
+ ``uint64_t`` must begin on an 8-byte boundary.
+
+- Minimum alignment: The offset of each region, as well as the beginning of the
+ payload, must also be a multiple of the overall minimum alignment value. This
+ value is determined by the greater of ``ARCH_KMALLOC_MINALIGN`` and
+ ``KASAN_GRANULE_SIZE`` (which is represented by ``KFUZZTEST_POISON_SIZE`` in
+ ``/include/linux/kfuzztest.h``). This minimum alignment ensures that all
+ function inputs respect C calling conventions.
+
+- Padding: The space between the end of one region's data and the beginning of
+ the next must be sufficient for padding. The padding must also be at least
+ the same minimum alignment value mentioned above. This is crucial for KASAN
+ builds, as it allows KFuzzTest to poison this unused space enabling precise
+ detection of out-of-bounds memory accesses between adjacent buffers.
+
+The minimum alignment value is architecture-dependent and is exposed to
+userspace via the read-only file
+``/sys/kernel/debug/kfuzztest/_config/minalign``. The framework relies on
+userspace tooling to construct the payload correctly, adhering to all three of
+these rules for every region.
+
+KFuzzTest Bridge Tool
+=====================
+
+The ``kfuzztest-bridge`` program is a userspace utility that encodes a random
+byte stream into the structured binary format expected by a KFuzzTest harness.
+It allows users to describe the target's input structure textually, making it
+easy to perform smoke tests or connect harnesses to blob-based fuzzing engines.
+
+This tool is intended to be simple, both in usage and implementation. Its
+structure and DSL are sufficient for simpler use-cases. For more advanced
+coverage-guided fuzzing it is recommended to use
+`syzkaller <https://github.com/google/syzkaller>` which implements deeper
+support for KFuzzTest targets.
+
+Usage
+-----
+
+The tool can be built with ``make tools/testing/kfuzztest-bridge``. In the case
+of libc incompatibilities, the tool will have to be linked statically or built
+on the target system.
+
+Example:
+
+.. code-block:: sh
+
+ ./tools/testing/kfuzztest-bridge \
+ "foo { u32 ptr[bar] }; bar { ptr[data] len[data, u64]}; data { arr[u8, 42] };" \
+ "my-fuzz-target" /dev/urandom
+
+The command takes three arguments
+
+1. A string describing the input structure (see `Textual Format`_ sub-section).
+2. The name of the target test, which corresponds to its directory in
+ ``/sys/kernel/debug/kfuzztest/``.
+3. A path to a file providing a stream of random data, such as
+ ``/dev/urandom``.
+
+The structure string in the example corresponds to the following C data
+structures:
+
+.. code-block:: c
+
+ struct foo {
+ u32 a;
+ struct bar *b;
+ };
+
+ struct bar {
+ struct data *d;
+ u64 data_len; /* Equals 42. */
+ };
+
+ struct data {
+ char arr[42];
+ };
+
+Textual Format
+--------------
+
+The textual format is a human-readable representation of the region-based binary
+format used by KFuzzTest. It is described by the following grammar:
+
+.. code-block:: text
+
+ schema ::= region ( ";" region )* [";"]
+ region ::= identifier "{" type ( " " type )* "}"
+ type ::= primitive | pointer | array | length | string
+ primitive ::= "u8" | "u16" | "u32" | "u64"
+ pointer ::= "ptr" "[" identifier "]"
+ array ::= "arr" "[" primitive "," integer "]"
+ length ::= "len" "[" identifier "," primitive "]"
+ string ::= "str" "[" integer "]"
+ identifier ::= [a-zA-Z_][a-zA-Z1-9_]*
+ integer ::= [0-9]+
+
+Pointers must reference a named region.
+
+To fuzz a raw buffer, the buffer must be defined in its own region, as shown
+below:
+
+.. code-block:: c
+
+ struct my_struct {
+ char *buf;
+ size_t buflen;
+ };
+
+This would correspond to the following textual description:
+
+.. code-block:: text
+
+ my_struct { ptr[buf] len[buf, u64] }; buf { arr[u8, n] };
+
+Here, ``n`` is some integer value defining the size of the byte array inside of
+the ``buf`` region.
diff --git a/tools/testing/kfuzztest-bridge/input_parser.c b/tools/testing/kfuzztest-bridge/input_parser.c
index b1fd8ba5217e..feaa59de49d7 100644
--- a/tools/testing/kfuzztest-bridge/input_parser.c
+++ b/tools/testing/kfuzztest-bridge/input_parser.c
@@ -16,6 +16,8 @@
* and its corresponding length encoded over 8 bytes, where `buf` itself
* contains a 42-byte array.
*
+ * The full grammar is documented in Documentation/dev-tools/kfuzztest.rst.
+ *
* Copyright 2025 Google LLC
*/
#include <errno.h>
--
2.51.0.470.ga7dc726c21-goog
^ permalink raw reply [flat|nested] 30+ messages in thread* [PATCH v2 06/10] kfuzztest: add KFuzzTest sample fuzz targets
2025-09-19 14:57 [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework Ethan Graham
` (4 preceding siblings ...)
2025-09-19 14:57 ` [PATCH v2 05/10] kfuzztest: add ReST documentation Ethan Graham
@ 2025-09-19 14:57 ` Ethan Graham
2025-09-19 14:57 ` [PATCH v2 07/10] crypto: implement KFuzzTest targets for PKCS7 and RSA parsing Ethan Graham
` (5 subsequent siblings)
11 siblings, 0 replies; 30+ messages in thread
From: Ethan Graham @ 2025-09-19 14:57 UTC (permalink / raw)
To: ethangraham, glider
Cc: andreyknvl, andy, brauner, brendan.higgins, davem, davidgow,
dhowells, dvyukov, elver, herbert, ignat, jack, jannh, johannes,
kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel, linux-mm,
lukas, rmoar, shuah, sj, tarasmadan
From: Ethan Graham <ethangraham@google.com>
Add two simple fuzz target samples to demonstrate the KFuzzTest API and
provide basic self-tests for the framework.
These examples showcase how a developer can define a fuzz target using
the FUZZ_TEST(), constraint, and annotation macros, and serve as runtime
sanity checks for the core logic. For example, they test that
out-of-bounds memory accesses into poisoned padding regions are
correctly detected in a KASAN build.
These have been tested by writing syzkaller-generated inputs into their
debugfs 'input' files and verifying that the correct KASAN reports were
triggered.
Signed-off-by: Ethan Graham <ethangraham@google.com>
Acked-by: Alexander Potapenko <glider@google.com>
---
PR v2:
- Fix build issues pointed out by the kernel test robot <lkp@intel.com>.
---
---
samples/Kconfig | 7 ++
samples/Makefile | 1 +
samples/kfuzztest/Makefile | 3 +
samples/kfuzztest/overflow_on_nested_buffer.c | 71 +++++++++++++++++++
samples/kfuzztest/underflow_on_buffer.c | 59 +++++++++++++++
5 files changed, 141 insertions(+)
create mode 100644 samples/kfuzztest/Makefile
create mode 100644 samples/kfuzztest/overflow_on_nested_buffer.c
create mode 100644 samples/kfuzztest/underflow_on_buffer.c
diff --git a/samples/Kconfig b/samples/Kconfig
index 6e072a5f1ed8..5209dd9d7a5c 100644
--- a/samples/Kconfig
+++ b/samples/Kconfig
@@ -320,6 +320,13 @@ config SAMPLE_HUNG_TASK
Reading these files with multiple processes triggers hung task
detection by holding locks for a long time (256 seconds).
+config SAMPLE_KFUZZTEST
+ bool "Build KFuzzTest sample targets"
+ depends on KFUZZTEST
+ help
+ Build KFuzzTest sample targets that serve as selftests for input
+ deserialization and inter-region redzone poisoning logic.
+
source "samples/rust/Kconfig"
source "samples/damon/Kconfig"
diff --git a/samples/Makefile b/samples/Makefile
index 07641e177bd8..3a0e7f744f44 100644
--- a/samples/Makefile
+++ b/samples/Makefile
@@ -44,4 +44,5 @@ obj-$(CONFIG_SAMPLE_DAMON_WSSE) += damon/
obj-$(CONFIG_SAMPLE_DAMON_PRCL) += damon/
obj-$(CONFIG_SAMPLE_DAMON_MTIER) += damon/
obj-$(CONFIG_SAMPLE_HUNG_TASK) += hung_task/
+obj-$(CONFIG_SAMPLE_KFUZZTEST) += kfuzztest/
obj-$(CONFIG_SAMPLE_TSM_MR) += tsm-mr/
diff --git a/samples/kfuzztest/Makefile b/samples/kfuzztest/Makefile
new file mode 100644
index 000000000000..4f8709876c9e
--- /dev/null
+++ b/samples/kfuzztest/Makefile
@@ -0,0 +1,3 @@
+# SPDX-License-Identifier: GPL-2.0-only
+
+obj-$(CONFIG_SAMPLE_KFUZZTEST) += overflow_on_nested_buffer.o underflow_on_buffer.o
diff --git a/samples/kfuzztest/overflow_on_nested_buffer.c b/samples/kfuzztest/overflow_on_nested_buffer.c
new file mode 100644
index 000000000000..2f1c3ff9f750
--- /dev/null
+++ b/samples/kfuzztest/overflow_on_nested_buffer.c
@@ -0,0 +1,71 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * This file contains a KFuzzTest example target that ensures that a buffer
+ * overflow on a nested region triggers a KASAN OOB access report.
+ *
+ * Copyright 2025 Google LLC
+ */
+
+/**
+ * DOC: test_overflow_on_nested_buffer
+ *
+ * This test uses a struct with two distinct dynamically allocated buffers.
+ * It checks that KFuzzTest's memory layout correctly poisons the memory
+ * regions and that KASAN can detect an overflow when reading one byte past the
+ * end of the first buffer (`a`).
+ *
+ * It can be invoked with kfuzztest-bridge using the following command:
+ *
+ * ./kfuzztest-bridge \
+ * "nested_buffers { ptr[a] len[a, u64] ptr[b] len[b, u64] }; \
+ * a { arr[u8, 64] }; b { arr[u8, 64] };" \
+ * "test_overflow_on_nested_buffer" /dev/urandom
+ *
+ * The first argument describes the C struct `nested_buffers` and specifies that
+ * both `a` and `b` are pointers to arrays of 64 bytes.
+ */
+#include <linux/kfuzztest.h>
+
+static void overflow_on_nested_buffer(const char *a, size_t a_len, const char *b, size_t b_len)
+{
+ size_t i;
+ pr_info("a = [%px, %px)", a, a + a_len);
+ pr_info("b = [%px, %px)", b, b + b_len);
+
+ /* Ensure that all bytes in arg->b are accessible. */
+ for (i = 0; i < b_len; i++)
+ READ_ONCE(b[i]);
+ /*
+ * Check that all bytes in arg->a are accessible, and provoke an OOB on
+ * the first byte to the right of the buffer which will trigger a KASAN
+ * report.
+ */
+ for (i = 0; i <= a_len; i++)
+ READ_ONCE(a[i]);
+}
+
+struct nested_buffers {
+ const char *a;
+ size_t a_len;
+ const char *b;
+ size_t b_len;
+};
+
+/**
+ * The KFuzzTest input format specifies that struct nested buffers should
+ * be expanded as:
+ *
+ * | a | b | pad[8] | *a | pad[8] | *b |
+ *
+ * where the padded regions are poisoned. We expect to trigger a KASAN report by
+ * overflowing one byte into the `a` buffer.
+ */
+FUZZ_TEST(test_overflow_on_nested_buffer, struct nested_buffers)
+{
+ KFUZZTEST_EXPECT_NOT_NULL(nested_buffers, a);
+ KFUZZTEST_EXPECT_NOT_NULL(nested_buffers, b);
+ KFUZZTEST_ANNOTATE_LEN(nested_buffers, a_len, a);
+ KFUZZTEST_ANNOTATE_LEN(nested_buffers, b_len, b);
+
+ overflow_on_nested_buffer(arg->a, arg->a_len, arg->b, arg->b_len);
+}
diff --git a/samples/kfuzztest/underflow_on_buffer.c b/samples/kfuzztest/underflow_on_buffer.c
new file mode 100644
index 000000000000..02704a1bfebb
--- /dev/null
+++ b/samples/kfuzztest/underflow_on_buffer.c
@@ -0,0 +1,59 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * This file contains a KFuzzTest example target that ensures that a buffer
+ * underflow on a region triggers a KASAN OOB access report.
+ *
+ * Copyright 2025 Google LLC
+ */
+
+/**
+ * DOC: test_underflow_on_buffer
+ *
+ * This test ensures that the region between the metadata struct and the
+ * dynamically allocated buffer is poisoned. It provokes a one-byte underflow
+ * on the buffer, which should be caught by KASAN.
+ *
+ * It can be invoked with kfuzztest-bridge using the following command:
+ *
+ * ./kfuzztest-bridge \
+ * "some_buffer { ptr[buf] len[buf, u64]}; buf { arr[u8, 128] };" \
+ * "test_underflow_on_buffer" /dev/urandom
+ *
+ * The first argument describes the C struct `some_buffer` and specifies that
+ * `buf` is a pointer to an array of 128 bytes. The second argument is the test
+ * name, and the third is a seed file.
+ */
+#include <linux/kfuzztest.h>
+
+static void underflow_on_buffer(char *buf, size_t buflen)
+{
+ size_t i;
+
+ pr_info("buf = [%px, %px)", buf, buf + buflen);
+
+ /* First ensure that all bytes in arg->b are accessible. */
+ for (i = 0; i < buflen; i++)
+ READ_ONCE(buf[i]);
+ /*
+ * Provoke a buffer overflow on the first byte preceding b, triggering
+ * a KASAN report.
+ */
+ READ_ONCE(*((char *)buf - 1));
+}
+
+struct some_buffer {
+ char *buf;
+ size_t buflen;
+};
+
+/**
+ * Tests that the region between struct some_buffer and the expanded *buf field
+ * is correctly poisoned by accessing the first byte before *buf.
+ */
+FUZZ_TEST(test_underflow_on_buffer, struct some_buffer)
+{
+ KFUZZTEST_EXPECT_NOT_NULL(some_buffer, buf);
+ KFUZZTEST_ANNOTATE_LEN(some_buffer, buflen, buf);
+
+ underflow_on_buffer(arg->buf, arg->buflen);
+}
--
2.51.0.470.ga7dc726c21-goog
^ permalink raw reply [flat|nested] 30+ messages in thread* [PATCH v2 07/10] crypto: implement KFuzzTest targets for PKCS7 and RSA parsing
2025-09-19 14:57 [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework Ethan Graham
` (5 preceding siblings ...)
2025-09-19 14:57 ` [PATCH v2 06/10] kfuzztest: add KFuzzTest sample fuzz targets Ethan Graham
@ 2025-09-19 14:57 ` Ethan Graham
2025-09-19 14:57 ` [PATCH v2 08/10] drivers/auxdisplay: add a KFuzzTest for parse_xy() Ethan Graham
` (4 subsequent siblings)
11 siblings, 0 replies; 30+ messages in thread
From: Ethan Graham @ 2025-09-19 14:57 UTC (permalink / raw)
To: ethangraham, glider
Cc: andreyknvl, andy, brauner, brendan.higgins, davem, davidgow,
dhowells, dvyukov, elver, herbert, ignat, jack, jannh, johannes,
kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel, linux-mm,
lukas, rmoar, shuah, sj, tarasmadan
From: Ethan Graham <ethangraham@google.com>
Add KFuzzTest targets for pkcs7_parse_message, rsa_parse_pub_key, and
rsa_parse_priv_key to serve as real-world examples of how the framework
is used.
These functions are ideal candidates for KFuzzTest as they perform
complex parsing of user-controlled data but are not directly exposed at
the syscall boundary. This makes them difficult to exercise with
traditional fuzzing tools and showcases the primary strength of the
KFuzzTest framework: providing an interface to fuzz internal functions.
To validate the effectiveness of the framework on these new targets, we
injected two artificial bugs and let syzkaller fuzz the targets in an
attempt to catch them.
The first of these was calling the asn1 decoder with an incorrect input
from pkcs7_parse_message, like so:
- ret = asn1_ber_decoder(&pkcs7_decoder, ctx, data, datalen);
+ ret = asn1_ber_decoder(&pkcs7_decoder, ctx, data, datalen + 1);
The second was bug deeper inside of asn1_ber_decoder itself, like so:
- for (len = 0; n > 0; n--)
+ for (len = 0; n >= 0; n--)
syzkaller was able to trigger these bugs, and the associated KASAN
slab-out-of-bounds reports, within seconds.
The targets are defined within crypto/asymmetric-keys/tests.
Signed-off-by: Ethan Graham <ethangraham@google.com>
Reviewed-by: Ignat Korchagin <ignat@cloudflare.com>
---
PR v2:
- Make fuzz targets also depend on the KConfig options needed for the
functions they are fuzzing, CONFIG_PKCS7_MESSAGE_PARSER and
CONFIG_CRYPTO_RSA respectively.
- Fix build issues pointed out by the kernel test robot <lkp@intel.com>.
- Account for return value of pkcs7_parse_message, and free resources if
the function call succeeds.
PR v1:
- Change the fuzz target build to depend on CONFIG_KFUZZTEST=y,
eliminating the need for a separate config option for each individual
file as suggested by Ignat Korchagin.
- Remove KFUZZTEST_EXPECT_LE on the length of the `key` field inside of
the fuzz targets. A maximum length is now set inside of the core input
parsing logic.
RFC v2:
- Move KFuzzTest targets outside of the source files into dedicated
_kfuzz.c files under /crypto/asymmetric_keys/tests/ as suggested by
Ignat Korchagin and Eric Biggers.
---
---
crypto/asymmetric_keys/Makefile | 2 +
crypto/asymmetric_keys/tests/Makefile | 4 ++
crypto/asymmetric_keys/tests/pkcs7_kfuzz.c | 26 +++++++++++++
.../asymmetric_keys/tests/rsa_helper_kfuzz.c | 38 +++++++++++++++++++
4 files changed, 70 insertions(+)
create mode 100644 crypto/asymmetric_keys/tests/Makefile
create mode 100644 crypto/asymmetric_keys/tests/pkcs7_kfuzz.c
create mode 100644 crypto/asymmetric_keys/tests/rsa_helper_kfuzz.c
diff --git a/crypto/asymmetric_keys/Makefile b/crypto/asymmetric_keys/Makefile
index bc65d3b98dcb..77b825aee6b2 100644
--- a/crypto/asymmetric_keys/Makefile
+++ b/crypto/asymmetric_keys/Makefile
@@ -67,6 +67,8 @@ obj-$(CONFIG_PKCS7_TEST_KEY) += pkcs7_test_key.o
pkcs7_test_key-y := \
pkcs7_key_type.o
+obj-y += tests/
+
#
# Signed PE binary-wrapped key handling
#
diff --git a/crypto/asymmetric_keys/tests/Makefile b/crypto/asymmetric_keys/tests/Makefile
new file mode 100644
index 000000000000..023d6a65fb89
--- /dev/null
+++ b/crypto/asymmetric_keys/tests/Makefile
@@ -0,0 +1,4 @@
+pkcs7-kfuzz-y := $(and $(CONFIG_KFUZZTEST),$(CONFIG_PKCS7_MESSAGE_PARSER))
+rsa-helper-kfuzz-y := $(and $(CONFIG_KFUZZTEST),$(CONFIG_CRYPTO_RSA))
+obj-$(pkcs7-kfuzz-y) += pkcs7_kfuzz.o
+obj-$(rsa-helper-kfuzz-y) += rsa_helper_kfuzz.o
diff --git a/crypto/asymmetric_keys/tests/pkcs7_kfuzz.c b/crypto/asymmetric_keys/tests/pkcs7_kfuzz.c
new file mode 100644
index 000000000000..c801f6b59de2
--- /dev/null
+++ b/crypto/asymmetric_keys/tests/pkcs7_kfuzz.c
@@ -0,0 +1,26 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * PKCS#7 parser KFuzzTest target
+ *
+ * Copyright 2025 Google LLC
+ */
+#include <crypto/pkcs7.h>
+#include <linux/kfuzztest.h>
+
+struct pkcs7_parse_message_arg {
+ const void *data;
+ size_t datalen;
+};
+
+FUZZ_TEST(test_pkcs7_parse_message, struct pkcs7_parse_message_arg)
+{
+ struct pkcs7_message *msg;
+
+ KFUZZTEST_EXPECT_NOT_NULL(pkcs7_parse_message_arg, data);
+ KFUZZTEST_ANNOTATE_ARRAY(pkcs7_parse_message_arg, data);
+ KFUZZTEST_ANNOTATE_LEN(pkcs7_parse_message_arg, datalen, data);
+
+ msg = pkcs7_parse_message(arg->data, arg->datalen);
+ if (msg && !IS_ERR(msg))
+ kfree(msg);
+}
diff --git a/crypto/asymmetric_keys/tests/rsa_helper_kfuzz.c b/crypto/asymmetric_keys/tests/rsa_helper_kfuzz.c
new file mode 100644
index 000000000000..bd29ed5e8c82
--- /dev/null
+++ b/crypto/asymmetric_keys/tests/rsa_helper_kfuzz.c
@@ -0,0 +1,38 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * RSA key extract helper KFuzzTest targets
+ *
+ * Copyright 2025 Google LLC
+ */
+#include <linux/kfuzztest.h>
+#include <crypto/internal/rsa.h>
+
+struct rsa_parse_pub_key_arg {
+ const void *key;
+ size_t key_len;
+};
+
+FUZZ_TEST(test_rsa_parse_pub_key, struct rsa_parse_pub_key_arg)
+{
+ KFUZZTEST_EXPECT_NOT_NULL(rsa_parse_pub_key_arg, key);
+ KFUZZTEST_ANNOTATE_ARRAY(rsa_parse_pub_key_arg, key);
+ KFUZZTEST_ANNOTATE_LEN(rsa_parse_pub_key_arg, key_len, key);
+
+ struct rsa_key out;
+ rsa_parse_pub_key(&out, arg->key, arg->key_len);
+}
+
+struct rsa_parse_priv_key_arg {
+ const void *key;
+ size_t key_len;
+};
+
+FUZZ_TEST(test_rsa_parse_priv_key, struct rsa_parse_priv_key_arg)
+{
+ KFUZZTEST_EXPECT_NOT_NULL(rsa_parse_priv_key_arg, key);
+ KFUZZTEST_ANNOTATE_ARRAY(rsa_parse_priv_key_arg, key);
+ KFUZZTEST_ANNOTATE_LEN(rsa_parse_priv_key_arg, key_len, key);
+
+ struct rsa_key out;
+ rsa_parse_priv_key(&out, arg->key, arg->key_len);
+}
--
2.51.0.470.ga7dc726c21-goog
^ permalink raw reply [flat|nested] 30+ messages in thread* [PATCH v2 08/10] drivers/auxdisplay: add a KFuzzTest for parse_xy()
2025-09-19 14:57 [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework Ethan Graham
` (6 preceding siblings ...)
2025-09-19 14:57 ` [PATCH v2 07/10] crypto: implement KFuzzTest targets for PKCS7 and RSA parsing Ethan Graham
@ 2025-09-19 14:57 ` Ethan Graham
2025-09-19 15:07 ` Alexander Potapenko
` (2 more replies)
2025-09-19 14:57 ` [PATCH v2 09/10] fs/binfmt_script: add KFuzzTest target for load_script Ethan Graham
` (3 subsequent siblings)
11 siblings, 3 replies; 30+ messages in thread
From: Ethan Graham @ 2025-09-19 14:57 UTC (permalink / raw)
To: ethangraham, glider
Cc: andreyknvl, andy, brauner, brendan.higgins, davem, davidgow,
dhowells, dvyukov, elver, herbert, ignat, jack, jannh, johannes,
kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel, linux-mm,
lukas, rmoar, shuah, sj, tarasmadan
From: Ethan Graham <ethangraham@google.com>
Add a KFuzzTest fuzzer for the parse_xy() function, located in a new
file under /drivers/auxdisplay/tests.
To validate the correctness and effectiveness of this KFuzzTest target,
a bug was injected into parse_xy() like so:
drivers/auxdisplay/charlcd.c:179
- s = p;
+ s = p + 1;
Although a simple off-by-one bug, it requires a specific input sequence
in order to trigger it, thus demonstrating the power of pairing
KFuzzTest with a coverage-guided fuzzer like syzkaller.
Signed-off-by: Ethan Graham <ethangraham@google.com>
---
drivers/auxdisplay/charlcd.c | 8 ++++++++
drivers/auxdisplay/tests/charlcd_kfuzz.c | 20 ++++++++++++++++++++
2 files changed, 28 insertions(+)
create mode 100644 drivers/auxdisplay/tests/charlcd_kfuzz.c
diff --git a/drivers/auxdisplay/charlcd.c b/drivers/auxdisplay/charlcd.c
index 09020bb8ad15..e079b5a9c93c 100644
--- a/drivers/auxdisplay/charlcd.c
+++ b/drivers/auxdisplay/charlcd.c
@@ -682,3 +682,11 @@ EXPORT_SYMBOL_GPL(charlcd_unregister);
MODULE_DESCRIPTION("Character LCD core support");
MODULE_LICENSE("GPL");
+
+/*
+ * When CONFIG_KFUZZTEST is enabled, we include this _kfuzz.c file to ensure
+ * that KFuzzTest targets are built.
+ */
+#ifdef CONFIG_KFUZZTEST
+#include "tests/charlcd_kfuzz.c"
+#endif /* CONFIG_KFUZZTEST */
diff --git a/drivers/auxdisplay/tests/charlcd_kfuzz.c b/drivers/auxdisplay/tests/charlcd_kfuzz.c
new file mode 100644
index 000000000000..28ce7069c65c
--- /dev/null
+++ b/drivers/auxdisplay/tests/charlcd_kfuzz.c
@@ -0,0 +1,20 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * charlcd KFuzzTest target
+ *
+ * Copyright 2025 Google LLC
+ */
+#include <linux/kfuzztest.h>
+
+struct parse_xy_arg {
+ const char *s;
+};
+
+FUZZ_TEST(test_parse_xy, struct parse_xy_arg)
+{
+ unsigned long x, y;
+
+ KFUZZTEST_EXPECT_NOT_NULL(parse_xy_arg, s);
+ KFUZZTEST_ANNOTATE_STRING(parse_xy_arg, s);
+ parse_xy(arg->s, &x, &y);
+}
--
2.51.0.470.ga7dc726c21-goog
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH v2 08/10] drivers/auxdisplay: add a KFuzzTest for parse_xy()
2025-09-19 14:57 ` [PATCH v2 08/10] drivers/auxdisplay: add a KFuzzTest for parse_xy() Ethan Graham
@ 2025-09-19 15:07 ` Alexander Potapenko
2025-09-20 10:53 ` Andy Shevchenko
2025-09-24 9:28 ` kernel test robot
2 siblings, 0 replies; 30+ messages in thread
From: Alexander Potapenko @ 2025-09-19 15:07 UTC (permalink / raw)
To: Ethan Graham
Cc: ethangraham, andreyknvl, andy, brauner, brendan.higgins, davem,
davidgow, dhowells, dvyukov, elver, herbert, ignat, jack, jannh,
johannes, kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel,
linux-mm, lukas, rmoar, shuah, sj, tarasmadan
On Fri, Sep 19, 2025 at 4:58 PM Ethan Graham <ethan.w.s.graham@gmail.com> wrote:
>
> From: Ethan Graham <ethangraham@google.com>
>
> Add a KFuzzTest fuzzer for the parse_xy() function, located in a new
> file under /drivers/auxdisplay/tests.
>
> To validate the correctness and effectiveness of this KFuzzTest target,
> a bug was injected into parse_xy() like so:
>
> drivers/auxdisplay/charlcd.c:179
> - s = p;
> + s = p + 1;
>
> Although a simple off-by-one bug, it requires a specific input sequence
> in order to trigger it, thus demonstrating the power of pairing
> KFuzzTest with a coverage-guided fuzzer like syzkaller.
>
> Signed-off-by: Ethan Graham <ethangraham@google.com>
Acked-by: Alexander Potapenko <glider@google.com>
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH v2 08/10] drivers/auxdisplay: add a KFuzzTest for parse_xy()
2025-09-19 14:57 ` [PATCH v2 08/10] drivers/auxdisplay: add a KFuzzTest for parse_xy() Ethan Graham
2025-09-19 15:07 ` Alexander Potapenko
@ 2025-09-20 10:53 ` Andy Shevchenko
2025-09-20 12:08 ` Alexander Potapenko
2025-09-24 9:28 ` kernel test robot
2 siblings, 1 reply; 30+ messages in thread
From: Andy Shevchenko @ 2025-09-20 10:53 UTC (permalink / raw)
To: Ethan Graham
Cc: ethangraham, glider, andreyknvl, andy, brauner, brendan.higgins,
davem, davidgow, dhowells, dvyukov, elver, herbert, ignat, jack,
jannh, johannes, kasan-dev, kees, kunit-dev, linux-crypto,
linux-kernel, linux-mm, lukas, rmoar, shuah, sj, tarasmadan
On Fri, Sep 19, 2025 at 5:58 PM Ethan Graham <ethan.w.s.graham@gmail.com> wrote:
>
> From: Ethan Graham <ethangraham@google.com>
>
> Add a KFuzzTest fuzzer for the parse_xy() function, located in a new
> file under /drivers/auxdisplay/tests.
>
> To validate the correctness and effectiveness of this KFuzzTest target,
> a bug was injected into parse_xy() like so:
>
> drivers/auxdisplay/charlcd.c:179
> - s = p;
> + s = p + 1;
>
> Although a simple off-by-one bug, it requires a specific input sequence
> in order to trigger it, thus demonstrating the power of pairing
> KFuzzTest with a coverage-guided fuzzer like syzkaller.
...
> --- a/drivers/auxdisplay/charlcd.c
> +++ b/drivers/auxdisplay/charlcd.c
> @@ -682,3 +682,11 @@ EXPORT_SYMBOL_GPL(charlcd_unregister);
>
> MODULE_DESCRIPTION("Character LCD core support");
> MODULE_LICENSE("GPL");
> +
> +/*
> + * When CONFIG_KFUZZTEST is enabled, we include this _kfuzz.c file to ensure
> + * that KFuzzTest targets are built.
> + */
> +#ifdef CONFIG_KFUZZTEST
> +#include "tests/charlcd_kfuzz.c"
> +#endif /* CONFIG_KFUZZTEST */
No, NAK. We don't want to see these in each and every module. Please,
make sure that nothing, except maybe Kconfig, is modified in this
folder (yet, you may add a _separate_ test module, as you already have
done in this patch).
--
With Best Regards,
Andy Shevchenko
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH v2 08/10] drivers/auxdisplay: add a KFuzzTest for parse_xy()
2025-09-20 10:53 ` Andy Shevchenko
@ 2025-09-20 12:08 ` Alexander Potapenko
2025-09-20 12:47 ` Lukas Wunner
0 siblings, 1 reply; 30+ messages in thread
From: Alexander Potapenko @ 2025-09-20 12:08 UTC (permalink / raw)
To: Andy Shevchenko
Cc: Ethan Graham, ethangraham, andreyknvl, andy, brauner,
brendan.higgins, davem, davidgow, dhowells, dvyukov, elver,
herbert, ignat, jack, jannh, johannes, kasan-dev, kees,
kunit-dev, linux-crypto, linux-kernel, linux-mm, lukas, rmoar,
shuah, sj, tarasmadan
On Sat, Sep 20, 2025 at 12:54 PM Andy Shevchenko
<andy.shevchenko@gmail.com> wrote:
>
> On Fri, Sep 19, 2025 at 5:58 PM Ethan Graham <ethan.w.s.graham@gmail.com> wrote:
> >
> > From: Ethan Graham <ethangraham@google.com>
> >
> > Add a KFuzzTest fuzzer for the parse_xy() function, located in a new
> > file under /drivers/auxdisplay/tests.
> >
> > To validate the correctness and effectiveness of this KFuzzTest target,
> > a bug was injected into parse_xy() like so:
> >
> > drivers/auxdisplay/charlcd.c:179
> > - s = p;
> > + s = p + 1;
> >
> > Although a simple off-by-one bug, it requires a specific input sequence
> > in order to trigger it, thus demonstrating the power of pairing
> > KFuzzTest with a coverage-guided fuzzer like syzkaller.
>
> ...
>
> > --- a/drivers/auxdisplay/charlcd.c
> > +++ b/drivers/auxdisplay/charlcd.c
> > @@ -682,3 +682,11 @@ EXPORT_SYMBOL_GPL(charlcd_unregister);
> >
> > MODULE_DESCRIPTION("Character LCD core support");
> > MODULE_LICENSE("GPL");
> > +
> > +/*
> > + * When CONFIG_KFUZZTEST is enabled, we include this _kfuzz.c file to ensure
> > + * that KFuzzTest targets are built.
> > + */
> > +#ifdef CONFIG_KFUZZTEST
> > +#include "tests/charlcd_kfuzz.c"
> > +#endif /* CONFIG_KFUZZTEST */
>
> No, NAK. We don't want to see these in each and every module. Please,
> make sure that nothing, except maybe Kconfig, is modified in this
> folder (yet, you may add a _separate_ test module, as you already have
> done in this patch).
This is one of the cases in which we can't go without changing the
original code, because parse_xy() is a static function.
Including the test into the source is not the only option, we could as
well make the function visible unconditionally, or introduce a macro
similar to VISIBLE_IF_KUNIT.
Do you prefer any of those?
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH v2 08/10] drivers/auxdisplay: add a KFuzzTest for parse_xy()
2025-09-20 12:08 ` Alexander Potapenko
@ 2025-09-20 12:47 ` Lukas Wunner
2025-09-21 18:25 ` Andy Shevchenko
0 siblings, 1 reply; 30+ messages in thread
From: Lukas Wunner @ 2025-09-20 12:47 UTC (permalink / raw)
To: Alexander Potapenko
Cc: Andy Shevchenko, Ethan Graham, ethangraham, andreyknvl, andy,
brauner, brendan.higgins, davem, davidgow, dhowells, dvyukov,
elver, herbert, ignat, jack, jannh, johannes, kasan-dev, kees,
kunit-dev, linux-crypto, linux-kernel, linux-mm, rmoar, shuah,
sj, tarasmadan
On Sat, Sep 20, 2025 at 02:08:01PM +0200, Alexander Potapenko wrote:
> On Sat, Sep 20, 2025 at 12:54 PM Andy Shevchenko <andy.shevchenko@gmail.com> wrote:
> > On Fri, Sep 19, 2025 at 5:58 PM Ethan Graham <ethan.w.s.graham@gmail.com> wrote:
> > > +++ b/drivers/auxdisplay/charlcd.c
> > > @@ -682,3 +682,11 @@ EXPORT_SYMBOL_GPL(charlcd_unregister);
> > >
> > > MODULE_DESCRIPTION("Character LCD core support");
> > > MODULE_LICENSE("GPL");
> > > +
> > > +/*
> > > + * When CONFIG_KFUZZTEST is enabled, we include this _kfuzz.c file to ensure
> > > + * that KFuzzTest targets are built.
> > > + */
> > > +#ifdef CONFIG_KFUZZTEST
> > > +#include "tests/charlcd_kfuzz.c"
> > > +#endif /* CONFIG_KFUZZTEST */
> >
> > No, NAK. We don't want to see these in each and every module. Please,
> > make sure that nothing, except maybe Kconfig, is modified in this
> > folder (yet, you may add a _separate_ test module, as you already have
> > done in this patch).
>
> This is one of the cases in which we can't go without changing the
> original code, because parse_xy() is a static function.
> Including the test into the source is not the only option, we could as
> well make the function visible unconditionally, or introduce a macro
> similar to VISIBLE_IF_KUNIT.
> Do you prefer any of those?
Just add something like this to drivers/auxdisplay/Makefile:
ifeq ($(CONFIG_KFUZZTEST),y)
CFLAGS_charlcd.o := -include $(src)/tests/charlcd_kfuzz.c
endif
Alternatively, if the file in tests/ always has the same name
as the source file but with "_kfuzz.c" suffix, consider amending
scripts/Makefile.build to always include the "_kfuzz.c" file
if it exists and CONFIG_KFUZZTEST=y, thus avoiding the need
to amend all the individual Makefiles in the tree.
Thanks,
Lukas
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH v2 08/10] drivers/auxdisplay: add a KFuzzTest for parse_xy()
2025-09-20 12:47 ` Lukas Wunner
@ 2025-09-21 18:25 ` Andy Shevchenko
0 siblings, 0 replies; 30+ messages in thread
From: Andy Shevchenko @ 2025-09-21 18:25 UTC (permalink / raw)
To: Lukas Wunner
Cc: Alexander Potapenko, Ethan Graham, ethangraham, andreyknvl, andy,
brauner, brendan.higgins, davem, davidgow, dhowells, dvyukov,
elver, herbert, ignat, jack, jannh, johannes, kasan-dev, kees,
kunit-dev, linux-crypto, linux-kernel, linux-mm, rmoar, shuah,
sj, tarasmadan
On Sat, Sep 20, 2025 at 3:47 PM Lukas Wunner <lukas@wunner.de> wrote:
> On Sat, Sep 20, 2025 at 02:08:01PM +0200, Alexander Potapenko wrote:
> > On Sat, Sep 20, 2025 at 12:54 PM Andy Shevchenko <andy.shevchenko@gmail.com> wrote:
> > > On Fri, Sep 19, 2025 at 5:58 PM Ethan Graham <ethan.w.s.graham@gmail.com> wrote:
...
> > > > +/*
> > > > + * When CONFIG_KFUZZTEST is enabled, we include this _kfuzz.c file to ensure
> > > > + * that KFuzzTest targets are built.
> > > > + */
> > > > +#ifdef CONFIG_KFUZZTEST
> > > > +#include "tests/charlcd_kfuzz.c"
> > > > +#endif /* CONFIG_KFUZZTEST */
> > >
> > > No, NAK. We don't want to see these in each and every module. Please,
> > > make sure that nothing, except maybe Kconfig, is modified in this
> > > folder (yet, you may add a _separate_ test module, as you already have
> > > done in this patch).
> >
> > This is one of the cases in which we can't go without changing the
> > original code, because parse_xy() is a static function.
> > Including the test into the source is not the only option, we could as
> > well make the function visible unconditionally, or introduce a macro
> > similar to VISIBLE_IF_KUNIT.
> > Do you prefer any of those?
>
> Just add something like this to drivers/auxdisplay/Makefile:
>
> ifeq ($(CONFIG_KFUZZTEST),y)
> CFLAGS_charlcd.o := -include $(src)/tests/charlcd_kfuzz.c
> endif
>
> Alternatively, if the file in tests/ always has the same name
> as the source file but with "_kfuzz.c" suffix, consider amending
> scripts/Makefile.build to always include the "_kfuzz.c" file
> if it exists and CONFIG_KFUZZTEST=y, thus avoiding the need
> to amend all the individual Makefiles in the tree.
Thanks, Lukas, for the ideas. Yes, something like this would be acceptable.
--
With Best Regards,
Andy Shevchenko
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH v2 08/10] drivers/auxdisplay: add a KFuzzTest for parse_xy()
2025-09-19 14:57 ` [PATCH v2 08/10] drivers/auxdisplay: add a KFuzzTest for parse_xy() Ethan Graham
2025-09-19 15:07 ` Alexander Potapenko
2025-09-20 10:53 ` Andy Shevchenko
@ 2025-09-24 9:28 ` kernel test robot
2 siblings, 0 replies; 30+ messages in thread
From: kernel test robot @ 2025-09-24 9:28 UTC (permalink / raw)
To: Ethan Graham, ethangraham, glider
Cc: llvm, oe-kbuild-all, andreyknvl, andy, brauner, brendan.higgins,
davem, davidgow, dhowells, dvyukov, elver, herbert, ignat, jack,
jannh, johannes, kasan-dev, kees, kunit-dev, linux-crypto,
linux-kernel, linux-mm, lukas, rmoar, shuah, sj, tarasmadan
Hi Ethan,
kernel test robot noticed the following build errors:
[auto build test ERROR on akpm-mm/mm-nonmm-unstable]
[also build test ERROR on herbert-cryptodev-2.6/master herbert-crypto-2.6/master linus/master v6.17-rc7 next-20250923]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Ethan-Graham/mm-kasan-implement-kasan_poison_range/20250919-225911
base: https://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm.git mm-nonmm-unstable
patch link: https://lore.kernel.org/r/20250919145750.3448393-9-ethan.w.s.graham%40gmail.com
patch subject: [PATCH v2 08/10] drivers/auxdisplay: add a KFuzzTest for parse_xy()
config: x86_64-randconfig-008-20250924 (https://download.01.org/0day-ci/archive/20250924/202509241655.GL49TRF9-lkp@intel.com/config)
compiler: clang version 20.1.8 (https://github.com/llvm/llvm-project 87f0227cb60147a26a1eeb4fb06e3b505e9c7261)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20250924/202509241655.GL49TRF9-lkp@intel.com/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202509241655.GL49TRF9-lkp@intel.com/
All errors (new ones prefixed by >>, old ones prefixed by <<):
>> ERROR: modpost: "kfuzztest_parse_and_relocate" [drivers/auxdisplay/charlcd.ko] undefined!
>> ERROR: modpost: "record_invocation" [drivers/auxdisplay/charlcd.ko] undefined!
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH v2 09/10] fs/binfmt_script: add KFuzzTest target for load_script
2025-09-19 14:57 [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework Ethan Graham
` (7 preceding siblings ...)
2025-09-19 14:57 ` [PATCH v2 08/10] drivers/auxdisplay: add a KFuzzTest for parse_xy() Ethan Graham
@ 2025-09-19 14:57 ` Ethan Graham
2025-09-19 15:07 ` Alexander Potapenko
2025-09-19 19:19 ` Kees Cook
2025-09-19 14:57 ` [PATCH v2 10/10] MAINTAINERS: add maintainer information for KFuzzTest Ethan Graham
` (2 subsequent siblings)
11 siblings, 2 replies; 30+ messages in thread
From: Ethan Graham @ 2025-09-19 14:57 UTC (permalink / raw)
To: ethangraham, glider
Cc: andreyknvl, andy, brauner, brendan.higgins, davem, davidgow,
dhowells, dvyukov, elver, herbert, ignat, jack, jannh, johannes,
kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel, linux-mm,
lukas, rmoar, shuah, sj, tarasmadan
From: Ethan Graham <ethangraham@google.com>
Add a KFuzzTest target for the load_script function to serve as a
real-world example of the framework's usage.
The load_script function is responsible for parsing the shebang line
(`#!`) of script files. This makes it an excellent candidate for
KFuzzTest, as it involves parsing user-controlled data within the
binary loading path, which is not directly exposed as a system call.
The provided fuzz target in fs/tests/binfmt_script_kfuzz.c illustrates
how to fuzz a function that requires more involved setup - here, we only
let the fuzzer generate input for the `buf` field of struct linux_bprm,
and manually set the other fields with sensible values inside of the
FUZZ_TEST body.
To demonstrate the effectiveness of the fuzz target, a buffer overflow
bug was injected in the load_script function like so:
- buf_end = bprm->buf + sizeof(bprm->buf) - 1;
+ buf_end = bprm->buf + sizeof(bprm->buf) + 1;
Which was caught in around 40 seconds by syzkaller simultaneously
fuzzing four other targets, a realistic use case where targets are
continuously fuzzed. It also requires that the fuzzer be smart enough to
generate an input starting with `#!`.
While this bug is shallow, the fact that the bug is caught quickly and
with minimal additional code can potentially be a source of confidence
when modifying existing implementations or writing new functions.
Signed-off-by: Ethan Graham <ethangraham@google.com>
---
PR v2:
- Introduce cleanup logic in the load_script fuzz target.
---
---
fs/binfmt_script.c | 8 +++++
fs/tests/binfmt_script_kfuzz.c | 58 ++++++++++++++++++++++++++++++++++
2 files changed, 66 insertions(+)
create mode 100644 fs/tests/binfmt_script_kfuzz.c
diff --git a/fs/binfmt_script.c b/fs/binfmt_script.c
index 637daf6e4d45..c09f224d6d7e 100644
--- a/fs/binfmt_script.c
+++ b/fs/binfmt_script.c
@@ -157,3 +157,11 @@ core_initcall(init_script_binfmt);
module_exit(exit_script_binfmt);
MODULE_DESCRIPTION("Kernel support for scripts starting with #!");
MODULE_LICENSE("GPL");
+
+/*
+ * When CONFIG_KFUZZTEST is enabled, we include this _kfuzz.c file to ensure
+ * that KFuzzTest targets are built.
+ */
+#ifdef CONFIG_KFUZZTEST
+#include "tests/binfmt_script_kfuzz.c"
+#endif /* CONFIG_KFUZZTEST */
diff --git a/fs/tests/binfmt_script_kfuzz.c b/fs/tests/binfmt_script_kfuzz.c
new file mode 100644
index 000000000000..26397a465270
--- /dev/null
+++ b/fs/tests/binfmt_script_kfuzz.c
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * binfmt_script loader KFuzzTest target
+ *
+ * Copyright 2025 Google LLC
+ */
+#include <linux/binfmts.h>
+#include <linux/kfuzztest.h>
+#include <linux/slab.h>
+#include <linux/sched/mm.h>
+
+struct load_script_arg {
+ char buf[BINPRM_BUF_SIZE];
+};
+
+FUZZ_TEST(test_load_script, struct load_script_arg)
+{
+ struct linux_binprm bprm = {};
+ char *arg_page;
+
+ arg_page = (char *)get_zeroed_page(GFP_KERNEL);
+ if (!arg_page)
+ return;
+
+ memcpy(bprm.buf, arg->buf, sizeof(bprm.buf));
+ /*
+ * `load_script` calls remove_arg_zero, which expects argc != 0. A
+ * static value of 1 is sufficient for fuzzing.
+ */
+ bprm.argc = 1;
+ bprm.p = (unsigned long)arg_page + PAGE_SIZE;
+ bprm.filename = kstrdup("fuzz_script", GFP_KERNEL);
+ if (!bprm.filename)
+ goto cleanup;
+ bprm.interp = kstrdup(bprm.filename, GFP_KERNEL);
+ if (!bprm.interp)
+ goto cleanup;
+
+ bprm.mm = mm_alloc();
+ if (!bprm.mm)
+ goto cleanup;
+
+ /*
+ * Call the target function. We expect it to fail and return an error
+ * (e.g., at open_exec), which is fine. The goal is to survive the
+ * initial parsing logic without crashing.
+ */
+ load_script(&bprm);
+
+cleanup:
+ if (bprm.mm)
+ mmput(bprm.mm);
+ if (bprm.interp)
+ kfree(bprm.interp);
+ if (bprm.filename)
+ kfree(bprm.filename);
+ free_page((unsigned long)arg_page);
+}
--
2.51.0.470.ga7dc726c21-goog
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH v2 09/10] fs/binfmt_script: add KFuzzTest target for load_script
2025-09-19 14:57 ` [PATCH v2 09/10] fs/binfmt_script: add KFuzzTest target for load_script Ethan Graham
@ 2025-09-19 15:07 ` Alexander Potapenko
2025-09-19 19:19 ` Kees Cook
1 sibling, 0 replies; 30+ messages in thread
From: Alexander Potapenko @ 2025-09-19 15:07 UTC (permalink / raw)
To: Ethan Graham
Cc: ethangraham, andreyknvl, andy, brauner, brendan.higgins, davem,
davidgow, dhowells, dvyukov, elver, herbert, ignat, jack, jannh,
johannes, kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel,
linux-mm, lukas, rmoar, shuah, sj, tarasmadan
On Fri, Sep 19, 2025 at 4:58 PM Ethan Graham <ethan.w.s.graham@gmail.com> wrote:
>
> From: Ethan Graham <ethangraham@google.com>
>
> Add a KFuzzTest target for the load_script function to serve as a
> real-world example of the framework's usage.
>
> The load_script function is responsible for parsing the shebang line
> (`#!`) of script files. This makes it an excellent candidate for
> KFuzzTest, as it involves parsing user-controlled data within the
> binary loading path, which is not directly exposed as a system call.
>
> The provided fuzz target in fs/tests/binfmt_script_kfuzz.c illustrates
> how to fuzz a function that requires more involved setup - here, we only
> let the fuzzer generate input for the `buf` field of struct linux_bprm,
> and manually set the other fields with sensible values inside of the
> FUZZ_TEST body.
>
> To demonstrate the effectiveness of the fuzz target, a buffer overflow
> bug was injected in the load_script function like so:
>
> - buf_end = bprm->buf + sizeof(bprm->buf) - 1;
> + buf_end = bprm->buf + sizeof(bprm->buf) + 1;
>
> Which was caught in around 40 seconds by syzkaller simultaneously
> fuzzing four other targets, a realistic use case where targets are
> continuously fuzzed. It also requires that the fuzzer be smart enough to
> generate an input starting with `#!`.
>
> While this bug is shallow, the fact that the bug is caught quickly and
> with minimal additional code can potentially be a source of confidence
> when modifying existing implementations or writing new functions.
>
> Signed-off-by: Ethan Graham <ethangraham@google.com>
Acked-by: Alexander Potapenko <glider@google.com>
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH v2 09/10] fs/binfmt_script: add KFuzzTest target for load_script
2025-09-19 14:57 ` [PATCH v2 09/10] fs/binfmt_script: add KFuzzTest target for load_script Ethan Graham
2025-09-19 15:07 ` Alexander Potapenko
@ 2025-09-19 19:19 ` Kees Cook
1 sibling, 0 replies; 30+ messages in thread
From: Kees Cook @ 2025-09-19 19:19 UTC (permalink / raw)
To: Ethan Graham
Cc: ethangraham, glider, andreyknvl, andy, brauner, brendan.higgins,
davem, davidgow, dhowells, dvyukov, elver, herbert, ignat, jack,
jannh, johannes, kasan-dev, kunit-dev, linux-crypto,
linux-kernel, linux-mm, lukas, rmoar, shuah, sj, tarasmadan
On Fri, Sep 19, 2025 at 02:57:49PM +0000, Ethan Graham wrote:
> From: Ethan Graham <ethangraham@google.com>
>
> Add a KFuzzTest target for the load_script function to serve as a
> real-world example of the framework's usage.
>
> The load_script function is responsible for parsing the shebang line
> (`#!`) of script files. This makes it an excellent candidate for
> KFuzzTest, as it involves parsing user-controlled data within the
> binary loading path, which is not directly exposed as a system call.
>
> The provided fuzz target in fs/tests/binfmt_script_kfuzz.c illustrates
> how to fuzz a function that requires more involved setup - here, we only
> let the fuzzer generate input for the `buf` field of struct linux_bprm,
> and manually set the other fields with sensible values inside of the
> FUZZ_TEST body.
>
> To demonstrate the effectiveness of the fuzz target, a buffer overflow
> bug was injected in the load_script function like so:
>
> - buf_end = bprm->buf + sizeof(bprm->buf) - 1;
> + buf_end = bprm->buf + sizeof(bprm->buf) + 1;
>
> Which was caught in around 40 seconds by syzkaller simultaneously
> fuzzing four other targets, a realistic use case where targets are
> continuously fuzzed. It also requires that the fuzzer be smart enough to
> generate an input starting with `#!`.
>
> While this bug is shallow, the fact that the bug is caught quickly and
> with minimal additional code can potentially be a source of confidence
> when modifying existing implementations or writing new functions.
>
> Signed-off-by: Ethan Graham <ethangraham@google.com>
>
> ---
> PR v2:
> - Introduce cleanup logic in the load_script fuzz target.
> ---
> ---
> fs/binfmt_script.c | 8 +++++
> fs/tests/binfmt_script_kfuzz.c | 58 ++++++++++++++++++++++++++++++++++
> 2 files changed, 66 insertions(+)
> create mode 100644 fs/tests/binfmt_script_kfuzz.c
>
> diff --git a/fs/binfmt_script.c b/fs/binfmt_script.c
> index 637daf6e4d45..c09f224d6d7e 100644
> --- a/fs/binfmt_script.c
> +++ b/fs/binfmt_script.c
> @@ -157,3 +157,11 @@ core_initcall(init_script_binfmt);
> module_exit(exit_script_binfmt);
> MODULE_DESCRIPTION("Kernel support for scripts starting with #!");
> MODULE_LICENSE("GPL");
> +
> +/*
> + * When CONFIG_KFUZZTEST is enabled, we include this _kfuzz.c file to ensure
> + * that KFuzzTest targets are built.
> + */
> +#ifdef CONFIG_KFUZZTEST
> +#include "tests/binfmt_script_kfuzz.c"
> +#endif /* CONFIG_KFUZZTEST */
> diff --git a/fs/tests/binfmt_script_kfuzz.c b/fs/tests/binfmt_script_kfuzz.c
> new file mode 100644
> index 000000000000..26397a465270
> --- /dev/null
> +++ b/fs/tests/binfmt_script_kfuzz.c
> @@ -0,0 +1,58 @@
> +// SPDX-License-Identifier: GPL-2.0-or-later
> +/*
> + * binfmt_script loader KFuzzTest target
> + *
> + * Copyright 2025 Google LLC
> + */
> +#include <linux/binfmts.h>
> +#include <linux/kfuzztest.h>
> +#include <linux/slab.h>
> +#include <linux/sched/mm.h>
> +
> +struct load_script_arg {
> + char buf[BINPRM_BUF_SIZE];
> +};
> +
> +FUZZ_TEST(test_load_script, struct load_script_arg)
> +{
> + struct linux_binprm bprm = {};
> + char *arg_page;
> +
> + arg_page = (char *)get_zeroed_page(GFP_KERNEL);
> + if (!arg_page)
> + return;
> +
> + memcpy(bprm.buf, arg->buf, sizeof(bprm.buf));
> + /*
> + * `load_script` calls remove_arg_zero, which expects argc != 0. A
> + * static value of 1 is sufficient for fuzzing.
> + */
> + bprm.argc = 1;
> + bprm.p = (unsigned long)arg_page + PAGE_SIZE;
> + bprm.filename = kstrdup("fuzz_script", GFP_KERNEL);
> + if (!bprm.filename)
> + goto cleanup;
> + bprm.interp = kstrdup(bprm.filename, GFP_KERNEL);
> + if (!bprm.interp)
> + goto cleanup;
> +
> + bprm.mm = mm_alloc();
> + if (!bprm.mm)
> + goto cleanup;
> +
> + /*
> + * Call the target function. We expect it to fail and return an error
> + * (e.g., at open_exec), which is fine. The goal is to survive the
> + * initial parsing logic without crashing.
> + */
> + load_script(&bprm);
> +
> +cleanup:
> + if (bprm.mm)
> + mmput(bprm.mm);
> + if (bprm.interp)
> + kfree(bprm.interp);
> + if (bprm.filename)
> + kfree(bprm.filename);
> + free_page((unsigned long)arg_page);
> +}
Yay fuzzing hooks! I'm excited about this series overall, but I'm not
a fan of this "manual" init/clean up of bprm.
If you're going to set up a bprm that passes through load_script(), it
needs to be both prepared correctly (alloc_bprm) and cleaned up correctly
(free_bprm). Otherwise, you may be fuzzing impossible states created by
the fuzztest setup. And having a second init/cleanup path in here makes
future refactoring work more of a burden/fragile.
But this is also kind of not a great example of fuzztest utility because
load_script _is_ actually directly accessible from syscalls: it is trivial
to externally fuzz load_script by just writing the buffer to a file and
execve'ing it. :)
-Kees
--
Kees Cook
^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH v2 10/10] MAINTAINERS: add maintainer information for KFuzzTest
2025-09-19 14:57 [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework Ethan Graham
` (8 preceding siblings ...)
2025-09-19 14:57 ` [PATCH v2 09/10] fs/binfmt_script: add KFuzzTest target for load_script Ethan Graham
@ 2025-09-19 14:57 ` Ethan Graham
2025-09-24 8:32 ` SeongJae Park
2025-09-19 15:04 ` [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework Alexander Potapenko
2025-09-24 12:52 ` Johannes Berg
11 siblings, 1 reply; 30+ messages in thread
From: Ethan Graham @ 2025-09-19 14:57 UTC (permalink / raw)
To: ethangraham, glider
Cc: andreyknvl, andy, brauner, brendan.higgins, davem, davidgow,
dhowells, dvyukov, elver, herbert, ignat, jack, jannh, johannes,
kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel, linux-mm,
lukas, rmoar, shuah, sj, tarasmadan
From: Ethan Graham <ethangraham@google.com>
Add myself as maintainer and Alexander Potapenko as reviewer for
KFuzzTest.
Signed-off-by: Ethan Graham <ethangraham@google.com>
Acked-by: Alexander Potapenko <glider@google.com>
---
MAINTAINERS | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/MAINTAINERS b/MAINTAINERS
index 6dcfbd11efef..14972e3e9d6a 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -13641,6 +13641,14 @@ F: include/linux/kfifo.h
F: lib/kfifo.c
F: samples/kfifo/
+KFUZZTEST
+M: Ethan Graham <ethan.w.s.graham@gmail.com>
+R: Alexander Potapenko <glider@google.com>
+F: include/linux/kfuzztest.h
+F: lib/kfuzztest/
+F: Documentation/dev-tools/kfuzztest.rst
+F: tools/kfuzztest-bridge/
+
KGDB / KDB /debug_core
M: Jason Wessel <jason.wessel@windriver.com>
M: Daniel Thompson <danielt@kernel.org>
--
2.51.0.470.ga7dc726c21-goog
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH v2 10/10] MAINTAINERS: add maintainer information for KFuzzTest
2025-09-19 14:57 ` [PATCH v2 10/10] MAINTAINERS: add maintainer information for KFuzzTest Ethan Graham
@ 2025-09-24 8:32 ` SeongJae Park
0 siblings, 0 replies; 30+ messages in thread
From: SeongJae Park @ 2025-09-24 8:32 UTC (permalink / raw)
To: Ethan Graham
Cc: SeongJae Park, ethangraham, glider, andreyknvl, andy, brauner,
brendan.higgins, davem, davidgow, dhowells, dvyukov, elver,
herbert, ignat, jack, jannh, johannes, kasan-dev, kees,
kunit-dev, linux-crypto, linux-kernel, linux-mm, lukas, rmoar,
shuah, tarasmadan
Hello Ethan,
On Fri, 19 Sep 2025 14:57:50 +0000 Ethan Graham <ethan.w.s.graham@gmail.com> wrote:
> From: Ethan Graham <ethangraham@google.com>
>
> Add myself as maintainer and Alexander Potapenko as reviewer for
> KFuzzTest.
>
> Signed-off-by: Ethan Graham <ethangraham@google.com>
> Acked-by: Alexander Potapenko <glider@google.com>
> ---
> MAINTAINERS | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
> diff --git a/MAINTAINERS b/MAINTAINERS
> index 6dcfbd11efef..14972e3e9d6a 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -13641,6 +13641,14 @@ F: include/linux/kfifo.h
> F: lib/kfifo.c
> F: samples/kfifo/
>
> +KFUZZTEST
> +M: Ethan Graham <ethan.w.s.graham@gmail.com>
> +R: Alexander Potapenko <glider@google.com>
> +F: include/linux/kfuzztest.h
> +F: lib/kfuzztest/
> +F: Documentation/dev-tools/kfuzztest.rst
> +F: tools/kfuzztest-bridge/
I found you moved kfuzztest-bridge to tools/testing/ on this version, accepting
my suggestion. Thank you for that.
Nevertheless, so I think the above line should also be updated like below.
F: tools/testing/kfuzztest-bridge
Thanks,
SJ
> +
> KGDB / KDB /debug_core
> M: Jason Wessel <jason.wessel@windriver.com>
> M: Daniel Thompson <danielt@kernel.org>
> --
> 2.51.0.470.ga7dc726c21-goog
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework
2025-09-19 14:57 [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework Ethan Graham
` (9 preceding siblings ...)
2025-09-19 14:57 ` [PATCH v2 10/10] MAINTAINERS: add maintainer information for KFuzzTest Ethan Graham
@ 2025-09-19 15:04 ` Alexander Potapenko
2025-09-24 12:52 ` Johannes Berg
11 siblings, 0 replies; 30+ messages in thread
From: Alexander Potapenko @ 2025-09-19 15:04 UTC (permalink / raw)
To: Ethan Graham, shuah
Cc: ethangraham, andreyknvl, andy, brauner, brendan.higgins, davem,
davidgow, dhowells, dvyukov, elver, herbert, ignat, jack, jannh,
johannes, kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel,
linux-mm, lukas, rmoar, sj, tarasmadan
On Fri, Sep 19, 2025 at 4:58 PM Ethan Graham <ethan.w.s.graham@gmail.com> wrote:
>
> From: Ethan Graham <ethangraham@google.com>
>
> This patch series introduces KFuzzTest, a lightweight framework for
> creating in-kernel fuzz targets for internal kernel functions.
Hi Shuah,
Since these are all fundamentally test code, I was wondering if the
selftests tree would be the appropriate path for merging them?
If you agree, would you be open to picking them up once the review is done?
Thank you!
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework
2025-09-19 14:57 [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework Ethan Graham
` (10 preceding siblings ...)
2025-09-19 15:04 ` [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework Alexander Potapenko
@ 2025-09-24 12:52 ` Johannes Berg
2025-09-25 8:35 ` Ethan Graham
11 siblings, 1 reply; 30+ messages in thread
From: Johannes Berg @ 2025-09-24 12:52 UTC (permalink / raw)
To: Ethan Graham, ethangraham, glider
Cc: andreyknvl, andy, brauner, brendan.higgins, davem, davidgow,
dhowells, dvyukov, elver, herbert, ignat, jack, jannh, kasan-dev,
kees, kunit-dev, linux-crypto, linux-kernel, linux-mm, lukas,
rmoar, shuah, sj, tarasmadan
On Fri, 2025-09-19 at 14:57 +0000, Ethan Graham wrote:
>
> This patch series introduces KFuzzTest, a lightweight framework for
> creating in-kernel fuzz targets for internal kernel functions.
>
> The primary motivation for KFuzzTest is to simplify the fuzzing of
> low-level, relatively stateless functions (e.g., data parsers, format
> converters) that are difficult to exercise effectively from the syscall
> boundary. It is intended for in-situ fuzzing of kernel code without
> requiring that it be built as a separate userspace library or that its
> dependencies be stubbed out. Using a simple macro-based API, developers
> can add a new fuzz target with minimal boilerplate code.
So ... I guess I understand the motivation to make this easy for
developers, but I'm not sure I'm happy to have all of this effectively
depend on syzkaller.
You spelled out the process to actually declare a fuzz test, but you
never spelled out the process to actually run fuzzing against it. For
the record, and everyone else who might be reading, here's my
understanding:
- the FUZZ_TEST() macro declares some magic in the Linux binary,
including the name of the struct that describes the necessary input
- there's a parser in syzkaller (and not really usable standalone) that
can parse the vmlinux binary (and doesn't handle modules) and
generates descriptions for the input from it
- I _think_ that the bridge tool uses these descriptions, though the
example you have in the documentation just says "use this command for
this test" and makes no representation as to how the first argument
to the bridge tool is created, it just appears out of thin air
- the bridge tool will then parse the description and use some random
data to create the serialised data that's deserialized in the kernel
and then passed to the test
- side note: did that really have to be a custom serialization
format? I don't see any discussion on that, there are different
formats that exist already, I'd think?
- the test runs now, and may or may not crash, as you'd expect
I was really hoping to integrate this with ARCH=um and other fuzzers[1],
but ... I don't really think it's entirely feasible. I can basically
only require hard-coding the input description like the bridge tool
does, but that doesn't scale, or attempt to extract a few thousand lines
of code from syzkaller to extract the data...
[1] in particular honggfuzz as I wrote earlier, due to the coverage
feedback format issues with afl++, but if I were able to use clang
right now I could probably also make afl++ work in a similar way
by adding support for --fsanitize-coverage=trace-pc-guard first.
I'm not even saying that you had many choices here, but it's definitely
annoying, at least to me, that all this infrastructure is effectively
dependent on syzkaller due to all of this. At the same time, yes, I get
that parsing dwarf and getting a description out is not an easy feat,
and without the infrastructure already in syzkaller it'd take more than
the ~1.1kLOC (and even that is not small) it has now.
I guess the biggest question to me is ultimately why all that is
necessary? Right now, there's only the single example kfuzztest that
even uses this infrastructure beyond a single linear buffer [2]. Where
is all that complexity even worth it? It's expressly intended for
simpler pieces of code that parse something ("data parsers, format
converters").
[2] admittedly the auxdisplay one is slightly different and uses a
string, but that's pretty much equivalent
johannes
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework
2025-09-24 12:52 ` Johannes Berg
@ 2025-09-25 8:35 ` Ethan Graham
2025-10-24 8:37 ` Johannes Berg
0 siblings, 1 reply; 30+ messages in thread
From: Ethan Graham @ 2025-09-25 8:35 UTC (permalink / raw)
To: Johannes Berg
Cc: ethangraham, glider, andreyknvl, andy, brauner, brendan.higgins,
davem, davidgow, dhowells, dvyukov, elver, herbert, ignat, jack,
jannh, kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel,
linux-mm, lukas, rmoar, shuah, sj, tarasmadan
On Wed, Sep 24, 2025 at 2:52 PM Johannes Berg <johannes@sipsolutions.net> wrote:
>
> On Fri, 2025-09-19 at 14:57 +0000, Ethan Graham wrote:
> >
> > This patch series introduces KFuzzTest, a lightweight framework for
> > creating in-kernel fuzz targets for internal kernel functions.
> >
> > The primary motivation for KFuzzTest is to simplify the fuzzing of
> > low-level, relatively stateless functions (e.g., data parsers, format
> > converters) that are difficult to exercise effectively from the syscall
> > boundary. It is intended for in-situ fuzzing of kernel code without
> > requiring that it be built as a separate userspace library or that its
> > dependencies be stubbed out. Using a simple macro-based API, developers
> > can add a new fuzz target with minimal boilerplate code.
>
> So ... I guess I understand the motivation to make this easy for
> developers, but I'm not sure I'm happy to have all of this effectively
> depend on syzkaller.
I would argue that it only depends on syzkaller because it is currently
the only fuzzer that implements support for KFuzzTest. The communication
interface itself is agnostic.
> You spelled out the process to actually declare a fuzz test, but you
> never spelled out the process to actually run fuzzing against it. For
Running the fuzzing is more of a tooling concern, and so instructions
were left out here. For the interested, the syzkaller flow is described
on GitHub: https://github.com/google/syzkaller/blob/master/docs/kfuzztest.md
> the record, and everyone else who might be reading, here's my
> understanding:
>
> - the FUZZ_TEST() macro declares some magic in the Linux binary,
> including the name of the struct that describes the necessary input
>
> - there's a parser in syzkaller (and not really usable standalone) that
> can parse the vmlinux binary (and doesn't handle modules) and
> generates descriptions for the input from it
>
> - I _think_ that the bridge tool uses these descriptions, though the
> example you have in the documentation just says "use this command for
> this test" and makes no representation as to how the first argument
> to the bridge tool is created, it just appears out of thin air
syzkaller doesn't use the bridge tool at all. Since a KFuzzTest target is
invoked when you write encoded data into its debugfs input file, any
fuzzer that is able to do this is able to fuzz it - this is what syzkaller
does. The bridge tool was added to provide an out-of-the-box tool
for fuzzing KFuzzTest targets with arbitrary data that doesn't depend
on syzkaller at all.
In the provided examples, the kfuzztest-bridge descriptions were
hand-written, but it's also feasible to generate them with the ELF
metadata in vmlinux. It would be easy to implement support for
this in syzkaller, but then we would depend on an external tool
for autogenerating these descriptions which we wanted to avoid.
>
> - the bridge tool will then parse the description and use some random
> data to create the serialised data that's deserialized in the kernel
> and then passed to the test
This is exactly right. It's not used by syzkaller, but this is how it's
intended to work when it's used as a standalone tool, or for bridging
between KFuzzTest targets and an arbitrary fuzzer that doesn't
implement the required encoding logic.
> - side note: did that really have to be a custom serialization
> format? I don't see any discussion on that, there are different
> formats that exist already, I'd think?
>
> - the test runs now, and may or may not crash, as you'd expect
>
> I was really hoping to integrate this with ARCH=um and other fuzzers[1],
> but ... I don't really think it's entirely feasible. I can basically
> only require hard-coding the input description like the bridge tool
> does, but that doesn't scale, or attempt to extract a few thousand lines
> of code from syzkaller to extract the data...
I would argue that integrating with other fuzzers is feasible, but it does
require some if not a lot of work depending on the level of support. syzkaller
already did most of the heavy lifting with smart input generation and mutation
for kernel functions, so the changes needed for KFuzzTest were mainly:
- Dynamically discovering targets, but you could just as easily write a
syzkaller description for them.
- Encoding logic for the input format.
Assuming a fuzzer is able to generate C-struct inputs for a kernel function,
the only further requirement is being able to encode the input and write
it into the debugfs input file. The ELF data extraction is a nice-to-have
for sure, but it's not a strict requirement.
>
> [1] in particular honggfuzz as I wrote earlier, due to the coverage
> feedback format issues with afl++, but if I were able to use clang
> right now I could probably also make afl++ work in a similar way
> by adding support for --fsanitize-coverage=trace-pc-guard first.
>
>
> I'm not even saying that you had many choices here, but it's definitely
> annoying, at least to me, that all this infrastructure is effectively
> dependent on syzkaller due to all of this. At the same time, yes, I get
> that parsing dwarf and getting a description out is not an easy feat,
> and without the infrastructure already in syzkaller it'd take more than
> the ~1.1kLOC (and even that is not small) it has now.
>
>
> I guess the biggest question to me is ultimately why all that is
> necessary? Right now, there's only the single example kfuzztest that
> even uses this infrastructure beyond a single linear buffer [2]. Where
> is all that complexity even worth it? It's expressly intended for
> simpler pieces of code that parse something ("data parsers, format
> converters").
You're right that the provided examples don't leverage the feature of
being able to pass more complex nested data into the kernel. Perhaps
for a future iteration, it might be worth adding a target for a function
that takes more complex input. What do you think?
I'm not sure how much of the kernel complexity really could be reduced
if we decided to support only simpler inputs (e.g., linear buffers).
It would certainly simplify the fuzzer implementation, but the kernel
code would likely be similar if not the same.
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework
2025-09-25 8:35 ` Ethan Graham
@ 2025-10-24 8:37 ` Johannes Berg
2025-10-28 17:38 ` Alexander Potapenko
0 siblings, 1 reply; 30+ messages in thread
From: Johannes Berg @ 2025-10-24 8:37 UTC (permalink / raw)
To: Ethan Graham
Cc: ethangraham, glider, andreyknvl, andy, brauner, brendan.higgins,
davem, davidgow, dhowells, dvyukov, elver, herbert, ignat, jack,
jannh, kasan-dev, kees, kunit-dev, linux-crypto, linux-kernel,
linux-mm, lukas, rmoar, shuah, sj, tarasmadan
Hi Ethan, all,
Sorry, my current foray into fuzzing got preempted by other things ...
> > So ... I guess I understand the motivation to make this easy for
> > developers, but I'm not sure I'm happy to have all of this effectively
> > depend on syzkaller.
>
> I would argue that it only depends on syzkaller because it is currently
> the only fuzzer that implements support for KFuzzTest. The communication
> interface itself is agnostic.
Yeah I can see how you could argue that. However, syzkaller is also
effectively the only fuzzer now that supports what you later call "smart
input generation", and adding it to any other fuzzer is really not
straight-forward, at least to me. No other fuzzer seems to really have
felt a need to have this, and there are ... dozens?
> > the record, and everyone else who might be reading, here's my
> > understanding:
> >
> > - the FUZZ_TEST() macro declares some magic in the Linux binary,
> > including the name of the struct that describes the necessary input
> >
> > - there's a parser in syzkaller (and not really usable standalone) that
> > can parse the vmlinux binary (and doesn't handle modules) and
> > generates descriptions for the input from it
> >
> > - I _think_ that the bridge tool uses these descriptions, though the
> > example you have in the documentation just says "use this command for
> > this test" and makes no representation as to how the first argument
> > to the bridge tool is created, it just appears out of thin air
>
> syzkaller doesn't use the bridge tool at all.
Right.
> Since a KFuzzTest target is
> invoked when you write encoded data into its debugfs input file, any
> fuzzer that is able to do this is able to fuzz it - this is what syzkaller
> does. The bridge tool was added to provide an out-of-the-box tool
> for fuzzing KFuzzTest targets with arbitrary data that doesn't depend
> on syzkaller at all.
Yes, I understand, I guess it just feels a bit like a fig-leaf to me to
paper over "you need syzkaller" because there's no way to really
(efficiently) use it for fuzzing.
> In the provided examples, the kfuzztest-bridge descriptions were
> hand-written, but it's also feasible to generate them with the ELF
> metadata in vmlinux. It would be easy to implement support for
> this in syzkaller, but then we would depend on an external tool
> for autogenerating these descriptions which we wanted to avoid.
Oh, I get that you wouldn't necessarily want to have a dependency on
syzkaller in the kernel example code, but in a sense my argument is that
there's no such tool at all since syzkaller cannot output anything, and
then you need to write all the descriptions by hand. Which is fine for
an _example_ but really doesn't scale to actually running fuzzing.
So then we're mostly back to "you need syzkaller to run fuzzing against
this", which at least to me isn't a great situation.
> > - the bridge tool will then parse the description and use some random
> > data to create the serialised data that's deserialized in the kernel
> > and then passed to the test
>
> This is exactly right. It's not used by syzkaller, but this is how it's
> intended to work when it's used as a standalone tool, or for bridging
> between KFuzzTest targets and an arbitrary fuzzer that doesn't
> implement the required encoding logic.
Yeah I guess, but that still requires hand-coding the descriptions (or
writing a separate parser), and notably doesn't work with a sort of in-
process fuzzing I was envisioning for ARCH=um. Which ought to be much
faster, and even combinable with fork() as I alluded to in earlier
emails.
> > I was really hoping to integrate this with ARCH=um and other fuzzers[1],
> > but ... I don't really think it's entirely feasible. I can basically
> > only require hard-coding the input description like the bridge tool
> > does, but that doesn't scale, or attempt to extract a few thousand lines
> > of code from syzkaller to extract the data...
>
> I would argue that integrating with other fuzzers is feasible, but it does
> require some if not a lot of work depending on the level of support. syzkaller
> already did most of the heavy lifting with smart input generation and mutation
> for kernel functions, so the changes needed for KFuzzTest were mainly:
>
> - Dynamically discovering targets, but you could just as easily write a
> syzkaller description for them.
> - Encoding logic for the input format.
>
> Assuming a fuzzer is able to generate C-struct inputs for a kernel function,
> the only further requirement is being able to encode the input and write
> it into the debugfs input file. The ELF data extraction is a nice-to-have
> for sure, but it's not a strict requirement.
I mean, yeah, I guess but ... Is there a fuzzer that is able generate
such input? I haven't seen one. And running the bridge tool separately
is going to be rather expensive (vs. in-process like I'm thinking
about), and some form of data extraction is needed to make this scale at
all.
Sure, I can do it all manually for a single test, but is it really a
good idea that syzkaller is the only thing that could possibly run this
at scale?
> > I guess the biggest question to me is ultimately why all that is
> > necessary? Right now, there's only the single example kfuzztest that
> > even uses this infrastructure beyond a single linear buffer [2]. Where
> > is all that complexity even worth it? It's expressly intended for
> > simpler pieces of code that parse something ("data parsers, format
> > converters").
>
> You're right that the provided examples don't leverage the feature of
> being able to pass more complex nested data into the kernel. Perhaps
> for a future iteration, it might be worth adding a target for a function
> that takes more complex input. What do you think?
Well, I guess my thought is that there isn't actually going to be a good
example that really _requires_ all this flexibility. We're going to want
to test (mostly?) functions that consume untrusted data, but untrusted
data tends to come in the form of a linear blob, via the network, from a
file, from userspace, etc. Pretty much only the syscall boundary has
highly structured untrusted data, but syzkaller already fuzzes that and
we're not likely to write special kfuzztests for syscalls?
> I'm not sure how much of the kernel complexity really could be reduced
> if we decided to support only simpler inputs (e.g., linear buffers).
> It would certainly simplify the fuzzer implementation, but the kernel
> code would likely be similar if not the same.
Well, you wouldn't need the whole custom serialization format and
deserialization code for a start, nor the linker changes around
KFUZZTEST_TABLE since run-time discovery would likely be sufficient,
though of course those are trivial. And the deserialization is almost
half of the overall infrastructure code?
Anyway, I don't really know what to do. Maybe this has even landed by
now ;-) I certainly would've preferred something that was easier to use
with other fuzzers and in-process fuzzing in ARCH=um, but then that'd
now mean I need to plug it in at a completely different level, or write
a DWARF parser and serializer if I don't want to have to hand-code each
target.
I really do want to do fuzz testing on wifi, but with kfuzztest it
basically means I rely on syzbot to actually run it or have to run
syzkaller myself, rather than being able to integrate it with other
fuzzers say in ARCH=um. Personally, I think it'd be worthwhile to have
that, but I don't see how to integrate it well with this infrastructure.
Also, more generally, it seems unlikely that _anyone_ would ever do
this, and then it's basically only syzbot that will ever run it.
johannes
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH v2 0/10] KFuzzTest: a new kernel fuzzing framework
2025-10-24 8:37 ` Johannes Berg
@ 2025-10-28 17:38 ` Alexander Potapenko
0 siblings, 0 replies; 30+ messages in thread
From: Alexander Potapenko @ 2025-10-28 17:38 UTC (permalink / raw)
To: Johannes Berg
Cc: Ethan Graham, ethangraham, andreyknvl, andy, brauner,
brendan.higgins, davem, davidgow, dhowells, dvyukov, elver,
herbert, ignat, jack, jannh, kasan-dev, kees, kunit-dev,
linux-crypto, linux-kernel, linux-mm, lukas, rmoar, shuah, sj,
tarasmadan
On Fri, Oct 24, 2025 at 10:38 AM Johannes Berg
<johannes@sipsolutions.net> wrote:
>
> Hi Ethan, all,
Hi Johannes,
> > I would argue that it only depends on syzkaller because it is currently
> > the only fuzzer that implements support for KFuzzTest. The communication
> > interface itself is agnostic.
>
> Yeah I can see how you could argue that. However, syzkaller is also
> effectively the only fuzzer now that supports what you later call "smart
> input generation", and adding it to any other fuzzer is really not
> straight-forward, at least to me. No other fuzzer seems to really have
> felt a need to have this, and there are ... dozens?
Structure-aware fuzzing is not unique to syzkaller, nor are domain
constraints for certain values.
https://github.com/google/fuzztest is one example of a fuzzer that
supports both.
libFuzzer also supports custom mutators
(https://github.com/google/fuzzing/blob/master/docs/structure-aware-fuzzing.md)
> > Since a KFuzzTest target is
> > invoked when you write encoded data into its debugfs input file, any
> > fuzzer that is able to do this is able to fuzz it - this is what syzkaller
> > does. The bridge tool was added to provide an out-of-the-box tool
> > for fuzzing KFuzzTest targets with arbitrary data that doesn't depend
> > on syzkaller at all.
>
> Yes, I understand, I guess it just feels a bit like a fig-leaf to me to
> paper over "you need syzkaller" because there's no way to really
> (efficiently) use it for fuzzing.
When designing KFuzzTest, we anticipated two potential user scenarios:
1. The code author develops the fuzz test and runs it locally to
ensure its sanity and catch obvious errors.
2. The fuzz test lands upstream and syzkaller runs it continuously.
Ethan initially developed tools for both scenarios on the syzkaller
side, prioritizing simplicity of use over the diversity of potential
non-default fuzzing engines.
However, because smoke testing does not require a syzkaller
dependency, he added the bridge utility (I believe David Gow suggested
it).
That utility is easy to use for smoke testing, as it requires only a
one-line structure description.
I understand it may not be suitable for users who want to extensively
fuzz a particular test on their own machine without involving
syzkaller.
I agree we can do a better job by implementing some of the following options:
1. For tests without nested structures, or for tests that request it
explicitly, allow a simpler input format via a separate debugfs file.
2. Export the constraints/annotations via debugfs in a string format
so that fuzzers do not need vmlinux access to obtain them.
3. Export the fuzz test input structure as a string. (We've looked
into this and deemed it infeasible because test inputs may reference C
structures, and we don't have a reflection mechanism that would allow
us to dump the contents of existing structs).
> > This is exactly right. It's not used by syzkaller, but this is how it's
> > intended to work when it's used as a standalone tool, or for bridging
> > between KFuzzTest targets and an arbitrary fuzzer that doesn't
> > implement the required encoding logic.
>
> Yeah I guess, but that still requires hand-coding the descriptions (or
> writing a separate parser), and notably doesn't work with a sort of in-
> process fuzzing I was envisioning for ARCH=um. Which ought to be much
> faster, and even combinable with fork() as I alluded to in earlier
> emails.
Can you describe the interface between the fuzz test and the fuzzing
engine that you have in mind?
For ARCH=um, if you don't need structure awareness, I think the
easiest solution would be to make FUZZ_TEST wrap the code into
something akin to LLVMFuzzerTestOneInput()
(https://llvm.org/docs/LibFuzzer.html) that would directly pass random
data into the function under test. The debugfs interface is probably
excessive in this case.
But let's say we want to run in-kernel fuzzing with e.g. AFL++ - will
a simplified debugfs interface solve the problem?
What special cases can we omit to simplify the interface?
> I mean, yeah, I guess but ... Is there a fuzzer that is able generate
> such input? I haven't seen one. And running the bridge tool separately
> is going to be rather expensive (vs. in-process like I'm thinking
> about), and some form of data extraction is needed to make this scale at
> all.
>
> Sure, I can do it all manually for a single test, but is it really a
> good idea that syzkaller is the only thing that could possibly run this
> at scale?
Adding more fuzzing engines will not automatically allow us to run
this at scale.
For that, we'll need a continuous fuzzing system to manage the kernels
and corpora, report bugs, find reproducers, and bisect the causes.
I don't think building one for another fuzzing engine will be worth it.
That said, we can help developers better fuzz their code during local
runs by not always requiring the serialization format.
> > You're right that the provided examples don't leverage the feature of
> > being able to pass more complex nested data into the kernel. Perhaps
> > for a future iteration, it might be worth adding a target for a function
> > that takes more complex input. What do you think?
>
> Well, I guess my thought is that there isn't actually going to be a good
> example that really _requires_ all this flexibility. We're going to want
> to test (mostly?) functions that consume untrusted data, but untrusted
> data tends to come in the form of a linear blob, via the network, from a
> file, from userspace, etc. Pretty much only the syscall boundary has
> highly structured untrusted data, but syzkaller already fuzzes that and
> we're not likely to write special kfuzztests for syscalls?
We are not limited to fuzzing parsers of untrusted data. The idea
behind KFuzzTest is to validate that a piece of code can cope with any
input satisfying the constraints.
We could just as well fuzz a sorting algorithm or the bitops.
E.g. Will Deacon had the idea of fuzzing a hypervisor, which
potentially has many parameters, not all of which are necessarily
blobs.
> > I'm not sure how much of the kernel complexity really could be reduced
> > if we decided to support only simpler inputs (e.g., linear buffers).
> > It would certainly simplify the fuzzer implementation, but the kernel
> > code would likely be similar if not the same.
>
> Well, you wouldn't need the whole custom serialization format and
> deserialization code for a start, nor the linker changes around
> KFUZZTEST_TABLE since run-time discovery would likely be sufficient,
> though of course those are trivial. And the deserialization is almost
> half of the overall infrastructure code?
We could indeed organize the code so that simpler test cases (e.g. the
examples provided in this series) do not require the custom
serialization format.
I am still not convinced the whole serialization idea is useless, but
perhaps having a simplified version will unblock more users.
>
> Anyway, I don't really know what to do. Maybe this has even landed by
> now ;-) I certainly would've preferred something that was easier to use
> with other fuzzers and in-process fuzzing in ARCH=um, but then that'd
> now mean I need to plug it in at a completely different level, or write
> a DWARF parser and serializer if I don't want to have to hand-code each
> target.
>
> I really do want to do fuzz testing on wifi, but with kfuzztest it
> basically means I rely on syzbot to actually run it or have to run
> syzkaller myself, rather than being able to integrate it with other
> fuzzers say in ARCH=um. Personally, I think it'd be worthwhile to have
> that, but I don't see how to integrate it well with this infrastructure.
Can you please share some potential entry points you have in mind?
Understanding which functions you want to fuzz will help us simplify the format.
Thank you for your input!
> Also, more generally, it seems unlikely that _anyone_ would ever do
> this, and then it's basically only syzbot that will ever run it.
>
> johannes
>
> --
> You received this message because you are subscribed to the Google Groups "kasan-dev" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to kasan-dev+unsubscribe@googlegroups.com.
> To view this discussion visit https://groups.google.com/d/msgid/kasan-dev/438ff89e22a815c81406c3c8761a951b0c7e6916.camel%40sipsolutions.net.
--
Alexander Potapenko
Software Engineer
Google Germany GmbH
Erika-Mann-Straße, 33
80636 München
Geschäftsführer: Paul Manicle, Liana Sebastian
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg
^ permalink raw reply [flat|nested] 30+ messages in thread