Re: [Qemu-devel] [PATCH v6 1/8] linux-user: Add support for semtimedop() syscall

[Aleksandar Markovic]

On Wed, Sep 4, 2019 at 3:11 PM Aleksandar Markovic <
address@hidden> wrote:

> From: Aleksandar Rikalo <address@hidden>
>
> Add support for semtimedop() emulation. It is based on invocation
> of safe_semtimedop().
>
> Conversion is left out of safe_semtimedop(), since other safe_xxx()
> usually don't contain similar conversions.
>
> Signed-off-by: Aleksandar Rikalo <address@hidden>
> Signed-off-by: Aleksandar Markovic <address@hidden>
> ---
>

Hi, Laurent.

Please note that Aleksandar R. proposed also a code segment mentioned
in this message:

https://lists.gnu.org/archive/html/qemu-devel/2019-07/msg07080.html

... but wasn't sure if he got all bits and peaces right. Please provide your
judgement - and I will modify the patch accordingly.

Thanks,
aleksandar


>  linux-user/syscall.c | 40 ++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 40 insertions(+)
>
> diff --git a/linux-user/syscall.c b/linux-user/syscall.c
> index 8367cb1..b5bc6e4 100644
> --- a/linux-user/syscall.c
> +++ b/linux-user/syscall.c
> @@ -6649,7 +6649,43 @@ static inline abi_long host_to_target_statx(struct
> target_statx *host_stx,
>      return 0;
>  }
>  #endif
> +#ifdef TARGET_NR_semtimedop
> +static inline abi_long do_semtimedop(int semid, abi_long ptr, unsigned
> nsops,
> +                                     abi_long timeout)
> +{
> +    struct sembuf *sops;
> +    struct timespec ts, *pts;
> +    abi_long ret;
> +
> +    if (timeout) {
> +        pts = &ts;
> +        if (target_to_host_timespec(pts, timeout)) {
> +            return -TARGET_EFAULT;
> +        }
> +    } else {
> +        pts = NULL;
> +    }
>
> +    sops = g_malloc(sizeof(struct sembuf) * nsops);
> +    if (sops == NULL) {
> +        return -TARGET_EFAULT;
> +    }
> +
> +    if (target_to_host_sembuf(sops, ptr, nsops)) {
> +        g_free(sops);
> +        return -TARGET_EFAULT;
> +    }
> +
> +#ifdef __NR_semtimedop
> +    ret = get_errno(safe_semtimedop(semid, sops, nsops, pts));
> +#else
> +    ret = -TARGET_ENOSYS;
> +#endif
> +    g_free(sops);
> +
> +    return ret;
> +}
> +#endif
>
>  /* ??? Using host futex calls even when target atomic operations
>     are not really atomic probably breaks things.  However implementing
> @@ -9193,6 +9229,10 @@ static abi_long do_syscall1(void *cpu_env, int num,
> abi_long arg1,
>      case TARGET_NR_semop:
>          return do_semop(arg1, arg2, arg3);
>  #endif
> +#ifdef TARGET_NR_semtimedop
> +    case TARGET_NR_semtimedop:
> +        return do_semtimedop(arg1, arg2, arg3, arg4);
> +#endif
>  #ifdef TARGET_NR_semctl
>      case TARGET_NR_semctl:
>          return do_semctl(arg1, arg2, arg3, arg4);
> --
> 2.7.4
>
>
>