freebsd-dev/sys/net80211/ieee80211_power.c
Adrian Chadd 5cda6006e4 Bring over my initial work from the net80211 TX locking branch.
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.
2013-03-08 20:23:55 +00:00

555 lines
14 KiB
C

/*-
* Copyright (c) 2002-2008 Sam Leffler, Errno Consulting
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
/*
* IEEE 802.11 power save support.
*/
#include "opt_wlan.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/socket.h>
#include <net/if.h>
#include <net/if_media.h>
#include <net/ethernet.h>
#include <net80211/ieee80211_var.h>
#include <net/bpf.h>
static void ieee80211_update_ps(struct ieee80211vap *, int);
static int ieee80211_set_tim(struct ieee80211_node *, int);
static MALLOC_DEFINE(M_80211_POWER, "80211power", "802.11 power save state");
void
ieee80211_power_attach(struct ieee80211com *ic)
{
}
void
ieee80211_power_detach(struct ieee80211com *ic)
{
}
void
ieee80211_power_vattach(struct ieee80211vap *vap)
{
if (vap->iv_opmode == IEEE80211_M_HOSTAP ||
vap->iv_opmode == IEEE80211_M_IBSS) {
/* NB: driver should override */
vap->iv_update_ps = ieee80211_update_ps;
vap->iv_set_tim = ieee80211_set_tim;
}
vap->iv_node_ps = ieee80211_node_pwrsave;
vap->iv_sta_ps = ieee80211_sta_pwrsave;
}
void
ieee80211_power_latevattach(struct ieee80211vap *vap)
{
/*
* Allocate these only if needed. Beware that we
* know adhoc mode doesn't support ATIM yet...
*/
if (vap->iv_opmode == IEEE80211_M_HOSTAP) {
vap->iv_tim_len = howmany(vap->iv_max_aid,8) * sizeof(uint8_t);
vap->iv_tim_bitmap = (uint8_t *) malloc(vap->iv_tim_len,
M_80211_POWER, M_NOWAIT | M_ZERO);
if (vap->iv_tim_bitmap == NULL) {
printf("%s: no memory for TIM bitmap!\n", __func__);
/* XXX good enough to keep from crashing? */
vap->iv_tim_len = 0;
}
}
}
void
ieee80211_power_vdetach(struct ieee80211vap *vap)
{
if (vap->iv_tim_bitmap != NULL) {
free(vap->iv_tim_bitmap, M_80211_POWER);
vap->iv_tim_bitmap = NULL;
}
}
void
ieee80211_psq_init(struct ieee80211_psq *psq, const char *name)
{
memset(psq, 0, sizeof(*psq));
psq->psq_maxlen = IEEE80211_PS_MAX_QUEUE;
IEEE80211_PSQ_INIT(psq, name); /* OS-dependent setup */
}
void
ieee80211_psq_cleanup(struct ieee80211_psq *psq)
{
#if 0
psq_drain(psq); /* XXX should not be needed? */
#else
KASSERT(psq->psq_len == 0, ("%d frames on ps q", psq->psq_len));
#endif
IEEE80211_PSQ_DESTROY(psq); /* OS-dependent cleanup */
}
/*
* Return the highest priority frame in the ps queue.
*/
struct mbuf *
ieee80211_node_psq_dequeue(struct ieee80211_node *ni, int *qlen)
{
struct ieee80211_psq *psq = &ni->ni_psq;
struct ieee80211_psq_head *qhead;
struct mbuf *m;
IEEE80211_PSQ_LOCK(psq);
qhead = &psq->psq_head[0];
again:
if ((m = qhead->head) != NULL) {
if ((qhead->head = m->m_nextpkt) == NULL)
qhead->tail = NULL;
KASSERT(qhead->len > 0, ("qhead len %d", qhead->len));
qhead->len--;
KASSERT(psq->psq_len > 0, ("psq len %d", psq->psq_len));
psq->psq_len--;
m->m_nextpkt = NULL;
}
if (m == NULL && qhead == &psq->psq_head[0]) {
/* Algol-68 style for loop */
qhead = &psq->psq_head[1];
goto again;
}
if (qlen != NULL)
*qlen = psq->psq_len;
IEEE80211_PSQ_UNLOCK(psq);
return m;
}
/*
* Reclaim an mbuf from the ps q. If marked with M_ENCAP
* we assume there is a node reference that must be relcaimed.
*/
static void
psq_mfree(struct mbuf *m)
{
if (m->m_flags & M_ENCAP) {
struct ieee80211_node *ni = (void *) m->m_pkthdr.rcvif;
ieee80211_free_node(ni);
}
m->m_nextpkt = NULL;
m_freem(m);
}
/*
* Clear any frames queued in the power save queue.
* The number of frames that were present is returned.
*/
static int
psq_drain(struct ieee80211_psq *psq)
{
struct ieee80211_psq_head *qhead;
struct mbuf *m;
int qlen;
IEEE80211_PSQ_LOCK(psq);
qlen = psq->psq_len;
qhead = &psq->psq_head[0];
again:
while ((m = qhead->head) != NULL) {
qhead->head = m->m_nextpkt;
psq_mfree(m);
}
qhead->tail = NULL;
qhead->len = 0;
if (qhead == &psq->psq_head[0]) { /* Algol-68 style for loop */
qhead = &psq->psq_head[1];
goto again;
}
psq->psq_len = 0;
IEEE80211_PSQ_UNLOCK(psq);
return qlen;
}
/*
* Clear any frames queued in the power save queue.
* The number of frames that were present is returned.
*/
int
ieee80211_node_psq_drain(struct ieee80211_node *ni)
{
return psq_drain(&ni->ni_psq);
}
/*
* Age frames on the power save queue. The aging interval is
* 4 times the listen interval specified by the station. This
* number is factored into the age calculations when the frame
* is placed on the queue. We store ages as time differences
* so we can check and/or adjust only the head of the list.
* If a frame's age exceeds the threshold then discard it.
* The number of frames discarded is returned so the caller
* can check if it needs to adjust the tim.
*/
int
ieee80211_node_psq_age(struct ieee80211_node *ni)
{
struct ieee80211_psq *psq = &ni->ni_psq;
int discard = 0;
if (psq->psq_len != 0) {
#ifdef IEEE80211_DEBUG
struct ieee80211vap *vap = ni->ni_vap;
#endif
struct ieee80211_psq_head *qhead;
struct mbuf *m;
IEEE80211_PSQ_LOCK(psq);
qhead = &psq->psq_head[0];
again:
while ((m = qhead->head) != NULL &&
M_AGE_GET(m) < IEEE80211_INACT_WAIT) {
IEEE80211_NOTE(vap, IEEE80211_MSG_POWER, ni,
"discard frame, age %u", M_AGE_GET(m));
if ((qhead->head = m->m_nextpkt) == NULL)
qhead->tail = NULL;
KASSERT(qhead->len > 0, ("qhead len %d", qhead->len));
qhead->len--;
KASSERT(psq->psq_len > 0, ("psq len %d", psq->psq_len));
psq->psq_len--;
psq_mfree(m);
discard++;
}
if (qhead == &psq->psq_head[0]) { /* Algol-68 style for loop */
qhead = &psq->psq_head[1];
goto again;
}
if (m != NULL)
M_AGE_SUB(m, IEEE80211_INACT_WAIT);
IEEE80211_PSQ_UNLOCK(psq);
IEEE80211_NOTE(vap, IEEE80211_MSG_POWER, ni,
"discard %u frames for age", discard);
IEEE80211_NODE_STAT_ADD(ni, ps_discard, discard);
}
return discard;
}
/*
* Handle a change in the PS station occupancy.
*/
static void
ieee80211_update_ps(struct ieee80211vap *vap, int nsta)
{
KASSERT(vap->iv_opmode == IEEE80211_M_HOSTAP ||
vap->iv_opmode == IEEE80211_M_IBSS,
("operating mode %u", vap->iv_opmode));
}
/*
* Indicate whether there are frames queued for a station in power-save mode.
*/
static int
ieee80211_set_tim(struct ieee80211_node *ni, int set)
{
struct ieee80211vap *vap = ni->ni_vap;
struct ieee80211com *ic = ni->ni_ic;
uint16_t aid;
int changed;
KASSERT(vap->iv_opmode == IEEE80211_M_HOSTAP ||
vap->iv_opmode == IEEE80211_M_IBSS,
("operating mode %u", vap->iv_opmode));
aid = IEEE80211_AID(ni->ni_associd);
KASSERT(aid < vap->iv_max_aid,
("bogus aid %u, max %u", aid, vap->iv_max_aid));
IEEE80211_LOCK(ic);
changed = (set != (isset(vap->iv_tim_bitmap, aid) != 0));
if (changed) {
if (set) {
setbit(vap->iv_tim_bitmap, aid);
vap->iv_ps_pending++;
} else {
clrbit(vap->iv_tim_bitmap, aid);
vap->iv_ps_pending--;
}
/* NB: we know vap is in RUN state so no need to check */
vap->iv_update_beacon(vap, IEEE80211_BEACON_TIM);
}
IEEE80211_UNLOCK(ic);
return changed;
}
/*
* Save an outbound packet for a node in power-save sleep state.
* The new packet is placed on the node's saved queue, and the TIM
* is changed, if necessary.
*/
int
ieee80211_pwrsave(struct ieee80211_node *ni, struct mbuf *m)
{
struct ieee80211_psq *psq = &ni->ni_psq;
struct ieee80211vap *vap = ni->ni_vap;
struct ieee80211com *ic = ni->ni_ic;
struct ieee80211_psq_head *qhead;
int qlen, age;
IEEE80211_PSQ_LOCK(psq);
if (psq->psq_len >= psq->psq_maxlen) {
psq->psq_drops++;
IEEE80211_PSQ_UNLOCK(psq);
IEEE80211_NOTE(vap, IEEE80211_MSG_ANY, ni,
"pwr save q overflow, drops %d (size %d)",
psq->psq_drops, psq->psq_len);
#ifdef IEEE80211_DEBUG
if (ieee80211_msg_dumppkts(vap))
ieee80211_dump_pkt(ni->ni_ic, mtod(m, caddr_t),
m->m_len, -1, -1);
#endif
psq_mfree(m);
return ENOSPC;
}
/*
* Tag the frame with it's expiry time and insert it in
* the appropriate queue. The aging interval is 4 times
* the listen interval specified by the station. Frames
* that sit around too long are reclaimed using this
* information.
*/
/* TU -> secs. XXX handle overflow? */
age = IEEE80211_TU_TO_MS((ni->ni_intval * ic->ic_bintval) << 2) / 1000;
/*
* Encapsulated frames go on the high priority queue,
* other stuff goes on the low priority queue. We use
* this to order frames returned out of the driver
* ahead of frames we collect in ieee80211_start.
*/
if (m->m_flags & M_ENCAP)
qhead = &psq->psq_head[0];
else
qhead = &psq->psq_head[1];
if (qhead->tail == NULL) {
struct mbuf *mh;
qhead->head = m;
/*
* Take care to adjust age when inserting the first
* frame of a queue and the other queue already has
* frames. We need to preserve the age difference
* relationship so ieee80211_node_psq_age works.
*/
if (qhead == &psq->psq_head[1]) {
mh = psq->psq_head[0].head;
if (mh != NULL)
age-= M_AGE_GET(mh);
} else {
mh = psq->psq_head[1].head;
if (mh != NULL) {
int nage = M_AGE_GET(mh) - age;
/* XXX is clamping to zero good 'nuf? */
M_AGE_SET(mh, nage < 0 ? 0 : nage);
}
}
} else {
qhead->tail->m_nextpkt = m;
age -= M_AGE_GET(qhead->head);
}
KASSERT(age >= 0, ("age %d", age));
M_AGE_SET(m, age);
m->m_nextpkt = NULL;
qhead->tail = m;
qhead->len++;
qlen = ++(psq->psq_len);
IEEE80211_PSQ_UNLOCK(psq);
IEEE80211_NOTE(vap, IEEE80211_MSG_POWER, ni,
"save frame with age %d, %u now queued", age, qlen);
if (qlen == 1 && vap->iv_set_tim != NULL)
vap->iv_set_tim(ni, 1);
return 0;
}
/*
* Move frames from the ps q to the vap's send queue
* and/or the driver's send queue; and kick the start
* method for each, as appropriate. Note we're careful
* to preserve packet ordering here.
*/
static void
pwrsave_flushq(struct ieee80211_node *ni)
{
struct ieee80211_psq *psq = &ni->ni_psq;
struct ieee80211com *ic = ni->ni_ic;
struct ieee80211vap *vap = ni->ni_vap;
struct ieee80211_psq_head *qhead;
struct ifnet *parent, *ifp;
struct mbuf *parent_q = NULL, *ifp_q = NULL;
struct mbuf *m;
IEEE80211_NOTE(vap, IEEE80211_MSG_POWER, ni,
"flush ps queue, %u packets queued", psq->psq_len);
IEEE80211_PSQ_LOCK(psq);
qhead = &psq->psq_head[0]; /* 802.11 frames */
if (qhead->head != NULL) {
/* XXX could dispatch through vap and check M_ENCAP */
parent = vap->iv_ic->ic_ifp;
/* XXX need different driver interface */
/* XXX bypasses q max and OACTIVE */
parent_q = qhead->head;
qhead->head = qhead->tail = NULL;
qhead->len = 0;
} else
parent = NULL;
qhead = &psq->psq_head[1]; /* 802.3 frames */
if (qhead->head != NULL) {
ifp = vap->iv_ifp;
/* XXX need different driver interface */
/* XXX bypasses q max and OACTIVE */
ifp_q = qhead->head;
qhead->head = qhead->tail = NULL;
qhead->len = 0;
} else
ifp = NULL;
psq->psq_len = 0;
IEEE80211_PSQ_UNLOCK(psq);
/* NB: do this outside the psq lock */
/* XXX packets might get reordered if parent is OACTIVE */
/* parent frames, should be encapsulated */
if (parent != NULL) {
while (parent_q != NULL) {
m = parent_q;
parent_q = m->m_nextpkt;
/* must be encapsulated */
KASSERT((m->m_flags & M_ENCAP),
("%s: parentq with non-M_ENCAP frame!\n",
__func__));
/*
* For encaped frames, we need to free the node
* reference upon failure.
*/
if (ieee80211_parent_transmit(ic, m) != 0)
ieee80211_free_node(ni);
}
}
/* VAP frames, aren't encapsulated */
if (ifp != NULL) {
while (ifp_q != NULL) {
m = ifp_q;
ifp_q = m->m_nextpkt;
KASSERT((!(m->m_flags & M_ENCAP)),
("%s: vapq with M_ENCAP frame!\n", __func__));
(void) ieee80211_vap_transmit(vap, m);
}
}
}
/*
* Handle station power-save state change.
*/
void
ieee80211_node_pwrsave(struct ieee80211_node *ni, int enable)
{
struct ieee80211vap *vap = ni->ni_vap;
int update;
update = 0;
if (enable) {
if ((ni->ni_flags & IEEE80211_NODE_PWR_MGT) == 0) {
vap->iv_ps_sta++;
update = 1;
}
ni->ni_flags |= IEEE80211_NODE_PWR_MGT;
IEEE80211_NOTE(vap, IEEE80211_MSG_POWER, ni,
"power save mode on, %u sta's in ps mode", vap->iv_ps_sta);
if (update)
vap->iv_update_ps(vap, vap->iv_ps_sta);
} else {
if (ni->ni_flags & IEEE80211_NODE_PWR_MGT) {
vap->iv_ps_sta--;
update = 1;
}
ni->ni_flags &= ~IEEE80211_NODE_PWR_MGT;
IEEE80211_NOTE(vap, IEEE80211_MSG_POWER, ni,
"power save mode off, %u sta's in ps mode", vap->iv_ps_sta);
/* NB: order here is intentional so TIM is clear before flush */
if (vap->iv_set_tim != NULL)
vap->iv_set_tim(ni, 0);
if (update) {
/* NB if no sta's in ps, driver should flush mc q */
vap->iv_update_ps(vap, vap->iv_ps_sta);
}
if (ni->ni_psq.psq_len != 0)
pwrsave_flushq(ni);
}
}
/*
* Handle power-save state change in station mode.
*/
void
ieee80211_sta_pwrsave(struct ieee80211vap *vap, int enable)
{
struct ieee80211_node *ni = vap->iv_bss;
if (!((enable != 0) ^ ((ni->ni_flags & IEEE80211_NODE_PWR_MGT) != 0)))
return;
IEEE80211_NOTE(vap, IEEE80211_MSG_POWER, ni,
"sta power save mode %s", enable ? "on" : "off");
if (!enable) {
ni->ni_flags &= ~IEEE80211_NODE_PWR_MGT;
ieee80211_send_nulldata(ieee80211_ref_node(ni));
/*
* Flush any queued frames; we can do this immediately
* because we know they'll be queued behind the null
* data frame we send the ap.
* XXX can we use a data frame to take us out of ps?
*/
if (ni->ni_psq.psq_len != 0)
pwrsave_flushq(ni);
} else {
ni->ni_flags |= IEEE80211_NODE_PWR_MGT;
ieee80211_send_nulldata(ieee80211_ref_node(ni));
}
}