MSDU count is not reliable in all cases.Hence it is better
to check whether next_link_desc is valid or not and then
free the MSDUs in the next descriptor and subsequently
release both the current and next link descriptor back to HW.
Change-Id: Ie5ea3be547f9aed6673293221f6b47661cd92cce
ring_desc is being validated in dp_rx_err_process before passing
the ring_desc to dp_rx_defrag_store_fragment.
Hence removing the additional checks for ring_desc.
Change-Id: Ib863ea4d512075beed58f09ff6167aa2a556efea
CRs-Fixed: 2771408
There's no declaration for API dp_peer_qos_stats_notify().
Define the API as "static inline".
Change-Id: Id6769d680b3304cebf00b2348fb5b9a3bc9c7368
CRs-Fixed: 2773655
Add statistics if the packets of the flow are coming on
a different REO than the one configured. This would help
in debugging the race conditions where the packets of the
same flow arrive on different REO rings.
Change-Id: I73714c4ad64e00df06c72ac4ae5f7567adeab505
CRs-Fixed: 2771719
Macro MAX_MONITOR_HEADER is used to specify the max. headroom size for
different buffer rings. With the monitor frag feature, macro
DP_RX_MON_MAX_MONITOR_HEADER is required to specify the headroom size for
the monitor destination buffers as per the the approch being used.
Change-Id: Ic4c0132292008942d7d443235e5f54d7eaa77e74
CRs-Fixed: 2772048
Add the support to coalesc the tx head pointer
register writes to reduce the number of PCIe
writes.
Change-Id: I4fea22464c2bf6930bac42dab72c1c2fc5fa18e2
CRs-Fixed: 2769017
The 11g and 11a connections are latency critical and
hence the tx coalescing cannot be applied on traffic
running for these connections.
Add a CDP API to save the number of low latency
connection in the dtapath vdev handle. The Software
latency manager can use this information to decide
and skip the tx register write coalescing for packet
transmission on low latency connections.
Change-Id: Iff51867447e9d4537a3637cb65d94e184e4249ef
CRs-Fixed: 2769025
The variable 'ret' in ‘__dp_ipa_tx_buf_smmu_mapping()’ maybe
used uninitialized. Set one default value for it.
CRs-Fixed: 2771822
Change-Id: Ibf4bf6aa2076de84b3a47122ff4f3cae055cd0f7
Change log level of print from qdf_err to dp_err in
dp_peer_qos_stats_notify to avoid console flooding
Change-Id: I214942135826a6b94e6cf0b34b69d2c662a35e7b
Currently the stale cookie and invalid cookie
counters are being logged inter-changeably.
Fix the logging for stale and invalid cookie counter.
Change-Id: I43346fe961ee750a3baa1bf8c2da2551fb06e973
CRs-Fixed: 2770877
1. Add flag to support hlos id override feature in dp vdev
2. Update tid from nbuf->priority in dp_tx_send
3. Update tid to nbuf->priority in dp_rx_process
Change-Id: I66e8d77733a667f3f60b77e0d7bb444f7c5ad93d
dp_tx_pdev_detach API takes care of flushing
tx descriptors and multicast enhancement mem
cleanup
dp_pdev_deinit takes care of cleaning up the
above actions and dp_tx_pdev_detach is simply
a redundant piece of code
Hence remove dp_tx_pdev_detach API and call
dp_tx_desc_flush directly from dp_pdev_deinit
CRs-Fixed: 2769303
Change-Id: Ic798793e81eefc7dc905ca7552fdca13e9fddd8c
Currently when IPA is enabled, all RX frames will be routed to IPA
ring REO2SW4 by default, there is no REO2HOST option. once IPA RX/TX
is slow, then likely will give RX back pressure to wlan RXDMA. per
FW feature request, when IPA RX is slow or stuck, route RX frames to
wlan host to mitigate issue, thus provide REO2SW1 ring as REO_DST
option into HWIO_REO_R0_DESTINATION_RING_CTRL_IX_0.
To make sure if without FW dynamic REO_DST ring switching change,
current host change should not give chance that RX frames
route to host when IPA is enabled, select CTRL_IX_0 index 2 to
set REO_DST option, the reason is: if RX hash is enabled, CTRL_IX_2
and CTRL_IX_3 index 16~31 is REO_DST option filled with REO2SW4,
if RX hash is disabled, then CTRL_IX_0 index 1 and 4 filled with
REO2SW4 is REO_DST option configured by
WMI_HOST_PEER_SET_DEFAULT_ROUTING, so CTRL_IX_0 index 2
is only used for FW dynamic ring switching.
Change-Id: I6af12e07472dbdd4ab6d97f24e9fd4082c4e7328
CRs-Fixed: 2767844
Dont pass las_msdu as a param to restitch func for frag
approach as it doesn't contain valid value and is not used
in the restitch api.
Change-Id: I974a8ae5e6ce4a9bf04985cbd37e103700f48301
peer wlanstats context is used only for RDK Plume stats
but the name is misleading as wlan statistics
Hence rename peer's wlanstats_ctx to rdkstats_ctx and
the APIs the APIs around the same
CRs-Fixed: 2767192
Change-Id: I9e8955554b94fa29a3964fed729a95a4db67c8f6
Enable data path polling mode based on INI
dp_poll_mode_enable. Configure all data path
interfacing rings (UMAC/LMAC) and Monitor rings
in polling mode when NSS offload is disabled.
By default, poll mode is disabled.
Change-Id: I0e51207042811f517a423eb7276e3f33c5313450
Skip the recording of rx error ring and rx reinject
history if the memory allocation for the history
fails during the attach.
Change-Id: Ifa74937d0c37e6013ccaec851dab7d9e53057ac2
CRs-Fixed: 2767579
Register IPA ready callback with IPA driver. The callback
sets ipa_is_ready flag in the host driver and kick starts
the ipa init sequence as part of which the Tx buffers are
mapped to IPA.
None of the IPA APIs are invoked until IPA ready registration
is complete.
Change-Id: I6570b2b347052164a274fbc22358ebf0719dcabf
CRs-Fixed: 2735107
In the scanerio where peer teardown and peer map happen
in parallel, AST entries could get added to the peer AST
list after cleanup happens as part of peer teardown. AST
cleanup as part of peer unmap handling would not remove
the entry from the peer AST list. This is because
dp_peer_get_ref_by_id returns NULL as peer state is moved
to logically deleted state.
Fix is to not take peer ref by id and use the peer
argument passed in which already holds ref in the
peer unmap handling.
Change-Id: I844d3f07e9eec962d44bbfba851ad46f350e4f97
CRs-Fixed: 2766779
In the scenario where htt packets fail to be transmitted to
FW, they are added in both endpoint tx queue and misc list.
On receiving FW down indication before these packets are sent
to FW, endpoint tx queue is flushed and nbufs are freed. Post
this, the aforementioned nbufs would be unmapped/freed again in
htt_htc_misc_pkt_pool_free causing panic.
Fix is to free the nbuf via misc pool free only when the magic
pattern is set.
Change-Id: Ie523a7ca0054eb4104a107b3bddd50cb9b585275
CRs-Fixed: 2762829
Debugfs support for HTT stats under compile
time flag "HTT_STATS_DEBUGFS_SUPPORT".
Change-Id: I44c4f11265fdb3b9d53a9ac545e2945854bb9a57
CRs-Fixed: 2760947
Add logic to pre-allocate DP consistent memory and reuse later
on wifi on/off. This prevents memory fragmentation.
Change-Id: I1e5eb0da39950a5d028dd46d38ff715f283e53ff
CRs-Fixed: 2740889
Currently the rx tid spinlock is destroyed from
peer delete (based on peer state) or peer unmap handler.
This can lead to a race condition when the peer is
deleted before the peer map command is processed/received.
Fix the above race condition by destroying the rx tid
spinlock only when the peer handle is destroyed/freed.
Change-Id: Iaf7ccea11a95732c1aa20e66af6dd4a9a66517c5
CRs-Fixed: 2763533
In Tx data path get the peer reference and use
until there is a peer_id mismatch. This is to avoid
per packet atomic operations
Change-Id: I59cdbef19001c694654f0967381a9242788d9cbc
As part of the VoW IGMP improvements, which deals with
conversion of igmp packets to unicast packets, the
following changes are done when the new feature is enabled:
1. IGMP/MLD packets with special addresses (addresses not
part of any multicast group) will be converted to all
existing clients of the AP. These packets will be
directly fed to igmp multicast to unicast conversion
module.
2. The IGMP/MLD packets with group specific addresses will
first go to multicast enhancement module, where they
will be allowed to pass and will be converted to unicast.
3. The new feature will be enabled only when multicast
enhancement feature is enabled, which will be ensured
at the configuration level.
FR : 61063
Change-Id: I2fd5a67766c806432fbe1a12cb9654a3600100f5
In roam HO failure, peer unmap indication is processed
even before Host initiates the peer cleanup. In the cleanup,
as part of cdp_peer_teardown, AST entries are freed up without
removing it from the peer AST list since get_peer_by_ref_id
returns NULL peer pointer as peer is moved to inactive state.
Fix is to delete peer AST entries as part of peer unmap
indication so that the entries are removed from the AST list.
Change-Id: I1b04517e8cc392b33bb439501e61e3d5534d8307
CRs-Fixed: 2763825
Add support to update the count of number of Punctured Preamble
PPDU's transmitted by the AP.
Change-Id: Id85a490e9e80566e639f48190a96d3107c6e5c61
CRs-fixed: 2755288
STATUS_BUFFER_DONE tlv written in first word for a status
buffer indicates that DMA is done for that status ring entry.
In existing implementation, for a status ring entry if
STATUS_BUFFER_DONE tlv is not written by HW, we poll on to status ring
entry until DMA is done by HW.
During lmac reset it may happnen that HW will not write STATUS_BUFFER_DONE
tlv in status buffer, in that case we end up polling infinitely leading
to backpressure on monitor status ring.
As per MAC team's suggestion, when HP + 1 entry is peeked and if DMA
is not done and if HP + 2 entry's DMA done is set,
replenish HP + 1 entry and start processing in next interrupt.
If HP + 2 entry's DMA done is not set,
poll onto HP + 1 entry DMA done to be set.
CRs-Fixed: 2740988
Change-Id: Ieef667f0bb4a47e74fc320c93243c637409f47f0
For QCN9000, full monitor is controlled through,
a. full_mon_mode ini param
b. QCA_SUPPORT_FULL_MON compile time macro
Currently enabling full monitor is not taken care under
macro QCA_SUPPORT_FULL_MON.
As monitor mode is not supported for 256M profile,
QCA_SUPPORT_FULL_MON is disabled.
Enable full monitor mode when both ini and macro are enabled.
Change-Id: Id5be5f7edc4739423fd74e28bb71a8f3339d87b3
CRs-Fixed: 2744622
In cdp_soc_deinit peer hash table is freed to add
check to avoid NULL access from CDP APIs which may do
peer hash find
Change-Id: I47c4126c34bb64d43112a4f8e268da1a98f4124d
This change will add a build flag support to MSCS wherein
the functions will be called based on the status of the flag
created. This flag will be enabled by default.
Change-Id: I1c234c710d3b37f534ed2df7b37343cbd65ab930
CRs-Fixed: 2757257
Sometimes frames are queued to dp_rx_thread wait queue
during WoW suspend in progress. At this time dp_rx_thread
is in SUSPEND state and these frames can't deliver to Linux
Stack, for ICMP case, if no other frames come to wake up
system, ICMP timeout will happen. Add a RX suspend check
in WoW to avoid suspend when frame comes before WMI_WOW_ENABLE
sends to FW
Change-Id: I756d3127f75da8f0e6e303c1250c01ce0c6ac573
CRs-Fixed: 2756263
In tx desc store vdev_id to avoid unprotected access of
vdev memory.
In tx path pass VDEV as argument to all APIs
In completion path get vdev from peer as vdev reference
in peer is protected with reference, in places where peer
is invalid get vdev object using vdev_id stored in tx desc
with a reference count
Change-Id: I364ebb51143218872c0bc6f856772ab1df00bb17
CE2 status ring full condition results in WMI req timeout
in host resulting in self recovery being triggered.
Set IRQ affinity of CE IRQs to gold cores for defconfig
builds only. This is to alleviate the processing of all
DP rings and CE rings on CPU0.
CE2 status ring will be reaped fast enough to ensure host
has enough credits from FW to send out the WMI reqs to FW.
Change-Id: Ifef338faac17b44eae811d23ec41d89bde1b29c7
CRs-Fixed: 2738317
