[PATCH v3 0/2] perf trace: Add more tests for BTF-augmented perf trace

[PATCH v3 0/2] perf trace: Add more tests for BTF-augmented perf trace

[Howard Chu]

Changes in v3:
 - Add vmlinux BTF check, and skip the tests if it doesn't exist

v1, v2:

The previous version of the perf trace BTF general augmentation tests
didn't pass Shellcheck (thanks to Arnaldo Carvalho de Melo
<acme@kernel.org> for pointing this out), this version uses bash instead
of POSIX shell to pass Shellcheck.

This patch series also adds documentation for the new option
--force-btf, which is used in the tests.

Link: https://lore.kernel.org/linux-perf-users/Zt9yiQq-n-W6I274@x1/

Howard Chu (2):
  perf trace: Add tests for BTF general augmentation
  perf docs: Add documentation for --force-btf option

 tools/perf/Documentation/perf-trace.txt     |  5 ++
 tools/perf/tests/shell/trace_btf_general.sh | 93 +++++++++++++++++++++
 2 files changed, 98 insertions(+)
 create mode 100755 tools/perf/tests/shell/trace_btf_general.sh

-- 
2.43.0

[PATCH v3 1/2] perf trace: Add tests for BTF general augmentation

[Howard Chu]

Currently, we only have perf trace augmentation tests for enum
arguments. This patch adds tests for more general syscall arguments,
such as struct pointers, strings, and buffers.

These tests utilize the perf config system to configure the perf trace
output, as suggested by Arnaldo Carvalho de Melo <acme@kernel.org>

Suggested-by: Arnaldo Carvalho de Melo <acme@kernel.org>
Signed-off-by: Howard Chu <howardchu95@gmail.com>
---
 tools/perf/tests/shell/trace_btf_general.sh | 93 +++++++++++++++++++++
 1 file changed, 93 insertions(+)
 create mode 100755 tools/perf/tests/shell/trace_btf_general.sh

diff --git a/tools/perf/tests/shell/trace_btf_general.sh b/tools/perf/tests/shell/trace_btf_general.sh
new file mode 100755
index 000000000000..bef07bad42bb
--- /dev/null
+++ b/tools/perf/tests/shell/trace_btf_general.sh
@@ -0,0 +1,93 @@
+#!/bin/bash
+# perf trace BTF general tests
+# SPDX-License-Identifier: GPL-2.0
+
+err=0
+set -e
+
+. "$(dirname $0)"/lib/probe.sh
+
+file1=$(mktemp /tmp/file1_XXXX)
+file2=$(echo $file1 | sed 's/file1/file2/g')
+
+buffer="buffer content"
+perf_config_tmp=$(mktemp /tmp/.perfconfig_XXXXX)
+
+trap cleanup EXIT TERM INT HUP
+
+check_vmlinux() {
+  echo "Checking if vmlinux BTF exists"
+  if ! ls /sys/kernel/btf/vmlinux 1>/dev/null 2>&1
+  then
+    echo "Skipped due to missing vmlinux BTF"
+    err=2
+  fi
+}
+
+trace_test_string() {
+  echo "Testing perf trace's string augmentation"
+  if ! perf trace -e renameat* --max-events=1 -- mv ${file1} ${file2} 2>&1 | \
+    grep -q -E "^mv/[0-9]+ renameat(2)?\(.*, \"${file1}\", .*, \"${file2}\", .*\) += +[0-9]+$"
+  then
+    echo "String augmentation test failed"
+    err=1
+  fi
+}
+
+trace_test_buffer() {
+  echo "Testing perf trace's buffer augmentation"
+  # echo will insert a newline (\10) at the end of the buffer
+  if ! perf trace -e write --max-events=1 -- echo "${buffer}" 2>&1 | \
+    grep -q -E "^echo/[0-9]+ write\([0-9]+, ${buffer}.*, [0-9]+\) += +[0-9]+$"
+  then
+    echo "Buffer augmentation test failed"
+    err=1
+  fi
+}
+
+trace_test_struct_btf() {
+  echo "Testing perf trace's struct augmentation"
+  if ! perf trace -e clock_nanosleep --force-btf --max-events=1 -- sleep 1 2>&1 | \
+    grep -q -E "^sleep/[0-9]+ clock_nanosleep\(0, 0, \{1,\}, 0x[0-9a-f]+\) += +[0-9]+$"
+  then
+    echo "BTF struct augmentation test failed"
+    err=1
+  fi
+}
+
+cleanup() {
+  rm -rf ${file1} ${file2} ${perf_config_tmp}
+}
+
+trap_cleanup() {
+  echo "Unexpected signal in ${FUNCNAME[1]}"
+  cleanup
+  exit 1
+}
+
+# don't overwrite user's perf config
+trace_config() {
+  export PERF_CONFIG=${perf_config_tmp}
+  perf config trace.show_arg_names=false trace.show_duration=false \
+    trace.show_timestamp=false trace.args_alignment=0
+}
+
+skip_if_no_perf_trace || exit 2
+
+check_vmlinux
+
+trace_config
+
+trace_test_string
+
+if [ $err = 0 ]; then
+  trace_test_buffer
+fi
+
+if [ $err = 0 ]; then
+  trace_test_struct_btf
+fi
+
+cleanup
+
+exit $err
-- 
2.43.0

[PATCH v3 2/2] perf docs: Add documentation for --force-btf option

[Howard Chu]

The --force-btf option is intended for debugging purposes and is
currently undocumented. Add documentation for it.

Signed-off-by: Howard Chu <howardchu95@gmail.com>
---
 tools/perf/Documentation/perf-trace.txt | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/tools/perf/Documentation/perf-trace.txt b/tools/perf/Documentation/perf-trace.txt
index 6e0cc50bbc13..fb3d2af33844 100644
--- a/tools/perf/Documentation/perf-trace.txt
+++ b/tools/perf/Documentation/perf-trace.txt
@@ -241,6 +241,11 @@ the thread executes on the designated CPUs. Default is to monitor all CPUs.
 	printing using the existing 'perf trace' syscall arg beautifiers to map integer
 	arguments to strings (pid to comm, syscall id to syscall name, etc).
 
+--force-btf::
+	Use btf_dump to pretty print syscall argument data, instead of using hand-crafted pretty
+	printers. This option is intended for testing BTF integration in perf trace. btf_dump-based
+	pretty-printing serves as a fallback to hand-crafted pretty printers, as the latter can
+	better pretty-print integer flags and struct pointers.
 
 PAGEFAULTS
 ----------
-- 
2.43.0

Re: [PATCH v3 1/2] perf trace: Add tests for BTF general augmentation

[Namhyung Kim]

Hi Howard,

On Wed, Dec 11, 2024 at 02:49:26PM -0800, Howard Chu wrote:
> Currently, we only have perf trace augmentation tests for enum
> arguments. This patch adds tests for more general syscall arguments,
> such as struct pointers, strings, and buffers.
> 
> These tests utilize the perf config system to configure the perf trace
> output, as suggested by Arnaldo Carvalho de Melo <acme@kernel.org>
> 
> Suggested-by: Arnaldo Carvalho de Melo <acme@kernel.org>
> Signed-off-by: Howard Chu <howardchu95@gmail.com>
> ---
>  tools/perf/tests/shell/trace_btf_general.sh | 93 +++++++++++++++++++++
>  1 file changed, 93 insertions(+)
>  create mode 100755 tools/perf/tests/shell/trace_btf_general.sh
> 
> diff --git a/tools/perf/tests/shell/trace_btf_general.sh b/tools/perf/tests/shell/trace_btf_general.sh
> new file mode 100755
> index 000000000000..bef07bad42bb
> --- /dev/null
> +++ b/tools/perf/tests/shell/trace_btf_general.sh
> @@ -0,0 +1,93 @@
> +#!/bin/bash
> +# perf trace BTF general tests
> +# SPDX-License-Identifier: GPL-2.0
> +
> +err=0
> +set -e
> +
> +. "$(dirname $0)"/lib/probe.sh
> +
> +file1=$(mktemp /tmp/file1_XXXX)
> +file2=$(echo $file1 | sed 's/file1/file2/g')
> +
> +buffer="buffer content"
> +perf_config_tmp=$(mktemp /tmp/.perfconfig_XXXXX)
> +
> +trap cleanup EXIT TERM INT HUP
> +
> +check_vmlinux() {
> +  echo "Checking if vmlinux BTF exists"
> +  if ! ls /sys/kernel/btf/vmlinux 1>/dev/null 2>&1

Normally we do

  if [ ! -f /sys/kernel/btf/vmlinux ]

> +  then
> +    echo "Skipped due to missing vmlinux BTF"
> +    err=2

This can be overwritten by trace_test_string.


> +  fi
> +}
> +
> +trace_test_string() {
> +  echo "Testing perf trace's string augmentation"
> +  if ! perf trace -e renameat* --max-events=1 -- mv ${file1} ${file2} 2>&1 | \
> +    grep -q -E "^mv/[0-9]+ renameat(2)?\(.*, \"${file1}\", .*, \"${file2}\", .*\) += +[0-9]+$"

Does this work without BTF support?

Thanks
Namhyung


> +  then
> +    echo "String augmentation test failed"
> +    err=1
> +  fi
> +}
> +
> +trace_test_buffer() {
> +  echo "Testing perf trace's buffer augmentation"
> +  # echo will insert a newline (\10) at the end of the buffer
> +  if ! perf trace -e write --max-events=1 -- echo "${buffer}" 2>&1 | \
> +    grep -q -E "^echo/[0-9]+ write\([0-9]+, ${buffer}.*, [0-9]+\) += +[0-9]+$"
> +  then
> +    echo "Buffer augmentation test failed"
> +    err=1
> +  fi
> +}
> +
> +trace_test_struct_btf() {
> +  echo "Testing perf trace's struct augmentation"
> +  if ! perf trace -e clock_nanosleep --force-btf --max-events=1 -- sleep 1 2>&1 | \
> +    grep -q -E "^sleep/[0-9]+ clock_nanosleep\(0, 0, \{1,\}, 0x[0-9a-f]+\) += +[0-9]+$"
> +  then
> +    echo "BTF struct augmentation test failed"
> +    err=1
> +  fi
> +}
> +
> +cleanup() {
> +  rm -rf ${file1} ${file2} ${perf_config_tmp}
> +}
> +
> +trap_cleanup() {
> +  echo "Unexpected signal in ${FUNCNAME[1]}"
> +  cleanup
> +  exit 1
> +}
> +
> +# don't overwrite user's perf config
> +trace_config() {
> +  export PERF_CONFIG=${perf_config_tmp}
> +  perf config trace.show_arg_names=false trace.show_duration=false \
> +    trace.show_timestamp=false trace.args_alignment=0
> +}
> +
> +skip_if_no_perf_trace || exit 2
> +
> +check_vmlinux
> +
> +trace_config
> +
> +trace_test_string
> +
> +if [ $err = 0 ]; then
> +  trace_test_buffer
> +fi
> +
> +if [ $err = 0 ]; then
> +  trace_test_struct_btf
> +fi
> +
> +cleanup
> +
> +exit $err
> -- 
> 2.43.0
> 

Re: [PATCH v3 1/2] perf trace: Add tests for BTF general augmentation

[Howard Chu]

Hello Namhyung,

On Fri, Dec 13, 2024 at 4:42 PM Namhyung Kim <namhyung@kernel.org> wrote:
>
> Hi Howard,
>
> On Wed, Dec 11, 2024 at 02:49:26PM -0800, Howard Chu wrote:
> > Currently, we only have perf trace augmentation tests for enum
> > arguments. This patch adds tests for more general syscall arguments,
> > such as struct pointers, strings, and buffers.
> >
> > These tests utilize the perf config system to configure the perf trace
> > output, as suggested by Arnaldo Carvalho de Melo <acme@kernel.org>
> >
> > Suggested-by: Arnaldo Carvalho de Melo <acme@kernel.org>
> > Signed-off-by: Howard Chu <howardchu95@gmail.com>
> > ---
> >  tools/perf/tests/shell/trace_btf_general.sh | 93 +++++++++++++++++++++
> >  1 file changed, 93 insertions(+)
> >  create mode 100755 tools/perf/tests/shell/trace_btf_general.sh
> >
> > diff --git a/tools/perf/tests/shell/trace_btf_general.sh b/tools/perf/tests/shell/trace_btf_general.sh
> > new file mode 100755
> > index 000000000000..bef07bad42bb
> > --- /dev/null
> > +++ b/tools/perf/tests/shell/trace_btf_general.sh
> > @@ -0,0 +1,93 @@
> > +#!/bin/bash
> > +# perf trace BTF general tests
> > +# SPDX-License-Identifier: GPL-2.0
> > +
> > +err=0
> > +set -e
> > +
> > +. "$(dirname $0)"/lib/probe.sh
> > +
> > +file1=$(mktemp /tmp/file1_XXXX)
> > +file2=$(echo $file1 | sed 's/file1/file2/g')
> > +
> > +buffer="buffer content"
> > +perf_config_tmp=$(mktemp /tmp/.perfconfig_XXXXX)
> > +
> > +trap cleanup EXIT TERM INT HUP
> > +
> > +check_vmlinux() {
> > +  echo "Checking if vmlinux BTF exists"
> > +  if ! ls /sys/kernel/btf/vmlinux 1>/dev/null 2>&1
>
> Normally we do
>
>   if [ ! -f /sys/kernel/btf/vmlinux ]

Sure, I'll change the check to this.

>
> > +  then
> > +    echo "Skipped due to missing vmlinux BTF"
> > +    err=2
>
> This can be overwritten by trace_test_string.

Oops my bad sorry.

>
>
> > +  fi
> > +}
> > +
> > +trace_test_string() {
> > +  echo "Testing perf trace's string augmentation"
> > +  if ! perf trace -e renameat* --max-events=1 -- mv ${file1} ${file2} 2>&1 | \
> > +    grep -q -E "^mv/[0-9]+ renameat(2)?\(.*, \"${file1}\", .*, \"${file2}\", .*\) += +[0-9]+$"
>
> Does this work without BTF support?

Yes, this works without BTF with Arnaldo's patch adding
sys_enter_renameat2. Before the patch, it wouldn't work because the
second filename would not be displayed properly. I understand this may
seem redundant, but now in perf trace all the data is collected
through the BTF general collector (if BTF is present), so this
trace_test_string() tests the new code path that uses BTF information
to pretty-print string arguments, since all data collection goes
through it.

Thanks,
Howard