> -----Original Message----- > From: Jakub Kicinski <kuba@xxxxxxxxxx> > Sent: Tuesday, December 6, 2022 9:24 AM > To: Leon Romanovsky <leon@xxxxxxxxxx> > Cc: Veerasenareddy Burru <vburru@xxxxxxxxxxx>; David S. Miller > <davem@xxxxxxxxxxxxx>; Eric Dumazet <edumazet@xxxxxxxxxx>; Paolo > Abeni <pabeni@xxxxxxxxxx>; netdev@xxxxxxxxxxxxxxx; linux- > kernel@xxxxxxxxxxxxxxx; Liron Himi <lironh@xxxxxxxxxxx>; Abhijit Ayarekar > <aayarekar@xxxxxxxxxxx>; Sathesh B Edara <sedara@xxxxxxxxxxx>; > Satananda Burla <sburla@xxxxxxxxxxx>; linux-doc@xxxxxxxxxxxxxxx > Subject: Re: [EXT] Re: [PATCH net-next v2 2/9] octeon_ep: poll for control > messages > > On Tue, 6 Dec 2022 10:58:47 +0200 Leon Romanovsky wrote: > > > Polling for control messages every 100ms? Sure. > > > > > > You say "valid in netdev" so perhaps you can educate us where/why it > > > would not be? > > > > It doesn't seem right to me that idle device burns CPU cycles, while > > it supports interrupts. If it needs "listen to FW", it will be much > > nicer to install interrupts immediately and don't wait for netdev. > > No doubt, if there is an alternative we can push for it to be implemented. I > guess this being yet another "IPU" there could be possible workarounds in > FW? As always with IPUs - hard to tell :/ > > If there is no alternative - it is what it is. > It's up to customers to buy good HW. > > That said, looking at what this set does - how are the VFs configured? > That's the showstopper for the series in my mind. VFs are created by writing to sriov_numvfs.