UM2019 Extended BPF: A New Type of Software
•
36 likes•33,241 views
BPF (Berkeley Packet Filter) has evolved from a limited virtual machine for efficient packet filtering to a new type of software called extended BPF. Extended BPF allows for custom, efficient, and production-safe performance analysis tools and observability programs to be run in the Linux kernel through BPF. It enables new event-based applications running as BPF programs attached to various kernel events like kprobes, uprobes, tracepoints, sockets, and more. Major companies like Facebook, Google, and Netflix are using BPF programs for tasks like intrusion detection, container security, firewalling, and observability with over 150,000 AWS instances running BPF programs. BPF provides a new program model and security features compared
Report
Share
UM2019 Extended BPF: A New Type of Software
- 1. Extended BPF A New Type of Software Brendan Gregg UbuntuMasters Oct 2019
- 2. BPF
- 3. Kernel Applications System Calls Hardware 50 Years, one (dominant) OS model
- 4. Hardware Supervisor Applications Ring 1 Privilege Ring 0 Ring 2 ... Origins: Multics, 1960s
- 5. Kernel User-mode Applications System Calls Hardware Modern Linux: A new OS model Kernel-mode Applications (BPF) BPF Helper Calls
- 6. 50 Years, one process state model SwappingKernel User Runnable Wait Block Sleep Idle schedule resource I/O acquire lock sleep wait for work Off-CPU On-CPU wakeup acquired wakeup work arrives preemption or time quantum expired swap out swap in Linux groups most sleep states
- 7. BPF program state model Loaded Enabled event fires program ended Off-CPU On-CPU BPF attach Kernel helpers Spinning spin lock
- 9. Kernel Recipes 2019, Alexei Starovoitov ~40 active BPF programs on every Facebook server
- 10. >150k AWS EC2 Ubuntu server instances ~34% US Internet traffic at night >130M subscribers ~14 active BPF programs on every instance (so far)
- 11. Kernel User-mode Applications Hardware Events (incl. clock) Modern Linux: Event-based Applications Kernel-mode Applications (BPF) Scheduler Kernel Events U.E.
- 12. Smaller Kernel User-mode Applications Hardware Modern Linux is becoming Microkernel-ish Kernel-mode Services & Drivers BPF BPF BPF The word “microkernel” has already been invoked by Jonathan Corbet, Thomas Graf, Greg Kroah-Hartman, ...
- 14. BPF
- 15. BPF 1992: Berkeley Packet Filter A limited virtual machine for efficient packet filters # tcpdump -d host 127.0.0.1 and port 80 (000) ldh [12] (001) jeq #0x800 jt 2 jf 18 (002) ld [26] (003) jeq #0x7f000001 jt 6 jf 4 (004) ld [30] (005) jeq #0x7f000001 jt 6 jf 18 (006) ldb [23] (007) jeq #0x84 jt 10 jf 8 (008) jeq #0x6 jt 10 jf 9 (009) jeq #0x11 jt 10 jf 18 (010) ldh [20] (011) jset #0x1fff jt 18 jf 12 (012) ldxb 4*([14]&0xf) (013) ldh [x + 14] (014) jeq #0x50 jt 17 jf 15 (015) ldh [x + 16] (016) jeq #0x50 jt 17 jf 18 (017) ret #262144 (018) ret #0
- 16. BPF 2019: aka extended BPF bpftrace BPF microconference XDP & Facebook Katran, Google KRSI, Netflix flowsrus, and many more bpfconf
- 17. BPF 2019 Kernel kprobes uprobes tracepoints sockets SDN Configuration User-Defined BPF Programs … Event TargetsRuntime perf_events BPF actions BPF verifier DDoS Mitigation Intrusion Detection Container Security Observability Firewalls Device Drivers
- 18. BPF is now a technology name, and no longer an acronym
- 19. BPF Internals 11 Registers Map Storage (Mbytes) Machine Code Execution BPF Helpers JIT Compiler BPF Instructions Rest of Kernel Events BPF Context Verifier Interpreter
- 20. Is BPF Turing complete?
- 21. A New Type of Software Execution model User defined Compil- ation Security Failure mode Resource access User task yes any user based abort syscall, fault Kernel task no static none panic direct BPF event yes JIT, CO-RE verified, JIT error message restricted helpers
- 22. Example Use Case: BPF Observability
- 23. BPF enables a new class of custom, efficient, and production safe performance analysis tools
- 24. BPF Perf Tools
- 25. Ubuntu Install # apt install bcc # apt install bpftrace BCC (BPF Compiler Collection): complex tools bpftrace: custom tools (Ubuntu 19.04+) These are default installs at Netflix, Facebook, etc.
- 26. Example: BCC tcplife Which processes are connecting to which port?
- 27. Example: BCC tcplife # ./tcplife PID COMM LADDR LPORT RADDR RPORT TX_KB RX_KB MS 22597 recordProg 127.0.0.1 46644 127.0.0.1 28527 0 0 0.23 3277 redis-serv 127.0.0.1 28527 127.0.0.1 46644 0 0 0.28 22598 curl 100.66.3.172 61620 52.205.89.26 80 0 1 91.79 22604 curl 100.66.3.172 44400 52.204.43.121 80 0 1 121.38 22624 recordProg 127.0.0.1 46648 127.0.0.1 28527 0 0 0.22 3277 redis-serv 127.0.0.1 28527 127.0.0.1 46648 0 0 0.27 22647 recordProg 127.0.0.1 46650 127.0.0.1 28527 0 0 0.21 3277 redis-serv 127.0.0.1 28527 127.0.0.1 46650 0 0 0.26 [...] Which processes are connecting to which port?
- 28. Example: BCC tcplife # tcplife -h ./usage: tcplife.py [-h] [-T] [-t] [-w] [-s] [-p PID] [-L LOCALPORT] [-D REMOTEPORT] Trace the lifespan of TCP sessions and summarize optional arguments: -h, --help show this help message and exit -T, --time include time column on output (HH:MM:SS) -t, --timestamp include timestamp on output (seconds) -w, --wide wide column output (fits IPv6 addresses) -s, --csv comma separated values output -p PID, --pid PID trace this PID only -L LOCALPORT, --localport LOCALPORT comma-separated list of local ports to trace. -D REMOTEPORT, --remoteport REMOTEPORT comma-separated list of remote ports to trace. examples: ./tcplife # trace all TCP connect()s ./tcplife -t # include time column (HH:MM:SS) [...]
- 29. Example: BCC biolatency What is the distribution of disk I/O latency? Per second?
- 30. Example: BCC biolatency # ./biolatency -mT 1 5 Tracing block device I/O... Hit Ctrl-C to end. 06:20:16 msecs : count distribution 0 -> 1 : 36 |**************************************| 2 -> 3 : 1 |* | 4 -> 7 : 3 |*** | 8 -> 15 : 17 |***************** | 16 -> 31 : 33 |********************************** | 32 -> 63 : 7 |******* | 64 -> 127 : 6 |****** | 06:20:17 msecs : count distribution 0 -> 1 : 96 |************************************ | 2 -> 3 : 25 |********* | 4 -> 7 : 29 |*********** | [...] What is the distribution of disk I/O latency? Per second?
- 32. Example: bpftrace readahead Is readahead polluting the cache?
- 33. Example: bpftrace readahead # readahead.bt Attaching 5 probes... ^C Readahead unused pages: 128 Readahead used page age (ms): @age_ms: [1] 2455 |@@@@@@@@@@@@@@@ | [2, 4) 8424 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [4, 8) 4417 |@@@@@@@@@@@@@@@@@@@@@@@@@@@ | [8, 16) 7680 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ | [16, 32) 4352 |@@@@@@@@@@@@@@@@@@@@@@@@@@ | [32, 64) 0 | | [64, 128) 0 | | [128, 256) 384 |@@ | Is readahead polluting the cache?
- 34. #!/usr/local/bin/bpftrace kprobe:__do_page_cache_readahead { @in_readahead[tid] = 1; } kretprobe:__do_page_cache_readahead { @in_readahead[tid] = 0; } kretprobe:__page_cache_alloc /@in_readahead[tid]/ { @birth[retval] = nsecs; @rapages++; } kprobe:mark_page_accessed /@birth[arg0]/ { @age_ms = hist((nsecs - @birth[arg0]) / 1000000); delete(@birth[arg0]); @rapages--; } END { printf("nReadahead unused pages: %dn", @rapages); printf("nReadahead used page age (ms):n"); print(@age_ms); clear(@age_ms); clear(@birth); clear(@in_readahead); clear(@rapages); }
- 35. Observability Challenges Broken off-CPU flame graph (no frame pointer) libc no frame pointer JIT function tracing
- 36. Many of our perf wins are from CPU flame graphs not CLI tracing Reality Check
- 37. Java JVM Kernel GC CPU Flame Graphs Alphabetical frame sort (A - Z) Stackdepth(0-max)
- 38. BPF-based CPU Flame Graphs perf record perf script stackcollapse-perf.pl flamegraph.pl perf.data flamegraph.pl profile.py Linux 4.9Linux 2.6
- 41. bpftool # bpftool perf pid 1765 fd 6: prog_id 26 kprobe func blk_account_io_start offset 0 pid 1765 fd 8: prog_id 27 kprobe func blk_account_io_done offset 0 pid 1765 fd 11: prog_id 28 kprobe func sched_fork offset 0 pid 1765 fd 15: prog_id 29 kprobe func ttwu_do_wakeup offset 0 pid 1765 fd 17: prog_id 30 kprobe func wake_up_new_task offset 0 pid 1765 fd 19: prog_id 31 kprobe func finish_task_switch offset 0 pid 1765 fd 26: prog_id 33 tracepoint inet_sock_set_state pid 21993 fd 6: prog_id 232 uprobe filename /proc/self/exe offset 1781927 pid 21993 fd 8: prog_id 233 uprobe filename /proc/self/exe offset 1781920 pid 21993 fd 15: prog_id 234 kprobe func blk_account_io_done offset 0 pid 21993 fd 17: prog_id 235 kprobe func blk_account_io_start offset 0 pid 25440 fd 8: prog_id 262 kprobe func blk_mq_start_request offset 0 pid 25440 fd 10: prog_id 263 kprobe func blk_account_io_done offset 0 # bpftool perf pid 1765 fd 6: prog_id 26 kprobe func blk_account_io_start offset 0 pid 1765 fd 8: prog_id 27 kprobe func blk_account_io_done offset 0 pid 1765 fd 11: prog_id 28 kprobe func sched_fork offset 0 pid 1765 fd 15: prog_id 29 kprobe func ttwu_do_wakeup offset 0 pid 1765 fd 17: prog_id 30 kprobe func wake_up_new_task offset 0 pid 1765 fd 19: prog_id 31 kprobe func finish_task_switch offset 0 pid 1765 fd 26: prog_id 33 tracepoint inet_sock_set_state pid 21993 fd 6: prog_id 232 uprobe filename /proc/self/exe offset 1781927 pid 21993 fd 8: prog_id 233 uprobe filename /proc/self/exe offset 1781920 pid 21993 fd 15: prog_id 234 kprobe func blk_account_io_done offset 0 pid 21993 fd 17: prog_id 235 kprobe func blk_account_io_start offset 0 pid 25440 fd 8: prog_id 262 kprobe func blk_mq_start_request offset 0 pid 25440 fd 10: prog_id 263 kprobe func blk_account_io_done offset 0 PID BPF ID Event
- 42. # bpftool prog dump jited id 263 int trace_req_done(struct pt_regs * ctx): 0xffffffffc082dc6f: ; struct request *req = ctx->di; 0: push %rbp 1: mov %rsp,%rbp 4: sub $0x38,%rsp b: sub $0x28,%rbp f: mov %rbx,0x0(%rbp) 13: mov %r13,0x8(%rbp) 17: mov %r14,0x10(%rbp) 1b: mov %r15,0x18(%rbp) 1f: xor %eax,%eax 21: mov %rax,0x20(%rbp) 25: mov 0x70(%rdi),%rdi ; struct request *req = ctx->di; 29: mov %rdi,-0x8(%rbp) ; tsp = bpf_map_lookup_elem((void *)bpf_pseudo_fd(1, -1), &req); 2d: movabs $0xffff96e680ab0000,%rdi 37: mov %rbp,%rsi 3a: add $0xfffffffffffffff8,%rsi ; tsp = bpf_map_lookup_elem((void *)bpf_pseudo_fd(1, -1), &req); 3e: callq 0xffffffffc39a49c1
- 43. LPC 2019, Arnaldo Carvalho de Melo CPU profiling of BPF programs
- 44. “We should be able to single-step execution... We should be able to take a core dump of all state.” – David S. Miller, LSFMM 2019 UNIVAC 1 1951
- 45. Future
- 46. Future Predictions More device drivers, incl. USB on BPF (ghk) Monitoring agents Intrusion detection systems TCP congestion controls CPU & container schedulers FS readahead policies CDN accelerator
- 47. Take Aways BPF is a new software type Start using BPF perf tools on Ubuntu: bcc, bpftrace
- 48. Thanks BPF: Alexei Starovoitov, Daniel Borkmann, David S. Miller, Linus Torvalds, BPF community BCC: Brenden Blanco, Yonghong Song, Sasha Goldsthein, BCC community bpftrace: Alastair Robertson, Matheus Marchini, Dan Xu, bpftrace community Canonical: BPF support, and libc-fp (thanks in advance) All photos credit myself; except slide 2 (Netflix) and 9 (KernelRecipes)