Skip to contentSkip to conversation

selftests: mm: stop relying on killall in charge_reserved_hugetlb

PATCH In review CaoRuichuang 2 participants 2 messages Apr 5 – Apr 9

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 selftests: mm: stop relying on killall in charge_reserved_hugetlb

charge_reserved_hugetlb.sh tears down background writers with killall -2 --wait write_to_hugetlbfs. That depends on killall from psmisc, which is not installed on minimal Ubuntu images, so the test fails in cleanup with "killall: command not found".

Track the writer PIDs we start ourselves and signal them directly during cleanup instead. Make write_hugetlb_memory.sh exec write_to_hugetlbfs so the recorded PID names the long-lived test process, and redirect async output straight to the temporary log file instead of going through a tee pipeline.

---
 .../selftests/mm/charge_reserved_hugetlb.sh   | 25 +++++++++++++++----
 .../selftests/mm/write_hugetlb_memory.sh      |  2 +-
 2 files changed, 21 insertions(+), 6 deletions(-)
tools/testing/selftests/mm/charge_reserved_hugetlb.sh +20 -5
index 447769657..757df4878 100755
@@ -44,6 +44,8 @@ else
 fi
 export cgroup_path
 
+write_pids=()
+
 function cleanup() {
   if [[ $cgroup2 ]]; then
     echo $$ >$cgroup_path/cgroup.procs
@@ -193,10 +195,12 @@ function write_hugetlbfs_and_get_usage() {
     [[ "$private" == "-r" ]] && [[ "$expect_failure" != 1 ]]; then
 
     bash write_hugetlb_memory.sh "$size" "$populate" "$write" \
-      "$cgroup" "$path" "$method" "$private" "-l" "$reserve" 2>&1 | tee $output &
+      "$cgroup" "$path" "$method" "$private" "-l" "$reserve" \
+      >"$output" 2>&1 &
 
     local write_result=$?
     local write_pid=$!
+    write_pids+=("$write_pid")
 
     until grep -q -i "DONE" $output; do
       echo waiting for DONE signal.
@@ -261,10 +265,21 @@ function write_hugetlbfs_and_get_usage() {
 function cleanup_hugetlb_memory() {
   set +e
   local cgroup="$1"
-  if [[ "$(pgrep -f write_to_hugetlbfs)" != "" ]]; then
-    echo killing write_to_hugetlbfs
-    killall -2 --wait write_to_hugetlbfs
-    wait_for_hugetlb_memory_to_get_depleted $cgroup
+  local write_pid
+
+  if (( ${#write_pids[@]} )); then
+    for write_pid in "${write_pids[@]}"; do
+      if kill -0 "$write_pid" 2>/dev/null; then
+        echo killing write_to_hugetlbfs pid "$write_pid"
+        kill -2 "$write_pid"
+        wait "$write_pid"
+      fi
+    done
+    write_pids=()
+
+    if [[ -n "$cgroup" ]]; then
+      wait_for_hugetlb_memory_to_get_depleted "$cgroup"
+    fi
   fi
   set -e
 
tools/testing/selftests/mm/write_hugetlb_memory.sh +1 -1
index 3d2d2eb9d..49164bbfc 100755
@@ -19,5 +19,5 @@ echo $$ > ${cgroup_path:-/dev/cgroup/memory}/"$cgroup"/cgroup.procs
 echo "Method is $method"
 
 set +e
-./write_to_hugetlbfs -p "$path" -s "$size" "$write" "$populate" -m "$method" \
+exec ./write_to_hugetlbfs -p "$path" -s "$size" "$write" "$populate" -m "$method" \
       "$private" "$want_sleep" "$reserve"
Signature · 1 line
2.39.5 (Apple Git-154)
reviewer
#1 lore raw

Hi,

Quoted 4 lines · CaoRuichuang

charge_reserved_hugetlb.sh tears down background writers with killall -2 --wait write_to_hugetlbfs. That depends on killall from psmisc, which is not installed on minimal Ubuntu images, so the test fails in cleanup with "killall: command not found".

Can't we just skip the test if killall is not available just like we skip it if it's not run as root?

Quoted 79 lines

Track the writer PIDs we start ourselves and signal them directly during cleanup instead. Make write_hugetlb_memory.sh exec write_to_hugetlbfs so the recorded PID names the long-lived test process, and redirect async output straight to the temporary log file instead of going through a tee pipeline.

---
 .../selftests/mm/charge_reserved_hugetlb.sh   | 25 +++++++++++++++----
 .../selftests/mm/write_hugetlb_memory.sh      |  2 +-
 2 files changed, 21 insertions(+), 6 deletions(-)
tools/testing/selftests/mm/charge_reserved_hugetlb.sh +20 -5
index 447769657..757df4878 100755
@@ -44,6 +44,8 @@ else
 fi
 export cgroup_path
 
+write_pids=()
+
 function cleanup() {
   if [[ $cgroup2 ]]; then
     echo $$ >$cgroup_path/cgroup.procs
@@ -193,10 +195,12 @@ function write_hugetlbfs_and_get_usage() {
     [[ "$private" == "-r" ]] && [[ "$expect_failure" != 1 ]]; then
 
     bash write_hugetlb_memory.sh "$size" "$populate" "$write" \
-      "$cgroup" "$path" "$method" "$private" "-l" "$reserve" 2>&1 | tee $output &
+      "$cgroup" "$path" "$method" "$private" "-l" "$reserve" \
+      >"$output" 2>&1 &
 
     local write_result=$?
     local write_pid=$!
+    write_pids+=("$write_pid")
 
     until grep -q -i "DONE" $output; do
       echo waiting for DONE signal.
@@ -261,10 +265,21 @@ function write_hugetlbfs_and_get_usage() {
 function cleanup_hugetlb_memory() {
   set +e
   local cgroup="$1"
-  if [[ "$(pgrep -f write_to_hugetlbfs)" != "" ]]; then
-    echo killing write_to_hugetlbfs
-    killall -2 --wait write_to_hugetlbfs
-    wait_for_hugetlb_memory_to_get_depleted $cgroup
+  local write_pid
+
+  if (( ${#write_pids[@]} )); then
+    for write_pid in "${write_pids[@]}"; do
+      if kill -0 "$write_pid" 2>/dev/null; then
+        echo killing write_to_hugetlbfs pid "$write_pid"
+        kill -2 "$write_pid"
+        wait "$write_pid"
+      fi
+    done
+    write_pids=()
+
+    if [[ -n "$cgroup" ]]; then
+      wait_for_hugetlb_memory_to_get_depleted "$cgroup"
+    fi
   fi
   set -e
 
tools/testing/selftests/mm/write_hugetlb_memory.sh +1 -1
index 3d2d2eb9d..49164bbfc 100755
@@ -19,5 +19,5 @@ echo $$ > ${cgroup_path:-/dev/cgroup/memory}/"$cgroup"/cgroup.procs
 echo "Method is $method"
 
 set +e
-./write_to_hugetlbfs -p "$path" -s "$size" "$write" "$populate" -m "$method" \
+exec ./write_to_hugetlbfs -p "$path" -s "$size" "$write" "$populate" -m "$method" \
       "$private" "$want_sleep" "$reserve"
Signature · 1 line
2.39.5 (Apple Git-154)
Signature · 2 lines
Sincerely yours,
Mike.