* [PATCH bpf-next v12 1/5] bpf: adjust BPF JIT dependency to BPF_SYSCALL
2025-03-19 16:36 [PATCH bpf-next v12 0/5] bpf: add cpu time counter kfuncs Vadim Fedorenko
@ 2025-03-19 16:36 ` Vadim Fedorenko
2025-03-19 16:36 ` [PATCH bpf-next v12 2/5] bpf: add bpf_get_cpu_time_counter kfunc Vadim Fedorenko
` (3 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Vadim Fedorenko @ 2025-03-19 16:36 UTC (permalink / raw)
To: Borislav Petkov, Alexei Starovoitov, Daniel Borkmann,
Andrii Nakryiko, Eduard Zingerman, Thomas Gleixner, Yonghong Song,
Vadim Fedorenko, Mykola Lysenko
Cc: x86, bpf, Peter Zijlstra, Vadim Fedorenko, Martin KaFai Lau
BPF JIT is moving towards optimizing kfuncs and it was long overdue to
switch the dependency. Let's do it now to simplify other patches in the
series.
Signed-off-by: Vadim Fedorenko <vadfed@meta.com>
---
kernel/bpf/Kconfig | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/kernel/bpf/Kconfig b/kernel/bpf/Kconfig
index 17067dcb4386..528d37819570 100644
--- a/kernel/bpf/Kconfig
+++ b/kernel/bpf/Kconfig
@@ -41,7 +41,7 @@ config BPF_SYSCALL
config BPF_JIT
bool "Enable BPF Just In Time compiler"
- depends on BPF
+ depends on BPF_SYSCALL
depends on HAVE_CBPF_JIT || HAVE_EBPF_JIT
select EXECMEM
help
--
2.47.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH bpf-next v12 2/5] bpf: add bpf_get_cpu_time_counter kfunc
2025-03-19 16:36 [PATCH bpf-next v12 0/5] bpf: add cpu time counter kfuncs Vadim Fedorenko
2025-03-19 16:36 ` [PATCH bpf-next v12 1/5] bpf: adjust BPF JIT dependency to BPF_SYSCALL Vadim Fedorenko
@ 2025-03-19 16:36 ` Vadim Fedorenko
2025-03-19 16:36 ` [PATCH bpf-next v12 3/5] bpf: add bpf_cpu_time_counter_to_ns helper Vadim Fedorenko
` (2 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Vadim Fedorenko @ 2025-03-19 16:36 UTC (permalink / raw)
To: Borislav Petkov, Alexei Starovoitov, Daniel Borkmann,
Andrii Nakryiko, Eduard Zingerman, Thomas Gleixner, Yonghong Song,
Vadim Fedorenko, Mykola Lysenko
Cc: x86, bpf, Peter Zijlstra, Vadim Fedorenko, Martin KaFai Lau
New kfunc to return ARCH-specific timecounter. The main reason to
implement this kfunc is to avoid extra overhead of benchmark
measurements, which are usually done by a pair of bpf_ktime_get_ns()
at the beginnig and at the end of the code block under benchmark.
When fully JITed this function doesn't implement conversion to the
monotonic clock and saves some CPU cycles by receiving timecounter
values in single-digit amount of instructions. The delta values can be
translated into nanoseconds using kfunc introduced in the next patch.
For x86_64 BPF JIT converts this kfunc into rdtsc ordered call. Other
architectures will get JIT implementation too if supported. The fallback
is to get CLOCK_MONOTONIC_RAW value in ns.
JIT version of the function uses "LFENCE; RDTSC" variant because it
doesn't care about cookie value returned by "RDTSCP" and it doesn't want
to trash RCX value. LFENCE option provides the same ordering guarantee as
RDTSCP variant.
The simplest use-case is added in 5th patch, where we calculate the time
spent by bpf_get_ns_current_pid_tgid() kfunc. More complex example is to
use session cookie to store timecounter value at kprobe/uprobe using
kprobe.session/uprobe.session, and calculate the difference at
kretprobe/uretprobe.
Acked-by: Eduard Zingerman <eddyz87@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Yonghong Song <yonghong.song@linux.dev>
Signed-off-by: Vadim Fedorenko <vadfed@meta.com>
---
arch/x86/net/bpf_jit_comp.c | 43 +++++++++++++++++++++++++++++++++++
arch/x86/net/bpf_jit_comp32.c | 1 +
include/linux/bpf.h | 3 +++
include/linux/filter.h | 1 +
kernel/bpf/core.c | 11 +++++++++
kernel/bpf/helpers.c | 11 +++++++++
kernel/bpf/verifier.c | 4 +++-
7 files changed, 73 insertions(+), 1 deletion(-)
diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c
index d3491cc0898b..284696d69df4 100644
--- a/arch/x86/net/bpf_jit_comp.c
+++ b/arch/x86/net/bpf_jit_comp.c
@@ -15,6 +15,7 @@
#include <asm/ftrace.h>
#include <asm/set_memory.h>
#include <asm/nospec-branch.h>
+#include <asm/timer.h>
#include <asm/text-patching.h>
#include <asm/unwind.h>
#include <asm/cfi.h>
@@ -2254,6 +2255,38 @@ st: if (is_imm8(insn->off))
case BPF_JMP | BPF_CALL: {
u8 *ip = image + addrs[i - 1];
+ if (insn->src_reg == BPF_PSEUDO_KFUNC_CALL &&
+ imm32 == BPF_CALL_IMM(bpf_get_cpu_time_counter) &&
+ bpf_jit_inlines_kfunc_call(imm32)) {
+ /* The default implementation of this kfunc uses
+ * ktime_get_raw_ns() which effectively is implemented as
+ * `(u64)rdtsc_ordered() & S64_MAX`. For JIT We skip
+ * masking part because we assume it's not needed in BPF
+ * use case (two measurements close in time).
+ * Original code for rdtsc_ordered() uses sequence:
+ * 'rdtsc; nop; nop; nop' to patch it into
+ * 'lfence; rdtsc' or 'rdtscp' depending on CPU features.
+ * JIT uses 'lfence; rdtsc' variant because BPF program
+ * doesn't care about cookie provided by rdtscp in RCX.
+ * Save RDX because RDTSC will use EDX:EAX to return u64
+ */
+ emit_mov_reg(&prog, true, AUX_REG, BPF_REG_3);
+ if (cpu_feature_enabled(X86_FEATURE_LFENCE_RDTSC))
+ EMIT_LFENCE();
+ EMIT2(0x0F, 0x31);
+
+ /* shl RDX, 32 */
+ maybe_emit_1mod(&prog, BPF_REG_3, true);
+ EMIT3(0xC1, add_1reg(0xE0, BPF_REG_3), 32);
+ /* or RAX, RDX */
+ maybe_emit_mod(&prog, BPF_REG_0, BPF_REG_3, true);
+ EMIT2(0x09, add_2reg(0xC0, BPF_REG_0, BPF_REG_3));
+ /* restore RDX from R11 */
+ emit_mov_reg(&prog, true, BPF_REG_3, AUX_REG);
+
+ break;
+ }
+
func = (u8 *) __bpf_call_base + imm32;
if (src_reg == BPF_PSEUDO_CALL && tail_call_reachable) {
LOAD_TAIL_CALL_CNT_PTR(stack_depth);
@@ -3865,3 +3898,13 @@ bool bpf_jit_supports_timed_may_goto(void)
{
return true;
}
+
+/* x86-64 JIT can inline kfunc */
+bool bpf_jit_inlines_kfunc_call(s32 imm)
+{
+ if (imm == BPF_CALL_IMM(bpf_get_cpu_time_counter) &&
+ cpu_feature_enabled(X86_FEATURE_TSC) &&
+ using_native_sched_clock() && sched_clock_stable())
+ return true;
+ return false;
+}
diff --git a/arch/x86/net/bpf_jit_comp32.c b/arch/x86/net/bpf_jit_comp32.c
index de0f9e5f9f73..68511888eb27 100644
--- a/arch/x86/net/bpf_jit_comp32.c
+++ b/arch/x86/net/bpf_jit_comp32.c
@@ -16,6 +16,7 @@
#include <asm/set_memory.h>
#include <asm/nospec-branch.h>
#include <asm/asm-prototypes.h>
+#include <asm/timer.h>
#include <linux/bpf.h>
/*
diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index 973a88d9b52b..6cf9138b2437 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -3389,6 +3389,9 @@ void bpf_user_rnd_init_once(void);
u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
+/* Inlined kfuncs */
+u64 bpf_get_cpu_time_counter(void);
+
#if defined(CONFIG_NET)
bool bpf_sock_common_is_valid_access(int off, int size,
enum bpf_access_type type,
diff --git a/include/linux/filter.h b/include/linux/filter.h
index 590476743f7a..2fbfa1bc3f49 100644
--- a/include/linux/filter.h
+++ b/include/linux/filter.h
@@ -1128,6 +1128,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog);
void bpf_jit_compile(struct bpf_prog *prog);
bool bpf_jit_needs_zext(void);
bool bpf_jit_inlines_helper_call(s32 imm);
+bool bpf_jit_inlines_kfunc_call(s32 imm);
bool bpf_jit_supports_subprog_tailcalls(void);
bool bpf_jit_supports_percpu_insn(void);
bool bpf_jit_supports_kfunc_call(void);
diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
index ba6b6118cf50..6ac61b9083ce 100644
--- a/kernel/bpf/core.c
+++ b/kernel/bpf/core.c
@@ -3040,6 +3040,17 @@ bool __weak bpf_jit_inlines_helper_call(s32 imm)
return false;
}
+/* Return true if the JIT inlines the call to the kfunc corresponding to
+ * the imm.
+ *
+ * The verifier will not patch the insn->imm for the call to the helper if
+ * this returns true.
+ */
+bool __weak bpf_jit_inlines_kfunc_call(s32 imm)
+{
+ return false;
+}
+
/* Return TRUE if the JIT backend supports mixing bpf2bpf and tailcalls. */
bool __weak bpf_jit_supports_subprog_tailcalls(void)
{
diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c
index ddaa41a70676..26f71e2438d2 100644
--- a/kernel/bpf/helpers.c
+++ b/kernel/bpf/helpers.c
@@ -3195,6 +3195,16 @@ __bpf_kfunc void bpf_local_irq_restore(unsigned long *flags__irq_flag)
local_irq_restore(*flags__irq_flag);
}
+__bpf_kfunc u64 bpf_get_cpu_time_counter(void)
+{
+ /* CLOCK_MONOTONIC_RAW is the closest analogue to what is implemented
+ * in JIT. The access time is the same as for CLOCK_MONOTONIC, but the
+ * slope of 'raw' is not affected by NTP adjustments, and with stable
+ * TSC it can provide less jitter in short term measurements.
+ */
+ return ktime_get_raw_fast_ns();
+}
+
__bpf_kfunc_end_defs();
BTF_KFUNCS_START(generic_btf_ids)
@@ -3295,6 +3305,7 @@ BTF_ID_FLAGS(func, bpf_iter_kmem_cache_next, KF_ITER_NEXT | KF_RET_NULL | KF_SLE
BTF_ID_FLAGS(func, bpf_iter_kmem_cache_destroy, KF_ITER_DESTROY | KF_SLEEPABLE)
BTF_ID_FLAGS(func, bpf_local_irq_save)
BTF_ID_FLAGS(func, bpf_local_irq_restore)
+BTF_ID_FLAGS(func, bpf_get_cpu_time_counter)
BTF_KFUNCS_END(common_btf_ids)
static const struct btf_kfunc_id_set common_kfunc_set = {
diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 9f8cbd5c61bc..aea1040b4462 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -17077,7 +17077,9 @@ static bool get_call_summary(struct bpf_verifier_env *env, struct bpf_insn *call
/* error would be reported later */
return false;
cs->num_params = btf_type_vlen(meta.func_proto);
- cs->fastcall = meta.kfunc_flags & KF_FASTCALL;
+ cs->fastcall = (meta.kfunc_flags & KF_FASTCALL) ||
+ (meta.btf == btf_vmlinux &&
+ bpf_jit_inlines_kfunc_call(call->imm));
cs->is_void = btf_type_is_void(btf_type_by_id(meta.btf, meta.func_proto->type));
return true;
}
--
2.47.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH bpf-next v12 3/5] bpf: add bpf_cpu_time_counter_to_ns helper
2025-03-19 16:36 [PATCH bpf-next v12 0/5] bpf: add cpu time counter kfuncs Vadim Fedorenko
2025-03-19 16:36 ` [PATCH bpf-next v12 1/5] bpf: adjust BPF JIT dependency to BPF_SYSCALL Vadim Fedorenko
2025-03-19 16:36 ` [PATCH bpf-next v12 2/5] bpf: add bpf_get_cpu_time_counter kfunc Vadim Fedorenko
@ 2025-03-19 16:36 ` Vadim Fedorenko
2025-03-19 16:36 ` [PATCH bpf-next v12 4/5] selftests/bpf: add selftest to check bpf_get_cpu_time_counter jit Vadim Fedorenko
2025-03-19 16:36 ` [PATCH bpf-next v12 5/5] selftests/bpf: add usage example for cpu time counter kfuncs Vadim Fedorenko
4 siblings, 0 replies; 6+ messages in thread
From: Vadim Fedorenko @ 2025-03-19 16:36 UTC (permalink / raw)
To: Borislav Petkov, Alexei Starovoitov, Daniel Borkmann,
Andrii Nakryiko, Eduard Zingerman, Thomas Gleixner, Yonghong Song,
Vadim Fedorenko, Mykola Lysenko
Cc: x86, bpf, Peter Zijlstra, Vadim Fedorenko, Martin KaFai Lau
The new helper should be used to convert deltas of values
received by bpf_get_cpu_time_counter() into nanoseconds. It is not
designed to do full conversion of time counter values to
CLOCK_MONOTONIC_RAW nanoseconds and cannot guarantee monotonicity of 2
independent values, but rather to convert the difference of 2 close
enough values of CPU timestamp counter into nanoseconds.
This function is JITted into just several instructions and adds as
low overhead as possible and perfectly suits benchmark use-cases.
When the kfunc is not JITted it returns the value provided as argument
because the kfunc in previous patch will return values in nanoseconds
and can be optimized by verifier.
Reviewed-by: Eduard Zingerman <eddyz87@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Vadim Fedorenko <vadfed@meta.com>
---
arch/x86/net/bpf_jit_comp.c | 29 ++++++++++++++++++++++++++++-
arch/x86/net/bpf_jit_comp32.c | 1 +
include/linux/bpf.h | 1 +
kernel/bpf/helpers.c | 6 ++++++
kernel/bpf/verifier.c | 9 ++++++++-
5 files changed, 44 insertions(+), 2 deletions(-)
diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c
index 284696d69df4..8ff8d7436fc9 100644
--- a/arch/x86/net/bpf_jit_comp.c
+++ b/arch/x86/net/bpf_jit_comp.c
@@ -9,6 +9,7 @@
#include <linux/filter.h>
#include <linux/if_vlan.h>
#include <linux/bpf.h>
+#include <linux/clocksource.h>
#include <linux/memory.h>
#include <linux/sort.h>
#include <asm/extable.h>
@@ -2287,6 +2288,31 @@ st: if (is_imm8(insn->off))
break;
}
+ if (insn->src_reg == BPF_PSEUDO_KFUNC_CALL &&
+ imm32 == BPF_CALL_IMM(bpf_cpu_time_counter_to_ns) &&
+ bpf_jit_inlines_kfunc_call(imm32)) {
+ struct cyc2ns_data data;
+ u32 mult, shift;
+
+ /* stable TSC runs with fixed frequency and
+ * transformation coefficients are also fixed
+ */
+ cyc2ns_read_begin(&data);
+ mult = data.cyc2ns_mul;
+ shift = data.cyc2ns_shift;
+ cyc2ns_read_end();
+ /* imul RAX, RDI, mult */
+ maybe_emit_mod(&prog, BPF_REG_1, BPF_REG_0, true);
+ EMIT2_off32(0x69, add_2reg(0xC0, BPF_REG_1, BPF_REG_0),
+ mult);
+
+ /* shr RAX, shift (which is less than 64) */
+ maybe_emit_1mod(&prog, BPF_REG_0, true);
+ EMIT3(0xC1, add_1reg(0xE8, BPF_REG_0), shift);
+
+ break;
+ }
+
func = (u8 *) __bpf_call_base + imm32;
if (src_reg == BPF_PSEUDO_CALL && tail_call_reachable) {
LOAD_TAIL_CALL_CNT_PTR(stack_depth);
@@ -3902,7 +3928,8 @@ bool bpf_jit_supports_timed_may_goto(void)
/* x86-64 JIT can inline kfunc */
bool bpf_jit_inlines_kfunc_call(s32 imm)
{
- if (imm == BPF_CALL_IMM(bpf_get_cpu_time_counter) &&
+ if ((imm == BPF_CALL_IMM(bpf_get_cpu_time_counter) ||
+ imm == BPF_CALL_IMM(bpf_cpu_time_counter_to_ns)) &&
cpu_feature_enabled(X86_FEATURE_TSC) &&
using_native_sched_clock() && sched_clock_stable())
return true;
diff --git a/arch/x86/net/bpf_jit_comp32.c b/arch/x86/net/bpf_jit_comp32.c
index 68511888eb27..83176a07fc08 100644
--- a/arch/x86/net/bpf_jit_comp32.c
+++ b/arch/x86/net/bpf_jit_comp32.c
@@ -12,6 +12,7 @@
#include <linux/netdevice.h>
#include <linux/filter.h>
#include <linux/if_vlan.h>
+#include <linux/clocksource.h>
#include <asm/cacheflush.h>
#include <asm/set_memory.h>
#include <asm/nospec-branch.h>
diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index 6cf9138b2437..fc03a3805b36 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -3391,6 +3391,7 @@ u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
/* Inlined kfuncs */
u64 bpf_get_cpu_time_counter(void);
+u64 bpf_cpu_time_counter_to_ns(u64 counter);
#if defined(CONFIG_NET)
bool bpf_sock_common_is_valid_access(int off, int size,
diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c
index 26f71e2438d2..a176bd5a33d0 100644
--- a/kernel/bpf/helpers.c
+++ b/kernel/bpf/helpers.c
@@ -3205,6 +3205,11 @@ __bpf_kfunc u64 bpf_get_cpu_time_counter(void)
return ktime_get_raw_fast_ns();
}
+__bpf_kfunc u64 bpf_cpu_time_counter_to_ns(u64 counter)
+{
+ return counter;
+}
+
__bpf_kfunc_end_defs();
BTF_KFUNCS_START(generic_btf_ids)
@@ -3306,6 +3311,7 @@ BTF_ID_FLAGS(func, bpf_iter_kmem_cache_destroy, KF_ITER_DESTROY | KF_SLEEPABLE)
BTF_ID_FLAGS(func, bpf_local_irq_save)
BTF_ID_FLAGS(func, bpf_local_irq_restore)
BTF_ID_FLAGS(func, bpf_get_cpu_time_counter)
+BTF_ID_FLAGS(func, bpf_cpu_time_counter_to_ns, KF_FASTCALL)
BTF_KFUNCS_END(common_btf_ids)
static const struct btf_kfunc_id_set common_kfunc_set = {
diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index aea1040b4462..3a908cf24e45 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -12007,6 +12007,7 @@ enum special_kfunc_type {
KF_bpf_iter_num_destroy,
KF_bpf_set_dentry_xattr,
KF_bpf_remove_dentry_xattr,
+ KF_bpf_cpu_time_counter_to_ns,
};
BTF_SET_START(special_kfunc_set)
@@ -12040,6 +12041,7 @@ BTF_ID(func, bpf_iter_css_task_new)
BTF_ID(func, bpf_set_dentry_xattr)
BTF_ID(func, bpf_remove_dentry_xattr)
#endif
+BTF_ID(func, bpf_cpu_time_counter_to_ns)
BTF_SET_END(special_kfunc_set)
BTF_ID_LIST(special_kfunc_list)
@@ -12096,6 +12098,7 @@ BTF_ID(func, bpf_remove_dentry_xattr)
BTF_ID_UNUSED
BTF_ID_UNUSED
#endif
+BTF_ID(func, bpf_cpu_time_counter_to_ns)
static bool is_kfunc_ret_null(struct bpf_kfunc_call_arg_meta *meta)
{
@@ -21246,6 +21249,9 @@ static int fixup_kfunc_call(struct bpf_verifier_env *env, struct bpf_insn *insn,
if (!bpf_jit_supports_far_kfunc_call())
insn->imm = BPF_CALL_IMM(desc->addr);
+ /* if JIT will inline kfunc verifier shouldn't change the code */
+ if (bpf_jit_inlines_kfunc_call(insn->imm))
+ return 0;
if (insn->off)
return 0;
if (desc->func_id == special_kfunc_list[KF_bpf_obj_new_impl] ||
@@ -21310,7 +21316,8 @@ static int fixup_kfunc_call(struct bpf_verifier_env *env, struct bpf_insn *insn,
__fixup_collection_insert_kfunc(&env->insn_aux_data[insn_idx], struct_meta_reg,
node_offset_reg, insn, insn_buf, cnt);
} else if (desc->func_id == special_kfunc_list[KF_bpf_cast_to_kern_ctx] ||
- desc->func_id == special_kfunc_list[KF_bpf_rdonly_cast]) {
+ desc->func_id == special_kfunc_list[KF_bpf_rdonly_cast] ||
+ desc->func_id == special_kfunc_list[KF_bpf_cpu_time_counter_to_ns]) {
insn_buf[0] = BPF_MOV64_REG(BPF_REG_0, BPF_REG_1);
*cnt = 1;
} else if (is_bpf_wq_set_callback_impl_kfunc(desc->func_id)) {
--
2.47.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH bpf-next v12 4/5] selftests/bpf: add selftest to check bpf_get_cpu_time_counter jit
2025-03-19 16:36 [PATCH bpf-next v12 0/5] bpf: add cpu time counter kfuncs Vadim Fedorenko
` (2 preceding siblings ...)
2025-03-19 16:36 ` [PATCH bpf-next v12 3/5] bpf: add bpf_cpu_time_counter_to_ns helper Vadim Fedorenko
@ 2025-03-19 16:36 ` Vadim Fedorenko
2025-03-19 16:36 ` [PATCH bpf-next v12 5/5] selftests/bpf: add usage example for cpu time counter kfuncs Vadim Fedorenko
4 siblings, 0 replies; 6+ messages in thread
From: Vadim Fedorenko @ 2025-03-19 16:36 UTC (permalink / raw)
To: Borislav Petkov, Alexei Starovoitov, Daniel Borkmann,
Andrii Nakryiko, Eduard Zingerman, Thomas Gleixner, Yonghong Song,
Vadim Fedorenko, Mykola Lysenko
Cc: x86, bpf, Peter Zijlstra, Vadim Fedorenko, Martin KaFai Lau
bpf_get_cpu_time_counter() is replaced with rdtsc instruction on x86_64.
Add tests to check that JIT works as expected. When JIT is not
supported, bpf_cpu_time_counter_to_ns() can be inlined by verifier.
Acked-by: Eduard Zingerman <eddyz87@gmail.com>
Signed-off-by: Vadim Fedorenko <vadfed@meta.com>
---
.../selftests/bpf/prog_tests/verifier.c | 2 +
.../selftests/bpf/progs/verifier_cpu_cycles.c | 120 ++++++++++++++++++
2 files changed, 122 insertions(+)
create mode 100644 tools/testing/selftests/bpf/progs/verifier_cpu_cycles.c
diff --git a/tools/testing/selftests/bpf/prog_tests/verifier.c b/tools/testing/selftests/bpf/prog_tests/verifier.c
index e66a57970d28..d5e7e302a344 100644
--- a/tools/testing/selftests/bpf/prog_tests/verifier.c
+++ b/tools/testing/selftests/bpf/prog_tests/verifier.c
@@ -102,6 +102,7 @@
#include "verifier_xdp_direct_packet_access.skel.h"
#include "verifier_bits_iter.skel.h"
#include "verifier_lsm.skel.h"
+#include "verifier_cpu_cycles.skel.h"
#include "irq.skel.h"
#define MAX_ENTRIES 11
@@ -236,6 +237,7 @@ void test_verifier_bits_iter(void) { RUN(verifier_bits_iter); }
void test_verifier_lsm(void) { RUN(verifier_lsm); }
void test_irq(void) { RUN(irq); }
void test_verifier_mtu(void) { RUN(verifier_mtu); }
+void test_verifier_cpu_cycles(void) { RUN(verifier_cpu_cycles); }
static int init_test_val_map(struct bpf_object *obj, char *map_name)
{
diff --git a/tools/testing/selftests/bpf/progs/verifier_cpu_cycles.c b/tools/testing/selftests/bpf/progs/verifier_cpu_cycles.c
new file mode 100644
index 000000000000..26c02010ccf1
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/verifier_cpu_cycles.c
@@ -0,0 +1,120 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2025 Meta Inc. */
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+#include "bpf_misc.h"
+
+extern u64 bpf_cpu_time_counter_to_ns(u64 cycles) __weak __ksym;
+extern u64 bpf_get_cpu_time_counter(void) __weak __ksym;
+
+SEC("syscall")
+__arch_x86_64
+__xlated("0: call kernel-function")
+__naked int bpf_rdtsc(void)
+{
+ asm volatile(
+ "call %[bpf_get_cpu_time_counter];"
+ "exit"
+ :
+ : __imm(bpf_get_cpu_time_counter)
+ : __clobber_all
+ );
+}
+
+SEC("syscall")
+__arch_x86_64
+/* program entry for bpf_rdtsc_jit_x86_64(), regular function prologue */
+__jited(" endbr64")
+__jited(" nopl (%rax,%rax)")
+__jited(" nopl (%rax)")
+__jited(" pushq %rbp")
+__jited(" movq %rsp, %rbp")
+__jited(" endbr64")
+/* save RDX in R11 as it will be overwritten */
+__jited(" movq %rdx, %r11")
+/* lfence may not be executed depending on cpu features */
+__jited(" {{(lfence|)}}")
+__jited(" rdtsc")
+/* combine EDX:EAX into RAX */
+__jited(" shlq ${{(32|0x20)}}, %rdx")
+__jited(" orq %rdx, %rax")
+/* restore RDX from R11 */
+__jited(" movq %r11, %rdx")
+__jited(" leave")
+__naked int bpf_rdtsc_jit_x86_64(void)
+{
+ asm volatile(
+ "call %[bpf_get_cpu_time_counter];"
+ "exit"
+ :
+ : __imm(bpf_get_cpu_time_counter)
+ : __clobber_all
+ );
+}
+
+SEC("syscall")
+__arch_arm64
+__xlated("0: r1 = 42")
+__xlated("1: r0 = r1")
+__naked int bpf_cyc2ns_arm(void)
+{
+ asm volatile(
+ "r1=0x2a;"
+ "call %[bpf_cpu_time_counter_to_ns];"
+ "exit"
+ :
+ : __imm(bpf_cpu_time_counter_to_ns)
+ : __clobber_all
+ );
+}
+
+SEC("syscall")
+__arch_x86_64
+__xlated("0: r1 = 42")
+__xlated("1: call kernel-function")
+__naked int bpf_cyc2ns(void)
+{
+ asm volatile(
+ "r1=0x2a;"
+ "call %[bpf_cpu_time_counter_to_ns];"
+ "exit"
+ :
+ : __imm(bpf_cpu_time_counter_to_ns)
+ : __clobber_all
+ );
+}
+
+SEC("syscall")
+__arch_x86_64
+/* program entry for bpf_rdtsc_jit_x86_64(), regular function prologue */
+__jited(" endbr64")
+__jited(" nopl (%rax,%rax)")
+__jited(" nopl (%rax)")
+__jited(" pushq %rbp")
+__jited(" movq %rsp, %rbp")
+__jited(" endbr64")
+/* save RDX in R11 as it will be overwritten */
+__jited(" movabsq $0x2a2a2a2a2a, %rdi")
+__jited(" imulq ${{.*}}, %rdi, %rax")
+__jited(" shrq ${{.*}}, %rax")
+__jited(" leave")
+__naked int bpf_cyc2ns_jit_x86(void)
+{
+ asm volatile(
+ "r1=0x2a2a2a2a2a ll;"
+ "call %[bpf_cpu_time_counter_to_ns];"
+ "exit"
+ :
+ : __imm(bpf_cpu_time_counter_to_ns)
+ : __clobber_all
+ );
+}
+
+void rdtsc(void)
+{
+ bpf_get_cpu_time_counter();
+ bpf_cpu_time_counter_to_ns(42);
+}
+
+char _license[] SEC("license") = "GPL";
--
2.47.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH bpf-next v12 5/5] selftests/bpf: add usage example for cpu time counter kfuncs
2025-03-19 16:36 [PATCH bpf-next v12 0/5] bpf: add cpu time counter kfuncs Vadim Fedorenko
` (3 preceding siblings ...)
2025-03-19 16:36 ` [PATCH bpf-next v12 4/5] selftests/bpf: add selftest to check bpf_get_cpu_time_counter jit Vadim Fedorenko
@ 2025-03-19 16:36 ` Vadim Fedorenko
4 siblings, 0 replies; 6+ messages in thread
From: Vadim Fedorenko @ 2025-03-19 16:36 UTC (permalink / raw)
To: Borislav Petkov, Alexei Starovoitov, Daniel Borkmann,
Andrii Nakryiko, Eduard Zingerman, Thomas Gleixner, Yonghong Song,
Vadim Fedorenko, Mykola Lysenko
Cc: x86, bpf, Peter Zijlstra, Vadim Fedorenko, Martin KaFai Lau
The selftest provides an example of how to measure the latency of bpf
kfunc/helper call using time stamp counter and how to convert measured
value into nanoseconds.
Signed-off-by: Vadim Fedorenko <vadfed@meta.com>
---
.../bpf/prog_tests/test_cpu_cycles.c | 35 +++++++++++++++++++
.../selftests/bpf/progs/test_cpu_cycles.c | 25 +++++++++++++
2 files changed, 60 insertions(+)
create mode 100644 tools/testing/selftests/bpf/prog_tests/test_cpu_cycles.c
create mode 100644 tools/testing/selftests/bpf/progs/test_cpu_cycles.c
diff --git a/tools/testing/selftests/bpf/prog_tests/test_cpu_cycles.c b/tools/testing/selftests/bpf/prog_tests/test_cpu_cycles.c
new file mode 100644
index 000000000000..067307f0c4c2
--- /dev/null
+++ b/tools/testing/selftests/bpf/prog_tests/test_cpu_cycles.c
@@ -0,0 +1,35 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2025 Meta Inc. */
+
+#include <test_progs.h>
+#include "test_cpu_cycles.skel.h"
+
+static void cpu_cycles(void)
+{
+ LIBBPF_OPTS(bpf_test_run_opts, opts);
+ struct test_cpu_cycles *skel;
+ int err, pfd;
+
+ skel = test_cpu_cycles__open_and_load();
+ if (!ASSERT_OK_PTR(skel, "test_cpu_cycles open and load"))
+ return;
+
+ pfd = bpf_program__fd(skel->progs.bpf_cpu_cycles);
+ if (!ASSERT_GT(pfd, 0, "test_cpu_cycles fd"))
+ goto fail;
+
+ err = bpf_prog_test_run_opts(pfd, &opts);
+ if (!ASSERT_OK(err, "test_cpu_cycles test run"))
+ goto fail;
+
+ ASSERT_NEQ(skel->bss->cycles, 0, "test_cpu_cycles 0 cycles");
+ ASSERT_NEQ(skel->bss->ns, 0, "test_cpu_cycles 0 ns");
+fail:
+ test_cpu_cycles__destroy(skel);
+}
+
+void test_cpu_cycles(void)
+{
+ if (test__start_subtest("cpu_cycles"))
+ cpu_cycles();
+}
diff --git a/tools/testing/selftests/bpf/progs/test_cpu_cycles.c b/tools/testing/selftests/bpf/progs/test_cpu_cycles.c
new file mode 100644
index 000000000000..3c428f3be831
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/test_cpu_cycles.c
@@ -0,0 +1,25 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2025 Meta Inc. */
+
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+
+extern u64 bpf_cpu_time_counter_to_ns(u64 cycles) __weak __ksym;
+extern u64 bpf_get_cpu_time_counter(void) __weak __ksym;
+
+__u64 cycles, ns;
+
+SEC("syscall")
+int bpf_cpu_cycles(void)
+{
+ struct bpf_pidns_info pidns;
+ __u64 start;
+
+ start = bpf_get_cpu_time_counter();
+ bpf_get_ns_current_pid_tgid(0, 0, &pidns, sizeof(struct bpf_pidns_info));
+ cycles = bpf_get_cpu_time_counter() - start;
+ ns = bpf_cpu_time_counter_to_ns(cycles);
+ return 0;
+}
+
+char _license[] SEC("license") = "GPL";
--
2.47.1
^ permalink raw reply related [flat|nested] 6+ messages in thread