On 2/6/17 4:03 PM, David Ahern wrote:
>> This doesn't look like a good solution, you loose the value of unspec.
>>
>> Just to clarify. You added a custom protocol value to netlink.
>> And then you are using upstream iproute2 source to display the value.
>
> no. I am saying the string displayed for protocol '0' is changing. This
> is all within iproute2 code and files; it has 2 strings for protocol 0:
>
> lib/rt_names.c:
> static char *rtnl_rtprot_tab[256] = {
> [RTPROT_UNSPEC] = "none",
>
> and the rt_protos file above shows "unspec"
>
> The presence of a custom protocol triggers the rt_protos file to be read:
>
> const char *rtnl_rtprot_n2a(int id, char *buf, int len)
> {
> if (id < 0 || id >= 256) {
> snprintf(buf, len, "%u", id);
> return buf;
> }
> if (!rtnl_rtprot_tab[id]) {
> if (!rtnl_rtprot_init)
> rtnl_rtprot_initialize();
>
>
> Reading the file changes the string in rtnl_rtprot_tab for
> RTPROT_UNSPEC. Both string values -- "none" and "unspec" come from
> iproute2, so my point is that string is inconsistent within iproute2.
>
You rejected the patch in patchworks. Do you understand my point above?
This is an iproute2 problem. It is the existence of a custom protocol
that triggers the iproute2 bug.
