mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH] wifi: brcmfmac: disable TX beamforming on BCM4364
@ 2026-11-04 15:16 Andre Eikmeyer
  2026-11-16 20:37 ` Arend van Spriel
  2026-11-18 14:10 ` [PATCH wireless-next v2] wifi: brcmfmac: turn off TX beamforming on Apple BCM4364 Andre Eikmeyer
  0 siblings, 2 replies; 5+ messages in thread
From: Andre Eikmeyer @ 2026-11-04 15:16 UTC (permalink / raw)
  To: arend.vanspriel
  Cc: linux-wireless, brcm80211, brcm80211-dev-list.pdl, linux-kernel,
	Andre Eikmeyer, Yossef Osman, Aditya Garg

With "txbf" enabled, the BCM4364 firmware (9.30.514.0.32.5.94, as shipped
by Apple for the MacBookPro15,1) transmits a single beamformed spatial
stream (nrate VHT MCS9 NSS1 TXBF) towards a two stream AP. With "txbf"
cleared it picks two streams (VHT MCS9 NSS2) on its own. The firmware
enables "txbf" by default, so the driver has to clear it.

Measured with iperf3 against a wired server on the same AP:

  0.5 m from the AP, -36 dBm:  351 -> 623 Mbit/s uplink
  -70 dBm:                      96-130 -> 149-175 Mbit/s uplink

Downlink and latency are unchanged. I did not notice any refression.

Tested-by: Yossef Osman <Yossef.Osman@proton.me>
Tested-by: Aditya Garg <aditya.garg@linux.dev>
Signed-off-by: Andre Eikmeyer <dev@deq.rocks>
---
The KaiT2en community runs this patch with its kernels, no regression
was reported.

 .../net/wireless/broadcom/brcm80211/brcmfmac/common.c    | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/common.c b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/common.c
index b6027d83e6a..b09abd5fee3 100644
--- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/common.c
+++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/common.c
@@ -10,6 +10,7 @@
 #include <linux/firmware.h>
 #include <brcmu_wifi.h>
 #include <brcmu_utils.h>
+#include <brcm_hw_ids.h>
 #include "core.h"
 #include "bus.h"
 #include "debug.h"
@@ -450,8 +451,12 @@ int brcmf_c_preinit_dcmds(struct brcmf_if *ifp)
 		goto done;
 	}
 
-	/* Enable tx beamforming, errors can be ignored (not supported) */
-	(void)brcmf_fil_iovar_int_set(ifp, "txbf", 1);
+	/* Enable tx beamforming, errors can be ignored (not supported).
+	 * BCM4364 firmware enables it by default and then transmits a single
+	 * beamformed stream instead of two, halving its uplink throughput.
+	 */
+	(void)brcmf_fil_iovar_int_set(ifp, "txbf",
+				      drvr->bus_if->chip != BRCM_CC_4364_CHIP_ID);
 done:
 	return err;
 }

base-commit: 991a73f4909f1348ffb4b38765cdb522a775e4c2

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] wifi: brcmfmac: disable TX beamforming on BCM4364
  2026-11-04 15:16 [PATCH] wifi: brcmfmac: disable TX beamforming on BCM4364 Andre Eikmeyer
@ 2026-11-16 20:37 ` Arend van Spriel
  2026-11-17 17:50   ` Aditya Garg
  2026-11-18 14:10 ` [PATCH wireless-next v2] wifi: brcmfmac: turn off TX beamforming on Apple BCM4364 Andre Eikmeyer
  1 sibling, 1 reply; 5+ messages in thread
From: Arend van Spriel @ 2026-11-16 20:37 UTC (permalink / raw)
  To: Andre Eikmeyer
  Cc: arend.vanspriel, linux-wireless, brcm80211,
	brcm80211-dev-list.pdl, linux-kernel, Yossef Osman, Aditya Garg

On Wed, 4 Nov 2026 16:16:23 +0100, Andre Eikmeyer <dev@deq.rocks> wrote:
> With "txbf" enabled, the BCM4364 firmware (9.30.514.0.32.5.94, as shipped
> by Apple for the MacBookPro15,1) transmits a single beamformed spatial
> stream (nrate VHT MCS9 NSS1 TXBF) towards a two stream AP. With "txbf"
> cleared it picks two streams (VHT MCS9 NSS2) on its own. The firmware
> enables "txbf" by default, so the driver has to clear it.
>
> [...]
>
> Tested-by: Yossef Osman <Yossef.Osman@proton.me>
> Tested-by: Aditya Garg <aditya.garg@linux.dev>
> Signed-off-by: Andre Eikmeyer <dev@deq.rocks>

[...]

> @@ -450,8 +451,12 @@ int brcmf_c_preinit_dcmds(struct brcmf_if *ifp)
>  		goto done;
>  	}
>
> -	/* Enable tx beamforming, errors can be ignored (not supported) */
> -	(void)brcmf_fil_iovar_int_set(ifp, "txbf", 1);
> +	/* Enable tx beamforming, errors can be ignored (not supported).
> +	 * BCM4364 firmware enables it by default and then transmits a single
> +	 * beamformed stream instead of two, halving its uplink throughput.
> +	 */
> +	(void)brcmf_fil_iovar_int_set(ifp, "txbf",
> +				      drvr->bus_if->chip != BRCM_CC_4364_CHIP_ID);
>  done:
>  	return err;
>  }

The chip ID check has the same problem as in the ARP/ND offload patch [1]:
the issue is described as specific to the firmware shipped by Apple
(9.30.514.0.32.5.94), but the quirk applies to every BCM4364. Please
restrict it as suggested there.

Also, the comment says "Enable tx beamforming" while the code now
conditionally disables it. Please make the code self-explanatory, e.g. a
local variable or an explicit if (chip == BRCM_CC_4364_CHIP_ID) with the
txbf value spelled out, instead of passing a bool expression as the
iovar value.

Minor: typo "refression" in the commit message, and the subject lacks
the target tree (wireless-next) which wifibot flagged.

Since this is a throughput quirk (not a fix) I would also like to know
whether TX beamforming is functional at all on the BCM4364 or just
mis-selected by rate control, so the commit message can state the actual
root cause.

[1] https://lore.kernel.org/all/20261104151620.70671-1-dev@deq.rocks/

Regards,
Arend

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] wifi: brcmfmac: disable TX beamforming on BCM4364
  2026-11-16 20:37 ` Arend van Spriel
@ 2026-11-17 17:50   ` Aditya Garg
  2026-11-17 19:46     ` Arend van Spriel
  0 siblings, 1 reply; 5+ messages in thread
From: Aditya Garg @ 2026-11-17 17:50 UTC (permalink / raw)
  To: Arend van Spriel
  Cc: Andre Eikmeyer, arend.vanspriel, linux-wireless, brcm80211,
	brcm80211-dev-list.pdl, linux-kernel, Yossef Osman



> On 17 Nov 2026, at 2:07 AM, Arend van Spriel <arend.vanspriel@broadcom.com> wrote:
> 
> On Wed, 4 Nov 2026 16:16:23 +0100, Andre Eikmeyer <dev@deq.rocks> wrote:
>> With "txbf" enabled, the BCM4364 firmware (9.30.514.0.32.5.94, as shipped
>> by Apple for the MacBookPro15,1) transmits a single beamformed spatial
>> stream (nrate VHT MCS9 NSS1 TXBF) towards a two stream AP. With "txbf"
>> cleared it picks two streams (VHT MCS9 NSS2) on its own. The firmware
>> enables "txbf" by default, so the driver has to clear it.
>> 
>> [...]
>> 
>> Tested-by: Yossef Osman <Yossef.Osman@proton.me>
>> Tested-by: Aditya Garg <aditya.garg@linux.dev>
>> Signed-off-by: Andre Eikmeyer <dev@deq.rocks>
> 
> [...]
> 
>> @@ -450,8 +451,12 @@ int brcmf_c_preinit_dcmds(struct brcmf_if *ifp)
>>        goto done;
>>    }
>> 
>> -    /* Enable tx beamforming, errors can be ignored (not supported) */
>> -    (void)brcmf_fil_iovar_int_set(ifp, "txbf", 1);
>> +    /* Enable tx beamforming, errors can be ignored (not supported).
>> +     * BCM4364 firmware enables it by default and then transmits a single
>> +     * beamformed stream instead of two, halving its uplink throughput.
>> +     */
>> +    (void)brcmf_fil_iovar_int_set(ifp, "txbf",
>> +                      drvr->bus_if->chip != BRCM_CC_4364_CHIP_ID);
>> done:
>>    return err;
>> }
> 
> The chip ID check has the same problem as in the ARP/ND offload patch [1]:
> the issue is described as specific to the firmware shipped by Apple
> (9.30.514.0.32.5.94), but the quirk applies to every BCM4364. Please
> restrict it as suggested there.
> 
> Also, the comment says "Enable tx beamforming" while the code now
> conditionally disables it. Please make the code self-explanatory, e.g. a
> local variable or an explicit if (chip == BRCM_CC_4364_CHIP_ID) with the
> txbf value spelled out, instead of passing a bool expression as the
> iovar value.
> 
> Minor: typo "refression" in the commit message, and the subject lacks
> the target tree (wireless-next) which wifibot flagged.
> 
> Since this is a throughput quirk (not a fix) I would also like to know
> whether TX beamforming is functional at all on the BCM4364 or just
> mis-selected by rate control, so the commit message can state the actual
> root cause.

Unrelated but Airdrop conversation seems to have been stalled :(.

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] wifi: brcmfmac: disable TX beamforming on BCM4364
  2026-11-17 17:50   ` Aditya Garg
@ 2026-11-17 19:46     ` Arend van Spriel
  0 siblings, 0 replies; 5+ messages in thread
From: Arend van Spriel @ 2026-11-17 19:46 UTC (permalink / raw)
  To: Aditya Garg
  Cc: Andre Eikmeyer, linux-wireless, brcm80211,
	brcm80211-dev-list.pdl, linux-kernel, Yossef Osman

On 17/11/2026 18:50, Aditya Garg wrote:
> 
> 
>> On 17 Nov 2026, at 2:07 AM, Arend van Spriel <arend.vanspriel@broadcom.com> wrote:
>>
>> On Wed, 4 Nov 2026 16:16:23 +0100, Andre Eikmeyer <dev@deq.rocks> wrote:
>>> With "txbf" enabled, the BCM4364 firmware (9.30.514.0.32.5.94, as shipped
>>> by Apple for the MacBookPro15,1) transmits a single beamformed spatial
>>> stream (nrate VHT MCS9 NSS1 TXBF) towards a two stream AP. With "txbf"
>>> cleared it picks two streams (VHT MCS9 NSS2) on its own. The firmware
>>> enables "txbf" by default, so the driver has to clear it.
>>>
>>> [...]
>>>
>>> Tested-by: Yossef Osman <Yossef.Osman@proton.me>
>>> Tested-by: Aditya Garg <aditya.garg@linux.dev>
>>> Signed-off-by: Andre Eikmeyer <dev@deq.rocks>
>>
>> [...]
>>
>>> @@ -450,8 +451,12 @@ int brcmf_c_preinit_dcmds(struct brcmf_if *ifp)
>>>         goto done;
>>>     }
>>>
>>> -    /* Enable tx beamforming, errors can be ignored (not supported) */
>>> -    (void)brcmf_fil_iovar_int_set(ifp, "txbf", 1);
>>> +    /* Enable tx beamforming, errors can be ignored (not supported).
>>> +     * BCM4364 firmware enables it by default and then transmits a single
>>> +     * beamformed stream instead of two, halving its uplink throughput.
>>> +     */
>>> +    (void)brcmf_fil_iovar_int_set(ifp, "txbf",
>>> +                      drvr->bus_if->chip != BRCM_CC_4364_CHIP_ID);
>>> done:
>>>     return err;
>>> }
>>
>> The chip ID check has the same problem as in the ARP/ND offload patch [1]:
>> the issue is described as specific to the firmware shipped by Apple
>> (9.30.514.0.32.5.94), but the quirk applies to every BCM4364. Please
>> restrict it as suggested there.
>>
>> Also, the comment says "Enable tx beamforming" while the code now
>> conditionally disables it. Please make the code self-explanatory, e.g. a
>> local variable or an explicit if (chip == BRCM_CC_4364_CHIP_ID) with the
>> txbf value spelled out, instead of passing a bool expression as the
>> iovar value.
>>
>> Minor: typo "refression" in the commit message, and the subject lacks
>> the target tree (wireless-next) which wifibot flagged.
>>
>> Since this is a throughput quirk (not a fix) I would also like to know
>> whether TX beamforming is functional at all on the BCM4364 or just
>> mis-selected by rate control, so the commit message can state the actual
>> root cause.
> 
> Unrelated but Airdrop conversation seems to have been stalled :(.

Indeed unrelated so let me respond to your earlier email that was more 
on the subject.

Regards,
Arend

^ permalink raw reply	[flat|nested] 5+ messages in thread

* [PATCH wireless-next v2] wifi: brcmfmac: turn off TX beamforming on Apple BCM4364
  2026-11-04 15:16 [PATCH] wifi: brcmfmac: disable TX beamforming on BCM4364 Andre Eikmeyer
  2026-11-16 20:37 ` Arend van Spriel
@ 2026-11-18 14:10 ` Andre Eikmeyer
  1 sibling, 0 replies; 5+ messages in thread
From: Andre Eikmeyer @ 2026-11-18 14:10 UTC (permalink / raw)
  To: arend.vanspriel
  Cc: linux-wireless, brcm80211, brcm80211-dev-list.pdl, linux-kernel,
	Andre Eikmeyer, Yossef Osman, Aditya Garg

brcmfmac sets "txbf" to 1 at init. The Apple firmware for the BCM4364
(9.30.514.0.32.5.94 on the MacBookPro15,1) has it on by default as
well, and then rate control settles on a single beamformed spatial
stream (nrate VHT MCS9 NSS1 TXBF) towards a two stream AP.

Beamforming more than one stream does not work on this firmware. With
"txbf" on and the rate fixed to two or three streams, the firmware takes
the rate but the uplink stalls. With "txbf" off, rate control picks two
streams (VHT MCS9 NSS2) on its own.

Uplink measured with iperf3 against a wired server was:

  0.5 m from the AP, -36 dBm:  351 -> 623 Mbit/s
  -70 dBm:                      96-130 -> 149-175 Mbit/s

Downlink and latency remain unchanged.

Set "txbf" to 0 on the BCM4364 when the board type is an Apple one.

Tested-by: Yossef Osman <Yossef.Osman@proton.me>
Tested-by: Aditya Garg <aditya.garg@linux.dev>
Signed-off-by: Andre Eikmeyer <dev@deq.rocks>
---
Changes in v2:
- Restrict the quirk to Apple boards with settings->board_type
- Spell out the txbf value with an explicit if and fix the comment
- State the cause. Beamforming more than one stream does not work
- Fix the typo
- Target wireless-next
- v1: https://lore.kernel.org/all/20261104151623.70695-1-dev@deq.rocks/

The T2 Linux community runs this patch with its kernels for quite a while
now, no regression was reported.

 .../wireless/broadcom/brcm80211/brcmfmac/common.c | 15 +++++++++++++--
 1 file changed, 13 insertions(+), 2 deletions(-)

diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/common.c b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/common.c
index 93458c2fbc2e..f8ff6217de16 100644
--- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/common.c
+++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/common.c
@@ -10,6 +10,7 @@
 #include <linux/firmware.h>
 #include <brcmu_wifi.h>
 #include <brcmu_utils.h>
+#include <brcm_hw_ids.h>
 #include "core.h"
 #include "bus.h"
 #include "debug.h"
@@ -275,6 +276,7 @@ int brcmf_c_preinit_dcmds(struct brcmf_if *ifp)
 	struct brcmf_rev_info *ri;
 	char *clmver;
 	char *ptr;
+	s32 txbf;
 	s32 err;
 
 	if (is_valid_ether_addr(ifp->mac_addr)) {
@@ -432,8 +434,17 @@ int brcmf_c_preinit_dcmds(struct brcmf_if *ifp)
 		goto done;
 	}
 
-	/* Enable tx beamforming, errors can be ignored (not supported) */
-	(void)brcmf_fil_iovar_int_set(ifp, "txbf", 1);
+	/* Enable tx beamforming, errors can be ignored (not supported).
+	 * The Apple firmware for the BCM4364 has it on by default and then
+	 * beamforms a single spatial stream, as beamforming more than one
+	 * stream does not work. Turn it off there, so two streams are used.
+	 */
+	txbf = 1;
+	if (drvr->bus_if->chip == BRCM_CC_4364_CHIP_ID &&
+	    drvr->settings->board_type &&
+	    strstarts(drvr->settings->board_type, "apple,"))
+		txbf = 0;
+	(void)brcmf_fil_iovar_int_set(ifp, "txbf", txbf);
 done:
 	return err;
 }

base-commit: 5dbf2415ca9821a611d09308e3b29d2c37e7cd77
-- 
2.56.0


^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2026-11-18 14:10 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-11-04 15:16 [PATCH] wifi: brcmfmac: disable TX beamforming on BCM4364 Andre Eikmeyer
2026-11-16 20:37 ` Arend van Spriel
2026-11-17 17:50   ` Aditya Garg
2026-11-17 19:46     ` Arend van Spriel
2026-11-18 14:10 ` [PATCH wireless-next v2] wifi: brcmfmac: turn off TX beamforming on Apple BCM4364 Andre Eikmeyer

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®