bpf: Do not zero-extend kfunc return values

In BPF all global functions, and BPF helpers return a 64-bit
value. For kfunc calls, this is not the case, and they can return
e.g. 32-bit values.

The return register R0 for kfuncs calls can therefore be marked as
subreg_def != DEF_NOT_SUBREG. In general, if a register is marked with
subreg_def != DEF_NOT_SUBREG, some archs (where bpf_jit_needs_zext()
returns true) require the verifier to insert explicit zero-extension
instructions.

For kfuncs calls, however, the caller should do sign/zero extension
for return values. In other words, the compiler is responsible to
insert proper instructions, not the verifier.

An example, provided by Yonghong Song:

$ cat t.c
extern unsigned foo(void);
unsigned bar1(void) {
     return foo();
}
unsigned bar2(void) {
     if (foo()) return 10; else return 20;
}

$ clang -target bpf -mcpu=v3 -O2 -c t.c && llvm-objdump -d t.o
t.o:    file format elf64-bpf

Disassembly of section .text:

0000000000000000 <bar1>:
	0:       85 10 00 00 ff ff ff ff call -0x1
	1:       95 00 00 00 00 00 00 00 exit

0000000000000010 <bar2>:
	2:       85 10 00 00 ff ff ff ff call -0x1
	3:       bc 01 00 00 00 00 00 00 w1 = w0
	4:       b4 00 00 00 14 00 00 00 w0 = 0x14
	5:       16 01 01 00 00 00 00 00 if w1 == 0x0 goto +0x1 <LBB1_2>
	6:       b4 00 00 00 0a 00 00 00 w0 = 0xa

0000000000000038 <LBB1_2>:
	7:       95 00 00 00 00 00 00 00 exit

If the return value of 'foo()' is used in the BPF program, the proper
zero-extension will be done.

Currently, the verifier correctly marks, say, a 32-bit return value as
subreg_def != DEF_NOT_SUBREG, but will fail performing the actual
zero-extension, due to a verifier bug in
opt_subreg_zext_lo32_rnd_hi32(). load_reg is not properly set to R0,
and the following path will be taken:

		if (WARN_ON(load_reg == -1)) {
			verbose(env, "verifier bug. zext_dst is set, but no reg is defined\n");
			return -EFAULT;
		}

A longer discussion from v1 can be found in the link below.

Correct the verifier by avoiding doing explicit zero-extension of R0
for kfunc calls. Note that R0 will still be marked as a sub-register
for return values smaller than 64-bit.

Fixes: 83a2881903 ("bpf: Account for BPF_FETCH in insn_has_def32()")
Link: https://lore.kernel.org/bpf/20221202103620.1915679-1-bjorn@kernel.org/
Suggested-by: Yonghong Song <yhs@meta.com>
Signed-off-by: Björn Töpel <bjorn@rivosinc.com>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/r/20221207103540.396496-1-bjorn@kernel.org
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

kernel/bpf/verifier.c

@@ -14836,6 +14836,10 @@ static int opt_subreg_zext_lo32_rnd_hi32(struct bpf_verifier_env *env,
 		if (!bpf_jit_needs_zext() && !is_cmpxchg_insn(&insn))
 			continue;
 
+		/* Zero-extension is done by the caller. */
+		if (bpf_pseudo_kfunc_call(&insn))
+			continue;
+
 		if (WARN_ON(load_reg == -1)) {
 			verbose(env, "verifier bug. zext_dst is set, but no reg is defined\n");
 			return -EFAULT;