This patchset implements a new TX lock, covering both the per-VAP (and
thus per-node) TX locking and the serialisation through to the underlying
physical device.
This implements the hard requirement that frames to the underlying physical
device are scheduled to the underlying device in the same order that they
are processed at the VAP layer. This includes adding extra encapsulation
state (such as sequence numbers and CCMP IV numbers.) Any order mismatch
here will result in dropped packets at the receiver.
There are multiple transmit contexts from the upper protocol layers as well
as the "raw" interface via the management and BPF transmit paths.
All of these need to be correctly serialised or bad behaviour will result
under load.
The specifics:
* add a new TX IC lock - it will eventually just be used for serialisation
to the underlying physical device but for now it's used for both the
VAP encapsulation/serialisation and the physical device dispatch.
This lock is specifically non-recursive.
* Methodize the parent transmit, vap transmit and ic_raw_xmit function
pointers; use lock assertions in the parent/vap transmit routines.
* Add a lock assertion in ieee80211_encap() - the TX lock must be held
here to guarantee sensible behaviour.
* Refactor out the packet sending code from ieee80211_start() - now
ieee80211_start() is just a loop over the ifnet queue and it dispatches
each VAP packet send through ieee80211_start_pkt().
Yes, I will likely rename ieee80211_start_pkt() to something that
better reflects its status as a VAP packet transmit path. More on
that later.
* Add locking around the management and BAR TX sending - to ensure that
encapsulation and TX are done hand-in-hand.
* Add locking in the mesh code - again, to ensure that encapsulation
and mesh transmit are done hand-in-hand.
* Add locking around the power save queue and ageq handling, when
dispatching to the parent interface.
* Add locking around the WDS handoff.
* Add a note in the mesh dispatch code that the TX path needs to be
re-thought-out - right now it's doing a direct parent device transmit
rather than going via the vap layer. It may "work", but it's likely
incorrect (as it bypasses any possible per-node power save and
aggregation handling.)
Why not a per-VAP or per-node lock?
Because in order to ensure per-VAP ordering, we'd have to hold the
VAP lock across parent->if_transmit(). There are a few problems
with this:
* There's some state being setup during each driver transmit - specifically,
the encryption encap / CCMP IV setup. That should eventually be dragged
back into the encapsulation phase but for now it lives in the driver TX path.
This should be locked.
* Two drivers (ath, iwn) re-use the node->ni_txseqs array in order to
allocate sequence numbers when doing transmit aggregation. This should
also be locked.
* Drivers may have multiple frames queued already - so when one calls
if_transmit(), it may end up dispatching multiple frames for different
VAPs/nodes, each needing a different lock when handling that particular
end destination.
So to be "correct" locking-wise, we'd end up needing to grab a VAP or
node lock inside the driver TX path when setting up crypto / AMPDU sequence
numbers, and we may already _have_ a TX lock held - mostly for the same
destination vap/node, but sometimes it'll be for others. That could lead
to LORs and thus deadlocks.
So for now, I'm sticking with an IC TX lock. It has the advantage of
papering over the above and it also has the added advantage that I can
assert that it's being held when doing a parent device transmit.
I'll look at splitting the locks out a bit more later on.
General outstanding net80211 TX path issues / TODO:
* Look into separating out the VAP serialisation and the IC handoff.
It's going to be tricky as parent->if_transmit() doesn't give me the
opportunity to split queuing from driver dispatch. See above.
* Work with monthadar to fix up the mesh transmit path so it doesn't go via
the parent interface when retransmitting frames.
* Push the encryption handling back into the driver, if it's at all
architectually sane to do so. I know it's possible - it's what mac80211
in Linux does.
* Make ieee80211_raw_xmit() queue a frame into VAP or parent queue rather
than doing a short-cut direct into the driver. There are QoS issues
here - you do want your management frames to be encapsulated and pushed
onto the stack sooner than the (large, bursty) amount of data frames
that are queued. But there has to be a saner way to do this.
* Fragments are still broken - drivers need to be upgraded to an if_transmit()
implementation and then fragmentation handling needs to be properly fixed.
Tested:
* STA - AR5416, AR9280, Intel 5300 abgn wifi
* Hostap - AR5416, AR9160, AR9280
* Mesh - some testing by monthadar@, more to come.
net80211 devices and vaps.
* vnet sets vnet0 during kldload and device probe/attach, but not for
the hotplug event. Thus, plugging in a NIC causes things to panic.
So, add a CURVNET_SET(vnet0) for now during the attach phase, until
the hotplug code is taught to set CURVNET_SET(vnet0).
* there's also no implied detach vnet context - so teach the detach
path about ifp->if_vnet.
* When creating/deleting vaps, also set the vnet context appropriately.
These can be done at any time.
Now, the problems!
* ieee80211.c is supposed to be OS-portable code, with no OS-specific stuff
like vnet. That should be fixed.
* When the device hotplug code gets taught about CURVNET_SET(vnet0), the
device vnet set can go away; but the VAP vnet set still needs to be there.
* .. and there still is the question about potentially adding an implied
CURVNET_SET(ifp->if_vnet) on if_free(), since any/all devices may end up
being detached by a hotplug event in today's world. That's going to be
a topic of a subsequent commit.
Currently, a channel width change updates the 802.11n HT info data in
net80211 but it doesn't trigger any device changes. So the device
driver may decide that HT40 frames can be transmitted but the last
device channel set only had HT20 set.
Now, a task is scheduled so a hardware reset or change isn't done
during any active ongoing RX. It also means that it's serialised
with the other task operations (eg channel change.)
This isn't the final incantation of this work, see below.
For now, any unmodified drivers will simply receive a channel
change log entry. A subsequent patch to ath(4) will introduce
some basic channel change handling (by resetting the NIC.)
Other NICs may need to update their rate control information.
TODO:
* There's still a small window at the present moment where the
channel width has been updated but the task hasn't been fired.
The final version of this should likely pass in a channel width
field to the driver and let the driver atomically do whatever
it needs to before changing the channel.
PR: kern/166286
This makes it much easier to determine whether an event occurs in the
net80211 taskqueue (which was called "ath0 taskq") or the ath driver
taskqueue (which is also called "ath0 taskq".)
Initialize ic_rxstream/ic_txstream with 2, for compatibility reasons.
Introduce 4 new HTC flags, which are used in addition to ic_rxstream
and ic_txstream to compute the hc_mcsset content and also for initializing
ni_htrates. The number of spatial streams is enough to determine support
for MCS0-31 but not for MCS32-76 as well as some TX parameters in the
hc_mcsset field.
adjust the IEEE80211_HTRATE_MAXSIZE constant, only MCS0 - 76 are valid
the other bits in the mcsset IE (77 - 127) are either reserved or used
for TX parameters.
the IEEE80211_C_RATECTL flag set, default to NONE for all drivers. Only if
a driver calls ieee80211_ratectl_init() check if the NONE algo is still
selected and try to use AMRR in that case. Drivers are still free to use
any other algo by calling ieee80211_ratectl_set() prior to the
ieee80211_ratectl_init() call.
After this change it is now safe to assume that a ratectl algo is always
available and selected, which renders the IEEE80211_C_RATECTL flag pretty
much useless. Therefore revert r211314 and 211546.
Reviewed by: rpaulo
MFC after: 2 weeks
queue length. The default value for this parameter is 50, which is
quite low for many of today's uses and the only way to modify this
parameter right now is to edit if_var.h file. Also add read-only
sysctl with the same name, so that it's possible to retrieve the
current value.
MFC after: 1 month
This framework allows drivers to abstract the rate control algorithm and
just feed the framework with the usable parameters. The rate control
framework will now deal with passing the parameters to the selected
algorithm. Right now we have AMRR (the default) and RSSADAPT but there's
no way to select one with ifconfig, yet.
The objective is to have more rate control algorithms in the net80211
stack so all drivers[0] can use it. Ideally, we'll have the well-known
sample rate control algorithm in the net80211 at some point so all
drivers can use it (not just ath).
[0] all drivers that do rate control in software, that is.
Reviewed by: bschmidt, thompsa, weyongo
MFC after: 1 months
the parent wirless interface. If the user passed in a mac address or it was
autogenerated then flag this to avoid trashing it on update.
This will fix wlan+lagg in a post vap world.
band-aid for the general problem that if_link_state_change can be
called between if_detach and if_free leaving a task queued that has
been free'd.
Spotted by: thompsa
Reviewed by: rwatson
Approved by: re (rwatson)
frequency w/o regulatory issues, do this by hooking if_transmit and
if_output with routines that discard all transmits
Reviewed by: thompsa, cbzimmer (intent)
Approved by: re (kensmith)
net80211 wireless stack. This work is based on the March 2009 D3.0 draft
standard. This standard is expected to become final next year.
This includes two main net80211 modules, ieee80211_mesh.c
which deals with peer link management, link metric calculation,
routing table control and mesh configuration and ieee80211_hwmp.c
which deals with the actually routing process on the mesh network.
HWMP is the mandatory routing protocol on by the mesh standard, but
others, such as RA-OLSR, can be implemented.
Authentication and encryption are not implemented.
There are several scripts under tools/tools/net80211/scripts that can be
used to test different mesh network topologies and they also teach you
how to setup a mesh vap (for the impatient: ifconfig wlan0 create
wlandev ... wlanmode mesh).
A new build option is available: IEEE80211_SUPPORT_MESH and it's enabled
by default on GENERIC kernels for i386, amd64, sparc64 and pc98.
Drivers that support mesh networks right now are: ath, ral and mwl.
More information at: http://wiki.freebsd.org/WifiMesh
Please note that this work is experimental. Also, please note that
bridging a mesh vap with another network interface is not yet supported.
Many thanks to the FreeBSD Foundation for sponsoring this project and to
Sam Leffler for his support.
Also, I would like to thank Gateworks Corporation for sending me a
Cambria board which was used during the development of this project.
Reviewed by: sam
Approved by: re (kensmith)
Obtained from: projects/mesh11s
long-term work before they can be serviced. Packets are tagged and
assigned an age (in seconds) at the point they are added to the
queue. If a packet is not retrieved before it's age expires it is
reclaimed. Tagging can take two forms: a reference to an ieee80211_node
(as happens in the tx path) or an opaque token in cases where there
is no reference or the node structure is not stable (i.e. it's going
to be destroyed).
o add ic_stageq to replace the per-node wds staging queue used for
dynamic wds
o add ieee80211_mac_hash for building ageq tokens; this computes a
32-bit hash from an 802.11 mac address (copied from the bridge)
o while here fix a stray ';' noticed in IEEE80211_PSQ_INIT
Reviewed by: rpaulo
Approved by: re (kensmith)
rather than pointers, requiring callers to properly dispose of those
references. The following routines now return references:
ifaddr_byindex
ifa_ifwithaddr
ifa_ifwithbroadaddr
ifa_ifwithdstaddr
ifa_ifwithnet
ifaof_ifpforaddr
ifa_ifwithroute
ifa_ifwithroute_fib
rt_getifa
rt_getifa_fib
IFP_TO_IA
ip_rtaddr
in6_ifawithifp
in6ifa_ifpforlinklocal
in6ifa_ifpwithaddr
in6_ifadd
carp_iamatch6
ip6_getdstifaddr
Remove unused macro which didn't have required referencing:
IFP_TO_IA6
This closes many small races in which changes to interface
or address lists while an ifaddr was in use could lead to use of freed
memory (etc). In a few cases, add missing if_addr_list locking
required to safely acquire references.
Because of a lack of deep copying support, we accept a race in which
an in6_ifaddr pointed to by mbuf tags and extracted with
ip6_getdstifaddr() doesn't hold a reference while in transmit. Once
we have mbuf tag deep copy support, this can be fixed.
Reviewed by: bz
Obtained from: Apple, Inc. (portions)
MFC after: 6 weeks (portions)
bridge do a better job.
o move ether_ifdetach to the top of ieee80211_detach
o do not clear if_softc at the top of ieee80211_detach; we no longer need
this because we are safeguarded against calls coming back through if_ioctl
o simplify the bpf tracker now that we don't null if_softc
This also fixes an issue where having a bpf consumer active when a vap
is destroyed would cause a crash because bpf referenced free'd memory.
Reviewed by: imp
o track # bpf taps on monitor mode vaps instead of # monitor mode vaps
o spam monitor mode taps on tx/rx
o fix ieee80211_radiotap_rx_all to dispatch frames only if the vap is up
o while here print radiotap (and superg) state in show com
o replace DLT_IEEE802_11 support in net80211 with DLT_IEEE802_11_RADIO
and remove explicit bpf support from wireless drivers; drivers now
use ieee80211_radiotap_attach to setup shared data structures that
hold the radiotap header for each packet tx/rx
o remove rx timestamp from the rx path; it was used only by the tdma support
for debugging and was mostly useless due to it being 32-bits and mostly
unavailable
o track DLT_IEEE80211_RADIO bpf attachments and maintain per-vap and
per-com state when there are active taps
o track the number of monitor mode vaps
o use bpf tap and monitor mode vap state to decide when to collect radiotap
state and dispatch frames; drivers no longer explicitly directly check
bpf state or use bpf calls to tap frames
o handle radiotap state updates on channel change in net80211; drivers
should not do this (unless they bypass net80211 which is almost always
a mistake)
o update various drivers to be more consistent/correct in handling radiotap
o update ral to include TSF in radiotap'd frames
o add promisc mode callback to wi
Reviewed by: cbzimmer, rpaulo, thompsa
sleepable context for net80211 driver callbacks. This removes the need for USB
and firmware based drivers to roll their own code to defer the chip programming
for state changes, scan requests, channel changes and mcast/promisc updates.
When a driver callback completes the hardware state is now guaranteed to have
been updated and is in sync with net80211 layer.
This nukes around 1300 lines of code from the wireless device drivers making
them more readable and less race prone.
The net80211 layer has been updated as follows
- all state/channel changes are serialised on the taskqueue.
- ieee80211_new_state() always queues and can now be called from any context
- scanning runs from a single taskq function and executes to completion. driver
callbacks are synchronous so the channel, phy mode and rx filters are
guaranteed to be set in hardware before probe request frames are
transmitted.
Help and contributions from Sam Leffler.
Reviewed by: sam
o remove ic_myaddr from ieee80211com
o change ieee80211_ifattach to take the mac address of the physical device
and use that to setup the lladdr.
o replace all references to ic_myaddr in drivers by IF_LLADDR
o related cleanups (e.g. kill dead code)
PR: kern/133178
Reviewed by: thompsa, rpaulo
and xmit parameters. This makes it possible to use tdma on fractional
channels.
o add IEEE80211_MODE_HALF and IEEE80211_MODE_QUARTER; note these are
band-agnostic (may need revisiting)
o setup all default rates in ic_sup_rates instead of doing it only
for active modes; we need these to calculate the default tx parameters
which are not recalculated after a regulatory update (can't just
recalculate after installing a new channel list because we might
clobber user settings)
o remove special case code in ieee80211_get_suprates; this is now
a candidate for an inline or removal
o add various entries for new modes (roaming+tx params, wme, rate
mapping, scan set setup, country ie construction, tdma, basic rates)
Note these modes are intentionally not visible through if_media.
parent interface tasks to complete. This had been added to the ioctl path but
it is also need elsewhere like detach so its safe to teardown.
Reported by: Hans Petter Selasky
Submitted by: sam
turbo option in addition to the mode bits; otherwise if the current
channel is a turbo mode channel we'll form an invalid media setting
and the ifmedia_set operation in vap_attach will panic.
While here use C99-style initialization for an array indexed by mode;
this makes it consistent w/ other usage and avoids breakage if we
should ever change the set of modes.
o add net80211 support for a tdma vap that is built on top of the
existing adhoc-demo support
o add tdma scheduling of frame transmission to the ath driver; it's
conceivable other devices might be capable of this too in which case
they can make use of the 802.11 protocol additions etc.
o add minor bits to user tools that need to know: ifconfig to setup and
configure, new statistics in athstats, and new debug mask bits
While the architecture can support >2 slots in a TDMA BSS the current
design is intended (and tested) for only 2 slots.
Sponsored by: Intel
really was meant to be 256. Adjust usage accordingly and replace
bogus usage of this value in checking IEEE channel #'s.
NB: this causes an ABI change; ifconfig must be recompiled
o yank useless code for setting fixed rate through media opts: this
mechanism didn't scale to HT rates and couldn't handle multiple bands;
fixed tx rates are set with the IEEE80211_IOC_TXPARAMS ioctl
o construct a name for the com lock as done for other locks
o pass the device name to IEEE80211_LOCK_INIT so the mtx name
is constructed as foo_com_lock
o introduce *_LOCK_OBJ macro's to hide the lock contents and
minimize redundant code
o add IEEE80211_C_STA capability to indicate sta mode is supported
(was previously assumed) and mark drivers as capable
o add ieee80211_opcap array to map an opmode to the equivalent capability bit
o move IEEE80211_C_OPMODE definition to where capabilities are defined so it's
clear it should be kept in sync (on future additions)
o check device capabilities in clone create before trying to create a vap;
this makes driver checks unneeded
o make error codes return on failed clone request unique
o temporarily add console printfs on clone request failures to aid in
debugging; these will move under DIAGNOSTIC or similar before release