Re: [PATCH bpf-next v4 06/11] bpf: Add BPF_FETCH field / create atomic_fetch_add instruction

From: Yonghong Song <yhs@fb.com>
To: Brendan Jackman <jackmanb@google.com>, <bpf@vger.kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	KP Singh <kpsingh@chromium.org>,
	Florent Revest <revest@chromium.org>,
	<linux-kernel@vger.kernel.org>, Jann Horn <jannh@google.com>
Subject: Re: [PATCH bpf-next v4 06/11] bpf: Add BPF_FETCH field / create atomic_fetch_add instruction
Date: Mon, 7 Dec 2020 17:41:05 -0800	[thread overview]
Message-ID: <4163e34b-754a-5607-c28a-4c575a2cc6e5@fb.com> (raw)
In-Reply-To: <20201207160734.2345502-7-jackmanb@google.com>

On 12/7/20 8:07 AM, Brendan Jackman wrote:
> The BPF_FETCH field can be set in bpf_insn.imm, for BPF_ATOMIC
> instructions, in order to have the previous value of the
> atomically-modified memory location loaded into the src register
> after an atomic op is carried out.
> 
> Suggested-by: Yonghong Song <yhs@fb.com>
> Signed-off-by: Brendan Jackman <jackmanb@google.com>
> ---
>   arch/x86/net/bpf_jit_comp.c    |  4 ++++
>   include/linux/filter.h         |  1 +
>   include/uapi/linux/bpf.h       |  3 +++
>   kernel/bpf/core.c              | 13 +++++++++++++
>   kernel/bpf/disasm.c            |  7 +++++++
>   kernel/bpf/verifier.c          | 33 ++++++++++++++++++++++++---------
>   tools/include/linux/filter.h   | 11 +++++++++++
>   tools/include/uapi/linux/bpf.h |  3 +++
>   8 files changed, 66 insertions(+), 9 deletions(-)
> 
> diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c
[...]

> index f345f12c1ff8..4e0100ba52c2 100644
> --- a/tools/include/linux/filter.h
> +++ b/tools/include/linux/filter.h
> @@ -173,6 +173,7 @@
>    * Atomic operations:
>    *
>    *   BPF_ADD                  *(uint *) (dst_reg + off16) += src_reg
> + *   BPF_ADD | BPF_FETCH      src_reg = atomic_fetch_add(dst_reg + off16, src_reg);
>    */
>   
>   #define BPF_ATOMIC64(OP, DST, SRC, OFF)				\
> @@ -201,6 +202,16 @@
>   		.off   = OFF,					\
>   		.imm   = BPF_ADD })
>   
> +/* Atomic memory add with fetch, src_reg = atomic_fetch_add(dst_reg + off, src_reg); */
> +
> +#define BPF_ATOMIC_FETCH_ADD(SIZE, DST, SRC, OFF)		\
> +	((struct bpf_insn) {					\
> +		.code  = BPF_STX | BPF_SIZE(SIZE) | BPF_ATOMIC,	\
> +		.dst_reg = DST,					\
> +		.src_reg = SRC,					\
> +		.off   = OFF,					\
> +		.imm   = BPF_ADD | BPF_FETCH })

Not sure whether it is a good idea or not to fold this into BPF_ATOMIC 
macro. At least you can define BPF_ATOMIC macro and
#define BPF_ATOMIC_FETCH_ADD(SIZE, DST, SRC, OFF)		\
     BPF_ATOMIC(SIZE, DST, SRC, OFF, BPF_ADD | BPF_FETCH)

to avoid too many code duplications?

> +
>   /* Memory store, *(uint *) (dst_reg + off16) = imm32 */
>   
>   #define BPF_ST_MEM(SIZE, DST, OFF, IMM)				\
> diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h
> index 98161e2d389f..d5389119291e 100644
> --- a/tools/include/uapi/linux/bpf.h
> +++ b/tools/include/uapi/linux/bpf.h
> @@ -44,6 +44,9 @@
>   #define BPF_CALL	0x80	/* function call */
>   #define BPF_EXIT	0x90	/* function return */
>   
> +/* atomic op type fields (stored in immediate) */
> +#define BPF_FETCH	0x01	/* fetch previous value into src reg */
> +
>   /* Register numbers */
>   enum {
>   	BPF_REG_0 = 0,
> 