netdev - Re: [PATCH v2 net-next] arp: flush arp cache on IFF

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]

Message-ID: <1369162454.2615.60.camel@bwh-desktop.uk.solarflarecom.com>
Date:	Tue, 21 May 2013 19:54:14 +0100
From:	Ben Hutchings <bhutchings@...arflare.com>
To:	Timo Teras <timo.teras@....fi>
CC:	David Miller <davem@...emloft.net>, <netdev@...r.kernel.org>,
	<kaber@...sh.net>
Subject: Re: [PATCH v2 net-next] arp: flush arp cache on IFF_NOARP change

On Tue, 2013-05-21 at 21:29 +0300, Timo Teras wrote:
> On Tue, 21 May 2013 17:25:19 +0100
> Ben Hutchings <bhutchings@...arflare.com> wrote:
> 
> > On Tue, 2013-05-21 at 13:23 +0300, Timo Teräs wrote:
> > > IFF_NOARP affects what kind of neighbor entries are created
> > > (nud NOARP or nud INCOMPLETE). If the flag changes, flush the arp
> > > cache to refresh all entries.
> > > 
> > > Signed-off-by: Timo Teräs <timo.teras@....fi>
> > > ---
> > > > This patch makes no sense at all.
> > > > 
> > > > The state bit in ->priv_flags is a boolean stating whether the
> > > > notified should do something or not.
> > > > 
> > > > But you're setting it to match what IFF_NOARP is.
> > > >
> > > > You should set it any time IFF_NOARP _changes_, and then clear
> > > > the bit when the notifier clears the neighbour entries.
> > > 
> > > IFF_NOARP_CHANGED is set according to "changes = dev->flags ^
> > > old_flags;" which reflect the change. But I agree that the clearing
> > > out bit was misplaced. This is especially true as it seems
> > > NETDEV_CHANGE can be notified from another place too.
> > [...]
> > 
> > None of the other persistent flags have a transient flag for changes;
> > why should IFF_NOARP?  Why not cache the IFF_NOARP state in the
> > in_device and then compare and update that in arp_netdev_event().
> 
> This was the earlier suggestion I got, so I followed that. NOARP flag
> is also used in IPv6 side (quick look would indicate that ndisc code
> needs similar fix), so it would be useful to have this generally
> available.

Sorry, I missed that IFF_NOARP is not really just about ARP.

> Would it be worthwhile to consider adding "flags_changed" to struct
> net_device or some other mechanism add this info about all flags?

It's inelegant to put transient data associated with an event in a
persistent data structure.  On the other hand, having every user cache
the old state is pretty awful as well.

Really, netdev notifiers should be changed to accept a structure that
encapsulates the changes rather than just a pointer to the net_device.
But making such a change would be an enormous pain and error-prone
because notifier functions aren't type-safe.

As an interim solution, I think either the general priv_flags_changed or
old_priv_flags would be preferable to defining extra transient flags.

Ben.

-- 
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html