The change enables offloading TCP checksum calculation to firmware.
There are still some issues with the checksum offload so better to
disable it by default until the issues are resolved.

To enable TCP checksum offload for tx and rx paths, use
the ethtool as follows:
ethtool -K <interface> tx on
ethtool -K <interface> rx on

To disable TCP checksum offload, for tx and rx paths, use
the ethtool as follows:
ethtool -K <interface> tx off
ethtool -K <interface> rx off

kvalo: indentation changes
Signed-off-by: NRishi Panjwani <rpanjwan@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This is needed to fix current hostapd/wpa_supplicant AP operations for
frame registration. P2P GO mode already advertised these, but AP mode
was forgotten and could not be used after the hostapd/wpa_supplicant
frame registration changes.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

hostapd/wpa_supplicant AP mode uses this operation to flush the station
entries. Implement this in ath6kl to avoid unnecessary warnings from
NL80211_CMD_DEL_STATION failing.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

cfg80211 could pass a NULL net_device to the driver via the
set_channel() callback, when it receives a request to set the
device's channel. Not handling this case properly results in this panic:

BUG: unable to handle kernel NULL pointer dereference at 0000000000000cb0
IP: [<ffffffffa0374e49>] ath6kl_cfg80211_ready+0x9/0x70 [ath6kl_sdio]

Call Trace:
[<ffffffffa0374ed7>] ath6kl_set_channel+0x27/0x90 [ath6kl_sdio]
[<ffffffffa04dfd5f>] cfg80211_set_freq+0xff/0x1d0 [cfg80211]
[<ffffffffa04d3275>] ? nl80211_set_wiphy+0x85/0x660 [cfg80211]
[<ffffffffa04d3198>] __nl80211_set_channel.isra.39+0x118/0x140 [cfg80211]
[<ffffffffa04d34f3>] nl80211_set_wiphy+0x303/0x660 [cfg80211]
[<ffffffff813678d7>] ? rtnl_lock+0x17/0x20
[<ffffffffa04ca165>] ? nl80211_pre_doit+0xb5/0x150 [cfg80211]
[<ffffffff81382a45>] genl_rcv_msg+0x1d5/0x250
[<ffffffff81382870>] ? genl_rcv+0x40/0x40
[<ffffffff81381be9>] netlink_rcv_skb+0xa9/0xd0
[<ffffffff81382855>] genl_rcv+0x25/0x40
[<ffffffff811354f0>] ? might_fault+0x40/0x90
[<ffffffff81381519>] netlink_unicast+0x2d9/0x320
[<ffffffff813818e6>] netlink_sendmsg+0x2c6/0x320
[<ffffffff81343800>] ? sock_update_classid+0xb0/0x110
[<ffffffff8133f52e>] sock_sendmsg+0x10e/0x130
[<ffffffff81169933>] ? mem_cgroup_update_page_stat+0x193/0x250
[<ffffffff811354f0>] ? might_fault+0x40/0x90
[<ffffffff811354f0>] ? might_fault+0x40/0x90
[<ffffffff81135539>] ? might_fault+0x89/0x90
[<ffffffff811354f0>] ? might_fault+0x40/0x90
[<ffffffff8134d126>] ? verify_iovec+0x56/0xd0
[<ffffffff8133f9e6>] __sys_sendmsg+0x396/0x3b0
[<ffffffff8108e823>] ? up_read+0x23/0x40
[<ffffffff81040e68>] ? do_page_fault+0x208/0x4e0
[<ffffffff81193151>] ? vfsmount_lock_local_unlock+0x21/0x60
[<ffffffff811947d0>] ? mntput_no_expire+0x30/0xe0
[<ffffffff8119489f>] ? mntput+0x1f/0x30
[<ffffffff81342149>] sys_sendmsg+0x49/0x90
[<ffffffff81425482>] system_call_fastpath+0x16/0x1b
Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

As firmware starting from 3.2.0.12 has some API changes and doesn't work
with older versions of ath6kl we need to bump up the API version. This
way we don't break anything.

Also store which version of API is used and print that during boot:

ath6kl: ar6003 hw 2.1.1 sdio fw 3.2.0.13 api 3
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Currently ath6kl has just hardcoded paths to each firmware file. Change
this more dynamic by separating the the directory and file name from each
other. That way it's easier to dynamically create full paths to firmware and
code looks better. And now it's possible to remove a function needed by
devicetree code.

While at it add a structure inside struct ath6kl_hw to contain all
firmware names. I deliberately omitted board file support as
those will be handled later.

This is needed for firmware API 3.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Firmware ARP module requires own IP addr in order to respond
to the outside world when the target is in WOW suspend state.

At present, firmware ARP module has capability to hold 2 IP addr.
So, WOW mode will be disabled if the total IP addr configured in
net_dev for our device is greater than firmware limit (MAX_IP_ADDRS)
which is 2 at this moment.
Signed-off-by: NRaja Mani <rmani@qca.qualcomm.com>
Signed-off-by: NThirumalai Pachamuthu <tpachamu@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

In this way, caller is free to pass only the value of IP addr
to configure.

In addition to this,
  * 'ips' variable data type in struct wmi_set_ip_cmd is changed
     from __le32 to __be32 in order to match network byte order.
  *  ipv4_is_multicast() is used to validate multicast ip addr.
  *  New argument if_idx is added to supply correct vif index
     to ath6kl_wmi_cmd_send().

This will be used in the next patch.
Signed-off-by: NRaja Mani <rmani@qca.qualcomm.com>
Signed-off-by: NThirumalai Pachamuthu <tpachamu@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Empty IE buffer means that the new association is not supposed to
include extra IEs. Make sure any previously configured (Re)Association
Request frame IEs get cleared in such a case. This is based on a patch
from Shuibing.

Cc: Dai Shuibing <shuibing@qca.qualcomm.com>
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This is to avoid the scenario where the bss entry of the AP got
expired when reporting roaming event to current AP. As the bss
entry for the current bss is available in driver, pass this bss
to cfg80211. This fixes WARNING: at net/wireless/sme.c:586.

This patch depends on the following patch in cfg80211
"cfg80211: Fix race in bss timeout".
Reported-by: NKalle Valo <kvalo@qca.qualcomm.com>
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This patch enables support for doing P2P management operations like device
discovery on top of a station interface. After group formation, the station
interface will become a P2P GO/client interface as the case may be.

This feature requires modifications to a couple of existing WMI structures and
therefore new command ids and structures have been defined in order to be
compatible with older firmware versions and other chips. The exception here is
the wmi_connect_cmd. Adding a new field to the end of the structure will not
cause any issues with previous firmware versions since firmware only checks for
minimum length of the command. The other structures are of variable length,
hence it was not possible to add new fields to the end.

The new command ids have to be added to the end of enum wmi_cmd_id, so it has
updated to match the firmware.

The driver will support both the 'old' and the 'new' commands for a while by
checking the firmware capabilities.
Signed-off-by: NAarthi Thiruvengadam <athiruve@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

AR6003 family use uart_tx=8 and refclk=26Mhz by default, and AR6004 family
uses different uart_tx pin and could also support various xtal source,
moves these per hw configuration.
Signed-off-by: NRyan Hsu <ryanhsu@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

ath6kl firmware supports scheduled scan functionality with the wow ssid
filter. But the firmware does not send any events after scan results
so I had to add a timer which notifies about new scan results.

Sched scan needs firmware version 3.2.0.6 or later. If firmware doesn't
support sched scan the driver will not enable the feature.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

For some strange reason I used ALIGN() to calculate index to the
buffer. That is totally bogus and wouldn't work when it tried to read
the second bit. Fix it by removing the ALIGN() altogether.

Also check that ie_len is not too short.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

According to the firmware engineers WOW_FILTER_SSID is actually the
second bit, not the first.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

During suspend we need to stop all vifs, not just the first.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This way it's possible to keep all disconnect logic in one function and
easier to add new functionality, like stopping scheduled scan.

There are some changes to commands called during network interface close,
but there should not be any visible changes in functionality.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

There's a workaround in ath6kl_cfg80211_stop() which emits disconnected
even when sme_state was disconnected. This is legacy from the old
staging driver and I can't repoduce the old problem anymore. I assume the
bug got fixed while the driver was cleaned up so let's get
rid of the hack.

This makes it possible to call ath6kl_cfg80211_stop from ath6kl_close()
which happens in a followup patch.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

It really should not return a void pointer. Also remove useless casts
from its users.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This is needed to allow WAPI AP to configure the initial group key
to the target in the same way as is done with TKIP/CCMP. This
fixes broadcast data frame delivery with the initial group key.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The maximum number of clients which ath6kl can support in AP mode
is 10. The limitation of 8 connections is only for older chipsets
which ath6kl does not support.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

It was only initialised but not used anywhere. Also remove two defines
which ended up unused after this change.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

ath6kl assumed cfg80211 passed to us power in dBm but it is in mBm.

Cc: Kalle Valo <kvalo@qca.qualcomm.com>
Signed-off-by: NLuis R. Rodriguez <mcgrof@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The return value of snprintf() is the number of bytes which would
have been copied if there was enough space, but we want the number of
bytes actually copied.  The scnprintf() function does this.

Also in theory, a %u can take take 10 digits so we may as well make
the buffer larger as well.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Currently mac80211 implements these for all devices,
but given restrictions of some devices that isn't
really true, so prepare for being able to remove the
capability for some mac80211 devices.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Acked-by: NKalle Valo <kvalo@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Firmware crashes while starting Soft AP in 32 bit x86 platform.
The reason is that the single dma buffer (ar_sdio->dma_buffer)
is used in ath6kl_sdio_read_write_sync() for unaligned buffer
handling and this function is called in the multiple context
at the same time. So, finally hits dma buffer corruption and
firmware crash.

Mutex is used to protect dma buffer to avoid data corruption.
Spin lock can not used to fix this issue since mmc stack
read/write calls may for sleep.

Observed this issue with recently commited patch
"ath6kl: Claim sdio function only at appropriate places"
861dd058

kvalo: change name of mutex to more descriptive and add a comment
about what it protects
Signed-off-by: NRaja Mani <rmani@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

There is an unlock missing on this error path.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The maximum number of supported virtual interfaces are 3.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Currently the max number of vifs which can be used for non-p2p
mode is determined in ath6kl_core_alloc(). But the maximum
supported vifs are parsed from firmware IE in ath6kl_fetch_fw_api2()
which would happen after ath6kl_core_alloc().
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This modparam was introduced to enable non-p2p mode operation
on two virtual interfaces. It does not seem to be necessary to
have a separate module parameter to do that. Instead, this option
can be enabled when any one of the interfaces is not going to be
used for p2p (ath6kl_p2p).
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The ath6kl responds to probe-requests in HW while operating as an AP. It
supports offloading exclusions to support the WPS, WPS2, P2P and
802.11u protocols.
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

When operating AP interface is brough down the mode is reset to
STA. This STA will be reconfigured into AP mode when the interface
is brought up again. This sequence will be successful only when
change_virtual_intf() returns with no error, but there is a
check in this callback which does the type change only when
that interface is active. This callback does nothing more
than saving the new interface type to vif and wdev, so the
sanity check for interface state and wmi state is not necessary.
This makes the AP interface functional again after interface down/up.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Connect control flag CONNECT_WPS_FLAG has to be cleared
by default even if the driver receives zero length IE
from cfg80211.

Otherwise this flag would be always set after WPS exchange
which would lead wpa_supplicant to fail to connect with
the received WPS credentials. This issue is observed only
in OPEN security.

kvalo: use cfg80211 instead of CFG in the commit log
Signed-off-by: NRaja Mani <rmani@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Not all firmwares support multiple vifs and we need to read the limit from
the firmware image.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Needed when detecting how many vifs firmware supports.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

When adding ar6004 SDIO support I forgot to add corresponding
MODULE_FIRMWARE() definitions to sdio.c.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Part of ath6kl uses "REV3" style of naming hardware versions and elsewhere
"hw 2.1.1" is used instead for the same version. This is confusing, use
the latter term everywhere.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

To make it easier to print name for each hardware type. Also move the hw
info print to ath6kl_init_hw_start() which is more logical place for it.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Board data address can change between firmwares so we need to read that
from the firmware image.

Also fix debug log for the patch address to print the address in hex.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>