Implement the switch_vif_chanctx operation with support for a
single-vif and SWAP mode.  The REASSIGN mode and multi-vifs are not
supported yet.

This operation needs to implement 4 steps, namely unassign, remove,
add and assign the chanctx.  In order to do this, split out these
operations into locked and non-locked parts, thus allowing us to call
them while locked.

Additionally, in order to allow us to restart the hardware when
something fails, add a boolean to the iwl_mvm_nic_restart() function
that tells whether the restart was triggered by a FW error or
something else.
Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

It turns out that adding the update type argument was pointless as
quota update is never called from the add_interface() callback.
Therefore, IWL_MVM_QUOTA_UPDATE_TYPE_NEW isn't actually needed and
then only a "disabled_vif" argument is needed for the upcoming CSA
work.

Remove the whole enum iwl_mvm_quota_update_type and pass the right
arguments (always NULL for disabled vif right now) to the function
in all current call sites.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Reviewed-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

Simplify the quota iterator by not passing the update type,
it only needs to know whether or not to skip an interface.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Reviewed-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

The firmware currently deals with zero quota for a given
binding, but it seems odd to send that down. Make sure
that we don't do that.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

There are some cases where we can currently send zero quota
for a valid binding, e.g. if we update while an interface is
bound to a channel context but not yet acting as an AP.

Avoid this by reordering the checks.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

In some cases (e.g. when we're doing a channel switch), we may need to
disable the quota of a vif temporarily.  In order to do so, add an
argument to the iwl_mvm_update_quotas() function to tell if the passed
vif is a new one or if it should be disregarded.
Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

when driver takes the MAC address from the HW section and
it isn't valid - print an error.
Signed-off-by: Eran Harary <eran.harary@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

Signed-off-by: David Spinadel <david.spinadel@intel.com>
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

Initialize LMAC scan command.
Fix EBS flag to be dependant on TLV flg and fix other bugs.
Signed-off-by: David Spinadel <david.spinadel@intel.com>
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

Add new scan API that uses the same command 0x51 for both regular and
sched scan.
Signed-off-by: David Spinadel <david.spinadel@intel.com>
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

These file are removed recursively anyway, so there's no
point saving them just to redundantly remove them later.
Signed-off-by: Eliad Peller <eliadx.peller@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

iwl_mvm_rs_rate_init() is called multiple times to re-init
the rate scaling statistics (e.g. after some idle time).

It clears all the lq_sta sta, including some fields that
shouldn't be cleared (e.g. debugfs pointers). Fix it
by adding a new 'persistent' sub-struct, and
avoid clearing it on (re-)init.

Move the initialization of the persistent fields to
rs_alloc_sta instead.
Signed-off-by: Eliad Peller <eliadx.peller@intel.com>
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

Signed-off-by: Eran Harary <eran.harary@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

When TxTxCo-Running is active, we can relax the constraints
on the rate control.
When RxRxCo-Running is active, we can relax the constrains
on SMPS.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

Fix the debugfs hook to make it able to display the data
with the old firmware API.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

Tell the firmware if TTC should be enabled when switching
to High Band.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

No need to send the big BT_COEX_CMD command, we have now
a much thiner command that updates only what is needed.
Adapt the code to that.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

No need to send the big BT_COEX_CMD command, we have now
a much thiner command that updates only what is needed.
Adapt the code to that.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

No need to send the big BT_COEX_CMD command, we have now
a much thiner command that updates only what is needed.
Adapt the code to that, and open the patch to the updates.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

Start the new BT Coex implementation.
Don't react to notifications for now - only the initial
configuration is implemented. The rest will happen in next
patches.
Since coex.c now uses the new the new structures in all
functions, we need to adapt the code to compile, even if it
doesn't run yet.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

A new API is coming. This new API is not backward
compatible. So we need to keep the old commands to be able
to work with the former API.
Move all the current code into a new file: coex_legacy.
If a firmware with the new API is detected, we currently
just bail out since the implementation of the new API will
come in future patches.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

These flags are not used by the firmware anyway.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

This is matches the convention of the other structures.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

In family 8000 products the MAC address in the OTP could be in either:
- WFPM address
- PCIE address
In sdio product we should read it from the WFPM, in pcie product we
should read it from the PCIe location.
This is relevant only from otp version 0xE08 and above.
While at it, fix the bytes order in version 0xE08.
Signed-off-by: Eran Harary <eran.harary@intel.com>
Reviewed-by: Liad Kaufman <liad.kaufman@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

This patch makes sure there're no target accesses in the add
interface flow before d0i3 exit completes.
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

The memory was not zeroed - fix that. Also update the
iwl_fw_error_dump_info structure.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

Its content can move to the caller.
While at it, move iwl_mvm_fw_error_rxf_dump to caller.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

Instead of reading all the data in the context of the
interrupt thread, collect the data in the restart flow
before the actual restart takes place so that the device
still has all the information.
Remove iwl_mvm_fw_error_sram_dump and move its content to
iwl_mvm_fw_error_dump.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

The chunks of data do not need to be multipliers of 4 nor
4-bytes aligned.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

Firmware folks seem say that this flag can make trouble.
Drop it. The advantage of CTS to self is that it slightly
reduces the cost of the protection, but make the protection
less reliable.

Cc: <stable@vger.kernel.org> [3.13+]
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

We should always prefer to use full RTS protection. Using
CTS to self gives a meaningless improvement, but this flow
is much harder for the firmware which is likely to have
issues with it.

CC: <stable@vger.kernel.org>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

This adds some cores with 0x2057 radio which will be supported soon as
well as core 40 that I missed in the earlier firmware patch.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

It doesn't include any device (radio revision) specific code yet, so it
isn't really usable. As the commit says, it's just some generic code.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Enable initialization and update calibration code to fix:
b43-phy0 ERROR: Radio 0x2057 rcal timeout
b43-phy0 debug: Radio 0x2057 rccal timeout
b43-phy0 debug: Radio 0x2057 rccal timeout
b43-phy0 ERROR: Radio 0x2057 rcal timeout
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

There are several exit path from the PCI probe function.
Some of them, that are taken in case of errors, forget to set the "err"
variable, that is returned by the probe function.
This can lead to the kernel thinking the probe function succeeds while it
didn't, and this in turn causes extra calls to the "remove" function.

This patch fix this problem by ensuring "err" variable is assigned to a proper
non-zero value in each exit path.
Signed-off-by: Andrea Merello <andrea.merello@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

All boards supported by this driver could work using PIO or MMIO for accessing
registers.
This driver tries to access HW by using MMIO, and, if this fails for somewhat
reason, the driver tries to fall back to PIO mode.

MMIO-mode is straightforward on all boards.
PIO-mode is straightforward on rtl8180 only.

On rtl8185 and rtl8187se boards not all registers are directly available in PIO
mode (they are paged).

On rtl8185 there are two pages and it is known how to switch page.
PIO mode works, except for only one access to a register out of default page,
recently added by me in the initialization code with patch:
rtl818x_pci: Fix rtl8185 excessive IFS after CTS-to-self
This can be easily fixed to work in both cases (MMIO and PIO).

On rtl8187se, for a number of reasons, there is much more work to do to fix PIO
access.
PIO access is currently broken on rtl8187se, and it never worked.

This patch fixes the said register write for rtl8185 and makes the driver to
fail cleanly if PIO mode is attempted with rtl8187se boards.

While doing this, I converted also a couple of printk(KERN_ERR) to dev_err(), in
order to make checkpatch happy.
Signed-off-by: Andrea Merello <andrea.merello@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Currently the driver configures mac80211 to provide two rates for each TX frame:
One initial rate and one alternate fallback rate, each one with its retry count.

HW does not support fully this: rtl8180 doesn't have support for rate scaling at
all, and rtl8185/rtl8187SE supports it in a way that does not fit with mac80211:
The HW does automatically fall back to the next lower rate, and only a lower
limit can be specified, so the HW may TX also on rates in between the two rates
specified by mac80211.  Furthermore only the total TX retry count can be
specified for each packet, while the number of TX attempts before scaling rate
can be configured only globally (not per each packet).

Currently the driver sets the HW auto rate fallback mechanism to quickly scale
rate after a couple of retries, and it uses the alternate rate requested by
mac80211 as fallback limit rate (and it does this even wrongly).

The HW indeed will behave differently than what mac80211 mandates, that is
probably undesirable, and the reported TX retry count may not refer to what
mac80211 thinks, and this could fool mac80211.

This patch makes the driver to declare to mac80211 to support only one rate
configuration for each packet, and it does disable the HW auto rate fallback
mechanism, relying only on SW and letting mac80211 to do all by itself.

This should ensure correct operation and fairness respect to mac80211.
Indeed here tests with iperf do not show significant performance differences.
Signed-off-by: Andrea Merello <andrea.merello@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>