The API is used to set (or reset) a received interface of a mbuf.
They are counterpart of m_get_rcvif, which will come in another
commit, hide internal of rcvif operation, and reduce the diff of
the upcoming change.
No functional change.
Note that there is an issue that ioctls for an interface and a destruction
of the interface can run in parallel and it causes race conditions on
bridge as well (it rarely happens). The issue will be addressed in the
interface common code (if.c).
Thanks to introducing softint-based if_input, the entire bridge code now
never run in hardware interrupt context. So we can simplify the code.
- Remove spin mutexes
- They were needed because some code of bridge could run in
hardware interrupt context
- We now need only an adaptive mutex for each shared object
(a member list and a forwarding table)
- Remove pktqueue
- bridge_input is already in softint, using another softint
(for bridge_forward) is useless
- Packet distribution should be down at device drivers
This change enables lockless accesses to bridge member lists.
See locking notes in a comment to know how pserialize and
mutexes are used.
This change also provides support for softint-based interrupt
handling; pserialize readers can run in both HW interrupt and
softint contexts.
As usual, pserialize is used only when NET_MPSAFE on.
- Introduce BRIDGE_MPSAFE
- It's enabled only when NET_MPSAFE is defined
in if.h or the kernel config
- Add iflist and rtlist mutex locks
- Locking iflist is performance sensitive,
so it's not used when !BRIDGE_MPSAFE
- Add bif object reference counting
- It enables fine-grain locking for bridge member lists
by allowing to not hold a lock during touching a bif
- bridge_release_member is added to decrement the
reference count
- A condition variable is added to do bridge_delete_member
gracefully
- Add if_bridgeif to ifnet
- It's a shortcut to a bif object of a bridge member
- It reduces a bif lookup cost and so lock contention on iflist
- Make bridgestp MPSAFE too
The network stack of NetBSD is well organized and
layered. A packet reception is processed from a
lower layer to an upper layer one by one. However,
ether_input and bridge_input are not structured so.
bridge_input is called inside ether_input.
The new structure replaces ifnet#if_input of a bridge
member with bridge_input when the member is attached.
So a packet goes straight on a packet reception via
a bridge, bridge_input => ether_input => ip_input.
The change is part of a patch of Lloyd Parkes submitted
in PR 48104. Unlike the patch, the change doesn't
intend to change the behavior of the packet processing.
Another patch will fix PR 48104.
* A sign extension error creating the bridge ID corrupted the
priority (always making it the maximum).
* Do not catch STP packets on an interface for which STP is not
enabled -- it's a violation of the spec, and causes STP to fail on
neighboring bridges.
* An optimization to bstp_input() -- some information is already
known when we call it.
contributed anonymously.
identify sockaddr_dl abuse that remains in the kernel, especially
the potential for overwriting memory past the end of a sockaddr_dl
with, e.g., memcpy(LLADDR(), ...).
Use sockaddr_dl_setaddr() in a few places.
