Huge thanks to our Platinum Members Endace and LiveAction,
and our Silver Member Veeam, for supporting the Wireshark Foundation and project.

Wireshark-dev: Re: [Wireshark-dev] Correct and efficient way of displaying bit fields?

From: Guy Harris <guy@xxxxxxxxxxxx>
Date: Sat, 8 Oct 2011 13:12:55 -0700
On Oct 7, 2011, at 2:22 PM, Kaul wrote:

> I'm struggling for some time now with displaying bitfields, I'm sure there must be something I'm overlooking, or it's just a bit difficult to do in Wireshark.
> 
> I have a 32bit, little endian field, which I'd like to parse the bits (as set/not set):
> Example:
> 05 00 00 00
> 
> 1 0 0 0 .... Feature A - set
> 0 0 0 0 ... Feature B - not set
> 0 0 1 0 ... Feature C - Set

That's 0xA, not 0x5; presumably, with the uppermost bit shown on the left and the lowermost bit shown on the right - the convention I've seen almost universally used - it'd be

0 0 0 1 .... Feature A - set
0 0 0 0 ... Feature B - not set
0 1 0 0 ... Feature C - Set

> 1. Do I really have to create a hf_xxx for each? And use something like proto_tree_add_bits_item() ? I was hoping to do it in a single proto_tree_add_xxx() and pass it a single HF that would hold a VALS(...) which will describe all the attributes.

Assuming the field is 4 bits, you could create a single item for those four bits, with 16 values.  In that case, that particular example would as something such as

	Feature set: 0x5 (feature A present, feature B not present, feature C present, feature D not present)

Is that what you want?

There is nothing in Wireshark that lets you show each flag as a single Boolean item with one field - you have (at least) four Booleans, hence you would need (at least) four fields.

> 2. How do I take into consideration the endianess?

By not using bit offsets.

> Best I could do so far, it works but it's ugly and not maintainable, is:

        #define COMMON_CAP_AUTH_SELECT  0x00000001
        #define COMMON_CAP_AUTH_SPICE   0x00000002
        #define COMMON_CAP_AUTH_SASL    0x00000004

                ...

        proto_tree_add_item(tree, hf_common_cap_auth_select, tvb, offset, 4, ENC_LITTLE_ENDIAN);
        proto_tree_add_item(tree, hf_common_cap_auth_spice, tvb, offset, 4, ENC_LITTLE_ENDIAN);
        proto_tree_add_item(tree, hf_common_cap_auth_sasl, tvb, offset, 4, ENC_LITTLE_ENDIAN);

                ...

        { &hf_common_cap_auth_select,
          { "Auth Selection", "spice.common_cap_auth_select",
            FT_BOOLEAN, 32, TFS(&tfs_set_notset), COMMON_CAP_AUTH_SELECT,
            NULL, HFILL }
        },
        { &hf_common_cap_auth_spice,
          { "Auth Spice", "spice.common_cap_auth_spice",
            FT_BOOLEAN, 32, TFS(&tfs_set_notset), COMMON_CAP_AUTH_SPICE,
            NULL, HFILL }
        },
        { &hf_common_cap_auth_sasl,
          { "Auth SASL", "spice.common_cap_auth_sasl",
            FT_BOOLEAN, 32, TFS(&tfs_set_notset), COMMON_CAP_AUTH_SASL,
            NULL, HFILL }
        },

> If I look at how it's done in packet-tcp.c, then it's again quite a bit of manual labour, this time with proto_tree_add_boolean() - per each single bit!

Well, yeah, each bit is a single Boolean field, each of which a user might want to check for in a filter expression, so of *course* there will be one call per bit, to put an item for that bit into the protocol tree!

And as for manual labor, well, the whole dissector was constructed with a lot of manual labor; the best way to get rid of the manual labor is to have a packet description language such as, for example, the wsgd language:

	http://wsgd.free.fr

and have something read that and generate code (or tables processed by an interpreter, or whatever).  Worrying only about fields that happen to take only one bit is worrying about a very minor concern.