On Thu, Jul 02, 2020 at 12:27 PM CEST, Lorenz Bauer wrote:
> On Thu, 2 Jul 2020 at 10:24, Jakub Sitnicki <jakub@xxxxxxxxxxxxxx> wrote:
>>
>> Run a BPF program before looking up a listening socket on the receive path.
>> Program selects a listening socket to yield as result of socket lookup by
>> calling bpf_sk_assign() helper and returning BPF_REDIRECT (7) code.
>>
>> Alternatively, program can also fail the lookup by returning with
>> BPF_DROP (1), or let the lookup continue as usual with BPF_OK (0) on
>> return. Other return values are treated the same as BPF_OK.
>
> I'd prefer if other values were treated as BPF_DROP, with other semantics
> unchanged. Otherwise we won't be able to introduce new semantics
> without potentially breaking user code.
That might be surprising or even risky. If you attach a badly written
program that say returns a negative value, it will drop all TCP SYNs and
UDP traffic.
>
>>
>> This lets the user match packets with listening sockets freely at the last
>> possible point on the receive path, where we know that packets are destined
>> for local delivery after undergoing policing, filtering, and routing.
>>
>> With BPF code selecting the socket, directing packets destined to an IP
>> range or to a port range to a single socket becomes possible.
>>
>> In case multiple programs are attached, they are run in series in the order
>> in which they were attached. The end result gets determined from return
>> code from each program according to following rules.
>>
>> 1. If any program returned BPF_REDIRECT and selected a valid socket, this
>> socket will be used as result of the lookup.
>> 2. If more than one program returned BPF_REDIRECT and selected a socket,
>> last selection takes effect.
>> 3. If any program returned BPF_DROP and none returned BPF_REDIRECT, the
>> socket lookup will fail with -ECONNREFUSED.
>> 4. If no program returned neither BPF_DROP nor BPF_REDIRECT, socket lookup
>> continues to htable-based lookup.
>>
>> Suggested-by: Marek Majkowski <marek@xxxxxxxxxxxxxx>
>> Signed-off-by: Jakub Sitnicki <jakub@xxxxxxxxxxxxxx>
>> ---
>>
>> Notes:
>> v3:
>> - Use a static_key to minimize the hook overhead when not used. (Alexei)
>> - Adapt for running an array of attached programs. (Alexei)
>> - Adapt for optionally skipping reuseport selection. (Martin)
>>
>> include/linux/bpf.h | 29 ++++++++++++++++++++++++++++
>> include/linux/filter.h | 39 ++++++++++++++++++++++++++++++++++++++
>> kernel/bpf/net_namespace.c | 32 ++++++++++++++++++++++++++++++-
>> net/core/filter.c | 2 ++
>> net/ipv4/inet_hashtables.c | 31 ++++++++++++++++++++++++++++++
>> 5 files changed, 132 insertions(+), 1 deletion(-)
>>
[...]
>> diff --git a/kernel/bpf/net_namespace.c b/kernel/bpf/net_namespace.c
>> index 090166824ca4..a7768feb3ade 100644
>> --- a/kernel/bpf/net_namespace.c
>> +++ b/kernel/bpf/net_namespace.c
>> @@ -25,6 +25,28 @@ struct bpf_netns_link {
>> /* Protects updates to netns_bpf */
>> DEFINE_MUTEX(netns_bpf_mutex);
>>
>> +static void netns_bpf_attach_type_disable(enum netns_bpf_attach_type type)
>
> Nit: maybe netns_bpf_attach_type_dec()? Disable sounds like it happens
> unconditionally.
attach_type_dec()/_inc() seems a bit cryptic, since it's not the attach
type we are incrementing/decrementing.
But I was considering _need()/_unneed(), which would follow an existing
example, if you think that improves things.
>
>> +{
>> + switch (type) {
>> + case NETNS_BPF_SK_LOOKUP:
>> + static_branch_dec(&bpf_sk_lookup_enabled);
>> + break;
>> + default:
>> + break;
>> + }
>> +}
>> +
>> +static void netns_bpf_attach_type_enable(enum netns_bpf_attach_type type)
>> +{
>> + switch (type) {
>> + case NETNS_BPF_SK_LOOKUP:
>> + static_branch_inc(&bpf_sk_lookup_enabled);
>> + break;
>> + default:
>> + break;
>> + }
>> +}
>> +
>> /* Must be called with netns_bpf_mutex held. */
>> static void netns_bpf_run_array_detach(struct net *net,
>> enum netns_bpf_attach_type type)
>> @@ -93,6 +115,9 @@ static void bpf_netns_link_release(struct bpf_link *link)
>> if (!net)
>> goto out_unlock;
>>
>> + /* Mark attach point as unused */
>> + netns_bpf_attach_type_disable(type);
>> +
>> /* Remember link position in case of safe delete */
>> idx = link_index(net, type, net_link);
>> list_del(&net_link->node);
>> @@ -416,6 +441,9 @@ static int netns_bpf_link_attach(struct net *net, struct bpf_link *link,
>> lockdep_is_held(&netns_bpf_mutex));
>> bpf_prog_array_free(run_array);
>>
>> + /* Mark attach point as used */
>> + netns_bpf_attach_type_enable(type);
>> +
>> out_unlock:
>> mutex_unlock(&netns_bpf_mutex);
>> return err;
>> @@ -491,8 +519,10 @@ static void __net_exit netns_bpf_pernet_pre_exit(struct net *net)
>> mutex_lock(&netns_bpf_mutex);
>> for (type = 0; type < MAX_NETNS_BPF_ATTACH_TYPE; type++) {
>> netns_bpf_run_array_detach(net, type);
>> - list_for_each_entry(net_link, &net->bpf.links[type], node)
>> + list_for_each_entry(net_link, &net->bpf.links[type], node) {
>> net_link->net = NULL; /* auto-detach link */
>> + netns_bpf_attach_type_disable(type);
>> + }
>> if (net->bpf.progs[type])
>> bpf_prog_put(net->bpf.progs[type]);
>> }
>> diff --git a/net/core/filter.c b/net/core/filter.c
[...]
