"ath6kl: Defer wiphy and netdev registration till the end of ath6kl_core_init()"
causes kernel panic by accessing the unallocated debug resources during
boot time. To fix this, split the debug initialization funtion into two,
one initializes the debug resource and the other takes care of debugfs
initialization. When this issue shows up the kernel crash dump would
look like

 ath6kl_debug_fwlog_event+0x9c/0x10a
 [<c10666c9>] register_lock_class+0x57/0x288
 [<c1065cd3>] ? trace_hardirqs_on+0xb/0xd
 [<f801f4c9>] ? ath6kl_debug_fwlog_event+0x9c/0x10a
 [<c1066a8a>] __lock_acquire+0x96/0xbe5
 [<c106007b>] ? alarmtimer_suspend+0x80/0x127
 [<c10258da>] ? vprintk+0x394/0x3b1
 [<f801f4c9>] ? ath6kl_debug_fwlog_event+0x9c/0x10a
 [<c10676b3>] lock_acquire+0xda/0xf9
 [<f801f4c9>] ? ath6kl_debug_fwlog_event+0x9c/0x10a
 [<c1532ce3>] _raw_spin_lock+0x28/0x58
 [<f801f4c9>] ? ath6kl_debug_fwlog_event+0x9c/0x10a
 [<f801f4c9>] ath6kl_debug_fwlog_event+0x9c/0x10a
 [<f80310a4>] ath6kl_wmi_control_rx+0x69d/0xb50 [ath6kl_core]
 [<f802d2e1>] ? ath6kl_rx+0x3c/0x839 [ath6kl_core]
 [<f802d35d>] ath6kl_rx+0xb8/0x839 [ath6kl_core]
 [<c104b81e>] ? local_clock+0x2d/0x4e
 [<c102a0af>] ? _local_bh_enable_ip+0x94/0x98
 [<f802bfc0>] ? ath6kl_alloc_amsdu_rxbuf+0xb7/0xb7
 [<f8023b28>] ath6kl_htc_rxmsg_pending_handler+0x891/0x988 [ath6kl_core]
 [<f802bf00>] ? ath6kl_refill_amsdu_rxbufs+0x89/0x92
 [<f802d2a5>] ? aggr_timeout+0xed/0xed [ath6kl_core]
 [<f802bfc0>] ? ath6kl_alloc_amsdu_rxbuf+0xb7/0xb7
 [<f802c420>] ? ath6kl_tx_complete+0x376/0x376 [ath6kl_core]
 [<f8020e92>] ath6kl_hif_intr_bh_handler+0xf7/0x33e
 [<c138ab00>] ? mmc_host_disable+0x15/0x3a
 [<f8123b5c>] ath6kl_sdio_irq_handler+0x3c/0x90 [ath6kl_sdio]
 [<c1392f56>] sdio_irq_thread+0xb6/0x29c
 [<c1392ea0>] ? sdio_claim_irq+0x1cb/0x1cb
 [<c103d4c0>] kthread+0x67/0x6c
 [<c103d459>] ? __init_kthread_worker+0x42/0x42
 [<c153903a>] kernel_thread_helper+0x6/0xd
 BUG: unable to handle kernel NULL pointer dereference at
EIP: [<f801f4d4>] ath6kl_debug_fwlog_event+0xa7/0x10a

kvalo: rename new function to ath6kl_debug_init_fs() and add a comment
why it's needed
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>

No need to copy received local buffer content to bounce buffer
(DMA buffer) while performing sync READ operation from the chip.

It's applicable for only WRITE operation.
Signed-off-by: NRaja Mani <rmani@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The commit "ath6kl: Use a mutex_lock to avoid
race in diabling and handling irq" introduces a
state where ath6kl_sdio_irq_handler() would be waiting
to claim the sdio function for receive indefinitely
when things happen in the following order.

ath6kl_sdio_irq_handler()
	- aquires mtx_irq
	- sdio_release_host()
					ath6kl_sdio_irq_disable()
						- sdio_claim_host()
						- sleep on mtx_irq
	ath6kl_hif_intr_bh_handler()
		- (indefinitely) wait for the sdio
		  function to be released to exclusively claim
		  it again for receive operation.

Fix this by replacing the mtx_irq with an atomic
variable and a wait_queue.

kvalo: add ath6kl_sdio_is_on_irq() due to open parenthesis alignment
Signed-off-by: NRaja Mani <rmani@qca.qualcomm.com>
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Collect residue firmware logs following firmware assert.
Firmware sends logs to host once the 1500 byte log buffer
has been filled. At time of assert, there could be residue
logs lying in the firmware. This patch pulls those residue
logs. This would give the full picture of the firmware.
Signed-off-by: NEtay Luz <eluz@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

firmware debug utility expects firmware log record size
to be 1500 bytes. This patch ensures that the
firmware record will be exactly 1500 bytes.

kvalo: remove trailing space
Signed-off-by: NEtay Luz <eluz@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The ath6kl FW does not distinguish between different types of hidden
SSIDs (empty or null), so treat all cfg80211 requests for hidden ssid
the same.
Signed-off-by: NThomas Pedersen <c_tpeder@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Remove check so that IE in firmware is cleared if not set. Without this fix, any
previously set IE will be used incorrectly in the next frame. For example,
consider the scenario where a P2P device scan is followed by a regular station
scan. The P2P IE set by the P2P scan needs to be cleared, otherwise the station
scan will contain the P2P IE.

kvalo: indentation fixes
Signed-off-by: NAarthi Thiruvengadam <athiruve@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

As regulatory events are processed even before the wiphy is registered,
calling regulatory_hint() at early stage should be fixed.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This makes the wiphy and initial netdev registration the last step
in dev initialization. Apart from the fact that this looks right,
it can also be used to setup wiphy with the regulatory information
received from firmware after uploading the firmware. Also it fixes
a FIXME in ath6kl_core_init() where mac address is copied into
netdev->dev_addr, ath6kl_interface_add() takes care of this as well.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

When the wmi event is vif specific, the validation of vif index
is taken care in ath6kl_wmi_proc_events_iface(). This also avoids
the need for a netdev to be registered while receiving initial events
like "target_ready" and "regulatory domain".
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Split the wmi event processing into the one which needs to be
vif specific and the reset. This is a step towards avoiding
the need for wiphy and a netdev registration before getting
any message from firmware.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Move netdev->hw_features setting from ath6kl_core_init() to
init_netdev() so that it is done for every interface.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

There are some code which initializes various wiphy members
left outside ath6kl_cfg80211_init(), in ath6kl_core_init().
Move them into a single palce.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Register driver's ht capabilities in wiphy to avoid failures
in setting ht channels from hostapd.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Without this change, SDIO shuts down due to CRC error during
data communication to the firmware in some of the platform.
for example, scan request issued to the firmware doesn't
return scan completed events and cause the socket interface
to always return -16 (device busy).

SDIO pad drive strength should be reduced for hw2.1.1 board
to avoid such errors.
Signed-off-by: NRaja Mani <rmani@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This makes the device get into scan state as soon as a
scan request is received, instead of blocking the scan
request till the traffic comes down. It is  necessary
for the deterministic foreground scan particularly when
having multiple vif operating where, if the scan is non-
deterministic, scan on one interface will not start
as long as there are traffic on the other interface. This
change passes 50 msec as foreground scan interval to fix this.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

htc_packet and htc_packet->buf_start are separately allocated
for endpoint 0. This is different for other endpoints where
packets are allocated as skb where htc_packet is skb->head
and they are freed properly. Free htc_packet and htc_packet->buf_start
separatly for endpoint 0.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Rx buffers should be allocated for control and best effort endpoints only
after the enpoints connection is esablished. But this is done before the
endpoint connection is complete, we don't even the control and BE endpoints
that time. Move the buffer allocation after endpoint connection is over,
after ath6kl_init_hw_start(). Found in review, never seen any real issue
with this.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

In ath6kl_cfg80211_add_key(), params is checked for non-NULL
but this pointer has been deferenced many times before this check. This
gives the following smatch warning. add_key() can never carry NULL as
params, remove this bogus check.

ath6kl_cfg80211_add_key(86) warn: variable dereferenced before check 'params'
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This fixes smatch warning "inconsistent returns sem:&ar->sem".
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Tx packets will begin to drop when there are multiple traffic priorities
and the current traffic is not the highest priority and the remaining
cookies drop below a certain number, which is fixed for all AC. It is
possilbe that lower priority AC have more traffic which will consume
more cookies and lock out higher priority AC from having any. Assign
each endpoint (AC) with a different Tx-packet-drop threshold so lower
priority AC is more likely to drop packets and the cookies become more
available to higher priority AC.
Signed-off-by: NChilam Ng <chilamng@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

PS buffering of unicast Action frames that are sent in a context
of a BSS. In AP mode when the recepient station goes to powersave
and PS_POLL flag is not set, we would buffer the frames. Send out
unicast mgmt bufferred frame when PS_POLL is received.

This fixes a bug in P2P GO behavior when sending a GO Discoverability
Request to a client that is in sleep mode.

kvalo: indentation fixes
Signed-off-by: NThirumalai Pachamuthu <tpachamu@qca.qualcomm.com>
Signed-off-by: NNaveen Gangadharan <ngangadh@qca.qualcomm.com>
Signed-off-by: NAarthi Thiruvengadam <athiruve@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The AP/GO mode API isn't very clearly defined, it
has "set beacon" and "new beacon" etc.

Modify the API to the following:
 * start AP -- all settings
 * change beacon -- new beacon data
 * stop AP -- stop AP mode operation

This also reflects in the nl80211 API, rename
the commands there correspondingly (but keep
the old names for compatibility.)

Overall, this makes it much clearer what's going
on in the API.

Kalle developed the ath6kl changes, I created
the rest of the patch.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

The firmware crash dump printout was wrong as it was using incorrect
offsets.

kvalo: improve commit log, change the "%d:" to print word indexes, not bytes
Signed-off-by: NNaveen Gangadharan <ngangadh@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

AR6003 2.1.1 supports both 1792 and 2048 byte board files.
Add support for 2048 byte board file.

kvalo: add ath6kl prefix to the title
Signed-off-by: NPrasanna Kumar <kumarpra@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Tx bundling is the more efficient use of SDIO bus and allows more packet
transfers with fewer bus transactions, and is a way to improve overall
throughput. However, Tx bundling has only 4 scatter request resources available.
When there are multiple traffic streams of different priorities, it's possible
that lower priority traffic may hog all the scatter requests and lock out the
higher prioirty traffic from bundling.
Tx bundling is now enabled per AC. When an AC do a scatter request and
the remaining scatter request resources is lower than a configurable
threshold, it will disable Tx bundling for all AC's of lower priorities.
When an AC has Tx bundling disabled and has no Tx bundles sent in a
consecutive and configurable number of packets, Tx bundling will be re-enabled
for that AC.
Signed-off-by: NChilam Ng <chilamng@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Update license header with the copyright to Qualcomm Atheros, Inc.
for the year 2011-2012.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

When debugging firmware issues it's not always enough to get
the latest firmware logs, sometimes we need to get logs from a longer
period. To make this possible, add a debugfs file named fwlog_block. When
reading from this file ath6kl will send firmware logs whenever available
and otherwise it will block and wait for new logs.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Currently firmware logs are stored in a circular buffer, but this was
not very flexible and fragile. It's a lot easier to store logs to struct
skbuffs and store them in a skb queue. Also this makes it possible
to easily increase the buffer size, even dynamically if we so want (but
that's not yet supported).

From user space point of view nothing should change.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

ath6kl: initialize the 'nominal_phy' field in the 'wmi_create_pstream_cmd' struct for create_qos command

The nominal_phy field is uninitialized. Initialize it to min_phy_rate for
create_qos.

kvalo: simplified the equation as checkpatch complained for a too long line
Signed-off-by: NChilam Ng <chilamng@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Both Luis and John reported that they see a compiler warning:

drivers/net/wireless/ath/ath6kl/init.c: In function 'ath6kl_init_hw_params':
drivers/net/wireless/ath/ath6kl/init.c:1377:26: warning: ‘hw’
may be used uninitialized in this function

Oddly enough I have never seen it. But AFAICT the code is correct and
hw is not used uninitalized so add uninitialized_var() to inform that to
the compiler.
Reported-by: NLuis R. Rodriguez <mcgrof@qca.qualcomm.com>
Reported-by: NJohn W. Linville <linville@tuxdriver.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Correct spelling "spported" to "supported" in
drivers/net/wireless/ath/ath6kl/cfg80211.c
Signed-off-by: NMasanari Iida <standby24x7@gmail.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Set the host pm flag MMC_PM_WAKE_SDIO_IRQ to allow host
to disable the sdc2_clk and sdc2_h_clk,so that the MSM device
enter into TCXO shutdown.
Signed-off-by: NSantosh Sajjan <ssajjan@codeaurora.org>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Error code ENOTCONN is more suitable than EINVAL to report
when the driver is not in connected state in ath6kl_wow_suspend().

I found this during code review.
Signed-off-by: NRaja Mani <rmani@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

For every WMI_SET_HOST_SLEEP_MODE_CMDID command (send from the host),
the firmware sends WMI_SET_HOST_SLEEP_MODE_CMD_PROCESSED_EVENTID as
an acknowledgement to the host.

In order to being sync with the firmware, the host has to wait for
WMI_SET_HOST_SLEEP_MODE_CMD_PROCESSED_EVENT event before going to
the suspend state. This patch ensures ath6kl_wow_suspend() waits
until it gets this event after sending set host sleep mode command.

This patch adds,
 * New command WMI_SET_HOST_SLEEP_MODE_CMD_PROCESSED_EVENTID in
   WMI event table.
 * New WMI function ath6kl_wmi_host_sleep_mode_cmd_prcd_evt_rx()
   to process the event.
 * New flag HOST_SLEEP_MODE_CMD_PROCESSED in VIF flags to record
   the arrival of the event.
Signed-off-by: NRaja Mani <rmani@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

First preference is given to the user configured WOW patterns.
If the user doesn't configure any patterns (for ex, via iw command),
the default patterns will be configured based on the current mode
(vif->nw_type) while going to WOW suspend.

Summary of changes:

  * ath6kl_wow_ap() is added to configure the below default
    patterns when the system enters into WOW suspend in AP mode.
      + Unicast IP, EAPOL-like and ARP packet pattern
      + ARP packet pattern
      + mDNS/SSDP/LLMNR pattern
      + DHCP broadcast pattern

 * ath6kl_wow_sta() is added to configure the below default
   patterns when the system enters into WOW suspend in STA mode.
      + Unicast packet pattern
      + mDNS/SSDP/LLMNR pattern

  * Move the user provided WOW patterns configuration code
    from ath6kl_wow_suspend() to a separate function called
    ath6kl_wow_usr().

  * Two argument variable's ('filter' and 'mask) data type in
    ath6kl_wmi_add_wow_pattern_cmd() are changed from 'u8 *' to
    'const u8 *'. This is needed to make all pattern and mask
    arrays to be 'static const u8' in the caller function.

  * New conditional check is added to make sure user
    configured pattern count is within the limit (WOW_MAX_FILTERS_PER_LIST).
Signed-off-by: NRaja Mani <rmani@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Using this patch, the user can bypass existing auto
suspend mode selection logic and force ath6kl to enter
into the suspend mode what he/she wants.

If the user doesn't choose any suspend mode while doing
insmod of the driver, auto suspend mode selection logic
will kick in and choose suspend mode based on the host
SDIO controller capability.

Generic module parameter is required to specify suspend
mode including Deep Sleep and WOW while doing insmod.
Renaming existing mod param variable suspend_cutpower
would be sufficient to meet this requirement.

New module parameter suspend_mode can take any one of
the below suspend state,
   1. cut power
   2. deep sleep
   3. wow
Signed-off-by: NRaja Mani <rmani@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

"ath6kl: Define a structure for connection specific aggregation information"
introduces this. In aggr_conn_init(), vif->aggr_cntxt is assigned to
aggr_conn->aggr_info, but vif->aggr_cntxt is not initialized at this
point, this would end up accessing an invalid pointer in aggregation
receive path. Fix this by passing the correct aggr_info to aggr_conn_init().
The panic trace would look like.

[<ffffffff8159e02e>] panic+0xa1/0x1c6
[<ffffffff8103773d>] ? kmsg_dump+0xfd/0x160
[<ffffffff815a2f6a>] oops_end+0xea/0xf0
[<ffffffff8102b95d>] no_context+0x11d/0x2d0
[<ffffffff8102bc5d>] __bad_area_nosemaphore+0x14d/0x230
[<ffffffff815a5c4d>] ? do_page_fault+0x30d/0x520
[<ffffffff8102bd53>] bad_area_nosemaphore+0x13/0x20
[<ffffffff815a5cfd>] do_page_fault+0x3bd/0x520
[<ffffffff8108bd60>] ? __lock_acquire+0x320/0x1680
[<ffffffff812e3a9d>] ? trace_hardirqs_off_thunk+0x3a/0x3c
[<ffffffff815a2385>] page_fault+0x25/0x30
[<ffffffffa0487a5f>] ? aggr_slice_amsdu+0xdf/0x170 [ath6kl_core]
[<ffffffffa0487bac>] aggr_deque_frms+0xbc/0x190 [ath6kl_core]
[<ffffffffa0488404>] ath6kl_rx+0x3e4/0xae0 [ath6kl_core]
[<ffffffffa047ae77>] ath6kl_htc_rxmsg_pending_handler+0x8b7/0xf10 [ath6kl_core]
[<ffffffffa00c82f0>] ? mmc_do_release_host+0x70/0x90 [mmc_core]
[<ffffffffa00c833a>] ? mmc_release_host+0x2a/0x50 [mmc_core]
[<ffffffffa04865c0>] ? ath6kl_alloc_amsdu_rxbuf+0x140/0x140 [ath6kl_core]
[<ffffffffa0477772>] ath6kl_hif_intr_bh_handler+0x362/0x510 [ath6kl_core]
[<ffffffffa01f1000>] ath6kl_sdio_irq_handler+0x60/0xb0 [ath6kl_sdio]
[<ffffffffa00d30bc>] sdio_irq_thread+0xec/0x320 [mmc_core]
[<ffffffffa00d2fd0>] ? sdio_claim_irq+0x220/0x220 [mmc_core]
[<ffffffffa00d2fd0>] ? sdio_claim_irq+0x220/0x220 [mmc_core]
[<ffffffff8105b21e>] kthread+0xbe/0xd0
[<ffffffff815ab574>] kernel_thread_helper+0x4/0x10
[<ffffffff815a2174>] ? retint_restore_args+0x13/0x13
[<ffffffff8105b160>] ? __init_kthread_worker+0x70/0x70
[<ffffffff815ab570>] ? gs_change+0x13/0x13
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Testmode (TCMD and ART) was not enabled when fw-2.bin or fw-3.bin files
were available, fix that by fetching testmode file just after the
board file but before rest of the firmware files are fetched.

I also added testmode field to struct ath6kl and moved the module parameter
to core.c. Now all module parameters are grouped in one place.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Currently rx aggregation related states are maintained per
vif, but this will not properly work when operating in AP mode.
Aggregation is completely broken when more than one
11n stations are connected to AP mode vif. Fix this issue
by keeping station specific aggregation state in sta_list.
Signed-off-by: NVasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>