Merge branch 'bpf-prep-patches-for-static-stack-liveness'

Alexei Starovoitov says:

====================
bpf: Prep patches for static stack liveness.

v4->v5:
- minor test fixup

v3->v4:
- fixed invalid recursion detection when calback is called multiple times

v3: https://lore.kernel.org/bpf/20260402212856.86606-1-alexei.starovoitov@gmail.com/

v2->v3:
- added recursive call detection
- fixed ubsan warning
- removed double declaration in the header
- added Acks

v2: https://lore.kernel.org/bpf/20260402061744.10885-1-alexei.starovoitov@gmail.com/

v1->v2:
. fixed bugs spotted by Eduard, Mykyta, claude and gemini
. fixed selftests that were failing in unpriv
. gemini(sashiko) found several precision improvements in patch 6,
  but they made no difference in real programs.

v1: https://lore.kernel.org/bpf/20260401021635.34636-1-alexei.starovoitov@gmail.com/
First 6 prep patches for static stack liveness.

. do src/dst_reg validation early and remove defensive checks

. sort subprog in topo order. We wanted to do this long ago
  to process global subprogs this way and in other cases.

. Add constant folding pass that computes map_ptr, subprog_idx,
  loads from readonly maps, and other constants that fit into 32-bit

. Use these constants to eliminate dead code. Replace predicted
  conditional branches with "jmp always". That reduces JIT prog size.

. Add two helpers that return access size from their arguments.
====================

Link: https://patch.msgid.link/20260403024422.87231-1-alexei.starovoitov@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

7 files changed, 252 insertions, 15 deletions

diff --git a/tools/testing/selftests/bpf/prog_tests/verifier.c b/tools/testing/selftests/bpf/prog_tests/verifier.c
index bcf01cb4cfe4..1ac366fd4dae 100644
--- a/tools/testing/selftests/bpf/prog_tests/verifier.c
+++ b/tools/testing/selftests/bpf/prog_tests/verifier.c
@@ -93,6 +93,7 @@
 #include "verifier_stack_ptr.skel.h"
 #include "verifier_store_release.skel.h"
 #include "verifier_subprog_precision.skel.h"
+#include "verifier_subprog_topo.skel.h"
 #include "verifier_subreg.skel.h"
 #include "verifier_tailcall.skel.h"
 #include "verifier_tailcall_jit.skel.h"
@@ -238,6 +239,7 @@ void test_verifier_spin_lock(void)            { RUN(verifier_spin_lock); }
 void test_verifier_stack_ptr(void)            { RUN(verifier_stack_ptr); }
 void test_verifier_store_release(void)        { RUN(verifier_store_release); }
 void test_verifier_subprog_precision(void)    { RUN(verifier_subprog_precision); }
+void test_verifier_subprog_topo(void)        { RUN(verifier_subprog_topo); }
 void test_verifier_subreg(void)               { RUN(verifier_subreg); }
 void test_verifier_tailcall(void)             { RUN(verifier_tailcall); }
 void test_verifier_tailcall_jit(void)         { RUN(verifier_tailcall_jit); }
diff --git a/tools/testing/selftests/bpf/progs/verifier_loops1.c b/tools/testing/selftests/bpf/progs/verifier_loops1.c
index fbdde80e7b90..d248ce877f14 100644
--- a/tools/testing/selftests/bpf/progs/verifier_loops1.c
+++ b/tools/testing/selftests/bpf/progs/verifier_loops1.c
@@ -138,8 +138,7 @@ l0_%=:	exit;						\
 SEC("tracepoint")
 __description("bounded recursion")
 __failure
-/* verifier limitation in detecting max stack depth */
-__msg("the call stack of 8 frames is too deep !")
+__msg("recursive call from")
 __naked void bounded_recursion(void)
 {
 	asm volatile ("					\
diff --git a/tools/testing/selftests/bpf/progs/verifier_scalar_ids.c b/tools/testing/selftests/bpf/progs/verifier_scalar_ids.c
index 58c7704d61cd..a5b8753ce52c 100644
--- a/tools/testing/selftests/bpf/progs/verifier_scalar_ids.c
+++ b/tools/testing/selftests/bpf/progs/verifier_scalar_ids.c
@@ -592,10 +592,10 @@ __naked void check_ids_in_regsafe_2(void)
  */
 SEC("socket")
 __success __log_level(2)
-__msg("11: (1d) if r3 == r4 goto pc+0")
+__msg("14: (1d) if r3 == r4 goto pc+0")
 __msg("frame 0: propagating r3,r4")
-__msg("11: safe")
-__msg("processed 15 insns")
+__msg("14: safe")
+__msg("processed 18 insns")
 __flag(BPF_F_TEST_STATE_FREQ)
 __naked void no_scalar_id_for_const(void)
 {
@@ -605,6 +605,7 @@ __naked void no_scalar_id_for_const(void)
 	"if r0 > 7 goto l0_%=;"
 	/* possibly generate same scalar ids for r3 and r4 */
 	"r1 = 0;"
+	"r1 ^= r1;" /* prevent bpf_prune_dead_branches from folding the branch */
 	"r1 = r1;"
 	"r3 = r1;"
 	"r4 = r1;"
@@ -612,7 +613,9 @@ __naked void no_scalar_id_for_const(void)
 "l0_%=:"
 	/* possibly generate different scalar ids for r3 and r4 */
 	"r1 = 0;"
+	"r1 ^= r1;"
 	"r2 = 0;"
+	"r2 ^= r2;"
 	"r3 = r1;"
 	"r4 = r2;"
 "l1_%=:"
@@ -628,10 +631,10 @@ __naked void no_scalar_id_for_const(void)
 /* Same as no_scalar_id_for_const() but for 32-bit values */
 SEC("socket")
 __success __log_level(2)
-__msg("11: (1e) if w3 == w4 goto pc+0")
+__msg("14: (1e) if w3 == w4 goto pc+0")
 __msg("frame 0: propagating r3,r4")
-__msg("11: safe")
-__msg("processed 15 insns")
+__msg("14: safe")
+__msg("processed 18 insns")
 __flag(BPF_F_TEST_STATE_FREQ)
 __naked void no_scalar_id_for_const32(void)
 {
@@ -641,6 +644,7 @@ __naked void no_scalar_id_for_const32(void)
 	"if r0 > 7 goto l0_%=;"
 	/* possibly generate same scalar ids for r3 and r4 */
 	"w1 = 0;"
+	"w1 ^= w1;" /* prevent bpf_prune_dead_branches from folding the branch */
 	"w1 = w1;"
 	"w3 = w1;"
 	"w4 = w1;"
@@ -648,11 +652,13 @@ __naked void no_scalar_id_for_const32(void)
 "l0_%=:"
 	/* possibly generate different scalar ids for r3 and r4 */
 	"w1 = 0;"
+	"w1 ^= w1;"
 	"w2 = 0;"
+	"w2 ^= w2;"
 	"w3 = w1;"
 	"w4 = w2;"
 "l1_%=:"
-	/* predictable jump, marks r1 and r2 precise */
+	/* predictable jump, marks r3 and r4 precise */
 	"if w3 == w4 goto +0;"
 	"r0 = 0;"
 	"exit;"
diff --git a/tools/testing/selftests/bpf/progs/verifier_subprog_topo.c b/tools/testing/selftests/bpf/progs/verifier_subprog_topo.c
new file mode 100644
index 000000000000..e2b9d14bbc3d
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/verifier_subprog_topo.c
@@ -0,0 +1,226 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2026 Meta Platforms, Inc. and affiliates. */
+
+#include <linux/bpf.h>
+#include <bpf/bpf_helpers.h>
+#include "bpf_misc.h"
+
+/* linear chain main -> A -> B */
+__naked __noinline __used
+static unsigned long linear_b(void)
+{
+	asm volatile (
+		"r0 = 42;"
+		"exit;"
+	);
+}
+
+__naked __noinline __used
+static unsigned long linear_a(void)
+{
+	asm volatile (
+		"call linear_b;"
+		"exit;"
+	);
+}
+
+SEC("?raw_tp")
+__success __log_level(2)
+__msg("topo_order[0] = linear_b")
+__msg("topo_order[1] = linear_a")
+__msg("topo_order[2] = topo_linear")
+__naked int topo_linear(void)
+{
+	asm volatile (
+		"call linear_a;"
+		"exit;"
+	);
+}
+
+/* diamond main -> A, main -> B, A -> C, B -> C */
+__naked __noinline __used
+static unsigned long diamond_c(void)
+{
+	asm volatile (
+		"r0 = 1;"
+		"exit;"
+	);
+}
+
+__naked __noinline __used
+static unsigned long diamond_b(void)
+{
+	asm volatile (
+		"call diamond_c;"
+		"exit;"
+	);
+}
+
+__naked __noinline __used
+static unsigned long diamond_a(void)
+{
+	asm volatile (
+		"call diamond_c;"
+		"exit;"
+	);
+}
+
+SEC("?raw_tp")
+__success __log_level(2)
+__msg("topo_order[0] = diamond_c")
+__msg("topo_order[3] = topo_diamond")
+__naked int topo_diamond(void)
+{
+	asm volatile (
+		"call diamond_a;"
+		"call diamond_b;"
+		"exit;"
+	);
+}
+
+/* main -> global_a (global) -> static_leaf (static, leaf) */
+__naked __noinline __used
+static unsigned long static_leaf(void)
+{
+	asm volatile (
+		"r0 = 7;"
+		"exit;"
+	);
+}
+
+__noinline __used
+int global_a(int x)
+{
+	return static_leaf();
+}
+
+SEC("?raw_tp")
+__success __log_level(2)
+__msg("topo_order[0] = static_leaf")
+__msg("topo_order[1] = global_a")
+__msg("topo_order[2] = topo_mixed")
+__naked int topo_mixed(void)
+{
+	asm volatile (
+		"r1 = 0;"
+		"call global_a;"
+		"exit;"
+	);
+}
+
+/*
+ * shared static callee from global and main:
+ * main -> shared_leaf (static)
+ * main -> global_b (global) -> shared_leaf (static)
+ */
+__naked __noinline __used
+static unsigned long shared_leaf(void)
+{
+	asm volatile (
+		"r0 = 99;"
+		"exit;"
+	);
+}
+
+__noinline __used
+int global_b(int x)
+{
+	return shared_leaf();
+}
+
+SEC("?raw_tp")
+__success __log_level(2)
+__msg("topo_order[0] = shared_leaf")
+__msg("topo_order[1] = global_b")
+__msg("topo_order[2] = topo_shared")
+__naked int topo_shared(void)
+{
+	asm volatile (
+		"call shared_leaf;"
+		"r1 = 0;"
+		"call global_b;"
+		"exit;"
+	);
+}
+
+/* duplicate calls to the same subprog */
+__naked __noinline __used
+static unsigned long dup_leaf(void)
+{
+	asm volatile (
+		"r0 = 0;"
+		"exit;"
+	);
+}
+
+SEC("?raw_tp")
+__success __log_level(2)
+__msg("topo_order[0] = dup_leaf")
+__msg("topo_order[1] = topo_dup_calls")
+__naked int topo_dup_calls(void)
+{
+	asm volatile (
+		"call dup_leaf;"
+		"call dup_leaf;"
+		"exit;"
+	);
+}
+
+/* main calls bpf_loop() with loop_cb as the callback */
+static int loop_cb(int idx, void *ctx)
+{
+	return 0;
+}
+
+SEC("?raw_tp")
+__success __log_level(2)
+__msg("topo_order[0] = loop_cb")
+__msg("topo_order[1] = topo_loop_cb")
+int topo_loop_cb(void)
+{
+	bpf_loop(1, loop_cb, NULL, 0);
+	return 0;
+}
+
+/*
+ * bpf_loop callback calling another subprog
+ * main -> bpf_loop(callback=loop_cb2) -> loop_cb2 -> loop_cb2_leaf
+ */
+__naked __noinline __used
+static unsigned long loop_cb2_leaf(void)
+{
+	asm volatile (
+		"r0 = 0;"
+		"exit;"
+	);
+}
+
+static int loop_cb2(int idx, void *ctx)
+{
+	return loop_cb2_leaf();
+}
+
+SEC("?raw_tp")
+__success __log_level(2)
+__msg("topo_order[0] = loop_cb2_leaf")
+__msg("topo_order[1] = loop_cb2")
+__msg("topo_order[2] = topo_loop_cb_chain")
+int topo_loop_cb_chain(void)
+{
+	bpf_loop(1, loop_cb2, NULL, 0);
+	return 0;
+}
+
+/* no calls (single subprog) */
+SEC("?raw_tp")
+__success __log_level(2)
+__msg("topo_order[0] = topo_no_calls")
+__naked int topo_no_calls(void)
+{
+	asm volatile (
+		"r0 = 0;"
+		"exit;"
+	);
+}
+
+char _license[] SEC("license") = "GPL";
diff --git a/tools/testing/selftests/bpf/progs/verifier_unpriv.c b/tools/testing/selftests/bpf/progs/verifier_unpriv.c
index 8ee1243e62a8..c16f8382cf17 100644
--- a/tools/testing/selftests/bpf/progs/verifier_unpriv.c
+++ b/tools/testing/selftests/bpf/progs/verifier_unpriv.c
@@ -584,7 +584,7 @@ __naked void alu32_mov_u32_const(void)
 {
 	asm volatile ("					\
 	w7 = 0;						\
-	w7 &= 1;					\
+	w7 ^= w7;					\
 	w0 = w7;					\
 	if r0 == 0 goto l0_%=;				\
 	r0 = *(u64*)(r7 + 0);				\
@@ -894,7 +894,9 @@ __naked void unpriv_spectre_v1_and_v4_simple(void)
 {
 	asm volatile ("					\
 	r8 = 0;						\
+	r8 ^= r8;					\
 	r9 = 0;						\
+	r9 ^= r9;					\
 	r0 = r10;					\
 	r1 = 0;						\
 	r2 = r10;					\
@@ -932,7 +934,9 @@ __naked void unpriv_ldimm64_spectre_v1_and_v4_simple(void)
 {
 	asm volatile ("					\
 	r8 = 0;						\
+	r8 ^= r8;					\
 	r9 = 0;						\
+	r9 ^= r9;					\
 	r0 = r10;					\
 	r1 = 0;						\
 	r2 = r10;					\
diff --git a/tools/testing/selftests/bpf/verifier/calls.c b/tools/testing/selftests/bpf/verifier/calls.c
index 29e57f0e56c3..c3164b9b2be5 100644
--- a/tools/testing/selftests/bpf/verifier/calls.c
+++ b/tools/testing/selftests/bpf/verifier/calls.c
@@ -455,7 +455,7 @@
 	BPF_EXIT_INSN(),
 	},
 	.prog_type = BPF_PROG_TYPE_TRACEPOINT,
-	.errstr = "the call stack of 9 frames is too deep",
+	.errstr = "recursive call",
 	.result = REJECT,
 },
 {
@@ -812,7 +812,7 @@
 	BPF_EXIT_INSN(),
 	},
 	.prog_type = BPF_PROG_TYPE_TRACEPOINT,
-	.errstr = "the call stack of 9 frames is too deep",
+	.errstr = "recursive call",
 	.result = REJECT,
 },
 {
@@ -824,7 +824,7 @@
 	BPF_EXIT_INSN(),
 	},
 	.prog_type = BPF_PROG_TYPE_TRACEPOINT,
-	.errstr = "the call stack of 9 frames is too deep",
+	.errstr = "recursive call",
 	.result = REJECT,
 },
 {
diff --git a/tools/testing/selftests/bpf/verifier/junk_insn.c b/tools/testing/selftests/bpf/verifier/junk_insn.c
index 89d690f1992a..7d10b0a48f51 100644
--- a/tools/testing/selftests/bpf/verifier/junk_insn.c
+++ b/tools/testing/selftests/bpf/verifier/junk_insn.c
@@ -28,7 +28,7 @@
 {
 	"junk insn4",
 	.insns = {
-	BPF_RAW_INSN(-1, -1, -1, -1, -1),
+	BPF_RAW_INSN(-1, 0, 0, -1, -1),
 	BPF_EXIT_INSN(),
 	},
 	.errstr = "unknown opcode ff",
@@ -37,7 +37,7 @@
 {
 	"junk insn5",
 	.insns = {
-	BPF_RAW_INSN(0x7f, -1, -1, -1, -1),
+	BPF_RAW_INSN(0x7f, 0, 0, -1, -1),
 	BPF_EXIT_INSN(),
 	},
 	.errstr = "BPF_ALU uses reserved fields",