- Linux-kselftest-mirror - lists.linaro.org

[PATCH net v2 0/2] selftests: mptcp: connect: cover alt modes

by Matthieu Baerts (NGI0)

mptcp_connect.sh can be executed manually with "-m <MODE>" and "-C" to make sure everything works as expected when using "mmap" and "sendfile" modes instead of "poll", and with the MPTCP checksum support. These modes should be validated, but they are not when the selftests are executed via the kselftest helpers. It means that most CIs validating these selftests, like NIPA for the net development trees and LKFT for the stable ones, are not covering these modes. To fix that, new test programs have been added, simply calling mptcp_connect.sh with the right parameters. The first patch can be backported up to v5.6, and the second one up to v5.14. Signed-off-by: Matthieu Baerts (NGI0) <matttbe(a)kernel.org> --- Changes in v2: - force using a different prefix in the subtests to avoid having the same test names in all mptcp_connect*.sh selftests. - Link to v1: https://lore.kernel.org/r/20250714-net-mptcp-sft-connect-alt-v1-0-bf1c5abbe… --- Matthieu Baerts (NGI0) (2): selftests: mptcp: connect: also cover alt modes selftests: mptcp: connect: also cover checksum tools/testing/selftests/net/mptcp/Makefile | 3 ++- tools/testing/selftests/net/mptcp/mptcp_connect_checksum.sh | 5 +++++ tools/testing/selftests/net/mptcp/mptcp_connect_mmap.sh | 5 +++++ tools/testing/selftests/net/mptcp/mptcp_connect_sendfile.sh | 5 +++++ 4 files changed, 17 insertions(+), 1 deletion(-) --- base-commit: b640daa2822a39ff76e70200cb2b7b892b896dce change-id: 20250714-net-mptcp-sft-connect-alt-c1aaf073ef4e Best regards, -- Matthieu Baerts (NGI0) <matttbe(a)kernel.org>

4 months, 3 weeks

4
12
0 0

Re: [PATCH v3 5/7] perf test: Introduce storing logs for shell tests

by Ian Rogers

On Mon, Jul 21, 2025 at 6:27 AM Jakub Brnak <jbrnak(a)redhat.com> wrote: > > From: Veronika Molnarova <vmolnaro(a)redhat.com> > > Create temporary directories for storing log files for shell tests > that could help while debugging. The log files are necessary for > perftool testsuite test cases also. If the variable KEEP_TEST_LOGS > is set keep the logs, else delete them. Is there perhaps a kunit equivalent of log files so we could keep the implementations as similar as possible? Thanks, Ian > Signed-off-by: Michael Petlan <mpetlan(a)redhat.com> > Signed-off-by: Veronika Molnarova <vmolnaro(a)redhat.com> > Signed-off-by: Jakub Brnak <jbrnak(a)redhat.com> > --- > tools/perf/tests/builtin-test.c | 90 ++++++++++++++++++++++++++++++++ > tools/perf/tests/tests-scripts.c | 3 ++ > tools/perf/tests/tests-scripts.h | 1 + > 3 files changed, 94 insertions(+) > > diff --git a/tools/perf/tests/builtin-test.c b/tools/perf/tests/builtin-test.c > index 4e3d2f779b01..89b180798224 100644 > --- a/tools/perf/tests/builtin-test.c > +++ b/tools/perf/tests/builtin-test.c > @@ -6,6 +6,7 @@ > */ > #include <ctype.h> > #include <fcntl.h> > +#include <ftw.h> > #include <errno.h> > #ifdef HAVE_BACKTRACE_SUPPORT > #include <execinfo.h> > @@ -282,6 +283,86 @@ static bool test_exclusive(const struct test_suite *t, int test_case) > return t->test_cases[test_case].exclusive; > } > > +static int delete_file(const char *fpath, const struct stat *sb __maybe_unused, > + int typeflag, struct FTW *ftwbuf) > +{ > + int rv = -1; > + > + /* Stop traversal if going too deep */ > + if (ftwbuf->level > 5) { > + pr_err("Tree traversal reached level %d, stopping.", ftwbuf->level); > + return rv; > + } > + > + /* Remove only expected directories */ > + if (typeflag == FTW_D || typeflag == FTW_DP){ > + const char *dirname = fpath + ftwbuf->base; > + > + if (strcmp(dirname, "logs") && strcmp(dirname, "examples") && > + strcmp(dirname, "header_tar") && strncmp(dirname, "perf_", 5)) { > + pr_err("Unknown directory %s", dirname); > + return rv; > + } > + } > + > + /* Attempt to remove the file */ > + rv = remove(fpath); > + if (rv) > + pr_err("Failed to remove file: %s", fpath); > + > + return rv; > +} > + > +static bool create_logs(struct test_suite *t, int pass){ > + bool store_logs = t->priv && ((struct shell_info*)(t->priv))->store_logs; > + if (pass == 1 && (!test_exclusive(t, 0) || sequential || dont_fork)) { > + /* Sequential and non-exclusive tests run on the first pass. */ > + return store_logs; > + } > + else if (pass != 1 && test_exclusive(t, 0) && !sequential && !dont_fork) { > + /* Exclusive tests without sequential run on the second pass. */ > + return store_logs; > + } > + return false; > +} > + > +static char *setup_shell_logs(const char *name) > +{ > + char template[PATH_MAX]; > + char *temp_dir; > + > + if (snprintf(template, PATH_MAX, "/tmp/perf_test_%s.XXXXXX", name) < 0) { > + pr_err("Failed to create log dir template"); > + return NULL; /* Skip the testsuite */ > + } > + > + temp_dir = mkdtemp(template); > + if (temp_dir) { > + setenv("PERFSUITE_RUN_DIR", temp_dir, 1); > + return strdup(temp_dir); > + } > + else { > + pr_err("Failed to create the temporary directory"); > + } > + > + return NULL; /* Skip the testsuite */ > +} > + > +static void cleanup_shell_logs(char *dirname) > +{ > + char *keep_logs = getenv("PERFTEST_KEEP_LOGS"); > + > + /* Check if logs should be kept or do cleanup */ > + if (dirname) { > + if (!keep_logs || strcmp(keep_logs, "y") != 0) { > + nftw(dirname, delete_file, 8, FTW_DEPTH | FTW_PHYS); > + } > + free(dirname); > + } > + > + unsetenv("PERFSUITE_RUN_DIR"); > +} > + > static bool perf_test__matches(const char *desc, int suite_num, int argc, const char *argv[]) > { > int i; > @@ -626,6 +707,7 @@ static int __cmd_test(struct test_suite **suites, int argc, const char *argv[], > for (struct test_suite **t = suites; *t; t++, curr_suite++) { > int curr_test_case; > bool suite_matched = false; > + char *tmpdir = NULL; > > if (!perf_test__matches(test_description(*t, -1), curr_suite, argc, argv)) { > /* > @@ -655,6 +737,13 @@ static int __cmd_test(struct test_suite **suites, int argc, const char *argv[], > } > > for (unsigned int run = 0; run < runs_per_test; run++) { > + /* Setup temporary log directories for shell test suites */ > + if (create_logs(*t, pass)) { > + tmpdir = setup_shell_logs((*t)->desc); > + > + if (tmpdir == NULL) /* Couldn't create log dir, skip test suite */ > + ((struct shell_info*)((*t)->priv))->has_setup = FAILED_SETUP; > + } > test_suite__for_each_test_case(*t, curr_test_case) { > if (!suite_matched && > !perf_test__matches(test_description(*t, curr_test_case), > @@ -667,6 +756,7 @@ static int __cmd_test(struct test_suite **suites, int argc, const char *argv[], > goto err_out; > } > } > + cleanup_shell_logs(tmpdir); > } > if (!sequential) { > /* Parallel mode starts tests but doesn't finish them. Do that now. */ > diff --git a/tools/perf/tests/tests-scripts.c b/tools/perf/tests/tests-scripts.c > index d680a878800f..d4e382898a30 100644 > --- a/tools/perf/tests/tests-scripts.c > +++ b/tools/perf/tests/tests-scripts.c > @@ -251,6 +251,7 @@ static struct test_suite* prepare_test_suite(int dir_fd) > > test_info->base_path = strdup_check(dirpath); /* Absolute path to dir */ > test_info->has_setup = NO_SETUP; > + test_info->store_logs = false; > > test_suite->priv = test_info; > test_suite->desc = NULL; > @@ -427,6 +428,8 @@ static void append_suits_in_dir(int dir_fd, > continue; > } > > + /* Store logs for testsuite is sub-directories */ > + ((struct shell_info*)(test_suite->priv))->store_logs = true; > if (is_test_script(fd, SHELL_SETUP)) { /* Check for setup existance */ > char *desc = shell_test__description(fd, SHELL_SETUP); > test_suite->desc = desc; /* Set the suite name by the setup description */ > diff --git a/tools/perf/tests/tests-scripts.h b/tools/perf/tests/tests-scripts.h > index da4dcd26140c..41da0a175e4e 100644 > --- a/tools/perf/tests/tests-scripts.h > +++ b/tools/perf/tests/tests-scripts.h > @@ -16,6 +16,7 @@ enum shell_setup { > struct shell_info { > const char *base_path; > enum shell_setup has_setup; > + bool store_logs; > }; > > struct test_suite **create_script_test_suites(void); > -- > 2.50.1 >

4 months, 3 weeks

1
0
0 0

[PATCH v4 0/6] binder: Set up KUnit tests for alloc

by Tiffany Yang

Hello, binder_alloc_selftest provides a robust set of checks for the binder allocator, but it rarely runs because it must hook into a running binder process and block all other binder threads until it completes. The test itself is a good candidate for conversion to KUnit, and it can be further isolated from user processes by using a test-specific lru freelist instead of the global one. This series converts the selftest to KUnit to make it less burdensome to run and to set up a foundation for unit testing future binder_alloc changes. Thanks, Tiffany Tiffany Yang (6): binder: Fix selftest page indexing binder: Store lru freelist in binder_alloc kunit: test: Export kunit_attach_mm() binder: Scaffolding for binder_alloc KUnit tests binder: Convert binder_alloc selftests to KUnit binder: encapsulate individual alloc test cases drivers/android/Kconfig | 15 +- drivers/android/Makefile | 2 +- drivers/android/binder.c | 10 +- drivers/android/binder_alloc.c | 39 +- drivers/android/binder_alloc.h | 14 +- drivers/android/binder_alloc_selftest.c | 306 ----------- drivers/android/binder_internal.h | 4 + drivers/android/tests/.kunitconfig | 7 + drivers/android/tests/Makefile | 6 + drivers/android/tests/binder_alloc_kunit.c | 572 +++++++++++++++++++++ include/kunit/test.h | 12 + lib/kunit/user_alloc.c | 4 +- 12 files changed, 651 insertions(+), 340 deletions(-) delete mode 100644 drivers/android/binder_alloc_selftest.c create mode 100644 drivers/android/tests/.kunitconfig create mode 100644 drivers/android/tests/Makefile create mode 100644 drivers/android/tests/binder_alloc_kunit.c -- 2.50.0.727.gbf7dc18ff4-goog

4 months, 3 weeks

4
13
0 0

[PATCH] selftests/bpf: Add LPM trie microbenchmarks

by Matt Fleming

From: Matt Fleming <mfleming(a)cloudflare.com> Add benchmarks for the standard set of operations: lookup, update, delete. Also, include a benchmark for trie_free() which is known to have terrible performance for maps with many entries. Benchmarks operate on tries without gaps in the key range, i.e. each test begins with a trie with valid keys in the range [0, nr_entries). This is intended to cause maximum branching when traversing the trie. All measurements are recorded inside the kernel to remove syscall overhead. Most benchmarks run an XDP program to generate stats but free needs to collect latencies using fentry/fexit on map_free_deferred() because it's not possible to use fentry directly on lpm_trie.c since commit c83508da5620 ("bpf: Avoid deadlock caused by nested kprobe and fentry bpf programs") and there's no way to create/destroy a map from within an XDP program. Here is example output from an AMD EPYC 9684X 96-Core machine for each of the benchmarks using a trie with 10K entries and a 32-bit prefix length, e.g. $ ./bench lpm-trie-$op \ --prefix_len=32 \ --producers=1 \ --nr_entries=10000 lookup: throughput 7.423 ± 0.023 M ops/s ( 7.423M ops/prod), latency 134.710 ns/op update: throughput 2.643 ± 0.015 M ops/s ( 2.643M ops/prod), latency 378.310 ns/op delete: throughput 0.712 ± 0.008 M ops/s ( 0.712M ops/prod), latency 1405.152 ns/op free: throughput 0.574 ± 0.003 K ops/s ( 0.574K ops/prod), latency 1.743 ms/op Signed-off-by: Matt Fleming <mfleming(a)cloudflare.com> --- tools/testing/selftests/bpf/Makefile | 2 + tools/testing/selftests/bpf/bench.c | 10 + tools/testing/selftests/bpf/bench.h | 1 + .../selftests/bpf/benchs/bench_lpm_trie_map.c | 345 ++++++++++++++++++ .../selftests/bpf/progs/lpm_trie_bench.c | 175 +++++++++ .../selftests/bpf/progs/lpm_trie_map.c | 19 + 6 files changed, 552 insertions(+) create mode 100644 tools/testing/selftests/bpf/benchs/bench_lpm_trie_map.c create mode 100644 tools/testing/selftests/bpf/progs/lpm_trie_bench.c create mode 100644 tools/testing/selftests/bpf/progs/lpm_trie_map.c diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile index 910d8d6402ef..10a5f1d0fa41 100644 --- a/tools/testing/selftests/bpf/Makefile +++ b/tools/testing/selftests/bpf/Makefile @@ -815,6 +815,7 @@ $(OUTPUT)/bench_bpf_hashmap_lookup.o: $(OUTPUT)/bpf_hashmap_lookup.skel.h $(OUTPUT)/bench_htab_mem.o: $(OUTPUT)/htab_mem_bench.skel.h $(OUTPUT)/bench_bpf_crypto.o: $(OUTPUT)/crypto_bench.skel.h $(OUTPUT)/bench_sockmap.o: $(OUTPUT)/bench_sockmap_prog.skel.h +$(OUTPUT)/bench_lpm_trie_map.o: $(OUTPUT)/lpm_trie_bench.skel.h $(OUTPUT)/lpm_trie_map.skel.h $(OUTPUT)/bench.o: bench.h testing_helpers.h $(BPFOBJ) $(OUTPUT)/bench: LDLIBS += -lm $(OUTPUT)/bench: $(OUTPUT)/bench.o \ @@ -836,6 +837,7 @@ $(OUTPUT)/bench: $(OUTPUT)/bench.o \ $(OUTPUT)/bench_htab_mem.o \ $(OUTPUT)/bench_bpf_crypto.o \ $(OUTPUT)/bench_sockmap.o \ + $(OUTPUT)/bench_lpm_trie_map.o \ # $(call msg,BINARY,,$@) $(Q)$(CC) $(CFLAGS) $(LDFLAGS) $(filter %.a %.o,$^) $(LDLIBS) -o $@ diff --git a/tools/testing/selftests/bpf/bench.c b/tools/testing/selftests/bpf/bench.c index ddd73d06a1eb..fd15f60fd5a8 100644 --- a/tools/testing/selftests/bpf/bench.c +++ b/tools/testing/selftests/bpf/bench.c @@ -284,6 +284,7 @@ extern struct argp bench_htab_mem_argp; extern struct argp bench_trigger_batch_argp; extern struct argp bench_crypto_argp; extern struct argp bench_sockmap_argp; +extern struct argp bench_lpm_trie_map_argp; static const struct argp_child bench_parsers[] = { { &bench_ringbufs_argp, 0, "Ring buffers benchmark", 0 }, @@ -299,6 +300,7 @@ static const struct argp_child bench_parsers[] = { { &bench_trigger_batch_argp, 0, "BPF triggering benchmark", 0 }, { &bench_crypto_argp, 0, "bpf crypto benchmark", 0 }, { &bench_sockmap_argp, 0, "bpf sockmap benchmark", 0 }, + { &bench_lpm_trie_map_argp, 0, "LPM trie map benchmark", 0 }, {}, }; @@ -558,6 +560,10 @@ extern const struct bench bench_htab_mem; extern const struct bench bench_crypto_encrypt; extern const struct bench bench_crypto_decrypt; extern const struct bench bench_sockmap; +extern const struct bench bench_lpm_trie_lookup; +extern const struct bench bench_lpm_trie_update; +extern const struct bench bench_lpm_trie_delete; +extern const struct bench bench_lpm_trie_free; static const struct bench *benchs[] = { &bench_count_global, @@ -625,6 +631,10 @@ static const struct bench *benchs[] = { &bench_crypto_encrypt, &bench_crypto_decrypt, &bench_sockmap, + &bench_lpm_trie_lookup, + &bench_lpm_trie_update, + &bench_lpm_trie_delete, + &bench_lpm_trie_free, }; static void find_benchmark(void) diff --git a/tools/testing/selftests/bpf/bench.h b/tools/testing/selftests/bpf/bench.h index 005c401b3e22..bea323820ffb 100644 --- a/tools/testing/selftests/bpf/bench.h +++ b/tools/testing/selftests/bpf/bench.h @@ -46,6 +46,7 @@ struct bench_res { unsigned long gp_ns; unsigned long gp_ct; unsigned int stime; + unsigned long duration_ns; }; struct bench { diff --git a/tools/testing/selftests/bpf/benchs/bench_lpm_trie_map.c b/tools/testing/selftests/bpf/benchs/bench_lpm_trie_map.c new file mode 100644 index 000000000000..ddd7d3669e70 --- /dev/null +++ b/tools/testing/selftests/bpf/benchs/bench_lpm_trie_map.c @@ -0,0 +1,345 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2025 Cloudflare */ + +/* + * All of these benchmarks operate on tries with keys in the range + * [0, args.nr_entries), i.e. there are no gaps or partially filled + * branches of the trie for any key < args.nr_entries. + * + * This gives an idea of worst-case behaviour. + */ + +#include <argp.h> +#include <linux/time64.h> +#include <linux/if_ether.h> +#include "lpm_trie_bench.skel.h" +#include "lpm_trie_map.skel.h" +#include "bench.h" +#include "testing_helpers.h" + +static struct ctx { + struct lpm_trie_bench *bench; +} ctx; + +static struct { + __u32 nr_entries; + __u32 prefixlen; +} args = { + .nr_entries = 10000, + .prefixlen = 32, +}; + +enum { + ARG_NR_ENTRIES = 9000, + ARG_PREFIX_LEN, +}; + +static const struct argp_option opts[] = { + { "nr_entries", ARG_NR_ENTRIES, "NR_ENTRIES", 0, + "Number of unique entries in the LPM trie" }, + { "prefix_len", ARG_PREFIX_LEN, "PREFIX_LEN", 0, + "Number of prefix bits to use in the LPM trie" }, + {}, +}; + +static error_t lpm_parse_arg(int key, char *arg, struct argp_state *state) +{ + long ret; + + switch (key) { + case ARG_NR_ENTRIES: + ret = strtol(arg, NULL, 10); + if (ret < 1 || ret > UINT_MAX) { + fprintf(stderr, "Invalid nr_entries count."); + argp_usage(state); + } + args.nr_entries = ret; + break; + case ARG_PREFIX_LEN: + ret = strtol(arg, NULL, 10); + if (ret < 1 || ret > UINT_MAX) { + fprintf(stderr, "Invalid prefix_len value."); + argp_usage(state); + } + args.prefixlen = ret; + break; + default: + return ARGP_ERR_UNKNOWN; + } + return 0; +} + +const struct argp bench_lpm_trie_map_argp = { + .options = opts, + .parser = lpm_parse_arg, +}; + +static void __lpm_validate(void) +{ + if (env.consumer_cnt != 0) { + fprintf(stderr, "benchmark doesn't support consumer!\n"); + exit(1); + } + + if ((1UL << args.prefixlen) < args.nr_entries) { + fprintf(stderr, "prefix_len value too small for nr_entries!\n"); + exit(1); + }; +} + +enum { OP_LOOKUP = 1, OP_UPDATE, OP_DELETE, OP_FREE }; + +static void lpm_delete_validate(void) +{ + __lpm_validate(); + + if (env.producer_cnt != 1) { + fprintf(stderr, + "lpm-trie-delete requires a single producer!\n"); + exit(1); + } +} + +static void lpm_free_validate(void) +{ + __lpm_validate(); + + if (env.producer_cnt != 1) { + fprintf(stderr, "lpm-trie-free requires a single producer!\n"); + exit(1); + } +} + +static void fill_map(int map_fd) +{ + int i, err; + + for (i = 0; i < args.nr_entries; i++) { + struct trie_key { + __u32 prefixlen; + __u32 data; + } key = { args.prefixlen, i }; + __u32 val = 1; + + err = bpf_map_update_elem(map_fd, &key, &val, BPF_NOEXIST); + if (err) { + fprintf(stderr, "failed to add key %d to map: %d\n", + key.data, -err); + exit(1); + } + } +} + +static void __lpm_setup(void) +{ + ctx.bench = lpm_trie_bench__open_and_load(); + if (!ctx.bench) { + fprintf(stderr, "failed to open skeleton\n"); + exit(1); + } + + ctx.bench->bss->nr_entries = args.nr_entries; + ctx.bench->bss->prefixlen = args.prefixlen; + + if (lpm_trie_bench__attach(ctx.bench)) { + fprintf(stderr, "failed to attach skeleton\n"); + exit(1); + } +} + +static void lpm_setup(void) +{ + int fd; + + __lpm_setup(); + + fd = bpf_map__fd(ctx.bench->maps.trie_map); + fill_map(fd); +} + +static void lpm_lookup_setup(void) +{ + lpm_setup(); + + ctx.bench->bss->op = OP_LOOKUP; +} + +static void lpm_update_setup(void) +{ + lpm_setup(); + + ctx.bench->bss->op = OP_UPDATE; +} + +static void lpm_delete_setup(void) +{ + lpm_setup(); + + ctx.bench->bss->op = OP_DELETE; +} + +static void lpm_free_setup(void) +{ + __lpm_setup(); + ctx.bench->bss->op = OP_FREE; +} + +static void lpm_measure(struct bench_res *res) +{ + res->hits = atomic_swap(&ctx.bench->bss->hits, 0); + res->duration_ns = atomic_swap(&ctx.bench->bss->duration_ns, 0); +} + +/* For LOOKUP, UPDATE, and DELETE */ +static void *lpm_producer(void *unused __always_unused) +{ + int err; + char in[ETH_HLEN]; /* unused */ + + LIBBPF_OPTS(bpf_test_run_opts, opts, .data_in = in, + .data_size_in = sizeof(in), .repeat = 1, ); + + while (true) { + int fd = bpf_program__fd(ctx.bench->progs.run_bench); + err = bpf_prog_test_run_opts(fd, &opts); + if (err) { + fprintf(stderr, "failed to run BPF prog: %d\n", err); + exit(1); + } + + if (opts.retval < 0) { + fprintf(stderr, "BPF prog returned error: %d\n", + opts.retval); + exit(1); + } + + if (ctx.bench->bss->op == OP_DELETE && opts.retval == 1) { + /* trie_map needs to be refilled */ + fill_map(bpf_map__fd(ctx.bench->maps.trie_map)); + } + } + + return NULL; +} + +static void *lpm_free_producer(void *unused __always_unused) +{ + while (true) { + struct lpm_trie_map *skel; + + skel = lpm_trie_map__open_and_load(); + if (!skel) { + fprintf(stderr, "failed to open skeleton\n"); + exit(1); + } + + fill_map(bpf_map__fd(skel->maps.trie_free_map)); + lpm_trie_map__destroy(skel); + } + + return NULL; +} + +static __always_inline double duration_ms(struct bench_res *res) +{ + if (!res->hits) + return 0.0; + + return res->duration_ns / res->hits / NSEC_PER_MSEC; +} + +static void free_ops_report_progress(int iter, struct bench_res *res, + long delta_ns) +{ + double hits_per_sec, hits_per_prod; + double rate_divisor = 1000.0; + char rate = 'K'; + + hits_per_sec = res->hits / (res->duration_ns / (double)NSEC_PER_SEC) / + rate_divisor; + hits_per_prod = hits_per_sec / env.producer_cnt; + + printf("Iter %3d (%7.3lfus): ", iter, + (delta_ns - NSEC_PER_SEC) / 1000.0); + printf("hits %8.3lf%c/s (%7.3lf%c/prod)\n", hits_per_sec, rate, + hits_per_prod, rate); +} + +static void free_ops_report_final(struct bench_res res[], int res_cnt) +{ + double hits_mean = 0.0, hits_stddev = 0.0; + double lat_divisor = 1000000.0; + double rate_divisor = 1000.0; + const char *unit = "ms"; + double latency = 0.0; + char rate = 'K'; + int i; + + for (i = 0; i < res_cnt; i++) { + double val = res[i].hits / rate_divisor / + (res[i].duration_ns / (double)NSEC_PER_SEC); + hits_mean += val / (0.0 + res_cnt); + latency += res[i].duration_ns / res[i].hits / (0.0 + res_cnt); + } + + if (res_cnt > 1) { + for (i = 0; i < res_cnt; i++) { + double val = + res[i].hits / rate_divisor / + (res[i].duration_ns / (double)NSEC_PER_SEC); + hits_stddev += (hits_mean - val) * (hits_mean - val) / + (res_cnt - 1.0); + } + + hits_stddev = sqrt(hits_stddev); + } + printf("Summary: throughput %8.3lf \u00B1 %5.3lf %c ops/s (%7.3lf%c ops/prod), ", + hits_mean, hits_stddev, rate, hits_mean / env.producer_cnt, + rate); + printf("latency %8.3lf %s/op\n", + latency / lat_divisor / env.producer_cnt, unit); +} + +const struct bench bench_lpm_trie_lookup = { + .name = "lpm-trie-lookup", + .argp = &bench_lpm_trie_map_argp, + .validate = __lpm_validate, + .setup = lpm_lookup_setup, + .producer_thread = lpm_producer, + .measure = lpm_measure, + .report_progress = ops_report_progress, + .report_final = ops_report_final, +}; + +const struct bench bench_lpm_trie_update = { + .name = "lpm-trie-update", + .argp = &bench_lpm_trie_map_argp, + .validate = __lpm_validate, + .setup = lpm_update_setup, + .producer_thread = lpm_producer, + .measure = lpm_measure, + .report_progress = ops_report_progress, + .report_final = ops_report_final, +}; + +const struct bench bench_lpm_trie_delete = { + .name = "lpm-trie-delete", + .argp = &bench_lpm_trie_map_argp, + .validate = lpm_delete_validate, + .setup = lpm_delete_setup, + .producer_thread = lpm_producer, + .measure = lpm_measure, + .report_progress = ops_report_progress, + .report_final = ops_report_final, +}; + +const struct bench bench_lpm_trie_free = { + .name = "lpm-trie-free", + .argp = &bench_lpm_trie_map_argp, + .validate = lpm_free_validate, + .setup = lpm_free_setup, + .producer_thread = lpm_free_producer, + .measure = lpm_measure, + .report_progress = free_ops_report_progress, + .report_final = free_ops_report_final, +}; diff --git a/tools/testing/selftests/bpf/progs/lpm_trie_bench.c b/tools/testing/selftests/bpf/progs/lpm_trie_bench.c new file mode 100644 index 000000000000..c335718cc240 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/lpm_trie_bench.c @@ -0,0 +1,175 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2025 Cloudflare */ + +#include <vmlinux.h> +#include <bpf/bpf_tracing.h> +#include <bpf/bpf_helpers.h> +#include <bpf/bpf_core_read.h> +#include "bpf_misc.h" + +#define BPF_OBJ_NAME_LEN 16U +#define MAX_ENTRIES 100000000 +#define NR_LOOPS 10000 + +struct trie_key { + __u32 prefixlen; + __u32 data; +}; + +char _license[] SEC("license") = "GPL"; + +struct { + __uint(type, BPF_MAP_TYPE_HASH); + __uint(max_entries, 512); + __type(key, struct bpf_map *); + __type(value, __u64); +} latency_free_start SEC(".maps"); + +/* Filled by userspace. See fill_map() in bench_lpm_trie_map.c */ +struct { + __uint(type, BPF_MAP_TYPE_LPM_TRIE); + __type(key, struct trie_key); + __type(value, __u32); + __uint(map_flags, BPF_F_NO_PREALLOC); + __uint(max_entries, MAX_ENTRIES); +} trie_map SEC(".maps"); + +long hits; +long duration_ns; + +/* Configured from userspace */ +__u64 nr_entries; +__u32 prefixlen; +__u8 op; + +static __always_inline void atomic_inc(long *cnt) +{ + __atomic_add_fetch(cnt, 1, __ATOMIC_SEQ_CST); +} + +static __always_inline long atomic_swap(long *cnt, long val) +{ + return __atomic_exchange_n(cnt, val, __ATOMIC_SEQ_CST); +} + +SEC("fentry/bpf_map_free_deferred") +int BPF_PROG(trie_free_entry, struct work_struct *work) +{ + struct bpf_map *map = container_of(work, struct bpf_map, work); + const char *name; + u32 map_type; + __u64 val; + + map_type = BPF_CORE_READ(map, map_type); + if (map_type != BPF_MAP_TYPE_LPM_TRIE) + return 0; + + /* + * Ideally we'd have access to the map ID but that's already + * freed before we enter trie_free(). + */ + name = BPF_CORE_READ(map, name); + if (bpf_strncmp(name, BPF_OBJ_NAME_LEN, "trie_free_map")) + return 0; + + val = bpf_ktime_get_ns(); + bpf_map_update_elem(&latency_free_start, &map, &val, BPF_ANY); + + return 0; +} + +SEC("fexit/bpf_map_free_deferred") +int BPF_PROG(trie_free_exit, struct work_struct *work) +{ + struct bpf_map *map = container_of(work, struct bpf_map, work); + __u64 *val; + + val = bpf_map_lookup_elem(&latency_free_start, &map); + if (val) { + __sync_add_and_fetch(&duration_ns, bpf_ktime_get_ns() - *val); + atomic_inc(&hits); + bpf_map_delete_elem(&latency_free_start, &map); + } + + return 0; +} + +static void gen_random_key(struct trie_key *key) +{ + key->prefixlen = prefixlen; + key->data = bpf_get_prandom_u32() % nr_entries; +} + +static int lookup(__u32 index, __u32 *unused) +{ + struct trie_key key; + + gen_random_key(&key); + bpf_map_lookup_elem(&trie_map, &key); + return 0; +} + +static int update(__u32 index, __u32 *unused) +{ + struct trie_key key; + u32 val = bpf_get_prandom_u32(); + + gen_random_key(&key); + bpf_map_update_elem(&trie_map, &key, &val, BPF_EXIST); + return 0; +} + +long deleted_entries; +long refill; + +static int delete (__u32 index, __u32 *unused) +{ + struct trie_key key = { + .data = deleted_entries, + .prefixlen = prefixlen, + }; + + bpf_map_delete_elem(&trie_map, &key); + atomic_inc(&deleted_entries); + + /* Do we need to refill the map? */ + if (deleted_entries >= nr_entries) { + atomic_swap(&refill, 1); + atomic_swap(&deleted_entries, 0); + return 1; + } + + return 0; +} + +SEC("xdp") +int BPF_PROG(run_bench) +{ + u64 start, delta; + bool need_refill = false; + + start = bpf_ktime_get_ns(); + + switch (op) { + case 1: + bpf_loop(NR_LOOPS, lookup, NULL, 0); + break; + case 2: + bpf_loop(NR_LOOPS, update, NULL, 0); + break; + case 3: + bpf_loop(NR_LOOPS, delete, NULL, 0); + need_refill = atomic_swap(&refill, 0); + break; + default: + bpf_printk("invalid benchmark operation\n"); + return -1; + } + + delta = bpf_ktime_get_ns() - start; + + __sync_add_and_fetch(&hits, NR_LOOPS); + __sync_add_and_fetch(&duration_ns, delta); + + return need_refill; +} diff --git a/tools/testing/selftests/bpf/progs/lpm_trie_map.c b/tools/testing/selftests/bpf/progs/lpm_trie_map.c new file mode 100644 index 000000000000..2ab43e2cd6c6 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/lpm_trie_map.c @@ -0,0 +1,19 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ +#include <linux/bpf.h> +#include <bpf/bpf_helpers.h> +#include <bpf/bpf_tracing.h> + +#define MAX_ENTRIES 100000000 + +struct trie_key { + __u32 prefixlen; + __u32 data; +}; + +struct { + __uint(type, BPF_MAP_TYPE_LPM_TRIE); + __type(key, struct trie_key); + __type(value, __u32); + __uint(map_flags, BPF_F_NO_PREALLOC); + __uint(max_entries, MAX_ENTRIES); +} trie_free_map SEC(".maps"); -- 2.34.1

4 months, 3 weeks

3
4
0 0

[PATCH bpf-next v2] selftests/bpf: Add LPM trie microbenchmarks

by Matt Fleming

From: Matt Fleming <mfleming(a)cloudflare.com> Add benchmarks for the standard set of operations: lookup, update, delete. Also, include a benchmark for trie_free() which is known to have terrible performance for maps with many entries. Benchmarks operate on tries without gaps in the key range, i.e. each test begins with a trie with valid keys in the range [0, nr_entries). This is intended to cause maximum branching when traversing the trie. All measurements are recorded inside the kernel to remove syscall overhead. Most benchmarks run an XDP program to generate stats but free needs to collect latencies using fentry/fexit on map_free_deferred() because it's not possible to use fentry directly on lpm_trie.c since commit c83508da5620 ("bpf: Avoid deadlock caused by nested kprobe and fentry bpf programs") and there's no way to create/destroy a map from within an XDP program. Here is example output from an AMD EPYC 9684X 96-Core machine for each of the benchmarks using a trie with 10K entries and a 32-bit prefix length, e.g. $ ./bench lpm-trie-$op \ --prefix_len=32 \ --producers=1 \ --nr_entries=10000 lookup: throughput 7.423 ± 0.023 M ops/s ( 7.423M ops/prod), latency 134.710 ns/op update: throughput 2.643 ± 0.015 M ops/s ( 2.643M ops/prod), latency 378.310 ns/op delete: throughput 0.712 ± 0.008 M ops/s ( 0.712M ops/prod), latency 1405.152 ns/op free: throughput 0.574 ± 0.003 K ops/s ( 0.574K ops/prod), latency 1.743 ms/op Tested-by: Jesper Dangaard Brouer <hawk(a)kernel.org> Reviewed-by: Jesper Dangaard Brouer <hawk(a)kernel.org> Signed-off-by: Matt Fleming <mfleming(a)cloudflare.com> --- Changes in v2: - Add Jesper's Tested-by and Revewied-by tags - Remove use of atomic_*() in favour of __sync_add_and_fetch() - Use a file-local 'deleted_entries' in the DELETE op benchmark and add a comment explaining why non-atomic accesses are safe. - Bump 'hits' with the number of bpf_loop() loops actually executed tools/testing/selftests/bpf/Makefile | 2 + tools/testing/selftests/bpf/bench.c | 10 + tools/testing/selftests/bpf/bench.h | 1 + .../selftests/bpf/benchs/bench_lpm_trie_map.c | 337 ++++++++++++++++++ .../selftests/bpf/progs/lpm_trie_bench.c | 171 +++++++++ .../selftests/bpf/progs/lpm_trie_map.c | 19 + 6 files changed, 540 insertions(+) create mode 100644 tools/testing/selftests/bpf/benchs/bench_lpm_trie_map.c create mode 100644 tools/testing/selftests/bpf/progs/lpm_trie_bench.c create mode 100644 tools/testing/selftests/bpf/progs/lpm_trie_map.c diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile index 910d8d6402ef..10a5f1d0fa41 100644 --- a/tools/testing/selftests/bpf/Makefile +++ b/tools/testing/selftests/bpf/Makefile @@ -815,6 +815,7 @@ $(OUTPUT)/bench_bpf_hashmap_lookup.o: $(OUTPUT)/bpf_hashmap_lookup.skel.h $(OUTPUT)/bench_htab_mem.o: $(OUTPUT)/htab_mem_bench.skel.h $(OUTPUT)/bench_bpf_crypto.o: $(OUTPUT)/crypto_bench.skel.h $(OUTPUT)/bench_sockmap.o: $(OUTPUT)/bench_sockmap_prog.skel.h +$(OUTPUT)/bench_lpm_trie_map.o: $(OUTPUT)/lpm_trie_bench.skel.h $(OUTPUT)/lpm_trie_map.skel.h $(OUTPUT)/bench.o: bench.h testing_helpers.h $(BPFOBJ) $(OUTPUT)/bench: LDLIBS += -lm $(OUTPUT)/bench: $(OUTPUT)/bench.o \ @@ -836,6 +837,7 @@ $(OUTPUT)/bench: $(OUTPUT)/bench.o \ $(OUTPUT)/bench_htab_mem.o \ $(OUTPUT)/bench_bpf_crypto.o \ $(OUTPUT)/bench_sockmap.o \ + $(OUTPUT)/bench_lpm_trie_map.o \ # $(call msg,BINARY,,$@) $(Q)$(CC) $(CFLAGS) $(LDFLAGS) $(filter %.a %.o,$^) $(LDLIBS) -o $@ diff --git a/tools/testing/selftests/bpf/bench.c b/tools/testing/selftests/bpf/bench.c index ddd73d06a1eb..fd15f60fd5a8 100644 --- a/tools/testing/selftests/bpf/bench.c +++ b/tools/testing/selftests/bpf/bench.c @@ -284,6 +284,7 @@ extern struct argp bench_htab_mem_argp; extern struct argp bench_trigger_batch_argp; extern struct argp bench_crypto_argp; extern struct argp bench_sockmap_argp; +extern struct argp bench_lpm_trie_map_argp; static const struct argp_child bench_parsers[] = { { &bench_ringbufs_argp, 0, "Ring buffers benchmark", 0 }, @@ -299,6 +300,7 @@ static const struct argp_child bench_parsers[] = { { &bench_trigger_batch_argp, 0, "BPF triggering benchmark", 0 }, { &bench_crypto_argp, 0, "bpf crypto benchmark", 0 }, { &bench_sockmap_argp, 0, "bpf sockmap benchmark", 0 }, + { &bench_lpm_trie_map_argp, 0, "LPM trie map benchmark", 0 }, {}, }; @@ -558,6 +560,10 @@ extern const struct bench bench_htab_mem; extern const struct bench bench_crypto_encrypt; extern const struct bench bench_crypto_decrypt; extern const struct bench bench_sockmap; +extern const struct bench bench_lpm_trie_lookup; +extern const struct bench bench_lpm_trie_update; +extern const struct bench bench_lpm_trie_delete; +extern const struct bench bench_lpm_trie_free; static const struct bench *benchs[] = { &bench_count_global, @@ -625,6 +631,10 @@ static const struct bench *benchs[] = { &bench_crypto_encrypt, &bench_crypto_decrypt, &bench_sockmap, + &bench_lpm_trie_lookup, + &bench_lpm_trie_update, + &bench_lpm_trie_delete, + &bench_lpm_trie_free, }; static void find_benchmark(void) diff --git a/tools/testing/selftests/bpf/bench.h b/tools/testing/selftests/bpf/bench.h index 005c401b3e22..bea323820ffb 100644 --- a/tools/testing/selftests/bpf/bench.h +++ b/tools/testing/selftests/bpf/bench.h @@ -46,6 +46,7 @@ struct bench_res { unsigned long gp_ns; unsigned long gp_ct; unsigned int stime; + unsigned long duration_ns; }; struct bench { diff --git a/tools/testing/selftests/bpf/benchs/bench_lpm_trie_map.c b/tools/testing/selftests/bpf/benchs/bench_lpm_trie_map.c new file mode 100644 index 000000000000..435b5c7ceee9 --- /dev/null +++ b/tools/testing/selftests/bpf/benchs/bench_lpm_trie_map.c @@ -0,0 +1,337 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2025 Cloudflare */ + +/* + * All of these benchmarks operate on tries with keys in the range + * [0, args.nr_entries), i.e. there are no gaps or partially filled + * branches of the trie for any key < args.nr_entries. + * + * This gives an idea of worst-case behaviour. + */ + +#include <argp.h> +#include <linux/time64.h> +#include <linux/if_ether.h> +#include "lpm_trie_bench.skel.h" +#include "lpm_trie_map.skel.h" +#include "bench.h" +#include "testing_helpers.h" + +static struct ctx { + struct lpm_trie_bench *bench; +} ctx; + +static struct { + __u32 nr_entries; + __u32 prefixlen; +} args = { + .nr_entries = 10000, + .prefixlen = 32, +}; + +enum { + ARG_NR_ENTRIES = 9000, + ARG_PREFIX_LEN, +}; + +static const struct argp_option opts[] = { + { "nr_entries", ARG_NR_ENTRIES, "NR_ENTRIES", 0, + "Number of unique entries in the LPM trie" }, + { "prefix_len", ARG_PREFIX_LEN, "PREFIX_LEN", 0, + "Number of prefix bits to use in the LPM trie" }, + {}, +}; + +static error_t lpm_parse_arg(int key, char *arg, struct argp_state *state) +{ + long ret; + + switch (key) { + case ARG_NR_ENTRIES: + ret = strtol(arg, NULL, 10); + if (ret < 1 || ret > UINT_MAX) { + fprintf(stderr, "Invalid nr_entries count."); + argp_usage(state); + } + args.nr_entries = ret; + break; + case ARG_PREFIX_LEN: + ret = strtol(arg, NULL, 10); + if (ret < 1 || ret > UINT_MAX) { + fprintf(stderr, "Invalid prefix_len value."); + argp_usage(state); + } + args.prefixlen = ret; + break; + default: + return ARGP_ERR_UNKNOWN; + } + return 0; +} + +const struct argp bench_lpm_trie_map_argp = { + .options = opts, + .parser = lpm_parse_arg, +}; + +static void __lpm_validate(void) +{ + if (env.consumer_cnt != 0) { + fprintf(stderr, "benchmark doesn't support consumer!\n"); + exit(1); + } + + if ((1UL << args.prefixlen) < args.nr_entries) { + fprintf(stderr, "prefix_len value too small for nr_entries!\n"); + exit(1); + }; +} + +enum { OP_LOOKUP = 1, OP_UPDATE, OP_DELETE, OP_FREE }; + +static void lpm_delete_validate(void) +{ + __lpm_validate(); + + if (env.producer_cnt != 1) { + fprintf(stderr, + "lpm-trie-delete requires a single producer!\n"); + exit(1); + } +} + +static void lpm_free_validate(void) +{ + __lpm_validate(); + + if (env.producer_cnt != 1) { + fprintf(stderr, "lpm-trie-free requires a single producer!\n"); + exit(1); + } +} + +static void fill_map(int map_fd) +{ + int i, err; + + for (i = 0; i < args.nr_entries; i++) { + struct trie_key { + __u32 prefixlen; + __u32 data; + } key = { args.prefixlen, i }; + __u32 val = 1; + + err = bpf_map_update_elem(map_fd, &key, &val, BPF_NOEXIST); + if (err) { + fprintf(stderr, "failed to add key %d to map: %d\n", + key.data, -err); + exit(1); + } + } +} + +static void __lpm_setup(void) +{ + ctx.bench = lpm_trie_bench__open_and_load(); + if (!ctx.bench) { + fprintf(stderr, "failed to open skeleton\n"); + exit(1); + } + + ctx.bench->bss->nr_entries = args.nr_entries; + ctx.bench->bss->prefixlen = args.prefixlen; + + if (lpm_trie_bench__attach(ctx.bench)) { + fprintf(stderr, "failed to attach skeleton\n"); + exit(1); + } +} + +static void lpm_setup(void) +{ + int fd; + + __lpm_setup(); + + fd = bpf_map__fd(ctx.bench->maps.trie_map); + fill_map(fd); +} + +static void lpm_lookup_setup(void) +{ + lpm_setup(); + + ctx.bench->bss->op = OP_LOOKUP; +} + +static void lpm_update_setup(void) +{ + lpm_setup(); + + ctx.bench->bss->op = OP_UPDATE; +} + +static void lpm_delete_setup(void) +{ + lpm_setup(); + + ctx.bench->bss->op = OP_DELETE; +} + +static void lpm_free_setup(void) +{ + __lpm_setup(); + ctx.bench->bss->op = OP_FREE; +} + +static void lpm_measure(struct bench_res *res) +{ + res->hits = atomic_swap(&ctx.bench->bss->hits, 0); + res->duration_ns = atomic_swap(&ctx.bench->bss->duration_ns, 0); +} + +/* For LOOKUP, UPDATE, and DELETE */ +static void *lpm_producer(void *unused __always_unused) +{ + int err; + char in[ETH_HLEN]; /* unused */ + + LIBBPF_OPTS(bpf_test_run_opts, opts, .data_in = in, + .data_size_in = sizeof(in), .repeat = 1, ); + + while (true) { + int fd = bpf_program__fd(ctx.bench->progs.run_bench); + err = bpf_prog_test_run_opts(fd, &opts); + if (err) { + fprintf(stderr, "failed to run BPF prog: %d\n", err); + exit(1); + } + + if (opts.retval < 0) { + fprintf(stderr, "BPF prog returned error: %d\n", + opts.retval); + exit(1); + } + + if (ctx.bench->bss->op == OP_DELETE && opts.retval == 1) { + /* trie_map needs to be refilled */ + fill_map(bpf_map__fd(ctx.bench->maps.trie_map)); + } + } + + return NULL; +} + +static void *lpm_free_producer(void *unused __always_unused) +{ + while (true) { + struct lpm_trie_map *skel; + + skel = lpm_trie_map__open_and_load(); + if (!skel) { + fprintf(stderr, "failed to open skeleton\n"); + exit(1); + } + + fill_map(bpf_map__fd(skel->maps.trie_free_map)); + lpm_trie_map__destroy(skel); + } + + return NULL; +} + +static void free_ops_report_progress(int iter, struct bench_res *res, + long delta_ns) +{ + double hits_per_sec, hits_per_prod; + double rate_divisor = 1000.0; + char rate = 'K'; + + hits_per_sec = res->hits / (res->duration_ns / (double)NSEC_PER_SEC) / + rate_divisor; + hits_per_prod = hits_per_sec / env.producer_cnt; + + printf("Iter %3d (%7.3lfus): ", iter, + (delta_ns - NSEC_PER_SEC) / 1000.0); + printf("hits %8.3lf%c/s (%7.3lf%c/prod)\n", hits_per_sec, rate, + hits_per_prod, rate); +} + +static void free_ops_report_final(struct bench_res res[], int res_cnt) +{ + double hits_mean = 0.0, hits_stddev = 0.0; + double lat_divisor = 1000000.0; + double rate_divisor = 1000.0; + const char *unit = "ms"; + double latency = 0.0; + char rate = 'K'; + int i; + + for (i = 0; i < res_cnt; i++) { + double val = res[i].hits / rate_divisor / + (res[i].duration_ns / (double)NSEC_PER_SEC); + hits_mean += val / (0.0 + res_cnt); + latency += res[i].duration_ns / res[i].hits / (0.0 + res_cnt); + } + + if (res_cnt > 1) { + for (i = 0; i < res_cnt; i++) { + double val = + res[i].hits / rate_divisor / + (res[i].duration_ns / (double)NSEC_PER_SEC); + hits_stddev += (hits_mean - val) * (hits_mean - val) / + (res_cnt - 1.0); + } + + hits_stddev = sqrt(hits_stddev); + } + printf("Summary: throughput %8.3lf \u00B1 %5.3lf %c ops/s (%7.3lf%c ops/prod), ", + hits_mean, hits_stddev, rate, hits_mean / env.producer_cnt, + rate); + printf("latency %8.3lf %s/op\n", + latency / lat_divisor / env.producer_cnt, unit); +} + +const struct bench bench_lpm_trie_lookup = { + .name = "lpm-trie-lookup", + .argp = &bench_lpm_trie_map_argp, + .validate = __lpm_validate, + .setup = lpm_lookup_setup, + .producer_thread = lpm_producer, + .measure = lpm_measure, + .report_progress = ops_report_progress, + .report_final = ops_report_final, +}; + +const struct bench bench_lpm_trie_update = { + .name = "lpm-trie-update", + .argp = &bench_lpm_trie_map_argp, + .validate = __lpm_validate, + .setup = lpm_update_setup, + .producer_thread = lpm_producer, + .measure = lpm_measure, + .report_progress = ops_report_progress, + .report_final = ops_report_final, +}; + +const struct bench bench_lpm_trie_delete = { + .name = "lpm-trie-delete", + .argp = &bench_lpm_trie_map_argp, + .validate = lpm_delete_validate, + .setup = lpm_delete_setup, + .producer_thread = lpm_producer, + .measure = lpm_measure, + .report_progress = ops_report_progress, + .report_final = ops_report_final, +}; + +const struct bench bench_lpm_trie_free = { + .name = "lpm-trie-free", + .argp = &bench_lpm_trie_map_argp, + .validate = lpm_free_validate, + .setup = lpm_free_setup, + .producer_thread = lpm_free_producer, + .measure = lpm_measure, + .report_progress = free_ops_report_progress, + .report_final = free_ops_report_final, +}; diff --git a/tools/testing/selftests/bpf/progs/lpm_trie_bench.c b/tools/testing/selftests/bpf/progs/lpm_trie_bench.c new file mode 100644 index 000000000000..1a138e21e156 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/lpm_trie_bench.c @@ -0,0 +1,171 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2025 Cloudflare */ + +#include <vmlinux.h> +#include <bpf/bpf_tracing.h> +#include <bpf/bpf_helpers.h> +#include <bpf/bpf_core_read.h> +#include "bpf_misc.h" + +#define BPF_OBJ_NAME_LEN 16U +#define MAX_ENTRIES 100000000 +#define NR_LOOPS 10000 + +struct trie_key { + __u32 prefixlen; + __u32 data; +}; + +char _license[] SEC("license") = "GPL"; + +struct { + __uint(type, BPF_MAP_TYPE_HASH); + __uint(max_entries, 512); + __type(key, struct bpf_map *); + __type(value, __u64); +} latency_free_start SEC(".maps"); + +/* Filled by userspace. See fill_map() in bench_lpm_trie_map.c */ +struct { + __uint(type, BPF_MAP_TYPE_LPM_TRIE); + __type(key, struct trie_key); + __type(value, __u32); + __uint(map_flags, BPF_F_NO_PREALLOC); + __uint(max_entries, MAX_ENTRIES); +} trie_map SEC(".maps"); + +long hits; +long duration_ns; + +/* Configured from userspace */ +__u32 nr_entries; +__u32 prefixlen; +__u8 op; + +SEC("fentry/bpf_map_free_deferred") +int BPF_PROG(trie_free_entry, struct work_struct *work) +{ + struct bpf_map *map = container_of(work, struct bpf_map, work); + const char *name; + u32 map_type; + __u64 val; + + map_type = BPF_CORE_READ(map, map_type); + if (map_type != BPF_MAP_TYPE_LPM_TRIE) + return 0; + + /* + * Ideally we'd have access to the map ID but that's already + * freed before we enter trie_free(). + */ + name = BPF_CORE_READ(map, name); + if (bpf_strncmp(name, BPF_OBJ_NAME_LEN, "trie_free_map")) + return 0; + + val = bpf_ktime_get_ns(); + bpf_map_update_elem(&latency_free_start, &map, &val, BPF_ANY); + + return 0; +} + +SEC("fexit/bpf_map_free_deferred") +int BPF_PROG(trie_free_exit, struct work_struct *work) +{ + struct bpf_map *map = container_of(work, struct bpf_map, work); + __u64 *val; + + val = bpf_map_lookup_elem(&latency_free_start, &map); + if (val) { + __sync_add_and_fetch(&duration_ns, bpf_ktime_get_ns() - *val); + __sync_add_and_fetch(&hits, 1); + bpf_map_delete_elem(&latency_free_start, &map); + } + + return 0; +} + +static void gen_random_key(struct trie_key *key) +{ + key->prefixlen = prefixlen; + key->data = bpf_get_prandom_u32() % nr_entries; +} + +static int lookup(__u32 index, __u32 *unused) +{ + struct trie_key key; + + gen_random_key(&key); + bpf_map_lookup_elem(&trie_map, &key); + return 0; +} + +static int update(__u32 index, __u32 *unused) +{ + struct trie_key key; + u32 val = bpf_get_prandom_u32(); + + gen_random_key(&key); + bpf_map_update_elem(&trie_map, &key, &val, BPF_EXIST); + return 0; +} + +static __u32 deleted_entries; + +static int delete (__u32 index, bool *need_refill) +{ + struct trie_key key = { + .data = deleted_entries, + .prefixlen = prefixlen, + }; + + bpf_map_delete_elem(&trie_map, &key); + + /* Do we need to refill the map? */ + if (++deleted_entries == nr_entries) { + /* + * Atomicity isn't required because DELETE only supports + * one producer running concurrently. What we need is a + * way to track how many entries have been deleted from + * the trie between consecutive invocations of the BPF + * prog because a single bpf_loop() call might not + * delete all entries, e.g. when NR_LOOPS < nr_entries. + */ + deleted_entries = 0; + *need_refill = true; + return 1; + } + + return 0; +} + +SEC("xdp") +int BPF_PROG(run_bench) +{ + bool need_refill = false; + u64 start, delta; + int loops; + + start = bpf_ktime_get_ns(); + + switch (op) { + case 1: + loops = bpf_loop(NR_LOOPS, lookup, NULL, 0); + break; + case 2: + loops = bpf_loop(NR_LOOPS, update, NULL, 0); + break; + case 3: + loops = bpf_loop(NR_LOOPS, delete, &need_refill, 0); + break; + default: + bpf_printk("invalid benchmark operation\n"); + return -1; + } + + delta = bpf_ktime_get_ns() - start; + + __sync_add_and_fetch(&duration_ns, delta); + __sync_add_and_fetch(&hits, loops); + + return need_refill; +} diff --git a/tools/testing/selftests/bpf/progs/lpm_trie_map.c b/tools/testing/selftests/bpf/progs/lpm_trie_map.c new file mode 100644 index 000000000000..2ab43e2cd6c6 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/lpm_trie_map.c @@ -0,0 +1,19 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ +#include <linux/bpf.h> +#include <bpf/bpf_helpers.h> +#include <bpf/bpf_tracing.h> + +#define MAX_ENTRIES 100000000 + +struct trie_key { + __u32 prefixlen; + __u32 data; +}; + +struct { + __uint(type, BPF_MAP_TYPE_LPM_TRIE); + __type(key, struct trie_key); + __type(value, __u32); + __uint(map_flags, BPF_F_NO_PREALLOC); + __uint(max_entries, MAX_ENTRIES); +} trie_free_map SEC(".maps"); -- 2.34.1

4 months, 3 weeks

1
0
0 0

[PATCH nf-next v4 0/2] Add IPIP flowtable SW acceleratio

by Lorenzo Bianconi

Introduce SW acceleration for IPIP tunnels in the netfilter flowtable infrastructure. --- Changes in v4: - Use the hash value of the saddr, daddr and protocol of outer IP header as encapsulation id. - Link to v3: https://lore.kernel.org/r/20250703-nf-flowtable-ipip-v3-0-880afd319b9f@kern… Changes in v3: - Add outer IP header sanity checks - target nf-next tree instead of net-next - Link to v2: https://lore.kernel.org/r/20250627-nf-flowtable-ipip-v2-0-c713003ce75b@kern… Changes in v2: - Introduce IPIP flowtable selftest - Link to v1: https://lore.kernel.org/r/20250623-nf-flowtable-ipip-v1-1-2853596e3941@kern… --- Lorenzo Bianconi (2): net: netfilter: Add IPIP flowtable SW acceleration selftests: netfilter: nft_flowtable.sh: Add IPIP flowtable selftest include/linux/netdevice.h | 1 + net/ipv4/ipip.c | 25 +++++++++++ net/netfilter/nf_flow_table_ip.c | 48 +++++++++++++++++++++- net/netfilter/nft_flow_offload.c | 1 + .../selftests/net/netfilter/nft_flowtable.sh | 40 ++++++++++++++++++ 5 files changed, 113 insertions(+), 2 deletions(-) --- base-commit: d61f6cb6f6ef3c70d2ccc0d9c85c508cb8017da9 change-id: 20250623-nf-flowtable-ipip-1b3d7b08d067 Best regards, -- Lorenzo Bianconi <lorenzo(a)kernel.org>

4 months, 3 weeks

2
7
0 0

[PATCH v2 0/2] rust: minor idiomatic fixes to doctest generator

by Tamir Duberstein

Please see individual commit messages. Signed-off-by: Tamir Duberstein <tamird(a)gmail.com> --- Changes in v2: - rustfmt. - Alice's RB. - Add second patch to emit information in panic rather than separately to stderr. - Link to v1: https://lore.kernel.org/r/20250527-idiomatic-match-slice-v1-1-34b0b1d1d58c@… --- Tamir Duberstein (2): rust: replace length checks with match rust: emit path candidates in panic message scripts/rustdoc_test_gen.rs | 33 +++++++++++++++++---------------- 1 file changed, 17 insertions(+), 16 deletions(-) --- base-commit: 1ce98bb2bb30713ec4374ef11ead0d7d3e856766 change-id: 20250527-idiomatic-match-slice-26a79d100e4d Best regards, -- Tamir Duberstein <tamird(a)gmail.com>

4 months, 3 weeks

2
8
0 0

[PATCH v2 00/15] selftests/futex: Refactor tests to use kselftest_harness.h

by André Almeida

This patch series refactors all futex selftests to use kselftest_harness.h instead of futex's logging.h, as discussed here [1]. This allows to remove a lot of boilerplate code and to simplify some parts of the test logic, mainly when the test needs to exit early. The result of this is more than 500 lines removed from tools/testing/selftests/futex/. Also, this enables new tests to use kselftest.h features like ASSERT_s and such. There are some caveats around this refactor: - logging.h had verbosity levels, while kselftest_harness.h doesn't. I created a new print function called ksft_print_dbg_msg() that prints the message if the user uses the -d flag, so now there's an equivalent of this feature. - futex_requeue_pi test accepted command line arguments to be used as test parameters (e.g. ./futex_requeue_pi -b -l -t 500000). This doesn't work with kselftest_harness.h because there's no straightforward way to send command line arguments to the test. I used FIXTURE_VARIANT() to achieve the same result, but now the parameters live inside of the test file, instead of on functional/run.sh. This increased a little bit the number of test cases for futex_requeue_pi, from 22 to 24. - test_harness_run() calls mmap(MAP_SHARED) before running the test and this has caused a side effect on test futex_numa_mpol.c. This test also calls mmap() and then try to access an address out of boundaries of this mapped memory for a "Memory out of range" subtest, where the kernel should return -EACCESS. After the refactor, the test address might be fall inside the first memory mapped region, thus being a valid address and succeeding the syscall, making the test fail. To fix that, I created a small "buffer zone" with mmap(PROT_NONE) between both mmaps. I have compared the results of run.sh before and after this patchset and didn't find any regression from the test results. Thanks, André [1] https://lore.kernel.org/lkml/87ecv6p364.ffs@tglx/ --- Changes in v2: - Rebased on top of tip/master - Dropped priv_hash global test variant now that this feature was dropped - Added include <stdbool.h> in the first patch - Link to v1: https://lore.kernel.org/r/20250704-tonyk-robust_test_cleanup-v1-0-c0ff4f24c… --- André Almeida (15): selftests: kselftest: Create ksft_print_dbg_msg() selftests/futex: Refactor futex_requeue_pi with kselftest_harness.h selftests/futex: Refactor futex_requeue_pi_mismatched_ops with kselftest_harness.h selftests/futex: Refactor futex_requeue_pi_signal_restart with kselftest_harness.h selftests/futex: Refactor futex_wait_timeout with kselftest_harness.h selftests/futex: Refactor futex_wait_wouldblock with kselftest_harness.h selftests/futex: Refactor futex_wait_unitialized_heap with kselftest_harness.h selftests/futex: Refactor futex_wait_private_mapped_file with kselftest_harness.h selftests/futex: Refactor futex_wait with kselftest_harness.h selftests/futex: Refactor futex_requeue with kselftest_harness.h selftests/futex: Refactor futex_waitv with kselftest_harness.h selftests/futex: Refactor futex_priv_hash with kselftest_harness.h selftests/futex: Refactor futex_numa_mpol with kselftest_harness.h selftests/futex: Drop logging.h include from futex_numa selftests/futex: Remove logging.h file tools/testing/selftests/futex/functional/Makefile | 3 +- .../selftests/futex/functional/futex_numa.c | 3 +- .../selftests/futex/functional/futex_numa_mpol.c | 57 ++--- .../selftests/futex/functional/futex_priv_hash.c | 49 +--- .../selftests/futex/functional/futex_requeue.c | 76 ++---- .../selftests/futex/functional/futex_requeue_pi.c | 261 ++++++++++----------- .../functional/futex_requeue_pi_mismatched_ops.c | 80 ++----- .../functional/futex_requeue_pi_signal_restart.c | 129 +++------- .../selftests/futex/functional/futex_wait.c | 103 +++----- .../functional/futex_wait_private_mapped_file.c | 83 ++----- .../futex/functional/futex_wait_timeout.c | 139 +++++------ .../functional/futex_wait_uninitialized_heap.c | 76 ++---- .../futex/functional/futex_wait_wouldblock.c | 75 ++---- .../selftests/futex/functional/futex_waitv.c | 98 ++++---- tools/testing/selftests/futex/functional/run.sh | 62 +---- tools/testing/selftests/futex/include/logging.h | 148 ------------ tools/testing/selftests/kselftest.h | 14 ++ tools/testing/selftests/kselftest_harness.h | 13 +- 18 files changed, 465 insertions(+), 1004 deletions(-) --- base-commit: ed0272f0675f31642c3d445a596b544de9db405b change-id: 20250703-tonyk-robust_test_cleanup-d1f3406365d9 Best regards, -- André Almeida <andrealmeid(a)igalia.com>

4 months, 3 weeks

1
15
0 0

[PATCH v8 0/6] use per-vma locks for /proc/pid/maps reads

by Suren Baghdasaryan

Reading /proc/pid/maps requires read-locking mmap_lock which prevents any other task from concurrently modifying the address space. This guarantees coherent reporting of virtual address ranges, however it can block important updates from happening. Oftentimes /proc/pid/maps readers are low priority monitoring tasks and them blocking high priority tasks results in priority inversion. Locking the entire address space is required to present fully coherent picture of the address space, however even current implementation does not strictly guarantee that by outputting vmas in page-size chunks and dropping mmap_lock in between each chunk. Address space modifications are possible while mmap_lock is dropped and userspace reading the content is expected to deal with possible concurrent address space modifications. Considering these relaxed rules, holding mmap_lock is not strictly needed as long as we can guarantee that a concurrently modified vma is reported either in its original form or after it was modified. This patchset switches from holding mmap_lock while reading /proc/pid/maps to taking per-vma locks as we walk the vma tree. This reduces the contention with tasks modifying the address space because they would have to contend for the same vma as opposed to the entire address space. Previous version of this patchset [1] tried to perform /proc/pid/maps reading under RCU, however its implementation is quite complex and the results are worse than the new version because it still relied on mmap_lock speculation which retries if any part of the address space gets modified. New implementaion is both simpler and results in less contention. Note that similar approach would not work for /proc/pid/smaps reading as it also walks the page table and that's not RCU-safe. Paul McKenney's designed a test [2] to measure mmap/munmap latencies while concurrently reading /proc/pid/maps. The test has a pair of processes scanning /proc/PID/maps, and another process unmapping and remapping 4K pages from a 128MB range of anonymous memory. At the end of each 10 second run, the latency of each mmap() or munmap() operation is measured, and for each run the maximum and mean latency is printed. The map/unmap process is started first, its PID is passed to the scanners, and then the map/unmap process waits until both scanners are running before starting its timed test. The scanners keep scanning until the specified /proc/PID/maps file disappears. The latest results from Paul: Stock mm-unstable, all of the runs had maximum latencies in excess of 0.5 milliseconds, and with 80% of the runs' latencies exceeding a full millisecond, and ranging up beyond 4 full milliseconds. In contrast, 99% of the runs with this patch series applied had maximum latencies of less than 0.5 milliseconds, with the single outlier at only 0.608 milliseconds. From a median-performance (as opposed to maximum-latency) viewpoint, this patch series also looks good, with stock mm weighing in at 11 microseconds and patch series at 6 microseconds, better than a 2x improvement. Before the change: ./run-proc-vs-map.sh --nsamples 100 --rawdata -- --busyduration 2 0.011 0.008 0.521 0.011 0.008 0.552 0.011 0.008 0.590 0.011 0.008 0.660 ... 0.011 0.015 2.987 0.011 0.015 3.038 0.011 0.016 3.431 0.011 0.016 4.707 After the change: ./run-proc-vs-map.sh --nsamples 100 --rawdata -- --busyduration 2 0.006 0.005 0.026 0.006 0.005 0.029 0.006 0.005 0.034 0.006 0.005 0.035 ... 0.006 0.006 0.421 0.006 0.006 0.423 0.006 0.006 0.439 0.006 0.006 0.608 The patchset also adds a number of tests to check for /proc/pid/maps data coherency. They are designed to detect any unexpected data tearing while performing some common address space modifications (vma split, resize and remap). Even before these changes, reading /proc/pid/maps might have inconsistent data because the file is read page-by-page with mmap_lock being dropped between the pages. An example of user-visible inconsistency can be that the same vma is printed twice: once before it was modified and then after the modifications. For example if vma was extended, it might be found and reported twice. What is not expected is to see a gap where there should have been a vma both before and after modification. This patchset increases the chances of such tearing, therefore it's even more important now to test for unexpected inconsistencies. In [3] Lorenzo identified the following possible vma merging/splitting scenarios: Merges with changes to existing vmas: 1 Merge both - mapping a vma over another one and between two vmas which can be merged after this replacement; 2. Merge left full - mapping a vma at the end of an existing one and completely over its right neighbor; 3. Merge left partial - mapping a vma at the end of an existing one and partially over its right neighbor; 4. Merge right full - mapping a vma before the start of an existing one and completely over its left neighbor; 5. Merge right partial - mapping a vma before the start of an existing one and partially over its left neighbor; Merges without changes to existing vmas: 6. Merge both - mapping a vma into a gap between two vmas which can be merged after the insertion; 7. Merge left - mapping a vma at the end of an existing one; 8. Merge right - mapping a vma before the start end of an existing one; Splits 9. Split with new vma at the lower address; 10. Split with new vma at the higher address; If such merges or splits happen concurrently with the /proc/maps reading we might report a vma twice, once before the modification and once after it is modified: Case 1 might report overwritten and previous vma along with the final merged vma; Case 2 might report previous and the final merged vma; Case 3 might cause us to retry once we detect the temporary gap caused by shrinking of the right neighbor; Case 4 might report overritten and the final merged vma; Case 5 might cause us to retry once we detect the temporary gap caused by shrinking of the left neighbor; Case 6 might report previous vma and the gap along with the final marged vma; Case 7 might report previous and the final merged vma; Case 8 might report the original gap and the final merged vma covering the gap; Case 9 might cause us to retry once we detect the temporary gap caused by shrinking of the original vma at the vma start; Case 10 might cause us to retry once we detect the temporary gap caused by shrinking of the original vma at the vma end; In all these cases the retry mechanism prevents us from reporting possible temporary gaps. Changes since v7 [4]: - Refactored tests to use kselftest harness, per David Hildenbrand and Lorenzo Stoakes - Removed PROCMAP_QUERY selftest, per David Hildenbrand and Lorenzo Stoakes - Added Acked-by, per David Hildenbrand - Replaced sentinels values with named definitions, per Vlastimil Babka - Added Reviewed-by, per Vlastimil Babka !!! NOTES FOR APPLYING THE PATCHSET !!! Applies cleanly over mm-unstable after reverting v7 version of this patchset (from 94951ab6fe6f to e47914e6c28f in mm-unstable). [1] https://lore.kernel.org/all/20250418174959.1431962-1-surenb@google.com/ [2] https://github.com/paulmckrcu/proc-mmap_sem-test [3] https://lore.kernel.org/all/e1863f40-39ab-4e5b-984a-c48765ffde1c@lucifer.lo… [4] https://lore.kernel.org/all/20250716030557.1547501-1-surenb@google.com/ Suren Baghdasaryan (6): selftests/proc: add /proc/pid/maps tearing from vma split test selftests/proc: extend /proc/pid/maps tearing test to include vma resizing selftests/proc: extend /proc/pid/maps tearing test to include vma remapping selftests/proc: add verbose mode for /proc/pid/maps tearing tests fs/proc/task_mmu: remove conversion of seq_file position to unsigned fs/proc/task_mmu: read proc/pid/maps under per-vma lock fs/proc/internal.h | 5 + fs/proc/task_mmu.c | 158 +++- include/linux/mmap_lock.h | 11 + mm/madvise.c | 3 +- mm/mmap_lock.c | 93 +++ tools/testing/selftests/proc/.gitignore | 1 + tools/testing/selftests/proc/Makefile | 1 + tools/testing/selftests/proc/proc-maps-race.c | 741 ++++++++++++++++++ 8 files changed, 997 insertions(+), 16 deletions(-) create mode 100644 tools/testing/selftests/proc/proc-maps-race.c -- 2.50.0.727.gbf7dc18ff4-goog

4 months, 3 weeks

2
7
0 0

Start Campaign Outreach with Fresh Data from GSX 2025

by Ben Graham

Hi , Interested in getting the GSX 2025 attendee list? Expo Name: Global Security Exchange (GSX) 2025 Total Number of records: 17,000 records List includes: Company Name, Contact Name, Job Title, Mailing Address, Phone, Emails, etc. Are you considering buying these leads? If yes, I can send you the pricing information. Awaiting your message Regards Ben Graham Demand Generation Manager US Marketing Data Inc., Please reply with REMOVE if you don't wish to receive further emails

4 months, 3 weeks

1
0
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-kselftest-mirror