Kan Yan <kyan@xxxxxxxxxx> writes: > Patchset v6 works for me with ath10k driver. AQL does its job as > expected and tests show very significant reduction in latency in > congested environment. The txq stuck issue in patchset v4 got fixed. Awesome! Thank you for testing! > However, the device's total pending airtime count still underflows > sometimes. Even though it doesn't cause apparent side effect, there is > some issue with the pending airtime update and needs further > debugging. Huh, it *under*flows? That's... interesting. Cloned SKBs? Or maybe the tx_time_est field is being set in some other place? Could I get you to add a dump_stack() to the underflow test so we can get an idea of where that happens? -Toke
