[PATCH bpf-next 0/2] RISC-V JIT support for bpf_get_current_task/_btf

[PATCH bpf-next 0/2] RISC-V JIT support for bpf_get_current_task/_btf

[Varun R Mallya]

These two patches add support for the bpf_get_current_task and
bpf_get_current_task_btf kfuncs in RISC-V JIT and add a selftest.

The first patch adds support for cpu and feature detection on
the JIT disassembly helper function as RISC-V JITed code was not
being disassembled using `LLVMCreateDisasm` as it was missing the
"+c" CPU feature and JITed code contained RISC-V Compressed (C)
Extension. This patch generalizes that to detect CPU features and
enables testing on more RISC-V JIT work ahead.

The second patch, which actually adds this support has been benchmarked
on QEMU RISC-V and shows significant improvements.
It was benchmarked using a simple loop inside a bpf program that ran
bpf_get_current_task() and execution time was measured. It used
bpf_prog_test_run_opts() to repeatedly trigger the BPF program. The loop
ran in 1 second intervals and it kept firing bpf_prog_test_run_opts() as
fast as possible until a second had elapsed and then reported statistics.

Varun R Mallya (2):
  selftests/bpf: use host CPU features in JIT disassembler
  bpf, riscv: inline bpf_get_current_task() and
    bpf_get_current_task_btf()

 arch/riscv/net/bpf_jit_comp64.c                     |  9 +++++++++
 tools/testing/selftests/bpf/jit_disasm_helpers.c    | 13 +++++++++++--
 .../selftests/bpf/progs/verifier_jit_inline.c       |  2 ++
 3 files changed, 22 insertions(+), 2 deletions(-)

-- 
2.54.0

[PATCH bpf-next 1/2] selftests/bpf: use host CPU features in JIT disassembler

[Varun R Mallya]

Pass the host CPU name and feature string to
LLVMCreateDisasmCPUFeatures() instead of using LLVMCreateDisasm(), so
the disassembler correctly decodes CPU-specific instructions and
extensions such as RISC-V compressed and vector instructions.

Signed-off-by: Varun R Mallya <varunrmallya@gmail.com>
---
 tools/testing/selftests/bpf/jit_disasm_helpers.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/bpf/jit_disasm_helpers.c b/tools/testing/selftests/bpf/jit_disasm_helpers.c
index 364c557c5115..3558fe10e28c 100644
--- a/tools/testing/selftests/bpf/jit_disasm_helpers.c
+++ b/tools/testing/selftests/bpf/jit_disasm_helpers.c
@@ -96,10 +96,19 @@ static int disasm_one_func(FILE *text_out, uint8_t *image, __u32 len)
 	__u32 *label_pc, pc;
 	int i, cnt, err = 0;
 	char buf[64];
+	char *cpu, *features;
 
 	triple = LLVMGetDefaultTargetTriple();
-	ctx = LLVMCreateDisasm(triple, &labels, 0, NULL, lookup_symbol);
-	if (!ASSERT_OK_PTR(ctx, "LLVMCreateDisasm")) {
+
+	cpu = LLVMGetHostCPUName();
+	features = LLVMGetHostCPUFeatures();
+
+	ctx = LLVMCreateDisasmCPUFeatures(triple, cpu, features, &labels, 0, NULL, lookup_symbol);
+
+	LLVMDisposeMessage(cpu);
+	LLVMDisposeMessage(features);
+
+	if (!ASSERT_OK_PTR(ctx, "LLVMCreateDisasmCPUFeatures")) {
 		err = -EINVAL;
 		goto out;
 	}
-- 
2.54.0

[PATCH bpf-next 2/2] bpf, riscv: inline bpf_get_current_task() and bpf_get_current_task_btf()

[Varun R Mallya]

On RISC-V, the current task pointer is stored in the thread pointer
register (tp). Emit a single `mv a5, tp` instead of a full helper
call for BPF_FUNC_get_current_task and BPF_FUNC_get_current_task_btf.

Register bpf_jit_inlines_helper_call() entries for both helpers so the
verifier treats them as inlined, and add the expected `mv a5, tp`
annotation to the riscv64 selftests.

The following show changes before and after this patch.

Before patch:

      auipc  t1,0x817a    # load upper PC-relative address
      jalr   -2004(t1)    # call bpf_get_current_task helper
      mv     a5,a0        # move return value to BPF_REG_0

After patch:

      mv     a5,tp        # directly: a5 = current (tp = thread pointer)

Benchmark (bpf_prog_test_run wrapping bpf_get_current_task in loop,
batch=100, 10s, QEMU RISC-V):

              | runs/sec  | helper-calls/sec | ns/call
 -------------+-----------+------------------+---------
 Before patch |   173,490 |       17,349,090 |      57
 After patch  |   320,497 |       32,049,780 |      31
 -------------+-----------+------------------+---------
 Improvement  |   +84.7%  |          +84.7%  |  -45.6%

Signed-off-by: Varun R Mallya <varunrmallya@gmail.com>
---
 arch/riscv/net/bpf_jit_comp64.c                         | 9 +++++++++
 tools/testing/selftests/bpf/progs/verifier_jit_inline.c | 2 ++
 2 files changed, 11 insertions(+)

diff --git a/arch/riscv/net/bpf_jit_comp64.c b/arch/riscv/net/bpf_jit_comp64.c
index 2f1109dbf105..e2c70c70cca8 100644
--- a/arch/riscv/net/bpf_jit_comp64.c
+++ b/arch/riscv/net/bpf_jit_comp64.c
@@ -1808,6 +1808,13 @@ int bpf_jit_emit_insn(const struct bpf_insn *insn, struct rv_jit_context *ctx,
 			break;
 		}
 
+		/* Implement helper call to bpf_get_current_task/_btf() inline */
+		if (insn->src_reg == 0 && (insn->imm == BPF_FUNC_get_current_task ||
+					   insn->imm == BPF_FUNC_get_current_task_btf)) {
+			emit_mv(bpf_to_rv_reg(BPF_REG_0, ctx), RV_REG_TP, ctx);
+			break;
+		}
+
 		mark_call(ctx);
 		ret = bpf_jit_get_func_addr(ctx->prog, insn, extra_pass,
 					    &addr, &fixed_addr);
@@ -2138,6 +2145,8 @@ bool bpf_jit_inlines_helper_call(s32 imm)
 {
 	switch (imm) {
 	case BPF_FUNC_get_smp_processor_id:
+	case BPF_FUNC_get_current_task:
+	case BPF_FUNC_get_current_task_btf:
 		return true;
 	default:
 		return false;
diff --git a/tools/testing/selftests/bpf/progs/verifier_jit_inline.c b/tools/testing/selftests/bpf/progs/verifier_jit_inline.c
index 885ff69a3a62..76d80605ec7f 100644
--- a/tools/testing/selftests/bpf/progs/verifier_jit_inline.c
+++ b/tools/testing/selftests/bpf/progs/verifier_jit_inline.c
@@ -10,6 +10,8 @@ __arch_x86_64
 __jited("	addq	%gs:{{.*}}, %rax")
 __arch_arm64
 __jited("	mrs	x8, SP_EL0")
+__arch_riscv64
+__jited("	mv	a5, tp")
 int inline_bpf_get_current_task(void)
 {
 	bpf_get_current_task();
-- 
2.54.0

Re: [PATCH bpf-next 2/2] bpf, riscv: inline bpf_get_current_task() and bpf_get_current_task_btf()

[Björn Töpel]

On Tue, 2 Jun 2026 at 22:59, Varun R Mallya <varunrmallya@gmail.com> wrote:
>
> On RISC-V, the current task pointer is stored in the thread pointer
> register (tp). Emit a single `mv a5, tp` instead of a full helper
> call for BPF_FUNC_get_current_task and BPF_FUNC_get_current_task_btf.
>
> Register bpf_jit_inlines_helper_call() entries for both helpers so the
> verifier treats them as inlined, and add the expected `mv a5, tp`
> annotation to the riscv64 selftests.
>
> The following show changes before and after this patch.
>
> Before patch:
>
>       auipc  t1,0x817a    # load upper PC-relative address
>       jalr   -2004(t1)    # call bpf_get_current_task helper
>       mv     a5,a0        # move return value to BPF_REG_0
>
> After patch:
>
>       mv     a5,tp        # directly: a5 = current (tp = thread pointer)
>
> Benchmark (bpf_prog_test_run wrapping bpf_get_current_task in loop,
> batch=100, 10s, QEMU RISC-V):
>
>               | runs/sec  | helper-calls/sec | ns/call
>  -------------+-----------+------------------+---------
>  Before patch |   173,490 |       17,349,090 |      57
>  After patch  |   320,497 |       32,049,780 |      31
>  -------------+-----------+------------------+---------
>  Improvement  |   +84.7%  |          +84.7%  |  -45.6%
>
> Signed-off-by: Varun R Mallya <varunrmallya@gmail.com>

Acked-by: Björn Töpel <bjorn@kernel.org>

Re: [PATCH bpf-next 1/2] selftests/bpf: use host CPU features in JIT disassembler

[Björn Töpel]

On Tue, 2 Jun 2026 at 22:59, Varun R Mallya <varunrmallya@gmail.com> wrote:
>
> Pass the host CPU name and feature string to
> LLVMCreateDisasmCPUFeatures() instead of using LLVMCreateDisasm(), so
> the disassembler correctly decodes CPU-specific instructions and
> extensions such as RISC-V compressed and vector instructions.
>
> Signed-off-by: Varun R Mallya <varunrmallya@gmail.com>

Reviewed-by: Björn Töpel <bjorn@kernel.org>

Re: [PATCH bpf-next 0/2] RISC-V JIT support for bpf_get_current_task/_btf

[patchwork-bot+netdevbpf]

Hello:

This series was applied to bpf/bpf-next.git (master)
by Alexei Starovoitov <ast@kernel.org>:

On Wed,  3 Jun 2026 02:28:45 +0530 you wrote:
> These two patches add support for the bpf_get_current_task and
> bpf_get_current_task_btf kfuncs in RISC-V JIT and add a selftest.
> 
> The first patch adds support for cpu and feature detection on
> the JIT disassembly helper function as RISC-V JITed code was not
> being disassembled using `LLVMCreateDisasm` as it was missing the
> "+c" CPU feature and JITed code contained RISC-V Compressed (C)
> Extension. This patch generalizes that to detect CPU features and
> enables testing on more RISC-V JIT work ahead.
> 
> [...]

Here is the summary with links:
  - [bpf-next,1/2] selftests/bpf: use host CPU features in JIT disassembler
    https://git.kernel.org/bpf/bpf-next/c/557d0cc3f252
  - [bpf-next,2/2] bpf, riscv: inline bpf_get_current_task() and bpf_get_current_task_btf()
    https://git.kernel.org/bpf/bpf-next/c/6d13ddb1d465

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html