| Age | Commit message (Collapse) | Author |
|
The unlocked station statistics path returns -EAGAIN whenever the driver
has no atomic statistics op, even if it has no statistics op at all.
For drivers without any statistics op, which is most of them, every
get/dump therefore falls back to the wiphy mutex and the unlocked path
never helps.
Only return -EAGAIN when the driver has a locked op but no atomic one.
The link variant also checked the wrong op before calling the atomic
link callback, ignored its return value, and lost its entry tracepoint.
Fix those too.
Fixes: ab5d6dd9c9b0 ("wifi: mac80211: implement unlocked STA statistics")
Signed-off-by: David Carlier <devnexen@gmail.com>
Link: https://patch.msgid.link/20261031200133.473693-1-devnexen@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
If ieee80211_process_adv_ttlm() disconnects, and some other
changes were also detected in the beacon, the link update
to the driver is still done, leading to use-after-free. Fix
this by returning, like other functions, whether or not the
update was accepted by the function.
Cc: stable@vger.kernel.org
Fixes: 702e80470a33 ("wifi: mac80211: support handling of advertised TID-to-link mapping")
Reviewed-by: Benjamin Berg <benjamin.berg@intel.com>
Link: https://patch.msgid.link/20261102220313.3cf5e709417b.I339967c7e614057253aa68c005d9ae220577f99a@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Delaying the OCB station worker while frames arrive from new peers can
retain an unbounded number of station allocations.
The receive-side limit counts installed stations, while new allocations
wait on incomplete_stations until process context inserts them. The list
lock does not cover the interval between allocation and list insertion.
Reserve one of 128 pending slots per interface before allocation. Release
the reservation on allocation failure, after the worker inserts or frees
the station, and when leave frees the pending list.
LLM assistance was used to find the bug and review.
Fixes: 239281f803e2 ("mac80211: 802.11p OCB mode support")
Cc: stable@vger.kernel.org
Signed-off-by: Zhao Li <enderaoelyther@gmail.com>
Link: https://patch.msgid.link/20261028142302.16634-3-enderaoelyther@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Frames from successive unknown OCB peers can make the deferred worker
install more than 128 stations.
The receive path checks local->num_sta before allocating each station,
but pending stations are not included in that count. Several allocations
can therefore pass before the worker inserts any of them, and the worker
then drains the list without rechecking the limit.
Recheck the installed-station limit in the worker under the wiphy mutex
and free entries that no longer fit.
LLM assistance was used to find the bug and review.
Fixes: 239281f803e2 ("mac80211: 802.11p OCB mode support")
Cc: stable@vger.kernel.org
Signed-off-by: Zhao Li <enderaoelyther@gmail.com>
Link: https://patch.msgid.link/20261028142302.16634-2-enderaoelyther@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Delaying the IBSS station worker while frames arrive from new peers can
retain an unbounded number of station allocations.
The receive-side limit counts installed stations, while new allocations
wait on incomplete_stations until process context inserts them. The list
lock does not cover the interval between allocation and list insertion.
Reserve one of 128 pending slots per interface before allocation. Release
the reservation on allocation failure, after the worker inserts or frees
the station, and when disconnect frees the pending list.
LLM assistance was used to find the bug and review.
Fixes: 8bf11d8d0811 ("mac80211: delay IBSS station insertion")
Cc: stable@vger.kernel.org
Signed-off-by: Zhao Li <enderaoelyther@gmail.com>
Link: https://patch.msgid.link/20261028142235.15428-3-enderaoelyther@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Frames from successive unknown IBSS peers can make the deferred worker
install more than 128 stations.
The receive path checks local->num_sta before allocating each station,
but pending stations are not included in that count. Several allocations
can therefore pass before the worker inserts any of them, and the worker
then drains the list without rechecking the limit.
Recheck the installed-station limit in the worker under the wiphy mutex
and free entries that no longer fit.
LLM assistance was used to find the bug and review.
Fixes: 8bf11d8d0811 ("mac80211: delay IBSS station insertion")
Cc: stable@vger.kernel.org
Signed-off-by: Zhao Li <enderaoelyther@gmail.com>
Link: https://patch.msgid.link/20261028142235.15428-2-enderaoelyther@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
An invalid EHT guard interval can select the wrong airtime group and, at
320 MHz, read beyond airtime_mcs_groups.
Guard interval 3 advances EHT_GROUP_IDX() into the next bandwidth group
at narrower widths. With one spatial stream and 320 MHz, it evaluates to
264, one past the 264-entry table. RX rate parsing rejects the value, but
the common TX-status validator has no EHT check and last_rate_info can
carry it to the airtime helper.
Reject values above NL80211_RATE_INFO_EHT_GI_3_2 in the helper. This also
protects exported ieee80211_calc_rx_airtime() callers that do not pass
through the TX-status validator.
LLM assistance was used to find the bug and review.
Fixes: 3702a33216e6 ("wifi: mac80211: introduce EHT rate support in AQL airtime")
Cc: stable@vger.kernel.org
Signed-off-by: Zhao Li <enderaoelyther@gmail.com>
Link: https://patch.msgid.link/20261028142212.14523-1-enderaoelyther@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
A failed MLO link update can leave the interface with the requested
dormant-links bitmap instead of its previous one. This can disable active
links or trigger a bitmap warning during rollback.
The error path restores the old valid-links bitmap but passes the new
dormant_links argument to ieee80211_set_vif_links_bitmaps().
Save the old dormant bitmap before changing link state and use it when
restoring the old configuration.
LLM assistance was used to find the bug and review.
Fixes: 6d543b34dbcf ("wifi: mac80211: Support disabled links during association")
Cc: stable@vger.kernel.org
Signed-off-by: Zhao Li <enderaoelyther@gmail.com>
Link: https://patch.msgid.link/20261028142123.13072-1-enderaoelyther@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Injecting a short frame with IEEE80211_RADIOTAP_F_FCS on a monitor
interface can pass a truncated 802.11 header to transmit handlers.
ieee80211_monitor_start_xmit() validates the linear 802.11 header before
parsing radiotap. Radiotap parsing then removes the four-byte FCS without
rechecking the remaining header length. Software encryption can subtract
the larger header length and pass the resulting negative length as an
unsigned size.
Recheck the required linear header length after pulling the radiotap
header and before handing the frame to the transmit path. Require at
least addr1, which the transmit path reads even for extension frames
whose header is shorter.
LLM assistance was used to find the bug and review.
Fixes: bddc0c411a45 ("mac80211: Fix NULL ptr deref for injected rate info")
Cc: stable@vger.kernel.org
Signed-off-by: Zhao Li <enderaoelyther@gmail.com>
Link: https://patch.msgid.link/20261028142101.11994-1-enderaoelyther@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Removing an MLO link can leave work queued on its link container after
the container is freed.
ieee80211_tear_down_links() stops each link before the RCU grace period.
A reader that fetched the link earlier can therefore queue work after
ieee80211_link_stop() has canceled it. ieee80211_request_smps() and
ieee80211_csa_finish() can both reach this window.
Record each removed link while removing its keys and debugfs state. Wait
for RCU readers, then stop the recorded links so cancellation includes
work queued by those readers. Free the detached keys after the grace
period as before.
LLM assistance was used to find the bug and review.
Fixes: efe9c2bfd1a8 ("wifi: mac80211: isolate driver from inactive links")
Cc: stable@vger.kernel.org
Signed-off-by: Zhao Li <enderaoelyther@gmail.com>
Link: https://patch.msgid.link/20261028141708.5532-1-enderaoelyther@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Resolve the APs station in case a beacon is received for a
non-transmitted BSS. This is important because the station has the key
information for beacon protection. Without this change, mac80211 will
not verify the beacon when connected to a non-transmitted BSSID.
Most hardware is not affected as the frames are already verified. But
for example the virtual hwsim driver is.
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Link: https://patch.msgid.link/20261027144558.84c3f2f79b01.I055996ff039ab5c2f441c93190809e598dbea333@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Set the link_id in the SKB status to the invalid value of 0xf when the
link ID is not valid. Also update the test in ieee80211_rx_handlers to
correctly check this case. This avoids switching to an invalid link
during RX of a single frame (when updating the link is not needed).
It also means that the SKB link_id now has a more reasonable value as it
is possible to differentiate the non-valid case from the case where the
value is set to 0.
Add back the appropriate checks before trying to use the link_id for
access to arrays. These were removed in commit 4db13371e217 ("wifi:
mac80211: refactor RX link_id and station handling").
Finally, this also fixes the link_sta lookup to use the correct link ID,
rather than an old value from rx->link_id.
Fixes: 2d03b83ef762 ("wifi: mac80211: fix non-MLD link ID handling")
Fixes: 4db13371e217 ("wifi: mac80211: refactor RX link_id and station handling")
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Link: https://patch.msgid.link/20261027144558.7d7ed492e70a.Iaa63d80b6f13f871d1ee3cf1ce27118f72850ac0@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Rolling back an MLO link addition can free link containers while RCU
readers still hold pointers to them. ieee80211_vif_update_links()
publishes new links before checking for duplicate addresses and calling
drv_change_vif_links(). If either fails, it restores the old pointers
and frees the new containers without waiting for readers.
During a non-MLO to MLO transition, the published link 0 replaces the
deflink, so existing link 0 users can acquire the new container. Readers
such as ieee80211_csa_finish() and
ieee80211_obss_color_collision_notify() can also queue work embedded in
the link. If they queue after ieee80211_link_stop() cancels that work,
it remains queued when the container is freed.
Wait for RCU readers after restoring the old pointers and before
stopping the new links, so their work is queued before it is cancelled.
Fixes: d3e2439b0f33 ("wifi: mac80211: fix link manipulation")
Cc: stable@vger.kernel.org # 7.2.x
Signed-off-by: Zhao Li <enderaoelyther@gmail.com>
Link: https://patch.msgid.link/20261026152654.80697-1-enderaoelyther@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
ieee80211_ibss_disconnect() leaves CSA active, as ieee80211_stop_mesh()
did until commit ac7472a24bd4 ("wifi: mac80211: abort chanswitch when
leaving a mesh"). An IBSS left during its countdown and joined again
refuses every new channel switch with -EBUSY.
Abort the channel switch the same way.
Fixes: cd7760e62c2a ("mac80211: add support for CSA in IBSS mode")
Reported-by: Luka Gejak <luka.gejak@linux.dev>
Closes: https://lore.kernel.org/linux-wireless/796598480d47c5bf474d11828769e864b8209558@linux.dev/
Signed-off-by: Mehmet Fide <mehmet.fide@screeningeagle.com>
Reviewed-by: Luka Gejak <luka.gejak@linux.dev>
Link: https://patch.msgid.link/20261025193615.1123718-1-mehmet.fide@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
If mesh fast-xmit cannot be used due to aggregation not
being fully set up yet, there's no need to linearize the
SKB. Avoid that, which addresses a stale pointer issue
in the caller.
Change-Id: I6b93e6c7587faf8749f8366ed574ee2b79fd47a5
Fixes: 8b0f5cb6bc7c ("wifi: mac80211: use mesh header cache to speed up mesh forwarding")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless
Johannes Berg says:
====================
A few fixes to start with:
- wext: fix notifier netns interaction
- rsi: avoid TID OOB access
- mac80211: reload pointer, fix A-MPDU flags
- brcmfmac: fix DMA regression
- cfg80211: fix regdb validation
* tag 'wireless-2026-11-02' of https://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless:
wifi: mac80211: reinstate IEEE80211_TX_CTL_AMPDU flag in ieee80211_tx_prep_agg()
wifi: cfg80211: fix out-of-bounds read in valid_rule()
wifi: rsi: Skip TX aggregation setup for non-QoS frames
wifi: wext: Limit notifier event flushes to the device's netns
wifi: brcmfmac: map msgbuf RX packet IDs bidirectional
wifi: mac80211: reload TDLS header pointer after skb_linearize()
====================
Link: https://patch.msgid.link/20261102104649.727621-3-johannes@sipsolutions.net
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
The commit 592234e941f1 ("wifi: mac80211: Fix iTXQ AMPDU fragmentation
handling") changed the way handling of the IEEE80211_TX_CTL_AMPDU flag
is done in mac80211, so the only place it is set is in
ieee80211_tx_dequeue().
This means that any frames that are sent to the driver via drv_tx() will
no longer have the flag set. In particular, this affects packets going
through ieee80211_tx_pending(). Evan reported that this can result in
the ath9k driver failing to account such packets in its block ack
window, eventually overrunning the window and stalling TX.
Fix this by reinstating the flag in ieee80211_tx_prep_agg() when inside
a valid block ack window. This should be harmless for the TXQ path, and
will reinstate the previous behaviour for drivers that use the flag as a
signal to track the packet in the window.
Fixes: 592234e941f1 ("wifi: mac80211: Fix iTXQ AMPDU fragmentation handling")
Cc: stable@vger.kernel.org
Reported-by: Evan Severson <eseverson@gmail.com>
Link: https://lore.kernel.org/all/20261014025734.2288693-1-eseverson@gmail.com/
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
Tested-by: Evan Severson <eseverson@gmail.com>
Link: https://patch.msgid.link/20261023110052.650519-1-toke@toke.dk
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
ieee80211_process_tdls_channel_switch() sets tf = skb->data before it
calls skb_linearize(), and reads tf->action_code afterwards.
ieee80211_rx_h_data() queues TDLS channel switch requests and
responses to the interface after pulling only
offsetof(struct ieee80211_tdls_data, u) bytes, so the skb can still be
non-linear when the interface work handles it. Drivers that build
paged RX skbs do this for every frame that does not fit into the skb
head, e.g. iwlwifi (mvm), which is also the only mac80211 driver
that sets NL80211_FEATURE_TDLS_CHANNEL_SWITCH. If the remaining data
does not fit into the tailroom, skb_linearize() moves the data to a
new head and frees the old one, and the action code is then read from
freed memory. Depending on what that memory holds by then, the frame
is handled as before or ends in the WARN_ON_ONCE() of the default
case.
The frame only has to be a data frame with the TDLS ethertype from an
authorized peer, e.g. relayed by the AP from another station in the
BSS; the read happens before the sender is checked to be a TDLS peer.
Load tf after skb_linearize(), as the request and response handlers
already do.
Found by static analysis tool CodeQL.
Fixes: 8a4d32f30d11 ("mac80211: add TDLS channel-switch Rx flow")
Cc: stable@vger.kernel.org
Assisted-by: LLM codeql
Signed-off-by: Weigang He <geoffreyhe2@gmail.com>
Link: https://patch.msgid.link/20261021145807.3018338-1-geoffreyhe2@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
The count of a netdev_hw_addr_list is changed in several places of
dev_addr_lists.c and a few drivers read it directly. The next patch
needs to account every change of the count of dev->mc.
Add __hw_addr_count_add(), __hw_addr_count_inc(), __hw_addr_count_dec()
and __hw_addr_count_reset(), use them for every change of the count
and rename the field to _count so that a direct write stands out.
Readers keep using netdev_hw_addr_list_count() and the netdev_uc_count()
and netdev_mc_count() helpers, the few that read the field directly
are converted. No functional change.
Signed-off-by: Yuyang Huang <sigefriedhyy@gmail.com>
Reviewed-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Link: https://patch.msgid.link/20261012011707.3378-3-sigefriedhyy@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
The code in mac80211 assumes that BSS entries are
allocated together with the struct ieee80211_bss
internal data, but the test didn't do that, which
with clang 23 resulted in a KASAN splat here. Fix
the code to properly have both structs.
Fix the allocation while at it, should be the ies
struct size, not the cbss.
Link: https://patch.msgid.link/20261013151329.b8135eb40430.I97a6cb689fbc19903c78e2f6dfe37bdc2117b605@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Beacon protection appends an MME to beacons, but the skb
allocation does not reserve tailroom for it. If the available
tailroom is too small, the crypto path drops the beacon.
Two factors often mask this: the 256-byte TIM reservation is
rarely fully used, and dev_alloc_skb() may round up the
allocation, leaving extra tailroom. Neither guarantees enough
room for the MME.
Reserve enough room for the largest MME.
Fixes: 0a3a84360b37 ("mac80211: Beacon protection using the new BIGTK (AP)")
Signed-off-by: Chien Wong <m@xv97.com>
Link: https://patch.msgid.link/20261012140738.70178-3-m@xv97.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
BIP-CMAC-256 and BIP-GMAC append a 26-byte MME but mac80211
currently reserves only 18 bytes of encryption tailroom.
Generating 26-byte MME may happen to work because the actual skb
tailroom can exceed the requested amount. For example, dev_alloc_skb()
rounds the allocation to kmalloc size buckets or alignment boundaries,
and the extra space may cover the 8-byte shortfall.
Reserve enough tailroom for the 26-byte MME.
Fixes: 56c52da2d554 ("mac80111: Add BIP-CMAC-256 cipher")
Signed-off-by: Chien Wong <m@xv97.com>
Link: https://patch.msgid.link/20261012140738.70178-2-m@xv97.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
If we fail to register the inet/inet6 address notifiers,
wiphy_unregister() is called before removing interface,
but the default interface already exists at this point.
This could also lead to a deadlock as wiphy_unregister()
then waits for the interface to go away.
Change the order to remove the interfaces first.
Assisted-by: LLM
Link: https://patch.msgid.link/20261011113308.c2cb32bf3160.Ib46993671ae3030950360ea378a507cef2bd4117@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Back when we supported algorithm switching at runtime, this
was needed for the open_count check. Now we don't have that
any more, so the check and RTNL section can go away.
Link: https://patch.msgid.link/20261011113308.0c5d35e2afea.I81141b28351e7c1ca23770616364346c33d7e08e@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Implement the new cfg80211 unlocked STA statistics. If
the driver has an operation, then we need to always do
the locked again, so return -EAGAIN unconditionally if
it has a locked but not atomic operation, but drivers
can implement the new atomic methods to provide needed
data atomically (and can also return -EAGAIN).
This should, given a driver without any ops or adding
the atomic ops, reduce wiphy mutex contention.
Link: https://patch.msgid.link/20261009201417.706683ff9d81.Ieb92e4c6599a198a116fa6a4f7944219c6d32932@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Refactor the code a bit to pass around link-STA pointers
instead of looking them up again and again. This will
simplify the next changes, since otherwise all of these
lookups would have to be converted to RCU-safe ones. It
also simplifies the code generally.
Link: https://patch.msgid.link/20261009201417.32dfeaa3c44d.I7ea45bf80dd493311670f8fdd018b5c4b770a4be@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
The wiphy mutex was only required for the cookie counter,
but that's been in cfg80211 for a while. The rest is just
datapath, so can work under RCU, just need to be careful
to handle interface type changes.
Link: https://patch.msgid.link/20261009201417.dffb8da303aa.Ica8bba30018704a7c776f2ef08e6f2571cd8f8c4@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
The comment about wiphy mutex is stale since the cookie
cleanup, and everything else can be under RCU, we just
have to drop the warning in case it races AP stop.
Link: https://patch.msgid.link/20261009201417.4501c07ff31e.Ic87a72afcaea6b9a8c92f7f063bf33a2744a7992@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Implement the mgmt_tx_unlocked() method and accept frames
that are transmitted on the operating channel with it, as
they don't need further handling and can go out directly.
For off-channel etc. just return -EAGAIN to fall back to
the locked version.
Link: https://patch.msgid.link/20261009201417.6491f3bdeb4e.Ib10158e2729f89719b551dadb8384e89b9dce688@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Refactor ieee80211_mgmt_tx() into on-channel and frame
building helper functions so we can later implement the
new mgmt_tx_unlocked() method.
Link: https://patch.msgid.link/20261009201417.a84e223f4998.I4848c3eeb2aaf327712233f0bdb38bb48bdbbba6@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Implement the new update_beacon() method to update beacon
and related templates without the wiphy mutex. If anything
else were to change return -EAGAIN, same if there are any
active countdowns, and during HW restart. Eventually, the
idea is to replace countdowns by beacon update requests so
hostapd can handle everything instead.
This doesn't enable it for any drivers since it depends on
the driver behaviour, if it fetches each beacon and doesn't
use the other templates it can just set the feature flag
(NL80211_EXT_FEATURE_UPDATE_BEACON), otherwise it must
implement the new update_beacon() mac80211 method.
The regular beacon update may need to take head, tail and
MBSSID/RNR elements from the old beacon, so while that's
going on mark "bss_update_in_progress" and make lockless
updates return -EAGAIN during that time, so that the old
data remains valid/identical.
For MBSSID, the lockless path also can't copy old data,
so that also has to return -EAGAIN.
All those -EAGAIN cases fall back to cfg80211 and do the
update with the lock held, where it can then be handled.
Userspace should ideally ensure it doesn't hit that path.
Assisted-by: LLM
Link: https://patch.msgid.link/20261009201417.a9433ea65d8e.Icfbcd157ac2bdeccfcc74a9f4c3997d3010ec0a9@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Refactor allocating/filling beacon data to a new function,
so we can reuse it for lockless beacon updates. Also split
out the bssid_indicator calculation, lockless updates will
need to ensure it didn't change.
Assisted-by: LLM
Link: https://patch.msgid.link/20261009201417.ba48af051dfa.If468a87e4c71f8fe8d0f467525a86e7921496e9a@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
The AP's beacon, presp etc. templates are only updated under
wiphy mutex right now. Prepare for being able to update them
without the wiphy mutex by using a (per-link) spinlock for
them.
Link: https://patch.msgid.link/20261009201417.f9414c150e5d.Iba8a15c4d37e441642b6df68c33ef8c97340e9a9@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
A 4-address AP_VLAN always sets A2 to sdata->vif.addr, which on an MLD
AP is the AP MLD address. For a non-MLO station, A2 must be the link
address it associated to.
Use the link address for a non-MLO station in the 4-address case as
well, like the AP case already does in ieee80211_build_hdr() and
ieee80211_check_fast_xmit(). Share the address selection between both
cases and both TX paths.
Fixes: 3e0278b717b0 ("wifi: mac80211: select link when transmitting to non-MLO stations")
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Link: https://patch.msgid.link/20261007105249.1234601-1-nbd@nbd.name
[use guard]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
An open debugfs file retains its file_operations pointer in the
debugfs proxy state. However, hwflags_ops in net/mac80211/debugfs.c
has no .owner set to THIS_MODULE.
As a result, full_proxy_open_regular() calls fops_get() which succeeds
without taking a reference to the mac80211 module. When CONFIG_MAC80211=m
and wireless debugfs is enabled, a userspace process can open the
hwflags debugfs file and keep the file descriptor open while the module
is unloaded (rmmod mac80211).
When the userspace process subsequently closes the descriptor,
full_proxy_release() accesses the already unmapped file_operations
structure, triggering a kernel oops in full_proxy_release():
BUG: unable to handle page fault for address: fffffbfff8056f03
#PF: supervisor read access in kernel mode
Oops: 0000 [#1] SMP KASAN NOPTI
RIP: 0010:full_proxy_release+0x8e/0x140
Call Trace:
<TASK>
__fput+0x39f/0xa60
fput_close_sync+0xff/0x200
__x64_sys_close+0x8c/0xf0
do_syscall_64+0x115/0x6a0
entry_SYSCALL_64_after_hwframe+0x77/0x7f
</TASK>
The issue was discovered via manual code audit of
net/mac80211/debugfs.c.
Set .owner = THIS_MODULE on hwflags_ops so that debugfs pins the
module until all open file descriptors have been closed.
Signed-off-by: Bui Viet Dung <dungvn2345@gmail.com>
Link: https://patch.msgid.link/20261008052412.541780-1-dungvn2345@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
ieee80211_inform_bss() currently calls ieee802_11_parse_elems to
parse the beacon elements, which internally sets the params->mode
to IEEE80211_CONN_MODE_HIGHEST. This means any S1G specific parsing
is bypassed entirely (in the beacon case it is the shorter TIM)
and the beacon is marked as corrupt:
wlan0: operating as S1G STA
wlan0: associating to AP 0c:bf:74:00:30:fd with corrupt beacon
wlan0: associate with 0c:bf:74:00:30:fd (try 1/3)
In this case, we can simply call ieee802_11_parse_elems_full and
build the params locally (similar to most mlme paths) such that we
can set the conn mode to ensure the S1G specific parsing occurs.
Once this is done, the beacon is no longer considered corrupt:
wlan0: operating as S1G STA
wlan0: associate with 0c:bf:74:00:30:fd (try 1/3)
Signed-off-by: Lachlan Hodges <lachlan.hodges@morsemicro.com>
Link: https://patch.msgid.link/20261008030939.474493-3-lachlan.hodges@morsemicro.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
An S1G TIM, unlike the regular TIM, can exclude both the bitmap
control and the PVB. As a result it's minimum length is actually 2.
Fix this by allowing a shorter TIM for S1G connections.
Suggested-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Lachlan Hodges <lachlan.hodges@morsemicro.com>
Link: https://patch.msgid.link/20261008030939.474493-2-lachlan.hodges@morsemicro.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
If TDLS setup packets are created by AF_PACKET or so, then
the payload may not be present. Since we only care about
those mac80211 generated (which are linear), just replace
the skb->len check with a skb_headlen() check to avoid any
out-of-bounds accesses.
Assisted-by: LLM
Link: https://patch.msgid.link/20261005224950.c3cf023d5191.Id537db8aca13e6df999938a3bc4cc6ce2a28f733@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
For injection we only check the header is in the skb
head, but e.g. ACK/CTS are really short, and we then
try to use A2 anyway. Require A1 (no S1G extension
frames) and use A2 only when present.
Assisted-by: LLM
Link: https://patch.msgid.link/20261005224950.0f3c3e88abde.I03477f2444a9f2cb2888d00320ac7c4504f150a3@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
These structs are placed after the tail, but that can make
them unaligned. Change the order so the structs are after
one another, and then followed by the head/tail/elements.
Assisted-by: LLM
Fixes: 2b3171c6fe0a ("mac80211: MBSSID beacon handling in AP mode")
Link: https://patch.msgid.link/20261005224950.2f554b2c7c10.Ie961eb6b1435f956446e052936258758b966c67f@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
The mesh and MPP path generation counters (for netlink)
are updated wrongly and sometimes unlocked. Fix those
issues.
Assisted-by: LLM
Link: https://patch.msgid.link/20261005224950.373dd5c1e46a.I29b95bc3f62b846084700599a261a203a6e459b4@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
With radiotap injection the HT/VHT MCSes can be just
about anything, but that can later get problematic.
Limit them to the rates the device reports, which is
presumably going to be a subset of what mac80211 can
do, except for HT MCS 32 which some devices report as
supported, but then TX status can't deal with it.
If it's not supported just ignore the config.
Assisted-by: LLM
Fixes: dfdfc2beb0dd ("mac80211: Parse legacy and HT rate in injected frames")
Fixes: 646e76bb5daf ("mac80211: parse VHT info in injected frames")
Link: https://patch.msgid.link/20261005224950.68e210b8487a.I2ee0e4664c54cf31002eb9e4f5190fd05ff4130f@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
The radiotap iterator initialization will already parse
the presence bitmaps, so it accesses a good part of the
header. This might not be present in the linear part of
the SKB, we haven't checked it at this point. Check the
lengths first.
Assisted-by: LLM
Fixes: cb17ed29a7a5 ("mac80211: parse radiotap header when selecting Tx queue")
Link: https://patch.msgid.link/20261005224950.b5c2c31790d2.I091422db411b595bde4dba2de69334cd2aeb33b0@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
We never kill the wake_txqs_tasklet so it could be running
while the hardware is being freed. Kill it, late enough to
never be able to restart afterwards.
Assisted-by: LLM
Fixes: 21a5d4c3a45c ("mac80211: add stop/start logic for software TXQs")
Link: https://patch.msgid.link/20261005224950.3fdb8d0ff377.I851e79ac625c4926e211642ee7a7e458728b0ad3@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
The mesh paths walk list used under RCU, so we must use
hlist_add_head_rcu() to publish an entry. Fix one place
that didn't do that.
Assisted-by: LLM
Fixes: b4c3fbe63601 ("mac80211: Use linked list instead of rhashtable walk for mesh tables")
Link: https://patch.msgid.link/20261005224950.e0db543145e1.Ieaf30d3546e8ff63ccf400f6fb26e027c218721f@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
In practice, we probably shouldn't end up with a fragmented
frame that's protected, filled into the cache, finished or
dropped and then the next frame is unprotected, since only
unicast data frames can be fragmented.
However, tools complain and it's more correct to be able to
do this (e.g. an unencrypted fragmented EAPOL _could_, at
least in theory, happen), so clear the flag when the entry
is reused.
Assisted-by: LLM
Fixes: 7e44a0b597f0 ("mac80211: prevent attacks on TKIP/WEP as well")
Link: https://patch.msgid.link/20261005224950.c3e4106f34d5.I866c0862db031c3a810d2333acdd22e2733d2b8c@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Cross-merge networking fixes after downstream PR (net-7.3-rc7).
Conflicts:
include/net/strparser.h
0984ebc631792 ("strparser: make sure __strp_recv isn't running before tearing down the parser")
02fd0a2111374 ("net: strparser: removed the aborted bit and aborts counter")
https://lore.kernel.org/asYfmLsfy-SCxXjr@sirena.co.uk
Adjacent changes:
Documentation/networking/strparser.rst
5f193cdabf25 ("docs: networking: strparser: remove general mode")
0984ebc63179 ("strparser: make sure __strp_recv isn't running before tearing down the parser")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
ieee80211_fill_rate_info() copies the driver's rate_info bandwidth into
the rx_status as is, but ieee80211_get_rate_duration() only handles
20/40/80/160/320 MHz. When a driver reports another width, e.g. an
HE/EHT RU allocation, it hits WARN_ON_ONCE() and returns 0 without
setting *overhead. ieee80211_rate_expected_tx_airtime() then goes on
using the uninitialized overhead, as Dan's static checker report
points out.
Seen on a Microsoft Surface Pro 11 (WCN7850, ath12k) while
NetworkManager dumps station info:
WARNING: net/mac80211/airtime.c:532 at ieee80211_get_rate_duration.isra.0+0xc0/0x180 [mac80211]
ieee80211_get_rate_duration.isra.0+0xc0/0x180 [mac80211]
sta_estimate_expected_throughput.isra.0+0x38/0x64 [mac80211]
sta_set_sinfo+0x7fc/0x8e4 [mac80211]
ieee80211_dump_station+0x4c/0x84 [mac80211]
nl80211_dump_station+0x594/0x7e8 [cfg80211]
Only accept widths that ieee80211_get_rate_duration() handles in
ieee80211_fill_rate_info(). Other widths then give no estimate instead
of a warning. Also return 0 from ieee80211_rate_expected_tx_airtime()
when no duration could be computed, as ieee80211_calc_rx_airtime()
already does.
Tested on the Surface Pro 11 with next-20260929: Wi-Fi works and
"expected throughput" is still reported (about 1049 Mbit/s for a
1297 Mbit/s HE 160 MHz tx rate). The warning itself is rare (twice in
several days of use), so this does not prove it is gone, only that the
change does not break the normal path.
Fixes: 094dc1619cb0 ("wifi: mac80211: factor out part of ieee80211_calc_expected_tx_airtime")
Reported-by: Dan Carpenter <error27@gmail.com>
Closes: https://lore.kernel.org/all/aneGlCMVnqRhlf6p@stanley.mountain/
Assisted-by: LLM
Signed-off-by: François Roux <franzelfranzel@gmail.com>
Link: https://patch.msgid.link/20261006083751.4015-1-franzelfranzel@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
In 4-address mode, ieee80211_build_hdr() and ieee80211_check_fast_xmit()
set A1 to sdata->deflink.u.mgd.bssid, which is not valid for an MLO
connection.
Use sdata->vif.cfg.ap_addr instead, like the 3-address case does.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Link: https://patch.msgid.link/20260930084819.3000572-2-nbd@nbd.name
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
When netdev registration fails or an interface cannot be moved out of a
dying network namespace, the netdev core calls ->ndo_uninit()
(ieee80211_uninit) with RTNL held.
However, ieee80211_uninit() calls ieee80211_teardown_sdata() without
holding wiphy_lock. During teardown, ieee80211_free_keys() invokes
wiphy_delayed_work_cancel(), which triggers a lockdep warning:
WARNING: CPU: 0 PID: 6112 at net/wireless/core.c:1839 wiphy_delayed_work_cancel+0x85/0xb0
...
Call Trace:
ieee80211_free_keys+0x31a/0x3c0 net/mac80211/key.c:380
ieee80211_teardown_sdata+0x2b8/0x10f0 net/mac80211/iface.c:928
ieee80211_uninit+0x3f/0x50 net/mac80211/iface.c:954
rollback_registered_many+0xa5c/0xef0 net/core/dev.c:10065
In the normal deletion path (ieee80211_del_virtual_intf), both RTNL and
wiphy_lock are held when ieee80211_teardown_sdata() runs, and the sdata
is removed from local->interfaces.
Change ieee80211_unlist_sdata() to return a boolean indicating whether the
sdata was present in the list, and acquire wiphy_lock in ieee80211_uninit()
only if the interface is actually being unlisted here.
Reported-by: syzbot+e9cbd080a7801f06cf08@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=e9cbd080a7801f06cf08
Tested-by: syzbot+e9cbd080a7801f06cf08@syzkaller.appspotmail.com
Signed-off-by: Raj Ojha <rajojha047@gmail.com>
Link: https://patch.msgid.link/20261003112741.52-1-rajojha047@gmail.com
[dissolve ieee80211_unlist_sdata() to avoid weird locking]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|