IOMMU map and unmap RX buffers into LPASS SMMU context
bank for LPASS to access RX buffers in direct link
usecases.
Change-Id: Ie72a008309e3abdf3fbc7198bb5d3af5e9497180
CRs-Fixed: 3356568
Currently HP update can skip for few packets if the system suspend
is happening in parallel with tx. This could lead to SMMU fault if
the skipped HP update goes after tx desc force free.
Remove the tx pause check in HP updates and have the check before
calling dp_tx_send in case of intra-bss forward case. No special
handling is needed for non intra-bss case as we don't expect packets
from network stack after suspend.
Change-Id: Id4f2cefcc0a14e7c16438b9bda5cb1d55eb05050
CRs-Fixed: 3327818
Changes dp_rx_buffers_no_map_replenish API to
miminize the contention on RX refill ring lock
Change-Id: Ie55f034c6d01abbab21707460cd0f20548b35ba2
CRs-Fixed: 3313886
We allocate extra buffers than requested when watermark is less
than critical low threshold. While allocating extra buffers we
should check for boundary conditions so that total allocated
buffers is always within the range of available entries.
Change-Id: I2f24c6157cb2afcc8bc5ca84b7e0308588a4d99d
CRs-Fixed: 3307051
get the HP of the RXDMA ring without incrementing
This will ensure in case of nbuf failures the ring
HP remains unchanged
Change-Id: I69ec9207a44a4c50484933797326e962ad2d4a5c
CRs-Fixed: 3309394
For STA MLO connection, AP might send M1 right after assoc response
on assoc link, but at this monment, host has not sent wmi_peer_assoc
for the assoc link and no osif_rx is registered to dp_vdev, MLD peer
will not be initialized in FW side and no HTT MLO peer map msg to host,
then this M1 frame will be detected as no valid txrx_peer and dropped
in host as dp_vdev->osif_rx is NULL as well.
Store this M1 frame to cached buffer queue until CP register the osif_rx
to DP and then flush them back to stack.
Change-Id: Ie84fa9c39db75fe77b9fd61dc1bf46f2fa737df7
CRs-Fixed: 3289587
Raw mode mpdus in REO end up resulting in NULL_QUEUE_DESCRIPTOR error.
Raw mode mpdus are not expected in REO, free them without further
processing.
Change-Id: Ica920caecf90a1107ce17836051e1019c9dfc994
CRs-Fixed: 3282836
In MLO mode, there are 2 link peers and 1 MLD peer. TX/RX stats
is kept in txrx_peer, which only exists in MLD peer. This change
aims at fixing HDD failing to get rx rates info when in MLO mode.
Change-Id: If476216a6b0d1c006fbc0d45a3cbfbb0be79efd3
CRs-Fixed: 3287090
Debug framework for tracking the SMMU mapping and unmapping of
skb buffer when IPA offload is enable and IPA_SMMU_NBUF_TRACKER
is define.
CRs-Fixed: 3255423
change-Id: I8966097945b33cb866f4642b98b3d1a4beabfad7
1. Add stats to indicate REO ring size and Rx desc pool size
2. Add low threshold interrupt stats
3. Update SW2RxDMA ring size max to 16K
Change-Id: If84b88bc08e447774ab445df36b9f2f2219356b8
CRs-Fixed: 3286940
Only in 11b mode, exclusive of other legacy modes(11ag) will
the NSS obtained from the TLV header be invalid, which is 0
and cannot be resolved. This change will set nss to 1 in 11b.
Change-Id: I93e8bdb125764fb0247265da328929ecc837fb84
CRs-Fixed: 3275793
Driver is able to get rates info per MSDU, which can
exclude special frames like ARP, DHCP, etc. This change
is to refill the rates info parsing by rateflags from FW,
including legacy rate, MCS, NSS, BW, preamble, GI, etc.
Change-Id: I129d5f0dc48758d1a92ac0c3622c1761cdae8286
CRs-Fixed: 3255467
When nbuf allocation failure happens, there is no retry scheme.
So RX buffers in refill ring may shrink and not got enlarged.
This change is aimed to replenish more RX buffers when watermark
is less than critical low threshold.
Change-Id: I201b9e252ba08ba1bff47e0b5ec819a45f1b1ddf
CRs-Fixed: 3245915
skb queue is not always mapped to the proper CPU
hence it is good to map the ring_id to CPU through get_cpu
API.
Change-Id: I3bf76452cc1fcb2033ef1a8640c4dd31ecba61b3
CRs-Fixed: 3255144
DP uses multi page allocation for tx/rx descriptor.
ID and offset mask of decriptor is based on 4096 which
couples with Kernel's MMU PAGE_SIZE. This cause trouble
when deploy driver on none-4K page size kernel.
Set qdf_dp_blockmem_size to 4096 so that DP won't
depend on kernel page size.
Change-Id: I17f5c10b394e8709e6b4b153f3dd094cf792787f
CRs-Fixed: 3235246
The list of SW descriptors that we create from the HW descriptors
in DP Tx completion path is corrupted in intra-bss path. This is causing a
crash when corrupted descriptor is accessed.
The cause of corruption seems to be due to incorrect rings being accessed.
Use cpu index to fetch tx. ring IDs.
Change-Id: I5342b7c4e24df983900fde79f36df8395bf628cb
CRs-Fixed: 3223420
For QCN9224 splitphy mode do not rely on rx_desc_pool_id
to get to the VDEV as desc pool is common across PDEVs
Change-Id: I2552878bd56696a8df463b35e2b273ed4e591050
CRs-Fixed: 3220690
Currently while sending nbuf to rx fate logs
we are not removing rx tlv before sending nbuf
to logger which is causing issue of incorrect
packet type issue.
To Fix issue make changes to make sure that
rx tlv is removed and nbuf length is set
Change-Id: Ib56c8e009398b3410c43be910feb4048bfb06b54
CRs-Fixed: 3241828
To avoid unmeaningful rate fluctuation, when report
rx linkspeed to upper layer, exclude special frames:
arp/ssdp/dhcp/eapol/ipv6 NA/NS/DHCPV6 in low rate.
Change-Id: I4bc49a808a02b4bc8c687f70690922045d65e739
CRs-Fixed: 3198484
Currently 11be msdu rate statistic is not supported yet.
Another issue is: HW reported pkt type value 6, while host
SW defined 11be type DOT11_BE (value 5), they are not matching.
DOT11_BE value 5 is referred in monitor mode and otherwhere,
so do map and conversion in host sw side.
Change-Id: Ic5c2dd2a35cbe0ecd6430f007b6e7e02ece40998
CRs-Fixed: 3230900
1. Remove prints in per packet path
2. Add prefecth for skb shinfo
3. Avoid qdf_get_cpu() in DP tx
Change-Id: I86d196cc7865a2d5b3a5a6098f6f7d84397a60d0
CRs-Fixed: 3227568
For non-self sta peer (sta_self_peer is set to 0) and in error cases
like (WMA peer create failure), peer->txrx_peer is NULL which
indicates DP peer setup has not happened and therefore there are
no rx cached frames flush. Having a qdf_assert_always() if sta_self_peer
is set to 0 is incorrect. This a valid scenario.
Hence remove the irrelevant qdf_assert_always().
Change-Id: I1ab18dd681534bb9257be388a74adedc529a67cf
CRs-Fixed: 3212253
Use microsecond timestamping for the tx descriptors
for capturing the latencies more accurately. It will help
to determine the appropriate bucket.
Change-Id: I216e083a7c89b01b6f2f384c1c0a85ca323d3a3e
CRs-Fixed: 3165153
In VoIP cases, no or minimal packet drop is expected during
roaming. Currently, data packets are dropped in host if they
are received after roam sync indication from firmware and before
peer map indication is received due to invalid peer.
This change allows rx UDP data packets during roaming before
peer map indication is received from FW if peer's auth_status
received in roam_sync_indication from FW is authorised.
Change-Id: Ic518b7cd0aef48a2b0f8c923a2e20838b07f3d1f
CRs-Fixed: 3168603
When TX completion is released from FW, dp_tx_notify_completion()
will not be invoked, then TX DCHP ACK will miss
hdd_softap_notify_tx_compl_cbk() which then
WMI_PEER_CRIT_PROTO_HINT_ENABLED with value 0 will not send to FW.
(1) move dp_tx_notify_completion() from dp_tx_comp_process_desc_list()
to dp_tx_comp_process_tx_status(), so that both TX completion release
source FW or TQM case will call it.
(2) clear TX notify flag for intra-bss forwarding BC/MC to avoid
unnecessary TX completion notify.
(3) Set ts.status for release source FW case otherwise it will be value
0 always.
Change-Id: I2bf9900a3d16ba162a83b061f0b96e7d2f79423a
CRs-Fixed: 3178423
This feature can be enabled runtime using sysfs interface.
Support is added to dump and clear the histogram stats.
The lower delay regions has to be more granular to indicate any
medium related issues for time sensitive XR applications.
Change-Id: I0a44a54d12d92ce016de349810cb2bedebaf9a58
CRs-Fixed: 2981006
Few APs are sending bar frame right after sending a
packet but the packet is not received and its retry
is received with the same sequence number. As SSN
is moved ahead due to bar frame, this packet is being
dropped.
To fix this, allow all frames to get delivered to stack.
Change-Id: I17f7126c8d318415e88b037b72563cf53cb14066
CRs-Fixed: 3151908
Add feature support to tag first packet that wakes up HOST from WoW.
rx_pkt_tlv.rx_msdu_end.reserved_1a field is used by TARGET to meet
such request.
Change-Id: I3d37e13e8cff49bc4f622d3070a19e4c4be56417
CRs-Fixed: 3137621
Intrabss check for NAWDS was added as a temporary fix to resolve the
issue of selfbridge enteries present in AST table. The actual reason
is loopback pkt received with one's own bridge mac. To avoid this, add
a check for not sending out any packet with an invalid peer id in the
ast table.
Change-Id: Ia4c520bcc8b7077f0b484a0bc40c4d26db77c3f4
CRs-Fixed: 3135142
Flush dp cached frames only for mld peers and legacy peers, as
link peers don't need it.
Change-Id: I19a27c5810a1fec6a62bfe5a6377073dc382fb4c
CRs-Fixed: 3131405
WAR for wrong length in first msdu in wbm rx error path while
handling scatter gather buffers. Get the msdu length from the
last buffer instead.
Change-Id: I869391ef5ca7005dce972013679c2143be836ecb
Move the stats parameters from the dp_peer which are used
in per packet path to txrx_peer
Change-Id: Ieb68b6950740791b37bbd2bfdc4815c3d7bc4834
CRs-Fixed: 3095637
Move the parameters from the dp_peer which are used
in per packet path to txrx_peer and params related
to monitor path to monitor peer.
Initialize txrx_peer and monitor peers.
Change-Id: I0449c67c1cd47deb76bc89b5ddc64174f6694eb0
CRs-Fixed: 3095637
simple Alloc is being used in RX path which avoids
certain debug logic. during free of nbuf we should
avoid this debug logic else it will report it as
double free, this will be triggered only when debug
is enabled
Change-Id: Iadb40071fb733cc4de3291784df5075d5a099a8e
use dma_inv_range call instead of dma map and unmap
APIs during replenish and reap. This complete code is
under a compile time macro.
before change: 4160Mbps @ 96.5% (core 3)
after change: 4160Mbps @ 90.5% (core 3)
Change-Id: I61bca349a369ace06cd86f353880108cee013cb1
