Internet on FIRE

Grrr, committing too fast.
#include "alloca.h" -> #include "lib/alloca.h"

#ifdef ALLOCA_H

This was a potential memory leak, but not the one Feela observed.

This one could happen when an import filter of some protocol modified
the rte (so that it would be rte_cow()ed) and later rejected it.

Fix %<something>I format strings.

Fixes to the progdoc.

Better description of the route distribution process.

Documentation.

Routing table garbage collector gets really called.

Asynchronous feeding of protocols.

Changed semantics of the rt_update hook. The attribute list we pass now
contains all attributes, not just the temporary ones. This avoids having
to merge the lists inside protocols or doing searches on both of them.

Also, do filtering of routes properly. (I'd like to avoid it, but it's...

Implemented `show route count' which is `show route stats' with
exception that it doesn't print the routes themselves.

Fixed `show route primary'.

Implemented `show route <...> stats'.

Fixed silly bug in previous commit.

Removed a lot of unused variables.

Please try compiling your code with --enable-warnings to see them. (The
unused parameter warnings are usually bogus, the unused variable ones
are very useful, but gcc is unable to control them separately.)

Event handlers no longer return re-queue flag. Instead of using it, just
call ev_schedule() on the same handler which should work perfectly now.

When tracing, always print incoming part of the trace before the outgoing one.

Avoid buffer overflows in `show routes' command.

Use FF_FORCE_TMPATTR where appropriate.

Protocol tracing: Don't dump core on filtered out routes.

Great cleanup of debug messages. LOCAL_DEBUG turned off in most modules,
several debug() calls converted to DBG.

Added debugging of communication between protocols and routing tables.
Just ask for "debug routes" if you want to see the routes and "debug filters"
if you want even the rejected ones.

If the user has specified identical preferences for instances of different
protocols, break the tie by comparing addresses, so we keep the ordering
unambiguous.

First attempt on dynamic reconfiguration. There are still lots of bugs
and problems to solve, but the hardest part works.

Except for special protocols (nowadays only the kernel syncer), don't
export host and link scope routes.

Events now return a value. If it's non-zero, the event is re-queued
for processing in next event cycle. This can be used to prevent background
actions (hint: user commands) from hogging the CPU for too long time.

Allow announces of rte's to protocols in FS_FEEDING state.
Else, we would get chicken-egg problems in the table-to-table protocol.

Added extra argument to rt_update hook which contains a pointer to the
temporary attribute list.

Removed TOS support. This simplifies many things a lot.

More changes to the kernel syncer.

o Now compatible with filtering.
o Learning of kernel routes supported only on CONFIG_SELF_CONSCIOUS
systems (on the others it's impossible to get it semantically correct).
o Learning now stores all of its routes in a separate fib and selects...

rte_update: Check sanity of incoming entries. Throw out (and log) all routes
to bogus prefixes and non-local routes to host scope addresses.

EFence helped to find using of already free rte's in rt_prune().

Moved sanity check of protocol state during annoucements to rte_announce.

struct proto again contains instance name (a copy of proto->cf->name).

die() -> bug() where appropriate.

Minor rte/rta interface changes:

o  rte can now contain a pointer to both cached and uncached rta. Protocols
     which don't need their own attribute caching can now just fill-in a rta,
     link it to rte without any calls to attribute cache and call rte_update()...

rte_update: Doesn't loop forever when multiple routes point to the same
destination.

Added pointer to network to RTE. The complications with passing NET separately
aren't worth 4 bytes per RTE.

rte_discard and rte_dump don't need net * as parameter.

After contemplating about RIP route timeouts for a long time, I've implemented
protocol callbacks for route insertion and deletion from the central table.
RIP should maintain its own per-protocol queue of existing routes, scan it
periodically and call rte_discard() for routes that have timed out.

Each protocol now hears even its own routes and needs to make its own
loop detection. This is needed since both RIP and OSPF handle multiple
neighbors and they need to redistribute routes learned from each neighbor
to the remaining ones.

Route update hook now gets network prefix as well as updated
route attributes.

Handle route deletion without segfaults. A bit more debug dumps.

Added debug dump function, but it's still empty :(

Parts of routing table code. Data structure declarations should be
complete now.