Takes the same approach as when adding a new address - we no longer
announce the new lladdr right away but we announce the result.
This will either be RTM_ADD or RTM_MISS.
RTM_DELETE is only sent if we have a lladdr assigned OR gc'ed.
This results in less messages via route(4) and tells us when a new
lladdr has been added (RTM_ADD), changed (RTM_CHANGE), deleted (RTM_DELETED)
or has failed to been resolved (RTM_MISS). The latter case can be
interpreted as unreachable.
Taking softnet_lock there can cause a locking error with nfs sosend, so we don't.
Having only KERNEL_LOCK is enough because now the routing table is protected by
KERNEL_LOCK that was introduced by the fix for PR 53043.
PR kern/54227 from Paul Ripke
tree. Otherwise a 5.2 version of getifaddrs(2) gets errors.
This makes the 5.2 version of ifconfig(8) behave the same on both
NetBSD-8 and -current. HOWEVER, both of them print nothing (for
``ifconfig -l'' command) so there's still a bug somewhere.
As reported originally by der Mouse.
In the function the routing table could be accessed without any locks, which was
unsafe. Actually, on netbsd-7, a kernel panic happened(*). The situation of
locking hasn't changed since netbsd-7 so we still need to hold the big locks on
-current (and netbsd-8) too.
Note that if NET_MPSAFE is enabled, the routing table is protected by its own
lock and we don't need the locks.
Reported and tested on netbsd-7 by sborrill@
(*) http://mail-index.netbsd.org/tech-net/2018/11/08/msg007153.html
<subsystem>_<function>_<version>_hook
NFCI
XXX Note that although this introduces a change in the kernel-to-
XXX module interface, we are NOT bumping the kernel version number.
XXX We will bump the version number once the interface stabilizes.
This change sweeps remaining lock decisions based on if locked or not by moving
utility functions of rtentry updates from rtsock.c and ensuring holding the
rt_lock. It also improves the atomicity of a update of a rtentry.
When waiting for a route update to finish, a waiter has to release its reference
to the route to avoid a deadlock. Because a updater tries to wait for references
to a target route (except for a reference by the updater itself) to be released.
An ifp returned from route_output_get_ifa was supposed to be protected
by a returned ifa; if the ifa belongs to ifp, holding the ifa prevents
the ifp from being freed. However route_output_get_ifa can return an ifp
to which a returned ifa doesn't belong. So we need to take a reference
to a returning ifp separately.
It reduces C&P codes such as "#ifndef NET_MPSAFE KERNEL_LOCK(1, NULL); ..."
scattered all over the source code and makes it easy to identify remaining
KERNEL_LOCK and/or softnet_lock that are held even if NET_MPSAFE.
No functional change
keysock now protects itself by its own mutex, which means that
the rawcb list is protected by two different mutexes (keysock's one
and softnet_lock for rtsock), of course it's useless.
Fix the situation by having a discrete rawcb list for each.
route(8) passes a sockaddr for netmask that is truncated with its
prefixlen. However the kernel basically doesn't expect such format
and may read beyond the data. So restore the original length of the
the data at the beginning of the kernel for the rest components.
Failures of ATF tests such as route_flags_blackhole6 should
be fixed.
kmem_alloc() with KM_SLEEP
kmem_zalloc() with KM_SLEEP
percpu_alloc()
pserialize_create()
psref_class_create()
all of these paths include an assertion that the allocation has not failed,
so callers should not assert that again.
un-wanted route(4) messages.
Inspired by the ROUTE_MSGFILTER equivalent in OpenBSD,
but with an API which allows the full range of potential message types.
This change needs a tweak in route_output_change to unbreak route
change commands (e.g., route change -inet6 default -reject).
PR kern/52077 (s-yamaguchi@IIJ and ozaki-r@)
skrll