On Mon, 2018-03-05 at 12:46 +0100, Oliver Neukum wrote: > On Mon, 2018-03-05 at 08:45 +0100, Marek Szyprowski wrote: > > Hi Oliver, > > > > On 2018-02-27 17:07, Oliver Neukum wrote: > > > Am Dienstag, den 27.02.2018, 07:13 -0800 schrieb Eric Dumazet: > > > > On Tue, 2018-02-27 at 07:09 -0800, Eric Dumazet wrote: > > > > > > > > > > Note that for this one, it seems we also could perform stats > > > > > updates in > > > > > BH context, since skb is queued via defer_bh() > > > > > > > > > > But simplicity wins I guess. > > > > > > > > Thinking more about this, I am not sure we have any guarantee > > > > that TX > > > > and RX can not run on multiple cpus. > > > > > > > > Using an unique syncp is not going to be safe, even if we make > > > > lockdep > > > > happy enough with the local_irq save/restore. > > > > > > Unfortunately you are right. It is not guaranteed for some > > > hardware. > > > > Does it mean that the fix proposed by Eric is not the proper > > solution? > > For asix it should work, but asix is unlikely to be the only driver > with that issue. 32 bit recieves less testing nowadays.
Yes, although the lockdep part could be enforced in 64bit if we really care. I will send a patch using two different sync (one for RX, one for TX)
