Change REO queue descriptor delete sequence to flush base and
extension descriptors separately, since flushing base descriptor
alone results in extensions flushed by HW much later,
sometimes after freeing descriptor memory, causing use-after-free
corruptions. Also increase REO command and status ring sizes
since number of commands per RX TID flush has increased with
this change.
Fix interrupt batch count setting for non pkt/buffer rings,
which includes REO status ring.
Also remove callbacks during reo command list destroy,
done as part of SOC detach. This should not be done since
callbacks will access stale data structures that are freed
as part of PDEV/SOC detach.
Change-Id: I4337454d3f53cc0b3c3014ef07cc2ba13cfebb18
CRs-Fixed: 2175989
Currently while reaping the rxda destination ring in dp_rx_mon_mpdu_pop,
we end up un-mapping the nbuf twice when -
1. msdu_ppdu_id > ppdu_id
2. msdu_ppdu_id = ppdu_id
For case 1, we exit from the dp_rx_mon_mpdu_pop without reaping the
descriptor completely, but we unmap the nbuf corresponding to it.
Next, when case 2 happens, we end up un-mapping the nbuf again to reap
the same descriptor.
Introduce a flag in the rx_desc to keep track of un-mapping the nbuf. If
the nbuf is already unmapped, donot unmap it.
Change-Id: Icb64fbc00312d6d0e6d41f7b475eb1285530c3d0
CRs-Fixed: 2185301
In monitor mode, when we try to stitch msdus together into an ampdu
to be sent to the stack (dp_rx_mon_restitch_mpdu_from_msdus),
we aggregate them together via frag_list. In the current code,
for the first msdu in which the frag_list is populated (if there are
more than one msdus to be sent to stack), the next pointer (skb->next)
is not set to NULL. This causes the HDD code to process the next
pointer leading to issues double free issues.
Set the next pointer for the first msdu(with frag_list) to NULL.
Change-Id: I60d1d463c7bbb602e4b199fbb691340ad6247bc5
CRs-Fixed: 2173405
Add APIs to be used by extended DP modules to attach/detach
their respective handles with Core DP module.
Change-Id: I63cf0883f7462a11b49666bda697c7e872d7b925
The buffer was not being freed after invalid peer handler.
Error are handled and returned but buffer free was not taken care.
Change-Id: I100cdc27e6ca1af3dde7ce6561f5dfb83b87127f
On Flushing txdesc resource on pdev detach caused null ptr
issue, resolved it by adding dummy function to not free desc
resources for mcl, as it is freed in vdev detach.
Change-Id: I9ab777571623c926b4ecbf9c95ba0837101eba26
CRs-Fixed: 2179488
Add support to transmit the original multicast frame (Control
frames) instead of dropping it in Tx path, if no snoop entries are
found for multicast address.
Change-Id: Ic8caeaf4b092f9368c8c2d76632941eb8d7bb9c8
Added CDP callback APIs so the DBDC repeater module
outside of cmn_dev can add, delete and update the AST
entries. Also with this change the sending of wds cmds
to fw is tied to adding AST, deleteing AST and updating
the AST entries.
Change-Id: I48d31962236860f50e16b771abaf3b8825b798df
ppdu stats timestamp,duration,channel etc were not
updated and corrected the same
Also RSSI was incorrectly getting updated for
mcast frames also, fix both of the above issues
Change-Id: Id8446c6c0cfe8761f8c524fc7af0ff5bdd8b96aa
Inactivity timer initialization is moved from dp
attach to soc attach as the timer is being
initialized multiple times in case of multiple
radios
Change-Id: I85d1d8710ccc7531c43ec82378c3ca0b65d26cf9
1. Assert if length received from FW is not equal to
the expected length
2. Count the number of PPDU stats received by host and
print in "iwpriv athX txrx_stats 260"
3. Optimize the PPDU stats tag mask sent to FW for
enhanced stats enabling
Change-Id: I60019de93bb3914955a03546504efc820c960746
CRs-Fixed: 2170704
Assert happened in rx path because rx reo rings setup
is done and after that default routing information is
updated, if any time elasped any buffer received in
exception ring falsely. Setting route information before
intialize.
Change-Id: I48c7292db0823d3eac639429f956dff1fcc41cae
When removing entries on reo_cmd_list, TAILQ_FOREACH used
instead of TAILQ_FOREACH_SAFE, that will only remove the first entry
and leak the others.
CRs-Fixed: 2184882
Change-Id: Ica2677925af34e62f98f4a8d2044c9c3feb69ba9
To expose cdp_pdev_stats to athstats, move cdp_stats from
cdp_txrx_cmn_struct.h to cdp_txrx_stats_struct.h and include
cdp_txrx_stats_struct.h wherever needed
Change-Id: I0e86527d47f5c6056a3bfbf5e8151594879427f3
CRs-Fixed: 2183725
with fragmented pkts re-injected back to REO destination
ring as frag_list, we need to linearize these nbufs while
handling intra-bss forwarding cases.
Change-Id: Id692974ac5d80f369fafae39aec32e2066dffb0b
On Firmware auto recovery , txdesc resources
might not be freed if f/w has not sent the
completion. Freeing those buffers in pdev
detach
Change-Id: I7d62258347f1fc9566fc8230564bcaf458ea4cc4
STATUS_DONE is missing sometimes in monitor status ring,
causing stalled processing in host and backpressure to RXDMA.
Adding a WAR to skip such entries.
Also increased the size of monitor link descriptor ring since
multiple descriptors are used for large MPDUs.
Change-Id: I556b4196482ed738471afc1f7e7b73cf3f8fbc51
Set update_peer_cache, learning_frame bits in meta header for all
the frames that are sent to firmware.
Change-Id: Id8dac0007cc87f792859e5108ddfa77746166b49
CRs-fixed: 2180223
Set link descriptor memory address to NULL after freeing to ensure
it is not freed again. This is seen with soc_detach called during
attach failures.
Change-Id: I7338cb8c64fcb652c95f42bcb9e998a6f043cecf
CRs-Fixed: 2160703
When TLVs come from TAC_thread last tlv sent by FW is
HTT_PPDU_STATS_SCH_CMD_STATUS_TLV
and in Tqm thread HTT_PPDU_STATS_USR_COMPLTN_ACK_BA_STATUS_TLV is the last
TLV to be sent. HTT_PPDU_STATS_USR_COMPLTN_ACK_BA_STATUS_TLV though can
come either in tqm thread or tac thread based on tid type.
Add change to process HTT_PPDU_STATS_USR_COMPLTN_ACK_BA_STATUS_TLV when it
comes from tqm thread and pass the populated buffer to the appropriate APIs
Change-Id: I7a064aba4e3c6814b63dcf4bb97cfb3be0194c39
CRs-Fixed: 2180477
The map\unmap checking logic need to be revised. Before that, it
is needed this hotfix to unblock other test.
CRs-Fixed: 2175684
Change-Id: Ife8147526307924079be9081eaf17323f67b87c0
pass psoc/pdev object as ctrl pointers to instead of legacy objects.
It introduces new APIs to get feature pointers of psoc/pdev.
DP attach API param modified to pass ctrl psoc (psoc object)
Change-Id: I3d83f28c3e72e5ee984a4220b9d5cf3c4522b501
CRs-Fixed: 2177109
Add API to return the msdu link descriptor from obtained from NSS back to
host link descriptor pool.
CRs-Fixed: 2178692
Change-Id: Ib4f967ea887e013d6461ba760a45dbda6fd92baa
dp_tx_send_exception is an exception path that transmits frames
through FW by setting DP_TX_DESC_FLAG_TO_FW
Change-Id: Id6abfcd2bf060c4fe3aaca91634f3be501530a7e
CRs-Fixed: 2176814
Move the peer Rx activity to PPDU_stats_handler
which is common for both NSS offload and Host datapath modes.
Change-Id: I06c3dd14d117cc291e3b19076ddfc7f3be00e602
CRs-Fixed: 2176385
1. find_first_bit API returns Size value if it doesn’t find any bit set.
This was causing VOICE AC counter to go high even for TID 0
2. Reset stats was not resetting the stats immediately
3. Added checks to avoid spurious stats increment
4. Detect data/mgmt pkt to avoid spurious stats increment
Change-Id: Ic9c5695fc7bdf58d1237e7e20bb79378a02a21f6
Add temporary fix to drop all encrypted fragments
until they can be handled properly.
Change-Id: I87a9c52d6359be8182c28d3fa047b9a65f3dd129
CRs-Fixed: 2174750
Add tid to the defrag waitlist & timeout if further fragments are not
received.
Change-Id: Iff2d2c23fe796cf70bcc6aa43ab02a308a33ee9f
CRs-Fixed: 2174750
- Handle WDS case during encapsulation
- Drop non-head duplicate fragments from list
- Get REO DST ring id from msdu_info of incoming packet
Change-Id: I7e4c9c211548aafe9b4ba88fb12890f38aced324
CRs-Fixed: 2174750
REO queue setup is getting triggered from processing
pending null queue exceptions in WBM release ring, after
sending peer delete command to FW, but before peer unmap
is received, causing target assert. Mark peer deletion
before sending peer delete command to avoid this.
Change-Id: I2d85d01e049cb37d9321beeee8e41f44e40e68ad
CRs-Fixed: 2155420
Add ppdu tid tlv value assignment in cdp dp stats layer
ppdu tid value is used by apstats command it was earlier
showing zeroes
Change-Id: Id24b08d2424dcc09424624d84044b3963bd81b36
cck rates were getting updated for ofdm rates
and nss index were getting wrongly incremented
fixed the same
Corrected sgi and preamble too
Change-Id: I2dabb685de6e0bd4ba5094845cde228b3c32e2a9
* Update the pkt_type counters correctly within the array bound
* Add support for NSS counters on Tx side
* Update ampdu flag in PPDU struct
* Add support for ampdu flag in hal_ppdu struct
Change-Id: I79ea52727124ea4be4d82912cb2513ea58e2af10
Update mon_rx_status fields for HE case. Also
added rs_fcs_err field to structure and populate
its value.
Change-Id: I4681543c7d6c6c7ac0d666cd93072caaee285585
The nbuf is not freed for all possible branches in rx_wbm_error_pr
cess function. Add fix to ensure the nbuf free for all conditions.
CRs-Fixed: 2171899
Change-Id: I8e40fc7dc5afb0be0257244e1e704a0200ffacc0
Remove dependency of extender ap on VAP layer and radio(ic) layer data
structure
Change-Id: Ib336f518847abd1b29f5ea3cfd1a508dc2cd14e8
CRs-Fixed: 2148247
The pointer to peer is dereferenced after memory free in peer
_unref_delete function. Fix the issue by changing function
call sequence.
CRs-Fixed: 2167698
Change-Id: Iebd8f958bde4964c16a3c312d037e8f06e319350
Sending all frames to firmware will have high throughput impact.
Queue mesh frames also to TCL/TQM and send cloned frame to firmware
only if meta header information is changed to update meta header
information. Increased meta info size as per new ext2 header.
Change-Id: I89ff27994477a5637979a9cd6a397286144b2535
CRs-Fixed: 2162643
