This patch implement event notification SCTP_SENDER_DRY_EVENT.
SCTP Socket API Extensions:
6.1.9. SCTP_SENDER_DRY_EVENT
When the SCTP stack has no more user data to send or retransmit, this
notification is given to the user. Also, at the time when a user app
subscribes to this event, if there is no data to be sent or
retransmit, the stack will immediately send up this notification.
iwlwifi: sanity check before counting number of tfds can be free
we use skb->data after calling ieee80211_tx_status_irqsafe(), which
could free skb instantly.
On current kernels I do not observe practical problems related with
bug, but on 2.6.35.y it cause random system hangs when stressing
wireless link, making bisection of other problems impossible.
Johannes Berg [Tue, 19 Apr 2011 18:44:04 +0000 (20:44 +0200)]
mac80211: fix SMPS debugfs locking
The locking with SMPS requests means that the
debugs file should lock the mgd mutex, not the
iflist mutex. Calls to __ieee80211_request_smps()
need to hold that mutex, so add an assertion.
This has always been wrong, but for some reason
never been noticed, probably because the locking
error only happens while unassociated.
sctp: move chunk from retransmit queue to abandoned list
If there is still data waiting to retransmit and remain in
retransmit queue, while doing the next retransmit, if the
chunk is abandoned, we should move it to abandoned list.
sctp: fix to check the source address of COOKIE-ECHO chunk
SCTP does not check whether the source address of COOKIE-ECHO
chunk is the original address of INIT chunk or part of the any
address parameters saved in COOKIE in CLOSED state. So even if
the COOKIE-ECHO chunk is from any address but with correct COOKIE,
the COOKIE-ECHO chunk still be accepted. If the COOKIE is not from
a valid address, the assoc should not be established.
sctp: bail from sctp_endpoint_lookup_assoc() if not bound
The sctp_endpoint_lookup_assoc() function uses a port hash
to lookup the association and then checks to see if any of
them are on the current endpoint. However, if the current
endpoint is not bound, there can't be any associations on
it, thus we can bail early.
sctp: remove redundant check when walking through a list of TLV parameters
When pos.v <= (void *)chunk + end - ntohs(pos.p->length) and
ntohs(pos.p->length) >= sizeof(sctp_paramhdr_t) these two expressions are all true,
pos.v <= (void *)chunk + end - sizeof(sctp_paramhdr_t) *must* be true.
This patch removes this kind of redundant check.
It's same to _sctp_walk_errors macro.
The slave member of struct aggregator does not necessarily point
to a slave which is part of the aggregator. It points to the
slave structure containing the aggregator structure, while
completely different slaves (or no slaves at all) may be part of
the aggregator.
The agg_device_up() function wrongly uses agg->slave to find the state
of the aggregator. Use agg->lag_ports->slave instead. The bug has
been introduced by commit 4cd6fe1c6483cde93e2ec91f58b7af9c9eea51ad
("bonding: fix link down handling in 802.3ad mode").
Michał Mirosław [Tue, 19 Apr 2011 06:13:10 +0000 (06:13 +0000)]
net: tun: convert to hw_features
This changes offload setting behaviour to what I think is correct:
- offloads set via ethtool mean what admin wants to use (by default
he wants 'em all)
- offloads set via ioctl() mean what userspace is expecting to get
(this limits which admin wishes are granted)
- TUN_NOCHECKSUM is ignored, as it might cause broken packets when
forwarded (ip_summed == CHECKSUM_UNNECESSARY means that checksum
was verified, not that it can be ignored)
If TUN_NOCHECKSUM is implemented, it should set skb->csum_* and
skb->ip_summed (= CHECKSUM_PARTIAL) for known protocols and let others
be verified by kernel when necessary.
sctp: Release all routes when processing acks ADD_IP or DEL_IP
When processing an ACK for ADD_IP parameter, we only release
the routes on non-active transports. This can cause a wrong
source address to be used. We can release the routes and
cause new route lookups and source address selection so that
new addresses can be used as source. Additionally, we don't need
to lookup routes for all transports at the same time. We can let
the transmit code path update the cached route when the transport
actually sends something.
We allow 0 port when adding new addresses. It only
makes sence to allow 0 port when removing addresses.
When removing the currently bound port will be used
when the port in the address is set to 0.
sctp: teach CACC algorithm about removed transports
When we have have to remove a transport due to ASCONF, we move
the data to a new active path. This can trigger CACC algorithm
to not mark that data as missing when SACKs arrive. This is
because the transport passed to the CACC algorithm is the one
this data is sitting on, not the one it was sent on (that one
may be gone). So, by sending the original transport (even if
it's NULL), we may start marking data as missing.
Ivo van Doorn [Mon, 18 Apr 2011 13:35:12 +0000 (15:35 +0200)]
rt2x00: Implement get_antenna and set_antenna callback functions
Implement the get_antenna and set_antenna callback functions, which will
allow clients to control the antenna for all non-11n hardware (Antenna handling
in rt2800 is still a bit magical, so we can't use the set_antenna for those drivers
yet).
To best support the set_antenna callback some modifications are needed in the
diversity handling. We should never look at the default antenna settings to determine
if software diversity is enabled. Instead we should set the diversity flag when
possible, which will allow the link_tuner to automatically pick up the tuning.
Ivo van Doorn [Mon, 18 Apr 2011 13:34:22 +0000 (15:34 +0200)]
rt2x00: Optimize register access in rt2800usb
All register reads/writes in rt2800usb were previously done with
rt2800_register_read/rt2800_register_write. These however indirectly
call rt2x00usb_register_read/rt2x00usb_register_write which adds an
additional overhead of at least one call and several move instructions
to each register access.
Replacing the calls to rt2800_register_read/rt2800_register_write with
direct calls to rt2x00usb_register_read/rt2x00usb_register_write gets
rid of quite a number of instructions in the drivers hotpaths (IRQ
handling and txdone handling).
For consistency replace all references to rt2800_register_read/write
with the rt2x00usb_register_read/write variants.
Helmut Schaa [Mon, 18 Apr 2011 13:34:01 +0000 (15:34 +0200)]
rt2x00: Optimize register access in rt2800pci
All register reads/writes in rt2800pci were previously done with
rt2800_register_read/rt2800_register_write. These however indirectly
call rt2x00pci_register_read/rt2x00pci_register_write which adds an
additional overhead of at least one call and several move instructions
to each register access.
Replacing the calls to rt2800_register_read/rt2800_register_write with
direct calls to rt2x00pci_register_read/rt2x00pci_register_write gets
rid of quite a number of instructions in the drivers hotpaths (IRQ
handling and txdone handling).
For consistency replace all references to rt2800_register_read/write
with the rt2x00pci_register_read/write variants.
The two functions that are in rt2x00ht.c can be much better placed
closer to the places where the call-sites of these functions are (one
in rt2x00config.c and one in rt2x00queue.c) allowing us to make these
functions static.
Also, conditional compilations doesn't seem to be necessary anymore as
802.11n support is quite common nowadays.
This makes the code a bit easier readable and searchable.
rt2x00: RT33xx device support is no longer experimental.
The rt33xx devices support for both PCI and USB devices has been in
the tree for a couple of months now, and seems to be functional and
not in a worse shape than the support for rt28xx and rt30xx devices.
No longer mark it as experimental and enable the support for these
devices by default.
Both USB and PCI drivers allow a system administrator to dynamically add
USB/PCI IDs to the device table that a driver supports via the
/sys/bus/{usb,pci,pci_express}/drivers/<driver-name>/new_id files.
However, for the rt2x00 drivers using this method currently crashes the
system with a NULL pointer failure.
This is due to the set-up of rt2x00 where the probe functions require a
rt2x00_ops structure in the driver_info field of the probed device. As
this field is empty for the dynamically added devices this fails for
these devices.
Fix this by introducing driver-specific probe wrappers that do nothing
but calling the bus-specific probe functions with the rt2x00_ops structure
as an argument, rather than depending on the driver_info field.
Helmut Schaa [Mon, 18 Apr 2011 13:31:31 +0000 (15:31 +0200)]
rt2x00: Always inline rt2x00pci_enable_interrupt
This allows the compiler to perform the necessary bitfield calculations
during compile time instead of run time and thus reduces the number of
instructions to run during each tasklet invocation. This should improve
performance in the RX hotpath.
This comes at the cost of a slight increase in the module size (for
example rt2800pci):
Before:
text data bss dec hex filename
14133 832 4 14969 3a79 drivers/net/wireless/rt2x00/rt2800pci.ko
After:
text data bss dec hex filename
14149 832 4 14985 3a89 drivers/net/wireless/rt2x00/rt2800pci.ko
Ivo van Doorn [Mon, 18 Apr 2011 13:31:02 +0000 (15:31 +0200)]
rt2x00: Decrease association time for USB devices
When powersaving is enabled, assocaition times are very high
(for WPA2 networks, the time can easily be around the 3 seconds).
This is caused, because the flushing of the queues takes
too much time. Without the flushing callback mac80211 assumes
a timeout of 100ms while scanning. Limit all flush waiting
loops to the same maximum.
We can apply this maximum by passing the drop status to the
driver, which makes sure the driver performs extra actions
during the waiting for the queue to become empty.
After these changes, association times fall within the
healthy range of ~0.6 seconds with powersaving enabled.
The difference between association time between powersaving
enabled and disabled is now only ~0.1 second (which can also
be due to the measuring method).
TX status is reported by the hardware when a packet has been
sent (or after TX failed after possible retries), which is some
time after the DMA completion. Since the rt2800usb hardware can
not signal interrupts we have to use a timer, otherwise the
TX status would only be read by the next packet's TX DMA
completion, or by the watchdog thread.
Add a timestamp to each queue entry which is updated whenever
the status of the entry changes, and remove the per-queue
timestamps. The previous check was incorrect and caused both
false positives and false negatives.
With the corrected check it comes apparent that the TX status
usually times out on rt2800usb unless there is sufficient traffic
(i.e. the next TX will complete the previous TX status).
Trying to fix the "TX status report missed" warnings
by reading the TX_STA_FIFO entries as quickly as possible.
The TX_STA_FIFO is too small in hardware, thus reading
it only from the workqueue is too slow and entries get lost.
Start an asynchronous read of the TX_STA_FIFO directly from
the TX URB completion callback (atomic context, thus it cannot
use the blocking rt2800_register_read()). If the async
read returns a valid FIFO entry, it is pushed into a larger
FIFO inside struct rt2x00_dev, until rt2800_txdone() picks
it up.
A .tx_dma_done callback is added to struct rt2x00lib_ops
to trigger the async read from the URB completion callback.
Helmut Schaa [Mon, 18 Apr 2011 13:28:04 +0000 (15:28 +0200)]
rt2x00: Use correct TBTT_SYNC config in AP mode
This seems to fix problems with some powersaving clients since a
positive value in TBTT_SYNC_CFG_TBTT_ADJUST introduces beacon skew,
which is not wanted in AP mode.
Also update the rest of the TBTT_SYNC config according to the
legacy drivers in AP mode.
Helmut Schaa [Mon, 18 Apr 2011 13:27:43 +0000 (15:27 +0200)]
rt2x00: Make rt2x00_queue_entry_for_each more flexible
Allow passing a void pointer to rt2x00_queue_entry_for_each which in
turn in provided to the callback function.
Furthermore, allow the callback function to stop processing by returning
true. And also notify the caller of rt2x00_queue_entry_for_each if the
loop was canceled by the callback.
No functional changes, just preparation for an upcoming patch.
Ivo van Doorn [Mon, 18 Apr 2011 13:27:06 +0000 (15:27 +0200)]
rt2x00: Split rt2x00dev->flags
The number of flags defined for the rt2x00dev->flags field,
has been growing over the years. Currently we are approaching
the maximum number of bits which are available in the field.
A secondary problem, is that one part of the field are initialized only
during boot, because the driver requirements are initialized or device
requirements are loaded from the EEPROM. In both cases, the flags are
fixed and will not change during device operation. The other flags are
the device state, and will change frequently. So far this resulted in the fact
that for some flags, the atomic bit accessors are used, while for the others
the non-atomic variants are used.
By splitting the flags up into a "flags" and "cap_flags" we can put all flags
which are fixed inside "cap_flags". This field can then be read non-atomically.
In the "flags" field we keep the device state, which is going to be read atomically.
This adds more room for more flags in the future, and sanitizes the field access methods.
Helmut Schaa [Mon, 18 Apr 2011 13:26:37 +0000 (15:26 +0200)]
rt2x00: Fix stuck queue in tx failure case
Since commit 0b7fde54f94979edc67bbf86b5adba702ebfefe8 "rt2x00: Protect
queue control with mutex" rt2x00 used rt2x00queue_pause_queue for
stopping a tx queue in mac80211. But in case of a failure in the tx
path rt2x00 still called ieee80211_stop_queue which stopped the queue
but prevented rt2x00queue_unpause_queue to wake the queue up again
resulting in a stuck tx queue.
Fix this by also using rt2x00queue_pause_queue in case of tx failures.
rt2x00: Enable WLAN LED on Ralink SoC (rt305x) devices
This patch adds WLAN LED support to the mac80211 rt2x00 driver for
Ralink SoC (rt305x) devices. The current WLAN LED drivers in
rt2800lib.c set the LED brightness via an MCU request, but do nothing
for SoC. This patch checks for SoC and sets the register to enable the
WLAN LED (instead of an MCU request). This enables the WLAN LED for
RT305x devices.
Felix Fietkau [Sun, 17 Apr 2011 21:28:10 +0000 (23:28 +0200)]
ath9k: assign keycache slots to unencrypted stations
Frame filtering relies on having a valid destination index (keycache slot),
to keep track of the destination. Assigning a keycache slot (configured
to unencrypted, with no key data attached) improves powersave handling in
AP mode with no encryption.
The dummy keycache entry for a station is cleared, when a real key gets
added.
Felix Fietkau [Sun, 17 Apr 2011 21:28:09 +0000 (23:28 +0200)]
ath9k: fix powersave frame filtering/buffering in AP mode
This patch fixes a long standing issue of pending packets in the queue being
sent (and retransmitted many times) to sleeping stations.
This was made worse by aggregation through driver-internal retransmitting
of A-MDPU subframes.
Previously the hardware tx filter was cleared unconditionally for every
single packet - with this patch it uses the IEEE80211_TX_CTL_CLEAR_PS_FILT
for unaggregated frames.
A sta_notify driver op is added to stop aggregation for stations when they
enter powersave mode. Subframes stay buffered inside the driver, to ensure
that the BlockAck window keeps a sane state.
Since the driver uses software aggregation, the clearing of the tx filter
needs to be handled by the driver instead of mac80211 for aggregated frames.
While leaving the oper channel, beacon generation is stopped
by mac80211 and beacon slots are marked as inactive.
During the scan, ath9k configures beacon timers
based on IEEE80211_CONF_OFFCHANNEL which inturn generates
beacon alert even though bslot is inactive.
ath9k fails to disable beacon alert while moving to offchannel
if none of the beacon slot is active. This is causing beacon
transmission on foreign channel. This patch enables swba
based on active bslots.
This issue was reported with two vifs (AP+STA) and triggered
scan in STA vif in unassociated state.
Felix Fietkau [Sun, 17 Apr 2011 15:45:00 +0000 (17:45 +0200)]
mac80211: add a function for setting the TIM bit for a specific station
This allows a driver to buffer frames for a PS station and tell mac80211
to wake it up even though mac80211 does not have any buffered frames for
it.
This is necessary for properly handling aggregation related buffering,
in ath9k, because the driver needs to keep its frames in order to keep
track of the Block-ACK window.
1) removal of unnecessary mwifiex_device structure
2) avoid passing adapter pointer to mwifiex_init_sw()
3) remove local variable drv_mode_info in mwifiex_add_card()
4) type change in mwifiex_bss_attr to match mwifiex_private
5) removal of more wordy comments
Sergei Shtylyov [Fri, 15 Apr 2011 15:23:11 +0000 (19:23 +0400)]
iwlegacy: use pci_dev->revision
Commit be663ab67077fac8e23eb8e231a8c1c94cb32e54 (iwlwifi: split the drivers for
agn and legacy devices 3945/4965) added code to read the 4965's revision ID from
the PCI configuration register while it's already stored by PCI subsystem in the
'revision' field of 'struct pci_dev'...
ath: fix 0x6C for beaconing/passive scan flags based on country IE
The 0x6C regulatory domain is just like the 0x6A regulatory
domain but differs in that 0x6C will allow adhoc and active
scan on its channels only if we are associated to an AP
with a country IE that allows those channels. The
ath_reg_apply_beaconing_flags() does just this -- we respect
the manufacturer's intent on only enabling beaconing modes
of operation if and only if blessed by the country IE.
Felix Fietkau [Thu, 14 Apr 2011 22:41:43 +0000 (00:41 +0200)]
ath9k: fix the return value of ath_stoprecv
The patch 'ath9k_hw: fix stopping rx DMA during resets' added code to detect
a condition where rx DMA was stopped, but the MAC failed to enter the idle
state. This condition requires a hardware reset, however the return value
of ath_stoprecv was 'true' in that case, which allowed it to skip the reset
when issuing a fast channel change.
netfilter: ipset: Fix the order of listing of sets
A restoreable saving of sets requires that list:set type of sets
come last and the code part which should have taken into account
the ordering was broken. The patch fixes the listing order.
Michał Mirosław [Mon, 18 Apr 2011 13:31:20 +0000 (13:31 +0000)]
net: fix section mismatches
Fix build warnings like the following:
WARNING: drivers/net/built-in.o(.data+0x12434): Section mismatch in reference from the variable madgemc_driver to the variable .init.data:madgemc_adapter_ids
And add some consts to EISA device ID tables along the way.
- the MSS value is actually contained in a 11 bits wide (0x7ff) field.
The extra bit in the former MSSMask did encompass the TSO command
bit ("LargeSend") as well (0xfff). Oops.
- the Tx descriptor layout is not the same through the whole chipset
family. The 8169 documentation, the 8168c documentation and Realtek's
drivers (8.020.00, 1.019.00, 6.014.00) highlight two layouts:
1. 8169, 8168 up to 8168b (included) and 8101
2. {8102e, 8168c} and beyond
- notwithstanding the "first descriptor" and "last descriptor" bits, the
same Tx descriptor content is enforced when a packet consists of several
descriptors. The chipsets are documented to require it.
Credits go to David Dillow <[email protected]> for the original patch.
projects / linux.git / log