- Oct 01, 2023
-
-
Felix Fietkau authored
It turns out that the code in mt7603_rx_pse_busy() does not detect actual hardware hangs, it only checks for busy conditions in PSE. A reset should only be performed if these conditions are true and if there is no rx activity as well. Reset the counter whenever a rx interrupt occurs. In order to also deal with a fully loaded CPU that leaves interrupts disabled with continuous NAPI polling, also check for pending rx interrupts in the function itself. Fixes: c8846e10 ("mt76: add driver for MT7603E and MT7628/7688") Signed-off-by:
Felix Fietkau <nbd@nbd.name>
-
- Sep 30, 2023
-
-
Johannes Berg authored
Looks like I lost that by accident, add it back. Fixes: 076fc877 ("wifi: cfg80211: remove wdev mutex") Signed-off-by:
Johannes Berg <johannes.berg@intel.com>
-
Johannes Berg authored
Somehow, I managed to botch this and pretty much completely break wifi. My original patch did contain these changes, but I seem to have lost them before sending to the list. Fix it now. Reported-and-tested-by:
Kalle Valo <kvalo@kernel.org> Fixes: 6c02fab7 ("wifi: mac80211: split ieee80211_drop_unencrypted_mgmt() return value") Signed-off-by:
-
- Sep 29, 2023
-
-
git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.gitKalle Valo authored
ath.git patches for v6.7. Major changes: ath12k * WCN7850: enable 320 MHz channels in 6 GHz band * WCN7850: hardware rfkill support * WCN7850: enable IEEE80211_HW_SINGLE_SCAN_ON_ALL_BANDS to make scan faster ath11k * add chip id board name while searching board-2.bin
-
Dmitry Antipov authored
Since 'u8' comparison against zero is always false, drop the corresponding branches of AMPDU_MIN_SPACE adjustment within 'rtlxxxx_set_hw_reg()' for rtl8188ee, rtl8192ce, rtl8192de, rtl8723ae, rtl8723be, and rtl8821ae. Compile tested only. Found by Linux Verification Center (linuxtesting.org) with SVACE. Signed-off-by:
Dmitry Antipov <dmantipov@yandex.ru> Acked-by:
Ping-Ke Shih <pkshih@realtek.com> Signed-off-by:
-
Zong-Zhe Yang authored
MCC (multi-channel concurrency) supports two combinations as below. * P2P-GO + STA * P2P-GC + STA We add the corresponding ieee80211_iface_limit for it into ieee80211_iface_combination. Besides, for multiple channels, it must run with mac80211 chanctx. So, only with it, ieee80211_iface_combination can allow MCC case. Signed-off-by:
Zong-Zhe Yang <kevin_yang@realtek.com> Signed-off-by:
-
Zong-Zhe Yang authored
We are going to allow RTL8852C to support MCC (multi-channel concurrency). So, we increase 8852c::support_chanctx_num up to 2. Signed-off-by:
-
Zong-Zhe Yang authored
During (TDMA-based) MCC (multi-channel concurrency), the below two cases might not have a good behavior on channel usage. * ROC (remain on channel) * HW scan So, we tend to separate them from MCC. The two cases would expect to operate the channel to which they want. However, during MCC, channels are scheduled by FW MCC state mechanism. So, channels cannot be controlled explicitly. To avoid the two cases from operating wrong channels with chance, we pause MCC (essentially stop FW MCC) once the two cases are coming. And then, we proceed MCC again (essentially restart FW MCC) once the two cases finish. Signed-off-by:
-
Zong-Zhe Yang authored
Under TDMA-based MCC (multi-channel concurrency), there are two roles, reference and auxiliary. We arrange MCC timeline based on time domain of reference role. Then, we calculate NoA start time according to MCC timeline. Besides, when MCC runs GO+STA mode, we plan an offset between GO time domain and STA time domain to make their TBTTs have a time gap. However, if GO is auxiliary role instead of reference role, NoA start time is described by STA time domain instead of GO time domain. To fix this, we apply the offset mentioned above to NoA start time to convert time domain from STA to GO. Signed-off-by:
-
Colin Ian King authored
The check on vga_gain[ch_idx] being less than zero is always false since vga_gain is a u8 array. Clean up the code by removing the check and the following assignment. Cleans up clang scan build warning: drivers/net/wireless/ralink/rt2x00/rt2800lib.c:9704:26: warning: result of comparison of unsigned expression < 0 is always false [-Wtautological-unsigned-zero-compare] Signed-off-by:
Colin Ian King <colin.i.king@gmail.com> Signed-off-by:
-
- Sep 28, 2023
-
-
Kalle Valo authored
Commit e8c18412 ("wifi: cfg80211: annotate iftype_data pointer with sparse") added sparse checks for struct ieee80211_sband_iftype_data handling which immediately found an issue in ath11k: drivers/net/wireless/ath/ath11k/mac.c:7952:22: warning: incorrect type in argument 1 (different address spaces) drivers/net/wireless/ath/ath11k/mac.c:7952:22: expected struct ieee80211_sta_he_cap const *he_cap drivers/net/wireless/ath/ath11k/mac.c:7952:22: got struct ieee80211_sta_he_cap const [noderef] __iftype_data * The problem here is that we are accessing sband->iftype_data directly even though we should use for_each_sband_iftype_data() or similar. Fortunately there's ieee80211_get_he_iftype_cap_vif() which is just what we need here so use it to get HE capabilities. Tested-on: WCN6855 hw2.0 PCI WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.23 Reported-by:
Johannes Berg <johannes@sipsolutions.net> Signed-off-by:
Kalle Valo <quic_kvalo@quicinc.com> Link: https://lore.kernel.org/r/20230927142708.2897504-2-kvalo@kernel.org
-
Dmitry Antipov authored
When compiling with clang 16.0.6, I've noticed the following: drivers/net/wireless/ath/ath11k/mac.c:8903:12: warning: stack frame size (1032) exceeds limit (1024) in 'ath11k_mac_op_remain_on_channel' [-Wframe-larger-than] static int ath11k_mac_op_remain_on_channel(struct ieee80211_hw *hw, ^ 68/1032 (6.59%) spills, 964/1032 (93.41%) variables So switch to kzalloc()'ed instance of 'struct scan_req_params' like it's done in 'ath11k_mac_op_hw_scan()'. Compile tested only. Signed-off-by:Jeff Johnson <quic_jjohnson@quicinc.com> Signed-off-by:
-
Kang Yang authored
WCN7850 and QCN9274 currently use the same structure rx_msdu_end_qcn9274 for msdu_end. But content of msdu_end on WCN7850 is different from that of QCN9274. Need to update it for WCN7850, otherwise will get the wrong values when using it. For example, TID is no longer in WCN7850's msdu_end. But ath12k_dp_rx_process_err() and ath12k_dp_rx_process_wbm_err() still get TID from msdu_end. So an uncertain value will be used in these two functions on WCN7850. Therefore, add new structure rx_msdu_end_wcn7850 for WCN7850. Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0-03427-QCAHMTSWPL_V1.0_V2.0_SILICONZ-1.15378.4 Signed-off-by:
Kang Yang <quic_kangyang@quicinc.com> Acked-by:
-
Sidhanta Sahu authored
Currently, firmware does not like when beacon mode is set as staggered mode for more than one beaconing vifs. Beacon mode for multiple beaconing (transmitted) vifs are expected to be in burst mode. So set beacon mode to burst mode by default. Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1 Signed-off-by:
Sidhanta Sahu <quic_sidhanta@quicinc.com> Signed-off-by:
-
Aloka Dixit authored
Mac80211 does not set flags BSS_CHANGED_FILS_DISCOVERY and BSS_CHANGED_UNSOL_BCAST_PROBE_RESP if there are no updates to FILS discovery and unsolicited broadcast probe response transmission configurations respectively. For BSS change operations such as channel switch, this results in the transmissions getting stopped because the driver does not send WMI command to firmware if the flags are not set. Remove the checks for the flags to always send the existing configuration to firmware. Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1 Signed-off-by:
Aloka Dixit <quic_alokad@quicinc.com> Acked-by:
-
Wu Yunchuan authored
No need cast (void *) to (struct ath11k_base *), struct hal_rx_msdu_link *), (struct ath11k_buffer_addr *) or other types. Signed-off-by:
Wu Yunchuan <yunchuan@nfschina.com> Acked-by:
-
Wen Gong authored
Currently mac80211 will send 3 scan request for each scan of WCN7850, they are 2.4 GHz/5 GHz/6 GHz band scan. Firmware of WCN7850 will cache the RNR IE(Reduced Neighbor Report element) which exist in the beacon of 2.4 GHz/5 GHz of the AP which is co-located with 6 GHz, and then use the cache to scan in 6 GHz band scan if the 6 GHz scan is in the same scan with the 2.4 GHz/5 GHz band, this will helpful to search more AP of 6 GHz. Also it will decrease the time cost of scan because firmware will use dual-band scan for the 2.4 GHz/5 GHz, it means the 2.4 GHz and 5 GHz scans are doing simultaneously. Set the flag IEEE80211_HW_SINGLE_SCAN_ON_ALL_BANDS for WCN7850 since it supports 2.4 GHz/5 GHz/6 GHz in a single wiphy/ieee80211_hw. This does not impact QCN9274, because it is not single_phy, so does not have 2.4 GHz/5 GHz/6 GHz in the same wiphy, then it does not match the condition and then IEEE80211_HW_SINGLE_SCAN_ON_ALL_BANDS will not set for QCN9274. Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0-03427-QCAHMTSWPL_V1.0_V2.0_SILICONZ-1.15378.4 Signed-off-by:
Wen Gong <quic_wgong@quicinc.com> Acked-by:
-
Wen Gong authored
Some firmware versions for WCN7850 report the default regdomain with alpha code "na" by default when load as a world wide regdomain, ath12k should treat it as a world wide alpha code. Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0-03427-QCAHMTSWPL_V1.0_V2.0_SILICONZ-1.15378.4 Signed-off-by:
-
Wen Gong authored
ath12k prints "Received scan event for unknown vdev" when doing the following test: 1. trigger scan 2. wait 0.2 second 3. iw reg set is issued or 11d scan complete event is sent from firmware Reason is: When iw reg set is issues or the 11d scan complete event is received, the new country code will be set to the firmware, and the new regdomain info indicated to ath12k, then the new channel list will be sent to the firmware. The firmware will cancel the current scan after receiving WMI_SCAN_CHAN_LIST_CMDID which is used for the new channel list, and the state of ath12k is ATH12K_SCAN_RUNNING, then ath12k_get_ar_on_scan_abort() returns NULL and ath12k_scan_event() returns at this point and does not indicate scan completion to mac80211. Indicate scan completion to mac80211 and get rid of the "Received scan event for unknown vdev" print for the above case. Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0-03427-QCAHMTSWPL_V1.0_V2.0_SILICONZ-1.15378.4 Signed-off-by:
-
Wen Gong authored
Scan failure can not be recovered from when running a loop of the following steps: 1. run scan: "iw wlan scan". 2. run command: echo assert > /sys/kernel/debug/ath12k/wcn7850\ hw2.0/simulate_fw_crash immediately after step 1. result: scan failed and can not recover even when wlan recovery succeeds: command failed: Device or resource busy (-16) reason: When scan arrives, WMI_START_SCAN_CMDID is sent to the firmware and function ath12k_mac_op_hw_scan() returns, then simulate_fw_crash arrives and the scan started event does not arrive, and then it starts to do recovery of wlan. __ath12k_mac_scan_finish() which is called from ath12k_core_halt() is one step of recovery, it will not call ieee80211_scan_completed() by logic currently because the scan state is ATH12K_SCAN_STARTING. Thus it leads the scan not being completed in mac80211, and leads all consecutive scans failing with -EBUSY in nl80211_trigger_scan even after wlan recovery success. Indicate scan complete with aborted flag to mac80211 for ATH12K_SCAN_STARTING to allow recovery from scan failed with "Device or resource busy (-16)" after wlan recovery. Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0-03427-QCAHMTSWPL_V1.0_V2.0_SILICONZ-1.15378.4 Signed-off-by:
-
Wen Gong authored
When firmware crashed while channel switch running, recovery starts in ath12k. Then ieee80211_sta_connection_lost() will be called by function ieee80211_restart_work() in mac80211. And then many WMI command timeout because firmware is crashed. Each WMI command cost 3 seconds, then the total time will be large and leads recovery fail. Hence change to set value ATH12K_FLAG_CRASH_FLUSH early and then ath12k_wmi_cmd_send() will not wait 3 seconds, then recovery will be started quickly and success. Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0-03427-QCAHMTSWPL_V1.0_V2.0_SILICONZ-1.15378.4 Signed-off-by:
-
Lingbo Kong authored
When hardware rfkill is enabled in the firmware, it will report the capability using WMI_SYS_CAP_INFO_RFKILL bit in the WMI_SERVICE_READY event to the host. Currently ath12k does not process this service capability. In order to support this, update ath12k to check if the capability is enabled, if so, send the GPIO information to firmware. When the firmware detects hardware rfkill is enabled by the user, it will report it using WMI_RFKILL_STATE_CHANGE_EVENTID. When ath12k receive the event, it will set the value of rfkill_radio_on based on whether radio_state is equal to WMI_RFKILL_RADIO_STATE_ON, then send WMI_PDEV_PARAM_RFKILL_ENABLE to firmware. Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0-03427-QCAHMTSWPL_V1.0_V2.0_SILICONZ-1.15378.4 Signed-off-by:
Lingbo Kong <quic_lingbok@quicinc.com> Acked-by:
-
Dmitry Antipov authored
Use 'kstrtoul_from_user()' in 'ath11k_write_file_spectral_count()' and 'ath11k_write_file_spectral_bins()' Signed-off-by:
-
- Sep 26, 2023
-
-
Johannes Berg authored
Make __ieee80211_data_to_8023() return more individual drop reasons instead of just doing RX_DROP_U_INVALID_8023. Signed-off-by: -
Johannes Berg authored
This has many different reasons, split the return value into the individual reasons for better traceability. Also, since symbolic tracing doesn't work for these, add a few comments for the numbering. Signed-off-by: -
Johannes Berg authored
Convert all instances of RX_DROP_UNUSABLE to indicate a better reason, and then remove RX_DROP_UNUSABLE. Signed-off-by: -
Johannes Berg authored
If we just check "result & RX_DROP_UNUSABLE", this really only works by accident, because SKB_DROP_REASON_SUBSYS_MAC80211_UNUSABLE got to have the value 1, and SKB_DROP_REASON_SUBSYS_MAC80211_MONITOR is 2. Fix this to really check the entire subsys mask for the value, so it doesn't matter what the subsystem value is. Fixes: 7f4e0970 ("wifi: mac80211: report all unusable beacon frames") Signed-off-by:
-
Johannes Berg authored
Add the local_state_change request to the deauth trace for easier debugging. Signed-off-by:
-
- Sep 25, 2023
-
-
Vinayak Yadawad authored
Introduce new feature flags for OWE offload that driver can advertise to indicate kernel/application space to avoid DH IE handling. When this flag is advertised, the driver/device will take care of DH IE inclusion and processing of peer DH IE to generate PMK. Signed-off-by:
Vinayak Yadawad <vinayak.yadawad@broadcom.com> Link: https://lore.kernel.org/r/f891cce4b52c939dfc6b71bb2f73e560e8cad287.1695374530.git.vinayak.yadawad@broadcom.com Signed-off-by:
-
Po-Hao Huang authored
Define UL-bandwidth values of trigger frame according to 802.11 std. Signed-off-by:
Po-Hao Huang <phhuang@realtek.com> Signed-off-by:
-
Mukesh Sisodiya authored
Add the support for prph register RF details and map it to get the RF ID of NIC. Signed-off-by:
Mukesh Sisodiya <mukesh.sisodiya@intel.com> Signed-off-by:
Gregory Greenman <gregory.greenman@intel.com> Link: https://lore.kernel.org/r/20230921110727.ccfc4868111f.I94dd75fc82443facf571f2fe8e23c50e9053a35a@changeid Signed-off-by:
-
Johannes Berg authored
The return value of this function is assigned, but then unused. Check for errors here. Signed-off-by:
-
Johannes Berg authored
Pull up the injection rate control one layer, and let it control the antenna settings as well. Since mac80211 has already checked that enough antennas are configured, and we only have two bits, it's enough to just copy the data over. Signed-off-by:
-
Johannes Berg authored
Refactor the injection and other frame TX rate handling to always return the injection rate directly, by factoring the legay rate portion out into a new function called in the two relevant places (injection and non-injection). Signed-off-by:
-
Johannes Berg authored
The register here is device specific, so we need to gate the reading/checking to apply only on AX210 family. Signed-off-by:
-
Johannes Berg authored
If the NIC is already dead, as detected by the transport then there's no point to try to grab the NIC access and time out, we can just fail fast. This may speed up recovery. Signed-off-by:
-
Yedidya Benshimol authored
This new version of wolan_info_notif supports the handling of bigtk during d3, this patch holds parsing of the new notif version, adding new keys and updating ipn of existing keys during the resume flow. Signed-off-by:
Yedidya Benshimol <yedidya.ben.shimol@intel.com> Signed-off-by:
-
Johannes Berg authored
There's a race with runtime PM getting enabled by userspace: - we rescan the PCI bus - this creates the new PCI device including its sysfs representation - udev sees the new device, and the (OS-specific?) scripting enables runtime PM by writing to power/control; this can happen _before_ the next step - this will runtime-suspend the device which saves the config space, including the BAR0 that wasn't assigned yet - the bus rescan assigns resources to the devices and writes them to the config space of the device (but not the runtime-pm saved copy) - the driver binds and this disallows runtime PM, so the device is resumed, restoring the (incomplete!) config space - the driver cannot work due to BAR0 not being configured Fixing the actual race is hard and deep in the PCI layer, though probably should be done for upstream as well; perhaps runtime PM should only be allowed after resource assignment, or some other TBD way. Work around this in the driver for now by simply (re-)assigning BAR0 when the driver initializes, if it's unset. Signed-off-by:
-
Anjaneyulu authored
China 2022 regulations are enabled by default. Disable only when disabled in BIOS or the firmware don't support this capability. If the firmware has this capability, read BIOS configuration data in function 4 using ACPI API and send GRP_REGULATORY_LARI_CONFIG_CHANGE_CMD to the firmware. Any error while reading BIOS data results in enablement of china 2022 regulations. Signed-off-by:
Anjaneyulu <pagadala.yesu.anjaneyulu@intel.com> Signed-off-by:
-
Johannes Berg authored
During HW restart, STA link changes happen while the link-sta is already allocated (had been prior to the restart). Adjust the allocation and checks to handle that. Signed-off-by:
-
