Issuing multiple JOIN commands to the wl12xx's firmware, while
we're associated, might have undesired implications, so the driver
prints a message when that happens, and warn developers who check
out the source.
Update the commentary in order to consider the one valid scenario
where this can happen: roaming.
Cautiously keep the message for now, until we either gain confidence
there are no unintentional JOIN-while-associated events, or until
we move to the new multi-role fw who solves this multiple-join issue
for good.
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
The wl12xx device normally drops all frames coming from BSSID
it is not joined with.
This behavior is configured today by the wl12xx driver in response
to a handful of ieee80211_bss_change and ieee80211_conf_changed
notification flags, such as BSS_CHANGED_ASSOC, BSS_CHANGED_BSSID,
IEEE80211_CONF_CHANGE_IDLE, etc..
This breaks when we roam to a new BSSID, where authentication frames
are sent before any BSS_CHANGED/CONF_CHANGED flags are received.
When this happens the hardware silently drops the authentication
responses, and the roaming fails.
Ideally this aggressive filtering behavior of the device should be disabled
upon a notification from mac80211. Such notification will take place
after multi-channel support will be added: mac80211 will likely send a
remain-on-channel notification to drivers when entering sensitive
states (like authentication), otherwise the firmware might jump to
different channels (to serve a different role).
Until those notifications materialize, disable the hw BSSID filter
when authentication requests are sent, so roaming would work.
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
Because of the hardware recovery mechanism, its possible the
__wl1271_op_remove_interface is called twice. Currently, this leads to a
kernel crash even before a kernel WARNing can be issued.
Fix this.
Signed-off-by: Juuso Oikarinen <juuso.oikarinen@nokia.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
ELP (Extremely/Enhanced Low Power, or something like that ;)) refers to
the powerstate of the 12xx chip, in which very low power is consumed,
and no commands (from the host) can be issued until the chip is woken up.
Wakeup/sleep commands must be protected by a wl->mutex, so it's generally
a good idea to call wakeup/sleep along with the mutex lock/unlock (where
needed). However, in some places the wl12xx driver calls wakeup/sleep in
some "inner" functions. This result in some "nested" wakeup/sleep calls
which might end up letting the chip go to sleep prematurely (e.g. during
event handling).
Fix it by rearranging the elp calls to come along with mutex_lock/unlock.
Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
Initialize the channel and band from mac80211 conf even when the FW is
not yet loaded. This mitigates a bug in AP-mode where the channel was
never changed from its initial setting after FW boot and was therefore
never configured to FW.
Reported-by: Alexander Boukaty <alexanderb@ti.com>
Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
We were using an array of booleans to mark the channels we had already
scanned. This was causing a sparse error, because bool is not a type
with defined size. To fix this, use bitmasks instead, which is much
cleaner anyway.
Thanks Johannes Berg for the idea.
Signed-off-by: Luciano Coelho <coelho@ti.com>
Fix the following sparse warnings:
drivers/net/wireless/wl12xx/main.c:1129:5: warning: symbol '__wl1271_plt_stop' was not declared. Should it be static?
drivers/net/wireless/wl12xx/main.c:2988:5: warning: symbol 'wl1271_op_ampdu_action' was not declared. Should it be static?
Both functions should be static.
Signed-off-by: Luciano Coelho <coelho@ti.com>
The Soft Gemini BT load ratio value has changed its meaning with FW
version 6.1.0.0.310. It now means the passive scan compensation
percentage during A2DP EDR. Instead of 50, we need to use 200.
Fix the SG configuration accordingly.
Signed-off-by: Luciano Coelho <coelho@ti.com>
The firmware requires dummy packets to be sent using TID 7
(WL1271_TID_MGMT). Instead of hardcoding it in the tx_fill_hdr()
function, set it when creating the packet itself.
This requires Eliad's fix to set the actual TID in the TX descriptor.
Cc: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
When passing a tx frame, the driver incorrectly set desc->tid
with the ac instead of the actual tid.
It has some serious implications when using 802.11n + QoS,
as the fw starts a BlockAck with the wrong tid (which finally
cause beacon loss and disconnection / some fw crash)
Fix it by using the actual tid stored in skb->priority.
Reported-by: Shahar Levi <shahar_levi@ti.com>
Signed-off-by: Eliad Peller <eliad@wizery.com>
Reviewed-by: Juuso Oikarinen <juuso.oikarinen@nokia.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
The interface list maintained in main.c is not mutex protected. This could
cause issues, as the list is accessed from notifier chains.
Signed-off-by: Juuso Oikarinen <juuso.oikarinen@nokia.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
On wl128x based devices, when TX packets are aggregated, each packet
size must be aligned to the SDIO block size, and sent using block mode
transfers.
The block size is set to 256 bytes, which is less than the maximum
possible byte transfer. Thus, if two small packets (< 256 bytes) are
aggregated, the aggregation buffer size would be 512, and will be sent
using byte mode transfers. This can have undesired side effects.
Fix this by setting the MMC_QUIRK_BLKSZ_FOR_BYTE_MODE mmc card quirk.
For 127x chips this has no effect, as the block size is set to 512
bytes.
Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
Allow early termination of 50 consecutive beacons.
This value is the recommended one by the 12xx's system/RF team,
and tests show that power consumption is improved as expected.
Reported-by: Ruthy Zaphir <ruthyz@ti.com>
Tested-by: Danil Shalumov <danils@ti.com>
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
The wl12xx chip supports one Rx STBC spatial stream. Announce this in
the HT capabilities info field.
Signed-off-by: Shahar Levi <shahar_levi@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
The rx-status passed to mac80211 along with each received frame contains the
band on which the frame was received. Under certain circumstances, this band
information may be incorrect, causing in worst case a WARNING from mac80211,
and causes the received frame to be dropped.
This scenario mainly occurs when performing connected-mode scans, when the
received scan results are from the other band than the one currently
associated to.
[Since desc_band doesn't exist anymore, use status->band in the later
call to ieee80211_channel_to_frequency() to fix compilation -- Luca]
Signed-off-by: Juuso Oikarinen <juuso.oikarinen@nokia.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
All the new firmware versions (>=6.1.3.50.58 for STA and >=6.2.0.0.47
for AP) use 1 spare TX block. We still want to support older
firmwares that require 2 spare blocks, so added a quirk to handle the
difference.
Also implemented a generic way of setting quirks that depend on the
firmware revision.
Signed-off-by: Luciano Coelho <coelho@ti.com>
Add support to wl128x chip via chip id
Signed-off-by: Shahar Levi <shahar_levi@ti.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
Choose a different FW for AP-mode wl127x and wl128x chips, base on chip
ID at boot time.
Signed-off-by: Arik Nemtsov <arik@wizery.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
Support sending dummy packet to wl128x FW as results of
dummy packet event. That is part of dynamic TX mem blocks mechanism.
Only send dummy packet when not in AP mode.
[Even though the DUMMY_PACKET_EVENT_ID and the
STA_REMOVE_COMPLETE_EVENT_ID events are defined to the same value, we
need to treat them separately in the code. Keep the check and enable
STA_REMOVE_COMPLETE_EVENT_ID for AP mode and DUMMY_PACKET_EVENT_ID for
STA. Moved one warning to a cleaner place. Use WL1271_TID_MGMT for
dummy packets -- Luca]
Signed-off-by: Shahar Levi <shahar_levi@ti.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
Reduced bus transactions in the Tx & Rx path.
[Removed unnecessary check wl->chip.id != CHIP_ID_1283_PG20 when
checking the quirk -- Luca]
Signed-off-by: Shahar Levi <shahar_levi@ti.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
Separate the memory configuration to chip-specific structures and
implement dynamic memory for wl128x.
This feature allows us to move TX memory blocks to the RX pool when
the RX path is overloaded.
Thanks for Arik Nemtsov <arik@wizery.com> for helping simplify the
wl1271_fw_status() code.
[Rewrote the commit subject and message for clarity; improved some
comments and changed "spare" to "padding" for consistency; added a
FIXME for the AP memory configuration -- Luca]
Signed-off-by: Shahar Levi <shahar_levi@ti.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
Boot sequence support FREF clock and TCXO clock.
WL128x has two clocks input - TCXO and FREF.
TCXO is the main clock of the device, while FREF is used to sync
between the GPS and the cellular modem.
Auto-detection checks where TCXO is 32.736MHz or 16.368MHz, in that
case the FREF will be used as the WLAN/BT main clock.
[Use clock enumeration as defined in linux/wl12xx.h; remove
unnecessary else block in wl128x_switch_fref; remove unnecessary
change in main.c; remove some unnecessary debug prints and comments;
fix potential use of uninitialized value (pll_config) -- Luca]
Signed-off-by: Shahar Levi <shahar_levi@ti.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
Take care of FW & NVS with the auto-detection between wl127x and
wl128x.
[Moved some common code outside if statements and added notes about
NVS structure assumptions; Fixed a bug when checking the nvs size: if
the size was incorrect, the local nvs variable was set to NULL, it
should be wl->nvs instead. -- Luca]
[Merged with potential buffer overflow fix -- Luca]
Signed-off-by: Shahar Levi <shahar_levi@ti.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
New general and radio parameters structures and functions.
Implemented as separate functions due to auto-detection
between wl127x and wl128x.
Signed-off-by: Shahar Levi <shahar_levi@ti.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
Add the the set_block_size op in the SDIO and in the SPI modules.
Since it is only used with SDIO, just explicitly set the op to NULL in
spi.c
Signed-off-by: Luciano Coelho <coelho@ti.com>
New acx command that sets: Rx fifo enable reduced bus transactions
in RX path. Tx bus transactions padding to SDIO block size that
improve preference in Tx and essential for working with SDIO HS (48Mhz).
The max SDIO block size is 256 when working with Tx bus transactions
padding to SDIO block.
Add new ops to SDIO & SPI that handles the win size change in case of
transactions padding (relevant only for SDIO).
[Fix endianess issues; simplify sdio-specific block_size handling;
minor changes in comments; use "aligned_len" in one calculation
instead of "pad" to avoid confusion -- Luca]
Signed-off-by: Shahar Levi <shahar_levi@ti.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
Definitions to support wl128x:
- New FW file name
- Chip ID
- New PLL Configuration Algorithm macros that will be used at wl128x
boot stage
- Rename NVS macro name: wl127x and wl128x are using the same NVS
file name. However, the ini parameters between them are
different. The driver will validate the correct NVS size in
wl1271_boot_upload_nvs().
[Cleaned up some of the definitions. -- Luca]
Signed-off-by: Shahar Levi <shahar_levi@ti.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
In order to prevent overran of IRQ polarity via FW the polarity setting move after
FW download and before IRQ enable.
Signed-off-by: Shahar Levi <shahar_levi@ti.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
In queue_requests_store, the code looks like
if (rl->count[BLK_RW_SYNC] >= q->nr_requests) {
blk_set_queue_full(q, BLK_RW_SYNC);
} else if (rl->count[BLK_RW_SYNC]+1 <= q->nr_requests) {
blk_clear_queue_full(q, BLK_RW_SYNC);
wake_up(&rl->wait[BLK_RW_SYNC]);
}
If we don't satify the situation of "if", we can get that
rl->count[BLK_RW_SYNC} < q->nr_quests. It is the same as
rl->count[BLK_RW_SYNC]+1 <= q->nr_requests.
All the "else" should satisfy the "else if" check so it isn't
needed actually.
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
We do not call blk_trace_remove_sysfs() in err return path
if kobject_add() fails. This path fixes it.
Cc: stable@kernel.org
Signed-off-by: Liu Yuan <tailai.ly@taobao.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
We don't pass in a 'force_kblockd' anymore, get rid of the
stsale comment.
Reported-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
We are currently using this flag to check whether it's safe
to call into ->request_fn(). If it is set, we punt to kblockd.
But we get a lot of false positives and excessive punts to
kblockd, which hurts performance.
The only real abuser of this infrastructure is SCSI. So export
the async queue run and convert SCSI over to use that. There's
room for improvement in that SCSI need not always use the async
call, but this fixes our performance issue and they can fix that
up in due time.
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
The `try-cc' user-defined function was in tools/perf/feature-tests.mak;
this commit moves it to tools/perf/config/utilities.mak.
Signed-off-by: Michael Witten <mfwitten@gmail.com>
Link: http://lkml.kernel.org/n/tip-bqhwcuxsrve0iodn6q4ejaoi@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Currently, Python 3 is not supported by perf's code; this
can cause the build to fail for systems that have Python 3
installed as the default python:
python{,-config}
The Correct Solution is to write compatibility code so that
Python 3 works out-of-the-box.
However, users often have an ancillary Python 2 installed:
python2{,-config}
Therefore, a quick fix is to allow the user to specify those
ancillary paths as the python binaries that Makefile should
use, thereby avoiding Python 3 altogether; as an added benefit,
the Python binaries may be installed in non-standard locations
without the need for updating any PATH variable.
This commit adds the ability to set PYTHON and/or PYTHON_CONFIG
either as environment variables or as make variables on the
command line; the paths may be relative, and usually only PYTHON
is necessary in order for PYTHON_CONFIG to be defined implicitly.
Some rudimentary error checking is performed when the user
explicitly specifies a value for any of these variables.
In addition, this commit introduces significantly robust makefile
infrastructure for working with paths and communicating with the
shell; it's currently only used for handling Python, but I hope
it will prove useful in refactoring the makefiles.
Thanks to:
Raghavendra D Prabhu <rprabhu@wnohang.net>
for motivating this patch.
Acked-by: Raghavendra D Prabhu <rprabhu@wnohang.net>
Link: http://lkml.kernel.org/r/e987828e-87ec-4973-95e7-47f10f5d9bab-mfwitten@gmail.com
Signed-off-by: Michael Witten <mfwitten@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
One more installment on an area that is mostly dormant.
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Vladis Kletnieks reported a new RCU debug warning in the scheduler.
Since commit dce840a087 ("sched: Dynamically allocate sched_domain/
sched_group data-structures") the sched_domain trees are protected by
RCU instead of RCU-sched.
This means that we need to include rcu_read_lock() protection when we
iterate them since disabling preemption doesn't suffice anymore.
Reported-by: Valdis.Kletnieks@vt.edu
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1302882741.2388.241.camel@twins
Signed-off-by: Ingo Molnar <mingo@elte.hu>
When a task in a taskgroup sleeps, pick_next_task starts all the way back at
the root and picks the task/taskgroup with the min vruntime across all
runnable tasks.
But when there are many frequently sleeping tasks across different taskgroups,
it makes better sense to stay with same taskgroup for its slice period (or
until all tasks in the taskgroup sleeps) instead of switching cross taskgroup
on each sleep after a short runtime.
This helps specifically where taskgroups corresponds to a process with
multiple threads. The change reduces the number of CR3 switches in this case.
Example:
Two taskgroups with 2 threads each which are running for 2ms and
sleeping for 1ms. Looking at sched:sched_switch shows:
BEFORE: taskgroup_1 threads [5004, 5005], taskgroup_2 threads [5016, 5017]
cpu-soaker-5004 [003] 3683.391089
cpu-soaker-5016 [003] 3683.393106
cpu-soaker-5005 [003] 3683.395119
cpu-soaker-5017 [003] 3683.397130
cpu-soaker-5004 [003] 3683.399143
cpu-soaker-5016 [003] 3683.401155
cpu-soaker-5005 [003] 3683.403168
cpu-soaker-5017 [003] 3683.405170
AFTER: taskgroup_1 threads [21890, 21891], taskgroup_2 threads [21934, 21935]
cpu-soaker-21890 [003] 865.895494
cpu-soaker-21935 [003] 865.897506
cpu-soaker-21934 [003] 865.899520
cpu-soaker-21935 [003] 865.901532
cpu-soaker-21934 [003] 865.903543
cpu-soaker-21935 [003] 865.905546
cpu-soaker-21891 [003] 865.907548
cpu-soaker-21890 [003] 865.909560
cpu-soaker-21891 [003] 865.911571
cpu-soaker-21890 [003] 865.913582
cpu-soaker-21891 [003] 865.915594
cpu-soaker-21934 [003] 865.917606
Similar problem is there when there are multiple taskgroups and say a task A
preempts currently running task B of taskgroup_1. On schedule, pick_next_task
can pick an unrelated task on taskgroup_2. Here it would be better to give some
preference to task B on pick_next_task.
A simple (may be extreme case) benchmark I tried was tbench with 2 tbench
client processes with 2 threads each running on a single CPU. Avg throughput
across 5 50 sec runs was:
BEFORE: 105.84 MB/sec
AFTER: 112.42 MB/sec
Signed-off-by: Venkatesh Pallipadi <venki@google.com>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1302802253-25760-1-git-send-email-venki@google.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Make set_*_buddy() work on non-task sched_entity, to facilitate the
use of next_buddy to cache a group entity in cases where one of the
tasks within that entity sleeps or gets preempted.
set_skip_buddy() was incorrectly comparing the policy of task that is
yielding to be not equal to SCHED_IDLE. Yielding should happen even
when task yielding is SCHED_IDLE. This change removes the policy check
on the yielding task.
Signed-off-by: Venkatesh Pallipadi <venki@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1302744070-30079-2-git-send-email-venki@google.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Using ALTERNATIVE() when checking for X86_FEATURE_PERFCTR_CORE avoids
an extra pointer chase and data cache hit.
Signed-off-by: Robert Richter <robert.richter@amd.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1302913676-14352-4-git-send-email-robert.richter@amd.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Depending on the unit mask settings some FPU events may be scheduled
only on cpu counter #3. This patch fixes this.
Signed-off-by: Robert Richter <robert.richter@amd.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@googlemail.com>
Link: http://lkml.kernel.org/r/1302913676-14352-3-git-send-email-robert.richter@amd.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
With AMD cpu family 15h a unit mask was introduced for the Data Cache
Miss event (0x041/L1-dcache-load-misses). We need to enable bit 0
(first data cache miss or streaming store to a 64 B cache line) of
this mask to proper count data cache misses.
Now we set this bit for all families and models. In case a PMU does
not implement a unit mask for event 0x041 the bit is ignored.
Signed-off-by: Andre Przywara <andre.przywara@amd.com>
Signed-off-by: Robert Richter <robert.richter@amd.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1302913676-14352-2-git-send-email-robert.richter@amd.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Ignore the return value from wattrset since we ignore the return
value in nconf.gui.c as well.
scripts/kconfig/nconf.c: In function 'print_function_line':
scripts/kconfig/nconf.c:376: warning: value computed is not used
scripts/kconfig/nconf.c:380: warning: value computed is not used
scripts/kconfig/nconf.c:387: warning: value computed is not used
scripts/kconfig/nconf.c: In function 'show_menu':
scripts/kconfig/nconf.c:956: warning: value computed is not used
scripts/kconfig/nconf.c:961: warning: value computed is not used
scripts/kconfig/nconf.c:963: warning: value computed is not used
scripts/kconfig/nconf.c:965: warning: value computed is not used
Cc: Nir Tzachar <nir.tzachar@gmail.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Michal Marek <mmarek@suse.cz>
For some configurations of CONFIG_PREEMPT that is not true. So
get rid of __call_for_each_cic() and always uses the explicitly
rcu_read_lock() protected call_for_each_cic() instead.
This fixes a potential bug related to IO scheduler removal or
online switching.
Thanks to Paul McKenney for clarifying this.
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>