Some APs (e.g. TP-LINK TL-WA801N) are disabling aggregation (downlink
to station) when U-APSD is enabled, resulting in low throughput.
Add a module parameter to allow disabling U-APSD support in the driver.

Also re-enable U-APSD for -9 firmware since the firmare issues were
fixed in this release.

There are devices that won't support U-APSD even with newer
firmware, so bring the TLV flag back to detect those.
Signed-off-by: NMatt Chen <matt.chen@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

New transport need to configure internal memory based on
the data in the (enlarged) alive notification from the
firmware. Add a transport API for this.
Signed-off-by: NEran Harary <eran.harary@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

In some situations (see comment) it makes sense to enable both
chains (if available) to get better throughput by having chain
diversity available.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

Fix a bug in nvm_read_section function if size of the section
is a multiple of 2K:

   - if the size of the section is *not* multiple of 2K,
   then we will have: read(2K) - return 2K ... read(2K) - return 2K
   read(2K) - return the rest (in bytes) and exit the while loop.
   - else, if the size of the section is a multiple of 2K,
   then we have: read(2K) - return 2K read(2K) - return 2K read(2K) -
   return 2K read(2K) - return 0 and exit the while with an error.

We should not return an error in the latter case, because it
might well be that the section was completely read.

Also, we try now to read all the sections as this is needed
for new devices.
Signed-off-by: NEran Harary <eran.harary@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

These values are used to give preference to WiFi according
to a certain pattern.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

setting SKIP_OVER_DTIM when skip_dtim_periods is 0
causes a ucode assert. set the flag only if needed.
Signed-off-by: NEliad Peller <eliadx.peller@intel.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

This fixes:

=================================
[ INFO: inconsistent lock state ]
3.14.3+ #5 Tainted: G           O
---------------------------------
inconsistent {SOFTIRQ-ON-W} -> {IN-SOFTIRQ-W} usage.
swapper/3/0 [HC0[0]:SC1[3]:HE1:SE0] takes:
 (&(&txq->lock)->rlock){+.?...}, at: [<ffffffffa059803c>] iwl_pcie_enqueue_hcmd+0x12c/0x1000 [iwlwifi]
{SOFTIRQ-ON-W} state was registered at:
  [<ffffffff810d9071>] __lock_acquire+0x5f1/0x13b0
  [<ffffffff810d9ee0>] lock_acquire+0xb0/0x1f0
  [<ffffffff817ef80e>] _raw_spin_lock+0x3e/0x80
  [<ffffffffa0598f7a>] iwl_pcie_txq_check_wrptrs+0x6a/0xb0 [iwlwifi]
  [<ffffffffa0594b5a>] iwl_pcie_irq_handler+0xdba/0x2670 [iwlwifi]
  [<ffffffff810ef1e0>] irq_thread_fn+0x20/0x50
  [<ffffffff810ef77f>] irq_thread+0x11f/0x150
  [<ffffffff810a04f0>] kthread+0xf0/0x110
  [<ffffffff817fa4bc>] ret_from_fork+0x7c/0xb0
irq event stamp: 1142192
hardirqs last  enabled at (1142192): [<ffffffff817efb6c>] _raw_spin_unlock_irq+0x2c/0x40
hardirqs last disabled at (1142191): [<ffffffff817ef9ef>] _raw_spin_lock_irq+0x1f/0x80
softirqs last  enabled at (1142188): [<ffffffff81079082>] _local_bh_enable+0x22/0x50
softirqs last disabled at (1142189): [<ffffffff8107ad35>] irq_exit+0xe5/0xf0

other info that might help us debug this:
 Possible unsafe locking scenario:

       CPU0
       ----
  lock(&(&txq->lock)->rlock);
  <Interrupt>
    lock(&(&txq->lock)->rlock);

Fixes: ea68f460 ("iwlwifi: pcie: clarify TX queue need_update handling")
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

When calling IWL_DEBUG_*() dev_dbg() is currently used, and this
could result in the log being ignored no matter the internal
loglevel, since dev_dbg() may get turned on or off based on the
dynamic debug mechanism.

Replace the dev_dbg() with dev_printk() since dynamic printing
is pointless as we use our own debug level mechanism and there
is just a single dev_dbg() call in the code.
Signed-off-by: NLiad Kaufman <liad.kaufman@intel.com>
[rephrase commit message a bit]
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

CMD_SYNC is really 0 which is confusing:

if (cmd.flags & CMD_SYNC) is always false.
Fix this by simply removing its definition.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

There's an extra newline that shouldn't be there - remove it.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

If a (debug) message is missing a newline, cause build error
so we don't keep reintroducing this problem.

This also prevents using a non-constant argument as the format
string which is a good idea anyway (e.g. for security reasons.)
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

A binding object associates virtual MACs to a Phy.
Binding object can ask (and get) air time, thus
allowing several MACs to share the medium.
A binding object can be in either one of 3 states:
1) standalone - A single active MAC
2) Different Channel Mode (DCM) - There are 2 active
   MACs or more (not necessarily just clients), but each
   has its own channel.
3) Same Channel Mode (SCM) - There are 2 active
   MACs or more (not necessarily just clients), that
   share the same channel.
This patch introduces an additional condition for enabling uAPSD:
uAPSD should be disabled for a client in DCM.
The reason for that is a low latency issue in Miracast.
Signed-off-by: NAvri Altman <avri.altman@intel.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

If two clients are active while each one is alone on its channel,
and the applicable flags are set, power management should be enabled.
This condition is referred to as Different Channel Mode (DCM).
Up to now, in DCM power management was supported only on BSS.
This patch adds support for P2P power management in DCM as well.
Signed-off-by: NAvri Altman <avri.altman@intel.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

There was a deadlock in monitor mode when we were setting the
channel if the channel was not 1.

======================================================
[ INFO: possible circular locking dependency detected ]
3.14.3 #4 Not tainted
-------------------------------------------------------
iw/3323 is trying to acquire lock:
 (&local->chanctx_mtx){+.+.+.}, at: [<ffffffffa062e2f2>] ieee80211_vif_release_channel+0x42/0xb0 [mac80211]

but task is already holding lock:
 (&local->iflist_mtx){+.+...}, at: [<ffffffffa0609e0a>] ieee80211_set_monitor_channel+0x5a/0x1b0 [mac80211]

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #2 (&local->iflist_mtx){+.+...}:
       [<ffffffff810d95bb>] __lock_acquire+0xb3b/0x13b0
       [<ffffffff810d9ee0>] lock_acquire+0xb0/0x1f0
       [<ffffffff817eb9c8>] mutex_lock_nested+0x78/0x4f0
       [<ffffffffa06225cf>] ieee80211_iterate_active_interfaces+0x2f/0x60 [mac80211]
       [<ffffffffa0518189>] iwl_mvm_recalc_multicast+0x49/0xa0 [iwlmvm]
       [<ffffffffa051822e>] iwl_mvm_configure_filter+0x4e/0x70 [iwlmvm]
       [<ffffffffa05e6d43>] ieee80211_configure_filter+0x153/0x5f0 [mac80211]
       [<ffffffffa05e71f5>] ieee80211_reconfig_filter+0x15/0x20 [mac80211]
       [snip]

-> #1 (&mvm->mutex){+.+.+.}:
       [<ffffffff810d95bb>] __lock_acquire+0xb3b/0x13b0
       [<ffffffff810d9ee0>] lock_acquire+0xb0/0x1f0
       [<ffffffff817eb9c8>] mutex_lock_nested+0x78/0x4f0
       [<ffffffffa0517246>] iwl_mvm_add_chanctx+0x56/0xe0 [iwlmvm]
       [<ffffffffa062ca1e>] ieee80211_new_chanctx+0x13e/0x410 [mac80211]
       [<ffffffffa062d953>] ieee80211_vif_use_channel+0x1c3/0x5a0 [mac80211]
       [<ffffffffa06035ab>] ieee80211_add_virtual_monitor+0x1ab/0x6b0 [mac80211]
       [<ffffffffa06052ea>] ieee80211_do_open+0xe6a/0x15a0 [mac80211]
       [<ffffffffa0605a79>] ieee80211_open+0x59/0x60 [mac80211]
       [snip]

-> #0 (&local->chanctx_mtx){+.+.+.}:
       [<ffffffff810d6cb7>] check_prevs_add+0x977/0x980
       [<ffffffff810d95bb>] __lock_acquire+0xb3b/0x13b0
       [<ffffffff810d9ee0>] lock_acquire+0xb0/0x1f0
       [<ffffffff817eb9c8>] mutex_lock_nested+0x78/0x4f0
       [<ffffffffa062e2f2>] ieee80211_vif_release_channel+0x42/0xb0 [mac80211]
       [<ffffffffa0609ec3>] ieee80211_set_monitor_channel+0x113/0x1b0 [mac80211]
       [<ffffffffa058fb37>] cfg80211_set_monitor_channel+0x77/0x2b0 [cfg80211]
       [<ffffffffa056e0b2>] __nl80211_set_channel+0x122/0x140 [cfg80211]
       [<ffffffffa0581374>] nl80211_set_wiphy+0x284/0xaf0 [cfg80211]
       [snip]

other info that might help us debug this:

Chain exists of:
  &local->chanctx_mtx --> &mvm->mutex --> &local->iflist_mtx

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(&local->iflist_mtx);
                               lock(&mvm->mutex);
                               lock(&local->iflist_mtx);
  lock(&local->chanctx_mtx);

 *** DEADLOCK ***

This deadlock actually occurs:
INFO: task iw:3323 blocked for more than 120 seconds.
      Not tainted 3.14.3 #4
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
iw              D ffff8800c8afcd80  4192  3323   3322 0x00000000
 ffff880078fdb7e0 0000000000000046 ffff8800c8afcd80 ffff880078fdbfd8
 00000000001d5540 00000000001d5540 ffff8801141b0000 ffff8800c8afcd80
 ffff880078ff9e38 ffff880078ff9e38 ffff880078ff9e40 0000000000000246
Call Trace:
 [<ffffffff817ea841>] schedule_preempt_disabled+0x31/0x80
 [<ffffffff817ebaed>] mutex_lock_nested+0x19d/0x4f0
 [<ffffffffa06225cf>] ? ieee80211_iterate_active_interfaces+0x2f/0x60 [mac80211]
 [<ffffffffa06225cf>] ? ieee80211_iterate_active_interfaces+0x2f/0x60 [mac80211]
 [<ffffffffa052a680>] ? iwl_mvm_power_mac_update_mode+0xc0/0xc0 [iwlmvm]
 [<ffffffffa06225cf>] ieee80211_iterate_active_interfaces+0x2f/0x60 [mac80211]
 [<ffffffffa0529357>] _iwl_mvm_power_update_binding+0x27/0x80 [iwlmvm]
 [<ffffffffa0516eb1>] iwl_mvm_unassign_vif_chanctx+0x81/0xc0 [iwlmvm]
 [<ffffffffa062d3ff>] __ieee80211_vif_release_channel+0xdf/0x470 [mac80211]
 [<ffffffffa062e2fa>] ieee80211_vif_release_channel+0x4a/0xb0 [mac80211]
 [<ffffffffa0609ec3>] ieee80211_set_monitor_channel+0x113/0x1b0 [mac80211]
 [<ffffffffa058fb37>] cfg80211_set_monitor_channel+0x77/0x2b0 [cfg80211]
 [<ffffffffa056e0b2>] __nl80211_set_channel+0x122/0x140 [cfg80211]
 [<ffffffffa0581374>] nl80211_set_wiphy+0x284/0xaf0 [cfg80211]

This fixes https://bugzilla.kernel.org/show_bug.cgi?id=75541

Cc: <stable@vger.kernel.org> [3.13+]
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

Stop sched scan on bss_info_changed if !idle to avoid sched scan
during association.
Signed-off-by: NDavid Spinadel <david.spinadel@intel.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

Add an option to notify mac80211 about sched scan complete from
iwl_mvm_stop_sched_scan(), to enable this notification when the
stop was called from within the driver.
Signed-off-by: NDavid Spinadel <david.spinadel@intel.com>
Reviewed-by: NAlexander Bondar <alexander.bondar@intel.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

in family 8000: additional channel 14 added to the low
band list. now the number of channels in the low band
is 15.
Signed-off-by: NEran Harary <eran.harary@intel.com>
Reviewed-by: NDor Shaish <dor.shaish@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

Some races with the hardware can happen when we take
ownership of the device. Don't give up after the first try.

Cc: <stable@vger.kernel.org>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

The 8000 family products need a file on the file system
which is used as NVM. This file is a must, if no filename
is supplied as module parameter, use a default filename.
Signed-off-by: NEran Harary <eran.harary@intel.com>
Reviewed-by: NDor Shaish <dor.shaish@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

This is not needed since the PCIe layer disables the
bottom halves before it calls the op_mode.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

All messages should have a trailing newline, add all the
missing ones. Also make all messages constants, replacing
the single one that pointlessly used a variable.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

In case RFKILL is in KILL position, the NIC will issue an
interrupt straight away. This interrupt won't be sent
because it is masked in the hardware.
But if our interrupt service routine is called for another
reason (SHARED_IRQ), then we'll look at the interrupt cause
and service it. This can cause bad things if we are not
ready yet.
Explicitly clean the interrupt cause register to make sure
we won't service anything before we are ready to.

Cc: <stable@vger.kernel.org> [3.14]
Reported-and-tested-by: NAlexander Monakov <amonakov@gmail.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

After fixes to the rs algorithm reenable MCS9.
Signed-off-by: NEyal Shapira <eyalx.shapira@intel.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

Since the declaration of iwl_mvm_fw_error_rxf_dump and
iwl_mvm_fw_error_sram_dump is under
ifdef CONFIG_IWLWIFI_DEBUGFS, do the same for their
implementation.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

This allows mac80211 to flush the Tx queues before it sends
critical management frames.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

Publish WIPHY_FLAG_HAS_CHANNEL_SWITCH if the fw supports
newly introduced IWL_UCODE_TLV_API_CSA_FLOW.
When CSA starts, save the switching vif inside mvm and during the CSA period
configure fw with a new beacon after each beacon transmission in order to
update the csa counters.
Also, handle correctly the CSA unbind-bind flow which is triggered by mac80211
when the actual channel switch happens.
Signed-off-by: NAndrei Otcheretianski <andrei.otcheretianski@intel.com>
Reviewed-by: NLuciano Coelho <luciano.coelho@intel.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

Some ampdu actions change queues by direct target access.

Since the bus is disabled in d0i3, make sure to exit d0i3
before handling these actions.
Signed-off-by: NEliad Peller <eliadx.peller@intel.com>
Reviewed-by: NArik Nemtsov <arik@wizery.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

When a firmware error occurs, capture the last 32 commands
(which are still in memory) in the error dump debugfs file.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

Currently the FW doesn't support sched scan while associated,
Prevent it.
Signed-off-by: NDavid Spinadel <david.spinadel@intel.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Reviewed-by: NLuciano Coelho <luciano.coelho@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

This variable always tracks a constant value (256) so there's
no need to have it. Removing it simplifies code generation,
reducing the .text size (by about 240 bytes on x86-64.)
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

The 'reset' argument is clearly a boolean, so use bool instead
of u8 with 0/1 values.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

If there's no LED on the system, it doesn't make a lot of sense
to include close to 4k of LED-related code (mostly in mac80211),
so instead of forcing LED support into the kernel, don't build
iwlwifi/mac80211 LED support if there's no LED class support.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Reviewed-by: NEliadX Peller <eliad@wizery.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

The old code checks if hw_section_num is valid while the
right thing to do is to check if section[hw_section_num].data
is valid.
Signed-off-by: NEran Harary <eran.harary@intel.com>
Reviewed-by: NLiad Kaufman <liad.kaufman@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

For family 8000 products, the driver should take the MAC
address from the mac_override section and only if this
section is empty it should take it from the HW section.
Signed-off-by: NEran Harary <eran.harary@intel.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

Read first the nvm sections from the OTP, then read the nvm
sections from the external file and override the OTP values
(if there were any values in the OTP).
Signed-off-by: NEran Harary <eran.harary@intel.com>
Reviewed-by: NDor Shaish <dor.shaish@intel.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

A few devices aren't allowed to be powered up at driver
load time. Add "power_up_nic_in_init" flag to iwl_cfg
structure to customize the load flow according to the
device.
Signed-off-by: NEran Harary <eran.harary@intel.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

OTP size changed in family 7000 and in family 8000.
Signed-off-by: NEran Harary <eran.harary@intel.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

There's no need to use a variable just to return the result
of another function call at the end of a function, remove it.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

The expected throughput table used when estimating a new column
depends on the channel bandwidth. When switching from legacy to
siso or mimo the wrong expected throughput table was used as it
was chosen based on the the channel bandwidth in legacy which is
always 20Mhz. Instead it should be chosen based on the current sta
bandwidth which could be also 40Mhz or 80Mhz.
When enabling MCS9 this bug leads to a problem where the max
expected throughput of the siso or mimo column is 0 as MCS9 isn't
supported in 20Mhz. This in turns prevents switching to siso or mimo.

Fix this by using the sta bandwidth when deciding which table to choose.
Signed-off-by: NEyal Shapira <eyalx.shapira@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

TPC is a power optimization which should only be attempted when
we're in the "stay in column" state after we've locked onto an optimal
rate.
Signed-off-by: NEyal Shapira <eyalx.shapira@intel.com>
Reviewed-by: NEliad Peller <eliad@wizery.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>