Skip to contentSkip to conversation

wifi: rtw88: Fix USB devices not transmitting beacons

PATCH Superseded Bitterblue Smith 4 participants 16 messages Jul 28, 2024 – Aug 21, 2024 Realtek wireless driver (rtw88)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 wifi: rtw88: Fix USB devices not transmitting beacons

All USB devices supported by rtw88 have the same problem: they don't transmit beacons in AP mode. The cause appears to be clearing BIT_EN_BCNQ_DL of REG_FWHW_TXQ_CTRL before uploading the beacon reserved page, so don't clear the bit for USB devices.

Tested with RTL8811CU and RTL8723DU.

---
I'm not sure if SDIO devices have the same problem. Also, should BIT_EN_BCNQ_DL ever be cleared in this function? The relevant code is disabled in all these drivers:

https://github.com/morrownr/8821cu-20210916/blob/3eacc28b721950b51b0249508cc31e6e54988a0c/hal/rtl8821c/rtl8821c_ops.c#L1603

https://github.com/tomaspinho/rtl8821ce/blob/f119398d868b1a3395f40c1df2e08b57b2c882cd/hal/rtl8821c/rtl8821c_ops.c#L1668

https://github.com/alireza-tabatabaee/RTL8821CS/blob/2d8a102dd7399a243f11b13314c299e509cc4ef1/rtl8821cs/hal/rtl8821c/rtl8821c_ops.c#L1603

https://github.com/morrownr/88x2bu-20210702/blob/3c8f59b5b05e5090e8593da1940a8d3ccac3f878/hal/rtl8822b/rtl8822b_ops.c#L1771

https://github.com/thales-dis-dr/rtl8822CU/blob/4182c79e0c5362dcea46088dab9fed27795b5579/hal/rtl8822c/rtl8822c_ops.c#L1941

---
 drivers/net/wireless/realtek/rtw88/fw.c | 13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)
drivers/net/wireless/realtek/rtw88/fw.c +8 -5
index ab7d414d0ba6..782f3776e0a0 100644
@@ -1468,10 +1468,12 @@ int rtw_fw_write_data_rsvd_page(struct rtw_dev *rtwdev, u16 pg_addr,
 	val |= BIT_ENSWBCN >> 8;
 	rtw_write8(rtwdev, REG_CR + 1, val);
 
-	val = rtw_read8(rtwdev, REG_FWHW_TXQ_CTRL + 2);
-	bckp[1] = val;
-	val &= ~(BIT_EN_BCNQ_DL >> 16);
-	rtw_write8(rtwdev, REG_FWHW_TXQ_CTRL + 2, val);
+	if (rtw_hci_type(rtwdev) != RTW_HCI_TYPE_USB) {
+		val = rtw_read8(rtwdev, REG_FWHW_TXQ_CTRL + 2);
+		bckp[1] = val;
+		val &= ~(BIT_EN_BCNQ_DL >> 16);
+		rtw_write8(rtwdev, REG_FWHW_TXQ_CTRL + 2, val);
+	}
 
 	ret = rtw_hci_write_data_rsvd_page(rtwdev, buf, size);
 	if (ret) {
@@ -1496,7 +1498,8 @@ int rtw_fw_write_data_rsvd_page(struct rtw_dev *rtwdev, u16 pg_addr,
 	rsvd_pg_head = rtwdev->fifo.rsvd_boundary;
 	rtw_write16(rtwdev, REG_FIFOPAGE_CTRL_2,
 		    rsvd_pg_head | BIT_BCN_VALID_V1);
-	rtw_write8(rtwdev, REG_FWHW_TXQ_CTRL + 2, bckp[1]);
+	if (rtw_hci_type(rtwdev) != RTW_HCI_TYPE_USB)
+		rtw_write8(rtwdev, REG_FWHW_TXQ_CTRL + 2, bckp[1]);
 	rtw_write8(rtwdev, REG_CR + 1, bckp[0]);
 
 	return ret;
Signature · 1 line
2.45.2
#1 lore raw

PATCH 2/2 wifi: rtw88: Avoid using macid 0 in AP mode

In AP mode, the firmware stops transmitting beacons if it receives H2C_CMD_RA_INFO for macid 0.

Leave macid 0 unused in AP mode. Macid 1 is already reserved for broadcast/multicast. Assign macid 2 to the first connected client.

Tested with RTL8811CU and RTL8723DU.

---
Tagging this for inclusion only in 6.6+ because that's the oldest stable kernel with USB support, and presumably the PCI devices don't have this problem. My RTL8822CE doesn't.

I don't know if SDIO devices also have this problem.

---
 drivers/net/wireless/realtek/rtw88/main.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
drivers/net/wireless/realtek/rtw88/main.c +4 -3
index b3a089b4f707..e0b1ccc3759c 100644
@@ -316,11 +316,12 @@ static void rtw_ips_work(struct work_struct *work)
 	mutex_unlock(&rtwdev->mutex);
 }
 
-static u8 rtw_acquire_macid(struct rtw_dev *rtwdev)
+static u8 rtw_acquire_macid(struct rtw_dev *rtwdev, struct ieee80211_vif *vif)
 {
 	unsigned long mac_id;
 
-	mac_id = find_first_zero_bit(rtwdev->mac_id_map, RTW_MAX_MAC_ID_NUM);
+	mac_id = find_next_zero_bit(rtwdev->mac_id_map, RTW_MAX_MAC_ID_NUM,
+				    vif->type == NL80211_IFTYPE_AP ? 2 : 0);
 	if (mac_id < RTW_MAX_MAC_ID_NUM)
 		set_bit(mac_id, rtwdev->mac_id_map);
 
@@ -345,7 +346,7 @@ int rtw_sta_add(struct rtw_dev *rtwdev, struct ieee80211_sta *sta,
 	struct rtw_vif *rtwvif = (struct rtw_vif *)vif->drv_priv;
 	int i;
 
-	si->mac_id = rtw_acquire_macid(rtwdev);
+	si->mac_id = rtw_acquire_macid(rtwdev, vif);
 	if (si->mac_id >= RTW_MAX_MAC_ID_NUM)
 		return -ENOSPC;
 
Signature · 1 line
2.45.2
maintainer
#2 lore raw

RE: [PATCH 2/2] wifi: rtw88: Avoid using macid 0 in AP mode

Quoted 6 lines · Bitterblue Smith

In AP mode, the firmware stops transmitting beacons if it receives H2C_CMD_RA_INFO for macid 0.

Leave macid 0 unused in AP mode. Macid 1 is already reserved for broadcast/multicast. Assign macid 2 to the first connected client.

Seemingly we missed to set mac_id in TX desc for a long time.

+#define RTW_TX_DESC_W1_MACID GENMASK(7, 0)
#define RTW_TX_DESC_W1_QSEL GENMASK(12, 8)
#define RTW_TX_DESC_W1_RATE_ID GENMASK(20, 16)

The mac_id should be from rtwvif->mac_id or si->mac_id according to operating mode and role.

And I suppose mac_id assignment for AP is mac_id 0 for broadcast/multicast, and other mac_id can be used by connected stations regularly.

#3 lore raw

Re: [PATCH 2/2] wifi: rtw88: Avoid using macid 0 in AP mode

Quoted 14 lines · Ping-Ke Shih
Quoted 6 lines · Bitterblue Smith

In AP mode, the firmware stops transmitting beacons if it receives H2C_CMD_RA_INFO for macid 0.

Leave macid 0 unused in AP mode. Macid 1 is already reserved for broadcast/multicast. Assign macid 2 to the first connected client.

Seemingly we missed to set mac_id in TX desc for a long time.

+#define RTW_TX_DESC_W1_MACID GENMASK(7, 0)
#define RTW_TX_DESC_W1_QSEL GENMASK(12, 8)
#define RTW_TX_DESC_W1_RATE_ID GENMASK(20, 16)

Quoted 6 lines

The mac_id should be from rtwvif->mac_id or si->mac_id according to operating mode and role.

And I suppose mac_id assignment for AP is mac_id 0 for broadcast/multicast, and other mac_id can be used by connected stations regularly.

What about the concurrent AP + station scenario? Will the station vif use the next available macid, whatever that is? Just wondering, I don't use concurrent mode.

Also, do you mean that you will do all this? It's not clear to me.

maintainer
#4 lore raw

RE: [PATCH 2/2] wifi: rtw88: Avoid using macid 0 in AP mode

Quoted 18 lines · Bitterblue Smith
Quoted 17 lines · Ping-Ke Shih
Quoted 6 lines · Bitterblue Smith

In AP mode, the firmware stops transmitting beacons if it receives H2C_CMD_RA_INFO for macid 0.

Leave macid 0 unused in AP mode. Macid 1 is already reserved for broadcast/multicast. Assign macid 2 to the first connected client.

Seemingly we missed to set mac_id in TX desc for a long time.

+#define RTW_TX_DESC_W1_MACID GENMASK(7, 0)
#define RTW_TX_DESC_W1_QSEL GENMASK(12, 8)
#define RTW_TX_DESC_W1_RATE_ID GENMASK(20, 16)

The mac_id should be from rtwvif->mac_id or si->mac_id according to operating mode and role.

Quoted 7 lines
Quoted 3 lines

And I suppose mac_id assignment for AP is mac_id 0 for broadcast/multicast, and other mac_id can be used by connected stations regularly.

What about the concurrent AP + station scenario? Will the station vif use the next available macid, whatever that is? Just wondering, I don't use concurrent mode.

My basic idea is to assign mac_id to each vif when adding vif. For station mode, sta->mac_id will reuse vif->mac_id. For AP mode, I will dynamically allocate an sta->mac_id to a station, and vif->mac_id is to send broadcast/multicast packets that are not belong to a sta. For example,

                  vif->mac_id      sta->mac_id
vif0 (STA mode)        0               0
vif1 (AP mode)         1               2...
Quoted 2 lines

Also, do you mean that you will do all this? It's not clear to me.

I can do it. Or are you interested in this?

#5 lore raw

Re: [PATCH 2/2] wifi: rtw88: Avoid using macid 0 in AP mode

Quoted 38 lines · Ping-Ke Shih
Quoted 18 lines · Bitterblue Smith
Quoted 17 lines · Ping-Ke Shih
Quoted 6 lines · Bitterblue Smith

In AP mode, the firmware stops transmitting beacons if it receives H2C_CMD_RA_INFO for macid 0.

Leave macid 0 unused in AP mode. Macid 1 is already reserved for broadcast/multicast. Assign macid 2 to the first connected client.

Seemingly we missed to set mac_id in TX desc for a long time.

+#define RTW_TX_DESC_W1_MACID GENMASK(7, 0)
#define RTW_TX_DESC_W1_QSEL GENMASK(12, 8)
#define RTW_TX_DESC_W1_RATE_ID GENMASK(20, 16)

The mac_id should be from rtwvif->mac_id or si->mac_id according to operating mode and role.

Quoted 7 lines
Quoted 3 lines

And I suppose mac_id assignment for AP is mac_id 0 for broadcast/multicast, and other mac_id can be used by connected stations regularly.

What about the concurrent AP + station scenario? Will the station vif use the next available macid, whatever that is? Just wondering, I don't use concurrent mode.

My basic idea is to assign mac_id to each vif when adding vif. For station mode, sta->mac_id will reuse vif->mac_id. For AP mode, I will dynamically allocate an sta->mac_id to a station, and vif->mac_id is to send broadcast/multicast packets that are not belong to a sta. For example,

                  vif->mac_id      sta->mac_id
vif0 (STA mode)        0               0
vif1 (AP mode)         1               2...
Quoted 1 line
Quoted 4 lines
Quoted 1 line

Also, do you mean that you will do all this? It's not clear to me.

I can do it. Or are you interested in this?

No, no, it's all yours. :)

#6 lore raw

Re: [PATCH 2/2] wifi: rtw88: Avoid using macid 0 in AP mode

Quoted 39 lines · Bitterblue Smith
Quoted 38 lines · Ping-Ke Shih
Quoted 18 lines · Bitterblue Smith
Quoted 17 lines · Ping-Ke Shih
Quoted 6 lines · Bitterblue Smith

In AP mode, the firmware stops transmitting beacons if it receives H2C_CMD_RA_INFO for macid 0.

Leave macid 0 unused in AP mode. Macid 1 is already reserved for broadcast/multicast. Assign macid 2 to the first connected client.

Seemingly we missed to set mac_id in TX desc for a long time.

+#define RTW_TX_DESC_W1_MACID GENMASK(7, 0)
#define RTW_TX_DESC_W1_QSEL GENMASK(12, 8)
#define RTW_TX_DESC_W1_RATE_ID GENMASK(20, 16)

The mac_id should be from rtwvif->mac_id or si->mac_id according to operating mode and role.

Quoted 7 lines
Quoted 3 lines

And I suppose mac_id assignment for AP is mac_id 0 for broadcast/multicast, and other mac_id can be used by connected stations regularly.

What about the concurrent AP + station scenario? Will the station vif use the next available macid, whatever that is? Just wondering, I don't use concurrent mode.

My basic idea is to assign mac_id to each vif when adding vif. For station mode, sta->mac_id will reuse vif->mac_id. For AP mode, I will dynamically allocate an sta->mac_id to a station, and vif->mac_id is to send broadcast/multicast packets that are not belong to a sta. For example,

                   vif->mac_id      sta->mac_id
vif0 (STA mode)        0               0
vif1 (AP mode)         1               2...
Quoted 1 line
Quoted 7 lines
Quoted 4 lines
Quoted 1 line

Also, do you mean that you will do all this? It's not clear to me.

I can do it. Or are you interested in this?

No, no, it's all yours. :)

It would be nice to have either this submitted fix or alternative one merged soon since the AP mode is completely broken otherwise (at least on RTL8821CU).

Thanks,
Maciej

maintainer
#7 lore raw

RE: [PATCH 2/2] wifi: rtw88: Avoid using macid 0 in AP mode

Quoted 42 lines · Maciej S. Szmigiero
Quoted 41 lines · Bitterblue Smith
Quoted 40 lines · Ping-Ke Shih
Quoted 18 lines · Bitterblue Smith
Quoted 17 lines · Ping-Ke Shih
Quoted 6 lines · Bitterblue Smith

In AP mode, the firmware stops transmitting beacons if it receives H2C_CMD_RA_INFO for macid 0.

Leave macid 0 unused in AP mode. Macid 1 is already reserved for broadcast/multicast. Assign macid 2 to the first connected client.

Seemingly we missed to set mac_id in TX desc for a long time.

+#define RTW_TX_DESC_W1_MACID GENMASK(7, 0)
#define RTW_TX_DESC_W1_QSEL GENMASK(12, 8)
#define RTW_TX_DESC_W1_RATE_ID GENMASK(20, 16)

The mac_id should be from rtwvif->mac_id or si->mac_id according to operating mode and role.

Quoted 7 lines
Quoted 3 lines

And I suppose mac_id assignment for AP is mac_id 0 for broadcast/multicast, and other mac_id can be used by connected stations regularly.

What about the concurrent AP + station scenario? Will the station vif use the next available macid, whatever that is? Just wondering, I don't use concurrent mode.

My basic idea is to assign mac_id to each vif when adding vif. For station mode, sta->mac_id will reuse vif->mac_id. For AP mode, I will dynamically allocate an sta->mac_id to a station, and vif->mac_id is to send broadcast/multicast packets that are not belong to a sta. For example,

                   vif->mac_id      sta->mac_id
vif0 (STA mode)        0               0
vif1 (AP mode)         1               2...
Quoted 2 lines

Also, do you mean that you will do all this? It's not clear to me.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

I can do it. Or are you interested in this?

No, no, it's all yours. :)

It would be nice to have either this submitted fix or alternative one merged soon since the AP mode is completely broken otherwise (at least on RTL8821CU).

I have sent a patch [2] to replace this one, but still needs patch 1/2 "wifi: rtw88: Fix USB devices not transmitting beacons". [1]

I have tested [1] + [2] can work on PCI devices.
Can anyone help to test if [1] + [2] also works on USB devices?

[1] https://lore.kernel.org/linux-wireless/[email protected]/
[2] https://lore.kernel.org/linux-wireless/[email protected]/T/#u

#8 lore raw

Re: [PATCH 2/2] wifi: rtw88: Avoid using macid 0 in AP mode

Quoted 49 lines · Ping-Ke Shih
Quoted 48 lines · Maciej S. Szmigiero
Quoted 46 lines · Bitterblue Smith
Quoted 38 lines · Ping-Ke Shih
Quoted 18 lines · Bitterblue Smith
Quoted 17 lines · Ping-Ke Shih
Quoted 6 lines · Bitterblue Smith

In AP mode, the firmware stops transmitting beacons if it receives H2C_CMD_RA_INFO for macid 0.

Leave macid 0 unused in AP mode. Macid 1 is already reserved for broadcast/multicast. Assign macid 2 to the first connected client.

Seemingly we missed to set mac_id in TX desc for a long time.

+#define RTW_TX_DESC_W1_MACID GENMASK(7, 0)

   #define RTW_TX_DESC_W1_QSEL GENMASK(12, 8)
   #define RTW_TX_DESC_W1_RATE_ID GENMASK(20, 16)

The mac_id should be from rtwvif->mac_id or si->mac_id according to operating mode and role.

Quoted 7 lines
Quoted 3 lines

And I suppose mac_id assignment for AP is mac_id 0 for broadcast/multicast, and other mac_id can be used by connected stations regularly.

What about the concurrent AP + station scenario? Will the station vif use the next available macid, whatever that is? Just wondering, I don't use concurrent mode.

My basic idea is to assign mac_id to each vif when adding vif. For station mode, sta->mac_id will reuse vif->mac_id. For AP mode, I will dynamically allocate an sta->mac_id to a station, and vif->mac_id is to send broadcast/multicast packets that are not belong to a sta. For example,

                    vif->mac_id      sta->mac_id
vif0 (STA mode)        0               0
vif1 (AP mode)         1               2...
Quoted 1 line
Quoted 4 lines
Quoted 1 line

Also, do you mean that you will do all this? It's not clear to me.

I can do it. Or are you interested in this?

No, no, it's all yours. :)

Quoted 8 lines
Quoted 2 lines

It would be nice to have either this submitted fix or alternative one merged soon since the AP mode is completely broken otherwise (at least on RTL8821CU).

I have sent a patch [2] to replace this one, but still needs patch 1/2 "wifi: rtw88: Fix USB devices not transmitting beacons". [1]

I have tested [1] + [2] can work on PCI devices.
Can anyone help to test if [1] + [2] also works on USB devices?

I tested these patches on USB RTL8821CU and can confirm that the AP mode seems to work as good with your patch [2] as it did with the Bitterblue's one.

The only issue with either your or Bitterblue's patches is that I occasionally get the following messages in the kernel log in the AP mode:

Quoted 3 lines

kernel: [T1234] rtw_8821cu 4-1.2:1.2: error beacon valid
kernel: [T1234] rtw_8821cu 4-1.2:1.2: failed to download drv rsvd page
kernel: [T1234] rtw_8821cu 4-1.2:1.2: failed to download beacon

Around the time these messages are logged ping RTT of a connected (battery powered) STA climbs into multiple seconds range so I guess there might be something wrong with beacon DTIM update here.

Quoted 2 lines

[1] https://lore.kernel.org/linux-wireless/[email protected]/
[2] https://lore.kernel.org/linux-wireless/[email protected]/T/#u

Thanks,
Maciej

maintainer
#9 lore raw

RE: [PATCH 2/2] wifi: rtw88: Avoid using macid 0 in AP mode

Hi Bitterblue,

Quoted 5 lines · Maciej S. Szmigiero

I tested these patches on USB RTL8821CU and can confirm that the AP mode seems to work as good with your patch [2] as it did with the Bitterblue's one.

The only issue with either your or Bitterblue's patches is that I occasionally get the following messages in the kernel log in the AP mode:

Quoted 7 lines
Quoted 3 lines

kernel: [T1234] rtw_8821cu 4-1.2:1.2: error beacon valid
kernel: [T1234] rtw_8821cu 4-1.2:1.2: failed to download drv rsvd page
kernel: [T1234] rtw_8821cu 4-1.2:1.2: failed to download beacon

Around the time these messages are logged ping RTT of a connected (battery powered) STA climbs into multiple seconds range so I guess there might be something wrong with beacon DTIM update here.

Can you also see this in your side? It seems like download beacon for change of DTIM might be failed occasionally.

Ping-Ke

#10 lore raw

Re: [PATCH 2/2] wifi: rtw88: Avoid using macid 0 in AP mode

Quoted 12 lines · Ping-Ke Shih

Hi Bitterblue,

Quoted 9 lines · Maciej S. Szmigiero

I tested these patches on USB RTL8821CU and can confirm that the AP mode seems to work as good with your patch [2] as it did with the Bitterblue's one.

The only issue with either your or Bitterblue's patches is that I occasionally get the following messages in the kernel log in the AP mode:

Quoted 3 lines

kernel: [T1234] rtw_8821cu 4-1.2:1.2: error beacon valid
kernel: [T1234] rtw_8821cu 4-1.2:1.2: failed to download drv rsvd page
kernel: [T1234] rtw_8821cu 4-1.2:1.2: failed to download beacon

Quoted 9 lines
Quoted 3 lines

Around the time these messages are logged ping RTT of a connected (battery powered) STA climbs into multiple seconds range so I guess there might be something wrong with beacon DTIM update here.

Can you also see this in your side? It seems like download beacon for change of DTIM might be failed occasionally.

Ping-Ke

Yes, I have seen it with RTL8812AU. It only fails sometimes. I don't know what is wrong.

maintainer
#11 lore raw

RE: [PATCH 2/2] wifi: rtw88: Avoid using macid 0 in AP mode

In reply to Bitterblue Smith

Hi Bitterblue Smith,

Quoted 5 lines
Quoted 2 lines

Also, do you mean that you will do all this? It's not clear to me.

I can do it. Or are you interested in this?

FYI.
I have made a draft version, and will send it out afterward.

maintainer
#12 lore raw
In reply to Bitterblue Smith
Quoted 5 lines · Bitterblue Smith

All USB devices supported by rtw88 have the same problem: they don't transmit beacons in AP mode. The cause appears to be clearing BIT_EN_BCNQ_DL of REG_FWHW_TXQ_CTRL before uploading the beacon reserved page, so don't clear the bit for USB devices.

Quoted 8 lines

Tested with RTL8811CU and RTL8723DU.

---
I'm not sure if SDIO devices have the same problem. Also, should BIT_EN_BCNQ_DL ever be cleared in this function? The relevant code is disabled in all these drivers:

I checked internal the latest driver, which codes are changed again. We will check flow internally and prepare a patch for this problem.

This patch gives us good hints to address problem. Thanks.

maintainer
#13 lore raw
In reply to Bitterblue Smith
Quoted 1 line · Bitterblue Smith

I'm not sure if SDIO devices have the same problem.

A SDIO user has confirmed this change can make SDIO devices send out beacon, but still have other problems though.

By v2, I will take this patch and change condition

    if (rtw_hci_type(rtwdev) != RTW_HCI_TYPE_USB)
to
    if (rtw_hci_type(rtwdev) == RTW_HCI_TYPE_PCIE)
#14 lore raw
Quoted 3 lines

A SDIO user has confirmed this change can make SDIO devices send out beacon, but still have other problems though.

I am the other user. :-)
This is also a test message, sorry for spamming.

#15 lore raw
In reply to Ping-Ke Shih
Quoted 11 lines · Ping-Ke Shih
Quoted 1 line · Bitterblue Smith

I'm not sure if SDIO devices have the same problem.

A SDIO user has confirmed this change can make SDIO devices send out beacon, but still have other problems though.

By v2, I will take this patch and change condition

    if (rtw_hci_type(rtwdev) != RTW_HCI_TYPE_USB)
to
    if (rtw_hci_type(rtwdev) == RTW_HCI_TYPE_PCIE)

That's great! I will send v2.