Re: [LTP] semctl: updates for multi worker testing.

[linuxtestproject.agent@gmail.com]

Hi Stephen,

On Tue, 14 Apr 2026, Stephen Bertram wrote:
> semctl: updates for multi worker testing.
>
> Changed semctl01, which required updates to
> reduce interference and provide an allowance
> for EINVAL and EIDRM failures.

The body should say *why* multi-worker support is needed — what harness
or runner exercises this in parallel, or what failure prompted the change.

> -static int sem_id = -1;
> -static int sem_index;
> +static __thread int sem_id = -1;
> +static __thread int sem_index;
>  static struct semid_ds buf;
>  static struct seminfo ipc_buf;
>  static unsigned short array[PSEMS];
>  static struct sembuf sops;
>  static int pid_arr[NCHILD];

buf, ipc_buf, array, sops, pid_arr are still plain globals. Making only
sem_id/sem_index TLS leaves all the helper functions that read/write
these globals unsafe for concurrent use, so the parallelisation is
incomplete.

[...]

> +		if (sid == -1)
> +			sem_id = sid = SAFE_SEMGET(IPC_PRIVATE, PSEMS,
> +						   IPC_CREAT | IPC_EXCL | SEM_RA);

cleanup() runs in the main thread and only sees the main thread's TLS
sem_id. Worker threads that hit this path will leak their semaphore sets.
Either track all allocated IDs in a shared, mutex-protected list, or drop
__thread and keep the test single-threaded.

Also note struct tcases::semid is now always NULL and never read; the
field can be removed.

---
Note:

Our agent completed the review of the patch. The full review can be
found at: https://patchwork.ozlabs.org/project/ltp/list/?series=499874

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.

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