Merge branch 'libbpf: auto-resize relocatable LOAD/STORE instructions'

Andrii Nakryiko says:

====================
Patch set implements logic in libbpf to auto-adjust memory size (1-, 2-, 4-,
8-bytes) of load/store (LD/ST/STX) instructions which have BPF CO-RE field
offset relocation associated with it. In practice this means transparent
handling of 32-bit kernels, both pointer and unsigned integers. Signed
integers are not relocatable with zero-extending loads/stores, so libbpf
poisons them and generates a warning. If/when BPF gets support for
sign-extending loads/stores, it would be possible to automatically relocate
them as well.

All the details are contained in patch #2 comments and commit message.
Patch #3 is a simple change in libbpf to make advanced testing with custom BTF
easier. Patch #4 validates correct uses of auto-resizable loads, as well as
check that libbpf fails invalid uses. Patch #1 skips CO-RE relocation for
programs that had bpf_program__set_autoload(prog, false) set on them, reducing
warnings and noise.

v2->v3:
  - fix copyright (Alexei);
v1->v2:
  - more consistent names for instruction mem size convertion routines (Alexei);
  - extended selftests to use relocatable STX instructions (Alexei);
  - added a fix for skipping CO-RE relocation for non-loadable programs.

Cc: Luka Perkov <luka.perkov@sartura.hr>
Cc: Tony Ambardar <tony.ambardar@gmail.com>
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

tools/lib/bpf/libbpf.c

@@ -5040,16 +5040,19 @@ static int bpf_core_spec_match(struct bpf_core_spec *local_spec,
 static int bpf_core_calc_field_relo(const struct bpf_program *prog,
 				    const struct bpf_core_relo *relo,
 				    const struct bpf_core_spec *spec,
-				    __u32 *val, bool *validate)
+				    __u32 *val, __u32 *field_sz, __u32 *type_id,
+				    bool *validate)
 {
 	const struct bpf_core_accessor *acc;
 	const struct btf_type *t;
-	__u32 byte_off, byte_sz, bit_off, bit_sz;
+	__u32 byte_off, byte_sz, bit_off, bit_sz, field_type_id;
 	const struct btf_member *m;
 	const struct btf_type *mt;
 	bool bitfield;
 	__s64 sz;
 
+	*field_sz = 0;
+
 	if (relo->kind == BPF_FIELD_EXISTS) {
 		*val = spec ? 1 : 0;
 		return 0;
@@ -5065,6 +5068,12 @@ static int bpf_core_calc_field_relo(const struct bpf_program *prog,
 	if (!acc->name) {
 		if (relo->kind == BPF_FIELD_BYTE_OFFSET) {
 			*val = spec->bit_offset / 8;
+			/* remember field size for load/store mem size */
+			sz = btf__resolve_size(spec->btf, acc->type_id);
+			if (sz < 0)
+				return -EINVAL;
+			*field_sz = sz;
+			*type_id = acc->type_id;
 		} else if (relo->kind == BPF_FIELD_BYTE_SIZE) {
 			sz = btf__resolve_size(spec->btf, acc->type_id);
 			if (sz < 0)
@@ -5081,7 +5090,7 @@ static int bpf_core_calc_field_relo(const struct bpf_program *prog,
 	}
 
 	m = btf_members(t) + acc->idx;
-	mt = skip_mods_and_typedefs(spec->btf, m->type, NULL);
+	mt = skip_mods_and_typedefs(spec->btf, m->type, &field_type_id);
 	bit_off = spec->bit_offset;
 	bit_sz = btf_member_bitfield_size(t, acc->idx);
 
@@ -5101,7 +5110,7 @@ static int bpf_core_calc_field_relo(const struct bpf_program *prog,
 			byte_off = bit_off / 8 / byte_sz * byte_sz;
 		}
 	} else {
-		sz = btf__resolve_size(spec->btf, m->type);
+		sz = btf__resolve_size(spec->btf, field_type_id);
 		if (sz < 0)
 			return -EINVAL;
 		byte_sz = sz;
@@ -5119,6 +5128,10 @@ static int bpf_core_calc_field_relo(const struct bpf_program *prog,
 	switch (relo->kind) {
 	case BPF_FIELD_BYTE_OFFSET:
 		*val = byte_off;
+		if (!bitfield) {
+			*field_sz = byte_sz;
+			*type_id = field_type_id;
+		}
 		break;
 	case BPF_FIELD_BYTE_SIZE:
 		*val = byte_sz;
@@ -5219,6 +5232,19 @@ struct bpf_core_relo_res
 	bool poison;
 	/* some relocations can't be validated against orig_val */
 	bool validate;
+	/* for field byte offset relocations or the forms:
+	 *     *(T *)(rX + <off>) = rY
+	 *     rX = *(T *)(rY + <off>),
+	 * we remember original and resolved field size to adjust direct
+	 * memory loads of pointers and integers; this is necessary for 32-bit
+	 * host kernel architectures, but also allows to automatically
+	 * relocate fields that were resized from, e.g., u32 to u64, etc.
+	 */
+	bool fail_memsz_adjust;
+	__u32 orig_sz;
+	__u32 orig_type_id;
+	__u32 new_sz;
+	__u32 new_type_id;
 };
 
 /* Calculate original and target relocation values, given local and target
@@ -5240,10 +5266,56 @@ static int bpf_core_calc_relo(const struct bpf_program *prog,
 	res->new_val = 0;
 	res->poison = false;
 	res->validate = true;
+	res->fail_memsz_adjust = false;
+	res->orig_sz = res->new_sz = 0;
+	res->orig_type_id = res->new_type_id = 0;
 
 	if (core_relo_is_field_based(relo->kind)) {
-		err = bpf_core_calc_field_relo(prog, relo, local_spec, &res->orig_val, &res->validate);
-		err = err ?: bpf_core_calc_field_relo(prog, relo, targ_spec, &res->new_val, NULL);
+		err = bpf_core_calc_field_relo(prog, relo, local_spec,
+					       &res->orig_val, &res->orig_sz,
+					       &res->orig_type_id, &res->validate);
+		err = err ?: bpf_core_calc_field_relo(prog, relo, targ_spec,
+						      &res->new_val, &res->new_sz,
+						      &res->new_type_id, NULL);
+		if (err)
+			goto done;
+		/* Validate if it's safe to adjust load/store memory size.
+		 * Adjustments are performed only if original and new memory
+		 * sizes differ.
+		 */
+		res->fail_memsz_adjust = false;
+		if (res->orig_sz != res->new_sz) {
+			const struct btf_type *orig_t, *new_t;
+
+			orig_t = btf__type_by_id(local_spec->btf, res->orig_type_id);
+			new_t = btf__type_by_id(targ_spec->btf, res->new_type_id);
+
+			/* There are two use cases in which it's safe to
+			 * adjust load/store's mem size:
+			 *   - reading a 32-bit kernel pointer, while on BPF
+			 *   size pointers are always 64-bit; in this case
+			 *   it's safe to "downsize" instruction size due to
+			 *   pointer being treated as unsigned integer with
+			 *   zero-extended upper 32-bits;
+			 *   - reading unsigned integers, again due to
+			 *   zero-extension is preserving the value correctly.
+			 *
+			 * In all other cases it's incorrect to attempt to
+			 * load/store field because read value will be
+			 * incorrect, so we poison relocated instruction.
+			 */
+			if (btf_is_ptr(orig_t) && btf_is_ptr(new_t))
+				goto done;
+			if (btf_is_int(orig_t) && btf_is_int(new_t) &&
+			    btf_int_encoding(orig_t) != BTF_INT_SIGNED &&
+			    btf_int_encoding(new_t) != BTF_INT_SIGNED)
+				goto done;
+
+			/* mark as invalid mem size adjustment, but this will
+			 * only be checked for LDX/STX/ST insns
+			 */
+			res->fail_memsz_adjust = true;
+		}
 	} else if (core_relo_is_type_based(relo->kind)) {
 		err = bpf_core_calc_type_relo(relo, local_spec, &res->orig_val);
 		err = err ?: bpf_core_calc_type_relo(relo, targ_spec, &res->new_val);
@@ -5252,6 +5324,7 @@ static int bpf_core_calc_relo(const struct bpf_program *prog,
 		err = err ?: bpf_core_calc_enumval_relo(relo, targ_spec, &res->new_val);
 	}
 
+done:
 	if (err == -EUCLEAN) {
 		/* EUCLEAN is used to signal instruction poisoning request */
 		res->poison = true;
@@ -5291,6 +5364,28 @@ static bool is_ldimm64(struct bpf_insn *insn)
 	return insn->code == (BPF_LD | BPF_IMM | BPF_DW);
 }
 
+static int insn_bpf_size_to_bytes(struct bpf_insn *insn)
+{
+	switch (BPF_SIZE(insn->code)) {
+	case BPF_DW: return 8;
+	case BPF_W: return 4;
+	case BPF_H: return 2;
+	case BPF_B: return 1;
+	default: return -1;
+	}
+}
+
+static int insn_bytes_to_bpf_size(__u32 sz)
+{
+	switch (sz) {
+	case 8: return BPF_DW;
+	case 4: return BPF_W;
+	case 2: return BPF_H;
+	case 1: return BPF_B;
+	default: return -1;
+	}
+}
+
 /*
  * Patch relocatable BPF instruction.
  *
@@ -5300,10 +5395,13 @@ static bool is_ldimm64(struct bpf_insn *insn)
  * spec, and is checked before patching instruction. If actual insn->imm value
  * is wrong, bail out with error.
  *
- * Currently three kinds of BPF instructions are supported:
+ * Currently supported classes of BPF instruction are:
  * 1. rX = <imm> (assignment with immediate operand);
  * 2. rX += <imm> (arithmetic operations with immediate operand);
- * 3. rX = <imm64> (load with 64-bit immediate value).
+ * 3. rX = <imm64> (load with 64-bit immediate value);
+ * 4. rX = *(T *)(rY + <off>), where T is one of {u8, u16, u32, u64};
+ * 5. *(T *)(rX + <off>) = rY, where T is one of {u8, u16, u32, u64};
+ * 6. *(T *)(rX + <off>) = <imm>, where T is one of {u8, u16, u32, u64}.
  */
 static int bpf_core_patch_insn(struct bpf_program *prog,
 			       const struct bpf_core_relo *relo,
@@ -5327,6 +5425,7 @@ static int bpf_core_patch_insn(struct bpf_program *prog,
 	class = BPF_CLASS(insn->code);
 
 	if (res->poison) {
+poison:
 		/* poison second part of ldimm64 to avoid confusing error from
 		 * verifier about "unknown opcode 00"
 		 */
@@ -5369,10 +5468,39 @@ static int bpf_core_patch_insn(struct bpf_program *prog,
 				prog->name, relo_idx, insn_idx, new_val);
 			return -ERANGE;
 		}
+		if (res->fail_memsz_adjust) {
+			pr_warn("prog '%s': relo #%d: insn #%d (LDX/ST/STX) accesses field incorrectly. "
+				"Make sure you are accessing pointers, unsigned integers, or fields of matching type and size.\n",
+				prog->name, relo_idx, insn_idx);
+			goto poison;
+		}
+
 		orig_val = insn->off;
 		insn->off = new_val;
 		pr_debug("prog '%s': relo #%d: patched insn #%d (LDX/ST/STX) off %u -> %u\n",
 			 prog->name, relo_idx, insn_idx, orig_val, new_val);
+
+		if (res->new_sz != res->orig_sz) {
+			int insn_bytes_sz, insn_bpf_sz;
+
+			insn_bytes_sz = insn_bpf_size_to_bytes(insn);
+			if (insn_bytes_sz != res->orig_sz) {
+				pr_warn("prog '%s': relo #%d: insn #%d (LDX/ST/STX) unexpected mem size: got %d, exp %u\n",
+					prog->name, relo_idx, insn_idx, insn_bytes_sz, res->orig_sz);
+				return -EINVAL;
+			}
+
+			insn_bpf_sz = insn_bytes_to_bpf_size(res->new_sz);
+			if (insn_bpf_sz < 0) {
+				pr_warn("prog '%s': relo #%d: insn #%d (LDX/ST/STX) invalid new mem size: %u\n",
+					prog->name, relo_idx, insn_idx, res->new_sz);
+				return -EINVAL;
+			}
+
+			insn->code = BPF_MODE(insn->code) | insn_bpf_sz | BPF_CLASS(insn->code);
+			pr_debug("prog '%s': relo #%d: patched insn #%d (LDX/ST/STX) mem_sz %u -> %u\n",
+				 prog->name, relo_idx, insn_idx, res->orig_sz, res->new_sz);
+		}
 		break;
 	case BPF_LD: {
 		__u64 imm;
@@ -5714,7 +5842,7 @@ bpf_object__relocate_core(struct bpf_object *obj, const char *targ_btf_path)
 		return 0;
 
 	if (targ_btf_path)
-		targ_btf = btf__parse_elf(targ_btf_path, NULL);
+		targ_btf = btf__parse(targ_btf_path, NULL);
 	else
 		targ_btf = obj->btf_vmlinux;
 	if (IS_ERR_OR_NULL(targ_btf)) {
@@ -5765,6 +5893,11 @@ bpf_object__relocate_core(struct bpf_object *obj, const char *targ_btf_path)
 				err = -EINVAL;
 				goto out;
 			}
+			/* no need to apply CO-RE relocation if the program is
+			 * not going to be loaded
+			 */
+			if (!prog->load)
+				continue;
 
 			err = bpf_core_apply_relo(prog, rec, i, obj->btf,
 						  targ_btf, cand_cache);

tools/testing/selftests/bpf/prog_tests/core_autosize.c

+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+
+#include <test_progs.h>
+#include <bpf/btf.h>
+
+/* real layout and sizes according to test's (32-bit) BTF
+ * needs to be defined before skeleton is included */
+struct test_struct___real {
+	unsigned int ptr; /* can't use `void *`, it is always 8 byte in BPF target */
+	unsigned int val2;
+	unsigned long long val1;
+	unsigned short val3;
+	unsigned char val4;
+	unsigned char _pad;
+};
+
+#include "test_core_autosize.skel.h"
+
+static int duration = 0;
+
+static struct {
+	unsigned long long ptr_samesized;
+	unsigned long long val1_samesized;
+	unsigned long long val2_samesized;
+	unsigned long long val3_samesized;
+	unsigned long long val4_samesized;
+	struct test_struct___real output_samesized;
+
+	unsigned long long ptr_downsized;
+	unsigned long long val1_downsized;
+	unsigned long long val2_downsized;
+	unsigned long long val3_downsized;
+	unsigned long long val4_downsized;
+	struct test_struct___real output_downsized;
+
+	unsigned long long ptr_probed;
+	unsigned long long val1_probed;
+	unsigned long long val2_probed;
+	unsigned long long val3_probed;
+	unsigned long long val4_probed;
+
+	unsigned long long ptr_signed;
+	unsigned long long val1_signed;
+	unsigned long long val2_signed;
+	unsigned long long val3_signed;
+	unsigned long long val4_signed;
+	struct test_struct___real output_signed;
+} out;
+
+void test_core_autosize(void)
+{
+	char btf_file[] = "/tmp/core_autosize.btf.XXXXXX";
+	int err, fd = -1, zero = 0;
+	int char_id, short_id, int_id, long_long_id, void_ptr_id, id;
+	struct test_core_autosize* skel = NULL;
+	struct bpf_object_load_attr load_attr = {};
+	struct bpf_program *prog;
+	struct bpf_map *bss_map;
+	struct btf *btf = NULL;
+	size_t written;
+	const void *raw_data;
+	__u32 raw_sz;
+	FILE *f = NULL;
+
+	btf = btf__new_empty();
+	if (!ASSERT_OK_PTR(btf, "empty_btf"))
+		return;
+	/* Emit the following struct with 32-bit pointer size:
+	 *
+	 * struct test_struct {
+	 *     void *ptr;
+	 *     unsigned long val2;
+	 *     unsigned long long val1;
+	 *     unsigned short val3;
+	 *     unsigned char val4;
+	 *     char: 8;
+	 * };
+	 *
+	 * This struct is going to be used as the "kernel BTF" for this test.
+	 * It's equivalent memory-layout-wise to test_struct__real above.
+	 */
+
+	/* force 32-bit pointer size */
+	btf__set_pointer_size(btf, 4);
+
+	char_id = btf__add_int(btf, "unsigned char", 1, 0);
+	ASSERT_EQ(char_id, 1, "char_id");
+	short_id = btf__add_int(btf, "unsigned short", 2, 0);
+	ASSERT_EQ(short_id, 2, "short_id");
+	/* "long unsigned int" of 4 byte size tells BTF that sizeof(void *) == 4 */
+	int_id = btf__add_int(btf, "long unsigned int", 4, 0);
+	ASSERT_EQ(int_id, 3, "int_id");
+	long_long_id = btf__add_int(btf, "unsigned long long", 8, 0);
+	ASSERT_EQ(long_long_id, 4, "long_long_id");
+	void_ptr_id = btf__add_ptr(btf, 0);
+	ASSERT_EQ(void_ptr_id, 5, "void_ptr_id");
+
+	id = btf__add_struct(btf, "test_struct", 20 /* bytes */);
+	ASSERT_EQ(id, 6, "struct_id");
+	err = btf__add_field(btf, "ptr", void_ptr_id, 0, 0);
+	err = err ?: btf__add_field(btf, "val2", int_id, 32, 0);
+	err = err ?: btf__add_field(btf, "val1", long_long_id, 64, 0);
+	err = err ?: btf__add_field(btf, "val3", short_id, 128, 0);
+	err = err ?: btf__add_field(btf, "val4", char_id, 144, 0);
+	ASSERT_OK(err, "struct_fields");
+
+	fd = mkstemp(btf_file);
+	if (CHECK(fd < 0, "btf_tmp", "failed to create file: %d\n", fd))
+		goto cleanup;
+	f = fdopen(fd, "w");
+	if (!ASSERT_OK_PTR(f, "btf_fdopen"))
+		goto cleanup;
+
+	raw_data = btf__get_raw_data(btf, &raw_sz);
+	if (!ASSERT_OK_PTR(raw_data, "raw_data"))
+		goto cleanup;
+	written = fwrite(raw_data, 1, raw_sz, f);
+	if (CHECK(written != raw_sz, "btf_write", "written: %zu, errno: %d\n", written, errno))
+		goto cleanup;
+	fflush(f);
+	fclose(f);
+	f = NULL;
+	close(fd);
+	fd = -1;
+
+	/* open and load BPF program with custom BTF as the kernel BTF */
+	skel = test_core_autosize__open();
+	if (!ASSERT_OK_PTR(skel, "skel_open"))
+		return;
+
+	/* disable handle_signed() for now */
+	prog = bpf_object__find_program_by_name(skel->obj, "handle_signed");
+	if (!ASSERT_OK_PTR(prog, "prog_find"))
+		goto cleanup;
+	bpf_program__set_autoload(prog, false);
+
+	load_attr.obj = skel->obj;
+	load_attr.target_btf_path = btf_file;
+	err = bpf_object__load_xattr(&load_attr);
+	if (!ASSERT_OK(err, "prog_load"))
+		goto cleanup;
+
+	prog = bpf_object__find_program_by_name(skel->obj, "handle_samesize");
+	if (!ASSERT_OK_PTR(prog, "prog_find"))
+		goto cleanup;
+	skel->links.handle_samesize = bpf_program__attach(prog);
+	if (!ASSERT_OK_PTR(skel->links.handle_samesize, "prog_attach"))
+		goto cleanup;
+
+	prog = bpf_object__find_program_by_name(skel->obj, "handle_downsize");
+	if (!ASSERT_OK_PTR(prog, "prog_find"))
+		goto cleanup;
+	skel->links.handle_downsize = bpf_program__attach(prog);
+	if (!ASSERT_OK_PTR(skel->links.handle_downsize, "prog_attach"))
+		goto cleanup;
+
+	prog = bpf_object__find_program_by_name(skel->obj, "handle_probed");
+	if (!ASSERT_OK_PTR(prog, "prog_find"))
+		goto cleanup;
+	skel->links.handle_probed = bpf_program__attach(prog);
+	if (!ASSERT_OK_PTR(skel->links.handle_probed, "prog_attach"))
+		goto cleanup;
+
+	usleep(1);
+
+	bss_map = bpf_object__find_map_by_name(skel->obj, "test_cor.bss");
+	if (!ASSERT_OK_PTR(bss_map, "bss_map_find"))
+		goto cleanup;
+
+	err = bpf_map_lookup_elem(bpf_map__fd(bss_map), &zero, (void *)&out);
+	if (!ASSERT_OK(err, "bss_lookup"))
+		goto cleanup;
+
+	ASSERT_EQ(out.ptr_samesized, 0x01020304, "ptr_samesized");
+	ASSERT_EQ(out.val1_samesized, 0x1020304050607080, "val1_samesized");
+	ASSERT_EQ(out.val2_samesized, 0x0a0b0c0d, "val2_samesized");
+	ASSERT_EQ(out.val3_samesized, 0xfeed, "val3_samesized");
+	ASSERT_EQ(out.val4_samesized, 0xb9, "val4_samesized");
+	ASSERT_EQ(out.output_samesized.ptr, 0x01020304, "ptr_samesized");
+	ASSERT_EQ(out.output_samesized.val1, 0x1020304050607080, "val1_samesized");
+	ASSERT_EQ(out.output_samesized.val2, 0x0a0b0c0d, "val2_samesized");
+	ASSERT_EQ(out.output_samesized.val3, 0xfeed, "val3_samesized");
+	ASSERT_EQ(out.output_samesized.val4, 0xb9, "val4_samesized");
+
+	ASSERT_EQ(out.ptr_downsized, 0x01020304, "ptr_downsized");
+	ASSERT_EQ(out.val1_downsized, 0x1020304050607080, "val1_downsized");
+	ASSERT_EQ(out.val2_downsized, 0x0a0b0c0d, "val2_downsized");
+	ASSERT_EQ(out.val3_downsized, 0xfeed, "val3_downsized");
+	ASSERT_EQ(out.val4_downsized, 0xb9, "val4_downsized");
+	ASSERT_EQ(out.output_downsized.ptr, 0x01020304, "ptr_downsized");
+	ASSERT_EQ(out.output_downsized.val1, 0x1020304050607080, "val1_downsized");
+	ASSERT_EQ(out.output_downsized.val2, 0x0a0b0c0d, "val2_downsized");
+	ASSERT_EQ(out.output_downsized.val3, 0xfeed, "val3_downsized");
+	ASSERT_EQ(out.output_downsized.val4, 0xb9, "val4_downsized");
+
+	ASSERT_EQ(out.ptr_probed, 0x01020304, "ptr_probed");
+	ASSERT_EQ(out.val1_probed, 0x1020304050607080, "val1_probed");
+	ASSERT_EQ(out.val2_probed, 0x0a0b0c0d, "val2_probed");
+	ASSERT_EQ(out.val3_probed, 0xfeed, "val3_probed");
+	ASSERT_EQ(out.val4_probed, 0xb9, "val4_probed");
+
+	test_core_autosize__destroy(skel);
+	skel = NULL;
+
+	/* now re-load with handle_signed() enabled, it should fail loading */
+	skel = test_core_autosize__open();
+	if (!ASSERT_OK_PTR(skel, "skel_open"))
+		return;
+
+	load_attr.obj = skel->obj;
+	load_attr.target_btf_path = btf_file;
+	err = bpf_object__load_xattr(&load_attr);
+	if (!ASSERT_ERR(err, "bad_prog_load"))
+		goto cleanup;
+
+cleanup:
+	if (f)
+		fclose(f);
+	if (fd >= 0)
+		close(fd);
+	remove(btf_file);
+	btf__free(btf);
+	test_core_autosize__destroy(skel);
+}

tools/testing/selftests/bpf/progs/test_core_autosize.c

+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+
+#include <linux/bpf.h>
+#include <stdint.h>
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_core_read.h>
+
+char _license[] SEC("license") = "GPL";
+
+/* fields of exactly the same size */
+struct test_struct___samesize {
+	void *ptr;
+	unsigned long long val1;
+	unsigned int val2;
+	unsigned short val3;
+	unsigned char val4;
+} __attribute((preserve_access_index));
+
+/* unsigned fields that have to be downsized by libbpf */
+struct test_struct___downsize {
+	void *ptr;
+	unsigned long val1;
+	unsigned long val2;
+	unsigned long val3;
+	unsigned long val4;
+	/* total sz: 40 */
+} __attribute__((preserve_access_index));
+
+/* fields with signed integers of wrong size, should be rejected */
+struct test_struct___signed {
+	void *ptr;
+	long val1;
+	long val2;
+	long val3;
+	long val4;
+} __attribute((preserve_access_index));
+
+/* real layout and sizes according to test's (32-bit) BTF */
+struct test_struct___real {
+	unsigned int ptr; /* can't use `void *`, it is always 8 byte in BPF target */
+	unsigned int val2;
+	unsigned long long val1;
+	unsigned short val3;
+	unsigned char val4;
+	unsigned char _pad;
+	/* total sz: 20 */
+};
+
+struct test_struct___real input = {
+	.ptr = 0x01020304,
+	.val1 = 0x1020304050607080,
+	.val2 = 0x0a0b0c0d,
+	.val3 = 0xfeed,
+	.val4 = 0xb9,
+	._pad = 0xff, /* make sure no accidental zeros are present */
+};
+
+unsigned long long ptr_samesized = 0;
+unsigned long long val1_samesized = 0;
+unsigned long long val2_samesized = 0;
+unsigned long long val3_samesized = 0;
+unsigned long long val4_samesized = 0;
+struct test_struct___real output_samesized = {};
+
+unsigned long long ptr_downsized = 0;
+unsigned long long val1_downsized = 0;
+unsigned long long val2_downsized = 0;
+unsigned long long val3_downsized = 0;
+unsigned long long val4_downsized = 0;
+struct test_struct___real output_downsized = {};
+
+unsigned long long ptr_probed = 0;
+unsigned long long val1_probed = 0;
+unsigned long long val2_probed = 0;
+unsigned long long val3_probed = 0;
+unsigned long long val4_probed = 0;
+
+unsigned long long ptr_signed = 0;
+unsigned long long val1_signed = 0;
+unsigned long long val2_signed = 0;
+unsigned long long val3_signed = 0;
+unsigned long long val4_signed = 0;
+struct test_struct___real output_signed = {};
+
+SEC("raw_tp/sys_exit")
+int handle_samesize(void *ctx)
+{
+	struct test_struct___samesize *in = (void *)&input;
+	struct test_struct___samesize *out = (void *)&output_samesized;
+
+	ptr_samesized = (unsigned long long)in->ptr;
+	val1_samesized = in->val1;
+	val2_samesized = in->val2;
+	val3_samesized = in->val3;
+	val4_samesized = in->val4;
+
+	out->ptr = in->ptr;
+	out->val1 = in->val1;
+	out->val2 = in->val2;
+	out->val3 = in->val3;
+	out->val4 = in->val4;
+
+	return 0;
+}
+
+SEC("raw_tp/sys_exit")
+int handle_downsize(void *ctx)
+{
+	struct test_struct___downsize *in = (void *)&input;
+	struct test_struct___downsize *out = (void *)&output_downsized;
+
+	ptr_downsized = (unsigned long long)in->ptr;
+	val1_downsized = in->val1;
+	val2_downsized = in->val2;
+	val3_downsized = in->val3;
+	val4_downsized = in->val4;
+
+	out->ptr = in->ptr;
+	out->val1 = in->val1;
+	out->val2 = in->val2;
+	out->val3 = in->val3;
+	out->val4 = in->val4;
+
+	return 0;
+}
+
+SEC("raw_tp/sys_enter")
+int handle_probed(void *ctx)
+{
+	struct test_struct___downsize *in = (void *)&input;
+	__u64 tmp;
+
+	tmp = 0;
+	bpf_core_read(&tmp, bpf_core_field_size(in->ptr), &in->ptr);
+	ptr_probed = tmp;
+
+	tmp = 0;
+	bpf_core_read(&tmp, bpf_core_field_size(in->val1), &in->val1);
+	val1_probed = tmp;
+
+	tmp = 0;
+	bpf_core_read(&tmp, bpf_core_field_size(in->val2), &in->val2);
+	val2_probed = tmp;
+
+	tmp = 0;
+	bpf_core_read(&tmp, bpf_core_field_size(in->val3), &in->val3);
+	val3_probed = tmp;
+
+	tmp = 0;
+	bpf_core_read(&tmp, bpf_core_field_size(in->val4), &in->val4);
+	val4_probed = tmp;
+
+	return 0;
+}
+
+SEC("raw_tp/sys_enter")
+int handle_signed(void *ctx)
+{
+	struct test_struct___signed *in = (void *)&input;
+	struct test_struct___signed *out = (void *)&output_signed;
+
+	val2_signed = in->val2;
+	val3_signed = in->val3;
+	val4_signed = in->val4;
+
+	out->val2= in->val2;
+	out->val3= in->val3;
+	out->val4= in->val4;
+
+	return 0;
+}