This flow happens when we get a failed single Tx response
on an AMPDU queue. In this case, the frame won't be sent
any more. So we need to move the window on the recipient
side. This is done by a BAR.

Now if we are in the following case: 10, 12 and 13 are ACKed
and 11 isn't.

 10  11  12  13.
 V   X   V   V

Then, 11 will be sent 16 times as an MPDU (as oppsed to
A-MPDU). If this failed, we are entering the flow described
above. So we need to send a BAR with ssn = 12.
But in this case, the scheduler will tell us to free frames
up to 13 (included).

So, it is perfectly possible to get a failed single Tx
response on an AMPDU queue that makes the scheduler's ssn
jump by more than 1 single packet.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

Make the rssi more accurate by taking in count per-chain AGC
values. Without this, the RSSI reports inaccurate values.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

Since the device is being restarted, all the Rx / Tx Block
Ack sessions are been wiped out by the driver. So ignore
the requests from mac80211 that stops Tx agg while
reconfiguring the device.

Note that stopping a non-existing Rx BA session is harmless,
so just honor mac80211's request.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

This fix removes the override of calibration request values sent
to the FW.
Due to that, the sending of default values to now implemented
calibrations is removed.
Signed-off-by: Dor Shaish <dor.shaish@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

The phy_cfg is given from the TLV value and does not have to be
built by us.
Signed-off-by: Dor Shaish <dor.shaish@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

We must set the valid TX antennas number in the ucode before
sending the phy_cfg_cmd and request for calibrations.
Signed-off-by: Dor Shaish <dor.shaish@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

This situation is clearly an error situation and the only
way to recover is to restart the driver / fw.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Reviewed-by: Ilan Peer <ilan.peer@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

When setting a monitor interface up or down, the idle state needs to be
recalculated, otherwise the hardware will just stay in its previous idle
state.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

The VHT MCSes we advertise to the AP were supposed to
be restricted to the AP, but due to a bug in the logic
mac80211 will advertise rates to the AP that aren't
even supported by the local device. To fix this skip
any adjustment if the NSS isn't supported at all.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

Introduced with de74a1d9
"mac80211: fix WPA with VLAN on AP side with ps-sta".
Apparently overwrites the sdata pointer with non-valid data in
the case of mesh.
Fix this by checking for IFTYPE_AP_VLAN.
Signed-off-by: Marco Porsch <marco@cozybit.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

Given a device with many channels capabilities the wiphy
information can still overflow even though its size in
3.9 was reduced to 3.8 levels. For new userspace and
kernel 3.10 we're going to implement a new "split dump"
protocol that can use multiple messages per wiphy.

For now though, add a workaround to be able to send more
information to userspace. Since generic netlink doesn't
have a way to set the minimum dump size globally, and we
wouldn't really want to set it globally anyway, increase
the size only when needed, as described in the comments.
As userspace might not be prepared for large buffers, we
can only use 4k.

Also increase the size for the get_wiphy command.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

Recently in commit 8a964f44
("iwlwifi: always copy first 16 bytes of commands") we fixed
the problem that the hardware writes back to the command and
that could overwrite parts of the data that was still needed
and would thus be corrupted.

Investigating this problem more closely we found that this
write-back isn't really ordered very well with respect to
other DMA traffic. Therefore, it sometimes happened that the
write-back occurred after unmapping the command again which
is clearly an issue and could corrupt the next allocation
that goes to that spot, or (better) cause IOMMU faults.

To fix this, allocate coherent memory for the first 16 bytes
of each command, containing the write-back part, and use it
for all queues. All the dynamic DMA mappings only need to be
TO_DEVICE then. This ensures that even when the write-back
happens "too late" it can't hit memory that has been freed
or a mapping that doesn't exist any more.

Since now the actual command is no longer modified, we can
also remove CMD_WANT_HCMD and get rid of the DMA sync that
was necessary to update the scratch pointer.
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

Supporting 8K A-MSDU means that we need to allocate order 1
pages for every Rx packet. Even when there is no traffic.
This adds stress on the memory manager. The handling of
compound pages is also less trivial for the memory manager
and not using them will make the allocation code run faster
although I didn't really measure.
Eric also pointed out that having huge buffers with little
data in them is not very nice towards the TCP stack since
the truesize of the skb is huge. This doesn't allow TCP
to have a big Rx window.
See https://patchwork.kernel.org/patch/2167711/ for details.

Note that very few vendors will actually send A-MSDU.
Disable this feature by default.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

The IWL_MAX_CMD_TFDS name for this constant is wrong, the
constant really indicates how many TBs we can use in the
driver for a single command TFD, rename the constant and
also add a comment explaining it.
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

The reason we mapped them bidirectionally was that not doing
so had caused IOMMU exceptions, due to the fact that the HW
writes back into the command. Now that the first part of the
command including the write-back part is always in the first
buffer, we don't need to map the remaining buffer(s) bidi
and can get rid of the special-casing for commands.

This is a requisite patch for another one to fix DMA mapping.
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

After Felix's patch it was still broken in case you
used more than just a single monitor interface. Fix
it better now.
Reported-by: Sujith Manoharan <sujith@msujith.org>
Tested-by: Sujith Manoharan <sujith@msujith.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

For SD8688, FUNC_INIT command is queued before fw_ready flag is
set. This causes the following crash as lbs_thread blocks any
command if fw_ready is not set.

[  209.338953] [<c0502248>] (__schedule+0x610/0x764) from [<bf20ae24>] (__lbs_cmd+0xb8/0x130 [libertas])
[  209.348340] [<bf20ae24>] (__lbs_cmd+0xb8/0x130 [libertas]) from [<bf222474>] (if_sdio_finish_power_on+0xec/0x1b0 [libertas_sdio])
[  209.360136] [<bf222474>] (if_sdio_finish_power_on+0xec/0x1b0 [libertas_sdio]) from [<bf2226c4>] (if_sdio_power_on+0x18c/0x20c [libertas_sdio])
[  209.373052] [<bf2226c4>] (if_sdio_power_on+0x18c/0x20c [libertas_sdio]) from [<bf222944>] (if_sdio_probe+0x200/0x31c [libertas_sdio])
[  209.385316] [<bf222944>] (if_sdio_probe+0x200/0x31c [libertas_sdio]) from [<bf01d820>] (sdio_bus_probe+0x94/0xfc [mmc_core])
[  209.396748] [<bf01d820>] (sdio_bus_probe+0x94/0xfc [mmc_core]) from [<c02e729c>] (driver_probe_device+0x12c/0x348)
[  209.407214] [<c02e729c>] (driver_probe_device+0x12c/0x348) from [<c02e7530>] (__driver_attach+0x78/0x9c)
[  209.416798] [<c02e7530>] (__driver_attach+0x78/0x9c) from [<c02e5658>] (bus_for_each_dev+0x50/0x88)
[  209.425946] [<c02e5658>] (bus_for_each_dev+0x50/0x88) from [<c02e6810>] (bus_add_driver+0x108/0x268)
[  209.435180] [<c02e6810>] (bus_add_driver+0x108/0x268) from [<c02e782c>] (driver_register+0xa4/0x134)
[  209.444426] [<c02e782c>] (driver_register+0xa4/0x134) from [<bf22601c>] (if_sdio_init_module+0x1c/0x3c [libertas_sdio])
[  209.455339] [<bf22601c>] (if_sdio_init_module+0x1c/0x3c [libertas_sdio]) from [<c00085b8>] (do_one_initcall+0x98/0x174)
[  209.466236] [<c00085b8>] (do_one_initcall+0x98/0x174) from [<c0076504>] (load_module+0x1c5c/0x1f80)
[  209.475390] [<c0076504>] (load_module+0x1c5c/0x1f80) from [<c007692c>] (sys_init_module+0x104/0x128)
[  209.484632] [<c007692c>] (sys_init_module+0x104/0x128) from [<c0008c40>] (ret_fast_syscall+0x0/0x38)

Fix it by setting fw_ready flag prior to queuing FUNC_INIT command.

Cc: <stable@vger.kernel.org> # 3.5+
Reported-by: Lubomir Rintel <lkundrak@v3.sk>
Tested-by: Lubomir Rintel <lkundrak@v3.sk>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

If CONFIG_MAC80211_MESH is not set, cfg80211 will now allow advertising
interface combinations with NL80211_IFTYPE_MESH_POINT present.
Add appropriate ifdefs to avoid running into errors.

Cc: stable@vger.kernel.org
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Maximum delay for waking up card is 50 ms. Because of typo in
counter, this delay goes to 500ms. This patch fixes the bug.

Cc: <stable@vger.kernel.org> # 3.2+
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

On many different chips, important aspects of the MAC state are not
fully cleared by a warm reset. This can show up as tx/rx hangs, those
annoying "DMA failed to stop in 10 ms..." messages or other quirks.

On AR933x, the chip can occasionally get stuck in a way that only a
driver unload/reload or a reboot would bring it back to life.

With this patch, a full reset is issued when bringing the chip out of
FULL-SLEEP state (after idle), or if either Rx or Tx was not shut down
properly. This makes the DMA related error messages disappear completely
in my tests on AR933x, and the chip does not get stuck anymore.

Cc: stable@vger.kernel.org
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

This spin lock was not initialized.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

The ath9k commit 2ef16755
(ath9k: fix signal strength reporting issues) fixed an issue where the
reported per-frame signal strength reported to mac80211 was being
overwritten with an internal average. The same issue is also present
in ath9k_htc.
In addition to preventing the driver from overwriting the value, this
commit also ensures that the internal average (which is used for ANI)
only tracks beacons of the AP that we're connected to.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Cc: stable@vger.kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>

RSSI is being stored internally as s8 in several places. The indication
of an unset RSSI value, ATH_RSSI_DUMMY_MARKER, was supposed to have been
set to 127, but ended up being set to 0x127 because of a code cleanup
mistake. This could lead to invalid signal strength values in a few
places.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Cc: stable@vger.kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>

This is another case of data increasing the size of the
wiphy information significantly with a new feature, for
now remove this as well.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

The PIC was supposed to be a small signature appended to the
PhyDB data, but the signature isn't really static and thus
attempting to check it just causes the warnings spuriously
so remove them.
Signed-off-by: Dor Shaish <dor.shaish@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

The wakeup packet in the status response is padded out
to a multiple of 4 bytes by the firmware for transfer
to the host, take that into account when checking the
length of the command.

Also, the reported wakeup packet includes the FCS but
the userspace API doesn't, so remove that. If it is a
data packet it is reported as an 802.3 packet but I
forgot to take into account and remove the encryption
head/tail, fix all of that as well.
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

Some mlme work structs are not cancelled on disassociation
nor interface deletion, which leads to them running after
the memory has been freed

There is not a clean way to cancel these in the disassociation
logic because they must be canceled outside of the ifmgd->mtx
lock, so just cancel them in mgd_stop logic that tears down
the station.

This fixes the crashes we see in 3.7.9+.  The crash stack
trace itself isn't so helpful, but this warning gives
more useful info:

WARNING: at /home/greearb/git/linux-3.7.dev.y/lib/debugobjects.c:261 debug_print_object+0x7c/0x8d()
ODEBUG: free active (active state 0) object type: work_struct hint: ieee80211_sta_monitor_work+0x0/0x14 [mac80211]
Modules linked in: [...]
Pid: 14743, comm: iw Tainted: G         C O 3.7.9+ #11
Call Trace:
 [<ffffffff81087ef8>] warn_slowpath_common+0x80/0x98
 [<ffffffff81087fa4>] warn_slowpath_fmt+0x41/0x43
 [<ffffffff812a2608>] debug_print_object+0x7c/0x8d
 [<ffffffff812a2bca>] debug_check_no_obj_freed+0x95/0x1c3
 [<ffffffff8114cc69>] slab_free_hook+0x70/0x79
 [<ffffffff8114ea3e>] kfree+0x62/0xb7
 [<ffffffff8149f465>] netdev_release+0x39/0x3e
 [<ffffffff8136ad67>] device_release+0x52/0x8a
 [<ffffffff812937db>] kobject_release+0x121/0x158
 [<ffffffff81293612>] kobject_put+0x4c/0x50
 [<ffffffff8148f0d7>] netdev_run_todo+0x25c/0x27e

Cc: stable@vger.kernel.org
Signed-off-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

Re-order the quiesce code so that timers are always
stopped before work-items are flushed. This was not
the problem I saw, but I think it may still be more
correct.
Signed-off-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

When stations are removed while packets are in the queue,
we drain the queues first, and then remove the stations.
If this happens in AP mode while the interface is removed
the MAC context might be removed from the firmware before
we removed the station(s), resulting in a SYSASSERT 3421.
This is because we remove the MAC context from the FW in
stop_ap(), but only flush the station drain work later in
remove_interface().

Refactor the code a bit to have a common MAC context
removal preparation first to solve this.
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

The FH hardware will always write back to the scratch field
in commands, even host commands not just TX commands, which
can overwrite parts of the command. This is problematic if
the command is re-used (with IWL_HCMD_DFL_NOCOPY) and can
cause calibration issues.

Address this problem by always putting at least the first
16 bytes into the buffer we also use for the command header
and therefore make the DMA engine write back into this.

For commands that are smaller than 16 bytes also always map
enough memory for the DMA engine to write back to.

Cc: stable@vger.kernel.org
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

When not using channel contexts with only monitor mode interfaces being
active, report local->monitor_chandef to userspace.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

When the driver does not want a monitor mode VIF, no channel context is
allocated for it. This causes ieee80211_recalc_idle to put the hardware
into idle mode if only a monitor mode is active, breaking injection.

Fix this by checking local->monitors in addition to active channel
contexts.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

Commit 6c17b77b (mac80211: Fix tx queue
handling during scans) contains a bug that causes off-channel frames to
get queued when they should be handed down to the driver for transmit.
Prevent this from happening.
Reported-by: Fabio Rossi <rossi.f@inwind.it>
Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

Unicast frame with unknown forwarding information always trigger
the path discovery assuming destination is always located inside the
MBSS. This patch allows the forwarding to look for mesh gate if path
discovery inside the MBSS has failed.
Reported-by: Cedric Voncken <cedric.voncken@acksys.fr>
Signed-off-by: Chun-Yeow Yeoh <yeohchunyeow@gmail.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

Just like the radar information, the TCP WoWLAN capability
data can increase the wiphy information and make it too
big. Remove the TCP WoWLAN information; no driver supports
it and new userspace tools will be required as well.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

The wiphy information is getting very close to being too
much for a typical netlink dump message and adding the
radar attributes to channels and interface combinations
can push it over the limit, which means userspace gets no
information whatsoever. Therefore, remove these again for
now, no driver actually supports radar detection anyway
and a modified userspace is required as well.

We're working on a solution that will allow userspace to
request splitting the information across multiple netlink
messages, which will allow us to add this back.

Cc: Simon Wunderlich <simon.wunderlich@s2003.tu-chemnitz.de>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

The ieee80211_beacon_add_tim() function might be called
by drivers with BHs enabled, which causes a potential
deadlock if TX happens at the same time and attempts to
lock the tim_lock as well. Use spin_lock_bh to fix it.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>