Re: Follow-up to "net: drop bad gso csum_start and offset in virtio_net_hdr" - backport for 5.15 needed

[Willem de Bruijn <willemdebruijn.kernel@gmail.com>]

Greg KH wrote:
> On Mon, Sep 09, 2024 at 12:05:17AM -0400, Willem de Bruijn wrote:
> > On Tue, Sep 3, 2024 at 4:03 AM Greg KH <gregkh@linuxfoundation.org> wrote:
> > >
> > > On Tue, Sep 03, 2024 at 09:37:30AM +0200, Christian Theune wrote:
> > > > Hi,
> > > >
> > > > the issue was so far handled in https://lore.kernel.org/regressions/ZsyMzW-4ee_U8NoX@eldamar.lan/T/#m390d6ef7b733149949fb329ae1abffec5cefb99b and https://bugzilla.kernel.org/show_bug.cgi?id=219129
> > > >
> > > > I haven’t seen any communication whether a backport for 5.15 is already in progress, so I thought I’d follow up here.
> > >
> > > Someone needs to send a working set of patches to apply.
> > 
> > The following stack of patches applies cleanly to 5.15.166
> > (original SHA1s, git log order, so inverse of order to apply):
> > 
> > 89add40066f9e net: drop bad gso csum_start and offset in virtio_net_hdr
> > 9840036786d9 gso: fix dodgy bit handling for GSO_UDP_L4
> > fc8b2a619469 net: more strict VIRTIO_NET_HDR_GSO_UDP_L4 validation
> > 
> > All three are already present in 6.1.109
> > 
> > Please let me know if I should send that stack using git send-email,
> > or whether this is sufficient into to backport.
> 
> I just tried it, they do not apply cleanly here for me at all :(
> 
> > The third commit has one Fixes referencing them:
> > 
> > 1382e3b6a350 net: change maximum number of UDP segments to 128
> > 
> > This simple -2/+2 line patch unfortunately cannot be backported
> > without conflicts without backporting non-stable feature changes.
> > There is a backport to 6.1.y, but that also won't apply cleanly to
> > 5.15.166 without backporting a feature (e2a4392b61f6 "udp: introduce
> > udp->udp_flags"), which itself does not apply cleanly.
> > 
> > So simplest is probably to fix up this commit and send it using git
> > send-email. I can do that as part of the stack with the above 3
> > patches, or stand-alone if the above can be cherry-picked by SHA1.
> 
> Please send me a set of working, and tested, patches and we will be glad
> to consider it.

Done:

https://lore.kernel.org/stable/20240909182506.270136-1-willemdebruijn.kernel@gmail.com/T/#m0086f42d20bfc4d6226a8cf379590032edfbbe21

The following worked fine for me. I hope I did not overlook anything.
Compile tested only. But as said, the same patches have landed in
6.1-stable.

	git fetch linux-stable linux-5.15.y
	git checkout linux-stable/linux-5.15.y
	git cherry-pick fc8b2a619469
	git cherry-pick 9840036786d9
	git cherry-pick 89add40066f9e
	make defconfig
	make -j 64

Unfortunately, the key commit itself has a bug report against it. I am
working on that right now.

But as the existing patch that it refers to in its Fixes has landed in
all stable kernels and is causing reports, it is a backporting case of
damned if we do, damned if we don't.

I intend to have a fix ready ASAP for netdev-net/main. If all stable
branches have all backports, it should apply cleanly everywhere. Just
not ready to be included in this series from the start.