Implement key installation and lookup (on TX and RX)
for MLO, so we can use multiple GTKs/IGTKs/BIGTKs.
Co-authored-by: NIlan Peer <ilan.peer@intel.com>
Signed-off-by: NIlan Peer <ilan.peer@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

We'll need the link e.g. for decrypt, and shouldn't be
looking it up all the time later, so track it in the RX
data.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

For drivers using software encryption for multicast TX, such
as mac80211_hwsim, mac80211 needs to duplicate the multicast
frames on each link, if MLO is enabled. Do this, but don't
just make it dependent on the key but provide a separate flag
for drivers to opt out of this.

This is not very efficient, I expect that drivers will do it
in firmware/hardware or at least with DMA engine assistence,
so this is mostly for hwsim.

To make this work, also implement the SNS11 sequence number
space that an AP MLD shall have, and modify the API to the
__ieee80211_subif_start_xmit() function to always require the
link ID bits to be set.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

There are a couple of new things that should be possible
with MLO:
 * selecting the link to transmit to a station by link ID,
   which a previous patch added to the nl80211 API
 * selecting the link by frequency, similarly
 * allowing transmittion to an MLD without specifying any
   channel or link ID, with MLD addresses

Enable these use cases. Also fix the address comparison
in client mode to use the AP (MLD) address.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

It might seem a bit pointless to do a multi-link operation
connection with just a single link, but this is already a
big change, so for now, limit MLO connections to a single
link.

Extending that to multiple links will require
 * work on parsing the multi-link element with STA profile
   properly, including element fragmentation;
 * checking the per-link status in the multi-link element
 * implementing logic to have active/inactive links to let
   drivers decide which links should be active;
 * implementing multicast RX deduplication;
 * and likely more.

For now this is still useful since it lets us do multi-link
connections for the purposes of testing APIs and the higher
layers such as wpa_supplicant.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Add the necessary API to parse the multi-link element in
the future. For now, link only to the element when found
so we can use it in the client-side code later.

Later, we'll need to fill this in to deal with element
fragmentation, parse the STA profile, etc.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

The flag here is currently per interface, but the way we
set and clear it means it should be per link, so change
it.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Recalculate min channel context for the given or all interface
links, depending on the caller. For a station state change, we
need to recalculate all of them since we don't know which link
(or multiple) it might be on.
Signed-off-by: NAndrei Otcheretianski <andrei.otcheretianski@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Allow link source address on TX.
Signed-off-by: NAndrei Otcheretianski <andrei.otcheretianski@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

In case of authentication with a legacy station, link addressed EAPOL
frames should be sent. Support it.
Signed-off-by: NAndrei Otcheretianski <andrei.otcheretianski@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

The way this works is that you add all the element data,
keeping a pointer to the length field of the element.
Then call this helper function, which will fragment the
element if there was more than 255 bytes in the element,
memmove()ing the data back if needed.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

For MLO, we may not have a full chandef here later, so change
the API to pass only the width.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

When parsing a frame containing a multi-BSSID element, we
need to know both the transmitted and non-transmitted BSSID
so we can parse it correctly.

Unfortunately, in quite a number of cases, we got this wrong
and were passing the wrong BSSID or useless information:
 * the mgmt->bssid from a frame is only the transmitted
   BSSID if the frame is a beacon
 * passing just one of the parameters as non-NULL isn't
   useful and ignored

In those case where we need to parse for a specific BSS we
always have a BSS structure pointer, representing the BSS
we need, whether transmitted or not. Thus, pass that pointer
to the parsing function instead of the two BSSIDs.

Also fix two bugs:
 * we need to re-parse all the elements for the other BSS
   when iterating the non-transmitted BSSes in scan
 * we need to parse for the correct BSS when setting up
   the channel data in client code

Fixes: 78ac51f8 ("mac80211: support multi-bssid")
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

This value should be per link, since a TDLS connection is
only established on a given link.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Refactor the element parsing into a version that has
a parameter struct so we can add more parameters more
easily in the future.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

This requires a few more changes.

While at it, also add a warning to ieee80211_get_sband()
to avoid it being used when there are multiple links.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

If we decide to stop tracking QoS/WMM parameters, then
this should be a per-link decision. Move the flag to
the link instead.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Do the first adjustments in the client-side code to pass
the link pointer (instead of sdata) to most places etc.

This is just preparation, so the real MLO patches become
smaller.

Note that this isn't complete, notably there are still
quite a few references to sta->deflink and sta->sta.deflink.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Remove the IEEE80211_STA_RESET_SIGNAL_AVE flag and use
a bool instead, but invert the polarity (now calling it
tracking_signal_avg) so we don't have to initialize it,
and put that into the link instead.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Take the link into account in the QoS settings (EDCA parameters)
APIs.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

In station/client mode, the link data needs a bit more
initialization and destruction than just zero-init and
kfree() respectively, implement that.

This required some shuffling of the link data handling
in general, as we should set it up in setup and do the
teardown in teardown, otherwise we're asymmetric in
case of interface type changes.

Also stop using kfree_rcu(), we cannot guarantee that
nothing is scheduling things that live within the link
(e.g. the u.mgd.request_smps_work) until we're sure it
cannot be referenced anymore, therefore synchronize
instead. This isn't very efficient, but we can always
optimize it later.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

This function can be static.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Separate out the connection downgrade flags from the ifmgd->flags
and put them into the link information instead. While at it, make
them a separate sparse type so we don't get confused about where
they belong and have static checking on correct handling.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Align the mac80211 implementation with P802.11be_D1.5.
Signed-off-by: NIlan Peer <ilan.peer@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Since links can be added and removed dynamically, we need to
somehow protect the sdata->link[] and vif->link_conf[] array
pointers from disappearing when accessing them without locks.
RCU-ify the pointers to achieve this, which requires quite a
bit of rework.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Similar to ieee80211_get_sband but get the sband of the link_conf.
Signed-off-by: NShaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

AP SMPS was removed and not needed anymore. Remove the leftovers.
Signed-off-by: NAndrei Otcheretianski <andrei.otcheretianski@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

When checking whether or not to accept a frame in RX,
take into account the configured link addresses. Also
look up the link station correctly.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

In order to maintain fairness, the amount of queueing needs to be limited
beyond the simple per-station AQL budget, otherwise the driver can simply
repeatedly do scheduling rounds until all queues that have not used their
AQL budget become eligble.

To be conservative, use the high AQL limit for the first txq and add half
of the low AQL for each subsequent queue.
Signed-off-by: NFelix Fietkau <nbd@nbd.name>
Acked-by: NToke Høiland-Jørgensen <toke@toke.dk>
Link: https://lore.kernel.org/r/20220625212411.36675-5-nbd@nbd.nameSigned-off-by: NJohannes Berg <johannes.berg@intel.com>

This allows proper deficit accounting to ensure that they don't carry their
deficit until the next time they become active
Signed-off-by: NFelix Fietkau <nbd@nbd.name>
Acked-by: NToke Høiland-Jørgensen <toke@toke.dk>
Link: https://lore.kernel.org/r/20220625212411.36675-4-nbd@nbd.nameSigned-off-by: NJohannes Berg <johannes.berg@intel.com>

This reverts commits 6a789ba6 and
2433647b.

The virtual time scheduler code has a number of issues:
- queues slowed down by hardware/firmware powersave handling were not properly
  handled.
- on ath10k in push-pull mode, tx queues that the driver tries to pull from
  were starved, causing excessive latency
- delay between tx enqueue and reported airtime use were causing excessively
  bursty tx behavior

The bursty behavior may also be present on the round-robin scheduler, but there
it is much easier to fix without introducing additional regressions
Signed-off-by: NFelix Fietkau <nbd@nbd.name>
Acked-by: NToke Høiland-Jørgensen <toke@toke.dk>
Link: https://lore.kernel.org/r/20220625212411.36675-1-nbd@nbd.nameSigned-off-by: NJohannes Berg <johannes.berg@intel.com>

Maintain a hash table of link-sta addresses so we can find
them for management frames etc. where addresses haven't
been replaced by the drivers to the MLD address yet.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

We need to be able to access these in a race-free way under
traffic while adding/removing them, so RCU-ify the pointers.
This requires passing a link_sta to a lot of functions so
we don't have to do the RCU handling everywhere.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

This is only used in rx.c, so move it into the file.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

If the interface is an MLD, then we don't know which band
the frame will be transmitted on, and we don't know how to
look up the band. Set the band information to zero in that
case, the driver cannot rely on it anyway.

No longer inline ieee80211_tx_skb_tid() since it's even
bigger now.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Add the necessary infrastructure, including a new driver
method, to add/remove links to/from an interface.

Also add the missing link address to bss_conf (which we
use as link_conf too), and fill it, in station mode for
now just randomly, in AP mode we get the address from
cfg80211 since the link must be created with an address
first.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

For MLD, we might eventually not really know the band on status,
but some code assumes it's there. Move the sband lookup deep to
the code that actually needs it, to make it clear where exactly
it's needed and for what purposes.

For rate control, at least initially we won't support it in MLO,
so that won't be an issue.

For TX monitoring, we may have to elide the rate and/or rely on
ieee80211_tx_status_ext() for rate information.

This also simplifies the function prototypes.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Add the link_id parameter and adjust the code accordingly.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Start making some SMPS related code MLD-aware. This isn't
really done yet, but again cuts down our 'deflink' reliance.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Update ieee80211_ht_cap_ie_to_sta_ht_cap() to handle per-link
data.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>