[dev] New "printfrr()" functionality available in FRR master

David Lamparter equinox at diac24.net
Fri Jun 7 06:43:43 EDT 2019

Hi all,

we've recently merged (https://github.com/FRRouting/frr/pull/4333) new
string formatting code which allows printing network/FRR-specific data
types more easily.  It's inspired by the Linux kernel and works by
*appending* "sub-formats" after %p/%d printf specifiers.  The idea is
to remain compatible with normal %p/%d calling conventions (so compiler
warnings keep working), but the code knows more about the
pointer/integer that is passed.

The idea is that instead of doing:

if (debug) {
  char buf[];
  zlog_debug("%s", inet_ntop(AF_INET, &addr, buf, sizeof(buf)));

you can do:

if (debug)
  zlog_debug("%pI4", &addr);

For example, "%pI4" is used to print a "struct in_addr *" as A.B.C.D,
with the inet_ntop() call happening inside of printfrr() instead of
separately beforehand.

Note that for the compiler this just looks like "%p" and you're passing
any random pointer.  printfrr knows to interpret the "I4" and print the
pointer contents as an IPv4 address instead.

This is documented under doc/developer/logging.rst, cf.
Extended formats are registered with "printfrr_ext_autoreg", so you can
also grep for that to see what's available.  Currently, we have:

%pI4 (struct in_addr *)
%pI6 (struct in_addr6 *)
%pFX (struct prefix *)
%pSG4 (struct prefix_sg *)
%pNH (struct nexthop *)
%pRN (struct route_node *)

They aren't put to much use yet, I'm planning to slowly introduce them
and simplify existing code.  I also expect that we add more printers as
needed & we see fit.

NB: all the zlog_* and vty_out functions use printfrr internally, so
these extension work there.  However, if you call snprintf(), you need
to change that to snprintfrr()!

I also have a work-in-progress gcc plugin that provides additional
compiler warnings for these extensions.  However, it needs a patched
version of the gcc-python wrapper and isn't particularly easy to use, so
I'm still pondering how to best integrate that.  In any case, this is an
extension purely for improved warnings and is not needed at all to use
this functionality.



More information about the dev mailing list