[PATCH v3 2/3] wifi: mac80211: restructure tx profile retrieval for MLO MBSSID

[Jeff Johnson]

On 3/5/2025 10:39 AM, Aloka Dixit wrote:
> On 2/28/2025 5:44 AM, Johannes Berg wrote:
>> On Thu, 2025-02-13 at 14:50 -0800, Aloka Dixit wrote:
>>> From: Rameshkumar Sundaram <rameshkumar.sundaram at oss.qualcomm.com>
>>>
>>> For MBSSID, each vif (struct ieee80211_vif) stores another vif
>>> pointer for the transmitting profile of MBSSID set. This won't
>>> suffice for MLO as there may be multiple links, each of which can
>>> be part of different MBSSID sets. Hence the information needs to
>>> be stored per-link. Additionally, the transmitted profile itself
>>> may be part of an MLD hence storing vif will not suffice either.
>>> Fix MLO by storing an instance of struct ieee80211_bss_conf
>>> for each link.
>>>
>>> Modify following operations to reflect the above structure updates:
>>> - channel switch completion
>>> - BSS color change completion
>>> - Removing nontransmitted links in ieee80211_stop_mbssid()
>>> - drivers retrieving the transmitted link for beacon templates.
>>>
>>> Signed-off-by: Rameshkumar Sundaram <rameshkumar.sundaram at oss.qualcomm.com>
>>> Co-developed-by: Muna Sinada <muna.sinada at oss.qualcomm.com>
>>> Signed-off-by: Muna Sinada <muna.sinada at oss.qualcomm.com>
>>> Co-developed-by: Aloka Dixit <aloka.dixit at oss.qualcomm.com>
>>> Signed-off-by: Aloka Dixit <aloka.dixit at oss.qualcomm.com>
>>> ---
>>>   drivers/net/wireless/ath/ath11k/mac.c         | 20 +++--
>>>   drivers/net/wireless/ath/ath12k/mac.c         | 33 ++++++---
>>
>> This probably shouldn't come with so many driver changes in the first
>> place, but in any case the changes don't apply on my tree.
>>
>> johannes
>>
> 
> I will refactor some of the driver changes in a separate patch in next 
> version.
> 
> This series isn't applying cleanly because it depends on another 
> refactoring series I had mentioned in the cover letter:
> https://patchwork.kernel.org/project/linux-wireless/list/?series=932457&archive=both&state=*
> 
> I know it is a big no-no for mac80211 changes to depend on the driver 
> but sent both for the sake of completeness. It also explains why the 
> driver refactoring was required in the first place, because without that 
> series this current patch was even uglier trying to access tx conf at 
> 2-3 places, RCU accesses everywhere.
> 
> Anyway, the pre-req is now in ath.git, let's revisit this one once it is 
> merged in wireless-next.

My PR today will pull those driver changes into wireless-next.

/jeff