[PATCH] bpf: remove trunctaiton warning in BPF_LD_IMM64_RAW

[PATCH] bpf: remove trunctaiton warning in BPF_LD_IMM64_RAW

[Ben Dooks]

When building W=1, the following warningds get generated by lib/test_bpf.c
due to the BPF_LD_IMM64_RAW() truncating a 64bit to 32bit value without
a specific mask to do so.

lib/test_bpf.c:6441:25: warning: cast truncates bits from constant value (ffffffff0000 becomes ffff0000)
lib/test_bpf.c:6442:25: warning: cast truncates bits from constant value (ffffffff0000 becomes ffff0000)
lib/test_bpf.c:6473:25: warning: cast truncates bits from constant value (123456789abcdef becomes 89abcdef)
lib/test_bpf.c:6489:25: warning: cast truncates bits from constant value (123456789abcdef becomes 89abcdef)
lib/test_bpf.c:6490:25: warning: cast truncates bits from constant value (123456780a0c0e0 becomes 80a0c0e0)
lib/test_bpf.c:6599:25: warning: cast truncates bits from constant value (123456789abcdef becomes 89abcdef)
lib/test_bpf.c:6637:25: warning: cast truncates bits from constant value (ffffffff0000 becomes ffff0000)

(more warnings truncated)

Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>
---
 include/linux/filter.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/linux/filter.h b/include/linux/filter.h
index c99bc3df2d28..c1ff6511f365 100644
--- a/include/linux/filter.h
+++ b/include/linux/filter.h
@@ -238,7 +238,7 @@ static inline bool insn_is_zext(const struct bpf_insn *insn)
 		.dst_reg = DST,					\
 		.src_reg = SRC,					\
 		.off   = 0,					\
-		.imm   = (__u32) (IMM) }),			\
+		.imm   = (__u32) ((IMM) & (__u64)0xffffffff) }), \
 	((struct bpf_insn) {					\
 		.code  = 0, /* zero is reserved opcode */	\
 		.dst_reg = 0,					\
-- 
2.37.2.352.g3c44437643

Re: [PATCH] bpf: remove trunctaiton warning in BPF_LD_IMM64_RAW

[Stanislav Fomichev]

On 03/14, Ben Dooks wrote:
> When building W=1, the following warningds get generated by lib/test_bpf.c
> due to the BPF_LD_IMM64_RAW() truncating a 64bit to 32bit value without
> a specific mask to do so.
> 
> lib/test_bpf.c:6441:25: warning: cast truncates bits from constant value (ffffffff0000 becomes ffff0000)
> lib/test_bpf.c:6442:25: warning: cast truncates bits from constant value (ffffffff0000 becomes ffff0000)
> lib/test_bpf.c:6473:25: warning: cast truncates bits from constant value (123456789abcdef becomes 89abcdef)
> lib/test_bpf.c:6489:25: warning: cast truncates bits from constant value (123456789abcdef becomes 89abcdef)
> lib/test_bpf.c:6490:25: warning: cast truncates bits from constant value (123456780a0c0e0 becomes 80a0c0e0)
> lib/test_bpf.c:6599:25: warning: cast truncates bits from constant value (123456789abcdef becomes 89abcdef)
> lib/test_bpf.c:6637:25: warning: cast truncates bits from constant value (ffffffff0000 becomes ffff0000)
> 
> (more warnings truncated)
> 
> Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>

I wonder how many other places (in the kernel) we have where we silently
cast u64 to u32.. Surprised that's the only warning you get :-)

Acked-by: Stanislav Fomichev <sdf@google.com>

Re: [PATCH] bpf: remove trunctaiton warning in BPF_LD_IMM64_RAW

[Alexei Starovoitov]

On Thu, Mar 14, 2024 at 3:32 AM Ben Dooks <ben.dooks@codethink.co.uk> wrote:
>
> When building W=1, the following warningds get generated by lib/test_bpf.c
> due to the BPF_LD_IMM64_RAW() truncating a 64bit to 32bit value without
> a specific mask to do so.
>
> lib/test_bpf.c:6441:25: warning: cast truncates bits from constant value (ffffffff0000 becomes ffff0000)
> lib/test_bpf.c:6442:25: warning: cast truncates bits from constant value (ffffffff0000 becomes ffff0000)
> lib/test_bpf.c:6473:25: warning: cast truncates bits from constant value (123456789abcdef becomes 89abcdef)
> lib/test_bpf.c:6489:25: warning: cast truncates bits from constant value (123456789abcdef becomes 89abcdef)
> lib/test_bpf.c:6490:25: warning: cast truncates bits from constant value (123456780a0c0e0 becomes 80a0c0e0)
> lib/test_bpf.c:6599:25: warning: cast truncates bits from constant value (123456789abcdef becomes 89abcdef)
> lib/test_bpf.c:6637:25: warning: cast truncates bits from constant value (ffffffff0000 becomes ffff0000)
>
> (more warnings truncated)
>
> Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>
> ---
>  include/linux/filter.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/include/linux/filter.h b/include/linux/filter.h
> index c99bc3df2d28..c1ff6511f365 100644
> --- a/include/linux/filter.h
> +++ b/include/linux/filter.h
> @@ -238,7 +238,7 @@ static inline bool insn_is_zext(const struct bpf_insn *insn)
>                 .dst_reg = DST,                                 \
>                 .src_reg = SRC,                                 \
>                 .off   = 0,                                     \
> -               .imm   = (__u32) (IMM) }),                      \
> +               .imm   = (__u32) ((IMM) & (__u64)0xffffffff) }), \

What compiler is this?
I don't think it's a good idea to uglify this macro just
to shut up a compiler.
As Stan said there are plenty of other places where
such _explicit_ u32 cast is used.
We're not going to mess with all of them.

pw-bot: cr