| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
commit 62a8ff67eba52dae9b107e1fb8827054ed00a265 upstream.
In certain scenarios a normal MSDU can be received as an A-MSDU when
the A-MSDU present bit of a QoS header gets flipped during reception.
Since this bit is unauthenticated, the hardware crypto engine can pass
the frame to the driver without any error indication.
This could result in processing unintended subframes collected in the
A-MSDU list. Hence, validate A-MSDU list by checking if the first frame
has a valid subframe header.
Comparing the non-aggregated MSDU and an A-MSDU, the fields of the first
subframe DA matches the LLC/SNAP header fields of a normal MSDU.
In order to avoid processing such frames, add a validation to
filter such A-MSDU frames where the first subframe header DA matches
with the LLC/SNAP header pattern.
Tested-on: QCA9984 hw1.0 PCI 10.4-3.10-00047
Cc: stable@vger.kernel.org
Signed-off-by: Sriram R <srirrama@codeaurora.org>
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Link: https://lore.kernel.org/r/20210511200110.e6f5eb7b9847.I38a77ae26096862527a5eab73caebd7346af8b66@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 99f41b8e43b8b4b31262adb8ac3e69088fff1289 ]
When STBC is enabled, NSTS_SU value need to be accounted for VHT NSS
calculation for SU case.
Without this fix, 1SS + STBC enabled case was reported wrongly as 2SS
in radiotap header on monitor mode capture.
Tested-on: QCA9984 10.4-3.10-00047
Signed-off-by: Sathishkumar Muruganandam <murugana@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597392971-3897-1-git-send-email-murugana@codeaurora.org
Signed-off-by: Sasha Levin <sashal@kernel.org>
|
|
[ Upstream commit bad60b8d1a7194df38fd7fe4b22f3f4dcf775099 ]
The idx in __ath10k_htt_rx_ring_fill_n function lives in
consistent dma region writable by the device. Malfunctional
or malicious device could manipulate such idx to have a OOB
write. Either by
htt->rx_ring.netbufs_ring[idx] = skb;
or by
ath10k_htt_set_paddrs_ring(htt, paddr, idx);
The idx can also be negative as it's signed, giving a large
memory space to write to.
It's possibly exploitable by corruptting a legit pointer with
a skb pointer. And then fill skb with payload as rougue object.
Part of the log here. Sometimes it appears as UAF when writing
to a freed memory by chance.
[ 15.594376] BUG: unable to handle page fault for address: ffff887f5c1804f0
[ 15.595483] #PF: supervisor write access in kernel mode
[ 15.596250] #PF: error_code(0x0002) - not-present page
[ 15.597013] PGD 0 P4D 0
[ 15.597395] Oops: 0002 [#1] SMP KASAN PTI
[ 15.597967] CPU: 0 PID: 82 Comm: kworker/u2:2 Not tainted 5.6.0 #69
[ 15.598843] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
BIOS rel-1.12.1-0-ga5cab58e9a3f-prebuilt.qemu.org 04/01/2014
[ 15.600438] Workqueue: ath10k_wq ath10k_core_register_work [ath10k_core]
[ 15.601389] RIP: 0010:__ath10k_htt_rx_ring_fill_n
(linux/drivers/net/wireless/ath/ath10k/htt_rx.c:173) ath10k_core
Signed-off-by: Zekun Shen <bruceshenzk@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200623221105.3486-1-bruceshenzk@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
|
|
[ Upstream commit 168f75f11fe68455e0d058a818ebccfc329d8685 ]
While debugging driver crashes related to a buggy firmware
crashing under load, I noticed that ath10k_htt_rx_ring_free
could be called without being under lock. I'm not sure if this
is the root cause of the crash or not, but it seems prudent to
protect it.
Originally tested on 4.16+ kernel with ath10k-ct 10.4 firmware
running on 9984 NIC.
Signed-off-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit 7eccb738fce57cbe53ed903ccf43f9ab257b15b3 upstream.
Rx data frames notified through HTT_T2H_MSG_TYPE_RX_IND and
HTT_T2H_MSG_TYPE_RX_FRAG_IND expect PN/TSC check to be done
on host (mac80211) rather than firmware. Rebuild cipher header
in every received data frames (that are notified through those
HTT interfaces) from the rx_hdr_status tlv available in the
rx descriptor of the first msdu. Skip setting RX_FLAG_IV_STRIPPED
flag for the packets which requires mac80211 PN/TSC check support
and set appropriate RX_FLAG for stripped crypto tail. Hw QCA988X,
QCA9887, QCA99X0, QCA9984, QCA9888 and QCA4019 currently need the
rebuilding of cipher header to perform PN/TSC check for replay
attack.
Please note that removing crypto tail for CCMP-256, GCMP and GCMP-256 ciphers
in raw mode needs to be fixed. Since Rx with these ciphers in raw
mode does not work in the current form even without this patch and
removing crypto tail for these chipers needs clean up, raw mode related
issues in CCMP-256, GCMP and GCMP-256 can be addressed in follow up
patches.
Tested-by: Manikanta Pubbisetty <mpubbise@qti.qualcomm.com>
Signed-off-by: Vasanthakumar Thiagarajan <vthiagar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Chipset from QCA99X0 onwards (QCA99X0, QCA9984, QCA4019 & future)
rx_hdr_status is not padded to align in 4-byte boundary. Define a
new hw_params field to handle different alignment behaviour between
different hw. This patch fixes improper retrieval of rfc1042 header
with QCA4019. This patch along with "ath10k: Properly remove padding
from the start of rx payload" will fix traffic failure in ethernet
decap mode for QCA4019.
Signed-off-by: Vasanthakumar Thiagarajan <vthiagar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
In QCA99X0 (QCA99X0, QCA9984, QCA9888 and QCA4019) family chips,
hw adds padding at the begining of the rx payload to make L3
header 4-byte aligned. In the chips doing this type of padding,
the number of bytes padded will be indicated through msdu_end:info1.
Define a hw_rx_desc_ops wrapper to retrieve the number of padded
bytes and use this while doing undecap. This should fix padding
related issues with ethernt decap format with QCA99X0, QCA9984,
QCA9888 and QCA4019 hw.
Signed-off-by: Vasanthakumar Thiagarajan <vthiagar@qti.qualcomm.com>
[Rename operations to hw_ops for other purposes]
Signed-off-by: Benjamin Berg <benjamin@sipsolutions.net>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
Commit 3c97f5de1f28 ("ath10k: implement NAPI support") conflicts with
ath-current. To avoid any merge problems merge ath-current to ath-next already now.
Conflicts:
drivers/net/wireless/ath/ath10k/htt_rx.c
|
|
commit 7a0adc83f34d ("ath10k: improve tx scheduling") is causing
severe throughput drop in multi client mode. This issue is originally
reported in veriwave setup with 50 clients with TCP downlink traffic.
While increasing number of clients, the average throughput drops
gradually. With 50 clients, the combined peak throughput is decreased
to 98 Mbps whereas reverting given commit restored it to 550 Mbps.
Processing txqs for every tx completion is causing overhead. Ideally for
management frame tx completion, pending txqs processing can be avoided.
The change partly reverts the commit "ath10k: improve tx scheduling".
Processing pending txqs after all skbs tx completion will yeild enough
room to burst tx frames.
Fixes: 7a0adc83f34d ("ath10k: improve tx scheduling")
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
Add NAPI support for rx and tx completion. NAPI poll is scheduled
from interrupt handler. The design is as below
- on interrupt
- schedule napi and mask interrupts
- on poll
- process all pipes (no actual Tx/Rx)
- process Rx within budget
- if quota exceeds budget reschedule napi poll by returning budget
- process Tx completions and update budget if necessary
- process Tx fetch indications (pull-push)
- push any other pending Tx (if possible)
- before resched or napi completion replenish htt rx ring buffer
- if work done < budget, complete napi poll and unmask interrupts
This change also get rid of two tasklets (intr_tq and txrx_compl_task).
Measured peak throughput with NAPI on IPQ4019 platform in controlled
environment. No noticeable reduction in throughput is seen and also
observed improvements in CPU usage. Approx. 15% CPU usage got reduced
in UDP uplink case.
DL: AP DUT Tx
UL: AP DUT Rx
IPQ4019 (avg. cpu usage %)
========
TOT +NAPI
=========== =============
TCP DL 644 Mbps (42%) 645 Mbps (36%)
TCP UL 673 Mbps (30%) 675 Mbps (26%)
UDP DL 682 Mbps (49%) 680 Mbps (49%)
UDP UL 720 Mbps (28%) 717 Mbps (11%)
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
With the %pK format specifier we hide the kernel addresses
with the help of kptr_restrict sysctl.
In this patch, %p is changed to %pK in the driver code.
The sysctl is documented in Documentation/sysctl/kernel.txt.
Signed-off-by: Maharaja Kennadyrajan <c_mkenna@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
On handling amsdu on rx path, get the rx_status from htt context. Without this
fix, we are seeing warnings when running DBDC traffic like this.
WARNING: CPU: 0 PID: 0 at net/mac80211/rx.c:4105 ieee80211_rx_napi+0x88/0x7d8 [mac80211]()
[ 1715.878248] CPU: 0 PID: 0 Comm: swapper/0 Tainted: G W 3.18.21 #1
[ 1715.878273] [<c001d3f4>] (unwind_backtrace) from [<c001a4b0>] (show_stack+0x10/0x14)
[ 1715.878293] [<c001a4b0>] (show_stack) from [<c01bee64>] (dump_stack+0x70/0xbc)
[ 1715.878315] [<c01bee64>] (dump_stack) from [<c002a61c>] (warn_slowpath_common+0x64/0x88)
[ 1715.878339] [<c002a61c>] (warn_slowpath_common) from [<c002a6d0>] (warn_slowpath_null+0x18/0x20)
[ 1715.878395] [<c002a6d0>] (warn_slowpath_null) from [<bf4caa98>] (ieee80211_rx_napi+0x88/0x7d8 [mac80211])
[ 1715.878474] [<bf4caa98>] (ieee80211_rx_napi [mac80211]) from [<bf568658>] (ath10k_htt_t2h_msg_handler+0xb48/0xbfc [ath10k_core])
[ 1715.878535] [<bf568658>] (ath10k_htt_t2h_msg_handler [ath10k_core]) from [<bf568708>] (ath10k_htt_t2h_msg_handler+0xbf8/0xbfc [ath10k_core])
[ 1715.878597] [<bf568708>] (ath10k_htt_t2h_msg_handler [ath10k_core]) from [<bf569160>] (ath10k_htt_txrx_compl_task+0xa54/0x1170 [ath10k_core])
[ 1715.878639] [<bf569160>] (ath10k_htt_txrx_compl_task [ath10k_core]) from [<c002db14>] (tasklet_action+0xb4/0x130)
[ 1715.878659] [<c002db14>] (tasklet_action) from [<c002d110>] (__do_softirq+0xe0/0x210)
[ 1715.878678] [<c002d110>] (__do_softirq) from [<c002d4b4>] (irq_exit+0x84/0xe0)
[ 1715.878700] [<c002d4b4>] (irq_exit) from [<c005a544>] (__handle_domain_irq+0x98/0xd0)
[ 1715.878722] [<c005a544>] (__handle_domain_irq) from [<c00085f4>] (gic_handle_irq+0x38/0x5c)
[ 1715.878741] [<c00085f4>] (gic_handle_irq) from [<c0009680>] (__irq_svc+0x40/0x74)
[ 1715.878753] Exception stack(0xc05f9f50 to 0xc05f9f98)
[ 1715.878767] 9f40: ffffffed 00000000 00399e1e c000a220
[ 1715.878786] 9f60: 00000000 c05f6780 c05f8000 00000000 c05f5db8 ffffffed c05f8000 c04d1980
[ 1715.878802] 9f80: 00000000 c05f9f98 c0018110 c0018114 60000013 ffffffff
[ 1715.878822] [<c0009680>] (__irq_svc) from [<c0018114>] (arch_cpu_idle+0x2c/0x50)
[ 1715.878844] [<c0018114>] (arch_cpu_idle) from [<c00530d4>] (cpu_startup_entry+0x108/0x234)
[ 1715.878866] [<c00530d4>] (cpu_startup_entry) from [<c05c7be0>] (start_kernel+0x33c/0x3b8)
[ 1715.878879] ---[ end trace 6d5e1cc0fef8ed6a ]---
[ 1715.878899] ------------[ cut here ]------------
Fixes: 18235664e7f9 ("ath10k: cleanup amsdu processing for rx indication")
Signed-off-by: Ashok Raj Nagarajan <arnagara@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
ath.git patches for 4.8. Major changes:
ath10k
* enable support for QCA9888
|
|
It is expected that all pktlog events for 10.4 firmware based solutions
should come through CE8 where as in case of 10.2 firmware based solutions,
it should come through one of the HTT events (HTT_T2H_MSG_TYPE_PKTLOG).
But from experiments with 10.4 based solutions, it is observed that pktlog
event for ATH_PKTLOG_TYPE_TX_MSDU_ID is coming through HTT pktlog event.
Currently, we always parse with 10.2 pktlog header which will lead to
pktlog decoding issues (payload length mismatch exceptions)
For trace points, it is required to provide only the payload size. So
fixing this by simplifying the payload size calculation without the use of
ath10k pktlog headers.
While there, remove the unused ath10k pktlog headers.
Signed-off-by: Ashok Raj Nagarajan <arnagara@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
The below warning message seems to hit occasionally with the following
combination (IPQ4019 + ACS scan) where we receive packets as a self peer
when hostapd does ACS when we bring up AP mode . ath10k has the below
fall back mechanism to fetch current operating channel in rx (it will
check for the next channel tracking variable if the current one is NULL)
[scan channel] --> [rx channel] --> [peer channel] -->
[vdev channel] --> [any vdev channel] --> [target oper channel]
'scan channel' and 'target operating channel' are directly fetched from
firmware events. All the others should be updated by mac80211.
During ACS scan we wouldn't have a valid channel context
assigned from mac80211 ('ar->rx_channel'), and also relying on
('ar->scan_channel') is not helpful (it becomes NULL when it goes to
BSS channel and also when the scan event is completed). In short we
cannot always rely on these two channel tracking variables.
'Target Operating Channel' (ar->tgt_oper_chan) seems to keep track of
the current operating even while we are doing ACS scan and etc. Hence
remove this un-necessary warning message and continue with
target_operating channel. At the worst case scenario when the target
operating channel is invalid (NULL) we already have an ath10k warning
message to notify we really don't have a proper channel configured in
rx to update the rx status("no channel configured; ignoring frame(s)!")
WARNING: CPU: 0 PID: 0 at ath/ath10k/htt_rx.c:803
[<c0318838>] (warn_slowpath_null) from [<bf4a0104>]
(ath10k_htt_rx_h_channel+0xe0/0x1b8 [ath10k_core])
[<bf4a0104>] (ath10k_htt_rx_h_channel [ath10k_core]) from
[<bf4a025c>] (ath10k_htt_rx_h_ppdu+0x80/0x288 [ath10k_core])
[<bf4a025c>] (ath10k_htt_rx_h_ppdu [ath10k_core]) from
[<bf4a1a9c>] (ath10k_htt_txrx_compl_task+0x724/0x9d4 [ath10k_core])
[<bf4a1a9c>] (ath10k_htt_txrx_compl_task [ath10k_core])
Fixes:3b0499e9ce42 ("ath10k: reduce warning messages during rx without proper channel context")
Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
Several cases of overlapping changes, except the packet scheduler
conflicts which deal with the addition of the free list parameter
to qdisc_enqueue().
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
commit 5c86d97bcc1d ("ath10k: combine txrx and replenish task")
introduced deadlock while processing rx in order indication message
for qca6174 based devices. While merging replenish and txrx tasklets,
replenish task should be called out of htt rx ring locking since it
is also try to acquire the same lock.
Unfortunately this issue is not exposed by other solutions (qca988x,
qca99x0 & qca4019), as rx_in_ord_ind message is specific to qca6174
based devices. This patch fixes
=============================================
[ INFO: possible recursive locking detected ]
4.7.0-rc2-wt-ath+ #1353 Tainted: G E
---------------------------------------------
swapper/3/0 is trying to acquire lock:
(&(&htt->rx_ring.lock)->rlock){+.-...}, at: [<f8d7ef19>]
ath10k_htt_rx_msdu_buff_replenish+0x29/0x90 [ath10k_core]
but task is already holding lock:
(&(&htt->rx_ring.lock)->rlock){+.-...}, at: [<f8d82cab>]
ath10k_htt_txrx_compl_task+0x21b/0x250 [ath10k_core]
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0
----
lock(&(&htt->rx_ring.lock)->rlock);
lock(&(&htt->rx_ring.lock)->rlock);
*** DEADLOCK ***
May be due to missing lock nesting notation
1 lock held by swapper/3/0:
#0: (&(&htt->rx_ring.lock)->rlock){+.-...}, at: [<f8d82cab>]
ath10k_htt_txrx_compl_task+0x21b/0x250 [ath10k_core]
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=119151
Fixes: 5c86d97bcc1d ("ath10k: combine txrx and replenish task")
Reported-by: Mike Lothian <mike@fireburn.co.uk>
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
Remove unused inline function phy_mode_to_band.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
Legacy rate packets may not necessarily be having a rx status
flag of '0' always, for example management frame have flags
like RX_FLAG_ONLY_MONITOR / RX_FLAG_MACTIME_END also set
Just check 'VHT' and 'HT' flags are not set , and simply clasify it as
legacy rate packets
Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
WARN_ON_ONCE when we receive packets for self peer when mac80211
had not assigned a proper channel context. This scenario happens
in QCA4019 when we start the AP via hostapd in background and start
it once again in the background without killing the previous instance!
This happens intermittently when we start / stop hostapd in a while loop
(incase the hostapd is not properly killed). This results in mac80211
chancontext to be unassigned, while the self peer continuous receive
packets in target operating channel. This results in lot of call traces
in the rx path. Make this as a WARN_ON_ONCE to avoid flooding the console
which result in rebooting low memory systems, while still reporting the
warning once that we are receiving packets in target operating channel and
to indicate that something is happening which is not the expected result.
WARNING: CPU: 0 PID: 0 at ath/ath10k/htt_rx.c:803
[<c0318838>] (warn_slowpath_null) from [<bf4a0104>]
(ath10k_htt_rx_h_channel+0xe0/0x1b8 [ath10k_core])
[<bf4a0104>] (ath10k_htt_rx_h_channel [ath10k_core]) from
[<bf4a025c>] (ath10k_htt_rx_h_ppdu+0x80/0x288 [ath10k_core])
[<bf4a025c>] (ath10k_htt_rx_h_ppdu [ath10k_core]) from
[<bf4a1a9c>] (ath10k_htt_txrx_compl_task+0x724/0x9d4 [ath10k_core])
[<bf4a1a9c>] (ath10k_htt_txrx_compl_task [ath10k_core])
Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
Recent changes revolving around implementing
wake_tx_queue support introduced a significant
performance regressions on some (slower, uni-proc)
systems.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
ath.git patches for 4.7. Major changes:
ath10k
* implement set_tsf() for 10.2.4 branch
* remove rare MSI range support
* remove deprecated firmware API 1 support
ath9k
* add module parameter to invert LED polarity
wcn36xx
* fixes to get the driver properly working on Dragonboard 410c
|
|
Preparation for testmode.c to use ath10k_core_fetch_board_data_api_n().
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
This enum is already perfectly aliased to enum nl80211_band, and
the only reason for it is that we get IEEE80211_NUM_BANDS out of
it. There's no really good reason to not declare the number of
bands in nl80211 though, so do that and remove the cfg80211 one.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/wireless-drivers-next
Kalle Valo says:
====================
wireless-drivers patches for 4.7
Major changes:
iwlwifi
* support for Link Quality measurement
* more work 9000 devices and MSIx
* continuation of the Dynamic Queue Allocation work
* make the paging less memory hungry
* 9000 new Rx path
* removal of IWLWIFI_UAPSD Kconfig option
ath10k
* implement push-pull tx model using mac80211 software queuing support
* enable scan in AP mode (NL80211_FEATURE_AP_SCAN)
wil6210
* add basic PBSS (Personal Basic Service Set) support
* add initial P2P support
* add oob_mode module parameter
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Decrement num_mpdus_ready only when rx amsdu is processed successfully.
Not doing so, will result in leak and impact stabilty under low memory
cases. Also commit 3128b3d8a2b9 ("ath10k: speedup htt rx descriptor
processing for rx_ind") missed to removed unused skb list rx_q.
Fixes: 3128b3d8a2b9 ("ath10k: speedup htt rx descriptor processing for rx_ind")
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
When HW crypto is used, there's no need for the CCMP/GCMP MIC to
be available to mac80211, and the hardware might have removed it
already after checking. The MIC is also useless to have when the
frame is already decrypted, so allow indicating that it's not
present.
Since we are running out of bits in mac80211_rx_flags, make
the flags field a u64.
Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Since tx completion and rx indication processing are moved out
of txrx tasklet and rx ring lock contention also removed from txrx
for rx_ind messages, it would be efficient to combine both replenish
and txrx tasks. Refill threshold is adjusted for both AP135 and AP148
(low and high end systems). With this adjustment in AP135, TCP DL is
improved from 603 Mbps to 620 Mbps and UDP DL is improved from 758 Mbps
to 803 Mbps. Also no watchdog are observed on UDP BiDi.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
Except qca61x4 family chips (qca6164, qca6174), copy engine 5 is used
for receiving target to host htt messages. In follow up patch, CE5
descriptors will be reused. In such case, same API can not be used as
htc layer callback where the response messages will be freed at the end.
Hence register new API for HTC layer that free up received message and
keep the message handler common for both HTC and HIF layers.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
In follow up patch, htt rx descriptors will be reused instead of
dealloc and refill. To achieve that htt rx indication messages
should not be deferred and should be processed in pci tasklet itself.
Also from rx indication message, mpdu_count alone is used. So it is
maintained as atomic variable and all rx amsdu handlers are done
processed from txrx tasklet. This change get rid of rx_compl_q usage.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
Make amsdu handlers (i.e amsdu_pop and rx_h_handler) common to both
rx_ind and frag_ind htt events. It is sufficient to hold rx_ring lock
for amsdu_pop alone and no need to hold it until the packets are
delivered to mac80211. This helps to reduce rx_lock contention as well.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
The fw descriptor was never used and probably never will be. It makes
little sense to maintain support for it. Remove it and simplify rx
processing. This will make it easier to optimize rx processing later
as well.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
To optmize CPU usage htt rx descriptors will be reused instead of
refilling it for htt rx copy engine (CE5). To support that all htt rx
indications should be proecssed at same context. Instead of queueing
actual indication message, queue copied message for txrx processing.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
To optimize CPU usage htt rx descriptors will be reused instead of
refilling it for htt rx copy engine (CE5). To support that all htt rx
indications should be processed at same context. FIFO queue is used
to maintain tx completion status for each msdu. This helps to retain
the order of tx completion.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
Frames that are transmitted via MGMT_TX are using reserved descriptor
slots in firmware. This limitation is for the htt_mgmt_tx path itself,
not for mgmt frames per se. In 16 MBSSID scenario, these reserved slots
will be easy exhausted due to frequent probe responses. So for 10.4
based solutions, probe responses are limited by a threshold (24).
management tx path is separate for all except tlv based solutions. Since
tlv solutions (qca6174 & qca9377) do not support 16 AP interfaces, it is
safe to move management descriptor limitation check under mgmt_tx
function. Though CPU improvement is negligible, unlikely conditions or
never hit conditions in hot path can be avoided on data transmission.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
Whenever firmware is configuring operating channel during scan or
home channel, channel change event will be indicated to host. In some
cases (device probe/ last vdev down), target will be configured to
default channel whereas host is unaware of target's operating channel.
This leads to packet drop due to unknown channel and kernel log will be
filled up with "no channel configured; ignoring frame(s)!". Fix that
by handling HTT_T2H_MSG_TYPE_CHAN_CHANGE event.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
Until now only WMI originating mgmt frames were
reported to mac80211. Management frames on HTT
were basically dropped (except frames which looked
like management but had FCS error).
To allow sniffing all frames (including offloaded
frames) without interfering with mac80211
operation and states a new rx_flag was introduced
and is not being used to distinguish frames and
classify them for mac80211.
Signed-off-by: Grzegorz Bajorski <grzegorz.bajorski@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
The current/old tx path design was that host, at
its own leisure, pushed tx frames to the device.
For HTT there was ~1000-1400 msdu queue depth.
After reaching that limit the driver would request
mac80211 to stop queues. There was little control
over what packets got in there as far as
DA/RA was considered so it was rather easy to
starve per-station traffic flows.
With MU-MIMO this became a significant problem
because the queue depth was insufficient to buffer
frames from multiple clients (which could have
different signal quality and capabilities) in an
efficient fashion.
Hence the new tx path in 10.4 was introduced: a
pull-push mode.
Firmware and host can share tx queue state via
DMA. The state is logically a 2 dimensional array
addressed via peer_id+tid pair. Each entry is a
counter (either number of bytes or packets. Host
keeps it updated and firmware uses it for
scheduling Tx pull requests to host.
This allows MU-MIMO to become a lot more effective
with 10+ clients.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
This implements very basic support for software
queueing. It also contains some knobs that will be
patched later.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
This merely adds some parsing, generation and
sanity checks with placeholders for real
code/functionality to be added later.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
On multicore systems, it is possible that txrx tasket can run
in parallel with pci tasklet (i.e smp affinity of ath10k irq is
assigned to multiple CPUs). Feeding and consuming from the same
rx completion list leads to txrx tasklet runs for longer period.
Prevent this by processing a snapshot of rx queue by moving list
into temporary list. Consecutive received frames will be processed
in next batch.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
Received frame indications are queued into a skb list and latest
processed by txrx tasklet. This skb queue is protected by htt rx lock.
Since the entire rx processing till delivering frame to mac80211 and
replenish tasks are processed under rx_lock protection, there might be
some delay in queuing newly received rx frame into that list on
multicore systems. Optimize this by using skb list lock while accessing
rx completion queue instead of htt rx lock.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
Currently, we are providing wrong payload data of pktlog to trace points.
Data we receive from FW through copy engine 8 contains pktlog data alone.
We don't need to parse anything in driver before handing it to trace
points.
Fixes: afb0bf7f530b ("ath10k: add support for pktlog in QCA99X0")
Signed-off-by: Ashok Raj Nagarajan <arnagara@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
This allows to use the new firmware which
implements the new tx data path. Without this
patch firmware supporting new tx path stops
responding shortly after booting.
This patch doesn't implement the entire pull-push
logic available in the new firmware. This will be
done in subsequent patches.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
New names make a bit more sense.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
On multicore systems it was possible for the txrx
tasklet to keep on running for long periods of
time on a single CPU due to tx completion
processing. Another CPU could feed the running
tasklet for an indefinite amount of time.
The tasklet is now guaranteed to run a finite
amount of time and is limited by HTT CE Rx ring
depth.
This improves behavior when RPS is used on target
system and might improve TCP handling as well.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
Allocations from the DMA zone were originally added for legacy ISA
stuff, or PCI devices that have specific limitations in their DMA
addressing capabilities. It has no place in ath10k, which can do
full 32-bit DMA.
Fixes memory allocation errors on some platforms.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
The QCA99X0 4 addresses RX packets pad 2 bytes at the beginning of
MSDU instead the end of ieee80211 header to keep alignment. The currently RX
data path can't parse the header correctly in this case. This patch fixes it
for QCA99X0.
Signed-off-by: Yanbo Li <yanbol@qca.qualcomm.com>
[kvalo@qca.qualcomm.com: checkpatch fixes and naming changes]
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
The RX rate idx is not correct for 11G mode OFDM packet.
Because the bitrate table start with CCK index instead of OFDM.
Signed-off-by: Yanbo Li <yanbol@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
|
This patch adds pktlog support for 10.4 fw versions.
Signed-off-by: Vivek Natarajan <nataraja@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
|
