diff mbox series

[1/5] mt76x02: correct set bssid for STA

Message ID 1541758814-2046-2-git-send-email-sgruszka@redhat.com (mailing list archive)
State Changes Requested
Delegated to: Kalle Valo
Headers show
Series mt76x02: bssid setup corrections for STA | expand

Commit Message

Stanislaw Gruszka Nov. 9, 2018, 10:20 a.m. UTC
Use MT_MAC_APC_BSSID0_H_EN bit to indicatate we are
setting BSSID for STA.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
---
 drivers/net/wireless/mediatek/mt76/mt76x02_mac.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)

Comments

Lorenzo Bianconi Nov. 9, 2018, 11:09 a.m. UTC | #1
> Use MT_MAC_APC_BSSID0_H_EN bit to indicatate we are
> setting BSSID for STA.
> 
> Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
> ---
>  drivers/net/wireless/mediatek/mt76/mt76x02_mac.c | 12 +++++++++---
>  1 file changed, 9 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
> index 59b336e34cb5..cfeae5586897 100644
> --- a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
> +++ b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
> @@ -826,10 +826,16 @@ void mt76x02_mac_work(struct work_struct *work)
>  
>  void mt76x02_mac_set_bssid(struct mt76x02_dev *dev, u8 idx, const u8 *addr)
>  {
> +	u32 bssid_h = FIELD_PREP(MT_MAC_APC_BSSID_H_ADDR,
> +				 get_unaligned_le16(addr + 4));
> +	u32 bssid_l = get_unaligned_le32(addr);
> +
> +	if (idx > 7)
> +		bssid_h |= MT_MAC_APC_BSSID0_H_EN;

This bit is to enable APClient mode and it is valid just for register 0x1094
(MT_MAC_APC_BSSID_{L,h}(0)). Moreover IIRC APClient has been disabled because
it causes a performance degradation.

> +
>  	idx &= 7;
> -	mt76_wr(dev, MT_MAC_APC_BSSID_L(idx), get_unaligned_le32(addr));
> -	mt76_rmw_field(dev, MT_MAC_APC_BSSID_H(idx), MT_MAC_APC_BSSID_H_ADDR,
> -		       get_unaligned_le16(addr + 4));
> +	mt76_wr(dev, MT_MAC_APC_BSSID_L(idx), bssid_l);
> +	mt76_wr(dev, MT_MAC_APC_BSSID_H(idx), bssid_h);
>  }
>  
>  static int
> -- 
> 1.9.3
>
Stanislaw Gruszka Nov. 9, 2018, 11:49 a.m. UTC | #2
On Fri, Nov 09, 2018 at 12:09:50PM +0100, Lorenzo Bianconi wrote:
> > Use MT_MAC_APC_BSSID0_H_EN bit to indicatate we are
> > setting BSSID for STA.
> > 
> > Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
> > ---
> >  drivers/net/wireless/mediatek/mt76/mt76x02_mac.c | 12 +++++++++---
> >  1 file changed, 9 insertions(+), 3 deletions(-)
> > 
> > diff --git a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
> > index 59b336e34cb5..cfeae5586897 100644
> > --- a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
> > +++ b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
> > @@ -826,10 +826,16 @@ void mt76x02_mac_work(struct work_struct *work)
> >  
> >  void mt76x02_mac_set_bssid(struct mt76x02_dev *dev, u8 idx, const u8 *addr)
> >  {
> > +	u32 bssid_h = FIELD_PREP(MT_MAC_APC_BSSID_H_ADDR,
> > +				 get_unaligned_le16(addr + 4));
> > +	u32 bssid_l = get_unaligned_le32(addr);
> > +
> > +	if (idx > 7)
> > +		bssid_h |= MT_MAC_APC_BSSID0_H_EN;
> 
> This bit is to enable APClient mode and it is valid just for register 0x1094
> (MT_MAC_APC_BSSID_{L,h}(0)). Moreover IIRC APClient has been disabled because
> it causes a performance degradation.

IIRC MT_MAC_APC_BSSID0_H_EN bit enable BSSID8-15 for STA usage, how we then
can program to device those BSSIDs ?

And if this bit is disabled, shouldn't we use MAC_BSSID_DW for setting BSSID
in STA mode ?

Thanks
Stanislaw
Felix Fietkau Nov. 9, 2018, 11:52 a.m. UTC | #3
On 2018-11-09 12:09, Lorenzo Bianconi wrote:
>> Use MT_MAC_APC_BSSID0_H_EN bit to indicatate we are
>> setting BSSID for STA.
>> 
>> Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
>> ---
>>  drivers/net/wireless/mediatek/mt76/mt76x02_mac.c | 12 +++++++++---
>>  1 file changed, 9 insertions(+), 3 deletions(-)
>> 
>> diff --git a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
>> index 59b336e34cb5..cfeae5586897 100644
>> --- a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
>> +++ b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
>> @@ -826,10 +826,16 @@ void mt76x02_mac_work(struct work_struct *work)
>>  
>>  void mt76x02_mac_set_bssid(struct mt76x02_dev *dev, u8 idx, const u8 *addr)
>>  {
>> +	u32 bssid_h = FIELD_PREP(MT_MAC_APC_BSSID_H_ADDR,
>> +				 get_unaligned_le16(addr + 4));
>> +	u32 bssid_l = get_unaligned_le32(addr);
>> +
>> +	if (idx > 7)
>> +		bssid_h |= MT_MAC_APC_BSSID0_H_EN;
> 
> This bit is to enable APClient mode and it is valid just for register 0x1094
> (MT_MAC_APC_BSSID_{L,h}(0)). Moreover IIRC APClient has been disabled because
> it causes a performance degradation.
The performance degradation was with MAC_ADDR_EXT, not AP-Client.
This patch should be fine.

- Felix
Lorenzo Bianconi Nov. 9, 2018, 12:50 p.m. UTC | #4
> On 2018-11-09 12:09, Lorenzo Bianconi wrote:
> >> Use MT_MAC_APC_BSSID0_H_EN bit to indicatate we are
> >> setting BSSID for STA.
> >> 
> >> Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
> >> ---
> >>  drivers/net/wireless/mediatek/mt76/mt76x02_mac.c | 12 +++++++++---
> >>  1 file changed, 9 insertions(+), 3 deletions(-)
> >> 
> >> diff --git a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
> >> index 59b336e34cb5..cfeae5586897 100644
> >> --- a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
> >> +++ b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
> >> @@ -826,10 +826,16 @@ void mt76x02_mac_work(struct work_struct *work)
> >>  
> >>  void mt76x02_mac_set_bssid(struct mt76x02_dev *dev, u8 idx, const u8 *addr)
> >>  {
> >> +	u32 bssid_h = FIELD_PREP(MT_MAC_APC_BSSID_H_ADDR,
> >> +				 get_unaligned_le16(addr + 4));
> >> +	u32 bssid_l = get_unaligned_le32(addr);
> >> +
> >> +	if (idx > 7)
> >> +		bssid_h |= MT_MAC_APC_BSSID0_H_EN;
> > 
> > This bit is to enable APClient mode and it is valid just for register 0x1094
> > (MT_MAC_APC_BSSID_{L,h}(0)). Moreover IIRC APClient has been disabled because
> > it causes a performance degradation.
> The performance degradation was with MAC_ADDR_EXT, not AP-Client.
> This patch should be fine.

Ah, now I remember sorry, you are right :)
Anyway IIRC MT_MAC_APC_BSSID0_H_EN is just for AP_CLIENT_BSSID0 (at least for
76x2), for others we need to use BIT(31) but just if APClient is disabled, right?
Moreover BSSID filter is currently disabled (BIT(3) in RX_FILTER_CFG).

# echo 0x1400 > /sys/kernel/debug/ieee80211/phy0/mt76/regidx
# cat /sys/kernel/debug/ieee80211/phy0/mt76/regval
0x00017f97

Have you tried to enabled it?

Regards,
Lorenzo

> 
> - Felix
Stanislaw Gruszka Nov. 9, 2018, 1:26 p.m. UTC | #5
On Fri, Nov 09, 2018 at 01:50:51PM +0100, Lorenzo Bianconi wrote:
> > On 2018-11-09 12:09, Lorenzo Bianconi wrote:
> > >> Use MT_MAC_APC_BSSID0_H_EN bit to indicatate we are
> > >> setting BSSID for STA.
> > >> 
> > >> Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
> > >> ---
> > >>  drivers/net/wireless/mediatek/mt76/mt76x02_mac.c | 12 +++++++++---
> > >>  1 file changed, 9 insertions(+), 3 deletions(-)
> > >> 
> > >> diff --git a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
> > >> index 59b336e34cb5..cfeae5586897 100644
> > >> --- a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
> > >> +++ b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
> > >> @@ -826,10 +826,16 @@ void mt76x02_mac_work(struct work_struct *work)
> > >>  
> > >>  void mt76x02_mac_set_bssid(struct mt76x02_dev *dev, u8 idx, const u8 *addr)
> > >>  {
> > >> +	u32 bssid_h = FIELD_PREP(MT_MAC_APC_BSSID_H_ADDR,
> > >> +				 get_unaligned_le16(addr + 4));
> > >> +	u32 bssid_l = get_unaligned_le32(addr);
> > >> +
> > >> +	if (idx > 7)
> > >> +		bssid_h |= MT_MAC_APC_BSSID0_H_EN;
> > > 
> > > This bit is to enable APClient mode and it is valid just for register 0x1094
> > > (MT_MAC_APC_BSSID_{L,h}(0)). Moreover IIRC APClient has been disabled because
> > > it causes a performance degradation.
> > The performance degradation was with MAC_ADDR_EXT, not AP-Client.
> > This patch should be fine.
> 
> Ah, now I remember sorry, you are right :)
> Anyway IIRC MT_MAC_APC_BSSID0_H_EN is just for AP_CLIENT_BSSID0 (at least for
> 76x2), for others we need to use BIT(31) but just if APClient is disabled, right?
> Moreover BSSID filter is currently disabled (BIT(3) in RX_FILTER_CFG).
> 
> # echo 0x1400 > /sys/kernel/debug/ieee80211/phy0/mt76/regidx
> # cat /sys/kernel/debug/ieee80211/phy0/mt76/regval
> 0x00017f97
> 
> Have you tried to enabled it?

I didn't try this yet, but I think it should be enabled to support
mulit BSS and MAC address change. 

However I still don't know how to set multiple bssid for STA,
Use BIT(31) or BIT(16) as indicator or maybe 
just write BSSID to MT_MAC_APC_BSSID_{H,L} and hardware will
recognize this is not AP BSSID ?

Thanks
Stanislaw
Lorenzo Bianconi Nov. 9, 2018, 1:42 p.m. UTC | #6
> On Fri, Nov 09, 2018 at 01:50:51PM +0100, Lorenzo Bianconi wrote:
> > > On 2018-11-09 12:09, Lorenzo Bianconi wrote:
> > > >> Use MT_MAC_APC_BSSID0_H_EN bit to indicatate we are
> > > >> setting BSSID for STA.
> > > >> 
> > > >> Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
> > > >> ---
> > > >>  drivers/net/wireless/mediatek/mt76/mt76x02_mac.c | 12 +++++++++---
> > > >>  1 file changed, 9 insertions(+), 3 deletions(-)
> > > >> 
> > > >> diff --git a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
> > > >> index 59b336e34cb5..cfeae5586897 100644
> > > >> --- a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
> > > >> +++ b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
> > > >> @@ -826,10 +826,16 @@ void mt76x02_mac_work(struct work_struct *work)
> > > >>  
> > > >>  void mt76x02_mac_set_bssid(struct mt76x02_dev *dev, u8 idx, const u8 *addr)
> > > >>  {
> > > >> +	u32 bssid_h = FIELD_PREP(MT_MAC_APC_BSSID_H_ADDR,
> > > >> +				 get_unaligned_le16(addr + 4));
> > > >> +	u32 bssid_l = get_unaligned_le32(addr);
> > > >> +
> > > >> +	if (idx > 7)
> > > >> +		bssid_h |= MT_MAC_APC_BSSID0_H_EN;
> > > > 
> > > > This bit is to enable APClient mode and it is valid just for register 0x1094
> > > > (MT_MAC_APC_BSSID_{L,h}(0)). Moreover IIRC APClient has been disabled because
> > > > it causes a performance degradation.
> > > The performance degradation was with MAC_ADDR_EXT, not AP-Client.
> > > This patch should be fine.
> > 
> > Ah, now I remember sorry, you are right :)
> > Anyway IIRC MT_MAC_APC_BSSID0_H_EN is just for AP_CLIENT_BSSID0 (at least for
> > 76x2), for others we need to use BIT(31) but just if APClient is disabled, right?
> > Moreover BSSID filter is currently disabled (BIT(3) in RX_FILTER_CFG).
> > 
> > # echo 0x1400 > /sys/kernel/debug/ieee80211/phy0/mt76/regidx
> > # cat /sys/kernel/debug/ieee80211/phy0/mt76/regval
> > 0x00017f97
> > 
> > Have you tried to enabled it?
> 
> I didn't try this yet, but I think it should be enabled to support
> mulit BSS and MAC address change. 

I agree we can use multi-BSS for mac filtering in AP mode but my gut feeling in
STA the hw uses the value in MT_MAC_ADDR_DW{0,1} for unicast packet filtering
(BIT(2) in RX_FILTER_CFG). Moreover I think there is a relation between
MT_MAC_ADDR_DW and BSSID values in multiAP mode.

Regards,
Lorenzo

> 
> However I still don't know how to set multiple bssid for STA,
> Use BIT(31) or BIT(16) as indicator or maybe 
> just write BSSID to MT_MAC_APC_BSSID_{H,L} and hardware will
> recognize this is not AP BSSID ?
> 
> Thanks
> Stanislaw
Felix Fietkau Nov. 9, 2018, 1:47 p.m. UTC | #7
On 2018-11-09 14:42, Lorenzo Bianconi wrote:
>> On Fri, Nov 09, 2018 at 01:50:51PM +0100, Lorenzo Bianconi wrote:
>> > > On 2018-11-09 12:09, Lorenzo Bianconi wrote:
>> > > >> Use MT_MAC_APC_BSSID0_H_EN bit to indicatate we are
>> > > >> setting BSSID for STA.
>> > > >> 
>> > > >> Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
>> > > >> ---
>> > > >>  drivers/net/wireless/mediatek/mt76/mt76x02_mac.c | 12 +++++++++---
>> > > >>  1 file changed, 9 insertions(+), 3 deletions(-)
>> > > >> 
>> > > >> diff --git a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
>> > > >> index 59b336e34cb5..cfeae5586897 100644
>> > > >> --- a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
>> > > >> +++ b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
>> > > >> @@ -826,10 +826,16 @@ void mt76x02_mac_work(struct work_struct *work)
>> > > >>  
>> > > >>  void mt76x02_mac_set_bssid(struct mt76x02_dev *dev, u8 idx, const u8 *addr)
>> > > >>  {
>> > > >> +	u32 bssid_h = FIELD_PREP(MT_MAC_APC_BSSID_H_ADDR,
>> > > >> +				 get_unaligned_le16(addr + 4));
>> > > >> +	u32 bssid_l = get_unaligned_le32(addr);
>> > > >> +
>> > > >> +	if (idx > 7)
>> > > >> +		bssid_h |= MT_MAC_APC_BSSID0_H_EN;
>> > > > 
>> > > > This bit is to enable APClient mode and it is valid just for register 0x1094
>> > > > (MT_MAC_APC_BSSID_{L,h}(0)). Moreover IIRC APClient has been disabled because
>> > > > it causes a performance degradation.
>> > > The performance degradation was with MAC_ADDR_EXT, not AP-Client.
>> > > This patch should be fine.
>> > 
>> > Ah, now I remember sorry, you are right :)
>> > Anyway IIRC MT_MAC_APC_BSSID0_H_EN is just for AP_CLIENT_BSSID0 (at least for
>> > 76x2), for others we need to use BIT(31) but just if APClient is disabled, right?
>> > Moreover BSSID filter is currently disabled (BIT(3) in RX_FILTER_CFG).
>> > 
>> > # echo 0x1400 > /sys/kernel/debug/ieee80211/phy0/mt76/regidx
>> > # cat /sys/kernel/debug/ieee80211/phy0/mt76/regval
>> > 0x00017f97
>> > 
>> > Have you tried to enabled it?
>> 
>> I didn't try this yet, but I think it should be enabled to support
>> mulit BSS and MAC address change. 
> 
> I agree we can use multi-BSS for mac filtering in AP mode but my gut feeling in
> STA the hw uses the value in MT_MAC_ADDR_DW{0,1} for unicast packet filtering
> (BIT(2) in RX_FILTER_CFG). Moreover I think there is a relation between
> MT_MAC_ADDR_DW and BSSID values in multiAP mode.
I'm pretty sure that value also gets used for local ACK/BA filtering. We
definitely need to set it up properly.
In the past, I observed cases where a misconfigured MAC address led to
no tx status being reported anymore.

- Felix
diff mbox series

Patch

diff --git a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
index 59b336e34cb5..cfeae5586897 100644
--- a/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
+++ b/drivers/net/wireless/mediatek/mt76/mt76x02_mac.c
@@ -826,10 +826,16 @@  void mt76x02_mac_work(struct work_struct *work)
 
 void mt76x02_mac_set_bssid(struct mt76x02_dev *dev, u8 idx, const u8 *addr)
 {
+	u32 bssid_h = FIELD_PREP(MT_MAC_APC_BSSID_H_ADDR,
+				 get_unaligned_le16(addr + 4));
+	u32 bssid_l = get_unaligned_le32(addr);
+
+	if (idx > 7)
+		bssid_h |= MT_MAC_APC_BSSID0_H_EN;
+
 	idx &= 7;
-	mt76_wr(dev, MT_MAC_APC_BSSID_L(idx), get_unaligned_le32(addr));
-	mt76_rmw_field(dev, MT_MAC_APC_BSSID_H(idx), MT_MAC_APC_BSSID_H_ADDR,
-		       get_unaligned_le16(addr + 4));
+	mt76_wr(dev, MT_MAC_APC_BSSID_L(idx), bssid_l);
+	mt76_wr(dev, MT_MAC_APC_BSSID_H(idx), bssid_h);
 }
 
 static int