mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH] wifi: nl80211: don't WARN on a missing self-managed regdom
@ 2026-10-06 16:17 François Roux
  2026-10-07 12:33 ` Diederik de Haas
  2026-10-07 14:55 ` Johannes Berg
  0 siblings, 2 replies; 5+ messages in thread
From: François Roux @ 2026-10-06 16:17 UTC (permalink / raw)
  To: Johannes Berg
  Cc: Diederik de Haas, Emmanuel Grumbach, linux-wireless, linux-kernel

nl80211_get_reg_do() has a WARN_ON() for a self-managed wiphy without a
private regdomain. Userspace can trigger it with a plain
NL80211_CMD_GET_REG request.

Drivers can register such a wiphy first and set its regdomain later.
ath12k, for example, calls ieee80211_register_hw() and only then applies
the regdomain with ath12k_regd_update(). iwd queries the regdomain of a
new wiphy as soon as it shows up, so it can hit that window at boot:

  WARNING: net/wireless/nl80211.c:10828 at nl80211_get_reg_do+0x1d0/0x1fc [cfg80211], CPU#5: iwd/721
  Call trace:
   nl80211_get_reg_do+0x1d0/0x1fc [cfg80211] (P)
   genl_family_rcv_msg_doit+0xd0/0x138
   genl_rcv_msg+0x58/0xd4
   netlink_rcv_skb+0x5c/0x140
   genl_rcv+0x34/0x4c
   netlink_unicast+0x31c/0x39c
   netlink_sendmsg+0x188/0x400

The same trace has been reported with an Intel BE200 (iwlwifi). The dump
variant, nl80211_get_reg_dump(), already skips a wiphy without a
regdomain without warning.

Keep returning -EINVAL, but drop the WARN_ON(). The same change was
posted in 2021 but was never applied.

Seen on a Microsoft Surface Pro 11 (WCN7850, ath12k) in 3 of 4 boots
with next-20260929.

Fixes: 1bdd716cbcca ("cfg80211: return private regdom for self-managed devices")
Reported-by: Diederik de Haas <diederik@cknow-tech.com>
Closes: https://lore.kernel.org/all/DL5P72WYQVOQ.23G9D01OY3BAA@cknow-tech.com/
Link: https://lore.kernel.org/all/iwlwifi.20210409123755.ba2ea961f4ae.I8fde32d3196e860efa3b4ec464c42194195b42ec@changeid/
Assisted-by: LLM
Signed-off-by: François Roux <franzelfranzel@gmail.com>
---
 net/wireless/nl80211.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/net/wireless/nl80211.c b/net/wireless/nl80211.c
index a18240455..2589491af 100644
--- a/net/wireless/nl80211.c
+++ b/net/wireless/nl80211.c
@@ -10824,8 +10824,12 @@ static int nl80211_get_reg_do(struct sk_buff *skb, struct genl_info *info)
 
 		regdom = get_wiphy_regdom(wiphy);
 
-		/* a self-managed-reg device must have a private regdom */
-		if (WARN_ON(!regdom && self_managed)) {
+		/*
+		 * A self-managed-reg device gets its private regdom from the
+		 * driver, which may happen only after the wiphy is registered.
+		 * Userspace can ask before that, so don't WARN.
+		 */
+		if (!regdom && self_managed) {
 			err = -EINVAL;
 			goto nla_put_failure_rcu;
 		}

base-commit: 6474fa070f2b8013b4b87350b775b8c3be6e8aac
-- 
2.56.0


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

* Re: [PATCH] wifi: nl80211: don't WARN on a missing self-managed regdom
  2026-10-06 16:17 [PATCH] wifi: nl80211: don't WARN on a missing self-managed regdom François Roux
@ 2026-10-07 12:33 ` Diederik de Haas
  2026-10-07 14:55 ` Johannes Berg
  1 sibling, 0 replies; 5+ messages in thread
From: Diederik de Haas @ 2026-10-07 12:33 UTC (permalink / raw)
  To: François Roux, Johannes Berg
  Cc: Diederik de Haas, Emmanuel Grumbach, linux-wireless, linux-kernel

Hi François, 

On Tue Oct 6, 2026 at 6:17 PM CEST, François Roux wrote:
> nl80211_get_reg_do() has a WARN_ON() for a self-managed wiphy without a
> private regdomain. Userspace can trigger it with a plain
> NL80211_CMD_GET_REG request.
>
> Drivers can register such a wiphy first and set its regdomain later.
> ath12k, for example, calls ieee80211_register_hw() and only then applies
> the regdomain with ath12k_regd_update(). iwd queries the regdomain of a
> new wiphy as soon as it shows up, so it can hit that window at boot:
>
>   WARNING: net/wireless/nl80211.c:10828 at nl80211_get_reg_do+0x1d0/0x1fc [cfg80211], CPU#5: iwd/721
>   Call trace:
>    nl80211_get_reg_do+0x1d0/0x1fc [cfg80211] (P)
>    genl_family_rcv_msg_doit+0xd0/0x138
>    genl_rcv_msg+0x58/0xd4
>    netlink_rcv_skb+0x5c/0x140
>    genl_rcv+0x34/0x4c
>    netlink_unicast+0x31c/0x39c
>    netlink_sendmsg+0x188/0x400
>
> The same trace has been reported with an Intel BE200 (iwlwifi). The dump
> variant, nl80211_get_reg_dump(), already skips a wiphy without a
> regdomain without warning.
>
> Keep returning -EINVAL, but drop the WARN_ON(). The same change was
> posted in 2021 but was never applied.
>
> Seen on a Microsoft Surface Pro 11 (WCN7850, ath12k) in 3 of 4 boots
> with next-20260929.

Thanks for this patch :-)

I added it to my 7.3-rc6 based kernel and booted into it.
With that kernel but without this patch, I (still) got the stack trace
as I reported earlier. But with this patch applied, I only see the following
in dmesg when I start ``iwd.service``:

```sh
[   99.031552] ACPI: <n/a>: failed to evaluate _DSM bf0212f2-788f-c64d-a5b3-1f738e285ade rev:0 func:0 (0x1001)
[   99.619075] wlan0: authenticate with 20:05:b7:00:29:19 (local address=f8:cf:52:97:97:58)
[   99.619887] wlan0: send auth to 20:05:b7:00:29:19 (try 1/3)
[   99.622752] wlan0: authenticated
[   99.624881] wlan0: associate with 20:05:b7:00:29:19 (try 1/3)
[   99.627724] wlan0: RX AssocResp from 20:05:b7:00:29:19 (capab=0x111 status=0 aid=1)
[   99.631005] wlan0: associated
[   99.657741] wlan0: Limiting TX power to 20 (23 - 3) dBm as advertised by 20:05:b7:00:29:19
```

Thus no longer a stack trace. So feel free to include my

Tested-by: Diederik de Haas <diederik@cknow-tech.com>  # NanoPC-T6 LTS + Intel BE200NGW

Cheers,
  Diederik

> Fixes: 1bdd716cbcca ("cfg80211: return private regdom for self-managed devices")
> Reported-by: Diederik de Haas <diederik@cknow-tech.com>
> Closes: https://lore.kernel.org/all/DL5P72WYQVOQ.23G9D01OY3BAA@cknow-tech.com/
> Link: https://lore.kernel.org/all/iwlwifi.20210409123755.ba2ea961f4ae.I8fde32d3196e860efa3b4ec464c42194195b42ec@changeid/
> Assisted-by: LLM
> Signed-off-by: François Roux <franzelfranzel@gmail.com>
> ---
>  net/wireless/nl80211.c | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/net/wireless/nl80211.c b/net/wireless/nl80211.c
> index a18240455..2589491af 100644
> --- a/net/wireless/nl80211.c
> +++ b/net/wireless/nl80211.c
> @@ -10824,8 +10824,12 @@ static int nl80211_get_reg_do(struct sk_buff *skb, struct genl_info *info)
>  
>  		regdom = get_wiphy_regdom(wiphy);
>  
> -		/* a self-managed-reg device must have a private regdom */
> -		if (WARN_ON(!regdom && self_managed)) {
> +		/*
> +		 * A self-managed-reg device gets its private regdom from the
> +		 * driver, which may happen only after the wiphy is registered.
> +		 * Userspace can ask before that, so don't WARN.
> +		 */
> +		if (!regdom && self_managed) {
>  			err = -EINVAL;
>  			goto nla_put_failure_rcu;
>  		}
>
> base-commit: 6474fa070f2b8013b4b87350b775b8c3be6e8aac



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

* Re: [PATCH] wifi: nl80211: don't WARN on a missing self-managed regdom
  2026-10-06 16:17 [PATCH] wifi: nl80211: don't WARN on a missing self-managed regdom François Roux
  2026-10-07 12:33 ` Diederik de Haas
@ 2026-10-07 14:55 ` Johannes Berg
  2026-10-07 15:18   ` Diederik de Haas
  1 sibling, 1 reply; 5+ messages in thread
From: Johannes Berg @ 2026-10-07 14:55 UTC (permalink / raw)
  To: François Roux
  Cc: Diederik de Haas, Emmanuel Grumbach, linux-wireless, linux-kernel

On Tue, 2026-10-06 at 18:17 +0200, François Roux wrote:
> nl80211_get_reg_do() has a WARN_ON() for a self-managed wiphy without a
> private regdomain. Userspace can trigger it with a plain
> NL80211_CMD_GET_REG request.
> 
> Drivers can register such a wiphy first and set its regdomain later.
> ath12k, for example, calls ieee80211_register_hw() and only then applies
> the regdomain with ath12k_regd_update(). iwd queries the regdomain of a
> new wiphy as soon as it shows up, so it can hit that window at boot:
> 
>   WARNING: net/wireless/nl80211.c:10828 at nl80211_get_reg_do+0x1d0/0x1fc [cfg80211], CPU#5: iwd/721

I'm not a fan of this, I think the drivers should have the regdom ready.

> The same trace has been reported with an Intel BE200 (iwlwifi).

I _thought_ we fixed that issue.

Anyway I think we need to fix the bugs, not paper over them. If it comes
to it we can even put a WARN_ON there to reject wiphy registration w/o a
regdomain. There should be one. As it is, iwd will get empty data or
something, that makes no sense.

johannes

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

* Re: [PATCH] wifi: nl80211: don't WARN on a missing self-managed regdom
  2026-10-07 14:55 ` Johannes Berg
@ 2026-10-07 15:18   ` Diederik de Haas
  2026-10-07 15:20     ` Johannes Berg
  0 siblings, 1 reply; 5+ messages in thread
From: Diederik de Haas @ 2026-10-07 15:18 UTC (permalink / raw)
  To: Johannes Berg, François Roux
  Cc: Diederik de Haas, Emmanuel Grumbach, linux-wireless, linux-kernel

On Wed Oct 7, 2026 at 4:55 PM CEST, Johannes Berg wrote:
> On Tue, 2026-10-06 at 18:17 +0200, François Roux wrote:
>> nl80211_get_reg_do() has a WARN_ON() for a self-managed wiphy without a
>> private regdomain. Userspace can trigger it with a plain
>> NL80211_CMD_GET_REG request.
>> 
>> Drivers can register such a wiphy first and set its regdomain later.
>> ath12k, for example, calls ieee80211_register_hw() and only then applies
>> the regdomain with ath12k_regd_update(). iwd queries the regdomain of a
>> new wiphy as soon as it shows up, so it can hit that window at boot:
>> 
>>   WARNING: net/wireless/nl80211.c:10828 at nl80211_get_reg_do+0x1d0/0x1fc [cfg80211], CPU#5: iwd/721
>
> I'm not a fan of this, I think the drivers should have the regdom ready.
>
>> The same trace has been reported with an Intel BE200 (iwlwifi).
>
> I _thought_ we fixed that issue.

If that fix is supposed to be in 7.3-rc6 (from Linus), then no.
If that's a recent fix/patch which is f.e. slated for -next/7.4(-rc1),
then can you point me to it? Happy to test it.

Before I reported the/my issue with Intel BE200 [1] I did search a bit
on the linux-wireless ML and saw some patches where ACPI was an (important?)
part. That's why I (so) explicitly mentioned that my system does NOT have
an ACPI based firmware (but U-Boot (with TF-A)).

Cheers,
  Diederik

[1] https://lore.kernel.org/linux-wireless/DL5P72WYQVOQ.23G9D01OY3BAA@cknow-tech.com/

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

* Re: [PATCH] wifi: nl80211: don't WARN on a missing self-managed regdom
  2026-10-07 15:18   ` Diederik de Haas
@ 2026-10-07 15:20     ` Johannes Berg
  0 siblings, 0 replies; 5+ messages in thread
From: Johannes Berg @ 2026-10-07 15:20 UTC (permalink / raw)
  To: Diederik de Haas, François Roux
  Cc: Emmanuel Grumbach, linux-wireless, linux-kernel

On Wed, 2026-10-07 at 17:18 +0200, Diederik de Haas wrote:
> On Wed Oct 7, 2026 at 4:55 PM CEST, Johannes Berg wrote:
> > On Tue, 2026-10-06 at 18:17 +0200, François Roux wrote:
> > > nl80211_get_reg_do() has a WARN_ON() for a self-managed wiphy without a
> > > private regdomain. Userspace can trigger it with a plain
> > > NL80211_CMD_GET_REG request.
> > > 
> > > Drivers can register such a wiphy first and set its regdomain later.
> > > ath12k, for example, calls ieee80211_register_hw() and only then applies
> > > the regdomain with ath12k_regd_update(). iwd queries the regdomain of a
> > > new wiphy as soon as it shows up, so it can hit that window at boot:
> > > 
> > >   WARNING: net/wireless/nl80211.c:10828 at nl80211_get_reg_do+0x1d0/0x1fc [cfg80211], CPU#5: iwd/721
> > 
> > I'm not a fan of this, I think the drivers should have the regdom ready.
> > 
> > > The same trace has been reported with an Intel BE200 (iwlwifi).
> > 
> > I _thought_ we fixed that issue.
> 
> If that fix is supposed to be in 7.3-rc6 (from Linus), then no.
> If that's a recent fix/patch which is f.e. slated for -next/7.4(-rc1),
> then can you point me to it? Happy to test it.

Well, it looks like it got *reverted*.

6895d74c11d8 ("wifi: iwlwifi: mld: initialize regulatory early")
f81aa834bfa9 ("wifi: iwlwifi: mld: Move regulatory domain initialization")

johannes

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

end of thread, other threads:[~2026-10-07 15:20 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-10-06 16:17 [PATCH] wifi: nl80211: don't WARN on a missing self-managed regdom François Roux
2026-10-07 12:33 ` Diederik de Haas
2026-10-07 14:55 ` Johannes Berg
2026-10-07 15:18   ` Diederik de Haas
2026-10-07 15:20     ` Johannes Berg

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®