Re: [PATCH] ipc/msg: add msgsnd_timed and msgrcv_timed syscall for system V message queue

[Arnd Bergmann <arnd@kernel.org>]

On Sat, Feb 27, 2021 at 7:52 AM Eric Gao <eric.tech@foxmail.com> wrote:
>
> sometimes, we need the msgsnd or msgrcv syscall can return after a limited
> time, so that the business thread do not be blocked here all the time. In
> this case, I add the msgsnd_timed and msgrcv_timed syscall that with time
> parameter, which has a unit of ms.
>
> Signed-off-by: Eric Gao <eric.tech@foxmail.com>

I have no opinion on whether we want or need this, but I'll have a look
at the implementation, to see if the ABI makes sense.

> index 8fd8c17..42b7db5 100644
> --- a/arch/mips/kernel/syscalls/syscall_n32.tbl
> +++ b/arch/mips/kernel/syscalls/syscall_n32.tbl
> @@ -381,3 +381,5 @@
>  440    n32     process_madvise                 sys_process_madvise
>  441    n32     epoll_pwait2                    compat_sys_epoll_pwait2
>  442    n32     mount_setattr                   sys_mount_setattr
> +443    n32     msgrcv_timed                    sys_msgrcv_timed
> +444    n32     msgsnd_timed                    sys_msgsnd_timed
> diff --git a/arch/mips/kernel/syscalls/syscall_o32.tbl b/arch/mips/kernel/syscalls/syscall_o32.tbl
> index 090d29c..0f1f6ee 100644
> --- a/arch/mips/kernel/syscalls/syscall_o32.tbl
> +++ b/arch/mips/kernel/syscalls/syscall_o32.tbl
> @@ -430,3 +430,5 @@
>  440    o32     process_madvise                 sys_process_madvise
>  441    o32     epoll_pwait2                    sys_epoll_pwait2                compat_sys_epoll_pwait2
>  442    o32     mount_setattr                   sys_mount_setattr
> +443    o32     msgrcv_timed                    sys_msgrcv_timed
> +444    o32     msgsnd_timed                    sys_msgsnd_timed

I think mips n32 and o32 both need to use the compat version when running on
a 64-bit kernel, while your patch makes them use the native version.

> @@ -905,7 +906,15 @@ static long do_msgsnd(int msqid, long mtype, void __user *mtext,
>
>                 ipc_unlock_object(&msq->q_perm);
>                 rcu_read_unlock();
> -               schedule();
> +
> +               /* sometimes, we need msgsnd syscall return after a given time */
> +               if (timeoutms <= 0) {
> +                       schedule();
> +               } else {
> +                       timeoutms = schedule_timeout(timeoutms);
> +                       if (timeoutms == 0)
> +                               timeoutflag = true;
> +               }

I wonder if this should be schedule_timeout_interruptible() or at least
schedule_timeout_killable() instead of schedule_timeout(). If it should,
this should probably be done as a separate change.

> +COMPAT_SYSCALL_DEFINE5(msgsnd_timed, int, msqid, compat_uptr_t, msgp,
> +                      compat_ssize_t, msgsz, int, msgflg, compat_long_t, timeoutms)
> +{
> +       struct compat_msgbuf __user *up = compat_ptr(msgp);
> +       compat_long_t mtype;
> +
> +       timeoutms = (timeoutms + 9) / 10;
> +
> +       if (get_user(mtype, &up->mtype))
> +               return -EFAULT;
> +
> +       return do_msgsnd(msqid, mtype, up->mtext, (ssize_t)msgsz, msgflg, (long)timeoutms);
> +}

My preference would be to simplify both the timed and non-timed version by
moving the get_user() into do_msgsnd() and using in_compat_task() to pick
the right type. Same for the receive side of course. If you do this,
watch out for
x32 support, which uses the 64-bit version.

       Arnd