Kalle Valo says:
====================
wireless-drivers-next patches for v5.10
First set of patches for v5.10. Most noteworthy here is ath11k getting
initial support for QCA6390 and IPQ6018 devices. But most of the
patches are cleanup: W=1 warning fixes, fallthrough keywords, DMA API
changes and tasklet API changes.
Major changes:
ath10k
* support SDIO firmware codedumps
* support station specific TID configurations
ath11k
* add support for IPQ6018
* add support for QCA6390 PCI devices
ath9k
* add support for NL80211_EXT_FEATURE_CAN_REPLACE_PTK0 to improve PTK0
rekeying
wcn36xx
* add support for TX ack
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
clang static analyzer reports this problem
mac.c:6204:2: warning: Attempt to free released memory
kfree(ar->mac.sbands[NL80211_BAND_2GHZ].channels);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The channels pointer is allocated in ath11k_mac_setup_channels_rates()
When it fails midway, it cleans up the memory it has already allocated.
So the error handling needs to skip freeing the memory.
There is a second problem.
ath11k_mac_setup_channels_rates(), allocates 3 channels. err_free
misses releasing ar->mac.sbands[NL80211_BAND_6GHZ].channels
Fixes: d5c65159f2 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200906212625.17059-1-trix@redhat.com
The return error check on the call to ath11k_pci_get_user_msi_assignment is
missing. If an error does occur, num_vectors is still set to zero and
later on a division by zero can occur when variable vector is being
calculated. Fix this by adding an error check after the call.
Addresses-Coverity: ("Division or modulo by zero")
Fixes: d4ecb90b38 ("ath11k: enable DP interrupt setup for QCA6390")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200819111452.52419-1-colin.king@canonical.com
In ath11k_qmi_prepare_bdf_download(), ath11k_core_firmware_request() is
called, but the returned pointer is not checked for errors. Rather the
variable ret (which will always be zero) is checked by mistake. Fix
this and replace the various gotos with simple returns for clarity.
While we are at it, move the call to memset, as variable bd is not used
on all code paths.
Fixes: 7b57b2ddec ("ath11k: create a common function to request all firmware files")
Signed-off-by: Alex Dewar <alex.dewar90@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200825143040.233619-1-alex.dewar90@gmail.com
The only usage of ath11k_thermal_ops is to pass its address to
thermal_cooling_device_register() which takes a const pointer. Make it
const to allow the compiler to put it in read-only memory.
Signed-off-by: Rikard Falkeborn <rikard.falkeborn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ath.git patches for v5.10. Major changes:
ath11k
* add support for QCA6390 PCI devices
wcn36xx
* add support for TX ack
ath9k
* add support for NL80211_EXT_FEATURE_CAN_REPLACE_PTK0 to improve PTK0
rekeying
For QCA6390, normal power up and power down can't bring MHI
to a workable state. This happens especially in warm reboot
and rmmod and insmod. Host needs to write a few registers to
bring MHI to normal state.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Carl Huang <cjhuang@codeauroro.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597555891-26112-10-git-send-email-kvalo@codeaurora.org
For QCA6390, wbm2sw1 is used for other purpose rather than
tx completion ring. So use TCL_DATA_RING 0 only for QCA6390.
Add MISC_CAPS_TCL_0_ONLY to control it.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Carl Huang <cjhuang@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597555891-26112-9-git-send-email-kvalo@codeaurora.org
QCA6390 doesn't enable V2 map and ummap event, so the addr search
flags and type is different from IPQ8074. Assign correct search flags
and type for QCA6390.
Without this change, ping sometimes fails. With this change, now ping
is always successful.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Carl Huang <cjhuang@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597555891-26112-7-git-send-email-kvalo@codeaurora.org
For QCA6390, it processes the reg chan list event only for phy0,
and it goes to fallback if the phy_id is not valid. For a valid
phy_id but not 0, just discard the event.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Carl Huang <cjhuang@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597555891-26112-5-git-send-email-kvalo@codeaurora.org
For QCA6390, only one pdev is created and this pdev manages both lmacs, thus
both rxdmas. So host needs to initialize all rxdma related rings for one pdev.
Another difference is for QCA6390, host fills rxbuf to firmware and firmware
further fills the rxbuf to rxbuf ring for each rxdma.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Carl Huang <cjhuang@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597555891-26112-4-git-send-email-kvalo@codeaurora.org
For QCA6390, it has 2 lmacs and thus 2 rxdmas. However, each rxdma has rxdma0
only, and doesn't have rxdma1. So for QCA6390, don't initialize rxdma1 related
rings such as rx_mon_buf_ring, rx_mon_dst_ring and rx_mon_desc_ring.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Carl Huang <cjhuang@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597555891-26112-3-git-send-email-kvalo@codeaurora.org
QCA6390 uses MSI interrupt, so need to configure msi_add and
msi_data to dp srngs. As there are so many DP srngs, so need
to group them. Each group shares one MSI interrupt.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Carl Huang <cjhuang@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597555891-26112-2-git-send-email-kvalo@codeaurora.org
For QCA6390, host puts hardware to Dual Band Simultaneous (DBS) mode by default
so both 2G and 5G bands can be used. Otherwise only the 5G band can be used.
QCA6390 doesn't provide band_to_mac configuration and firmware will do the
band_to_mac map.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Carl Huang <cjhuang@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597576599-8857-12-git-send-email-kvalo@codeaurora.org
This macro is evil as it's accesses ab variable in a hidden way. It's better
for readibility to access ab->hw_params.ce_count directly.
This is done in a separate patch to keep the patches simple. No functional changes.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597576599-8857-6-git-send-email-kvalo@codeaurora.org
This macro is evil as it's accesses ab variable in a hidden way. It's better
for readibility to access ab->hw_params.host_ce_config directly.
This is done in a separate patch to keep the patches simple. No functional changes.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597576599-8857-5-git-send-email-kvalo@codeaurora.org
QCA6390 uses only 9 Copy Engines while IPQ8074 may use 12, make it possible to
change CE configuration dynamically via hw_params.
The defines for host_ce_config_wlan and CE_COUNT are temporary solutions, they
will be removed in the following patches to keep things simple.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Carl Huang <cjhuang@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597576599-8857-4-git-send-email-kvalo@codeaurora.org
Now some of the HAL register macros access ab variable in a hidden way, make ab
variable visible in the macro by adding it as an argument.
This is done in a separate patch to keep the patches simple. No functional changes.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597576599-8857-3-git-send-email-kvalo@codeaurora.org
On QCA6390 reg_start and reg_size values are different from IPQ8074 so we need
to change the values runtime. As we can't modify a static const variable
hw_srng_config directly, instead use it as a template, copy it and modify the
copy with correct values.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597389030-13887-12-git-send-email-kvalo@codeaurora.org
On x86 and other non-qcom platforms, host needs to explicitly tell the firmware
to use the internal sleep clock. Some QCA6390 modules have OTP burnt with
external sleep clock selected, and these modules can't work expectedly unless
firmware selects internal sleep clock.
Add a field to hw_params to support this difference.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Carl Huang <cjhuang@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597389030-13887-11-git-send-email-kvalo@codeaurora.org
The decoded_size is wrongly assigned in ath11k_qmi_msg_handlers and it results
in out of boundary access in qmi_decode. The correct decoded_size should be
calculated from the related ind_msg structure.
This issue is exposed with QCA6390 because it needs 11 small memory chunks
which are stored in qmi_wlanfw_request_mem_ind_msg_v01 and hence the
decoded_size exceeds the wrongly assigend decoded_size.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Carl Huang <cjhuang@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597389030-13887-9-git-send-email-kvalo@codeaurora.org
On x86 it's sometimes difficult to allocate a large contigous DMA
memory, so instead allocate blocks of small chunk memory.
In ath11k_qmi_msg_mem_request_cb() the error handling was cleaned up to avoid
an unused variable warning. Also changed the test from (ret < 0) to just (ret)
as the functions don't return any positive values.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Carl Huang <cjhuang@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597389030-13887-8-git-send-email-kvalo@codeaurora.org
CE srng is used for control path and CE srng processing is done using tasklet
bottom half. Setup ce tasklet initialization and scheduling for control path.
Needed for PCI support.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Govind Singh <govinds@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597389030-13887-7-git-send-email-kvalo@codeaurora.org
QMI service instance id is used for qmi service lookup, IPQ8074 and QCA6390
uses different instance id for service lookup. Fill appropriate QMI service
instance id for respective targets.
Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.1.0.1-01238-QCAHKSWPL_SILICONZ-2
Signed-off-by: Govind Singh <govinds@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1597389030-13887-4-git-send-email-kvalo@codeaurora.org
