Re: [PATCH bpf-next 3/3] bpf: arraymap: Skip boundscheck during inlining when possible

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Mon, 2025-01-20 at 21:35 -0700, Daniel Xu wrote:

[...]

Hi Daniel,

> @@ -221,11 +221,13 @@ static int array_map_gen_lookup(struct bpf_map *map,
>  
>  	*insn++ = BPF_ALU64_IMM(BPF_ADD, map_ptr, offsetof(struct bpf_array, value));
>  	*insn++ = BPF_LDX_MEM(BPF_W, ret, index, 0);
> -	if (!map->bypass_spec_v1) {
> -		*insn++ = BPF_JMP_IMM(BPF_JGE, ret, map->max_entries, 4);
> -		*insn++ = BPF_ALU32_IMM(BPF_AND, ret, array->index_mask);
> -	} else {
> -		*insn++ = BPF_JMP_IMM(BPF_JGE, ret, map->max_entries, 3);
> +	if (!inbounds) {
> +		if (!map->bypass_spec_v1) {
> +			*insn++ = BPF_JMP_IMM(BPF_JGE, ret, map->max_entries, 4);
> +			*insn++ = BPF_ALU32_IMM(BPF_AND, ret, array->index_mask);
> +		} else {
> +			*insn++ = BPF_JMP_IMM(BPF_JGE, ret, map->max_entries, 3);
> +		}
>  	}
>  
>  	if (is_power_of_2(elem_size)) {

Note that below this hunk there is the following code:

	*insn++ = BPF_JMP_IMM(BPF_JA, 0, 0, 1);
	*insn++ = BPF_MOV64_IMM(ret, 0);
	return insn - insn_buf;

This part becomes redundant after your change. E.g. here is jit
listing for an_array_with_a_32bit_constant_0_no_nullness selftest:

JITED:
=============
func #0:
0:	f3 0f 1e fa                         	endbr64
4:	0f 1f 44 00 00                      	nopl	(%rax,%rax)
9:	0f 1f 00                            	nopl	(%rax)
c:	55                                  	pushq	%rbp
d:	48 89 e5                            	movq	%rsp, %rbp
10:	f3 0f 1e fa                         	endbr64
14:	48 81 ec 08 00 00 00                	subq	$0x8, %rsp
1b:	31 ff                               	xorl	%edi, %edi
1d:	89 7d fc                            	movl	%edi, -0x4(%rbp)
20:	48 89 ee                            	movq	%rbp, %rsi
23:	48 83 c6 fc                         	addq	$-0x4, %rsi
27:	48 bf 00 70 58 06 81 88 ff ff       	movabsq	$-0x777ef9a79000, %rdi
31:	48 81 c7 d8 01 00 00                	addq	$0x1d8, %rdi
38:	8b 46 00                            	movl	(%rsi), %eax
3b:	48 6b c0 30                         	imulq	$0x30, %rax, %rax
3f:	48 01 f8                            	addq	%rdi, %rax
42:	eb 02                               	jmp	L0             //
44:	31 c0                               	xorl	%eax, %eax     // never executed
46:	bf 04 00 00 00                      L0:	movl	$0x4, %edi     //
4b:	89 78 00                            	movl	%edi, (%rax)
4e:	b8 04 00 00 00                      	movl	$0x4, %eax
53:	c9                                  	leave
54:	e9 22 38 50 c3                      	jmp	0xffffffffc350387b

Also note that there are __arch_x86_64 and __jited tags for selftests.
These allow to match against disassembly of the generated binary code.
(See verifier_tailcall_jit.c for an example).
I think it would be good to add a test matching jited code for this feature.

[...]






[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux