Merge tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next

Daniel Borkmann says:

====================
pull-request: bpf-next 2024-03-25

We've added 38 non-merge commits during the last 13 day(s) which contain
a total of 50 files changed, 867 insertions(+), 274 deletions(-).

The main changes are:

1) Add the ability to specify and retrieve BPF cookie also for raw
   tracepoint programs in order to ease migration from classic to raw
   tracepoints, from Andrii Nakryiko.

2) Allow the use of bpf_get_{ns_,}current_pid_tgid() helper for all
   program types and add additional BPF selftests, from Yonghong Song.

3) Several improvements to bpftool and its build, for example, enabling
   libbpf logs when loading pid_iter in debug mode, from Quentin Monnet.

4) Check the return code of all BPF-related set_memory_*() functions during
   load and bail out in case they fail, from Christophe Leroy.

5) Avoid a goto in regs_refine_cond_op() such that the verifier can
   be better integrated into Agni tool which doesn't support backedges
   yet, from Harishankar Vishwanathan.

6) Add a small BPF trie perf improvement by always inlining
   longest_prefix_match, from Jesper Dangaard Brouer.

7) Small BPF selftest refactor in bpf_tcp_ca.c to utilize start_server()
   helper instead of open-coding it, from Geliang Tang.

8) Improve test_tc_tunnel.sh BPF selftest to prevent client connect
   before the server bind, from Alessandro Carminati.

9) Fix BPF selftest benchmark for older glibc and use syscall(SYS_gettid)
   instead of gettid(), from Alan Maguire.

10) Implement a backward-compatible method for struct_ops types with
    additional fields which are not present in older kernels,
    from Kui-Feng Lee.

11) Add a small helper to check if an instruction is addr_space_cast
    from as(0) to as(1) and utilize it in x86-64 JIT, from Puranjay Mohan.

12) Small cleanup to remove unnecessary error check in
    bpf_struct_ops_map_update_elem, from Martin KaFai Lau.

13) Improvements to libbpf fd validity checks for BPF map/programs,
    from Mykyta Yatsenko.

* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (38 commits)
  selftests/bpf: Fix flaky test btf_map_in_map/lookup_update
  bpf: implement insn_is_cast_user() helper for JITs
  bpf: Avoid get_kernel_nofault() to fetch kprobe entry IP
  selftests/bpf: Use start_server in bpf_tcp_ca
  bpf: Sync uapi bpf.h to tools directory
  libbpf: Add new sec_def "sk_skb/verdict"
  selftests/bpf: Mark uprobe trigger functions with nocf_check attribute
  selftests/bpf: Use syscall(SYS_gettid) instead of gettid() wrapper in bench
  bpf-next: Avoid goto in regs_refine_cond_op()
  bpftool: Clean up HOST_CFLAGS, HOST_LDFLAGS for bootstrap bpftool
  selftests/bpf: scale benchmark counting by using per-CPU counters
  bpftool: Remove unnecessary source files from bootstrap version
  bpftool: Enable libbpf logs when loading pid_iter in debug mode
  selftests/bpf: add raw_tp/tp_btf BPF cookie subtests
  libbpf: add support for BPF cookie for raw_tp/tp_btf programs
  bpf: support BPF cookie in raw tracepoint (raw_tp, tp_btf) programs
  bpf: pass whole link instead of prog when triggering raw tracepoint
  bpf: flatten bpf_probe_register call chain
  selftests/bpf: Prevent client connect before server bind in test_tc_tunnel.sh
  selftests/bpf: Add a sk_msg prog bpf_get_ns_current_pid_tgid() test
  ...
====================

Link: https://lore.kernel.org/r/20240325233940.7154-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

1 files changed, 40 insertions, 8 deletions

diff --git a/tools/testing/selftests/bpf/benchs/bench_trigger.c b/tools/testing/selftests/bpf/benchs/bench_trigger.c
index ace0d1011a8e..d66eddacd642 100644
--- a/tools/testing/selftests/bpf/benchs/bench_trigger.c
+++ b/tools/testing/selftests/bpf/benchs/bench_trigger.c
@@ -1,15 +1,47 @@
 // SPDX-License-Identifier: GPL-2.0
 /* Copyright (c) 2020 Facebook */
+#define _GNU_SOURCE
+#include <unistd.h>
 #include "bench.h"
 #include "trigger_bench.skel.h"
 #include "trace_helpers.h"
 
+/* adjust slot shift in inc_hits() if changing */
+#define MAX_BUCKETS 256
+
+#pragma GCC diagnostic ignored "-Wattributes"
+
 /* BPF triggering benchmarks */
 static struct trigger_ctx {
 	struct trigger_bench *skel;
 } ctx;
 
-static struct counter base_hits;
+static struct counter base_hits[MAX_BUCKETS];
+
+static __always_inline void inc_counter(struct counter *counters)
+{
+	static __thread int tid = 0;
+	unsigned slot;
+
+	if (unlikely(tid == 0))
+		tid = syscall(SYS_gettid);
+
+	/* multiplicative hashing, it's fast */
+	slot = 2654435769U * tid;
+	slot >>= 24;
+
+	atomic_inc(&base_hits[slot].value); /* use highest byte as an index */
+}
+
+static long sum_and_reset_counters(struct counter *counters)
+{
+	int i;
+	long sum = 0;
+
+	for (i = 0; i < MAX_BUCKETS; i++)
+		sum += atomic_swap(&counters[i].value, 0);
+	return sum;
+}
 
 static void trigger_validate(void)
 {
@@ -23,14 +55,14 @@ static void *trigger_base_producer(void *input)
 {
 	while (true) {
 		(void)syscall(__NR_getpgid);
-		atomic_inc(&base_hits.value);
+		inc_counter(base_hits);
 	}
 	return NULL;
 }
 
 static void trigger_base_measure(struct bench_res *res)
 {
-	res->hits = atomic_swap(&base_hits.value, 0);
+	res->hits = sum_and_reset_counters(base_hits);
 }
 
 static void *trigger_producer(void *input)
@@ -42,7 +74,7 @@ static void *trigger_producer(void *input)
 
 static void trigger_measure(struct bench_res *res)
 {
-	res->hits = atomic_swap(&ctx.skel->bss->hits, 0);
+	res->hits = sum_and_reset_counters(ctx.skel->bss->hits);
 }
 
 static void setup_ctx(void)
@@ -137,7 +169,7 @@ static void trigger_fmodret_setup(void)
  * GCC doesn't generate stack setup preample for these functions due to them
  * having no input arguments and doing nothing in the body.
  */
-__weak void uprobe_target_nop(void)
+__nocf_check __weak void uprobe_target_nop(void)
 {
 	asm volatile ("nop");
 }
@@ -146,7 +178,7 @@ __weak void opaque_noop_func(void)
 {
 }
 
-__weak int uprobe_target_push(void)
+__nocf_check __weak int uprobe_target_push(void)
 {
 	/* overhead of function call is negligible compared to uprobe
 	 * triggering, so this shouldn't affect benchmark results much
@@ -155,7 +187,7 @@ __weak int uprobe_target_push(void)
 	return 1;
 }
 
-__weak void uprobe_target_ret(void)
+__nocf_check __weak void uprobe_target_ret(void)
 {
 	asm volatile ("");
 }
@@ -164,7 +196,7 @@ static void *uprobe_base_producer(void *input)
 {
 	while (true) {
 		uprobe_target_nop();
-		atomic_inc(&base_hits.value);
+		inc_counter(base_hits);
 	}
 	return NULL;
 }