Currently AMSDU and AMPDU aggregation is enabled for TID 0 to
TID 5 streams. Lets enable it for remaining two streams also.
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Currently AMPDU aggregation is preferred over AMSDU. AMSDU
aggregation is performed only if AMPDU streams in firmware
are full.
This patch adds simultaneous AMSDU and AMPDU aggregation
support. This mechanism helps to improve throughput.
AMSDU is enabled only for 8897 chipsets which supports 4K
transmit buffer. User can disable AMSDU using
'disable_tx_amsdu' module parameter.
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Write io memory to clean PCIe buffer only when PCIe device is
present else this results into crash because of invalid memory
access.
Cc: <stable@vger.kernel.org> # 3.9+
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch adds AMPDU support for TDLS link. We have set 11n
capabilities including AMPDU parameters during ENABLE_LINK.
We set a variable in RA list to indicate this as TDLS link.
This patch uses these capabilities to know if AMPDU is allowed
on TDLS link and enables AMPDU aggregation for TX and RX reording
support for RA list for this peer.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
1. If data packet is addressed to TDLS peer for which link is
established, mark these packets with TDLS flag so that FW can
send them on direct link instead of sending via AP.
2. If data packet is addressed to TDLS peer and TDLS setup is
underway, move these packets to TDLS queue.
3. If this packet is TDLS setup packet, do not block it.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
While TDLS link is being setup, few packets from this station to
peer station may be buffered at AP. It may happen that once TDLS
link is setup, packets sent from station to peer on direct link
get delivered before traffic from AP arrives at peer station.
This results into packet reordering issue at peer station.
To avoid this, we hold data packets destined to TDLS peer during
TDLS setup. These packets are moved to temperory TDLS TX queue.
Upon successful TDLS setup, they are moved to RA list created for
this peer. Upon failure, packets are moved back to AP's RA list
for that particular TID.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch adds support for parsing TDLS action frames during
station receive handler.
Peer station capabilities are stored into station node.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
tos_to_tid_inv values are needed even during TDLS restore
operations. Currently tos_to_tid_inv is part of wmm.c and
is declared static.
Make it part of private structure so that it can be used in
other files as well.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch fixes a bug where we were checking for AP's AMPDU param
setting even when transmitting traffic to associated station.
Patch adds provision to pass additional parameter ra_list pointer
to function which checks if AMPDU is allowed. If current BSS type is
AP, we check station's AMPDU params else we check AP's AMPDU params.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Before we loop for next iteration we adjust the buffer pointer and
"resp_len":
curr += (tlv_len + sizeof(tlv_hdr->header));
resp_len -= (tlv_len + sizeof(tlv_hdr->header));
If "resp_len" gets set to negative then it counts as a high positive
value.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The variables cancel_scan_cmd, enable_data, hs_activate and valid are
only assigned the values true and false. Change its type to bool.
The simplified semantic patch that find this problem is as
follows (http://coccinelle.lip6.fr/):
@exists@
type T;
identifier b;
@@
- T
+ bool
b = ...;
... when any
b = \(true\|false\)
Signed-off-by: Peter Senna Tschudin <peter.senna@gmail.com>
Acked-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Bug 60815 - Interface hangs in mwifiex_usb
https://bugzilla.kernel.org/show_bug.cgi?id=60815
We have 4 bytes of interface header for packets delivered to SDIO
and PCIe, but not for USB interface.
In Tx AMSDU case, currently 4 bytes of garbage data is unnecessarily
appended for USB packets. This sometimes leads to a firmware hang,
because it may not interpret the data packet correctly.
Problem is fixed by removing this redundant headroom for USB.
Cc: <stable@vger.kernel.org> # 3.5+
Tested-by: Dmitry Khromov <icechrome@gmail.com>
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Maintain total outstanding packet count for RA list instead of total
outstanding size as packet count metric seems more reasonable for
checking threshold etc.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
pkt_count is used to determine if BA can be formed on this RA list
by comparing it with randomly generated BA threshold. The pkt_count
variable name here is ambiguous and does not reflect its usage
correctly. Rename it to ba_pkt_count.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch disables tx_amsdu support in mwifiex by default.
tx_amdsu support can be enabled via module parameter at load time.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Rotate bss prio list, so the bss next to the one served, will come first
in the list of bss' with equal priority. This way we pick bss nodes in a
round robin fashion. Using list rotation instead of a cur ptr simplifies
iteration to calling list_for_each_entry. List rotation is done via
list_move, where the head itself is temporarily removed and then
re-inserted after the bss just served.
Signed-off-by: Andreas Fenkart <andreas.fenkart@streamunlimited.com>
Acked-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
After a packet is successfully transmitted, ra list is rotated, so the ra
next to the one transmitted, will be the first in the list. This way we
pick the ra' in a round robin fashion. This significantly simplifies
iteration in mwifiex_wmm_get_highest_priolist_ptr to a call to
list_for_each_entry.
List rotation is done via list_move, where the head itself is temporarily
removed and then re-inserted after the item just transferred.
Signed-off-by: Andreas Fenkart <andreas.fenkart@streamunlimited.com>
Acked-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Not locking ra_list when dequeuing packets creates race conditions.
When adding a packet 'tx_pkts_queued' is modified before setting
highest_priority_queue. If in-between the main loop starts, it will
see a packet queued (tx_pkts_queued > 0) but will not find it, since
max prio is not set yet. Depending on the scheduling, the thread
trying to add the packet could complete and restore the situation.
But this is not something to rely on.
Another race condition exists, if a new packet, exceeding current
max prio is added. If concurrently a packet is dequeued, the newly
set max prio will be overwritten with the value of the dequeued
packet. This can occur, because selecting a packet and modifying
the max prio is not atomic. The result in an infinite loop unless,
a new packet is added that has at least the priority of the hidden
packet.
Same applies to bss_prio_tbl. Forward iteration is no proper
lock-free technique and provides no protection from calls to
list_del. Although BSS are currently not added/removed dynamically,
this must not be the case in the future. Hence always hold proper
locks when accessing those lists.
Signed-off-by: Andreas Fenkart <andreas.fenkart@streamunlimited.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Using NO_PKT_PRIO_TID and tx_pkts_queued to check for an empty
state, can lead to a contradictory state, resulting in an
infinite loop. Currently queueing and dequeuing of packets is
not synchronized, and can happen concurrently. While tx_pkts_queued
is incremented when adding a packet, max prio is set to NO_PKT when
the WMM list is empty. If a packet is added right after the check
for empty, but before setting max prio to NO_PKT, that packet is
trapped and creates an infinite loop.
Because of the new packet, tx_pkts_queued is at least 1, indicating
wmm lists are not empty. Opposing that max prio is NO_PKT, which
means "skip this wmm queue, it has no packets". The infinite loop
results, because the main loop checks the wmm lists for not empty
via tx_pkts_queued, but for dequeing it uses max_prio to see if it
can skip current list. This will never end, unless a new packet is
added which will restore max prio to the level of the trapped packet.
The solution here is to rely on tx_pkts_queued solely for checking
wmm queue to be empty, and drop the NO_PKT define. It does not
address the locking issue.
Signed-off-by: Andreas Fenkart <andreas.fenkart@streamunlimited.com>
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
ra_list_spinlock is used to protect struct mwifiex_wmm_desc and
embedded structures such as ra_list. tid_tbl_lock while more fine
grained, is not used but in one function. That function is not
called reentrantly. To protect ra_list from concurrent modification
ra_list_spinlock must be held.
Signed-off-by: Andreas Fenkart <andreas.fenkart@streamunlimited.com>
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
adapter->bss_prio_tbl list has already been checked in outer loop.
The inner loop works with priv_tmp->wmm.tid_tbl_ptr list. Also the
lock taken, gives hint that this is likely a copy-paste error.
Signed-off-by: Andreas Fenkart <andreas.fenkart@streamunlimited.com>
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch fixes an issue where RX throughput values observed
were substantially lower than TX counterparts for PCIe8897 STA.
PCIe8897 supports larger rx_win_size. After changing these values
we see big improvement for TX and RX throughput values.
Different tx_win_size and rx_win_size are used for AP mode.
All BA setup related initialization has been moved to separate
function.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Sagar Bijwe <bsagar@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>
alloc failures already get standardized OOM
messages and a dump_stack.
Convert kzalloc's with multiplies to kcalloc.
Convert kmalloc's with multiplies to kmalloc_array.
Remove now unused variables.
Remove unnecessary memset after kzalloc->kcalloc.
Whitespace cleanups for these changes.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is an enhacement to mwifiex_pcie driver to use
map/unmap PCI memory APIs. This reduces one memcpy each in TX
path and RX path, and enhances throughput.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch adds handler to clean PCIe TX rings after disconnect
or bss stop is called for PCIe based mwifiex driver.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch adds support for multiple TX queues inside mwifiex
driver. Four different queues according to WMM access categories
are defined for each virtual interface. When a packet is
received from netdev for transmission, tx pending count for
particular queue is incremented and if tx pending count has
reached upper water-mark, this queue is stopped instead of
stopping all queues. Similarly when a packet is successfully
transmitted from device, tx pending count is decremented per
queue and if pending count falls below lower water-mark, queue
operations are again resumed. This ensures that not all
tranmission is blocked if traffic with particular TOS value
suddenly increases.
Also wake all queues after association/IBSS_join/uAP_BSS_start
to enable traffic on all queues.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Implement mgmt_tx in cfg80211 ops through data path.
Advertise probe resp offload and skip to send probe resp in AP
or GO mode.
Signed-off-by: Stone Piao <piaoyun@marvell.com>
Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: Kiran Divekar <dkiran@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Remove unnecessary blank lines and extra tab.
Signed-off-by: Stone Piao <piaoyun@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
"adapter->bss_prio_tbl[j].bss_prio_head" points to linked list of
interfaces with priority 'j'. "bss_prio_tbl[j].bss_prio_cur" is
supposed to point to next interface every time the routine for
dequeuing the packet is called. This ensures that each interface
gets fair chance.
Currently we have AP and station interfaces with priority '0'.
Therefore "adapter->bss_prio_tbl[0].bss_prio_cur" should
alternately point to AP and station nodes. Since "bss_prio_cur"
is not correctly updated, for each packet picked for AP, two
packets are picked for station interface.
This patch fixes the problem by correctly updating "bss_prio_cur".
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch adds support for handling BA request and BA setup
events for AP interface.
RA list is marked as either 11n enabled or disabled from station's
capabilities in association request. BA setup is initiated only
after some specific number of packets for particular RA list are
transmitted.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Kiran Divekar <dkiran@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
After association, STA will go through eapol handshake with WPS
enabled AP. It's observed that WPS handshake fails with some 11n
AP. The reason for the failure is that the eapol packet is sent
via 11n frame aggregation.
The eapol packet should be sent directly without 11n aggregation.
This patch fixes the problem by adding WPS session control while
dequeuing Tx packets for transmission.
Cc: "3.4.y" <stable@vger.kernel.org>
Signed-off-by: Stone Piao <piaoyun@marvell.com>
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Currently we check the sequence number of last packet received
against start_win. If a sequence hole is detected, start_win is
updated to next sequence number.
Since the rx sequence number is initialized to 0, a corner case
exists when BA setup happens immediately after association. As
0 is a valid sequence number, start_win gets increased to 1
incorrectly. This causes the first packet with sequence number 0
being dropped.
Initialize rx sequence number as 0xffff and skip adjusting
start_win if the sequence number remains 0xffff. The sequence
number will be updated once the first packet is received.
Cc: "3.0.y, 3.1.y, 3.2.y, 3.3.y, 3.4.y" <stable@vger.kernel.org>
Signed-off-by: Stone Piao <piaoyun@marvell.com>
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Kiran Divekar <dkiran@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Reserve space for one more priv structure.
This will be used by AP interface.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: Kiran Divekar <dkiran@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch supports Avastar 88W8797 chipset with USB interface.
The corresponding firmware image file is located at:
"mrvl/usb8797_uapsta.bin"
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Kiran Divekar <dkiran@marvell.com>
Signed-off-by: Frank Huang <frankh@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This saves some space and adds better readability.
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>
Pass mwifiex_private pointer directly to wmm_add_buf_txqueue()
instead of passing adapter and then deriving priv again in
mwifiex_get_priv_by_id(). This reduces a 'for' loop in TX path.
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Current implementation, for retrieving priv from adapter,
uses bss_index. In multi interface environment supporting
different types, bss_index may not be unique.
Use bss_type along with bss_num to retrieve the priv.
bss_index is removed with this change.
Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: Kiran Divekar <dkiran@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The commit "mwifiex: remove list traversal.."(fcf2176c87..)
wrongly modifies AMSDU aggregation check. Due to this even though
packet size for iperf traffic is already large, we unnecessarily
try to aggregate them which adds some delay. If Tx iperf is started
on UUT for 30 seconds, UUT keeps sending Tx packets for few more
seconds.
That commit is reverted to fix the problem.
Also, MIN_NUM_AMSDU check is moved inside the loop to optimize the
loop.
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>
For ad-hoc mode, RA is created for each peer connected. In case of
multicast traffic new RA will be created for each multicast
address. While processing Tx packets we have to go through this
RA list. We can avoid some RA nodes by sharing same RA for both
multicast and broadcast packets.
Therefore "memset(0xff)" is used to treat multicast packet as broadcast
one while choosing RA.
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>
Instead of counting the number of packets in txq
for particular RA list before AMSDU creation,
maintain a counter which will keep track of the
same.
This will reduce some MIPS while generating AMSDU
traffic as we only have to check the counter instead
of traversing through skb list.
Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
If we go through the entire for loop in
mwifiex_wmm_get_highest_priolist_ptr() and don't find any packets,
set highest_queued_prio to NO_PKT_PRIO_TID (< LOW_PRIO_TID).
Signed-off-by: Marc Yang <yangyang@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
add checks to mwifiex_wmm_process_tx() loop so it doesn't re-enter
mwifiex_dequeue_tx_packet() to find it can't send.
Signed-off-by: Marc Yang <yangyang@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch adds highest_queued_prio to track priority of packets as
they are enqueued so that mwifiex_wmm_get_highest_priolist_ptr()
starts checking at the first level where we have packets, instead of
the highest. The function also lowers priority value to the level
where first packet is found.
Signed-off-by: Marc Yang <yangyang@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch adds tx_pkts_queued to track number of packets being
enqueued & dequeued so that mwifiex_wmm_lists_empty() evaluation
is lightweight.
Signed-off-by: Marc Yang <yangyang@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Skip initialization of local variables with some default values
if the values are not going to be used further down the code path.
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>
Amitkumar Karwar