Re: [PATCH net-next v1 04/15] psp: add support for socket-less psp associations

From
Todd Bert Freed <hidden>
Date
2026-11-11 18:24:09
Also in
bpf, linux-doc, linux-kselftest, linux-rdma, lkml
Reply to
Daniel Zahka, 2026-11-11
I'm not sure the kernel actually needs any bookkeeping at all for these
generic rx spi allocations. You could probably just pass the spi, dev
id, key, and probably rotation count (as a device key generation of
sorts) to the user without creating a kernel object. For socket
psp_assocs, the rx state is consulted on the datapath before queueing to
the socket. For tx state in all psp_assocs, there is the possibility
that an SADB handle needs to be cleaned up. Neither of those apply here
or in the rest of the series as I understand it.

I figure the tc ingress rule can just take the spi and dev id by value
rather than needing a reference to an object that we create here.
That is how it works in this series. The tc psp decap rule (ingress) does not
make reference to an rx assoc ; instead, it should have flower matches on
psp_spi, psp_ver, and psp_dev. That effectively ties it to packets from just
one rx assoc.

So why keep track of rx assoc objects at all? Two considerations.

1/ don't accept packets that correspond to a stale association. That is, even
if the device decrypted the packet, it should be dropped if the device key
which minted the assoc's SPI is two rotations behind. For psp socket offload
that is enforced in __psp_sk_rx_policy_check() with the pas->generation
check.

The spec permits SPIs to be reused immediately after two rotations. So the
device could decrypt a packet with a SPI that is more than two rotations old
in that case, or if a previous packet is being replayed, for example.

2/ hardware offload. the solution to #1 should be able to be fully offloaded
to the device. psp tunnel mode will be hw offload only, and is implemented
in the eswitch with connect-x.

We could add a psp_gen flower match to replicate the pas->generation
check, but the device has no access to the pas and can't offload that.

Keeping track of the rx assoc means that the kernel can take action on the
"this assoc is now stale" event. Possibly by enumerating the tc rules and
either deleting them, or marking them stale for userspace to reap.

-Todd
Thread (38 messages) 38 messages, 4 authors, last reply 1d ago
Keyboard shortcuts
hback out one level
jnext message in thread / next series-diff position
kprevious message in thread / previous series-diff position
ldrill in
Escclose help / fold thread tree
?toggle this help