[PATCH] net: tls, update curr on splice as well

[PATCH] net: tls, update curr on splice as well

[John Fastabend]

commit c5a595000e2677e865a39f249c056bc05d6e55fd upstream.

Backport of upstream fix for tls on 6.1 and lower kernels.
The curr pointer must also be updated on the splice similar to how
we do this for other copy types.

Cc: stable@vger.kernel.org # 6.1.x-
Reported-by: Jann Horn <jannh@google.com>
Fixes: d829e9c4112b ("tls: convert to generic sk_msg interface")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
---
 net/tls/tls_sw.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/net/tls/tls_sw.c b/net/tls/tls_sw.c
index 2e60bf06adff..0323040d34bc 100644
--- a/net/tls/tls_sw.c
+++ b/net/tls/tls_sw.c
@@ -1225,6 +1225,8 @@ static int tls_sw_do_sendpage(struct sock *sk, struct page *page,
 		}
 
 		sk_msg_page_add(msg_pl, page, copy, offset);
+		msg_pl->sg.copybreak = 0;
+		msg_pl->sg.curr = msg_pl->sg.end;
 		sk_mem_charge(sk, copy);
 
 		offset += copy;
-- 
2.33.0

[missing stable fix on 5.x] [PATCH] net: tls, update curr on splice as well

[Jann Horn]

On Thu, Dec 14, 2023 at 8:45 PM John Fastabend <john.fastabend@gmail.com> wrote:
> commit c5a595000e2677e865a39f249c056bc05d6e55fd upstream.
>
> Backport of upstream fix for tls on 6.1 and lower kernels.
> The curr pointer must also be updated on the splice similar to how
> we do this for other copy types.
>
> Cc: stable@vger.kernel.org # 6.1.x-

I think this Cc marker was wrong - the commit message says "on 6.1 and
lower kernels", but this marker seems to say "6.1 and *newer*
kernels". The current status is that this issue is fixed on 6.6.7 and
6.1.69, but not on the 5.x stable kernels.

> Reported-by: Jann Horn <jannh@google.com>
> Fixes: d829e9c4112b ("tls: convert to generic sk_msg interface")
> Signed-off-by: John Fastabend <john.fastabend@gmail.com>
> ---
>  net/tls/tls_sw.c | 2 ++
>  1 file changed, 2 insertions(+)
>
> diff --git a/net/tls/tls_sw.c b/net/tls/tls_sw.c
> index 2e60bf06adff..0323040d34bc 100644
> --- a/net/tls/tls_sw.c
> +++ b/net/tls/tls_sw.c
> @@ -1225,6 +1225,8 @@ static int tls_sw_do_sendpage(struct sock *sk, struct page *page,
>                 }
>
>                 sk_msg_page_add(msg_pl, page, copy, offset);
> +               msg_pl->sg.copybreak = 0;
> +               msg_pl->sg.curr = msg_pl->sg.end;
>                 sk_mem_charge(sk, copy);
>
>                 offset += copy;
> --
> 2.33.0
>

Re: [missing stable fix on 5.x] [PATCH] net: tls, update curr on splice as well

[Greg Kroah-Hartman]

On Mon, Jan 08, 2024 at 02:10:31PM +0100, Jann Horn wrote:
> On Thu, Dec 14, 2023 at 8:45 PM John Fastabend <john.fastabend@gmail.com> wrote:
> > commit c5a595000e2677e865a39f249c056bc05d6e55fd upstream.
> >
> > Backport of upstream fix for tls on 6.1 and lower kernels.
> > The curr pointer must also be updated on the splice similar to how
> > we do this for other copy types.
> >
> > Cc: stable@vger.kernel.org # 6.1.x-
> 
> I think this Cc marker was wrong - the commit message says "on 6.1 and
> lower kernels", but this marker seems to say "6.1 and *newer*
> kernels". The current status is that this issue is fixed on 6.6.7 and
> 6.1.69, but not on the 5.x stable kernels.

Then can someone provide a working backport to those kernels?  Right
now, this one does not apply at all there.

thanks,

greg k-h

Re: [missing stable fix on 5.x] [PATCH] net: tls, update curr on splice as well

[John Fastabend]

Greg Kroah-Hartman wrote:
> On Mon, Jan 08, 2024 at 02:10:31PM +0100, Jann Horn wrote:
> > On Thu, Dec 14, 2023 at 8:45 PM John Fastabend <john.fastabend@gmail.com> wrote:
> > > commit c5a595000e2677e865a39f249c056bc05d6e55fd upstream.
> > >
> > > Backport of upstream fix for tls on 6.1 and lower kernels.
> > > The curr pointer must also be updated on the splice similar to how
> > > we do this for other copy types.
> > >
> > > Cc: stable@vger.kernel.org # 6.1.x-
> > 
> > I think this Cc marker was wrong - the commit message says "on 6.1 and
> > lower kernels", but this marker seems to say "6.1 and *newer*
> > kernels". The current status is that this issue is fixed on 6.6.7 and
> > 6.1.69, but not on the 5.x stable kernels.
> 
> Then can someone provide a working backport to those kernels?  Right
> now, this one does not apply at all there.

I'll put something together now and run it through some tests then
send it. I'm not sure where I got the notation 6.1.x- to mean <=6.1,
but I don't see it in the stable-kernel-rules.rst at the moment
so maybe I made it up. Sorry about that.

> 
> thanks,
> 
> greg k-h