Re: [PATCH v2,net] hv_netvsc: Fix tx_table init in rndis_set_subchannel()

[Stephen Hemminger]

On Sun, 15 Dec 2019 16:38:00 +0000
Haiyang Zhang <haiyangz@xxxxxxxxxxxxx> wrote:

> > -----Original Message-----
> > From: Jakub Kicinski <jakub.kicinski@xxxxxxxxxxxxx>
> > Sent: Saturday, December 14, 2019 2:30 PM
> > To: Haiyang Zhang <haiyangz@xxxxxxxxxxxxx>
> > Cc: sashal@xxxxxxxxxx; linux-hyperv@xxxxxxxxxxxxxxx; netdev@xxxxxxxxxxxxxxx;
> > KY Srinivasan <kys@xxxxxxxxxxxxx>; Stephen Hemminger
> > <sthemmin@xxxxxxxxxxxxx>; olaf@xxxxxxxxx; vkuznets
> > <vkuznets@xxxxxxxxxx>; davem@xxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx
> > Subject: Re: [PATCH v2,net] hv_netvsc: Fix tx_table init in rndis_set_subchannel()
> > 
> > On Wed, 11 Dec 2019 14:26:27 -0800, Haiyang Zhang wrote:  
> > > Host can provide send indirection table messages anytime after RSS is
> > > enabled by calling rndis_filter_set_rss_param(). So the host provided
> > > table values may be overwritten by the initialization in
> > > rndis_set_subchannel().
> > >
> > > To prevent this problem, move the tx_table initialization before calling
> > > rndis_filter_set_rss_param().
> > >
> > > Fixes: a6fb6aa3cfa9 ("hv_netvsc: Set tx_table to equal weight after  
> > subchannels open")  
> > > Signed-off-by: Haiyang Zhang <haiyangz@xxxxxxxxxxxxx>  
> > 
> > Applied, but there are two more problems with this code:
> >  - you should not reset the indirection table if it was configured by
> >    the user to something other than the default (use the
> >    netif_is_rxfh_configured() helper to check for that)  
> 
> For Send indirection table (tx_table) ethtool doesn't have the option 
> to set it, and it's usually provided by the host. So we always initialize 
> it...
> But, yes, for Receive indirection table (rx_table), I will make a fix, so 
> it will be set to default only for new devices, or changing the number 
> of channels; otherwise it will remain the same during operations like 
> changing MTU, ringparam.
> 
> 
> >  - you should use the ethtool_rxfh_indir_default() wrapper  
> For rx_table, we already use it:
>                 rndis_device->rx_table[i] = ethtool_rxfh_indir_default(
> For tx_table, I know it's the same operation (%, mod), but this wrapper 
> function's name is for rx_table. Should we use it for tx_table too?
> 
> > 
> > Please fix the former problem in the net tree, and after net is merged
> > into linux/master and net-next in a week or two please follow up with
> > the fix for the latter for net-next.  
> 
> Sure.
> 
> Thanks,
> - Haiyang
> 
As Haiyang said, this send indirection table is unique to Hyper-V it is not part of
any of the other device models. It is not supported by ethtool. It would not be
appropriate to repurpose the existing indirection tool; the device already uses
the receive indirection table for RSS.