Re: [PATCH net] docs: net: fix documentation on .ndo_get_stats

[Vladimir Oltean <olteanv@gmail.com>]

On Mon, Jan 04, 2021 at 09:53:09AM -0800, Jakub Kicinski wrote:
> On Mon, 4 Jan 2021 12:42:27 +0200 Vladimir Oltean wrote:
> > On Wed, Dec 30, 2020 at 07:45:24PM -0800, Jakub Kicinski wrote:
> > > Fix calling context.
> > >
> > > Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> > > ---
> > >  Documentation/networking/netdevices.rst | 4 ++--
> > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/Documentation/networking/netdevices.rst b/Documentation/networking/netdevices.rst
> > > index 5a85fcc80c76..a80676f5477d 100644
> > > --- a/Documentation/networking/netdevices.rst
> > > +++ b/Documentation/networking/netdevices.rst
> > > @@ -64,8 +64,8 @@ struct net_device synchronization rules
> > >  	Context: process
> > >
> > >  ndo_get_stats:
> > > -	Synchronization: dev_base_lock rwlock.
> > > -	Context: nominally process, but don't sleep inside an rwlock
> > > +	Synchronization: rtnl_lock() semaphore, or RCU.
> > > +	Context: atomic
> > >
> > >  ndo_start_xmit:
> > >  	Synchronization: __netif_tx_lock spinlock.
> >
> > And what happened to dev_base_lock? Did it suddenly go away?
>
> I thought all callers switched to RCU. You investigated this in depth,
> did I miss something? I'm sending this correction because I have a
> series which adds to other sections of this file and this jumped out
> to me as incorrect.

Well, there's netstat_show from net/core/net-sysfs.c still. I couldn't
figure why that lock exists, it doesn't seem to protect something in
particular.