Re: [PATCH] t_ofd_locks: fix initialization sequence

[stsp <stsp2@yandex.ru>]

06.07.2023 13:41, Murphy Zhou пишет:
> Signed-off-by: Stas Sergeev <stsp2@yandex.ru>
> ---
>   src/t_ofd_locks.c | 73 ++++++++++++++++++++++++++++++-----------------
>   1 file changed, 47 insertions(+), 26 deletions(-)
>
> diff --git a/src/t_ofd_locks.c b/src/t_ofd_locks.c
> index e77f2659..daa6f96c 100644
> --- a/src/t_ofd_locks.c
> +++ b/src/t_ofd_locks.c
> @@ -297,6 +297,7 @@ int main(int argc, char **argv)
>                        semid = semget(semkey, 2, IPC_CREAT|IPC_EXCL);
>                        if (semid < 0 && errno == EEXIST) {
>                                /* remove sem set after one round of test */
> +                             semid = semget(semkey, 2, IPC_CREAT);
>                                if (semctl(semid, 2, IPC_RMID, semu) == -1)
> Good catch. This RMID is useless unless we have got the existing
> semaphore. According to SEMGET(2), seems should be:
>
>      semid = semget(semkey, 2, 0);
>
> to obtain an existing semaphore?

Yes, I just wanted to avoid the purely
theoretical condition when someone
else removed this sem right before we
did second semget(). So I added IPC_CREAT
just as a safety measure.
Should I remove it?

> The while loop makes sure we get the semaphore before continuing
> the test. It's been some time, I'm not sure but now I really can't see
> this really hurts.
What while loop do you mean and what
doesn't hurt? Does the rest of the patch
look ok?