In case we need to move the scheduler write pointer by
steps of 0x40, 0x80 or 0xc0, the scheduler gets stuck.
This leads to hardware error interrupts with status:
0x5A5A5A5A or alike.

In order to work around this, detect in the transport
layer that we are going to hit this case and tell iwlmvm
to increment the sequence number of the packets. This
allows to keep the requirement that the WiFi sequence
number is in sync with the index in the scheduler Tx queue
and it also allows to avoid the problematic sequence.
This means that from time to time, we will start a queue
from ssn + 1, but that shouldn't be a problem since we
don't switch to new queues for AMPDU now that we have
DQA which allows to keep the same queue while toggling
the AMPDU state.

This bug has been fixed on 9000 devices and up.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

A TID may not have traffic but still have a BA agreement
active (or being setup / torn down) since a BA agreement
can be triggered by a debugfs hook.
Just avoid to consider such a TID as inactive to make the
logic safer.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

If the hardware is stuck, we can't read any of the memory we need to
dump it, so we end up printing only 0xa5a5a5a5, which is useless.

To solve this, poke the hardware by triggering a reset and re-enabling
the clocks if we detect a HW error.
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

When we started using threaded irqs, all the opmode calls were changed
to be called with local_bh disabled.  The reason for this was it was
that mac80211 needs that.  When we are handling FW errors, mac80211 is
not involved, so we don't need it.
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

In BSS mode in the disconnection flow, mac80211 removes
the AP station before the vif is set to unassociated.
Our firmware wants it the other way around: first set
the vif as unassociated, and then remove the AP station.

In order to bridge between those two different behaviors,
iwlmvm doesn't remove the station from the firmware when
mac80211 removes it, but only after the vif is set to
unassociated. The implementation is in
iwl_mvm_bss_info_changed_station:

if (assoc state was modified && mvmvif->ap_sta_id is VALID
    && assoc state is now UNASSC)
	remove_the_station_from_the_firmware()

During the recovery flow, mac80211 re-adds the AP station
and then reconfigures the vif. Since the vif is not
associated, and then, we enter the if above (which was
intended to be taken in the disconnection flow only) and
remove the station we just added. This defeats the
recovery flow.

Fix this by not removing the AP station in this flow if
we are in recovery flow.

Cc: stable@vger.kernel.org
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

We allocate nvm_data in iwl_mvm_nvm_get_from_fw().  If something goes
wrong after the allocation (i.e. if no valid MAC address is valid), we
should free nvm_data before returning an error.
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

This function just calls kfree(), so it only obscures the code without
bringing any benefits.  Remove it.
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

Properly document the transmit buffer bits using an enum and
kernel-doc documentation.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

Add kernel-doc documentation for enum iwl_debug_cmds, linking the
structures used by the commands.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

This improves documentation, since kernel-doc can't deal with the
union well.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

To stop and start the FW monitor in the 7000 HW
family we need to use a different bit, otherwise
after stopping it for the first time - it won't
get restarted. Use the correct bitmask.

Note: This fix is only for DRAM collection mode.
      For other modes, an additional fix will be
      needed.
Signed-off-by: Liad Kaufman <liad.kaufman@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

Fix the enum link by adding the missing & and provide the link
to the TX response documentation.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

When the command name is printed on command completion, the wrong
group is used, leading to the wrong name being printed. Fix this
by using the group ID without inappropriately mangling it through
iwl_cmd_groupid() - it's already a u8. Also, while at it, use it
from the same place as the command ID, everything else is just
confusing.

Fixes: ab02165c ("iwlwifi: add wide firmware command infrastructure for TX")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

When sending non-linear SKBs that should be included in the regular
TX tracing completely (and not be pushed into the tx_data tracing),
the (tracing) code didn't correctly take the fact that they were
non-linear into account and added only the skb head portion.

This probably never really triggered, since those frames we want
traced fully are most likely linear anyway, but the code gets easier
to understand and we lose an argument to the tracing function, so
overall fixing this is better.

Fixes: 206eea78 ("iwlwifi: pcie: support frag SKBs")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

There's no need to calculate the data_len outside of the tracepoint,
since it's always skb->len - hdr_len, which are both available inside.
Simplify the callers and move the calculation in.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

Show the name of the member (scanned_channels) that provides the
length with some better markup.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

Use kstrtou32_from_user() in debugfs instead of open-coding it.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

On A000 HW, the SCD rdptr has only 8 bits allocated
for it, thus when checking if a queue is full, or
when checking if the SSN is equal to the TID's
next_reclaimed, A000 HW should trim the SSN.

Fix this by "normalizing" the SSN to wrap around
0xFF when comparing to the next_reclaimed on A000
HW.
Signed-off-by: Liad Kaufman <liad.kaufman@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

We don't actually care about the value at all, just making sure
that we can successfully parse a single integer value, but that's
entirely pointless - remove it.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

When toggling the RF-kill pin quickly in succession, the driver can
get rather confused because it might be in the process of shutting
down, expecting all commands to go through quickly due to rfkill,
but the transport already thinks the device is accessible again,
even though it previously shut it down. This leads to bugs, and I
even observed a kernel panic.

Avoid this by making the PCIe code only report that the radio is
enabled again after the higher layers actually decided to shut it
off.

This also pulls out this common RF-kill checking code into a common
function called by both transport generations and also moves it to
the direct method - in the internal helper we don't really care
about the RF-kill status anymore since we won't report it up until
the stop anyway.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

If we happen to be in or get into the queue sync when RF-kill
is asserted, we return from there and warn since there are
still queue sync notifications outstanding. These can't ever
come though, because we're in RF-kill, so don't WARN then.

While at it, also move the warning to the appropriate place,
if the request is not synchronous then we shouldn't warn, but
currently always will.

To make it fast, also trigger the waitq when on rfkill assert.

Fixes: 0636b938 ("iwlwifi: mvm: implement driver RX queues sync command")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

In order to debug "hardware" RF-kill flows, add a low-level hook to
allow changing the "hardware" RF-kill from debugfs.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

There's no point in duplicating exactly the same code here
for legacy and MSI-X interrupts, so pull it out into a new
function to call in both places.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

In case that rate's antenna is wrong at the init stage, it's
very hard to say what went wrong. Add debug data to the already
existing WARN_ON_ONCE.
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

Some static checkers (e.g. smatch) complain about the logic, saying that
resp_cp might be leaked. Clearly that isn't true, but making the logic
easier to follow does not result in any significant code changes and makes
the code more readable by moving the NULL check closer to its source.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

The firmware moved the development from a0 MAC to z0.
z0 is using the same RFID and device ID as a0 so we only
need to switch the name.
Signed-off-by: Mordechai Goodstein <mordechay.goodstein@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

New FW versions require the D0I3_END_CMD to be sent as the first
command to the FW in the resume flow.  If the TLV is set, send that
command first, otherwise keep the original behavior (i.e. send last).
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

There is no need to send D0I3_END_CMD as ASYNC during the system
resume flow.  Additionally, the other flags used are meaningless in
this case (they were just copied from the runtime resume flow), so
remove them all.
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

Shaul reported that when iwlmvm was sending beacons, it didn't properly
also take ownership of the probe responses. This is because the whole
mac80211 callback (tx_last_beacon) wasn't implemented. Fix that to make
IBSS discovery work better.
Reported-by: Shaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

Letting the preprocessor/compiler generate the shift/mask by itself
is a win for readability, so use bitfield.h for some registers.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

It's safer to use scnprintf() here because the buffer might
be too short for the full format strings. In most cases
this isn't true because of external limits on the values.

In one case, this fixes a stack data leak.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

Avoid one kind of symbol shadowing another in iwl_mvm_flush_sta()
by renaming the function parameter.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

The queue ID should never be 512 either, so correct the check
to be >= instead of just >.

Fixes: 310181ec ("iwlwifi: move to TVQM mode")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

Since we exit if buf->num_stored is 0, there's no need to
check it again later.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

When sending a Tx Command with a Tx packet, we allocate the
Tx command separately from the payload of the packet.
The WiFi MAC header is then copied into the buffer that was
allocated for the Tx Command. This means that this buffer
needs to be big enough to contain both. This is why it is
allocated with iwl_trans_alloc_tx_cmd which returns a
pointer to a newly allocated not zeroed struct
iwl_device_cmd.

The Tx command has a few bit fields and hence it needs to
be zeroed, but all the rest of the buffer doesn't need to
be zeroed since it will either be memcopy'ed with the MAC
header, or not even sent to the device.
This means that we don't need to zero all the
iwl_device_cmd structure, but rather only the size of
the iwl_tx_cmd structure.

Since sizeof(iwl_tx_cmd) - sizeof(iwl_tx_cmd) is about
260 bytes, this can avoid touching 4 cache lines for each
packet.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

Install GTKs on AP side for new TX API.
Don't add IV space, it's added by the HW.

While at that fix GCMP abnd GCMP-256 GTK installation
which work similarly to the new TX API.
Signed-off-by: David Spinadel <david.spinadel@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

If FW isn't alive, trying to collect debug data will
result in errors both in driver and in the collected
data, so just warn and leave the collecting function
in this case.
Signed-off-by: Liad Kaufman <liad.kaufman@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

Contrary to what some of the comments say, if rfkill was
asserted the transport will return -ERFKILL instead of
success, if CMD_WANT_SKB was set, so it's not necessary
to check cmd.resp_pkt for being NULL if the return code
was success.

Validate that this is true in iwl_trans_send_cmd().

Most of the other code modifications were done with the
following spatch:
    @@
    struct iwl_host_cmd cmd;
    identifier pkt;
    @@
    <...
    (
    pkt = cmd.resp_pkt;
    ...
    -if (!pkt) { ... }
    |
    pkt = cmd.resp_pkt;
    ...
    -if (WARN_ON(!pkt)) { ... }
    |
    -if (!cmd.resp_pkt) { ... }
    )
    ...>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

In DQA mode, there is no need to wait for the TXQ to
clear out after getting a DELBA, since traffic can
continue running on the queue.
Signed-off-by: Liad Kaufman <liad.kaufman@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>