bpf: Prevent bpf program recursion for raw tracepoint probes
Subsystems (5)
Jiri Olsa
reviewerPATCH v2 1/1 bpf: Prevent bpf program recursion for raw tracepoint probes
We got report from sysbot [1] about warnings that were caused by bpf program attached to contention_begin raw tracepoint triggering the same tracepoint by using bpf_trace_printk helper that takes trace_printk_lock lock.
Call Trace: <TASK> ? trace_event_raw_event_bpf_trace_printk+0x5f/0x90 bpf_trace_printk+0x2b/0xe0 bpf_prog_a9aec6167c091eef_prog+0x1f/0x24 bpf_trace_run2+0x26/0x90 native_queued_spin_lock_slowpath+0x1c6/0x2b0 _raw_spin_lock_irqsave+0x44/0x50 bpf_trace_printk+0x3f/0xe0 bpf_prog_a9aec6167c091eef_prog+0x1f/0x24 bpf_trace_run2+0x26/0x90 native_queued_spin_lock_slowpath+0x1c6/0x2b0 _raw_spin_lock_irqsave+0x44/0x50 bpf_trace_printk+0x3f/0xe0 bpf_prog_a9aec6167c091eef_prog+0x1f/0x24 bpf_trace_run2+0x26/0x90 native_queued_spin_lock_slowpath+0x1c6/0x2b0 _raw_spin_lock_irqsave+0x44/0x50 bpf_trace_printk+0x3f/0xe0 bpf_prog_a9aec6167c091eef_prog+0x1f/0x24 bpf_trace_run2+0x26/0x90 native_queued_spin_lock_slowpath+0x1c6/0x2b0 _raw_spin_lock_irqsave+0x44/0x50 __unfreeze_partials+0x5b/0x160 ...
The can be reproduced by attaching bpf program as raw tracepoint on contention_begin tracepoint. The bpf prog calls bpf_trace_printk helper. Then by running perf bench the spin lock code is forced to take slow path and call contention_begin tracepoint.
Fixing this by skipping execution of the bpf program if it's already running, Using bpf prog 'active' field, which is being currently used by trampoline programs for the same reason.
Moving bpf_prog_inc_misses_counter to syscall.c because trampoline.c is compiled in just for CONFIG_BPF_JIT option.
[1] https://lore.kernel.org/bpf/YxhFe3EwqchC%2FfYf@krava/T/#t
---
v2 changes:
- moved inc_misses_counter from trampoline to syscall object to fix compilation fails found by kernel test robot
- added Stanislav's ack
include/linux/bpf.h | 6 ++++++ kernel/bpf/syscall.c | 11 +++++++++++ kernel/bpf/trampoline.c | 15 ++------------- kernel/trace/bpf_trace.c | 6 ++++++ 4 files changed, 25 insertions(+), 13 deletions(-)
include/linux/bpf.h +6 -0
@@ -2029,6 +2029,8 @@ static inline bool has_current_bpf_ctx(void) {return !!current->bpf_ctx; } + +void notrace bpf_prog_inc_misses_counter(struct bpf_prog *prog);#else /* !CONFIG_BPF_SYSCALL */ static inline struct bpf_prog *bpf_prog_get (u32 ufd) {
@@ -2251,6 +2253,10 @@ static inline bool has_current_bpf_ctx(void) {return false; } + +static inline void bpf_prog_inc_misses_counter (struct bpf_prog *prog) +{ +}#endif /* CONFIG_BPF_SYSCALL */ void __bpf_free_used_btfs(struct bpf_prog_aux *aux,
kernel/bpf/syscall.c +11 -0
@@ -2098,6 +2098,17 @@ struct bpf_prog_kstats {u64 misses; }; +void notrace bpf_prog_inc_misses_counter(struct bpf_prog *prog) +{ +struct bpf_prog_stats *stats; +unsigned int flags; + + stats =this_cpu_ptr (prog->stats); + flags =u64_stats_update_begin_irqsave (&stats->syncp); +u64_stats_inc (&stats->misses); +u64_stats_update_end_irqrestore (&stats->syncp, flags); +} +static void bpf_prog_get_stats (const struct bpf_prog *prog,struct bpf_prog_kstats *stats) {
kernel/bpf/trampoline.c +2 -13
@@ -863,17 +863,6 @@ static __always_inline u64 notrace bpf_prog_start_time(void)return start; } -static void notrace inc_misses_counter(struct bpf_prog *prog) -{ -struct bpf_prog_stats *stats; -unsigned int flags; - - stats =this_cpu_ptr (prog->stats); - flags =u64_stats_update_begin_irqsave (&stats->syncp); -u64_stats_inc (&stats->misses); -u64_stats_update_end_irqrestore (&stats->syncp, flags); -} -/* The logic is similar to bpf_prog_run(), but with an explicit * rcu_read_lock() and migrate_disable() which are required * for the trampoline. The macro is split into
@@ -896,7 +885,7 @@ u64 notrace __bpf_prog_enter(struct bpf_prog *prog, struct bpf_tramp_run_ctx *ru run_ctx->saved_run_ctx =bpf_set_run_ctx (&run_ctx->run_ctx);if (unlikely (this_cpu_inc_return (*(prog->active)) !=1 )) { -inc_misses_counter (prog); +bpf_prog_inc_misses_counter (prog);return 0 ; }return bpf_prog_start_time ();
@@ -967,7 +956,7 @@ u64 notrace __bpf_prog_enter_sleepable(struct bpf_prog *prog, struct bpf_tramp_rmight_fault ();if (unlikely (this_cpu_inc_return (*(prog->active)) !=1 )) { -inc_misses_counter (prog); +bpf_prog_inc_misses_counter (prog);return 0 ; }
kernel/trace/bpf_trace.c +6 -0
@@ -2042,9 +2042,15 @@ static __always_inlinevoid __bpf_trace_run (struct bpf_prog *prog,u64 *args) {cant_sleep (); +if (unlikely (this_cpu_inc_return (*(prog->active)) !=1 )) { +bpf_prog_inc_misses_counter (prog); + goto out; + }rcu_read_lock (); (void )bpf_prog_run (prog, args);rcu_read_unlock (); +out: +this_cpu_dec (*(prog->active)); }#define UNPACK (...) __VA_ARGS__
Signature · 1 line
2.37.3
This patch was applied to bpf/bpf-next.git (master) by Alexei Starovoitov <[email protected]>:
Hello:
This patch was applied to bpf/bpf-next.git (master) by Alexei Starovoitov <[email protected]>:
Quoted 31 lines · On Fri, 16 Sep 2022 09:19:14 +0200 you wrote:
We got report from sysbot [1] about warnings that were caused by bpf program attached to
contention_beginraw tracepoint triggering the same tracepoint by usingbpf_trace_printkhelper that takestrace_printk_locklock.Call Trace: <TASK> ? trace_event_raw_event_bpf_trace_printk+0x5f/0x90 bpf_trace_printk+0x2b/0xe0 bpf_prog_a9aec6167c091eef_prog+0x1f/0x24 bpf_trace_run2+0x26/0x90 native_queued_spin_lock_slowpath+0x1c6/0x2b0 _raw_spin_lock_irqsave+0x44/0x50 bpf_trace_printk+0x3f/0xe0 bpf_prog_a9aec6167c091eef_prog+0x1f/0x24 bpf_trace_run2+0x26/0x90 native_queued_spin_lock_slowpath+0x1c6/0x2b0 _raw_spin_lock_irqsave+0x44/0x50 bpf_trace_printk+0x3f/0xe0 bpf_prog_a9aec6167c091eef_prog+0x1f/0x24 bpf_trace_run2+0x26/0x90 native_queued_spin_lock_slowpath+0x1c6/0x2b0 _raw_spin_lock_irqsave+0x44/0x50 bpf_trace_printk+0x3f/0xe0 bpf_prog_a9aec6167c091eef_prog+0x1f/0x24 bpf_trace_run2+0x26/0x90 native_queued_spin_lock_slowpath+0x1c6/0x2b0 _raw_spin_lock_irqsave+0x44/0x50 __unfreeze_partials+0x5b/0x160 ...
Quoted 1 line
[...]
Here is the summary with links:
- [PATCHv2,bpf-next] bpf: Prevent bpf program recursion for raw tracepoint probes
https://git.kernel.org/bpf/bpf-next/c/05b24ff9b2cf
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html