The complete profile bit together with the NSTR link pair
present bit indicate whether or not the NSTR bitmap is,
the NSTR bitmap size just indicates how big it is.
Fixes: 7b6f08771b ("wifi: ieee80211: Support validating ML station profile length")
Fixes: 5c1f97537b ("wifi: mac80211: store BSS param change count from assoc response")
Reported-by: Dan Carpenter <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
All the configurations for the various Bz/Gl devices
are basically identical, except for Gl A-step and the
firmware filename prefixes.
Add some infrastructure to auto-generate the firmware
filename prefix based on the detected MAC step and
RF name/step, and remove all the unneeded configs.
This reduces the size of the iwlwifi module by ~9k:
517582 27111 560 545253 851e5 drivers/net/wireless/intel/iwlwifi/iwlwifi.ko
526885 27083 560 554528 87620 drivers/net/wireless/intel/iwlwifi/iwlwifi.ko
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Gregory Greenman <[email protected]>
Link: https://lore.kernel.org/r/20230621130443.1dc121ba338f.I07d651516eb82cbaded4724ef30558a50f2fa866@changeid
Signed-off-by: Johannes Berg <[email protected]>
Roee reported various hard-to-debug crashes with pings in
EHT aggregation scenarios. Enabling KASAN showed that we
access the BAID allocation out of bounds, and looking at
the code a bit shows that since the reorder buffer entry
(struct iwl_mvm_reorder_buf_entry) is 128 bytes if debug
such as lockdep is enabled, then staring from an agg size
512 we overflow the size calculation, and allocate a much
smaller structure than we should, causing slab corruption
once we initialize this.
Fix this by simply using u32 instead of u16.
Reported-by: Roee Goldfiner <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Gregory Greenman <[email protected]>
Link: https://lore.kernel.org/r/20230620125813.f428c856030d.I2c2bb808e945adb71bc15f5b2bac2d8957ea90eb@changeid
Signed-off-by: Johannes Berg <[email protected]>
There were crashes reported in this code, and the timer_shutdown()
warning in one of the previous patches indicates that the timeout
timer for the AP response (addba_resp_timer) is still armed while
we're stopping the aggregation session.
After a very long deliberation of the code, so far the only way I
could find that might cause this would be the following sequence:
- session start requested
- session start indicated to driver, but driver returns
IEEE80211_AMPDU_TX_START_DELAY_ADDBA
- session stop requested, sets HT_AGG_STATE_WANT_STOP
- session stop worker runs ___ieee80211_stop_tx_ba_session(),
sets HT_AGG_STATE_STOPPING
From here on, the order doesn't matter exactly, but:
1. driver calls ieee80211_start_tx_ba_cb_irqsafe(),
setting HT_AGG_STATE_START_CB
2. driver calls ieee80211_stop_tx_ba_cb_irqsafe(),
setting HT_AGG_STATE_STOP_CB
3. the worker will run ieee80211_start_tx_ba_cb() for
HT_AGG_STATE_START_CB
4. the worker will run ieee80211_stop_tx_ba_cb() for
HT_AGG_STATE_STOP_CB
(the order could also be 1./3./2./4.)
This will cause ieee80211_start_tx_ba_cb() to send out the AddBA
request frame to the AP and arm the timer, but we're already in
the middle of stopping and so the ieee80211_stop_tx_ba_cb() will
no longer assume it needs to stop anything.
Prevent this by checking for WANT_STOP/STOPPING in the start CB,
and warn if we're sending a frame on a stopping session.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Gregory Greenman <[email protected]>
Link: https://lore.kernel.org/r/20230618214436.e5b52777462a.I0b2ed6658e81804279f5d7c9c1918cb1f6626bf2@changeid
Signed-off-by: Johannes Berg <[email protected]>
Since regulatory disconnect was added, OCB and NAN interface
types were added, which made it completely unusable for any
driver that allowed OCB/NAN. Add OCB/NAN (though NAN doesn't
do anything, we don't have any info) and also remove all the
logic that opts out, so it won't be broken again if/when new
interface types are added.
Fixes: 6e0bd6c35b ("cfg80211: 802.11p OCB mode handling")
Fixes: cb3b7d8765 ("cfg80211: add start / stop NAN commands")
Signed-off-by: Johannes Berg <[email protected]>
Link: https://lore.kernel.org/r/20230616222844.2794d1625a26.I8e78a3789a29e6149447b3139df724a6f1b46fc3@changeid
Signed-off-by: Johannes Berg <[email protected]>
There are some locking changes that will later otherwise
cause conflicts, so merge wireless into wireless-next to
avoid those.
Signed-off-by: Johannes Berg <[email protected]>
When we restart the firmware, we shouldn't keep old debug data
around. Since the "allocate" function might not reallocate the
memory blocks (they're only freed when we unbind from the device),
clear the memory to have a clean slate for debug data.
This is a bit more complex since we normally don't enter into
the allocation function, but duplicating the logic didn't seem
enticing either, so rework the allocation a bit to always go
into the individual block allocation, but there clear if it's
already allocated, rather than allocating again.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Gregory Greenman <[email protected]>
Link: https://lore.kernel.org/r/20230613155501.30b82881cfbf.I39520aff8ac95ee64d39dc5913525a1efd7995fa@changeid
Signed-off-by: Johannes Berg <[email protected]>
Given the existing code in iwl_dbg_tlv_update_drams(), the
following can happen and cause firmware asserts, and even
the device to become unusable:
* We set the magic so the firmware will use the data;
* we try to fill multiple allocation IDs, with at least
one successful, but - crucially - one failing and thus
not touching the data;
* we don't clear the data since there was one success.
This doesn't seem like much of a problem just yet, however,
what happens now is that the allocation ID(s) that failed
are not initialized.
There are two additional things to know:
* we never free these allocations across FW restart or
interface down/up etc., in fact we never free them until
the driver is unbound from the device (e.g. unloaded)
* the firmware uses the DRAM info structure for real debug
data when it has used it completely
Given that, and the fact that we never initialize the data
on restart, we can be unlucky and end up with an allocation
that looks for the most part valid (valid ID, valid number
of buffers, etc.) but has bad sizes - causing the firmware
to throw an assert we can never recover from.
Fixing the code to have the entire buffers cleared (which
we should do so old debug data isn't sticking around) is a
bit more complex, so as a first step make the actual code
that fills the information more robust by clearing the
structure first, and filling the magic values only if it
actually succeeded for one, rather than doing it the other
way around.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Gregory Greenman <[email protected]>
Link: https://lore.kernel.org/r/20230613155501.87cf5528f4bc.I26ac907a4162297808b33467fc7f5d8177474a34@changeid
Signed-off-by: Johannes Berg <[email protected]>
There's quite a bit of code accessing sband iftype data
(HE, HE 6 GHz, EHT) and we always need to remember to use
the ieee80211_vif_type_p2p() helper. Add new helpers to
directly get it from the sband/vif rather than having to
call ieee80211_vif_type_p2p().
Convert most code with the following spatch:
@@
expression vif, sband;
@@
-ieee80211_get_he_iftype_cap(sband, ieee80211_vif_type_p2p(vif))
+ieee80211_get_he_iftype_cap_vif(sband, vif)
@@
expression vif, sband;
@@
-ieee80211_get_eht_iftype_cap(sband, ieee80211_vif_type_p2p(vif))
+ieee80211_get_eht_iftype_cap_vif(sband, vif)
@@
expression vif, sband;
@@
-ieee80211_get_he_6ghz_capa(sband, ieee80211_vif_type_p2p(vif))
+ieee80211_get_he_6ghz_capa_vif(sband, vif)
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Gregory Greenman <[email protected]>
Link: https://lore.kernel.org/r/20230604120651.db099f49e764.Ie892966c49e22c7b7ee1073bc684f142debfdc84@changeid
Signed-off-by: Johannes Berg <[email protected]>
For W=2, we can enable more kernel-doc warnings,
such as missing return value descriptions etc.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Masahiro Yamada <[email protected]>
The kernel-doc script currently reports a number of issues
only in "verbose" mode, but that's initialized from V=1
(via KBUILD_VERBOSE), so if you use KDOC_WERROR=1 then
adding V=1 might actually break the build. This is rather
unexpected.
Change kernel-doc to not change its behaviour wrt. errors
(or warnings) when verbose mode is enabled, but rather add
separate warning flags (and -Wall) for it. Allow enabling
those flags via environment/make variables in the kernel's
build system for easier user use, but to not have to parse
them in the script itself.
Signed-off-by: Johannes Berg <[email protected]>
Acked-by: Jonathan Corbet <[email protected]>
Signed-off-by: Masahiro Yamada <[email protected]>
Move the beacon loss work that might cause a disconnect
and the CSA disconnect work to be wiphy work, so we hold
the wiphy lock for them.
Signed-off-by: Johannes Berg <[email protected]>
Channel switch obviously must be handled per link, and we
have a (potential) deadlock when canceling that work. Use
the new delayed wiphy work to handle this instead and get
rid of the explicit timer that way too.
Signed-off-by: Johannes Berg <[email protected]>
SMPS requests are per link, and currently there's a potential
deadlock with canceling. Use the new wiphy work to handle SMPS
instead, so that the cancel cannot deadlock.
Signed-off-by: Johannes Berg <[email protected]>
Since we want to have wiphy_lock() for the unregistration
in the future, unregister also netdevs via cfg80211 now
to be able to hold the wiphy_lock() for it.
Signed-off-by: Johannes Berg <[email protected]>
We'll need this later to convert other works that might
be cancelled from here, so convert this one first.
Signed-off-by: Johannes Berg <[email protected]>
Add a work abstraction at the cfg80211 level that will always
hold the wiphy_lock() for any work executed and therefore also
can be canceled safely (without waiting) while holding that.
This improves on what we do now as with the new wiphy works we
don't have to worry about locking while cancelling them safely.
Also, don't let such works run while the device is suspended,
since they'll likely need to interact with the device. Flush
them before suspend though.
Signed-off-by: Johannes Berg <[email protected]>
Sending the wiphy out might cause calls to the driver,
notably get_txq_stats() and get_antenna(). These aren't
very important, since the normally have their own locks
and/or just send out static data, but if the contract
should be that the wiphy lock is always held, these are
also affected. Fix that.
Signed-off-by: Johannes Berg <[email protected]>
Missed this ioctl since it's in wext-sme.c where we
usually get via a front-level ioctl handler in the
other files, but it should also hold the wiphy lock
to align the locking contract towards the driver.
Signed-off-by: Johannes Berg <[email protected]>
This is a driver callback, and the driver should be able
to assume that it's called with the wiphy lock held. Move
the call up so that's true, it has no other effect since
the device is already unregistering and we cannot reach
this function through other paths.
Signed-off-by: Johannes Berg <[email protected]>
Most code paths in cfg80211 already hold the wiphy lock,
mostly by virtue of being called from nl80211, so make
the pmsr cleanup worker also hold it, aligning the
locking promises between different parts of cfg80211.
Signed-off-by: Johannes Berg <[email protected]>
Most code paths in cfg80211 already hold the wiphy lock,
mostly by virtue of being called from nl80211, so make
the auto-disconnect worker also hold it, aligning the
locking promises between different parts of cfg80211.
Signed-off-by: Johannes Berg <[email protected]>
There are a number of upcoming things in both the stack and
drivers that would otherwise conflict, so merge wireless to
wireless-next to be able to avoid those conflicts.
Signed-off-by: Johannes Berg <[email protected]>
This reverts commit b70813e4a8 ("wifi: iwlwifi: update response
for mcc_update command") since it causes a merge conflict, and it
seems easier to redo the patch later.
Signed-off-by: Johannes Berg <[email protected]>
This reverts commit 1bcbb1208e ("wifi: iwlwifi: mvm: FTM
initiator MLO support") as it causes a merge conflict, and
we can defer and re-do those changes later.
Signed-off-by: Johannes Berg <[email protected]>
This should use wiphy_lock() now instead of requiring the
RTNL, since __cfg80211_leave() via cfg80211_leave() is now
requiring that lock to be held.
Fixes: a05829a722 ("cfg80211: avoid holding the RTNL when calling the driver")
Signed-off-by: Johannes Berg <[email protected]>
This should use wiphy_lock() now instead of acquiring the
RTNL, since cfg80211_stop_sched_scan_req() now needs that.
Fixes: a05829a722 ("cfg80211: avoid holding the RTNL when calling the driver")
Signed-off-by: Johannes Berg <[email protected]>
If we have a reconfig failure in the driver, then we need
to shut down the network interface(s) at the network stack
level through cfg80211, which can result in a lot of those
"Failed check-sdata-in-driver check, ..." warnings, since
interfaces are considered to not be in the driver when the
reconfiguration fails, but we still need to go through all
the shutdown flow.
Avoid many of these warnings by storing the fact that the
stack experienced a reconfiguration failure and not doing
the warning in that case.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Gregory Greenman <[email protected]>
Link: https://lore.kernel.org/r/20230604120651.3750c4ae6e76.I9e80d6026f59263c008a1a68f6cd6891ca0b93b0@changeid
Signed-off-by: Johannes Berg <[email protected]>
When we flush stations, we first take them off the list
and then destroy them one by one. If we do the different
mode recalculations while destroying them, we cause the
following scenario:
- STA 1 has 80 MHz - min chanctx width is now 80 MHz
- STA 2 has 80 MHz
- empty STA list
- destroy STA 2
- recalc min chanctx width -> results in 20 MHz as
the STA list is already empty
This is broken, since as far as the driver is concerned
STA 1 still exists at this point, and this causes issues
at least with iwlwifi.
Fix - and also optimize - this by doing the recalc of
min chanctx width (and also P2P PS) only after all the
stations were removed.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Gregory Greenman <[email protected]>
Link: https://lore.kernel.org/r/20230604120651.48d262b6b42d.Ia15532657c17535c28ec0c5df263b65f0f80663c@changeid
Signed-off-by: Johannes Berg <[email protected]>
If the station disables all links it's in powersave and
we shouldn't transmit anything to it, but we don't handle
that correctly yet. For now, just avoid the warning, once
we really add support for this case we can revert to the
old warning.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Gregory Greenman <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Johannes Berg <[email protected]>
Johannes Berg