Messages in this thread | | | From | Alexander Duyck <> | Date | Sun, 21 Nov 2021 13:16:33 -0800 | Subject | Re: [igb] netconsole triggers warning in netpoll_poll_dev |
| |
On Sun, Nov 21, 2021 at 3:44 AM Danielle Ratson <danieller@nvidia.com> wrote: > > > > > > drivers/net/ethernet/intel/igb/igb_main.c | 12 ++++++++---- > > > > > 1 file changed, 8 insertions(+), 4 deletions(-) > > > > > > > > > > diff --git a/drivers/net/ethernet/intel/igb/igb_main.c > > > > > b/drivers/net/ethernet/intel/igb/igb_main.c > > > > > index 0cd37ad81b4e..b0a9bed14071 100644 > > > > > --- a/drivers/net/ethernet/intel/igb/igb_main.c > > > > > +++ b/drivers/net/ethernet/intel/igb/igb_main.c > > > > > @@ -7991,12 +7991,16 @@ static void igb_ring_irq_enable(struct > > > > igb_q_vector *q_vector) > > > > > **/ > > > > > static int igb_poll(struct napi_struct *napi, int budget) { > > > > > - struct igb_q_vector *q_vector = container_of(napi, > > > > > - struct igb_q_vector, > > > > > - napi); > > > > > - bool clean_complete = true; > > > > > + struct igb_q_vector *q_vector; > > > > > + bool clean_complete; > > > > > int work_done = 0; > > > > > > > > > > + /* if budget is zero, we have a special case for netconsole, so > > > > > + * make sure to set clean_complete to false in that case. > > > > > + */ > > > > > + clean_complete = !!budget; > > > > > + > > > > > + q_vector = container_of(napi, struct igb_q_vector, napi); > > > > > #ifdef CONFIG_IGB_DCA > > > > > if (q_vector->adapter->flags & IGB_FLAG_DCA_ENABLED) > > > > > igb_update_dca(q_vector); > > > > > > > > I'm not a big fan of moving the q_vector init as a part of this > > > > patch since it just means more backport work. > > > > > > > > That said the change itself should be harmless so I am good with it > > > > either way. > > > > > > > > Reviewed-by: Alexander Duyck <alexanderduyck@fb.com> > > > > > > Hi, > > > > > > I have lately added the netconsole module, and since then we see the > > same warning constantly in the logs. > > > I have tried to apply Jesse's patch but it didn't seem to solve the issue. > > > > > > Did anyone managed to solve the issue and can share with us? > > > > > > Thanks, > > > Danielle > > > > The one issue I can see is that it basically leaves the igb_poll call stuck in > > polling mode. > > > > The easiest fix for all of this in the in-kernel driver is to just get rid of the > > "min" at the end and instead just "return work_done;". The extra > > complication is only needed if you were to be polling multiple queues and > > that isn't the case here so we should simplify it and get rid of the buggy > > "budget - 1" return value. > > Thank you very much for your reply Alexander! > It seems to work well! > > Are you planning to send it upstream?
No, I was just suggesting it as a possible solution. Feel free to put together your own patch and submit it if it is working for you.
- Alex
| |