On Wed, 26 Jul 2023 16:01:23 +0530 Md Danish Anwar wrote:
I think MAX_SKB_FRAGS is OK. If the available pool = MAX_SKB_FRAGS we should be
able to wake the queue.
MAX_SKB_FRAGS only counts frags and you also need space to map the head, no?
In general we advise to wait until there's at least 2 * MAX_SKB_FRAGS
to avoid frequent sleep/wake cycles. But IDK how long your queue is,
maybe it's too much.
No I don't think any lock is required here. emac_set_port_state() aquires lock
before updating port status. Also emac_ndo_set_rx_mode_work() is scheduled by a
singlethreaded workqueue.
if (netif_running()) outside of any locks is usually a red flag, but if
you're confident it's fine it's fine :)