On Wed, Aug 21, 2024 at 8:31 AM David Hildenbrand <david@xxxxxxxxxx> wrote:
>
> Currently, running the charge_reserved_hugetlb.sh selftest we can
> sometimes observe something like:
>
> $ ./charge_reserved_hugetlb.sh -cgroup-v2
> ...
> write_result is 0
> After write:
> hugetlb_usage=0
> reserved_usage=10485760
> killing write_to_hugetlbfs
> Received 2.
> Deleting the memory
> Detach failure: Invalid argument
> umount: /mnt/huge: target is busy.
>
> Both cases are issues in the test.
>
> While the unmount error seems to be racy, it will make the test fail:
> $ ./run_vmtests.sh -t hugetlb
> ...
> # [FAIL]
> not ok 10 charge_reserved_hugetlb.sh -cgroup-v2 # exit=32
>
> The issue is that we are not waiting for the write_to_hugetlbfs process
> to quit. So it might still have a hugetlbfs file open, about which
> umount is not happy. Fix that by making "killall" wait for the process
> to quit.
>
> The other error ("Detach failure: Invalid argument") does not seem to
> result in a test error, but is misleading. Turns out write_to_hugetlbfs.c
> unconditionally tries to cleanup using shmdt(), even when we only
> mmap()'ed a hugetlb file. Even worse, shmaddr is never even set for the
> SHM case. Fix that as well.
>
> With this change it seems to work as expected.
>
> Fixes: 29750f71a9b4 ("hugetlb_cgroup: add hugetlb_cgroup reservation tests")
> Reported-by: Mario Casquero <mcasquer@xxxxxxxxxx>
> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> Cc: Shuah Khan <shuah@xxxxxxxxxx>
> Cc: Muchun Song <muchun.song@xxxxxxxxx>
> Cc: Mina Almasry <almasrymina@xxxxxxxxxx>
> Signed-off-by: David Hildenbrand <david@xxxxxxxxxx>
Initially I thought it could be nice to split fixes for the 2 issues
in separate patches in case one of them ends up needing a revert or
something, but probably not worth a respin. Fixes look good to me.
Reviewed-by: Mina Almasry <almasrymina@xxxxxxxxxx>
> ---
> .../selftests/mm/charge_reserved_hugetlb.sh | 2 +-
> .../testing/selftests/mm/write_to_hugetlbfs.c | 21 +++++++++++--------
> 2 files changed, 13 insertions(+), 10 deletions(-)
>
> diff --git a/tools/testing/selftests/mm/charge_reserved_hugetlb.sh b/tools/testing/selftests/mm/charge_reserved_hugetlb.sh
> index d680c00d2853a..67df7b47087f0 100755
> --- a/tools/testing/selftests/mm/charge_reserved_hugetlb.sh
> +++ b/tools/testing/selftests/mm/charge_reserved_hugetlb.sh
> @@ -254,7 +254,7 @@ function cleanup_hugetlb_memory() {
> local cgroup="$1"
> if [[ "$(pgrep -f write_to_hugetlbfs)" != "" ]]; then
> echo killing write_to_hugetlbfs
> - killall -2 write_to_hugetlbfs
> + killall -2 --wait write_to_hugetlbfs
This looks correct. I don't think I expected killall not to wait.
> wait_for_hugetlb_memory_to_get_depleted $cgroup
> fi
> set -e
> diff --git a/tools/testing/selftests/mm/write_to_hugetlbfs.c b/tools/testing/selftests/mm/write_to_hugetlbfs.c
> index 6a2caba19ee1d..1289d311efd70 100644
> --- a/tools/testing/selftests/mm/write_to_hugetlbfs.c
> +++ b/tools/testing/selftests/mm/write_to_hugetlbfs.c
> @@ -28,7 +28,7 @@ enum method {
>
> /* Global variables. */
> static const char *self;
> -static char *shmaddr;
> +static int *shmaddr;
> static int shmid;
>
> /*
> @@ -47,15 +47,17 @@ void sig_handler(int signo)
> {
> printf("Received %d.\n", signo);
> if (signo == SIGINT) {
> - printf("Deleting the memory\n");
> - if (shmdt((const void *)shmaddr) != 0) {
> - perror("Detach failure");
> + if (shmaddr) {
> + printf("Deleting the memory\n");
> + if (shmdt((const void *)shmaddr) != 0) {
> + perror("Detach failure");
> + shmctl(shmid, IPC_RMID, NULL);
> + exit(4);
> + }
> +
> shmctl(shmid, IPC_RMID, NULL);
> - exit(4);
> + printf("Done deleting the memory\n");
> }
> -
> - shmctl(shmid, IPC_RMID, NULL);
> - printf("Done deleting the memory\n");
This seems like a simple refactor to only delete when shmaddr is set,
looks fine to me.
> }
> exit(2);
> }
> @@ -211,7 +213,8 @@ int main(int argc, char **argv)
> shmctl(shmid, IPC_RMID, NULL);
> exit(2);
> }
> - printf("shmaddr: %p\n", ptr);
> + shmaddr = ptr;
> + printf("shmaddr: %p\n", shmaddr);
>
Setting shmaddr seems correct and an oversight. I don't see shmaddr
set anywhere in the current code.
> break;
> default:
> --
> 2.46.0
>
--
Thanks,
Mina
