Re: supporting extend 'open live capture' parametes

[Guy Harris <guy () alum mit edu>]

Andy Howell wrote:

> How about having a generic list of options? Something like:
>
> typedef enum {
>   END_OF_OPTS,
>   PARAM_1,
>   PARAM_2,
> } pcap_opts;
>
> typedef union {
>   void           *p;
>   unsigned int   u;
> } pcap_opt_value;

That assumes that an option will either be an "unsigned int" or a pointer.

I'd be inclined to have something more like

        typedef struct {
                bpf_u_int32     option;
                bpf_u_int32     optlen;
        } pcap_opthdr;

with the buffer being a sequence of pcap_opthdr structures followed by 
option values.  "option" is the code for the option type (with 0 being 
the "end of option list" option), and "optlen" is the length of the 
option value (so an application can skip options it doesn't understand).

"pcap_opthdr" is 64 bits long, so as long as the option buffer is 
aligned on a {32-bit,64-bit} boundary, the options will be aligned on 
that boundary as well (and if the option buffer is mallocated, it'll be 
aligned on the strictest boundary required for C data types).

I'd also have pcap_open_live_ex() take a "pcap_opthdr *" as an argument 
(for one thing, that means no whining from the compiler that we're 
casting a "char *" to a type requiring stricter alignment).
        
> We would probably want a 
> functions that returns the valid options for a device.

That's his "pcap_get_capabilities()" function.
-
This is the tcpdump-workers list.
Visit https://cod.sandelman.ca/ to unsubscribe.