[PATCH strace] Implement decoding of RXRPC_* control messages

[Dmitry V. Levin]

Hi,

On Fri, May 05, 2023 at 02:57:15PM +0100, David Howells wrote:
[...]
> @@ -212,6 +233,25 @@ print_cmsg_ip_origdstaddr(struct tcb *tcp, const void *cmsg_data,
>  	print_sockaddr(tcp, cmsg_data, addr_len);
>  }
>  
> +static void
> +print_cmsg_rxrpc_user_call_id(struct tcb *tcp, const void *cmsg_data,
> +			      const unsigned int data_len)
> +{
> +	const unsigned int *p = cmsg_data;
> +
> +	/* Size is "unsigned long" but we might be stracing a 32-bit program
> +	 * from a 64-bit strace or vice versa.
> +	 */
> +	if (data_len == 4)
> +		print_local_array_ex(tcp, p, 1, 4,
> +				     print_xint_array_member, NULL, 0, NULL, NULL);
> +	else if (data_len == 8)
> +		print_local_array_ex(tcp, p, 1, 8,
> +				     print_xint_array_member, NULL, 0, NULL, NULL);
> +	else
> +		print_cmsg_xint(tcp, cmsg_data, data_len);
> +}
> +

I suppose a switch statement would be more concise in this case.

> diff --git a/src/xlat/rxrpc_cmsg_types.in b/src/xlat/rxrpc_cmsg_types.in
> new file mode 100644
> index 000000000..7e045b5a6
> --- /dev/null
> +++ b/src/xlat/rxrpc_cmsg_types.in
> @@ -0,0 +1,12 @@
> +RXRPC_USER_CALL_ID	1
> +RXRPC_ABORT		2
> +RXRPC_ACK		3
> +RXRPC_NET_ERROR		5
> +RXRPC_BUSY		6
> +RXRPC_LOCAL_ERROR	7
> +RXRPC_NEW_CALL		8
> +RXRPC_EXCLUSIVE_CALL	10
> +RXRPC_UPGRADE_SERVICE	11
> +RXRPC_TX_LENGTH		12
> +RXRPC_SET_CALL_TIMEOUT	13
> +RXRPC_CHARGE_ACCEPT	14

Looks like this table would benefit from #value_indexed.

By the way, would it make sense to list RXRPC_ACCEPT here as well, even
though it was removed from the kernel by commit v4.9-rc1~127^2~247^2?

And the last but not least, do you have any ideas how this new decoder
could be tested?


-- 
ldv