Steering connections to sockets with BPF socket lookup hook Jakub Sitnicki, Cloudflare @jkbs0 @cloudflare October 28-29, 2020 Who am I? ● Software Engineer at Cloudflare Spectrum TCP/UDP reverse proxy are closed check VM IP What is socket lookup? raw PREROUTING filter INPUT conntrack routing decision mangle PREROUTING nat PREROUTING socket lookup socket receive buffer Application Protocol path for local delivery Service dispatch with BPF socket lookup packet metadata BPF program lookup result 010 101 010 struct bpf_sk_lookup { __u32 family; __u32 protocol; __u32 remote_ip4;

conferences Email: klaus.iglberger@gmx.deContent 3 Overview Name Lookup (Un-)Qualified Name Lookup Argument Dependent Lookup Two-Phase Lookup Template Argument Deduction SFINAE Overload Resolution Virtual Dispatch Deleting FunctionsContent 4 Overview Name Lookup (Un-)Qualified Name Lookup Argument Dependent Lookup Two-Phase Lookup Template Argument Deduction SFINAE Overload Resolution examples; … mention every single detail; … try to be bleeding edge.Overview 6Overview 7 Name Lookup Name Lookup Select all (visible) candidate functions with a certain name within the current scope. If

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 Consistent lookup vindexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 398 Lookup as Primary Vindex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 398 Non-Unique Lookup Vindexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410 Unique Lookup Vindexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 Consistent lookup vindexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 461 Lookup as Primary Vindex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 461 Non-Unique Lookup Vindexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474 Unique Lookup Vindexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Consistent lookup vindexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435 Lookup as Primary Vindex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435 18 Non-Unique Lookup Vindexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 448 Unique Lookup Vindexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Consistent lookup vindexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514 Lookup as Primary Vindex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515 Non-Unique Lookup Vindexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

CREATE TABLE customer_lookup ( id int NOT NULL, keyspace_id varbinary(256), primary key(id) ); The customer table is the main table we want to shard using country. The lookup table will help us do that keyspace_id. The lookup table is used to store the id to keyspace_id mapping. We connect it to the customer table as follows: We first define a lookup vindex: "customer_region_lookup": { "type": "co "consistent_lookup_unique", "params": { "table": "customer_lookup", "from": "id", "to": "keyspace_id" }, "owner": "customer" }, Then we create it as a vindex on the customer table: "customer": { "column_vindexes":

will examine each of these features in turn:  Function overloading and overload resolution  Name lookup  Default function arguments  Function templates  For each feature, we’ll explain first how it with an example that combines several of these features. 9 Outline  Function Overloading  Name Lookup  Default Function Arguments  Function Templates  Tying It All Together 10Back to Basics - Function (constand volatile), and  ref-qualifiers (&and &&). 33 Outline  Function Overloading  Name Lookup  Default Function Arguments  Function Templates  Tying It All Together 34Back to Basics - Function

subclass Writing a FileField subclass Custom Lookups A simple lookup example A simple transformer example Writing an efficient abs__lt lookup A bilateral transformer example Writing alternative implementations reference Model class reference Model Meta options Model instance reference QuerySet API reference Lookup API reference Query Expressions Conditional Expressions Database Functions Request and response types | Indexes | Meta options | Model class QuerySets: Making queries | QuerySet method reference | Lookup expressions Model instances: Instance methods | Accessing related objects Migrations: Introduction

subclass Writing a FileField subclass How to write custom lookups A lookup example A transformer example Writing an efficient abs__lt lookup A bilateral transformer example Writing alternative implementations reference Model class reference Model Meta options Model instance reference QuerySet API reference Lookup API reference Query Expressions Conditional Expressions Database Functions Paginator Paginator types | Indexes | Meta options | Model class QuerySets: Making queries | QuerySet method reference | Lookup expressions Model instances: Instance methods | Accessing related objects Migrations: Introduction