On Wed, 5 Jul 2023 06:20:26 +0000 Wei Fang wrote: > > > In addtion, this patch also updates the tx_stop_threshold and the > > > tx_wake_threshold of the tx ring. In previous logic, the value of > > > tx_stop_threshold is 217, however, the value of tx_wake_threshold is > > > 147, it does not make sense that tx_wake_threshold is less than > > > tx_stop_threshold. > > > > What do these actually mean? I could imagine that as the ring fills you don't > > want to stop until it is 217/512 full. There is then some hysteresis, such that it > > has to drop below 147/512 before more can be added? > > > You must have misunderstood, let me explain more clearly, the queue will be > stopped when the available BDs are less than tx_stop_threshold (217 BDs). And > the queue will be waked when the available BDs are greater than tx_wake_threshold > (147 BDs). So in most cases, the available BDs are greater than tx_wake_threshold > when the queue is stopped, the only effect is to delay packet sending. > In my opinion, tx_wake_threshold should be greater than tx_stop_threshold, we > should stop queue when the available BDs are not enough for a skb to be attached. > And wake the queue when the available BDs are sufficient for a skb. But you shouldn't restart the queue for a single packet either. Restarting for a single packet wastes CPU cycles as there will be much more stop / start operations. Two large packets seem like the absolute minimum reasonable wake threshold. Setting tx_stop_threshold to MAX_SKB_FRAGS doesn't seem right either, as you won't be able to accept a full TSO frame. Please split the change, the netdev_err_once() should be one patch and then the change to wake thresh a separate one. -- pw-bot: cr
