- 08 Mar, 2024 4 commits
-
-
Baochen Qiang authored
Currently host relies on CE interrupts to get notified that the service ready message is ready. This results in timeout issue if the interrupt is not fired, due to some unknown reasons. See below logs: [76321.937866] ath10k_pci 0000:02:00.0: wmi service ready event not received ... [76322.016738] ath10k_pci 0000:02:00.0: Could not init core: -110 And finally it causes WLAN interface bring up failure. Change to give it one more chance here by polling CE rings, before failing directly. Tested-on: QCA6174 hw3.2 PCI WLAN.RM.4.4.1-00157-QCARMSWPZ-1 Fixes: 5e3dd157 ("ath10k: mac80211 driver for Qualcomm Atheros 802.11ac CQA98xx devices") Reported-by:
James Prestwood <prestwoj@gmail.com> Tested-By: James Prestwood <prestwoj@gmail.com> # on QCA6174 hw3.2 Link: https://lore.kernel.org/linux-wireless/304ce305-fbe6-420e-ac2a-d61ae5e6ca1a@gmail.com/Signed-off-by:
Baochen Qiang <quic_bqiang@quicinc.com> Acked-by:
Jeff Johnson <quic_jjohnson@quicinc.com> Signed-off-by:
Kalle Valo <quic_kvalo@quicinc.com> Link: https://msgid.link/20240227030409.89702-1-quic_bqiang@quicinc.com
-
Jeff Johnson authored
Currently the soc_dp_stats debugfs file has the following permissions: # ls -l /sys/kernel/debug/ath11k/pci-0000:03:00.0/soc_dp_stats -rw------- 1 root root 0 Mar 4 15:04 /sys/kernel/debug/ath11k/pci-0000:03:00.0/soc_dp_stats However this file does not actually support write operations -- no .write() method is registered. Therefore use the correct permissions when creating the file. After the change: # ls -l /sys/kernel/debug/ath11k/pci-0000:03:00.0/soc_dp_stats -r-------- 1 root root 0 Mar 4 15:15 /sys/kernel/debug/ath11k/pci-0000:03:00.0/soc_dp_stats Tested-on: WCN6855 hw2.1 PCI WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.30 Signed-off-by:
-
Jeff Johnson authored
ath11k inherited struct wmi_start_scan_arg from ath10k. However, in ath11k, this struct is unused -- struct scan_req_params is used for this functionality. So remove the unused struct. No functional changes, compile tested only. Signed-off-by:
-
Jeff Johnson authored
ath12k inherited struct wmi_start_scan_arg from ath11k. However, in ath12k, this struct is unused -- struct ath12k_wmi_scan_req_arg is used for this functionality. So remove the unused struct. No functional changes, compile tested only. Signed-off-by:
-
- 05 Mar, 2024 25 commits
-
-
git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/athKalle Valo authored
ath.git patches for v6.9 Only some minor cleanup in ath11k and ath12k. Adding Jeff as the maintainer for ath10k, ath11k and ath12k DT bindings.
-
Chin-Yen Lee authored
Now WoWLAN firmware will disable PCIE DMA after driver call cfg_wake function, and it will lead to release offload packet fail because driver can't receive completion notification from firmware. We move release offload packet earlier to avoid this error. Signed-off-by:
Chin-Yen Lee <timlee@realtek.com> Signed-off-by:
Ping-Ke Shih <pkshih@realtek.com> Signed-off-by:
Kalle Valo <kvalo@kernel.org> Link: https://msgid.link/20240302005828.13666-8-pkshih@realtek.com
-
Chin-Yen Lee authored
The security engine is set for management frames by default for 802.11be chips, so no need to set it in WoWLAN flow. Signed-off-by:
-
Chin-Yen Lee authored
The setting during suspend or resume is different between different generation, so update it. Signed-off-by:
-
Chin-Yen Lee authored
The registers to configure mac function for WoWLAN mode that are different from different generation, so update them. Signed-off-by:
-
Chin-Yen Lee authored
The register of control and polling function for TX/RX DMA is different from different generation, so update them. Also rename polling_dma function to polling_dma_idle to avoid misunderstanding. Signed-off-by:
-
Chin-Yen Lee authored
The statue register is for driver to check if WoWLAN mode works or stops successfully. It is changed for new generation, so update it. Signed-off-by:
-
Chin-Yen Lee authored
The WoWLAN reason register is used for driver to get the wakeup reason for reporting to cfg80211, and it is different from chips. So put it into chip information. Signed-off-by:
-
Duoming Zhou authored
The kzalloc() in brcmf_pmksa_v3_op() will return null if the physical memory has run out. As a result, if we dereference the null value, the null pointer dereference bug will happen. Return -ENOMEM from brcmf_pmksa_v3_op() if kzalloc() fails for pmk_op. Fixes: a96202ac ("wifi: brcmfmac: cfg80211: Add support for PMKID_V3 operations") Acked-by:
Arend van Spriel <arend.vanspriel@broadcom.com> Signed-off-by:
Duoming Zhou <duoming@zju.edu.cn> Reviewed-by:
Kees Cook <keescook@chromium.org> Signed-off-by:
-
Ching-Te Ku authored
The 2 Bluetooth profiles (Hands free profile & Human interface device) have high duty transmission, it will affect the traffic of WiFi packet frequently. And once the WiFi traffic down to B/G mode, it will need a better success rate to recover the transmission rate. Add new policy option to solve the above situation. Signed-off-by:
Ching-Te Ku <ku920601@realtek.com> Signed-off-by:
-
Ching-Te Ku authored
If Bluetooth is working, it will update their info regularly. And the code will increase the counters while the info updating. Use this counter to judge is Bluetooth working or not. Don't need to set Bluetooth power or gain when it is not working. Signed-off-by:
-
Ching-Te Ku authored
Add return value to H2C function, and only record down the value while H2C command success, this can help us to check the real time status. Signed-off-by:
-
Ching-Te Ku authored
Wi-Fi firmware need some driver information to do decision or do some real-time control. Driver will update these information by H2C command. The chip 8922a H2C command index is different from before chips/branch, so need to assign the correct index to let firmware parsing. Signed-off-by:
-
Ching-Te Ku authored
Change structure member from bit field to normal variable to reduce unnecessary translation. Signed-off-by:
-
Ching-Te Ku authored
To avoid using bit fields for H2C command, rearrange the structure. And also patch the corresponding code for the using of this structure. No logic changes for existing chips. Signed-off-by:
-
Ping-Ke Shih authored
Under some circumstances, coexistence mechanism want to keep grant BT or WiFi, such as inquiry and WiFi is connecting, to ensure BT or WiFi can transmit or receive data in that period. Signed-off-by:
-
Ping-Ke Shih authored
When hardware grant BT initially but transition to grant WiFi, the PLT (polluted) bit is set to assist coexistence mechanism to debug if grant signal is expected. Signed-off-by:
-
Johannes Berg authored
The code here looks well thought-out, so it seems likely that the byte-swaps are actually needed, due to SPI bus and device behaviour. Restore the byte-swapping, in a way that doesn't result in sparse warnings. Fixes: 7ceade65 ("wifi: cw1200: fix __le16 sparse warnings") Signed-off-by:
Johannes Berg <johannes.berg@intel.com> Signed-off-by:
-
Javier Martinez Canillas authored
When these failures happen, the warning and call trace is printed which is excessive. Instead, just print the error but rate limited to prevent warns to unnecessarily pollute the kernel log buffer and make the serial console practically unusable. For example, on an AM625 BeaglePlay board where accessing a SDIO WiFi chip fails with an -110 (ETIMEDOUT) error: $ dmesg | grep "sdio write\|read failed (-110)" | wc -l 39 Signed-off-by:
Javier Martinez Canillas <javierm@redhat.com> Reviewed-by:
Breno Leitao <leitao@debian.org> Signed-off-by:
-
Bitterblue Smith authored
PCI drivers and USB drivers can both use the same function, read_efuse_byte(), and they can call it directly. rtl8192de was the only user. Tested only with the upcoming rtl8192du driver. Signed-off-by:
Bitterblue Smith <rtl8821cerfe2@gmail.com> Acked-by:
-
Bitterblue Smith authored
total_fa_cnt is supposed to include cck_fa_cnt and ofdm_fa_cnt, not just ofdm_fa_cnt. Fixes: 96036123 ("rtw88: 8821c: add false alarm statistics") Signed-off-by:
-
Bitterblue Smith authored
Tenda U9 V2.0, which contains RTL8811CU, is practically unusable because of frequent disconnections: Feb 23 14:46:45 ideapad2 wpa_supplicant[427]: wlp3s0f3u2: CTRL-EVENT-BEACON-LOSS Feb 23 14:46:46 ideapad2 wpa_supplicant[427]: wlp3s0f3u2: CTRL-EVENT-DISCONNECTED bssid=90:55:de:__:__:__ reason=4 locally_generated=1 Feb 23 14:46:52 ideapad2 wpa_supplicant[427]: wlp3s0f3u2: CTRL-EVENT-CONNECTED - Connection to 90:55:de:__:__:__ completed [id=0 id_str=] Feb 23 14:46:54 ideapad2 wpa_supplicant[427]: wlp3s0f3u2: CTRL-EVENT-BEACON-LOSS Feb 23 14:46:55 ideapad2 wpa_supplicant[427]: wlp3s0f3u2: CTRL-EVENT-DISCONNECTED bssid=90:55:de:__:__:__ reason=4 locally_generated=1 Feb 23 14:47:01 ideapad2 wpa_supplicant[427]: wlp3s0f3u2: CTRL-EVENT-CONNECTED - Connection to 90:55:de:__:__:__ completed [id=0 id_str=] Feb 23 14:47:04 ideapad2 wpa_supplicant[427]: wlp3s0f3u2: CTRL-EVENT-BEACON-LOSS Feb 23 14:47:05 ideapad2 wpa_supplicant[427]: wlp3s0f3u2: CTRL-EVENT-DISCONNECTED bssid=90:55:de:__:__:__ reason=4 locally_generated=1 This is caused by a mistake in the chip initialisation. This version of the chip requires loading an extra AGC table right after the main one, but the extra table is being loaded at the wrong time, in rtw_chip_board_info_setup(). Move the extra AGC table loading to the right place, in rtw_phy_load_tables(). The rtw_chip_board_info_setup() can only do "software" things, and rtw_phy_load_tables() can really do IO. Fixes: 5d6651fe ("rtw88: 8821c: support RFE type2 wifi NIC") Signed-off-by:
-
Bitterblue Smith authored
Clear bit 8 of REG_SYS_STATUS1 after MAC power on. Without this, some RTL8821CU and RTL8811CU cannot connect to any network: Feb 19 13:33:11 ideapad2 kernel: wlp3s0f3u2: send auth to 90:55:de:__:__:__ (try 1/3) Feb 19 13:33:13 ideapad2 kernel: wlp3s0f3u2: send auth to 90:55:de:__:__:__ (try 2/3) Feb 19 13:33:14 ideapad2 kernel: wlp3s0f3u2: send auth to 90:55:de:__:__:__ (try 3/3) Feb 19 13:33:15 ideapad2 kernel: wlp3s0f3u2: authentication with 90:55:de:__:__:__ timed out The RTL8822CU and RTL8822BU out-of-tree drivers do this as well, so do it for all three types of chips. Tested with RTL8811CU (Tenda U9 V2.0). Signed-off-by:
-
Bitterblue Smith authored
RTL8822CU, RTL8822BU, and RTL8821CU need an extra register write after reading and writing certain addresses. Without this, the firmware upload fails approximately more than 50% of the time. Tested with RTL8811CU (Tenda U9 V2.0) which is the same as RTL8821CU but without Bluetooth. Fixes: a82dfd33 ("wifi: rtw88: Add common USB chip support") Signed-off-by:
-
Nick Morrow authored
Add VID/PIDs that are known to be missing for this driver. Removed /* 8811CU */ and /* 8821CU */ as they are redundant since the file is specific to those chips. Removed /* TOTOLINK A650UA v3 */ as the manufacturer. It has a REALTEK VID so it may not be specific to this adapter. Verified and tested. Cc: stable@vger.kernel.org Signed-off-by:
Nick Morrow <morrownr@gmail.com> Signed-off-by:
Larry Finger <Larry.Finger@lwfinger.net> Acked-by:
-
- 04 Mar, 2024 11 commits
-
-
Johannes Berg authored
If a link is deactivated, we really cannot sustain any TDLS connections on that link any more. With the API now changed, fix this issue and remove TDLS connections. Reviewed-by:
Miriam Rachel Korenblit <miriam.rachel.korenblit@intel.com> Reviewed-by:
Emmanuel Grumbach <emmanuel.grumbach@intel.com> Link: https://msgid.link/20240228095719.a7dd812c37bf.I3474dbde79e9e7a539d47f6f81f32e6c3e459080@changeidSigned-off-by:
-
Johannes Berg authored
If a link does CSA, or if it changes SMPS mode, we need to drop the TDLS peers, but we really should drop them only on the affected link. Fix that. Reviewed-by:
-
Johannes Berg authored
Unify all the CSA handling, including handling of a beacon after the CSA, into ieee80211_sta_process_chanswitch(). The CRC of the beacon will change due to changes in the CSA/ECSA elements, so there's really no need to have the 'beacon after CSA' handling before the CRC processing or to change the beacon_crc_valid value here. Reviewed-by:
-
Johannes Berg authored
When doing CSA in multi-link, there really isn't a need to stop transmissions entirely. Add a feature flag for drivers to indicate they can handle quiet in CSA (be it by parsing themselves, or by implementing drv_pre_channel_switch()), to make that possible. Also clean up the csa_block_tx handling: it clearly cannot handle multi-link due to the way queues are stopped, move it to the sdata. Drivers should be doing it themselves for working properly during CSA in MLO anyway. Also rename it to indicate that it reflects TX was blocked at mac80211. Reviewed-by:
-
Johannes Berg authored
Pass the link conf to the abort_channel_switch driver method so the driver can handle things correctly. Reviewed-by:
-
Johannes Berg authored
For CSA to work correctly in multi-link scenarios, pass the link_id to the relevant callbacks. While at it, unify/deduplicate the tracing for them. Reviewed-by:
-
Johannes Berg authored
If we just want to determine the length of the fragmented data, we basically need the same logic, and really we want it to be _literally_ the same logic, so it cannot be out of sync in any way. Allow calling cfg80211_defragment_element() without an output buffer, where it then just returns the required output size. Also add this to the tests, just to exercise it, using the pre-calculated length to really do the defragmentation, which checks that this is sufficient. Reviewed-by:
Benjamin Berg <benjamin.berg@intel.com> Link: https://msgid.link/20240228095718.6d6565b9e3f2.Ib441903f4b8644ba04b1c766f90580ee6f54fc66@changeidSigned-off-by:
-
Johannes Berg authored
In mac80211 we'll need to look at reduced neighbor report entries for channel switch purposes, so export the iteration function to make that simpler. Reviewed-by:
-
Johannes Berg authored
Rework the data structures to hide element parsing internals from the users. Reviewed-by:
-
Johannes Berg authored
We're always using "scratch + len - pos", so we don't need to subtract here to calculate the remaining length. Remove the unnecessary subtraction. Reviewed-by:
Ilan Peer <ilan.peer@intel.com> Link: https://msgid.link/20240228094902.44e07cfa9e63.I7a9758fb9bc6b726aac49804f2f05cd521bc4128@changeidSigned-off-by:
-
Johannes Berg authored
Using the scratch buffer (without advancing it) here in the mlme.c code seems somewhat wrong, defragment the reconfig multi-link element already when parsing. This might be a bit more work in certain cases, but makes the whole thing more regular. Reviewed-by:
-
