On Tue, 12 Dec 2023 at 20:52, Daniel Xu <dxu@xxxxxxxxx> wrote:
>
> cc Kumar
>
> On Tue, Dec 12, 2023 at 09:17:02AM -0700, Daniel Xu wrote:
> > On Mon, Dec 11, 2023 at 04:25:06PM -0800, Eyal Birger wrote:
> > > On Mon, Dec 11, 2023 at 3:49 PM Daniel Xu <dxu@xxxxxxxxx> wrote:
> > > >
> > > > On Mon, Dec 11, 2023 at 03:13:07PM -0800, Eyal Birger wrote:
> > > > > On Mon, Dec 11, 2023 at 2:31 PM Daniel Xu <dxu@xxxxxxxxx> wrote:
> > > > > >
> > > > > > On Mon, Dec 11, 2023 at 01:39:25PM -0800, Eyal Birger wrote:
> > > > > > > Hi Daniel,
> > > > > > >
> > > > > > > Tiny nits below in case you respin this for other reasons:
> > > > > > >
> > > > > > > On Mon, Dec 11, 2023 at 12:20 PM Daniel Xu <dxu@xxxxxxxxx> wrote:
> > > > > > > >
> > > > > > > > This commit extends test_tunnel selftest to test the new XDP xfrm state
> > > > > > > > lookup kfunc.
> > > > > > > >
> > > > > > > > Co-developed-by: Antony Antony <antony.antony@xxxxxxxxxxx>
> > > > > > > > Signed-off-by: Antony Antony <antony.antony@xxxxxxxxxxx>
> > > > > > > > Signed-off-by: Daniel Xu <dxu@xxxxxxxxx>
> > > > > > > > ---
> > > > > > > > .../selftests/bpf/prog_tests/test_tunnel.c | 20 ++++++--
> > > > > > > > .../selftests/bpf/progs/test_tunnel_kern.c | 51 +++++++++++++++++++
> > > > > > > > 2 files changed, 67 insertions(+), 4 deletions(-)
> > > > > > > >
> > > > > > > > diff --git a/tools/testing/selftests/bpf/prog_tests/test_tunnel.c b/tools/testing/selftests/bpf/prog_tests/test_tunnel.c
> > > > > > > > index 2d7f8fa82ebd..fc804095d578 100644
> > > > > > > > --- a/tools/testing/selftests/bpf/prog_tests/test_tunnel.c
> > > > > > > > +++ b/tools/testing/selftests/bpf/prog_tests/test_tunnel.c
> > > > > > > > @@ -278,7 +278,7 @@ static int add_xfrm_tunnel(void)
> > > > > > > > SYS(fail,
> > > > > > > > "ip netns exec at_ns0 "
> > > > > > > > "ip xfrm state add src %s dst %s proto esp "
> > > > > > > > - "spi %d reqid 1 mode tunnel "
> > > > > > > > + "spi %d reqid 1 mode tunnel replay-window 42 "
> > > > > > > > "auth-trunc 'hmac(sha1)' %s 96 enc 'cbc(aes)' %s",
> > > > > > > > IP4_ADDR_VETH0, IP4_ADDR1_VETH1, XFRM_SPI_IN_TO_OUT, XFRM_AUTH, XFRM_ENC);
> > > > > > > > SYS(fail,
> > > > > > > > @@ -292,7 +292,7 @@ static int add_xfrm_tunnel(void)
> > > > > > > > SYS(fail,
> > > > > > > > "ip netns exec at_ns0 "
> > > > > > > > "ip xfrm state add src %s dst %s proto esp "
> > > > > > > > - "spi %d reqid 2 mode tunnel "
> > > > > > > > + "spi %d reqid 2 mode tunnel replay-window 42 "
> > > > > > >
> > > > > > > nit: why do you need to set the replay-window in both directions?
> > > > > >
> > > > > > No reason - probably just careless here.
> > > > > >
> > > > > > >
> > > > > > > > "auth-trunc 'hmac(sha1)' %s 96 enc 'cbc(aes)' %s",
> > > > > > > > IP4_ADDR1_VETH1, IP4_ADDR_VETH0, XFRM_SPI_OUT_TO_IN, XFRM_AUTH, XFRM_ENC);
> > > > > > > > SYS(fail,
> > > > > > > > @@ -313,7 +313,7 @@ static int add_xfrm_tunnel(void)
> > > > > > > > */
> > > > > > > > SYS(fail,
> > > > > > > > "ip xfrm state add src %s dst %s proto esp "
> > > > > > > > - "spi %d reqid 1 mode tunnel "
> > > > > > > > + "spi %d reqid 1 mode tunnel replay-window 42 "
> > > > > > > > "auth-trunc 'hmac(sha1)' %s 96 enc 'cbc(aes)' %s",
> > > > > > > > IP4_ADDR_VETH0, IP4_ADDR1_VETH1, XFRM_SPI_IN_TO_OUT, XFRM_AUTH, XFRM_ENC);
> > > > > > > > SYS(fail,
> > > > > > > > @@ -325,7 +325,7 @@ static int add_xfrm_tunnel(void)
> > > > > > > > /* root -> at_ns0 */
> > > > > > > > SYS(fail,
> > > > > > > > "ip xfrm state add src %s dst %s proto esp "
> > > > > > > > - "spi %d reqid 2 mode tunnel "
> > > > > > > > + "spi %d reqid 2 mode tunnel replay-window 42 "
> > > > > > > > "auth-trunc 'hmac(sha1)' %s 96 enc 'cbc(aes)' %s",
> > > > > > > > IP4_ADDR1_VETH1, IP4_ADDR_VETH0, XFRM_SPI_OUT_TO_IN, XFRM_AUTH, XFRM_ENC);
> > > > > > > > SYS(fail,
> > > > > > > > @@ -628,8 +628,10 @@ static void test_xfrm_tunnel(void)
> > > > > > > > {
> > > > > > > > DECLARE_LIBBPF_OPTS(bpf_tc_hook, tc_hook,
> > > > > > > > .attach_point = BPF_TC_INGRESS);
> > > > > > > > + LIBBPF_OPTS(bpf_xdp_attach_opts, opts);
> > > > > > > > struct test_tunnel_kern *skel = NULL;
> > > > > > > > struct nstoken *nstoken;
> > > > > > > > + int xdp_prog_fd;
> > > > > > > > int tc_prog_fd;
> > > > > > > > int ifindex;
> > > > > > > > int err;
> > > > > > > > @@ -654,6 +656,14 @@ static void test_xfrm_tunnel(void)
> > > > > > > > if (attach_tc_prog(&tc_hook, tc_prog_fd, -1))
> > > > > > > > goto done;
> > > > > > > >
> > > > > > > > + /* attach xdp prog to tunnel dev */
> > > > > > > > + xdp_prog_fd = bpf_program__fd(skel->progs.xfrm_get_state_xdp);
> > > > > > > > + if (!ASSERT_GE(xdp_prog_fd, 0, "bpf_program__fd"))
> > > > > > > > + goto done;
> > > > > > > > + err = bpf_xdp_attach(ifindex, xdp_prog_fd, XDP_FLAGS_REPLACE, &opts);
> > > > > > > > + if (!ASSERT_OK(err, "bpf_xdp_attach"))
> > > > > > > > + goto done;
> > > > > > > > +
> > > > > > > > /* ping from at_ns0 namespace test */
> > > > > > > > nstoken = open_netns("at_ns0");
> > > > > > > > err = test_ping(AF_INET, IP4_ADDR_TUNL_DEV1);
> > > > > > > > @@ -667,6 +677,8 @@ static void test_xfrm_tunnel(void)
> > > > > > > > goto done;
> > > > > > > > if (!ASSERT_EQ(skel->bss->xfrm_remote_ip, 0xac100164, "remote_ip"))
> > > > > > > > goto done;
> > > > > > > > + if (!ASSERT_EQ(skel->bss->xfrm_replay_window, 42, "replay_window"))
> > > > > > > > + goto done;
> > > > > > > >
> > > > > > > > done:
> > > > > > > > delete_xfrm_tunnel();
> > > > > > > > diff --git a/tools/testing/selftests/bpf/progs/test_tunnel_kern.c b/tools/testing/selftests/bpf/progs/test_tunnel_kern.c
> > > > > > > > index 3a59eb9c34de..c0dd38616562 100644
> > > > > > > > --- a/tools/testing/selftests/bpf/progs/test_tunnel_kern.c
> > > > > > > > +++ b/tools/testing/selftests/bpf/progs/test_tunnel_kern.c
> > > > > > > > @@ -30,6 +30,10 @@ int bpf_skb_set_fou_encap(struct __sk_buff *skb_ctx,
> > > > > > > > struct bpf_fou_encap *encap, int type) __ksym;
> > > > > > > > int bpf_skb_get_fou_encap(struct __sk_buff *skb_ctx,
> > > > > > > > struct bpf_fou_encap *encap) __ksym;
> > > > > > > > +struct xfrm_state *
> > > > > > > > +bpf_xdp_get_xfrm_state(struct xdp_md *ctx, struct bpf_xfrm_state_opts *opts,
> > > > > > > > + u32 opts__sz) __ksym;
> > > > > > > > +void bpf_xdp_xfrm_state_release(struct xfrm_state *x) __ksym;
> > > > > > > >
> > > > > > > > struct {
> > > > > > > > __uint(type, BPF_MAP_TYPE_ARRAY);
> > > > > > > > @@ -950,4 +954,51 @@ int xfrm_get_state(struct __sk_buff *skb)
> > > > > > > > return TC_ACT_OK;
> > > > > > > > }
> > > > > > > >
> > > > > > > > +volatile int xfrm_replay_window = 0;
> > > > > > > > +
> > > > > > > > +SEC("xdp")
> > > > > > > > +int xfrm_get_state_xdp(struct xdp_md *xdp)
> > > > > > > > +{
> > > > > > > > + struct bpf_xfrm_state_opts opts = {};
> > > > > > > > + struct xfrm_state *x = NULL;
> > > > > > > > + struct ip_esp_hdr *esph;
> > > > > > > > + struct bpf_dynptr ptr;
> > > > > > > > + u8 esph_buf[8] = {};
> > > > > > > > + u8 iph_buf[20] = {};
> > > > > > > > + struct iphdr *iph;
> > > > > > > > + u32 off;
> > > > > > > > +
> > > > > > > > + if (bpf_dynptr_from_xdp(xdp, 0, &ptr))
> > > > > > > > + goto out;
> > > > > > > > +
> > > > > > > > + off = sizeof(struct ethhdr);
> > > > > > > > + iph = bpf_dynptr_slice(&ptr, off, iph_buf, sizeof(iph_buf));
> > > > > > > > + if (!iph || iph->protocol != IPPROTO_ESP)
> > > > > > > > + goto out;
> > > > > > > > +
> > > > > > > > + off += sizeof(struct iphdr);
> > > > > > > > + esph = bpf_dynptr_slice(&ptr, off, esph_buf, sizeof(esph_buf));
> > > > > > > > + if (!esph)
> > > > > > > > + goto out;
> > > > > > > > +
> > > > > > > > + opts.netns_id = BPF_F_CURRENT_NETNS;
> > > > > > > > + opts.daddr.a4 = iph->daddr;
> > > > > > > > + opts.spi = esph->spi;
> > > > > > > > + opts.proto = IPPROTO_ESP;
> > > > > > > > + opts.family = AF_INET;
> > > > > > > > +
> > > > > > > > + x = bpf_xdp_get_xfrm_state(xdp, &opts, sizeof(opts));
> > > > > > > > + if (!x || opts.error)
> > > > > > >
> > > > > > > nit: how can opts.error be non zero if x == NULL?
> > > > > >
> > > > > > Ignoring the new -ENOENT case, it can't. Which is why I'm testing that
> > > > > > behavior here.
> > > > >
> > > > > I'm sorry, I don't understand.
> > > > >
> > > > > AFAICT, regardless of the -ENOENT change, I don't see
> > > > > how (!x) is false and (opt.error) is true, and so
> > > > > "if (!x || opts.error)" is always equivalent to "if (!x)".
> > > > >
> > > > > What am I missing?
> > > > > Eyal.
> > > >
> > > > The selftests are tests so my intention was to check edge cases here.
> > > > In normal operation it shouldn't be possible that
> > > > bpf_xdp_get_xfrm_state() returns non-NULL and also an error. Maybe
> > > > another way of writing this would be:
> > > >
> > > > if (!x)
> > > > goto out;
> > > > assert(opts.error == 0);
> > >
> > > I think this would convey the "edge case testing" notion better.
> > >
> > > >
> > > > If I'm trying to be too clever (or maybe just wrong) or it's pointless,
> > > > I can remove the `opts.error` condition.
> > >
> > > At least for me the tests also serve as references as to how the
> > > API is expected to be used, so I think it'd be clearer without
> > > signaling that opts.error could potentially be nonzero on success.
> > >
> > > An assertion would indeed make that clear.
> >
> > Sure, sounds good. I will check on the new bpf assert infra.
>
> Couldn't quite get bpf_assert() working. The following diff:
>
> diff --git a/tools/testing/selftests/bpf/progs/test_tunnel_kern.c b/tools/testing/selftests/bpf/progs/test_tunnel_kern.c
> index c0dd38616562..f00dba85ac5d 100644
> --- a/tools/testing/selftests/bpf/progs/test_tunnel_kern.c
> +++ b/tools/testing/selftests/bpf/progs/test_tunnel_kern.c
> @@ -8,8 +8,9 @@
> */
> #include "vmlinux.h"
> #include <bpf/bpf_core_read.h>
> -#include <bpf/bpf_helpers.h>
> #include <bpf/bpf_endian.h>
> +#include <bpf/bpf_helpers.h>
> +#include "bpf_experimental.h"
> #include "bpf_kfuncs.h"
> #include "bpf_tracing_net.h"
>
> @@ -988,8 +989,9 @@ int xfrm_get_state_xdp(struct xdp_md *xdp)
> opts.family = AF_INET;
>
> x = bpf_xdp_get_xfrm_state(xdp, &opts, sizeof(opts));
> - if (!x || opts.error)
> + if (!x)
> goto out;
> + bpf_assert_with(opts.error == 0, XDP_PASS);
>
> if (!x->replay_esn)
> goto out;
>
> results in:
>
> 57: (b7) r1 = 2 ; R1_w=2 refs=5
> 58: (85) call bpf_throw#115436
> calling kernel function bpf_throw is not allowed
>
I think this might be because bpf_throw is not registered for use by
BPF_PROG_TYPE_XDP. I would simply register the generic_kfunc_set for
this program type as well, since it's already done for TC.
> It looks like the above error comes from verifier.c:fetch_kfunc_meta,
> but I can run the exceptions selftests just fine with the same bzImage.
> So I'm thinking it's not a kfunc registration or BTF issue.
>
> Maybe it's cuz I'm holding onto KFUNC_ACQUIRE'd `x`? Not sure.
>
Yes, even once you enable this, this will fail for now. I am sending
out a series later this week that enables bpf_throw with acquired
references, but until then may I suggest the following:
#define bpf_assert_if(cond) for (int ___i = 0, ___j = (cond); !(___j) \
&& !___j; bpf_throw(), ___i++)
This will allow you to insert some cleanup code with an assertion.
Then in my series, I will convert this temporary bpf_assert_if back to
the normal bpf_assert.
It would look like:
bpf_assert_if(opts.error == 0) {
// Execute if assertion failed
bpf_xdp_xfrm_state_release(x);
}
Likewise for bpf_assert_with_if, you get the idea.
> So for now I think I'll drop checking opts.error.
>
> [...]
