2022-03-31 03:50:47

by Leon Romanovsky

[permalink] [raw]
Subject: [PATCH net] ixgbe: ensure IPsec VF<->PF compatibility

From: Leon Romanovsky <[email protected]>

The VF driver can forward any IPsec flags and such makes the function
is not extendable and prone to backward/forward incompatibility.

If new software runs on VF, it won't know that PF configured something
completely different as it "knows" only XFRM_OFFLOAD_INBOUND flag.

Fixes: eda0333ac293 ("ixgbe: add VF IPsec management")
Reviewed-by: Raed Salem <[email protected]>
Signed-off-by: Leon Romanovsky <[email protected]>
---
There is no simple fix for this VF/PF incompatibility as long as FW
doesn't filter/decline unsupported options when convey mailbox from VF
to PF.
---
drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
index e596e1a9fc75..236f244e3f65 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
@@ -903,7 +903,9 @@ int ixgbe_ipsec_vf_add_sa(struct ixgbe_adapter *adapter, u32 *msgbuf, u32 vf)
/* Tx IPsec offload doesn't seem to work on this
* device, so block these requests for now.
*/
- if (!(sam->flags & XFRM_OFFLOAD_INBOUND)) {
+ sam->flags = sam->flags & ~XFRM_OFFLOAD_IPV6;
+ if (!(sam->flags & XFRM_OFFLOAD_INBOUND) ||
+ sam->flags & ~XFRM_OFFLOAD_INBOUND) {
err = -EOPNOTSUPP;
goto err_out;
}
--
2.35.1


2022-03-31 04:54:06

by Shannon Nelson

[permalink] [raw]
Subject: Re: [PATCH net] ixgbe: ensure IPsec VF<->PF compatibility

On 3/30/22 4:01 AM, Leon Romanovsky wrote:
> From: Leon Romanovsky <[email protected]>
>
> The VF driver can forward any IPsec flags and such makes the function
> is not extendable and prone to backward/forward incompatibility.
>
> If new software runs on VF, it won't know that PF configured something
> completely different as it "knows" only XFRM_OFFLOAD_INBOUND flag.
>
> Fixes: eda0333ac293 ("ixgbe: add VF IPsec management")
> Reviewed-by: Raed Salem <[email protected]>
> Signed-off-by: Leon Romanovsky <[email protected]>
> ---
> There is no simple fix for this VF/PF incompatibility as long as FW
> doesn't filter/decline unsupported options when convey mailbox from VF
> to PF.
> ---
> drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
> index e596e1a9fc75..236f244e3f65 100644
> --- a/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
> +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
> @@ -903,7 +903,9 @@ int ixgbe_ipsec_vf_add_sa(struct ixgbe_adapter *adapter, u32 *msgbuf, u32 vf)
> /* Tx IPsec offload doesn't seem to work on this
> * device, so block these requests for now.
> */
> - if (!(sam->flags & XFRM_OFFLOAD_INBOUND)) {
> + sam->flags = sam->flags & ~XFRM_OFFLOAD_IPV6;
> + if (!(sam->flags & XFRM_OFFLOAD_INBOUND) ||
> + sam->flags & ~XFRM_OFFLOAD_INBOUND) {

So after stripping the IPV6 flag, you're checking to be sure that
INBOUND is the only flag enabled, right?
Could you use
    if (sam->flags != XFRM_OFFLOAD_INBOUND) {
instead?

sln

> err = -EOPNOTSUPP;
> goto err_out;
> }

2022-03-31 08:57:40

by Leon Romanovsky

[permalink] [raw]
Subject: Re: [PATCH net] ixgbe: ensure IPsec VF<->PF compatibility

On Wed, Mar 30, 2022 at 09:13:21AM -0700, Shannon Nelson wrote:
> On 3/30/22 4:01 AM, Leon Romanovsky wrote:
> > From: Leon Romanovsky <[email protected]>
> >
> > The VF driver can forward any IPsec flags and such makes the function
> > is not extendable and prone to backward/forward incompatibility.
> >
> > If new software runs on VF, it won't know that PF configured something
> > completely different as it "knows" only XFRM_OFFLOAD_INBOUND flag.
> >
> > Fixes: eda0333ac293 ("ixgbe: add VF IPsec management")
> > Reviewed-by: Raed Salem <[email protected]>
> > Signed-off-by: Leon Romanovsky <[email protected]>
> > ---
> > There is no simple fix for this VF/PF incompatibility as long as FW
> > doesn't filter/decline unsupported options when convey mailbox from VF
> > to PF.
> > ---
> > drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c | 4 +++-
> > 1 file changed, 3 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
> > index e596e1a9fc75..236f244e3f65 100644
> > --- a/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
> > +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
> > @@ -903,7 +903,9 @@ int ixgbe_ipsec_vf_add_sa(struct ixgbe_adapter *adapter, u32 *msgbuf, u32 vf)
> > /* Tx IPsec offload doesn't seem to work on this
> > * device, so block these requests for now.
> > */
> > - if (!(sam->flags & XFRM_OFFLOAD_INBOUND)) {
> > + sam->flags = sam->flags & ~XFRM_OFFLOAD_IPV6;
> > + if (!(sam->flags & XFRM_OFFLOAD_INBOUND) ||
> > + sam->flags & ~XFRM_OFFLOAD_INBOUND) {
>
> So after stripping the IPV6 flag, you're checking to be sure that INBOUND is
> the only flag enabled, right?
> Could you use
> ??? if (sam->flags != XFRM_OFFLOAD_INBOUND) {
> instead?

Sure, I'll send new version soon.

Thanks

>
> sln
>
> > err = -EOPNOTSUPP;
> > goto err_out;
> > }
>