Alexander Wetzel <alexander@xxxxxxxxxxxxxx> writes:
> Calling only __ieee80211_select_queue() for control port TX exposes
> drivers which do not support QoS to non-zero values in
> skb->queue_mapping and even can assign not available queues to
> info->hw_queue.
> This can cause issues for drivers like we did e.g. see in
> '746285cf81dc ("rtl818x: Prevent using not initialized queues")'.
>
> This also prevents a redundant call to __ieee80211_select_queue() when
> using control port TX with iTXQ (pull path).
> And it starts to prioritize 802.11 preauthentication frames
> (ETH_P_PREAUTH) on all TX paths.
>
> Pierre Asselin confirmed that this patch indeed prevents crashing his
> system without '746285cf81dc ("rtl818x: Prevent using not initialized
> queues")'.
>
> Tested-by: Pierre Asselin <pa@xxxxxxxxx>
> Signed-off-by: Alexander Wetzel <alexander@xxxxxxxxxxxxxx>
> ---
>
> Starting to prioritize ETH_P_PREAUTH was just added since I noticed that
> contradictory to at least my expectations control port does accept
> ETH_P_PREAUTH but handles these like a normal frame for the priority.
> That can be broken out or even drop, when needed.
>
> While looking at the code I also tripped over multiple other questions
> and I'll probably propose a much more invasive change how to handle
> the queue assignment. (End2end we seem to do some quite stupid things.)
>
> Additionally I really don't get why we call skb_get_hash() on queue
> assignment:
> I found the commit '180ac48ee62f ("mac80211: calculate skb hash early
> when using itxq")' but don't see why calculating the hash early is
> useful. Any hints here are appreciated. fq_flow_idx() seems to do that
> when needed and I can't find any other usage of the hash...
The commit message of that commit has a hint:
This avoids flow separation issues when using software encryption.
The idea being that the packet contents can change on encryption, but
skb->hash is preserved, so you want it to run before encryption happens
to keep flows in the same queue.
However, AFAICT ieee80211_tx_h_encrypt() is called after frames are
dequeued from the TXQs, so not actually sure this is needed. Adding
Felix, in the hope that he can explain the reasoning behind that commit :)
-Toke
