Add mailmap entries for some Qualcomm Wi-Fi developers now using
@oss.qualcomm.com credentials.
For ath12, add support for the qcom,calibration-variant Device Tree
property and add more infrastructure for the IPQ5332 platform.
Other than that, just an assortment of cleanups and minor bug fixes across
the "ath" family of drivers that aren't urgent enough to go through the
"current" process.
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQQ/mtSHzPUi16IfDEksFbugiYzLewUCaqqncgAKCRAsFbugiYzL
e/I7AP9XZ+pkgUYw1MUUBiey/Ov/bVQKd/DYU3jRr6Zz0lJw3AEA+ZpSxbef1AcK
txnHjzKtKNLm5dXsvZay9NOR3wVVYAM=
=M4Qu
-----END PGP SIGNATURE-----
Merge tag 'ath-next-20260916' of git://git.kernel.org/pub/scm/linux/kernel/git/ath/ath
Jeff Johnson says:
==================
ath.git patches for v7.4 (PR #1)
Add mailmap entries for some Qualcomm Wi-Fi developers now using
@oss.qualcomm.com credentials.
For ath12, add support for the qcom,calibration-variant Device Tree
property and add more infrastructure for the IPQ5332 platform.
Other than that, just an assortment of cleanups and minor bug fixes across
the "ath" family of drivers that aren't urgent enough to go through the
"current" process.
==================
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The station may be relevant for queuing and will also generally be
resolved in some cases. However, we want to be able to prevent looking
up the station based on the address.
Add a station parameter, which can be set to the correct station, to an
error value to prevent station lookup or to NULL to get the old
behaviour where the address is used to find the appropriate station.
Also disable the station lookup for ieee80211_tx_skb_tid_band already as
it does not make any sense to find a station when doing an off-channel
transmit.
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Link: https://patch.msgid.link/20260915151925.759b6144ddf9.Ib7be90db0d3712d14e7a292023ff3d922baef860@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
For MLD stations, userspace may need to explicitly transmit a frame to
a specific link address. In that case, it needs to ensure that no
address translation happens.
In the reverse case of an RX, userspace may need to know the link
address for a frame. By passing the information whether a STA is known
for the frame, userspace knows whether link translation happened and
can do the reverse lookup when needed.
This is important for flows where a STA is still registered but the
connection has been lost and it is returning again.
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Link: https://patch.msgid.link/20260915151925.f12af80b1972.I4cfd3db83ba3dc0267575f919ba613ea5d10abd5@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The code has grown over time and it was not correctly handling all
cases. Examples of issues are that for management frames we would match
the received address against the MLD address even though we should not
resolve the station in that case. Another issue was that for data frames
we would not correctly fall back to use link addresses when the driver
does not provide the pubsta already.
The new code still uses the same approach as before. For data frames,
assume that a valid station exists and interfaces do not need to be
iterated. On the other hand, for non-data frames (or if a data frame
without a station is received) all interfaces must be iterated.
Note that this rework makes mac80211 slightly more strict. For example,
previously mac80211 would have incorrectly accepted a data frame that
was transmitted on the wrong link.
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Link: https://patch.msgid.link/20260915151925.de2fc9a47273.Ie12ec077142c6a7fdbb58cdfc5660497cc75150e@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The link ID for one frequency may be different between VIFs. As such,
the sensible thing is to set the link ID later in the flow once the SKB
is duplicated for each VIF. As the link ID is not passed in from outside
mac80211 it is always valid and the corresponding bit can be dropped.
Also switch a few more places to pass the link STA as that is a natural
way to pass the link information around.
This fixes the per-link statistics when frames are reordered.
Note that this patch deliberately leaves some places unchanged and even
incorrect as this will be addressed by further refactorings.
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Link: https://patch.msgid.link/20260915151925.467593bc1105.I8abe3601d0335dca22d48bcee864817cfccf0de0@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
If a station is passed then the link ID also needs to be known. As such,
it is a more natural API to simply pass the link station directly rather
than pushing the link information into the RX status.
Furthermore, having the link ID in the RX status is not actually correct
because the link IDs are VIF specific and there may be multiple VIFs. In
the case of a station this relationship is clear, but then one may as
well use the link station.
This patch only changes the API and emulates the old (incorrect)
behaviour for now. The mac80211 RX code will be updated in later
patches.
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Link: https://patch.msgid.link/20260915151925.06f41565116a.I4a2d45609e94b52654b10ec572e59a45d09c41f4@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Using the link_sta is a natural way to pass both the STA and the link ID
information at the same time. Use that internally to the driver in
preparation to mac80211 changing its API and adopting the same method.
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Link: https://patch.msgid.link/20260915151925.0b1269375e9e.I41df561f8129fac1327a07729cf4734b8752bcd6@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
wireless.wiki.kernel.org has not existed for a while. Update all links to
point to the current rendered pages at wireless.docs.kernel.org.
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Link: https://patch.msgid.link/20260915-wireless-wiki-v1-1-21dc76684b44@oss.qualcomm.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
brcmf_bss_connect_done() receives the firmware event in @e but discards it
on the failure path. Every failed connect is reported to cfg80211 as
WLAN_STATUS_AUTH_TIMEOUT (16), whatever the firmware actually said, so
userspace only ever sees:
wlan0: CTRL-EVENT-ASSOC-REJECT bssid=00:00:00:00:00:00 status_code=16
The all-zero BSSID comes from the same place: conn_params is memset to
zero and profile->bssid has not been filled in when the station never
associated.
status_code=16 therefore carries no information about the cause. It is not
an AP response and it does not mean "authentication timed out" - it is the
only failure value this driver can produce. This is a recurring source of
confusion: [1] has been open since 2023 with more than twenty follow-ups
and no explanation of the code, and covers BCM4345/6, BCM43430 and
CYW43455 across several kernel versions. It supersedes [2], filed against
the firmware repository a day earlier and closed in favour of it.
The firmware's own status (BRCMF_E_STATUS_*) is more specific - FAIL,
TIMEOUT, NO_NETWORKS, ABORT and so on - and it is already in hand. Log it
so the cause can be narrowed down without rebuilding the kernel.
bphy_err() is used rather than brcmf_dbg() or brcmf_info(): it is the form
the rest of this file uses, and of the three it is the only one that is
both visible in a distribution kernel and bounded. brcmf_dbg(CONN)
expands to no_printk() unless CONFIG_BRCMDBG is set, which is what makes
the reports in [1] and [2] impossible to act on - the people hitting this
run stock kernels. brcmf_info() expands to a plain pr_info() in a
non-debug build and is not rate limited, and wpa_supplicant retries the
association every few seconds, so it would flood the log. bphy_err() is
guarded by net_ratelimit() there, so a station retrying against an
unreachable AP prints at most a few lines per second.
All three fields are printed because only some of them are meaningful on
each path into the failure branch. brcmf_is_nonetwork() keys off @status,
so that is the useful field for a join that never associated - the case in
[1]. brcmf_is_linkdown() keys off @event_code and @flags and does not look
at @status at all, so @status can read 0 there; on those events it is
@reason which carries the 802.11 reason code, as the mapping in
brcmf_map_fw_linkdown_reason() shows. Printing the three together lets
the reader tell which path was taken instead of guessing from one number.
The status reported to cfg80211 is left alone; changing it would alter what
userspace sees.
[1] https://github.com/RPi-Distro/firmware-nonfree/issues/38
[2] https://github.com/raspberrypi/firmware/issues/1829
Assisted-by: Claude:claude-opus-5
Signed-off-by: Ryohei Hashimoto <laurel.medalist12@gmail.com>
Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Link: https://patch.msgid.link/20260913165526.4426-1-laurel.medalist12@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Advertise NL80211_EXT_FEATURE_FAST_ROAM_OFFLOAD only when firmware
reports FBT or OKC support. That lets userspace provide PMK material for
firmware fast roaming without enabling the path on devices that cannot
complete the offloaded roam.
Assisted-by: GitHub-Copilot-CLI:gpt-5.5
Signed-off-by: Carella Chen <carella.chen@infineon.com>
Signed-off-by: Jason Huang <jason.huang2@infineon.com>
Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Link: https://patch.msgid.link/20260911065656.1269623-6-Jason.Huang2@infineon.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Firmware can complete FT or OKC roaming before the driver reports the
connect or roam event to cfg80211. Detect those successful offloaded
cases when the profile uses firmware 1X or roaming offload and the
association request carries PMK cache state, FT is in use, or OKC is
enabled.
After reporting the connect or roam event, call cfg80211_port_authorized()
so nl80211 emits the dedicated NL80211_CMD_PORT_AUTHORIZED event instead
of reusing the reserved NL80211_ATTR_PORT_AUTHORIZED flag in CONNECT/ROAM
notifications.
Assisted-by: GitHub-Copilot-CLI:gpt-5.5
Signed-off-by: Darren Li <hsin-hung.li@cypress.com>
Signed-off-by: Chung-Hsien Hsu <Chung-Hsien.Hsu@infineon.com>
Signed-off-by: Chi-hsien Lin <chi-hsien.lin@cypress.com>
Signed-off-by: Carella Chen <carella.chen@infineon.com>
Signed-off-by: Jason Huang <jason.huang2@infineon.com>
Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Link: https://patch.msgid.link/20260911065656.1269623-5-Jason.Huang2@infineon.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
FT and OKC roaming offload need PMK material in firmware even when
the initial 802.1X exchange is handled by userspace. Add a roaming
firmware-supplicant profile state for those connections and allow PMK
configuration while firmware FT or OKC state is active.
Program the OKC PMK through the okc_info_pmk iovar before setting the
regular PMK when firmware reports OKC state. Only select the roaming
firmware-supplicant state when firmware advertises FBT or OKC support;
devices without those capabilities keep the profile in the userspace
supplicant state and do not issue roaming-offload specific iovars on
every connect.
For FT-PSK with firmware supplicant support, keep the profile in the
PSK firmware-supplicant state.
Assisted-by: GitHub-Copilot-CLI:gpt-5.5
Signed-off-by: Darren Li <hsin-hung.li@cypress.com>
Signed-off-by: Chung-Hsien Hsu <Chung-Hsien.Hsu@infineon.com>
Signed-off-by: Chi-hsien Lin <chi-hsien.lin@cypress.com>
Co-developed-by: Carella Chen <carella.chen@infineon.com>
Signed-off-by: Carella Chen <carella.chen@infineon.com>
Co-developed-by: Shelley Yang <shelley.yang@infineon.com>
Signed-off-by: Shelley Yang <shelley.yang@infineon.com>
Signed-off-by: Jason Huang <jason.huang2@infineon.com>
Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Link: https://patch.msgid.link/20260911065656.1269623-4-Jason.Huang2@infineon.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Some firmware advertises Fast BSS Transition and Opportunistic Key
Caching support through the firmware capability string. Track those
capabilities so later roaming offload handling can be enabled only
when firmware reports support.
Assisted-by: GitHub-Copilot-CLI:gpt-5.5
Signed-off-by: Carella Chen <carella.chen@infineon.com>
Signed-off-by: Jason Huang <jason.huang2@infineon.com>
Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Link: https://patch.msgid.link/20260911065656.1269623-3-Jason.Huang2@infineon.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Drivers can offload FT or OKC roaming without necessarily advertising
full 802.1X 4-way handshake offload. Add an extended feature bit for
that capability and allow PMK configuration when it is present.
This gives userspace a way to provide PMK or PMK-R0 material needed by
fast roaming firmware implementations while keeping the existing 4-way
handshake offload capability unchanged.
Assisted-by: GitHub-Copilot-CLI:gpt-5.5
Signed-off-by: Carella Chen <carella.chen@infineon.com>
Signed-off-by: Jason Huang <jason.huang2@infineon.com>
Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Link: https://patch.msgid.link/20260911065656.1269623-2-Jason.Huang2@infineon.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The rt2x00 PCI and USB probe paths allocate EEPROM and RF storage with
plain kzalloc() and then free it from bus-specific teardown helpers. The
USB path also manages the CSR cache the same way. These buffers are
tied to the device lifetime, so the explicit free paths add probe and
disconnect cleanup without providing separate ownership.
Allocate the buffers with devm_kzalloc() before the mac80211 hardware is
allocated, then attach the resulting storage to struct rt2x00_dev after
the driver-private state exists. This lets driver detach and probe
failure rely on device-managed cleanup and removes the duplicated
bus-specific buffer freeing.
Assisted-by: Codex:GPT-5.5
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Acked-by: Stanislaw Gruszka <stf_xl@wp.pl>
Link: https://patch.msgid.link/20260904211628.83180-1-rosenp@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The 'secure link' feature is not supported by upstream (the reference
code relies on mbedTLS to implement it, which is not an option for the
Linux kernel).
If the firmware enforce the 'secure link' feature, wfx_probe() exits
early. However, err still holds the value returned by the previous call
to wait_for_completion_timeout(), which is the number of jiffies left
before the timeout, hence strictly positive (the zero case is handled
just above).
wfx_probe() therefore returns a positive value. The bus probe functions
forward it as-is and the driver core negates it (see really_probe()).
The device does not get bound, which is the expected outcome, but the
reported error code is a meaningless negative jiffies count instead of
an errno.
Return -EOPNOTSUPP, as already done a few lines above when the firmware
API version is not supported.
Copilot reviewed the code and wrote some parts of the log.
Reported-by: Alexander Sverdlin <alexander.sverdlin@siemens.com>
Closes: https://lore.kernel.org/all/cc4de4e40e3ca50e7288be2b9067802576567c25.camel@siemens.com/
Assisted-by: Copilot:claude-opus-5
Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com>
Reviewed-by: Alexander Sverdlin <alexander.sverdlin@siemens.com>
Link: https://patch.msgid.link/20260906204831.949692-3-jerome.pouiller@silabs.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
When the device reports that it is too hot, wfx_suspend_hot_dev() blocks
the Tx queues and arms cooling_timeout_work with a 10s delay. If the
device recovers a normal temperature before the delay expires, the work
is canceled. Else, the work declares the chip frozen and unblocks the Tx
queues.
However, this work is never canceled when the device is removed.
cooling_timeout_work is queued on the system workqueue, while struct
wfx_dev is released by wfx_free_common() (through ieee80211_free_hw()).
So, if the device is unbound during this 10s window, the work fires
after struct wfx_dev has been freed and dereferences it.
Cancel the work during the teardown. It has to be done after
wfx_bh_unregister(): the "device too hot" indication is processed by the
bh, so canceling the work earlier would allow the bh to rearm it. On the
other hand, the work calls wfx_tx_unlock(), which may in turn call
wfx_bh_request_tx(). So it has to be canceled before bh_wq is destroyed.
Note that the Tx queues are intentionally left blocked: the device is
going away, so there is nothing to unblock.
The issue was reported by the Sashiko review bot and the fix has been
written by Copilot (including the commit log). The use-case is difficult
to reproduce, so this code has not been tested. However, I don't believe
this patch could cause any regression.
Fixes: 1d52d29983 ("staging: wfx: add support for 'device too hot' indication")
Assisted-by: Sashiko:gemini-3.1-pro-preview
Assisted-by: Copilot:claude-opus-5
Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com>
Tested-by: Alexander Sverdlin <alexander.sverdlin@siemens.com>
Reviewed-by: Alexander Sverdlin <alexander.sverdlin@siemens.com>
Link: https://patch.msgid.link/20260906204831.949692-2-jerome.pouiller@silabs.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
In a new test framework for hwsim I'm working on,
tests can modify the frame as it passes through
the wmediumd APIs. If then the original frame is
reported on the monitor interface, things get a
bit confusing.
Overwrite the frame that will be reported to the
monitor with the frame reported for RX if both
transmitter and cookie attributes are present
(wmediumd doesn't set them).
Also add a HWSIM_ATTR_NO_MONITOR flag to allow a
frame to be dropped entirely from monitor report,
e.g. in case it was never delivered to any peer.
Link: https://patch.msgid.link/20260911154632.c06da6d18055.Iaba24acec4c5765fbad77376418b17d1b863791b@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
A round is ended when ieee80211_next_txq() meets a txq whose remembered
round number is the current one, meaning it has gone full circle.
When the AC exceeds the airtime limit, ieee80211_txq_schedule_start()
sets the round number to 0 to close the round. The next open round is
1. A txq from an earlier round 1 still remembers the "1", so the next
round stops on it and serves nothing.
Fix: use a separate open/closed flag and let the round number keep
counting.
Testing: ~4200 skipped selections per 20s without the fix and 20
skipped selections per 20s with the fix on ath11k (with my AQL series
applied) at BE 500/1000.
Fixes: 8e4bac0671 ("wifi: mac80211: add a per-PHY AQL limit to improve fairness")
Assisted-by: Claude:claude-opus-5
Signed-off-by: Julius Bairaktaris <julius@bairaktaris.de>
Reviewed-by: Toke Høiland-Jørgensen <toke@toke.dk>
Link: https://patch.msgid.link/20260908144155.756569-1-julius@bairaktaris.de
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
mesh_path_add() has returned the new (or the already existing) struct
mesh_path, or an ERR_PTR(), since commit ae76eef027 ("mac80211: return
new mpath from mesh_path_add()"), but its kernel-doc still says
"Returns: 0 on success". Describe the pointer.
Fixes: ae76eef027 ("mac80211: return new mpath from mesh_path_add()")
Assisted-by: LLM
Signed-off-by: Karl Mehltretter <kmehltretter@gmail.com>
Link: https://patch.msgid.link/20260912072840.50710-1-kmehltretter@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
ieee80211_tx_control_port() can be called from nl80211_tx_control_port()
in normal process context via netlink sendmsg(), where BH/preemption
is not disabled. dev_sw_netstats_tx_add() uses this_cpu_ptr() internally,
which requires preemption to be disabled, triggering a
"BUG: using smp_processor_id() in preemptible code" warning.
Fix this by moving the local_bh_disable()/local_bh_enable() section to
also cover dev_sw_netstats_tx_add() and ieee80211_tpt_led_trig_tx().
Fixes: d5a014204a ("wifi: mac80211: tx: simplify control port frame transmission")
Reported-by: syzbot+d979bd35c8a76fd1b6f5@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=d979bd35c8a76fd1b6f5
Signed-off-by: Deepanshu Kartikey <kartikey406@gmail.com>
Link: https://patch.msgid.link/20260910041047.24437-1-kartikey406@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
hwsim_mcast_config_msg() falls back to genlmsg_multicast() when it has no
genl_info, and that only reaches init_net. A radio in another namespace
therefore has its DEL_RADIO announced in init_net, while its own namespace
hears nothing.
Multicast to wiphy_net() instead, the net that GET and DUMP already filter
on. Tested with a config-group listener in each namespace, on a base and a
patched kernel.
Assisted-by: Claude:claude-opus-5 codeql
Signed-off-by: Maoyi Xie <maoyixie.tju@gmail.com>
Link: https://patch.msgid.link/20260904140300.3402053-1-maoyixie.tju@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
When bringing a monitor interface up and down again, the
refcount is leaked for active monitor interfaces, so the
hardware filter is never reset.
Move the decrement out so it mirrors the increment.
Signed-off-by: Devin Wittmayer <lucid_duck@justthetip.ca>
Link: https://patch.msgid.link/20260904202817.12564-1-lucid_duck@justthetip.ca
[rewrite commit message better]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
IPQ5332 platform uses a MultiPD architecture where multiple firmware
instances share common read-only code segments. The driver currently
loads only the per-UserPD firmware, which leaves the shared read-only
firmware unloaded.
Add support for loading the shared read-only firmware once before the
UserPDs are powered up. The shared segment contains common code that is
identical across all UserPDs and reduces the total firmware memory
requirement.
Store the reserved memory mapping in struct ath12k_ahb_rproc_info so it can
be shared by all UserPDs and mapped only once. Keep the mapping valid
across UserPD power cycles, and unmap it when the last UserPD is removed.
Shut down the shared firmware from the rproc deconfiguration path when the
last UserPD is removed.
Also, pass NULL for the reloc_base parameter because the driver uses fixed
memory regions configured through device tree and does not need the
adjusted physical address after relocation.
Tested-on: IPQ5332 hw1.0 AHB WLAN.WBE.1.6-01275-QCAHKSWPL_SILICONZ-1
Signed-off-by: Aaradhana Sahu <aaradhana.sahu@oss.qualcomm.com>
Reviewed-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Reviewed-by: Vasanthakumar Thiagarajan <vasanthakumar.thiagarajan@oss.qualcomm.com>
Link: https://patch.msgid.link/20260811054316.518494-1-aaradhana.sahu@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Commit c4f5187364 ("wifi: ath9k: Remove redundant rcu_read_lock/unlock()
in spin_lock") removed the explicit rcu_read_lock()/rcu_read_unlock() pair
from ath_txq_schedule(), relying on spin_lock_bh() to provide an implicit
RCU read-side critical section.
That is correct on kernels without CONFIG_DEBUG_LOCK_ALLOC, where
rcu_read_lock_any_held() falls back to !preemptible() as a proxy, and
spin_lock_bh() disables preemption.
However, rcu_dereference() in ath_merge_ratetbl() checks
rcu_read_lock_held(), which under CONFIG_DEBUG_LOCK_ALLOC only returns
true when lock_is_held(&rcu_lock_map) — set exclusively by an explicit
rcu_read_lock(). A spin_lock_bh() does not set rcu_lock_map, so on a
CONFIG_DEBUG_LOCK_ALLOC + CONFIG_PROVE_RCU kernel, any call path that
reaches ath_merge_ratetbl() with a non-NULL sta will produce a lockdep
splat.
The correct fix is to use rcu_dereference_bh(), whose validity check
calls rcu_read_lock_bh_held(), which returns true whenever BH is disabled
(in_softirq() || irqs_disabled()). This matches the actual protection
at all callers of ath_set_rates() that pass a non-NULL sta: they all hold
a spin_lock_bh() (either sc->chan_lock or txq->axq_lock).
Fixes: c4f5187364 ("wifi: ath9k: Remove redundant rcu_read_lock/unlock() in spin_lock")
Assisted-by: LLM
Acked-by: Toke Høiland-Jørgensen <toke@toke.dk>
Link: https://patch.msgid.link/20260813-ath9k-rcu-fix-v1-1-424665a1ce60@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
A monitor PPDU may span multiple status buffers. However,
__ath12k_wifi7_dp_mon_process_ring() clears mon_ppdu_info before
processing every buffer.
As a result, continuation buffers lose PPDU state collected from
earlier buffers, including ppdu_id and PHY metadata.
This can cause monitor RX processing to fail when
ath12k_wifi7_dp_rx_mon_dest_process() observes a PPDU ID mismatch
between status and destination rings, preventing MSDUs from being
delivered. It can also produce incomplete EHT metadata that triggers
warnings from mac80211:
Rate marked as an EHT rate but data is invalid: MCS:0, NSS:0
ath12k_wifi7_dp_mon_srng_process() already tracks status TLV
continuations via ppdu_continuation. Apply the same logic here and
only reinitialize mon_ppdu_info when starting a new PPDU.
mon_ppdu_info persists across ar->monitor_started transitions. If
monitor mode stops while a PPDU is mid-continuation, ppdu_continuation
stays set and could leak into the next monitor session. Reset it
when monitor mode starts.
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.1.c7-00108-QCAHMTSWPL_V1.0_V2.0_SILICONZ_UPSTREAM-3
Fixes: 78d3d907d0 ("wifi: ath12k: add support to reap and process monitor status ring")
Signed-off-by: Kang Yang <kang.yang@oss.qualcomm.com>
Reviewed-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Reviewed-by: Tamizh Chelvam Raja <tamizh.raja@oss.qualcomm.com>
Link: https://patch.msgid.link/20260828065534.1537-1-kang.yang@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
When external IRQ configuration fails, the driver does not release all
IRQs that were successfully requested before the failure. This can leak
IRQ resources during probe failure.
Free previously requested external IRQs when external IRQ configuration
fails.
Also remove the NAPI instance with netif_napi_del() before freeing the
associated netdev to properly clean up the NAPI resources.
Store the IRQ number only after request_irq() succeeds to avoid recording
an IRQ that was not successfully requested.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.6-01243-QCAHKSWPL_SILICONZ-1
Fixes: d889913205 ("wifi: ath12k: driver for Qualcomm Wi-Fi 7 devices")
Signed-off-by: Aaradhana Sahu <aaradhana.sahu@oss.qualcomm.com>
Reviewed-by: Rameshkumar Sundaram <rameshkumar.sundaram@oss.qualcomm.com>
Reviewed-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Link: https://patch.msgid.link/20260819110215.2485514-3-aaradhana.sahu@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
When CE IRQ configuration fails, the driver does not release all IRQs
that were successfully requested before the failure. This can leak IRQ
resources during probe failure.
Free the previously requested CE IRQs before returning from the error
path to ensure that partially initialized IRQ resources are properly
cleaned up during probe failure.
Factor out the CE IRQ cleanup into a helper to reuse the cleanup logic
during both error handling and driver teardown.
Also free CE IRQs when external IRQ configuration fails, before
returning from the error path.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.6-01243-QCAHKSWPL_SILICONZ-1
Fixes: d889913205 ("wifi: ath12k: driver for Qualcomm Wi-Fi 7 devices")
Signed-off-by: Aaradhana Sahu <aaradhana.sahu@oss.qualcomm.com>
Reviewed-by: Rameshkumar Sundaram <rameshkumar.sundaram@oss.qualcomm.com>
Reviewed-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Link: https://patch.msgid.link/20260819110215.2485514-2-aaradhana.sahu@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
ath12k_wifi7_hal_tx_cmd_ext_desc_setup() programs the low 32 bits of the
TX buffer physical address into HAL_TX_MSDU_EXT_INFO0_BUF_PTR_LO but
hardcodes the high bits field HAL_TX_MSDU_EXT_INFO1_BUF_PTR_HI to 0x0.
dma_addr_t is 64-bit on platforms with CONFIG_ARCH_DMA_ADDR_T_64BIT, so
whenever a TX buffer is mapped above the 4 GB boundary the upper bits of
paddr are dropped and the firmware is handed a truncated address,
leading to potential TX failures.
Write the upper bits of paddr into BUF_PTR_HI to fix this issue.
Issue found during code review, compile tested only.
Fixes: d889913205 ("wifi: ath12k: driver for Qualcomm Wi-Fi 7 devices")
Signed-off-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Reviewed-by: Rameshkumar Sundaram <rameshkumar.sundaram@oss.qualcomm.com>
Link: https://patch.msgid.link/20260818-ath12k-truncated-tx-dma-addr-v1-1-ead978969371@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
ath12k_wifi7_dp_tx() caches hdr, eth, and skb_cb from the skb before
calling ath12k_dp_tx_align_payload(). That function may shift skb->data
in place (when headroom or tailroom is sufficient) or reallocate the
buffer entirely via skb_realloc_headroom(), freeing the original skb.
In either case hdr, eth, and skb_cb are left pointing into stale memory.
After alignment, only hdr is refreshed, leaving eth and skb_cb stale.
skb_cb is written immediately after (storing DMA addresses), and eth is
re-read on every TCL ring retry via the tcl_ring_sel goto, so both
accesses are use-after-free or stale-pointer bugs depending on which
alignment path was taken.
Refresh eth (conditionally, to preserve the encap-mode distinction) and
skb_cb alongside hdr after ath12k_dp_tx_align_payload() returns, so all
three point into the live skb for all subsequent accesses.
Issue found during code review, compile tested only.
Fixes: 38055789d1 ("wifi: ath12k: use 128 bytes aligned iova in transmit path for WCN7850")
Signed-off-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Reviewed-by: Rameshkumar Sundaram <rameshkumar.sundaram@oss.qualcomm.com>
Link: https://patch.msgid.link/20260818-ath12k-uaf-for-aligned-tx-v1-1-d6ae195b15e7@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
ath12k_wifi7_dp_tx() maps the original MSDU into ti.paddr and
stores the address in skb_cb->paddr. When an extended MSDU descriptor
is used, the function later maps the ext descriptor and overwrites
ti.paddr and ti.data_len with the ext-desc DMA address and length.
If TCL ring allocation then fails, the error path first unmaps the
ext-desc DMA address from skb_cb->paddr_ext_desc. It then falls through
to fail_unmap_dma, which uses ti.paddr/ti.data_len for the original
MSDU unmap. At that point ti.paddr still refers to the ext-desc
mapping, so the ext descriptor is unmapped twice and the original MSDU
mapping is left mapped.
Fix the original MSDU unwind to use skb_cb->paddr and skb->len instead
of ti.paddr/ti.data_len. The ti fields cannot be used after the ext
descriptor is mapped because they are intentionally reused for the
buffer submitted to TCL. Also clear skb_cb->paddr_ext_desc after the
ext-desc unmap so a TCL ring retry cannot observe stale ext-desc DMA
state from the previous attempt.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.6-01243-QCAHKSWPL_SILICONZ-1
Fixes: 37a068fc9d ("wifi: ath12k: Handle error cases during extended skb allocation")
Signed-off-by: Rameshkumar Sundaram <rameshkumar.sundaram@oss.qualcomm.com>
Reviewed-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Link: https://patch.msgid.link/20260813-ext-msdu-fix-v1-1-c5e80ccd061a@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
On failure, ath12k_mac_setup_channels_rates() frees band channel
arrays but leaves the pointers set, or clears the wrong band after a
copy-paste typo (6 GHz free paired with a 2 GHz NULL). Clear the
matching sbands[].channels pointer after each kfree().
Compile tested only.
Signed-off-by: Linkai Gong <gonglinkai@kylinos.cn>
Reviewed-by: Rameshkumar Sundaram <rameshkumar.sundaram@oss.qualcomm.com>
Reviewed-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Link: https://patch.msgid.link/20260811020523.387433-1-gonglinkai@kylinos.cn
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Replace the userspace stdint types uint64_t/uint8_t with the kernel
types u64/u8 to conform to the surrounding ath12k code, which uses the
kernel types everywhere else.
No functional changes, compile tested only.
Signed-off-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Reviewed-by: Rameshkumar Sundaram <rameshkumar.sundaram@oss.qualcomm.com>
Link: https://patch.msgid.link/20260819-ath12k-cleanup-userspace-types-v1-1-161a1e8910f5@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Only the two QCN9274 hw_params advertise NL80211_IFTYPE_AP_VLAN; the
IPQ5332 entry does not. ath12k sets SW_CRYPTO_CONTROL, so mac80211 does
not add the mode on the driver's behalf either -- ieee80211_alloc_hw_nm()
only does that for drivers that leave SW_CRYPTO_CONTROL clear, leaving
the declaration to drivers that can transmit software-encrypted frames.
AP/VLAN is therefore absent from the wiphy, and creating an AP_VLAN
interface without NL80211_ATTR_4ADDR is rejected by
cfg80211_iftype_allowed():
nl80211: Failed to create interface <name>: -95 (Operation not supported)
That is the path hostapd uses for dynamic per-station VLANs, in
hostapd_vlan_if_add(). The 4-address (WDS) case is not affected, as
cfg80211_iftype_allowed() short-circuits it on WIPHY_FLAG_4ADDR_AP,
which mac80211 sets unconditionally at alloc time.
The QCN9274 entries already declare AP_VLAN, and the transmit path is
shared Wi-Fi 7 code with no per-chip or per-bus gating, so IPQ5332 can
deliver it as well. AP_VLAN is a software interface type, so no
interface combination changes are needed.
On a mixed-bus single-wiphy group the effect is wider still, since
ath12k_mac_get_ifmodes() intersects interface_modes across all radios:
one IPQ5332 masks AP_VLAN for the QCN9274 radios too.
Advertise AP_VLAN on IPQ5332 as QCN9274 does. With the mode advertised,
a non-4addr AP_VLAN interface can be created on an AP interface of an
IPQ5332 + 2x QCN9274 single-wiphy group:
# iw dev phy0.0-ap0 interface add tvlan0 type __ap_vlan
# iw dev tvlan0 info
Interface tvlan0
ifindex 19
wdev 0x9
addr 00:03:7f:12:0b:0b
type AP/VLAN
wiphy 0
Radios: 0 1 2
RADIUS-assigned dynamic VLANs were not exercised end to end.
Tested-on: IPQ5332 hw1.0 AHB WLAN.WBE.1.6-01270-QCAHKSWPL_SILICONZ-1
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.6-01243-QCAHKSWPL_SILICONZ-1
Signed-off-by: Kamil Bienkiewicz <perceivalpercy@gmail.com>
Reviewed-by: Rameshkumar Sundaram <rameshkumar.sundaram@oss.qualcomm.com>
Link: https://patch.msgid.link/20260828005126.1685832-1-perceivalpercy@gmail.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
In this function, we iterate over dp->reo_cmd_cache_flush_list using
list_for_each_entry_safe(), under dp->reo_cmd_lock for concurrent access,
and for each expired entries we :
- drop the lock
- call ath11k_dp_reo_cache_flush()
- kfree() the entry
- retake the lock to keep on iterating
list_for_each_entry_safe() protects us from deleting the entry during
iteration but doesn't protect for concurrent access. So another thread can
take the lock and modify the list in between and crash like below.
To fix the issue, move all entries that needs to be freed to a local list
while under the lock and then iterate over the list to free the entries
without holding the lock.
BUG: Unable to handle kernel paging request at virtual address 00000010ddbeef8c
Call trace:
ath11k_dp_rx_tid_del_func+0x164/0x3c8
ath11k_dp_process_reo_status+0x1d4/0x2fc
ath11k_dp_service_srng+0x334/0x338
ath11k_pcic_ext_grp_napi_poll+0x30/0xc0
__napi_poll+0x34/0x184
napi_threaded_poll+0xb4/0x1d8
kthread+0xdc/0xe0
ret_from_fork+0x10/0x20
Tested-on: QCN9074 hw1.0 PCI WLAN.HK.2.9.0.1-01977-QCAHKSWPL_SILICONZ-1
Fixes: d5c65159f2 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
Suggested-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Signed-off-by: Nicolas Escande <nico.escande@gmail.com>
Reviewed-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Reviewed-by: Rameshkumar Sundaram <rameshkumar.sundaram@oss.qualcomm.com>
Link: https://patch.msgid.link/20260813-ath11k-locking-v2-1-f4113a0bd0f4@gmail.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
The previous logic in ath11k_ce_rx_post_pipe() incorrectly required both
dest_ring and status_ring to be NULL in order to exit the function.
This caused the function to continue even if only one of the pointers
was NULL, potentially leading to null pointer dereferences in
ath11k_ce_rx_buf_enqueue_pipe().
Fix the condition by modifying the logic so that the function returns
early if either dest_ring or status_ring is NULL.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Fixes: d5c65159f2 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
Signed-off-by: Mikhail Lobanov <m.lobanov@rosalinux.ru>
Reviewed-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Reviewed-by: Rameshkumar Sundaram <rameshkumar.sundaram@oss.qualcomm.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Link: https://patch.msgid.link/20240909150824.28195-1-m.lobanov@rosalinux.ru
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Drivers without FAST_XMIT and hardware scan support currently drop TX
frames during software scanning while being off-channel. This happens
in ieee80211_tx_h_check_assoc() before frames reach internal TX queue.
Move the off-channel check to invoke_tx_handlers_late(), after
ieee80211_queue_skb(). Frames can then remain queued and be transmitted
when the interface returns on-channel. This how TX works currently when
code goes via ieee80211_xmit_fast() path.
Below is rt2x00 driver ping output example while scanning, before and
after the fix. After the fix, packets are no longer lost. Some have
increased latency, what is expected behavior.
BEFORE:
64 bytes from 192.168.0.1: icmp_seq=29 ttl=64 time=3.61 ms
64 bytes from 192.168.0.1: icmp_seq=31 ttl=64 time=3.33 ms # Missed 30
64 bytes from 192.168.0.1: icmp_seq=32 ttl=64 time=4.03 ms
64 bytes from 192.168.0.1: icmp_seq=34 ttl=64 time=3.10 ms # Missed 33
64 bytes from 192.168.0.1: icmp_seq=36 ttl=64 time=3.04 ms # Missed 35
64 bytes from 192.168.0.1: icmp_seq=38 ttl=64 time=3.05 ms # Missed 37
64 bytes from 192.168.0.1: icmp_seq=40 ttl=64 time=4.14 ms # Missed 39
64 bytes from 192.168.0.1: icmp_seq=41 ttl=64 time=4.50 ms
AFTER:
64 bytes from 192.168.0.1: icmp_seq=29 ttl=64 time=3.40 ms
64 bytes from 192.168.0.1: icmp_seq=30 ttl=64 time=6.10 ms
64 bytes from 192.168.0.1: icmp_seq=31 ttl=64 time=4.69 ms
64 bytes from 192.168.0.1: icmp_seq=32 ttl=64 time=4.49 ms
64 bytes from 192.168.0.1: icmp_seq=33 ttl=64 time=148 ms
64 bytes from 192.168.0.1: icmp_seq=34 ttl=64 time=6.83 ms
64 bytes from 192.168.0.1: icmp_seq=35 ttl=64 time=8.66 ms
64 bytes from 192.168.0.1: icmp_seq=36 ttl=64 time=13.0 ms
64 bytes from 192.168.0.1: icmp_seq=37 ttl=64 time=4.78 ms
64 bytes from 192.168.0.1: icmp_seq=38 ttl=64 time=3.85 ms
Cc: <stable@vger.kernel.org>
Signed-off-by: Stanislaw Gruszka <stf_xl@wp.pl>
Link: https://patch.msgid.link/20260908124651.2995-1-stf_xl@wp.pl
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Without this ieee80211_select_key_8023() is never called and wifi never
connects.
Fixes: 0e80db0866 ("wifi: mac80211: fix key selection for encap offload frames")
Signed-off-by: Bert Karwatzki <spasswolf@web.de>
Link: https://patch.msgid.link/20260908094341.12196-1-spasswolf@web.de
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
ieee80211_debugfs_rename_netdev() renames sdata->vif.debugfs_dir
without sharing synchronization with ieee80211_debugfs_recreate_netdev().
When MLO link updates switch an interface to or from MLO, the recreate
path can remove and rebuild that dentry while a concurrent netdev rename
still uses the stale pointer.
Serialize the rename helper with the same wiphy mutex that protects the
recreate path so the dentry cannot be removed while debugfs_change_name()
is operating on it.
Reported-by: Vega <vega@nebusec.ai>
Signed-off-by: Zhiling Zou <zhilinz@nebusec.ai>
Link: https://patch.msgid.link/76eecaf74ca2d4e4aedf46c3ee7505917433468b.1785729184.git.zhilinz@nebusec.ai
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
When packet sockets use PACKET_QDISC_BYPASS, mac80211 can parse the
radiotap header while selecting the queue and again before transmit.
If the header includes an FCS, both parses trim the skb. Validate the FCS
on both parses, but trim it only before transmit. This keeps queue
selection read-only and avoids trimming shared skb data twice.
Fixes: cb17ed29a7 ("mac80211: parse radiotap header when selecting Tx queue")
Signed-off-by: Mariano Baragiola <mbaragiola@linux.com>
Link: https://patch.msgid.link/20260817233521.2747027-1-mbaragiola@linux.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
As per IEEE Std 802.11-2024, subclause 26.17.2.2, a 6 GHz AP can transmit
a Beacon frame using a non-HT duplicate PPDU, so that stations scanning
only PSC channels can discover it. Currently, there is no mechanism
to request the driver to transmit non-HT duplicate Beacon frames.
In [1], changes were made to parse the Beacon from user-space to set
the non-HT duplicate flag in the kernel. However, it was suggested that
non-HT transmission be part of the Beacon TX rate settings. Hence, add
NL80211_TXRATE_6GHZ_NON_HT_DUP as a flag attribute under
NL80211_ATTR_TX_RATES to let user-space request the driver to transmit
non-HT duplicate Beacons.
Also add validation to reject the Beacon if non-HT duplicate Beacon
transmission is enabled when legacy rates are not configured.
[1] https://lore.kernel.org/all/1644914581-21682-1-git-send-email-quic_ramess@quicinc.com/
Signed-off-by: Manish Dharanenthiran <manish.dharanenthiran@oss.qualcomm.com>
Link: https://patch.msgid.link/20260805-dup-beacon-v1-1-68574869c896@oss.qualcomm.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The two scnprintf arguments for fq_overlimit and fq_overmemory are in the
wrong order, which results in the aqm debugfs file showing the overmemory
count under the fq_overlimit label and vice versa.
Fixes: 2a4e675d88 ("mac80211: Export fq memory limit information in debugfs")
Reviewed-by: Toke Høiland-Jørgensen <toke@toke.dk>
Assisted-by: Claude:claude-opus-5
Signed-off-by: Julius Bairaktaris <julius@bairaktaris.de>
Link: https://patch.msgid.link/20260904090626.92727-1-julius@bairaktaris.de
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
check_fwfile_format() reads each block header and trusts its data length
before proving that both fit in the firmware image. A truncated header or
oversized block can therefore pass malformed extents to the download path.
Require the remaining image to contain the complete header and payload
before accepting each block.
Signed-off-by: Pengpeng Hou <pengpeng@iscas.ac.cn>
Link: https://patch.msgid.link/20260830140420.26399-1-pengpeng@iscas.ac.cn
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
check_fwfile_format() reads each block header and trusts its data length
before proving that both fit in the firmware image. A truncated header or
oversized block can therefore pass malformed extents to the download path.
Require the remaining image to contain the complete header and payload
before accepting each block.
Signed-off-by: Pengpeng Hou <pengpeng@iscas.ac.cn>
Link: https://patch.msgid.link/20260830124810.97223-1-pengpeng@iscas.ac.cn
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
