Currently cdp ops are given pdev/vdev/peer
handle as its arguments, which is directly
accessed in those APIs. This can cause a
race-condition in access of the respective
handles if it has been deleted in parallel.
Hence as a part of cdp convergence, pass only
the pdev_id or vdev_id or peer mac address,
which will be used to get the respective handles,
and hence avoiding the unwanted access of the
handles if it has been deleted.
- ipa_get_resource
- ipa_set_doorbell_paddr
- ipa_set_active
- ipa_register_op_cb
- ipa_get_stat
- ipa_tx_data_frame
- ipa_uc_get_share_stats
- ipa_uc_set_quota
- ipa_enable_autonomy
- ipa_disable_autonomy
- ipa_setup
- ipa_enable_pipes
- ipa_disable_pipes
- ipa_rx_intrabss_fwd
Change-Id: I678d7a7de7132417ff6051b0fd6da5d14426d21e
CRs-Fixed: 2540861
For wlan_ipa_setup_iface, if found the net_dev is already there, but the
device_mode is different, will remove this obsolete record and update it
with new interface context for current IPA WLAN event.
Change-Id: Ib9ecd7f7b3b9603c37dc1d69cd0f3a2868c978ec
CRs-Fixed: 2567222
Currently the cdp apis are given pdev/vdev/peer
handle as its arguments, which is directly
accessed in those APIs. This can cause a
race-condition in access of the respective
handles if it has been deleted in parallel.
Hence as a part of cdp convergence, pass only
the pdev/vdev id or peer mac address, which will be
used to get the respective handles, and hence
avoiding the unwanted access of the handles if
it has been deleted.
Converged flowctl_ops
- flow_pool_map_handler
- flow_pool_unmap_handler
- dump_flow_pool_info
- tx_desc_thresh_reached
CRs-Fixed: 2539738
Change-Id: I9b11e66d8c2644f16e0f2464c323798a4cba007a
In DBS scenario(SAP + STA) when all SPA clients disconnect, while STA is
still connected, IPA pipes are disabled on the lithium target. At this
time, its possible that some packets TX from IPA over WLAN are still
pending. If these completions come after IPA pipes are disabled, it can
lead to a NOC error, since the GSI doorbell register for WBM2SW2 ring
may be clock gated (after IPA pipes are disabled).
To avoid this situation, wait for some time before disabling IPA pipes.
IPA pipes are disabled after a timeout, when system suspend call tries
to suspend the bus. A driver unload or a softap tear-down will also
disable the pipes.
Change-Id: I542049fa19d0dcf5c31d9b8a2d836388847dd6c1
CRs-Fixed: 2553670
Support wlan ipa clk voting for kona. Host should provide
bandiwdth levels to IPA driver for which IPA uc monitors
the levels. Once the threshold is reached IPA uc interrupts
the IPA driver and IPA driver informs host driver via callback
registered.
Change-Id: I9fd805d69858a413f20b9e55a9c02a82054c646b
CRs-Fixed: 2526300
When doing IPA initialization, WLAN driver registers a
notification callback for IPA uC ready. In the callback,
we directly set uc_loaded to true before scheduling a work
task to connect WLAN IPA pipes. However issue is that in
the work handler, uc_loaded is checked and if it is true,
we return without connecting WLAN IPA pipes.
Therefore properly set uc_loaded to true so that WLAN IPA
pipes can be connected correctly.
Change-Id: I27a14e627a1a75836f08dc5dec8c8b1d09725e10
CRs-Fixed: 2537625
Implement metering stats quota to support ipa offload sta plus
sap wifi sharing use cases.
Change-Id: Ic9d5ad817ffb4d671a43f3f3aebb2d8cce293873
CRs-Fixed: 2517696
Merge from 2517936 & 2490501.
SMMU fault is observed for WDI 1.0 MCC TX when SMMU is enabled.
Issue is that dma_addr passed from IPA driver is iova of IPA
domain. WLAN driver directly uses the iova to do transmission.
Since iova is not mapped in WLAN domain, SMMU fault occurs.
Fix is to map skb to WLAN domain for TX when SMMU is enabled.
After doing DMA map, an offset of frag_header and ipa_header
is added to the iova and skb->cb is updated with the new iova.
After TX completion, skb DMA unmap is done but with the iova
of the above offset included, which leads to unmapped iova error.
Fix is to recalculate the iova and update skb->cb with new iova,
which has frag_header and ipa_header length removed.
Change-Id: Ib0112d6a3de6dc9af43e5c48a49085c7d4f41a9b
CRs-Fixed: 2534845
During wlan IPA suspend, force the bandwidth voting to high
when CFG_DP_IPA_ENABLE_FORCE_VOTING is enabled.
Change-Id: I9b368081f9c0919f5fecfacf318ad1a714cc33cf
CRs-Fixed: 2521815
There is a race condition of vdev_to_iface, it is referenced in IPA W2I
callback routine, but it's updated after IPA enable operation. If frames
received right after IPA enabled, callback function will return failure
due to vdev_to_iface value is not filled.
Change-Id: Iff831b66771539d1de4af89e395821dc50d13af3
CRs-Fixed: 2517515
In __wlan_ipa_wlan_evt, for QDF_IPA_STA_DISCONNECT event,
ipa_ctx->vdev_to_iface[session_id] is reset back to
WLAN_IPA_MAX_SESSION only when ipa_ctx->sap_num_connected_sta
is greater than 0. So, in STA+SAP concurrency case, when
no client is connected to SAP, if STA is disconnected,
vdev_to_iface is not reset for that particular session id,
making it unusable for further events involving IPA iface
creation.
Fix is to move the code to reset vdev_to_iface outside the check
for ipa_ctx->sap_num_connected_sta in QDF_IPA_STA_DISCONNECT
scenario, so that vdev_to_iface is reset irrespective of SAP.
Change-Id: I06c79b62c540546cf4842f81d235f96b2965661e
CRs-Fixed: 2525709
Set the same IPA bandwidth for both the cons and prod pipes based
on total packet counts, to avoid overwrite voting by one another,
as per IPA-PM implementation.
Change-Id: I5f93e1ff91f9fbeb42ea8bd93792e288a073b6b1
CRs-Fixed: 2521811
If STA disconnect failed for any reason, cleanup IPA STA iface
if not already done.
Change-Id: I27ff33324bc4724e8470af9a0c434fa03e8aa5c3
CRs-Fixed: 2505563
For runtime PM if the bus is suspended driver need to consider
extra 6 sec time for bus resume.
Thus add 6 sec extra in WMI timeouts if runtime PM is supported.
Change-Id: I5515cc889a0315382bac11a33ea6f901b7af1c46
CRs-Fixed: 2507029
Local sta_id or peer_id is being cleaned up across DP, HDD
and PS/WMA. So, any references to local peer_id/sta_id will
be replaced by peer mac address and all interactions
between the layers will be based on peer mac address.
Change-Id: I78230bf1e1ec090e83245bff04953fa2e03b91cd
CRs-Fixed: 2504013
On MDM platform, low throughput is observed for SAP RX
terminated scenario. In case of SAP, SAP device itself
is added as a slave to the bridge device, which acts as
a router device.
For SAP RX terminated scenario, destination MAC address
of RX packets is the bridge device. However for WDI 3.0,
WLAN IPA has to check if destination MAC is a valid peer
for SAP intra-bss forwarding.
Currently in intra-bss forwarding, dp_find_peer_by_addr()
is used to look up if destination MAC address is a valid
peer. Issue is that in dp_find_peer_by_addr() peer
spinlock is grabbed via qdf_spin_lock_bh. IPA RX is in
process context and qdf_spin_lock_bh turns to spin_lock_bh,
which hurts netif_rx badly.
Fix is to check if RX packet is bridge device terminated
in IPA component module so as to bypass intra-bss
forwarding logic and to make the RX path BH lockless.
Note that if there's a BH lockless API to check if
destination MAC address is a valid peer, this check will
not be necessary.
Change-Id: I00fca3e75ae55ae1d2f591fddd00e49dfb634c5c
CRs-Fixed: 2494442
As for Rome, currently it usually just only support two
IPA interfaces offload, if use the default value 3, which
will cause the IPA pipe setup failure issue. So export
the max IPA interface numbers, and it's better to set
the appropriate value from build file according to
different requirement for different chipset and platform.
Meanwhile, it support SMMU on the sa415 platform, so
enable SMMU for it.
Change-Id: I2de31bcb4d38f5e7964d2cbdc2fc6f143eef510d
CRs-Fixed: 2480627
wlan_ipa_w2i_cb is notify cb for IPA RX terminated
scenario and therefore it is invoked on a per-packet
basis.
qdf_op_protect and qdf_op_unprotect leads to ~100mbps
througput degradation for IPA RX terminated scenario
on MDM platforms. Thus remove qdf_op_protect and
qdf_op_unprotect on MDM platforms as a temporary
workaround to unblock releases and meet KPI. Note that
this will be revisited.
Change-Id: I1cab0581288456662ee4aa5d63df1b0f1213436b
CRs-Fixed: 2469073
During resource unloading, if two next to next client connect
request comes, one is handled in pending event handler in
wlan_ipa_uc_op_cb callback after processing FW TX suspend response
and other one is handled in scheduler thread, this causes race
condition in sending IPA offload enable for STA iface. So send IPA
enable req to FW before queuing the client connect event to the
pending events.
Change-Id: Ifcabc2481e590e5e128434614c4b46fe53a34636
CRs-Fixed: 2460573
Abstract the callback hdd_softap_ipa_start_xmit by
hdd_softap_ipa_start_xmit, to match the return type of
wlan_ipa_softap_xmit pointer, to fix a CFI failure.
Change-Id: If1cb1a80801558d5c7831ec89ad5281b0bbe36f3
In STA-SAP config, IPA offload is enabled only when first
client is connected to SAP. This is OK for MSM platforms.
But for MDM platforms, this is not the case. There's STA
tethering cases, where traffic is routed between STA and
USB tethering devices while SAP is idle and no wifi clients
are connected to SAP. In such case, IPA offload needs to
be enabled for STA to promote performance.
Fix is to use available MDM_PLATFORM marco and STA only IPA
support is decided upon driver build time.
Change-Id: If2bc016aa941f2c0651a2669f1169e631208326b
CRs-Fixed: 2433608
Check for validity of vdev obtained from the CDP
cdp_get_vdev_from_vdev_id API.
Change-Id: Ie7fb99a0e08135bd3f2cc6b63070a4bd72df943a
CRs-Fixed: 2434583
Support WDI 3.0 SW path intra-bss forwarding. Major
difference for WDI 3.0 is the metadata info passed
from ipa driver in skb->cb[].
Previously intra-bss fwd decision is done by FW and
it passes fw_desc to IPA where IPA driver passes onto
WLAN driver. Now for WDI 3.0, FW is not involved in RX
path and SW path intra-bss fwd decision has to be done
in wlan driver.
Change-Id: I0b4b9bbf16440dd47b6f97460f8931ea3a79303c
CRs-Fixed: 2426857
When vdev is logically deleted, vdev reference will
be returned as NULL by objmgr.
Remove vdev sanity from QDF_BUG while sending IPA evt
in __wlan_ipa_wlan_evt.
Change-Id: Ie7f9697768d53c71a6bead377c35bbc3f62e808a
CRs-Fixed: 2433521
The only function called through this pointer type doesn't return
a value and the caller doesn't check for a return value. Change
the function pointer type to return void to fix a CFI failure due
to a type mismatch:
CFI failure (target: [<ffffff853ba7cee8>]
hdd_ipa_send_skb_to_network+0x0/0x4 [wlan]):
...
Call trace:
__ubsan_handle_cfi_check_fail+0x50/0x58
__cfi_check+0x31fa0/0x3d5a4 [wlan]
wlan_ipa_w2i_cb+0x5fc/0x8fc [wlan]
...
Change-Id: I94f6bb0fc06acebde12cf80294999d9419296c4a
CRs-Fixed: 2431600
ipa disable pipes request is honoured based on
ipa_pipes_down flag. However, there is a possible
synchronization issue if disable pipes request is
initiated from different contexts, as there is no
locking to protect this flag.
Add pipes_down_in_progress flag in ipa_ctx and
use this flag to ignore disable pipes requests if
one request is already in progress.
Change-Id: Iabf01249d4b6a551e03cea2ae96a0c6096f43ff8
CRs-Fixed: 2423155
qdf_delayed_work_t has been superseded by qdf_delayed_work. Replace all
instances as appropriate.
Change-Id: I83f0dcec050bc9f91624048bd171f9a86b0394b5
CRs-Fixed: 2420202
Resolve ipa compilation issue for targets with config
ipa unified api enable and have kernel version 4.9.
Change-Id: Ibf7c51e32850024ceeb154f18a272044482af5ba
CRs-Fixed: 2422172
Upon client connection to SAP, WLAN IPA will pass a message
containing client MAC address and ethernet header offset to
IPA driver. Per WDI 3.0, there's no frag_header and
ipa_header prepended in front of ethernet header. Thus
ethernet header length is passed to IPA driver as header
length during interface registration. But in per-client
message, offset passed to IPA driver still accounts for
length of frag_header and ipa_header. As a result packet
queued to event ring will have wrong ethernet header.
And this causes IPA to WLAN TX HW path not work as expected.
Fix is to define frag_header and ipa_header as 0 sized
structure for WDI 3.0 so that ethernet header offset will
be 0.
Change-Id: I58cdf78732ad2909bf89058f2a5bc24c42f841d2
CRs-Fixed: 2421315
Per the Linux Kernel coding style, as enforced by the kernel
checkpatch script, pointers should not be explicitly compared to
NULL. Therefore within ipa replace any such comparisons with logical
operations performed on the pointer itself.
Change-Id: I82e4bed12122ad5aea47650f6d75e72a52207883
CRs-Fixed: 2418401
Session_id in __wlan_ipa_wlan_evt is used as an index
to access the array ipa_ctx->vdev_to_iface[].
Add sanity check for session_id to avoid invalid access.
Change-Id: Ie6079924c4772c3fa892b4537120db136a899885
CRs-Fixed: 2395162
Currently second STA wlan ipa interface is not deleted in case
of STA-STA concurrency for second sta disconnection. At the time
of driver recovery first virtual interface is deleted and later
existing ipa interfaces are deleted. This leads to invalid access
of second STA ipa interface net dev at the time of deletion during
driver recovery. Fix this by deleting the interface at the time of
sta disconnect event.
Change-Id: I21e46fe5c6f0b0213d139050acf229cb5ab914eb
CRs-Fixed: 2376897
Excessive wlan ipa call back error logs when driver is in
transition stat which may leads to watchdog timeout. So rate
limit the logs.
Change-Id: I1398ea67abde53d3a5f0b5ab7c7ea1dd8553e4cc
CRs-Fixed: 2379522
Add cdp api to check if tx desc pool available descriptor
threshold has reached.
Change-Id: I12cc38b56a68ca7ee8f4560677fcdabccb38928b
CRs-Fixed: 2369244
In case of lithium dp there is no FW interaction in RX packet
processing and is HW REO ring based. So modify the ipa component
is fw activated api to return true if the ipa system pipes are
enabled.
Change-Id: I8cde4b0f696b75097e6b50f8b0c71c2b6fff6d6d
CRs-Fixed: 2374083
In case of lithium dp there is no FW interaction in RX packet
processing and is HW REO ring based. So in case of handle
first client connection and last client disconnection host
should not send HTT FW requests for pipe resume/enable at FW.
So modify the enable and disable ipa pipes host functions for
lithium dp.
Change-Id: I72661050fe1f85c13aca0c78b5c0369884ec2999
CRs-Fixed: 2369252
Almost all of the definitions in txrx.h are obsolete. Rather than
remove the obsolete definitions and leave behind a file containing
only the two definitions that are used, move those useful definitions
to ol_txrx_types.h and completely remove txrx.h.
A side effect of this change is that a couple of unrelated files no
longer build because some of their dependencies were being met as a
side effect by txrx.h including other files. To solve the build issues
explicitly include the necessary header files where they are needed.
Change-Id: Ic3e3b36939e4ab04831d41a1e230bfab3da086a2
CRs-Fixed: 2363231
This reverts commit 2459533, which fixes WDI 1.0 MCC SMMU fault
but it's only a partial fix. And we also need 2490501 to be a
complete fix. But 2490501 has regressions and leads to NULL
dereference, and it's reverted in 2517642 until a new complete
fix is ready.
Based on above, revert 2459533 and WDI 1.0 MCC SMMU datapath
will be a known issue.
Change-Id: I6444bfd8552906d889531b7e77ca6aa6b6fa7d6d
CRs-Fixed: 2352927
This reverts change I7d04e9b01fbd973601ee9e4e74c6bf4dd3d0f7bb
This patchset is causing wlan_ipa_nbuf_cb to be called even when ipa_ctx
is null. wlan_ipa_nbuf_cb should not be called for non-IPA packets. Also
there should be a NULL check for ipa_ctx in the function.
Reverting the change may lead to SMMU errors during AP-AP MCC mode.
CRs-Fixed: 2352812
Change-Id: I2f2382ca12a14a09ed124ccd18a7f450282c5975
Excessive failure log due driver is in transition state. This
is coming from stats callback function, rate limiting the logs.
Change-Id: Idcd5304c492fa538614ac45d4c7bb3a601f23bcc
CRs-Fixed: 2347721
For WDI 1.0 MCC TX, wlan driver is involved for TX and skb is
provided from IPA driver. With SMMU enabled, dma_addr is of
IPA domain. In such case, wlan driver needs to do DMA map
itself. After doing DMA map, an offset of frag_header
and ipa_header is added to the iova and skb->cb is updated
with the new iova. After TX completion, skb DMA unmap is done
but with the iova of the above offset included, which leads to
unmapped iova error.
Fix is to recalculate the iova and update skb->cb with new iova,
which has frag_header and ipa_header length removed.
Change-Id: I7d04e9b01fbd973601ee9e4e74c6bf4dd3d0f7bb
CRs-Fixed: 2338049
SMMU fault is observed for WDI 1.0 MCC TX when SMMU is enabled.
Issue is that dma_addr passed from IPA driver is iova of IPA
domain. WLAN driver directly uses the iova to do transmission.
Since iova is not mapped in WLAN domain, SMMU fault occurs.
Fix is to map skb to WLAN domain for TX when SMMU is enabled.
Change-Id: I4c029a1e4dfff149e2784f5b95fe1f3cf3e3719e
CRs-Fixed: 2318537
