netdev - Re: [PATCH v2 net-next 01/14] rtnetlink: prepare nla_put

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date: Sat, 24 Feb 2024 12:08:56 +0100
From: Eric Dumazet <edumazet@...gle.com>
To: Donald Hunter <donald.hunter@...il.com>
Cc: "David S . Miller" <davem@...emloft.net>, Jakub Kicinski <kuba@...nel.org>, 
	Paolo Abeni <pabeni@...hat.com>, netdev@...r.kernel.org, 
	Ido Schimmel <idosch@...dia.com>, Jiri Pirko <jiri@...dia.com>, eric.dumazet@...il.com
Subject: Re: [PATCH v2 net-next 01/14] rtnetlink: prepare nla_put_iflink() to
 run under RCU

On Sat, Feb 24, 2024 at 11:46 AM Donald Hunter <donald.hunter@...il.com> wrote:
>
> On Sat, 24 Feb 2024 at 08:21, Eric Dumazet <edumazet@...gle.com> wrote:
> >
> > On Fri, Feb 23, 2024 at 4:25 PM Donald Hunter <donald.hunter@...il.com> wrote:
> > >
> > > I notice that several of the *_get_iflink() implementations are wrapped
> > > with rcu_read_lock()/unlock() and many are not. Shouldn't this be done
> > > consistently for all?
> >
> > I do not understand the question, could you give one example of what
> > you saw so that I can comment ?
>
> I did include a snippet of your patch showing ipoib_get_iflink() which
> does not use rcu_read_lock() / unlock() and vxcan_get_iflink() which
> does. Sorry if that wasn't clear. My concern is that I'd expect all
> implementers of .ndo_get_iflink to need to be consistent, whether that
> is with or without the calls. Does it just mean that individual
> drivers are being overly cautious, or are protecting internal usage?
>
> No use of rcu_read_lock() / unlock() here:
>
> > index 7a5be705d71830d5bb3aa26a96a4463df03883a4..6f2a688fccbfb02ae7bdf3d55cca0e77fa9b56b4 100644
> > --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c
> > +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c
> > @@ -1272,10 +1272,10 @@ static int ipoib_get_iflink(const struct net_device *dev)
> >
> >       /* parent interface */
> >       if (!test_bit(IPOIB_FLAG_SUBINTERFACE, &priv->flags))
> > -             return dev->ifindex;
> > +             return READ_ONCE(dev->ifindex);

No need here, because dev is guaranteed to be alive during this call.

> >
> >       /* child/vlan interface */
> > -     return priv->parent->ifindex;
> > +     return READ_ONCE(priv->parent->ifindex);

Sure, no need for rcu_read_lock() here because priv->parent is stable
(can not change during lifetime)


> >  }
>
> And use of them here:
>
> > diff --git a/drivers/net/can/vxcan.c b/drivers/net/can/vxcan.c
> > index 98c669ad5141479b509ee924ddba3da6bca554cd..f7fabba707ea640cab8863e63bb19294e333ba2c 100644
> > --- a/drivers/net/can/vxcan.c
> > +++ b/drivers/net/can/vxcan.c
> > @@ -119,7 +119,7 @@ static int vxcan_get_iflink(const struct net_device *dev)
> >
> >       rcu_read_lock();
> >       peer = rcu_dereference(priv->peer);
> > -     iflink = peer ? peer->ifindex : 0;
> > +     iflink = peer ? READ_ONCE(peer->ifindex) : 0;
> >       rcu_read_unlock();
> >
> >       return iflink;
>
>
> > We do not need an rcu_read_lock() only to fetch dev->ifindex, if this
> > is what concerns you.
>
> In which case, it seems that no .ndo_get_iflink implementations should
> need the rcu_read_* calls?

rcu_read_lock() is needed in all cases a dereference is performed,
expecting RCU protection of the pointer.

In vxcan_get_iflink(), we access priv->peer, then peer->ifindex.

rcu_read_lock() is needed because of the second dereference, peer->ifindex.

Without rcu_read_lock(), peer could be freed before we get a chance to
read peer->ifindex.