Re: [PATCH] erspan: Initialize options_len before referencing options.

[Simon Horman]

On Fri, Dec 12, 2025 at 07:32:01AM +0000, Frode Nordahl wrote:
> The struct ip_tunnel_info has a flexible array member named
> options that is protected by a counted_by(options_len)
> attribute.
> 
> The compiler will use this information to enforce runtime bounds
> checking deployed by FORTIFY_SOURCE string helpers.
> 
> As laid out in the GCC documentation, the counter must be
> initialized before the first reference to the flexible array
> member.
> 
> In the normal case the ip_tunnel_info_opts_set() helper is used
> which would initialize options_len properly, however in the GRE
> ERSPAN code a partial update is done, preventing the use of the
> helper function.
> 
> Before this change the handling of ERSPAN traffic in GRE tunnels
> would cause a kernel panic when the kernel is compiled with
> GCC 15+ and having FORTIFY_SOURCE configured:
> 
> memcpy: detected buffer overflow: 4 byte write of buffer size 0
> 
> Call Trace:
>  <IRQ>
>  __fortify_panic+0xd/0xf
>  erspan_rcv.cold+0x68/0x83
>  ? ip_route_input_slow+0x816/0x9d0
>  gre_rcv+0x1b2/0x1c0
>  gre_rcv+0x8e/0x100
>  ? raw_v4_input+0x2a0/0x2b0
>  ip_protocol_deliver_rcu+0x1ea/0x210
>  ip_local_deliver_finish+0x86/0x110
>  ip_local_deliver+0x65/0x110
>  ? ip_rcv_finish_core+0xd6/0x360
>  ip_rcv+0x186/0x1a0
> 
> Link: https://gcc.gnu.org/onlinedocs/gcc/Common-Variable-Attributes.html#index-counted_005fby-variable-attribute
> Reported-at: https://launchpad.net/bugs/2129580
> Fixes: bb5e62f2d547 ("net: Add options as a flexible array to struct ip_tunnel_info")
> Signed-off-by: Frode Nordahl <fnordahl@xxxxxxxxxx>

Hi Frode,

Thanks for your patch (and nice to see you recently in Prague :).

Overall this looks good to me but I have some minor feedback.


Firstly, the cited patch seems to cover more than erspan.
So I'm wondering if you took at look at other cases where
this might occur? No problem either way, but if so it might
be worth mentioning in the commit message.


Regarding the comments in the code. I am wondering if the are necessary
as the information is also contained in the commit message. And if the
source documented every such case then things could get rather verbose.

If you do feel strongly about it keeping it then could I ask that
(other than the URL) it is line-wrapped trimmed to 80 columns wide or less,
as is still preferred for Networking (but confusingly not all Kernel) code.


As a fix for code present in net this should be targeted at that tree.
It's best to do so explicitly like this:

Subject: [PATCH net] ...

And it's probably also best to CC stable@xxxxxxxxxxxxxxx.
That practice isn't as widespread as perhaps it should be for Networking code.
But it does seem worth mentioning.

...