Currently the code to support Multipass on SAP is
present along with the code to support WDS. Hence with
the code in its current state, we will not be able to
enable Multipass support without enabling WDS.
Move the multipass support code out of the WDS support
code, to be able to enable Multipass for chipsets which
do not use WDS.
Change-Id: Iaafa8dc4f16314d9e3e160fe01251c3684adbf67
CRs-Fixed: 3468548
If the ingress net device is NULL, return gracefully
from the API to check if the skb is from rmnet
Change-Id: I81b8cea7f311fe581e83f2310155b83b28f2fd32
CRs-Fixed: 3445050
In the case of WCN6450, WBM HW block is removed in the UMAC.
TX completions come via HTT messages. Add logic to handle
HTT TX completion messages from the firmware.
Changes are specific to WCN6450 and hence implement the logic
in the arch specific code.
Change-Id: I447020354ce26e8948e4b49648c434fb2ed302cd
CRs-Fixed: 3381814
This code helps to initialize and deinitialize new SOFTUMAC
based Rhine architecture.
Change-Id: I374140ccb3b31e9351c6e683c77d81a5a876472a
CRs-Fixed: 3382913
1) Enable MLO Link Stats Support
2) Add Stats counter to capture Invalid Link ID
3) Fix the FW Recovery crash while ol_stats
disable.
Change-Id: I1b5263ae024579c5ec2018f8082f0d03ffdf0030
CRs-Fixed: 3431319
Add support to handle the extended Tx completion
statistics during direct switch mode.
Change-Id: If1b6906d5d2b3156ead8ef039524f47e763cb543
CRs-Fixed: 3431840
Add support to enable/disable MLO Link Peer stats through
ini and cfg80211tool enable_ol stats command
Change-Id: Id1229a149befa416d060e1b07eee150e6b295abf
CRs-Fixed: 3397721
Tx completion is processed in napi context, which should follow
napi's max quota, otherwise once process count is over napi quota,
it will trigger napi repoll in kernel, meanwhile by current design,
tx completion hard irq will be enabled when the 1st napi process
finished, in such case 2nd napi process will conflict with repoll process,
finally trigger kernel side napi list crash.
Since napi quota is used in serval cases, while for tx completion
process, we want to process tx completion count as many as possbile
in one cycle, set 64k transmit frames account for 1 napi work unit
to gain max tx kpi.
Change-Id: Ic24f131c90b90b0e118edffcab559ddf31779dcf
CRs-Fixed: 3421368
Reuse the tx descriptors released in tx completions
without releasing the associated skbs to reduce
the cpu utilization in direct switch mode.
Change-Id: I4ab3ac58977a626344877b8a818a4dbc4864aaf3
CRs-Fixed: 3393968
The kernel-doc script identified a large number of documentation
issues in dp/wifi3.0/dp_tx.[ch], so fix those issues. In addition,
there are a number of instances where public functions have their
implementation documented instead of having their interface
documented, so move that documentation.
Change-Id: I7a09c451f3f4af3dd606778c49cce81677f6fe58
CRs-Fixed: 3373159
The Tx delay computed is very large if the enqueue
timestamp is zero.
Compute Tx delay only when the enqueue timestamp is valid.
Change-Id: Ic4032a6f4a01fc661dc1155a4543cd7fc4216970
CRs-Fixed: 3354639
The following functions are currently public functions:
dp_tx_tso_cmn_desc_pool_alloc()
dp_tx_tso_cmn_desc_pool_init()
dp_tx_tso_cmn_desc_pool_deinit()
dp_tx_tso_cmn_desc_pool_free()
However these functions are only called from within the file where
they are defined, so make them static.
Change-Id: If57b36398e0cf6a33be2b85bc00d651fe5ed9a71
CRs-Fixed: 3371860
This change extracts several common code snippet and put
them into a new function. This can keep the code clean.
Change-Id: Ib13f06903db2efbf0e447436fb5abb1386cdd41d
CRs-Fixed: 3370243
Currently, we record TX musdu number and packet length, which falls into
three categories: from stack, from host and successfully transmitted. In
TSO mode, the msdu from stack will be divided into several fragments and
each of the fragment will be transmitted as an individual msdu. This change
will take TSO into account when it comes to the 3 forms of TX statstics.
Change-Id: Ie11cb7541b4e109609bf3104739a5e19f51dcc13
CRs-Fixed: 3361412
Special frames like EAPOL should not be dropped as it would cause
failure in connection. To prevent the dropping of higher priority
special frames, reserve a set of descriptors from the tx descriptor
pool.
Check the regular soc or pdev limit for the allocating tx descriptors
for the regular packets and use the reserved set of descriptors
only for the special packets.
CRs-Fixed: 3331359
Change-Id: I15532fb5e325ba6ad997090008bda8c52707a05b
The stats to indicate the packet drop for EAPOL packets increments twice
in an instance. Fix this double incrementing of stats counter.
CRs-Fixed: 3365584
Change-Id: I524fa1a196452089c94acf50b52db0b00f562317
1.Add check in Tx completion path to handle
mcast packets from all ML partner vdevs
2.Handle cases where a ML mcast vdev can be
deleted and started as NON-ML vdev and
viceversa by adding reset ML mcast conf.
3.Optimized Register configurations for Mcast
at Init/Delete and Start/Stop AP
Change-Id: Iab8a5a081e2e0377509574d362754d32c5b83df2
CRs-Fixed: 3350350
Suspect HW update WBM2SW ring HP, but the ring entry contents are
not updated accordingly which then host will fetch one stale ring entry,
this makes other TX packet are freed unexpectedlly.
Add change to detect this situation earlier, if HW cookie conversion is
done, then invalidate 2nd dword for upper 32bits of VA, so next time when
reap this ring entry contents to know if this desc is updated by
HW or not. if HW cookie conversion is not done, then compare the PA in
buff_addr_info with PA in current TX desc to check.
Change-Id: I351eb4f860216fc618ff28736d4832fcec45dcc5
CRs-Fixed: 3345935
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
buffer_timestamp in wbm2sw ring descriptor is 19 bits only with
unit 1024 us, so it can only accommodate 0x7FFF * 1024 us same as
29 bits value with unit us. but if the timestamp value is >
0x7FFF * 1024 us, then the value stored in buffer_timestamp is not
right as other higher bits bit29 ~ bit31 is lost.
restore bit29 ~bit31 value for buffer_timestamp to fix current issue.
Change-Id: Ie929f560365cc822883711133541772348775357
CRs-Fixed: 3336241
pkts from rmnet device are sent as SG, we see poor
throughput when transmitting these SG frames using
an MSDU extension descriptor. to resolve this issue
we identify if the frame is from RMNET device and if so
copy the headers from skb linear address to the frag
address and provide the frag address to HW for transmission
CRs-Fixed: 3337822
Change-Id: Ic1b7d9f861dcd4f838509772c45e47d3917fc59f
Reap mode desc from the Tx completion ring in case the number of
available entries are less than the half of the ring entries.
Change-Id: I3635b88c3e98cc8a5f30fc58cf5e353f6301d4ad
CRs-Fixed: 3325826
Add the stats to indicate the packet drop for EAPOL packets.
This will be useful for the debug of EAPOL packet drop.
CRs-Fixed: 3329895
Change-Id: I5e0abb4e574116a55de124c0012c17154c7c63da
Add support for:-
1. PPE VP entry attach and detach.
2. Per VAP PRI2TID Support
3. Dump the PPE VP HW entries.
4. Add tx completion handling for ppeds descriptors
Change-Id: I2a6d0be5bb556663a39a24d17b703877f3b5ad00
CRs-Fixed: 3276981
