Skip to contentSkip to conversation

cgroup_fj: don't disable controller that was already enabled

PATCH Applied Andrea Cervesato 4 participants 5 messages Jun 11 – Jun 12

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 cgroup_fj: don't disable controller that was already enabled

From: Andrea Cervesato <[email protected]>

On immutable distributions (and systemd-managed systems in general) the cpuset controller is typically already enabled at the root cgroup level. The cleanup tried to unconditionally disable it via subtree_control, which failed with EBUSY because other system cgroups were still using it, producing a spurious TWARN.

Save whether the controller was already enabled at root before cgroup_require and skip the root-level disable in cleanup when it was.

---
 testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh +9 -1
index 153d351d7936effb15527620f565acaaed7ac747..8b5e73466ba32d2c66a2805a03f4d2136c16509b 100755
@@ -57,6 +57,14 @@ create_subgroup()
 
 common_setup()
 {
+    # Check if the controller is already enabled at root before
+    # cgroup_require potentially enables it, so that cleanup does
+    # not disable a controller the system was already using.
+    if [ -f /sys/fs/cgroup/cgroup.subtree_control ] && \
+       grep -qw "$subsystem" /sys/fs/cgroup/cgroup.subtree_control; then
+        ctrl_was_enabled=1
+    fi
+
     cgroup_require "$subsystem"
     mount_point=$(cgroup_get_mountpoint "$subsystem")
     start_path=$(cgroup_get_test_path "$subsystem")
@@ -77,7 +85,7 @@ common_cleanup()
 
     cgroup_cleanup
 
-    if [ "$cgroup_version" = "2" ]; then
+    if [ "$cgroup_version" = "2" ] && [ "$ctrl_was_enabled" != "1" ]; then
         case "$subsystem" in
         cpu|io|memory|pids)
             :;;

---
base-commit: a375e8deed471723f5a0114c56eb48fe6f6f45d3
change-id: 20260611-fix_cgroup_immutable-2259f7591f8d

Best regards,

Signature · 5 lines
Andrea Cervesato <[email protected]>


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#1 lore raw

Hi Andrea,

Quoted 1 line · Andrea Cervesato

cgroup_fj: don't disable controller that was already enabled

Verdict: Reviewed

---
Note:

The agent can sometimes produce false positives although often its findings are genuine. If you find issues with the review, please comment this email or ignore the suggestions.

Regards,
LTP AI Reviewer

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by Wei Gao on v1 1/1
#2 lore raw
In reply to Andrea Cervesato
Quoted 50 lines · Andrea Cervesato

From: Andrea Cervesato <[email protected]>

On immutable distributions (and systemd-managed systems in general) the cpuset controller is typically already enabled at the root cgroup level. The cleanup tried to unconditionally disable it via subtree_control, which failed with EBUSY because other system cgroups were still using it, producing a spurious TWARN.

Save whether the controller was already enabled at root before cgroup_require and skip the root-level disable in cleanup when it was.

---
 testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh +9 -2
index 153d351d7936effb15527620f565acaaed7ac747..8b5e73466ba32d2c66a2805a03f4d2136c16509b 100755
@@ -57,6 +57,14 @@ create_subgroup()
 
 common_setup()
 {
+    # Check if the controller is already enabled at root before
+    # cgroup_require potentially enables it, so that cleanup does
+    # not disable a controller the system was already using.
+    if [ -f /sys/fs/cgroup/cgroup.subtree_control ] && \
+       grep -qw "$subsystem" /sys/fs/cgroup/cgroup.subtree_control; then
+        ctrl_was_enabled=1
+    fi
+
     cgroup_require "$subsystem"
     mount_point=$(cgroup_get_mountpoint "$subsystem")
     start_path=$(cgroup_get_test_path "$subsystem")
@@ -77,7 +85,7 @@ common_cleanup()
 
     cgroup_cleanup
 
-    if [ "$cgroup_version" = "2" ]; then
+    if [ "$cgroup_version" = "2" ] && [ "$ctrl_was_enabled" != "1" ]; then
         case "$subsystem" in
         cpu|io|memory|pids)
             :;;
 
---

base-commit: a375e8deed471723f5a0114c56eb48fe6f6f45d3
change-id: 20260611-fix_cgroup_immutable-2259f7591f8d

Quoted 3 lines

Best regards,

Signature · 1 line
Andrea Cervesato <[email protected]>
Quoted 4 lines
Signature · 2 lines
-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by
#3 lore raw
Quoted 52 lines · Wei Gao via ltp
Quoted 8 lines · Andrea Cervesato

From: Andrea Cervesato <[email protected]>

On immutable distributions (and systemd-managed systems in general) the cpuset controller is typically already enabled at the root cgroup level. The cleanup tried to unconditionally disable it via subtree_control, which failed with EBUSY because other system cgroups were still using it, producing a spurious TWARN.

Quoted 7 lines

Save whether the controller was already enabled at root before cgroup_require and skip the root-level disable in cleanup when it was.

---
 testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh | 10

+++++++++-

Review

Quoted 3 lines
 1 file changed, 9 insertions(+), 1 deletion(-)
(quoted diff) +0 -0

b/testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh

Quoted 1 line

index

153d351d7936effb15527620f565acaaed7ac747..8b5e73466ba32d2c66a2805a03f4d2136c16509b 100755

Quoted 18 lines
testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh +8 -0
@@ -57,6 +57,14 @@ create_subgroup()
 
 common_setup()
 {
+    # Check if the controller is already enabled at root before
+    # cgroup_require potentially enables it, so that cleanup does
+    # not disable a controller the system was already using.
+    if [ -f /sys/fs/cgroup/cgroup.subtree_control ] && \
+       grep -qw "$subsystem" /sys/fs/cgroup/cgroup.subtree_control; then
+        ctrl_was_enabled=1
+    fi
+
     cgroup_require "$subsystem"
     mount_point=$(cgroup_get_mountpoint "$subsystem")
     start_path=$(cgroup_get_test_path "$subsystem")

testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh common_cleanup() · line 88

Quoted 5 lines
testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh +1 -1
@@ -77,7 +85,7 @@ common_cleanup()
 
     cgroup_cleanup
 
-    if [ "$cgroup_version" = "2" ]; then
+    if [ "$cgroup_version" = "2" ] && [ "$ctrl_was_enabled" != "1" ];

then

Quoted 4 lines
testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh +0 -0
         case "$subsystem" in
         cpu|io|memory|pids)
             :;;
Quoted 9 lines
Quoted 7 lines

---
base-commit: a375e8deed471723f5a0114c56eb48fe6f6f45d3
change-id: 20260611-fix_cgroup_immutable-2259f7591f8d

Best regards,
--
Andrea Cervesato <[email protected]>

Feel free to add:

Quoted 8 lines
Quoted 4 lines

--
Mailing list info: https://lists.linux.it/listinfo/ltp

--
Mailing list info: https://lists.linux.it/listinfo/ltp

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp