the address belongs to.
This allows any ARP message we receieved from another interface to
be correctly dropped.
While here, move the protocol length check higher up the food chain.
It was created to copy FreeBSD, however actually the cache isn't
necessary. Remove it to simplify the code and reduce the cost to
maintain it (e.g., keep a consistency with a corresponding local
route).
inp can be NULL when receiving an IPv4 packet on an IPv4-mapped IPv6
address. In that case KASSERT(sotoinpcb(so) == inp) always fails.
Should fix PR kern/52304 (at least it fixes the same panic as the
report)
therefore don't pass in a stack allocated buffer
vmstat -mv shows pool(s) with broken name(s)
use the name argument passed into wqinput_create directly
which is a static string in all 4 callee cases
(workqueue_create/workqueue_init copies the name argument)
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.
An in_ifaddr object is initialized with refcnt=0 and the refcnt
is incremented when being enqueued to the lists. However before
enqueuing it, in_ifinit can hold and refelease a reference to
it, i.e., call ifaref and ifafree, resulting in that the object
is freed in ifafree because its refcnt is decremented to 0.
It can be reproduced by doing:
ifconfig tun0 create
ifconfig tun1 create
ifconfig tun0 10.1 10.2
ifconfig tun1 10.2 10.1
ifconfig # Cause a kernel panic (may depend on environmemts)
We need to initialize a created in_ifaddr object with refcnt=1
to make the object survive over in_ifinit.
The issue is found by ryo@
If the handler is delayed because of the indirection call via softint,
some operations are executed in reverse and may cause unexpected
behaviors. For example, due to the issue a GARP packet wasn't sent on
a transition from the BACKUP state to the MASTER state; this happened
because IN_IFF_DETACHED flag wasn't cleared on arpannounce, which
had been cleared in the link_state_change handler.
This fixes an issue reported by sborrill@ on tech-net:
http://mail-index.netbsd.org/tech-net/2017/03/14/msg006283.html
Note that we should modularize netipsec and reduce reverse symbol references
(referencing symbols of netipsec from net, netinet and netinet6) though,
the task needs lots of code changes. Prior to doing so, rumpifying it and
having ATF tests should be useful.
- Allow encapinit to be called twice (by ifinit and ipe4_attach)
- ifinit didn't call encapinit if IPSEC is enabled (ipe4_attach called
it instead), however, on a rump kernel ipe4_attach may not be called
even if IPSEC is enabled. So we need to allow ifinit to call it anyway
rjs