The TX/RX work structures must be able to run independently
of other workqueues. This is because mac80211 might use
the flush() callback function from various context, which depends
on the TX/RX work to complete while the main thread is blocked
(until the the TX queues are empty).
This should reduce the number of 'Queue %d failed to flush' warnings.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
During device removal all pending work and tasklets must
be guaranteed to be halted. So far only the txstatus_tasklet
was killed.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Enable the txstatus tasklet when interrupts are enabled and disable it
together with the interrupts. Also make the txstatus tasklet useful even
without the tx status FIFO and make use of the generic rt2x00 tasklet
initialization macro.
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
No functional changes, just preparation for moving interrupt handling to
tasklets. The tasklets are disabled by default. Drivers making use of
them need to enable the tasklets when the device state is set to IRQ_ON.
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Allow the beacondone and pretbtt functions to update the beacon from
atomic context by using the beacon update functions with caller locking.
This is a preparation for moving the periodic beacon handling into
tasklets that require atomic context.
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch allows to dynamically remove beaconing interfaces without
shutting beaconing down on all interfaces.
The only place to start and stop beaconing are now the start- and
stop_queue callbacks. Hence, we can remove some register writes during
interface bring up (config_intf) and only write the correct sync mode
to the register there.
When multiple beaconing interfaces are present we should enable
beaconing as soon as mac80211 enables beaconing on at least one of
them. The beacon queue gets stopped when the last beaconing
interface was stopped by mac80211. Therefore, introduce another
interface counter to keep track ot the number of enabled beaconing
interfaces and start or stop the beacon queue accordingly.
To allow single interfaces to stop beaconing, add a new driver
callback clear_beacon to clear a single interface's beacon without
affecting the other interfaces. Don't overload the clear_entry callback
for clearing beacons as that would introduce additional overhead
(check for each TX queue) into the clear_entry callback which is used
on the drivers TX/RX hotpaths.
Furthermore, the write beacon callback doesn't need to enable beaconing
anymore but since beaconing should be disabled while a new beacon is
written or cleared we still disable beacon generation and enable it
afterwards again in the driver specific callbacks. However, beacon
related interrupts should not be disabled/enabled here, that's solely
done from the start- and stop queue callbacks. It would be nice to stop
the beacon queue just before the beacon update and enable it afterwards
in rt2x00queue itself instead of the current implementation that relies
on the driver doing the right thing. However, since start- and
stop_queue are mutex protected we cannot use them for atomic beacon
updates.
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Extend channel to frequency mapping for 802.11j Japan 4.9GHz band, according to
IEEE802.11 section 17.3.8.3.2 and Annex J. Because there are now overlapping
channel numbers in the 2GHz and 5GHz band we can't map from channel to
frequency without knowing the band. This is no problem as in most contexts we
know the band. In places where we don't know the band (and WEXT compatibility)
we assume the 2GHz band for channels below 14.
This patch does not implement all channel to frequency mappings defined in
802.11, it's just an extension for 802.11j 20MHz channels. 5MHz and 10MHz
channels as well as 802.11y channels have been omitted.
The following drivers have been updated to reflect the API changes:
iwl-3945, iwl-agn, iwmc3200wifi, libertas, mwl8k, rt2x00, wl1251, wl12xx.
The drivers have been compile-tested only.
Signed-off-by: Bruno Randolf <br1@einfach.org>
Signed-off-by: Brian Prodoehl <bprodoehl@gmail.com>
Acked-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Instead of protecting delayed_flags with a spinlock use atomic bitops to
make the code more readable.
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
When DRIVER_REQUIRE_TXSTATUS_FIFO is set, intialize the
txstatus_fifo, but initialize rt2x00dev->txstatus_tasklet
only when both DRIVER_REQUIRE_TXSTATUS_FIFO and
rt2x00dev->ops->lib->txstatus_tasklet are set.
This allows the txstatus_fifo to be used by rt2800usb which
does not use txstatus_tasklet.
Signed-off-by: Johannes Stezenbach <js@sig21.net>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Add a queue entry flag ENTRY_DATA_STATUS_PENDING,
which can be used to indicate a queue entry has
returned from the hardware and is waiting for
status processing. Using this flag we can add
some extra sanity checks to prevent queue corruption.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Add the rt2x00_dmastart function to rt2x00lib which
marks the queue_entry as "owned by device", and increased
the Q_INDEX number.
This cleanups up the index handling by rt2x00lib which
at until so far used hackish approaches to keep the
RX queue index numbering sane.
The rt2x00pci.c changes are from Helmut Schaa
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Add a new command to the queue handlers: "flush",
this moves the flush() callback from mac80211
into rt2x00queue and adds support for flushing
the RX queue as well.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Add wrapper functions in rt2x00queue.c to
start & stop queues. This control must be protected
using a mutex.
Queues can also be paused which will halt the flow
of packets between the driver and mac80211. This doesn't
require a mutex protection.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
As part of the queue refactoring, change the queue callback
function names to have 3 different actions: start, kick & stop.
We can now also remove the STATE_RADIO_RX_ON/STATE_RADIO_RX_OFF
device_state flags, and replace the usage with using the
start_queue/stop_queue callback functions.
This streamlines the RX queue handling to the
similar approach as all other queues.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
All rt2x00 drivers except rt2800pci call ieee80211_tx_status() from
a workqueue, which causes "NOHZ: local_softirq_pending 08" messages.
To fix it, add ieee80211_tx_status_ni() similar to ieee80211_rx_ni()
which can be called from process context, and call it from
rt2x00lib_txdone(). For the rt2800pci special case a driver
flag is introduced.
Signed-off-by: Johannes Stezenbach <js@sig21.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Put the assignment of header_length after pull out extra tx headroom
Signed-off-by: RA-Jay Hung <jay_hung@ralinktech.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
As part of the queue refactoring, the rt2x00lib_toggle_rx
can be removed and replaced with the call directly to
the set_device_state callback function.
We can remove the STATE_RADIO_RX_ON_LINK and
STATE_RADIO_RX_OFF_LINK, as it was only used for
special behavior inside rt2x00lib rather then the
drivers.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
rt2x00dev.c:689: ERROR: spaces required around that '=' (ctx:WxV)
Signed-off-by: Mark Einon <mark.einon@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
By not scheduling the TX/RX completion worker threads
when Radio is disabled, or hardware has been unplugged,
the queues cannot be completely cleaned.
This causes crashes when the hardware has been unplugged while
the radio is still enabled.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
kill_urb guarentees that when the function returns, the URB has
been fully killed. This means we don't need the extra sleeping
after the call to kill_urb.
kill_urb can however also guarentee the submit_urb to fail, as
a result, we must catch the return value from submit_urb an
correctly mark the entry as owned by the driver, and the
status as broken.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
When the RX skb allocation failed, we should recycle
the previously allocated skbuffer. By calling return
we would kill the RX queue completely since the
entry would be invalidated.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Similar to the PLCP signal and bitrates values,
we should validate the MCS value from the RX descriptor
before sending it to mac80211.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
A lot of functions accept a struct rt2x00_dev combined with
either a struct queue_entry or struct data_queue argument.
This can be simplified by only passing on the queue/entry
argument.
In cases where rt2x00_dev and a sk_buff are send together,
we can send the queue_entry instead.
rt2x00usb_alloc_urb and rt2x00usb_free_urb have a bit
of vague naming. Instead they allocate all the data which
belongs to a rt2x00 data queue entry.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch changes the way tx status reports are handled by rt2800pci.
Previously rt2800pci would sometimes lose tx status reports as the
TX_STA_FIFO register is a fifo of 16 entries that can overflow in case
we don't read it often/fast enough. Since interrupts are disabled in the
device during the execution of the interrupt thread it happend sometimes
under high network and CPU load that processing took too long and a few
tx status reports were dropped by the hw.
To fix this issue the TX_STA_FIFO register is read directly in the
interrupt handler and stored in a kfifo which is large enough to hold
all status reports of all used tx queues.
To process the status reports a new tasklet txstatus_tasklet is used.
Using the already used interrupt thread is not possible since we don't
want to disable the TX_FIFO_STATUS interrupt while processing them and
it is not possible to schedule the interrupt thread multiple times for
execution. A tasklet instead can be scheduled multiple times which
allows to leave the TX_FIFO_STATUS interrupt enabled while a previously
scheduled tasklet is still executing.
In short: All other interrupts are handled in the interrupt thread as
before. Only the TX_FIFO_STATUS interrupt is partly handled in the
interrupt handler and finished in the according tasklet.
One drawback of this patch is that it duplicates some code from
rt2800lib. However, that can be cleaned up in the future once the
rt2800usb and rt2800pci tx status handling converge more.
Using this patch on a Ralink RT3052 embedded board gives me a reliable
wireless connection even under high CPU and network load. I've
transferred several gigabytes without any queue lockups.
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The watchdog for rt2800usb triggers frequently causing all URB's
to be canceled often enough to interrupt the normal TX flow.
More research indicated that not the URB upload to the USB host
were hanging, but instead the TX status reports.
To correctly detect what is going on, we introduce Q_INDEX_DMA_DONE
which is an index counter between Q_INDEX_DONE and Q_INDEX and indicates
if the frame has been transfered to the device.
This also requires the rt2x00queue timeout functions to be updated
to differentiate between a DMA timeout (time between Q_INDEX and
Q_INDEX_DMA_DONE timeout) and a STATUS timeout (time between
Q_INDEX_DMA_DONE and Q_INDEX_DONE timeout)
All Q_INDEX_DMA_DONE code was taken from the RFC from
Helmut Schaa <helmut.schaa@googlemail.com> for the implementation
for watchdog for rt2800pci.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
During initialization each driver reads the default TX power
for each individual channel. However mac80211 only accepts the
maximum value (which is also handled as default value).
As a result, the TX power of the device was being limited to
the default value, which is often quite low compared to the
real maximum acceptable value.
This patch allows each driver to set the maximum value on a
per-channel basis which is forwarded to mac80211. The default
value will be preserved for now, in case we want to update
mac80211 to differentiate between the maximum and default txpower.
This fixes bug complaining about limited TX power values like:
https://bugzilla.kernel.org/show_bug.cgi?id=16358
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
rt2x00 was keeping a copy of ieee80211_rx_status embedded
into the rt2x00_dev structure. For each RX frame, this structure
was copied into the skb->cb where mac80211 would handle it further.
However at the moment only the fields current band, and frequency
were updated. Whereas the band was already provided directly within
the rt2x00_dev structure. Save a memcpy action, and reduce memory
a bit, by adding a curr_freq field to rt2x00_dev, and completely
remove the ieee80211_rx_status structure from rt2x00_dev.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
At some points, some drivers can't report the full TX status
information. This can happen for the UNKNOWN state, or the
FAILURE state (in case the URB failed).
Add a wrapper function to simplify reporting the
empty TX information.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Move all TX and RX completion handling into a work structure,
which is handeled on the mac80211 workqueue. This simplifies
the code in rt2x00lib since it no longer needs to check if the
device is USB or PCI to decide which mac80211 function should be used.
In the watchdog some changes are needed since it can no longer rely
on the TX completion function to be run while looping through the
entries. (Both functions now work on the same workqueue, so this
would deadlock). So the watchdog now waits for the URB to return,
and handle the TX status report directly.
As a side-effect, the debugfs entry for the RX queue now correctly
displays the positions of the INDEX and INDEX_DONE counters. This
also implies that it is not possible to perform checks like queue_empty()
and queue_full() on the RX queue.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Conflicts:
drivers/vhost/net.c
net/bridge/br_device.c
Fix merge conflict in drivers/vhost/net.c with guidance from
Stephen Rothwell.
Revert the effects of net-2.6 commit 573201f36f
since net-next-2.6 has fixes that make bridge netpoll work properly thus
we don't need it disabled.
Signed-off-by: David S. Miller <davem@davemloft.net>
The rt2x00dev->intf_work workqueue is never initialized when a driver is
probed for a non-existent device (in this case rt2500usb). On such a
path we call rt2x00lib_remove_dev() to free any resources initialized
during the probe before we use INIT_WORK to initialize the workqueue.
This causes lockdep to get confused since the lock used in the workqueue
hasn't been initialized yet but is now being acquired during
cancel_work_sync() called by rt2x00lib_remove_dev().
Fix this by initializing the workqueue first before we attempt to probe
the device. This should make lockdep happy and avoid breaking any
assumptions about how the library cleans up after a probe fails.
phy0 -> rt2x00lib_probe_dev: Error - Failed to allocate device.
INFO: trying to register non-static key.
the code is fine but needs lockdep annotation.
turning off the locking correctness validator.
Pid: 2027, comm: modprobe Not tainted 2.6.35-rc5+ #60
Call Trace:
[<ffffffff8105fe59>] register_lock_class+0x152/0x31f
[<ffffffff81344a00>] ? usb_control_msg+0xd5/0x111
[<ffffffff81061bde>] __lock_acquire+0xce/0xcf4
[<ffffffff8105f6fd>] ? trace_hardirqs_off+0xd/0xf
[<ffffffff81492aef>] ? _raw_spin_unlock_irqrestore+0x33/0x41
[<ffffffff810628d5>] lock_acquire+0xd1/0xf7
[<ffffffff8104f037>] ? __cancel_work_timer+0x99/0x17e
[<ffffffff8104f06e>] __cancel_work_timer+0xd0/0x17e
[<ffffffff8104f037>] ? __cancel_work_timer+0x99/0x17e
[<ffffffff8104f136>] cancel_work_sync+0xb/0xd
[<ffffffffa0096675>] rt2x00lib_remove_dev+0x25/0xb0 [rt2x00lib]
[<ffffffffa0096bf7>] rt2x00lib_probe_dev+0x380/0x3ed [rt2x00lib]
[<ffffffff811d78a7>] ? __raw_spin_lock_init+0x31/0x52
[<ffffffffa00bbd2c>] ? T.676+0xe/0x10 [rt2x00usb]
[<ffffffffa00bbe4f>] rt2x00usb_probe+0x121/0x15e [rt2x00usb]
[<ffffffff813468bd>] usb_probe_interface+0x151/0x19e
[<ffffffff812ea08e>] driver_probe_device+0xa7/0x136
[<ffffffff812ea167>] __driver_attach+0x4a/0x66
[<ffffffff812ea11d>] ? __driver_attach+0x0/0x66
[<ffffffff812e96ca>] bus_for_each_dev+0x54/0x89
[<ffffffff812e9efd>] driver_attach+0x19/0x1b
[<ffffffff812e9b64>] bus_add_driver+0xb4/0x204
[<ffffffff812ea41b>] driver_register+0x98/0x109
[<ffffffff813465dd>] usb_register_driver+0xb2/0x173
[<ffffffffa00ca000>] ? rt2500usb_init+0x0/0x20 [rt2500usb]
[<ffffffffa00ca01e>] rt2500usb_init+0x1e/0x20 [rt2500usb]
[<ffffffff81000203>] do_one_initcall+0x6d/0x17a
[<ffffffff8106cae8>] sys_init_module+0x9c/0x1e0
[<ffffffff8100296b>] system_call_fastpath+0x16/0x1b
Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Updating the beacon on pre tbtt instead of beacondone allows much lower
latency in regard to TIM updates. Hence, use the pre tbtt interrupt for
updating the beacon in rt2800pci (older devices don't provide a pre tbtt
interrupt).
Also, add a new driver flag to indicate if a driver has pre tbtt support
or not and implement the according behavior in rt2x00lib.
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Although mac80211 buffers broad- and mutlicast frames for us in AP mode
we still have to send them out after a DTIM beacon. Implement this
behavior by sending out the buffered frames when the beacondone
interrupt is processed.
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Since all rt2x00 PCI drivers use threaded interrupts now we don't need
to schedule a work just to update the beacon. The only place where the
beacon still gets updated in atomic context is from the set_tim
callback. Hence, move the work scheduling there.
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Use threaded interrupts for all rt2x00 PCI devices.
This has several generic advantages:
- Reduce the time we spend in hard irq context
- Use non-atmic mac80211 functions for rx/tx
Furthermore implementing broad- and multicast buffering will be
much easier in process context while maintaining low latency and
updating the beacon just before transmission (pre tbtt interrupt)
can also be done in process context.
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Implement watchdog monitoring for USB devices (PCI support can
be added later). This will determine if URBs being uploaded to
the hardware are actually returning. Both rt2500usb and rt2800usb
have shown that URBs being uploaded can remain hanging without
being released by the hardware.
By using this watchdog, a queue can be reset when this occurs.
For rt2800usb it has been tested that the connection is preserved
even though this interruption.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Now that the write_tx_data functions are merged, also merge the relevant
parts of the txdone handling into common code, rather than {usb,pci}
specific code.
Signed-off-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
In order to implement tx mpdu aggregation we only have to implement
the ampdu_action callback such that mac80211 allows negotiation of
blockack sessions.
The hardware will handle everything on its own as long as the ampdu
flag in the TXWI struct is set up correctly and we translate the tx
status correctly.
For now, refuse requests to start rx aggregation.
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
In some corner cases the reported tx rates/retries didn't match the really
used ones.
The hardware lowers the tx rate on each consecutive retry by 1 (but won't
fall back from MCS to legacy rates) _until_ it reaches the lowest one.
In case the frame wasn't sent succesful the number of retries is 7 and if
a rate index <7 was used the previous code reported negative rate indexes
which were then ignored by the rate control algorithm and mac80211.
Instead, report the remaining number of retries to have happened with
the lowest rate (index 0). This should give the rate control algorithm
slightly more accurate information about the used tx rates/retries.
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
TXDONE_FALLBACK doesn't express if the frame was sent successful or not. It
only tells us that the hw used fallback rates for retries. Hence, don't use
TXDONE_FALLBACK as success indicator.
Before this patch we reported success to the rate control algorithm which
was wrong in a number of cases and might have lead to improper tx rate
selections.
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Instead of fiddling with the skb->data pointer and thereby risking
out of bounds accesses, properly reserve the space needed in an
skb for descriptors.
Signed-off-by: Gertjan van Wingerde <gwingerde@gmail.com>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
In particular, several occurances of funny versions of 'success',
'unknown', 'therefore', 'acknowledge', 'argument', 'achieve', 'address',
'beginning', 'desirable', 'separate' and 'necessary' are fixed.
Signed-off-by: Daniel Mack <daniel@caiaq.de>
Cc: Joe Perches <joe@perches.com>
Cc: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Current rt2x00 drivers may result in a "ieee80211_tx_status: headroom too
small" error message when a frame needs to be properly aligned before
transmitting it.
This is because the space needed to ensure proper alignment isn't
requested from mac80211.
Fix this by adding sufficient amount of alignment space to the amount
of headroom requested for TX frames.
Reported-by: David Ellingsworth <david@identd.dyndns.org>
Signed-off-by: Gertjan van Wingerde <gwingerde@gmail.com>
Acked-by: Ivo van Doorn <ivdoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Ivo van Doorn