perf test: waiting.sh: Replace timestamp polling with sleep
Adrian Hunter
reviewerPATCH 1/1 perf test: waiting.sh: Replace timestamp polling with sleep
The waiting helpers implement timeouts using:
date +%s%1N
This relies on GNU coreutils date truncating %N to the specified width, so %1N yields tenths of a second.
Rust coreutils (uutils) interprets the width differently and does not truncate the nanoseconds field. Consequently "date +%1N" returns all nine nanosecond digits, causing the elapsed-time calculation to be done in nanoseconds while timeout values remain in tenths of a second.
As a result, timeout comparisons succeed immediately and the waiting helpers time out on their first iteration. This causes test_intel_pt.sh to fail on systems using uutils "date".
Avoid implementation-specific date formatting entirely. Instead, wait for 100 ms on each iteration and count the timeout down. Besides fixing the portability issue, this removes the busy-waiting behaviour in wait_for_perf_to_start(), which could otherwise consume CPU while waiting for perf record to start.
Since the timeout is now based on repeated sleeps, it is only approximate. Update the comments accordingly. Also make is_running() wait for exactly the documented number of tenths by changing its timeout test from -gt to the new logic, and quote tm_out in the modified code.
--- tools/perf/tests/shell/lib/waiting.sh | 34 +++++++++++++-------------- 1 file changed, 16 insertions(+), 18 deletions(-)
tools/perf/tests/shell/lib/waiting.sh +16 -18
@@ -1,77 +1,75 @@ #!/bin/bash # SPDX-License-Identifier: GPL-2.0 -tenths=date\ +%s%1N - # Wait for PID $1 to have $2 number of threads started -# Time out after $3 tenths of a second or 5 seconds if $3 is "" +# Time out after approx. $3 tenths of a second or 5 seconds if $3 is "" wait_for_threads() { tm_out=$3 ; [ -n "${tm_out}" ] || tm_out=50 - start_time=$($tenths) while [ -e "/proc/$1/task" ] ; do th_cnt=$(find "/proc/$1/task" -mindepth 1 -maxdepth 1 -printf x | wc -c) if [ "${th_cnt}" -ge "$2" ] ; then return 0 fi - # Wait at most tm_out tenths of a second - if [ $(($($tenths) - start_time)) -ge $tm_out ] ; then + if [ "${tm_out}" -le 0 ] ; then echo "PID $1 does not have $2 threads" return 1 fi + sleep 0.1 + tm_out=$((tm_out - 1)) done return 1 } # Wait for perf record -vvv 2>$2 with PID $1 to start by looking at file $2 # It depends on capturing perf record debug message "perf record has started" -# Time out after $3 tenths of a second or 5 seconds if $3 is "" +# Time out after approx. $3 tenths of a second or 5 seconds if $3 is "" wait_for_perf_to_start() { tm_out=$3 ; [ -n "${tm_out}" ] || tm_out=50 echo "Waiting for \"perf record has started\" message" - start_time=$($tenths) while [ -e "/proc/$1" ] ; do if grep -q "perf record has started" "$2" ; then echo OK break fi - # Wait at most tm_out tenths of a second - if [ $(($($tenths) - start_time)) -ge $tm_out ] ; then + if [ "${tm_out}" -le 0 ] ; then echo "perf recording did not start" return 1 fi + sleep 0.1 + tm_out=$((tm_out - 1)) done return 0 } # Wait for process PID %1 to exit -# Time out after $2 tenths of a second or 5 seconds if $2 is "" +# Time out after approx. $2 tenths of a second or 5 seconds if $2 is "" wait_for_process_to_exit() { tm_out=$2 ; [ -n "${tm_out}" ] || tm_out=50 - start_time=$($tenths) while [ -e "/proc/$1" ] ; do - # Wait at most tm_out tenths of a second - if [ $(($($tenths) - start_time)) -ge $tm_out ] ; then + if [ "${tm_out}" -le 0 ] ; then echo "PID $1 did not exit as expected" return 1 fi + sleep 0.1 + tm_out=$((tm_out - 1)) done return 0 } -# Check if PID $1 is still running after $2 tenths of a second +# Check if PID $1 is still running after approx. $2 tenths of a second # or 0.3 seconds if $2 is "" is_running() { tm_out=$2 ; [ -n "${tm_out}" ] || tm_out=3 - start_time=$($tenths) while [ -e "/proc/$1" ] ; do - # Check for at least tm_out tenths of a second - if [ $(($($tenths) - start_time)) -gt $tm_out ] ; then + if [ "${tm_out}" -le 0 ] ; then return 0 fi + sleep 0.1 + tm_out=$((tm_out - 1)) done echo "PID $1 exited prematurely" return 1
Signature · 1 line
2.53.0
[email protected]
AI reviewSashiko has reviewed this patch and found no issues. It looks great!
Quoted 22 lines
The waiting helpers implement timeouts using:
date +%s%1N
This relies on GNU coreutils date truncating %N to the specified width, so %1N yields tenths of a second.
Rust coreutils (uutils) interprets the width differently and does not truncate the nanoseconds field. Consequently "date +%1N" returns all nine nanosecond digits, causing the elapsed-time calculation to be done in nanoseconds while timeout values remain in tenths of a second.
As a result, timeout comparisons succeed immediately and the waiting helpers time out on their first iteration. This causes
test_intel_pt.shto fail on systems using uutils "date".Avoid implementation-specific date formatting entirely. Instead, wait for 100 ms on each iteration and count the timeout down. Besides fixing the portability issue, this removes the busy-waiting behaviour in
wait_for_perf_to_start(), which could otherwise consume CPU while waiting for perf record to start.
Quoted 6 lines
Since the timeout is now based on repeated sleeps, it is only approximate. Update the comments accordingly. Also make
is_running()wait for exactly the documented number of tenths by changing its timeout test from -gt to the new logic, and quotetm_outin the modified code.Signed-off-by Adrian Hunter <[email protected]>
Sashiko has reviewed this patch and found no issues. It looks great!
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
Quoted 23 lines · Adrian Hunter
The waiting helpers implement timeouts using:
date +%s%1N
This relies on GNU coreutils date truncating %N to the specified width, so %1N yields tenths of a second.
Rust coreutils (uutils) interprets the width differently and does not truncate the nanoseconds field. Consequently "date +%1N" returns all nine nanosecond digits, causing the elapsed-time calculation to be done in nanoseconds while timeout values remain in tenths of a second.
As a result, timeout comparisons succeed immediately and the waiting helpers time out on their first iteration. This causes
test_intel_pt.shto fail on systems using uutils "date".Avoid implementation-specific date formatting entirely. Instead, wait for 100 ms on each iteration and count the timeout down. Besides fixing the portability issue, this removes the busy-waiting behaviour in
wait_for_perf_to_start(), which could otherwise consume CPU while waiting for perf record to start.
Quoted 6 lines
Since the timeout is now based on repeated sleeps, it is only approximate. Update the comments accordingly. Also make
is_running()wait for exactly the documented number of tenths by changing its timeout test from -gt to the new logic, and quotetm_outin the modified code.Signed-off-by Adrian Hunter <[email protected]>
Thanks,
Ian
Quoted 105 lines
--- tools/perf/tests/shell/lib/waiting.sh | 34 +++++++++++++-------------- 1 file changed, 16 insertions(+), 18 deletions(-)tools/perf/tests/shell/lib/waiting.sh +16
-19@@ -1,77 +1,75 @@ #!/bin/bash # SPDX-License-Identifier: GPL-2.0 -tenths=date\ +%s%1N - # Wait for PID $1 to have $2 number of threads started -# Time out after $3 tenths of a second or 5 seconds if $3 is "" +# Time out after approx. $3 tenths of a second or 5 seconds if $3 is "" wait_for_threads() { tm_out=$3 ; [ -n "${tm_out}" ] || tm_out=50 - start_time=$($tenths) while [ -e "/proc/$1/task" ] ; do th_cnt=$(find "/proc/$1/task" -mindepth 1 -maxdepth 1 -printf x | wc -c) if [ "${th_cnt}" -ge "$2" ] ; then return 0 fi - # Wait at most tm_out tenths of a second - if [ $(($($tenths) - start_time)) -ge $tm_out ] ; then + if [ "${tm_out}" -le 0 ] ; then echo "PID $1 does not have $2 threads" return 1 fi + sleep 0.1 + tm_out=$((tm_out - 1)) done return 1 } # Wait for perf record -vvv 2>$2 with PID $1 to start by looking at file $2 # It depends on capturing perf record debug message "perf record has started" -# Time out after $3 tenths of a second or 5 seconds if $3 is "" +# Time out after approx. $3 tenths of a second or 5 seconds if $3 is "" wait_for_perf_to_start() { tm_out=$3 ; [ -n "${tm_out}" ] || tm_out=50 echo "Waiting for \"perf record has started\" message" - start_time=$($tenths) while [ -e "/proc/$1" ] ; do if grep -q "perf record has started" "$2" ; then echo OK break fi - # Wait at most tm_out tenths of a second - if [ $(($($tenths) - start_time)) -ge $tm_out ] ; then + if [ "${tm_out}" -le 0 ] ; then echo "perf recording did not start" return 1 fi + sleep 0.1 + tm_out=$((tm_out - 1)) done return 0 } # Wait for process PID %1 to exit -# Time out after $2 tenths of a second or 5 seconds if $2 is "" +# Time out after approx. $2 tenths of a second or 5 seconds if $2 is "" wait_for_process_to_exit() { tm_out=$2 ; [ -n "${tm_out}" ] || tm_out=50 - start_time=$($tenths) while [ -e "/proc/$1" ] ; do - # Wait at most tm_out tenths of a second - if [ $(($($tenths) - start_time)) -ge $tm_out ] ; then + if [ "${tm_out}" -le 0 ] ; then echo "PID $1 did not exit as expected" return 1 fi + sleep 0.1 + tm_out=$((tm_out - 1)) done return 0 } -# Check if PID $1 is still running after $2 tenths of a second +# Check if PID $1 is still running after approx. $2 tenths of a second # or 0.3 seconds if $2 is "" is_running() { tm_out=$2 ; [ -n "${tm_out}" ] || tm_out=3 - start_time=$($tenths) while [ -e "/proc/$1" ] ; do - # Check for at least tm_out tenths of a second - if [ $(($($tenths) - start_time)) -gt $tm_out ] ; then + if [ "${tm_out}" -le 0 ] ; then return 0 fi + sleep 0.1 + tm_out=$((tm_out - 1)) done echo "PID $1 exited prematurely" return 1 --2.53.0
Quoted 24 lines · Ian Rogers
Quoted 23 lines · Adrian Hunter
The waiting helpers implement timeouts using:
date +%s%1N
This relies on GNU coreutils date truncating %N to the specified width, so %1N yields tenths of a second.
Rust coreutils (uutils) interprets the width differently and does not truncate the nanoseconds field. Consequently "date +%1N" returns all nine nanosecond digits, causing the elapsed-time calculation to be done in nanoseconds while timeout values remain in tenths of a second.
As a result, timeout comparisons succeed immediately and the waiting helpers time out on their first iteration. This causes
test_intel_pt.shto fail on systems using uutils "date".Avoid implementation-specific date formatting entirely. Instead, wait for 100 ms on each iteration and count the timeout down. Besides fixing the portability issue, this removes the busy-waiting behaviour in
wait_for_perf_to_start(), which could otherwise consume CPU while waiting for perf record to start.
Quoted 8 lines
Quoted 6 lines
Since the timeout is now based on repeated sleeps, it is only approximate. Update the comments accordingly. Also make
is_running()wait for exactly the documented number of tenths by changing its timeout test from -gt to the new logic, and quotetm_outin the modified code.Signed-off-by Adrian Hunter <[email protected]>Reviewed-by Ian Rogers <[email protected]>
Thanks, applied to perf-tools-next, for v7.4.
- Arnaldo