On Tue, Feb 4, 2020 at 6:20 PM Björn Töpel <bjorn.topel@xxxxxxxxx> wrote: > > > > On 2020-02-04 10:46, Maciej Fijalkowski wrote: > > On Tue, Feb 04, 2020 at 04:41:31PM +0100, Björn Töpel wrote: > [...] > > > > So to summarize that: > > - return EAGAIN instead of ENETDOWN within ndo_xsk_wakeup implementations > > in i40e/ixgbe, > > - check for ENETDOWN in kick_tx() as well > > > > Do you guys agree? > > > > Well, start with latter. From some perspective, changing the return > value is ABI breakage, no? > > > What about timeout and syscall repetition? That's the last question. > > > > Hmm, timeout in what context? When to redo the syscall? Cameron, FYI: https://lore.kernel.org/netdev/20200205212147.GA5358@xxxxxxxxx/T/#r715e0a78ae13c0fc1ce67b3219e9a532b2bfd011 /Magnus
|