In platform_device_register_full() the err_alloc label is
misleading, we only ever jump to it if the pdev is NULL,
but it then proceeds to free it, which is a no-op.
Remove the label and simply exit the function immediately.
Signed-off-by: Johannes Berg <[email protected]>
Reviewed-by: Rafael J. Wysocki <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>
Merge net-next to resolve a conflict and to get the mac80211
rhashtable fixes so further patches can be applied on top.
Signed-off-by: Johannes Berg <[email protected]>
Sometimes, we may want to transport higher bandwidth data
through vendor events, and in that case sending it multicast
is a bad idea. Allow vendor events to be unicast.
Signed-off-by: Johannes Berg <[email protected]>
For non-offloaded rate control, implement VHT extended NSS
support by just ignoring 160 MHz on transmit if the peer
doesn't support the full NSS (2). This is pretty unlikely
and gets us 160 MHz RX as well as TX in most cases, since
the typical case for this workaround is 4x4 AP only doing
2x2 on 160 MHz.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
There's no point in this, we already do everything in a nested
fashion, and if we didn't we'd already crash in iwl_mvm_leds_exit()
etc. Just remove the bit.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
We support this, so we should advertise it. In fact, if we don't,
mac80211 will do it for us (as we advertise SUPPORTS_VHT_EXT_NSS_BW
to it), but that requires a memory reallocation which is wasteful:
ieee80211 phy0: copying sband (band 1) due to VHT EXT NSS BW flag
Set the flag here to avoid that.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
The IWL_MVM_INIT_STATUS_REG_HW_INIT_COMPLETE bit shouldn't be set,
and realistically we should complete all setup before we call the
ieee80211_register_hw() function. Fix this.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
In HE-TB PPDUs (labeled HE-TRIG in radiotap), we were overwriting
the data4.spatial_reuse_1 field with the spatial reuse data that
the firmware gives us for SU/MU PPDUs. Fix that by moving that,
we are already setting the data4.spatial_reuse_{1,2,3,4} fields
in the TB PPDU case.
Signed-off-by: Johannes Berg <[email protected]>
Fixes: 69f3ca8ed3 ("iwlwifi: mvm: show more HE radiotap data for TB PPDUs")
Signed-off-by: Luca Coelho <[email protected]>
We now no longer have any special code in
iwl_mvm_pass_packet_to_mac80211(), so don't
need to pass NO_PSDU packets through it.
Stop doing so and clean up the code there.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Add support for FTM initiator, i.e. peer measurements with FTM
if the firmware supports FTM.
Additionally, add two defines we depend on in
include/linux/ieee80211.h.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Avraham Stern <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Add support for FTM responder for hardware/firmware combinations
that advertise support for it.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Avraham Stern <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Fix cfg80211_gen_new_bssid() to not rely on u64 modulo arithmetic,
which isn't needed since we really just want to mask there. Also,
clean it up to calculate the mask only once and use GENMASK_ULL()
instead of open-coding the mask calculation.
Signed-off-by: Johannes Berg <[email protected]>
It's possible that the caller of cfg80211_classify8021d() uses the
value to index an array, like mac80211 in ieee80211_downgrade_queue().
Prevent speculation on the return value.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
Without recording the netlink port ID, we cannot return the
results or complete messages to userspace, nor will we be
able to abort if the socket is closed, so clearly we need
to fill the value.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
If a driver does any significant activity in its ibss_join method,
then it will very well expect that to be called during restart,
before any stations are added. Do that.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
Jouni reports that in some cases it is possible that getting
disconnected (or stopping AP, after previous patches) results
in further operations hitting the window within the regulatory
core restoring the regdomain to the defaults. The reason for
this is that we have to call out to CRDA or otherwise do some
asynchronous work, and thus can't do the restore atomically.
However, we've previously seen all the data we need to do the
restore, so we can hang on to that data and use it later for
the restore. This makes the whole thing happen within a single
locked section and thus atomic.
However, we can't *always* do this - there are unfortunately
cases where the restore needs to re-request, because this is
also used (abused?) as an error recovery process, so make the
new behaviour optional and only use it when doing a regular
restore as described above.
Reported-by: Jouni Malinen <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
This finally merges the multi-BSSID code. This is the result of a
long collaboration between the team at Qualcomm (Peng and Jouni)
and our team at Intel (mostly Sara).
Signed-off-by: Johannes Berg <[email protected]>
This makes for much simpler code, simply walk through all
the elements and check that the last one found ends with
the end of the data. This works because if any element is
malformed the walk is aborted, we end up with a mismatch.
Signed-off-by: Johannes Berg <[email protected]>
We currently have a number of helpers to find elements that just
return a u8 *, change those to return a struct element and add
inlines to deal with the u8 * compatibility.
Note that the match behaviour is changed to start the natch at
the data, so conversion from _ie_match to _elem_match need to
be done carefully.
Signed-off-by: Johannes Berg <[email protected]>
Use the new for_each_element() helper here, we cannot use
for_each_subelement() since we have a fixed 1 byte before
the subelements start.
While at it, also fix le16_to_cpup() to be get_unaligned_le16()
since we don't know anything about alignment.
Signed-off-by: Johannes Berg <[email protected]>
Rather than always iterating elements from frames with pure
u8 pointers, add a type "struct element" that encapsulates
the id/datalen/data format of them.
Then, add the element iteration macros
* for_each_element
* for_each_element_id
* for_each_element_extid
which take, as their first 'argument', such a structure and
iterate through a given u8 array interpreting it as elements.
While at it and since we'll need it, also add
* for_each_subelement
* for_each_subelement_id
* for_each_subelement_extid
which instead of taking data/length just take an outer element
and use its data/datalen.
Also add for_each_element_completed() to determine if any of
the loops above completed, i.e. it was able to parse all of
the elements successfully and no data remained.
Use for_each_element_id() in cfg80211_find_ie_match() as the
first user of this.
Signed-off-by: Johannes Berg <[email protected]>
When we destroy the interface we already hold the wdev->mtx
while calling cfg80211_pmsr_wdev_down(), which assumes this
isn't true and flushes the worker that takes the lock, thus
leading to a deadlock.
Fix this by refactoring the worker and calling its code in
cfg80211_pmsr_wdev_down() directly.
We still need to flush the work later to make sure it's not
still running and will crash, but it will not do anything.
Fixes: 9bb7e0f24e ("cfg80211: add peer measurement with FTM initiator API")
Signed-off-by: Johannes Berg <[email protected]>
When we *don't* have a MAC address attribute, we shouldn't
try to use this - this was intended to copy the local MAC
address instead, so fix it.
Fixes: 9bb7e0f24e ("cfg80211: add peer measurement with FTM initiator API")
Signed-off-by: Johannes Berg <[email protected]>
There's a small hardware bug in 22260 devices which thus require a
few more delays during initialization. Implement this workaround.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
If we have >=10 (logical) CPUs, our command size exceeds the
internal buffer size and the command fails; fix that by using
IWL_HCMD_DFL_NOCOPY for the command that's allocated anyway.
While at it, also fix the leak of cmd, and use struct_size()
to calculate its size.
Signed-off-by: Johannes Berg <[email protected]>
Fixes: 8edbfaa198 ("iwlwifi: mvm: configure multi RX queue")
Signed-off-by: Luca Coelho <[email protected]>
The typical sequence of setting INIT_DONE and then waiting
for clock stabilisation is going to need a new workarounds,
so first of all refactor it.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Hiding the HE radiotap data for further processing of the SKB just
caused another bug when adding the L-SIG data. Simply stop doing
this and adjust the skb->data pointer accordingly when we need to
get the 802.11 header.
While at it, also verify and fix the data alignment, we need to add
2 bytes padding with the vendor data to ensure the whole length of
all radiotap headers is a multiple of 4.
Signed-off-by: Johannes Berg <[email protected]>
Fixes: 6721039d5b ("iwlwifi: mvm: add L-SIG length to radiotap")
Signed-off-by: Luca Coelho <[email protected]>
Even if they're static const, there's no need to duplicate
the structs every time they're included and used. Move them
to an appropriate C file instead.
Also remove useless parentheses along the way.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Expose the trigger-based PPDU SIG-A bandwidth to radiotap in
the newly defined bits thereof.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Use %u instead of using %d which looks signed but then won't
get signed output if using an unsigned variable. It doesn't
matter much, but be consistent.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
In iwl_mvm_sta_alloc_queue_tvqm(), we know that we have a
station, so no need to check it.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Fix the missing and malformed documentation that kernel-doc and
sphinx warn about. While at it, also add some things to the docs
to fix missing links.
Sadly, the only way I could find to fix this was to add some
trailing whitespace.
Signed-off-by: Johannes Berg <[email protected]>
Since we now prevent regulatory restore during STA disconnect
if concurrent AP interfaces are active, we need to reschedule
this check when the AP state changes. This fixes never doing
a restore when an AP is the last interface to stop. Or to put
it another way: we need to re-check after anything we check
here changes.
Cc: [email protected]
Fixes: 113f3aaa81 ("cfg80211: Prevent regulatory restore during STA disconnect in concurrent interfaces")
Signed-off-by: Johannes Berg <[email protected]>
In typical cases, there's no need to pass both the maxattr
and the policy array pointer, as the maxattr should just be
ARRAY_SIZE(policy) - 1. Therefore, to be less error prone,
just remove the maxattr argument from the default macros
and deduce the size accordingly.
Leave the original macros with a leading underscore to use
here and in case somebody needs to pass a policy pointer
where the policy isn't declared in the same place and thus
ARRAY_SIZE() cannot be used.
Signed-off-by: Johannes Berg <[email protected]>
Merge net-next so that we get the changes from net, which would
otherwise conflict with the NLA_POLICY_NESTED/_ARRAY changes.
Signed-off-by: Johannes Berg <[email protected]>
Same story as before, these use struct ifreq and thus need
to be read with the shorter version to not cause faults.
Cc: [email protected]
Fixes: f92d4fc953 ("kill bond_ioctl()")
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This reverts commit 1cebf8f143 ("socket: fix struct ifreq
size in compat ioctl"), it's a bugfix for another commit that
I'll revert next.
This is not a 'perfect' revert, I'm keeping some coding style
intact rather than revert to the state with indentation errors.
Cc: [email protected]
Fixes: 1cebf8f143 ("socket: fix struct ifreq size in compat ioctl")
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
In order to make more sense out of the captured radiotap data e.g.
when the configured AID changes, add the currently configured AID
to the radiotap data as a vendor extension field.
This is made race-free by updating the included value from inside
the RX path (using a notification wait) for the command response
from the firmware, which thus means it's serialized with frame RX.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
The reorder buffer is bypassed because the firmware won't have
any BA sessions, document this.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Implement CSI (channel estimation matrix) reporting in the mvm
driver, if the firmware has the capability.
Currently only a debugfs API is provided as the API is still
under discussion.
For now, RX aggregation must be disabled to use this feature
on data frames as we haven't found a good way to attach the
data to A-MPDUs, given complexities with multi-queue.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Add the location/time-of-flight/FTM APIs that we'll use in
follow-up patches to implement FTM responder and initiator.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
That's what we pass, and we don't want/need any negative values.
Found by sparse/smatch.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
If we're in the else branch of checking "tt->state == IWL_TI_CT_KILL"
so there's no point in checking "tt->state != IWL_TI_CT_KILL" again.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
This is an ancient (~2015) implementation that no longer matches
the firmware in any way, and most likely never worked. Remove all
of it so it can be reintroduced properly.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
The current code assigns the reference, and then goes to increment
it if the toggle bit has changed. That way, we get
Toggle 0 0 0 0 1 1 1 1
ID 1 1 1 1 1 2 2 2
Fix that by assigning the post-toggle ID to get
Toggle 0 0 0 0 1 1 1 1
ID 1 1 1 1 2 2 2 2
Reported-by: Danny Alexander <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
Fixes: fbe4112791 ("iwlwifi: mvm: update mpdu metadata API")
Signed-off-by: Luca Coelho <[email protected]>
These inlines just check the module parameters, so they don't
need a configuration parameter and can move to a better place.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Clean up the LDBG config command to not be called "continuous
recording", and while at it actually remove the continuous
recording implementation completely since it was only used for
store & forward architectures.
This also fixes a bug at least in iwl_fw_dbg_buffer_allocation()
because what's now "__le32 type" (matching the firmware) used to
be "__le16 enable_recording", so the buffer allocation config
sub-struct would erroneously have started at the wrong offset.
In the other cases this didn't actually lead to a bug as other
bytes in pad[] were all zeroes, so accessing the 16-bit value as
a 32-bit value wouldn't make a difference (in little endian.)
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
These files have a long history of code changes, but analysing
the remaining code leads to having only a few changes that are
not already owned by Intel, notably from
- Andy Lutomirski <[email protected]>
- Joonwoo Park <[email protected]>
- Kirtika Ruchandani <[email protected]>
- Rajat Jain <[email protected]>
- Stanislaw Gruszka <[email protected]>
remaining in the code today.
Note that
- I myself was working for Intel and for any possibly code
that might be before my employment there give permission
- Wizery employees were working for Intel
More specifically, we identified the following commits that
(partially may) remain today:
25c03d8e8c Joonwoo Park <[email protected]> ("iwlwifi: do not schedule tasklet when rcv unused irq")
f36d04abe6 Stanislaw Gruszka <[email protected]> ("iwlwifi: use dma_alloc_coherent")
387f3381f7 Stanislaw Gruszka <[email protected]> ("iwlwifi: fix dma mappings and skbs leak")
2624e96ce1 Stanislaw Gruszka <[email protected]> ("iwlwifi: fix possible data overwrite in hcmd callback")
bfe4b80e9f Stanislaw Gruszka <[email protected]> ("iwlwifi: always check if got h/w access before write")
d536c32b45 Andy Lutomirski <[email protected]> ("iwlwifi: pcie: log when waking the NIC for hcmd submission fails")
a6d24fad00 Rajat Jain <[email protected]> ("iwlwifi: pcie: dump registers when HW becomes inaccessible")
fb12777ab5 Kirtika Ruchandani <[email protected]> ("iwlwifi: Add more call-sites for pcie reg dumper")
3a73a30049 Stanislaw Gruszka <[email protected]> ("iwlwifi: cleanup/fix memory barriers")
aa5affbacb Stanislaw Gruszka <[email protected]> ("iwlwifi: dump stack when fail to gain access to the device")
Align the licenses with their permission to clean up and to
make it all identical.
CC: Joonwoo Park <[email protected]>
CC: Stanislaw Gruszka <[email protected]>
CC: Andy Lutomirski <[email protected]>
CC: Rajat Jain <[email protected]>
CC: Kirtika Ruchandani <[email protected]>
Acked-by: Johannes Berg <[email protected]>
Acked-by: Kirtika Ruchandani <[email protected]>
Acked-by: Stanislaw Gruszka <[email protected]>
Acked-by: Joonwoo Park <[email protected]>
Acked-by: Rajat Jain <[email protected]>
Acked-by: Andy Lutomirski <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Due to a general shortage of RX API bits, the firmware is going
to reuse this bit on non-CCK frames to mean something else. Use
it only on CCK frames to prepare for that change.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
With just %16, it means 16 characters padding, but we really
don't want to print "0x 1F4547B", but instead want to
have this filled with zeroes, so we need the 0.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Clearly, I missed this when trying out the previously
merged patches. Remove the spurious variable now.
Signed-off-by: Johannes Berg <[email protected]>
Digging through the ioctls with Al because of the previous
patches, we found that on 64-bit decnet's dn_dev_ioctl()
is wrong, because struct ifreq::ifr_ifru is actually 24
bytes (not 16 as expected from struct sockaddr) due to the
ifru_map and ifru_settings members.
Clearly, decnet expects the ioctl to be called with a struct
like
struct ifreq_dn {
char ifr_name[IFNAMSIZ];
struct sockaddr_dn ifr_addr;
};
since it does
struct ifreq *ifr = ...;
struct sockaddr_dn *sdn = (struct sockaddr_dn *)&ifr->ifr_addr;
This means that DN_IFREQ_SIZE is too big for what it wants on
64-bit, as it is
sizeof(struct ifreq) - sizeof(struct sockaddr) +
sizeof(struct sockaddr_dn)
This assumes that sizeof(struct sockaddr) is the size of ifr_ifru
but that isn't true.
Fix this to use offsetof(struct ifreq, ifr_ifru).
This indeed doesn't really matter much - the result is that we
copy in/out 8 bytes more than we should on 64-bit platforms. In
case the "struct ifreq_dn" lands just on the end of a page though
it might lead to faults.
As far as I can tell, it has been like this forever, so it seems
very likely that nobody cares.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
These are unused by both firmware and driver, remove them.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
On any failure, including if we crash the firmware or get garbage
data, we currently ignore this and pretend the OTP was empty.
Clearly, this isn't typically the case.
In cases other than the firmware saying it can't read the requested
section, or the section having ended, make the access actually fail
and trickle the error up through the layers to fail init.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
syzbot reported an out-of-bounds read when passing certain
malformed messages into nl80211. The specific place where
this happened isn't interesting, the problem is that nested
policy parsing was referring to the wrong maximum attribute
and thus the policy wasn't long enough.
Fix this by referring to the correct attribute. Since this
is really not necessary, I'll come up with a separate patch
to just pass the policy instead of both, in the common case
we can infer the maxattr from the size of the policy array.
Reported-by: [email protected]
Cc: [email protected]
Fixes: 9bb7e0f24e ("cfg80211: add peer measurement with FTM initiator API")
Signed-off-by: Johannes Berg <[email protected]>
Under certain circumstances, radios created via netlink could
have n_limits be zero and no possible interface types, which
makes no sense. Reject this early to prevent a WARN_ON() in
cfg80211.
Fixes: 99e3a44bac ("mac80211_hwsim: allow setting iftype support")
Reported-by: [email protected]
Signed-off-by: Johannes Berg <[email protected]>
Fix two bugs in ieee80211_get_vht_max_nss():
* the spec says we should round down
(reported by Nissim)
* there's a double condition, the first one is wrong,
supp_width == 0 / ext_nss_bw == 2 is valid in 80+80
(found by smatch)
Fixes: b0aa75f0b1 ("ieee80211: add new VHT capability fields/parsing")
Reported-by: Nissim Bendanan <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
If validate_pae_over_nl80211() were to fail in nl80211_crypto_settings(),
we might leak the 'connkeys' allocation. Fix this.
Fixes: 64bf3d4bc2 ("nl80211: Add CONTROL_PORT_OVER_NL80211 attribute")
Signed-off-by: Johannes Berg <[email protected]>
Due to the alignment handling, it actually matters where in the code
we add the 4 bytes for the presence bitmap to the length; the first
field is the timestamp with 8 byte alignment so we need to add the
space for the extra vendor namespace presence bitmap *before* we do
any alignment for the fields.
Move the presence bitmap length accounting to the right place to fix
the alignment for the data properly.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
In the iwlwifi conversion, we sometimes call this from outside
of the wake_tx_queue() method, and in those cases must be in an
RCU critical section. Document this requirement.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
At the place where this code lives now, the skb can never be
NULL, so we can remove the pointless NULL check.
It seems to exist because this code was moved around a few times
and originally came from a place where it could in fact be NULL.
Signed-off-by: Johannes Berg <[email protected]>
This isn't really a problem now, but it means that the function
has a few NULL checks that are only relevant when coming from
the initial interface added in mac80211, and that's confusing.
Just pass non-NULL (but equivalently empty) in that case and
remove all the NULL checks.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
The older code and current userspace assumed that this data
is the content of the Measurement Report element, starting
with the Measurement Token. Clarify this in the documentation.
Signed-off-by: Johannes Berg <[email protected]>
The pointer and corresponding length is always set in pairs
in cfg80211, so no need to have this strange defensive check
that also confuses static checkers. Clean it up.
Signed-off-by: Johannes Berg <[email protected]>
All the queue management code runs under mvm->mutex, so there are
only very few cases of accessing the data structures without it:
* TX path, which doesn't take any locks anyway
* iwl_mvm_wake_sw_queue() and iwl_mvm_stop_sw_queue() where we
just (atomically) read a bitmap, so the lock isn't needed.
Therefore, we can remove the spinlock. This enables some cleanup
in the ugly locking in iwl_mvm_inactivity_check().
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
When we mark a TID as no longer having a queue, there's no
guarantee the TX path isn't using this txq_id right now,
having accessed it just before we reset the value. To fix
this, add synchronize_net() when we change the TIDs from
having a queue to not having one, so that we can then be
sure that the TX path is no longer accessing that queue.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
In some cases, like in the rsi driver hardware scan offload, there
may be scenarios in which hardware scan might not be available or
desirable.
Allow drivers to cope with this by letting them fall back to software
scan by returning the special value 1 from the hardware scan method.
Requested-by: Sushant Kumar Mishra <[email protected]>
Requested-by: Siva Rebbagondla <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
There's nothing much for mac80211 to do, so only pass through
the requests with minimal checks and tracing. The driver must
call cfg80211's results APIs.
Signed-off-by: Johannes Berg <[email protected]>
Add a new "peer measurement" API, that can be used to measure
certain things related to a peer. Right now, only implement
FTM (flight time measurement) over it, but the idea is that
it'll be extensible to also support measuring the necessary
things to calculate e.g. angle-of-arrival for WiGig.
The API is structured to have a generic list of peers and
channels to measure with/on, and then for each of those a
set of measurements (again, only FTM right now) to perform.
Results are sent to the requesting socket, including a final
complete message.
Closing the controlling netlink socket will abort a running
measurement.
v3:
- add a bit to report "final" for partial results
- remove list keeping etc. and just unicast out the results
to the requester (big code reduction ...)
- also send complete message unicast, and as a result
remove the multicast group
- separate out struct cfg80211_pmsr_ftm_request_peer
from struct cfg80211_pmsr_request_peer
- document timeout == 0 if no timeout
- disallow setting timeout nl80211 attribute to 0,
must not include attribute for no timeout
- make MAC address randomization optional
- change num bursts exponent default to 0 (1 burst, rather
rather than the old default of 15==don't care)
v4:
- clarify NL80211_ATTR_TIMEOUT documentation
v5:
- remove unnecessary nl80211 multicast/family changes
- remove partial results bit/flag, final is sufficient
- add max_bursts_exponent, max_ftms_per_burst to capability
- rename "frames per burst" -> "FTMs per burst"
v6:
- rename cfg80211_pmsr_free_wdev() to cfg80211_pmsr_wdev_down()
and call it in leave, so the device can't go down with any
pending measurements
v7:
- wording fixes (Lior)
- fix ftm.max_bursts_exponent to allow having the limit of 0 (Lior)
v8:
- copyright statements
- minor coding style fixes
- fix error path leak
Signed-off-by: Johannes Berg <[email protected]>
Add a helper function nl_set_extack_cookie_u64() to use a u64 as
the netlink extended ACK cookie, to avoid having to open-code it
in any users of the cookie.
A u64 should be sufficient for most subsystems though we allow
for up to 20 bytes right now. This also matches the cookies in
nl80211 where I intend to use this.
Signed-off-by: Johannes Berg <[email protected]>
Acked-by: David S. Miller <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
We have a bool and an __le16 called qos, rename the inner __le16
to 'qoshdr' to make it more obvious and to avoid sparse warnings.
Signed-off-by: Johannes Berg <[email protected]>
We have a macro here that uses an inner variable 'i' that
also exists in the outer scope - use '_i' in the macro.
Signed-off-by: Johannes Berg <[email protected]>
We have a pointer called 'tidstats' that shadows a bool function
argument with the same name, but we actually only use it once so
just remove the pointer.
Signed-off-by: Johannes Berg <[email protected]>
This variable shadows something that gets generated inside
the tracing macros, which causes sparse to warn. Avoid it
so sparse output is more readable, even if it doesn't seem
to cause any trouble.
Signed-off-by: Johannes Berg <[email protected]>
This variable shadows something that gets generated inside
the tracing macros, which causes sparse to warn. Avoid it
so sparse output is more readable, even if it doesn't seem
to cause any trouble.
Signed-off-by: Johannes Berg <[email protected]>
If the FTM responder settings are changed simultaneously with
the CSA beacon, the buffer size allocated isn't sufficient and
we'll have a heap overrun. Fix this.
While at it, also clean up the ftm_responder assignment, doing
it only if ftm_responder is non-zero is valid as it's 0 to start
with, but not really useful to understand the code.
Fixes: bc847970f4 ("mac80211: support FTM responder configuration/statistics")
Signed-off-by: Johannes Berg <[email protected]>
When FTM is enabled, doing a CSA will unexpectedly lose it since
the value of ftm_responder may be initialized to 0 instead of -1,
so fix that.
Fixes: 81e54d08d9 ("cfg80211: support FTM responder configuration/statistics")
Signed-off-by: Johannes Berg <[email protected]>
We already have BUILD_BUG_ON_ZERO() which I just hadn't found
before, so we should use it here instead of open-coding another
implementation thereof.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
I'm not really sure exactly _why_ I've been carrying a note
for what's probably _years_ to check that we don't do this,
but we clearly do reflect frames back to the station itself
if it sends such.
One way or the other, it's useless since the station doesn't
really need the AP to talk to itself, so suppress it.
While at it, clarify some of the logic by removing skb->data
references in favour of the destination address (pointer) we
already have separately.
Signed-off-by: Johannes Berg <[email protected]>
Instead of open-coding a lot of calls to is_valid_ie_attr(),
add this validation directly to the policy, now that we can.
Signed-off-by: Johannes Berg <[email protected]>
Many range checks can be done in the policy, move them
there. A few in mesh are added in the code (taken out of
the macros) because they don't fit into the s16 range in
the policy validation.
Signed-off-by: Johannes Berg <[email protected]>
We don't really need this state: instead of having an inactive
state where we can awaken zombie queues again if needed, just
keep them in their normal state unless a new queue is actually
needed and there's no other way of getting one.
We do this here by making the inactivity check not free queues
unless instructed that we now really need to allocate one to a
specific station, and in that case it'll just free the queue
immediately, without doing any inactivity step inbetween.
The only downside is a little bit more processing in this case,
but the code complexity is lower.
Additionally, this fixes a corner case: due to the way the code
worked, we could only ever reuse an inactive queue if it was
the reserved queue for a station, as iwl_mvm_find_free_queue()
would never consider returning an inactive queue.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
We want to call iwl_mvm_inactivity_check() from here in the
next patch, so need to move the code down to be able to.
Fix a minor checkpatch complaint while at it.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
This function is only used in the file where it's declared,
so just make it static.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Instead of iterating all the queues after having potentially
changed some queue configurations, rechecking if that was done,
mark the ones that do need a TID change explicitly in a bitmap
and use that to send the change to the firmware.
While at it, also rename iwl_mvm_change_queue_owner() to
iwl_mvm_change_queue_tid() since that's more obvious - the
"kind" of owner isn't immediately clear right now.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
We set the queue to this state, only to pretty much immediately
move it out of it again. However, we can't even hit any of the
code that checks if the queue is reconfiguring, because all of
this happens under mvm->mutex and we hold the all the way from
marking the queue as RECONFIGURING to marking it as READY again.
Additionally, the queue that became RECONFIGURING would've been
in SHARED state before, and it can safely stay in that state. In
case of errors, it previously would have stayed in RECONFIGURING
which it could never have left again.
Remove the state entirely and just track the queues that need to
be reconfigured in a separate, local, bitmap.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
We currently reconfigure the queues after the inactivity check,
but only in one of the two callers. This might leave queues in
a state where the TID owner is wrong, if called when reserving
a queue for a new station.
Clean this up and do the reconfiguration inside the inactivity
check function. This requires changing the locking, but one of
the two places already holds the mvm mutex and the other easily
can.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
If TVQM is used we skip over this, move the code into a new
function to get rid of the label.
Signed-off-by: Johannes Berg <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
Johannes Berg