[PATCH bpf-next v3 1/2] bpf, riscv: Add BPF stack arguments support for RV64 JIT

Pu Lehui pulehui at huawei.com
Mon Aug 17 20:48:14 PDT 2026


On 2026/8/17 9:52, Pu Lehui wrote:
> Hi Feng Jiang,
> 
> On 2026/8/14 9:29, Feng Jiang wrote:
>> Add bpf_jit_supports_stack_args() for the RV64 JIT so BPF subprograms
>> and kfuncs can receive more than 5 arguments via the stack
>> (BPF_REG_PARAMS / r11).
>>
>> For BPF-to-BPF calls the caller writes outgoing arguments at the
>> bottom of its frame (SP-relative). The callee reads them with
>> FP-relative loads. Its FP is set to the caller SP in the prologue,
>> so the offsets match.
>>
>> The RISC-V ABI puts arguments 6-8 in A5-A7 and arguments 9+ at
>> SP+0. Before each kfunc call, load arguments 6-8 from the outgoing
>> area into A5-A7 and copy any remaining arguments down so argument 9
>> lands at SP+0.
> 
> bpf2bpf calls pass extra arguments entirely on the stack, whereas kfuncs 
> pass them via a mix of registers and the stack. This inconsistency not 
> only wastes stack slots in kfunc scenarios but is also error-prone and 
> easy to overlook during maintenance.
> 
> To make argument passing more straightforward and consistent, I think we 
> can align it with the standard riscv calling convention: use a0-a7 for 
> the first 8 arguments and pass any remaining arguments on the stack.
> 
> To support this, the current bpf-to-rv register mapping needs to be 
> revised:
> 
> T0 -> BPF_REG_0 (ftrace use T0 too, but it might not be an issue / needs 
> verification)
> T6 -> TCC
> T7 -> BPF_REG_AX (rarely used, so let's move to t7)
> 
> Note that this remapping affects multiple places and requires careful 
> auditing.

Discussed offline with Feng Jiang; I will take over refactoring this 
patch and post an updated version once it's ready.

> 
>>
>> A5 (BPF_REG_0), A6 (TCC) and A7 are safe to clobber here: R0 is not
>> live before a call, TCC is backed up on the stack, and A7 is unused
>> by the JIT.
> 
> 
> 
>>
>> Limit the existing kfunc sign-extension loop to MAX_BPF_FUNC_REG_ARGS
>> iterations; otherwise idx >= 5 resolves to S1-S5 (BPF_R6-FP) and
>> corrupts callee-saved registers.
>>
>> Signed-off-by: Feng Jiang <jiangfeng at kylinos.cn>
>> ---
>>   arch/riscv/net/bpf_jit.h        |  1 +
>>   arch/riscv/net/bpf_jit_comp64.c | 79 
>> +++++++++++++++++++++++++++++++++++++++--
>>   arch/riscv/net/bpf_jit_core.c   |  4 +++
>>   3 files changed, 81 insertions(+), 3 deletions(-)
>>
>> diff --git a/arch/riscv/net/bpf_jit.h b/arch/riscv/net/bpf_jit.h
>> index 419b9d795f2a..9eb4e149505d 100644
>> --- a/arch/riscv/net/bpf_jit.h
>> +++ b/arch/riscv/net/bpf_jit.h
>> @@ -82,6 +82,7 @@ struct rv_jit_context {
>>       unsigned long flags;
>>       int stack_size;
>>       int tcc_offset;
>> +    u16 stack_arg_size;
>>       u64 arena_vm_start;
>>       u64 user_vm_start;
>>   };
>> diff --git a/arch/riscv/net/bpf_jit_comp64.c 
>> b/arch/riscv/net/bpf_jit_comp64.c
>> index 6b9972b07c1b..58cb3e5ff6b4 100644
>> --- a/arch/riscv/net/bpf_jit_comp64.c
>> +++ b/arch/riscv/net/bpf_jit_comp64.c
>> @@ -1815,18 +1815,43 @@ int bpf_jit_emit_insn(const struct bpf_insn 
>> *insn, struct rv_jit_context *ctx,
>>           if (insn->src_reg == BPF_PSEUDO_KFUNC_CALL) {
>>               const struct btf_func_model *fm;
>> -            int idx;
>> +            int idx, nargs;
>>               fm = bpf_jit_find_kfunc_model(ctx->prog, insn);
>>               if (!fm)
>>                   return -EINVAL;
>> -            for (idx = 0; idx < fm->nr_args; idx++) {
>> +            nargs = min_t(int, fm->nr_args, MAX_BPF_FUNC_REG_ARGS);
>> +            for (idx = 0; idx < nargs; idx++) {
>>                   u8 reg = bpf_to_rv_reg(BPF_REG_1 + idx, ctx);
>>                   if (fm->arg_size[idx] == sizeof(int))
>>                       emit_sextw(reg, reg, ctx);
>>               }
>> +
>> +            /* BPF stack args -> RISC-V ABI: args 6-8 in A5-A7, 9+ at 
>> SP+0 */
>> +            if (fm->nr_args > MAX_BPF_FUNC_REG_ARGS) {
>> +                int n_stack = fm->nr_args - MAX_BPF_FUNC_REG_ARGS;
>> +                int n_reg = min_t(int, n_stack,
>> +                          RV_MAX_REG_ARGS - MAX_BPF_FUNC_REG_ARGS);
>> +
>> +                for (idx = 0; idx < n_reg; idx++) {
>> +                    int sz = fm->arg_size[MAX_BPF_FUNC_REG_ARGS + idx];
>> +
>> +                    emit_ld(RV_REG_A5 + idx, idx * 8, RV_REG_SP, ctx);
>> +                    if (sz == sizeof(int))
>> +                        emit_sextw(RV_REG_A5 + idx, RV_REG_A5 + idx, 
>> ctx);
>> +                }
>> +
>> +                for (idx = n_reg; idx < n_stack; idx++) {
>> +                    int sz = fm->arg_size[MAX_BPF_FUNC_REG_ARGS + idx];
>> +
>> +                    emit_ld(RV_REG_T1, idx * 8, RV_REG_SP, ctx);
>> +                    if (sz == sizeof(int))
>> +                        emit_sextw(RV_REG_T1, RV_REG_T1, ctx);
>> +                    emit_sd(RV_REG_SP, (idx - n_reg) * 8, RV_REG_T1, 
>> ctx);
>> +                }
>> +            }
> 
> looks a bit complicated, how about the following after redefine regmap:
> 
> ```
> for (idx = 0; idx < fm->nr_args; idx++) {
>    if (idx < RV_MAX_REG_ARGS) {
>      u8 reg = RV_REG_A0 + idx;
> 
>      sign_extend(reg, reg, xxx)
>    } else {
>      load from stack to T1
>      sign_extend
>      retore from T1 to orig stack slot
>    }
> }
> ```
> 
>>           }
>>           /* restore TCC to RV_REG_TCC before bpf2bpf call */
>> @@ -1891,6 +1916,21 @@ int bpf_jit_emit_insn(const struct bpf_insn 
>> *insn, struct rv_jit_context *ctx,
>>       case BPF_LDX | BPF_MEM | BPF_H:
>>       case BPF_LDX | BPF_MEM | BPF_W:
>>       case BPF_LDX | BPF_MEM | BPF_DW:
>> +        if (insn->src_reg == BPF_REG_PARAMS) {
> 
> pls use is_stack_arg_ldx/st/stx instead.
> 
>> +            int idx = off / 8 - 1;
>> +
>> +            if (is_12b_int(idx * 8)) {
>> +                emit_ldx_insn(rd, idx * 8, RV_REG_FP, BPF_SIZE(code), 
>> false, ctx);
>> +            } else {
>> +                emit_imm(RV_REG_T1, idx * 8, ctx);
>> +                emit_add(RV_REG_T1, RV_REG_T1, RV_REG_FP, ctx);
>> +                emit_ldx_insn(rd, 0, RV_REG_T1, BPF_SIZE(code), 
>> false, ctx);
>> +            }
>> +            if (BPF_SIZE(code) != BPF_DW && insn_is_zext(&insn[1]))
>> +                return 1;
>> +            break;
> 
> This piece of code was refactored not long ago. Let's see if we can 
> merge it into the logic below or fold it into emit_ldx—let's give it a try.
> 
>> +        }
>> +        fallthrough;
>>       case BPF_LDX | BPF_PROBE_MEM | BPF_B:
>>       case BPF_LDX | BPF_PROBE_MEM | BPF_H:
>>       case BPF_LDX | BPF_PROBE_MEM | BPF_W:
>> @@ -1938,6 +1978,20 @@ int bpf_jit_emit_insn(const struct bpf_insn 
>> *insn, struct rv_jit_context *ctx,
>>       case BPF_ST | BPF_MEM | BPF_H:
>>       case BPF_ST | BPF_MEM | BPF_W:
>>       case BPF_ST | BPF_MEM | BPF_DW:
>> +        if (insn->dst_reg == BPF_REG_PARAMS) {
>> +            int idx = -off / 8 - 1;
>> +
>> +            emit_imm(RV_REG_T1, imm, ctx);
>> +            if (is_12b_int(idx * 8)) {
>> +                emit_stx_insn(RV_REG_SP, idx * 8, RV_REG_T1, 
>> BPF_SIZE(code), ctx);
>> +            } else {
>> +                emit_imm(RV_REG_T2, idx * 8, ctx);
>> +                emit_add(RV_REG_T2, RV_REG_SP, RV_REG_T2, ctx);
>> +                emit_stx_insn(RV_REG_T2, 0, RV_REG_T1, 
>> BPF_SIZE(code), ctx);
>> +            }
>> +            break;
>> +        }
>> +        fallthrough;
> 
> ditto
> 
>>       /* ST | PROBE_MEM32: *(size *)(dst + RV_REG_ARENA + off) = imm */
>>       case BPF_ST | BPF_PROBE_MEM32 | BPF_B:
>>       case BPF_ST | BPF_PROBE_MEM32 | BPF_H:
>> @@ -1960,6 +2014,19 @@ int bpf_jit_emit_insn(const struct bpf_insn 
>> *insn, struct rv_jit_context *ctx,
>>       case BPF_STX | BPF_MEM | BPF_H:
>>       case BPF_STX | BPF_MEM | BPF_W:
>>       case BPF_STX | BPF_MEM | BPF_DW:
>> +        if (insn->dst_reg == BPF_REG_PARAMS) {
>> +            int idx = -off / 8 - 1;
>> +
>> +            if (is_12b_int(idx * 8)) {
>> +                emit_stx_insn(RV_REG_SP, idx * 8, rs, BPF_SIZE(code), 
>> ctx);
>> +            } else {
>> +                emit_imm(RV_REG_T1, idx * 8, ctx);
>> +                emit_add(RV_REG_T1, RV_REG_SP, RV_REG_T1, ctx);
>> +                emit_stx_insn(RV_REG_T1, 0, rs, BPF_SIZE(code), ctx);
>> +            }
>> +            break;
>> +        }
>> +        fallthrough;
> 
> ditto
> 
>>       /* STX | PROBE_MEM32: *(size *)(dst + RV_REG_ARENA + off) = src */
>>       case BPF_STX | BPF_PROBE_MEM32 | BPF_B:
>>       case BPF_STX | BPF_PROBE_MEM32 | BPF_H:
>> @@ -2036,6 +2103,7 @@ void bpf_jit_build_prologue(struct 
>> rv_jit_context *ctx, bool is_subprog)
>>       stack_adjust = round_up(stack_adjust, STACK_ALIGN);
>>       stack_adjust += bpf_stack_adjust;
>> +    stack_adjust += ctx->stack_arg_size;
>>       store_offset = stack_adjust - 8;
>> @@ -2093,7 +2161,7 @@ void bpf_jit_build_prologue(struct 
>> rv_jit_context *ctx, bool is_subprog)
>>       emit_addi(RV_REG_FP, RV_REG_SP, stack_adjust, ctx);
>>       if (bpf_stack_adjust)
>> -        emit_addi(RV_REG_S5, RV_REG_SP, bpf_stack_adjust, ctx);
>> +        emit_addi(RV_REG_S5, RV_REG_SP, ctx->stack_arg_size + 
>> bpf_stack_adjust, ctx);
>>       ctx->stack_size = stack_adjust;
>> @@ -2171,3 +2239,8 @@ bool bpf_jit_supports_timed_may_goto(void)
>>   {
>>       return true;
>>   }
>> +
>> +bool bpf_jit_supports_stack_args(void)
>> +{
>> +    return true;
>> +}
>> diff --git a/arch/riscv/net/bpf_jit_core.c 
>> b/arch/riscv/net/bpf_jit_core.c
>> index cbfcd287ea16..844a0f3e0fa9 100644
>> --- a/arch/riscv/net/bpf_jit_core.c
>> +++ b/arch/riscv/net/bpf_jit_core.c
>> @@ -72,6 +72,10 @@ struct bpf_prog *bpf_int_jit_compile(struct 
>> bpf_verifier_env *env, struct bpf_pr
>>       ctx->arena_vm_start = 
>> bpf_arena_get_kern_vm_start(prog->aux->arena);
>>       ctx->user_vm_start = bpf_arena_get_user_vm_start(prog->aux->arena);
>>       ctx->prog = prog;
>> +
>> +    ctx->stack_arg_size = round_up(bpf_out_stack_arg_cnt(env, prog) *
>> +                       sizeof(u64), STACK_ALIGN);
> 
> It's not u64—since every stack slot is aligned to the pointer size 
> (sizeof(long)), I think using 8 is fine.
> 
>> +
>>       ctx->offset = kvzalloc_objs(int, prog->len);
>>       if (!ctx->offset)
>>           goto out_offset;
>>
> 
> _______________________________________________
> linux-riscv mailing list
> linux-riscv at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv



More information about the linux-riscv mailing list