The net lock should be held to keep the assumption, how about this patch: https://github.com/apache/incubator-nuttx/pull/6368
On Sun, Jun 5, 2022 at 5:02 PM Michael Jung <michael.j...@secore.ly> wrote: > Hello everybody, > > With pull request #6330 there is a call to net_driver->d_txavail() from > tcp_timer_expiry(), which is scheduled on the low priority work queue. To > my understanding this means that net_driver->tx_avail() is called without > the network being locked, correct? > > At least in net/driver/skeleton.c the assumption that the network is locked > is explicitly stated for skel_txavail(). If my understanding is correct, > then we have at least some documentation inconsistency here, but this could > also be a serialization bug. > > Please advise > Michael > On Sun, Jun 5, 2022 at 5:02 PM Michael Jung <michael.j...@secore.ly> wrote: > Hello everybody, > > With pull request #6330 there is a call to net_driver->d_txavail() from > tcp_timer_expiry(), which is scheduled on the low priority work queue. To > my understanding this means that net_driver->tx_avail() is called without > the network being locked, correct? > > At least in net/driver/skeleton.c the assumption that the network is locked > is explicitly stated for skel_txavail(). If my understanding is correct, > then we have at least some documentation inconsistency here, but this could > also be a serialization bug. > > Please advise > Michael >
