David Hildenbrand (Red Hat) david@kernel.org wrote:
On 12/21/25 09:58, Li Wang wrote:
The hugetlb cgroup usage wait loops in charge_reserved_hugetlb.sh were unbounded and could hang forever if the expected cgroup file value never appears (e.g. due to bugs, timing issues, or unexpected behavior).
Did you actually hit that in practice? Just wondering.
Yes.
On an aarch64 64k setup with 512MB hugepages, the test failed earlier (hugetlbfs got mounted with an effective size of 0 due to size=256M), so write_to_hugetlbfs couldn’t allocate the expected pages. After that, the script’s wait loops never observed the target value, so they spun forever.
Detail see below logs.
--- Error log --- # uname -r 6.12.0-xxx.el10.aarch64+64k
# ls /sys/kernel/mm/hugepages/hugepages-* hugepages-16777216kB/ hugepages-2048kB/ hugepages-524288kB/
#./charge_reserved_hugetlb.sh -cgroup-v2 # ----------------------------------------- ... # nr hugepages = 10 # writing cgroup limit: 5368709120 # writing reseravation limit: 5368709120 ... # write_to_hugetlbfs: Error mapping the file: Cannot allocate memory # Waiting for hugetlb memory reservation to reach size 2684354560. # 0 # Waiting for hugetlb memory reservation to reach size 2684354560. # 0 # Waiting for hugetlb memory reservation to reach size 2684354560. # 0 # Waiting for hugetlb memory reservation to reach size 2684354560. # 0 # Waiting for hugetlb memory reservation to reach size 2684354560. # 0 # Waiting for hugetlb memory reservation to reach size 2684354560. # 0 ...
Introduce a small helper, wait_for_file_value(), and use it for:
- waiting for reservation usage to drop to 0,
- waiting for reservation usage to reach a given size,
- waiting for fault usage to reach a given size.
This makes the waits consistent and adds a hard timeout (120 tries with 0.5s sleep) so the test fails instead of stalling indefinitely.
Signed-off-by: Li Wang liwang@redhat.com Cc: David Hildenbrand david@kernel.org Cc: Mark Brown broonie@kernel.org Cc: Shuah Khan shuah@kernel.org Cc: Waiman Long longman@redhat.com
.../selftests/mm/charge_reserved_hugetlb.sh | 47 ++++++++++--------- 1 file changed, 26 insertions(+), 21 deletions(-)
diff --git a/tools/testing/selftests/mm/charge_reserved_hugetlb.sh b/tools/testing/selftests/mm/charge_reserved_hugetlb.sh index e1fe16bcbbe8..249a5776c074 100755 --- a/tools/testing/selftests/mm/charge_reserved_hugetlb.sh +++ b/tools/testing/selftests/mm/charge_reserved_hugetlb.sh @@ -100,7 +100,7 @@ function setup_cgroup() { echo writing cgroup limit: "$cgroup_limit" echo "$cgroup_limit" >$cgroup_path/$name/hugetlb.${MB}MB.$fault_limit_file
- echo writing reseravation limit: "$reservation_limit"
- echo writing reservation limit: "$reservation_limit" echo "$reservation_limit" > \ $cgroup_path/$name/hugetlb.${MB}MB.$reservation_limit_file
@@ -112,41 +112,46 @@ function setup_cgroup() { fi }
+function wait_for_file_value() {
- local path="$1"
- local expect="$2"
- local max_tries="120"
- local i cur
I would just move "cur" into the loop; I don't see a reason to print it on the error path when you printed the value on the last "Waiting" line?
local cur="$(cat "$path")"
+1
Also, not sure if you really need the "local i" here.
What if the path does not exist, do we want to catch that earlier and bail out instead of letting "cat" fail here?
Yes, we can add a file check before the "cat" loop.
- for ((i=1; i<=max_tries; i++)); do
- cur="$(cat "$path")"
- if [[ "$cur" == "$expect" ]]; then
return 0- fi
- echo "Waiting for $path to become '$expect' (current: '$cur') (try $i/$max_tries)"
- sleep 0.5
Any reason we don't go for the more intuitive "wait 1s" - max 60s wait?
Sure, the total loop time are same.