Now that we allow the first Beacon frame after each connection to
be processed at the host, we can figure out the DTIM period and
expose it with Beacon interval through STA info BSS parameters
to user space.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Since we may end up using a dummy BSS entry when roaming, allow one Beacon
frame -based bssinfo from the current BSS to be processed prior to starting
to filter all bssinfo events. This allows cfg80211 BSS table to be filled
with proper data in the roaming case where the full Beacon data may not
have been present at the time of roamed event.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

There is no point in unconditionally converting the bssinfo header
to the old version since only the new header is being used and the
driver can as well read the values from it when needed. Leaving out
the conversion saves some extra memory copying.

In addition, use the calculated "rssi" value snr - 95 dBm to get the
proper value in cfg80211 BSS table (i.e., something that more or less
matches with the value used in STA info).
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Now that the scan results are reported directly to the cfg80211 BSS
table there is no need for maintaining this internal node table
implementation for scan results. Remove the definitions and node
table functions.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The internal node table in ath6kl was not really used for any useful
purpose. It was just used to collect scan results during a scan and then
provide them in a burst to cfg80211 at the completion of the scan. There
is no point in doing this since cfg80211 is perfectly capable of
maintaining the BSS table and the BSS inform messages are sent in
separate function calls anyway.

This provides more complete information in the cfg80211 BSS table since
this allows Beacon and Probe Response frames to be distinguished and IEs
from them reported separately.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Since ath6kl does not actually update cfg80211 BSS table when this
event occurs, there is not much need for removing the entries from
the internal table that is not really used or exposed.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

ath6kl does not actually update cfg80211 BSS table when this update
occurs, so there is not much need in updating the internal table
that is not used or exposed.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This event has been deprecated and there is no need for ath6kl to
include code for processing it.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This fixes the BSS "update" just before the connected or roamed event.
The previous implementation was completely broken: it forced a hardcoded
signal strength and IEs from Association _Request_ frame instead of any
Beacon information. This broke various things, including PMKSA caching.

The current workaround for creating a dummy BSS entry before the roamed
event is not exactly ideal, but that is quite a bit better than the
previous state. As a future improvement, cfg80211 could potentially be
extended to allow this type of use or ath6kl could delay sending the
roamed event before receiving a BSS info event.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This fixes the following panic observed on card removal.

BUG: unable to handle kernel paging request at f86e22ac
EIP is at wiphy_update_regulatory+0x252/0x590 [cfg80211]
Call Trace:
 [<f92fb9e5>] set_regdom+0x165/0x600 [cfg80211]
 [<c02057ba>] ? __kmalloc+0x10a/0x190
 [<c0366557>] ? nla_parse+0xb7/0xd0
 [<f9300b92>] ? T.1400+0x12/0x20 [cfg80211]
 [<f9300c84>] nl80211_set_reg+0xe4/0x270 [cfg80211]
 [<f92fe560>] ? nl80211_pre_doit+0x0/0x160 [cfg80211]
 [<c050bdeb>] genl_rcv_msg+0x23b/0x280
 [<c050bbb0>] ? genl_rcv_msg+0x0/0x280
 [<c050ab76>] netlink_rcv_skb+0x86/0xb0
 [<c050bb80>] ? genl_rcv+0x0/0x30
 [<c050bb9c>] genl_rcv+0x1c/0x30
Signed-off-by: NVivek Natarajan <nataraja@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Configuration of the WEP keys needs to be delayed until the AP mode has
been properly started at the target. Partial support for delaying the
WEP key configuration was already in place in the driver, but the actual
part of deciding when to do this was missing.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

After a successful completion of disconnect() driver needs to set
it's sme_state to SME_DISCONNECTED to be in sync with cfg80211
state.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Target already tries with different authentication mechanism
when authentication type is configured to NL80211_AUTHTYPE_AUTOMATIC.
Remove this piece of code from driver. Having this code in driver
even affects auto + WEP authentication in some cases.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The new firmware format includes capability bits which make it
possible to check what features the firmware supports. Add infrastructure
to read the capabilities. For now it only provides
ATH6KL_FW_CAPABILITY_HOST_P2P which is not even used anywhere yet, but that
will be added later.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

A new version of firmware needs different reserved ram size so read that
from the firmware image.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

It's actually possible to read the firmware start address from hardware,
that way there's no need to hardcode the address in hardware.

Thanks to Chilam Ng for the idea.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Less if clauses this way and again easier to override the values.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Currently the load address was calculated everytime when it was needed,
and with a mess if clauses. Simplify this by adding a field to struct
ath6kl for each address and choose the address with simple switch
statements.

Also move the code just after target version is retrieved. That way it's
easier to override the values later in the boot process.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

In the new format all the format images are embedded into one file.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Brent reported that ath6kl busy loops if firmware doesn't boot for some
reason (in this case he was using an older firmware which wasn't supported
by ath6kl).

Investigation revealed that this was even on purpose,
ath6kl_bmi_get_rx_lkahd() had a parameter to disable the timeout check,
which is extremely evil. I didn't find any reason why the timeout needs
to be disabled so I just removed the feature. The function already busyloops
a maximum of one second if it doesn't get an answer, even that's too long.
If something takes longer than that a more friendly approach is needed.
Reported-by: NBrent Taylor <btaylor1@motorolasolutions.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

In preparation for the new firmware image format.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

When no default board data file is present query the device tree for a
board-id setting to identify the board data to use.  If the FDT lacks the
necesary info fall back to the previous behaviour of using a compile-time
board filename.
Signed-off-by: NSam Leffler <sleffler@chromium.org>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

For now, use a module parameter (ath6kl_p2p) to allow P2P support
to be enabled. This is needed since there is no mechanism for
enabling the P2P mode more dynamically for a single netdev.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Similarly like with the tx path, unify naming in htc rx path. No functional
changes.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This is to make it easier follow tx code path inside htc.No functional
changes.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Process the regulatory code from eeprom and pass the
country information to cfg80211.

kvalo: add space between struct name and *
Signed-off-by: NVivek Natarajan <nataraja@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

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

This cleans up the connect event parsing by defining a union in
struct wmi_connect_event to match with the three possible sets of
fields that the target uses depending on which type of connect
event is being indicated. In addition, two AP cases are now
separated from ath6kl_connect_event() so that correct field names
can be used to make it actually possible to understand what the
code is doing.

The bug hiding in the previous mess was in parsing the AID incorrectly
when processing the new station connecting event in AP mode. The fix
here for that is also fixing TIM updates for PS buffering to use the
correct AID.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

There is a 2-octet reserved field between the flag and aid fields. Fix
that to make the target actually behave as requested.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Limit the length of the channel list to WMI_MAX_CHANNELS to avoid
rejection of the request in wmi.c. Since there is not really much
point in using a specific list of more than 32 channels, drop the
channel list if more channels are specified and scan all channels.

Fix cfg80211 scan API use: ar->scan_req must be set only if returning
success from scan() handler. The previous version would result in use
of freed memory and likely kernel panic should the scan request fail
to be sent to the target.
Signed-off-by: NJouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

To write a value to register:
echo <register_offset>=<register_value> > <degfs_root>/ieee80211/phyX/ath6kl/reg_write

kvalo: rename file to reg_write to follow the style of other debugfs files
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-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>

So that this can be called from debug.c when adding support
to write chip register.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Firmware initiates roaming only after it reaches a rssi of 20.
This lower rssi threshold can be modified through a wmi command
to modify the roaming behavior.

kvalo: rename debugfs functions and move comment about rssi units next to
ath6kl_wmi_set_roam_lrssi_cmd()
Signed-off-by: NVivek Natarajan <nataraja@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Need to make sure the chip address for which we need the value
si endian safe.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

To dump a particular register:

    echo <reg_addr> > <debugfs_root>/ieee80211/phyX/ath6kl/reg_addr

To dump the entire register set:

    echo 0 > <debugfs_root>/ieee80211/phyX/ath6kl/reg_addr

Register values will be available at:

    cat <debugfs_root>/ieee80211/phyX/ath6kl/reg_dump

kvalo: commit log cleanup, renamed few functions, removed a warning
message
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Firmare sends the logs only when it's internal ring buffer is full. But
if firmware crashes we need to retrieve the latest logs through diagnose
window. This is now done everytime the debugfs file is read.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Just to make them a bit easier to read and unify naming. 32 suffix
in the function name means that it will be a 32 bit transfer. If there's
no number a buffer is transfered instead.

Use void pointers to get rid of ugly casts.

Don't provide target address as a pointer, pass it by value. Same for
the value used in write32().
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Firmware log parameters can be controlled now with help of fwlog_mask
debugfs file.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Firmware sends binary logs with WMIX_DBGLOG_EVENTID event. Create
a buffer which stores the latest logs and which can be copied from
fwlog debugfs file with cp command.

To save memory firmware log support is enabled only when CONFIG_ATH6KL_DEBUG
is enabled.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>