Re: net/core/sock.c lacks some SO_TIMESTAMPING_NEW support

[Willem de Bruijn <willemdebruijn.kernel@gmail.com>]

Thomas Lange wrote:
> It seems that net/core/sock.c is missing support for SO_TIMESTAMPING_NEW in
> some paths.
> 
> I cross compile for a 32bit ARM system using Yocto 4.3.1, which seems to have
> 64bit time by default. This maps SO_TIMESTAMPING to SO_TIMESTAMPING_NEW which
> is expected AFAIK.
> 
> However, this breaks my application (Chrony) that sends SO_TIMESTAMPING as
> a cmsg:
> 
> sendmsg(4, {msg_name={sa_family=AF_INET, sin_port=htons(123), sin_addr=inet_addr("172.16.11.22")}, msg_namelen=16, msg_iov=[{iov_base="#\0\6 \0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., iov_len=48}], msg_iovlen=1, msg_control=[{cmsg_len=16, cmsg_level=SOL_SOCKET, cmsg_type=SO_TIMESTAMPING_NEW, cmsg_data=???}], msg_controllen=16, msg_flags=0}, 0) = -1 EINVAL (Invalid argument)
> 
> This is because __sock_cmsg_send() does not support SO_TIMESTAMPING_NEW as-is.
> 
> This patch seems to fix things and the packet is transmitted:
> 
> diff --git a/net/core/sock.c b/net/core/sock.c
> index 16584e2dd648..a56ec1d492c9 100644
> --- a/net/core/sock.c
> +++ b/net/core/sock.c
> @@ -2821,6 +2821,7 @@ int __sock_cmsg_send(struct sock *sk, struct cmsghdr *cmsg,
>                  sockc->mark = *(u32 *)CMSG_DATA(cmsg);
>                  break;
>          case SO_TIMESTAMPING_OLD:
> +       case SO_TIMESTAMPING_NEW:
>                  if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
>                          return -EINVAL;
> 
> However, looking through the module, it seems that sk_getsockopt() has no
> support for SO_TIMESTAMPING_NEW either, but sk_setsockopt() has.

Good point. Adding the author to see if this was a simple oversight or
there was a rationale at the time for leaving it out.

> Testing seems to confirm this:
> 
> setsockopt(4, SOL_SOCKET, SO_TIMESTAMPING_NEW, [1048], 4) = 0
> getsockopt(4, SOL_SOCKET, SO_TIMESTAMPING_NEW, 0x7ed5db20, [4]) = -1 ENOPROTOOPT (Protocol not available)
> 
> Patching sk_getsockopt() is not as obvious to me though.
> 
> I used a custom 6.6 kernel for my tests.
> The relevant code seems unchanged in net-next.git though.
> 
> /Thomas