This is called with spinlocks held so we have to use GFP_ATOMIC. It's
the sc_pcu_lock in ath9k_stop() that's the issue. The call tree looks
like this:
ath9k_stop()
ath_prepare_reset()
ath_stoprecv()
ath_flushrecv()
ath_rx_tasklet()
ath9k_dfs_process_phyerr()
pd->add_pulse() => dpd_add_pulse()
channel_detector_get()
channel_detector_create()
pri_detector_init()
channel_detector_create() uses GFP_ATOMIC as well.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Tested-by: Zefir Kurtisi <zefir.kurtisi@neratec.com>
Acked-by: Zefir Kurtisi <zefir.kurtisi@neratec.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The DFS pattern detector was initially planned to reside on
a higher layer and used generic pr_*() logging functions.
Being part of ath9k, use ath_dbg() instead and make DFS log
ouput selectable via ATH_DBG_DFS (0x20000) at runtime.
This patch does not contain functional modifications.
Signed-off-by: Zefir Kurtisi <zefir.kurtisi@neratec.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Pretty much everywhere that uses a trace definition
header that's not in include/trace/events/ uses the
make system for the include path rather than putting
it into the sources, so do that in ath5k as well.
This came up during backporting work (where this is
required), but since all other drivers do it this
way upstream it seemed applicable.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Inside one FIFO slot queue, EDMA chipsets have the same link pointer
re-read race condition as older chipsets, so the same buffer holding
logic needs to be used in order to avoid use-after-free bugs.
Unlike on older chips, it can be skipped for the end of the queue.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
If AR_CRCErr, AR_PHYErr, AR_DecryptCRCErr or AR_MichaelErr is indicated
in the rx status word, but AR_RxFrameOK is also set, the descriptor
contents are typically invalid. This can show up as a warning about
invalid MCS rates in a frame. Even with those checks in place, a
descriptor with invalid MCS rates can still sometimes make it through to
the driver (mostly on older hardware like AR91xx).
Detect such errors in the last descriptor of a frame and discard the
whole frame if present.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
As the comment in ath_get_next_rx_buf indicates, if a descriptor with
the done bit set follows one with the done bit cleared, both descriptors
should be discarded, however the driver is not doing that yet.
To fix this, use the rs->rs_more flag as an indicator that the following
frame should be discarded. This also helps with the split buffer case:
if the first part of the frame is discarded, the following parts need to
be discarded as well, since they contain no valid header or usable data.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Instead of leaving the buffer without skb and breaking out of the loop
(which could leak the rx buffer), use the common error path.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Banks 0-3,7 are neither modified at run time, nor SREV dependent.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
There are two sets of initvals for this RF bank, one with TPC support and
one without.
The TPC one always gets used, so remove the other one to avoid confusion.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
These are the remaining knobs in ath9k to support DFS:
* mark AR9280 and AR9580 as DFS tested
* synchronize DFS regulatory domain to reg notifyer
* set required RX filter flags for radar detection
* process radar PHY errors at DFS detector
* notify DFS master on radar detection
DFS support requires CONFIG_ATH9K_DFS_CERTIFIED to be set.
Signed-off-by: Zefir Kurtisi <zefir.kurtisi@neratec.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This helps testing DFS without radar generating
equipment and is required for certification.
Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Signed-off-by: Zefir Kurtisi <zefir.kurtisi@neratec.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Signed-off-by: Zefir Kurtisi <zefir.kurtisi@neratec.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Should make the warning messages more useful.
Signed-off-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
ath_complete_reset will not be called if ath9k_hw_reset
is unsuccessful, so we need to re-enable intertupts to
balence the previous ath_prepare_reset call. Also schedule a
reset as a best effort method to recover the chip from
whatever state caused the channel change failure.
Fixes https://bugzilla.kernel.org/show_bug.cgi?id=55771
Signed-off-by: Robert Shade <robert.shade@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The messages are currently hard coding "1ms", which does not match
the actual timeout being used.
Signed-off-by: Robert Shade <robert.shade@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Marco Fonseca reported a issue with his carl9170 device:
"I'm seeing a problem with the carl driver. If I change channels
repeatedly on the 2.4ghz band, monitoring (e.g. tcpdump) will
eventually halt. I've seen this on various versions of the carl
driver/firmware (both from 1.9.4 to 1.9.7)"
<http://marc.info/?l=linux-wireless&m=136381302428113>
The culprit was identified as "fast channel change feature" which
according to Adrian Chadd is: "... notoriously unreliable and
really only fully debugged on some very later chips."
<http://marc.info/?l=linux-wireless&m=136416984531380>
Therefore, this patch removes the fast channel change feature.
The phy will now always have to go through a cold reset when
changing channels, but it should no longer become deaf.
Cc: Marco Fonseca <marco@tampabay.rr.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The spectral data provided via relay-fs introduces a buffering
latency given by the subbuf_size. To meet the requirements for
delay-sensitive applications (like real-time spectral plotter),
reduce subbuf_size and increase n_subbufs.
Signed-off-by: Zefir Kurtisi <zefir.kurtisi@neratec.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The following issue was reported.
WARNING: at net/mac80211/util.c:599 ieee80211_can_queue_work.isra.7+0x32/0x40 [mac80211]()
Hardware name: iMac12,1
queueing ieee80211 work while going to suspend
Pid: 0, comm: swapper/0 Tainted: PF O 3.8.2-206.fc18.x86_64 #1
Call Trace: Mar 16 09:39:17 Parags-iMac kernel: [ 3993.642992] <IRQ>
[<ffffffff8105e61f>] warn_slowpath_common+0x7f/0xc0
[<ffffffffa0581420>] ? ath_start_rx_poll+0x70/0x70 [ath9k]
<ffffffff8105e716>] warn_slowpath_fmt+0x46/0x50
[<ffffffffa045b542>] ieee80211_can_queue_work.isra.7+0x32/0x40
Fix this by avoiding to queue the work if our device has
already been marked as suspended or stopped.
Reported-by: Parag Warudkar <parag.lkml@gmail.com>
Tested-by: Parag Warudkar <parag.lkml@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: Luis R. Rodriguez <mcgrof@do-not-panic.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Drivers that don't use chanctxes cannot perform VHT association because
they still use a "backward compatibility" pair of {ieee80211_channel,
nl80211_channel_type} in ieee80211_conf and ieee80211_local.
Signed-off-by: Karl Beldan <karl.beldan@rivierawaves.com>
[fix kernel-doc]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
My new tracing code for ath6kl introduced these warnings on 64-bit:
trace.h:38:1: warning: format '%d' expects argument of type 'int',
but argument 4 has type 'size_t' [-Wformat]
trace.h:61:1: warning: format '%d' expects argument of type 'int',
but argument 4 has type 'size_t' [-Wformat]
trace.h:84:1: warning: format '%d' expects argument of type 'int',
but argument 6 has type 'size_t' [-Wformat]
trace.h:119:1: warning: format '%d' expects argument of type 'int',
but argument 7 has type 'size_t' [-Wformat]
trace.h:173:1: warning: format '%d' expects argument of type 'int',
but argument 3 has type 'size_t' [-Wformat]
trace.h:193:1: warning: format '%d' expects argument of type 'int',
but argument 5 has type 'size_t' [-Wformat]
trace.h:221:1: warning: format '%d' expects argument of type 'int',
but argument 5 has type 'size_t' [-Wformat]
Fix them by using %zd.
Reported-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The beacon and multicast-buffer queues are managed by the beacon
tasklet, and the generic tx path hang check does not help in any way
here. Running it on those queues anyway can introduce some race
conditions leading to unnecessary chip resets.
Cc: stable@vger.kernel.org
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The commit 'ath9k_hw: fix calibration issues on chainmask that don't
include chain 0' changed the hardware chainmask to the chip chainmask
for the duration of the calibration, but the revert to user
configuration in the reset path runs too early.
That causes some issues with limiting the number of antennas (including
spurious failure in hardware-generated packets).
Fix this by reverting the chainmask after the essential parts of the
calibration that need the workaround, and before NF calibration is run.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Reported-by: Wojciech Dubowik <Wojciech.Dubowik@neratec.com>
Tested-by: Wojciech Dubowik <Wojciech.Dubowik@neratec.com>
Cc: stable@vger.kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
There are a number of situations in which mac80211 only
really needs to flush queues for one virtual interface,
and in fact during this frames might be transmitted on
other virtual interfaces. Calculate and pass a queue
bitmap to the driver so it knows which queues to flush.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Either first 3 bytes of the first received tcp segment or last one
over MTU size file can be loss due to the byte alignment problem.
Although ATH6KL_HTC_ALIGN_BYTES was defined for 'extra bytes for htc header
alignment' in the patch "Fix buffer alignment for scatter-gather
I/O"(1df94a857), there exists the bytes loss issue which means that it will be
truncated 3 bytes in the transmitted file contents if a file which has over MTU
size is transferred through TCP/IP stack. It doesn't look like TCP/IP stack
bug of 3.5 or the latest version of kernel but the byte alignment issue. This
patch is to use the roundup() function for the byte alignment rather than the
predefined ATH6KL_HTC_ALIGN_BYTES.
kvalo: fixed indentation
Signed-off-by: Myoungje Kim <mjei78@gmail.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Dan found a check from ath6kl_rx() which doesn't make any sense at all:
" 1327 if (status || !(skb->data + HTC_HDR_LENGTH)) {
^^^^^^^^^^^^^^^^^^^^^^^^^^
skb->data is a pointer. This pointer math is always going to be false.
Should it be testing "packet->act_len < HTC_HDR_LENGTH" or something?"
I don't know what the check really was supposed to do, but I think Dan's guess
is right. Fix it accordingly.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Dan reported that smatch found a possible issue in ath6kl_wmi_beginscan_cmd()
where we might access sc->supp_rates beyond the end. It shouldn't happen as
ar->wiphy->bands always have just the first two bands set, but add an extra
check just to be sure.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Now all log messages are sent through the tracing infrastruture as well.
Tracing point doesn't follow debug_mask module parameter, instead it sends
all debug messages, so once you enable ath6kl_log_dbg tracing point you will
get a lot of messages. Needs to be discussed if this is sensible or not.
The overhead should be small enough and we anyway include debug level as
well so it's easy to filter in user space.
I wasn't really sure what to do with ath6kl_dbg_dump() and for now decided
that it also sends the buffer to user space. But most likely in the future
ath6kl_dbg_dump() should go away in favor of using proper tracing points, but
we will see.
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Add tracing points for htc layer, just dumping the packets to user space.
I wasn't really sure what to do with the status value, it might not always
be accurate, but I included it anyway.
I skipped htc_pipe (and usb) implementation for now. Need to add those
tracepoints later.
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Add a tracing point for hif irq and dump the register content to user space.
This is in hif.c as we could use the same code also with SPI but, as ath6kl
doesn't SPI and most likely never will be, this is used just by SDIO so
name the trace point as ath6kl_sdio_irq to make it easier to manage filters.
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Add basic tracing infrastructure support to ath6kl and which can be
enabled with CONFIG_ATH6KL_TRACING.
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Julien reported that ar6004 usb device fails to initialise
after host has been rebooted and power is still on for the ar6004 device. He
found out that doing a cold reset fixes the issue.
I wasn't sure what would be the best way to detect if target needs a reset so I
settled on checking a timeout from htc_wait_recv_ctrl_message().
Reported-by: Julien Massot <jmassot@aldebaran-robotics.com>
Tested-by: Julien Massot <jmassot@aldebaran-robotics.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
It was annoying to debug usb warm reboot initialisation problems as many usb
related functions just ignored errors and it wasn't obvious from the kernel
logs what was failing. Fix all that so that error messages are printed and
errors are handled properly.
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Printin the firmware capabilities during the first firmware boot makes it easier to find out what
features firmware supports.
Obligatory screenshot:
[21025.678481] ath6kl: ar6003 hw 2.1.1 sdio fw 3.2.0.144 api 3
[21025.678667] ath6kl: firmware supports: sched-scan,sta-p2pdev-duplex,rsn-cap-override
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Firmware got support for temperature measurement.
There are 2 temperature sensors: MAC and radio
"not available" temperature - reported by FW as 0 or ~0
Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Adjust driver for changes in the FW API.
Noticeable changes in the FW are:
- temperature sensing
- infrastructure for multiple connections
- infrastructure for P2P
- signal strength indication
This commit introduces only changes that are required to support same functionality
as previous firmware, no new features.
Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
New firmware allows for shorter SW reset procedure.
After SW reset, FW raises "fw done" IRQ, at this
moment mailbox control structures are initialized, driver caches it.
New status bit wil_status_reset_done introduced to track completion
of the reset. It is set by "fw ready" irq, and required for WMI rx flow
to access control structures.
WMI Tx flow protected by other status bit, wil_status_fwready. It can't
be set before wil_status_reset_done is set by design.
Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>