From: Carsten Haitzler <[email protected]>
This adds a test harness and tests to run perf record and examine the
resuling output when coresight is enabled on arm64 and check the
resulting quality of the output as part of perf test. These tests use
various tools to produce output from perf record then measure some key
specific aspects of that data to see if the data exists at all and
contains key aspects such as measuring some data for every thread of
a test or produces sufficient data for large exeuction runs of a large
executable. etc.
Signed-off-by: Carsten Haitzler <[email protected]>
---
MAINTAINERS | 4 +
tools/perf/.gitignore | 6 +-
tools/perf/Documentation/arm-coresight.txt | 140 ++++++++++++++++++
tools/perf/Makefile.perf | 14 +-
tools/perf/tests/shell/coresight/Makefile | 30 ++++
.../tests/shell/coresight/Makefile.miniconfig | 23 +++
.../shell/coresight/asm_pure_loop/.gitignore | 1 +
.../shell/coresight/asm_pure_loop/Makefile | 30 ++++
.../coresight/asm_pure_loop/asm_pure_loop.S | 28 ++++
.../shell/coresight/memcpy_thread/.gitignore | 1 +
.../shell/coresight/memcpy_thread/Makefile | 29 ++++
.../coresight/memcpy_thread/memcpy_thread.c | 79 ++++++++++
.../shell/coresight/thread_loop/.gitignore | 1 +
.../shell/coresight/thread_loop/Makefile | 29 ++++
.../shell/coresight/thread_loop/thread_loop.c | 86 +++++++++++
.../coresight/unroll_loop_thread/.gitignore | 1 +
.../coresight/unroll_loop_thread/Makefile | 29 ++++
.../unroll_loop_thread/unroll_loop_thread.c | 74 +++++++++
.../tests/shell/coresight_asm_pure_loop.sh | 18 +++
.../shell/coresight_memcpy_thread_16k_10.sh | 18 +++
.../coresight_thread_loop_check_tid_10.sh | 19 +++
.../coresight_thread_loop_check_tid_2.sh | 19 +++
.../shell/coresight_unroll_loop_thread_10.sh | 18 +++
tools/perf/tests/shell/lib/coresight.sh | 130 ++++++++++++++++
24 files changed, 823 insertions(+), 4 deletions(-)
create mode 100644 tools/perf/Documentation/arm-coresight.txt
create mode 100644 tools/perf/tests/shell/coresight/Makefile
create mode 100644 tools/perf/tests/shell/coresight/Makefile.miniconfig
create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/Makefile
create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
create mode 100644 tools/perf/tests/shell/coresight/thread_loop/.gitignore
create mode 100644 tools/perf/tests/shell/coresight/thread_loop/Makefile
create mode 100644 tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
create mode 100755 tools/perf/tests/shell/coresight_asm_pure_loop.sh
create mode 100755 tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
create mode 100755 tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
create mode 100755 tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
create mode 100755 tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
create mode 100644 tools/perf/tests/shell/lib/coresight.sh
diff --git a/MAINTAINERS b/MAINTAINERS
index 673c7124ca82..18cc20609f2e 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -1918,10 +1918,14 @@ F: drivers/hwtracing/coresight/*
F: include/dt-bindings/arm/coresight-cti-dt.h
F: include/linux/coresight*
F: samples/coresight/*
+F: tools/perf/Documentation/arm-coresight.txt
F: tools/perf/arch/arm/util/auxtrace.c
F: tools/perf/arch/arm/util/cs-etm.c
F: tools/perf/arch/arm/util/cs-etm.h
F: tools/perf/arch/arm/util/pmu.c
+F: tools/perf/tests/shell/coresight_*
+F: tools/perf/tests/shell/tools/Makefile
+F: tools/perf/tests/shell/tools/coresight/*
F: tools/perf/util/cs-etm-decoder/*
F: tools/perf/util/cs-etm.*
diff --git a/tools/perf/.gitignore b/tools/perf/.gitignore
index 20b8ab984d5f..138c679ecacd 100644
--- a/tools/perf/.gitignore
+++ b/tools/perf/.gitignore
@@ -15,8 +15,9 @@ perf*.1
perf*.xml
perf*.html
common-cmds.h
-perf.data
-perf.data.old
+perf*.data
+perf*.data.old
+stats-*.csv
output.svg
perf-archive
perf-with-kcore
@@ -30,6 +31,7 @@ config.mak.autogen
*-flex.*
*.pyc
*.pyo
+*.stdout
.config-detected
util/intel-pt-decoder/inat-tables.c
arch/*/include/generated/
diff --git a/tools/perf/Documentation/arm-coresight.txt b/tools/perf/Documentation/arm-coresight.txt
new file mode 100644
index 000000000000..3a9e6c573c58
--- /dev/null
+++ b/tools/perf/Documentation/arm-coresight.txt
@@ -0,0 +1,140 @@
+Arm Coresight Support
+=====================
+
+Coresight is a feature of some Arm based processors that allows for
+debugging. One of the things it can do is trace every instruction
+executed and remotely expose that information in a hardware compressed
+stream. Perf is able to locally access that stream and store it to the
+output perf data files. This stream can then be later decoded to give the
+instructions that were traced for debugging or profiling purposes. You
+can log such data with a perf record command like:
+
+ perf record -e cs_etm//u testbinary
+
+This would run some test binary (testbinary) until it exits and record
+a perf.data trace file. That file would have AUX sections if coresight
+is working correctly. You can dump the content of this file as
+readable text with a command like:
+
+ perf report --stdio --dump -i perf.data
+
+You should find some sections of this file have AUX data blocks like:
+
+ 0x1e78 [0x30]: PERF_RECORD_AUXTRACE size: 0x11dd0 offset: 0 ref: 0x1b614fc1061b0ad1 idx: 0 tid: 531230 cpu: -1
+
+ . ... CoreSight ETM Trace data: size 73168 bytes
+ Idx:0; ID:10; I_ASYNC : Alignment Synchronisation.
+ Idx:12; ID:10; I_TRACE_INFO : Trace Info.; INFO=0x0 { CC.0 }
+ Idx:17; ID:10; I_ADDR_L_64IS0 : Address, Long, 64 bit, IS0.; Addr=0x0000000000000000;
+ Idx:26; ID:10; I_TRACE_ON : Trace On.
+ Idx:27; ID:10; I_ADDR_CTXT_L_64IS0 : Address & Context, Long, 64 bit, IS0.; Addr=0x0000FFFFB6069140; Ctxt: AArch64,EL0, NS;
+ Idx:38; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
+ Idx:39; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
+ Idx:40; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
+ Idx:41; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEN
+ ...
+
+If you see these above, then your system is tracing coresight data
+correctly.
+
+To compile perf with coresight support in the perf directory do
+
+ make CORESIGHT=1
+
+This will compile the perf tool with coresight support as well as
+build some small test binaries for perf test. This requires you also
+be compiling for 64bit Arm (ARM64/aarch64). The tools run as part of
+perf coresight tracing are in tests/shell/tools/coresight.
+
+You will also want coresight support enabled in your kernel config.
+Ensure it is enabled with:
+
+ CONFIG_CORESIGHT=y
+
+There are various other coresight options you probably also want
+enabled like:
+
+ CONFIG_CORESIGHT_LINKS_AND_SINKS=y
+ CONFIG_CORESIGHT_LINK_AND_SINK_TMC=y
+ CONFIG_CORESIGHT_CATU=y
+ CONFIG_CORESIGHT_SINK_TPIU=y
+ CONFIG_CORESIGHT_SINK_ETBV10=y
+ CONFIG_CORESIGHT_SOURCE_ETM4X=y
+ CONFIG_CORESIGHT_STM=y
+ CONFIG_CORESIGHT_CPU_DEBUG=y
+ CONFIG_CORESIGHT_CTI=y
+ CONFIG_CORESIGHT_CTI_INTEGRATION_REGS=y
+
+Please refer to the kernel configuration help for more information.
+
+Perf test - Verify kernel and userspace perf coresight work
+===========================================================
+
+When you run perf test, it will do a lot of self tests. Some of those
+tests will cover Coresight (only if enabled and on ARM64). You
+generally would run perf test from the tools/perf directory in the
+kernel tree. Some tests will check some internal perf support like:
+
+ Check Arm CoreSight trace data recording and synthesized samples
+
+Some others will actually use perf record and some test binaries that
+are in tests/shell/tools/coresight and will collect traces to ensure a
+minimum level of functionality is met. The scripts that launch these
+tests are in tests/shell. These will all look like:
+
+ Coresight / Memcpy 1M 25 Threads
+ Coresight / Unroll Loop Thread 2
+ ...
+
+These perf record tests will not run if the tool binaries do not exist
+in tests/shell/tools/coresight/*/ and will be skipped. If you do not
+have coresight support in hardware then either do not build perf with
+coresight support or remove these binaries in order to not have these
+tests fail and have them skip instead.
+
+These tests will log historical results in the current working
+directory (e.g. tools/perf) and will be named stats-*.csv like:
+
+ stats-asm_pure_loop-out.csv
+ stats-bubble_sort-random.csv
+ ...
+
+These statistic files log some aspects of the AUX data sections in
+the perf data output counting some numbers of certain encodings (a
+good way to know that it's working in a very simple way). One problem
+with coresight is that given a large enough amount of data needing to
+be logged, some of it can be lost due to the processor not waking up
+in time to read out all the data from buffers etc.. You will notice
+that the amount of data collected can vary a lot per run of perf test.
+If you wish to see how this changes over time, simply run perf test
+multiple times and all these csv files will have more and more data
+appended to it that you can later examine, graph and otherwise use to
+figure out if things have become worse or better.
+
+Be aware that amny of these tests take quite a while to run, specifically
+in processing the perf data file and dumping contents to then examine what
+is inside.
+
+You can change where these csv logs are stored by setting the
+PERF_TEST_CORESIGHT_STATDIR environment variable before running perf
+test like:
+
+ export PERF_TEST_CORESIGHT_STATDIR=/var/tmp
+ perf test
+
+They will also store resulting perf output data in the current
+directory for later inspection like:
+
+ perf-memcpy-1m.data
+ perf-thread_loop-2th.data
+ ...
+
+You can alter where the perf data files are stored by setting the
+PERF_TEST_CORESIGHT_DATADIR environment variable such as:
+
+ PERF_TEST_CORESIGHT_DATADIR=/var/tmp
+ perf test
+
+You may wish to set these above environment variables if you which to
+keep the output of tests outside of the current working directory for
+longer term storage and examination.
diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
index ac861e42c8f7..b97db83992e0 100644
--- a/tools/perf/Makefile.perf
+++ b/tools/perf/Makefile.perf
@@ -630,7 +630,15 @@ sync_file_range_tbls := $(srctree)/tools/perf/trace/beauty/sync_file_range.sh
$(sync_file_range_arrays): $(linux_uapi_dir)/fs.h $(sync_file_range_tbls)
$(Q)$(SHELL) '$(sync_file_range_tbls)' $(linux_uapi_dir) > $@
-all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS)
+TESTS_CORESIGHT_DIR := $(srctree)/tools/perf/tests/shell/coresight
+
+tests-coresight-targets: FORCE
+ $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR)
+
+tests-coresight-targets-clean:
+ $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR) clean
+
+all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS) tests-coresight-targets
# Create python binding output directory if not already present
_dummy := $(shell [ -d '$(OUTPUT)python' ] || mkdir -p '$(OUTPUT)python')
@@ -1020,6 +1028,7 @@ install-tests: all install-gtk
$(INSTALL) tests/shell/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell'; \
$(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'; \
$(INSTALL) tests/shell/lib/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'
+ $(Q)$(MAKE) -C tests/shell/coresight install-tests
install-bin: install-tools install-tests install-traceevent-plugins
@@ -1088,7 +1097,7 @@ endif # BUILD_BPF_SKEL
bpf-skel-clean:
$(call QUIET_CLEAN, bpf-skel) $(RM) -r $(SKEL_TMP_OUT) $(SKELETONS)
-clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean
+clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean tests-coresight-targets-clean
$(call QUIET_CLEAN, core-objs) $(RM) $(LIBPERF_A) $(OUTPUT)perf-archive $(OUTPUT)perf-with-kcore $(OUTPUT)perf-iostat $(LANG_BINDINGS)
$(Q)find $(if $(OUTPUT),$(OUTPUT),.) -name '*.o' -delete -o -name '\.*.cmd' -delete -o -name '\.*.d' -delete
$(Q)$(RM) $(OUTPUT).config-detected
@@ -1155,5 +1164,6 @@ FORCE:
.PHONY: shell_compatibility_test please_set_SHELL_PATH_to_a_more_modern_shell
.PHONY: $(GIT-HEAD-PHONY) TAGS tags cscope FORCE prepare
.PHONY: libtraceevent_plugins archheaders
+.PHONY: $(TESTS_CORESIGHT_TARGETS)
endif # force_fixdep
diff --git a/tools/perf/tests/shell/coresight/Makefile b/tools/perf/tests/shell/coresight/Makefile
new file mode 100644
index 000000000000..dda99aeac158
--- /dev/null
+++ b/tools/perf/tests/shell/coresight/Makefile
@@ -0,0 +1,30 @@
+# SPDX-License-Identifier: GPL-2.0-only
+# Carsten Haitzler <[email protected]>, 2021
+include ../../../../../tools/scripts/Makefile.include
+include ../../../../../tools/scripts/Makefile.arch
+include ../../../../../tools/scripts/utilities.mak
+
+SUBDIRS = \
+ asm_pure_loop \
+ thread_loop \
+ memcpy_thread \
+ unroll_loop_thread
+
+all: $(SUBDIRS)
+$(SUBDIRS):
+ $(Q)$(MAKE) -C $@
+
+INSTALLDIRS = $(SUBDIRS:%=install-%)
+
+install-tests: $(INSTALLDIRS)
+$(INSTALLDIRS):
+ $(Q)$(MAKE) -C $(@:install-%=%) install-tests
+
+CLEANDIRS = $(SUBDIRS:%=clean-%)
+
+clean: $(CLEANDIRS)
+$(CLEANDIRS):
+ $(Q)$(MAKE) -C $(@:clean-%=%) clean >/dev/null
+
+.PHONY: all clean $(SUBDIRS) $(CLEANDIRS) $(INSTALLDIRS)
+
diff --git a/tools/perf/tests/shell/coresight/Makefile.miniconfig b/tools/perf/tests/shell/coresight/Makefile.miniconfig
new file mode 100644
index 000000000000..893c12685fed
--- /dev/null
+++ b/tools/perf/tests/shell/coresight/Makefile.miniconfig
@@ -0,0 +1,23 @@
+# SPDX-License-Identifier: GPL-2.0-only
+# Carsten Haitzler <[email protected]>, 2021
+
+ifndef DESTDIR
+prefix ?= $(HOME)
+endif
+
+DESTDIR_SQ = $(subst ','\'',$(DESTDIR))
+perfexecdir = libexec/perf-core
+perfexec_instdir = $(perfexecdir)
+
+ifneq ($(filter /%,$(firstword $(perfexecdir))),)
+perfexec_instdir = $(perfexecdir)
+else
+perfexec_instdir = $(prefix)/$(perfexecdir)
+endif
+
+perfexec_instdir_SQ = $(subst ','\'',$(perfexec_instdir))
+INSTALL = install
+
+include ../../../../../scripts/Makefile.include
+include ../../../../../scripts/Makefile.arch
+include ../../../../../scripts/utilities.mak
diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
new file mode 100644
index 000000000000..468673ac32e8
--- /dev/null
+++ b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
@@ -0,0 +1 @@
+asm_pure_loop
diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
new file mode 100644
index 000000000000..10c5a60cb71c
--- /dev/null
+++ b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
@@ -0,0 +1,30 @@
+# SPDX-License-Identifier: GPL-2.0
+# Carsten Haitzler <[email protected]>, 2021
+
+include ../Makefile.miniconfig
+
+BIN=asm_pure_loop
+LIB=
+
+all: $(BIN)
+
+$(BIN): $(BIN).S
+ifdef CORESIGHT
+ifeq ($(ARCH),arm64)
+ $(Q)$(CC) $(BIN).S -nostdlib -static -o $(BIN) $(LIB)
+endif
+endif
+
+install-tests: all
+ifdef CORESIGHT
+ifeq ($(ARCH),arm64)
+ $(call QUIET_INSTALL, tests) \
+ $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
+ $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
+endif
+endif
+
+clean:
+ $(Q)$(RM) -f $(BIN)
+
+.PHONY: all clean install-tests
diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
new file mode 100644
index 000000000000..75cf084a927d
--- /dev/null
+++ b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
@@ -0,0 +1,28 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Tamas Zsoldos <[email protected]>, 2021 */
+
+.globl _start
+_start:
+ mov x0, 0x0000ffff
+ mov x1, xzr
+loop:
+ nop
+ nop
+ cbnz x1, noskip
+ nop
+ nop
+ adrp x2, skip
+ add x2, x2, :lo12:skip
+ br x2
+ nop
+ nop
+noskip:
+ nop
+ nop
+skip:
+ sub x0, x0, 1
+ cbnz x0, loop
+
+ mov x0, #0
+ mov x8, #93 // __NR_exit syscall
+ svc #0
diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
new file mode 100644
index 000000000000..f8217e56091e
--- /dev/null
+++ b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
@@ -0,0 +1 @@
+memcpy_thread
diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/Makefile b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile
new file mode 100644
index 000000000000..e2604cfae74b
--- /dev/null
+++ b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile
@@ -0,0 +1,29 @@
+# SPDX-License-Identifier: GPL-2.0
+# Carsten Haitzler <[email protected]>, 2021
+include ../Makefile.miniconfig
+
+BIN=memcpy_thread
+LIB=-pthread
+
+all: $(BIN)
+
+$(BIN): $(BIN).c
+ifdef CORESIGHT
+ifeq ($(ARCH),arm64)
+ $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
+endif
+endif
+
+install-tests: all
+ifdef CORESIGHT
+ifeq ($(ARCH),arm64)
+ $(call QUIET_INSTALL, tests) \
+ $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
+ $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
+endif
+endif
+
+clean:
+ $(Q)$(RM) -f $(BIN)
+
+.PHONY: all clean install-tests
diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
new file mode 100644
index 000000000000..a7e169d1bf64
--- /dev/null
+++ b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
@@ -0,0 +1,79 @@
+// SPDX-License-Identifier: GPL-2.0
+// Carsten Haitzler <[email protected]>, 2021
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <string.h>
+#include <pthread.h>
+
+struct args {
+ unsigned long loops;
+ unsigned long size;
+ pthread_t th;
+ void *ret;
+};
+
+static void *thrfn(void *arg)
+{
+ struct args *a = arg;
+ unsigned long i, len = a->loops;
+ unsigned char *src, *dst;
+
+ src = malloc(a->size * 1024);
+ dst = malloc(a->size * 1024);
+ if ((!src) || (!dst)) {
+ printf("ERR: Can't allocate memory\n");
+ exit(1);
+ }
+ for (i = 0; i < len; i++)
+ memcpy(dst, src, a->size * 1024);
+}
+
+static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
+{
+ pthread_t t;
+ pthread_attr_t attr;
+
+ pthread_attr_init(&attr);
+ pthread_create(&t, &attr, fn, arg);
+ return t;
+}
+
+int main(int argc, char **argv)
+{
+ unsigned long i, len, size, thr;
+ pthread_t threads[256];
+ struct args args[256];
+ long long v;
+
+ if (argc < 4) {
+ printf("ERR: %s [copysize Kb] [numthreads] [numloops (hundreds)]\n", argv[0]);
+ exit(1);
+ }
+
+ v = atoll(argv[1]);
+ if ((v < 1) || (v > (1024 * 1024))) {
+ printf("ERR: max memory 1GB (1048576 KB)\n");
+ exit(1);
+ }
+ size = v;
+ thr = atol(argv[2]);
+ if ((thr < 1) || (thr > 256)) {
+ printf("ERR: threads 1-256\n");
+ exit(1);
+ }
+ v = atoll(argv[3]);
+ if ((v < 1) || (v > 40000000000ll)) {
+ printf("ERR: loops 1-40000000000 (hundreds)\n");
+ exit(1);
+ }
+ len = v * 100;
+ for (i = 0; i < thr; i++) {
+ args[i].loops = len;
+ args[i].size = size;
+ args[i].th = new_thr(thrfn, &(args[i]));
+ }
+ for (i = 0; i < thr; i++)
+ pthread_join(args[i].th, &(args[i].ret));
+ return 0;
+}
diff --git a/tools/perf/tests/shell/coresight/thread_loop/.gitignore b/tools/perf/tests/shell/coresight/thread_loop/.gitignore
new file mode 100644
index 000000000000..6d4c33eaa9e8
--- /dev/null
+++ b/tools/perf/tests/shell/coresight/thread_loop/.gitignore
@@ -0,0 +1 @@
+thread_loop
diff --git a/tools/perf/tests/shell/coresight/thread_loop/Makefile b/tools/perf/tests/shell/coresight/thread_loop/Makefile
new file mode 100644
index 000000000000..424df4e8b0e6
--- /dev/null
+++ b/tools/perf/tests/shell/coresight/thread_loop/Makefile
@@ -0,0 +1,29 @@
+# SPDX-License-Identifier: GPL-2.0
+# Carsten Haitzler <[email protected]>, 2021
+include ../Makefile.miniconfig
+
+BIN=thread_loop
+LIB=-pthread
+
+all: $(BIN)
+
+$(BIN): $(BIN).c
+ifdef CORESIGHT
+ifeq ($(ARCH),arm64)
+ $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
+endif
+endif
+
+install-tests: all
+ifdef CORESIGHT
+ifeq ($(ARCH),arm64)
+ $(call QUIET_INSTALL, tests) \
+ $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
+ $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
+endif
+endif
+
+clean:
+ $(Q)$(RM) -f $(BIN)
+
+.PHONY: all clean install-tests
diff --git a/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
new file mode 100644
index 000000000000..c0158fac7d0b
--- /dev/null
+++ b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
@@ -0,0 +1,86 @@
+// SPDX-License-Identifier: GPL-2.0
+// Carsten Haitzler <[email protected]>, 2021
+
+// define this for gettid()
+#define _GNU_SOURCE
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <string.h>
+#include <pthread.h>
+#include <sys/syscall.h>
+#ifndef SYS_gettid
+// gettid is 178 on arm64
+# define SYS_gettid 178
+#endif
+#define gettid() syscall(SYS_gettid)
+
+struct args {
+ unsigned int loops;
+ pthread_t th;
+ void *ret;
+};
+
+static void *thrfn(void *arg)
+{
+ struct args *a = arg;
+ int i = 0, len = a->loops;
+
+ if (getenv("SHOW_TID")) {
+ unsigned long long tid = gettid();
+
+ printf("%llu\n", tid);
+ }
+ asm volatile(
+ "loop:\n"
+ "add %[i], %[i], #1\n"
+ "cmp %[i], %[len]\n"
+ "blt loop\n"
+ : /* out */
+ : /* in */ [i] "r" (i), [len] "r" (len)
+ : /* clobber */
+ );
+ return (void *)(long)i;
+}
+
+static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
+{
+ pthread_t t;
+ pthread_attr_t attr;
+
+ pthread_attr_init(&attr);
+ pthread_create(&t, &attr, fn, arg);
+ return t;
+}
+
+int main(int argc, char **argv)
+{
+ unsigned int i, len, thr;
+ pthread_t threads[256];
+ struct args args[256];
+
+ if (argc < 3) {
+ printf("ERR: %s [numthreads] [numloops (millions)]\n", argv[0]);
+ exit(1);
+ }
+
+ thr = atoi(argv[1]);
+ if ((thr < 1) || (thr > 256)) {
+ printf("ERR: threads 1-256\n");
+ exit(1);
+ }
+ len = atoi(argv[2]);
+ if ((len < 1) || (len > 4000)) {
+ printf("ERR: max loops 4000 (millions)\n");
+ exit(1);
+ }
+ len *= 1000000;
+ for (i = 0; i < thr; i++) {
+ args[i].loops = len;
+ args[i].th = new_thr(thrfn, &(args[i]));
+ }
+ for (i = 0; i < thr; i++)
+ pthread_join(args[i].th, &(args[i].ret));
+ return 0;
+}
diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
new file mode 100644
index 000000000000..2cb4e996dbf3
--- /dev/null
+++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
@@ -0,0 +1 @@
+unroll_loop_thread
diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
new file mode 100644
index 000000000000..45ab2be8be92
--- /dev/null
+++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
@@ -0,0 +1,29 @@
+# SPDX-License-Identifier: GPL-2.0
+# Carsten Haitzler <[email protected]>, 2021
+include ../Makefile.miniconfig
+
+BIN=unroll_loop_thread
+LIB=-pthread
+
+all: $(BIN)
+
+$(BIN): $(BIN).c
+ifdef CORESIGHT
+ifeq ($(ARCH),arm64)
+ $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
+endif
+endif
+
+install-tests: all
+ifdef CORESIGHT
+ifeq ($(ARCH),arm64)
+ $(call QUIET_INSTALL, tests) \
+ $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
+ $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
+endif
+endif
+
+clean:
+ $(Q)$(RM) -f $(BIN)
+
+.PHONY: all clean install-tests
diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
new file mode 100644
index 000000000000..cb9d22c7dfb9
--- /dev/null
+++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
@@ -0,0 +1,74 @@
+// SPDX-License-Identifier: GPL-2.0
+// Carsten Haitzler <[email protected]>, 2021
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <string.h>
+#include <pthread.h>
+
+struct args {
+ pthread_t th;
+ unsigned int in, out;
+ void *ret;
+};
+
+static void *thrfn(void *arg)
+{
+ struct args *a = arg;
+ unsigned int i, in = a->in;
+
+ for (i = 0; i < 10000; i++) {
+ asm volatile (
+// force an unroll of thia add instruction so we can test long runs of code
+#define SNIP1 "add %[in], %[in], #1\n"
+// 10
+#define SNIP2 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1
+// 100
+#define SNIP3 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2
+// 1000
+#define SNIP4 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3
+// 10000
+#define SNIP5 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4
+// 100000
+ SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5
+ : /* out */
+ : /* in */ [in] "r" (in)
+ : /* clobber */
+ );
+ }
+}
+
+static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
+{
+ pthread_t t;
+ pthread_attr_t attr;
+
+ pthread_attr_init(&attr);
+ pthread_create(&t, &attr, fn, arg);
+ return t;
+}
+
+int main(int argc, char **argv)
+{
+ unsigned int i, thr;
+ pthread_t threads[256];
+ struct args args[256];
+
+ if (argc < 2) {
+ printf("ERR: %s [numthreads]\n", argv[0]);
+ exit(1);
+ }
+
+ thr = atoi(argv[1]);
+ if ((thr > 256) || (thr < 1)) {
+ printf("ERR: threads 1-256\n");
+ exit(1);
+ }
+ for (i = 0; i < thr; i++) {
+ args[i].in = rand();
+ args[i].th = new_thr(thrfn, &(args[i]));
+ }
+ for (i = 0; i < thr; i++)
+ pthread_join(args[i].th, &(args[i].ret));
+ return 0;
+}
diff --git a/tools/perf/tests/shell/coresight_asm_pure_loop.sh b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
new file mode 100755
index 000000000000..3f0dbefcad50
--- /dev/null
+++ b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
@@ -0,0 +1,18 @@
+#!/bin/sh -e
+# Coresight / ASM Pure Loop
+
+# SPDX-License-Identifier: GPL-2.0
+# Carsten Haitzler <[email protected]>, 2021
+
+TEST="asm_pure_loop"
+. $(dirname $0)/lib/coresight.sh
+ARGS=""
+DATV="out"
+DATA="$DATD/perf-$TEST-$DATV.data"
+
+perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
+
+perf_dump_aux_verify "$DATA" 10 10 10
+
+err=$?
+exit $err
diff --git a/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh b/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
new file mode 100755
index 000000000000..8972af835016
--- /dev/null
+++ b/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
@@ -0,0 +1,18 @@
+#!/bin/sh -e
+# Coresight / Memcpy 16k 10 Threads
+
+# SPDX-License-Identifier: GPL-2.0
+# Carsten Haitzler <[email protected]>, 2021
+
+TEST="memcpy_thread"
+. $(dirname $0)/lib/coresight.sh
+ARGS="16 10 1"
+DATV="16k_10"
+DATA="$DATD/perf-$TEST-$DATV.data"
+
+perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
+
+perf_dump_aux_verify "$DATA" 10 10 10
+
+err=$?
+exit $err
diff --git a/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh b/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
new file mode 100755
index 000000000000..5b468901f89b
--- /dev/null
+++ b/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
@@ -0,0 +1,19 @@
+#!/bin/sh -e
+# Coresight / Thread Loop 10 Threads - Check TID
+
+# SPDX-License-Identifier: GPL-2.0
+# Carsten Haitzler <[email protected]>, 2021
+
+TEST="thread_loop"
+. $(dirname $0)/lib/coresight.sh
+ARGS="10 1"
+DATV="check-tid-10th"
+DATA="$DATD/perf-$TEST-$DATV.data"
+STDO="$DATD/perf-$TEST-$DATV.stdout"
+
+SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO
+
+perf_dump_aux_tid_verify "$DATA" "$STDO"
+
+err=$?
+exit $err
diff --git a/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh b/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
new file mode 100755
index 000000000000..f8b7abd3aa03
--- /dev/null
+++ b/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
@@ -0,0 +1,19 @@
+#!/bin/sh -e
+# Coresight / Thread Loop 2 Threads - Check TID
+
+# SPDX-License-Identifier: GPL-2.0
+# Carsten Haitzler <[email protected]>, 2021
+
+TEST="thread_loop"
+. $(dirname $0)/lib/coresight.sh
+ARGS="2 20"
+DATV="check-tid-2th"
+DATA="$DATD/perf-$TEST-$DATV.data"
+STDO="$DATD/perf-$TEST-$DATV.stdout"
+
+SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO
+
+perf_dump_aux_tid_verify "$DATA" "$STDO"
+
+err=$?
+exit $err
diff --git a/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh b/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
new file mode 100755
index 000000000000..c985dfb025c2
--- /dev/null
+++ b/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
@@ -0,0 +1,18 @@
+#!/bin/sh -e
+# Coresight / Unroll Loop Thread 10
+
+# SPDX-License-Identifier: GPL-2.0
+# Carsten Haitzler <[email protected]>, 2021
+
+TEST="unroll_loop_thread"
+. $(dirname $0)/lib/coresight.sh
+ARGS="10"
+DATV="10"
+DATA="$DATD/perf-$TEST-$DATV.data"
+
+perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
+
+perf_dump_aux_verify "$DATA" 10 10 10
+
+err=$?
+exit $err
diff --git a/tools/perf/tests/shell/lib/coresight.sh b/tools/perf/tests/shell/lib/coresight.sh
new file mode 100644
index 000000000000..6a611b073f02
--- /dev/null
+++ b/tools/perf/tests/shell/lib/coresight.sh
@@ -0,0 +1,130 @@
+# SPDX-License-Identifier: GPL-2.0
+# Carsten Haitzler <[email protected]>, 2021
+
+# This is sourced from a driver script so no need for #!/bin... etc. at the
+# top - the assumption below is that it runs as part of sourcing after the
+# test sets up some basic env vars to say what it is.
+
+# perf record options for the perf tests to use
+PERFRECMEM="-m ,128M"
+PERFRECOPT="$PERFRECMEM -e cs_etm//u"
+
+# These tests need to be run as root or coresight won't allow large buffers
+# and will not collect proper data
+UID=`id -u`
+if test "$UID" -ne 0; then
+ echo "Not running as root... skip"
+ exit 2
+fi
+
+TOOLS=$(dirname $0)
+DIR="$TOOLS/coresight/$TEST"
+BIN="$DIR/$TEST"
+# If the test tool/binary does not exist and is executable then skip the test
+if ! test -x "$BIN"; then exit 2; fi
+DATD="."
+# If the data dir env is set then make the data dir use that instead of ./
+if test -n "$PERF_TEST_CORESIGHT_DATADIR"; then
+ DATD="$PERF_TEST_CORESIGHT_DATADIR";
+fi
+# If the stat dir env is set then make the data dir use that instead of ./
+STATD="."
+if test -n "$PERF_TEST_CORESIGHT_STATDIR"; then
+ STATD="$PERF_TEST_CORESIGHT_STATDIR";
+fi
+
+# Called if the test fails - error code 2
+err() {
+ echo "$1"
+ exit 1
+}
+
+# Check that some statistics from our perf
+check_val_min() {
+ STATF="$4"
+ if test "$2" -lt "$3"; then
+ echo ", FAILED" >> "$STATF"
+ err "Sanity check number of $1 is too low ($2 < $3)"
+ fi
+}
+
+perf_dump_aux_verify() {
+ # Some basic checking that the AUX chunk contains some sensible data
+ # to see that we are recording something and at least a minimum
+ # amount of it. We should almost always see F3 atoms in just about
+ # anything but certainly we will see some trace info and async atom
+ # chunks.
+ DUMP="$DATD/perf-tmp-aux-dump.txt"
+ perf report --stdio --dump -i "$1" | \
+ grep -o -e I_ATOM_F3 -e I_ASYNC -e I_TRACE_INFO > "$DUMP"
+ # Simply count how many of these atoms we find to see that we are
+ # producing a reasonable amount of data - exact checks are not sane
+ # as this is a lossy process where we may lose some blocks and the
+ # compiler may produce different code depending on the compiler and
+ # optimization options, so this is rough just to see if we're
+ # either missing almost all the data or all of it
+ ATOM_F3_NUM=`grep I_ATOM_F3 "$DUMP" | wc -l`
+ ATOM_ASYNC_NUM=`grep I_ASYNC "$DUMP" | wc -l`
+ ATOM_TRACE_INFO_NUM=`grep I_TRACE_INFO "$DUMP" | wc -l`
+ rm -f "$DUMP"
+
+ # Arguments provide minimums for a pass
+ CHECK_F3_MIN="$2"
+ CHECK_ASYNC_MIN="$3"
+ CHECK_TRACE_INFO_MIN="$4"
+
+ # Write out statistics, so over time you can track results to see if
+ # there is a pattern - for example we have less "noisy" results that
+ # produce more consistent amounts of data each run, to see if over
+ # time any techinques to minimize data loss are having an effect or
+ # not
+ STATF="$STATD/stats-$TEST-$DATV.csv"
+ if ! test -f "$STATF"; then
+ echo "ATOM F3 Count, Minimum, ATOM ASYNC Count, Minimum, TRACE INFO Count, Minimum" > "$STATF"
+ fi
+ echo -n "$ATOM_F3_NUM, $CHECK_F3_MIN, $ATOM_ASYNC_NUM, $CHECK_ASYNC_MIN, $ATOM_TRACE_INFO_NUM, $CHECK_TRACE_INFO_MIN" >> "$STATF"
+
+ # Actually check to see if we passed or failed.
+ check_val_min "ATOM_F3" "$ATOM_F3_NUM" "$CHECK_F3_MIN" "$STATF"
+ check_val_min "ASYNC" "$ATOM_ASYNC_NUM" "$CHECK_ASYNC_MIN" "$STATF"
+ check_val_min "TRACE_INFO" "$ATOM_TRACE_INFO_NUM" "$CHECK_TRACE_INFO_MIN" "$STATF"
+ echo ", Ok" >> "$STATF"
+}
+
+perf_dump_aux_tid_verify() {
+ # Specifically crafted test will produce a list of Tread ID's to
+ # stdout that need to be checked to see that they have had trace
+ # info collected in AUX blocks in the perf data. This will go
+ # through all the TID's that are listed as CID=0xabcdef and see
+ # that all the Thread IDs the test tool reports are in the perf
+ # data AUX chunks
+
+ # The TID test tools will print a TID per stdout line that are being
+ # tested
+ TIDS=`cat "$2"`
+ # Scan the perf report to find the TIDs that are actually CID in hex
+ # and build a list of the ones found
+ FOUND_TIDS=`perf report --stdio --dump -i "$1" | \
+ grep -o "CID=0x[0-9a-z]\+" | sed 's/CID=//g' | \
+ uniq | sort | uniq`
+
+ # Iterate over the list of TIDs that the test says it has and find
+ # them in the TIDs found in the perf report
+ MISSING=""
+ for TID2 in $TIDS; do
+ FOUND=""
+ for TIDHEX in $FOUND_TIDS; do
+ TID=`printf "%i" $TIDHEX`
+ if test "$TID" -eq "$TID2"; then
+ FOUND="y"
+ break
+ fi
+ done
+ if test -z "$FOUND"; then
+ MISSING="$MISSING $TID"
+ fi
+ done
+ if test -n "$MISSING"; then
+ err "Thread IDs $MISSING not found in perf AUX data"
+ fi
+}
--
2.32.0
Hi Carsten,
On Wed, Mar 09, 2022 at 12:28:59PM +0000, [email protected] wrote:
> From: Carsten Haitzler <[email protected]>
>
> This adds a test harness and tests to run perf record and examine the
> resuling output when coresight is enabled on arm64 and check the
> resulting quality of the output as part of perf test. These tests use
> various tools to produce output from perf record then measure some key
> specific aspects of that data to see if the data exists at all and
> contains key aspects such as measuring some data for every thread of
> a test or produces sufficient data for large exeuction runs of a large
> executable. etc.
>
> Signed-off-by: Carsten Haitzler <[email protected]>
> ---
> MAINTAINERS | 4 +
> tools/perf/.gitignore | 6 +-
> tools/perf/Documentation/arm-coresight.txt | 140 ++++++++++++++++++
> tools/perf/Makefile.perf | 14 +-
> tools/perf/tests/shell/coresight/Makefile | 30 ++++
> .../tests/shell/coresight/Makefile.miniconfig | 23 +++
> .../shell/coresight/asm_pure_loop/.gitignore | 1 +
> .../shell/coresight/asm_pure_loop/Makefile | 30 ++++
> .../coresight/asm_pure_loop/asm_pure_loop.S | 28 ++++
> .../shell/coresight/memcpy_thread/.gitignore | 1 +
> .../shell/coresight/memcpy_thread/Makefile | 29 ++++
> .../coresight/memcpy_thread/memcpy_thread.c | 79 ++++++++++
> .../shell/coresight/thread_loop/.gitignore | 1 +
> .../shell/coresight/thread_loop/Makefile | 29 ++++
> .../shell/coresight/thread_loop/thread_loop.c | 86 +++++++++++
> .../coresight/unroll_loop_thread/.gitignore | 1 +
> .../coresight/unroll_loop_thread/Makefile | 29 ++++
> .../unroll_loop_thread/unroll_loop_thread.c | 74 +++++++++
> .../tests/shell/coresight_asm_pure_loop.sh | 18 +++
> .../shell/coresight_memcpy_thread_16k_10.sh | 18 +++
> .../coresight_thread_loop_check_tid_10.sh | 19 +++
> .../coresight_thread_loop_check_tid_2.sh | 19 +++
> .../shell/coresight_unroll_loop_thread_10.sh | 18 +++
> tools/perf/tests/shell/lib/coresight.sh | 130 ++++++++++++++++
Very big change... Why squash all patches form previous verion to this
single one big patch? Usually the format with small patches is much
better for reviewing.
And I cannot apply cleanly on perf core branch:
https://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git branch: perf/core.
> 24 files changed, 823 insertions(+), 4 deletions(-)
> create mode 100644 tools/perf/Documentation/arm-coresight.txt
> create mode 100644 tools/perf/tests/shell/coresight/Makefile
> create mode 100644 tools/perf/tests/shell/coresight/Makefile.miniconfig
> create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
> create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
> create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
> create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
> create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/Makefile
> create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
> create mode 100644 tools/perf/tests/shell/coresight/thread_loop/.gitignore
> create mode 100644 tools/perf/tests/shell/coresight/thread_loop/Makefile
> create mode 100644 tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
> create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
> create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
> create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
> create mode 100755 tools/perf/tests/shell/coresight_asm_pure_loop.sh
> create mode 100755 tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
> create mode 100755 tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
> create mode 100755 tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
> create mode 100755 tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
> create mode 100644 tools/perf/tests/shell/lib/coresight.sh
>
> diff --git a/MAINTAINERS b/MAINTAINERS
> index 673c7124ca82..18cc20609f2e 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -1918,10 +1918,14 @@ F: drivers/hwtracing/coresight/*
> F: include/dt-bindings/arm/coresight-cti-dt.h
> F: include/linux/coresight*
> F: samples/coresight/*
> +F: tools/perf/Documentation/arm-coresight.txt
> F: tools/perf/arch/arm/util/auxtrace.c
> F: tools/perf/arch/arm/util/cs-etm.c
> F: tools/perf/arch/arm/util/cs-etm.h
> F: tools/perf/arch/arm/util/pmu.c
> +F: tools/perf/tests/shell/coresight_*
> +F: tools/perf/tests/shell/tools/Makefile
> +F: tools/perf/tests/shell/tools/coresight/*
> F: tools/perf/util/cs-etm-decoder/*
> F: tools/perf/util/cs-etm.*
>
> diff --git a/tools/perf/.gitignore b/tools/perf/.gitignore
> index 20b8ab984d5f..138c679ecacd 100644
> --- a/tools/perf/.gitignore
> +++ b/tools/perf/.gitignore
> @@ -15,8 +15,9 @@ perf*.1
> perf*.xml
> perf*.html
> common-cmds.h
> -perf.data
> -perf.data.old
> +perf*.data
> +perf*.data.old
> +stats-*.csv
> output.svg
> perf-archive
> perf-with-kcore
> @@ -30,6 +31,7 @@ config.mak.autogen
> *-flex.*
> *.pyc
> *.pyo
> +*.stdout
> .config-detected
> util/intel-pt-decoder/inat-tables.c
> arch/*/include/generated/
> diff --git a/tools/perf/Documentation/arm-coresight.txt b/tools/perf/Documentation/arm-coresight.txt
> new file mode 100644
> index 000000000000..3a9e6c573c58
> --- /dev/null
> +++ b/tools/perf/Documentation/arm-coresight.txt
> @@ -0,0 +1,140 @@
> +Arm Coresight Support
> +=====================
> +
> +Coresight is a feature of some Arm based processors that allows for
> +debugging. One of the things it can do is trace every instruction
> +executed and remotely expose that information in a hardware compressed
> +stream.
Maybe here need to sync a bit for the terminology in
Documentation/trace/coresight/coresight.rst.
Something like:
"Coresight is a feature of some Arm based processors that allows for
debugging. One of the things it can do is trace instruction path and
expose that information in a hardware compressed stream for either
debugger or HW assisted tracing locally.
See Documentation/trace/coresight/coresight.rst for details."
> Perf is able to locally access that stream and store it to the
> +output perf data files. This stream can then be later decoded to give the
> +instructions that were traced for debugging or profiling purposes. You
> +can log such data with a perf record command like:
> +
> + perf record -e cs_etm//u testbinary
> +
> +This would run some test binary (testbinary) until it exits and record
> +a perf.data trace file. That file would have AUX sections if coresight
> +is working correctly. You can dump the content of this file as
> +readable text with a command like:
> +
> + perf report --stdio --dump -i perf.data
> +
> +You should find some sections of this file have AUX data blocks like:
> +
> + 0x1e78 [0x30]: PERF_RECORD_AUXTRACE size: 0x11dd0 offset: 0 ref: 0x1b614fc1061b0ad1 idx: 0 tid: 531230 cpu: -1
> +
> + . ... CoreSight ETM Trace data: size 73168 bytes
> + Idx:0; ID:10; I_ASYNC : Alignment Synchronisation.
> + Idx:12; ID:10; I_TRACE_INFO : Trace Info.; INFO=0x0 { CC.0 }
> + Idx:17; ID:10; I_ADDR_L_64IS0 : Address, Long, 64 bit, IS0.; Addr=0x0000000000000000;
> + Idx:26; ID:10; I_TRACE_ON : Trace On.
> + Idx:27; ID:10; I_ADDR_CTXT_L_64IS0 : Address & Context, Long, 64 bit, IS0.; Addr=0x0000FFFFB6069140; Ctxt: AArch64,EL0, NS;
> + Idx:38; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
> + Idx:39; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
> + Idx:40; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
> + Idx:41; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEN
> + ...
> +
> +If you see these above, then your system is tracing coresight data
> +correctly.
> +
> +To compile perf with coresight support in the perf directory do
> +
> + make CORESIGHT=1
It is inaccurate that if we don't mention openCSD lib.
> +
> +This will compile the perf tool with coresight support as well as
> +build some small test binaries for perf test. This requires you also
> +be compiling for 64bit Arm (ARM64/aarch64). The tools run as part of
> +perf coresight tracing are in tests/shell/tools/coresight.
For build perf tool, I think above paragraphs are duplicate with the
document Documentation/trace/coresight/coresight.rst. Can we simply
say:
"The details for building perf tool with support Arm Coresight can be
found in the "HOWTO.md" file of the openCSD gitHub repository:
https://github.com/Linaro/opencsd.
And "HOWTO.md" file gives the information and examples for how to use
perf tool to record and report Coresight trace data. It's the
prerequisite for this perf Coresight test."
> +You will also want coresight support enabled in your kernel config.
> +Ensure it is enabled with:
> +
> + CONFIG_CORESIGHT=y
> +
> +There are various other coresight options you probably also want
> +enabled like:
> +
> + CONFIG_CORESIGHT_LINKS_AND_SINKS=y
> + CONFIG_CORESIGHT_LINK_AND_SINK_TMC=y
> + CONFIG_CORESIGHT_CATU=y
> + CONFIG_CORESIGHT_SINK_TPIU=y
> + CONFIG_CORESIGHT_SINK_ETBV10=y
> + CONFIG_CORESIGHT_SOURCE_ETM4X=y
> + CONFIG_CORESIGHT_STM=y
> + CONFIG_CORESIGHT_CPU_DEBUG=y
> + CONFIG_CORESIGHT_CTI=y
> + CONFIG_CORESIGHT_CTI_INTEGRATION_REGS=y
> +
> +Please refer to the kernel configuration help for more information.
I prefer to remove these kernel configuration since they are not
inconsistent on different platforms (e.g. ETBV10, ETM4X, etc), and
some configurations might not necessary (e.g. CPU_DEBUG).
> +Perf test - Verify kernel and userspace perf coresight work
> +===========================================================
> +
> +When you run perf test, it will do a lot of self tests. Some of those
> +tests will cover Coresight (only if enabled and on ARM64). You
> +generally would run perf test from the tools/perf directory in the
> +kernel tree. Some tests will check some internal perf support like:
> +
> + Check Arm CoreSight trace data recording and synthesized samples
> +
> +Some others will actually use perf record and some test binaries that
> +are in tests/shell/tools/coresight and will collect traces to ensure a
> +minimum level of functionality is met. The scripts that launch these
> +tests are in tests/shell. These will all look like:
> +
> + Coresight / Memcpy 1M 25 Threads
> + Coresight / Unroll Loop Thread 2
> + ...
Please update based on the latest test case names, at my side, I can
see the testing case like:
Coresight / ASM Pure Loop
Coresight / Memcpy 16k 10 Threads
Coresight / Thread Loop 10 Threads - Check TID
Coresight / Thread Loop 2 Threads - Check TID
Coresight / Unroll Loop Thread 10
> +
> +These perf record tests will not run if the tool binaries do not exist
> +in tests/shell/tools/coresight/*/ and will be skipped. If you do not
> +have coresight support in hardware then either do not build perf with
> +coresight support or remove these binaries in order to not have these
> +tests fail and have them skip instead.
> +
> +These tests will log historical results in the current working
> +directory (e.g. tools/perf) and will be named stats-*.csv like:
> +
> + stats-asm_pure_loop-out.csv
> + stats-bubble_sort-random.csv
> + ...
> +
> +These statistic files log some aspects of the AUX data sections in
> +the perf data output counting some numbers of certain encodings (a
> +good way to know that it's working in a very simple way). One problem
> +with coresight is that given a large enough amount of data needing to
> +be logged, some of it can be lost due to the processor not waking up
> +in time to read out all the data from buffers etc.. You will notice
> +that the amount of data collected can vary a lot per run of perf test.
> +If you wish to see how this changes over time, simply run perf test
> +multiple times and all these csv files will have more and more data
> +appended to it that you can later examine, graph and otherwise use to
> +figure out if things have become worse or better.
I am confused by this narrative. Does it try to remind that the final
testing result (pass or fail) is not stable? Or should we run for
multiple times so have more chance to capture issues?
> +Be aware that amny of these tests take quite a while to run, specifically
s/amny/many
> +in processing the perf data file and dumping contents to then examine what
> +is inside.
> +
> +You can change where these csv logs are stored by setting the
> +PERF_TEST_CORESIGHT_STATDIR environment variable before running perf
> +test like:
> +
> + export PERF_TEST_CORESIGHT_STATDIR=/var/tmp
> + perf test
> +
> +They will also store resulting perf output data in the current
> +directory for later inspection like:
> +
> + perf-memcpy-1m.data
> + perf-thread_loop-2th.data
> + ...
> +
> +You can alter where the perf data files are stored by setting the
> +PERF_TEST_CORESIGHT_DATADIR environment variable such as:
> +
> + PERF_TEST_CORESIGHT_DATADIR=/var/tmp
> + perf test
> +
> +You may wish to set these above environment variables if you which to
> +keep the output of tests outside of the current working directory for
> +longer term storage and examination.
> diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
> index ac861e42c8f7..b97db83992e0 100644
> --- a/tools/perf/Makefile.perf
> +++ b/tools/perf/Makefile.perf
> @@ -630,7 +630,15 @@ sync_file_range_tbls := $(srctree)/tools/perf/trace/beauty/sync_file_range.sh
> $(sync_file_range_arrays): $(linux_uapi_dir)/fs.h $(sync_file_range_tbls)
> $(Q)$(SHELL) '$(sync_file_range_tbls)' $(linux_uapi_dir) > $@
>
> -all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS)
> +TESTS_CORESIGHT_DIR := $(srctree)/tools/perf/tests/shell/coresight
> +
> +tests-coresight-targets: FORCE
> + $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR)
> +
> +tests-coresight-targets-clean:
> + $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR) clean
> +
> +all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS) tests-coresight-targets
>
> # Create python binding output directory if not already present
> _dummy := $(shell [ -d '$(OUTPUT)python' ] || mkdir -p '$(OUTPUT)python')
> @@ -1020,6 +1028,7 @@ install-tests: all install-gtk
> $(INSTALL) tests/shell/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell'; \
> $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'; \
> $(INSTALL) tests/shell/lib/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'
> + $(Q)$(MAKE) -C tests/shell/coresight install-tests
>
> install-bin: install-tools install-tests install-traceevent-plugins
>
> @@ -1088,7 +1097,7 @@ endif # BUILD_BPF_SKEL
> bpf-skel-clean:
> $(call QUIET_CLEAN, bpf-skel) $(RM) -r $(SKEL_TMP_OUT) $(SKELETONS)
>
> -clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean
> +clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean tests-coresight-targets-clean
> $(call QUIET_CLEAN, core-objs) $(RM) $(LIBPERF_A) $(OUTPUT)perf-archive $(OUTPUT)perf-with-kcore $(OUTPUT)perf-iostat $(LANG_BINDINGS)
> $(Q)find $(if $(OUTPUT),$(OUTPUT),.) -name '*.o' -delete -o -name '\.*.cmd' -delete -o -name '\.*.d' -delete
> $(Q)$(RM) $(OUTPUT).config-detected
> @@ -1155,5 +1164,6 @@ FORCE:
> .PHONY: shell_compatibility_test please_set_SHELL_PATH_to_a_more_modern_shell
> .PHONY: $(GIT-HEAD-PHONY) TAGS tags cscope FORCE prepare
> .PHONY: libtraceevent_plugins archheaders
> +.PHONY: $(TESTS_CORESIGHT_TARGETS)
I don't find other places using TESTS_CORESIGHT_TARGETS. Is this
redundant?
> endif # force_fixdep
> diff --git a/tools/perf/tests/shell/coresight/Makefile b/tools/perf/tests/shell/coresight/Makefile
> new file mode 100644
> index 000000000000..dda99aeac158
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/Makefile
> @@ -0,0 +1,30 @@
> +# SPDX-License-Identifier: GPL-2.0-only
> +# Carsten Haitzler <[email protected]>, 2021
> +include ../../../../../tools/scripts/Makefile.include
> +include ../../../../../tools/scripts/Makefile.arch
> +include ../../../../../tools/scripts/utilities.mak
> +
> +SUBDIRS = \
> + asm_pure_loop \
> + thread_loop \
> + memcpy_thread \
> + unroll_loop_thread
> +
> +all: $(SUBDIRS)
> +$(SUBDIRS):
> + $(Q)$(MAKE) -C $@
> +
> +INSTALLDIRS = $(SUBDIRS:%=install-%)
> +
> +install-tests: $(INSTALLDIRS)
> +$(INSTALLDIRS):
> + $(Q)$(MAKE) -C $(@:install-%=%) install-tests
> +
> +CLEANDIRS = $(SUBDIRS:%=clean-%)
> +
> +clean: $(CLEANDIRS)
> +$(CLEANDIRS):
> + $(Q)$(MAKE) -C $(@:clean-%=%) clean >/dev/null
> +
> +.PHONY: all clean $(SUBDIRS) $(CLEANDIRS) $(INSTALLDIRS)
> +
> diff --git a/tools/perf/tests/shell/coresight/Makefile.miniconfig b/tools/perf/tests/shell/coresight/Makefile.miniconfig
> new file mode 100644
> index 000000000000..893c12685fed
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/Makefile.miniconfig
> @@ -0,0 +1,23 @@
> +# SPDX-License-Identifier: GPL-2.0-only
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +ifndef DESTDIR
> +prefix ?= $(HOME)
> +endif
> +
> +DESTDIR_SQ = $(subst ','\'',$(DESTDIR))
> +perfexecdir = libexec/perf-core
> +perfexec_instdir = $(perfexecdir)
> +
> +ifneq ($(filter /%,$(firstword $(perfexecdir))),)
> +perfexec_instdir = $(perfexecdir)
> +else
> +perfexec_instdir = $(prefix)/$(perfexecdir)
> +endif
> +
> +perfexec_instdir_SQ = $(subst ','\'',$(perfexec_instdir))
> +INSTALL = install
> +
> +include ../../../../../scripts/Makefile.include
> +include ../../../../../scripts/Makefile.arch
> +include ../../../../../scripts/utilities.mak
> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
> new file mode 100644
> index 000000000000..468673ac32e8
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
> @@ -0,0 +1 @@
> +asm_pure_loop
Do we really need there '.gitignore' files under the folder
'tools/perf/tests/shell/coresight/'.
> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
> new file mode 100644
> index 000000000000..10c5a60cb71c
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
> @@ -0,0 +1,30 @@
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +include ../Makefile.miniconfig
> +
> +BIN=asm_pure_loop
> +LIB=
Remove the unused variable 'LIB='.
> +
> +all: $(BIN)
> +
> +$(BIN): $(BIN).S
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(Q)$(CC) $(BIN).S -nostdlib -static -o $(BIN) $(LIB)
> +endif
> +endif
> +
> +install-tests: all
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(call QUIET_INSTALL, tests) \
> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
> +endif
> +endif
> +
> +clean:
> + $(Q)$(RM) -f $(BIN)
> +
> +.PHONY: all clean install-tests
There have four sub folders under tools/perf/tests/shell/coresight:
asm_pure_loop
memcpy_thread
thread_loop
unroll_loop_thread
And every folder has its own Makefile and every Makefile is quite
close to each other. I am just wandering if it's possible to
remove the 4 Makefiles in these four sub folders, and simply use
tools/perf/tests/shell/coresight/Makefile as the central place to
build these assistant programs.
> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
> new file mode 100644
> index 000000000000..75cf084a927d
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
> @@ -0,0 +1,28 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/* Tamas Zsoldos <[email protected]>, 2021 */
> +
> +.globl _start
> +_start:
> + mov x0, 0x0000ffff
> + mov x1, xzr
> +loop:
> + nop
> + nop
> + cbnz x1, noskip
> + nop
> + nop
> + adrp x2, skip
> + add x2, x2, :lo12:skip
> + br x2
> + nop
> + nop
> +noskip:
> + nop
> + nop
> +skip:
> + sub x0, x0, 1
> + cbnz x0, loop
> +
> + mov x0, #0
> + mov x8, #93 // __NR_exit syscall
> + svc #0
I tested the case "ASM Pure Loop" on my Juno board, and it complaints:
root@debian:/mnt/export/arm-linux-kernel/tools/perf# ./perf test -v 76
76: Coresight / ASM Pure Loop :
--- start ---
test child forked, pid 9063
failed to mmap with 12 (Cannot allocate memory)
test child finished with -1
---- end ----
Coresight / ASM Pure Loop: FAILED!
Since I only setup the 1GB memory for the Linux kernel, it fails to
allocate AUX ring buffer with the size 256MB. So I manully change
the buffer size to 8MB in tools/perf/tests/shell/lib/coresight.sh:
PERFRECMEM="-m ,8M"
So finally I can see the test case is passed:
root@debian:/mnt/export/arm-linux-kernel/tools/perf# ./perf test -v 76
76: Coresight / ASM Pure Loop :
--- start ---
test child forked, pid 9481
-m ,8M -e cs_etm//u
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.681 MB ./perf-asm_pure_loop-out.data ]
test child finished with 0
---- end ----
Coresight / ASM Pure Loop: Ok
Do you think we really need to use 256MiB as the AUX buffer size?
IIRC, it means we allocate 256MiB per CPU for this case, on the other
hand, you could see the final perf data file size is small (0.681
MiB).
Seems to me, it's not necessary to allocate so big buffer for
the test, and I tried to run below 4 cases with 8MiB, all of them can
pass the testing :)
> diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
> new file mode 100644
> index 000000000000..f8217e56091e
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
> @@ -0,0 +1 @@
> +memcpy_thread
> diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/Makefile b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile
> new file mode 100644
> index 000000000000..e2604cfae74b
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile
> @@ -0,0 +1,29 @@
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +include ../Makefile.miniconfig
> +
> +BIN=memcpy_thread
> +LIB=-pthread
> +
> +all: $(BIN)
> +
> +$(BIN): $(BIN).c
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
> +endif
> +endif
> +
> +install-tests: all
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(call QUIET_INSTALL, tests) \
> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
> +endif
> +endif
> +
> +clean:
> + $(Q)$(RM) -f $(BIN)
> +
> +.PHONY: all clean install-tests
> diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
> new file mode 100644
> index 000000000000..a7e169d1bf64
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
> @@ -0,0 +1,79 @@
> +// SPDX-License-Identifier: GPL-2.0
> +// Carsten Haitzler <[email protected]>, 2021
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <unistd.h>
> +#include <string.h>
> +#include <pthread.h>
> +
> +struct args {
> + unsigned long loops;
> + unsigned long size;
> + pthread_t th;
> + void *ret;
> +};
> +
> +static void *thrfn(void *arg)
> +{
> + struct args *a = arg;
> + unsigned long i, len = a->loops;
> + unsigned char *src, *dst;
> +
> + src = malloc(a->size * 1024);
> + dst = malloc(a->size * 1024);
> + if ((!src) || (!dst)) {
> + printf("ERR: Can't allocate memory\n");
> + exit(1);
> + }
> + for (i = 0; i < len; i++)
> + memcpy(dst, src, a->size * 1024);
> +}
> +
> +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
> +{
> + pthread_t t;
> + pthread_attr_t attr;
> +
> + pthread_attr_init(&attr);
> + pthread_create(&t, &attr, fn, arg);
> + return t;
> +}
> +
> +int main(int argc, char **argv)
> +{
> + unsigned long i, len, size, thr;
> + pthread_t threads[256];
> + struct args args[256];
> + long long v;
> +
> + if (argc < 4) {
> + printf("ERR: %s [copysize Kb] [numthreads] [numloops (hundreds)]\n", argv[0]);
> + exit(1);
> + }
> +
> + v = atoll(argv[1]);
> + if ((v < 1) || (v > (1024 * 1024))) {
> + printf("ERR: max memory 1GB (1048576 KB)\n");
> + exit(1);
> + }
> + size = v;
> + thr = atol(argv[2]);
> + if ((thr < 1) || (thr > 256)) {
> + printf("ERR: threads 1-256\n");
> + exit(1);
> + }
> + v = atoll(argv[3]);
> + if ((v < 1) || (v > 40000000000ll)) {
> + printf("ERR: loops 1-40000000000 (hundreds)\n");
> + exit(1);
> + }
> + len = v * 100;
> + for (i = 0; i < thr; i++) {
> + args[i].loops = len;
> + args[i].size = size;
> + args[i].th = new_thr(thrfn, &(args[i]));
> + }
> + for (i = 0; i < thr; i++)
> + pthread_join(args[i].th, &(args[i].ret));
> + return 0;
> +}
> diff --git a/tools/perf/tests/shell/coresight/thread_loop/.gitignore b/tools/perf/tests/shell/coresight/thread_loop/.gitignore
> new file mode 100644
> index 000000000000..6d4c33eaa9e8
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/thread_loop/.gitignore
> @@ -0,0 +1 @@
> +thread_loop
> diff --git a/tools/perf/tests/shell/coresight/thread_loop/Makefile b/tools/perf/tests/shell/coresight/thread_loop/Makefile
> new file mode 100644
> index 000000000000..424df4e8b0e6
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/thread_loop/Makefile
> @@ -0,0 +1,29 @@
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +include ../Makefile.miniconfig
> +
> +BIN=thread_loop
> +LIB=-pthread
> +
> +all: $(BIN)
> +
> +$(BIN): $(BIN).c
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
> +endif
> +endif
> +
> +install-tests: all
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(call QUIET_INSTALL, tests) \
> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
> +endif
> +endif
> +
> +clean:
> + $(Q)$(RM) -f $(BIN)
> +
> +.PHONY: all clean install-tests
> diff --git a/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
> new file mode 100644
> index 000000000000..c0158fac7d0b
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
> @@ -0,0 +1,86 @@
> +// SPDX-License-Identifier: GPL-2.0
> +// Carsten Haitzler <[email protected]>, 2021
> +
> +// define this for gettid()
> +#define _GNU_SOURCE
> +
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <unistd.h>
> +#include <string.h>
> +#include <pthread.h>
> +#include <sys/syscall.h>
> +#ifndef SYS_gettid
> +// gettid is 178 on arm64
> +# define SYS_gettid 178
> +#endif
> +#define gettid() syscall(SYS_gettid)
> +
> +struct args {
> + unsigned int loops;
> + pthread_t th;
> + void *ret;
> +};
> +
> +static void *thrfn(void *arg)
> +{
> + struct args *a = arg;
> + int i = 0, len = a->loops;
> +
> + if (getenv("SHOW_TID")) {
> + unsigned long long tid = gettid();
> +
> + printf("%llu\n", tid);
> + }
> + asm volatile(
> + "loop:\n"
> + "add %[i], %[i], #1\n"
> + "cmp %[i], %[len]\n"
> + "blt loop\n"
> + : /* out */
> + : /* in */ [i] "r" (i), [len] "r" (len)
> + : /* clobber */
> + );
> + return (void *)(long)i;
> +}
> +
> +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
> +{
> + pthread_t t;
> + pthread_attr_t attr;
> +
> + pthread_attr_init(&attr);
> + pthread_create(&t, &attr, fn, arg);
> + return t;
> +}
> +
> +int main(int argc, char **argv)
> +{
> + unsigned int i, len, thr;
> + pthread_t threads[256];
> + struct args args[256];
> +
> + if (argc < 3) {
> + printf("ERR: %s [numthreads] [numloops (millions)]\n", argv[0]);
> + exit(1);
> + }
> +
> + thr = atoi(argv[1]);
> + if ((thr < 1) || (thr > 256)) {
> + printf("ERR: threads 1-256\n");
> + exit(1);
> + }
> + len = atoi(argv[2]);
> + if ((len < 1) || (len > 4000)) {
> + printf("ERR: max loops 4000 (millions)\n");
> + exit(1);
> + }
> + len *= 1000000;
> + for (i = 0; i < thr; i++) {
> + args[i].loops = len;
> + args[i].th = new_thr(thrfn, &(args[i]));
> + }
> + for (i = 0; i < thr; i++)
> + pthread_join(args[i].th, &(args[i].ret));
> + return 0;
> +}
> diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
> new file mode 100644
> index 000000000000..2cb4e996dbf3
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
> @@ -0,0 +1 @@
> +unroll_loop_thread
> diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
> new file mode 100644
> index 000000000000..45ab2be8be92
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
> @@ -0,0 +1,29 @@
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +include ../Makefile.miniconfig
> +
> +BIN=unroll_loop_thread
> +LIB=-pthread
> +
> +all: $(BIN)
> +
> +$(BIN): $(BIN).c
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
> +endif
> +endif
> +
> +install-tests: all
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(call QUIET_INSTALL, tests) \
> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
> +endif
> +endif
> +
> +clean:
> + $(Q)$(RM) -f $(BIN)
> +
> +.PHONY: all clean install-tests
> diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
> new file mode 100644
> index 000000000000..cb9d22c7dfb9
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
> @@ -0,0 +1,74 @@
> +// SPDX-License-Identifier: GPL-2.0
> +// Carsten Haitzler <[email protected]>, 2021
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <unistd.h>
> +#include <string.h>
> +#include <pthread.h>
> +
> +struct args {
> + pthread_t th;
> + unsigned int in, out;
> + void *ret;
> +};
> +
> +static void *thrfn(void *arg)
> +{
> + struct args *a = arg;
> + unsigned int i, in = a->in;
> +
> + for (i = 0; i < 10000; i++) {
> + asm volatile (
> +// force an unroll of thia add instruction so we can test long runs of code
> +#define SNIP1 "add %[in], %[in], #1\n"
> +// 10
> +#define SNIP2 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1
> +// 100
> +#define SNIP3 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2
> +// 1000
> +#define SNIP4 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3
> +// 10000
> +#define SNIP5 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4
> +// 100000
> + SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5
> + : /* out */
> + : /* in */ [in] "r" (in)
> + : /* clobber */
> + );
> + }
> +}
> +
> +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
> +{
> + pthread_t t;
> + pthread_attr_t attr;
> +
> + pthread_attr_init(&attr);
> + pthread_create(&t, &attr, fn, arg);
> + return t;
> +}
> +
> +int main(int argc, char **argv)
> +{
> + unsigned int i, thr;
> + pthread_t threads[256];
> + struct args args[256];
> +
> + if (argc < 2) {
> + printf("ERR: %s [numthreads]\n", argv[0]);
> + exit(1);
> + }
> +
> + thr = atoi(argv[1]);
> + if ((thr > 256) || (thr < 1)) {
> + printf("ERR: threads 1-256\n");
> + exit(1);
> + }
> + for (i = 0; i < thr; i++) {
> + args[i].in = rand();
> + args[i].th = new_thr(thrfn, &(args[i]));
> + }
> + for (i = 0; i < thr; i++)
> + pthread_join(args[i].th, &(args[i].ret));
> + return 0;
> +}
> diff --git a/tools/perf/tests/shell/coresight_asm_pure_loop.sh b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
> new file mode 100755
> index 000000000000..3f0dbefcad50
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
> @@ -0,0 +1,18 @@
> +#!/bin/sh -e
> +# Coresight / ASM Pure Loop
> +
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +TEST="asm_pure_loop"
> +. $(dirname $0)/lib/coresight.sh
> +ARGS=""
> +DATV="out"
> +DATA="$DATD/perf-$TEST-$DATV.data"
> +
> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
> +
> +perf_dump_aux_verify "$DATA" 10 10 10
> +
> +err=$?
> +exit $err
Can we organize the shell scripts by moving them into the folder
tools/perf/tests/shell/coresight?
coresight_asm_pure_loop.sh
coresight_memcpy_thread_16k_10.sh
coresight_thread_loop_check_tid_10.sh
coresight_thread_loop_check_tid_2.sh
coresight_unroll_loop_thread_10.sh
And we even can consider to move script test_arm_coresight.sh into
the folder tools/perf/tests/shell/coresight and change its
name as 'coresight_smoke_test.sh'.
Thanks,
Leo
> diff --git a/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh b/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
> new file mode 100755
> index 000000000000..8972af835016
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
> @@ -0,0 +1,18 @@
> +#!/bin/sh -e
> +# Coresight / Memcpy 16k 10 Threads
> +
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +TEST="memcpy_thread"
> +. $(dirname $0)/lib/coresight.sh
> +ARGS="16 10 1"
> +DATV="16k_10"
> +DATA="$DATD/perf-$TEST-$DATV.data"
> +
> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
> +
> +perf_dump_aux_verify "$DATA" 10 10 10
> +
> +err=$?
> +exit $err
> diff --git a/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh b/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
> new file mode 100755
> index 000000000000..5b468901f89b
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
> @@ -0,0 +1,19 @@
> +#!/bin/sh -e
> +# Coresight / Thread Loop 10 Threads - Check TID
> +
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +TEST="thread_loop"
> +. $(dirname $0)/lib/coresight.sh
> +ARGS="10 1"
> +DATV="check-tid-10th"
> +DATA="$DATD/perf-$TEST-$DATV.data"
> +STDO="$DATD/perf-$TEST-$DATV.stdout"
> +
> +SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO
> +
> +perf_dump_aux_tid_verify "$DATA" "$STDO"
> +
> +err=$?
> +exit $err
> diff --git a/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh b/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
> new file mode 100755
> index 000000000000..f8b7abd3aa03
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
> @@ -0,0 +1,19 @@
> +#!/bin/sh -e
> +# Coresight / Thread Loop 2 Threads - Check TID
> +
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +TEST="thread_loop"
> +. $(dirname $0)/lib/coresight.sh
> +ARGS="2 20"
> +DATV="check-tid-2th"
> +DATA="$DATD/perf-$TEST-$DATV.data"
> +STDO="$DATD/perf-$TEST-$DATV.stdout"
> +
> +SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO
> +
> +perf_dump_aux_tid_verify "$DATA" "$STDO"
> +
> +err=$?
> +exit $err
> diff --git a/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh b/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
> new file mode 100755
> index 000000000000..c985dfb025c2
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
> @@ -0,0 +1,18 @@
> +#!/bin/sh -e
> +# Coresight / Unroll Loop Thread 10
> +
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +TEST="unroll_loop_thread"
> +. $(dirname $0)/lib/coresight.sh
> +ARGS="10"
> +DATV="10"
> +DATA="$DATD/perf-$TEST-$DATV.data"
> +
> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
> +
> +perf_dump_aux_verify "$DATA" 10 10 10
> +
> +err=$?
> +exit $err
> diff --git a/tools/perf/tests/shell/lib/coresight.sh b/tools/perf/tests/shell/lib/coresight.sh
> new file mode 100644
> index 000000000000..6a611b073f02
> --- /dev/null
> +++ b/tools/perf/tests/shell/lib/coresight.sh
> @@ -0,0 +1,130 @@
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +# This is sourced from a driver script so no need for #!/bin... etc. at the
> +# top - the assumption below is that it runs as part of sourcing after the
> +# test sets up some basic env vars to say what it is.
> +
> +# perf record options for the perf tests to use
> +PERFRECMEM="-m ,128M"
> +PERFRECOPT="$PERFRECMEM -e cs_etm//u"
> +
> +# These tests need to be run as root or coresight won't allow large buffers
> +# and will not collect proper data
> +UID=`id -u`
> +if test "$UID" -ne 0; then
> + echo "Not running as root... skip"
> + exit 2
> +fi
> +
> +TOOLS=$(dirname $0)
> +DIR="$TOOLS/coresight/$TEST"
> +BIN="$DIR/$TEST"
> +# If the test tool/binary does not exist and is executable then skip the test
> +if ! test -x "$BIN"; then exit 2; fi
> +DATD="."
> +# If the data dir env is set then make the data dir use that instead of ./
> +if test -n "$PERF_TEST_CORESIGHT_DATADIR"; then
> + DATD="$PERF_TEST_CORESIGHT_DATADIR";
> +fi
> +# If the stat dir env is set then make the data dir use that instead of ./
> +STATD="."
> +if test -n "$PERF_TEST_CORESIGHT_STATDIR"; then
> + STATD="$PERF_TEST_CORESIGHT_STATDIR";
> +fi
> +
> +# Called if the test fails - error code 2
> +err() {
> + echo "$1"
> + exit 1
> +}
> +
> +# Check that some statistics from our perf
> +check_val_min() {
> + STATF="$4"
> + if test "$2" -lt "$3"; then
> + echo ", FAILED" >> "$STATF"
> + err "Sanity check number of $1 is too low ($2 < $3)"
> + fi
> +}
> +
> +perf_dump_aux_verify() {
> + # Some basic checking that the AUX chunk contains some sensible data
> + # to see that we are recording something and at least a minimum
> + # amount of it. We should almost always see F3 atoms in just about
> + # anything but certainly we will see some trace info and async atom
> + # chunks.
> + DUMP="$DATD/perf-tmp-aux-dump.txt"
> + perf report --stdio --dump -i "$1" | \
> + grep -o -e I_ATOM_F3 -e I_ASYNC -e I_TRACE_INFO > "$DUMP"
> + # Simply count how many of these atoms we find to see that we are
> + # producing a reasonable amount of data - exact checks are not sane
> + # as this is a lossy process where we may lose some blocks and the
> + # compiler may produce different code depending on the compiler and
> + # optimization options, so this is rough just to see if we're
> + # either missing almost all the data or all of it
> + ATOM_F3_NUM=`grep I_ATOM_F3 "$DUMP" | wc -l`
> + ATOM_ASYNC_NUM=`grep I_ASYNC "$DUMP" | wc -l`
> + ATOM_TRACE_INFO_NUM=`grep I_TRACE_INFO "$DUMP" | wc -l`
> + rm -f "$DUMP"
> +
> + # Arguments provide minimums for a pass
> + CHECK_F3_MIN="$2"
> + CHECK_ASYNC_MIN="$3"
> + CHECK_TRACE_INFO_MIN="$4"
> +
> + # Write out statistics, so over time you can track results to see if
> + # there is a pattern - for example we have less "noisy" results that
> + # produce more consistent amounts of data each run, to see if over
> + # time any techinques to minimize data loss are having an effect or
> + # not
> + STATF="$STATD/stats-$TEST-$DATV.csv"
> + if ! test -f "$STATF"; then
> + echo "ATOM F3 Count, Minimum, ATOM ASYNC Count, Minimum, TRACE INFO Count, Minimum" > "$STATF"
> + fi
> + echo -n "$ATOM_F3_NUM, $CHECK_F3_MIN, $ATOM_ASYNC_NUM, $CHECK_ASYNC_MIN, $ATOM_TRACE_INFO_NUM, $CHECK_TRACE_INFO_MIN" >> "$STATF"
> +
> + # Actually check to see if we passed or failed.
> + check_val_min "ATOM_F3" "$ATOM_F3_NUM" "$CHECK_F3_MIN" "$STATF"
> + check_val_min "ASYNC" "$ATOM_ASYNC_NUM" "$CHECK_ASYNC_MIN" "$STATF"
> + check_val_min "TRACE_INFO" "$ATOM_TRACE_INFO_NUM" "$CHECK_TRACE_INFO_MIN" "$STATF"
> + echo ", Ok" >> "$STATF"
> +}
> +
> +perf_dump_aux_tid_verify() {
> + # Specifically crafted test will produce a list of Tread ID's to
> + # stdout that need to be checked to see that they have had trace
> + # info collected in AUX blocks in the perf data. This will go
> + # through all the TID's that are listed as CID=0xabcdef and see
> + # that all the Thread IDs the test tool reports are in the perf
> + # data AUX chunks
> +
> + # The TID test tools will print a TID per stdout line that are being
> + # tested
> + TIDS=`cat "$2"`
> + # Scan the perf report to find the TIDs that are actually CID in hex
> + # and build a list of the ones found
> + FOUND_TIDS=`perf report --stdio --dump -i "$1" | \
> + grep -o "CID=0x[0-9a-z]\+" | sed 's/CID=//g' | \
> + uniq | sort | uniq`
> +
> + # Iterate over the list of TIDs that the test says it has and find
> + # them in the TIDs found in the perf report
> + MISSING=""
> + for TID2 in $TIDS; do
> + FOUND=""
> + for TIDHEX in $FOUND_TIDS; do
> + TID=`printf "%i" $TIDHEX`
> + if test "$TID" -eq "$TID2"; then
> + FOUND="y"
> + break
> + fi
> + done
> + if test -z "$FOUND"; then
> + MISSING="$MISSING $TID"
> + fi
> + done
> + if test -n "$MISSING"; then
> + err "Thread IDs $MISSING not found in perf AUX data"
> + fi
> +}
> --
> 2.32.0
>
On 4/10/22 09:30, Leo Yan wrote:
> Hi Carsten,
>
> On Wed, Mar 09, 2022 at 12:28:59PM +0000, [email protected] wrote:
>> From: Carsten Haitzler <[email protected]>
>>
>> This adds a test harness and tests to run perf record and examine the
>> resuling output when coresight is enabled on arm64 and check the
>> resulting quality of the output as part of perf test. These tests use
>> various tools to produce output from perf record then measure some key
>> specific aspects of that data to see if the data exists at all and
>> contains key aspects such as measuring some data for every thread of
>> a test or produces sufficient data for large exeuction runs of a large
>> executable. etc.
>>
>> Signed-off-by: Carsten Haitzler <[email protected]>
>> ---
>> MAINTAINERS | 4 +
>> tools/perf/.gitignore | 6 +-
>> tools/perf/Documentation/arm-coresight.txt | 140 ++++++++++++++++++
>> tools/perf/Makefile.perf | 14 +-
>> tools/perf/tests/shell/coresight/Makefile | 30 ++++
>> .../tests/shell/coresight/Makefile.miniconfig | 23 +++
>> .../shell/coresight/asm_pure_loop/.gitignore | 1 +
>> .../shell/coresight/asm_pure_loop/Makefile | 30 ++++
>> .../coresight/asm_pure_loop/asm_pure_loop.S | 28 ++++
>> .../shell/coresight/memcpy_thread/.gitignore | 1 +
>> .../shell/coresight/memcpy_thread/Makefile | 29 ++++
>> .../coresight/memcpy_thread/memcpy_thread.c | 79 ++++++++++
>> .../shell/coresight/thread_loop/.gitignore | 1 +
>> .../shell/coresight/thread_loop/Makefile | 29 ++++
>> .../shell/coresight/thread_loop/thread_loop.c | 86 +++++++++++
>> .../coresight/unroll_loop_thread/.gitignore | 1 +
>> .../coresight/unroll_loop_thread/Makefile | 29 ++++
>> .../unroll_loop_thread/unroll_loop_thread.c | 74 +++++++++
>> .../tests/shell/coresight_asm_pure_loop.sh | 18 +++
>> .../shell/coresight_memcpy_thread_16k_10.sh | 18 +++
>> .../coresight_thread_loop_check_tid_10.sh | 19 +++
>> .../coresight_thread_loop_check_tid_2.sh | 19 +++
>> .../shell/coresight_unroll_loop_thread_10.sh | 18 +++
>> tools/perf/tests/shell/lib/coresight.sh | 130 ++++++++++++++++
>
> Very big change... Why squash all patches form previous verion to this
> single one big patch? Usually the format with small patches is much
> better for reviewing.
I was asked to re-jig the tree and in doing so I also ended up cutting
down the size a lot so this just makes more sense together as a "here
are the tests" as adding infra without any tests makes no sense and the
tests themelves are self-contained in their own directories and source
files and "drivign scripts" thus it's essentially patch 1 appended to
patch 2 to patch 3 etc. and still broken up in the patch file by file.
> And I cannot apply cleanly on perf core branch:
> https://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git branch: perf/core.
I generated this based on tip. I can re-do it based on the above.
>> 24 files changed, 823 insertions(+), 4 deletions(-)
>> create mode 100644 tools/perf/Documentation/arm-coresight.txt
>> create mode 100644 tools/perf/tests/shell/coresight/Makefile
>> create mode 100644 tools/perf/tests/shell/coresight/Makefile.miniconfig
>> create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
>> create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
>> create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
>> create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
>> create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/Makefile
>> create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
>> create mode 100644 tools/perf/tests/shell/coresight/thread_loop/.gitignore
>> create mode 100644 tools/perf/tests/shell/coresight/thread_loop/Makefile
>> create mode 100644 tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
>> create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
>> create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
>> create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
>> create mode 100755 tools/perf/tests/shell/coresight_asm_pure_loop.sh
>> create mode 100755 tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
>> create mode 100755 tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
>> create mode 100755 tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
>> create mode 100755 tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
>> create mode 100644 tools/perf/tests/shell/lib/coresight.sh
>>
>> diff --git a/MAINTAINERS b/MAINTAINERS
>> index 673c7124ca82..18cc20609f2e 100644
>> --- a/MAINTAINERS
>> +++ b/MAINTAINERS
>> @@ -1918,10 +1918,14 @@ F: drivers/hwtracing/coresight/*
>> F: include/dt-bindings/arm/coresight-cti-dt.h
>> F: include/linux/coresight*
>> F: samples/coresight/*
>> +F: tools/perf/Documentation/arm-coresight.txt
>> F: tools/perf/arch/arm/util/auxtrace.c
>> F: tools/perf/arch/arm/util/cs-etm.c
>> F: tools/perf/arch/arm/util/cs-etm.h
>> F: tools/perf/arch/arm/util/pmu.c
>> +F: tools/perf/tests/shell/coresight_*
>> +F: tools/perf/tests/shell/tools/Makefile
>> +F: tools/perf/tests/shell/tools/coresight/*
>> F: tools/perf/util/cs-etm-decoder/*
>> F: tools/perf/util/cs-etm.*
>>
>> diff --git a/tools/perf/.gitignore b/tools/perf/.gitignore
>> index 20b8ab984d5f..138c679ecacd 100644
>> --- a/tools/perf/.gitignore
>> +++ b/tools/perf/.gitignore
>> @@ -15,8 +15,9 @@ perf*.1
>> perf*.xml
>> perf*.html
>> common-cmds.h
>> -perf.data
>> -perf.data.old
>> +perf*.data
>> +perf*.data.old
>> +stats-*.csv
>> output.svg
>> perf-archive
>> perf-with-kcore
>> @@ -30,6 +31,7 @@ config.mak.autogen
>> *-flex.*
>> *.pyc
>> *.pyo
>> +*.stdout
>> .config-detected
>> util/intel-pt-decoder/inat-tables.c
>> arch/*/include/generated/
>> diff --git a/tools/perf/Documentation/arm-coresight.txt b/tools/perf/Documentation/arm-coresight.txt
>> new file mode 100644
>> index 000000000000..3a9e6c573c58
>> --- /dev/null
>> +++ b/tools/perf/Documentation/arm-coresight.txt
>> @@ -0,0 +1,140 @@
>> +Arm Coresight Support
>> +=====================
>> +
>> +Coresight is a feature of some Arm based processors that allows for
>> +debugging. One of the things it can do is trace every instruction
>> +executed and remotely expose that information in a hardware compressed
>> +stream.
>
> Maybe here need to sync a bit for the terminology in
> Documentation/trace/coresight/coresight.rst.
>
> Something like:
>
> "Coresight is a feature of some Arm based processors that allows for
> debugging. One of the things it can do is trace instruction path and
> expose that information in a hardware compressed stream for either
> debugger or HW assisted tracing locally.
>
> See Documentation/trace/coresight/coresight.rst for details."
Sure. Can do that.
>> Perf is able to locally access that stream and store it to the
>> +output perf data files. This stream can then be later decoded to give the
>> +instructions that were traced for debugging or profiling purposes. You
>> +can log such data with a perf record command like:
>> +
>> + perf record -e cs_etm//u testbinary
>> +
>> +This would run some test binary (testbinary) until it exits and record
>> +a perf.data trace file. That file would have AUX sections if coresight
>> +is working correctly. You can dump the content of this file as
>> +readable text with a command like:
>> +
>> + perf report --stdio --dump -i perf.data
>> +
>> +You should find some sections of this file have AUX data blocks like:
>> +
>> + 0x1e78 [0x30]: PERF_RECORD_AUXTRACE size: 0x11dd0 offset: 0 ref: 0x1b614fc1061b0ad1 idx: 0 tid: 531230 cpu: -1
>> +
>> + . ... CoreSight ETM Trace data: size 73168 bytes
>> + Idx:0; ID:10; I_ASYNC : Alignment Synchronisation.
>> + Idx:12; ID:10; I_TRACE_INFO : Trace Info.; INFO=0x0 { CC.0 }
>> + Idx:17; ID:10; I_ADDR_L_64IS0 : Address, Long, 64 bit, IS0.; Addr=0x0000000000000000;
>> + Idx:26; ID:10; I_TRACE_ON : Trace On.
>> + Idx:27; ID:10; I_ADDR_CTXT_L_64IS0 : Address & Context, Long, 64 bit, IS0.; Addr=0x0000FFFFB6069140; Ctxt: AArch64,EL0, NS;
>> + Idx:38; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
>> + Idx:39; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
>> + Idx:40; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
>> + Idx:41; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEN
>> + ...
>> +
>> +If you see these above, then your system is tracing coresight data
>> +correctly.
>> +
>> +To compile perf with coresight support in the perf directory do
>> +
>> + make CORESIGHT=1
>
> It is inaccurate that if we don't mention openCSD lib.
Do you mean I need to mention that you need the opencsd library
installed too?
>> +
>> +This will compile the perf tool with coresight support as well as
>> +build some small test binaries for perf test. This requires you also
>> +be compiling for 64bit Arm (ARM64/aarch64). The tools run as part of
>> +perf coresight tracing are in tests/shell/tools/coresight.
>
> For build perf tool, I think above paragraphs are duplicate with the
> document Documentation/trace/coresight/coresight.rst. Can we simply
> say:
>
> "The details for building perf tool with support Arm Coresight can be
> found in the "HOWTO.md" file of the openCSD gitHub repository:
> https://github.com/Linaro/opencsd.
I can. I put this here as I didn't go clone OpencCSD first but used my
distro OpenCSD packages and thus of course didn't have the documentation
in front of me. I spent some time wondering why it wasn't building with
coresight support even though it detected OpenCSD when I compiled... I
didn't expect to have to go to some separate project git repository and
read docs there on how to build the perf tool here in the kernel. I
wrote this because it was an actual problem I hit and it's a lot less
frustrating to "end users" to give them the information they need in the
relevant place they need it instead of sending them around to other
project trees. Building perf with coresight support is handled by the
perf tree int he kernel, not OpenCSD, thus IMHO that is where the
documentation belongs - alongside the thing that determines how to build
something.
> And "HOWTO.md" file gives the information and examples for how to use
> perf tool to record and report Coresight trace data. It's the
> prerequisite for this perf Coresight test."
>
>> +You will also want coresight support enabled in your kernel config.
>> +Ensure it is enabled with:
>> +
>> + CONFIG_CORESIGHT=y
>> +
>> +There are various other coresight options you probably also want
>> +enabled like:
>> +
>> + CONFIG_CORESIGHT_LINKS_AND_SINKS=y
>> + CONFIG_CORESIGHT_LINK_AND_SINK_TMC=y
>> + CONFIG_CORESIGHT_CATU=y
>> + CONFIG_CORESIGHT_SINK_TPIU=y
>> + CONFIG_CORESIGHT_SINK_ETBV10=y
>> + CONFIG_CORESIGHT_SOURCE_ETM4X=y
>> + CONFIG_CORESIGHT_STM=y
>> + CONFIG_CORESIGHT_CPU_DEBUG=y
>> + CONFIG_CORESIGHT_CTI=y
>> + CONFIG_CORESIGHT_CTI_INTEGRATION_REGS=y
>> +
>> +Please refer to the kernel configuration help for more information.
>
> I prefer to remove these kernel configuration since they are not
> inconsistent on different platforms (e.g. ETBV10, ETM4X, etc), and
> some configurations might not necessary (e.g. CPU_DEBUG).
Certainly there should be some documentation on which kernel configs you
might want to turn on then? Imagine someone new comes along and doesn't
have any idea what to possible enable at all and manages to build perf
with coresight support (as above) then finds it doesn't work because
they didn't enable enough config in the kernel? Sure - could probably
trim these down a bit but the point here is to alert the user to there
being a range of coresight config options that you need to turn on that
you likely will find are not turned on. They certainly are not turned on
on distro kernels and a lot of the time when you have a platform that
already boots/works you start with your distro kernel config file
because you want everything enabled so it actually boots. I've learned
the hard way to do this as you manage to forget to turn on some MMC
driver or some other feature and your boot hangs or doesn't find rootfs etc.
What would you recommend then as a "turn these on and coresight will
almost certainly work for you on your given hardware " then?
>> +Perf test - Verify kernel and userspace perf coresight work
>> +===========================================================
>> +
>> +When you run perf test, it will do a lot of self tests. Some of those
>> +tests will cover Coresight (only if enabled and on ARM64). You
>> +generally would run perf test from the tools/perf directory in the
>> +kernel tree. Some tests will check some internal perf support like:
>> +
>> + Check Arm CoreSight trace data recording and synthesized samples
>> +
>> +Some others will actually use perf record and some test binaries that
>> +are in tests/shell/tools/coresight and will collect traces to ensure a
>> +minimum level of functionality is met. The scripts that launch these
>> +tests are in tests/shell. These will all look like:
>> +
>> + Coresight / Memcpy 1M 25 Threads
>> + Coresight / Unroll Loop Thread 2
>> + ...
>
> Please update based on the latest test case names, at my side, I can
> see the testing case like:
>
> Coresight / ASM Pure Loop
> Coresight / Memcpy 16k 10 Threads
> Coresight / Thread Loop 10 Threads - Check TID
> Coresight / Thread Loop 2 Threads - Check TID
> Coresight / Unroll Loop Thread 10
Oh sorry - yeah. I wrote the docs based on the earlier tests. Will fix.
>> +
>> +These perf record tests will not run if the tool binaries do not exist
>> +in tests/shell/tools/coresight/*/ and will be skipped. If you do not
>> +have coresight support in hardware then either do not build perf with
>> +coresight support or remove these binaries in order to not have these
>> +tests fail and have them skip instead.
>> +
>> +These tests will log historical results in the current working
>> +directory (e.g. tools/perf) and will be named stats-*.csv like:
>> +
>> + stats-asm_pure_loop-out.csv
>> + stats-bubble_sort-random.csv
>> + ...
>> +
>> +These statistic files log some aspects of the AUX data sections in
>> +the perf data output counting some numbers of certain encodings (a
>> +good way to know that it's working in a very simple way). One problem
>> +with coresight is that given a large enough amount of data needing to
>> +be logged, some of it can be lost due to the processor not waking up
>> +in time to read out all the data from buffers etc.. You will notice
>> +that the amount of data collected can vary a lot per run of perf test.
>> +If you wish to see how this changes over time, simply run perf test
>> +multiple times and all these csv files will have more and more data
>> +appended to it that you can later examine, graph and otherwise use to
>> +figure out if things have become worse or better.
>
> I am confused by this narrative. Does it try to remind that the final
> testing result (pass or fail) is not stable? Or should we run for
> multiple times so have more chance to capture issues?
That is correct. I thought I was clear that it's lossy. That is actually
the case. I have tests here that actually fail because there is no data
collected from some threads at all (missing CID blocks for some of the
threads that run in the test). The point is to have tests that may be
failing now but in future will improve. I lowered the minimum bar to
pass for most tests to have "at least just a little data" but most tests
show highly variable amount of captured data. the csv files are there to
over-time give you a good idea of the stability of the captured data.
>> +Be aware that amny of these tests take quite a while to run, specifically
>
> s/amny/many
Indeed. will fix.
>> +in processing the perf data file and dumping contents to then examine what
>> +is inside.
>> +
>> +You can change where these csv logs are stored by setting the
>> +PERF_TEST_CORESIGHT_STATDIR environment variable before running perf
>> +test like:
>> +
>> + export PERF_TEST_CORESIGHT_STATDIR=/var/tmp
>> + perf test
>> +
>> +They will also store resulting perf output data in the current
>> +directory for later inspection like:
>> +
>> + perf-memcpy-1m.data
>> + perf-thread_loop-2th.data
>> + ...
>> +
>> +You can alter where the perf data files are stored by setting the
>> +PERF_TEST_CORESIGHT_DATADIR environment variable such as:
>> +
>> + PERF_TEST_CORESIGHT_DATADIR=/var/tmp
>> + perf test
>> +
>> +You may wish to set these above environment variables if you which to
>> +keep the output of tests outside of the current working directory for
>> +longer term storage and examination.
>> diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
>> index ac861e42c8f7..b97db83992e0 100644
>> --- a/tools/perf/Makefile.perf
>> +++ b/tools/perf/Makefile.perf
>> @@ -630,7 +630,15 @@ sync_file_range_tbls := $(srctree)/tools/perf/trace/beauty/sync_file_range.sh
>> $(sync_file_range_arrays): $(linux_uapi_dir)/fs.h $(sync_file_range_tbls)
>> $(Q)$(SHELL) '$(sync_file_range_tbls)' $(linux_uapi_dir) > $@
>>
>> -all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS)
>> +TESTS_CORESIGHT_DIR := $(srctree)/tools/perf/tests/shell/coresight
>> +
>> +tests-coresight-targets: FORCE
>> + $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR)
>> +
>> +tests-coresight-targets-clean:
>> + $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR) clean
>> +
>> +all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS) tests-coresight-targets
>>
>> # Create python binding output directory if not already present
>> _dummy := $(shell [ -d '$(OUTPUT)python' ] || mkdir -p '$(OUTPUT)python')
>> @@ -1020,6 +1028,7 @@ install-tests: all install-gtk
>> $(INSTALL) tests/shell/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell'; \
>> $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'; \
>> $(INSTALL) tests/shell/lib/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'
>> + $(Q)$(MAKE) -C tests/shell/coresight install-tests
>>
>> install-bin: install-tools install-tests install-traceevent-plugins
>>
>> @@ -1088,7 +1097,7 @@ endif # BUILD_BPF_SKEL
>> bpf-skel-clean:
>> $(call QUIET_CLEAN, bpf-skel) $(RM) -r $(SKEL_TMP_OUT) $(SKELETONS)
>>
>> -clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean
>> +clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean tests-coresight-targets-clean
>> $(call QUIET_CLEAN, core-objs) $(RM) $(LIBPERF_A) $(OUTPUT)perf-archive $(OUTPUT)perf-with-kcore $(OUTPUT)perf-iostat $(LANG_BINDINGS)
>> $(Q)find $(if $(OUTPUT),$(OUTPUT),.) -name '*.o' -delete -o -name '\.*.cmd' -delete -o -name '\.*.d' -delete
>> $(Q)$(RM) $(OUTPUT).config-detected
>> @@ -1155,5 +1164,6 @@ FORCE:
>> .PHONY: shell_compatibility_test please_set_SHELL_PATH_to_a_more_modern_shell
>> .PHONY: $(GIT-HEAD-PHONY) TAGS tags cscope FORCE prepare
>> .PHONY: libtraceevent_plugins archheaders
>> +.PHONY: $(TESTS_CORESIGHT_TARGETS)
>
> I don't find other places using TESTS_CORESIGHT_TARGETS. Is this
> redundant?
I'll check - it may have been left over from my previous patch set.
>> endif # force_fixdep
>> diff --git a/tools/perf/tests/shell/coresight/Makefile b/tools/perf/tests/shell/coresight/Makefile
>> new file mode 100644
>> index 000000000000..dda99aeac158
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/Makefile
>> @@ -0,0 +1,30 @@
>> +# SPDX-License-Identifier: GPL-2.0-only
>> +# Carsten Haitzler <[email protected]>, 2021
>> +include ../../../../../tools/scripts/Makefile.include
>> +include ../../../../../tools/scripts/Makefile.arch
>> +include ../../../../../tools/scripts/utilities.mak
>> +
>> +SUBDIRS = \
>> + asm_pure_loop \
>> + thread_loop \
>> + memcpy_thread \
>> + unroll_loop_thread
>> +
>> +all: $(SUBDIRS)
>> +$(SUBDIRS):
>> + $(Q)$(MAKE) -C $@
>> +
>> +INSTALLDIRS = $(SUBDIRS:%=install-%)
>> +
>> +install-tests: $(INSTALLDIRS)
>> +$(INSTALLDIRS):
>> + $(Q)$(MAKE) -C $(@:install-%=%) install-tests
>> +
>> +CLEANDIRS = $(SUBDIRS:%=clean-%)
>> +
>> +clean: $(CLEANDIRS)
>> +$(CLEANDIRS):
>> + $(Q)$(MAKE) -C $(@:clean-%=%) clean >/dev/null
>> +
>> +.PHONY: all clean $(SUBDIRS) $(CLEANDIRS) $(INSTALLDIRS)
>> +
>> diff --git a/tools/perf/tests/shell/coresight/Makefile.miniconfig b/tools/perf/tests/shell/coresight/Makefile.miniconfig
>> new file mode 100644
>> index 000000000000..893c12685fed
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/Makefile.miniconfig
>> @@ -0,0 +1,23 @@
>> +# SPDX-License-Identifier: GPL-2.0-only
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +ifndef DESTDIR
>> +prefix ?= $(HOME)
>> +endif
>> +
>> +DESTDIR_SQ = $(subst ','\'',$(DESTDIR))
>> +perfexecdir = libexec/perf-core
>> +perfexec_instdir = $(perfexecdir)
>> +
>> +ifneq ($(filter /%,$(firstword $(perfexecdir))),)
>> +perfexec_instdir = $(perfexecdir)
>> +else
>> +perfexec_instdir = $(prefix)/$(perfexecdir)
>> +endif
>> +
>> +perfexec_instdir_SQ = $(subst ','\'',$(perfexec_instdir))
>> +INSTALL = install
>> +
>> +include ../../../../../scripts/Makefile.include
>> +include ../../../../../scripts/Makefile.arch
>> +include ../../../../../scripts/utilities.mak
>> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
>> new file mode 100644
>> index 000000000000..468673ac32e8
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
>> @@ -0,0 +1 @@
>> +asm_pure_loop
>
> Do we really need there '.gitignore' files under the folder
> 'tools/perf/tests/shell/coresight/'.
Where would you rather have them to ignore the generated binary tools?
>> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
>> new file mode 100644
>> index 000000000000..10c5a60cb71c
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
>> @@ -0,0 +1,30 @@
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +include ../Makefile.miniconfig
>> +
>> +BIN=asm_pure_loop
>> +LIB=
>
> Remove the unused variable 'LIB='.
I have this because I wanted to have a simple template to be able to
re-use for more tests over time. It's so much easier to maintain and
extend if every makefile and tool follow a similar pattern and you can
almost copy & paste between them as they don't have "exceptions". You
really want me to remove this?
>> +
>> +all: $(BIN)
>> +
>> +$(BIN): $(BIN).S
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(Q)$(CC) $(BIN).S -nostdlib -static -o $(BIN) $(LIB)
>> +endif
>> +endif
>> +
>> +install-tests: all
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(call QUIET_INSTALL, tests) \
>> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
>> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
>> +endif
>> +endif
>> +
>> +clean:
>> + $(Q)$(RM) -f $(BIN)
>> +
>> +.PHONY: all clean install-tests
>
> There have four sub folders under tools/perf/tests/shell/coresight:
>
> asm_pure_loop
> memcpy_thread
> thread_loop
> unroll_loop_thread
>
> And every folder has its own Makefile and every Makefile is quite
> close to each other. I am just wandering if it's possible to
> remove the 4 Makefiles in these four sub folders, and simply use
> tools/perf/tests/shell/coresight/Makefile as the central place to
> build these assistant programs.
I did this so it's easier to etxent over time. having a single parent
makefile that over time accumulates little ugly "if's" and exceptions
makes longer-term maintenance and extending harder. I did it this way to
make this easy - make a copy of a dir - add that dir to a parent
makefile then modify the makefile as needed (but only as needed).
>> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
>> new file mode 100644
>> index 000000000000..75cf084a927d
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
>> @@ -0,0 +1,28 @@
>> +/* SPDX-License-Identifier: GPL-2.0 */
>> +/* Tamas Zsoldos <[email protected]>, 2021 */
>> +
>> +.globl _start
>> +_start:
>> + mov x0, 0x0000ffff
>> + mov x1, xzr
>> +loop:
>> + nop
>> + nop
>> + cbnz x1, noskip
>> + nop
>> + nop
>> + adrp x2, skip
>> + add x2, x2, :lo12:skip
>> + br x2
>> + nop
>> + nop
>> +noskip:
>> + nop
>> + nop
>> +skip:
>> + sub x0, x0, 1
>> + cbnz x0, loop
>> +
>> + mov x0, #0
>> + mov x8, #93 // __NR_exit syscall
>> + svc #0
>
> I tested the case "ASM Pure Loop" on my Juno board, and it complaints:
>
> root@debian:/mnt/export/arm-linux-kernel/tools/perf# ./perf test -v 76
> 76: Coresight / ASM Pure Loop :
> --- start ---
> test child forked, pid 9063
> failed to mmap with 12 (Cannot allocate memory)
> test child finished with -1
> ---- end ----
> Coresight / ASM Pure Loop: FAILED!
>
> Since I only setup the 1GB memory for the Linux kernel, it fails to
> allocate AUX ring buffer with the size 256MB. So I manully change
> the buffer size to 8MB in tools/perf/tests/shell/lib/coresight.sh:
>
> PERFRECMEM="-m ,8M"
>
> So finally I can see the test case is passed:
This is artificial isn't it? limiting to 1GB. You certainly have far
more memory than that available. My testse were on a system with 4GB and
I had no issues.
> root@debian:/mnt/export/arm-linux-kernel/tools/perf# ./perf test -v 76
> 76: Coresight / ASM Pure Loop :
> --- start ---
> test child forked, pid 9481
> -m ,8M -e cs_etm//u
> [ perf record: Woken up 1 times to write data ]
> [ perf record: Captured and wrote 0.681 MB ./perf-asm_pure_loop-out.data ]
> test child finished with 0
> ---- end ----
> Coresight / ASM Pure Loop: Ok
>
> Do you think we really need to use 256MiB as the AUX buffer size?
> IIRC, it means we allocate 256MiB per CPU for this case, on the other
> hand, you could see the final perf data file size is small (0.681
> MiB).
>
> Seems to me, it's not necessary to allocate so big buffer for
> the test, and I tried to run below 4 cases with 8MiB, all of them can
> pass the testing :)
I didn't think anyone with a system with coresight support that would be
running perf record locally would only have 1GB of ram... I knew junos
had 8GB and my dragonboard has 4GB ... so I know I was on the smaller
side. I thought a larger buffer == safer results (less chance of needing
to write out the buffer during capture). Admittdly I used 256Mb when my
tests ran for much longer and collected more data. I can try drop to 8
or 16gb and see.
>> diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
>> new file mode 100644
>> index 000000000000..f8217e56091e
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
>> @@ -0,0 +1 @@
>> +memcpy_thread
>> diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/Makefile b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile
>> new file mode 100644
>> index 000000000000..e2604cfae74b
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile
>> @@ -0,0 +1,29 @@
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +include ../Makefile.miniconfig
>> +
>> +BIN=memcpy_thread
>> +LIB=-pthread
>> +
>> +all: $(BIN)
>> +
>> +$(BIN): $(BIN).c
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
>> +endif
>> +endif
>> +
>> +install-tests: all
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(call QUIET_INSTALL, tests) \
>> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
>> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
>> +endif
>> +endif
>> +
>> +clean:
>> + $(Q)$(RM) -f $(BIN)
>> +
>> +.PHONY: all clean install-tests
>> diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
>> new file mode 100644
>> index 000000000000..a7e169d1bf64
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
>> @@ -0,0 +1,79 @@
>> +// SPDX-License-Identifier: GPL-2.0
>> +// Carsten Haitzler <[email protected]>, 2021
>> +#include <stdio.h>
>> +#include <stdlib.h>
>> +#include <unistd.h>
>> +#include <string.h>
>> +#include <pthread.h>
>> +
>> +struct args {
>> + unsigned long loops;
>> + unsigned long size;
>> + pthread_t th;
>> + void *ret;
>> +};
>> +
>> +static void *thrfn(void *arg)
>> +{
>> + struct args *a = arg;
>> + unsigned long i, len = a->loops;
>> + unsigned char *src, *dst;
>> +
>> + src = malloc(a->size * 1024);
>> + dst = malloc(a->size * 1024);
>> + if ((!src) || (!dst)) {
>> + printf("ERR: Can't allocate memory\n");
>> + exit(1);
>> + }
>> + for (i = 0; i < len; i++)
>> + memcpy(dst, src, a->size * 1024);
>> +}
>> +
>> +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
>> +{
>> + pthread_t t;
>> + pthread_attr_t attr;
>> +
>> + pthread_attr_init(&attr);
>> + pthread_create(&t, &attr, fn, arg);
>> + return t;
>> +}
>> +
>> +int main(int argc, char **argv)
>> +{
>> + unsigned long i, len, size, thr;
>> + pthread_t threads[256];
>> + struct args args[256];
>> + long long v;
>> +
>> + if (argc < 4) {
>> + printf("ERR: %s [copysize Kb] [numthreads] [numloops (hundreds)]\n", argv[0]);
>> + exit(1);
>> + }
>> +
>> + v = atoll(argv[1]);
>> + if ((v < 1) || (v > (1024 * 1024))) {
>> + printf("ERR: max memory 1GB (1048576 KB)\n");
>> + exit(1);
>> + }
>> + size = v;
>> + thr = atol(argv[2]);
>> + if ((thr < 1) || (thr > 256)) {
>> + printf("ERR: threads 1-256\n");
>> + exit(1);
>> + }
>> + v = atoll(argv[3]);
>> + if ((v < 1) || (v > 40000000000ll)) {
>> + printf("ERR: loops 1-40000000000 (hundreds)\n");
>> + exit(1);
>> + }
>> + len = v * 100;
>> + for (i = 0; i < thr; i++) {
>> + args[i].loops = len;
>> + args[i].size = size;
>> + args[i].th = new_thr(thrfn, &(args[i]));
>> + }
>> + for (i = 0; i < thr; i++)
>> + pthread_join(args[i].th, &(args[i].ret));
>> + return 0;
>> +}
>> diff --git a/tools/perf/tests/shell/coresight/thread_loop/.gitignore b/tools/perf/tests/shell/coresight/thread_loop/.gitignore
>> new file mode 100644
>> index 000000000000..6d4c33eaa9e8
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/thread_loop/.gitignore
>> @@ -0,0 +1 @@
>> +thread_loop
>> diff --git a/tools/perf/tests/shell/coresight/thread_loop/Makefile b/tools/perf/tests/shell/coresight/thread_loop/Makefile
>> new file mode 100644
>> index 000000000000..424df4e8b0e6
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/thread_loop/Makefile
>> @@ -0,0 +1,29 @@
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +include ../Makefile.miniconfig
>> +
>> +BIN=thread_loop
>> +LIB=-pthread
>> +
>> +all: $(BIN)
>> +
>> +$(BIN): $(BIN).c
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
>> +endif
>> +endif
>> +
>> +install-tests: all
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(call QUIET_INSTALL, tests) \
>> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
>> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
>> +endif
>> +endif
>> +
>> +clean:
>> + $(Q)$(RM) -f $(BIN)
>> +
>> +.PHONY: all clean install-tests
>> diff --git a/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
>> new file mode 100644
>> index 000000000000..c0158fac7d0b
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
>> @@ -0,0 +1,86 @@
>> +// SPDX-License-Identifier: GPL-2.0
>> +// Carsten Haitzler <[email protected]>, 2021
>> +
>> +// define this for gettid()
>> +#define _GNU_SOURCE
>> +
>> +#include <stdio.h>
>> +#include <stdlib.h>
>> +#include <unistd.h>
>> +#include <string.h>
>> +#include <pthread.h>
>> +#include <sys/syscall.h>
>> +#ifndef SYS_gettid
>> +// gettid is 178 on arm64
>> +# define SYS_gettid 178
>> +#endif
>> +#define gettid() syscall(SYS_gettid)
>> +
>> +struct args {
>> + unsigned int loops;
>> + pthread_t th;
>> + void *ret;
>> +};
>> +
>> +static void *thrfn(void *arg)
>> +{
>> + struct args *a = arg;
>> + int i = 0, len = a->loops;
>> +
>> + if (getenv("SHOW_TID")) {
>> + unsigned long long tid = gettid();
>> +
>> + printf("%llu\n", tid);
>> + }
>> + asm volatile(
>> + "loop:\n"
>> + "add %[i], %[i], #1\n"
>> + "cmp %[i], %[len]\n"
>> + "blt loop\n"
>> + : /* out */
>> + : /* in */ [i] "r" (i), [len] "r" (len)
>> + : /* clobber */
>> + );
>> + return (void *)(long)i;
>> +}
>> +
>> +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
>> +{
>> + pthread_t t;
>> + pthread_attr_t attr;
>> +
>> + pthread_attr_init(&attr);
>> + pthread_create(&t, &attr, fn, arg);
>> + return t;
>> +}
>> +
>> +int main(int argc, char **argv)
>> +{
>> + unsigned int i, len, thr;
>> + pthread_t threads[256];
>> + struct args args[256];
>> +
>> + if (argc < 3) {
>> + printf("ERR: %s [numthreads] [numloops (millions)]\n", argv[0]);
>> + exit(1);
>> + }
>> +
>> + thr = atoi(argv[1]);
>> + if ((thr < 1) || (thr > 256)) {
>> + printf("ERR: threads 1-256\n");
>> + exit(1);
>> + }
>> + len = atoi(argv[2]);
>> + if ((len < 1) || (len > 4000)) {
>> + printf("ERR: max loops 4000 (millions)\n");
>> + exit(1);
>> + }
>> + len *= 1000000;
>> + for (i = 0; i < thr; i++) {
>> + args[i].loops = len;
>> + args[i].th = new_thr(thrfn, &(args[i]));
>> + }
>> + for (i = 0; i < thr; i++)
>> + pthread_join(args[i].th, &(args[i].ret));
>> + return 0;
>> +}
>> diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
>> new file mode 100644
>> index 000000000000..2cb4e996dbf3
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
>> @@ -0,0 +1 @@
>> +unroll_loop_thread
>> diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
>> new file mode 100644
>> index 000000000000..45ab2be8be92
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
>> @@ -0,0 +1,29 @@
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +include ../Makefile.miniconfig
>> +
>> +BIN=unroll_loop_thread
>> +LIB=-pthread
>> +
>> +all: $(BIN)
>> +
>> +$(BIN): $(BIN).c
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
>> +endif
>> +endif
>> +
>> +install-tests: all
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(call QUIET_INSTALL, tests) \
>> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
>> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
>> +endif
>> +endif
>> +
>> +clean:
>> + $(Q)$(RM) -f $(BIN)
>> +
>> +.PHONY: all clean install-tests
>> diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
>> new file mode 100644
>> index 000000000000..cb9d22c7dfb9
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
>> @@ -0,0 +1,74 @@
>> +// SPDX-License-Identifier: GPL-2.0
>> +// Carsten Haitzler <[email protected]>, 2021
>> +#include <stdio.h>
>> +#include <stdlib.h>
>> +#include <unistd.h>
>> +#include <string.h>
>> +#include <pthread.h>
>> +
>> +struct args {
>> + pthread_t th;
>> + unsigned int in, out;
>> + void *ret;
>> +};
>> +
>> +static void *thrfn(void *arg)
>> +{
>> + struct args *a = arg;
>> + unsigned int i, in = a->in;
>> +
>> + for (i = 0; i < 10000; i++) {
>> + asm volatile (
>> +// force an unroll of thia add instruction so we can test long runs of code
>> +#define SNIP1 "add %[in], %[in], #1\n"
>> +// 10
>> +#define SNIP2 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1
>> +// 100
>> +#define SNIP3 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2
>> +// 1000
>> +#define SNIP4 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3
>> +// 10000
>> +#define SNIP5 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4
>> +// 100000
>> + SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5
>> + : /* out */
>> + : /* in */ [in] "r" (in)
>> + : /* clobber */
>> + );
>> + }
>> +}
>> +
>> +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
>> +{
>> + pthread_t t;
>> + pthread_attr_t attr;
>> +
>> + pthread_attr_init(&attr);
>> + pthread_create(&t, &attr, fn, arg);
>> + return t;
>> +}
>> +
>> +int main(int argc, char **argv)
>> +{
>> + unsigned int i, thr;
>> + pthread_t threads[256];
>> + struct args args[256];
>> +
>> + if (argc < 2) {
>> + printf("ERR: %s [numthreads]\n", argv[0]);
>> + exit(1);
>> + }
>> +
>> + thr = atoi(argv[1]);
>> + if ((thr > 256) || (thr < 1)) {
>> + printf("ERR: threads 1-256\n");
>> + exit(1);
>> + }
>> + for (i = 0; i < thr; i++) {
>> + args[i].in = rand();
>> + args[i].th = new_thr(thrfn, &(args[i]));
>> + }
>> + for (i = 0; i < thr; i++)
>> + pthread_join(args[i].th, &(args[i].ret));
>> + return 0;
>> +}
>> diff --git a/tools/perf/tests/shell/coresight_asm_pure_loop.sh b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
>> new file mode 100755
>> index 000000000000..3f0dbefcad50
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
>> @@ -0,0 +1,18 @@
>> +#!/bin/sh -e
>> +# Coresight / ASM Pure Loop
>> +
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +TEST="asm_pure_loop"
>> +. $(dirname $0)/lib/coresight.sh
>> +ARGS=""
>> +DATV="out"
>> +DATA="$DATD/perf-$TEST-$DATV.data"
>> +
>> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
>> +
>> +perf_dump_aux_verify "$DATA" 10 10 10
>> +
>> +err=$?
>> +exit $err
>
> Can we organize the shell scripts by moving them into the folder
> tools/perf/tests/shell/coresight?
We can - but it comes with a fair few more changes.
> coresight_asm_pure_loop.sh
> coresight_memcpy_thread_16k_10.sh
> coresight_thread_loop_check_tid_10.sh
> coresight_thread_loop_check_tid_2.sh
> coresight_unroll_loop_thread_10.sh
>
> And we even can consider to move script test_arm_coresight.sh into
> the folder tools/perf/tests/shell/coresight and change its
> name as 'coresight_smoke_test.sh'.
Indeed these other tests I left alone for now and had not thought about
how to marry these together yet - leaving this for another day and
another patch set rather than this patch set itself. That was my
thoguht. I was trying to make an "Easier to extend by just dropping a
test into a dir" setup here to make maintenance and expansion easier
over time (and thus encourage testing by having a simple repeatable test
infra to duplicate). I ended up with a dir per test tool you need to
build and a driver script in the tests/shell dir. I think this is
certainly worth considering but perhaps as a separate set of work to
marry these?
I piggybacked on the existing shell test infra but added a fair few more
scripts. To do what you suggest I'd need to modify the core shell test
code to walk subdirs recursively then looking for child scripts. The
problem is how does perf test's shell handling know about the coresight
subdir vs the lib subdir? Both contain *.sh shell scripts - the
difference is the ones in lib are not executable. Is this sufficiently
different? I could also open them to check the have #!/bin/... as the
first line. Hardcoding just a single coresight subdir just feels wrong
and hacky to me, thus the generic recursion solution I suggest here.
I can definitely see how extending to subdirs would make supporting
testing cleaner and divide things into their own domains (dirs).
> Thanks,
> Leo
>
>> diff --git a/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh b/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
>> new file mode 100755
>> index 000000000000..8972af835016
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
>> @@ -0,0 +1,18 @@
>> +#!/bin/sh -e
>> +# Coresight / Memcpy 16k 10 Threads
>> +
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +TEST="memcpy_thread"
>> +. $(dirname $0)/lib/coresight.sh
>> +ARGS="16 10 1"
>> +DATV="16k_10"
>> +DATA="$DATD/perf-$TEST-$DATV.data"
>> +
>> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
>> +
>> +perf_dump_aux_verify "$DATA" 10 10 10
>> +
>> +err=$?
>> +exit $err
>> diff --git a/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh b/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
>> new file mode 100755
>> index 000000000000..5b468901f89b
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
>> @@ -0,0 +1,19 @@
>> +#!/bin/sh -e
>> +# Coresight / Thread Loop 10 Threads - Check TID
>> +
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +TEST="thread_loop"
>> +. $(dirname $0)/lib/coresight.sh
>> +ARGS="10 1"
>> +DATV="check-tid-10th"
>> +DATA="$DATD/perf-$TEST-$DATV.data"
>> +STDO="$DATD/perf-$TEST-$DATV.stdout"
>> +
>> +SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO
>> +
>> +perf_dump_aux_tid_verify "$DATA" "$STDO"
>> +
>> +err=$?
>> +exit $err
>> diff --git a/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh b/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
>> new file mode 100755
>> index 000000000000..f8b7abd3aa03
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
>> @@ -0,0 +1,19 @@
>> +#!/bin/sh -e
>> +# Coresight / Thread Loop 2 Threads - Check TID
>> +
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +TEST="thread_loop"
>> +. $(dirname $0)/lib/coresight.sh
>> +ARGS="2 20"
>> +DATV="check-tid-2th"
>> +DATA="$DATD/perf-$TEST-$DATV.data"
>> +STDO="$DATD/perf-$TEST-$DATV.stdout"
>> +
>> +SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO
>> +
>> +perf_dump_aux_tid_verify "$DATA" "$STDO"
>> +
>> +err=$?
>> +exit $err
>> diff --git a/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh b/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
>> new file mode 100755
>> index 000000000000..c985dfb025c2
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
>> @@ -0,0 +1,18 @@
>> +#!/bin/sh -e
>> +# Coresight / Unroll Loop Thread 10
>> +
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +TEST="unroll_loop_thread"
>> +. $(dirname $0)/lib/coresight.sh
>> +ARGS="10"
>> +DATV="10"
>> +DATA="$DATD/perf-$TEST-$DATV.data"
>> +
>> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
>> +
>> +perf_dump_aux_verify "$DATA" 10 10 10
>> +
>> +err=$?
>> +exit $err
>> diff --git a/tools/perf/tests/shell/lib/coresight.sh b/tools/perf/tests/shell/lib/coresight.sh
>> new file mode 100644
>> index 000000000000..6a611b073f02
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/lib/coresight.sh
>> @@ -0,0 +1,130 @@
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +# This is sourced from a driver script so no need for #!/bin... etc. at the
>> +# top - the assumption below is that it runs as part of sourcing after the
>> +# test sets up some basic env vars to say what it is.
>> +
>> +# perf record options for the perf tests to use
>> +PERFRECMEM="-m ,128M"
>> +PERFRECOPT="$PERFRECMEM -e cs_etm//u"
>> +
>> +# These tests need to be run as root or coresight won't allow large buffers
>> +# and will not collect proper data
>> +UID=`id -u`
>> +if test "$UID" -ne 0; then
>> + echo "Not running as root... skip"
>> + exit 2
>> +fi
>> +
>> +TOOLS=$(dirname $0)
>> +DIR="$TOOLS/coresight/$TEST"
>> +BIN="$DIR/$TEST"
>> +# If the test tool/binary does not exist and is executable then skip the test
>> +if ! test -x "$BIN"; then exit 2; fi
>> +DATD="."
>> +# If the data dir env is set then make the data dir use that instead of ./
>> +if test -n "$PERF_TEST_CORESIGHT_DATADIR"; then
>> + DATD="$PERF_TEST_CORESIGHT_DATADIR";
>> +fi
>> +# If the stat dir env is set then make the data dir use that instead of ./
>> +STATD="."
>> +if test -n "$PERF_TEST_CORESIGHT_STATDIR"; then
>> + STATD="$PERF_TEST_CORESIGHT_STATDIR";
>> +fi
>> +
>> +# Called if the test fails - error code 2
>> +err() {
>> + echo "$1"
>> + exit 1
>> +}
>> +
>> +# Check that some statistics from our perf
>> +check_val_min() {
>> + STATF="$4"
>> + if test "$2" -lt "$3"; then
>> + echo ", FAILED" >> "$STATF"
>> + err "Sanity check number of $1 is too low ($2 < $3)"
>> + fi
>> +}
>> +
>> +perf_dump_aux_verify() {
>> + # Some basic checking that the AUX chunk contains some sensible data
>> + # to see that we are recording something and at least a minimum
>> + # amount of it. We should almost always see F3 atoms in just about
>> + # anything but certainly we will see some trace info and async atom
>> + # chunks.
>> + DUMP="$DATD/perf-tmp-aux-dump.txt"
>> + perf report --stdio --dump -i "$1" | \
>> + grep -o -e I_ATOM_F3 -e I_ASYNC -e I_TRACE_INFO > "$DUMP"
>> + # Simply count how many of these atoms we find to see that we are
>> + # producing a reasonable amount of data - exact checks are not sane
>> + # as this is a lossy process where we may lose some blocks and the
>> + # compiler may produce different code depending on the compiler and
>> + # optimization options, so this is rough just to see if we're
>> + # either missing almost all the data or all of it
>> + ATOM_F3_NUM=`grep I_ATOM_F3 "$DUMP" | wc -l`
>> + ATOM_ASYNC_NUM=`grep I_ASYNC "$DUMP" | wc -l`
>> + ATOM_TRACE_INFO_NUM=`grep I_TRACE_INFO "$DUMP" | wc -l`
>> + rm -f "$DUMP"
>> +
>> + # Arguments provide minimums for a pass
>> + CHECK_F3_MIN="$2"
>> + CHECK_ASYNC_MIN="$3"
>> + CHECK_TRACE_INFO_MIN="$4"
>> +
>> + # Write out statistics, so over time you can track results to see if
>> + # there is a pattern - for example we have less "noisy" results that
>> + # produce more consistent amounts of data each run, to see if over
>> + # time any techinques to minimize data loss are having an effect or
>> + # not
>> + STATF="$STATD/stats-$TEST-$DATV.csv"
>> + if ! test -f "$STATF"; then
>> + echo "ATOM F3 Count, Minimum, ATOM ASYNC Count, Minimum, TRACE INFO Count, Minimum" > "$STATF"
>> + fi
>> + echo -n "$ATOM_F3_NUM, $CHECK_F3_MIN, $ATOM_ASYNC_NUM, $CHECK_ASYNC_MIN, $ATOM_TRACE_INFO_NUM, $CHECK_TRACE_INFO_MIN" >> "$STATF"
>> +
>> + # Actually check to see if we passed or failed.
>> + check_val_min "ATOM_F3" "$ATOM_F3_NUM" "$CHECK_F3_MIN" "$STATF"
>> + check_val_min "ASYNC" "$ATOM_ASYNC_NUM" "$CHECK_ASYNC_MIN" "$STATF"
>> + check_val_min "TRACE_INFO" "$ATOM_TRACE_INFO_NUM" "$CHECK_TRACE_INFO_MIN" "$STATF"
>> + echo ", Ok" >> "$STATF"
>> +}
>> +
>> +perf_dump_aux_tid_verify() {
>> + # Specifically crafted test will produce a list of Tread ID's to
>> + # stdout that need to be checked to see that they have had trace
>> + # info collected in AUX blocks in the perf data. This will go
>> + # through all the TID's that are listed as CID=0xabcdef and see
>> + # that all the Thread IDs the test tool reports are in the perf
>> + # data AUX chunks
>> +
>> + # The TID test tools will print a TID per stdout line that are being
>> + # tested
>> + TIDS=`cat "$2"`
>> + # Scan the perf report to find the TIDs that are actually CID in hex
>> + # and build a list of the ones found
>> + FOUND_TIDS=`perf report --stdio --dump -i "$1" | \
>> + grep -o "CID=0x[0-9a-z]\+" | sed 's/CID=//g' | \
>> + uniq | sort | uniq`
>> +
>> + # Iterate over the list of TIDs that the test says it has and find
>> + # them in the TIDs found in the perf report
>> + MISSING=""
>> + for TID2 in $TIDS; do
>> + FOUND=""
>> + for TIDHEX in $FOUND_TIDS; do
>> + TID=`printf "%i" $TIDHEX`
>> + if test "$TID" -eq "$TID2"; then
>> + FOUND="y"
>> + break
>> + fi
>> + done
>> + if test -z "$FOUND"; then
>> + MISSING="$MISSING $TID"
>> + fi
>> + done
>> + if test -n "$MISSING"; then
>> + err "Thread IDs $MISSING not found in perf AUX data"
>> + fi
>> +}
>> --
>> 2.32.0
>>
Hi Carsten,
Sorry for late response.
On Thu, Apr 21, 2022 at 06:38:33PM +0100, Carsten Haitzler wrote:
[...]
> > Very big change... Why squash all patches form previous verion to this
> > single one big patch? Usually the format with small patches is much
> > better for reviewing.
>
> I was asked to re-jig the tree and in doing so I also ended up cutting down
> the size a lot so this just makes more sense together as a "here are the
> tests" as adding infra without any tests makes no sense and the tests
> themelves are self-contained in their own directories and source files and
> "drivign scripts" thus it's essentially patch 1 appended to patch 2 to patch
> 3 etc. and still broken up in the patch file by file.
I am not sure if I understand the meaning, seems to me you could
organize the patch series like:
- Patch for common files (e.g. script lib/coresight.sh or some
Makefile changes);
- Patches for enabling test cases, E.g.:
patch for asm_pure_loop;
patch for thread loop (include unroll loop);
patch for memcpy;
- Patch for documentation.
If this is not comfortable for you, at least we can use three patches:
- Patch for common file (e.g. script lib/coresight.sh);
- Patch for test cases;
- Patch for documentation.
[...]
> > > +If you see these above, then your system is tracing coresight data
> > > +correctly.
> > > +
> > > +To compile perf with coresight support in the perf directory do
> > > +
> > > + make CORESIGHT=1
> >
> > It is inaccurate that if we don't mention openCSD lib.
>
> Do you mean I need to mention that you need the opencsd library installed
> too?
Yes, otherwise, users might directly build perf without opencsd lib,
then finally they cannot use perf with Arm CoreSight.
> > > +This will compile the perf tool with coresight support as well as
> > > +build some small test binaries for perf test. This requires you also
> > > +be compiling for 64bit Arm (ARM64/aarch64). The tools run as part of
> > > +perf coresight tracing are in tests/shell/tools/coresight.
> >
> > For build perf tool, I think above paragraphs are duplicate with the
> > document Documentation/trace/coresight/coresight.rst. Can we simply
> > say:
> >
> > "The details for building perf tool with support Arm Coresight can be
> > found in the "HOWTO.md" file of the openCSD gitHub repository:
> > https://github.com/Linaro/opencsd.
>
> I can. I put this here as I didn't go clone OpencCSD first but used my
> distro OpenCSD packages and thus of course didn't have the documentation in
> front of me. I spent some time wondering why it wasn't building with
> coresight support even though it detected OpenCSD when I compiled... I
> didn't expect to have to go to some separate project git repository and read
> docs there on how to build the perf tool here in the kernel. I wrote this
> because it was an actual problem I hit and it's a lot less frustrating to
> "end users" to give them the information they need in the relevant place
> they need it instead of sending them around to other project trees. Building
> perf with coresight support is handled by the perf tree int he kernel, not
> OpenCSD, thus IMHO that is where the documentation belongs - alongside the
> thing that determines how to build something.
Understand.
> > And "HOWTO.md" file gives the information and examples for how to use
> > perf tool to record and report Coresight trace data. It's the
> > prerequisite for this perf Coresight test."
> >
> > > +You will also want coresight support enabled in your kernel config.
> > > +Ensure it is enabled with:
> > > +
> > > + CONFIG_CORESIGHT=y
> > > +
> > > +There are various other coresight options you probably also want
> > > +enabled like:
> > > +
> > > + CONFIG_CORESIGHT_LINKS_AND_SINKS=y
> > > + CONFIG_CORESIGHT_LINK_AND_SINK_TMC=y
> > > + CONFIG_CORESIGHT_CATU=y
> > > + CONFIG_CORESIGHT_SINK_TPIU=y
> > > + CONFIG_CORESIGHT_SINK_ETBV10=y
> > > + CONFIG_CORESIGHT_SOURCE_ETM4X=y
> > > + CONFIG_CORESIGHT_STM=y
> > > + CONFIG_CORESIGHT_CPU_DEBUG=y
> > > + CONFIG_CORESIGHT_CTI=y
> > > + CONFIG_CORESIGHT_CTI_INTEGRATION_REGS=y
> > > +
> > > +Please refer to the kernel configuration help for more information.
> >
> > I prefer to remove these kernel configuration since they are not
> > inconsistent on different platforms (e.g. ETBV10, ETM4X, etc), and
> > some configurations might not necessary (e.g. CPU_DEBUG).
>
> Certainly there should be some documentation on which kernel configs you
> might want to turn on then? Imagine someone new comes along and doesn't have
> any idea what to possible enable at all and manages to build perf with
> coresight support (as above) then finds it doesn't work because they didn't
> enable enough config in the kernel? Sure - could probably trim these down a
> bit but the point here is to alert the user to there being a range of
> coresight config options that you need to turn on that you likely will find
> are not turned on. They certainly are not turned on on distro kernels and a
> lot of the time when you have a platform that already boots/works you start
> with your distro kernel config file because you want everything enabled so
> it actually boots. I've learned the hard way to do this as you manage to
> forget to turn on some MMC driver or some other feature and your boot hangs
> or doesn't find rootfs etc.
So far, we will have two documents in Linux kernel:
- Documentation/trace/coresight/coresight.rst;
- tools/perf/Documentation/arm-coresight.txt.
We need to avoid overlap between these two files. I think we could use
the file Documentation/trace/coresight/coresight.rst to focus on
CoreSight driver module relates stuffs and
tools/perf/Documentation/arm-coresight.txt is more about the perf
usages.
But, the file Documentation/trace/coresight/coresight.rst doesn't give
any info for kernel configs, I think which would be a better place to
give information for building kernel modules.
> What would you recommend then as a "turn these on and coresight will almost
> certainly work for you on your given hardware " then?
This would be fine. Alternatively, we could add a section in the file
Documentation/trace/coresight/coresight.rst to describe how to build
CoreSight modules.
How you think for this? I also would like to get suggestions from
CoreSight maintainers Suzuki/Mathieu/Mike.
[...]
> > Please update based on the latest test case names, at my side, I can
> > see the testing case like:
> >
> > Coresight / ASM Pure Loop
> > Coresight / Memcpy 16k 10 Threads
> > Coresight / Thread Loop 10 Threads - Check TID
> > Coresight / Thread Loop 2 Threads - Check TID
> > Coresight / Unroll Loop Thread 10
>
> Oh sorry - yeah. I wrote the docs based on the earlier tests. Will fix.
Thanks.
> > > +
> > > +These perf record tests will not run if the tool binaries do not exist
> > > +in tests/shell/tools/coresight/*/ and will be skipped. If you do not
> > > +have coresight support in hardware then either do not build perf with
> > > +coresight support or remove these binaries in order to not have these
> > > +tests fail and have them skip instead.
> > > +
> > > +These tests will log historical results in the current working
> > > +directory (e.g. tools/perf) and will be named stats-*.csv like:
> > > +
> > > + stats-asm_pure_loop-out.csv
> > > + stats-bubble_sort-random.csv
> > > + ...
> > > +
> > > +These statistic files log some aspects of the AUX data sections in
> > > +the perf data output counting some numbers of certain encodings (a
> > > +good way to know that it's working in a very simple way). One problem
> > > +with coresight is that given a large enough amount of data needing to
> > > +be logged, some of it can be lost due to the processor not waking up
> > > +in time to read out all the data from buffers etc.. You will notice
> > > +that the amount of data collected can vary a lot per run of perf test.
> > > +If you wish to see how this changes over time, simply run perf test
> > > +multiple times and all these csv files will have more and more data
> > > +appended to it that you can later examine, graph and otherwise use to
> > > +figure out if things have become worse or better.
> >
> > I am confused by this narrative. Does it try to remind that the final
> > testing result (pass or fail) is not stable? Or should we run for
> > multiple times so have more chance to capture issues?
>
> That is correct. I thought I was clear that it's lossy. That is actually the
> case. I have tests here that actually fail because there is no data
> collected from some threads at all (missing CID blocks for some of the
> threads that run in the test). The point is to have tests that may be
> failing now but in future will improve. I lowered the minimum bar to pass
> for most tests to have "at least just a little data" but most tests show
> highly variable amount of captured data. the csv files are there to
> over-time give you a good idea of the stability of the captured data.
Okay, this would be fine for me. Though I am a bit worry that later if
users report a failure, then how we can tell them this is a bug or it's
just tracing quality issue?
[...]
> > > diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
> > > new file mode 100644
> > > index 000000000000..468673ac32e8
> > > --- /dev/null
> > > +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
> > > @@ -0,0 +1 @@
> > > +asm_pure_loop
> >
> > Do we really need there '.gitignore' files under the folder
> > 'tools/perf/tests/shell/coresight/'.
>
> Where would you rather have them to ignore the generated binary tools?
It's interesting that I wanted to find a case to object you, so I tried
to check the folder linux/samples/bpf, but it does use .gitignore file
to ignore built binaries :)
Adding .gitignore is in practice and this would be fine for me.
> > > diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
> > > new file mode 100644
> > > index 000000000000..10c5a60cb71c
> > > --- /dev/null
> > > +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
> > > @@ -0,0 +1,30 @@
> > > +# SPDX-License-Identifier: GPL-2.0
> > > +# Carsten Haitzler <[email protected]>, 2021
> > > +
> > > +include ../Makefile.miniconfig
> > > +
> > > +BIN=asm_pure_loop
> > > +LIB=
> >
> > Remove the unused variable 'LIB='.
>
> I have this because I wanted to have a simple template to be able to re-use
> for more tests over time. It's so much easier to maintain and extend if
> every makefile and tool follow a similar pattern and you can almost copy &
> paste between them as they don't have "exceptions". You really want me to
> remove this?
It's fine to keep it. Could you add a comment for this?
To be honest, I am not experienced for bash shell script, so I have no
idea why write like this way. If you think this is very common usage
in shell, then you could keep it and don't need to add comment.
[...]
> > There have four sub folders under tools/perf/tests/shell/coresight:
> >
> > asm_pure_loop
> > memcpy_thread
> > thread_loop
> > unroll_loop_thread
> >
> > And every folder has its own Makefile and every Makefile is quite
> > close to each other. I am just wandering if it's possible to
> > remove the 4 Makefiles in these four sub folders, and simply use
> > tools/perf/tests/shell/coresight/Makefile as the central place to
> > build these assistant programs.
>
> I did this so it's easier to etxent over time. having a single parent
> makefile that over time accumulates little ugly "if's" and exceptions makes
> longer-term maintenance and extending harder. I did it this way to make this
> easy - make a copy of a dir - add that dir to a parent makefile then modify
> the makefile as needed (but only as needed).
Okay, let's keep the saperate makefiles.
> > > diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
> > > new file mode 100644
> > > index 000000000000..75cf084a927d
> > > --- /dev/null
> > > +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
> > > @@ -0,0 +1,28 @@
> > > +/* SPDX-License-Identifier: GPL-2.0 */
> > > +/* Tamas Zsoldos <[email protected]>, 2021 */
> > > +
> > > +.globl _start
> > > +_start:
> > > + mov x0, 0x0000ffff
> > > + mov x1, xzr
> > > +loop:
> > > + nop
> > > + nop
> > > + cbnz x1, noskip
> > > + nop
> > > + nop
> > > + adrp x2, skip
> > > + add x2, x2, :lo12:skip
> > > + br x2
> > > + nop
> > > + nop
> > > +noskip:
> > > + nop
> > > + nop
> > > +skip:
> > > + sub x0, x0, 1
> > > + cbnz x0, loop
> > > +
> > > + mov x0, #0
> > > + mov x8, #93 // __NR_exit syscall
> > > + svc #0
> >
> > I tested the case "ASM Pure Loop" on my Juno board, and it complaints:
> >
> > root@debian:/mnt/export/arm-linux-kernel/tools/perf# ./perf test -v 76
> > 76: Coresight / ASM Pure Loop :
> > --- start ---
> > test child forked, pid 9063
> > failed to mmap with 12 (Cannot allocate memory)
> > test child finished with -1
> > ---- end ----
> > Coresight / ASM Pure Loop: FAILED!
> >
> > Since I only setup the 1GB memory for the Linux kernel, it fails to
> > allocate AUX ring buffer with the size 256MB. So I manully change
> > the buffer size to 8MB in tools/perf/tests/shell/lib/coresight.sh:
> >
> > PERFRECMEM="-m ,8M"
> >
> > So finally I can see the test case is passed:
>
> This is artificial isn't it? limiting to 1GB. You certainly have far more
> memory than that available. My testse were on a system with 4GB and I had no
> issues.
Please see below comment.
> > root@debian:/mnt/export/arm-linux-kernel/tools/perf# ./perf test -v 76
> > 76: Coresight / ASM Pure Loop :
> > --- start ---
> > test child forked, pid 9481
> > -m ,8M -e cs_etm//u
> > [ perf record: Woken up 1 times to write data ]
> > [ perf record: Captured and wrote 0.681 MB ./perf-asm_pure_loop-out.data ]
> > test child finished with 0
> > ---- end ----
> > Coresight / ASM Pure Loop: Ok
> >
> > Do you think we really need to use 256MiB as the AUX buffer size?
> > IIRC, it means we allocate 256MiB per CPU for this case, on the other
> > hand, you could see the final perf data file size is small (0.681
> > MiB).
> >
> > Seems to me, it's not necessary to allocate so big buffer for
> > the test, and I tried to run below 4 cases with 8MiB, all of them can
> > pass the testing :)
>
> I didn't think anyone with a system with coresight support that would be
> running perf record locally would only have 1GB of ram... I knew junos had
> 8GB and my dragonboard has 4GB ... so I know I was on the smaller side. I
> thought a larger buffer == safer results (less chance of needing to write
> out the buffer during capture). Admittdly I used 256Mb when my tests ran for
> much longer and collected more data. I can try drop to 8 or 16gb and see.
Yes, my Juno board has 8GB but I also have DB410c with 1GB with quad
coes [1]. I am still concern for 256MB buffer size, it's not friendly for
embedded system, and even not good for server. For example, if we run
this testing on Arm server with 96 cores (like Hisilicon D06 board),
then we need the buffer size is:
256MiB * 96 = 16GiB
I agree usually 16GiB is not a problem for server, but seems to me
it doesn't make much sense to consume huge memory resource for the
testing.
In other words, if set 8MiB (or 16MiB, 32MiB) buffer size and doesn't
see testing result regression, I think this would be good to decrease
the buffer size.
[1] https://www.96boards.org/product/dragonboard410c/
[...]
> > > diff --git a/tools/perf/tests/shell/coresight_asm_pure_loop.sh b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
> > > new file mode 100755
> > > index 000000000000..3f0dbefcad50
> > > --- /dev/null
> > > +++ b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
> > > @@ -0,0 +1,18 @@
> > > +#!/bin/sh -e
> > > +# Coresight / ASM Pure Loop
> > > +
> > > +# SPDX-License-Identifier: GPL-2.0
> > > +# Carsten Haitzler <[email protected]>, 2021
> > > +
> > > +TEST="asm_pure_loop"
> > > +. $(dirname $0)/lib/coresight.sh
> > > +ARGS=""
> > > +DATV="out"
> > > +DATA="$DATD/perf-$TEST-$DATV.data"
> > > +
> > > +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
> > > +
> > > +perf_dump_aux_verify "$DATA" 10 10 10
> > > +
> > > +err=$?
> > > +exit $err
> >
> > Can we organize the shell scripts by moving them into the folder
> > tools/perf/tests/shell/coresight?
>
> We can - but it comes with a fair few more changes.
>
> > coresight_asm_pure_loop.sh
> > coresight_memcpy_thread_16k_10.sh
> > coresight_thread_loop_check_tid_10.sh
> > coresight_thread_loop_check_tid_2.sh
> > coresight_unroll_loop_thread_10.sh
> >
> > And we even can consider to move script test_arm_coresight.sh into
> > the folder tools/perf/tests/shell/coresight and change its
> > name as 'coresight_smoke_test.sh'.
>
> Indeed these other tests I left alone for now and had not thought about how
> to marry these together yet - leaving this for another day and another patch
> set rather than this patch set itself. That was my thoguht. I was trying to
> make an "Easier to extend by just dropping a test into a dir" setup here to
> make maintenance and expansion easier over time (and thus encourage testing
> by having a simple repeatable test infra to duplicate). I ended up with a
> dir per test tool you need to build and a driver script in the tests/shell
> dir. I think this is certainly worth considering but perhaps as a separate
> set of work to marry these?
Okay, it would be fine to use separate set for moving the script
test_arm_coresight.sh, which is a simple case.
> I piggybacked on the existing shell test infra but added a fair few more
> scripts. To do what you suggest I'd need to modify the core shell test code
> to walk subdirs recursively then looking for child scripts. The problem is
> how does perf test's shell handling know about the coresight subdir vs the
> lib subdir?
Yeah, now I understand your point. How about file layout like below?
tools/perf/tests/shell/coresight_test_hub.sh
tools/perf/tests/shell/coresight/coresight_asm_pure_loop.sh
tools/perf/tests/shell/coresight/coresight_memcpy_thread_16k_10.sh
tools/perf/tests/shell/coresight/coresight_thread_loop_check_tid_10.sh
tools/perf/tests/shell/coresight/coresight_thread_loop_check_tid_2.sh
tools/perf/tests/shell/coresight/coresight_unroll_loop_thread_10.sh
So we use tools/perf/tests/shell/coresight_test_hub.sh as an interface
to hook with Perf test infrastructure, and then hub.sh file calls
testing scripts under the sub folder. Seems to me, this is also
friendly for later's extension.
> Both contain *.sh shell scripts - the difference is the ones in
> lib are not executable. Is this sufficiently different? I could also open
> them to check the have #!/bin/... as the first line. Hardcoding just a
> single coresight subdir just feels wrong and hacky to me, thus the generic
> recursion solution I suggest here.
Agreed. I also don't prefer this way.
> I can definitely see how extending to subdirs would make supporting testing
> cleaner and divide things into their own domains (dirs).
Thanks a lot for the work! The test cases are good for me (but I would
say Mike is the best person for reviewing testing trace data quaility),
I just want to make sure it's not hard for later maintenance.
Leo
On Thu, May 26, 2022 at 04:20:39PM +0800, Leo Yan wrote:
[...]
> Yes, my Juno board has 8GB but I also have DB410c with 1GB with quad
> coes [1]. I am still concern for 256MB buffer size, it's not friendly for
> embedded system, and even not good for server. For example, if we run
> this testing on Arm server with 96 cores (like Hisilicon D06 board),
> then we need the buffer size is:
>
> 256MiB * 96 = 16GiB
Correct for my bad math... here should be 256MiB * 96 = 24GiB.
On Wed, Mar 09, 2022 at 12:28:59PM +0000, [email protected] wrote:
> From: Carsten Haitzler <[email protected]>
>
> This adds a test harness and tests to run perf record and examine the
> resuling output when coresight is enabled on arm64 and check the
> resulting quality of the output as part of perf test. These tests use
> various tools to produce output from perf record then measure some key
> specific aspects of that data to see if the data exists at all and
> contains key aspects such as measuring some data for every thread of
> a test or produces sufficient data for large exeuction runs of a large
> executable. etc.
>
> Signed-off-by: Carsten Haitzler <[email protected]>
> ---
> MAINTAINERS | 4 +
> tools/perf/.gitignore | 6 +-
> tools/perf/Documentation/arm-coresight.txt | 140 ++++++++++++++++++
> tools/perf/Makefile.perf | 14 +-
> tools/perf/tests/shell/coresight/Makefile | 30 ++++
> .../tests/shell/coresight/Makefile.miniconfig | 23 +++
> .../shell/coresight/asm_pure_loop/.gitignore | 1 +
> .../shell/coresight/asm_pure_loop/Makefile | 30 ++++
> .../coresight/asm_pure_loop/asm_pure_loop.S | 28 ++++
> .../shell/coresight/memcpy_thread/.gitignore | 1 +
> .../shell/coresight/memcpy_thread/Makefile | 29 ++++
> .../coresight/memcpy_thread/memcpy_thread.c | 79 ++++++++++
> .../shell/coresight/thread_loop/.gitignore | 1 +
> .../shell/coresight/thread_loop/Makefile | 29 ++++
> .../shell/coresight/thread_loop/thread_loop.c | 86 +++++++++++
> .../coresight/unroll_loop_thread/.gitignore | 1 +
> .../coresight/unroll_loop_thread/Makefile | 29 ++++
> .../unroll_loop_thread/unroll_loop_thread.c | 74 +++++++++
> .../tests/shell/coresight_asm_pure_loop.sh | 18 +++
> .../shell/coresight_memcpy_thread_16k_10.sh | 18 +++
> .../coresight_thread_loop_check_tid_10.sh | 19 +++
> .../coresight_thread_loop_check_tid_2.sh | 19 +++
> .../shell/coresight_unroll_loop_thread_10.sh | 18 +++
> tools/perf/tests/shell/lib/coresight.sh | 130 ++++++++++++++++
> 24 files changed, 823 insertions(+), 4 deletions(-)
As Leo pointed out this is a big patch and hard to digest intellectually.
> create mode 100644 tools/perf/Documentation/arm-coresight.txt
> create mode 100644 tools/perf/tests/shell/coresight/Makefile
> create mode 100644 tools/perf/tests/shell/coresight/Makefile.miniconfig
> create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
> create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
> create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
> create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
> create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/Makefile
> create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
> create mode 100644 tools/perf/tests/shell/coresight/thread_loop/.gitignore
> create mode 100644 tools/perf/tests/shell/coresight/thread_loop/Makefile
> create mode 100644 tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
> create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
> create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
> create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
> create mode 100755 tools/perf/tests/shell/coresight_asm_pure_loop.sh
> create mode 100755 tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
> create mode 100755 tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
> create mode 100755 tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
> create mode 100755 tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
> create mode 100644 tools/perf/tests/shell/lib/coresight.sh
>
> diff --git a/MAINTAINERS b/MAINTAINERS
> index 673c7124ca82..18cc20609f2e 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -1918,10 +1918,14 @@ F: drivers/hwtracing/coresight/*
> F: include/dt-bindings/arm/coresight-cti-dt.h
> F: include/linux/coresight*
> F: samples/coresight/*
> +F: tools/perf/Documentation/arm-coresight.txt
> F: tools/perf/arch/arm/util/auxtrace.c
> F: tools/perf/arch/arm/util/cs-etm.c
> F: tools/perf/arch/arm/util/cs-etm.h
> F: tools/perf/arch/arm/util/pmu.c
> +F: tools/perf/tests/shell/coresight_*
> +F: tools/perf/tests/shell/tools/Makefile
> +F: tools/perf/tests/shell/tools/coresight/*
> F: tools/perf/util/cs-etm-decoder/*
> F: tools/perf/util/cs-etm.*
>
> diff --git a/tools/perf/.gitignore b/tools/perf/.gitignore
> index 20b8ab984d5f..138c679ecacd 100644
> --- a/tools/perf/.gitignore
> +++ b/tools/perf/.gitignore
> @@ -15,8 +15,9 @@ perf*.1
> perf*.xml
> perf*.html
> common-cmds.h
> -perf.data
> -perf.data.old
> +perf*.data
> +perf*.data.old
> +stats-*.csv
> output.svg
> perf-archive
> perf-with-kcore
> @@ -30,6 +31,7 @@ config.mak.autogen
> *-flex.*
> *.pyc
> *.pyo
> +*.stdout
> .config-detected
> util/intel-pt-decoder/inat-tables.c
> arch/*/include/generated/
> diff --git a/tools/perf/Documentation/arm-coresight.txt b/tools/perf/Documentation/arm-coresight.txt
> new file mode 100644
> index 000000000000..3a9e6c573c58
> --- /dev/null
> +++ b/tools/perf/Documentation/arm-coresight.txt
I think it would be best to keep all the coresight documentation under the
current coresight documentation repository[1]. That way all the information on
coresight can be found in a central place.
Some part of what is added by this patch is redundant with what is currently
available in [1]. Other parts are tests specific and should be added under
something like "coresight-perf-test.rst".
Thanks,
Mathieu
[1]. Documentation/trace/coresight/
> @@ -0,0 +1,140 @@
> +Arm Coresight Support
> +=====================
> +
> +Coresight is a feature of some Arm based processors that allows for
> +debugging. One of the things it can do is trace every instruction
> +executed and remotely expose that information in a hardware compressed
> +stream. Perf is able to locally access that stream and store it to the
> +output perf data files. This stream can then be later decoded to give the
> +instructions that were traced for debugging or profiling purposes. You
> +can log such data with a perf record command like:
> +
> + perf record -e cs_etm//u testbinary
> +
> +This would run some test binary (testbinary) until it exits and record
> +a perf.data trace file. That file would have AUX sections if coresight
> +is working correctly. You can dump the content of this file as
> +readable text with a command like:
> +
> + perf report --stdio --dump -i perf.data
> +
> +You should find some sections of this file have AUX data blocks like:
> +
> + 0x1e78 [0x30]: PERF_RECORD_AUXTRACE size: 0x11dd0 offset: 0 ref: 0x1b614fc1061b0ad1 idx: 0 tid: 531230 cpu: -1
> +
> + . ... CoreSight ETM Trace data: size 73168 bytes
> + Idx:0; ID:10; I_ASYNC : Alignment Synchronisation.
> + Idx:12; ID:10; I_TRACE_INFO : Trace Info.; INFO=0x0 { CC.0 }
> + Idx:17; ID:10; I_ADDR_L_64IS0 : Address, Long, 64 bit, IS0.; Addr=0x0000000000000000;
> + Idx:26; ID:10; I_TRACE_ON : Trace On.
> + Idx:27; ID:10; I_ADDR_CTXT_L_64IS0 : Address & Context, Long, 64 bit, IS0.; Addr=0x0000FFFFB6069140; Ctxt: AArch64,EL0, NS;
> + Idx:38; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
> + Idx:39; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
> + Idx:40; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
> + Idx:41; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEN
> + ...
> +
> +If you see these above, then your system is tracing coresight data
> +correctly.
> +
> +To compile perf with coresight support in the perf directory do
> +
> + make CORESIGHT=1
> +
> +This will compile the perf tool with coresight support as well as
> +build some small test binaries for perf test. This requires you also
> +be compiling for 64bit Arm (ARM64/aarch64). The tools run as part of
> +perf coresight tracing are in tests/shell/tools/coresight.
> +
> +You will also want coresight support enabled in your kernel config.
> +Ensure it is enabled with:
> +
> + CONFIG_CORESIGHT=y
> +
> +There are various other coresight options you probably also want
> +enabled like:
> +
> + CONFIG_CORESIGHT_LINKS_AND_SINKS=y
> + CONFIG_CORESIGHT_LINK_AND_SINK_TMC=y
> + CONFIG_CORESIGHT_CATU=y
> + CONFIG_CORESIGHT_SINK_TPIU=y
> + CONFIG_CORESIGHT_SINK_ETBV10=y
> + CONFIG_CORESIGHT_SOURCE_ETM4X=y
> + CONFIG_CORESIGHT_STM=y
> + CONFIG_CORESIGHT_CPU_DEBUG=y
> + CONFIG_CORESIGHT_CTI=y
> + CONFIG_CORESIGHT_CTI_INTEGRATION_REGS=y
> +
> +Please refer to the kernel configuration help for more information.
> +
> +Perf test - Verify kernel and userspace perf coresight work
> +===========================================================
> +
> +When you run perf test, it will do a lot of self tests. Some of those
> +tests will cover Coresight (only if enabled and on ARM64). You
> +generally would run perf test from the tools/perf directory in the
> +kernel tree. Some tests will check some internal perf support like:
> +
> + Check Arm CoreSight trace data recording and synthesized samples
> +
> +Some others will actually use perf record and some test binaries that
> +are in tests/shell/tools/coresight and will collect traces to ensure a
> +minimum level of functionality is met. The scripts that launch these
> +tests are in tests/shell. These will all look like:
> +
> + Coresight / Memcpy 1M 25 Threads
> + Coresight / Unroll Loop Thread 2
> + ...
> +
> +These perf record tests will not run if the tool binaries do not exist
> +in tests/shell/tools/coresight/*/ and will be skipped. If you do not
> +have coresight support in hardware then either do not build perf with
> +coresight support or remove these binaries in order to not have these
> +tests fail and have them skip instead.
> +
> +These tests will log historical results in the current working
> +directory (e.g. tools/perf) and will be named stats-*.csv like:
> +
> + stats-asm_pure_loop-out.csv
> + stats-bubble_sort-random.csv
> + ...
> +
> +These statistic files log some aspects of the AUX data sections in
> +the perf data output counting some numbers of certain encodings (a
> +good way to know that it's working in a very simple way). One problem
> +with coresight is that given a large enough amount of data needing to
> +be logged, some of it can be lost due to the processor not waking up
> +in time to read out all the data from buffers etc.. You will notice
> +that the amount of data collected can vary a lot per run of perf test.
> +If you wish to see how this changes over time, simply run perf test
> +multiple times and all these csv files will have more and more data
> +appended to it that you can later examine, graph and otherwise use to
> +figure out if things have become worse or better.
> +
> +Be aware that amny of these tests take quite a while to run, specifically
> +in processing the perf data file and dumping contents to then examine what
> +is inside.
> +
> +You can change where these csv logs are stored by setting the
> +PERF_TEST_CORESIGHT_STATDIR environment variable before running perf
> +test like:
> +
> + export PERF_TEST_CORESIGHT_STATDIR=/var/tmp
> + perf test
> +
> +They will also store resulting perf output data in the current
> +directory for later inspection like:
> +
> + perf-memcpy-1m.data
> + perf-thread_loop-2th.data
> + ...
> +
> +You can alter where the perf data files are stored by setting the
> +PERF_TEST_CORESIGHT_DATADIR environment variable such as:
> +
> + PERF_TEST_CORESIGHT_DATADIR=/var/tmp
> + perf test
> +
> +You may wish to set these above environment variables if you which to
> +keep the output of tests outside of the current working directory for
> +longer term storage and examination.
> diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
> index ac861e42c8f7..b97db83992e0 100644
> --- a/tools/perf/Makefile.perf
> +++ b/tools/perf/Makefile.perf
> @@ -630,7 +630,15 @@ sync_file_range_tbls := $(srctree)/tools/perf/trace/beauty/sync_file_range.sh
> $(sync_file_range_arrays): $(linux_uapi_dir)/fs.h $(sync_file_range_tbls)
> $(Q)$(SHELL) '$(sync_file_range_tbls)' $(linux_uapi_dir) > $@
>
> -all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS)
> +TESTS_CORESIGHT_DIR := $(srctree)/tools/perf/tests/shell/coresight
> +
> +tests-coresight-targets: FORCE
> + $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR)
> +
> +tests-coresight-targets-clean:
> + $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR) clean
> +
> +all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS) tests-coresight-targets
>
> # Create python binding output directory if not already present
> _dummy := $(shell [ -d '$(OUTPUT)python' ] || mkdir -p '$(OUTPUT)python')
> @@ -1020,6 +1028,7 @@ install-tests: all install-gtk
> $(INSTALL) tests/shell/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell'; \
> $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'; \
> $(INSTALL) tests/shell/lib/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'
> + $(Q)$(MAKE) -C tests/shell/coresight install-tests
>
> install-bin: install-tools install-tests install-traceevent-plugins
>
> @@ -1088,7 +1097,7 @@ endif # BUILD_BPF_SKEL
> bpf-skel-clean:
> $(call QUIET_CLEAN, bpf-skel) $(RM) -r $(SKEL_TMP_OUT) $(SKELETONS)
>
> -clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean
> +clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean tests-coresight-targets-clean
> $(call QUIET_CLEAN, core-objs) $(RM) $(LIBPERF_A) $(OUTPUT)perf-archive $(OUTPUT)perf-with-kcore $(OUTPUT)perf-iostat $(LANG_BINDINGS)
> $(Q)find $(if $(OUTPUT),$(OUTPUT),.) -name '*.o' -delete -o -name '\.*.cmd' -delete -o -name '\.*.d' -delete
> $(Q)$(RM) $(OUTPUT).config-detected
> @@ -1155,5 +1164,6 @@ FORCE:
> .PHONY: shell_compatibility_test please_set_SHELL_PATH_to_a_more_modern_shell
> .PHONY: $(GIT-HEAD-PHONY) TAGS tags cscope FORCE prepare
> .PHONY: libtraceevent_plugins archheaders
> +.PHONY: $(TESTS_CORESIGHT_TARGETS)
>
> endif # force_fixdep
> diff --git a/tools/perf/tests/shell/coresight/Makefile b/tools/perf/tests/shell/coresight/Makefile
> new file mode 100644
> index 000000000000..dda99aeac158
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/Makefile
> @@ -0,0 +1,30 @@
> +# SPDX-License-Identifier: GPL-2.0-only
> +# Carsten Haitzler <[email protected]>, 2021
> +include ../../../../../tools/scripts/Makefile.include
> +include ../../../../../tools/scripts/Makefile.arch
> +include ../../../../../tools/scripts/utilities.mak
> +
> +SUBDIRS = \
> + asm_pure_loop \
> + thread_loop \
> + memcpy_thread \
> + unroll_loop_thread
> +
> +all: $(SUBDIRS)
> +$(SUBDIRS):
> + $(Q)$(MAKE) -C $@
> +
> +INSTALLDIRS = $(SUBDIRS:%=install-%)
> +
> +install-tests: $(INSTALLDIRS)
> +$(INSTALLDIRS):
> + $(Q)$(MAKE) -C $(@:install-%=%) install-tests
> +
> +CLEANDIRS = $(SUBDIRS:%=clean-%)
> +
> +clean: $(CLEANDIRS)
> +$(CLEANDIRS):
> + $(Q)$(MAKE) -C $(@:clean-%=%) clean >/dev/null
> +
> +.PHONY: all clean $(SUBDIRS) $(CLEANDIRS) $(INSTALLDIRS)
> +
> diff --git a/tools/perf/tests/shell/coresight/Makefile.miniconfig b/tools/perf/tests/shell/coresight/Makefile.miniconfig
> new file mode 100644
> index 000000000000..893c12685fed
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/Makefile.miniconfig
> @@ -0,0 +1,23 @@
> +# SPDX-License-Identifier: GPL-2.0-only
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +ifndef DESTDIR
> +prefix ?= $(HOME)
> +endif
> +
> +DESTDIR_SQ = $(subst ','\'',$(DESTDIR))
> +perfexecdir = libexec/perf-core
> +perfexec_instdir = $(perfexecdir)
> +
> +ifneq ($(filter /%,$(firstword $(perfexecdir))),)
> +perfexec_instdir = $(perfexecdir)
> +else
> +perfexec_instdir = $(prefix)/$(perfexecdir)
> +endif
> +
> +perfexec_instdir_SQ = $(subst ','\'',$(perfexec_instdir))
> +INSTALL = install
> +
> +include ../../../../../scripts/Makefile.include
> +include ../../../../../scripts/Makefile.arch
> +include ../../../../../scripts/utilities.mak
> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
> new file mode 100644
> index 000000000000..468673ac32e8
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
> @@ -0,0 +1 @@
> +asm_pure_loop
> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
> new file mode 100644
> index 000000000000..10c5a60cb71c
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
> @@ -0,0 +1,30 @@
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +include ../Makefile.miniconfig
> +
> +BIN=asm_pure_loop
> +LIB=
> +
> +all: $(BIN)
> +
> +$(BIN): $(BIN).S
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(Q)$(CC) $(BIN).S -nostdlib -static -o $(BIN) $(LIB)
> +endif
> +endif
> +
> +install-tests: all
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(call QUIET_INSTALL, tests) \
> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
> +endif
> +endif
> +
> +clean:
> + $(Q)$(RM) -f $(BIN)
> +
> +.PHONY: all clean install-tests
> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
> new file mode 100644
> index 000000000000..75cf084a927d
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
> @@ -0,0 +1,28 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/* Tamas Zsoldos <[email protected]>, 2021 */
> +
> +.globl _start
> +_start:
> + mov x0, 0x0000ffff
> + mov x1, xzr
> +loop:
> + nop
> + nop
> + cbnz x1, noskip
> + nop
> + nop
> + adrp x2, skip
> + add x2, x2, :lo12:skip
> + br x2
> + nop
> + nop
> +noskip:
> + nop
> + nop
> +skip:
> + sub x0, x0, 1
> + cbnz x0, loop
> +
> + mov x0, #0
> + mov x8, #93 // __NR_exit syscall
> + svc #0
> diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
> new file mode 100644
> index 000000000000..f8217e56091e
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
> @@ -0,0 +1 @@
> +memcpy_thread
> diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/Makefile b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile
> new file mode 100644
> index 000000000000..e2604cfae74b
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile
> @@ -0,0 +1,29 @@
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +include ../Makefile.miniconfig
> +
> +BIN=memcpy_thread
> +LIB=-pthread
> +
> +all: $(BIN)
> +
> +$(BIN): $(BIN).c
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
> +endif
> +endif
> +
> +install-tests: all
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(call QUIET_INSTALL, tests) \
> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
> +endif
> +endif
> +
> +clean:
> + $(Q)$(RM) -f $(BIN)
> +
> +.PHONY: all clean install-tests
> diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
> new file mode 100644
> index 000000000000..a7e169d1bf64
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
> @@ -0,0 +1,79 @@
> +// SPDX-License-Identifier: GPL-2.0
> +// Carsten Haitzler <[email protected]>, 2021
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <unistd.h>
> +#include <string.h>
> +#include <pthread.h>
> +
> +struct args {
> + unsigned long loops;
> + unsigned long size;
> + pthread_t th;
> + void *ret;
> +};
> +
> +static void *thrfn(void *arg)
> +{
> + struct args *a = arg;
> + unsigned long i, len = a->loops;
> + unsigned char *src, *dst;
> +
> + src = malloc(a->size * 1024);
> + dst = malloc(a->size * 1024);
> + if ((!src) || (!dst)) {
> + printf("ERR: Can't allocate memory\n");
> + exit(1);
> + }
> + for (i = 0; i < len; i++)
> + memcpy(dst, src, a->size * 1024);
> +}
> +
> +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
> +{
> + pthread_t t;
> + pthread_attr_t attr;
> +
> + pthread_attr_init(&attr);
> + pthread_create(&t, &attr, fn, arg);
> + return t;
> +}
> +
> +int main(int argc, char **argv)
> +{
> + unsigned long i, len, size, thr;
> + pthread_t threads[256];
> + struct args args[256];
> + long long v;
> +
> + if (argc < 4) {
> + printf("ERR: %s [copysize Kb] [numthreads] [numloops (hundreds)]\n", argv[0]);
> + exit(1);
> + }
> +
> + v = atoll(argv[1]);
> + if ((v < 1) || (v > (1024 * 1024))) {
> + printf("ERR: max memory 1GB (1048576 KB)\n");
> + exit(1);
> + }
> + size = v;
> + thr = atol(argv[2]);
> + if ((thr < 1) || (thr > 256)) {
> + printf("ERR: threads 1-256\n");
> + exit(1);
> + }
> + v = atoll(argv[3]);
> + if ((v < 1) || (v > 40000000000ll)) {
> + printf("ERR: loops 1-40000000000 (hundreds)\n");
> + exit(1);
> + }
> + len = v * 100;
> + for (i = 0; i < thr; i++) {
> + args[i].loops = len;
> + args[i].size = size;
> + args[i].th = new_thr(thrfn, &(args[i]));
> + }
> + for (i = 0; i < thr; i++)
> + pthread_join(args[i].th, &(args[i].ret));
> + return 0;
> +}
> diff --git a/tools/perf/tests/shell/coresight/thread_loop/.gitignore b/tools/perf/tests/shell/coresight/thread_loop/.gitignore
> new file mode 100644
> index 000000000000..6d4c33eaa9e8
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/thread_loop/.gitignore
> @@ -0,0 +1 @@
> +thread_loop
> diff --git a/tools/perf/tests/shell/coresight/thread_loop/Makefile b/tools/perf/tests/shell/coresight/thread_loop/Makefile
> new file mode 100644
> index 000000000000..424df4e8b0e6
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/thread_loop/Makefile
> @@ -0,0 +1,29 @@
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +include ../Makefile.miniconfig
> +
> +BIN=thread_loop
> +LIB=-pthread
> +
> +all: $(BIN)
> +
> +$(BIN): $(BIN).c
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
> +endif
> +endif
> +
> +install-tests: all
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(call QUIET_INSTALL, tests) \
> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
> +endif
> +endif
> +
> +clean:
> + $(Q)$(RM) -f $(BIN)
> +
> +.PHONY: all clean install-tests
> diff --git a/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
> new file mode 100644
> index 000000000000..c0158fac7d0b
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
> @@ -0,0 +1,86 @@
> +// SPDX-License-Identifier: GPL-2.0
> +// Carsten Haitzler <[email protected]>, 2021
> +
> +// define this for gettid()
> +#define _GNU_SOURCE
> +
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <unistd.h>
> +#include <string.h>
> +#include <pthread.h>
> +#include <sys/syscall.h>
> +#ifndef SYS_gettid
> +// gettid is 178 on arm64
> +# define SYS_gettid 178
> +#endif
> +#define gettid() syscall(SYS_gettid)
> +
> +struct args {
> + unsigned int loops;
> + pthread_t th;
> + void *ret;
> +};
> +
> +static void *thrfn(void *arg)
> +{
> + struct args *a = arg;
> + int i = 0, len = a->loops;
> +
> + if (getenv("SHOW_TID")) {
> + unsigned long long tid = gettid();
> +
> + printf("%llu\n", tid);
> + }
> + asm volatile(
> + "loop:\n"
> + "add %[i], %[i], #1\n"
> + "cmp %[i], %[len]\n"
> + "blt loop\n"
> + : /* out */
> + : /* in */ [i] "r" (i), [len] "r" (len)
> + : /* clobber */
> + );
> + return (void *)(long)i;
> +}
> +
> +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
> +{
> + pthread_t t;
> + pthread_attr_t attr;
> +
> + pthread_attr_init(&attr);
> + pthread_create(&t, &attr, fn, arg);
> + return t;
> +}
> +
> +int main(int argc, char **argv)
> +{
> + unsigned int i, len, thr;
> + pthread_t threads[256];
> + struct args args[256];
> +
> + if (argc < 3) {
> + printf("ERR: %s [numthreads] [numloops (millions)]\n", argv[0]);
> + exit(1);
> + }
> +
> + thr = atoi(argv[1]);
> + if ((thr < 1) || (thr > 256)) {
> + printf("ERR: threads 1-256\n");
> + exit(1);
> + }
> + len = atoi(argv[2]);
> + if ((len < 1) || (len > 4000)) {
> + printf("ERR: max loops 4000 (millions)\n");
> + exit(1);
> + }
> + len *= 1000000;
> + for (i = 0; i < thr; i++) {
> + args[i].loops = len;
> + args[i].th = new_thr(thrfn, &(args[i]));
> + }
> + for (i = 0; i < thr; i++)
> + pthread_join(args[i].th, &(args[i].ret));
> + return 0;
> +}
> diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
> new file mode 100644
> index 000000000000..2cb4e996dbf3
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
> @@ -0,0 +1 @@
> +unroll_loop_thread
> diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
> new file mode 100644
> index 000000000000..45ab2be8be92
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
> @@ -0,0 +1,29 @@
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +include ../Makefile.miniconfig
> +
> +BIN=unroll_loop_thread
> +LIB=-pthread
> +
> +all: $(BIN)
> +
> +$(BIN): $(BIN).c
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
> +endif
> +endif
> +
> +install-tests: all
> +ifdef CORESIGHT
> +ifeq ($(ARCH),arm64)
> + $(call QUIET_INSTALL, tests) \
> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
> +endif
> +endif
> +
> +clean:
> + $(Q)$(RM) -f $(BIN)
> +
> +.PHONY: all clean install-tests
> diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
> new file mode 100644
> index 000000000000..cb9d22c7dfb9
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
> @@ -0,0 +1,74 @@
> +// SPDX-License-Identifier: GPL-2.0
> +// Carsten Haitzler <[email protected]>, 2021
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <unistd.h>
> +#include <string.h>
> +#include <pthread.h>
> +
> +struct args {
> + pthread_t th;
> + unsigned int in, out;
> + void *ret;
> +};
> +
> +static void *thrfn(void *arg)
> +{
> + struct args *a = arg;
> + unsigned int i, in = a->in;
> +
> + for (i = 0; i < 10000; i++) {
> + asm volatile (
> +// force an unroll of thia add instruction so we can test long runs of code
> +#define SNIP1 "add %[in], %[in], #1\n"
> +// 10
> +#define SNIP2 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1
> +// 100
> +#define SNIP3 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2
> +// 1000
> +#define SNIP4 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3
> +// 10000
> +#define SNIP5 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4
> +// 100000
> + SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5
> + : /* out */
> + : /* in */ [in] "r" (in)
> + : /* clobber */
> + );
> + }
> +}
> +
> +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
> +{
> + pthread_t t;
> + pthread_attr_t attr;
> +
> + pthread_attr_init(&attr);
> + pthread_create(&t, &attr, fn, arg);
> + return t;
> +}
> +
> +int main(int argc, char **argv)
> +{
> + unsigned int i, thr;
> + pthread_t threads[256];
> + struct args args[256];
> +
> + if (argc < 2) {
> + printf("ERR: %s [numthreads]\n", argv[0]);
> + exit(1);
> + }
> +
> + thr = atoi(argv[1]);
> + if ((thr > 256) || (thr < 1)) {
> + printf("ERR: threads 1-256\n");
> + exit(1);
> + }
> + for (i = 0; i < thr; i++) {
> + args[i].in = rand();
> + args[i].th = new_thr(thrfn, &(args[i]));
> + }
> + for (i = 0; i < thr; i++)
> + pthread_join(args[i].th, &(args[i].ret));
> + return 0;
> +}
> diff --git a/tools/perf/tests/shell/coresight_asm_pure_loop.sh b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
> new file mode 100755
> index 000000000000..3f0dbefcad50
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
> @@ -0,0 +1,18 @@
> +#!/bin/sh -e
> +# Coresight / ASM Pure Loop
> +
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +TEST="asm_pure_loop"
> +. $(dirname $0)/lib/coresight.sh
> +ARGS=""
> +DATV="out"
> +DATA="$DATD/perf-$TEST-$DATV.data"
> +
> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
> +
> +perf_dump_aux_verify "$DATA" 10 10 10
> +
> +err=$?
> +exit $err
> diff --git a/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh b/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
> new file mode 100755
> index 000000000000..8972af835016
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
> @@ -0,0 +1,18 @@
> +#!/bin/sh -e
> +# Coresight / Memcpy 16k 10 Threads
> +
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +TEST="memcpy_thread"
> +. $(dirname $0)/lib/coresight.sh
> +ARGS="16 10 1"
> +DATV="16k_10"
> +DATA="$DATD/perf-$TEST-$DATV.data"
> +
> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
> +
> +perf_dump_aux_verify "$DATA" 10 10 10
> +
> +err=$?
> +exit $err
> diff --git a/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh b/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
> new file mode 100755
> index 000000000000..5b468901f89b
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
> @@ -0,0 +1,19 @@
> +#!/bin/sh -e
> +# Coresight / Thread Loop 10 Threads - Check TID
> +
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +TEST="thread_loop"
> +. $(dirname $0)/lib/coresight.sh
> +ARGS="10 1"
> +DATV="check-tid-10th"
> +DATA="$DATD/perf-$TEST-$DATV.data"
> +STDO="$DATD/perf-$TEST-$DATV.stdout"
> +
> +SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO
> +
> +perf_dump_aux_tid_verify "$DATA" "$STDO"
> +
> +err=$?
> +exit $err
> diff --git a/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh b/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
> new file mode 100755
> index 000000000000..f8b7abd3aa03
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
> @@ -0,0 +1,19 @@
> +#!/bin/sh -e
> +# Coresight / Thread Loop 2 Threads - Check TID
> +
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +TEST="thread_loop"
> +. $(dirname $0)/lib/coresight.sh
> +ARGS="2 20"
> +DATV="check-tid-2th"
> +DATA="$DATD/perf-$TEST-$DATV.data"
> +STDO="$DATD/perf-$TEST-$DATV.stdout"
> +
> +SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO
> +
> +perf_dump_aux_tid_verify "$DATA" "$STDO"
> +
> +err=$?
> +exit $err
> diff --git a/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh b/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
> new file mode 100755
> index 000000000000..c985dfb025c2
> --- /dev/null
> +++ b/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
> @@ -0,0 +1,18 @@
> +#!/bin/sh -e
> +# Coresight / Unroll Loop Thread 10
> +
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +TEST="unroll_loop_thread"
> +. $(dirname $0)/lib/coresight.sh
> +ARGS="10"
> +DATV="10"
> +DATA="$DATD/perf-$TEST-$DATV.data"
> +
> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
> +
> +perf_dump_aux_verify "$DATA" 10 10 10
> +
> +err=$?
> +exit $err
> diff --git a/tools/perf/tests/shell/lib/coresight.sh b/tools/perf/tests/shell/lib/coresight.sh
> new file mode 100644
> index 000000000000..6a611b073f02
> --- /dev/null
> +++ b/tools/perf/tests/shell/lib/coresight.sh
> @@ -0,0 +1,130 @@
> +# SPDX-License-Identifier: GPL-2.0
> +# Carsten Haitzler <[email protected]>, 2021
> +
> +# This is sourced from a driver script so no need for #!/bin... etc. at the
> +# top - the assumption below is that it runs as part of sourcing after the
> +# test sets up some basic env vars to say what it is.
> +
> +# perf record options for the perf tests to use
> +PERFRECMEM="-m ,128M"
> +PERFRECOPT="$PERFRECMEM -e cs_etm//u"
> +
> +# These tests need to be run as root or coresight won't allow large buffers
> +# and will not collect proper data
> +UID=`id -u`
> +if test "$UID" -ne 0; then
> + echo "Not running as root... skip"
> + exit 2
> +fi
> +
> +TOOLS=$(dirname $0)
> +DIR="$TOOLS/coresight/$TEST"
> +BIN="$DIR/$TEST"
> +# If the test tool/binary does not exist and is executable then skip the test
> +if ! test -x "$BIN"; then exit 2; fi
> +DATD="."
> +# If the data dir env is set then make the data dir use that instead of ./
> +if test -n "$PERF_TEST_CORESIGHT_DATADIR"; then
> + DATD="$PERF_TEST_CORESIGHT_DATADIR";
> +fi
> +# If the stat dir env is set then make the data dir use that instead of ./
> +STATD="."
> +if test -n "$PERF_TEST_CORESIGHT_STATDIR"; then
> + STATD="$PERF_TEST_CORESIGHT_STATDIR";
> +fi
> +
> +# Called if the test fails - error code 2
> +err() {
> + echo "$1"
> + exit 1
> +}
> +
> +# Check that some statistics from our perf
> +check_val_min() {
> + STATF="$4"
> + if test "$2" -lt "$3"; then
> + echo ", FAILED" >> "$STATF"
> + err "Sanity check number of $1 is too low ($2 < $3)"
> + fi
> +}
> +
> +perf_dump_aux_verify() {
> + # Some basic checking that the AUX chunk contains some sensible data
> + # to see that we are recording something and at least a minimum
> + # amount of it. We should almost always see F3 atoms in just about
> + # anything but certainly we will see some trace info and async atom
> + # chunks.
> + DUMP="$DATD/perf-tmp-aux-dump.txt"
> + perf report --stdio --dump -i "$1" | \
> + grep -o -e I_ATOM_F3 -e I_ASYNC -e I_TRACE_INFO > "$DUMP"
> + # Simply count how many of these atoms we find to see that we are
> + # producing a reasonable amount of data - exact checks are not sane
> + # as this is a lossy process where we may lose some blocks and the
> + # compiler may produce different code depending on the compiler and
> + # optimization options, so this is rough just to see if we're
> + # either missing almost all the data or all of it
> + ATOM_F3_NUM=`grep I_ATOM_F3 "$DUMP" | wc -l`
> + ATOM_ASYNC_NUM=`grep I_ASYNC "$DUMP" | wc -l`
> + ATOM_TRACE_INFO_NUM=`grep I_TRACE_INFO "$DUMP" | wc -l`
> + rm -f "$DUMP"
> +
> + # Arguments provide minimums for a pass
> + CHECK_F3_MIN="$2"
> + CHECK_ASYNC_MIN="$3"
> + CHECK_TRACE_INFO_MIN="$4"
> +
> + # Write out statistics, so over time you can track results to see if
> + # there is a pattern - for example we have less "noisy" results that
> + # produce more consistent amounts of data each run, to see if over
> + # time any techinques to minimize data loss are having an effect or
> + # not
> + STATF="$STATD/stats-$TEST-$DATV.csv"
> + if ! test -f "$STATF"; then
> + echo "ATOM F3 Count, Minimum, ATOM ASYNC Count, Minimum, TRACE INFO Count, Minimum" > "$STATF"
> + fi
> + echo -n "$ATOM_F3_NUM, $CHECK_F3_MIN, $ATOM_ASYNC_NUM, $CHECK_ASYNC_MIN, $ATOM_TRACE_INFO_NUM, $CHECK_TRACE_INFO_MIN" >> "$STATF"
> +
> + # Actually check to see if we passed or failed.
> + check_val_min "ATOM_F3" "$ATOM_F3_NUM" "$CHECK_F3_MIN" "$STATF"
> + check_val_min "ASYNC" "$ATOM_ASYNC_NUM" "$CHECK_ASYNC_MIN" "$STATF"
> + check_val_min "TRACE_INFO" "$ATOM_TRACE_INFO_NUM" "$CHECK_TRACE_INFO_MIN" "$STATF"
> + echo ", Ok" >> "$STATF"
> +}
> +
> +perf_dump_aux_tid_verify() {
> + # Specifically crafted test will produce a list of Tread ID's to
> + # stdout that need to be checked to see that they have had trace
> + # info collected in AUX blocks in the perf data. This will go
> + # through all the TID's that are listed as CID=0xabcdef and see
> + # that all the Thread IDs the test tool reports are in the perf
> + # data AUX chunks
> +
> + # The TID test tools will print a TID per stdout line that are being
> + # tested
> + TIDS=`cat "$2"`
> + # Scan the perf report to find the TIDs that are actually CID in hex
> + # and build a list of the ones found
> + FOUND_TIDS=`perf report --stdio --dump -i "$1" | \
> + grep -o "CID=0x[0-9a-z]\+" | sed 's/CID=//g' | \
> + uniq | sort | uniq`
> +
> + # Iterate over the list of TIDs that the test says it has and find
> + # them in the TIDs found in the perf report
> + MISSING=""
> + for TID2 in $TIDS; do
> + FOUND=""
> + for TIDHEX in $FOUND_TIDS; do
> + TID=`printf "%i" $TIDHEX`
> + if test "$TID" -eq "$TID2"; then
> + FOUND="y"
> + break
> + fi
> + done
> + if test -z "$FOUND"; then
> + MISSING="$MISSING $TID"
> + fi
> + done
> + if test -n "$MISSING"; then
> + err "Thread IDs $MISSING not found in perf AUX data"
> + fi
> +}
> --
> 2.32.0
>
On Mon, 30 May 2022 at 10:27, Mathieu Poirier
<[email protected]> wrote:
>
> On Wed, Mar 09, 2022 at 12:28:59PM +0000, [email protected] wrote:
> > From: Carsten Haitzler <[email protected]>
> >
> > This adds a test harness and tests to run perf record and examine the
> > resuling output when coresight is enabled on arm64 and check the
> > resulting quality of the output as part of perf test. These tests use
> > various tools to produce output from perf record then measure some key
> > specific aspects of that data to see if the data exists at all and
> > contains key aspects such as measuring some data for every thread of
> > a test or produces sufficient data for large exeuction runs of a large
> > executable. etc.
> >
> > Signed-off-by: Carsten Haitzler <[email protected]>
> > ---
> > MAINTAINERS | 4 +
> > tools/perf/.gitignore | 6 +-
> > tools/perf/Documentation/arm-coresight.txt | 140 ++++++++++++++++++
> > tools/perf/Makefile.perf | 14 +-
> > tools/perf/tests/shell/coresight/Makefile | 30 ++++
> > .../tests/shell/coresight/Makefile.miniconfig | 23 +++
> > .../shell/coresight/asm_pure_loop/.gitignore | 1 +
> > .../shell/coresight/asm_pure_loop/Makefile | 30 ++++
> > .../coresight/asm_pure_loop/asm_pure_loop.S | 28 ++++
> > .../shell/coresight/memcpy_thread/.gitignore | 1 +
> > .../shell/coresight/memcpy_thread/Makefile | 29 ++++
> > .../coresight/memcpy_thread/memcpy_thread.c | 79 ++++++++++
> > .../shell/coresight/thread_loop/.gitignore | 1 +
> > .../shell/coresight/thread_loop/Makefile | 29 ++++
> > .../shell/coresight/thread_loop/thread_loop.c | 86 +++++++++++
> > .../coresight/unroll_loop_thread/.gitignore | 1 +
> > .../coresight/unroll_loop_thread/Makefile | 29 ++++
> > .../unroll_loop_thread/unroll_loop_thread.c | 74 +++++++++
> > .../tests/shell/coresight_asm_pure_loop.sh | 18 +++
> > .../shell/coresight_memcpy_thread_16k_10.sh | 18 +++
> > .../coresight_thread_loop_check_tid_10.sh | 19 +++
> > .../coresight_thread_loop_check_tid_2.sh | 19 +++
> > .../shell/coresight_unroll_loop_thread_10.sh | 18 +++
> > tools/perf/tests/shell/lib/coresight.sh | 130 ++++++++++++++++
> > 24 files changed, 823 insertions(+), 4 deletions(-)
>
> As Leo pointed out this is a big patch and hard to digest intellectually.
>
> > create mode 100644 tools/perf/Documentation/arm-coresight.txt
> > create mode 100644 tools/perf/tests/shell/coresight/Makefile
> > create mode 100644 tools/perf/tests/shell/coresight/Makefile.miniconfig
> > create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
> > create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
> > create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
> > create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
> > create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/Makefile
> > create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
> > create mode 100644 tools/perf/tests/shell/coresight/thread_loop/.gitignore
> > create mode 100644 tools/perf/tests/shell/coresight/thread_loop/Makefile
> > create mode 100644 tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
> > create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
> > create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
> > create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
> > create mode 100755 tools/perf/tests/shell/coresight_asm_pure_loop.sh
> > create mode 100755 tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
> > create mode 100755 tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
> > create mode 100755 tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
> > create mode 100755 tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
> > create mode 100644 tools/perf/tests/shell/lib/coresight.sh
> >
> > diff --git a/MAINTAINERS b/MAINTAINERS
> > index 673c7124ca82..18cc20609f2e 100644
> > --- a/MAINTAINERS
> > +++ b/MAINTAINERS
> > @@ -1918,10 +1918,14 @@ F: drivers/hwtracing/coresight/*
> > F: include/dt-bindings/arm/coresight-cti-dt.h
> > F: include/linux/coresight*
> > F: samples/coresight/*
> > +F: tools/perf/Documentation/arm-coresight.txt
> > F: tools/perf/arch/arm/util/auxtrace.c
> > F: tools/perf/arch/arm/util/cs-etm.c
> > F: tools/perf/arch/arm/util/cs-etm.h
> > F: tools/perf/arch/arm/util/pmu.c
> > +F: tools/perf/tests/shell/coresight_*
> > +F: tools/perf/tests/shell/tools/Makefile
> > +F: tools/perf/tests/shell/tools/coresight/*
> > F: tools/perf/util/cs-etm-decoder/*
> > F: tools/perf/util/cs-etm.*
> >
> > diff --git a/tools/perf/.gitignore b/tools/perf/.gitignore
> > index 20b8ab984d5f..138c679ecacd 100644
> > --- a/tools/perf/.gitignore
> > +++ b/tools/perf/.gitignore
> > @@ -15,8 +15,9 @@ perf*.1
> > perf*.xml
> > perf*.html
> > common-cmds.h
> > -perf.data
> > -perf.data.old
> > +perf*.data
> > +perf*.data.old
> > +stats-*.csv
> > output.svg
> > perf-archive
> > perf-with-kcore
> > @@ -30,6 +31,7 @@ config.mak.autogen
> > *-flex.*
> > *.pyc
> > *.pyo
> > +*.stdout
> > .config-detected
> > util/intel-pt-decoder/inat-tables.c
> > arch/*/include/generated/
> > diff --git a/tools/perf/Documentation/arm-coresight.txt b/tools/perf/Documentation/arm-coresight.txt
> > new file mode 100644
> > index 000000000000..3a9e6c573c58
> > --- /dev/null
> > +++ b/tools/perf/Documentation/arm-coresight.txt
>
> I think it would be best to keep all the coresight documentation under the
> current coresight documentation repository[1]. That way all the information on
> coresight can be found in a central place.
>
> Some part of what is added by this patch is redundant with what is currently
> available in [1]. Other parts are tests specific and should be added under
> something like "coresight-perf-test.rst".
>
> Thanks,
> Mathieu
>
> [1]. Documentation/trace/coresight/
>
I forgot... Please add a proper cover letter for this patchset.
>
> > @@ -0,0 +1,140 @@
> > +Arm Coresight Support
> > +=====================
> > +
> > +Coresight is a feature of some Arm based processors that allows for
> > +debugging. One of the things it can do is trace every instruction
> > +executed and remotely expose that information in a hardware compressed
> > +stream. Perf is able to locally access that stream and store it to the
> > +output perf data files. This stream can then be later decoded to give the
> > +instructions that were traced for debugging or profiling purposes. You
> > +can log such data with a perf record command like:
> > +
> > + perf record -e cs_etm//u testbinary
> > +
> > +This would run some test binary (testbinary) until it exits and record
> > +a perf.data trace file. That file would have AUX sections if coresight
> > +is working correctly. You can dump the content of this file as
> > +readable text with a command like:
> > +
> > + perf report --stdio --dump -i perf.data
> > +
> > +You should find some sections of this file have AUX data blocks like:
> > +
> > + 0x1e78 [0x30]: PERF_RECORD_AUXTRACE size: 0x11dd0 offset: 0 ref: 0x1b614fc1061b0ad1 idx: 0 tid: 531230 cpu: -1
> > +
> > + . ... CoreSight ETM Trace data: size 73168 bytes
> > + Idx:0; ID:10; I_ASYNC : Alignment Synchronisation.
> > + Idx:12; ID:10; I_TRACE_INFO : Trace Info.; INFO=0x0 { CC.0 }
> > + Idx:17; ID:10; I_ADDR_L_64IS0 : Address, Long, 64 bit, IS0.; Addr=0x0000000000000000;
> > + Idx:26; ID:10; I_TRACE_ON : Trace On.
> > + Idx:27; ID:10; I_ADDR_CTXT_L_64IS0 : Address & Context, Long, 64 bit, IS0.; Addr=0x0000FFFFB6069140; Ctxt: AArch64,EL0, NS;
> > + Idx:38; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
> > + Idx:39; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
> > + Idx:40; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
> > + Idx:41; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEN
> > + ...
> > +
> > +If you see these above, then your system is tracing coresight data
> > +correctly.
> > +
> > +To compile perf with coresight support in the perf directory do
> > +
> > + make CORESIGHT=1
> > +
> > +This will compile the perf tool with coresight support as well as
> > +build some small test binaries for perf test. This requires you also
> > +be compiling for 64bit Arm (ARM64/aarch64). The tools run as part of
> > +perf coresight tracing are in tests/shell/tools/coresight.
> > +
> > +You will also want coresight support enabled in your kernel config.
> > +Ensure it is enabled with:
> > +
> > + CONFIG_CORESIGHT=y
> > +
> > +There are various other coresight options you probably also want
> > +enabled like:
> > +
> > + CONFIG_CORESIGHT_LINKS_AND_SINKS=y
> > + CONFIG_CORESIGHT_LINK_AND_SINK_TMC=y
> > + CONFIG_CORESIGHT_CATU=y
> > + CONFIG_CORESIGHT_SINK_TPIU=y
> > + CONFIG_CORESIGHT_SINK_ETBV10=y
> > + CONFIG_CORESIGHT_SOURCE_ETM4X=y
> > + CONFIG_CORESIGHT_STM=y
> > + CONFIG_CORESIGHT_CPU_DEBUG=y
> > + CONFIG_CORESIGHT_CTI=y
> > + CONFIG_CORESIGHT_CTI_INTEGRATION_REGS=y
> > +
> > +Please refer to the kernel configuration help for more information.
> > +
> > +Perf test - Verify kernel and userspace perf coresight work
> > +===========================================================
> > +
> > +When you run perf test, it will do a lot of self tests. Some of those
> > +tests will cover Coresight (only if enabled and on ARM64). You
> > +generally would run perf test from the tools/perf directory in the
> > +kernel tree. Some tests will check some internal perf support like:
> > +
> > + Check Arm CoreSight trace data recording and synthesized samples
> > +
> > +Some others will actually use perf record and some test binaries that
> > +are in tests/shell/tools/coresight and will collect traces to ensure a
> > +minimum level of functionality is met. The scripts that launch these
> > +tests are in tests/shell. These will all look like:
> > +
> > + Coresight / Memcpy 1M 25 Threads
> > + Coresight / Unroll Loop Thread 2
> > + ...
> > +
> > +These perf record tests will not run if the tool binaries do not exist
> > +in tests/shell/tools/coresight/*/ and will be skipped. If you do not
> > +have coresight support in hardware then either do not build perf with
> > +coresight support or remove these binaries in order to not have these
> > +tests fail and have them skip instead.
> > +
> > +These tests will log historical results in the current working
> > +directory (e.g. tools/perf) and will be named stats-*.csv like:
> > +
> > + stats-asm_pure_loop-out.csv
> > + stats-bubble_sort-random.csv
> > + ...
> > +
> > +These statistic files log some aspects of the AUX data sections in
> > +the perf data output counting some numbers of certain encodings (a
> > +good way to know that it's working in a very simple way). One problem
> > +with coresight is that given a large enough amount of data needing to
> > +be logged, some of it can be lost due to the processor not waking up
> > +in time to read out all the data from buffers etc.. You will notice
> > +that the amount of data collected can vary a lot per run of perf test.
> > +If you wish to see how this changes over time, simply run perf test
> > +multiple times and all these csv files will have more and more data
> > +appended to it that you can later examine, graph and otherwise use to
> > +figure out if things have become worse or better.
> > +
> > +Be aware that amny of these tests take quite a while to run, specifically
> > +in processing the perf data file and dumping contents to then examine what
> > +is inside.
> > +
> > +You can change where these csv logs are stored by setting the
> > +PERF_TEST_CORESIGHT_STATDIR environment variable before running perf
> > +test like:
> > +
> > + export PERF_TEST_CORESIGHT_STATDIR=/var/tmp
> > + perf test
> > +
> > +They will also store resulting perf output data in the current
> > +directory for later inspection like:
> > +
> > + perf-memcpy-1m.data
> > + perf-thread_loop-2th.data
> > + ...
> > +
> > +You can alter where the perf data files are stored by setting the
> > +PERF_TEST_CORESIGHT_DATADIR environment variable such as:
> > +
> > + PERF_TEST_CORESIGHT_DATADIR=/var/tmp
> > + perf test
> > +
> > +You may wish to set these above environment variables if you which to
> > +keep the output of tests outside of the current working directory for
> > +longer term storage and examination.
> > diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
> > index ac861e42c8f7..b97db83992e0 100644
> > --- a/tools/perf/Makefile.perf
> > +++ b/tools/perf/Makefile.perf
> > @@ -630,7 +630,15 @@ sync_file_range_tbls := $(srctree)/tools/perf/trace/beauty/sync_file_range.sh
> > $(sync_file_range_arrays): $(linux_uapi_dir)/fs.h $(sync_file_range_tbls)
> > $(Q)$(SHELL) '$(sync_file_range_tbls)' $(linux_uapi_dir) > $@
> >
> > -all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS)
> > +TESTS_CORESIGHT_DIR := $(srctree)/tools/perf/tests/shell/coresight
> > +
> > +tests-coresight-targets: FORCE
> > + $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR)
> > +
> > +tests-coresight-targets-clean:
> > + $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR) clean
> > +
> > +all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS) tests-coresight-targets
> >
> > # Create python binding output directory if not already present
> > _dummy := $(shell [ -d '$(OUTPUT)python' ] || mkdir -p '$(OUTPUT)python')
> > @@ -1020,6 +1028,7 @@ install-tests: all install-gtk
> > $(INSTALL) tests/shell/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell'; \
> > $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'; \
> > $(INSTALL) tests/shell/lib/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'
> > + $(Q)$(MAKE) -C tests/shell/coresight install-tests
> >
> > install-bin: install-tools install-tests install-traceevent-plugins
> >
> > @@ -1088,7 +1097,7 @@ endif # BUILD_BPF_SKEL
> > bpf-skel-clean:
> > $(call QUIET_CLEAN, bpf-skel) $(RM) -r $(SKEL_TMP_OUT) $(SKELETONS)
> >
> > -clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean
> > +clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean tests-coresight-targets-clean
> > $(call QUIET_CLEAN, core-objs) $(RM) $(LIBPERF_A) $(OUTPUT)perf-archive $(OUTPUT)perf-with-kcore $(OUTPUT)perf-iostat $(LANG_BINDINGS)
> > $(Q)find $(if $(OUTPUT),$(OUTPUT),.) -name '*.o' -delete -o -name '\.*.cmd' -delete -o -name '\.*.d' -delete
> > $(Q)$(RM) $(OUTPUT).config-detected
> > @@ -1155,5 +1164,6 @@ FORCE:
> > .PHONY: shell_compatibility_test please_set_SHELL_PATH_to_a_more_modern_shell
> > .PHONY: $(GIT-HEAD-PHONY) TAGS tags cscope FORCE prepare
> > .PHONY: libtraceevent_plugins archheaders
> > +.PHONY: $(TESTS_CORESIGHT_TARGETS)
> >
> > endif # force_fixdep
> > diff --git a/tools/perf/tests/shell/coresight/Makefile b/tools/perf/tests/shell/coresight/Makefile
> > new file mode 100644
> > index 000000000000..dda99aeac158
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight/Makefile
> > @@ -0,0 +1,30 @@
> > +# SPDX-License-Identifier: GPL-2.0-only
> > +# Carsten Haitzler <[email protected]>, 2021
> > +include ../../../../../tools/scripts/Makefile.include
> > +include ../../../../../tools/scripts/Makefile.arch
> > +include ../../../../../tools/scripts/utilities.mak
> > +
> > +SUBDIRS = \
> > + asm_pure_loop \
> > + thread_loop \
> > + memcpy_thread \
> > + unroll_loop_thread
> > +
> > +all: $(SUBDIRS)
> > +$(SUBDIRS):
> > + $(Q)$(MAKE) -C $@
> > +
> > +INSTALLDIRS = $(SUBDIRS:%=install-%)
> > +
> > +install-tests: $(INSTALLDIRS)
> > +$(INSTALLDIRS):
> > + $(Q)$(MAKE) -C $(@:install-%=%) install-tests
> > +
> > +CLEANDIRS = $(SUBDIRS:%=clean-%)
> > +
> > +clean: $(CLEANDIRS)
> > +$(CLEANDIRS):
> > + $(Q)$(MAKE) -C $(@:clean-%=%) clean >/dev/null
> > +
> > +.PHONY: all clean $(SUBDIRS) $(CLEANDIRS) $(INSTALLDIRS)
> > +
> > diff --git a/tools/perf/tests/shell/coresight/Makefile.miniconfig b/tools/perf/tests/shell/coresight/Makefile.miniconfig
> > new file mode 100644
> > index 000000000000..893c12685fed
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight/Makefile.miniconfig
> > @@ -0,0 +1,23 @@
> > +# SPDX-License-Identifier: GPL-2.0-only
> > +# Carsten Haitzler <[email protected]>, 2021
> > +
> > +ifndef DESTDIR
> > +prefix ?= $(HOME)
> > +endif
> > +
> > +DESTDIR_SQ = $(subst ','\'',$(DESTDIR))
> > +perfexecdir = libexec/perf-core
> > +perfexec_instdir = $(perfexecdir)
> > +
> > +ifneq ($(filter /%,$(firstword $(perfexecdir))),)
> > +perfexec_instdir = $(perfexecdir)
> > +else
> > +perfexec_instdir = $(prefix)/$(perfexecdir)
> > +endif
> > +
> > +perfexec_instdir_SQ = $(subst ','\'',$(perfexec_instdir))
> > +INSTALL = install
> > +
> > +include ../../../../../scripts/Makefile.include
> > +include ../../../../../scripts/Makefile.arch
> > +include ../../../../../scripts/utilities.mak
> > diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
> > new file mode 100644
> > index 000000000000..468673ac32e8
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
> > @@ -0,0 +1 @@
> > +asm_pure_loop
> > diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
> > new file mode 100644
> > index 000000000000..10c5a60cb71c
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
> > @@ -0,0 +1,30 @@
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Carsten Haitzler <[email protected]>, 2021
> > +
> > +include ../Makefile.miniconfig
> > +
> > +BIN=asm_pure_loop
> > +LIB=
> > +
> > +all: $(BIN)
> > +
> > +$(BIN): $(BIN).S
> > +ifdef CORESIGHT
> > +ifeq ($(ARCH),arm64)
> > + $(Q)$(CC) $(BIN).S -nostdlib -static -o $(BIN) $(LIB)
> > +endif
> > +endif
> > +
> > +install-tests: all
> > +ifdef CORESIGHT
> > +ifeq ($(ARCH),arm64)
> > + $(call QUIET_INSTALL, tests) \
> > + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
> > + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
> > +endif
> > +endif
> > +
> > +clean:
> > + $(Q)$(RM) -f $(BIN)
> > +
> > +.PHONY: all clean install-tests
> > diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
> > new file mode 100644
> > index 000000000000..75cf084a927d
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
> > @@ -0,0 +1,28 @@
> > +/* SPDX-License-Identifier: GPL-2.0 */
> > +/* Tamas Zsoldos <[email protected]>, 2021 */
> > +
> > +.globl _start
> > +_start:
> > + mov x0, 0x0000ffff
> > + mov x1, xzr
> > +loop:
> > + nop
> > + nop
> > + cbnz x1, noskip
> > + nop
> > + nop
> > + adrp x2, skip
> > + add x2, x2, :lo12:skip
> > + br x2
> > + nop
> > + nop
> > +noskip:
> > + nop
> > + nop
> > +skip:
> > + sub x0, x0, 1
> > + cbnz x0, loop
> > +
> > + mov x0, #0
> > + mov x8, #93 // __NR_exit syscall
> > + svc #0
> > diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
> > new file mode 100644
> > index 000000000000..f8217e56091e
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
> > @@ -0,0 +1 @@
> > +memcpy_thread
> > diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/Makefile b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile
> > new file mode 100644
> > index 000000000000..e2604cfae74b
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile
> > @@ -0,0 +1,29 @@
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Carsten Haitzler <[email protected]>, 2021
> > +include ../Makefile.miniconfig
> > +
> > +BIN=memcpy_thread
> > +LIB=-pthread
> > +
> > +all: $(BIN)
> > +
> > +$(BIN): $(BIN).c
> > +ifdef CORESIGHT
> > +ifeq ($(ARCH),arm64)
> > + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
> > +endif
> > +endif
> > +
> > +install-tests: all
> > +ifdef CORESIGHT
> > +ifeq ($(ARCH),arm64)
> > + $(call QUIET_INSTALL, tests) \
> > + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
> > + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
> > +endif
> > +endif
> > +
> > +clean:
> > + $(Q)$(RM) -f $(BIN)
> > +
> > +.PHONY: all clean install-tests
> > diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
> > new file mode 100644
> > index 000000000000..a7e169d1bf64
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
> > @@ -0,0 +1,79 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +// Carsten Haitzler <[email protected]>, 2021
> > +#include <stdio.h>
> > +#include <stdlib.h>
> > +#include <unistd.h>
> > +#include <string.h>
> > +#include <pthread.h>
> > +
> > +struct args {
> > + unsigned long loops;
> > + unsigned long size;
> > + pthread_t th;
> > + void *ret;
> > +};
> > +
> > +static void *thrfn(void *arg)
> > +{
> > + struct args *a = arg;
> > + unsigned long i, len = a->loops;
> > + unsigned char *src, *dst;
> > +
> > + src = malloc(a->size * 1024);
> > + dst = malloc(a->size * 1024);
> > + if ((!src) || (!dst)) {
> > + printf("ERR: Can't allocate memory\n");
> > + exit(1);
> > + }
> > + for (i = 0; i < len; i++)
> > + memcpy(dst, src, a->size * 1024);
> > +}
> > +
> > +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
> > +{
> > + pthread_t t;
> > + pthread_attr_t attr;
> > +
> > + pthread_attr_init(&attr);
> > + pthread_create(&t, &attr, fn, arg);
> > + return t;
> > +}
> > +
> > +int main(int argc, char **argv)
> > +{
> > + unsigned long i, len, size, thr;
> > + pthread_t threads[256];
> > + struct args args[256];
> > + long long v;
> > +
> > + if (argc < 4) {
> > + printf("ERR: %s [copysize Kb] [numthreads] [numloops (hundreds)]\n", argv[0]);
> > + exit(1);
> > + }
> > +
> > + v = atoll(argv[1]);
> > + if ((v < 1) || (v > (1024 * 1024))) {
> > + printf("ERR: max memory 1GB (1048576 KB)\n");
> > + exit(1);
> > + }
> > + size = v;
> > + thr = atol(argv[2]);
> > + if ((thr < 1) || (thr > 256)) {
> > + printf("ERR: threads 1-256\n");
> > + exit(1);
> > + }
> > + v = atoll(argv[3]);
> > + if ((v < 1) || (v > 40000000000ll)) {
> > + printf("ERR: loops 1-40000000000 (hundreds)\n");
> > + exit(1);
> > + }
> > + len = v * 100;
> > + for (i = 0; i < thr; i++) {
> > + args[i].loops = len;
> > + args[i].size = size;
> > + args[i].th = new_thr(thrfn, &(args[i]));
> > + }
> > + for (i = 0; i < thr; i++)
> > + pthread_join(args[i].th, &(args[i].ret));
> > + return 0;
> > +}
> > diff --git a/tools/perf/tests/shell/coresight/thread_loop/.gitignore b/tools/perf/tests/shell/coresight/thread_loop/.gitignore
> > new file mode 100644
> > index 000000000000..6d4c33eaa9e8
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight/thread_loop/.gitignore
> > @@ -0,0 +1 @@
> > +thread_loop
> > diff --git a/tools/perf/tests/shell/coresight/thread_loop/Makefile b/tools/perf/tests/shell/coresight/thread_loop/Makefile
> > new file mode 100644
> > index 000000000000..424df4e8b0e6
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight/thread_loop/Makefile
> > @@ -0,0 +1,29 @@
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Carsten Haitzler <[email protected]>, 2021
> > +include ../Makefile.miniconfig
> > +
> > +BIN=thread_loop
> > +LIB=-pthread
> > +
> > +all: $(BIN)
> > +
> > +$(BIN): $(BIN).c
> > +ifdef CORESIGHT
> > +ifeq ($(ARCH),arm64)
> > + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
> > +endif
> > +endif
> > +
> > +install-tests: all
> > +ifdef CORESIGHT
> > +ifeq ($(ARCH),arm64)
> > + $(call QUIET_INSTALL, tests) \
> > + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
> > + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
> > +endif
> > +endif
> > +
> > +clean:
> > + $(Q)$(RM) -f $(BIN)
> > +
> > +.PHONY: all clean install-tests
> > diff --git a/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
> > new file mode 100644
> > index 000000000000..c0158fac7d0b
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
> > @@ -0,0 +1,86 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +// Carsten Haitzler <[email protected]>, 2021
> > +
> > +// define this for gettid()
> > +#define _GNU_SOURCE
> > +
> > +#include <stdio.h>
> > +#include <stdlib.h>
> > +#include <unistd.h>
> > +#include <string.h>
> > +#include <pthread.h>
> > +#include <sys/syscall.h>
> > +#ifndef SYS_gettid
> > +// gettid is 178 on arm64
> > +# define SYS_gettid 178
> > +#endif
> > +#define gettid() syscall(SYS_gettid)
> > +
> > +struct args {
> > + unsigned int loops;
> > + pthread_t th;
> > + void *ret;
> > +};
> > +
> > +static void *thrfn(void *arg)
> > +{
> > + struct args *a = arg;
> > + int i = 0, len = a->loops;
> > +
> > + if (getenv("SHOW_TID")) {
> > + unsigned long long tid = gettid();
> > +
> > + printf("%llu\n", tid);
> > + }
> > + asm volatile(
> > + "loop:\n"
> > + "add %[i], %[i], #1\n"
> > + "cmp %[i], %[len]\n"
> > + "blt loop\n"
> > + : /* out */
> > + : /* in */ [i] "r" (i), [len] "r" (len)
> > + : /* clobber */
> > + );
> > + return (void *)(long)i;
> > +}
> > +
> > +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
> > +{
> > + pthread_t t;
> > + pthread_attr_t attr;
> > +
> > + pthread_attr_init(&attr);
> > + pthread_create(&t, &attr, fn, arg);
> > + return t;
> > +}
> > +
> > +int main(int argc, char **argv)
> > +{
> > + unsigned int i, len, thr;
> > + pthread_t threads[256];
> > + struct args args[256];
> > +
> > + if (argc < 3) {
> > + printf("ERR: %s [numthreads] [numloops (millions)]\n", argv[0]);
> > + exit(1);
> > + }
> > +
> > + thr = atoi(argv[1]);
> > + if ((thr < 1) || (thr > 256)) {
> > + printf("ERR: threads 1-256\n");
> > + exit(1);
> > + }
> > + len = atoi(argv[2]);
> > + if ((len < 1) || (len > 4000)) {
> > + printf("ERR: max loops 4000 (millions)\n");
> > + exit(1);
> > + }
> > + len *= 1000000;
> > + for (i = 0; i < thr; i++) {
> > + args[i].loops = len;
> > + args[i].th = new_thr(thrfn, &(args[i]));
> > + }
> > + for (i = 0; i < thr; i++)
> > + pthread_join(args[i].th, &(args[i].ret));
> > + return 0;
> > +}
> > diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
> > new file mode 100644
> > index 000000000000..2cb4e996dbf3
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
> > @@ -0,0 +1 @@
> > +unroll_loop_thread
> > diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
> > new file mode 100644
> > index 000000000000..45ab2be8be92
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
> > @@ -0,0 +1,29 @@
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Carsten Haitzler <[email protected]>, 2021
> > +include ../Makefile.miniconfig
> > +
> > +BIN=unroll_loop_thread
> > +LIB=-pthread
> > +
> > +all: $(BIN)
> > +
> > +$(BIN): $(BIN).c
> > +ifdef CORESIGHT
> > +ifeq ($(ARCH),arm64)
> > + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
> > +endif
> > +endif
> > +
> > +install-tests: all
> > +ifdef CORESIGHT
> > +ifeq ($(ARCH),arm64)
> > + $(call QUIET_INSTALL, tests) \
> > + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
> > + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
> > +endif
> > +endif
> > +
> > +clean:
> > + $(Q)$(RM) -f $(BIN)
> > +
> > +.PHONY: all clean install-tests
> > diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
> > new file mode 100644
> > index 000000000000..cb9d22c7dfb9
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
> > @@ -0,0 +1,74 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +// Carsten Haitzler <[email protected]>, 2021
> > +#include <stdio.h>
> > +#include <stdlib.h>
> > +#include <unistd.h>
> > +#include <string.h>
> > +#include <pthread.h>
> > +
> > +struct args {
> > + pthread_t th;
> > + unsigned int in, out;
> > + void *ret;
> > +};
> > +
> > +static void *thrfn(void *arg)
> > +{
> > + struct args *a = arg;
> > + unsigned int i, in = a->in;
> > +
> > + for (i = 0; i < 10000; i++) {
> > + asm volatile (
> > +// force an unroll of thia add instruction so we can test long runs of code
> > +#define SNIP1 "add %[in], %[in], #1\n"
> > +// 10
> > +#define SNIP2 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1
> > +// 100
> > +#define SNIP3 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2
> > +// 1000
> > +#define SNIP4 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3
> > +// 10000
> > +#define SNIP5 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4
> > +// 100000
> > + SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5
> > + : /* out */
> > + : /* in */ [in] "r" (in)
> > + : /* clobber */
> > + );
> > + }
> > +}
> > +
> > +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
> > +{
> > + pthread_t t;
> > + pthread_attr_t attr;
> > +
> > + pthread_attr_init(&attr);
> > + pthread_create(&t, &attr, fn, arg);
> > + return t;
> > +}
> > +
> > +int main(int argc, char **argv)
> > +{
> > + unsigned int i, thr;
> > + pthread_t threads[256];
> > + struct args args[256];
> > +
> > + if (argc < 2) {
> > + printf("ERR: %s [numthreads]\n", argv[0]);
> > + exit(1);
> > + }
> > +
> > + thr = atoi(argv[1]);
> > + if ((thr > 256) || (thr < 1)) {
> > + printf("ERR: threads 1-256\n");
> > + exit(1);
> > + }
> > + for (i = 0; i < thr; i++) {
> > + args[i].in = rand();
> > + args[i].th = new_thr(thrfn, &(args[i]));
> > + }
> > + for (i = 0; i < thr; i++)
> > + pthread_join(args[i].th, &(args[i].ret));
> > + return 0;
> > +}
> > diff --git a/tools/perf/tests/shell/coresight_asm_pure_loop.sh b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
> > new file mode 100755
> > index 000000000000..3f0dbefcad50
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
> > @@ -0,0 +1,18 @@
> > +#!/bin/sh -e
> > +# Coresight / ASM Pure Loop
> > +
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Carsten Haitzler <[email protected]>, 2021
> > +
> > +TEST="asm_pure_loop"
> > +. $(dirname $0)/lib/coresight.sh
> > +ARGS=""
> > +DATV="out"
> > +DATA="$DATD/perf-$TEST-$DATV.data"
> > +
> > +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
> > +
> > +perf_dump_aux_verify "$DATA" 10 10 10
> > +
> > +err=$?
> > +exit $err
> > diff --git a/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh b/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
> > new file mode 100755
> > index 000000000000..8972af835016
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
> > @@ -0,0 +1,18 @@
> > +#!/bin/sh -e
> > +# Coresight / Memcpy 16k 10 Threads
> > +
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Carsten Haitzler <[email protected]>, 2021
> > +
> > +TEST="memcpy_thread"
> > +. $(dirname $0)/lib/coresight.sh
> > +ARGS="16 10 1"
> > +DATV="16k_10"
> > +DATA="$DATD/perf-$TEST-$DATV.data"
> > +
> > +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
> > +
> > +perf_dump_aux_verify "$DATA" 10 10 10
> > +
> > +err=$?
> > +exit $err
> > diff --git a/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh b/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
> > new file mode 100755
> > index 000000000000..5b468901f89b
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
> > @@ -0,0 +1,19 @@
> > +#!/bin/sh -e
> > +# Coresight / Thread Loop 10 Threads - Check TID
> > +
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Carsten Haitzler <[email protected]>, 2021
> > +
> > +TEST="thread_loop"
> > +. $(dirname $0)/lib/coresight.sh
> > +ARGS="10 1"
> > +DATV="check-tid-10th"
> > +DATA="$DATD/perf-$TEST-$DATV.data"
> > +STDO="$DATD/perf-$TEST-$DATV.stdout"
> > +
> > +SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO
> > +
> > +perf_dump_aux_tid_verify "$DATA" "$STDO"
> > +
> > +err=$?
> > +exit $err
> > diff --git a/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh b/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
> > new file mode 100755
> > index 000000000000..f8b7abd3aa03
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
> > @@ -0,0 +1,19 @@
> > +#!/bin/sh -e
> > +# Coresight / Thread Loop 2 Threads - Check TID
> > +
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Carsten Haitzler <[email protected]>, 2021
> > +
> > +TEST="thread_loop"
> > +. $(dirname $0)/lib/coresight.sh
> > +ARGS="2 20"
> > +DATV="check-tid-2th"
> > +DATA="$DATD/perf-$TEST-$DATV.data"
> > +STDO="$DATD/perf-$TEST-$DATV.stdout"
> > +
> > +SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO
> > +
> > +perf_dump_aux_tid_verify "$DATA" "$STDO"
> > +
> > +err=$?
> > +exit $err
> > diff --git a/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh b/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
> > new file mode 100755
> > index 000000000000..c985dfb025c2
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
> > @@ -0,0 +1,18 @@
> > +#!/bin/sh -e
> > +# Coresight / Unroll Loop Thread 10
> > +
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Carsten Haitzler <[email protected]>, 2021
> > +
> > +TEST="unroll_loop_thread"
> > +. $(dirname $0)/lib/coresight.sh
> > +ARGS="10"
> > +DATV="10"
> > +DATA="$DATD/perf-$TEST-$DATV.data"
> > +
> > +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
> > +
> > +perf_dump_aux_verify "$DATA" 10 10 10
> > +
> > +err=$?
> > +exit $err
> > diff --git a/tools/perf/tests/shell/lib/coresight.sh b/tools/perf/tests/shell/lib/coresight.sh
> > new file mode 100644
> > index 000000000000..6a611b073f02
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/lib/coresight.sh
> > @@ -0,0 +1,130 @@
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Carsten Haitzler <[email protected]>, 2021
> > +
> > +# This is sourced from a driver script so no need for #!/bin... etc. at the
> > +# top - the assumption below is that it runs as part of sourcing after the
> > +# test sets up some basic env vars to say what it is.
> > +
> > +# perf record options for the perf tests to use
> > +PERFRECMEM="-m ,128M"
> > +PERFRECOPT="$PERFRECMEM -e cs_etm//u"
> > +
> > +# These tests need to be run as root or coresight won't allow large buffers
> > +# and will not collect proper data
> > +UID=`id -u`
> > +if test "$UID" -ne 0; then
> > + echo "Not running as root... skip"
> > + exit 2
> > +fi
> > +
> > +TOOLS=$(dirname $0)
> > +DIR="$TOOLS/coresight/$TEST"
> > +BIN="$DIR/$TEST"
> > +# If the test tool/binary does not exist and is executable then skip the test
> > +if ! test -x "$BIN"; then exit 2; fi
> > +DATD="."
> > +# If the data dir env is set then make the data dir use that instead of ./
> > +if test -n "$PERF_TEST_CORESIGHT_DATADIR"; then
> > + DATD="$PERF_TEST_CORESIGHT_DATADIR";
> > +fi
> > +# If the stat dir env is set then make the data dir use that instead of ./
> > +STATD="."
> > +if test -n "$PERF_TEST_CORESIGHT_STATDIR"; then
> > + STATD="$PERF_TEST_CORESIGHT_STATDIR";
> > +fi
> > +
> > +# Called if the test fails - error code 2
> > +err() {
> > + echo "$1"
> > + exit 1
> > +}
> > +
> > +# Check that some statistics from our perf
> > +check_val_min() {
> > + STATF="$4"
> > + if test "$2" -lt "$3"; then
> > + echo ", FAILED" >> "$STATF"
> > + err "Sanity check number of $1 is too low ($2 < $3)"
> > + fi
> > +}
> > +
> > +perf_dump_aux_verify() {
> > + # Some basic checking that the AUX chunk contains some sensible data
> > + # to see that we are recording something and at least a minimum
> > + # amount of it. We should almost always see F3 atoms in just about
> > + # anything but certainly we will see some trace info and async atom
> > + # chunks.
> > + DUMP="$DATD/perf-tmp-aux-dump.txt"
> > + perf report --stdio --dump -i "$1" | \
> > + grep -o -e I_ATOM_F3 -e I_ASYNC -e I_TRACE_INFO > "$DUMP"
> > + # Simply count how many of these atoms we find to see that we are
> > + # producing a reasonable amount of data - exact checks are not sane
> > + # as this is a lossy process where we may lose some blocks and the
> > + # compiler may produce different code depending on the compiler and
> > + # optimization options, so this is rough just to see if we're
> > + # either missing almost all the data or all of it
> > + ATOM_F3_NUM=`grep I_ATOM_F3 "$DUMP" | wc -l`
> > + ATOM_ASYNC_NUM=`grep I_ASYNC "$DUMP" | wc -l`
> > + ATOM_TRACE_INFO_NUM=`grep I_TRACE_INFO "$DUMP" | wc -l`
> > + rm -f "$DUMP"
> > +
> > + # Arguments provide minimums for a pass
> > + CHECK_F3_MIN="$2"
> > + CHECK_ASYNC_MIN="$3"
> > + CHECK_TRACE_INFO_MIN="$4"
> > +
> > + # Write out statistics, so over time you can track results to see if
> > + # there is a pattern - for example we have less "noisy" results that
> > + # produce more consistent amounts of data each run, to see if over
> > + # time any techinques to minimize data loss are having an effect or
> > + # not
> > + STATF="$STATD/stats-$TEST-$DATV.csv"
> > + if ! test -f "$STATF"; then
> > + echo "ATOM F3 Count, Minimum, ATOM ASYNC Count, Minimum, TRACE INFO Count, Minimum" > "$STATF"
> > + fi
> > + echo -n "$ATOM_F3_NUM, $CHECK_F3_MIN, $ATOM_ASYNC_NUM, $CHECK_ASYNC_MIN, $ATOM_TRACE_INFO_NUM, $CHECK_TRACE_INFO_MIN" >> "$STATF"
> > +
> > + # Actually check to see if we passed or failed.
> > + check_val_min "ATOM_F3" "$ATOM_F3_NUM" "$CHECK_F3_MIN" "$STATF"
> > + check_val_min "ASYNC" "$ATOM_ASYNC_NUM" "$CHECK_ASYNC_MIN" "$STATF"
> > + check_val_min "TRACE_INFO" "$ATOM_TRACE_INFO_NUM" "$CHECK_TRACE_INFO_MIN" "$STATF"
> > + echo ", Ok" >> "$STATF"
> > +}
> > +
> > +perf_dump_aux_tid_verify() {
> > + # Specifically crafted test will produce a list of Tread ID's to
> > + # stdout that need to be checked to see that they have had trace
> > + # info collected in AUX blocks in the perf data. This will go
> > + # through all the TID's that are listed as CID=0xabcdef and see
> > + # that all the Thread IDs the test tool reports are in the perf
> > + # data AUX chunks
> > +
> > + # The TID test tools will print a TID per stdout line that are being
> > + # tested
> > + TIDS=`cat "$2"`
> > + # Scan the perf report to find the TIDs that are actually CID in hex
> > + # and build a list of the ones found
> > + FOUND_TIDS=`perf report --stdio --dump -i "$1" | \
> > + grep -o "CID=0x[0-9a-z]\+" | sed 's/CID=//g' | \
> > + uniq | sort | uniq`
> > +
> > + # Iterate over the list of TIDs that the test says it has and find
> > + # them in the TIDs found in the perf report
> > + MISSING=""
> > + for TID2 in $TIDS; do
> > + FOUND=""
> > + for TIDHEX in $FOUND_TIDS; do
> > + TID=`printf "%i" $TIDHEX`
> > + if test "$TID" -eq "$TID2"; then
> > + FOUND="y"
> > + break
> > + fi
> > + done
> > + if test -z "$FOUND"; then
> > + MISSING="$MISSING $TID"
> > + fi
> > + done
> > + if test -n "$MISSING"; then
> > + err "Thread IDs $MISSING not found in perf AUX data"
> > + fi
> > +}
> > --
> > 2.32.0
> >
On 5/30/22 17:27, Mathieu Poirier wrote:
> On Wed, Mar 09, 2022 at 12:28:59PM +0000, [email protected] wrote:
>> From: Carsten Haitzler <[email protected]>
>>
>> This adds a test harness and tests to run perf record and examine the
>> resuling output when coresight is enabled on arm64 and check the
>> resulting quality of the output as part of perf test. These tests use
>> various tools to produce output from perf record then measure some key
>> specific aspects of that data to see if the data exists at all and
>> contains key aspects such as measuring some data for every thread of
>> a test or produces sufficient data for large exeuction runs of a large
>> executable. etc.
>>
>> Signed-off-by: Carsten Haitzler <[email protected]>
>> ---
>> MAINTAINERS | 4 +
>> tools/perf/.gitignore | 6 +-
>> tools/perf/Documentation/arm-coresight.txt | 140 ++++++++++++++++++
>> tools/perf/Makefile.perf | 14 +-
>> tools/perf/tests/shell/coresight/Makefile | 30 ++++
>> .../tests/shell/coresight/Makefile.miniconfig | 23 +++
>> .../shell/coresight/asm_pure_loop/.gitignore | 1 +
>> .../shell/coresight/asm_pure_loop/Makefile | 30 ++++
>> .../coresight/asm_pure_loop/asm_pure_loop.S | 28 ++++
>> .../shell/coresight/memcpy_thread/.gitignore | 1 +
>> .../shell/coresight/memcpy_thread/Makefile | 29 ++++
>> .../coresight/memcpy_thread/memcpy_thread.c | 79 ++++++++++
>> .../shell/coresight/thread_loop/.gitignore | 1 +
>> .../shell/coresight/thread_loop/Makefile | 29 ++++
>> .../shell/coresight/thread_loop/thread_loop.c | 86 +++++++++++
>> .../coresight/unroll_loop_thread/.gitignore | 1 +
>> .../coresight/unroll_loop_thread/Makefile | 29 ++++
>> .../unroll_loop_thread/unroll_loop_thread.c | 74 +++++++++
>> .../tests/shell/coresight_asm_pure_loop.sh | 18 +++
>> .../shell/coresight_memcpy_thread_16k_10.sh | 18 +++
>> .../coresight_thread_loop_check_tid_10.sh | 19 +++
>> .../coresight_thread_loop_check_tid_2.sh | 19 +++
>> .../shell/coresight_unroll_loop_thread_10.sh | 18 +++
>> tools/perf/tests/shell/lib/coresight.sh | 130 ++++++++++++++++
>> 24 files changed, 823 insertions(+), 4 deletions(-)
>
> As Leo pointed out this is a big patch and hard to digest intellectually.
>
>> create mode 100644 tools/perf/Documentation/arm-coresight.txt
>> create mode 100644 tools/perf/tests/shell/coresight/Makefile
>> create mode 100644 tools/perf/tests/shell/coresight/Makefile.miniconfig
>> create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
>> create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
>> create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
>> create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
>> create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/Makefile
>> create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
>> create mode 100644 tools/perf/tests/shell/coresight/thread_loop/.gitignore
>> create mode 100644 tools/perf/tests/shell/coresight/thread_loop/Makefile
>> create mode 100644 tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
>> create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
>> create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
>> create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
>> create mode 100755 tools/perf/tests/shell/coresight_asm_pure_loop.sh
>> create mode 100755 tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
>> create mode 100755 tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
>> create mode 100755 tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
>> create mode 100755 tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
>> create mode 100644 tools/perf/tests/shell/lib/coresight.sh
>>
>> diff --git a/MAINTAINERS b/MAINTAINERS
>> index 673c7124ca82..18cc20609f2e 100644
>> --- a/MAINTAINERS
>> +++ b/MAINTAINERS
>> @@ -1918,10 +1918,14 @@ F: drivers/hwtracing/coresight/*
>> F: include/dt-bindings/arm/coresight-cti-dt.h
>> F: include/linux/coresight*
>> F: samples/coresight/*
>> +F: tools/perf/Documentation/arm-coresight.txt
>> F: tools/perf/arch/arm/util/auxtrace.c
>> F: tools/perf/arch/arm/util/cs-etm.c
>> F: tools/perf/arch/arm/util/cs-etm.h
>> F: tools/perf/arch/arm/util/pmu.c
>> +F: tools/perf/tests/shell/coresight_*
>> +F: tools/perf/tests/shell/tools/Makefile
>> +F: tools/perf/tests/shell/tools/coresight/*
>> F: tools/perf/util/cs-etm-decoder/*
>> F: tools/perf/util/cs-etm.*
>>
>> diff --git a/tools/perf/.gitignore b/tools/perf/.gitignore
>> index 20b8ab984d5f..138c679ecacd 100644
>> --- a/tools/perf/.gitignore
>> +++ b/tools/perf/.gitignore
>> @@ -15,8 +15,9 @@ perf*.1
>> perf*.xml
>> perf*.html
>> common-cmds.h
>> -perf.data
>> -perf.data.old
>> +perf*.data
>> +perf*.data.old
>> +stats-*.csv
>> output.svg
>> perf-archive
>> perf-with-kcore
>> @@ -30,6 +31,7 @@ config.mak.autogen
>> *-flex.*
>> *.pyc
>> *.pyo
>> +*.stdout
>> .config-detected
>> util/intel-pt-decoder/inat-tables.c
>> arch/*/include/generated/
>> diff --git a/tools/perf/Documentation/arm-coresight.txt b/tools/perf/Documentation/arm-coresight.txt
>> new file mode 100644
>> index 000000000000..3a9e6c573c58
>> --- /dev/null
>> +++ b/tools/perf/Documentation/arm-coresight.txt
>
> I think it would be best to keep all the coresight documentation under the
> current coresight documentation repository[1]. That way all the information on
> coresight can be found in a central place.
OK. I just added this here because this was relevant "get you going"
documentation for perf + coresight and testing and other relevant docs
were here including architecture specific ones like the intel docs. I'll
find a way of merging it with the above mentioned docs.
> Some part of what is added by this patch is redundant with what is currently
> available in [1]. Other parts are tests specific and should be added under
> something like "coresight-perf-test.rst".
Sure - some is, but most is not mentioned there (the vast majority). I'd
argue essentially all of it is relevant to someone getting started and
the documentation in the above location is not very friendly or useful
in that regard (which is why I wrote the documentation here to begin
with). As I can't really find anything it duplicates other than the
initial paragraph which I can remove, I'll merge the rest in and link it
in as above, unless you can be specific on what it really duplicates?
> Thanks,
> Mathieu
>
> [1]. Documentation/trace/coresight/
>
>
>> @@ -0,0 +1,140 @@
>> +Arm Coresight Support
>> +=====================
>> +
>> +Coresight is a feature of some Arm based processors that allows for
>> +debugging. One of the things it can do is trace every instruction
>> +executed and remotely expose that information in a hardware compressed
>> +stream. Perf is able to locally access that stream and store it to the
>> +output perf data files. This stream can then be later decoded to give the
>> +instructions that were traced for debugging or profiling purposes. You
>> +can log such data with a perf record command like:
>> +
>> + perf record -e cs_etm//u testbinary
>> +
>> +This would run some test binary (testbinary) until it exits and record
>> +a perf.data trace file. That file would have AUX sections if coresight
>> +is working correctly. You can dump the content of this file as
>> +readable text with a command like:
>> +
>> + perf report --stdio --dump -i perf.data
>> +
>> +You should find some sections of this file have AUX data blocks like:
>> +
>> + 0x1e78 [0x30]: PERF_RECORD_AUXTRACE size: 0x11dd0 offset: 0 ref: 0x1b614fc1061b0ad1 idx: 0 tid: 531230 cpu: -1
>> +
>> + . ... CoreSight ETM Trace data: size 73168 bytes
>> + Idx:0; ID:10; I_ASYNC : Alignment Synchronisation.
>> + Idx:12; ID:10; I_TRACE_INFO : Trace Info.; INFO=0x0 { CC.0 }
>> + Idx:17; ID:10; I_ADDR_L_64IS0 : Address, Long, 64 bit, IS0.; Addr=0x0000000000000000;
>> + Idx:26; ID:10; I_TRACE_ON : Trace On.
>> + Idx:27; ID:10; I_ADDR_CTXT_L_64IS0 : Address & Context, Long, 64 bit, IS0.; Addr=0x0000FFFFB6069140; Ctxt: AArch64,EL0, NS;
>> + Idx:38; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
>> + Idx:39; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
>> + Idx:40; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
>> + Idx:41; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEN
>> + ...
>> +
>> +If you see these above, then your system is tracing coresight data
>> +correctly.
>> +
>> +To compile perf with coresight support in the perf directory do
>> +
>> + make CORESIGHT=1
>> +
>> +This will compile the perf tool with coresight support as well as
>> +build some small test binaries for perf test. This requires you also
>> +be compiling for 64bit Arm (ARM64/aarch64). The tools run as part of
>> +perf coresight tracing are in tests/shell/tools/coresight.
>> +
>> +You will also want coresight support enabled in your kernel config.
>> +Ensure it is enabled with:
>> +
>> + CONFIG_CORESIGHT=y
>> +
>> +There are various other coresight options you probably also want
>> +enabled like:
>> +
>> + CONFIG_CORESIGHT_LINKS_AND_SINKS=y
>> + CONFIG_CORESIGHT_LINK_AND_SINK_TMC=y
>> + CONFIG_CORESIGHT_CATU=y
>> + CONFIG_CORESIGHT_SINK_TPIU=y
>> + CONFIG_CORESIGHT_SINK_ETBV10=y
>> + CONFIG_CORESIGHT_SOURCE_ETM4X=y
>> + CONFIG_CORESIGHT_STM=y
>> + CONFIG_CORESIGHT_CPU_DEBUG=y
>> + CONFIG_CORESIGHT_CTI=y
>> + CONFIG_CORESIGHT_CTI_INTEGRATION_REGS=y
>> +
>> +Please refer to the kernel configuration help for more information.
>> +
>> +Perf test - Verify kernel and userspace perf coresight work
>> +===========================================================
>> +
>> +When you run perf test, it will do a lot of self tests. Some of those
>> +tests will cover Coresight (only if enabled and on ARM64). You
>> +generally would run perf test from the tools/perf directory in the
>> +kernel tree. Some tests will check some internal perf support like:
>> +
>> + Check Arm CoreSight trace data recording and synthesized samples
>> +
>> +Some others will actually use perf record and some test binaries that
>> +are in tests/shell/tools/coresight and will collect traces to ensure a
>> +minimum level of functionality is met. The scripts that launch these
>> +tests are in tests/shell. These will all look like:
>> +
>> + Coresight / Memcpy 1M 25 Threads
>> + Coresight / Unroll Loop Thread 2
>> + ...
>> +
>> +These perf record tests will not run if the tool binaries do not exist
>> +in tests/shell/tools/coresight/*/ and will be skipped. If you do not
>> +have coresight support in hardware then either do not build perf with
>> +coresight support or remove these binaries in order to not have these
>> +tests fail and have them skip instead.
>> +
>> +These tests will log historical results in the current working
>> +directory (e.g. tools/perf) and will be named stats-*.csv like:
>> +
>> + stats-asm_pure_loop-out.csv
>> + stats-bubble_sort-random.csv
>> + ...
>> +
>> +These statistic files log some aspects of the AUX data sections in
>> +the perf data output counting some numbers of certain encodings (a
>> +good way to know that it's working in a very simple way). One problem
>> +with coresight is that given a large enough amount of data needing to
>> +be logged, some of it can be lost due to the processor not waking up
>> +in time to read out all the data from buffers etc.. You will notice
>> +that the amount of data collected can vary a lot per run of perf test.
>> +If you wish to see how this changes over time, simply run perf test
>> +multiple times and all these csv files will have more and more data
>> +appended to it that you can later examine, graph and otherwise use to
>> +figure out if things have become worse or better.
>> +
>> +Be aware that amny of these tests take quite a while to run, specifically
>> +in processing the perf data file and dumping contents to then examine what
>> +is inside.
>> +
>> +You can change where these csv logs are stored by setting the
>> +PERF_TEST_CORESIGHT_STATDIR environment variable before running perf
>> +test like:
>> +
>> + export PERF_TEST_CORESIGHT_STATDIR=/var/tmp
>> + perf test
>> +
>> +They will also store resulting perf output data in the current
>> +directory for later inspection like:
>> +
>> + perf-memcpy-1m.data
>> + perf-thread_loop-2th.data
>> + ...
>> +
>> +You can alter where the perf data files are stored by setting the
>> +PERF_TEST_CORESIGHT_DATADIR environment variable such as:
>> +
>> + PERF_TEST_CORESIGHT_DATADIR=/var/tmp
>> + perf test
>> +
>> +You may wish to set these above environment variables if you which to
>> +keep the output of tests outside of the current working directory for
>> +longer term storage and examination.
>> diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
>> index ac861e42c8f7..b97db83992e0 100644
>> --- a/tools/perf/Makefile.perf
>> +++ b/tools/perf/Makefile.perf
>> @@ -630,7 +630,15 @@ sync_file_range_tbls := $(srctree)/tools/perf/trace/beauty/sync_file_range.sh
>> $(sync_file_range_arrays): $(linux_uapi_dir)/fs.h $(sync_file_range_tbls)
>> $(Q)$(SHELL) '$(sync_file_range_tbls)' $(linux_uapi_dir) > $@
>>
>> -all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS)
>> +TESTS_CORESIGHT_DIR := $(srctree)/tools/perf/tests/shell/coresight
>> +
>> +tests-coresight-targets: FORCE
>> + $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR)
>> +
>> +tests-coresight-targets-clean:
>> + $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR) clean
>> +
>> +all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS) tests-coresight-targets
>>
>> # Create python binding output directory if not already present
>> _dummy := $(shell [ -d '$(OUTPUT)python' ] || mkdir -p '$(OUTPUT)python')
>> @@ -1020,6 +1028,7 @@ install-tests: all install-gtk
>> $(INSTALL) tests/shell/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell'; \
>> $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'; \
>> $(INSTALL) tests/shell/lib/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'
>> + $(Q)$(MAKE) -C tests/shell/coresight install-tests
>>
>> install-bin: install-tools install-tests install-traceevent-plugins
>>
>> @@ -1088,7 +1097,7 @@ endif # BUILD_BPF_SKEL
>> bpf-skel-clean:
>> $(call QUIET_CLEAN, bpf-skel) $(RM) -r $(SKEL_TMP_OUT) $(SKELETONS)
>>
>> -clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean
>> +clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean tests-coresight-targets-clean
>> $(call QUIET_CLEAN, core-objs) $(RM) $(LIBPERF_A) $(OUTPUT)perf-archive $(OUTPUT)perf-with-kcore $(OUTPUT)perf-iostat $(LANG_BINDINGS)
>> $(Q)find $(if $(OUTPUT),$(OUTPUT),.) -name '*.o' -delete -o -name '\.*.cmd' -delete -o -name '\.*.d' -delete
>> $(Q)$(RM) $(OUTPUT).config-detected
>> @@ -1155,5 +1164,6 @@ FORCE:
>> .PHONY: shell_compatibility_test please_set_SHELL_PATH_to_a_more_modern_shell
>> .PHONY: $(GIT-HEAD-PHONY) TAGS tags cscope FORCE prepare
>> .PHONY: libtraceevent_plugins archheaders
>> +.PHONY: $(TESTS_CORESIGHT_TARGETS)
>>
>> endif # force_fixdep
>> diff --git a/tools/perf/tests/shell/coresight/Makefile b/tools/perf/tests/shell/coresight/Makefile
>> new file mode 100644
>> index 000000000000..dda99aeac158
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/Makefile
>> @@ -0,0 +1,30 @@
>> +# SPDX-License-Identifier: GPL-2.0-only
>> +# Carsten Haitzler <[email protected]>, 2021
>> +include ../../../../../tools/scripts/Makefile.include
>> +include ../../../../../tools/scripts/Makefile.arch
>> +include ../../../../../tools/scripts/utilities.mak
>> +
>> +SUBDIRS = \
>> + asm_pure_loop \
>> + thread_loop \
>> + memcpy_thread \
>> + unroll_loop_thread
>> +
>> +all: $(SUBDIRS)
>> +$(SUBDIRS):
>> + $(Q)$(MAKE) -C $@
>> +
>> +INSTALLDIRS = $(SUBDIRS:%=install-%)
>> +
>> +install-tests: $(INSTALLDIRS)
>> +$(INSTALLDIRS):
>> + $(Q)$(MAKE) -C $(@:install-%=%) install-tests
>> +
>> +CLEANDIRS = $(SUBDIRS:%=clean-%)
>> +
>> +clean: $(CLEANDIRS)
>> +$(CLEANDIRS):
>> + $(Q)$(MAKE) -C $(@:clean-%=%) clean >/dev/null
>> +
>> +.PHONY: all clean $(SUBDIRS) $(CLEANDIRS) $(INSTALLDIRS)
>> +
>> diff --git a/tools/perf/tests/shell/coresight/Makefile.miniconfig b/tools/perf/tests/shell/coresight/Makefile.miniconfig
>> new file mode 100644
>> index 000000000000..893c12685fed
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/Makefile.miniconfig
>> @@ -0,0 +1,23 @@
>> +# SPDX-License-Identifier: GPL-2.0-only
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +ifndef DESTDIR
>> +prefix ?= $(HOME)
>> +endif
>> +
>> +DESTDIR_SQ = $(subst ','\'',$(DESTDIR))
>> +perfexecdir = libexec/perf-core
>> +perfexec_instdir = $(perfexecdir)
>> +
>> +ifneq ($(filter /%,$(firstword $(perfexecdir))),)
>> +perfexec_instdir = $(perfexecdir)
>> +else
>> +perfexec_instdir = $(prefix)/$(perfexecdir)
>> +endif
>> +
>> +perfexec_instdir_SQ = $(subst ','\'',$(perfexec_instdir))
>> +INSTALL = install
>> +
>> +include ../../../../../scripts/Makefile.include
>> +include ../../../../../scripts/Makefile.arch
>> +include ../../../../../scripts/utilities.mak
>> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
>> new file mode 100644
>> index 000000000000..468673ac32e8
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
>> @@ -0,0 +1 @@
>> +asm_pure_loop
>> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
>> new file mode 100644
>> index 000000000000..10c5a60cb71c
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
>> @@ -0,0 +1,30 @@
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +include ../Makefile.miniconfig
>> +
>> +BIN=asm_pure_loop
>> +LIB=
>> +
>> +all: $(BIN)
>> +
>> +$(BIN): $(BIN).S
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(Q)$(CC) $(BIN).S -nostdlib -static -o $(BIN) $(LIB)
>> +endif
>> +endif
>> +
>> +install-tests: all
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(call QUIET_INSTALL, tests) \
>> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
>> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
>> +endif
>> +endif
>> +
>> +clean:
>> + $(Q)$(RM) -f $(BIN)
>> +
>> +.PHONY: all clean install-tests
>> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
>> new file mode 100644
>> index 000000000000..75cf084a927d
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
>> @@ -0,0 +1,28 @@
>> +/* SPDX-License-Identifier: GPL-2.0 */
>> +/* Tamas Zsoldos <[email protected]>, 2021 */
>> +
>> +.globl _start
>> +_start:
>> + mov x0, 0x0000ffff
>> + mov x1, xzr
>> +loop:
>> + nop
>> + nop
>> + cbnz x1, noskip
>> + nop
>> + nop
>> + adrp x2, skip
>> + add x2, x2, :lo12:skip
>> + br x2
>> + nop
>> + nop
>> +noskip:
>> + nop
>> + nop
>> +skip:
>> + sub x0, x0, 1
>> + cbnz x0, loop
>> +
>> + mov x0, #0
>> + mov x8, #93 // __NR_exit syscall
>> + svc #0
>> diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
>> new file mode 100644
>> index 000000000000..f8217e56091e
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
>> @@ -0,0 +1 @@
>> +memcpy_thread
>> diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/Makefile b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile
>> new file mode 100644
>> index 000000000000..e2604cfae74b
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile
>> @@ -0,0 +1,29 @@
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +include ../Makefile.miniconfig
>> +
>> +BIN=memcpy_thread
>> +LIB=-pthread
>> +
>> +all: $(BIN)
>> +
>> +$(BIN): $(BIN).c
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
>> +endif
>> +endif
>> +
>> +install-tests: all
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(call QUIET_INSTALL, tests) \
>> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
>> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
>> +endif
>> +endif
>> +
>> +clean:
>> + $(Q)$(RM) -f $(BIN)
>> +
>> +.PHONY: all clean install-tests
>> diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
>> new file mode 100644
>> index 000000000000..a7e169d1bf64
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
>> @@ -0,0 +1,79 @@
>> +// SPDX-License-Identifier: GPL-2.0
>> +// Carsten Haitzler <[email protected]>, 2021
>> +#include <stdio.h>
>> +#include <stdlib.h>
>> +#include <unistd.h>
>> +#include <string.h>
>> +#include <pthread.h>
>> +
>> +struct args {
>> + unsigned long loops;
>> + unsigned long size;
>> + pthread_t th;
>> + void *ret;
>> +};
>> +
>> +static void *thrfn(void *arg)
>> +{
>> + struct args *a = arg;
>> + unsigned long i, len = a->loops;
>> + unsigned char *src, *dst;
>> +
>> + src = malloc(a->size * 1024);
>> + dst = malloc(a->size * 1024);
>> + if ((!src) || (!dst)) {
>> + printf("ERR: Can't allocate memory\n");
>> + exit(1);
>> + }
>> + for (i = 0; i < len; i++)
>> + memcpy(dst, src, a->size * 1024);
>> +}
>> +
>> +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
>> +{
>> + pthread_t t;
>> + pthread_attr_t attr;
>> +
>> + pthread_attr_init(&attr);
>> + pthread_create(&t, &attr, fn, arg);
>> + return t;
>> +}
>> +
>> +int main(int argc, char **argv)
>> +{
>> + unsigned long i, len, size, thr;
>> + pthread_t threads[256];
>> + struct args args[256];
>> + long long v;
>> +
>> + if (argc < 4) {
>> + printf("ERR: %s [copysize Kb] [numthreads] [numloops (hundreds)]\n", argv[0]);
>> + exit(1);
>> + }
>> +
>> + v = atoll(argv[1]);
>> + if ((v < 1) || (v > (1024 * 1024))) {
>> + printf("ERR: max memory 1GB (1048576 KB)\n");
>> + exit(1);
>> + }
>> + size = v;
>> + thr = atol(argv[2]);
>> + if ((thr < 1) || (thr > 256)) {
>> + printf("ERR: threads 1-256\n");
>> + exit(1);
>> + }
>> + v = atoll(argv[3]);
>> + if ((v < 1) || (v > 40000000000ll)) {
>> + printf("ERR: loops 1-40000000000 (hundreds)\n");
>> + exit(1);
>> + }
>> + len = v * 100;
>> + for (i = 0; i < thr; i++) {
>> + args[i].loops = len;
>> + args[i].size = size;
>> + args[i].th = new_thr(thrfn, &(args[i]));
>> + }
>> + for (i = 0; i < thr; i++)
>> + pthread_join(args[i].th, &(args[i].ret));
>> + return 0;
>> +}
>> diff --git a/tools/perf/tests/shell/coresight/thread_loop/.gitignore b/tools/perf/tests/shell/coresight/thread_loop/.gitignore
>> new file mode 100644
>> index 000000000000..6d4c33eaa9e8
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/thread_loop/.gitignore
>> @@ -0,0 +1 @@
>> +thread_loop
>> diff --git a/tools/perf/tests/shell/coresight/thread_loop/Makefile b/tools/perf/tests/shell/coresight/thread_loop/Makefile
>> new file mode 100644
>> index 000000000000..424df4e8b0e6
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/thread_loop/Makefile
>> @@ -0,0 +1,29 @@
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +include ../Makefile.miniconfig
>> +
>> +BIN=thread_loop
>> +LIB=-pthread
>> +
>> +all: $(BIN)
>> +
>> +$(BIN): $(BIN).c
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
>> +endif
>> +endif
>> +
>> +install-tests: all
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(call QUIET_INSTALL, tests) \
>> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
>> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
>> +endif
>> +endif
>> +
>> +clean:
>> + $(Q)$(RM) -f $(BIN)
>> +
>> +.PHONY: all clean install-tests
>> diff --git a/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
>> new file mode 100644
>> index 000000000000..c0158fac7d0b
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
>> @@ -0,0 +1,86 @@
>> +// SPDX-License-Identifier: GPL-2.0
>> +// Carsten Haitzler <[email protected]>, 2021
>> +
>> +// define this for gettid()
>> +#define _GNU_SOURCE
>> +
>> +#include <stdio.h>
>> +#include <stdlib.h>
>> +#include <unistd.h>
>> +#include <string.h>
>> +#include <pthread.h>
>> +#include <sys/syscall.h>
>> +#ifndef SYS_gettid
>> +// gettid is 178 on arm64
>> +# define SYS_gettid 178
>> +#endif
>> +#define gettid() syscall(SYS_gettid)
>> +
>> +struct args {
>> + unsigned int loops;
>> + pthread_t th;
>> + void *ret;
>> +};
>> +
>> +static void *thrfn(void *arg)
>> +{
>> + struct args *a = arg;
>> + int i = 0, len = a->loops;
>> +
>> + if (getenv("SHOW_TID")) {
>> + unsigned long long tid = gettid();
>> +
>> + printf("%llu\n", tid);
>> + }
>> + asm volatile(
>> + "loop:\n"
>> + "add %[i], %[i], #1\n"
>> + "cmp %[i], %[len]\n"
>> + "blt loop\n"
>> + : /* out */
>> + : /* in */ [i] "r" (i), [len] "r" (len)
>> + : /* clobber */
>> + );
>> + return (void *)(long)i;
>> +}
>> +
>> +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
>> +{
>> + pthread_t t;
>> + pthread_attr_t attr;
>> +
>> + pthread_attr_init(&attr);
>> + pthread_create(&t, &attr, fn, arg);
>> + return t;
>> +}
>> +
>> +int main(int argc, char **argv)
>> +{
>> + unsigned int i, len, thr;
>> + pthread_t threads[256];
>> + struct args args[256];
>> +
>> + if (argc < 3) {
>> + printf("ERR: %s [numthreads] [numloops (millions)]\n", argv[0]);
>> + exit(1);
>> + }
>> +
>> + thr = atoi(argv[1]);
>> + if ((thr < 1) || (thr > 256)) {
>> + printf("ERR: threads 1-256\n");
>> + exit(1);
>> + }
>> + len = atoi(argv[2]);
>> + if ((len < 1) || (len > 4000)) {
>> + printf("ERR: max loops 4000 (millions)\n");
>> + exit(1);
>> + }
>> + len *= 1000000;
>> + for (i = 0; i < thr; i++) {
>> + args[i].loops = len;
>> + args[i].th = new_thr(thrfn, &(args[i]));
>> + }
>> + for (i = 0; i < thr; i++)
>> + pthread_join(args[i].th, &(args[i].ret));
>> + return 0;
>> +}
>> diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
>> new file mode 100644
>> index 000000000000..2cb4e996dbf3
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
>> @@ -0,0 +1 @@
>> +unroll_loop_thread
>> diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
>> new file mode 100644
>> index 000000000000..45ab2be8be92
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
>> @@ -0,0 +1,29 @@
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +include ../Makefile.miniconfig
>> +
>> +BIN=unroll_loop_thread
>> +LIB=-pthread
>> +
>> +all: $(BIN)
>> +
>> +$(BIN): $(BIN).c
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
>> +endif
>> +endif
>> +
>> +install-tests: all
>> +ifdef CORESIGHT
>> +ifeq ($(ARCH),arm64)
>> + $(call QUIET_INSTALL, tests) \
>> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
>> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
>> +endif
>> +endif
>> +
>> +clean:
>> + $(Q)$(RM) -f $(BIN)
>> +
>> +.PHONY: all clean install-tests
>> diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
>> new file mode 100644
>> index 000000000000..cb9d22c7dfb9
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
>> @@ -0,0 +1,74 @@
>> +// SPDX-License-Identifier: GPL-2.0
>> +// Carsten Haitzler <[email protected]>, 2021
>> +#include <stdio.h>
>> +#include <stdlib.h>
>> +#include <unistd.h>
>> +#include <string.h>
>> +#include <pthread.h>
>> +
>> +struct args {
>> + pthread_t th;
>> + unsigned int in, out;
>> + void *ret;
>> +};
>> +
>> +static void *thrfn(void *arg)
>> +{
>> + struct args *a = arg;
>> + unsigned int i, in = a->in;
>> +
>> + for (i = 0; i < 10000; i++) {
>> + asm volatile (
>> +// force an unroll of thia add instruction so we can test long runs of code
>> +#define SNIP1 "add %[in], %[in], #1\n"
>> +// 10
>> +#define SNIP2 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1
>> +// 100
>> +#define SNIP3 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2
>> +// 1000
>> +#define SNIP4 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3
>> +// 10000
>> +#define SNIP5 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4
>> +// 100000
>> + SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5
>> + : /* out */
>> + : /* in */ [in] "r" (in)
>> + : /* clobber */
>> + );
>> + }
>> +}
>> +
>> +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
>> +{
>> + pthread_t t;
>> + pthread_attr_t attr;
>> +
>> + pthread_attr_init(&attr);
>> + pthread_create(&t, &attr, fn, arg);
>> + return t;
>> +}
>> +
>> +int main(int argc, char **argv)
>> +{
>> + unsigned int i, thr;
>> + pthread_t threads[256];
>> + struct args args[256];
>> +
>> + if (argc < 2) {
>> + printf("ERR: %s [numthreads]\n", argv[0]);
>> + exit(1);
>> + }
>> +
>> + thr = atoi(argv[1]);
>> + if ((thr > 256) || (thr < 1)) {
>> + printf("ERR: threads 1-256\n");
>> + exit(1);
>> + }
>> + for (i = 0; i < thr; i++) {
>> + args[i].in = rand();
>> + args[i].th = new_thr(thrfn, &(args[i]));
>> + }
>> + for (i = 0; i < thr; i++)
>> + pthread_join(args[i].th, &(args[i].ret));
>> + return 0;
>> +}
>> diff --git a/tools/perf/tests/shell/coresight_asm_pure_loop.sh b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
>> new file mode 100755
>> index 000000000000..3f0dbefcad50
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
>> @@ -0,0 +1,18 @@
>> +#!/bin/sh -e
>> +# Coresight / ASM Pure Loop
>> +
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +TEST="asm_pure_loop"
>> +. $(dirname $0)/lib/coresight.sh
>> +ARGS=""
>> +DATV="out"
>> +DATA="$DATD/perf-$TEST-$DATV.data"
>> +
>> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
>> +
>> +perf_dump_aux_verify "$DATA" 10 10 10
>> +
>> +err=$?
>> +exit $err
>> diff --git a/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh b/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
>> new file mode 100755
>> index 000000000000..8972af835016
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
>> @@ -0,0 +1,18 @@
>> +#!/bin/sh -e
>> +# Coresight / Memcpy 16k 10 Threads
>> +
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +TEST="memcpy_thread"
>> +. $(dirname $0)/lib/coresight.sh
>> +ARGS="16 10 1"
>> +DATV="16k_10"
>> +DATA="$DATD/perf-$TEST-$DATV.data"
>> +
>> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
>> +
>> +perf_dump_aux_verify "$DATA" 10 10 10
>> +
>> +err=$?
>> +exit $err
>> diff --git a/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh b/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
>> new file mode 100755
>> index 000000000000..5b468901f89b
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
>> @@ -0,0 +1,19 @@
>> +#!/bin/sh -e
>> +# Coresight / Thread Loop 10 Threads - Check TID
>> +
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +TEST="thread_loop"
>> +. $(dirname $0)/lib/coresight.sh
>> +ARGS="10 1"
>> +DATV="check-tid-10th"
>> +DATA="$DATD/perf-$TEST-$DATV.data"
>> +STDO="$DATD/perf-$TEST-$DATV.stdout"
>> +
>> +SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO
>> +
>> +perf_dump_aux_tid_verify "$DATA" "$STDO"
>> +
>> +err=$?
>> +exit $err
>> diff --git a/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh b/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
>> new file mode 100755
>> index 000000000000..f8b7abd3aa03
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
>> @@ -0,0 +1,19 @@
>> +#!/bin/sh -e
>> +# Coresight / Thread Loop 2 Threads - Check TID
>> +
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +TEST="thread_loop"
>> +. $(dirname $0)/lib/coresight.sh
>> +ARGS="2 20"
>> +DATV="check-tid-2th"
>> +DATA="$DATD/perf-$TEST-$DATV.data"
>> +STDO="$DATD/perf-$TEST-$DATV.stdout"
>> +
>> +SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO
>> +
>> +perf_dump_aux_tid_verify "$DATA" "$STDO"
>> +
>> +err=$?
>> +exit $err
>> diff --git a/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh b/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
>> new file mode 100755
>> index 000000000000..c985dfb025c2
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
>> @@ -0,0 +1,18 @@
>> +#!/bin/sh -e
>> +# Coresight / Unroll Loop Thread 10
>> +
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +TEST="unroll_loop_thread"
>> +. $(dirname $0)/lib/coresight.sh
>> +ARGS="10"
>> +DATV="10"
>> +DATA="$DATD/perf-$TEST-$DATV.data"
>> +
>> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
>> +
>> +perf_dump_aux_verify "$DATA" 10 10 10
>> +
>> +err=$?
>> +exit $err
>> diff --git a/tools/perf/tests/shell/lib/coresight.sh b/tools/perf/tests/shell/lib/coresight.sh
>> new file mode 100644
>> index 000000000000..6a611b073f02
>> --- /dev/null
>> +++ b/tools/perf/tests/shell/lib/coresight.sh
>> @@ -0,0 +1,130 @@
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Carsten Haitzler <[email protected]>, 2021
>> +
>> +# This is sourced from a driver script so no need for #!/bin... etc. at the
>> +# top - the assumption below is that it runs as part of sourcing after the
>> +# test sets up some basic env vars to say what it is.
>> +
>> +# perf record options for the perf tests to use
>> +PERFRECMEM="-m ,128M"
>> +PERFRECOPT="$PERFRECMEM -e cs_etm//u"
>> +
>> +# These tests need to be run as root or coresight won't allow large buffers
>> +# and will not collect proper data
>> +UID=`id -u`
>> +if test "$UID" -ne 0; then
>> + echo "Not running as root... skip"
>> + exit 2
>> +fi
>> +
>> +TOOLS=$(dirname $0)
>> +DIR="$TOOLS/coresight/$TEST"
>> +BIN="$DIR/$TEST"
>> +# If the test tool/binary does not exist and is executable then skip the test
>> +if ! test -x "$BIN"; then exit 2; fi
>> +DATD="."
>> +# If the data dir env is set then make the data dir use that instead of ./
>> +if test -n "$PERF_TEST_CORESIGHT_DATADIR"; then
>> + DATD="$PERF_TEST_CORESIGHT_DATADIR";
>> +fi
>> +# If the stat dir env is set then make the data dir use that instead of ./
>> +STATD="."
>> +if test -n "$PERF_TEST_CORESIGHT_STATDIR"; then
>> + STATD="$PERF_TEST_CORESIGHT_STATDIR";
>> +fi
>> +
>> +# Called if the test fails - error code 2
>> +err() {
>> + echo "$1"
>> + exit 1
>> +}
>> +
>> +# Check that some statistics from our perf
>> +check_val_min() {
>> + STATF="$4"
>> + if test "$2" -lt "$3"; then
>> + echo ", FAILED" >> "$STATF"
>> + err "Sanity check number of $1 is too low ($2 < $3)"
>> + fi
>> +}
>> +
>> +perf_dump_aux_verify() {
>> + # Some basic checking that the AUX chunk contains some sensible data
>> + # to see that we are recording something and at least a minimum
>> + # amount of it. We should almost always see F3 atoms in just about
>> + # anything but certainly we will see some trace info and async atom
>> + # chunks.
>> + DUMP="$DATD/perf-tmp-aux-dump.txt"
>> + perf report --stdio --dump -i "$1" | \
>> + grep -o -e I_ATOM_F3 -e I_ASYNC -e I_TRACE_INFO > "$DUMP"
>> + # Simply count how many of these atoms we find to see that we are
>> + # producing a reasonable amount of data - exact checks are not sane
>> + # as this is a lossy process where we may lose some blocks and the
>> + # compiler may produce different code depending on the compiler and
>> + # optimization options, so this is rough just to see if we're
>> + # either missing almost all the data or all of it
>> + ATOM_F3_NUM=`grep I_ATOM_F3 "$DUMP" | wc -l`
>> + ATOM_ASYNC_NUM=`grep I_ASYNC "$DUMP" | wc -l`
>> + ATOM_TRACE_INFO_NUM=`grep I_TRACE_INFO "$DUMP" | wc -l`
>> + rm -f "$DUMP"
>> +
>> + # Arguments provide minimums for a pass
>> + CHECK_F3_MIN="$2"
>> + CHECK_ASYNC_MIN="$3"
>> + CHECK_TRACE_INFO_MIN="$4"
>> +
>> + # Write out statistics, so over time you can track results to see if
>> + # there is a pattern - for example we have less "noisy" results that
>> + # produce more consistent amounts of data each run, to see if over
>> + # time any techinques to minimize data loss are having an effect or
>> + # not
>> + STATF="$STATD/stats-$TEST-$DATV.csv"
>> + if ! test -f "$STATF"; then
>> + echo "ATOM F3 Count, Minimum, ATOM ASYNC Count, Minimum, TRACE INFO Count, Minimum" > "$STATF"
>> + fi
>> + echo -n "$ATOM_F3_NUM, $CHECK_F3_MIN, $ATOM_ASYNC_NUM, $CHECK_ASYNC_MIN, $ATOM_TRACE_INFO_NUM, $CHECK_TRACE_INFO_MIN" >> "$STATF"
>> +
>> + # Actually check to see if we passed or failed.
>> + check_val_min "ATOM_F3" "$ATOM_F3_NUM" "$CHECK_F3_MIN" "$STATF"
>> + check_val_min "ASYNC" "$ATOM_ASYNC_NUM" "$CHECK_ASYNC_MIN" "$STATF"
>> + check_val_min "TRACE_INFO" "$ATOM_TRACE_INFO_NUM" "$CHECK_TRACE_INFO_MIN" "$STATF"
>> + echo ", Ok" >> "$STATF"
>> +}
>> +
>> +perf_dump_aux_tid_verify() {
>> + # Specifically crafted test will produce a list of Tread ID's to
>> + # stdout that need to be checked to see that they have had trace
>> + # info collected in AUX blocks in the perf data. This will go
>> + # through all the TID's that are listed as CID=0xabcdef and see
>> + # that all the Thread IDs the test tool reports are in the perf
>> + # data AUX chunks
>> +
>> + # The TID test tools will print a TID per stdout line that are being
>> + # tested
>> + TIDS=`cat "$2"`
>> + # Scan the perf report to find the TIDs that are actually CID in hex
>> + # and build a list of the ones found
>> + FOUND_TIDS=`perf report --stdio --dump -i "$1" | \
>> + grep -o "CID=0x[0-9a-z]\+" | sed 's/CID=//g' | \
>> + uniq | sort | uniq`
>> +
>> + # Iterate over the list of TIDs that the test says it has and find
>> + # them in the TIDs found in the perf report
>> + MISSING=""
>> + for TID2 in $TIDS; do
>> + FOUND=""
>> + for TIDHEX in $FOUND_TIDS; do
>> + TID=`printf "%i" $TIDHEX`
>> + if test "$TID" -eq "$TID2"; then
>> + FOUND="y"
>> + break
>> + fi
>> + done
>> + if test -z "$FOUND"; then
>> + MISSING="$MISSING $TID"
>> + fi
>> + done
>> + if test -n "$MISSING"; then
>> + err "Thread IDs $MISSING not found in perf AUX data"
>> + fi
>> +}
>> --
>> 2.32.0
>>
On 5/30/22 17:47, Mathieu Poirier wrote:
> On Mon, 30 May 2022 at 10:27, Mathieu Poirier
> <[email protected]> wrote:
>>
>> On Wed, Mar 09, 2022 at 12:28:59PM +0000, [email protected] wrote:
>>> From: Carsten Haitzler <[email protected]>
>>>
>>> This adds a test harness and tests to run perf record and examine the
>>> resuling output when coresight is enabled on arm64 and check the
>>> resulting quality of the output as part of perf test. These tests use
>>> various tools to produce output from perf record then measure some key
>>> specific aspects of that data to see if the data exists at all and
>>> contains key aspects such as measuring some data for every thread of
>>> a test or produces sufficient data for large exeuction runs of a large
>>> executable. etc.
>>>
>>> Signed-off-by: Carsten Haitzler <[email protected]>
>>> ---
>>> MAINTAINERS | 4 +
>>> tools/perf/.gitignore | 6 +-
>>> tools/perf/Documentation/arm-coresight.txt | 140 ++++++++++++++++++
>>> tools/perf/Makefile.perf | 14 +-
>>> tools/perf/tests/shell/coresight/Makefile | 30 ++++
>>> .../tests/shell/coresight/Makefile.miniconfig | 23 +++
>>> .../shell/coresight/asm_pure_loop/.gitignore | 1 +
>>> .../shell/coresight/asm_pure_loop/Makefile | 30 ++++
>>> .../coresight/asm_pure_loop/asm_pure_loop.S | 28 ++++
>>> .../shell/coresight/memcpy_thread/.gitignore | 1 +
>>> .../shell/coresight/memcpy_thread/Makefile | 29 ++++
>>> .../coresight/memcpy_thread/memcpy_thread.c | 79 ++++++++++
>>> .../shell/coresight/thread_loop/.gitignore | 1 +
>>> .../shell/coresight/thread_loop/Makefile | 29 ++++
>>> .../shell/coresight/thread_loop/thread_loop.c | 86 +++++++++++
>>> .../coresight/unroll_loop_thread/.gitignore | 1 +
>>> .../coresight/unroll_loop_thread/Makefile | 29 ++++
>>> .../unroll_loop_thread/unroll_loop_thread.c | 74 +++++++++
>>> .../tests/shell/coresight_asm_pure_loop.sh | 18 +++
>>> .../shell/coresight_memcpy_thread_16k_10.sh | 18 +++
>>> .../coresight_thread_loop_check_tid_10.sh | 19 +++
>>> .../coresight_thread_loop_check_tid_2.sh | 19 +++
>>> .../shell/coresight_unroll_loop_thread_10.sh | 18 +++
>>> tools/perf/tests/shell/lib/coresight.sh | 130 ++++++++++++++++
>>> 24 files changed, 823 insertions(+), 4 deletions(-)
>>
>> As Leo pointed out this is a big patch and hard to digest intellectually.
>>
>>> create mode 100644 tools/perf/Documentation/arm-coresight.txt
>>> create mode 100644 tools/perf/tests/shell/coresight/Makefile
>>> create mode 100644 tools/perf/tests/shell/coresight/Makefile.miniconfig
>>> create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
>>> create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
>>> create mode 100644 tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
>>> create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
>>> create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/Makefile
>>> create mode 100644 tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
>>> create mode 100644 tools/perf/tests/shell/coresight/thread_loop/.gitignore
>>> create mode 100644 tools/perf/tests/shell/coresight/thread_loop/Makefile
>>> create mode 100644 tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
>>> create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
>>> create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
>>> create mode 100644 tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
>>> create mode 100755 tools/perf/tests/shell/coresight_asm_pure_loop.sh
>>> create mode 100755 tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
>>> create mode 100755 tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
>>> create mode 100755 tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
>>> create mode 100755 tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
>>> create mode 100644 tools/perf/tests/shell/lib/coresight.sh
>>>
>>> diff --git a/MAINTAINERS b/MAINTAINERS
>>> index 673c7124ca82..18cc20609f2e 100644
>>> --- a/MAINTAINERS
>>> +++ b/MAINTAINERS
>>> @@ -1918,10 +1918,14 @@ F: drivers/hwtracing/coresight/*
>>> F: include/dt-bindings/arm/coresight-cti-dt.h
>>> F: include/linux/coresight*
>>> F: samples/coresight/*
>>> +F: tools/perf/Documentation/arm-coresight.txt
>>> F: tools/perf/arch/arm/util/auxtrace.c
>>> F: tools/perf/arch/arm/util/cs-etm.c
>>> F: tools/perf/arch/arm/util/cs-etm.h
>>> F: tools/perf/arch/arm/util/pmu.c
>>> +F: tools/perf/tests/shell/coresight_*
>>> +F: tools/perf/tests/shell/tools/Makefile
>>> +F: tools/perf/tests/shell/tools/coresight/*
>>> F: tools/perf/util/cs-etm-decoder/*
>>> F: tools/perf/util/cs-etm.*
>>>
>>> diff --git a/tools/perf/.gitignore b/tools/perf/.gitignore
>>> index 20b8ab984d5f..138c679ecacd 100644
>>> --- a/tools/perf/.gitignore
>>> +++ b/tools/perf/.gitignore
>>> @@ -15,8 +15,9 @@ perf*.1
>>> perf*.xml
>>> perf*.html
>>> common-cmds.h
>>> -perf.data
>>> -perf.data.old
>>> +perf*.data
>>> +perf*.data.old
>>> +stats-*.csv
>>> output.svg
>>> perf-archive
>>> perf-with-kcore
>>> @@ -30,6 +31,7 @@ config.mak.autogen
>>> *-flex.*
>>> *.pyc
>>> *.pyo
>>> +*.stdout
>>> .config-detected
>>> util/intel-pt-decoder/inat-tables.c
>>> arch/*/include/generated/
>>> diff --git a/tools/perf/Documentation/arm-coresight.txt b/tools/perf/Documentation/arm-coresight.txt
>>> new file mode 100644
>>> index 000000000000..3a9e6c573c58
>>> --- /dev/null
>>> +++ b/tools/perf/Documentation/arm-coresight.txt
>>
>> I think it would be best to keep all the coresight documentation under the
>> current coresight documentation repository[1]. That way all the information on
>> coresight can be found in a central place.
>>
>> Some part of what is added by this patch is redundant with what is currently
>> available in [1]. Other parts are tests specific and should be added under
>> something like "coresight-perf-test.rst".
>>
>> Thanks,
>> Mathieu
>>
>> [1]. Documentation/trace/coresight/
>>
>
> I forgot... Please add a proper cover letter for this patchset.
ok - sure. next round.
>>
>>> @@ -0,0 +1,140 @@
>>> +Arm Coresight Support
>>> +=====================
>>> +
>>> +Coresight is a feature of some Arm based processors that allows for
>>> +debugging. One of the things it can do is trace every instruction
>>> +executed and remotely expose that information in a hardware compressed
>>> +stream. Perf is able to locally access that stream and store it to the
>>> +output perf data files. This stream can then be later decoded to give the
>>> +instructions that were traced for debugging or profiling purposes. You
>>> +can log such data with a perf record command like:
>>> +
>>> + perf record -e cs_etm//u testbinary
>>> +
>>> +This would run some test binary (testbinary) until it exits and record
>>> +a perf.data trace file. That file would have AUX sections if coresight
>>> +is working correctly. You can dump the content of this file as
>>> +readable text with a command like:
>>> +
>>> + perf report --stdio --dump -i perf.data
>>> +
>>> +You should find some sections of this file have AUX data blocks like:
>>> +
>>> + 0x1e78 [0x30]: PERF_RECORD_AUXTRACE size: 0x11dd0 offset: 0 ref: 0x1b614fc1061b0ad1 idx: 0 tid: 531230 cpu: -1
>>> +
>>> + . ... CoreSight ETM Trace data: size 73168 bytes
>>> + Idx:0; ID:10; I_ASYNC : Alignment Synchronisation.
>>> + Idx:12; ID:10; I_TRACE_INFO : Trace Info.; INFO=0x0 { CC.0 }
>>> + Idx:17; ID:10; I_ADDR_L_64IS0 : Address, Long, 64 bit, IS0.; Addr=0x0000000000000000;
>>> + Idx:26; ID:10; I_TRACE_ON : Trace On.
>>> + Idx:27; ID:10; I_ADDR_CTXT_L_64IS0 : Address & Context, Long, 64 bit, IS0.; Addr=0x0000FFFFB6069140; Ctxt: AArch64,EL0, NS;
>>> + Idx:38; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
>>> + Idx:39; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
>>> + Idx:40; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEEEEEEEEEEEEEE
>>> + Idx:41; ID:10; I_ATOM_F6 : Atom format 6.; EEEEEEEEEEEN
>>> + ...
>>> +
>>> +If you see these above, then your system is tracing coresight data
>>> +correctly.
>>> +
>>> +To compile perf with coresight support in the perf directory do
>>> +
>>> + make CORESIGHT=1
>>> +
>>> +This will compile the perf tool with coresight support as well as
>>> +build some small test binaries for perf test. This requires you also
>>> +be compiling for 64bit Arm (ARM64/aarch64). The tools run as part of
>>> +perf coresight tracing are in tests/shell/tools/coresight.
>>> +
>>> +You will also want coresight support enabled in your kernel config.
>>> +Ensure it is enabled with:
>>> +
>>> + CONFIG_CORESIGHT=y
>>> +
>>> +There are various other coresight options you probably also want
>>> +enabled like:
>>> +
>>> + CONFIG_CORESIGHT_LINKS_AND_SINKS=y
>>> + CONFIG_CORESIGHT_LINK_AND_SINK_TMC=y
>>> + CONFIG_CORESIGHT_CATU=y
>>> + CONFIG_CORESIGHT_SINK_TPIU=y
>>> + CONFIG_CORESIGHT_SINK_ETBV10=y
>>> + CONFIG_CORESIGHT_SOURCE_ETM4X=y
>>> + CONFIG_CORESIGHT_STM=y
>>> + CONFIG_CORESIGHT_CPU_DEBUG=y
>>> + CONFIG_CORESIGHT_CTI=y
>>> + CONFIG_CORESIGHT_CTI_INTEGRATION_REGS=y
>>> +
>>> +Please refer to the kernel configuration help for more information.
>>> +
>>> +Perf test - Verify kernel and userspace perf coresight work
>>> +===========================================================
>>> +
>>> +When you run perf test, it will do a lot of self tests. Some of those
>>> +tests will cover Coresight (only if enabled and on ARM64). You
>>> +generally would run perf test from the tools/perf directory in the
>>> +kernel tree. Some tests will check some internal perf support like:
>>> +
>>> + Check Arm CoreSight trace data recording and synthesized samples
>>> +
>>> +Some others will actually use perf record and some test binaries that
>>> +are in tests/shell/tools/coresight and will collect traces to ensure a
>>> +minimum level of functionality is met. The scripts that launch these
>>> +tests are in tests/shell. These will all look like:
>>> +
>>> + Coresight / Memcpy 1M 25 Threads
>>> + Coresight / Unroll Loop Thread 2
>>> + ...
>>> +
>>> +These perf record tests will not run if the tool binaries do not exist
>>> +in tests/shell/tools/coresight/*/ and will be skipped. If you do not
>>> +have coresight support in hardware then either do not build perf with
>>> +coresight support or remove these binaries in order to not have these
>>> +tests fail and have them skip instead.
>>> +
>>> +These tests will log historical results in the current working
>>> +directory (e.g. tools/perf) and will be named stats-*.csv like:
>>> +
>>> + stats-asm_pure_loop-out.csv
>>> + stats-bubble_sort-random.csv
>>> + ...
>>> +
>>> +These statistic files log some aspects of the AUX data sections in
>>> +the perf data output counting some numbers of certain encodings (a
>>> +good way to know that it's working in a very simple way). One problem
>>> +with coresight is that given a large enough amount of data needing to
>>> +be logged, some of it can be lost due to the processor not waking up
>>> +in time to read out all the data from buffers etc.. You will notice
>>> +that the amount of data collected can vary a lot per run of perf test.
>>> +If you wish to see how this changes over time, simply run perf test
>>> +multiple times and all these csv files will have more and more data
>>> +appended to it that you can later examine, graph and otherwise use to
>>> +figure out if things have become worse or better.
>>> +
>>> +Be aware that amny of these tests take quite a while to run, specifically
>>> +in processing the perf data file and dumping contents to then examine what
>>> +is inside.
>>> +
>>> +You can change where these csv logs are stored by setting the
>>> +PERF_TEST_CORESIGHT_STATDIR environment variable before running perf
>>> +test like:
>>> +
>>> + export PERF_TEST_CORESIGHT_STATDIR=/var/tmp
>>> + perf test
>>> +
>>> +They will also store resulting perf output data in the current
>>> +directory for later inspection like:
>>> +
>>> + perf-memcpy-1m.data
>>> + perf-thread_loop-2th.data
>>> + ...
>>> +
>>> +You can alter where the perf data files are stored by setting the
>>> +PERF_TEST_CORESIGHT_DATADIR environment variable such as:
>>> +
>>> + PERF_TEST_CORESIGHT_DATADIR=/var/tmp
>>> + perf test
>>> +
>>> +You may wish to set these above environment variables if you which to
>>> +keep the output of tests outside of the current working directory for
>>> +longer term storage and examination.
>>> diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
>>> index ac861e42c8f7..b97db83992e0 100644
>>> --- a/tools/perf/Makefile.perf
>>> +++ b/tools/perf/Makefile.perf
>>> @@ -630,7 +630,15 @@ sync_file_range_tbls := $(srctree)/tools/perf/trace/beauty/sync_file_range.sh
>>> $(sync_file_range_arrays): $(linux_uapi_dir)/fs.h $(sync_file_range_tbls)
>>> $(Q)$(SHELL) '$(sync_file_range_tbls)' $(linux_uapi_dir) > $@
>>>
>>> -all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS)
>>> +TESTS_CORESIGHT_DIR := $(srctree)/tools/perf/tests/shell/coresight
>>> +
>>> +tests-coresight-targets: FORCE
>>> + $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR)
>>> +
>>> +tests-coresight-targets-clean:
>>> + $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR) clean
>>> +
>>> +all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS) tests-coresight-targets
>>>
>>> # Create python binding output directory if not already present
>>> _dummy := $(shell [ -d '$(OUTPUT)python' ] || mkdir -p '$(OUTPUT)python')
>>> @@ -1020,6 +1028,7 @@ install-tests: all install-gtk
>>> $(INSTALL) tests/shell/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell'; \
>>> $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'; \
>>> $(INSTALL) tests/shell/lib/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'
>>> + $(Q)$(MAKE) -C tests/shell/coresight install-tests
>>>
>>> install-bin: install-tools install-tests install-traceevent-plugins
>>>
>>> @@ -1088,7 +1097,7 @@ endif # BUILD_BPF_SKEL
>>> bpf-skel-clean:
>>> $(call QUIET_CLEAN, bpf-skel) $(RM) -r $(SKEL_TMP_OUT) $(SKELETONS)
>>>
>>> -clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean
>>> +clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean tests-coresight-targets-clean
>>> $(call QUIET_CLEAN, core-objs) $(RM) $(LIBPERF_A) $(OUTPUT)perf-archive $(OUTPUT)perf-with-kcore $(OUTPUT)perf-iostat $(LANG_BINDINGS)
>>> $(Q)find $(if $(OUTPUT),$(OUTPUT),.) -name '*.o' -delete -o -name '\.*.cmd' -delete -o -name '\.*.d' -delete
>>> $(Q)$(RM) $(OUTPUT).config-detected
>>> @@ -1155,5 +1164,6 @@ FORCE:
>>> .PHONY: shell_compatibility_test please_set_SHELL_PATH_to_a_more_modern_shell
>>> .PHONY: $(GIT-HEAD-PHONY) TAGS tags cscope FORCE prepare
>>> .PHONY: libtraceevent_plugins archheaders
>>> +.PHONY: $(TESTS_CORESIGHT_TARGETS)
>>>
>>> endif # force_fixdep
>>> diff --git a/tools/perf/tests/shell/coresight/Makefile b/tools/perf/tests/shell/coresight/Makefile
>>> new file mode 100644
>>> index 000000000000..dda99aeac158
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight/Makefile
>>> @@ -0,0 +1,30 @@
>>> +# SPDX-License-Identifier: GPL-2.0-only
>>> +# Carsten Haitzler <[email protected]>, 2021
>>> +include ../../../../../tools/scripts/Makefile.include
>>> +include ../../../../../tools/scripts/Makefile.arch
>>> +include ../../../../../tools/scripts/utilities.mak
>>> +
>>> +SUBDIRS = \
>>> + asm_pure_loop \
>>> + thread_loop \
>>> + memcpy_thread \
>>> + unroll_loop_thread
>>> +
>>> +all: $(SUBDIRS)
>>> +$(SUBDIRS):
>>> + $(Q)$(MAKE) -C $@
>>> +
>>> +INSTALLDIRS = $(SUBDIRS:%=install-%)
>>> +
>>> +install-tests: $(INSTALLDIRS)
>>> +$(INSTALLDIRS):
>>> + $(Q)$(MAKE) -C $(@:install-%=%) install-tests
>>> +
>>> +CLEANDIRS = $(SUBDIRS:%=clean-%)
>>> +
>>> +clean: $(CLEANDIRS)
>>> +$(CLEANDIRS):
>>> + $(Q)$(MAKE) -C $(@:clean-%=%) clean >/dev/null
>>> +
>>> +.PHONY: all clean $(SUBDIRS) $(CLEANDIRS) $(INSTALLDIRS)
>>> +
>>> diff --git a/tools/perf/tests/shell/coresight/Makefile.miniconfig b/tools/perf/tests/shell/coresight/Makefile.miniconfig
>>> new file mode 100644
>>> index 000000000000..893c12685fed
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight/Makefile.miniconfig
>>> @@ -0,0 +1,23 @@
>>> +# SPDX-License-Identifier: GPL-2.0-only
>>> +# Carsten Haitzler <[email protected]>, 2021
>>> +
>>> +ifndef DESTDIR
>>> +prefix ?= $(HOME)
>>> +endif
>>> +
>>> +DESTDIR_SQ = $(subst ','\'',$(DESTDIR))
>>> +perfexecdir = libexec/perf-core
>>> +perfexec_instdir = $(perfexecdir)
>>> +
>>> +ifneq ($(filter /%,$(firstword $(perfexecdir))),)
>>> +perfexec_instdir = $(perfexecdir)
>>> +else
>>> +perfexec_instdir = $(prefix)/$(perfexecdir)
>>> +endif
>>> +
>>> +perfexec_instdir_SQ = $(subst ','\'',$(perfexec_instdir))
>>> +INSTALL = install
>>> +
>>> +include ../../../../../scripts/Makefile.include
>>> +include ../../../../../scripts/Makefile.arch
>>> +include ../../../../../scripts/utilities.mak
>>> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
>>> new file mode 100644
>>> index 000000000000..468673ac32e8
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
>>> @@ -0,0 +1 @@
>>> +asm_pure_loop
>>> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
>>> new file mode 100644
>>> index 000000000000..10c5a60cb71c
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
>>> @@ -0,0 +1,30 @@
>>> +# SPDX-License-Identifier: GPL-2.0
>>> +# Carsten Haitzler <[email protected]>, 2021
>>> +
>>> +include ../Makefile.miniconfig
>>> +
>>> +BIN=asm_pure_loop
>>> +LIB=
>>> +
>>> +all: $(BIN)
>>> +
>>> +$(BIN): $(BIN).S
>>> +ifdef CORESIGHT
>>> +ifeq ($(ARCH),arm64)
>>> + $(Q)$(CC) $(BIN).S -nostdlib -static -o $(BIN) $(LIB)
>>> +endif
>>> +endif
>>> +
>>> +install-tests: all
>>> +ifdef CORESIGHT
>>> +ifeq ($(ARCH),arm64)
>>> + $(call QUIET_INSTALL, tests) \
>>> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
>>> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
>>> +endif
>>> +endif
>>> +
>>> +clean:
>>> + $(Q)$(RM) -f $(BIN)
>>> +
>>> +.PHONY: all clean install-tests
>>> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
>>> new file mode 100644
>>> index 000000000000..75cf084a927d
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
>>> @@ -0,0 +1,28 @@
>>> +/* SPDX-License-Identifier: GPL-2.0 */
>>> +/* Tamas Zsoldos <[email protected]>, 2021 */
>>> +
>>> +.globl _start
>>> +_start:
>>> + mov x0, 0x0000ffff
>>> + mov x1, xzr
>>> +loop:
>>> + nop
>>> + nop
>>> + cbnz x1, noskip
>>> + nop
>>> + nop
>>> + adrp x2, skip
>>> + add x2, x2, :lo12:skip
>>> + br x2
>>> + nop
>>> + nop
>>> +noskip:
>>> + nop
>>> + nop
>>> +skip:
>>> + sub x0, x0, 1
>>> + cbnz x0, loop
>>> +
>>> + mov x0, #0
>>> + mov x8, #93 // __NR_exit syscall
>>> + svc #0
>>> diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
>>> new file mode 100644
>>> index 000000000000..f8217e56091e
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore
>>> @@ -0,0 +1 @@
>>> +memcpy_thread
>>> diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/Makefile b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile
>>> new file mode 100644
>>> index 000000000000..e2604cfae74b
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile
>>> @@ -0,0 +1,29 @@
>>> +# SPDX-License-Identifier: GPL-2.0
>>> +# Carsten Haitzler <[email protected]>, 2021
>>> +include ../Makefile.miniconfig
>>> +
>>> +BIN=memcpy_thread
>>> +LIB=-pthread
>>> +
>>> +all: $(BIN)
>>> +
>>> +$(BIN): $(BIN).c
>>> +ifdef CORESIGHT
>>> +ifeq ($(ARCH),arm64)
>>> + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
>>> +endif
>>> +endif
>>> +
>>> +install-tests: all
>>> +ifdef CORESIGHT
>>> +ifeq ($(ARCH),arm64)
>>> + $(call QUIET_INSTALL, tests) \
>>> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
>>> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
>>> +endif
>>> +endif
>>> +
>>> +clean:
>>> + $(Q)$(RM) -f $(BIN)
>>> +
>>> +.PHONY: all clean install-tests
>>> diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
>>> new file mode 100644
>>> index 000000000000..a7e169d1bf64
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c
>>> @@ -0,0 +1,79 @@
>>> +// SPDX-License-Identifier: GPL-2.0
>>> +// Carsten Haitzler <[email protected]>, 2021
>>> +#include <stdio.h>
>>> +#include <stdlib.h>
>>> +#include <unistd.h>
>>> +#include <string.h>
>>> +#include <pthread.h>
>>> +
>>> +struct args {
>>> + unsigned long loops;
>>> + unsigned long size;
>>> + pthread_t th;
>>> + void *ret;
>>> +};
>>> +
>>> +static void *thrfn(void *arg)
>>> +{
>>> + struct args *a = arg;
>>> + unsigned long i, len = a->loops;
>>> + unsigned char *src, *dst;
>>> +
>>> + src = malloc(a->size * 1024);
>>> + dst = malloc(a->size * 1024);
>>> + if ((!src) || (!dst)) {
>>> + printf("ERR: Can't allocate memory\n");
>>> + exit(1);
>>> + }
>>> + for (i = 0; i < len; i++)
>>> + memcpy(dst, src, a->size * 1024);
>>> +}
>>> +
>>> +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
>>> +{
>>> + pthread_t t;
>>> + pthread_attr_t attr;
>>> +
>>> + pthread_attr_init(&attr);
>>> + pthread_create(&t, &attr, fn, arg);
>>> + return t;
>>> +}
>>> +
>>> +int main(int argc, char **argv)
>>> +{
>>> + unsigned long i, len, size, thr;
>>> + pthread_t threads[256];
>>> + struct args args[256];
>>> + long long v;
>>> +
>>> + if (argc < 4) {
>>> + printf("ERR: %s [copysize Kb] [numthreads] [numloops (hundreds)]\n", argv[0]);
>>> + exit(1);
>>> + }
>>> +
>>> + v = atoll(argv[1]);
>>> + if ((v < 1) || (v > (1024 * 1024))) {
>>> + printf("ERR: max memory 1GB (1048576 KB)\n");
>>> + exit(1);
>>> + }
>>> + size = v;
>>> + thr = atol(argv[2]);
>>> + if ((thr < 1) || (thr > 256)) {
>>> + printf("ERR: threads 1-256\n");
>>> + exit(1);
>>> + }
>>> + v = atoll(argv[3]);
>>> + if ((v < 1) || (v > 40000000000ll)) {
>>> + printf("ERR: loops 1-40000000000 (hundreds)\n");
>>> + exit(1);
>>> + }
>>> + len = v * 100;
>>> + for (i = 0; i < thr; i++) {
>>> + args[i].loops = len;
>>> + args[i].size = size;
>>> + args[i].th = new_thr(thrfn, &(args[i]));
>>> + }
>>> + for (i = 0; i < thr; i++)
>>> + pthread_join(args[i].th, &(args[i].ret));
>>> + return 0;
>>> +}
>>> diff --git a/tools/perf/tests/shell/coresight/thread_loop/.gitignore b/tools/perf/tests/shell/coresight/thread_loop/.gitignore
>>> new file mode 100644
>>> index 000000000000..6d4c33eaa9e8
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight/thread_loop/.gitignore
>>> @@ -0,0 +1 @@
>>> +thread_loop
>>> diff --git a/tools/perf/tests/shell/coresight/thread_loop/Makefile b/tools/perf/tests/shell/coresight/thread_loop/Makefile
>>> new file mode 100644
>>> index 000000000000..424df4e8b0e6
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight/thread_loop/Makefile
>>> @@ -0,0 +1,29 @@
>>> +# SPDX-License-Identifier: GPL-2.0
>>> +# Carsten Haitzler <[email protected]>, 2021
>>> +include ../Makefile.miniconfig
>>> +
>>> +BIN=thread_loop
>>> +LIB=-pthread
>>> +
>>> +all: $(BIN)
>>> +
>>> +$(BIN): $(BIN).c
>>> +ifdef CORESIGHT
>>> +ifeq ($(ARCH),arm64)
>>> + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
>>> +endif
>>> +endif
>>> +
>>> +install-tests: all
>>> +ifdef CORESIGHT
>>> +ifeq ($(ARCH),arm64)
>>> + $(call QUIET_INSTALL, tests) \
>>> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
>>> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
>>> +endif
>>> +endif
>>> +
>>> +clean:
>>> + $(Q)$(RM) -f $(BIN)
>>> +
>>> +.PHONY: all clean install-tests
>>> diff --git a/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
>>> new file mode 100644
>>> index 000000000000..c0158fac7d0b
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c
>>> @@ -0,0 +1,86 @@
>>> +// SPDX-License-Identifier: GPL-2.0
>>> +// Carsten Haitzler <[email protected]>, 2021
>>> +
>>> +// define this for gettid()
>>> +#define _GNU_SOURCE
>>> +
>>> +#include <stdio.h>
>>> +#include <stdlib.h>
>>> +#include <unistd.h>
>>> +#include <string.h>
>>> +#include <pthread.h>
>>> +#include <sys/syscall.h>
>>> +#ifndef SYS_gettid
>>> +// gettid is 178 on arm64
>>> +# define SYS_gettid 178
>>> +#endif
>>> +#define gettid() syscall(SYS_gettid)
>>> +
>>> +struct args {
>>> + unsigned int loops;
>>> + pthread_t th;
>>> + void *ret;
>>> +};
>>> +
>>> +static void *thrfn(void *arg)
>>> +{
>>> + struct args *a = arg;
>>> + int i = 0, len = a->loops;
>>> +
>>> + if (getenv("SHOW_TID")) {
>>> + unsigned long long tid = gettid();
>>> +
>>> + printf("%llu\n", tid);
>>> + }
>>> + asm volatile(
>>> + "loop:\n"
>>> + "add %[i], %[i], #1\n"
>>> + "cmp %[i], %[len]\n"
>>> + "blt loop\n"
>>> + : /* out */
>>> + : /* in */ [i] "r" (i), [len] "r" (len)
>>> + : /* clobber */
>>> + );
>>> + return (void *)(long)i;
>>> +}
>>> +
>>> +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
>>> +{
>>> + pthread_t t;
>>> + pthread_attr_t attr;
>>> +
>>> + pthread_attr_init(&attr);
>>> + pthread_create(&t, &attr, fn, arg);
>>> + return t;
>>> +}
>>> +
>>> +int main(int argc, char **argv)
>>> +{
>>> + unsigned int i, len, thr;
>>> + pthread_t threads[256];
>>> + struct args args[256];
>>> +
>>> + if (argc < 3) {
>>> + printf("ERR: %s [numthreads] [numloops (millions)]\n", argv[0]);
>>> + exit(1);
>>> + }
>>> +
>>> + thr = atoi(argv[1]);
>>> + if ((thr < 1) || (thr > 256)) {
>>> + printf("ERR: threads 1-256\n");
>>> + exit(1);
>>> + }
>>> + len = atoi(argv[2]);
>>> + if ((len < 1) || (len > 4000)) {
>>> + printf("ERR: max loops 4000 (millions)\n");
>>> + exit(1);
>>> + }
>>> + len *= 1000000;
>>> + for (i = 0; i < thr; i++) {
>>> + args[i].loops = len;
>>> + args[i].th = new_thr(thrfn, &(args[i]));
>>> + }
>>> + for (i = 0; i < thr; i++)
>>> + pthread_join(args[i].th, &(args[i].ret));
>>> + return 0;
>>> +}
>>> diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
>>> new file mode 100644
>>> index 000000000000..2cb4e996dbf3
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore
>>> @@ -0,0 +1 @@
>>> +unroll_loop_thread
>>> diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
>>> new file mode 100644
>>> index 000000000000..45ab2be8be92
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile
>>> @@ -0,0 +1,29 @@
>>> +# SPDX-License-Identifier: GPL-2.0
>>> +# Carsten Haitzler <[email protected]>, 2021
>>> +include ../Makefile.miniconfig
>>> +
>>> +BIN=unroll_loop_thread
>>> +LIB=-pthread
>>> +
>>> +all: $(BIN)
>>> +
>>> +$(BIN): $(BIN).c
>>> +ifdef CORESIGHT
>>> +ifeq ($(ARCH),arm64)
>>> + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB)
>>> +endif
>>> +endif
>>> +
>>> +install-tests: all
>>> +ifdef CORESIGHT
>>> +ifeq ($(ARCH),arm64)
>>> + $(call QUIET_INSTALL, tests) \
>>> + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)'; \
>>> + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/tools/$(BIN)/$(BIN)'
>>> +endif
>>> +endif
>>> +
>>> +clean:
>>> + $(Q)$(RM) -f $(BIN)
>>> +
>>> +.PHONY: all clean install-tests
>>> diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
>>> new file mode 100644
>>> index 000000000000..cb9d22c7dfb9
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c
>>> @@ -0,0 +1,74 @@
>>> +// SPDX-License-Identifier: GPL-2.0
>>> +// Carsten Haitzler <[email protected]>, 2021
>>> +#include <stdio.h>
>>> +#include <stdlib.h>
>>> +#include <unistd.h>
>>> +#include <string.h>
>>> +#include <pthread.h>
>>> +
>>> +struct args {
>>> + pthread_t th;
>>> + unsigned int in, out;
>>> + void *ret;
>>> +};
>>> +
>>> +static void *thrfn(void *arg)
>>> +{
>>> + struct args *a = arg;
>>> + unsigned int i, in = a->in;
>>> +
>>> + for (i = 0; i < 10000; i++) {
>>> + asm volatile (
>>> +// force an unroll of thia add instruction so we can test long runs of code
>>> +#define SNIP1 "add %[in], %[in], #1\n"
>>> +// 10
>>> +#define SNIP2 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1
>>> +// 100
>>> +#define SNIP3 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2
>>> +// 1000
>>> +#define SNIP4 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3
>>> +// 10000
>>> +#define SNIP5 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4
>>> +// 100000
>>> + SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5
>>> + : /* out */
>>> + : /* in */ [in] "r" (in)
>>> + : /* clobber */
>>> + );
>>> + }
>>> +}
>>> +
>>> +static pthread_t new_thr(void *(*fn) (void *arg), void *arg)
>>> +{
>>> + pthread_t t;
>>> + pthread_attr_t attr;
>>> +
>>> + pthread_attr_init(&attr);
>>> + pthread_create(&t, &attr, fn, arg);
>>> + return t;
>>> +}
>>> +
>>> +int main(int argc, char **argv)
>>> +{
>>> + unsigned int i, thr;
>>> + pthread_t threads[256];
>>> + struct args args[256];
>>> +
>>> + if (argc < 2) {
>>> + printf("ERR: %s [numthreads]\n", argv[0]);
>>> + exit(1);
>>> + }
>>> +
>>> + thr = atoi(argv[1]);
>>> + if ((thr > 256) || (thr < 1)) {
>>> + printf("ERR: threads 1-256\n");
>>> + exit(1);
>>> + }
>>> + for (i = 0; i < thr; i++) {
>>> + args[i].in = rand();
>>> + args[i].th = new_thr(thrfn, &(args[i]));
>>> + }
>>> + for (i = 0; i < thr; i++)
>>> + pthread_join(args[i].th, &(args[i].ret));
>>> + return 0;
>>> +}
>>> diff --git a/tools/perf/tests/shell/coresight_asm_pure_loop.sh b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
>>> new file mode 100755
>>> index 000000000000..3f0dbefcad50
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
>>> @@ -0,0 +1,18 @@
>>> +#!/bin/sh -e
>>> +# Coresight / ASM Pure Loop
>>> +
>>> +# SPDX-License-Identifier: GPL-2.0
>>> +# Carsten Haitzler <[email protected]>, 2021
>>> +
>>> +TEST="asm_pure_loop"
>>> +. $(dirname $0)/lib/coresight.sh
>>> +ARGS=""
>>> +DATV="out"
>>> +DATA="$DATD/perf-$TEST-$DATV.data"
>>> +
>>> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
>>> +
>>> +perf_dump_aux_verify "$DATA" 10 10 10
>>> +
>>> +err=$?
>>> +exit $err
>>> diff --git a/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh b/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
>>> new file mode 100755
>>> index 000000000000..8972af835016
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight_memcpy_thread_16k_10.sh
>>> @@ -0,0 +1,18 @@
>>> +#!/bin/sh -e
>>> +# Coresight / Memcpy 16k 10 Threads
>>> +
>>> +# SPDX-License-Identifier: GPL-2.0
>>> +# Carsten Haitzler <[email protected]>, 2021
>>> +
>>> +TEST="memcpy_thread"
>>> +. $(dirname $0)/lib/coresight.sh
>>> +ARGS="16 10 1"
>>> +DATV="16k_10"
>>> +DATA="$DATD/perf-$TEST-$DATV.data"
>>> +
>>> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
>>> +
>>> +perf_dump_aux_verify "$DATA" 10 10 10
>>> +
>>> +err=$?
>>> +exit $err
>>> diff --git a/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh b/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
>>> new file mode 100755
>>> index 000000000000..5b468901f89b
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight_thread_loop_check_tid_10.sh
>>> @@ -0,0 +1,19 @@
>>> +#!/bin/sh -e
>>> +# Coresight / Thread Loop 10 Threads - Check TID
>>> +
>>> +# SPDX-License-Identifier: GPL-2.0
>>> +# Carsten Haitzler <[email protected]>, 2021
>>> +
>>> +TEST="thread_loop"
>>> +. $(dirname $0)/lib/coresight.sh
>>> +ARGS="10 1"
>>> +DATV="check-tid-10th"
>>> +DATA="$DATD/perf-$TEST-$DATV.data"
>>> +STDO="$DATD/perf-$TEST-$DATV.stdout"
>>> +
>>> +SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO
>>> +
>>> +perf_dump_aux_tid_verify "$DATA" "$STDO"
>>> +
>>> +err=$?
>>> +exit $err
>>> diff --git a/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh b/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
>>> new file mode 100755
>>> index 000000000000..f8b7abd3aa03
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight_thread_loop_check_tid_2.sh
>>> @@ -0,0 +1,19 @@
>>> +#!/bin/sh -e
>>> +# Coresight / Thread Loop 2 Threads - Check TID
>>> +
>>> +# SPDX-License-Identifier: GPL-2.0
>>> +# Carsten Haitzler <[email protected]>, 2021
>>> +
>>> +TEST="thread_loop"
>>> +. $(dirname $0)/lib/coresight.sh
>>> +ARGS="2 20"
>>> +DATV="check-tid-2th"
>>> +DATA="$DATD/perf-$TEST-$DATV.data"
>>> +STDO="$DATD/perf-$TEST-$DATV.stdout"
>>> +
>>> +SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO
>>> +
>>> +perf_dump_aux_tid_verify "$DATA" "$STDO"
>>> +
>>> +err=$?
>>> +exit $err
>>> diff --git a/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh b/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
>>> new file mode 100755
>>> index 000000000000..c985dfb025c2
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/coresight_unroll_loop_thread_10.sh
>>> @@ -0,0 +1,18 @@
>>> +#!/bin/sh -e
>>> +# Coresight / Unroll Loop Thread 10
>>> +
>>> +# SPDX-License-Identifier: GPL-2.0
>>> +# Carsten Haitzler <[email protected]>, 2021
>>> +
>>> +TEST="unroll_loop_thread"
>>> +. $(dirname $0)/lib/coresight.sh
>>> +ARGS="10"
>>> +DATV="10"
>>> +DATA="$DATD/perf-$TEST-$DATV.data"
>>> +
>>> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
>>> +
>>> +perf_dump_aux_verify "$DATA" 10 10 10
>>> +
>>> +err=$?
>>> +exit $err
>>> diff --git a/tools/perf/tests/shell/lib/coresight.sh b/tools/perf/tests/shell/lib/coresight.sh
>>> new file mode 100644
>>> index 000000000000..6a611b073f02
>>> --- /dev/null
>>> +++ b/tools/perf/tests/shell/lib/coresight.sh
>>> @@ -0,0 +1,130 @@
>>> +# SPDX-License-Identifier: GPL-2.0
>>> +# Carsten Haitzler <[email protected]>, 2021
>>> +
>>> +# This is sourced from a driver script so no need for #!/bin... etc. at the
>>> +# top - the assumption below is that it runs as part of sourcing after the
>>> +# test sets up some basic env vars to say what it is.
>>> +
>>> +# perf record options for the perf tests to use
>>> +PERFRECMEM="-m ,128M"
>>> +PERFRECOPT="$PERFRECMEM -e cs_etm//u"
>>> +
>>> +# These tests need to be run as root or coresight won't allow large buffers
>>> +# and will not collect proper data
>>> +UID=`id -u`
>>> +if test "$UID" -ne 0; then
>>> + echo "Not running as root... skip"
>>> + exit 2
>>> +fi
>>> +
>>> +TOOLS=$(dirname $0)
>>> +DIR="$TOOLS/coresight/$TEST"
>>> +BIN="$DIR/$TEST"
>>> +# If the test tool/binary does not exist and is executable then skip the test
>>> +if ! test -x "$BIN"; then exit 2; fi
>>> +DATD="."
>>> +# If the data dir env is set then make the data dir use that instead of ./
>>> +if test -n "$PERF_TEST_CORESIGHT_DATADIR"; then
>>> + DATD="$PERF_TEST_CORESIGHT_DATADIR";
>>> +fi
>>> +# If the stat dir env is set then make the data dir use that instead of ./
>>> +STATD="."
>>> +if test -n "$PERF_TEST_CORESIGHT_STATDIR"; then
>>> + STATD="$PERF_TEST_CORESIGHT_STATDIR";
>>> +fi
>>> +
>>> +# Called if the test fails - error code 2
>>> +err() {
>>> + echo "$1"
>>> + exit 1
>>> +}
>>> +
>>> +# Check that some statistics from our perf
>>> +check_val_min() {
>>> + STATF="$4"
>>> + if test "$2" -lt "$3"; then
>>> + echo ", FAILED" >> "$STATF"
>>> + err "Sanity check number of $1 is too low ($2 < $3)"
>>> + fi
>>> +}
>>> +
>>> +perf_dump_aux_verify() {
>>> + # Some basic checking that the AUX chunk contains some sensible data
>>> + # to see that we are recording something and at least a minimum
>>> + # amount of it. We should almost always see F3 atoms in just about
>>> + # anything but certainly we will see some trace info and async atom
>>> + # chunks.
>>> + DUMP="$DATD/perf-tmp-aux-dump.txt"
>>> + perf report --stdio --dump -i "$1" | \
>>> + grep -o -e I_ATOM_F3 -e I_ASYNC -e I_TRACE_INFO > "$DUMP"
>>> + # Simply count how many of these atoms we find to see that we are
>>> + # producing a reasonable amount of data - exact checks are not sane
>>> + # as this is a lossy process where we may lose some blocks and the
>>> + # compiler may produce different code depending on the compiler and
>>> + # optimization options, so this is rough just to see if we're
>>> + # either missing almost all the data or all of it
>>> + ATOM_F3_NUM=`grep I_ATOM_F3 "$DUMP" | wc -l`
>>> + ATOM_ASYNC_NUM=`grep I_ASYNC "$DUMP" | wc -l`
>>> + ATOM_TRACE_INFO_NUM=`grep I_TRACE_INFO "$DUMP" | wc -l`
>>> + rm -f "$DUMP"
>>> +
>>> + # Arguments provide minimums for a pass
>>> + CHECK_F3_MIN="$2"
>>> + CHECK_ASYNC_MIN="$3"
>>> + CHECK_TRACE_INFO_MIN="$4"
>>> +
>>> + # Write out statistics, so over time you can track results to see if
>>> + # there is a pattern - for example we have less "noisy" results that
>>> + # produce more consistent amounts of data each run, to see if over
>>> + # time any techinques to minimize data loss are having an effect or
>>> + # not
>>> + STATF="$STATD/stats-$TEST-$DATV.csv"
>>> + if ! test -f "$STATF"; then
>>> + echo "ATOM F3 Count, Minimum, ATOM ASYNC Count, Minimum, TRACE INFO Count, Minimum" > "$STATF"
>>> + fi
>>> + echo -n "$ATOM_F3_NUM, $CHECK_F3_MIN, $ATOM_ASYNC_NUM, $CHECK_ASYNC_MIN, $ATOM_TRACE_INFO_NUM, $CHECK_TRACE_INFO_MIN" >> "$STATF"
>>> +
>>> + # Actually check to see if we passed or failed.
>>> + check_val_min "ATOM_F3" "$ATOM_F3_NUM" "$CHECK_F3_MIN" "$STATF"
>>> + check_val_min "ASYNC" "$ATOM_ASYNC_NUM" "$CHECK_ASYNC_MIN" "$STATF"
>>> + check_val_min "TRACE_INFO" "$ATOM_TRACE_INFO_NUM" "$CHECK_TRACE_INFO_MIN" "$STATF"
>>> + echo ", Ok" >> "$STATF"
>>> +}
>>> +
>>> +perf_dump_aux_tid_verify() {
>>> + # Specifically crafted test will produce a list of Tread ID's to
>>> + # stdout that need to be checked to see that they have had trace
>>> + # info collected in AUX blocks in the perf data. This will go
>>> + # through all the TID's that are listed as CID=0xabcdef and see
>>> + # that all the Thread IDs the test tool reports are in the perf
>>> + # data AUX chunks
>>> +
>>> + # The TID test tools will print a TID per stdout line that are being
>>> + # tested
>>> + TIDS=`cat "$2"`
>>> + # Scan the perf report to find the TIDs that are actually CID in hex
>>> + # and build a list of the ones found
>>> + FOUND_TIDS=`perf report --stdio --dump -i "$1" | \
>>> + grep -o "CID=0x[0-9a-z]\+" | sed 's/CID=//g' | \
>>> + uniq | sort | uniq`
>>> +
>>> + # Iterate over the list of TIDs that the test says it has and find
>>> + # them in the TIDs found in the perf report
>>> + MISSING=""
>>> + for TID2 in $TIDS; do
>>> + FOUND=""
>>> + for TIDHEX in $FOUND_TIDS; do
>>> + TID=`printf "%i" $TIDHEX`
>>> + if test "$TID" -eq "$TID2"; then
>>> + FOUND="y"
>>> + break
>>> + fi
>>> + done
>>> + if test -z "$FOUND"; then
>>> + MISSING="$MISSING $TID"
>>> + fi
>>> + done
>>> + if test -n "$MISSING"; then
>>> + err "Thread IDs $MISSING not found in perf AUX data"
>>> + fi
>>> +}
>>> --
>>> 2.32.0
>>>
On 5/26/22 09:20, Leo Yan wrote:
> Hi Carsten,
>
> Sorry for late response.
>
> On Thu, Apr 21, 2022 at 06:38:33PM +0100, Carsten Haitzler wrote:
>
> [...]
>
>>> Very big change... Why squash all patches form previous verion to this
>>> single one big patch? Usually the format with small patches is much
>>> better for reviewing.
>>
>> I was asked to re-jig the tree and in doing so I also ended up cutting down
>> the size a lot so this just makes more sense together as a "here are the
>> tests" as adding infra without any tests makes no sense and the tests
>> themelves are self-contained in their own directories and source files and
>> "drivign scripts" thus it's essentially patch 1 appended to patch 2 to patch
>> 3 etc. and still broken up in the patch file by file.
>
> I am not sure if I understand the meaning, seems to me you could
> organize the patch series like:
>
> - Patch for common files (e.g. script lib/coresight.sh or some
> Makefile changes);
> - Patches for enabling test cases, E.g.:
> patch for asm_pure_loop;
> patch for thread loop (include unroll loop);
> patch for memcpy;
> - Patch for documentation.
Each of those last patches are simple a patch with added "fluff" to add
a line to a makefile each time then entire new files. A single patch set
already divides each of these into their own files in the same patch are
already divided - the only difference is - do you get the added fluff
fiddling to add files to the makefile or just have them all added at
once and the rest is already broken up. If you scroll through a patch
with 3 files, or have 3 patches with 1 file in each - it's the same in
the end. It's divided either way, but creating a patch set where one
patch after the other adds a new test and each adds a line to the same
makefle is a more work as I can't just git add/commit the test files to
each patch. I have to first remove all tests. Then add the parent
makefile for the tools with 1 line in it for 1 tool and 1 set of patch
scripts. I then have to add the next one and edit this parent makefile
to add one line and so on - where it actually makes the total size of
all the patches bigger as this makefile that lists all the child dirs
keeps being modified only to just add a single line added to it and the
rest is all stand-alone new files anyway. As all the changes to this
parent makefile all modify the same context, they can't be separated
out and cherry-picked as they then get conflicts in the context anyways
so the patches can't be chosen separately without resolving conflicts
anyway, thus to me it makes no sense to do all the extra work for a
bigger total patch set that can't be split out without conflicts anyway...
But I kind of give up. It seems there is a view (that I don't understand
as over-all the data is more complex and larger to review when broken
out) that more data in N patches is preferable to less data in a single
patch (even though it's all broken out into different file blocks in the
same patch so you scroll down and basically in essence "cat *.patch" and
"cat onefile.patch" produce the same thing other than the "add one more
line to this makefile"). I'll break it up - it just is more busy-work to
create a bigger overall patch set with more changes to look at overall.
> If this is not comfortable for you, at least we can use three patches:
>
> - Patch for common file (e.g. script lib/coresight.sh);
> - Patch for test cases;
> - Patch for documentation.
The documentation I can see breaking out - though that wants to be moved
somewhere else in the tree now and not alongside the perf tool. I
generally go for the view of documentation belongs with the thing it
documents and if there is new information to document, it belongs with
the patch introducing what it documents. :)
> [...]
>
>>>> +If you see these above, then your system is tracing coresight data
>>>> +correctly.
>>>> +
>>>> +To compile perf with coresight support in the perf directory do
>>>> +
>>>> + make CORESIGHT=1
>>>
>>> It is inaccurate that if we don't mention openCSD lib.
>>
>> Do you mean I need to mention that you need the opencsd library installed
>> too?
>
> Yes, otherwise, users might directly build perf without opencsd lib,
> then finally they cannot use perf with Arm CoreSight.
OK - sure. Will document that OpenCSD is needed and where to possibly
get it (git, packages etc.).
>>>> +This will compile the perf tool with coresight support as well as
>>>> +build some small test binaries for perf test. This requires you also
>>>> +be compiling for 64bit Arm (ARM64/aarch64). The tools run as part of
>>>> +perf coresight tracing are in tests/shell/tools/coresight.
>>>
>>> For build perf tool, I think above paragraphs are duplicate with the
>>> document Documentation/trace/coresight/coresight.rst. Can we simply
>>> say:
>>>
>>> "The details for building perf tool with support Arm Coresight can be
>>> found in the "HOWTO.md" file of the openCSD gitHub repository:
>>> https://github.com/Linaro/opencsd.
>>
>> I can. I put this here as I didn't go clone OpencCSD first but used my
>> distro OpenCSD packages and thus of course didn't have the documentation in
>> front of me. I spent some time wondering why it wasn't building with
>> coresight support even though it detected OpenCSD when I compiled... I
>> didn't expect to have to go to some separate project git repository and read
>> docs there on how to build the perf tool here in the kernel. I wrote this
>> because it was an actual problem I hit and it's a lot less frustrating to
>> "end users" to give them the information they need in the relevant place
>> they need it instead of sending them around to other project trees. Building
>> perf with coresight support is handled by the perf tree int he kernel, not
>> OpenCSD, thus IMHO that is where the documentation belongs - alongside the
>> thing that determines how to build something.
>
> Understand.
I'll move this over to the place that Matthieu and you mention
(Documentation/trace/coresight), but I do think this kind of information
is necessary as above, so thanks. :)
>>> And "HOWTO.md" file gives the information and examples for how to use
>>> perf tool to record and report Coresight trace data. It's the
>>> prerequisite for this perf Coresight test."
>>>
>>>> +You will also want coresight support enabled in your kernel config.
>>>> +Ensure it is enabled with:
>>>> +
>>>> + CONFIG_CORESIGHT=y
>>>> +
>>>> +There are various other coresight options you probably also want
>>>> +enabled like:
>>>> +
>>>> + CONFIG_CORESIGHT_LINKS_AND_SINKS=y
>>>> + CONFIG_CORESIGHT_LINK_AND_SINK_TMC=y
>>>> + CONFIG_CORESIGHT_CATU=y
>>>> + CONFIG_CORESIGHT_SINK_TPIU=y
>>>> + CONFIG_CORESIGHT_SINK_ETBV10=y
>>>> + CONFIG_CORESIGHT_SOURCE_ETM4X=y
>>>> + CONFIG_CORESIGHT_STM=y
>>>> + CONFIG_CORESIGHT_CPU_DEBUG=y
>>>> + CONFIG_CORESIGHT_CTI=y
>>>> + CONFIG_CORESIGHT_CTI_INTEGRATION_REGS=y
>>>> +
>>>> +Please refer to the kernel configuration help for more information.
>>>
>>> I prefer to remove these kernel configuration since they are not
>>> inconsistent on different platforms (e.g. ETBV10, ETM4X, etc), and
>>> some configurations might not necessary (e.g. CPU_DEBUG).
>>
>> Certainly there should be some documentation on which kernel configs you
>> might want to turn on then? Imagine someone new comes along and doesn't have
>> any idea what to possible enable at all and manages to build perf with
>> coresight support (as above) then finds it doesn't work because they didn't
>> enable enough config in the kernel? Sure - could probably trim these down a
>> bit but the point here is to alert the user to there being a range of
>> coresight config options that you need to turn on that you likely will find
>> are not turned on. They certainly are not turned on on distro kernels and a
>> lot of the time when you have a platform that already boots/works you start
>> with your distro kernel config file because you want everything enabled so
>> it actually boots. I've learned the hard way to do this as you manage to
>> forget to turn on some MMC driver or some other feature and your boot hangs
>> or doesn't find rootfs etc.
>
> So far, we will have two documents in Linux kernel:
>
> - Documentation/trace/coresight/coresight.rst;
> - tools/perf/Documentation/arm-coresight.txt.
>
> We need to avoid overlap between these two files. I think we could use
> the file Documentation/trace/coresight/coresight.rst to focus on
> CoreSight driver module relates stuffs and
> tools/perf/Documentation/arm-coresight.txt is more about the perf
> usages.
That's the case here with this patch set. To me - I would look for the
perf docs where I put them - in the perf documentation directory, not
somewhere else.
> But, the file Documentation/trace/coresight/coresight.rst doesn't give
> any info for kernel configs, I think which would be a better place to
> give information for building kernel modules.
Correct. I do think you're right. I think that the "core docs" should
give this information and the perf tool docs should at least reference
these so you know to look there. I think I should at least put a stub
doc in the perf tool doc tree to tell someone to go look at the "core
docs" for more information.
>> What would you recommend then as a "turn these on and coresight will almost
>> certainly work for you on your given hardware " then?
>
> This would be fine. Alternatively, we could add a section in the file
> Documentation/trace/coresight/coresight.rst to describe how to build
> CoreSight modules.
I think that is good/best and then "link" the docs so someone knows
where to look for more information.
> How you think for this? I also would like to get suggestions from
> CoreSight maintainers Suzuki/Mathieu/Mike.
>
> [...]
>
>>> Please update based on the latest test case names, at my side, I can
>>> see the testing case like:
>>>
>>> Coresight / ASM Pure Loop
>>> Coresight / Memcpy 16k 10 Threads
>>> Coresight / Thread Loop 10 Threads - Check TID
>>> Coresight / Thread Loop 2 Threads - Check TID
>>> Coresight / Unroll Loop Thread 10
>>
>> Oh sorry - yeah. I wrote the docs based on the earlier tests. Will fix.
>
> Thanks.
>
>>>> +
>>>> +These perf record tests will not run if the tool binaries do not exist
>>>> +in tests/shell/tools/coresight/*/ and will be skipped. If you do not
>>>> +have coresight support in hardware then either do not build perf with
>>>> +coresight support or remove these binaries in order to not have these
>>>> +tests fail and have them skip instead.
>>>> +
>>>> +These tests will log historical results in the current working
>>>> +directory (e.g. tools/perf) and will be named stats-*.csv like:
>>>> +
>>>> + stats-asm_pure_loop-out.csv
>>>> + stats-bubble_sort-random.csv
>>>> + ...
>>>> +
>>>> +These statistic files log some aspects of the AUX data sections in
>>>> +the perf data output counting some numbers of certain encodings (a
>>>> +good way to know that it's working in a very simple way). One problem
>>>> +with coresight is that given a large enough amount of data needing to
>>>> +be logged, some of it can be lost due to the processor not waking up
>>>> +in time to read out all the data from buffers etc.. You will notice
>>>> +that the amount of data collected can vary a lot per run of perf test.
>>>> +If you wish to see how this changes over time, simply run perf test
>>>> +multiple times and all these csv files will have more and more data
>>>> +appended to it that you can later examine, graph and otherwise use to
>>>> +figure out if things have become worse or better.
>>>
>>> I am confused by this narrative. Does it try to remind that the final
>>> testing result (pass or fail) is not stable? Or should we run for
>>> multiple times so have more chance to capture issues?
>>
>> That is correct. I thought I was clear that it's lossy. That is actually the
>> case. I have tests here that actually fail because there is no data
>> collected from some threads at all (missing CID blocks for some of the
>> threads that run in the test). The point is to have tests that may be
>> failing now but in future will improve. I lowered the minimum bar to pass
>> for most tests to have "at least just a little data" but most tests show
>> highly variable amount of captured data. the csv files are there to
>> over-time give you a good idea of the stability of the captured data.
>
> Okay, this would be fine for me. Though I am a bit worry that later if
> users report a failure, then how we can tell them this is a bug or it's
> just tracing quality issue?
Well this is kind of both a bug and a quality issue. Reality is I have
tests that literally get no information in the perf trace for some
threads. Reality is we have what I'd call "Quality of trace" bugs and I
was doing these test as a way of beginning to explore that and quantify
it as that is not something the existing tests were doing. We want to
have tests that in future with new designs show the quality bug to be
fixed. It's kind of test driven development. Expose the quality bug,
document it as such and then in future show it to be fixed when that
happens (v9 should fix this...). I probably should document this as such
(as a quality bug and why such failures are to be expected) so I'll add
that to my TODO to make sure it's clear that these tests might fail and
it's a quality issue (I mention the CSV files that are generated as a
way to over-time track that quality).
> [...]
>
>>>> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
>>>> new file mode 100644
>>>> index 000000000000..468673ac32e8
>>>> --- /dev/null
>>>> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore
>>>> @@ -0,0 +1 @@
>>>> +asm_pure_loop
>>>
>>> Do we really need there '.gitignore' files under the folder
>>> 'tools/perf/tests/shell/coresight/'.
>>
>> Where would you rather have them to ignore the generated binary tools?
>
> It's interesting that I wanted to find a case to object you, so I tried
> to check the folder linux/samples/bpf, but it does use .gitignore file
> to ignore built binaries :)
>
> Adding .gitignore is in practice and this would be fine for me.
:)
>>>> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
>>>> new file mode 100644
>>>> index 000000000000..10c5a60cb71c
>>>> --- /dev/null
>>>> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile
>>>> @@ -0,0 +1,30 @@
>>>> +# SPDX-License-Identifier: GPL-2.0
>>>> +# Carsten Haitzler <[email protected]>, 2021
>>>> +
>>>> +include ../Makefile.miniconfig
>>>> +
>>>> +BIN=asm_pure_loop
>>>> +LIB=
>>>
>>> Remove the unused variable 'LIB='.
>>
>> I have this because I wanted to have a simple template to be able to re-use
>> for more tests over time. It's so much easier to maintain and extend if
>> every makefile and tool follow a similar pattern and you can almost copy &
>> paste between them as they don't have "exceptions". You really want me to
>> remove this?
>
> It's fine to keep it. Could you add a comment for this?
No problems! Can do.
> To be honest, I am not experienced for bash shell script, so I have no
> idea why write like this way. If you think this is very common usage
> in shell, then you could keep it and don't need to add comment.
Well this is a makefile above (not shell), but having done lots of
makefile and other higher level make tooling (autotools, cmake, meson)
and having to maintain larger trees with "We build the same kind of
thing N times but each module/tool needs slightly different
linking/includes/tooling" ... I have learned the value of using
templates that are the same but sometimes you have empty fields or
fields that change. When you have to fix/change a "design pattern" issue
it's easier to do when the files share more in common and it's easier
for someone to start a new tool/module/subdir by taking an existing
template and just modifying it as needed and everything they need is
there. It really makes maintenance so much easier in the long-run as you
expand the set of things something builds. Design patterns in your build
system really help. :)
> [...]
>
>
>>> There have four sub folders under tools/perf/tests/shell/coresight:
>>>
>>> asm_pure_loop
>>> memcpy_thread
>>> thread_loop
>>> unroll_loop_thread
>>>
>>> And every folder has its own Makefile and every Makefile is quite
>>> close to each other. I am just wandering if it's possible to
>>> remove the 4 Makefiles in these four sub folders, and simply use
>>> tools/perf/tests/shell/coresight/Makefile as the central place to
>>> build these assistant programs.
>>
>> I did this so it's easier to etxent over time. having a single parent
>> makefile that over time accumulates little ugly "if's" and exceptions makes
>> longer-term maintenance and extending harder. I did it this way to make this
>> easy - make a copy of a dir - add that dir to a parent makefile then modify
>> the makefile as needed (but only as needed).
>
> Okay, let's keep the saperate makefiles.
:)
>>>> diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
>>>> new file mode 100644
>>>> index 000000000000..75cf084a927d
>>>> --- /dev/null
>>>> +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S
>>>> @@ -0,0 +1,28 @@
>>>> +/* SPDX-License-Identifier: GPL-2.0 */
>>>> +/* Tamas Zsoldos <[email protected]>, 2021 */
>>>> +
>>>> +.globl _start
>>>> +_start:
>>>> + mov x0, 0x0000ffff
>>>> + mov x1, xzr
>>>> +loop:
>>>> + nop
>>>> + nop
>>>> + cbnz x1, noskip
>>>> + nop
>>>> + nop
>>>> + adrp x2, skip
>>>> + add x2, x2, :lo12:skip
>>>> + br x2
>>>> + nop
>>>> + nop
>>>> +noskip:
>>>> + nop
>>>> + nop
>>>> +skip:
>>>> + sub x0, x0, 1
>>>> + cbnz x0, loop
>>>> +
>>>> + mov x0, #0
>>>> + mov x8, #93 // __NR_exit syscall
>>>> + svc #0
>>>
>>> I tested the case "ASM Pure Loop" on my Juno board, and it complaints:
>>>
>>> root@debian:/mnt/export/arm-linux-kernel/tools/perf# ./perf test -v 76
>>> 76: Coresight / ASM Pure Loop :
>>> --- start ---
>>> test child forked, pid 9063
>>> failed to mmap with 12 (Cannot allocate memory)
>>> test child finished with -1
>>> ---- end ----
>>> Coresight / ASM Pure Loop: FAILED!
>>>
>>> Since I only setup the 1GB memory for the Linux kernel, it fails to
>>> allocate AUX ring buffer with the size 256MB. So I manully change
>>> the buffer size to 8MB in tools/perf/tests/shell/lib/coresight.sh:
>>>
>>> PERFRECMEM="-m ,8M"
>>>
>>> So finally I can see the test case is passed:
>>
>> This is artificial isn't it? limiting to 1GB. You certainly have far more
>> memory than that available. My testse were on a system with 4GB and I had no
>> issues.
>
> Please see below comment.
>
>>> root@debian:/mnt/export/arm-linux-kernel/tools/perf# ./perf test -v 76
>>> 76: Coresight / ASM Pure Loop :
>>> --- start ---
>>> test child forked, pid 9481
>>> -m ,8M -e cs_etm//u
>>> [ perf record: Woken up 1 times to write data ]
>>> [ perf record: Captured and wrote 0.681 MB ./perf-asm_pure_loop-out.data ]
>>> test child finished with 0
>>> ---- end ----
>>> Coresight / ASM Pure Loop: Ok
>>>
>>> Do you think we really need to use 256MiB as the AUX buffer size?
>>> IIRC, it means we allocate 256MiB per CPU for this case, on the other
>>> hand, you could see the final perf data file size is small (0.681
>>> MiB).
>>>
>>> Seems to me, it's not necessary to allocate so big buffer for
>>> the test, and I tried to run below 4 cases with 8MiB, all of them can
>>> pass the testing :)
>>
>> I didn't think anyone with a system with coresight support that would be
>> running perf record locally would only have 1GB of ram... I knew junos had
>> 8GB and my dragonboard has 4GB ... so I know I was on the smaller side. I
>> thought a larger buffer == safer results (less chance of needing to write
>> out the buffer during capture). Admittdly I used 256Mb when my tests ran for
>> much longer and collected more data. I can try drop to 8 or 16gb and see.
>
> Yes, my Juno board has 8GB but I also have DB410c with 1GB with quad
> coes [1]. I am still concern for 256MB buffer size, it's not friendly for
> embedded system, and even not good for server. For example, if we run
> this testing on Arm server with 96 cores (like Hisilicon D06 board),
> then we need the buffer size is:
>
> 256MiB * 96 = 16GiB
A bit more actually 24M - but I see you correct this already in a
following mail, but your point remains the same. :)
> I agree usually 16GiB is not a problem for server, but seems to me
> it doesn't make much sense to consume huge memory resource for the
> testing.
Even 24GB is not a problem IMHO if you have 96 cores, but ... I see your
point.
> In other words, if set 8MiB (or 16MiB, 32MiB) buffer size and doesn't
> see testing result regression, I think this would be good to decrease
> the buffer size.
I'll drop it down to something less - sure. Maybe between 16-64M and
that should mean everything you have there is able to do this without
issues. 1GB with 8 cores would only need 512M for buffers - you have a
lot left over. So 64M or less would be just fine I think (and with 4
cores... even more headroom).
> [1] https://www.96boards.org/product/dragonboard410c/
>
> [...]
>
>>>> diff --git a/tools/perf/tests/shell/coresight_asm_pure_loop.sh b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
>>>> new file mode 100755
>>>> index 000000000000..3f0dbefcad50
>>>> --- /dev/null
>>>> +++ b/tools/perf/tests/shell/coresight_asm_pure_loop.sh
>>>> @@ -0,0 +1,18 @@
>>>> +#!/bin/sh -e
>>>> +# Coresight / ASM Pure Loop
>>>> +
>>>> +# SPDX-License-Identifier: GPL-2.0
>>>> +# Carsten Haitzler <[email protected]>, 2021
>>>> +
>>>> +TEST="asm_pure_loop"
>>>> +. $(dirname $0)/lib/coresight.sh
>>>> +ARGS=""
>>>> +DATV="out"
>>>> +DATA="$DATD/perf-$TEST-$DATV.data"
>>>> +
>>>> +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS
>>>> +
>>>> +perf_dump_aux_verify "$DATA" 10 10 10
>>>> +
>>>> +err=$?
>>>> +exit $err
>>>
>>> Can we organize the shell scripts by moving them into the folder
>>> tools/perf/tests/shell/coresight?
>>
>> We can - but it comes with a fair few more changes.
>>
>>> coresight_asm_pure_loop.sh
>>> coresight_memcpy_thread_16k_10.sh
>>> coresight_thread_loop_check_tid_10.sh
>>> coresight_thread_loop_check_tid_2.sh
>>> coresight_unroll_loop_thread_10.sh
>>>
>>> And we even can consider to move script test_arm_coresight.sh into
>>> the folder tools/perf/tests/shell/coresight and change its
>>> name as 'coresight_smoke_test.sh'.
>>
>> Indeed these other tests I left alone for now and had not thought about how
>> to marry these together yet - leaving this for another day and another patch
>> set rather than this patch set itself. That was my thoguht. I was trying to
>> make an "Easier to extend by just dropping a test into a dir" setup here to
>> make maintenance and expansion easier over time (and thus encourage testing
>> by having a simple repeatable test infra to duplicate). I ended up with a
>> dir per test tool you need to build and a driver script in the tests/shell
>> dir. I think this is certainly worth considering but perhaps as a separate
>> set of work to marry these?
>
> Okay, it would be fine to use separate set for moving the script
> test_arm_coresight.sh, which is a simple case.
I think that breaking out test_arm_coresight.sh into sub-tests that are
stand-alone is better. I've had to deal with test suites that have one
test case and that test fails/succeeds BUT this test case actually tests
30+ API calls or something like that and one of those fail. Which one?
You don't immediately know. You have to now go digging through log files
to find out. Breaking this out into less efficient but easier to see
"one test per thing you test" is much better and long-run really saves
time and headaches, but this is "let's do this another day" but it is
why I am over-engineering the tests with the view that these will expand
in number and move over such tests.
>> I piggybacked on the existing shell test infra but added a fair few more
>> scripts. To do what you suggest I'd need to modify the core shell test code
>> to walk subdirs recursively then looking for child scripts. The problem is
>> how does perf test's shell handling know about the coresight subdir vs the
>> lib subdir?
>
> Yeah, now I understand your point. How about file layout like below?
>
> tools/perf/tests/shell/coresight_test_hub.sh
> tools/perf/tests/shell/coresight/coresight_asm_pure_loop.sh
> tools/perf/tests/shell/coresight/coresight_memcpy_thread_16k_10.sh
> tools/perf/tests/shell/coresight/coresight_thread_loop_check_tid_10.sh
> tools/perf/tests/shell/coresight/coresight_thread_loop_check_tid_2.sh
> tools/perf/tests/shell/coresight/coresight_unroll_loop_thread_10.sh
>
> So we use tools/perf/tests/shell/coresight_test_hub.sh as an interface
> to hook with Perf test infrastructure, and then hub.sh file calls
> testing scripts under the sub folder. Seems to me, this is also
> friendly for later's extension.
Aaaaah here I disagree. The current test infra will only see
"coresight_test_hub.sh" and then that all passes, fails or skips as a
single test (see above). ub-tests are hidden when you do "perf test".
I've had to deal with this before and it is (IMHO) a horrible way to go
as you spend time then re-running the tests by hand and digging through
other logs to see what failed rather than the toplevel tests indicating
right there pass, fail or skip. Over time it becomes a time-sink to hunt
these and mystifies new people who don't know where to dig for the
specific failure.
To break them up like you want into subdirs then requires me to go to
perf test (builtin-script.c) and change how the walking of test subdirs
works with shells. It seems to have some legacy code in there that walks
some "lang" dirs with "bin" dirs in them that seemingly do not exist
today in the kernel tree. The right way to do this if you want to break
this out into a coresight dir would be to modify how perf test walks
dirs and builds the list of tests. I'd have to walk all subdirs then
specifically filter for tests it knows will exist (executable, end in
.sh, maybe be named like parentdir/parentdir_XXX.sh like your above
suggestion of coresight/coresight_XXX.sh etc.). This now involves
re-jigging a lot more code and abstracting the walking into a single
function that walks once then stores the data in an array of struct
pef_shell_test or such which now is filtered (and sorted which we should
do which we don't now), and thus clean this all up, remove the
(seemingly) legacy lang/bin dir walk code (I need to dig through history
to find out where this came in and why it was there) and so on.
This certainly then makes this set of patches bigger and I need to break
out this "revamp" of the walking of test shell scripts a preparation
patch, but just saying: It certainly raises the bar for work and I don't
think having a single parent .sh that then walks all the children and
produces a single test output is right. I could popen this test and read
the output to generate multiple output tests but, IMHO, that is worse
and more complex than fixing the tree walk above to build an index of
all shell tests once etc. like:
struct script_file {
char *dir;
char *file;
char *label;
}
struct script_file *list_script_files(void);
... so we run this whenever we want a list of shell tests and just
iterate over it until a NULL member (dir, file, label are all NULL).
Simple and does the job but moves all the walking into a single function
instead of there being like 3 of them now.
>> Both contain *.sh shell scripts - the difference is the ones in
>> lib are not executable. Is this sufficiently different? I could also open
>> them to check the have #!/bin/... as the first line. Hardcoding just a
>> single coresight subdir just feels wrong and hacky to me, thus the generic
>> recursion solution I suggest here.
>
> Agreed. I also don't prefer this way.
>
>> I can definitely see how extending to subdirs would make supporting testing
>> cleaner and divide things into their own domains (dirs).
>
> Thanks a lot for the work! The test cases are good for me (but I would
> say Mike is the best person for reviewing testing trace data quaility),
> I just want to make sure it's not hard for later maintenance.
Sure. Let me at least make a start on the shell-test walk code and
cleaning it up. I've just been staring at this and trying to find a
"better way" but I'm pretty much at "Do it my current way with no subdir
and keep the patch simple" or "Make this a more complex a patch to try
streamline the code and abstract the shell tests better and include
sub-dir walking" as I describe above. Long term I think this is actually
a very good idea to clean this up and allow tests then in general to go
into domain-specific subdirs. It just brings forward this kind of work
ahead of getting extra tests in. I can't really see another sensible option.