Ping-Ke,
A user of my GitHub repo has reported a regression beginning with commit
26a125f550a3 ("wifi: rtw89: correct PS calculation for SUPPORTS_DYNAMIC_PS").
The user finds interference between wifi and the BT device that is part of the
package.
We were able to determine that loading rtw89_core with the "disable_ps_mode=y"
option does fix the problem. Our discussion of the problem can be seen at
https://github.com/lwfinger/rtw89/issues/262#issuecomment-1664658764.
Reading the patch in question, it seems that a vif count of 1 is not sufficient
to enable PS, and that it is also necessary to make sure that BT is not active
as well.
Thanks,
Larry
> -----Original Message-----
> From: Larry Finger <[email protected]> On Behalf Of Larry Finger
> Sent: Friday, August 4, 2023 6:14 AM
> To: Ping-Ke Shih <[email protected]>
> Cc: linux-wireless <[email protected]>
> Subject: Regression in rtw89
>
> Ping-Ke,
>
> A user of my GitHub repo has reported a regression beginning with commit
> 26a125f550a3 ("wifi: rtw89: correct PS calculation for SUPPORTS_DYNAMIC_PS").
> The user finds interference between wifi and the BT device that is part of the
> package.
>
> We were able to determine that loading rtw89_core with the "disable_ps_mode=y"
> option does fix the problem. Our discussion of the problem can be seen at
> https://github.com/lwfinger/rtw89/issues/262#issuecomment-1664658764.
>
As the commit message of the 26a125f550a3, PS was broken since v5.20, and that
patch fixes it. The correct bisection is from initial version of 8852BE + this patch.
(8852BE is introduced by kernel v6.2, and PS was broken at that time.)
As the discussion in github, it looks like a BT-coexistence problem, because
Both WiFi-only and BT-only work. Please update BT driver and firmware to have
below:
BT driver:
- Bluetooth: btrtl: Firmware format v2 support
- Bluetooth: btrtl: Add the support for RTL8851B
BT firmware: (copy all files from linux-firmware.git:/rtl_bt/)
- rtl_bt: Add firmware and config files for RTL8851B
Then, read btc_info in period of 2 seconds before and after entering PS.
A sample script to capture log
while [ true ]; do cat /sys/kernel/debug/ieee80211/phy0/rtw89/btc_info; date; sleep 2; done
And, use 'iw wlan0 set power_save on/off' to switch PS state. Also, please
point out the time when BT goes wrong.
> Reading the patch in question, it seems that a vif count of 1 is not sufficient
> to enable PS, and that it is also necessary to make sure that BT is not active
> as well.
'vif' count refers to WiFi vif that is normally one for regular use (STA mode only).
When it enters/leaves PS, rtw89_btc_ntfy_radio_state() is called to notify
BT-coexistence mechanism to reflect current WiFi PS state. By the way, things
related to BT are existing in coex.c.
Ping-Ke
On 8/3/23 19:54, Ping-Ke Shih wrote:
> As the commit message of the 26a125f550a3, PS was broken since v5.20, and that
> patch fixes it. The correct bisection is from initial version of 8852BE + this patch.
> (8852BE is introduced by kernel v6.2, and PS was broken at that time.)
>
> As the discussion in github, it looks like a BT-coexistence problem, because
> Both WiFi-only and BT-only work. Please update BT driver and firmware to have
> below:
> BT driver:
> - Bluetooth: btrtl: Firmware format v2 support
> - Bluetooth: btrtl: Add the support for RTL8851B
> BT firmware: (copy all files from linux-firmware.git:/rtl_bt/)
> - rtl_bt: Add firmware and config files for RTL8851B
>
> Then, read btc_info in period of 2 seconds before and after entering PS.
> A sample script to capture log
> while [ true ]; do cat /sys/kernel/debug/ieee80211/phy0/rtw89/btc_info; date; sleep 2; done
> And, use 'iw wlan0 set power_save on/off' to switch PS state. Also, please
> point out the time when BT goes wrong.
>
Ping-Ke,
Thanks for the quick response. The OP is using kernel 6.4.7, thus his btrtl
should be new enough. I copied the latest firmware in linux-firmware, and he
copied it into the appropriate locations. That did not help.
I then had him create a script to capture the log. He found that BT goes wrong
immediately as soon as power_save is set on. His logs were as follows:
*************************************************************************
First is when the problem started. It happened as soon as I enabled the
power_save feature on my wifi card.
=========================================
WL FW / BT FW 0.29.29.1 / NA
manual 0
=========================================
[bt_info] : raw_data[49 00 c5 00 04 7f] (type:auto/cnt:1576/same:1)
=========================================
========== [BTC COEX INFO (1)] ==========
[coex_version] : Coex:7.0.1(branch:19), WL_FW_coex:7.8.0(branch:29)(Match,
desired:5.5.0), BT_FW_coex:7(Match, desired:5)
[sub_module] : WL_FW:0.29.29.1, BT_FW:0x1fc6b20f(patch)
[hw_info] : cv:1, rfe_type:0x1, ant_iso:10, ant_pg:2, 3rd_coex:0,
dbcc:0, tx_num:2, rx_num:2
========== [WL Status] ==========
[status] : link_mode:9, rf_off:0, power_save:1,
scan:N(band:1/phy_map:0x0), connecting:N, roam:N, 4way:N, init_ok:Y
[port_0] : role=1(phy-0), connect=2(client_cnt=-1), mode=3,
center_ch=42, bw=2, mac_id=0, max_tx_time=0us, max_tx_retry=0
[port_0] : rssi=-59dBm(51), busy=0, dir=DL, tx[rate:260/busy_level:0],
rx[rate:280/busy_level:0/drop:0]
========== [BT Status] ==========
[status] : enable:Y, btg:Y, connect:Y, igno_wl:N, mailbox_avl:Y,
rfk_state:0x0
[profile] : profile:A2DP, multi-link:N, role:Master, ble-connect:N,
CQDDR:N, A2DP_active:Y, PAN_active:N
[link] : rssi:-53dBm, tx_rate:3M, acl_busy!!
afh_map[ffff_ffff_ffff_f3f1_fffe], wl_ch_map[en:0/ch:0/bw:32]
[stat_cnt] : retry:0, relink:1, rate_chg:11, reinit:0, reenable:1,
role-switch:0, afh:0, inq_page:1(inq:0/page:1), igno_wl:1
[A2DP] : type:Legacy, bit-pool:63, flush-time:0, vid:0xa,
Dev-name:0x2d4857, sut_pwr:0, golden-rx:0
[bt_info] : raw_data[49 00 c5 00 04 7f] (type:auto/cnt:1576/same:1)
[trx_req_cnt] : Hi-rx = 178, Hi-tx = 0, Lo-rx = 546, Lo-tx = 546
(bt_polut_wl_tx = 0)(BG:0-0/INIT:0-0/LE:0-0)
========== [Mechanism Status (Auto)] ==========
[status] : type:shared, reason:NTFY_RADIO_STATE(), action:WL_OFF(),
ant_path:5, run_cnt:2220
[dm_steps] : -> UPDATE_BT_INFO -> WL_5G -> OFF_EQ0 ->
NTFY_WL_STA -> WL_5G -> OFF_EQ0
[dm_steps] : -> UPDATE_BT_INFO -> WL_5G -> OFF_EQ0 ->
UPDATE_BT_INFO -> WL_5G -> OFF_EQ0
[dm_steps] : -> UPDATE_BT_INFO -> WL_5G -> OFF_EQ0 ->
UPDATE_BT_INFO -> WL_5G -> OFF_EQ0
[dm_steps] : -> UPDATE_BT_INFO -> WL_5G -> OFF_EQ0 ->
UPDATE_BT_INFO -> WL_5G -> OFF_EQ0
[dm_steps] : -> NTFY_WL_STA -> WL_5G -> OFF_EQ0 ->
NTFY_RADIO_STATE -> WL_OFF -> OFF_BT
[dm_flag] : wl_only:0, bt_only:0, igno_bt:1, free_run:0, wl_ps_ctrl:0,
wl_mimo_ps:0, leak_ap:0, fw_offload:N
[trx_ctrl] : wl_rssi_lvl:3, para_lvl:0, wl_tx_pwr:orig, wl_rx_lvl:0,
bt_tx_pwr_dec:0, bt_rx_lna:7(Ori-tbl), wl_btg_rx:0
[dm_ctrl] : wl_tx_limit[en:0/max_t:5280us/max_retry:31],
bt_slot_reg:0-TU, bt_scan_rx_low_pri:0
[tdma_policy] : type:0, rx_flow_ctrl:0, tx_pause:0, wl_toggle_n:0, leak_n:0,
ext_ctrl:0, policy_type:0
[slot_list] : OFF[100/0xe5555555/0], B2W[005/0xea5a5a5a/1],
W1[070/0xea5a5a5a/1], W2[015/0xea5a5a5a/1], W2B[015/0xea5a5a5a/1]
[slot_list] : B1[250/0xe5555555/0], B2[007/0xea5a5a5a/0],
B3[005/0xe5555555/0], B4[050/0xe5555555/0], LK[020/0xea5a5a5a/1]
[slot_list] : BLK[500/0x55555555/0], E2G[000/0xea5a5a5a/0],
E5G[000/0xffffffff/1], EBT[000/0xe5555555/0], ENULL[000/0xaaaaaaaa/1]
[slot_list] : WLK[250/0xea5a5a5a/0], W1FDD[050/0xffffffff/1],
B1FDD[050/0xffffdfff/1]
========== [HW Status] ==========
[scoreboard] : WL->BT:0x00004083(cnt:443), BT->WL:0x4001051b(total:219,
bt_update:14)
[reg] : 0_0xda24=0xffffffff, 0_0xda28=0xffffffff,
0_0xda2c=0xf0ffffff, 0_0xda30=0x00000008, 0_0xda4c=0x0000000d, 0_0xda10=0x00000100
[reg] : 0_0xda20=0x00000002, 0_0xda34=0x00000100,
0_0xcef4=0x00000000, 0_0x8424=0x00000000, 0_0xd200=0x00000000, 0_0xd220=0x00000000
[reg] : 1_0x0980=0x10182250, 1_0x4738=0x8325c500,
1_0x4688=0x1a10ff04, 1_0x4694=0x04fa2a2a
========== [Statistics] ==========
[summary] : h2c_cnt=886(fail:1), c2h_cnt=0
[notify_cnt] : total=821, show_coex_info=162, power_on=0, init_coex=1,
power_off=0, radio_state=398, role_info=2, wl_rfk=5, wl_sta=238
[notify_cnt] : scan_start=2, scan_finish=2, switch_band=2, special_pkt=5,
timer=4, control=0, customerize=0sam. 05 août 2023 10:18:38 CEST
*************************************************************************
Second is the output right before the previous one, when power_save was disabled
and no issues were present.
=========================================
WL FW / BT FW 0.29.29.1 / NA
manual 0
=========================================
[bt_info] : raw_data[49 00 c5 00 04 7f] (type:auto/cnt:1576/same:1)
=========================================
========== [BTC COEX INFO (1)] ==========
[coex_version] : Coex:7.0.1(branch:19), WL_FW_coex:7.8.0(branch:29)(Match,
desired:5.5.0), BT_FW_coex:7(Match, desired:5)
[sub_module] : WL_FW:0.29.29.1, BT_FW:0x1fc6b20f(patch)
[hw_info] : cv:1, rfe_type:0x1, ant_iso:10, ant_pg:2, 3rd_coex:0,
dbcc:0, tx_num:2, rx_num:2
========== [WL Status] ==========
[status] : link_mode:9, rf_off:0, power_save:0,
scan:N(band:1/phy_map:0x0), connecting:N, roam:N, 4way:N, init_ok:Y
[port_0] : role=1(phy-0), connect=2(client_cnt=-1), mode=3,
center_ch=42, bw=2, mac_id=0, max_tx_time=0us, max_tx_retry=0
[port_0] : rssi=-59dBm(51), busy=0, dir=DL, tx[rate:260/busy_level:0],
rx[rate:280/busy_level:0/drop:0]
========== [BT Status] ==========
[status] : enable:Y, btg:Y, connect:Y, igno_wl:N, mailbox_avl:Y,
rfk_state:0x0
[profile] : profile:A2DP, multi-link:N, role:Master, ble-connect:N,
CQDDR:N, A2DP_active:Y, PAN_active:N
[link] : rssi:-53dBm, tx_rate:3M, acl_busy!!
afh_map[ffff_ffff_ffff_f3f1_fffe], wl_ch_map[en:0/ch:0/bw:32]
[stat_cnt] : retry:0, relink:1, rate_chg:11, reinit:0, reenable:1,
role-switch:0, afh:0, inq_page:1(inq:0/page:1), igno_wl:1
[A2DP] : type:Legacy, bit-pool:63, flush-time:0, vid:0xa,
Dev-name:0x2d4857, sut_pwr:0, golden-rx:0
[bt_info] : raw_data[49 00 c5 00 04 7f] (type:auto/cnt:1576/same:1)
[trx_req_cnt] : Hi-rx = 178, Hi-tx = 0, Lo-rx = 546, Lo-tx = 546
(bt_polut_wl_tx = 0)(BG:0-0/INIT:0-0/LE:0-0)
========== [Mechanism Status (Auto)] ==========
[status] : type:shared, reason:NTFY_WL_STA(), action:WL_5G(),
ant_path:5, run_cnt:2219
[dm_steps] : -> UPDATE_BT_INFO -> WL_5G -> OFF_EQ0 ->
UPDATE_BT_INFO -> WL_5G -> OFF_EQ0
[dm_steps] : -> NTFY_WL_STA -> WL_5G -> OFF_EQ0 ->
UPDATE_BT_INFO -> WL_5G -> OFF_EQ0
[dm_steps] : -> UPDATE_BT_INFO -> WL_5G -> OFF_EQ0 ->
UPDATE_BT_INFO -> WL_5G -> OFF_EQ0
[dm_steps] : -> UPDATE_BT_INFO -> WL_5G -> OFF_EQ0 ->
UPDATE_BT_INFO -> WL_5G -> OFF_EQ0
[dm_steps] : -> UPDATE_BT_INFO -> WL_5G -> OFF_EQ0 ->
NTFY_WL_STA -> WL_5G -> OFF_EQ0
[dm_flag] : wl_only:0, bt_only:0, igno_bt:0, free_run:0, wl_ps_ctrl:0,
wl_mimo_ps:0, leak_ap:0, fw_offload:N
[trx_ctrl] : wl_rssi_lvl:3, para_lvl:0, wl_tx_pwr:orig, wl_rx_lvl:0,
bt_tx_pwr_dec:0, bt_rx_lna:7(Ori-tbl), wl_btg_rx:0
[dm_ctrl] : wl_tx_limit[en:0/max_t:5280us/max_retry:31],
bt_slot_reg:0-TU, bt_scan_rx_low_pri:0
[tdma_policy] : type:0, rx_flow_ctrl:0, tx_pause:0, wl_toggle_n:0, leak_n:0,
ext_ctrl:0, policy_type:2
[slot_list] : OFF[100/0xffffffff/1], B2W[005/0xea5a5a5a/1],
W1[070/0xea5a5a5a/1], W2[015/0xea5a5a5a/1], W2B[015/0xea5a5a5a/1]
[slot_list] : B1[250/0xe5555555/0], B2[007/0xea5a5a5a/0],
B3[005/0xe5555555/0], B4[050/0xe5555555/0], LK[020/0xea5a5a5a/1]
[slot_list] : BLK[500/0x55555555/0], E2G[000/0xea5a5a5a/0],
E5G[000/0xffffffff/1], EBT[000/0xe5555555/0], ENULL[000/0xaaaaaaaa/1]
[slot_list] : WLK[250/0xea5a5a5a/0], W1FDD[050/0xffffffff/1],
B1FDD[050/0xffffdfff/1]
========== [HW Status] ==========
[scoreboard] : WL->BT:0x00004083(cnt:443), BT->WL:0x4001051b(total:219,
bt_update:14)
[gnt_status] : pta_owner:WL, phy-0[gnt_wl:SW-1/gnt_bt:HW-1],
phy-1[gnt_wl:SW-1/gnt_bt:HW-1]
[reg] : 0_0xda24=0xffffffff, 0_0xda28=0xffffffff,
0_0xda2c=0xf0ffffff, 0_0xda30=0x00000008, 0_0xda4c=0x0000000d, 0_0xda10=0x00000100
[reg] : 0_0xda20=0x00000002, 0_0xda34=0x00000100,
0_0xcef4=0x00000000, 0_0x8424=0x00000000, 0_0xd200=0x00000000, 0_0xd220=0x00000000
[reg] : 1_0x0980=0x10182250, 1_0x4738=0x8325c500,
1_0x4688=0x1a10ff04, 1_0x4694=0x04fa2a2a
========== [Statistics] ==========
[summary] : h2c_cnt=884(fail:1, fw_recv:1083), c2h_cnt=0(fw_send:2137,
len:235), rpt_cnt=549(fw_send:548), rpt_map=0x64
[mailbox] : send_ok:192, send_fail:0, recv:48257,
A2DP_empty:46499(stop:0, tx:181882, ack:170362, nack:2061)
[RFK/LPS] : wl_rfk[req:5/go:5/reject:0/tout:0], bt_rfk[req:0],
AOAC[RF_on:2031/RF_off:2031]
[notify_cnt] : total=818, show_coex_info=160, power_on=0, init_coex=1,
power_off=0, radio_state=397, role_info=2, wl_rfk=5, wl_sta=238
[notify_cnt] : scan_start=2, scan_finish=2, switch_band=2, special_pkt=5,
timer=4, control=0, customerize=0sam. 05 août 2023 10:18:36 CEST
Thanks for your help.
Larry
> -----Original Message-----
> From: Larry Finger <[email protected]> On Behalf Of Larry Finger
> Sent: Sunday, August 6, 2023 2:24 AM
> To: Ping-Ke Shih <[email protected]>
> Cc: linux-wireless <[email protected]>
> Subject: Re: Regression in rtw89
>
> Thanks for the quick response. The OP is using kernel 6.4.7, thus his btrtl
> should be new enough. I copied the latest firmware in linux-firmware, and he
> copied it into the appropriate locations. That did not help.
I would like to reproduce this symptom in my side. Could you summarize the
WiFi driver/firmware and BT driver/firmware the OP uses now? We also need
to know the BT device? Do you have the device model name?
>
> I then had him create a script to capture the log. He found that BT goes wrong
> immediately as soon as power_save is set on. His logs were as follows:
Our coex team needs the full log in period of 2 seconds, not only two snapshot.
Though we are trying to reproduce, it may not get the symptom right now.
Could you provide the log fully?
connection --> enter PS (BT chopped audio) -> leave PS (BT audio as normal) ->
enter PS (BT chopped audio)
Thank you
Ping-Ke